diff --git a/sys/contrib/openzfs/META b/sys/contrib/openzfs/META index 8a257f0feb17..19a796050f5b 100644 --- a/sys/contrib/openzfs/META +++ b/sys/contrib/openzfs/META @@ -1,10 +1,10 @@ Meta: 1 Name: zfs Branch: 1.0 Version: 2.2.99 Release: 1 Release-Tags: relext License: CDDL Author: OpenZFS -Linux-Maximum: 6.7 +Linux-Maximum: 6.8 Linux-Minimum: 3.10 diff --git a/sys/contrib/openzfs/cmd/arcstat.in b/sys/contrib/openzfs/cmd/arcstat.in index 8df1c62f7e86..220f343b5b62 100755 --- a/sys/contrib/openzfs/cmd/arcstat.in +++ b/sys/contrib/openzfs/cmd/arcstat.in @@ -1,637 +1,674 @@ #!/usr/bin/env @PYTHON_SHEBANG@ # # Print out ZFS ARC Statistics exported via kstat(1) # For a definition of fields, or usage, use arcstat -v # # This script was originally a fork of the original arcstat.pl (0.1) # by Neelakanth Nadgir, originally published on his Sun blog on # 09/18/2007 # http://blogs.sun.com/realneel/entry/zfs_arc_statistics # # A new version aimed to improve upon the original by adding features # and fixing bugs as needed. This version was maintained by Mike # Harsch and was hosted in a public open source repository: # http://github.com/mharsch/arcstat # # but has since moved to the illumos-gate repository. # # This Python port was written by John Hixson for FreeNAS, introduced # in commit e2c29f: # https://github.com/freenas/freenas # # and has been improved by many people since. # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License, Version 1.0 only # (the "License"). You may not use this file except in compliance # with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Fields have a fixed width. Every interval, we fill the "v" # hash with its corresponding value (v[field]=value) using calculate(). # @hdr is the array of fields that needs to be printed, so we # just iterate over this array and print the values using our pretty printer. # # This script must remain compatible with Python 3.6+. # import sys import time import getopt import re import copy from signal import signal, SIGINT, SIGWINCH, SIG_DFL cols = { # HDR: [Size, Scale, Description] "time": [8, -1, "Time"], "hits": [4, 1000, "ARC hits per second"], "iohs": [4, 1000, "ARC I/O hits per second"], "miss": [4, 1000, "ARC misses per second"], "read": [4, 1000, "Total ARC accesses per second"], "hit%": [4, 100, "ARC hit percentage"], "ioh%": [4, 100, "ARC I/O hit percentage"], "miss%": [5, 100, "ARC miss percentage"], "dhit": [4, 1000, "Demand hits per second"], "dioh": [4, 1000, "Demand I/O hits per second"], "dmis": [4, 1000, "Demand misses per second"], "dh%": [3, 100, "Demand hit percentage"], "di%": [3, 100, "Demand I/O hit percentage"], "dm%": [3, 100, "Demand miss percentage"], "ddhit": [5, 1000, "Demand data hits per second"], "ddioh": [5, 1000, "Demand data I/O hits per second"], "ddmis": [5, 1000, "Demand data misses per second"], "ddh%": [4, 100, "Demand data hit percentage"], "ddi%": [4, 100, "Demand data I/O hit percentage"], "ddm%": [4, 100, "Demand data miss percentage"], "dmhit": [5, 1000, "Demand metadata hits per second"], "dmioh": [5, 1000, "Demand metadata I/O hits per second"], "dmmis": [5, 1000, "Demand metadata misses per second"], "dmh%": [4, 100, "Demand metadata hit percentage"], "dmi%": [4, 100, "Demand metadata I/O hit percentage"], "dmm%": [4, 100, "Demand metadata miss percentage"], "phit": [4, 1000, "Prefetch hits per second"], "pioh": [4, 1000, "Prefetch I/O hits per second"], "pmis": [4, 1000, "Prefetch misses per second"], "ph%": [3, 100, "Prefetch hits percentage"], "pi%": [3, 100, "Prefetch I/O hits percentage"], "pm%": [3, 100, "Prefetch miss percentage"], "pdhit": [5, 1000, "Prefetch data hits per second"], "pdioh": [5, 1000, "Prefetch data I/O hits per second"], "pdmis": [5, 1000, "Prefetch data misses per second"], "pdh%": [4, 100, "Prefetch data hits percentage"], "pdi%": [4, 100, "Prefetch data I/O hits percentage"], "pdm%": [4, 100, "Prefetch data miss percentage"], "pmhit": [5, 1000, "Prefetch metadata hits per second"], "pmioh": [5, 1000, "Prefetch metadata I/O hits per second"], "pmmis": [5, 1000, "Prefetch metadata misses per second"], "pmh%": [4, 100, "Prefetch metadata hits percentage"], "pmi%": [4, 100, "Prefetch metadata I/O hits percentage"], "pmm%": [4, 100, "Prefetch metadata miss percentage"], "mhit": [4, 1000, "Metadata hits per second"], "mioh": [4, 1000, "Metadata I/O hits per second"], "mmis": [4, 1000, "Metadata misses per second"], "mread": [5, 1000, "Metadata accesses per second"], "mh%": [3, 100, "Metadata hit percentage"], "mi%": [3, 100, "Metadata I/O hit percentage"], "mm%": [3, 100, "Metadata miss percentage"], "arcsz": [5, 1024, "ARC size"], "size": [5, 1024, "ARC size"], "c": [5, 1024, "ARC target size"], "mfu": [4, 1000, "MFU list hits per second"], "mru": [4, 1000, "MRU list hits per second"], "mfug": [4, 1000, "MFU ghost list hits per second"], "mrug": [4, 1000, "MRU ghost list hits per second"], "unc": [4, 1000, "Uncached list hits per second"], "eskip": [5, 1000, "evict_skip per second"], "el2skip": [7, 1000, "evict skip, due to l2 writes, per second"], "el2cach": [7, 1024, "Size of L2 cached evictions per second"], "el2el": [5, 1024, "Size of L2 eligible evictions per second"], "el2mfu": [6, 1024, "Size of L2 eligible MFU evictions per second"], "el2mru": [6, 1024, "Size of L2 eligible MRU evictions per second"], "el2inel": [7, 1024, "Size of L2 ineligible evictions per second"], "mtxmis": [6, 1000, "mutex_miss per second"], "dread": [5, 1000, "Demand accesses per second"], "ddread": [6, 1000, "Demand data accesses per second"], "dmread": [6, 1000, "Demand metadata accesses per second"], "pread": [5, 1000, "Prefetch accesses per second"], "pdread": [6, 1000, "Prefetch data accesses per second"], "pmread": [6, 1000, "Prefetch metadata accesses per second"], "l2hits": [6, 1000, "L2ARC hits per second"], "l2miss": [6, 1000, "L2ARC misses per second"], "l2read": [6, 1000, "Total L2ARC accesses per second"], "l2hit%": [6, 100, "L2ARC access hit percentage"], "l2miss%": [7, 100, "L2ARC access miss percentage"], "l2pref": [6, 1024, "L2ARC prefetch allocated size"], "l2mfu": [5, 1024, "L2ARC MFU allocated size"], "l2mru": [5, 1024, "L2ARC MRU allocated size"], "l2data": [6, 1024, "L2ARC data allocated size"], "l2meta": [6, 1024, "L2ARC metadata allocated size"], "l2pref%": [7, 100, "L2ARC prefetch percentage"], "l2mfu%": [6, 100, "L2ARC MFU percentage"], "l2mru%": [6, 100, "L2ARC MRU percentage"], "l2data%": [7, 100, "L2ARC data percentage"], "l2meta%": [7, 100, "L2ARC metadata percentage"], "l2asize": [7, 1024, "Actual (compressed) size of the L2ARC"], "l2size": [6, 1024, "Size of the L2ARC"], "l2bytes": [7, 1024, "Bytes read per second from the L2ARC"], "grow": [4, 1000, "ARC grow disabled"], "need": [5, 1024, "ARC reclaim need"], "free": [5, 1024, "ARC free memory"], "avail": [5, 1024, "ARC available memory"], "waste": [5, 1024, "Wasted memory due to round up to pagesize"], + "ztotal": [6, 1000, "zfetch total prefetcher calls per second"], + "zhits": [5, 1000, "zfetch stream hits per second"], + "zahead": [6, 1000, "zfetch hits ahead of streams per second"], + "zpast": [5, 1000, "zfetch hits behind streams per second"], + "zmisses": [7, 1000, "zfetch stream misses per second"], + "zmax": [4, 1000, "zfetch limit reached per second"], + "zfuture": [7, 1000, "zfetch stream future per second"], + "zstride": [7, 1000, "zfetch stream strides per second"], + "zissued": [7, 1000, "zfetch prefetches issued per second"], + "zactive": [7, 1000, "zfetch prefetches active per second"], } v = {} hdr = ["time", "read", "ddread", "ddh%", "dmread", "dmh%", "pread", "ph%", "size", "c", "avail"] xhdr = ["time", "mfu", "mru", "mfug", "mrug", "unc", "eskip", "mtxmis", "dread", "pread", "read"] +zhdr = ["time", "ztotal", "zhits", "zahead", "zpast", "zmisses", "zmax", + "zfuture", "zstride", "zissued", "zactive"] sint = 1 # Default interval is 1 second count = 1 # Default count is 1 hdr_intr = 20 # Print header every 20 lines of output opfile = None sep = " " # Default separator is 2 spaces l2exist = False cmd = ("Usage: arcstat [-havxp] [-f fields] [-o file] [-s string] [interval " "[count]]\n") cur = {} d = {} out = None kstat = None pretty_print = True if sys.platform.startswith('freebsd'): # Requires py-sysctl on FreeBSD import sysctl def kstat_update(): global kstat k = [ctl for ctl in sysctl.filter('kstat.zfs.misc.arcstats') if ctl.type != sysctl.CTLTYPE_NODE] if not k: sys.exit(1) kstat = {} for s in k: if not s: continue name, value = s.name, s.value # Trims 'kstat.zfs.misc.arcstats' from the name kstat[name[24:]] = int(value) elif sys.platform.startswith('linux'): def kstat_update(): global kstat - k = [line.strip() for line in open('/proc/spl/kstat/zfs/arcstats')] + k1 = [line.strip() for line in open('/proc/spl/kstat/zfs/arcstats')] - if not k: + k2 = ["zfetch_" + line.strip() for line in + open('/proc/spl/kstat/zfs/zfetchstats')] + + if k1 is None or k2 is None: sys.exit(1) - del k[0:2] + del k1[0:2] + del k2[0:2] + k = k1 + k2 kstat = {} for s in k: if not s: continue name, unused, value = s.split() kstat[name] = int(value) def detailed_usage(): sys.stderr.write("%s\n" % cmd) sys.stderr.write("Field definitions are as follows:\n") for key in cols: sys.stderr.write("%11s : %s\n" % (key, cols[key][2])) sys.stderr.write("\n") sys.exit(0) def usage(): sys.stderr.write("%s\n" % cmd) sys.stderr.write("\t -h : Print this help message\n") sys.stderr.write("\t -a : Print all possible stats\n") sys.stderr.write("\t -v : List all possible field headers and definitions" "\n") sys.stderr.write("\t -x : Print extended stats\n") + sys.stderr.write("\t -z : Print zfetch stats\n") sys.stderr.write("\t -f : Specify specific fields to print (see -v)\n") sys.stderr.write("\t -o : Redirect output to the specified file\n") sys.stderr.write("\t -s : Override default field separator with custom " "character or string\n") sys.stderr.write("\t -p : Disable auto-scaling of numerical fields\n") sys.stderr.write("\nExamples:\n") sys.stderr.write("\tarcstat -o /tmp/a.log 2 10\n") sys.stderr.write("\tarcstat -s \",\" -o /tmp/a.log 2 10\n") sys.stderr.write("\tarcstat -v\n") sys.stderr.write("\tarcstat -f time,hit%,dh%,ph%,mh% 1\n") sys.stderr.write("\n") sys.exit(1) def snap_stats(): global cur global kstat prev = copy.deepcopy(cur) kstat_update() cur = kstat for key in cur: if re.match(key, "class"): continue if key in prev: d[key] = cur[key] - prev[key] else: d[key] = cur[key] def prettynum(sz, scale, num=0): suffix = [' ', 'K', 'M', 'G', 'T', 'P', 'E', 'Z'] index = 0 save = 0 # Special case for date field if scale == -1: return "%s" % num # Rounding error, return 0 elif 0 < num < 1: num = 0 while abs(num) > scale and index < 5: save = num num = num / scale index += 1 if index == 0: return "%*d" % (sz, num) if abs(save / scale) < 10: return "%*.1f%s" % (sz - 1, num, suffix[index]) else: return "%*d%s" % (sz - 1, num, suffix[index]) def print_values(): global hdr global sep global v global pretty_print if pretty_print: fmt = lambda col: prettynum(cols[col][0], cols[col][1], v[col]) else: fmt = lambda col: str(v[col]) sys.stdout.write(sep.join(fmt(col) for col in hdr)) sys.stdout.write("\n") sys.stdout.flush() def print_header(): global hdr global sep global pretty_print if pretty_print: fmt = lambda col: "%*s" % (cols[col][0], col) else: fmt = lambda col: col sys.stdout.write(sep.join(fmt(col) for col in hdr)) sys.stdout.write("\n") def get_terminal_lines(): try: import fcntl import termios import struct data = fcntl.ioctl(sys.stdout.fileno(), termios.TIOCGWINSZ, '1234') sz = struct.unpack('hh', data) return sz[0] except Exception: pass def update_hdr_intr(): global hdr_intr lines = get_terminal_lines() if lines and lines > 3: hdr_intr = lines - 3 def resize_handler(signum, frame): update_hdr_intr() def init(): global sint global count global hdr global xhdr + global zhdr global opfile global sep global out global l2exist global pretty_print desired_cols = None aflag = False xflag = False hflag = False vflag = False + zflag = False i = 1 try: opts, args = getopt.getopt( sys.argv[1:], - "axo:hvs:f:p", + "axzo:hvs:f:p", [ "all", "extended", + "zfetch", "outfile", "help", "verbose", "separator", "columns", "parsable" ] ) except getopt.error as msg: sys.stderr.write("Error: %s\n" % str(msg)) usage() opts = None for opt, arg in opts: if opt in ('-a', '--all'): aflag = True if opt in ('-x', '--extended'): xflag = True if opt in ('-o', '--outfile'): opfile = arg i += 1 if opt in ('-h', '--help'): hflag = True if opt in ('-v', '--verbose'): vflag = True if opt in ('-s', '--separator'): sep = arg i += 1 if opt in ('-f', '--columns'): desired_cols = arg i += 1 if opt in ('-p', '--parsable'): pretty_print = False + if opt in ('-z', '--zfetch'): + zflag = True i += 1 argv = sys.argv[i:] sint = int(argv[0]) if argv else sint count = int(argv[1]) if len(argv) > 1 else (0 if len(argv) > 0 else 1) - if hflag or (xflag and desired_cols): + if hflag or (xflag and zflag) or ((zflag or xflag) and desired_cols): usage() if vflag: detailed_usage() if xflag: hdr = xhdr + if zflag: + hdr = zhdr + update_hdr_intr() # check if L2ARC exists snap_stats() l2_size = cur.get("l2_size") if l2_size: l2exist = True if desired_cols: hdr = desired_cols.split(",") invalid = [] incompat = [] for ele in hdr: if ele not in cols: invalid.append(ele) elif not l2exist and ele.startswith("l2"): sys.stdout.write("No L2ARC Here\n%s\n" % ele) incompat.append(ele) if len(invalid) > 0: sys.stderr.write("Invalid column definition! -- %s\n" % invalid) usage() if len(incompat) > 0: sys.stderr.write("Incompatible field specified! -- %s\n" % incompat) usage() if aflag: if l2exist: hdr = cols.keys() else: hdr = [col for col in cols.keys() if not col.startswith("l2")] if opfile: try: out = open(opfile, "w") sys.stdout = out except IOError: sys.stderr.write("Cannot open %s for writing\n" % opfile) sys.exit(1) def calculate(): global d global v global l2exist v = dict() v["time"] = time.strftime("%H:%M:%S", time.localtime()) v["hits"] = d["hits"] // sint v["iohs"] = d["iohits"] // sint v["miss"] = d["misses"] // sint v["read"] = v["hits"] + v["iohs"] + v["miss"] v["hit%"] = 100 * v["hits"] // v["read"] if v["read"] > 0 else 0 v["ioh%"] = 100 * v["iohs"] // v["read"] if v["read"] > 0 else 0 v["miss%"] = 100 - v["hit%"] - v["ioh%"] if v["read"] > 0 else 0 v["dhit"] = (d["demand_data_hits"] + d["demand_metadata_hits"]) // sint v["dioh"] = (d["demand_data_iohits"] + d["demand_metadata_iohits"]) // sint v["dmis"] = (d["demand_data_misses"] + d["demand_metadata_misses"]) // sint v["dread"] = v["dhit"] + v["dioh"] + v["dmis"] v["dh%"] = 100 * v["dhit"] // v["dread"] if v["dread"] > 0 else 0 v["di%"] = 100 * v["dioh"] // v["dread"] if v["dread"] > 0 else 0 v["dm%"] = 100 - v["dh%"] - v["di%"] if v["dread"] > 0 else 0 v["ddhit"] = d["demand_data_hits"] // sint v["ddioh"] = d["demand_data_iohits"] // sint v["ddmis"] = d["demand_data_misses"] // sint v["ddread"] = v["ddhit"] + v["ddioh"] + v["ddmis"] v["ddh%"] = 100 * v["ddhit"] // v["ddread"] if v["ddread"] > 0 else 0 v["ddi%"] = 100 * v["ddioh"] // v["ddread"] if v["ddread"] > 0 else 0 v["ddm%"] = 100 - v["ddh%"] - v["ddi%"] if v["ddread"] > 0 else 0 v["dmhit"] = d["demand_metadata_hits"] // sint v["dmioh"] = d["demand_metadata_iohits"] // sint v["dmmis"] = d["demand_metadata_misses"] // sint v["dmread"] = v["dmhit"] + v["dmioh"] + v["dmmis"] v["dmh%"] = 100 * v["dmhit"] // v["dmread"] if v["dmread"] > 0 else 0 v["dmi%"] = 100 * v["dmioh"] // v["dmread"] if v["dmread"] > 0 else 0 v["dmm%"] = 100 - v["dmh%"] - v["dmi%"] if v["dmread"] > 0 else 0 v["phit"] = (d["prefetch_data_hits"] + d["prefetch_metadata_hits"]) // sint v["pioh"] = (d["prefetch_data_iohits"] + d["prefetch_metadata_iohits"]) // sint v["pmis"] = (d["prefetch_data_misses"] + d["prefetch_metadata_misses"]) // sint v["pread"] = v["phit"] + v["pioh"] + v["pmis"] v["ph%"] = 100 * v["phit"] // v["pread"] if v["pread"] > 0 else 0 v["pi%"] = 100 * v["pioh"] // v["pread"] if v["pread"] > 0 else 0 v["pm%"] = 100 - v["ph%"] - v["pi%"] if v["pread"] > 0 else 0 v["pdhit"] = d["prefetch_data_hits"] // sint v["pdioh"] = d["prefetch_data_iohits"] // sint v["pdmis"] = d["prefetch_data_misses"] // sint v["pdread"] = v["pdhit"] + v["pdioh"] + v["pdmis"] v["pdh%"] = 100 * v["pdhit"] // v["pdread"] if v["pdread"] > 0 else 0 v["pdi%"] = 100 * v["pdioh"] // v["pdread"] if v["pdread"] > 0 else 0 v["pdm%"] = 100 - v["pdh%"] - v["pdi%"] if v["pdread"] > 0 else 0 v["pmhit"] = d["prefetch_metadata_hits"] // sint v["pmioh"] = d["prefetch_metadata_iohits"] // sint v["pmmis"] = d["prefetch_metadata_misses"] // sint v["pmread"] = v["pmhit"] + v["pmioh"] + v["pmmis"] v["pmh%"] = 100 * v["pmhit"] // v["pmread"] if v["pmread"] > 0 else 0 v["pmi%"] = 100 * v["pmioh"] // v["pmread"] if v["pmread"] > 0 else 0 v["pmm%"] = 100 - v["pmh%"] - v["pmi%"] if v["pmread"] > 0 else 0 v["mhit"] = (d["prefetch_metadata_hits"] + d["demand_metadata_hits"]) // sint v["mioh"] = (d["prefetch_metadata_iohits"] + d["demand_metadata_iohits"]) // sint v["mmis"] = (d["prefetch_metadata_misses"] + d["demand_metadata_misses"]) // sint v["mread"] = v["mhit"] + v["mioh"] + v["mmis"] v["mh%"] = 100 * v["mhit"] // v["mread"] if v["mread"] > 0 else 0 v["mi%"] = 100 * v["mioh"] // v["mread"] if v["mread"] > 0 else 0 v["mm%"] = 100 - v["mh%"] - v["mi%"] if v["mread"] > 0 else 0 v["arcsz"] = cur["size"] v["size"] = cur["size"] v["c"] = cur["c"] v["mfu"] = d["mfu_hits"] // sint v["mru"] = d["mru_hits"] // sint v["mrug"] = d["mru_ghost_hits"] // sint v["mfug"] = d["mfu_ghost_hits"] // sint v["unc"] = d["uncached_hits"] // sint v["eskip"] = d["evict_skip"] // sint v["el2skip"] = d["evict_l2_skip"] // sint v["el2cach"] = d["evict_l2_cached"] // sint v["el2el"] = d["evict_l2_eligible"] // sint v["el2mfu"] = d["evict_l2_eligible_mfu"] // sint v["el2mru"] = d["evict_l2_eligible_mru"] // sint v["el2inel"] = d["evict_l2_ineligible"] // sint v["mtxmis"] = d["mutex_miss"] // sint + v["ztotal"] = (d["zfetch_hits"] + d["zfetch_future"] + d["zfetch_stride"] + + d["zfetch_past"] + d["zfetch_misses"]) // sint + v["zhits"] = d["zfetch_hits"] // sint + v["zahead"] = (d["zfetch_future"] + d["zfetch_stride"]) // sint + v["zpast"] = d["zfetch_past"] // sint + v["zmisses"] = d["zfetch_misses"] // sint + v["zmax"] = d["zfetch_max_streams"] // sint + v["zfuture"] = d["zfetch_future"] // sint + v["zstride"] = d["zfetch_stride"] // sint + v["zissued"] = d["zfetch_io_issued"] // sint + v["zactive"] = d["zfetch_io_active"] // sint if l2exist: v["l2hits"] = d["l2_hits"] // sint v["l2miss"] = d["l2_misses"] // sint v["l2read"] = v["l2hits"] + v["l2miss"] v["l2hit%"] = 100 * v["l2hits"] // v["l2read"] if v["l2read"] > 0 else 0 v["l2miss%"] = 100 - v["l2hit%"] if v["l2read"] > 0 else 0 v["l2asize"] = cur["l2_asize"] v["l2size"] = cur["l2_size"] v["l2bytes"] = d["l2_read_bytes"] // sint v["l2pref"] = cur["l2_prefetch_asize"] v["l2mfu"] = cur["l2_mfu_asize"] v["l2mru"] = cur["l2_mru_asize"] v["l2data"] = cur["l2_bufc_data_asize"] v["l2meta"] = cur["l2_bufc_metadata_asize"] v["l2pref%"] = 100 * v["l2pref"] // v["l2asize"] v["l2mfu%"] = 100 * v["l2mfu"] // v["l2asize"] v["l2mru%"] = 100 * v["l2mru"] // v["l2asize"] v["l2data%"] = 100 * v["l2data"] // v["l2asize"] v["l2meta%"] = 100 * v["l2meta"] // v["l2asize"] v["grow"] = 0 if cur["arc_no_grow"] else 1 v["need"] = cur["arc_need_free"] v["free"] = cur["memory_free_bytes"] v["avail"] = cur["memory_available_bytes"] v["waste"] = cur["abd_chunk_waste_size"] def main(): global sint global count global hdr_intr i = 0 count_flag = 0 init() if count > 0: count_flag = 1 signal(SIGINT, SIG_DFL) signal(SIGWINCH, resize_handler) while True: if i == 0: print_header() snap_stats() calculate() print_values() if count_flag == 1: if count <= 1: break count -= 1 i = 0 if i >= hdr_intr else i + 1 time.sleep(sint) if out: out.close() if __name__ == '__main__': main() diff --git a/sys/contrib/openzfs/cmd/zfs/zfs_main.c b/sys/contrib/openzfs/cmd/zfs/zfs_main.c index 27665fd2de71..3fc2d2810495 100644 --- a/sys/contrib/openzfs/cmd/zfs/zfs_main.c +++ b/sys/contrib/openzfs/cmd/zfs/zfs_main.c @@ -1,9048 +1,9058 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2020 by Delphix. All rights reserved. * Copyright 2012 Milan Jurik. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. * Copyright (c) 2013 Steven Hartland. All rights reserved. * Copyright 2016 Igor Kozhukhov . * Copyright 2016 Nexenta Systems, Inc. * Copyright (c) 2019 Datto Inc. * Copyright (c) 2019, loli10K * Copyright 2019 Joyent, Inc. * Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef HAVE_IDMAP #include #include #endif /* HAVE_IDMAP */ #include "zfs_iter.h" #include "zfs_util.h" #include "zfs_comutil.h" #include "zfs_projectutil.h" libzfs_handle_t *g_zfs; static char history_str[HIS_MAX_RECORD_LEN]; static boolean_t log_history = B_TRUE; static int zfs_do_clone(int argc, char **argv); static int zfs_do_create(int argc, char **argv); static int zfs_do_destroy(int argc, char **argv); static int zfs_do_get(int argc, char **argv); static int zfs_do_inherit(int argc, char **argv); static int zfs_do_list(int argc, char **argv); static int zfs_do_mount(int argc, char **argv); static int zfs_do_rename(int argc, char **argv); static int zfs_do_rollback(int argc, char **argv); static int zfs_do_set(int argc, char **argv); static int zfs_do_upgrade(int argc, char **argv); static int zfs_do_snapshot(int argc, char **argv); static int zfs_do_unmount(int argc, char **argv); static int zfs_do_share(int argc, char **argv); static int zfs_do_unshare(int argc, char **argv); static int zfs_do_send(int argc, char **argv); static int zfs_do_receive(int argc, char **argv); static int zfs_do_promote(int argc, char **argv); static int zfs_do_userspace(int argc, char **argv); static int zfs_do_allow(int argc, char **argv); static int zfs_do_unallow(int argc, char **argv); static int zfs_do_hold(int argc, char **argv); static int zfs_do_holds(int argc, char **argv); static int zfs_do_release(int argc, char **argv); static int zfs_do_diff(int argc, char **argv); static int zfs_do_bookmark(int argc, char **argv); static int zfs_do_channel_program(int argc, char **argv); static int zfs_do_load_key(int argc, char **argv); static int zfs_do_unload_key(int argc, char **argv); static int zfs_do_change_key(int argc, char **argv); static int zfs_do_project(int argc, char **argv); static int zfs_do_version(int argc, char **argv); static int zfs_do_redact(int argc, char **argv); static int zfs_do_wait(int argc, char **argv); #ifdef __FreeBSD__ static int zfs_do_jail(int argc, char **argv); static int zfs_do_unjail(int argc, char **argv); #endif #ifdef __linux__ static int zfs_do_zone(int argc, char **argv); static int zfs_do_unzone(int argc, char **argv); #endif static int zfs_do_help(int argc, char **argv); /* * Enable a reasonable set of defaults for libumem debugging on DEBUG builds. */ #ifdef DEBUG const char * _umem_debug_init(void) { return ("default,verbose"); /* $UMEM_DEBUG setting */ } const char * _umem_logging_init(void) { return ("fail,contents"); /* $UMEM_LOGGING setting */ } #endif typedef enum { HELP_CLONE, HELP_CREATE, HELP_DESTROY, HELP_GET, HELP_INHERIT, HELP_UPGRADE, HELP_LIST, HELP_MOUNT, HELP_PROMOTE, HELP_RECEIVE, HELP_RENAME, HELP_ROLLBACK, HELP_SEND, HELP_SET, HELP_SHARE, HELP_SNAPSHOT, HELP_UNMOUNT, HELP_UNSHARE, HELP_ALLOW, HELP_UNALLOW, HELP_USERSPACE, HELP_GROUPSPACE, HELP_PROJECTSPACE, HELP_PROJECT, HELP_HOLD, HELP_HOLDS, HELP_RELEASE, HELP_DIFF, HELP_BOOKMARK, HELP_CHANNEL_PROGRAM, HELP_LOAD_KEY, HELP_UNLOAD_KEY, HELP_CHANGE_KEY, HELP_VERSION, HELP_REDACT, HELP_JAIL, HELP_UNJAIL, HELP_WAIT, HELP_ZONE, HELP_UNZONE, } zfs_help_t; typedef struct zfs_command { const char *name; int (*func)(int argc, char **argv); zfs_help_t usage; } zfs_command_t; /* * Master command table. Each ZFS command has a name, associated function, and * usage message. The usage messages need to be internationalized, so we have * to have a function to return the usage message based on a command index. * * These commands are organized according to how they are displayed in the usage * message. An empty command (one with a NULL name) indicates an empty line in * the generic usage message. */ static zfs_command_t command_table[] = { { "version", zfs_do_version, HELP_VERSION }, { NULL }, { "create", zfs_do_create, HELP_CREATE }, { "destroy", zfs_do_destroy, HELP_DESTROY }, { NULL }, { "snapshot", zfs_do_snapshot, HELP_SNAPSHOT }, { "rollback", zfs_do_rollback, HELP_ROLLBACK }, { "clone", zfs_do_clone, HELP_CLONE }, { "promote", zfs_do_promote, HELP_PROMOTE }, { "rename", zfs_do_rename, HELP_RENAME }, { "bookmark", zfs_do_bookmark, HELP_BOOKMARK }, { "program", zfs_do_channel_program, HELP_CHANNEL_PROGRAM }, { NULL }, { "list", zfs_do_list, HELP_LIST }, { NULL }, { "set", zfs_do_set, HELP_SET }, { "get", zfs_do_get, HELP_GET }, { "inherit", zfs_do_inherit, HELP_INHERIT }, { "upgrade", zfs_do_upgrade, HELP_UPGRADE }, { NULL }, { "userspace", zfs_do_userspace, HELP_USERSPACE }, { "groupspace", zfs_do_userspace, HELP_GROUPSPACE }, { "projectspace", zfs_do_userspace, HELP_PROJECTSPACE }, { NULL }, { "project", zfs_do_project, HELP_PROJECT }, { NULL }, { "mount", zfs_do_mount, HELP_MOUNT }, { "unmount", zfs_do_unmount, HELP_UNMOUNT }, { "share", zfs_do_share, HELP_SHARE }, { "unshare", zfs_do_unshare, HELP_UNSHARE }, { NULL }, { "send", zfs_do_send, HELP_SEND }, { "receive", zfs_do_receive, HELP_RECEIVE }, { NULL }, { "allow", zfs_do_allow, HELP_ALLOW }, { NULL }, { "unallow", zfs_do_unallow, HELP_UNALLOW }, { NULL }, { "hold", zfs_do_hold, HELP_HOLD }, { "holds", zfs_do_holds, HELP_HOLDS }, { "release", zfs_do_release, HELP_RELEASE }, { "diff", zfs_do_diff, HELP_DIFF }, { "load-key", zfs_do_load_key, HELP_LOAD_KEY }, { "unload-key", zfs_do_unload_key, HELP_UNLOAD_KEY }, { "change-key", zfs_do_change_key, HELP_CHANGE_KEY }, { "redact", zfs_do_redact, HELP_REDACT }, { "wait", zfs_do_wait, HELP_WAIT }, #ifdef __FreeBSD__ { "jail", zfs_do_jail, HELP_JAIL }, { "unjail", zfs_do_unjail, HELP_UNJAIL }, #endif #ifdef __linux__ { "zone", zfs_do_zone, HELP_ZONE }, { "unzone", zfs_do_unzone, HELP_UNZONE }, #endif }; #define NCOMMAND (sizeof (command_table) / sizeof (command_table[0])) zfs_command_t *current_command; static const char * get_usage(zfs_help_t idx) { switch (idx) { case HELP_CLONE: return (gettext("\tclone [-p] [-o property=value] ... " " \n")); case HELP_CREATE: return (gettext("\tcreate [-Pnpuv] [-o property=value] ... " "\n" "\tcreate [-Pnpsv] [-b blocksize] [-o property=value] ... " "-V \n")); case HELP_DESTROY: return (gettext("\tdestroy [-fnpRrv] \n" "\tdestroy [-dnpRrv] " "@[%][,...]\n" "\tdestroy #\n")); case HELP_GET: return (gettext("\tget [-rHp] [-d max] " "[-o \"all\" | field[,...]]\n" "\t [-t type[,...]] [-s source[,...]]\n" "\t <\"all\" | property[,...]> " "[filesystem|volume|snapshot|bookmark] ...\n")); case HELP_INHERIT: return (gettext("\tinherit [-rS] " " ...\n")); case HELP_UPGRADE: return (gettext("\tupgrade [-v]\n" "\tupgrade [-r] [-V version] <-a | filesystem ...>\n")); case HELP_LIST: return (gettext("\tlist [-Hp] [-r|-d max] [-o property[,...]] " "[-s property]...\n\t [-S property]... [-t type[,...]] " "[filesystem|volume|snapshot] ...\n")); case HELP_MOUNT: return (gettext("\tmount\n" "\tmount [-flvO] [-o opts] <-a|-R filesystem|" "filesystem>\n")); case HELP_PROMOTE: return (gettext("\tpromote \n")); case HELP_RECEIVE: return (gettext("\treceive [-vMnsFhu] " "[-o =] ... [-x ] ...\n" "\t \n" "\treceive [-vMnsFhu] [-o =] ... " "[-x ] ... \n" "\t [-d | -e] \n" "\treceive -A \n")); case HELP_RENAME: return (gettext("\trename [-f] " "\n" "\trename -p [-f] \n" "\trename -u [-f] \n" "\trename -r \n")); case HELP_ROLLBACK: return (gettext("\trollback [-rRf] \n")); case HELP_SEND: return (gettext("\tsend [-DLPbcehnpsVvw] " "[-i|-I snapshot]\n" "\t [-R [-X dataset[,dataset]...]] \n" "\tsend [-DnVvPLecw] [-i snapshot|bookmark] " "\n" "\tsend [-DnPpVvLec] [-i bookmark|snapshot] " "--redact \n" "\tsend [-nVvPe] -t \n" "\tsend [-PnVv] --saved filesystem\n")); case HELP_SET: return (gettext("\tset [-u] ... " " ...\n")); case HELP_SHARE: return (gettext("\tshare [-l] <-a [nfs|smb] | filesystem>\n")); case HELP_SNAPSHOT: return (gettext("\tsnapshot [-r] [-o property=value] ... " "@ ...\n")); case HELP_UNMOUNT: return (gettext("\tunmount [-fu] " "<-a | filesystem|mountpoint>\n")); case HELP_UNSHARE: return (gettext("\tunshare " "<-a [nfs|smb] | filesystem|mountpoint>\n")); case HELP_ALLOW: return (gettext("\tallow \n" "\tallow [-ldug] " "<\"everyone\"|user|group>[,...] [,...]\n" "\t \n" "\tallow [-ld] -e [,...] " "\n" "\tallow -c [,...] \n" "\tallow -s @setname [,...] " "\n")); case HELP_UNALLOW: return (gettext("\tunallow [-rldug] " "<\"everyone\"|user|group>[,...]\n" "\t [[,...]] \n" "\tunallow [-rld] -e [[,...]] " "\n" "\tunallow [-r] -c [[,...]] " "\n" "\tunallow [-r] -s @setname [[,...]] " "\n")); case HELP_USERSPACE: return (gettext("\tuserspace [-Hinp] [-o field[,...]] " "[-s field] ...\n" "\t [-S field] ... [-t type[,...]] " "\n")); case HELP_GROUPSPACE: return (gettext("\tgroupspace [-Hinp] [-o field[,...]] " "[-s field] ...\n" "\t [-S field] ... [-t type[,...]] " "\n")); case HELP_PROJECTSPACE: return (gettext("\tprojectspace [-Hp] [-o field[,...]] " "[-s field] ... \n" "\t [-S field] ... \n")); case HELP_PROJECT: return (gettext("\tproject [-d|-r] \n" "\tproject -c [-0] [-d|-r] [-p id] \n" "\tproject -C [-k] [-r] \n" "\tproject [-p id] [-r] [-s] \n")); case HELP_HOLD: return (gettext("\thold [-r] ...\n")); case HELP_HOLDS: return (gettext("\tholds [-rHp] ...\n")); case HELP_RELEASE: return (gettext("\trelease [-r] ...\n")); case HELP_DIFF: return (gettext("\tdiff [-FHth] " "[snapshot|filesystem]\n")); case HELP_BOOKMARK: return (gettext("\tbookmark " "\n")); case HELP_CHANNEL_PROGRAM: return (gettext("\tprogram [-jn] [-t ] " "[-m ]\n" "\t [lua args...]\n")); case HELP_LOAD_KEY: return (gettext("\tload-key [-rn] [-L ] " "<-a | filesystem|volume>\n")); case HELP_UNLOAD_KEY: return (gettext("\tunload-key [-r] " "<-a | filesystem|volume>\n")); case HELP_CHANGE_KEY: return (gettext("\tchange-key [-l] [-o keyformat=]\n" "\t [-o keylocation=] [-o pbkdf2iters=]\n" "\t \n" "\tchange-key -i [-l] \n")); case HELP_VERSION: return (gettext("\tversion\n")); case HELP_REDACT: return (gettext("\tredact " " ...\n")); case HELP_JAIL: return (gettext("\tjail \n")); case HELP_UNJAIL: return (gettext("\tunjail \n")); case HELP_WAIT: return (gettext("\twait [-t ] \n")); case HELP_ZONE: return (gettext("\tzone \n")); case HELP_UNZONE: return (gettext("\tunzone \n")); default: __builtin_unreachable(); } } void nomem(void) { (void) fprintf(stderr, gettext("internal error: out of memory\n")); exit(1); } /* * Utility function to guarantee malloc() success. */ void * safe_malloc(size_t size) { void *data; if ((data = calloc(1, size)) == NULL) nomem(); return (data); } static void * safe_realloc(void *data, size_t size) { void *newp; if ((newp = realloc(data, size)) == NULL) { free(data); nomem(); } return (newp); } static char * safe_strdup(const char *str) { char *dupstr = strdup(str); if (dupstr == NULL) nomem(); return (dupstr); } /* * Callback routine that will print out information for each of * the properties. */ static int usage_prop_cb(int prop, void *cb) { FILE *fp = cb; (void) fprintf(fp, "\t%-15s ", zfs_prop_to_name(prop)); if (zfs_prop_readonly(prop)) (void) fprintf(fp, " NO "); else (void) fprintf(fp, "YES "); if (zfs_prop_inheritable(prop)) (void) fprintf(fp, " YES "); else (void) fprintf(fp, " NO "); (void) fprintf(fp, "%s\n", zfs_prop_values(prop) ?: "-"); return (ZPROP_CONT); } /* * Display usage message. If we're inside a command, display only the usage for * that command. Otherwise, iterate over the entire command table and display * a complete usage message. */ static __attribute__((noreturn)) void usage(boolean_t requested) { int i; boolean_t show_properties = B_FALSE; FILE *fp = requested ? stdout : stderr; if (current_command == NULL) { (void) fprintf(fp, gettext("usage: zfs command args ...\n")); (void) fprintf(fp, gettext("where 'command' is one of the following:\n\n")); for (i = 0; i < NCOMMAND; i++) { if (command_table[i].name == NULL) (void) fprintf(fp, "\n"); else (void) fprintf(fp, "%s", get_usage(command_table[i].usage)); } (void) fprintf(fp, gettext("\nEach dataset is of the form: " "pool/[dataset/]*dataset[@name]\n")); } else { (void) fprintf(fp, gettext("usage:\n")); (void) fprintf(fp, "%s", get_usage(current_command->usage)); } if (current_command != NULL && (strcmp(current_command->name, "set") == 0 || strcmp(current_command->name, "get") == 0 || strcmp(current_command->name, "inherit") == 0 || strcmp(current_command->name, "list") == 0)) show_properties = B_TRUE; if (show_properties) { (void) fprintf(fp, "%s", gettext("\nThe following properties are supported:\n")); (void) fprintf(fp, "\n\t%-14s %s %s %s\n\n", "PROPERTY", "EDIT", "INHERIT", "VALUES"); /* Iterate over all properties */ (void) zprop_iter(usage_prop_cb, fp, B_FALSE, B_TRUE, ZFS_TYPE_DATASET); (void) fprintf(fp, "\t%-15s ", "userused@..."); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "groupused@..."); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "projectused@..."); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "userobjused@..."); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "groupobjused@..."); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "projectobjused@..."); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "userquota@..."); (void) fprintf(fp, "YES NO | none\n"); (void) fprintf(fp, "\t%-15s ", "groupquota@..."); (void) fprintf(fp, "YES NO | none\n"); (void) fprintf(fp, "\t%-15s ", "projectquota@..."); (void) fprintf(fp, "YES NO | none\n"); (void) fprintf(fp, "\t%-15s ", "userobjquota@..."); (void) fprintf(fp, "YES NO | none\n"); (void) fprintf(fp, "\t%-15s ", "groupobjquota@..."); (void) fprintf(fp, "YES NO | none\n"); (void) fprintf(fp, "\t%-15s ", "projectobjquota@..."); (void) fprintf(fp, "YES NO | none\n"); (void) fprintf(fp, "\t%-15s ", "written@"); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, "\t%-15s ", "written#"); (void) fprintf(fp, " NO NO \n"); (void) fprintf(fp, gettext("\nSizes are specified in bytes " "with standard units such as K, M, G, etc.\n")); (void) fprintf(fp, "%s", gettext("\nUser-defined properties " "can be specified by using a name containing a colon " "(:).\n")); (void) fprintf(fp, gettext("\nThe {user|group|project}" "[obj]{used|quota}@ properties must be appended with\n" "a user|group|project specifier of one of these forms:\n" " POSIX name (eg: \"matt\")\n" " POSIX id (eg: \"126829\")\n" " SMB name@domain (eg: \"matt@sun\")\n" " SMB SID (eg: \"S-1-234-567-89\")\n")); } else { (void) fprintf(fp, gettext("\nFor the property list, run: %s\n"), "zfs set|get"); (void) fprintf(fp, gettext("\nFor the delegated permission list, run: %s\n"), "zfs allow|unallow"); (void) fprintf(fp, gettext("\nFor further help on a command or topic, " "run: %s\n"), "zfs help []"); } /* * See comments at end of main(). */ if (getenv("ZFS_ABORT") != NULL) { (void) printf("dumping core by request\n"); abort(); } exit(requested ? 0 : 2); } /* * Take a property=value argument string and add it to the given nvlist. * Modifies the argument inplace. */ static boolean_t parseprop(nvlist_t *props, char *propname) { char *propval; if ((propval = strchr(propname, '=')) == NULL) { (void) fprintf(stderr, gettext("missing " "'=' for property=value argument\n")); return (B_FALSE); } *propval = '\0'; propval++; if (nvlist_exists(props, propname)) { (void) fprintf(stderr, gettext("property '%s' " "specified multiple times\n"), propname); return (B_FALSE); } if (nvlist_add_string(props, propname, propval) != 0) nomem(); return (B_TRUE); } /* * Take a property name argument and add it to the given nvlist. * Modifies the argument inplace. */ static boolean_t parsepropname(nvlist_t *props, char *propname) { if (strchr(propname, '=') != NULL) { (void) fprintf(stderr, gettext("invalid character " "'=' in property argument\n")); return (B_FALSE); } if (nvlist_exists(props, propname)) { (void) fprintf(stderr, gettext("property '%s' " "specified multiple times\n"), propname); return (B_FALSE); } if (nvlist_add_boolean(props, propname) != 0) nomem(); return (B_TRUE); } static int parse_depth(char *opt, int *flags) { char *tmp; int depth; depth = (int)strtol(opt, &tmp, 0); if (*tmp) { (void) fprintf(stderr, gettext("%s is not an integer\n"), optarg); usage(B_FALSE); } if (depth < 0) { (void) fprintf(stderr, gettext("Depth can not be negative.\n")); usage(B_FALSE); } *flags |= (ZFS_ITER_DEPTH_LIMIT|ZFS_ITER_RECURSE); return (depth); } #define PROGRESS_DELAY 2 /* seconds */ static const char *pt_reverse = "\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b"; static time_t pt_begin; static char *pt_header = NULL; static boolean_t pt_shown; static void start_progress_timer(void) { pt_begin = time(NULL) + PROGRESS_DELAY; pt_shown = B_FALSE; } static void set_progress_header(const char *header) { assert(pt_header == NULL); pt_header = safe_strdup(header); if (pt_shown) { (void) printf("%s: ", header); (void) fflush(stdout); } } static void update_progress(const char *update) { if (!pt_shown && time(NULL) > pt_begin) { int len = strlen(update); (void) printf("%s: %s%*.*s", pt_header, update, len, len, pt_reverse); (void) fflush(stdout); pt_shown = B_TRUE; } else if (pt_shown) { int len = strlen(update); (void) printf("%s%*.*s", update, len, len, pt_reverse); (void) fflush(stdout); } } static void finish_progress(const char *done) { if (pt_shown) { (void) puts(done); (void) fflush(stdout); } free(pt_header); pt_header = NULL; } static int zfs_mount_and_share(libzfs_handle_t *hdl, const char *dataset, zfs_type_t type) { zfs_handle_t *zhp = NULL; int ret = 0; zhp = zfs_open(hdl, dataset, type); if (zhp == NULL) return (1); /* * Volumes may neither be mounted or shared. Potentially in the * future filesystems detected on these volumes could be mounted. */ if (zfs_get_type(zhp) == ZFS_TYPE_VOLUME) { zfs_close(zhp); return (0); } /* * Mount and/or share the new filesystem as appropriate. We provide a * verbose error message to let the user know that their filesystem was * in fact created, even if we failed to mount or share it. * * If the user doesn't want the dataset automatically mounted, then * skip the mount/share step */ if (zfs_prop_valid_for_type(ZFS_PROP_CANMOUNT, type, B_FALSE) && zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_ON) { if (zfs_mount_delegation_check()) { (void) fprintf(stderr, gettext("filesystem " "successfully created, but it may only be " "mounted by root\n")); ret = 1; } else if (zfs_mount(zhp, NULL, 0) != 0) { (void) fprintf(stderr, gettext("filesystem " "successfully created, but not mounted\n")); ret = 1; } else if (zfs_share(zhp, NULL) != 0) { (void) fprintf(stderr, gettext("filesystem " "successfully created, but not shared\n")); ret = 1; } zfs_commit_shares(NULL); } zfs_close(zhp); return (ret); } /* * zfs clone [-p] [-o prop=value] ... * * Given an existing dataset, create a writable copy whose initial contents * are the same as the source. The newly created dataset maintains a * dependency on the original; the original cannot be destroyed so long as * the clone exists. * * The '-p' flag creates all the non-existing ancestors of the target first. */ static int zfs_do_clone(int argc, char **argv) { zfs_handle_t *zhp = NULL; boolean_t parents = B_FALSE; nvlist_t *props; int ret = 0; int c; if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) nomem(); /* check options */ while ((c = getopt(argc, argv, "o:p")) != -1) { switch (c) { case 'o': if (!parseprop(props, optarg)) { nvlist_free(props); return (1); } break; case 'p': parents = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); goto usage; } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing source dataset " "argument\n")); goto usage; } if (argc < 2) { (void) fprintf(stderr, gettext("missing target dataset " "argument\n")); goto usage; } if (argc > 2) { (void) fprintf(stderr, gettext("too many arguments\n")); goto usage; } /* open the source dataset */ if ((zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_SNAPSHOT)) == NULL) { nvlist_free(props); return (1); } if (parents && zfs_name_valid(argv[1], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) { /* * Now create the ancestors of the target dataset. If the * target already exists and '-p' option was used we should not * complain. */ if (zfs_dataset_exists(g_zfs, argv[1], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) { zfs_close(zhp); nvlist_free(props); return (0); } if (zfs_create_ancestors(g_zfs, argv[1]) != 0) { zfs_close(zhp); nvlist_free(props); return (1); } } /* pass to libzfs */ ret = zfs_clone(zhp, argv[1], props); /* create the mountpoint if necessary */ if (ret == 0) { if (log_history) { (void) zpool_log_history(g_zfs, history_str); log_history = B_FALSE; } ret = zfs_mount_and_share(g_zfs, argv[1], ZFS_TYPE_DATASET); } zfs_close(zhp); nvlist_free(props); return (!!ret); usage: ASSERT3P(zhp, ==, NULL); nvlist_free(props); usage(B_FALSE); return (-1); } /* * Return a default volblocksize for the pool which always uses more than * half of the data sectors. This primarily applies to dRAID which always * writes full stripe widths. */ static uint64_t default_volblocksize(zpool_handle_t *zhp, nvlist_t *props) { uint64_t volblocksize, asize = SPA_MINBLOCKSIZE; nvlist_t *tree, **vdevs; uint_t nvdevs; nvlist_t *config = zpool_get_config(zhp, NULL); if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 || nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &vdevs, &nvdevs) != 0) { return (ZVOL_DEFAULT_BLOCKSIZE); } for (int i = 0; i < nvdevs; i++) { nvlist_t *nv = vdevs[i]; uint64_t ashift, ndata, nparity; if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &ashift) != 0) continue; if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DRAID_NDATA, &ndata) == 0) { /* dRAID minimum allocation width */ asize = MAX(asize, ndata * (1ULL << ashift)); } else if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, &nparity) == 0) { /* raidz minimum allocation width */ if (nparity == 1) asize = MAX(asize, 2 * (1ULL << ashift)); else asize = MAX(asize, 4 * (1ULL << ashift)); } else { /* mirror or (non-redundant) leaf vdev */ asize = MAX(asize, 1ULL << ashift); } } /* * Calculate the target volblocksize such that more than half * of the asize is used. The following table is for 4k sectors. * * n asize blksz used | n asize blksz used * -------------------------+--------------------------------- * 1 4,096 8,192 100% | 9 36,864 32,768 88% * 2 8,192 8,192 100% | 10 40,960 32,768 80% * 3 12,288 8,192 66% | 11 45,056 32,768 72% * 4 16,384 16,384 100% | 12 49,152 32,768 66% * 5 20,480 16,384 80% | 13 53,248 32,768 61% * 6 24,576 16,384 66% | 14 57,344 32,768 57% * 7 28,672 16,384 57% | 15 61,440 32,768 53% * 8 32,768 32,768 100% | 16 65,536 65,636 100% * * This is primarily a concern for dRAID which always allocates * a full stripe width. For dRAID the default stripe width is * n=8 in which case the volblocksize is set to 32k. Ignoring * compression there are no unused sectors. This same reasoning * applies to raidz[2,3] so target 4 sectors to minimize waste. */ uint64_t tgt_volblocksize = ZVOL_DEFAULT_BLOCKSIZE; while (tgt_volblocksize * 2 <= asize) tgt_volblocksize *= 2; const char *prop = zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE); if (nvlist_lookup_uint64(props, prop, &volblocksize) == 0) { /* Issue a warning when a non-optimal size is requested. */ if (volblocksize < ZVOL_DEFAULT_BLOCKSIZE) { (void) fprintf(stderr, gettext("Warning: " "volblocksize (%llu) is less than the default " "minimum block size (%llu).\nTo reduce wasted " "space a volblocksize of %llu is recommended.\n"), (u_longlong_t)volblocksize, (u_longlong_t)ZVOL_DEFAULT_BLOCKSIZE, (u_longlong_t)tgt_volblocksize); } else if (volblocksize < tgt_volblocksize) { (void) fprintf(stderr, gettext("Warning: " "volblocksize (%llu) is much less than the " "minimum allocation\nunit (%llu), which wastes " "at least %llu%% of space. To reduce wasted " "space,\nuse a larger volblocksize (%llu is " "recommended), fewer dRAID data disks\n" "per group, or smaller sector size (ashift).\n"), (u_longlong_t)volblocksize, (u_longlong_t)asize, (u_longlong_t)((100 * (asize - volblocksize)) / asize), (u_longlong_t)tgt_volblocksize); } } else { volblocksize = tgt_volblocksize; fnvlist_add_uint64(props, prop, volblocksize); } return (volblocksize); } /* * zfs create [-Pnpv] [-o prop=value] ... fs * zfs create [-Pnpsv] [-b blocksize] [-o prop=value] ... -V vol size * * Create a new dataset. This command can be used to create filesystems * and volumes. Snapshot creation is handled by 'zfs snapshot'. * For volumes, the user must specify a size to be used. * * The '-s' flag applies only to volumes, and indicates that we should not try * to set the reservation for this volume. By default we set a reservation * equal to the size for any volume. For pools with SPA_VERSION >= * SPA_VERSION_REFRESERVATION, we set a refreservation instead. * * The '-p' flag creates all the non-existing ancestors of the target first. * * The '-n' flag is no-op (dry run) mode. This will perform a user-space sanity * check of arguments and properties, but does not check for permissions, * available space, etc. * * The '-u' flag prevents the newly created file system from being mounted. * * The '-v' flag is for verbose output. * * The '-P' flag is used for parseable output. It implies '-v'. */ static int zfs_do_create(int argc, char **argv) { zfs_type_t type = ZFS_TYPE_FILESYSTEM; zpool_handle_t *zpool_handle = NULL; nvlist_t *real_props = NULL; uint64_t volsize = 0; int c; boolean_t noreserve = B_FALSE; boolean_t bflag = B_FALSE; boolean_t parents = B_FALSE; boolean_t dryrun = B_FALSE; boolean_t nomount = B_FALSE; boolean_t verbose = B_FALSE; boolean_t parseable = B_FALSE; int ret = 1; nvlist_t *props; uint64_t intval; const char *strval; if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) nomem(); /* check options */ while ((c = getopt(argc, argv, ":PV:b:nso:puv")) != -1) { switch (c) { case 'V': type = ZFS_TYPE_VOLUME; if (zfs_nicestrtonum(g_zfs, optarg, &intval) != 0) { (void) fprintf(stderr, gettext("bad volume " "size '%s': %s\n"), optarg, libzfs_error_description(g_zfs)); goto error; } if (nvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLSIZE), intval) != 0) nomem(); volsize = intval; break; case 'P': verbose = B_TRUE; parseable = B_TRUE; break; case 'p': parents = B_TRUE; break; case 'b': bflag = B_TRUE; if (zfs_nicestrtonum(g_zfs, optarg, &intval) != 0) { (void) fprintf(stderr, gettext("bad volume " "block size '%s': %s\n"), optarg, libzfs_error_description(g_zfs)); goto error; } if (nvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), intval) != 0) nomem(); break; case 'n': dryrun = B_TRUE; break; case 'o': if (!parseprop(props, optarg)) goto error; break; case 's': noreserve = B_TRUE; break; case 'u': nomount = B_TRUE; break; case 'v': verbose = B_TRUE; break; case ':': (void) fprintf(stderr, gettext("missing size " "argument\n")); goto badusage; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); goto badusage; } } if ((bflag || noreserve) && type != ZFS_TYPE_VOLUME) { (void) fprintf(stderr, gettext("'-s' and '-b' can only be " "used when creating a volume\n")); goto badusage; } if (nomount && type != ZFS_TYPE_FILESYSTEM) { (void) fprintf(stderr, gettext("'-u' can only be " "used when creating a filesystem\n")); goto badusage; } argc -= optind; argv += optind; /* check number of arguments */ if (argc == 0) { (void) fprintf(stderr, gettext("missing %s argument\n"), zfs_type_to_name(type)); goto badusage; } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); goto badusage; } if (dryrun || type == ZFS_TYPE_VOLUME) { char msg[ZFS_MAX_DATASET_NAME_LEN * 2]; char *p; if ((p = strchr(argv[0], '/')) != NULL) *p = '\0'; zpool_handle = zpool_open(g_zfs, argv[0]); if (p != NULL) *p = '/'; if (zpool_handle == NULL) goto error; (void) snprintf(msg, sizeof (msg), dryrun ? gettext("cannot verify '%s'") : gettext("cannot create '%s'"), argv[0]); if (props && (real_props = zfs_valid_proplist(g_zfs, type, props, 0, NULL, zpool_handle, B_TRUE, msg)) == NULL) { zpool_close(zpool_handle); goto error; } } if (type == ZFS_TYPE_VOLUME) { const char *prop = zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE); uint64_t volblocksize = default_volblocksize(zpool_handle, real_props); if (volblocksize != ZVOL_DEFAULT_BLOCKSIZE && nvlist_lookup_string(props, prop, &strval) != 0) { char *tmp; if (asprintf(&tmp, "%llu", (u_longlong_t)volblocksize) == -1) nomem(); nvlist_add_string(props, prop, tmp); free(tmp); } /* * If volsize is not a multiple of volblocksize, round it * up to the nearest multiple of the volblocksize. */ if (volsize % volblocksize) { volsize = P2ROUNDUP_TYPED(volsize, volblocksize, uint64_t); if (nvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLSIZE), volsize) != 0) { nvlist_free(props); nomem(); } } } if (type == ZFS_TYPE_VOLUME && !noreserve) { uint64_t spa_version; zfs_prop_t resv_prop; spa_version = zpool_get_prop_int(zpool_handle, ZPOOL_PROP_VERSION, NULL); if (spa_version >= SPA_VERSION_REFRESERVATION) resv_prop = ZFS_PROP_REFRESERVATION; else resv_prop = ZFS_PROP_RESERVATION; volsize = zvol_volsize_to_reservation(zpool_handle, volsize, real_props); if (nvlist_lookup_string(props, zfs_prop_to_name(resv_prop), &strval) != 0) { if (nvlist_add_uint64(props, zfs_prop_to_name(resv_prop), volsize) != 0) { nvlist_free(props); nomem(); } } } if (zpool_handle != NULL) { zpool_close(zpool_handle); nvlist_free(real_props); } if (parents && zfs_name_valid(argv[0], type)) { /* * Now create the ancestors of target dataset. If the target * already exists and '-p' option was used we should not * complain. */ if (zfs_dataset_exists(g_zfs, argv[0], type)) { ret = 0; goto error; } if (verbose) { (void) printf(parseable ? "create_ancestors\t%s\n" : dryrun ? "would create ancestors of %s\n" : "create ancestors of %s\n", argv[0]); } if (!dryrun) { if (zfs_create_ancestors(g_zfs, argv[0]) != 0) { goto error; } } } if (verbose) { nvpair_t *nvp = NULL; (void) printf(parseable ? "create\t%s\n" : dryrun ? "would create %s\n" : "create %s\n", argv[0]); while ((nvp = nvlist_next_nvpair(props, nvp)) != NULL) { uint64_t uval; const char *sval; switch (nvpair_type(nvp)) { case DATA_TYPE_UINT64: VERIFY0(nvpair_value_uint64(nvp, &uval)); (void) printf(parseable ? "property\t%s\t%llu\n" : "\t%s=%llu\n", nvpair_name(nvp), (u_longlong_t)uval); break; case DATA_TYPE_STRING: VERIFY0(nvpair_value_string(nvp, &sval)); (void) printf(parseable ? "property\t%s\t%s\n" : "\t%s=%s\n", nvpair_name(nvp), sval); break; default: (void) fprintf(stderr, "property '%s' " "has illegal type %d\n", nvpair_name(nvp), nvpair_type(nvp)); abort(); } } } if (dryrun) { ret = 0; goto error; } /* pass to libzfs */ if (zfs_create(g_zfs, argv[0], type, props) != 0) goto error; if (log_history) { (void) zpool_log_history(g_zfs, history_str); log_history = B_FALSE; } if (nomount) { ret = 0; goto error; } ret = zfs_mount_and_share(g_zfs, argv[0], ZFS_TYPE_DATASET); error: nvlist_free(props); return (ret); badusage: nvlist_free(props); usage(B_FALSE); return (2); } /* * zfs destroy [-rRf] * zfs destroy [-rRd] * * -r Recursively destroy all children * -R Recursively destroy all dependents, including clones * -f Force unmounting of any dependents * -d If we can't destroy now, mark for deferred destruction * * Destroys the given dataset. By default, it will unmount any filesystems, * and refuse to destroy a dataset that has any dependents. A dependent can * either be a child, or a clone of a child. */ typedef struct destroy_cbdata { boolean_t cb_first; boolean_t cb_force; boolean_t cb_recurse; boolean_t cb_error; boolean_t cb_doclones; zfs_handle_t *cb_target; boolean_t cb_defer_destroy; boolean_t cb_verbose; boolean_t cb_parsable; boolean_t cb_dryrun; nvlist_t *cb_nvl; nvlist_t *cb_batchedsnaps; /* first snap in contiguous run */ char *cb_firstsnap; /* previous snap in contiguous run */ char *cb_prevsnap; int64_t cb_snapused; char *cb_snapspec; char *cb_bookmark; uint64_t cb_snap_count; } destroy_cbdata_t; /* * Check for any dependents based on the '-r' or '-R' flags. */ static int destroy_check_dependent(zfs_handle_t *zhp, void *data) { destroy_cbdata_t *cbp = data; const char *tname = zfs_get_name(cbp->cb_target); const char *name = zfs_get_name(zhp); if (strncmp(tname, name, strlen(tname)) == 0 && (name[strlen(tname)] == '/' || name[strlen(tname)] == '@')) { /* * This is a direct descendant, not a clone somewhere else in * the hierarchy. */ if (cbp->cb_recurse) goto out; if (cbp->cb_first) { (void) fprintf(stderr, gettext("cannot destroy '%s': " "%s has children\n"), zfs_get_name(cbp->cb_target), zfs_type_to_name(zfs_get_type(cbp->cb_target))); (void) fprintf(stderr, gettext("use '-r' to destroy " "the following datasets:\n")); cbp->cb_first = B_FALSE; cbp->cb_error = B_TRUE; } (void) fprintf(stderr, "%s\n", zfs_get_name(zhp)); } else { /* * This is a clone. We only want to report this if the '-r' * wasn't specified, or the target is a snapshot. */ if (!cbp->cb_recurse && zfs_get_type(cbp->cb_target) != ZFS_TYPE_SNAPSHOT) goto out; if (cbp->cb_first) { (void) fprintf(stderr, gettext("cannot destroy '%s': " "%s has dependent clones\n"), zfs_get_name(cbp->cb_target), zfs_type_to_name(zfs_get_type(cbp->cb_target))); (void) fprintf(stderr, gettext("use '-R' to destroy " "the following datasets:\n")); cbp->cb_first = B_FALSE; cbp->cb_error = B_TRUE; cbp->cb_dryrun = B_TRUE; } (void) fprintf(stderr, "%s\n", zfs_get_name(zhp)); } out: zfs_close(zhp); return (0); } static int destroy_batched(destroy_cbdata_t *cb) { int error = zfs_destroy_snaps_nvl(g_zfs, cb->cb_batchedsnaps, B_FALSE); fnvlist_free(cb->cb_batchedsnaps); cb->cb_batchedsnaps = fnvlist_alloc(); return (error); } static int destroy_callback(zfs_handle_t *zhp, void *data) { destroy_cbdata_t *cb = data; const char *name = zfs_get_name(zhp); int error; if (cb->cb_verbose) { if (cb->cb_parsable) { (void) printf("destroy\t%s\n", name); } else if (cb->cb_dryrun) { (void) printf(gettext("would destroy %s\n"), name); } else { (void) printf(gettext("will destroy %s\n"), name); } } /* * Ignore pools (which we've already flagged as an error before getting * here). */ if (strchr(zfs_get_name(zhp), '/') == NULL && zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) { zfs_close(zhp); return (0); } if (cb->cb_dryrun) { zfs_close(zhp); return (0); } /* * We batch up all contiguous snapshots (even of different * filesystems) and destroy them with one ioctl. We can't * simply do all snap deletions and then all fs deletions, * because we must delete a clone before its origin. */ if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT) { cb->cb_snap_count++; fnvlist_add_boolean(cb->cb_batchedsnaps, name); if (cb->cb_snap_count % 10 == 0 && cb->cb_defer_destroy) { error = destroy_batched(cb); if (error != 0) { zfs_close(zhp); return (-1); } } } else { error = destroy_batched(cb); if (error != 0 || zfs_unmount(zhp, NULL, cb->cb_force ? MS_FORCE : 0) != 0 || zfs_destroy(zhp, cb->cb_defer_destroy) != 0) { zfs_close(zhp); /* * When performing a recursive destroy we ignore errors * so that the recursive destroy could continue * destroying past problem datasets */ if (cb->cb_recurse) { cb->cb_error = B_TRUE; return (0); } return (-1); } } zfs_close(zhp); return (0); } static int destroy_print_cb(zfs_handle_t *zhp, void *arg) { destroy_cbdata_t *cb = arg; const char *name = zfs_get_name(zhp); int err = 0; if (nvlist_exists(cb->cb_nvl, name)) { if (cb->cb_firstsnap == NULL) cb->cb_firstsnap = strdup(name); if (cb->cb_prevsnap != NULL) free(cb->cb_prevsnap); /* this snap continues the current range */ cb->cb_prevsnap = strdup(name); if (cb->cb_firstsnap == NULL || cb->cb_prevsnap == NULL) nomem(); if (cb->cb_verbose) { if (cb->cb_parsable) { (void) printf("destroy\t%s\n", name); } else if (cb->cb_dryrun) { (void) printf(gettext("would destroy %s\n"), name); } else { (void) printf(gettext("will destroy %s\n"), name); } } } else if (cb->cb_firstsnap != NULL) { /* end of this range */ uint64_t used = 0; err = lzc_snaprange_space(cb->cb_firstsnap, cb->cb_prevsnap, &used); cb->cb_snapused += used; free(cb->cb_firstsnap); cb->cb_firstsnap = NULL; free(cb->cb_prevsnap); cb->cb_prevsnap = NULL; } zfs_close(zhp); return (err); } static int destroy_print_snapshots(zfs_handle_t *fs_zhp, destroy_cbdata_t *cb) { int err; assert(cb->cb_firstsnap == NULL); assert(cb->cb_prevsnap == NULL); err = zfs_iter_snapshots_sorted_v2(fs_zhp, 0, destroy_print_cb, cb, 0, 0); if (cb->cb_firstsnap != NULL) { uint64_t used = 0; if (err == 0) { err = lzc_snaprange_space(cb->cb_firstsnap, cb->cb_prevsnap, &used); } cb->cb_snapused += used; free(cb->cb_firstsnap); cb->cb_firstsnap = NULL; free(cb->cb_prevsnap); cb->cb_prevsnap = NULL; } return (err); } static int snapshot_to_nvl_cb(zfs_handle_t *zhp, void *arg) { destroy_cbdata_t *cb = arg; int err = 0; /* Check for clones. */ if (!cb->cb_doclones && !cb->cb_defer_destroy) { cb->cb_target = zhp; cb->cb_first = B_TRUE; err = zfs_iter_dependents_v2(zhp, 0, B_TRUE, destroy_check_dependent, cb); } if (err == 0) { if (nvlist_add_boolean(cb->cb_nvl, zfs_get_name(zhp))) nomem(); } zfs_close(zhp); return (err); } static int gather_snapshots(zfs_handle_t *zhp, void *arg) { destroy_cbdata_t *cb = arg; int err = 0; err = zfs_iter_snapspec_v2(zhp, 0, cb->cb_snapspec, snapshot_to_nvl_cb, cb); if (err == ENOENT) err = 0; if (err != 0) goto out; if (cb->cb_verbose) { err = destroy_print_snapshots(zhp, cb); if (err != 0) goto out; } if (cb->cb_recurse) err = zfs_iter_filesystems_v2(zhp, 0, gather_snapshots, cb); out: zfs_close(zhp); return (err); } static int destroy_clones(destroy_cbdata_t *cb) { nvpair_t *pair; for (pair = nvlist_next_nvpair(cb->cb_nvl, NULL); pair != NULL; pair = nvlist_next_nvpair(cb->cb_nvl, pair)) { zfs_handle_t *zhp = zfs_open(g_zfs, nvpair_name(pair), ZFS_TYPE_SNAPSHOT); if (zhp != NULL) { boolean_t defer = cb->cb_defer_destroy; int err; /* * We can't defer destroy non-snapshots, so set it to * false while destroying the clones. */ cb->cb_defer_destroy = B_FALSE; err = zfs_iter_dependents_v2(zhp, 0, B_FALSE, destroy_callback, cb); cb->cb_defer_destroy = defer; zfs_close(zhp); if (err != 0) return (err); } } return (0); } static int zfs_do_destroy(int argc, char **argv) { destroy_cbdata_t cb = { 0 }; int rv = 0; int err = 0; int c; zfs_handle_t *zhp = NULL; char *at, *pound; zfs_type_t type = ZFS_TYPE_DATASET; /* check options */ while ((c = getopt(argc, argv, "vpndfrR")) != -1) { switch (c) { case 'v': cb.cb_verbose = B_TRUE; break; case 'p': cb.cb_verbose = B_TRUE; cb.cb_parsable = B_TRUE; break; case 'n': cb.cb_dryrun = B_TRUE; break; case 'd': cb.cb_defer_destroy = B_TRUE; type = ZFS_TYPE_SNAPSHOT; break; case 'f': cb.cb_force = B_TRUE; break; case 'r': cb.cb_recurse = B_TRUE; break; case 'R': cb.cb_recurse = B_TRUE; cb.cb_doclones = B_TRUE; break; case '?': default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (argc == 0) { (void) fprintf(stderr, gettext("missing dataset argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } at = strchr(argv[0], '@'); pound = strchr(argv[0], '#'); if (at != NULL) { /* Build the list of snaps to destroy in cb_nvl. */ cb.cb_nvl = fnvlist_alloc(); *at = '\0'; zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) { nvlist_free(cb.cb_nvl); return (1); } cb.cb_snapspec = at + 1; if (gather_snapshots(zfs_handle_dup(zhp), &cb) != 0 || cb.cb_error) { rv = 1; goto out; } if (nvlist_empty(cb.cb_nvl)) { (void) fprintf(stderr, gettext("could not find any " "snapshots to destroy; check snapshot names.\n")); rv = 1; goto out; } if (cb.cb_verbose) { char buf[16]; zfs_nicebytes(cb.cb_snapused, buf, sizeof (buf)); if (cb.cb_parsable) { (void) printf("reclaim\t%llu\n", (u_longlong_t)cb.cb_snapused); } else if (cb.cb_dryrun) { (void) printf(gettext("would reclaim %s\n"), buf); } else { (void) printf(gettext("will reclaim %s\n"), buf); } } if (!cb.cb_dryrun) { if (cb.cb_doclones) { cb.cb_batchedsnaps = fnvlist_alloc(); err = destroy_clones(&cb); if (err == 0) { err = zfs_destroy_snaps_nvl(g_zfs, cb.cb_batchedsnaps, B_FALSE); } if (err != 0) { rv = 1; goto out; } } if (err == 0) { err = zfs_destroy_snaps_nvl(g_zfs, cb.cb_nvl, cb.cb_defer_destroy); } } if (err != 0) rv = 1; } else if (pound != NULL) { int err; nvlist_t *nvl; if (cb.cb_dryrun) { (void) fprintf(stderr, "dryrun is not supported with bookmark\n"); return (-1); } if (cb.cb_defer_destroy) { (void) fprintf(stderr, "defer destroy is not supported with bookmark\n"); return (-1); } if (cb.cb_recurse) { (void) fprintf(stderr, "recursive is not supported with bookmark\n"); return (-1); } /* * Unfortunately, zfs_bookmark() doesn't honor the * casesensitivity setting. However, we can't simply * remove this check, because lzc_destroy_bookmarks() * ignores non-existent bookmarks, so this is necessary * to get a proper error message. */ if (!zfs_bookmark_exists(argv[0])) { (void) fprintf(stderr, gettext("bookmark '%s' " "does not exist.\n"), argv[0]); return (1); } nvl = fnvlist_alloc(); fnvlist_add_boolean(nvl, argv[0]); err = lzc_destroy_bookmarks(nvl, NULL); if (err != 0) { (void) zfs_standard_error(g_zfs, err, "cannot destroy bookmark"); } nvlist_free(nvl); return (err); } else { /* Open the given dataset */ if ((zhp = zfs_open(g_zfs, argv[0], type)) == NULL) return (1); cb.cb_target = zhp; /* * Perform an explicit check for pools before going any further. */ if (!cb.cb_recurse && strchr(zfs_get_name(zhp), '/') == NULL && zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) { (void) fprintf(stderr, gettext("cannot destroy '%s': " "operation does not apply to pools\n"), zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use 'zfs destroy -r " "%s' to destroy all datasets in the pool\n"), zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use 'zpool destroy %s' " "to destroy the pool itself\n"), zfs_get_name(zhp)); rv = 1; goto out; } /* * Check for any dependents and/or clones. */ cb.cb_first = B_TRUE; if (!cb.cb_doclones && zfs_iter_dependents_v2(zhp, 0, B_TRUE, destroy_check_dependent, &cb) != 0) { rv = 1; goto out; } if (cb.cb_error) { rv = 1; goto out; } cb.cb_batchedsnaps = fnvlist_alloc(); if (zfs_iter_dependents_v2(zhp, 0, B_FALSE, destroy_callback, &cb) != 0) { rv = 1; goto out; } /* * Do the real thing. The callback will close the * handle regardless of whether it succeeds or not. */ err = destroy_callback(zhp, &cb); zhp = NULL; if (err == 0) { err = zfs_destroy_snaps_nvl(g_zfs, cb.cb_batchedsnaps, cb.cb_defer_destroy); } if (err != 0 || cb.cb_error == B_TRUE) rv = 1; } out: fnvlist_free(cb.cb_batchedsnaps); fnvlist_free(cb.cb_nvl); if (zhp != NULL) zfs_close(zhp); return (rv); } static boolean_t is_recvd_column(zprop_get_cbdata_t *cbp) { int i; zfs_get_column_t col; for (i = 0; i < ZFS_GET_NCOLS && (col = cbp->cb_columns[i]) != GET_COL_NONE; i++) if (col == GET_COL_RECVD) return (B_TRUE); return (B_FALSE); } /* * zfs get [-rHp] [-o all | field[,field]...] [-s source[,source]...] * < all | property[,property]... > < fs | snap | vol > ... * * -r recurse over any child datasets * -H scripted mode. Headers are stripped, and fields are separated * by tabs instead of spaces. * -o Set of fields to display. One of "name,property,value, * received,source". Default is "name,property,value,source". * "all" is an alias for all five. * -s Set of sources to allow. One of * "local,default,inherited,received,temporary,none". Default is * all six. * -p Display values in parsable (literal) format. * * Prints properties for the given datasets. The user can control which * columns to display as well as which property types to allow. */ /* * Invoked to display the properties for a single dataset. */ static int get_callback(zfs_handle_t *zhp, void *data) { char buf[ZFS_MAXPROPLEN]; char rbuf[ZFS_MAXPROPLEN]; zprop_source_t sourcetype; char source[ZFS_MAX_DATASET_NAME_LEN]; zprop_get_cbdata_t *cbp = data; nvlist_t *user_props = zfs_get_user_props(zhp); zprop_list_t *pl = cbp->cb_proplist; nvlist_t *propval; const char *strval; const char *sourceval; boolean_t received = is_recvd_column(cbp); for (; pl != NULL; pl = pl->pl_next) { char *recvdval = NULL; /* * Skip the special fake placeholder. This will also skip over * the name property when 'all' is specified. */ if (pl->pl_prop == ZFS_PROP_NAME && pl == cbp->cb_proplist) continue; if (pl->pl_prop != ZPROP_USERPROP) { if (zfs_prop_get(zhp, pl->pl_prop, buf, sizeof (buf), &sourcetype, source, sizeof (source), cbp->cb_literal) != 0) { if (pl->pl_all) continue; if (!zfs_prop_valid_for_type(pl->pl_prop, ZFS_TYPE_DATASET, B_FALSE)) { (void) fprintf(stderr, gettext("No such property '%s'\n"), zfs_prop_to_name(pl->pl_prop)); continue; } sourcetype = ZPROP_SRC_NONE; (void) strlcpy(buf, "-", sizeof (buf)); } if (received && (zfs_prop_get_recvd(zhp, zfs_prop_to_name(pl->pl_prop), rbuf, sizeof (rbuf), cbp->cb_literal) == 0)) recvdval = rbuf; zprop_print_one_property(zfs_get_name(zhp), cbp, zfs_prop_to_name(pl->pl_prop), buf, sourcetype, source, recvdval); } else if (zfs_prop_userquota(pl->pl_user_prop)) { sourcetype = ZPROP_SRC_LOCAL; if (zfs_prop_get_userquota(zhp, pl->pl_user_prop, buf, sizeof (buf), cbp->cb_literal) != 0) { sourcetype = ZPROP_SRC_NONE; (void) strlcpy(buf, "-", sizeof (buf)); } zprop_print_one_property(zfs_get_name(zhp), cbp, pl->pl_user_prop, buf, sourcetype, source, NULL); } else if (zfs_prop_written(pl->pl_user_prop)) { sourcetype = ZPROP_SRC_LOCAL; if (zfs_prop_get_written(zhp, pl->pl_user_prop, buf, sizeof (buf), cbp->cb_literal) != 0) { sourcetype = ZPROP_SRC_NONE; (void) strlcpy(buf, "-", sizeof (buf)); } zprop_print_one_property(zfs_get_name(zhp), cbp, pl->pl_user_prop, buf, sourcetype, source, NULL); } else { if (nvlist_lookup_nvlist(user_props, pl->pl_user_prop, &propval) != 0) { if (pl->pl_all) continue; sourcetype = ZPROP_SRC_NONE; strval = "-"; } else { strval = fnvlist_lookup_string(propval, ZPROP_VALUE); sourceval = fnvlist_lookup_string(propval, ZPROP_SOURCE); if (strcmp(sourceval, zfs_get_name(zhp)) == 0) { sourcetype = ZPROP_SRC_LOCAL; } else if (strcmp(sourceval, ZPROP_SOURCE_VAL_RECVD) == 0) { sourcetype = ZPROP_SRC_RECEIVED; } else { sourcetype = ZPROP_SRC_INHERITED; (void) strlcpy(source, sourceval, sizeof (source)); } } if (received && (zfs_prop_get_recvd(zhp, pl->pl_user_prop, rbuf, sizeof (rbuf), cbp->cb_literal) == 0)) recvdval = rbuf; zprop_print_one_property(zfs_get_name(zhp), cbp, pl->pl_user_prop, strval, sourcetype, source, recvdval); } } return (0); } static int zfs_do_get(int argc, char **argv) { zprop_get_cbdata_t cb = { 0 }; int i, c, flags = ZFS_ITER_ARGS_CAN_BE_PATHS; int types = ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK; char *fields; int ret = 0; int limit = 0; zprop_list_t fake_name = { 0 }; /* * Set up default columns and sources. */ cb.cb_sources = ZPROP_SRC_ALL; cb.cb_columns[0] = GET_COL_NAME; cb.cb_columns[1] = GET_COL_PROPERTY; cb.cb_columns[2] = GET_COL_VALUE; cb.cb_columns[3] = GET_COL_SOURCE; cb.cb_type = ZFS_TYPE_DATASET; /* check options */ while ((c = getopt(argc, argv, ":d:o:s:rt:Hp")) != -1) { switch (c) { case 'p': cb.cb_literal = B_TRUE; break; case 'd': limit = parse_depth(optarg, &flags); break; case 'r': flags |= ZFS_ITER_RECURSE; break; case 'H': cb.cb_scripted = B_TRUE; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case 'o': /* * Process the set of columns to display. We zero out * the structure to give us a blank slate. */ memset(&cb.cb_columns, 0, sizeof (cb.cb_columns)); i = 0; for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const col_subopts[] = { "name", "property", "value", "received", "source", "all" }; static const zfs_get_column_t col_subopt_col[] = { GET_COL_NAME, GET_COL_PROPERTY, GET_COL_VALUE, GET_COL_RECVD, GET_COL_SOURCE }; static const int col_subopt_flags[] = { 0, 0, 0, ZFS_ITER_RECVD_PROPS, 0 }; if (i == ZFS_GET_NCOLS) { (void) fprintf(stderr, gettext("too " "many fields given to -o " "option\n")); usage(B_FALSE); } for (c = 0; c < ARRAY_SIZE(col_subopts); ++c) if (strcmp(tok, col_subopts[c]) == 0) goto found; (void) fprintf(stderr, gettext("invalid column name '%s'\n"), tok); usage(B_FALSE); found: if (c >= 5) { if (i > 0) { (void) fprintf(stderr, gettext("\"all\" conflicts " "with specific fields " "given to -o option\n")); usage(B_FALSE); } memcpy(cb.cb_columns, col_subopt_col, sizeof (col_subopt_col)); flags |= ZFS_ITER_RECVD_PROPS; i = ZFS_GET_NCOLS; } else { cb.cb_columns[i++] = col_subopt_col[c]; flags |= col_subopt_flags[c]; } } break; case 's': cb.cb_sources = 0; for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const source_opt[] = { "local", "default", "inherited", "received", "temporary", "none" }; static const int source_flg[] = { ZPROP_SRC_LOCAL, ZPROP_SRC_DEFAULT, ZPROP_SRC_INHERITED, ZPROP_SRC_RECEIVED, ZPROP_SRC_TEMPORARY, ZPROP_SRC_NONE }; for (i = 0; i < ARRAY_SIZE(source_opt); ++i) if (strcmp(tok, source_opt[i]) == 0) { cb.cb_sources |= source_flg[i]; goto found2; } (void) fprintf(stderr, gettext("invalid source '%s'\n"), tok); usage(B_FALSE); found2:; } break; case 't': types = 0; flags &= ~ZFS_ITER_PROP_LISTSNAPS; for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const type_opts[] = { - "filesystem", "volume", - "snapshot", "snap", + "filesystem", + "fs", + "volume", + "vol", + "snapshot", + "snap", "bookmark", - "all" }; + "all" + }; static const int type_types[] = { - ZFS_TYPE_FILESYSTEM, ZFS_TYPE_VOLUME, - ZFS_TYPE_SNAPSHOT, ZFS_TYPE_SNAPSHOT, + ZFS_TYPE_FILESYSTEM, + ZFS_TYPE_FILESYSTEM, + ZFS_TYPE_VOLUME, + ZFS_TYPE_VOLUME, + ZFS_TYPE_SNAPSHOT, + ZFS_TYPE_SNAPSHOT, ZFS_TYPE_BOOKMARK, - ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK }; + ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK + }; for (i = 0; i < ARRAY_SIZE(type_opts); ++i) if (strcmp(tok, type_opts[i]) == 0) { types |= type_types[i]; goto found3; } (void) fprintf(stderr, gettext("invalid type '%s'\n"), tok); usage(B_FALSE); found3:; } break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing property " "argument\n")); usage(B_FALSE); } fields = argv[0]; /* * Handle users who want to get all snapshots or bookmarks * of a dataset (ex. 'zfs get -t snapshot refer '). */ if ((types == ZFS_TYPE_SNAPSHOT || types == ZFS_TYPE_BOOKMARK) && argc > 1 && (flags & ZFS_ITER_RECURSE) == 0 && limit == 0) { flags |= (ZFS_ITER_DEPTH_LIMIT | ZFS_ITER_RECURSE); limit = 1; } if (zprop_get_list(g_zfs, fields, &cb.cb_proplist, ZFS_TYPE_DATASET) != 0) usage(B_FALSE); argc--; argv++; /* * As part of zfs_expand_proplist(), we keep track of the maximum column * width for each property. For the 'NAME' (and 'SOURCE') columns, we * need to know the maximum name length. However, the user likely did * not specify 'name' as one of the properties to fetch, so we need to * make sure we always include at least this property for * print_get_headers() to work properly. */ if (cb.cb_proplist != NULL) { fake_name.pl_prop = ZFS_PROP_NAME; fake_name.pl_width = strlen(gettext("NAME")); fake_name.pl_next = cb.cb_proplist; cb.cb_proplist = &fake_name; } cb.cb_first = B_TRUE; /* run for each object */ ret = zfs_for_each(argc, argv, flags, types, NULL, &cb.cb_proplist, limit, get_callback, &cb); if (cb.cb_proplist == &fake_name) zprop_free_list(fake_name.pl_next); else zprop_free_list(cb.cb_proplist); return (ret); } /* * inherit [-rS] ... * * -r Recurse over all children * -S Revert to received value, if any * * For each dataset specified on the command line, inherit the given property * from its parent. Inheriting a property at the pool level will cause it to * use the default value. The '-r' flag will recurse over all children, and is * useful for setting a property on a hierarchy-wide basis, regardless of any * local modifications for each dataset. */ typedef struct inherit_cbdata { const char *cb_propname; boolean_t cb_received; } inherit_cbdata_t; static int inherit_recurse_cb(zfs_handle_t *zhp, void *data) { inherit_cbdata_t *cb = data; zfs_prop_t prop = zfs_name_to_prop(cb->cb_propname); /* * If we're doing it recursively, then ignore properties that * are not valid for this type of dataset. */ if (prop != ZPROP_INVAL && !zfs_prop_valid_for_type(prop, zfs_get_type(zhp), B_FALSE)) return (0); return (zfs_prop_inherit(zhp, cb->cb_propname, cb->cb_received) != 0); } static int inherit_cb(zfs_handle_t *zhp, void *data) { inherit_cbdata_t *cb = data; return (zfs_prop_inherit(zhp, cb->cb_propname, cb->cb_received) != 0); } static int zfs_do_inherit(int argc, char **argv) { int c; zfs_prop_t prop; inherit_cbdata_t cb = { 0 }; char *propname; int ret = 0; int flags = 0; boolean_t received = B_FALSE; /* check options */ while ((c = getopt(argc, argv, "rS")) != -1) { switch (c) { case 'r': flags |= ZFS_ITER_RECURSE; break; case 'S': received = B_TRUE; break; case '?': default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing property argument\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing dataset argument\n")); usage(B_FALSE); } propname = argv[0]; argc--; argv++; if ((prop = zfs_name_to_prop(propname)) != ZPROP_USERPROP) { if (zfs_prop_readonly(prop)) { (void) fprintf(stderr, gettext( "%s property is read-only\n"), propname); return (1); } if (!zfs_prop_inheritable(prop) && !received) { (void) fprintf(stderr, gettext("'%s' property cannot " "be inherited\n"), propname); if (prop == ZFS_PROP_QUOTA || prop == ZFS_PROP_RESERVATION || prop == ZFS_PROP_REFQUOTA || prop == ZFS_PROP_REFRESERVATION) { (void) fprintf(stderr, gettext("use 'zfs set " "%s=none' to clear\n"), propname); (void) fprintf(stderr, gettext("use 'zfs " "inherit -S %s' to revert to received " "value\n"), propname); } return (1); } if (received && (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION)) { (void) fprintf(stderr, gettext("'%s' property cannot " "be reverted to a received value\n"), propname); return (1); } } else if (!zfs_prop_user(propname)) { (void) fprintf(stderr, gettext("invalid property '%s'\n"), propname); usage(B_FALSE); } cb.cb_propname = propname; cb.cb_received = received; if (flags & ZFS_ITER_RECURSE) { ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET, NULL, NULL, 0, inherit_recurse_cb, &cb); } else { ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET, NULL, NULL, 0, inherit_cb, &cb); } return (ret); } typedef struct upgrade_cbdata { uint64_t cb_numupgraded; uint64_t cb_numsamegraded; uint64_t cb_numfailed; uint64_t cb_version; boolean_t cb_newer; boolean_t cb_foundone; char cb_lastfs[ZFS_MAX_DATASET_NAME_LEN]; } upgrade_cbdata_t; static int same_pool(zfs_handle_t *zhp, const char *name) { int len1 = strcspn(name, "/@"); const char *zhname = zfs_get_name(zhp); int len2 = strcspn(zhname, "/@"); if (len1 != len2) return (B_FALSE); return (strncmp(name, zhname, len1) == 0); } static int upgrade_list_callback(zfs_handle_t *zhp, void *data) { upgrade_cbdata_t *cb = data; int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION); /* list if it's old/new */ if ((!cb->cb_newer && version < ZPL_VERSION) || (cb->cb_newer && version > ZPL_VERSION)) { char *str; if (cb->cb_newer) { str = gettext("The following filesystems are " "formatted using a newer software version and\n" "cannot be accessed on the current system.\n\n"); } else { str = gettext("The following filesystems are " "out of date, and can be upgraded. After being\n" "upgraded, these filesystems (and any 'zfs send' " "streams generated from\n" "subsequent snapshots) will no longer be " "accessible by older software versions.\n\n"); } if (!cb->cb_foundone) { (void) puts(str); (void) printf(gettext("VER FILESYSTEM\n")); (void) printf(gettext("--- ------------\n")); cb->cb_foundone = B_TRUE; } (void) printf("%2u %s\n", version, zfs_get_name(zhp)); } return (0); } static int upgrade_set_callback(zfs_handle_t *zhp, void *data) { upgrade_cbdata_t *cb = data; int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION); int needed_spa_version; int spa_version; if (zfs_spa_version(zhp, &spa_version) < 0) return (-1); needed_spa_version = zfs_spa_version_map(cb->cb_version); if (needed_spa_version < 0) return (-1); if (spa_version < needed_spa_version) { /* can't upgrade */ (void) printf(gettext("%s: can not be " "upgraded; the pool version needs to first " "be upgraded\nto version %d\n\n"), zfs_get_name(zhp), needed_spa_version); cb->cb_numfailed++; return (0); } /* upgrade */ if (version < cb->cb_version) { char verstr[24]; (void) snprintf(verstr, sizeof (verstr), "%llu", (u_longlong_t)cb->cb_version); if (cb->cb_lastfs[0] && !same_pool(zhp, cb->cb_lastfs)) { /* * If they did "zfs upgrade -a", then we could * be doing ioctls to different pools. We need * to log this history once to each pool, and bypass * the normal history logging that happens in main(). */ (void) zpool_log_history(g_zfs, history_str); log_history = B_FALSE; } if (zfs_prop_set(zhp, "version", verstr) == 0) cb->cb_numupgraded++; else cb->cb_numfailed++; (void) strlcpy(cb->cb_lastfs, zfs_get_name(zhp), sizeof (cb->cb_lastfs)); } else if (version > cb->cb_version) { /* can't downgrade */ (void) printf(gettext("%s: can not be downgraded; " "it is already at version %u\n"), zfs_get_name(zhp), version); cb->cb_numfailed++; } else { cb->cb_numsamegraded++; } return (0); } /* * zfs upgrade * zfs upgrade -v * zfs upgrade [-r] [-V ] <-a | filesystem> */ static int zfs_do_upgrade(int argc, char **argv) { boolean_t all = B_FALSE; boolean_t showversions = B_FALSE; int ret = 0; upgrade_cbdata_t cb = { 0 }; int c; int flags = ZFS_ITER_ARGS_CAN_BE_PATHS; /* check options */ while ((c = getopt(argc, argv, "rvV:a")) != -1) { switch (c) { case 'r': flags |= ZFS_ITER_RECURSE; break; case 'v': showversions = B_TRUE; break; case 'V': if (zfs_prop_string_to_index(ZFS_PROP_VERSION, optarg, &cb.cb_version) != 0) { (void) fprintf(stderr, gettext("invalid version %s\n"), optarg); usage(B_FALSE); } break; case 'a': all = B_TRUE; break; case '?': default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if ((!all && !argc) && ((flags & ZFS_ITER_RECURSE) | cb.cb_version)) usage(B_FALSE); if (showversions && (flags & ZFS_ITER_RECURSE || all || cb.cb_version || argc)) usage(B_FALSE); if ((all || argc) && (showversions)) usage(B_FALSE); if (all && argc) usage(B_FALSE); if (showversions) { /* Show info on available versions. */ (void) printf(gettext("The following filesystem versions are " "supported:\n\n")); (void) printf(gettext("VER DESCRIPTION\n")); (void) printf("--- -----------------------------------------" "---------------\n"); (void) printf(gettext(" 1 Initial ZFS filesystem version\n")); (void) printf(gettext(" 2 Enhanced directory entries\n")); (void) printf(gettext(" 3 Case insensitive and filesystem " "user identifier (FUID)\n")); (void) printf(gettext(" 4 userquota, groupquota " "properties\n")); (void) printf(gettext(" 5 System attributes\n")); (void) printf(gettext("\nFor more information on a particular " "version, including supported releases,\n")); (void) printf("see the ZFS Administration Guide.\n\n"); ret = 0; } else if (argc || all) { /* Upgrade filesystems */ if (cb.cb_version == 0) cb.cb_version = ZPL_VERSION; ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_FILESYSTEM, NULL, NULL, 0, upgrade_set_callback, &cb); (void) printf(gettext("%llu filesystems upgraded\n"), (u_longlong_t)cb.cb_numupgraded); if (cb.cb_numsamegraded) { (void) printf(gettext("%llu filesystems already at " "this version\n"), (u_longlong_t)cb.cb_numsamegraded); } if (cb.cb_numfailed != 0) ret = 1; } else { /* List old-version filesystems */ boolean_t found; (void) printf(gettext("This system is currently running " "ZFS filesystem version %llu.\n\n"), ZPL_VERSION); flags |= ZFS_ITER_RECURSE; ret = zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM, NULL, NULL, 0, upgrade_list_callback, &cb); found = cb.cb_foundone; cb.cb_foundone = B_FALSE; cb.cb_newer = B_TRUE; ret |= zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM, NULL, NULL, 0, upgrade_list_callback, &cb); if (!cb.cb_foundone && !found) { (void) printf(gettext("All filesystems are " "formatted with the current version.\n")); } } return (ret); } /* * zfs userspace [-Hinp] [-o field[,...]] [-s field [-s field]...] * [-S field [-S field]...] [-t type[,...]] * filesystem | snapshot | path * zfs groupspace [-Hinp] [-o field[,...]] [-s field [-s field]...] * [-S field [-S field]...] [-t type[,...]] * filesystem | snapshot | path * zfs projectspace [-Hp] [-o field[,...]] [-s field [-s field]...] * [-S field [-S field]...] filesystem | snapshot | path * * -H Scripted mode; elide headers and separate columns by tabs. * -i Translate SID to POSIX ID. * -n Print numeric ID instead of user/group name. * -o Control which fields to display. * -p Use exact (parsable) numeric output. * -s Specify sort columns, descending order. * -S Specify sort columns, ascending order. * -t Control which object types to display. * * Displays space consumed by, and quotas on, each user in the specified * filesystem or snapshot. */ /* us_field_types, us_field_hdr and us_field_names should be kept in sync */ enum us_field_types { USFIELD_TYPE, USFIELD_NAME, USFIELD_USED, USFIELD_QUOTA, USFIELD_OBJUSED, USFIELD_OBJQUOTA }; static const char *const us_field_hdr[] = { "TYPE", "NAME", "USED", "QUOTA", "OBJUSED", "OBJQUOTA" }; static const char *const us_field_names[] = { "type", "name", "used", "quota", "objused", "objquota" }; #define USFIELD_LAST (sizeof (us_field_names) / sizeof (char *)) #define USTYPE_PSX_GRP (1 << 0) #define USTYPE_PSX_USR (1 << 1) #define USTYPE_SMB_GRP (1 << 2) #define USTYPE_SMB_USR (1 << 3) #define USTYPE_PROJ (1 << 4) #define USTYPE_ALL \ (USTYPE_PSX_GRP | USTYPE_PSX_USR | USTYPE_SMB_GRP | USTYPE_SMB_USR | \ USTYPE_PROJ) static int us_type_bits[] = { USTYPE_PSX_GRP, USTYPE_PSX_USR, USTYPE_SMB_GRP, USTYPE_SMB_USR, USTYPE_ALL }; static const char *const us_type_names[] = { "posixgroup", "posixuser", "smbgroup", "smbuser", "all" }; typedef struct us_node { nvlist_t *usn_nvl; uu_avl_node_t usn_avlnode; uu_list_node_t usn_listnode; } us_node_t; typedef struct us_cbdata { nvlist_t **cb_nvlp; uu_avl_pool_t *cb_avl_pool; uu_avl_t *cb_avl; boolean_t cb_numname; boolean_t cb_nicenum; boolean_t cb_sid2posix; zfs_userquota_prop_t cb_prop; zfs_sort_column_t *cb_sortcol; size_t cb_width[USFIELD_LAST]; } us_cbdata_t; static boolean_t us_populated = B_FALSE; typedef struct { zfs_sort_column_t *si_sortcol; boolean_t si_numname; } us_sort_info_t; static int us_field_index(const char *field) { for (int i = 0; i < USFIELD_LAST; i++) { if (strcmp(field, us_field_names[i]) == 0) return (i); } return (-1); } static int us_compare(const void *larg, const void *rarg, void *unused) { const us_node_t *l = larg; const us_node_t *r = rarg; us_sort_info_t *si = (us_sort_info_t *)unused; zfs_sort_column_t *sortcol = si->si_sortcol; boolean_t numname = si->si_numname; nvlist_t *lnvl = l->usn_nvl; nvlist_t *rnvl = r->usn_nvl; int rc = 0; boolean_t lvb, rvb; for (; sortcol != NULL; sortcol = sortcol->sc_next) { const char *lvstr = ""; const char *rvstr = ""; uint32_t lv32 = 0; uint32_t rv32 = 0; uint64_t lv64 = 0; uint64_t rv64 = 0; zfs_prop_t prop = sortcol->sc_prop; const char *propname = NULL; boolean_t reverse = sortcol->sc_reverse; switch (prop) { case ZFS_PROP_TYPE: propname = "type"; (void) nvlist_lookup_uint32(lnvl, propname, &lv32); (void) nvlist_lookup_uint32(rnvl, propname, &rv32); if (rv32 != lv32) rc = (rv32 < lv32) ? 1 : -1; break; case ZFS_PROP_NAME: propname = "name"; if (numname) { compare_nums: (void) nvlist_lookup_uint64(lnvl, propname, &lv64); (void) nvlist_lookup_uint64(rnvl, propname, &rv64); if (rv64 != lv64) rc = (rv64 < lv64) ? 1 : -1; } else { if ((nvlist_lookup_string(lnvl, propname, &lvstr) == ENOENT) || (nvlist_lookup_string(rnvl, propname, &rvstr) == ENOENT)) { goto compare_nums; } rc = strcmp(lvstr, rvstr); } break; case ZFS_PROP_USED: case ZFS_PROP_QUOTA: if (!us_populated) break; if (prop == ZFS_PROP_USED) propname = "used"; else propname = "quota"; (void) nvlist_lookup_uint64(lnvl, propname, &lv64); (void) nvlist_lookup_uint64(rnvl, propname, &rv64); if (rv64 != lv64) rc = (rv64 < lv64) ? 1 : -1; break; default: break; } if (rc != 0) { if (rc < 0) return (reverse ? 1 : -1); else return (reverse ? -1 : 1); } } /* * If entries still seem to be the same, check if they are of the same * type (smbentity is added only if we are doing SID to POSIX ID * translation where we can have duplicate type/name combinations). */ if (nvlist_lookup_boolean_value(lnvl, "smbentity", &lvb) == 0 && nvlist_lookup_boolean_value(rnvl, "smbentity", &rvb) == 0 && lvb != rvb) return (lvb < rvb ? -1 : 1); return (0); } static boolean_t zfs_prop_is_user(unsigned p) { return (p == ZFS_PROP_USERUSED || p == ZFS_PROP_USERQUOTA || p == ZFS_PROP_USEROBJUSED || p == ZFS_PROP_USEROBJQUOTA); } static boolean_t zfs_prop_is_group(unsigned p) { return (p == ZFS_PROP_GROUPUSED || p == ZFS_PROP_GROUPQUOTA || p == ZFS_PROP_GROUPOBJUSED || p == ZFS_PROP_GROUPOBJQUOTA); } static boolean_t zfs_prop_is_project(unsigned p) { return (p == ZFS_PROP_PROJECTUSED || p == ZFS_PROP_PROJECTQUOTA || p == ZFS_PROP_PROJECTOBJUSED || p == ZFS_PROP_PROJECTOBJQUOTA); } static inline const char * us_type2str(unsigned field_type) { switch (field_type) { case USTYPE_PSX_USR: return ("POSIX User"); case USTYPE_PSX_GRP: return ("POSIX Group"); case USTYPE_SMB_USR: return ("SMB User"); case USTYPE_SMB_GRP: return ("SMB Group"); case USTYPE_PROJ: return ("Project"); default: return ("Undefined"); } } static int userspace_cb(void *arg, const char *domain, uid_t rid, uint64_t space) { us_cbdata_t *cb = (us_cbdata_t *)arg; zfs_userquota_prop_t prop = cb->cb_prop; char *name = NULL; const char *propname; char sizebuf[32]; us_node_t *node; uu_avl_pool_t *avl_pool = cb->cb_avl_pool; uu_avl_t *avl = cb->cb_avl; uu_avl_index_t idx; nvlist_t *props; us_node_t *n; zfs_sort_column_t *sortcol = cb->cb_sortcol; unsigned type = 0; const char *typestr; size_t namelen; size_t typelen; size_t sizelen; int typeidx, nameidx, sizeidx; us_sort_info_t sortinfo = { sortcol, cb->cb_numname }; boolean_t smbentity = B_FALSE; if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) nomem(); node = safe_malloc(sizeof (us_node_t)); uu_avl_node_init(node, &node->usn_avlnode, avl_pool); node->usn_nvl = props; if (domain != NULL && domain[0] != '\0') { #ifdef HAVE_IDMAP /* SMB */ char sid[MAXNAMELEN + 32]; uid_t id; uint64_t classes; int err; directory_error_t e; smbentity = B_TRUE; (void) snprintf(sid, sizeof (sid), "%s-%u", domain, rid); if (prop == ZFS_PROP_GROUPUSED || prop == ZFS_PROP_GROUPQUOTA) { type = USTYPE_SMB_GRP; err = sid_to_id(sid, B_FALSE, &id); } else { type = USTYPE_SMB_USR; err = sid_to_id(sid, B_TRUE, &id); } if (err == 0) { rid = id; if (!cb->cb_sid2posix) { e = directory_name_from_sid(NULL, sid, &name, &classes); if (e != NULL) directory_error_free(e); if (name == NULL) name = sid; } } #else nvlist_free(props); free(node); return (-1); #endif /* HAVE_IDMAP */ } if (cb->cb_sid2posix || domain == NULL || domain[0] == '\0') { /* POSIX or -i */ if (zfs_prop_is_group(prop)) { type = USTYPE_PSX_GRP; if (!cb->cb_numname) { struct group *g; if ((g = getgrgid(rid)) != NULL) name = g->gr_name; } } else if (zfs_prop_is_user(prop)) { type = USTYPE_PSX_USR; if (!cb->cb_numname) { struct passwd *p; if ((p = getpwuid(rid)) != NULL) name = p->pw_name; } } else { type = USTYPE_PROJ; } } /* * Make sure that the type/name combination is unique when doing * SID to POSIX ID translation (hence changing the type from SMB to * POSIX). */ if (cb->cb_sid2posix && nvlist_add_boolean_value(props, "smbentity", smbentity) != 0) nomem(); /* Calculate/update width of TYPE field */ typestr = us_type2str(type); typelen = strlen(gettext(typestr)); typeidx = us_field_index("type"); if (typelen > cb->cb_width[typeidx]) cb->cb_width[typeidx] = typelen; if (nvlist_add_uint32(props, "type", type) != 0) nomem(); /* Calculate/update width of NAME field */ if ((cb->cb_numname && cb->cb_sid2posix) || name == NULL) { if (nvlist_add_uint64(props, "name", rid) != 0) nomem(); namelen = snprintf(NULL, 0, "%u", rid); } else { if (nvlist_add_string(props, "name", name) != 0) nomem(); namelen = strlen(name); } nameidx = us_field_index("name"); if (nameidx >= 0 && namelen > cb->cb_width[nameidx]) cb->cb_width[nameidx] = namelen; /* * Check if this type/name combination is in the list and update it; * otherwise add new node to the list. */ if ((n = uu_avl_find(avl, node, &sortinfo, &idx)) == NULL) { uu_avl_insert(avl, node, idx); } else { nvlist_free(props); free(node); node = n; props = node->usn_nvl; } /* Calculate/update width of USED/QUOTA fields */ if (cb->cb_nicenum) { if (prop == ZFS_PROP_USERUSED || prop == ZFS_PROP_GROUPUSED || prop == ZFS_PROP_USERQUOTA || prop == ZFS_PROP_GROUPQUOTA || prop == ZFS_PROP_PROJECTUSED || prop == ZFS_PROP_PROJECTQUOTA) { zfs_nicebytes(space, sizebuf, sizeof (sizebuf)); } else { zfs_nicenum(space, sizebuf, sizeof (sizebuf)); } } else { (void) snprintf(sizebuf, sizeof (sizebuf), "%llu", (u_longlong_t)space); } sizelen = strlen(sizebuf); if (prop == ZFS_PROP_USERUSED || prop == ZFS_PROP_GROUPUSED || prop == ZFS_PROP_PROJECTUSED) { propname = "used"; if (!nvlist_exists(props, "quota")) (void) nvlist_add_uint64(props, "quota", 0); } else if (prop == ZFS_PROP_USERQUOTA || prop == ZFS_PROP_GROUPQUOTA || prop == ZFS_PROP_PROJECTQUOTA) { propname = "quota"; if (!nvlist_exists(props, "used")) (void) nvlist_add_uint64(props, "used", 0); } else if (prop == ZFS_PROP_USEROBJUSED || prop == ZFS_PROP_GROUPOBJUSED || prop == ZFS_PROP_PROJECTOBJUSED) { propname = "objused"; if (!nvlist_exists(props, "objquota")) (void) nvlist_add_uint64(props, "objquota", 0); } else if (prop == ZFS_PROP_USEROBJQUOTA || prop == ZFS_PROP_GROUPOBJQUOTA || prop == ZFS_PROP_PROJECTOBJQUOTA) { propname = "objquota"; if (!nvlist_exists(props, "objused")) (void) nvlist_add_uint64(props, "objused", 0); } else { return (-1); } sizeidx = us_field_index(propname); if (sizeidx >= 0 && sizelen > cb->cb_width[sizeidx]) cb->cb_width[sizeidx] = sizelen; if (nvlist_add_uint64(props, propname, space) != 0) nomem(); return (0); } static void print_us_node(boolean_t scripted, boolean_t parsable, int *fields, int types, size_t *width, us_node_t *node) { nvlist_t *nvl = node->usn_nvl; char valstr[MAXNAMELEN]; boolean_t first = B_TRUE; int cfield = 0; int field; uint32_t ustype; /* Check type */ (void) nvlist_lookup_uint32(nvl, "type", &ustype); if (!(ustype & types)) return; while ((field = fields[cfield]) != USFIELD_LAST) { nvpair_t *nvp = NULL; data_type_t type; uint32_t val32 = -1; uint64_t val64 = -1; const char *strval = "-"; while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) if (strcmp(nvpair_name(nvp), us_field_names[field]) == 0) break; type = nvp == NULL ? DATA_TYPE_UNKNOWN : nvpair_type(nvp); switch (type) { case DATA_TYPE_UINT32: val32 = fnvpair_value_uint32(nvp); break; case DATA_TYPE_UINT64: val64 = fnvpair_value_uint64(nvp); break; case DATA_TYPE_STRING: strval = fnvpair_value_string(nvp); break; case DATA_TYPE_UNKNOWN: break; default: (void) fprintf(stderr, "invalid data type\n"); } switch (field) { case USFIELD_TYPE: if (type == DATA_TYPE_UINT32) strval = us_type2str(val32); break; case USFIELD_NAME: if (type == DATA_TYPE_UINT64) { (void) sprintf(valstr, "%llu", (u_longlong_t)val64); strval = valstr; } break; case USFIELD_USED: case USFIELD_QUOTA: if (type == DATA_TYPE_UINT64) { if (parsable) { (void) sprintf(valstr, "%llu", (u_longlong_t)val64); strval = valstr; } else if (field == USFIELD_QUOTA && val64 == 0) { strval = "none"; } else { zfs_nicebytes(val64, valstr, sizeof (valstr)); strval = valstr; } } break; case USFIELD_OBJUSED: case USFIELD_OBJQUOTA: if (type == DATA_TYPE_UINT64) { if (parsable) { (void) sprintf(valstr, "%llu", (u_longlong_t)val64); strval = valstr; } else if (field == USFIELD_OBJQUOTA && val64 == 0) { strval = "none"; } else { zfs_nicenum(val64, valstr, sizeof (valstr)); strval = valstr; } } break; } if (!first) { if (scripted) (void) putchar('\t'); else (void) fputs(" ", stdout); } if (scripted) (void) fputs(strval, stdout); else if (field == USFIELD_TYPE || field == USFIELD_NAME) (void) printf("%-*s", (int)width[field], strval); else (void) printf("%*s", (int)width[field], strval); first = B_FALSE; cfield++; } (void) putchar('\n'); } static void print_us(boolean_t scripted, boolean_t parsable, int *fields, int types, size_t *width, boolean_t rmnode, uu_avl_t *avl) { us_node_t *node; const char *col; int cfield = 0; int field; if (!scripted) { boolean_t first = B_TRUE; while ((field = fields[cfield]) != USFIELD_LAST) { col = gettext(us_field_hdr[field]); if (field == USFIELD_TYPE || field == USFIELD_NAME) { (void) printf(first ? "%-*s" : " %-*s", (int)width[field], col); } else { (void) printf(first ? "%*s" : " %*s", (int)width[field], col); } first = B_FALSE; cfield++; } (void) printf("\n"); } for (node = uu_avl_first(avl); node; node = uu_avl_next(avl, node)) { print_us_node(scripted, parsable, fields, types, width, node); if (rmnode) nvlist_free(node->usn_nvl); } } static int zfs_do_userspace(int argc, char **argv) { zfs_handle_t *zhp; zfs_userquota_prop_t p; uu_avl_pool_t *avl_pool; uu_avl_t *avl_tree; uu_avl_walk_t *walk; char *delim; char deffields[] = "type,name,used,quota,objused,objquota"; char *ofield = NULL; char *tfield = NULL; int cfield = 0; int fields[256]; int i; boolean_t scripted = B_FALSE; boolean_t prtnum = B_FALSE; boolean_t parsable = B_FALSE; boolean_t sid2posix = B_FALSE; int ret = 0; int c; zfs_sort_column_t *sortcol = NULL; int types = USTYPE_PSX_USR | USTYPE_SMB_USR; us_cbdata_t cb; us_node_t *node; us_node_t *rmnode; uu_list_pool_t *listpool; uu_list_t *list; uu_avl_index_t idx = 0; uu_list_index_t idx2 = 0; if (argc < 2) usage(B_FALSE); if (strcmp(argv[0], "groupspace") == 0) { /* Toggle default group types */ types = USTYPE_PSX_GRP | USTYPE_SMB_GRP; } else if (strcmp(argv[0], "projectspace") == 0) { types = USTYPE_PROJ; prtnum = B_TRUE; } while ((c = getopt(argc, argv, "nHpo:s:S:t:i")) != -1) { switch (c) { case 'n': if (types == USTYPE_PROJ) { (void) fprintf(stderr, gettext("invalid option 'n'\n")); usage(B_FALSE); } prtnum = B_TRUE; break; case 'H': scripted = B_TRUE; break; case 'p': parsable = B_TRUE; break; case 'o': ofield = optarg; break; case 's': case 'S': if (zfs_add_sort_column(&sortcol, optarg, c == 's' ? B_FALSE : B_TRUE) != 0) { (void) fprintf(stderr, gettext("invalid field '%s'\n"), optarg); usage(B_FALSE); } break; case 't': if (types == USTYPE_PROJ) { (void) fprintf(stderr, gettext("invalid option 't'\n")); usage(B_FALSE); } tfield = optarg; break; case 'i': if (types == USTYPE_PROJ) { (void) fprintf(stderr, gettext("invalid option 'i'\n")); usage(B_FALSE); } sid2posix = B_TRUE; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing dataset name\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } /* Use default output fields if not specified using -o */ if (ofield == NULL) ofield = deffields; do { if ((delim = strchr(ofield, ',')) != NULL) *delim = '\0'; if ((fields[cfield++] = us_field_index(ofield)) == -1) { (void) fprintf(stderr, gettext("invalid type '%s' " "for -o option\n"), ofield); return (-1); } if (delim != NULL) ofield = delim + 1; } while (delim != NULL); fields[cfield] = USFIELD_LAST; /* Override output types (-t option) */ if (tfield != NULL) { types = 0; do { boolean_t found = B_FALSE; if ((delim = strchr(tfield, ',')) != NULL) *delim = '\0'; for (i = 0; i < sizeof (us_type_bits) / sizeof (int); i++) { if (strcmp(tfield, us_type_names[i]) == 0) { found = B_TRUE; types |= us_type_bits[i]; break; } } if (!found) { (void) fprintf(stderr, gettext("invalid type " "'%s' for -t option\n"), tfield); return (-1); } if (delim != NULL) tfield = delim + 1; } while (delim != NULL); } if ((zhp = zfs_path_to_zhandle(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT)) == NULL) return (1); if (zfs_get_underlying_type(zhp) != ZFS_TYPE_FILESYSTEM) { (void) fprintf(stderr, gettext("operation is only applicable " "to filesystems and their snapshots\n")); zfs_close(zhp); return (1); } if ((avl_pool = uu_avl_pool_create("us_avl_pool", sizeof (us_node_t), offsetof(us_node_t, usn_avlnode), us_compare, UU_DEFAULT)) == NULL) nomem(); if ((avl_tree = uu_avl_create(avl_pool, NULL, UU_DEFAULT)) == NULL) nomem(); /* Always add default sorting columns */ (void) zfs_add_sort_column(&sortcol, "type", B_FALSE); (void) zfs_add_sort_column(&sortcol, "name", B_FALSE); cb.cb_sortcol = sortcol; cb.cb_numname = prtnum; cb.cb_nicenum = !parsable; cb.cb_avl_pool = avl_pool; cb.cb_avl = avl_tree; cb.cb_sid2posix = sid2posix; for (i = 0; i < USFIELD_LAST; i++) cb.cb_width[i] = strlen(gettext(us_field_hdr[i])); for (p = 0; p < ZFS_NUM_USERQUOTA_PROPS; p++) { if ((zfs_prop_is_user(p) && !(types & (USTYPE_PSX_USR | USTYPE_SMB_USR))) || (zfs_prop_is_group(p) && !(types & (USTYPE_PSX_GRP | USTYPE_SMB_GRP))) || (zfs_prop_is_project(p) && types != USTYPE_PROJ)) continue; cb.cb_prop = p; if ((ret = zfs_userspace(zhp, p, userspace_cb, &cb)) != 0) { zfs_close(zhp); return (ret); } } zfs_close(zhp); /* Sort the list */ if ((node = uu_avl_first(avl_tree)) == NULL) return (0); us_populated = B_TRUE; listpool = uu_list_pool_create("tmplist", sizeof (us_node_t), offsetof(us_node_t, usn_listnode), NULL, UU_DEFAULT); list = uu_list_create(listpool, NULL, UU_DEFAULT); uu_list_node_init(node, &node->usn_listnode, listpool); while (node != NULL) { rmnode = node; node = uu_avl_next(avl_tree, node); uu_avl_remove(avl_tree, rmnode); if (uu_list_find(list, rmnode, NULL, &idx2) == NULL) uu_list_insert(list, rmnode, idx2); } for (node = uu_list_first(list); node != NULL; node = uu_list_next(list, node)) { us_sort_info_t sortinfo = { sortcol, cb.cb_numname }; if (uu_avl_find(avl_tree, node, &sortinfo, &idx) == NULL) uu_avl_insert(avl_tree, node, idx); } uu_list_destroy(list); uu_list_pool_destroy(listpool); /* Print and free node nvlist memory */ print_us(scripted, parsable, fields, types, cb.cb_width, B_TRUE, cb.cb_avl); zfs_free_sort_columns(sortcol); /* Clean up the AVL tree */ if ((walk = uu_avl_walk_start(cb.cb_avl, UU_WALK_ROBUST)) == NULL) nomem(); while ((node = uu_avl_walk_next(walk)) != NULL) { uu_avl_remove(cb.cb_avl, node); free(node); } uu_avl_walk_end(walk); uu_avl_destroy(avl_tree); uu_avl_pool_destroy(avl_pool); return (ret); } /* * list [-Hp][-r|-d max] [-o property[,...]] [-s property] ... [-S property] * [-t type[,...]] [filesystem|volume|snapshot] ... * * -H Scripted mode; elide headers and separate columns by tabs * -p Display values in parsable (literal) format. * -r Recurse over all children * -d Limit recursion by depth. * -o Control which fields to display. * -s Specify sort columns, descending order. * -S Specify sort columns, ascending order. * -t Control which object types to display. * * When given no arguments, list all filesystems in the system. * Otherwise, list the specified datasets, optionally recursing down them if * '-r' is specified. */ typedef struct list_cbdata { boolean_t cb_first; boolean_t cb_literal; boolean_t cb_scripted; zprop_list_t *cb_proplist; } list_cbdata_t; /* * Given a list of columns to display, output appropriate headers for each one. */ static void print_header(list_cbdata_t *cb) { zprop_list_t *pl = cb->cb_proplist; char headerbuf[ZFS_MAXPROPLEN]; const char *header; int i; boolean_t first = B_TRUE; boolean_t right_justify; color_start(ANSI_BOLD); for (; pl != NULL; pl = pl->pl_next) { if (!first) { (void) printf(" "); } else { first = B_FALSE; } right_justify = B_FALSE; if (pl->pl_prop != ZPROP_USERPROP) { header = zfs_prop_column_name(pl->pl_prop); right_justify = zfs_prop_align_right(pl->pl_prop); } else { for (i = 0; pl->pl_user_prop[i] != '\0'; i++) headerbuf[i] = toupper(pl->pl_user_prop[i]); headerbuf[i] = '\0'; header = headerbuf; } if (pl->pl_next == NULL && !right_justify) (void) printf("%s", header); else if (right_justify) (void) printf("%*s", (int)pl->pl_width, header); else (void) printf("%-*s", (int)pl->pl_width, header); } color_end(); (void) printf("\n"); } /* * Decides on the color that the avail value should be printed in. * > 80% used = yellow * > 90% used = red */ static const char * zfs_list_avail_color(zfs_handle_t *zhp) { uint64_t used = zfs_prop_get_int(zhp, ZFS_PROP_USED); uint64_t avail = zfs_prop_get_int(zhp, ZFS_PROP_AVAILABLE); int percentage = (int)((double)avail / MAX(avail + used, 1) * 100); if (percentage > 20) return (NULL); else if (percentage > 10) return (ANSI_YELLOW); else return (ANSI_RED); } /* * Given a dataset and a list of fields, print out all the properties according * to the described layout. */ static void print_dataset(zfs_handle_t *zhp, list_cbdata_t *cb) { zprop_list_t *pl = cb->cb_proplist; boolean_t first = B_TRUE; char property[ZFS_MAXPROPLEN]; nvlist_t *userprops = zfs_get_user_props(zhp); nvlist_t *propval; const char *propstr; boolean_t right_justify; for (; pl != NULL; pl = pl->pl_next) { if (!first) { if (cb->cb_scripted) (void) putchar('\t'); else (void) fputs(" ", stdout); } else { first = B_FALSE; } if (pl->pl_prop == ZFS_PROP_NAME) { (void) strlcpy(property, zfs_get_name(zhp), sizeof (property)); propstr = property; right_justify = zfs_prop_align_right(pl->pl_prop); } else if (pl->pl_prop != ZPROP_USERPROP) { if (zfs_prop_get(zhp, pl->pl_prop, property, sizeof (property), NULL, NULL, 0, cb->cb_literal) != 0) propstr = "-"; else propstr = property; right_justify = zfs_prop_align_right(pl->pl_prop); } else if (zfs_prop_userquota(pl->pl_user_prop)) { if (zfs_prop_get_userquota(zhp, pl->pl_user_prop, property, sizeof (property), cb->cb_literal) != 0) propstr = "-"; else propstr = property; right_justify = B_TRUE; } else if (zfs_prop_written(pl->pl_user_prop)) { if (zfs_prop_get_written(zhp, pl->pl_user_prop, property, sizeof (property), cb->cb_literal) != 0) propstr = "-"; else propstr = property; right_justify = B_TRUE; } else { if (nvlist_lookup_nvlist(userprops, pl->pl_user_prop, &propval) != 0) propstr = "-"; else propstr = fnvlist_lookup_string(propval, ZPROP_VALUE); right_justify = B_FALSE; } /* * zfs_list_avail_color() needs ZFS_PROP_AVAILABLE + USED * - so we need another for() search for the USED part * - when no colors wanted, we can skip the whole thing */ if (use_color() && pl->pl_prop == ZFS_PROP_AVAILABLE) { zprop_list_t *pl2 = cb->cb_proplist; for (; pl2 != NULL; pl2 = pl2->pl_next) { if (pl2->pl_prop == ZFS_PROP_USED) { color_start(zfs_list_avail_color(zhp)); /* found it, no need for more loops */ break; } } } /* * If this is being called in scripted mode, or if this is the * last column and it is left-justified, don't include a width * format specifier. */ if (cb->cb_scripted || (pl->pl_next == NULL && !right_justify)) (void) fputs(propstr, stdout); else if (right_justify) (void) printf("%*s", (int)pl->pl_width, propstr); else (void) printf("%-*s", (int)pl->pl_width, propstr); if (pl->pl_prop == ZFS_PROP_AVAILABLE) color_end(); } (void) putchar('\n'); } /* * Generic callback function to list a dataset or snapshot. */ static int list_callback(zfs_handle_t *zhp, void *data) { list_cbdata_t *cbp = data; if (cbp->cb_first) { if (!cbp->cb_scripted) print_header(cbp); cbp->cb_first = B_FALSE; } print_dataset(zhp, cbp); return (0); } static int zfs_do_list(int argc, char **argv) { int c; char default_fields[] = "name,used,available,referenced,mountpoint"; int types = ZFS_TYPE_DATASET; boolean_t types_specified = B_FALSE; char *fields = default_fields; list_cbdata_t cb = { 0 }; int limit = 0; int ret = 0; zfs_sort_column_t *sortcol = NULL; int flags = ZFS_ITER_PROP_LISTSNAPS | ZFS_ITER_ARGS_CAN_BE_PATHS; /* check options */ while ((c = getopt(argc, argv, "HS:d:o:prs:t:")) != -1) { switch (c) { case 'o': fields = optarg; break; case 'p': cb.cb_literal = B_TRUE; flags |= ZFS_ITER_LITERAL_PROPS; break; case 'd': limit = parse_depth(optarg, &flags); break; case 'r': flags |= ZFS_ITER_RECURSE; break; case 'H': cb.cb_scripted = B_TRUE; break; case 's': if (zfs_add_sort_column(&sortcol, optarg, B_FALSE) != 0) { (void) fprintf(stderr, gettext("invalid property '%s'\n"), optarg); usage(B_FALSE); } break; case 'S': if (zfs_add_sort_column(&sortcol, optarg, B_TRUE) != 0) { (void) fprintf(stderr, gettext("invalid property '%s'\n"), optarg); usage(B_FALSE); } break; case 't': types = 0; types_specified = B_TRUE; flags &= ~ZFS_ITER_PROP_LISTSNAPS; for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const type_subopts[] = { "filesystem", "fs", "volume", "vol", "snapshot", "snap", "bookmark", "all" }; static const int type_types[] = { ZFS_TYPE_FILESYSTEM, ZFS_TYPE_FILESYSTEM, ZFS_TYPE_VOLUME, ZFS_TYPE_VOLUME, ZFS_TYPE_SNAPSHOT, ZFS_TYPE_SNAPSHOT, ZFS_TYPE_BOOKMARK, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK }; for (c = 0; c < ARRAY_SIZE(type_subopts); ++c) if (strcmp(tok, type_subopts[c]) == 0) { types |= type_types[c]; goto found3; } (void) fprintf(stderr, gettext("invalid type '%s'\n"), tok); usage(B_FALSE); found3:; } break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* * If "-o space" and no types were specified, don't display snapshots. */ if (strcmp(fields, "space") == 0 && types_specified == B_FALSE) types &= ~ZFS_TYPE_SNAPSHOT; /* * Handle users who want to list all snapshots or bookmarks * of the current dataset (ex. 'zfs list -t snapshot '). */ if ((types == ZFS_TYPE_SNAPSHOT || types == ZFS_TYPE_BOOKMARK) && argc > 0 && (flags & ZFS_ITER_RECURSE) == 0 && limit == 0) { flags |= (ZFS_ITER_DEPTH_LIMIT | ZFS_ITER_RECURSE); limit = 1; } /* * If the user specifies '-o all', the zprop_get_list() doesn't * normally include the name of the dataset. For 'zfs list', we always * want this property to be first. */ if (zprop_get_list(g_zfs, fields, &cb.cb_proplist, ZFS_TYPE_DATASET) != 0) usage(B_FALSE); cb.cb_first = B_TRUE; /* * If we are only going to list and sort by properties that are "fast" * then we can use "simple" mode and avoid populating the properties * nvlist. */ if (zfs_list_only_by_fast(cb.cb_proplist) && zfs_sort_only_by_fast(sortcol)) flags |= ZFS_ITER_SIMPLE; ret = zfs_for_each(argc, argv, flags, types, sortcol, &cb.cb_proplist, limit, list_callback, &cb); zprop_free_list(cb.cb_proplist); zfs_free_sort_columns(sortcol); if (ret == 0 && cb.cb_first && !cb.cb_scripted) (void) fprintf(stderr, gettext("no datasets available\n")); return (ret); } /* * zfs rename [-fu] * zfs rename [-f] -p * zfs rename [-u] -r * * Renames the given dataset to another of the same type. * * The '-p' flag creates all the non-existing ancestors of the target first. * The '-u' flag prevents file systems from being remounted during rename. */ static int zfs_do_rename(int argc, char **argv) { zfs_handle_t *zhp; renameflags_t flags = { 0 }; int c; int ret = 0; int types; boolean_t parents = B_FALSE; /* check options */ while ((c = getopt(argc, argv, "pruf")) != -1) { switch (c) { case 'p': parents = B_TRUE; break; case 'r': flags.recursive = B_TRUE; break; case 'u': flags.nounmount = B_TRUE; break; case 'f': flags.forceunmount = B_TRUE; break; case '?': default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing source dataset " "argument\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing target dataset " "argument\n")); usage(B_FALSE); } if (argc > 2) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if (flags.recursive && parents) { (void) fprintf(stderr, gettext("-p and -r options are mutually " "exclusive\n")); usage(B_FALSE); } if (flags.nounmount && parents) { (void) fprintf(stderr, gettext("-u and -p options are mutually " "exclusive\n")); usage(B_FALSE); } if (flags.recursive && strchr(argv[0], '@') == 0) { (void) fprintf(stderr, gettext("source dataset for recursive " "rename must be a snapshot\n")); usage(B_FALSE); } if (flags.nounmount) types = ZFS_TYPE_FILESYSTEM; else if (parents) types = ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME; else types = ZFS_TYPE_DATASET; if ((zhp = zfs_open(g_zfs, argv[0], types)) == NULL) return (1); /* If we were asked and the name looks good, try to create ancestors. */ if (parents && zfs_name_valid(argv[1], zfs_get_type(zhp)) && zfs_create_ancestors(g_zfs, argv[1]) != 0) { zfs_close(zhp); return (1); } ret = (zfs_rename(zhp, argv[1], flags) != 0); zfs_close(zhp); return (ret); } /* * zfs promote * * Promotes the given clone fs to be the parent */ static int zfs_do_promote(int argc, char **argv) { zfs_handle_t *zhp; int ret = 0; /* check options */ if (argc > 1 && argv[1][0] == '-') { (void) fprintf(stderr, gettext("invalid option '%c'\n"), argv[1][1]); usage(B_FALSE); } /* check number of arguments */ if (argc < 2) { (void) fprintf(stderr, gettext("missing clone filesystem" " argument\n")); usage(B_FALSE); } if (argc > 2) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } zhp = zfs_open(g_zfs, argv[1], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) return (1); ret = (zfs_promote(zhp) != 0); zfs_close(zhp); return (ret); } static int zfs_do_redact(int argc, char **argv) { char *snap = NULL; char *bookname = NULL; char **rsnaps = NULL; int numrsnaps = 0; argv++; argc--; if (argc < 3) { (void) fprintf(stderr, gettext("too few arguments\n")); usage(B_FALSE); } snap = argv[0]; bookname = argv[1]; rsnaps = argv + 2; numrsnaps = argc - 2; nvlist_t *rsnapnv = fnvlist_alloc(); for (int i = 0; i < numrsnaps; i++) { fnvlist_add_boolean(rsnapnv, rsnaps[i]); } int err = lzc_redact(snap, bookname, rsnapnv); fnvlist_free(rsnapnv); switch (err) { case 0: break; case ENOENT: { zfs_handle_t *zhp = zfs_open(g_zfs, snap, ZFS_TYPE_SNAPSHOT); if (zhp == NULL) { (void) fprintf(stderr, gettext("provided snapshot %s " "does not exist\n"), snap); } else { zfs_close(zhp); } for (int i = 0; i < numrsnaps; i++) { zhp = zfs_open(g_zfs, rsnaps[i], ZFS_TYPE_SNAPSHOT); if (zhp == NULL) { (void) fprintf(stderr, gettext("provided " "snapshot %s does not exist\n"), rsnaps[i]); } else { zfs_close(zhp); } } break; } case EEXIST: (void) fprintf(stderr, gettext("specified redaction bookmark " "(%s) provided already exists\n"), bookname); break; case ENAMETOOLONG: (void) fprintf(stderr, gettext("provided bookmark name cannot " "be used, final name would be too long\n")); break; case E2BIG: (void) fprintf(stderr, gettext("too many redaction snapshots " "specified\n")); break; case EINVAL: if (strchr(bookname, '#') != NULL) (void) fprintf(stderr, gettext( "redaction bookmark name must not contain '#'\n")); else (void) fprintf(stderr, gettext( "redaction snapshot must be descendent of " "snapshot being redacted\n")); break; case EALREADY: (void) fprintf(stderr, gettext("attempted to redact redacted " "dataset or with respect to redacted dataset\n")); break; case ENOTSUP: (void) fprintf(stderr, gettext("redaction bookmarks feature " "not enabled\n")); break; case EXDEV: (void) fprintf(stderr, gettext("potentially invalid redaction " "snapshot; full dataset names required\n")); break; case ESRCH: (void) fprintf(stderr, gettext("attempted to resume redaction " " with a mismatched redaction list\n")); break; default: (void) fprintf(stderr, gettext("internal error: %s\n"), strerror(errno)); } return (err); } /* * zfs rollback [-rRf] * * -r Delete any intervening snapshots before doing rollback * -R Delete any snapshots and their clones * -f ignored for backwards compatibility * * Given a filesystem, rollback to a specific snapshot, discarding any changes * since then and making it the active dataset. If more recent snapshots exist, * the command will complain unless the '-r' flag is given. */ typedef struct rollback_cbdata { uint64_t cb_create; uint8_t cb_younger_ds_printed; boolean_t cb_first; int cb_doclones; char *cb_target; int cb_error; boolean_t cb_recurse; } rollback_cbdata_t; static int rollback_check_dependent(zfs_handle_t *zhp, void *data) { rollback_cbdata_t *cbp = data; if (cbp->cb_first && cbp->cb_recurse) { (void) fprintf(stderr, gettext("cannot rollback to " "'%s': clones of previous snapshots exist\n"), cbp->cb_target); (void) fprintf(stderr, gettext("use '-R' to " "force deletion of the following clones and " "dependents:\n")); cbp->cb_first = 0; cbp->cb_error = 1; } (void) fprintf(stderr, "%s\n", zfs_get_name(zhp)); zfs_close(zhp); return (0); } /* * Report some snapshots/bookmarks more recent than the one specified. * Used when '-r' is not specified. We reuse this same callback for the * snapshot dependents - if 'cb_dependent' is set, then this is a * dependent and we should report it without checking the transaction group. */ static int rollback_check(zfs_handle_t *zhp, void *data) { rollback_cbdata_t *cbp = data; /* * Max number of younger snapshots and/or bookmarks to display before * we stop the iteration. */ const uint8_t max_younger = 32; if (cbp->cb_doclones) { zfs_close(zhp); return (0); } if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) { if (cbp->cb_first && !cbp->cb_recurse) { (void) fprintf(stderr, gettext("cannot " "rollback to '%s': more recent snapshots " "or bookmarks exist\n"), cbp->cb_target); (void) fprintf(stderr, gettext("use '-r' to " "force deletion of the following " "snapshots and bookmarks:\n")); cbp->cb_first = 0; cbp->cb_error = 1; } if (cbp->cb_recurse) { if (zfs_iter_dependents_v2(zhp, 0, B_TRUE, rollback_check_dependent, cbp) != 0) { zfs_close(zhp); return (-1); } } else { (void) fprintf(stderr, "%s\n", zfs_get_name(zhp)); cbp->cb_younger_ds_printed++; } } zfs_close(zhp); if (cbp->cb_younger_ds_printed == max_younger) { /* * This non-recursive rollback is going to fail due to the * presence of snapshots and/or bookmarks that are younger than * the rollback target. * We printed some of the offending objects, now we stop * zfs_iter_snapshot/bookmark iteration so we can fail fast and * avoid iterating over the rest of the younger objects */ (void) fprintf(stderr, gettext("Output limited to %d " "snapshots/bookmarks\n"), max_younger); return (-1); } return (0); } static int zfs_do_rollback(int argc, char **argv) { int ret = 0; int c; boolean_t force = B_FALSE; rollback_cbdata_t cb = { 0 }; zfs_handle_t *zhp, *snap; char parentname[ZFS_MAX_DATASET_NAME_LEN]; char *delim; uint64_t min_txg = 0; /* check options */ while ((c = getopt(argc, argv, "rRf")) != -1) { switch (c) { case 'r': cb.cb_recurse = 1; break; case 'R': cb.cb_recurse = 1; cb.cb_doclones = 1; break; case 'f': force = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing dataset argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } /* open the snapshot */ if ((snap = zfs_open(g_zfs, argv[0], ZFS_TYPE_SNAPSHOT)) == NULL) return (1); /* open the parent dataset */ (void) strlcpy(parentname, argv[0], sizeof (parentname)); verify((delim = strrchr(parentname, '@')) != NULL); *delim = '\0'; if ((zhp = zfs_open(g_zfs, parentname, ZFS_TYPE_DATASET)) == NULL) { zfs_close(snap); return (1); } /* * Check for more recent snapshots and/or clones based on the presence * of '-r' and '-R'. */ cb.cb_target = argv[0]; cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG); cb.cb_first = B_TRUE; cb.cb_error = 0; if (cb.cb_create > 0) min_txg = cb.cb_create; if ((ret = zfs_iter_snapshots_v2(zhp, 0, rollback_check, &cb, min_txg, 0)) != 0) goto out; if ((ret = zfs_iter_bookmarks_v2(zhp, 0, rollback_check, &cb)) != 0) goto out; if ((ret = cb.cb_error) != 0) goto out; /* * Rollback parent to the given snapshot. */ ret = zfs_rollback(zhp, snap, force); out: zfs_close(snap); zfs_close(zhp); if (ret == 0) return (0); else return (1); } /* * zfs set property=value ... { fs | snap | vol } ... * * Sets the given properties for all datasets specified on the command line. */ static int set_callback(zfs_handle_t *zhp, void *data) { zprop_set_cbdata_t *cb = data; int ret = zfs_prop_set_list_flags(zhp, cb->cb_proplist, cb->cb_flags); if (ret != 0 || libzfs_errno(g_zfs) != EZFS_SUCCESS) { switch (libzfs_errno(g_zfs)) { case EZFS_MOUNTFAILED: (void) fprintf(stderr, gettext("property may be set " "but unable to remount filesystem\n")); break; case EZFS_SHARENFSFAILED: (void) fprintf(stderr, gettext("property may be set " "but unable to reshare filesystem\n")); break; } } return (ret); } static int zfs_do_set(int argc, char **argv) { zprop_set_cbdata_t cb = { 0 }; int ds_start = -1; /* argv idx of first dataset arg */ int ret = 0; int i, c; /* check options */ while ((c = getopt(argc, argv, "u")) != -1) { switch (c) { case 'u': cb.cb_flags |= ZFS_SET_NOMOUNT; break; case '?': default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing arguments\n")); usage(B_FALSE); } if (argc < 2) { if (strchr(argv[0], '=') == NULL) { (void) fprintf(stderr, gettext("missing property=value " "argument(s)\n")); } else { (void) fprintf(stderr, gettext("missing dataset " "name(s)\n")); } usage(B_FALSE); } /* validate argument order: prop=val args followed by dataset args */ for (i = 0; i < argc; i++) { if (strchr(argv[i], '=') != NULL) { if (ds_start > 0) { /* out-of-order prop=val argument */ (void) fprintf(stderr, gettext("invalid " "argument order\n")); usage(B_FALSE); } } else if (ds_start < 0) { ds_start = i; } } if (ds_start < 0) { (void) fprintf(stderr, gettext("missing dataset name(s)\n")); usage(B_FALSE); } /* Populate a list of property settings */ if (nvlist_alloc(&cb.cb_proplist, NV_UNIQUE_NAME, 0) != 0) nomem(); for (i = 0; i < ds_start; i++) { if (!parseprop(cb.cb_proplist, argv[i])) { ret = -1; goto error; } } ret = zfs_for_each(argc - ds_start, argv + ds_start, 0, ZFS_TYPE_DATASET, NULL, NULL, 0, set_callback, &cb); error: nvlist_free(cb.cb_proplist); return (ret); } typedef struct snap_cbdata { nvlist_t *sd_nvl; boolean_t sd_recursive; const char *sd_snapname; } snap_cbdata_t; static int zfs_snapshot_cb(zfs_handle_t *zhp, void *arg) { snap_cbdata_t *sd = arg; char *name; int rv = 0; int error; if (sd->sd_recursive && zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) != 0) { zfs_close(zhp); return (0); } error = asprintf(&name, "%s@%s", zfs_get_name(zhp), sd->sd_snapname); if (error == -1) nomem(); fnvlist_add_boolean(sd->sd_nvl, name); free(name); if (sd->sd_recursive) rv = zfs_iter_filesystems_v2(zhp, 0, zfs_snapshot_cb, sd); zfs_close(zhp); return (rv); } /* * zfs snapshot [-r] [-o prop=value] ... * * Creates a snapshot with the given name. While functionally equivalent to * 'zfs create', it is a separate command to differentiate intent. */ static int zfs_do_snapshot(int argc, char **argv) { int ret = 0; int c; nvlist_t *props; snap_cbdata_t sd = { 0 }; boolean_t multiple_snaps = B_FALSE; if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) nomem(); if (nvlist_alloc(&sd.sd_nvl, NV_UNIQUE_NAME, 0) != 0) nomem(); /* check options */ while ((c = getopt(argc, argv, "ro:")) != -1) { switch (c) { case 'o': if (!parseprop(props, optarg)) { nvlist_free(sd.sd_nvl); nvlist_free(props); return (1); } break; case 'r': sd.sd_recursive = B_TRUE; multiple_snaps = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); goto usage; } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing snapshot argument\n")); goto usage; } if (argc > 1) multiple_snaps = B_TRUE; for (; argc > 0; argc--, argv++) { char *atp; zfs_handle_t *zhp; atp = strchr(argv[0], '@'); if (atp == NULL) goto usage; *atp = '\0'; sd.sd_snapname = atp + 1; zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) goto usage; if (zfs_snapshot_cb(zhp, &sd) != 0) goto usage; } ret = zfs_snapshot_nvl(g_zfs, sd.sd_nvl, props); nvlist_free(sd.sd_nvl); nvlist_free(props); if (ret != 0 && multiple_snaps) (void) fprintf(stderr, gettext("no snapshots were created\n")); return (ret != 0); usage: nvlist_free(sd.sd_nvl); nvlist_free(props); usage(B_FALSE); return (-1); } /* * Array of prefixes to exclude – * a linear search, even if executed for each dataset, * is plenty good enough. */ typedef struct zfs_send_exclude_arg { size_t count; const char **list; } zfs_send_exclude_arg_t; static boolean_t zfs_do_send_exclude(zfs_handle_t *zhp, void *context) { zfs_send_exclude_arg_t *excludes = context; const char *name = zfs_get_name(zhp); for (size_t i = 0; i < excludes->count; ++i) { size_t len = strlen(excludes->list[i]); if (strncmp(name, excludes->list[i], len) == 0 && memchr("/@", name[len], sizeof ("/@"))) return (B_FALSE); } return (B_TRUE); } /* * Send a backup stream to stdout. */ static int zfs_do_send(int argc, char **argv) { char *fromname = NULL; char *toname = NULL; char *resume_token = NULL; char *cp; zfs_handle_t *zhp; sendflags_t flags = { 0 }; int c, err; nvlist_t *dbgnv = NULL; char *redactbook = NULL; zfs_send_exclude_arg_t excludes = { 0 }; struct option long_options[] = { {"replicate", no_argument, NULL, 'R'}, {"skip-missing", no_argument, NULL, 's'}, {"redact", required_argument, NULL, 'd'}, {"props", no_argument, NULL, 'p'}, {"parsable", no_argument, NULL, 'P'}, {"dedup", no_argument, NULL, 'D'}, {"proctitle", no_argument, NULL, 'V'}, {"verbose", no_argument, NULL, 'v'}, {"dryrun", no_argument, NULL, 'n'}, {"large-block", no_argument, NULL, 'L'}, {"embed", no_argument, NULL, 'e'}, {"resume", required_argument, NULL, 't'}, {"compressed", no_argument, NULL, 'c'}, {"raw", no_argument, NULL, 'w'}, {"backup", no_argument, NULL, 'b'}, {"holds", no_argument, NULL, 'h'}, {"saved", no_argument, NULL, 'S'}, {"exclude", required_argument, NULL, 'X'}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, ":i:I:RsDpVvnPLeht:cwbd:SX:", long_options, NULL)) != -1) { switch (c) { case 'X': for (char *ds; (ds = strsep(&optarg, ",")) != NULL; ) { if (!zfs_name_valid(ds, ZFS_TYPE_DATASET) || strchr(ds, '/') == NULL) { (void) fprintf(stderr, gettext("-X %s: " "not a valid non-root dataset name" ".\n"), ds); usage(B_FALSE); } excludes.list = safe_realloc(excludes.list, sizeof (char *) * (excludes.count + 1)); excludes.list[excludes.count++] = ds; } break; case 'i': if (fromname) usage(B_FALSE); fromname = optarg; break; case 'I': if (fromname) usage(B_FALSE); fromname = optarg; flags.doall = B_TRUE; break; case 'R': flags.replicate = B_TRUE; break; case 's': flags.skipmissing = B_TRUE; break; case 'd': redactbook = optarg; break; case 'p': flags.props = B_TRUE; break; case 'b': flags.backup = B_TRUE; break; case 'h': flags.holds = B_TRUE; break; case 'P': flags.parsable = B_TRUE; break; case 'V': flags.progressastitle = B_TRUE; break; case 'v': flags.verbosity++; flags.progress = B_TRUE; break; case 'D': (void) fprintf(stderr, gettext("WARNING: deduplicated send is no " "longer supported. A regular,\n" "non-deduplicated stream will be generated.\n\n")); break; case 'n': flags.dryrun = B_TRUE; break; case 'L': flags.largeblock = B_TRUE; break; case 'e': flags.embed_data = B_TRUE; break; case 't': resume_token = optarg; break; case 'c': flags.compress = B_TRUE; break; case 'w': flags.raw = B_TRUE; flags.compress = B_TRUE; flags.embed_data = B_TRUE; flags.largeblock = B_TRUE; break; case 'S': flags.saved = B_TRUE; break; case ':': /* * If a parameter was not passed, optopt contains the * value that would normally lead us into the * appropriate case statement. If it's > 256, then this * must be a longopt and we should look at argv to get * the string. Otherwise it's just the character, so we * should use it directly. */ if (optopt <= UINT8_MAX) { (void) fprintf(stderr, gettext("missing argument for '%c' " "option\n"), optopt); } else { (void) fprintf(stderr, gettext("missing argument for '%s' " "option\n"), argv[optind - 1]); } free(excludes.list); usage(B_FALSE); break; case '?': default: /* * If an invalid flag was passed, optopt contains the * character if it was a short flag, or 0 if it was a * longopt. */ if (optopt != 0) { (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); } else { (void) fprintf(stderr, gettext("invalid option '%s'\n"), argv[optind - 1]); } free(excludes.list); usage(B_FALSE); } } if ((flags.parsable || flags.progressastitle) && flags.verbosity == 0) flags.verbosity = 1; if (excludes.count > 0 && !flags.replicate) { free(excludes.list); (void) fprintf(stderr, gettext("Cannot specify " "dataset exclusion (-X) on a non-recursive " "send.\n")); return (1); } argc -= optind; argv += optind; if (resume_token != NULL) { if (fromname != NULL || flags.replicate || flags.props || flags.backup || flags.holds || flags.saved || redactbook != NULL) { free(excludes.list); (void) fprintf(stderr, gettext("invalid flags combined with -t\n")); usage(B_FALSE); } if (argc > 0) { free(excludes.list); (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } } else { if (argc < 1) { free(excludes.list); (void) fprintf(stderr, gettext("missing snapshot argument\n")); usage(B_FALSE); } if (argc > 1) { free(excludes.list); (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } } if (flags.saved) { if (fromname != NULL || flags.replicate || flags.props || flags.doall || flags.backup || flags.holds || flags.largeblock || flags.embed_data || flags.compress || flags.raw || redactbook != NULL) { free(excludes.list); (void) fprintf(stderr, gettext("incompatible flags " "combined with saved send flag\n")); usage(B_FALSE); } if (strchr(argv[0], '@') != NULL) { free(excludes.list); (void) fprintf(stderr, gettext("saved send must " "specify the dataset with partially-received " "state\n")); usage(B_FALSE); } } if (flags.raw && redactbook != NULL) { free(excludes.list); (void) fprintf(stderr, gettext("Error: raw sends may not be redacted.\n")); return (1); } if (!flags.dryrun && isatty(STDOUT_FILENO)) { free(excludes.list); (void) fprintf(stderr, gettext("Error: Stream can not be written to a terminal.\n" "You must redirect standard output.\n")); return (1); } if (flags.saved) { zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_DATASET); if (zhp == NULL) { free(excludes.list); return (1); } err = zfs_send_saved(zhp, &flags, STDOUT_FILENO, resume_token); free(excludes.list); zfs_close(zhp); return (err != 0); } else if (resume_token != NULL) { free(excludes.list); return (zfs_send_resume(g_zfs, &flags, STDOUT_FILENO, resume_token)); } if (flags.skipmissing && !flags.replicate) { free(excludes.list); (void) fprintf(stderr, gettext("skip-missing flag can only be used in " "conjunction with replicate\n")); usage(B_FALSE); } /* * For everything except -R and -I, use the new, cleaner code path. */ if (!(flags.replicate || flags.doall)) { char frombuf[ZFS_MAX_DATASET_NAME_LEN]; if (fromname != NULL && (strchr(fromname, '#') == NULL && strchr(fromname, '@') == NULL)) { /* * Neither bookmark or snapshot was specified. Print a * warning, and assume snapshot. */ (void) fprintf(stderr, "Warning: incremental source " "didn't specify type, assuming snapshot. Use '@' " "or '#' prefix to avoid ambiguity.\n"); (void) snprintf(frombuf, sizeof (frombuf), "@%s", fromname); fromname = frombuf; } if (fromname != NULL && (fromname[0] == '#' || fromname[0] == '@')) { /* * Incremental source name begins with # or @. * Default to same fs as target. */ char tmpbuf[ZFS_MAX_DATASET_NAME_LEN]; (void) strlcpy(tmpbuf, fromname, sizeof (tmpbuf)); (void) strlcpy(frombuf, argv[0], sizeof (frombuf)); cp = strchr(frombuf, '@'); if (cp != NULL) *cp = '\0'; (void) strlcat(frombuf, tmpbuf, sizeof (frombuf)); fromname = frombuf; } zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_DATASET); if (zhp == NULL) { free(excludes.list); return (1); } err = zfs_send_one(zhp, fromname, STDOUT_FILENO, &flags, redactbook); free(excludes.list); zfs_close(zhp); return (err != 0); } if (fromname != NULL && strchr(fromname, '#')) { (void) fprintf(stderr, gettext("Error: multiple snapshots cannot be " "sent from a bookmark.\n")); free(excludes.list); return (1); } if (redactbook != NULL) { (void) fprintf(stderr, gettext("Error: multiple snapshots " "cannot be sent redacted.\n")); free(excludes.list); return (1); } if ((cp = strchr(argv[0], '@')) == NULL) { (void) fprintf(stderr, gettext("Error: " "Unsupported flag with filesystem or bookmark.\n")); free(excludes.list); return (1); } *cp = '\0'; toname = cp + 1; zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) { free(excludes.list); return (1); } /* * If they specified the full path to the snapshot, chop off * everything except the short name of the snapshot, but special * case if they specify the origin. */ if (fromname && (cp = strchr(fromname, '@')) != NULL) { char origin[ZFS_MAX_DATASET_NAME_LEN]; zprop_source_t src; (void) zfs_prop_get(zhp, ZFS_PROP_ORIGIN, origin, sizeof (origin), &src, NULL, 0, B_FALSE); if (strcmp(origin, fromname) == 0) { fromname = NULL; flags.fromorigin = B_TRUE; } else { *cp = '\0'; if (cp != fromname && strcmp(argv[0], fromname)) { zfs_close(zhp); free(excludes.list); (void) fprintf(stderr, gettext("incremental source must be " "in same filesystem\n")); usage(B_FALSE); } fromname = cp + 1; if (strchr(fromname, '@') || strchr(fromname, '/')) { zfs_close(zhp); free(excludes.list); (void) fprintf(stderr, gettext("invalid incremental source\n")); usage(B_FALSE); } } } if (flags.replicate && fromname == NULL) flags.doall = B_TRUE; err = zfs_send(zhp, fromname, toname, &flags, STDOUT_FILENO, excludes.count > 0 ? zfs_do_send_exclude : NULL, &excludes, flags.verbosity >= 3 ? &dbgnv : NULL); if (flags.verbosity >= 3 && dbgnv != NULL) { /* * dump_nvlist prints to stdout, but that's been * redirected to a file. Make it print to stderr * instead. */ (void) dup2(STDERR_FILENO, STDOUT_FILENO); dump_nvlist(dbgnv, 0); nvlist_free(dbgnv); } zfs_close(zhp); free(excludes.list); return (err != 0); } /* * Restore a backup stream from stdin. */ static int zfs_do_receive(int argc, char **argv) { int c, err = 0; recvflags_t flags = { 0 }; boolean_t abort_resumable = B_FALSE; nvlist_t *props; if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) nomem(); /* check options */ while ((c = getopt(argc, argv, ":o:x:dehMnuvFsAc")) != -1) { switch (c) { case 'o': if (!parseprop(props, optarg)) { nvlist_free(props); usage(B_FALSE); } break; case 'x': if (!parsepropname(props, optarg)) { nvlist_free(props); usage(B_FALSE); } break; case 'd': if (flags.istail) { (void) fprintf(stderr, gettext("invalid option " "combination: -d and -e are mutually " "exclusive\n")); usage(B_FALSE); } flags.isprefix = B_TRUE; break; case 'e': if (flags.isprefix) { (void) fprintf(stderr, gettext("invalid option " "combination: -d and -e are mutually " "exclusive\n")); usage(B_FALSE); } flags.istail = B_TRUE; break; case 'h': flags.skipholds = B_TRUE; break; case 'M': flags.forceunmount = B_TRUE; break; case 'n': flags.dryrun = B_TRUE; break; case 'u': flags.nomount = B_TRUE; break; case 'v': flags.verbose = B_TRUE; break; case 's': flags.resumable = B_TRUE; break; case 'F': flags.force = B_TRUE; break; case 'A': abort_resumable = B_TRUE; break; case 'c': flags.heal = B_TRUE; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* zfs recv -e (use "tail" name) implies -d (remove dataset "head") */ if (flags.istail) flags.isprefix = B_TRUE; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing snapshot argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if (abort_resumable) { if (flags.isprefix || flags.istail || flags.dryrun || flags.resumable || flags.nomount) { (void) fprintf(stderr, gettext("invalid option\n")); usage(B_FALSE); } char namebuf[ZFS_MAX_DATASET_NAME_LEN]; (void) snprintf(namebuf, sizeof (namebuf), "%s/%%recv", argv[0]); if (zfs_dataset_exists(g_zfs, namebuf, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) { zfs_handle_t *zhp = zfs_open(g_zfs, namebuf, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) { nvlist_free(props); return (1); } err = zfs_destroy(zhp, B_FALSE); zfs_close(zhp); } else { zfs_handle_t *zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) usage(B_FALSE); if (!zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) || zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN, NULL, 0, NULL, NULL, 0, B_TRUE) == -1) { (void) fprintf(stderr, gettext("'%s' does not have any " "resumable receive state to abort\n"), argv[0]); nvlist_free(props); zfs_close(zhp); return (1); } err = zfs_destroy(zhp, B_FALSE); zfs_close(zhp); } nvlist_free(props); return (err != 0); } if (isatty(STDIN_FILENO)) { (void) fprintf(stderr, gettext("Error: Backup stream can not be read " "from a terminal.\n" "You must redirect standard input.\n")); nvlist_free(props); return (1); } err = zfs_receive(g_zfs, argv[0], props, &flags, STDIN_FILENO, NULL); nvlist_free(props); return (err != 0); } /* * allow/unallow stuff */ /* copied from zfs/sys/dsl_deleg.h */ #define ZFS_DELEG_PERM_CREATE "create" #define ZFS_DELEG_PERM_DESTROY "destroy" #define ZFS_DELEG_PERM_SNAPSHOT "snapshot" #define ZFS_DELEG_PERM_ROLLBACK "rollback" #define ZFS_DELEG_PERM_CLONE "clone" #define ZFS_DELEG_PERM_PROMOTE "promote" #define ZFS_DELEG_PERM_RENAME "rename" #define ZFS_DELEG_PERM_MOUNT "mount" #define ZFS_DELEG_PERM_SHARE "share" #define ZFS_DELEG_PERM_SEND "send" #define ZFS_DELEG_PERM_RECEIVE "receive" #define ZFS_DELEG_PERM_ALLOW "allow" #define ZFS_DELEG_PERM_USERPROP "userprop" #define ZFS_DELEG_PERM_VSCAN "vscan" /* ??? */ #define ZFS_DELEG_PERM_USERQUOTA "userquota" #define ZFS_DELEG_PERM_GROUPQUOTA "groupquota" #define ZFS_DELEG_PERM_USERUSED "userused" #define ZFS_DELEG_PERM_GROUPUSED "groupused" #define ZFS_DELEG_PERM_USEROBJQUOTA "userobjquota" #define ZFS_DELEG_PERM_GROUPOBJQUOTA "groupobjquota" #define ZFS_DELEG_PERM_USEROBJUSED "userobjused" #define ZFS_DELEG_PERM_GROUPOBJUSED "groupobjused" #define ZFS_DELEG_PERM_HOLD "hold" #define ZFS_DELEG_PERM_RELEASE "release" #define ZFS_DELEG_PERM_DIFF "diff" #define ZFS_DELEG_PERM_BOOKMARK "bookmark" #define ZFS_DELEG_PERM_LOAD_KEY "load-key" #define ZFS_DELEG_PERM_CHANGE_KEY "change-key" #define ZFS_DELEG_PERM_PROJECTUSED "projectused" #define ZFS_DELEG_PERM_PROJECTQUOTA "projectquota" #define ZFS_DELEG_PERM_PROJECTOBJUSED "projectobjused" #define ZFS_DELEG_PERM_PROJECTOBJQUOTA "projectobjquota" #define ZFS_NUM_DELEG_NOTES ZFS_DELEG_NOTE_NONE static zfs_deleg_perm_tab_t zfs_deleg_perm_tbl[] = { { ZFS_DELEG_PERM_ALLOW, ZFS_DELEG_NOTE_ALLOW }, { ZFS_DELEG_PERM_CLONE, ZFS_DELEG_NOTE_CLONE }, { ZFS_DELEG_PERM_CREATE, ZFS_DELEG_NOTE_CREATE }, { ZFS_DELEG_PERM_DESTROY, ZFS_DELEG_NOTE_DESTROY }, { ZFS_DELEG_PERM_DIFF, ZFS_DELEG_NOTE_DIFF}, { ZFS_DELEG_PERM_HOLD, ZFS_DELEG_NOTE_HOLD }, { ZFS_DELEG_PERM_MOUNT, ZFS_DELEG_NOTE_MOUNT }, { ZFS_DELEG_PERM_PROMOTE, ZFS_DELEG_NOTE_PROMOTE }, { ZFS_DELEG_PERM_RECEIVE, ZFS_DELEG_NOTE_RECEIVE }, { ZFS_DELEG_PERM_RELEASE, ZFS_DELEG_NOTE_RELEASE }, { ZFS_DELEG_PERM_RENAME, ZFS_DELEG_NOTE_RENAME }, { ZFS_DELEG_PERM_ROLLBACK, ZFS_DELEG_NOTE_ROLLBACK }, { ZFS_DELEG_PERM_SEND, ZFS_DELEG_NOTE_SEND }, { ZFS_DELEG_PERM_SHARE, ZFS_DELEG_NOTE_SHARE }, { ZFS_DELEG_PERM_SNAPSHOT, ZFS_DELEG_NOTE_SNAPSHOT }, { ZFS_DELEG_PERM_BOOKMARK, ZFS_DELEG_NOTE_BOOKMARK }, { ZFS_DELEG_PERM_LOAD_KEY, ZFS_DELEG_NOTE_LOAD_KEY }, { ZFS_DELEG_PERM_CHANGE_KEY, ZFS_DELEG_NOTE_CHANGE_KEY }, { ZFS_DELEG_PERM_GROUPQUOTA, ZFS_DELEG_NOTE_GROUPQUOTA }, { ZFS_DELEG_PERM_GROUPUSED, ZFS_DELEG_NOTE_GROUPUSED }, { ZFS_DELEG_PERM_USERPROP, ZFS_DELEG_NOTE_USERPROP }, { ZFS_DELEG_PERM_USERQUOTA, ZFS_DELEG_NOTE_USERQUOTA }, { ZFS_DELEG_PERM_USERUSED, ZFS_DELEG_NOTE_USERUSED }, { ZFS_DELEG_PERM_USEROBJQUOTA, ZFS_DELEG_NOTE_USEROBJQUOTA }, { ZFS_DELEG_PERM_USEROBJUSED, ZFS_DELEG_NOTE_USEROBJUSED }, { ZFS_DELEG_PERM_GROUPOBJQUOTA, ZFS_DELEG_NOTE_GROUPOBJQUOTA }, { ZFS_DELEG_PERM_GROUPOBJUSED, ZFS_DELEG_NOTE_GROUPOBJUSED }, { ZFS_DELEG_PERM_PROJECTUSED, ZFS_DELEG_NOTE_PROJECTUSED }, { ZFS_DELEG_PERM_PROJECTQUOTA, ZFS_DELEG_NOTE_PROJECTQUOTA }, { ZFS_DELEG_PERM_PROJECTOBJUSED, ZFS_DELEG_NOTE_PROJECTOBJUSED }, { ZFS_DELEG_PERM_PROJECTOBJQUOTA, ZFS_DELEG_NOTE_PROJECTOBJQUOTA }, { NULL, ZFS_DELEG_NOTE_NONE } }; /* permission structure */ typedef struct deleg_perm { zfs_deleg_who_type_t dp_who_type; const char *dp_name; boolean_t dp_local; boolean_t dp_descend; } deleg_perm_t; /* */ typedef struct deleg_perm_node { deleg_perm_t dpn_perm; uu_avl_node_t dpn_avl_node; } deleg_perm_node_t; typedef struct fs_perm fs_perm_t; /* permissions set */ typedef struct who_perm { zfs_deleg_who_type_t who_type; const char *who_name; /* id */ char who_ug_name[256]; /* user/group name */ fs_perm_t *who_fsperm; /* uplink */ uu_avl_t *who_deleg_perm_avl; /* permissions */ } who_perm_t; /* */ typedef struct who_perm_node { who_perm_t who_perm; uu_avl_node_t who_avl_node; } who_perm_node_t; typedef struct fs_perm_set fs_perm_set_t; /* fs permissions */ struct fs_perm { const char *fsp_name; uu_avl_t *fsp_sc_avl; /* sets,create */ uu_avl_t *fsp_uge_avl; /* user,group,everyone */ fs_perm_set_t *fsp_set; /* uplink */ }; /* */ typedef struct fs_perm_node { fs_perm_t fspn_fsperm; uu_avl_t *fspn_avl; uu_list_node_t fspn_list_node; } fs_perm_node_t; /* top level structure */ struct fs_perm_set { uu_list_pool_t *fsps_list_pool; uu_list_t *fsps_list; /* list of fs_perms */ uu_avl_pool_t *fsps_named_set_avl_pool; uu_avl_pool_t *fsps_who_perm_avl_pool; uu_avl_pool_t *fsps_deleg_perm_avl_pool; }; static inline const char * deleg_perm_type(zfs_deleg_note_t note) { /* subcommands */ switch (note) { /* SUBCOMMANDS */ /* OTHER */ case ZFS_DELEG_NOTE_GROUPQUOTA: case ZFS_DELEG_NOTE_GROUPUSED: case ZFS_DELEG_NOTE_USERPROP: case ZFS_DELEG_NOTE_USERQUOTA: case ZFS_DELEG_NOTE_USERUSED: case ZFS_DELEG_NOTE_USEROBJQUOTA: case ZFS_DELEG_NOTE_USEROBJUSED: case ZFS_DELEG_NOTE_GROUPOBJQUOTA: case ZFS_DELEG_NOTE_GROUPOBJUSED: case ZFS_DELEG_NOTE_PROJECTUSED: case ZFS_DELEG_NOTE_PROJECTQUOTA: case ZFS_DELEG_NOTE_PROJECTOBJUSED: case ZFS_DELEG_NOTE_PROJECTOBJQUOTA: /* other */ return (gettext("other")); default: return (gettext("subcommand")); } } static int who_type2weight(zfs_deleg_who_type_t who_type) { int res; switch (who_type) { case ZFS_DELEG_NAMED_SET_SETS: case ZFS_DELEG_NAMED_SET: res = 0; break; case ZFS_DELEG_CREATE_SETS: case ZFS_DELEG_CREATE: res = 1; break; case ZFS_DELEG_USER_SETS: case ZFS_DELEG_USER: res = 2; break; case ZFS_DELEG_GROUP_SETS: case ZFS_DELEG_GROUP: res = 3; break; case ZFS_DELEG_EVERYONE_SETS: case ZFS_DELEG_EVERYONE: res = 4; break; default: res = -1; } return (res); } static int who_perm_compare(const void *larg, const void *rarg, void *unused) { (void) unused; const who_perm_node_t *l = larg; const who_perm_node_t *r = rarg; zfs_deleg_who_type_t ltype = l->who_perm.who_type; zfs_deleg_who_type_t rtype = r->who_perm.who_type; int lweight = who_type2weight(ltype); int rweight = who_type2weight(rtype); int res = lweight - rweight; if (res == 0) res = strncmp(l->who_perm.who_name, r->who_perm.who_name, ZFS_MAX_DELEG_NAME-1); if (res == 0) return (0); if (res > 0) return (1); else return (-1); } static int deleg_perm_compare(const void *larg, const void *rarg, void *unused) { (void) unused; const deleg_perm_node_t *l = larg; const deleg_perm_node_t *r = rarg; int res = strncmp(l->dpn_perm.dp_name, r->dpn_perm.dp_name, ZFS_MAX_DELEG_NAME-1); if (res == 0) return (0); if (res > 0) return (1); else return (-1); } static inline void fs_perm_set_init(fs_perm_set_t *fspset) { memset(fspset, 0, sizeof (fs_perm_set_t)); if ((fspset->fsps_list_pool = uu_list_pool_create("fsps_list_pool", sizeof (fs_perm_node_t), offsetof(fs_perm_node_t, fspn_list_node), NULL, UU_DEFAULT)) == NULL) nomem(); if ((fspset->fsps_list = uu_list_create(fspset->fsps_list_pool, NULL, UU_DEFAULT)) == NULL) nomem(); if ((fspset->fsps_named_set_avl_pool = uu_avl_pool_create( "named_set_avl_pool", sizeof (who_perm_node_t), offsetof( who_perm_node_t, who_avl_node), who_perm_compare, UU_DEFAULT)) == NULL) nomem(); if ((fspset->fsps_who_perm_avl_pool = uu_avl_pool_create( "who_perm_avl_pool", sizeof (who_perm_node_t), offsetof( who_perm_node_t, who_avl_node), who_perm_compare, UU_DEFAULT)) == NULL) nomem(); if ((fspset->fsps_deleg_perm_avl_pool = uu_avl_pool_create( "deleg_perm_avl_pool", sizeof (deleg_perm_node_t), offsetof( deleg_perm_node_t, dpn_avl_node), deleg_perm_compare, UU_DEFAULT)) == NULL) nomem(); } static inline void fs_perm_fini(fs_perm_t *); static inline void who_perm_fini(who_perm_t *); static inline void fs_perm_set_fini(fs_perm_set_t *fspset) { fs_perm_node_t *node = uu_list_first(fspset->fsps_list); while (node != NULL) { fs_perm_node_t *next_node = uu_list_next(fspset->fsps_list, node); fs_perm_t *fsperm = &node->fspn_fsperm; fs_perm_fini(fsperm); uu_list_remove(fspset->fsps_list, node); free(node); node = next_node; } uu_avl_pool_destroy(fspset->fsps_named_set_avl_pool); uu_avl_pool_destroy(fspset->fsps_who_perm_avl_pool); uu_avl_pool_destroy(fspset->fsps_deleg_perm_avl_pool); } static inline void deleg_perm_init(deleg_perm_t *deleg_perm, zfs_deleg_who_type_t type, const char *name) { deleg_perm->dp_who_type = type; deleg_perm->dp_name = name; } static inline void who_perm_init(who_perm_t *who_perm, fs_perm_t *fsperm, zfs_deleg_who_type_t type, const char *name) { uu_avl_pool_t *pool; pool = fsperm->fsp_set->fsps_deleg_perm_avl_pool; memset(who_perm, 0, sizeof (who_perm_t)); if ((who_perm->who_deleg_perm_avl = uu_avl_create(pool, NULL, UU_DEFAULT)) == NULL) nomem(); who_perm->who_type = type; who_perm->who_name = name; who_perm->who_fsperm = fsperm; } static inline void who_perm_fini(who_perm_t *who_perm) { deleg_perm_node_t *node = uu_avl_first(who_perm->who_deleg_perm_avl); while (node != NULL) { deleg_perm_node_t *next_node = uu_avl_next(who_perm->who_deleg_perm_avl, node); uu_avl_remove(who_perm->who_deleg_perm_avl, node); free(node); node = next_node; } uu_avl_destroy(who_perm->who_deleg_perm_avl); } static inline void fs_perm_init(fs_perm_t *fsperm, fs_perm_set_t *fspset, const char *fsname) { uu_avl_pool_t *nset_pool = fspset->fsps_named_set_avl_pool; uu_avl_pool_t *who_pool = fspset->fsps_who_perm_avl_pool; memset(fsperm, 0, sizeof (fs_perm_t)); if ((fsperm->fsp_sc_avl = uu_avl_create(nset_pool, NULL, UU_DEFAULT)) == NULL) nomem(); if ((fsperm->fsp_uge_avl = uu_avl_create(who_pool, NULL, UU_DEFAULT)) == NULL) nomem(); fsperm->fsp_set = fspset; fsperm->fsp_name = fsname; } static inline void fs_perm_fini(fs_perm_t *fsperm) { who_perm_node_t *node = uu_avl_first(fsperm->fsp_sc_avl); while (node != NULL) { who_perm_node_t *next_node = uu_avl_next(fsperm->fsp_sc_avl, node); who_perm_t *who_perm = &node->who_perm; who_perm_fini(who_perm); uu_avl_remove(fsperm->fsp_sc_avl, node); free(node); node = next_node; } node = uu_avl_first(fsperm->fsp_uge_avl); while (node != NULL) { who_perm_node_t *next_node = uu_avl_next(fsperm->fsp_uge_avl, node); who_perm_t *who_perm = &node->who_perm; who_perm_fini(who_perm); uu_avl_remove(fsperm->fsp_uge_avl, node); free(node); node = next_node; } uu_avl_destroy(fsperm->fsp_sc_avl); uu_avl_destroy(fsperm->fsp_uge_avl); } static void set_deleg_perm_node(uu_avl_t *avl, deleg_perm_node_t *node, zfs_deleg_who_type_t who_type, const char *name, char locality) { uu_avl_index_t idx = 0; deleg_perm_node_t *found_node = NULL; deleg_perm_t *deleg_perm = &node->dpn_perm; deleg_perm_init(deleg_perm, who_type, name); if ((found_node = uu_avl_find(avl, node, NULL, &idx)) == NULL) uu_avl_insert(avl, node, idx); else { node = found_node; deleg_perm = &node->dpn_perm; } switch (locality) { case ZFS_DELEG_LOCAL: deleg_perm->dp_local = B_TRUE; break; case ZFS_DELEG_DESCENDENT: deleg_perm->dp_descend = B_TRUE; break; case ZFS_DELEG_NA: break; default: assert(B_FALSE); /* invalid locality */ } } static inline int parse_who_perm(who_perm_t *who_perm, nvlist_t *nvl, char locality) { nvpair_t *nvp = NULL; fs_perm_set_t *fspset = who_perm->who_fsperm->fsp_set; uu_avl_t *avl = who_perm->who_deleg_perm_avl; zfs_deleg_who_type_t who_type = who_perm->who_type; while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) { const char *name = nvpair_name(nvp); data_type_t type = nvpair_type(nvp); uu_avl_pool_t *avl_pool = fspset->fsps_deleg_perm_avl_pool; deleg_perm_node_t *node = safe_malloc(sizeof (deleg_perm_node_t)); VERIFY(type == DATA_TYPE_BOOLEAN); uu_avl_node_init(node, &node->dpn_avl_node, avl_pool); set_deleg_perm_node(avl, node, who_type, name, locality); } return (0); } static inline int parse_fs_perm(fs_perm_t *fsperm, nvlist_t *nvl) { nvpair_t *nvp = NULL; fs_perm_set_t *fspset = fsperm->fsp_set; while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) { nvlist_t *nvl2 = NULL; const char *name = nvpair_name(nvp); uu_avl_t *avl = NULL; uu_avl_pool_t *avl_pool = NULL; zfs_deleg_who_type_t perm_type = name[0]; char perm_locality = name[1]; const char *perm_name = name + 3; who_perm_t *who_perm = NULL; assert('$' == name[2]); if (nvpair_value_nvlist(nvp, &nvl2) != 0) return (-1); switch (perm_type) { case ZFS_DELEG_CREATE: case ZFS_DELEG_CREATE_SETS: case ZFS_DELEG_NAMED_SET: case ZFS_DELEG_NAMED_SET_SETS: avl_pool = fspset->fsps_named_set_avl_pool; avl = fsperm->fsp_sc_avl; break; case ZFS_DELEG_USER: case ZFS_DELEG_USER_SETS: case ZFS_DELEG_GROUP: case ZFS_DELEG_GROUP_SETS: case ZFS_DELEG_EVERYONE: case ZFS_DELEG_EVERYONE_SETS: avl_pool = fspset->fsps_who_perm_avl_pool; avl = fsperm->fsp_uge_avl; break; default: assert(!"unhandled zfs_deleg_who_type_t"); } who_perm_node_t *found_node = NULL; who_perm_node_t *node = safe_malloc( sizeof (who_perm_node_t)); who_perm = &node->who_perm; uu_avl_index_t idx = 0; uu_avl_node_init(node, &node->who_avl_node, avl_pool); who_perm_init(who_perm, fsperm, perm_type, perm_name); if ((found_node = uu_avl_find(avl, node, NULL, &idx)) == NULL) { if (avl == fsperm->fsp_uge_avl) { uid_t rid = 0; struct passwd *p = NULL; struct group *g = NULL; const char *nice_name = NULL; switch (perm_type) { case ZFS_DELEG_USER_SETS: case ZFS_DELEG_USER: rid = atoi(perm_name); p = getpwuid(rid); if (p) nice_name = p->pw_name; break; case ZFS_DELEG_GROUP_SETS: case ZFS_DELEG_GROUP: rid = atoi(perm_name); g = getgrgid(rid); if (g) nice_name = g->gr_name; break; default: break; } if (nice_name != NULL) { (void) strlcpy( node->who_perm.who_ug_name, nice_name, 256); } else { /* User or group unknown */ (void) snprintf( node->who_perm.who_ug_name, sizeof (node->who_perm.who_ug_name), "(unknown: %d)", rid); } } uu_avl_insert(avl, node, idx); } else { node = found_node; who_perm = &node->who_perm; } assert(who_perm != NULL); (void) parse_who_perm(who_perm, nvl2, perm_locality); } return (0); } static inline int parse_fs_perm_set(fs_perm_set_t *fspset, nvlist_t *nvl) { nvpair_t *nvp = NULL; uu_avl_index_t idx = 0; while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) { nvlist_t *nvl2 = NULL; const char *fsname = nvpair_name(nvp); data_type_t type = nvpair_type(nvp); fs_perm_t *fsperm = NULL; fs_perm_node_t *node = safe_malloc(sizeof (fs_perm_node_t)); fsperm = &node->fspn_fsperm; VERIFY(DATA_TYPE_NVLIST == type); uu_list_node_init(node, &node->fspn_list_node, fspset->fsps_list_pool); idx = uu_list_numnodes(fspset->fsps_list); fs_perm_init(fsperm, fspset, fsname); if (nvpair_value_nvlist(nvp, &nvl2) != 0) return (-1); (void) parse_fs_perm(fsperm, nvl2); uu_list_insert(fspset->fsps_list, node, idx); } return (0); } static inline const char * deleg_perm_comment(zfs_deleg_note_t note) { const char *str = ""; /* subcommands */ switch (note) { /* SUBCOMMANDS */ case ZFS_DELEG_NOTE_ALLOW: str = gettext("Must also have the permission that is being" "\n\t\t\t\tallowed"); break; case ZFS_DELEG_NOTE_CLONE: str = gettext("Must also have the 'create' ability and 'mount'" "\n\t\t\t\tability in the origin file system"); break; case ZFS_DELEG_NOTE_CREATE: str = gettext("Must also have the 'mount' ability"); break; case ZFS_DELEG_NOTE_DESTROY: str = gettext("Must also have the 'mount' ability"); break; case ZFS_DELEG_NOTE_DIFF: str = gettext("Allows lookup of paths within a dataset;" "\n\t\t\t\tgiven an object number. Ordinary users need this" "\n\t\t\t\tin order to use zfs diff"); break; case ZFS_DELEG_NOTE_HOLD: str = gettext("Allows adding a user hold to a snapshot"); break; case ZFS_DELEG_NOTE_MOUNT: str = gettext("Allows mount/umount of ZFS datasets"); break; case ZFS_DELEG_NOTE_PROMOTE: str = gettext("Must also have the 'mount'\n\t\t\t\tand" " 'promote' ability in the origin file system"); break; case ZFS_DELEG_NOTE_RECEIVE: str = gettext("Must also have the 'mount' and 'create'" " ability"); break; case ZFS_DELEG_NOTE_RELEASE: str = gettext("Allows releasing a user hold which\n\t\t\t\t" "might destroy the snapshot"); break; case ZFS_DELEG_NOTE_RENAME: str = gettext("Must also have the 'mount' and 'create'" "\n\t\t\t\tability in the new parent"); break; case ZFS_DELEG_NOTE_ROLLBACK: str = gettext(""); break; case ZFS_DELEG_NOTE_SEND: str = gettext(""); break; case ZFS_DELEG_NOTE_SHARE: str = gettext("Allows sharing file systems over NFS or SMB" "\n\t\t\t\tprotocols"); break; case ZFS_DELEG_NOTE_SNAPSHOT: str = gettext(""); break; case ZFS_DELEG_NOTE_LOAD_KEY: str = gettext("Allows loading or unloading an encryption key"); break; case ZFS_DELEG_NOTE_CHANGE_KEY: str = gettext("Allows changing or adding an encryption key"); break; /* * case ZFS_DELEG_NOTE_VSCAN: * str = gettext(""); * break; */ /* OTHER */ case ZFS_DELEG_NOTE_GROUPQUOTA: str = gettext("Allows accessing any groupquota@... property"); break; case ZFS_DELEG_NOTE_GROUPUSED: str = gettext("Allows reading any groupused@... property"); break; case ZFS_DELEG_NOTE_USERPROP: str = gettext("Allows changing any user property"); break; case ZFS_DELEG_NOTE_USERQUOTA: str = gettext("Allows accessing any userquota@... property"); break; case ZFS_DELEG_NOTE_USERUSED: str = gettext("Allows reading any userused@... property"); break; case ZFS_DELEG_NOTE_USEROBJQUOTA: str = gettext("Allows accessing any userobjquota@... property"); break; case ZFS_DELEG_NOTE_GROUPOBJQUOTA: str = gettext("Allows accessing any \n\t\t\t\t" "groupobjquota@... property"); break; case ZFS_DELEG_NOTE_GROUPOBJUSED: str = gettext("Allows reading any groupobjused@... property"); break; case ZFS_DELEG_NOTE_USEROBJUSED: str = gettext("Allows reading any userobjused@... property"); break; case ZFS_DELEG_NOTE_PROJECTQUOTA: str = gettext("Allows accessing any projectquota@... property"); break; case ZFS_DELEG_NOTE_PROJECTOBJQUOTA: str = gettext("Allows accessing any \n\t\t\t\t" "projectobjquota@... property"); break; case ZFS_DELEG_NOTE_PROJECTUSED: str = gettext("Allows reading any projectused@... property"); break; case ZFS_DELEG_NOTE_PROJECTOBJUSED: str = gettext("Allows accessing any \n\t\t\t\t" "projectobjused@... property"); break; /* other */ default: str = ""; } return (str); } struct allow_opts { boolean_t local; boolean_t descend; boolean_t user; boolean_t group; boolean_t everyone; boolean_t create; boolean_t set; boolean_t recursive; /* unallow only */ boolean_t prt_usage; boolean_t prt_perms; char *who; char *perms; const char *dataset; }; static inline int prop_cmp(const void *a, const void *b) { const char *str1 = *(const char **)a; const char *str2 = *(const char **)b; return (strcmp(str1, str2)); } static void allow_usage(boolean_t un, boolean_t requested, const char *msg) { const char *opt_desc[] = { "-h", gettext("show this help message and exit"), "-l", gettext("set permission locally"), "-d", gettext("set permission for descents"), "-u", gettext("set permission for user"), "-g", gettext("set permission for group"), "-e", gettext("set permission for everyone"), "-c", gettext("set create time permission"), "-s", gettext("define permission set"), /* unallow only */ "-r", gettext("remove permissions recursively"), }; size_t unallow_size = sizeof (opt_desc) / sizeof (char *); size_t allow_size = unallow_size - 2; const char *props[ZFS_NUM_PROPS]; int i; size_t count = 0; FILE *fp = requested ? stdout : stderr; zprop_desc_t *pdtbl = zfs_prop_get_table(); const char *fmt = gettext("%-16s %-14s\t%s\n"); (void) fprintf(fp, gettext("Usage: %s\n"), get_usage(un ? HELP_UNALLOW : HELP_ALLOW)); (void) fprintf(fp, gettext("Options:\n")); for (i = 0; i < (un ? unallow_size : allow_size); i += 2) { const char *opt = opt_desc[i]; const char *optdsc = opt_desc[i + 1]; (void) fprintf(fp, gettext(" %-10s %s\n"), opt, optdsc); } (void) fprintf(fp, gettext("\nThe following permissions are " "supported:\n\n")); (void) fprintf(fp, fmt, gettext("NAME"), gettext("TYPE"), gettext("NOTES")); for (i = 0; i < ZFS_NUM_DELEG_NOTES; i++) { const char *perm_name = zfs_deleg_perm_tbl[i].z_perm; zfs_deleg_note_t perm_note = zfs_deleg_perm_tbl[i].z_note; const char *perm_type = deleg_perm_type(perm_note); const char *perm_comment = deleg_perm_comment(perm_note); (void) fprintf(fp, fmt, perm_name, perm_type, perm_comment); } for (i = 0; i < ZFS_NUM_PROPS; i++) { zprop_desc_t *pd = &pdtbl[i]; if (pd->pd_visible != B_TRUE) continue; if (pd->pd_attr == PROP_READONLY) continue; props[count++] = pd->pd_name; } props[count] = NULL; qsort(props, count, sizeof (char *), prop_cmp); for (i = 0; i < count; i++) (void) fprintf(fp, fmt, props[i], gettext("property"), ""); if (msg != NULL) (void) fprintf(fp, gettext("\nzfs: error: %s"), msg); exit(requested ? 0 : 2); } static inline const char * munge_args(int argc, char **argv, boolean_t un, size_t expected_argc, char **permsp) { if (un && argc == expected_argc - 1) *permsp = NULL; else if (argc == expected_argc) *permsp = argv[argc - 2]; else allow_usage(un, B_FALSE, gettext("wrong number of parameters\n")); return (argv[argc - 1]); } static void parse_allow_args(int argc, char **argv, boolean_t un, struct allow_opts *opts) { int uge_sum = opts->user + opts->group + opts->everyone; int csuge_sum = opts->create + opts->set + uge_sum; int ldcsuge_sum = csuge_sum + opts->local + opts->descend; int all_sum = un ? ldcsuge_sum + opts->recursive : ldcsuge_sum; if (uge_sum > 1) allow_usage(un, B_FALSE, gettext("-u, -g, and -e are mutually exclusive\n")); if (opts->prt_usage) { if (argc == 0 && all_sum == 0) allow_usage(un, B_TRUE, NULL); else usage(B_FALSE); } if (opts->set) { if (csuge_sum > 1) allow_usage(un, B_FALSE, gettext("invalid options combined with -s\n")); opts->dataset = munge_args(argc, argv, un, 3, &opts->perms); if (argv[0][0] != '@') allow_usage(un, B_FALSE, gettext("invalid set name: missing '@' prefix\n")); opts->who = argv[0]; } else if (opts->create) { if (ldcsuge_sum > 1) allow_usage(un, B_FALSE, gettext("invalid options combined with -c\n")); opts->dataset = munge_args(argc, argv, un, 2, &opts->perms); } else if (opts->everyone) { if (csuge_sum > 1) allow_usage(un, B_FALSE, gettext("invalid options combined with -e\n")); opts->dataset = munge_args(argc, argv, un, 2, &opts->perms); } else if (uge_sum == 0 && argc > 0 && strcmp(argv[0], "everyone") == 0) { opts->everyone = B_TRUE; argc--; argv++; opts->dataset = munge_args(argc, argv, un, 2, &opts->perms); } else if (argc == 1 && !un) { opts->prt_perms = B_TRUE; opts->dataset = argv[argc-1]; } else { opts->dataset = munge_args(argc, argv, un, 3, &opts->perms); opts->who = argv[0]; } if (!opts->local && !opts->descend) { opts->local = B_TRUE; opts->descend = B_TRUE; } } static void store_allow_perm(zfs_deleg_who_type_t type, boolean_t local, boolean_t descend, const char *who, char *perms, nvlist_t *top_nvl) { int i; char ld[2] = { '\0', '\0' }; char who_buf[MAXNAMELEN + 32]; char base_type = '\0'; char set_type = '\0'; nvlist_t *base_nvl = NULL; nvlist_t *set_nvl = NULL; nvlist_t *nvl; if (nvlist_alloc(&base_nvl, NV_UNIQUE_NAME, 0) != 0) nomem(); if (nvlist_alloc(&set_nvl, NV_UNIQUE_NAME, 0) != 0) nomem(); switch (type) { case ZFS_DELEG_NAMED_SET_SETS: case ZFS_DELEG_NAMED_SET: set_type = ZFS_DELEG_NAMED_SET_SETS; base_type = ZFS_DELEG_NAMED_SET; ld[0] = ZFS_DELEG_NA; break; case ZFS_DELEG_CREATE_SETS: case ZFS_DELEG_CREATE: set_type = ZFS_DELEG_CREATE_SETS; base_type = ZFS_DELEG_CREATE; ld[0] = ZFS_DELEG_NA; break; case ZFS_DELEG_USER_SETS: case ZFS_DELEG_USER: set_type = ZFS_DELEG_USER_SETS; base_type = ZFS_DELEG_USER; if (local) ld[0] = ZFS_DELEG_LOCAL; if (descend) ld[1] = ZFS_DELEG_DESCENDENT; break; case ZFS_DELEG_GROUP_SETS: case ZFS_DELEG_GROUP: set_type = ZFS_DELEG_GROUP_SETS; base_type = ZFS_DELEG_GROUP; if (local) ld[0] = ZFS_DELEG_LOCAL; if (descend) ld[1] = ZFS_DELEG_DESCENDENT; break; case ZFS_DELEG_EVERYONE_SETS: case ZFS_DELEG_EVERYONE: set_type = ZFS_DELEG_EVERYONE_SETS; base_type = ZFS_DELEG_EVERYONE; if (local) ld[0] = ZFS_DELEG_LOCAL; if (descend) ld[1] = ZFS_DELEG_DESCENDENT; break; default: assert(set_type != '\0' && base_type != '\0'); } if (perms != NULL) { char *curr = perms; char *end = curr + strlen(perms); while (curr < end) { char *delim = strchr(curr, ','); if (delim == NULL) delim = end; else *delim = '\0'; if (curr[0] == '@') nvl = set_nvl; else nvl = base_nvl; (void) nvlist_add_boolean(nvl, curr); if (delim != end) *delim = ','; curr = delim + 1; } for (i = 0; i < 2; i++) { char locality = ld[i]; if (locality == 0) continue; if (!nvlist_empty(base_nvl)) { if (who != NULL) (void) snprintf(who_buf, sizeof (who_buf), "%c%c$%s", base_type, locality, who); else (void) snprintf(who_buf, sizeof (who_buf), "%c%c$", base_type, locality); (void) nvlist_add_nvlist(top_nvl, who_buf, base_nvl); } if (!nvlist_empty(set_nvl)) { if (who != NULL) (void) snprintf(who_buf, sizeof (who_buf), "%c%c$%s", set_type, locality, who); else (void) snprintf(who_buf, sizeof (who_buf), "%c%c$", set_type, locality); (void) nvlist_add_nvlist(top_nvl, who_buf, set_nvl); } } } else { for (i = 0; i < 2; i++) { char locality = ld[i]; if (locality == 0) continue; if (who != NULL) (void) snprintf(who_buf, sizeof (who_buf), "%c%c$%s", base_type, locality, who); else (void) snprintf(who_buf, sizeof (who_buf), "%c%c$", base_type, locality); (void) nvlist_add_boolean(top_nvl, who_buf); if (who != NULL) (void) snprintf(who_buf, sizeof (who_buf), "%c%c$%s", set_type, locality, who); else (void) snprintf(who_buf, sizeof (who_buf), "%c%c$", set_type, locality); (void) nvlist_add_boolean(top_nvl, who_buf); } } } static int construct_fsacl_list(boolean_t un, struct allow_opts *opts, nvlist_t **nvlp) { if (nvlist_alloc(nvlp, NV_UNIQUE_NAME, 0) != 0) nomem(); if (opts->set) { store_allow_perm(ZFS_DELEG_NAMED_SET, opts->local, opts->descend, opts->who, opts->perms, *nvlp); } else if (opts->create) { store_allow_perm(ZFS_DELEG_CREATE, opts->local, opts->descend, NULL, opts->perms, *nvlp); } else if (opts->everyone) { store_allow_perm(ZFS_DELEG_EVERYONE, opts->local, opts->descend, NULL, opts->perms, *nvlp); } else { char *curr = opts->who; char *end = curr + strlen(curr); while (curr < end) { const char *who; zfs_deleg_who_type_t who_type = ZFS_DELEG_WHO_UNKNOWN; char *endch; char *delim = strchr(curr, ','); char errbuf[256]; char id[64]; struct passwd *p = NULL; struct group *g = NULL; uid_t rid; if (delim == NULL) delim = end; else *delim = '\0'; rid = (uid_t)strtol(curr, &endch, 0); if (opts->user) { who_type = ZFS_DELEG_USER; if (*endch != '\0') p = getpwnam(curr); else p = getpwuid(rid); if (p != NULL) rid = p->pw_uid; else if (*endch != '\0') { (void) snprintf(errbuf, sizeof (errbuf), gettext("invalid user %s\n"), curr); allow_usage(un, B_TRUE, errbuf); } } else if (opts->group) { who_type = ZFS_DELEG_GROUP; if (*endch != '\0') g = getgrnam(curr); else g = getgrgid(rid); if (g != NULL) rid = g->gr_gid; else if (*endch != '\0') { (void) snprintf(errbuf, sizeof (errbuf), gettext("invalid group %s\n"), curr); allow_usage(un, B_TRUE, errbuf); } } else { if (*endch != '\0') { p = getpwnam(curr); } else { p = getpwuid(rid); } if (p == NULL) { if (*endch != '\0') { g = getgrnam(curr); } else { g = getgrgid(rid); } } if (p != NULL) { who_type = ZFS_DELEG_USER; rid = p->pw_uid; } else if (g != NULL) { who_type = ZFS_DELEG_GROUP; rid = g->gr_gid; } else { (void) snprintf(errbuf, sizeof (errbuf), gettext("invalid user/group %s\n"), curr); allow_usage(un, B_TRUE, errbuf); } } (void) sprintf(id, "%u", rid); who = id; store_allow_perm(who_type, opts->local, opts->descend, who, opts->perms, *nvlp); curr = delim + 1; } } return (0); } static void print_set_creat_perms(uu_avl_t *who_avl) { const char *sc_title[] = { gettext("Permission sets:\n"), gettext("Create time permissions:\n"), NULL }; who_perm_node_t *who_node = NULL; int prev_weight = -1; for (who_node = uu_avl_first(who_avl); who_node != NULL; who_node = uu_avl_next(who_avl, who_node)) { uu_avl_t *avl = who_node->who_perm.who_deleg_perm_avl; zfs_deleg_who_type_t who_type = who_node->who_perm.who_type; const char *who_name = who_node->who_perm.who_name; int weight = who_type2weight(who_type); boolean_t first = B_TRUE; deleg_perm_node_t *deleg_node; if (prev_weight != weight) { (void) printf("%s", sc_title[weight]); prev_weight = weight; } if (who_name == NULL || strnlen(who_name, 1) == 0) (void) printf("\t"); else (void) printf("\t%s ", who_name); for (deleg_node = uu_avl_first(avl); deleg_node != NULL; deleg_node = uu_avl_next(avl, deleg_node)) { if (first) { (void) printf("%s", deleg_node->dpn_perm.dp_name); first = B_FALSE; } else (void) printf(",%s", deleg_node->dpn_perm.dp_name); } (void) printf("\n"); } } static void print_uge_deleg_perms(uu_avl_t *who_avl, boolean_t local, boolean_t descend, const char *title) { who_perm_node_t *who_node = NULL; boolean_t prt_title = B_TRUE; uu_avl_walk_t *walk; if ((walk = uu_avl_walk_start(who_avl, UU_WALK_ROBUST)) == NULL) nomem(); while ((who_node = uu_avl_walk_next(walk)) != NULL) { const char *who_name = who_node->who_perm.who_name; const char *nice_who_name = who_node->who_perm.who_ug_name; uu_avl_t *avl = who_node->who_perm.who_deleg_perm_avl; zfs_deleg_who_type_t who_type = who_node->who_perm.who_type; char delim = ' '; deleg_perm_node_t *deleg_node; boolean_t prt_who = B_TRUE; for (deleg_node = uu_avl_first(avl); deleg_node != NULL; deleg_node = uu_avl_next(avl, deleg_node)) { if (local != deleg_node->dpn_perm.dp_local || descend != deleg_node->dpn_perm.dp_descend) continue; if (prt_who) { const char *who = NULL; if (prt_title) { prt_title = B_FALSE; (void) printf("%s", title); } switch (who_type) { case ZFS_DELEG_USER_SETS: case ZFS_DELEG_USER: who = gettext("user"); if (nice_who_name) who_name = nice_who_name; break; case ZFS_DELEG_GROUP_SETS: case ZFS_DELEG_GROUP: who = gettext("group"); if (nice_who_name) who_name = nice_who_name; break; case ZFS_DELEG_EVERYONE_SETS: case ZFS_DELEG_EVERYONE: who = gettext("everyone"); who_name = NULL; break; default: assert(who != NULL); } prt_who = B_FALSE; if (who_name == NULL) (void) printf("\t%s", who); else (void) printf("\t%s %s", who, who_name); } (void) printf("%c%s", delim, deleg_node->dpn_perm.dp_name); delim = ','; } if (!prt_who) (void) printf("\n"); } uu_avl_walk_end(walk); } static void print_fs_perms(fs_perm_set_t *fspset) { fs_perm_node_t *node = NULL; char buf[MAXNAMELEN + 32]; const char *dsname = buf; for (node = uu_list_first(fspset->fsps_list); node != NULL; node = uu_list_next(fspset->fsps_list, node)) { uu_avl_t *sc_avl = node->fspn_fsperm.fsp_sc_avl; uu_avl_t *uge_avl = node->fspn_fsperm.fsp_uge_avl; int left = 0; (void) snprintf(buf, sizeof (buf), gettext("---- Permissions on %s "), node->fspn_fsperm.fsp_name); (void) printf("%s", dsname); left = 70 - strlen(buf); while (left-- > 0) (void) printf("-"); (void) printf("\n"); print_set_creat_perms(sc_avl); print_uge_deleg_perms(uge_avl, B_TRUE, B_FALSE, gettext("Local permissions:\n")); print_uge_deleg_perms(uge_avl, B_FALSE, B_TRUE, gettext("Descendent permissions:\n")); print_uge_deleg_perms(uge_avl, B_TRUE, B_TRUE, gettext("Local+Descendent permissions:\n")); } } static fs_perm_set_t fs_perm_set = { NULL, NULL, NULL, NULL }; struct deleg_perms { boolean_t un; nvlist_t *nvl; }; static int set_deleg_perms(zfs_handle_t *zhp, void *data) { struct deleg_perms *perms = (struct deleg_perms *)data; zfs_type_t zfs_type = zfs_get_type(zhp); if (zfs_type != ZFS_TYPE_FILESYSTEM && zfs_type != ZFS_TYPE_VOLUME) return (0); return (zfs_set_fsacl(zhp, perms->un, perms->nvl)); } static int zfs_do_allow_unallow_impl(int argc, char **argv, boolean_t un) { zfs_handle_t *zhp; nvlist_t *perm_nvl = NULL; nvlist_t *update_perm_nvl = NULL; int error = 1; int c; struct allow_opts opts = { 0 }; const char *optstr = un ? "ldugecsrh" : "ldugecsh"; /* check opts */ while ((c = getopt(argc, argv, optstr)) != -1) { switch (c) { case 'l': opts.local = B_TRUE; break; case 'd': opts.descend = B_TRUE; break; case 'u': opts.user = B_TRUE; break; case 'g': opts.group = B_TRUE; break; case 'e': opts.everyone = B_TRUE; break; case 's': opts.set = B_TRUE; break; case 'c': opts.create = B_TRUE; break; case 'r': opts.recursive = B_TRUE; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case 'h': opts.prt_usage = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check arguments */ parse_allow_args(argc, argv, un, &opts); /* try to open the dataset */ if ((zhp = zfs_open(g_zfs, opts.dataset, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) == NULL) { (void) fprintf(stderr, "Failed to open dataset: %s\n", opts.dataset); return (-1); } if (zfs_get_fsacl(zhp, &perm_nvl) != 0) goto cleanup2; fs_perm_set_init(&fs_perm_set); if (parse_fs_perm_set(&fs_perm_set, perm_nvl) != 0) { (void) fprintf(stderr, "Failed to parse fsacl permissions\n"); goto cleanup1; } if (opts.prt_perms) print_fs_perms(&fs_perm_set); else { (void) construct_fsacl_list(un, &opts, &update_perm_nvl); if (zfs_set_fsacl(zhp, un, update_perm_nvl) != 0) goto cleanup0; if (un && opts.recursive) { struct deleg_perms data = { un, update_perm_nvl }; if (zfs_iter_filesystems_v2(zhp, 0, set_deleg_perms, &data) != 0) goto cleanup0; } } error = 0; cleanup0: nvlist_free(perm_nvl); nvlist_free(update_perm_nvl); cleanup1: fs_perm_set_fini(&fs_perm_set); cleanup2: zfs_close(zhp); return (error); } static int zfs_do_allow(int argc, char **argv) { return (zfs_do_allow_unallow_impl(argc, argv, B_FALSE)); } static int zfs_do_unallow(int argc, char **argv) { return (zfs_do_allow_unallow_impl(argc, argv, B_TRUE)); } static int zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding) { int errors = 0; int i; const char *tag; boolean_t recursive = B_FALSE; const char *opts = holding ? "rt" : "r"; int c; /* check options */ while ((c = getopt(argc, argv, opts)) != -1) { switch (c) { case 'r': recursive = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 2) usage(B_FALSE); tag = argv[0]; --argc; ++argv; if (holding && tag[0] == '.') { /* tags starting with '.' are reserved for libzfs */ (void) fprintf(stderr, gettext("tag may not start with '.'\n")); usage(B_FALSE); } for (i = 0; i < argc; ++i) { zfs_handle_t *zhp; char parent[ZFS_MAX_DATASET_NAME_LEN]; const char *delim; char *path = argv[i]; delim = strchr(path, '@'); if (delim == NULL) { (void) fprintf(stderr, gettext("'%s' is not a snapshot\n"), path); ++errors; continue; } (void) strlcpy(parent, path, MIN(sizeof (parent), delim - path + 1)); zhp = zfs_open(g_zfs, parent, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) { ++errors; continue; } if (holding) { if (zfs_hold(zhp, delim+1, tag, recursive, -1) != 0) ++errors; } else { if (zfs_release(zhp, delim+1, tag, recursive) != 0) ++errors; } zfs_close(zhp); } return (errors != 0); } /* * zfs hold [-r] [-t] ... * * -r Recursively hold * * Apply a user-hold with the given tag to the list of snapshots. */ static int zfs_do_hold(int argc, char **argv) { return (zfs_do_hold_rele_impl(argc, argv, B_TRUE)); } /* * zfs release [-r] ... * * -r Recursively release * * Release a user-hold with the given tag from the list of snapshots. */ static int zfs_do_release(int argc, char **argv) { return (zfs_do_hold_rele_impl(argc, argv, B_FALSE)); } typedef struct holds_cbdata { boolean_t cb_recursive; const char *cb_snapname; nvlist_t **cb_nvlp; size_t cb_max_namelen; size_t cb_max_taglen; } holds_cbdata_t; #define STRFTIME_FMT_STR "%a %b %e %H:%M %Y" #define DATETIME_BUF_LEN (32) /* * */ static void print_holds(boolean_t scripted, int nwidth, int tagwidth, nvlist_t *nvl, boolean_t parsable) { int i; nvpair_t *nvp = NULL; const char *const hdr_cols[] = { "NAME", "TAG", "TIMESTAMP" }; const char *col; if (!scripted) { for (i = 0; i < 3; i++) { col = gettext(hdr_cols[i]); if (i < 2) (void) printf("%-*s ", i ? tagwidth : nwidth, col); else (void) printf("%s\n", col); } } while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) { const char *zname = nvpair_name(nvp); nvlist_t *nvl2; nvpair_t *nvp2 = NULL; (void) nvpair_value_nvlist(nvp, &nvl2); while ((nvp2 = nvlist_next_nvpair(nvl2, nvp2)) != NULL) { char tsbuf[DATETIME_BUF_LEN]; const char *tagname = nvpair_name(nvp2); uint64_t val = 0; time_t time; struct tm t; (void) nvpair_value_uint64(nvp2, &val); time = (time_t)val; (void) localtime_r(&time, &t); (void) strftime(tsbuf, DATETIME_BUF_LEN, gettext(STRFTIME_FMT_STR), &t); if (scripted) { if (parsable) { (void) printf("%s\t%s\t%ld\n", zname, tagname, (unsigned long)time); } else { (void) printf("%s\t%s\t%s\n", zname, tagname, tsbuf); } } else { if (parsable) { (void) printf("%-*s %-*s %ld\n", nwidth, zname, tagwidth, tagname, (unsigned long)time); } else { (void) printf("%-*s %-*s %s\n", nwidth, zname, tagwidth, tagname, tsbuf); } } } } } /* * Generic callback function to list a dataset or snapshot. */ static int holds_callback(zfs_handle_t *zhp, void *data) { holds_cbdata_t *cbp = data; nvlist_t *top_nvl = *cbp->cb_nvlp; nvlist_t *nvl = NULL; nvpair_t *nvp = NULL; const char *zname = zfs_get_name(zhp); size_t znamelen = strlen(zname); if (cbp->cb_recursive) { const char *snapname; char *delim = strchr(zname, '@'); if (delim == NULL) return (0); snapname = delim + 1; if (strcmp(cbp->cb_snapname, snapname)) return (0); } if (zfs_get_holds(zhp, &nvl) != 0) return (-1); if (znamelen > cbp->cb_max_namelen) cbp->cb_max_namelen = znamelen; while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) { const char *tag = nvpair_name(nvp); size_t taglen = strlen(tag); if (taglen > cbp->cb_max_taglen) cbp->cb_max_taglen = taglen; } return (nvlist_add_nvlist(top_nvl, zname, nvl)); } /* * zfs holds [-rHp] ... * * -r Lists holds that are set on the named snapshots recursively. * -H Scripted mode; elide headers and separate columns by tabs. * -p Display values in parsable (literal) format. */ static int zfs_do_holds(int argc, char **argv) { int c; boolean_t errors = B_FALSE; boolean_t scripted = B_FALSE; boolean_t recursive = B_FALSE; boolean_t parsable = B_FALSE; int types = ZFS_TYPE_SNAPSHOT; holds_cbdata_t cb = { 0 }; int limit = 0; int ret = 0; int flags = 0; /* check options */ while ((c = getopt(argc, argv, "rHp")) != -1) { switch (c) { case 'r': recursive = B_TRUE; break; case 'H': scripted = B_TRUE; break; case 'p': parsable = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (recursive) { types |= ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME; flags |= ZFS_ITER_RECURSE; } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) usage(B_FALSE); nvlist_t *nvl = fnvlist_alloc(); for (int i = 0; i < argc; ++i) { char *snapshot = argv[i]; const char *delim; const char *snapname; delim = strchr(snapshot, '@'); if (delim == NULL) { (void) fprintf(stderr, gettext("'%s' is not a snapshot\n"), snapshot); errors = B_TRUE; continue; } snapname = delim + 1; if (recursive) snapshot[delim - snapshot] = '\0'; cb.cb_recursive = recursive; cb.cb_snapname = snapname; cb.cb_nvlp = &nvl; /* * 1. collect holds data, set format options */ ret = zfs_for_each(1, argv + i, flags, types, NULL, NULL, limit, holds_callback, &cb); if (ret != 0) errors = B_TRUE; } /* * 2. print holds data */ print_holds(scripted, cb.cb_max_namelen, cb.cb_max_taglen, nvl, parsable); if (nvlist_empty(nvl)) (void) fprintf(stderr, gettext("no datasets available\n")); nvlist_free(nvl); return (errors); } #define CHECK_SPINNER 30 #define SPINNER_TIME 3 /* seconds */ #define MOUNT_TIME 1 /* seconds */ typedef struct get_all_state { char **ga_datasets; int ga_count; boolean_t ga_verbose; get_all_cb_t *ga_cbp; } get_all_state_t; static int get_one_dataset(zfs_handle_t *zhp, void *data) { static const char *const spin[] = { "-", "\\", "|", "/" }; static int spinval = 0; static int spincheck = 0; static time_t last_spin_time = (time_t)0; get_all_state_t *state = data; zfs_type_t type = zfs_get_type(zhp); if (state->ga_verbose) { if (--spincheck < 0) { time_t now = time(NULL); if (last_spin_time + SPINNER_TIME < now) { update_progress(spin[spinval++ % 4]); last_spin_time = now; } spincheck = CHECK_SPINNER; } } /* * Iterate over any nested datasets. */ if (zfs_iter_filesystems_v2(zhp, 0, get_one_dataset, data) != 0) { zfs_close(zhp); return (1); } /* * Skip any datasets whose type does not match. */ if ((type & ZFS_TYPE_FILESYSTEM) == 0) { zfs_close(zhp); return (0); } libzfs_add_handle(state->ga_cbp, zhp); assert(state->ga_cbp->cb_used <= state->ga_cbp->cb_alloc); return (0); } static int get_recursive_datasets(zfs_handle_t *zhp, void *data) { get_all_state_t *state = data; int len = strlen(zfs_get_name(zhp)); for (int i = 0; i < state->ga_count; ++i) { if (strcmp(state->ga_datasets[i], zfs_get_name(zhp)) == 0) return (get_one_dataset(zhp, data)); else if ((strncmp(state->ga_datasets[i], zfs_get_name(zhp), len) == 0) && state->ga_datasets[i][len] == '/') { (void) zfs_iter_filesystems_v2(zhp, 0, get_recursive_datasets, data); } } zfs_close(zhp); return (0); } static void get_all_datasets(get_all_state_t *state) { if (state->ga_verbose) set_progress_header(gettext("Reading ZFS config")); if (state->ga_datasets == NULL) (void) zfs_iter_root(g_zfs, get_one_dataset, state); else (void) zfs_iter_root(g_zfs, get_recursive_datasets, state); if (state->ga_verbose) finish_progress(gettext("done.")); } /* * Generic callback for sharing or mounting filesystems. Because the code is so * similar, we have a common function with an extra parameter to determine which * mode we are using. */ typedef enum { OP_SHARE, OP_MOUNT } share_mount_op_t; typedef struct share_mount_state { share_mount_op_t sm_op; boolean_t sm_verbose; int sm_flags; char *sm_options; enum sa_protocol sm_proto; /* only valid for OP_SHARE */ pthread_mutex_t sm_lock; /* protects the remaining fields */ uint_t sm_total; /* number of filesystems to process */ uint_t sm_done; /* number of filesystems processed */ int sm_status; /* -1 if any of the share/mount operations failed */ } share_mount_state_t; /* * Share or mount a dataset. */ static int share_mount_one(zfs_handle_t *zhp, int op, int flags, enum sa_protocol protocol, boolean_t explicit, const char *options) { char mountpoint[ZFS_MAXPROPLEN]; char shareopts[ZFS_MAXPROPLEN]; char smbshareopts[ZFS_MAXPROPLEN]; const char *cmdname = op == OP_SHARE ? "share" : "mount"; struct mnttab mnt; uint64_t zoned, canmount; boolean_t shared_nfs, shared_smb; assert(zfs_get_type(zhp) & ZFS_TYPE_FILESYSTEM); /* * Check to make sure we can mount/share this dataset. If we * are in the global zone and the filesystem is exported to a * local zone, or if we are in a local zone and the * filesystem is not exported, then it is an error. */ zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED); if (zoned && getzoneid() == GLOBAL_ZONEID) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "dataset is exported to a local zone\n"), cmdname, zfs_get_name(zhp)); return (1); } else if (!zoned && getzoneid() != GLOBAL_ZONEID) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "permission denied\n"), cmdname, zfs_get_name(zhp)); return (1); } /* * Ignore any filesystems which don't apply to us. This * includes those with a legacy mountpoint, or those with * legacy share options. */ verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mountpoint, sizeof (mountpoint), NULL, NULL, 0, B_FALSE) == 0); verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, shareopts, sizeof (shareopts), NULL, NULL, 0, B_FALSE) == 0); verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, smbshareopts, sizeof (smbshareopts), NULL, NULL, 0, B_FALSE) == 0); if (op == OP_SHARE && strcmp(shareopts, "off") == 0 && strcmp(smbshareopts, "off") == 0) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot share '%s': " "legacy share\n"), zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use exports(5) or " "smb.conf(5) to share this filesystem, or set " "the sharenfs or sharesmb property\n")); return (1); } /* * We cannot share or mount legacy filesystems. If the * shareopts is non-legacy but the mountpoint is legacy, we * treat it as a legacy share. */ if (strcmp(mountpoint, "legacy") == 0) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "legacy mountpoint\n"), cmdname, zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use %s(8) to " "%s this filesystem\n"), cmdname, cmdname); return (1); } if (strcmp(mountpoint, "none") == 0) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': no " "mountpoint set\n"), cmdname, zfs_get_name(zhp)); return (1); } /* * canmount explicit outcome * on no pass through * on yes pass through * off no return 0 * off yes display error, return 1 * noauto no return 0 * noauto yes pass through */ canmount = zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT); if (canmount == ZFS_CANMOUNT_OFF) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "'canmount' property is set to 'off'\n"), cmdname, zfs_get_name(zhp)); return (1); } else if (canmount == ZFS_CANMOUNT_NOAUTO && !explicit) { /* * When performing a 'zfs mount -a', we skip any mounts for * datasets that have 'noauto' set. Sharing a dataset with * 'noauto' set is only allowed if it's mounted. */ if (op == OP_MOUNT) return (0); if (op == OP_SHARE && !zfs_is_mounted(zhp, NULL)) { /* also purge it from existing exports */ zfs_unshare(zhp, mountpoint, NULL); return (0); } } /* * If this filesystem is encrypted and does not have * a loaded key, we can not mount it. */ if ((flags & MS_CRYPT) == 0 && zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF && zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS) == ZFS_KEYSTATUS_UNAVAILABLE) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "encryption key not loaded\n"), cmdname, zfs_get_name(zhp)); return (1); } /* * If this filesystem is inconsistent and has a receive resume * token, we can not mount it. */ if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) && zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN, NULL, 0, NULL, NULL, 0, B_TRUE) == 0) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "Contains partially-completed state from " "\"zfs receive -s\", which can be resumed with " "\"zfs send -t\"\n"), cmdname, zfs_get_name(zhp)); return (1); } if (zfs_prop_get_int(zhp, ZFS_PROP_REDACTED) && !(flags & MS_FORCE)) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot %s '%s': " "Dataset is not complete, was created by receiving " "a redacted zfs send stream.\n"), cmdname, zfs_get_name(zhp)); return (1); } /* * At this point, we have verified that the mountpoint and/or * shareopts are appropriate for auto management. If the * filesystem is already mounted or shared, return (failing * for explicit requests); otherwise mount or share the * filesystem. */ switch (op) { case OP_SHARE: { enum sa_protocol prot[] = {SA_PROTOCOL_NFS, SA_NO_PROTOCOL}; shared_nfs = zfs_is_shared(zhp, NULL, prot); *prot = SA_PROTOCOL_SMB; shared_smb = zfs_is_shared(zhp, NULL, prot); if ((shared_nfs && shared_smb) || (shared_nfs && strcmp(shareopts, "on") == 0 && strcmp(smbshareopts, "off") == 0) || (shared_smb && strcmp(smbshareopts, "on") == 0 && strcmp(shareopts, "off") == 0)) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot share " "'%s': filesystem already shared\n"), zfs_get_name(zhp)); return (1); } if (!zfs_is_mounted(zhp, NULL) && zfs_mount(zhp, NULL, flags) != 0) return (1); *prot = protocol; if (zfs_share(zhp, protocol == SA_NO_PROTOCOL ? NULL : prot)) return (1); } break; case OP_MOUNT: mnt.mnt_mntopts = (char *)(options ?: ""); if (!hasmntopt(&mnt, MNTOPT_REMOUNT) && zfs_is_mounted(zhp, NULL)) { if (!explicit) return (0); (void) fprintf(stderr, gettext("cannot mount " "'%s': filesystem already mounted\n"), zfs_get_name(zhp)); return (1); } if (zfs_mount(zhp, options, flags) != 0) return (1); break; } return (0); } /* * Reports progress in the form "(current/total)". Not thread-safe. */ static void report_mount_progress(int current, int total) { static time_t last_progress_time = 0; time_t now = time(NULL); char info[32]; /* display header if we're here for the first time */ if (current == 1) { set_progress_header(gettext("Mounting ZFS filesystems")); } else if (current != total && last_progress_time + MOUNT_TIME >= now) { /* too soon to report again */ return; } last_progress_time = now; (void) sprintf(info, "(%d/%d)", current, total); if (current == total) finish_progress(info); else update_progress(info); } /* * zfs_foreach_mountpoint() callback that mounts or shares one filesystem and * updates the progress meter. */ static int share_mount_one_cb(zfs_handle_t *zhp, void *arg) { share_mount_state_t *sms = arg; int ret; ret = share_mount_one(zhp, sms->sm_op, sms->sm_flags, sms->sm_proto, B_FALSE, sms->sm_options); pthread_mutex_lock(&sms->sm_lock); if (ret != 0) sms->sm_status = ret; sms->sm_done++; if (sms->sm_verbose) report_mount_progress(sms->sm_done, sms->sm_total); pthread_mutex_unlock(&sms->sm_lock); return (ret); } static void append_options(char *mntopts, char *newopts) { int len = strlen(mntopts); /* original length plus new string to append plus 1 for the comma */ if (len + 1 + strlen(newopts) >= MNT_LINE_MAX) { (void) fprintf(stderr, gettext("the opts argument for " "'%s' option is too long (more than %d chars)\n"), "-o", MNT_LINE_MAX); usage(B_FALSE); } if (*mntopts) mntopts[len++] = ','; (void) strcpy(&mntopts[len], newopts); } static enum sa_protocol sa_protocol_decode(const char *protocol) { for (enum sa_protocol i = 0; i < ARRAY_SIZE(sa_protocol_names); ++i) if (strcmp(protocol, sa_protocol_names[i]) == 0) return (i); (void) fputs(gettext("share type must be one of: "), stderr); for (enum sa_protocol i = 0; i < ARRAY_SIZE(sa_protocol_names); ++i) (void) fprintf(stderr, "%s%s", i != 0 ? ", " : "", sa_protocol_names[i]); (void) fputc('\n', stderr); usage(B_FALSE); } static int share_mount(int op, int argc, char **argv) { int do_all = 0; int recursive = 0; boolean_t verbose = B_FALSE; int c, ret = 0; char *options = NULL; int flags = 0; /* check options */ while ((c = getopt(argc, argv, op == OP_MOUNT ? ":aRlvo:Of" : "al")) != -1) { switch (c) { case 'a': do_all = 1; break; case 'R': recursive = 1; break; case 'v': verbose = B_TRUE; break; case 'l': flags |= MS_CRYPT; break; case 'o': if (*optarg == '\0') { (void) fprintf(stderr, gettext("empty mount " "options (-o) specified\n")); usage(B_FALSE); } if (options == NULL) options = safe_malloc(MNT_LINE_MAX + 1); /* option validation is done later */ append_options(options, optarg); break; case 'O': flags |= MS_OVERLAY; break; case 'f': flags |= MS_FORCE; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check number of arguments */ if (do_all || recursive) { enum sa_protocol protocol = SA_NO_PROTOCOL; if (op == OP_SHARE && argc > 0) { protocol = sa_protocol_decode(argv[0]); argc--; argv++; } if (argc != 0 && do_all) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if (argc == 0 && recursive) { (void) fprintf(stderr, gettext("no dataset provided\n")); usage(B_FALSE); } start_progress_timer(); get_all_cb_t cb = { 0 }; get_all_state_t state = { 0 }; if (argc == 0) { state.ga_datasets = NULL; state.ga_count = -1; } else { zfs_handle_t *zhp; for (int i = 0; i < argc; i++) { zhp = zfs_open(g_zfs, argv[i], ZFS_TYPE_FILESYSTEM); if (zhp == NULL) usage(B_FALSE); zfs_close(zhp); } state.ga_datasets = argv; state.ga_count = argc; } state.ga_verbose = verbose; state.ga_cbp = &cb; get_all_datasets(&state); if (cb.cb_used == 0) { free(options); return (0); } share_mount_state_t share_mount_state = { 0 }; share_mount_state.sm_op = op; share_mount_state.sm_verbose = verbose; share_mount_state.sm_flags = flags; share_mount_state.sm_options = options; share_mount_state.sm_proto = protocol; share_mount_state.sm_total = cb.cb_used; pthread_mutex_init(&share_mount_state.sm_lock, NULL); /* For a 'zfs share -a' operation start with a clean slate. */ if (op == OP_SHARE) zfs_truncate_shares(NULL); /* * libshare isn't mt-safe, so only do the operation in parallel * if we're mounting. Additionally, the key-loading option must * be serialized so that we can prompt the user for their keys * in a consistent manner. */ zfs_foreach_mountpoint(g_zfs, cb.cb_handles, cb.cb_used, share_mount_one_cb, &share_mount_state, op == OP_MOUNT && !(flags & MS_CRYPT)); zfs_commit_shares(NULL); ret = share_mount_state.sm_status; for (int i = 0; i < cb.cb_used; i++) zfs_close(cb.cb_handles[i]); free(cb.cb_handles); } else if (argc == 0) { FILE *mnttab; struct mnttab entry; if ((op == OP_SHARE) || (options != NULL)) { (void) fprintf(stderr, gettext("missing filesystem " "argument (specify -a for all)\n")); usage(B_FALSE); } /* * When mount is given no arguments, go through * /proc/self/mounts and display any active ZFS mounts. * We hide any snapshots, since they are controlled * automatically. */ if ((mnttab = fopen(MNTTAB, "re")) == NULL) { free(options); return (ENOENT); } while (getmntent(mnttab, &entry) == 0) { if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0 || strchr(entry.mnt_special, '@') != NULL) continue; (void) printf("%-30s %s\n", entry.mnt_special, entry.mnt_mountp); } (void) fclose(mnttab); } else { zfs_handle_t *zhp; if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if ((zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM)) == NULL) { ret = 1; } else { ret = share_mount_one(zhp, op, flags, SA_NO_PROTOCOL, B_TRUE, options); zfs_commit_shares(NULL); zfs_close(zhp); } } free(options); return (ret); } /* * zfs mount -a * zfs mount filesystem * * Mount all filesystems, or mount the given filesystem. */ static int zfs_do_mount(int argc, char **argv) { return (share_mount(OP_MOUNT, argc, argv)); } /* * zfs share -a [nfs | smb] * zfs share filesystem * * Share all filesystems, or share the given filesystem. */ static int zfs_do_share(int argc, char **argv) { return (share_mount(OP_SHARE, argc, argv)); } typedef struct unshare_unmount_node { zfs_handle_t *un_zhp; char *un_mountp; uu_avl_node_t un_avlnode; } unshare_unmount_node_t; static int unshare_unmount_compare(const void *larg, const void *rarg, void *unused) { (void) unused; const unshare_unmount_node_t *l = larg; const unshare_unmount_node_t *r = rarg; return (strcmp(l->un_mountp, r->un_mountp)); } /* * Convenience routine used by zfs_do_umount() and manual_unmount(). Given an * absolute path, find the entry /proc/self/mounts, verify that it's a * ZFS filesystem, and unmount it appropriately. */ static int unshare_unmount_path(int op, char *path, int flags, boolean_t is_manual) { zfs_handle_t *zhp; int ret = 0; struct stat64 statbuf; struct extmnttab entry; const char *cmdname = (op == OP_SHARE) ? "unshare" : "unmount"; ino_t path_inode; /* * Search for the given (major,minor) pair in the mount table. */ if (getextmntent(path, &entry, &statbuf) != 0) { if (op == OP_SHARE) { (void) fprintf(stderr, gettext("cannot %s '%s': not " "currently mounted\n"), cmdname, path); return (1); } (void) fprintf(stderr, gettext("warning: %s not in" "/proc/self/mounts\n"), path); if ((ret = umount2(path, flags)) != 0) (void) fprintf(stderr, gettext("%s: %s\n"), path, strerror(errno)); return (ret != 0); } path_inode = statbuf.st_ino; if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0) { (void) fprintf(stderr, gettext("cannot %s '%s': not a ZFS " "filesystem\n"), cmdname, path); return (1); } if ((zhp = zfs_open(g_zfs, entry.mnt_special, ZFS_TYPE_FILESYSTEM)) == NULL) return (1); ret = 1; if (stat64(entry.mnt_mountp, &statbuf) != 0) { (void) fprintf(stderr, gettext("cannot %s '%s': %s\n"), cmdname, path, strerror(errno)); goto out; } else if (statbuf.st_ino != path_inode) { (void) fprintf(stderr, gettext("cannot " "%s '%s': not a mountpoint\n"), cmdname, path); goto out; } if (op == OP_SHARE) { char nfs_mnt_prop[ZFS_MAXPROPLEN]; char smbshare_prop[ZFS_MAXPROPLEN]; verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0); verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, smbshare_prop, sizeof (smbshare_prop), NULL, NULL, 0, B_FALSE) == 0); if (strcmp(nfs_mnt_prop, "off") == 0 && strcmp(smbshare_prop, "off") == 0) { (void) fprintf(stderr, gettext("cannot unshare " "'%s': legacy share\n"), path); (void) fprintf(stderr, gettext("use exportfs(8) " "or smbcontrol(1) to unshare this filesystem\n")); } else if (!zfs_is_shared(zhp, NULL, NULL)) { (void) fprintf(stderr, gettext("cannot unshare '%s': " "not currently shared\n"), path); } else { ret = zfs_unshare(zhp, path, NULL); zfs_commit_shares(NULL); } } else { char mtpt_prop[ZFS_MAXPROPLEN]; verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mtpt_prop, sizeof (mtpt_prop), NULL, NULL, 0, B_FALSE) == 0); if (is_manual) { ret = zfs_unmount(zhp, NULL, flags); } else if (strcmp(mtpt_prop, "legacy") == 0) { (void) fprintf(stderr, gettext("cannot unmount " "'%s': legacy mountpoint\n"), zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use umount(8) " "to unmount this filesystem\n")); } else { ret = zfs_unmountall(zhp, flags); } } out: zfs_close(zhp); return (ret != 0); } /* * Generic callback for unsharing or unmounting a filesystem. */ static int unshare_unmount(int op, int argc, char **argv) { int do_all = 0; int flags = 0; int ret = 0; int c; zfs_handle_t *zhp; char nfs_mnt_prop[ZFS_MAXPROPLEN]; char sharesmb[ZFS_MAXPROPLEN]; /* check options */ while ((c = getopt(argc, argv, op == OP_SHARE ? ":a" : "afu")) != -1) { switch (c) { case 'a': do_all = 1; break; case 'f': flags |= MS_FORCE; break; case 'u': flags |= MS_CRYPT; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (do_all) { /* * We could make use of zfs_for_each() to walk all datasets in * the system, but this would be very inefficient, especially * since we would have to linearly search /proc/self/mounts for * each one. Instead, do one pass through /proc/self/mounts * looking for zfs entries and call zfs_unmount() for each one. * * Things get a little tricky if the administrator has created * mountpoints beneath other ZFS filesystems. In this case, we * have to unmount the deepest filesystems first. To accomplish * this, we place all the mountpoints in an AVL tree sorted by * the special type (dataset name), and walk the result in * reverse to make sure to get any snapshots first. */ FILE *mnttab; struct mnttab entry; uu_avl_pool_t *pool; uu_avl_t *tree = NULL; unshare_unmount_node_t *node; uu_avl_index_t idx; uu_avl_walk_t *walk; enum sa_protocol *protocol = NULL, single_protocol[] = {SA_NO_PROTOCOL, SA_NO_PROTOCOL}; if (op == OP_SHARE && argc > 0) { *single_protocol = sa_protocol_decode(argv[0]); protocol = single_protocol; argc--; argv++; } if (argc != 0) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if (((pool = uu_avl_pool_create("unmount_pool", sizeof (unshare_unmount_node_t), offsetof(unshare_unmount_node_t, un_avlnode), unshare_unmount_compare, UU_DEFAULT)) == NULL) || ((tree = uu_avl_create(pool, NULL, UU_DEFAULT)) == NULL)) nomem(); if ((mnttab = fopen(MNTTAB, "re")) == NULL) { uu_avl_destroy(tree); uu_avl_pool_destroy(pool); return (ENOENT); } while (getmntent(mnttab, &entry) == 0) { /* ignore non-ZFS entries */ if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0) continue; /* ignore snapshots */ if (strchr(entry.mnt_special, '@') != NULL) continue; if ((zhp = zfs_open(g_zfs, entry.mnt_special, ZFS_TYPE_FILESYSTEM)) == NULL) { ret = 1; continue; } /* * Ignore datasets that are excluded/restricted by * parent pool name. */ if (zpool_skip_pool(zfs_get_pool_name(zhp))) { zfs_close(zhp); continue; } switch (op) { case OP_SHARE: verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0); if (strcmp(nfs_mnt_prop, "off") != 0) break; verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0); if (strcmp(nfs_mnt_prop, "off") == 0) continue; break; case OP_MOUNT: /* Ignore legacy mounts */ verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0); if (strcmp(nfs_mnt_prop, "legacy") == 0) continue; /* Ignore canmount=noauto mounts */ if (zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_NOAUTO) continue; break; default: break; } node = safe_malloc(sizeof (unshare_unmount_node_t)); node->un_zhp = zhp; node->un_mountp = safe_strdup(entry.mnt_mountp); uu_avl_node_init(node, &node->un_avlnode, pool); if (uu_avl_find(tree, node, NULL, &idx) == NULL) { uu_avl_insert(tree, node, idx); } else { zfs_close(node->un_zhp); free(node->un_mountp); free(node); } } (void) fclose(mnttab); /* * Walk the AVL tree in reverse, unmounting each filesystem and * removing it from the AVL tree in the process. */ if ((walk = uu_avl_walk_start(tree, UU_WALK_REVERSE | UU_WALK_ROBUST)) == NULL) nomem(); while ((node = uu_avl_walk_next(walk)) != NULL) { const char *mntarg = NULL; uu_avl_remove(tree, node); switch (op) { case OP_SHARE: if (zfs_unshare(node->un_zhp, node->un_mountp, protocol) != 0) ret = 1; break; case OP_MOUNT: if (zfs_unmount(node->un_zhp, mntarg, flags) != 0) ret = 1; break; } zfs_close(node->un_zhp); free(node->un_mountp); free(node); } if (op == OP_SHARE) zfs_commit_shares(protocol); uu_avl_walk_end(walk); uu_avl_destroy(tree); uu_avl_pool_destroy(pool); } else { if (argc != 1) { if (argc == 0) (void) fprintf(stderr, gettext("missing filesystem argument\n")); else (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } /* * We have an argument, but it may be a full path or a ZFS * filesystem. Pass full paths off to unmount_path() (shared by * manual_unmount), otherwise open the filesystem and pass to * zfs_unmount(). */ if (argv[0][0] == '/') return (unshare_unmount_path(op, argv[0], flags, B_FALSE)); if ((zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM)) == NULL) return (1); verify(zfs_prop_get(zhp, op == OP_SHARE ? ZFS_PROP_SHARENFS : ZFS_PROP_MOUNTPOINT, nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0); switch (op) { case OP_SHARE: verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0); verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, sharesmb, sizeof (sharesmb), NULL, NULL, 0, B_FALSE) == 0); if (strcmp(nfs_mnt_prop, "off") == 0 && strcmp(sharesmb, "off") == 0) { (void) fprintf(stderr, gettext("cannot " "unshare '%s': legacy share\n"), zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use " "exports(5) or smb.conf(5) to unshare " "this filesystem\n")); ret = 1; } else if (!zfs_is_shared(zhp, NULL, NULL)) { (void) fprintf(stderr, gettext("cannot " "unshare '%s': not currently " "shared\n"), zfs_get_name(zhp)); ret = 1; } else if (zfs_unshareall(zhp, NULL) != 0) { ret = 1; } break; case OP_MOUNT: if (strcmp(nfs_mnt_prop, "legacy") == 0) { (void) fprintf(stderr, gettext("cannot " "unmount '%s': legacy " "mountpoint\n"), zfs_get_name(zhp)); (void) fprintf(stderr, gettext("use " "umount(8) to unmount this " "filesystem\n")); ret = 1; } else if (!zfs_is_mounted(zhp, NULL)) { (void) fprintf(stderr, gettext("cannot " "unmount '%s': not currently " "mounted\n"), zfs_get_name(zhp)); ret = 1; } else if (zfs_unmountall(zhp, flags) != 0) { ret = 1; } break; } zfs_close(zhp); } return (ret); } /* * zfs unmount [-fu] -a * zfs unmount [-fu] filesystem * * Unmount all filesystems, or a specific ZFS filesystem. */ static int zfs_do_unmount(int argc, char **argv) { return (unshare_unmount(OP_MOUNT, argc, argv)); } /* * zfs unshare -a * zfs unshare filesystem * * Unshare all filesystems, or a specific ZFS filesystem. */ static int zfs_do_unshare(int argc, char **argv) { return (unshare_unmount(OP_SHARE, argc, argv)); } static int find_command_idx(const char *command, int *idx) { int i; for (i = 0; i < NCOMMAND; i++) { if (command_table[i].name == NULL) continue; if (strcmp(command, command_table[i].name) == 0) { *idx = i; return (0); } } return (1); } static int zfs_do_diff(int argc, char **argv) { zfs_handle_t *zhp; int flags = 0; char *tosnap = NULL; char *fromsnap = NULL; char *atp, *copy; int err = 0; int c; struct sigaction sa; while ((c = getopt(argc, argv, "FHth")) != -1) { switch (c) { case 'F': flags |= ZFS_DIFF_CLASSIFY; break; case 'H': flags |= ZFS_DIFF_PARSEABLE; break; case 't': flags |= ZFS_DIFF_TIMESTAMP; break; case 'h': flags |= ZFS_DIFF_NO_MANGLE; break; default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("must provide at least one snapshot name\n")); usage(B_FALSE); } if (argc > 2) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } fromsnap = argv[0]; tosnap = (argc == 2) ? argv[1] : NULL; copy = NULL; if (*fromsnap != '@') copy = strdup(fromsnap); else if (tosnap) copy = strdup(tosnap); if (copy == NULL) usage(B_FALSE); if ((atp = strchr(copy, '@')) != NULL) *atp = '\0'; if ((zhp = zfs_open(g_zfs, copy, ZFS_TYPE_FILESYSTEM)) == NULL) { free(copy); return (1); } free(copy); /* * Ignore SIGPIPE so that the library can give us * information on any failure */ if (sigemptyset(&sa.sa_mask) == -1) { err = errno; goto out; } sa.sa_flags = 0; sa.sa_handler = SIG_IGN; if (sigaction(SIGPIPE, &sa, NULL) == -1) { err = errno; goto out; } err = zfs_show_diffs(zhp, STDOUT_FILENO, fromsnap, tosnap, flags); out: zfs_close(zhp); return (err != 0); } /* * zfs bookmark | * * Creates a bookmark with the given name from the source snapshot * or creates a copy of an existing source bookmark. */ static int zfs_do_bookmark(int argc, char **argv) { char *source, *bookname; char expbuf[ZFS_MAX_DATASET_NAME_LEN]; int source_type; nvlist_t *nvl; int ret = 0; int c; /* check options */ while ((c = getopt(argc, argv, "")) != -1) { switch (c) { case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); goto usage; } } argc -= optind; argv += optind; /* check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing source argument\n")); goto usage; } if (argc < 2) { (void) fprintf(stderr, gettext("missing bookmark argument\n")); goto usage; } source = argv[0]; bookname = argv[1]; if (strchr(source, '@') == NULL && strchr(source, '#') == NULL) { (void) fprintf(stderr, gettext("invalid source name '%s': " "must contain a '@' or '#'\n"), source); goto usage; } if (strchr(bookname, '#') == NULL) { (void) fprintf(stderr, gettext("invalid bookmark name '%s': " "must contain a '#'\n"), bookname); goto usage; } /* * expand source or bookname to full path: * one of them may be specified as short name */ { char **expand; char *source_short, *bookname_short; source_short = strpbrk(source, "@#"); bookname_short = strpbrk(bookname, "#"); if (source_short == source && bookname_short == bookname) { (void) fprintf(stderr, gettext( "either source or bookmark must be specified as " "full dataset paths")); goto usage; } else if (source_short != source && bookname_short != bookname) { expand = NULL; } else if (source_short != source) { strlcpy(expbuf, source, sizeof (expbuf)); expand = &bookname; } else if (bookname_short != bookname) { strlcpy(expbuf, bookname, sizeof (expbuf)); expand = &source; } else { abort(); } if (expand != NULL) { *strpbrk(expbuf, "@#") = '\0'; /* dataset name in buf */ (void) strlcat(expbuf, *expand, sizeof (expbuf)); *expand = expbuf; } } /* determine source type */ switch (*strpbrk(source, "@#")) { case '@': source_type = ZFS_TYPE_SNAPSHOT; break; case '#': source_type = ZFS_TYPE_BOOKMARK; break; default: abort(); } /* test the source exists */ zfs_handle_t *zhp; zhp = zfs_open(g_zfs, source, source_type); if (zhp == NULL) goto usage; zfs_close(zhp); nvl = fnvlist_alloc(); fnvlist_add_string(nvl, bookname, source); ret = lzc_bookmark(nvl, NULL); fnvlist_free(nvl); if (ret != 0) { const char *err_msg = NULL; char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create bookmark '%s'"), bookname); switch (ret) { case EXDEV: err_msg = "bookmark is in a different pool"; break; case ZFS_ERR_BOOKMARK_SOURCE_NOT_ANCESTOR: err_msg = "source is not an ancestor of the " "new bookmark's dataset"; break; case EEXIST: err_msg = "bookmark exists"; break; case EINVAL: err_msg = "invalid argument"; break; case ENOTSUP: err_msg = "bookmark feature not enabled"; break; case ENOSPC: err_msg = "out of space"; break; case ENOENT: err_msg = "dataset does not exist"; break; default: (void) zfs_standard_error(g_zfs, ret, errbuf); break; } if (err_msg != NULL) { (void) fprintf(stderr, "%s: %s\n", errbuf, dgettext(TEXT_DOMAIN, err_msg)); } } return (ret != 0); usage: usage(B_FALSE); return (-1); } static int zfs_do_channel_program(int argc, char **argv) { int ret, fd, c; size_t progsize, progread; nvlist_t *outnvl = NULL; uint64_t instrlimit = ZCP_DEFAULT_INSTRLIMIT; uint64_t memlimit = ZCP_DEFAULT_MEMLIMIT; boolean_t sync_flag = B_TRUE, json_output = B_FALSE; zpool_handle_t *zhp; /* check options */ while ((c = getopt(argc, argv, "nt:m:j")) != -1) { switch (c) { case 't': case 'm': { uint64_t arg; char *endp; errno = 0; arg = strtoull(optarg, &endp, 0); if (errno != 0 || *endp != '\0') { (void) fprintf(stderr, gettext( "invalid argument " "'%s': expected integer\n"), optarg); goto usage; } if (c == 't') { instrlimit = arg; } else { ASSERT3U(c, ==, 'm'); memlimit = arg; } break; } case 'n': { sync_flag = B_FALSE; break; } case 'j': { json_output = B_TRUE; break; } case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); goto usage; } } argc -= optind; argv += optind; if (argc < 2) { (void) fprintf(stderr, gettext("invalid number of arguments\n")); goto usage; } const char *poolname = argv[0]; const char *filename = argv[1]; if (strcmp(filename, "-") == 0) { fd = 0; filename = "standard input"; } else if ((fd = open(filename, O_RDONLY)) < 0) { (void) fprintf(stderr, gettext("cannot open '%s': %s\n"), filename, strerror(errno)); return (1); } if ((zhp = zpool_open(g_zfs, poolname)) == NULL) { (void) fprintf(stderr, gettext("cannot open pool '%s'\n"), poolname); if (fd != 0) (void) close(fd); return (1); } zpool_close(zhp); /* * Read in the channel program, expanding the program buffer as * necessary. */ progread = 0; progsize = 1024; char *progbuf = safe_malloc(progsize); do { ret = read(fd, progbuf + progread, progsize - progread); progread += ret; if (progread == progsize && ret > 0) { progsize *= 2; progbuf = safe_realloc(progbuf, progsize); } } while (ret > 0); if (fd != 0) (void) close(fd); if (ret < 0) { free(progbuf); (void) fprintf(stderr, gettext("cannot read '%s': %s\n"), filename, strerror(errno)); return (1); } progbuf[progread] = '\0'; /* * Any remaining arguments are passed as arguments to the lua script as * a string array: * { * "argv" -> [ "arg 1", ... "arg n" ], * } */ nvlist_t *argnvl = fnvlist_alloc(); fnvlist_add_string_array(argnvl, ZCP_ARG_CLIARGV, (const char **)argv + 2, argc - 2); if (sync_flag) { ret = lzc_channel_program(poolname, progbuf, instrlimit, memlimit, argnvl, &outnvl); } else { ret = lzc_channel_program_nosync(poolname, progbuf, instrlimit, memlimit, argnvl, &outnvl); } if (ret != 0) { /* * On error, report the error message handed back by lua if one * exists. Otherwise, generate an appropriate error message, * falling back on strerror() for an unexpected return code. */ const char *errstring = NULL; const char *msg = gettext("Channel program execution failed"); uint64_t instructions = 0; if (outnvl != NULL && nvlist_exists(outnvl, ZCP_RET_ERROR)) { const char *es = NULL; (void) nvlist_lookup_string(outnvl, ZCP_RET_ERROR, &es); if (es == NULL) errstring = strerror(ret); else errstring = es; if (ret == ETIME) { (void) nvlist_lookup_uint64(outnvl, ZCP_ARG_INSTRLIMIT, &instructions); } } else { switch (ret) { case EINVAL: errstring = "Invalid instruction or memory limit."; break; case ENOMEM: errstring = "Return value too large."; break; case ENOSPC: errstring = "Memory limit exhausted."; break; case ETIME: errstring = "Timed out."; break; case EPERM: errstring = "Permission denied. Channel " "programs must be run as root."; break; default: (void) zfs_standard_error(g_zfs, ret, msg); } } if (errstring != NULL) (void) fprintf(stderr, "%s:\n%s\n", msg, errstring); if (ret == ETIME && instructions != 0) (void) fprintf(stderr, gettext("%llu Lua instructions\n"), (u_longlong_t)instructions); } else { if (json_output) { (void) nvlist_print_json(stdout, outnvl); } else if (nvlist_empty(outnvl)) { (void) fprintf(stdout, gettext("Channel program fully " "executed and did not produce output.\n")); } else { (void) fprintf(stdout, gettext("Channel program fully " "executed and produced output:\n")); dump_nvlist(outnvl, 4); } } free(progbuf); fnvlist_free(outnvl); fnvlist_free(argnvl); return (ret != 0); usage: usage(B_FALSE); return (-1); } typedef struct loadkey_cbdata { boolean_t cb_loadkey; boolean_t cb_recursive; boolean_t cb_noop; char *cb_keylocation; uint64_t cb_numfailed; uint64_t cb_numattempted; } loadkey_cbdata_t; static int load_key_callback(zfs_handle_t *zhp, void *data) { int ret; boolean_t is_encroot; loadkey_cbdata_t *cb = data; uint64_t keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS); /* * If we are working recursively, we want to skip loading / unloading * keys for non-encryption roots and datasets whose keys are already * in the desired end-state. */ if (cb->cb_recursive) { ret = zfs_crypto_get_encryption_root(zhp, &is_encroot, NULL); if (ret != 0) return (ret); if (!is_encroot) return (0); if ((cb->cb_loadkey && keystatus == ZFS_KEYSTATUS_AVAILABLE) || (!cb->cb_loadkey && keystatus == ZFS_KEYSTATUS_UNAVAILABLE)) return (0); } cb->cb_numattempted++; if (cb->cb_loadkey) ret = zfs_crypto_load_key(zhp, cb->cb_noop, cb->cb_keylocation); else ret = zfs_crypto_unload_key(zhp); if (ret != 0) { cb->cb_numfailed++; return (ret); } return (0); } static int load_unload_keys(int argc, char **argv, boolean_t loadkey) { int c, ret = 0, flags = 0; boolean_t do_all = B_FALSE; loadkey_cbdata_t cb = { 0 }; cb.cb_loadkey = loadkey; while ((c = getopt(argc, argv, "anrL:")) != -1) { /* noop and alternate keylocations only apply to zfs load-key */ if (loadkey) { switch (c) { case 'n': cb.cb_noop = B_TRUE; continue; case 'L': cb.cb_keylocation = optarg; continue; default: break; } } switch (c) { case 'a': do_all = B_TRUE; cb.cb_recursive = B_TRUE; break; case 'r': flags |= ZFS_ITER_RECURSE; cb.cb_recursive = B_TRUE; break; default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (!do_all && argc == 0) { (void) fprintf(stderr, gettext("Missing dataset argument or -a option\n")); usage(B_FALSE); } if (do_all && argc != 0) { (void) fprintf(stderr, gettext("Cannot specify dataset with -a option\n")); usage(B_FALSE); } if (cb.cb_recursive && cb.cb_keylocation != NULL && strcmp(cb.cb_keylocation, "prompt") != 0) { (void) fprintf(stderr, gettext("alternate keylocation may only " "be 'prompt' with -r or -a\n")); usage(B_FALSE); } ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, NULL, NULL, 0, load_key_callback, &cb); if (cb.cb_noop || (cb.cb_recursive && cb.cb_numattempted != 0)) { (void) printf(gettext("%llu / %llu key(s) successfully %s\n"), (u_longlong_t)(cb.cb_numattempted - cb.cb_numfailed), (u_longlong_t)cb.cb_numattempted, loadkey ? (cb.cb_noop ? "verified" : "loaded") : "unloaded"); } if (cb.cb_numfailed != 0) ret = -1; return (ret); } static int zfs_do_load_key(int argc, char **argv) { return (load_unload_keys(argc, argv, B_TRUE)); } static int zfs_do_unload_key(int argc, char **argv) { return (load_unload_keys(argc, argv, B_FALSE)); } static int zfs_do_change_key(int argc, char **argv) { int c, ret; uint64_t keystatus; boolean_t loadkey = B_FALSE, inheritkey = B_FALSE; zfs_handle_t *zhp = NULL; nvlist_t *props = fnvlist_alloc(); while ((c = getopt(argc, argv, "lio:")) != -1) { switch (c) { case 'l': loadkey = B_TRUE; break; case 'i': inheritkey = B_TRUE; break; case 'o': if (!parseprop(props, optarg)) { nvlist_free(props); return (1); } break; default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (inheritkey && !nvlist_empty(props)) { (void) fprintf(stderr, gettext("Properties not allowed for inheriting\n")); usage(B_FALSE); } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("Missing dataset argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("Too many arguments\n")); usage(B_FALSE); } zhp = zfs_open(g_zfs, argv[argc - 1], ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) usage(B_FALSE); if (loadkey) { keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS); if (keystatus != ZFS_KEYSTATUS_AVAILABLE) { ret = zfs_crypto_load_key(zhp, B_FALSE, NULL); if (ret != 0) { nvlist_free(props); zfs_close(zhp); return (-1); } } /* refresh the properties so the new keystatus is visible */ zfs_refresh_properties(zhp); } ret = zfs_crypto_rewrap(zhp, props, inheritkey); if (ret != 0) { nvlist_free(props); zfs_close(zhp); return (-1); } nvlist_free(props); zfs_close(zhp); return (0); } /* * 1) zfs project [-d|-r] * List project ID and inherit flag of file(s) or directories. * -d: List the directory itself, not its children. * -r: List subdirectories recursively. * * 2) zfs project -C [-k] [-r] * Clear project inherit flag and/or ID on the file(s) or directories. * -k: Keep the project ID unchanged. If not specified, the project ID * will be reset as zero. * -r: Clear on subdirectories recursively. * * 3) zfs project -c [-0] [-d|-r] [-p id] * Check project ID and inherit flag on the file(s) or directories, * report the outliers. * -0: Print file name followed by a NUL instead of newline. * -d: Check the directory itself, not its children. * -p: Specify the referenced ID for comparing with the target file(s) * or directories' project IDs. If not specified, the target (top) * directory's project ID will be used as the referenced one. * -r: Check subdirectories recursively. * * 4) zfs project [-p id] [-r] [-s] * Set project ID and/or inherit flag on the file(s) or directories. * -p: Set the project ID as the given id. * -r: Set on subdirectories recursively. If not specify "-p" option, * it will use top-level directory's project ID as the given id, * then set both project ID and inherit flag on all descendants * of the top-level directory. * -s: Set project inherit flag. */ static int zfs_do_project(int argc, char **argv) { zfs_project_control_t zpc = { .zpc_expected_projid = ZFS_INVALID_PROJID, .zpc_op = ZFS_PROJECT_OP_DEFAULT, .zpc_dironly = B_FALSE, .zpc_keep_projid = B_FALSE, .zpc_newline = B_TRUE, .zpc_recursive = B_FALSE, .zpc_set_flag = B_FALSE, }; int ret = 0, c; if (argc < 2) usage(B_FALSE); while ((c = getopt(argc, argv, "0Ccdkp:rs")) != -1) { switch (c) { case '0': zpc.zpc_newline = B_FALSE; break; case 'C': if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) { (void) fprintf(stderr, gettext("cannot " "specify '-C' '-c' '-s' together\n")); usage(B_FALSE); } zpc.zpc_op = ZFS_PROJECT_OP_CLEAR; break; case 'c': if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) { (void) fprintf(stderr, gettext("cannot " "specify '-C' '-c' '-s' together\n")); usage(B_FALSE); } zpc.zpc_op = ZFS_PROJECT_OP_CHECK; break; case 'd': zpc.zpc_dironly = B_TRUE; /* overwrite "-r" option */ zpc.zpc_recursive = B_FALSE; break; case 'k': zpc.zpc_keep_projid = B_TRUE; break; case 'p': { char *endptr; errno = 0; zpc.zpc_expected_projid = strtoull(optarg, &endptr, 0); if (errno != 0 || *endptr != '\0') { (void) fprintf(stderr, gettext("project ID must be less than " "%u\n"), UINT32_MAX); usage(B_FALSE); } if (zpc.zpc_expected_projid >= UINT32_MAX) { (void) fprintf(stderr, gettext("invalid project ID\n")); usage(B_FALSE); } break; } case 'r': zpc.zpc_recursive = B_TRUE; /* overwrite "-d" option */ zpc.zpc_dironly = B_FALSE; break; case 's': if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) { (void) fprintf(stderr, gettext("cannot " "specify '-C' '-c' '-s' together\n")); usage(B_FALSE); } zpc.zpc_set_flag = B_TRUE; zpc.zpc_op = ZFS_PROJECT_OP_SET; break; default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (zpc.zpc_op == ZFS_PROJECT_OP_DEFAULT) { if (zpc.zpc_expected_projid != ZFS_INVALID_PROJID) zpc.zpc_op = ZFS_PROJECT_OP_SET; else zpc.zpc_op = ZFS_PROJECT_OP_LIST; } switch (zpc.zpc_op) { case ZFS_PROJECT_OP_LIST: if (zpc.zpc_keep_projid) { (void) fprintf(stderr, gettext("'-k' is only valid together with '-C'\n")); usage(B_FALSE); } if (!zpc.zpc_newline) { (void) fprintf(stderr, gettext("'-0' is only valid together with '-c'\n")); usage(B_FALSE); } break; case ZFS_PROJECT_OP_CHECK: if (zpc.zpc_keep_projid) { (void) fprintf(stderr, gettext("'-k' is only valid together with '-C'\n")); usage(B_FALSE); } break; case ZFS_PROJECT_OP_CLEAR: if (zpc.zpc_dironly) { (void) fprintf(stderr, gettext("'-d' is useless together with '-C'\n")); usage(B_FALSE); } if (!zpc.zpc_newline) { (void) fprintf(stderr, gettext("'-0' is only valid together with '-c'\n")); usage(B_FALSE); } if (zpc.zpc_expected_projid != ZFS_INVALID_PROJID) { (void) fprintf(stderr, gettext("'-p' is useless together with '-C'\n")); usage(B_FALSE); } break; case ZFS_PROJECT_OP_SET: if (zpc.zpc_dironly) { (void) fprintf(stderr, gettext("'-d' is useless for set project ID and/or " "inherit flag\n")); usage(B_FALSE); } if (zpc.zpc_keep_projid) { (void) fprintf(stderr, gettext("'-k' is only valid together with '-C'\n")); usage(B_FALSE); } if (!zpc.zpc_newline) { (void) fprintf(stderr, gettext("'-0' is only valid together with '-c'\n")); usage(B_FALSE); } break; default: ASSERT(0); break; } argv += optind; argc -= optind; if (argc == 0) { (void) fprintf(stderr, gettext("missing file or directory target(s)\n")); usage(B_FALSE); } for (int i = 0; i < argc; i++) { int err; err = zfs_project_handle(argv[i], &zpc); if (err && !ret) ret = err; } return (ret); } static int zfs_do_wait(int argc, char **argv) { boolean_t enabled[ZFS_WAIT_NUM_ACTIVITIES]; int error = 0, i; int c; /* By default, wait for all types of activity. */ for (i = 0; i < ZFS_WAIT_NUM_ACTIVITIES; i++) enabled[i] = B_TRUE; while ((c = getopt(argc, argv, "t:")) != -1) { switch (c) { case 't': /* Reset activities array */ memset(&enabled, 0, sizeof (enabled)); for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const col_subopts[ ZFS_WAIT_NUM_ACTIVITIES] = { "deleteq" }; for (i = 0; i < ARRAY_SIZE(col_subopts); ++i) if (strcmp(tok, col_subopts[i]) == 0) { enabled[i] = B_TRUE; goto found; } (void) fprintf(stderr, gettext("invalid activity '%s'\n"), tok); usage(B_FALSE); found:; } break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argv += optind; argc -= optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing 'filesystem' " "argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } zfs_handle_t *zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM); if (zhp == NULL) return (1); for (;;) { boolean_t missing = B_FALSE; boolean_t any_waited = B_FALSE; for (int i = 0; i < ZFS_WAIT_NUM_ACTIVITIES; i++) { boolean_t waited; if (!enabled[i]) continue; error = zfs_wait_status(zhp, i, &missing, &waited); if (error != 0 || missing) break; any_waited = (any_waited || waited); } if (error != 0 || missing || !any_waited) break; } zfs_close(zhp); return (error); } /* * Display version message */ static int zfs_do_version(int argc, char **argv) { (void) argc, (void) argv; return (zfs_version_print() != 0); } /* Display documentation */ static int zfs_do_help(int argc, char **argv) { char page[MAXNAMELEN]; if (argc < 3 || strcmp(argv[2], "zfs") == 0) strcpy(page, "zfs"); else if (strcmp(argv[2], "concepts") == 0 || strcmp(argv[2], "props") == 0) snprintf(page, sizeof (page), "zfs%s", argv[2]); else snprintf(page, sizeof (page), "zfs-%s", argv[2]); execlp("man", "man", page, NULL); fprintf(stderr, "couldn't run man program: %s", strerror(errno)); return (-1); } int main(int argc, char **argv) { int ret = 0; int i = 0; const char *cmdname; char **newargv; (void) setlocale(LC_ALL, ""); (void) setlocale(LC_NUMERIC, "C"); (void) textdomain(TEXT_DOMAIN); opterr = 0; /* * Make sure the user has specified some command. */ if (argc < 2) { (void) fprintf(stderr, gettext("missing command\n")); usage(B_FALSE); } cmdname = argv[1]; /* * The 'umount' command is an alias for 'unmount' */ if (strcmp(cmdname, "umount") == 0) cmdname = "unmount"; /* * The 'recv' command is an alias for 'receive' */ if (strcmp(cmdname, "recv") == 0) cmdname = "receive"; /* * The 'snap' command is an alias for 'snapshot' */ if (strcmp(cmdname, "snap") == 0) cmdname = "snapshot"; /* * Special case '-?' */ if ((strcmp(cmdname, "-?") == 0) || (strcmp(cmdname, "--help") == 0)) usage(B_TRUE); /* * Special case '-V|--version' */ if ((strcmp(cmdname, "-V") == 0) || (strcmp(cmdname, "--version") == 0)) return (zfs_do_version(argc, argv)); /* * Special case 'help' */ if (strcmp(cmdname, "help") == 0) return (zfs_do_help(argc, argv)); if ((g_zfs = libzfs_init()) == NULL) { (void) fprintf(stderr, "%s\n", libzfs_error_init(errno)); return (1); } zfs_save_arguments(argc, argv, history_str, sizeof (history_str)); libzfs_print_on_error(g_zfs, B_TRUE); zfs_setproctitle_init(argc, argv, environ); /* * Many commands modify input strings for string parsing reasons. * We create a copy to protect the original argv. */ newargv = safe_malloc((argc + 1) * sizeof (newargv[0])); for (i = 0; i < argc; i++) newargv[i] = strdup(argv[i]); newargv[argc] = NULL; /* * Run the appropriate command. */ libzfs_mnttab_cache(g_zfs, B_TRUE); if (find_command_idx(cmdname, &i) == 0) { current_command = &command_table[i]; ret = command_table[i].func(argc - 1, newargv + 1); } else if (strchr(cmdname, '=') != NULL) { verify(find_command_idx("set", &i) == 0); current_command = &command_table[i]; ret = command_table[i].func(argc, newargv); } else { (void) fprintf(stderr, gettext("unrecognized " "command '%s'\n"), cmdname); usage(B_FALSE); ret = 1; } for (i = 0; i < argc; i++) free(newargv[i]); free(newargv); if (ret == 0 && log_history) (void) zpool_log_history(g_zfs, history_str); libzfs_fini(g_zfs); /* * The 'ZFS_ABORT' environment variable causes us to dump core on exit * for the purposes of running ::findleaks. */ if (getenv("ZFS_ABORT") != NULL) { (void) printf("dumping core by request\n"); abort(); } return (ret); } /* * zfs zone nsfile filesystem * * Add or delete the given dataset to/from the namespace. */ #ifdef __linux__ static int zfs_do_zone_impl(int argc, char **argv, boolean_t attach) { zfs_handle_t *zhp; int ret; if (argc < 3) { (void) fprintf(stderr, gettext("missing argument(s)\n")); usage(B_FALSE); } if (argc > 3) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } zhp = zfs_open(g_zfs, argv[2], ZFS_TYPE_FILESYSTEM); if (zhp == NULL) return (1); ret = (zfs_userns(zhp, argv[1], attach) != 0); zfs_close(zhp); return (ret); } static int zfs_do_zone(int argc, char **argv) { return (zfs_do_zone_impl(argc, argv, B_TRUE)); } static int zfs_do_unzone(int argc, char **argv) { return (zfs_do_zone_impl(argc, argv, B_FALSE)); } #endif #ifdef __FreeBSD__ #include #include /* * Attach/detach the given dataset to/from the given jail */ static int zfs_do_jail_impl(int argc, char **argv, boolean_t attach) { zfs_handle_t *zhp; int jailid, ret; /* check number of arguments */ if (argc < 3) { (void) fprintf(stderr, gettext("missing argument(s)\n")); usage(B_FALSE); } if (argc > 3) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } jailid = jail_getid(argv[1]); if (jailid < 0) { (void) fprintf(stderr, gettext("invalid jail id or name\n")); usage(B_FALSE); } zhp = zfs_open(g_zfs, argv[2], ZFS_TYPE_FILESYSTEM); if (zhp == NULL) return (1); ret = (zfs_jail(zhp, jailid, attach) != 0); zfs_close(zhp); return (ret); } /* * zfs jail jailid filesystem * * Attach the given dataset to the given jail */ static int zfs_do_jail(int argc, char **argv) { return (zfs_do_jail_impl(argc, argv, B_TRUE)); } /* * zfs unjail jailid filesystem * * Detach the given dataset from the given jail */ static int zfs_do_unjail(int argc, char **argv) { return (zfs_do_jail_impl(argc, argv, B_FALSE)); } #endif diff --git a/sys/contrib/openzfs/cmd/zinject/zinject.c b/sys/contrib/openzfs/cmd/zinject/zinject.c index e9141fb4ba55..ed60cce3dd16 100644 --- a/sys/contrib/openzfs/cmd/zinject/zinject.c +++ b/sys/contrib/openzfs/cmd/zinject/zinject.c @@ -1,1338 +1,1435 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2015 by Delphix. All rights reserved. * Copyright (c) 2017, Intel Corporation. - * Copyright (c) 2024, Klara Inc. + * Copyright (c) 2023-2024, Klara Inc. */ /* * ZFS Fault Injector * * This userland component takes a set of options and uses libzpool to translate * from a user-visible object type and name to an internal representation. * There are two basic types of faults: device faults and data faults. * * * DEVICE FAULTS * * Errors can be injected into a particular vdev using the '-d' option. This * option takes a path or vdev GUID to uniquely identify the device within a * pool. There are four types of errors that can be injected, IO, ENXIO, * ECHILD, and EILSEQ. These can be controlled through the '-e' option and the * default is ENXIO. For EIO failures, any attempt to read data from the device * will return EIO, but a subsequent attempt to reopen the device will succeed. * For ENXIO failures, any attempt to read from the device will return EIO, but * any attempt to reopen the device will also return ENXIO. The EILSEQ failures * only apply to read operations (-T read) and will flip a bit after the device * has read the original data. * * For label faults, the -L option must be specified. This allows faults * to be injected into either the nvlist, uberblock, pad1, or pad2 region * of all the labels for the specified device. * * This form of the command looks like: * * zinject -d device [-e errno] [-L ] pool * * * DATA FAULTS * * We begin with a tuple of the form: * * * * type A string describing the type of data to target. Each type * implicitly describes how to interpret 'object'. Currently, * the following values are supported: * * data User data for a file * dnode Dnode for a file or directory * * The following MOS objects are special. Instead of injecting * errors on a particular object or blkid, we inject errors across * all objects of the given type. * * mos Any data in the MOS * mosdir object directory * config pool configuration * bpobj blkptr list * spacemap spacemap * metaslab metaslab * errlog persistent error log * * level Object level. Defaults to '0', not applicable to all types. If * a range is given, this corresponds to the indirect block * corresponding to the specific range. * * range A numerical range [start,end) within the object. Defaults to * the full size of the file. * * object A string describing the logical location of the object. For * files and directories (currently the only supported types), * this is the path of the object on disk. * * This is translated, via libzpool, into the following internal representation: * * * * These types should be self-explanatory. This tuple is then passed to the * kernel via a special ioctl() to initiate fault injection for the given * object. Note that 'type' is not strictly necessary for fault injection, but * is used when translating existing faults into a human-readable string. * * * The command itself takes one of the forms: * * zinject * zinject <-a | -u pool> * zinject -c * zinject [-q] <-t type> [-f freq] [-u] [-a] [-m] [-e errno] [-l level] * [-r range] * zinject [-f freq] [-a] [-m] [-u] -b objset:object:level:start:end pool * * With no arguments, the command prints all currently registered injection * handlers, with their numeric identifiers. * * The '-c' option will clear the given handler, or all handlers if 'all' is * specified. * * The '-e' option takes a string describing the errno to simulate. This must * be one of 'io', 'checksum', 'decompress', or 'decrypt'. In most cases this * will result in the same behavior, but RAID-Z will produce a different set of * ereports for this situation. * * The '-a', '-u', and '-m' flags toggle internal flush behavior. If '-a' is * specified, then the ARC cache is flushed appropriately. If '-u' is * specified, then the underlying SPA is unloaded. Either of these flags can be * specified independently of any other handlers. The '-m' flag automatically * does an unmount and remount of the underlying dataset to aid in flushing the * cache. * * The '-f' flag controls the frequency of errors injected, expressed as a * real number percentage between 0.0001 and 100. The default is 100. * * The this form is responsible for actually injecting the handler into the * framework. It takes the arguments described above, translates them to the * internal tuple using libzpool, and then issues an ioctl() to register the * handler. * * The final form can target a specific bookmark, regardless of whether a * human-readable interface has been designed. It allows developers to specify * a particular block by number. */ #include #include #include #include #include #include #include #include #include #include #undef verify /* both libzfs.h and zfs_context.h want to define this */ #include "zinject.h" libzfs_handle_t *g_zfs; int zfs_fd; static const char *const errtable[TYPE_INVAL] = { "data", "dnode", "mos", "mosdir", "metaslab", "config", "bpobj", "spacemap", "errlog", "uber", "nvlist", "pad1", "pad2" }; static err_type_t name_to_type(const char *arg) { int i; for (i = 0; i < TYPE_INVAL; i++) if (strcmp(errtable[i], arg) == 0) return (i); return (TYPE_INVAL); } static const char * type_to_name(uint64_t type) { switch (type) { case DMU_OT_OBJECT_DIRECTORY: return ("mosdir"); case DMU_OT_OBJECT_ARRAY: return ("metaslab"); case DMU_OT_PACKED_NVLIST: return ("config"); case DMU_OT_BPOBJ: return ("bpobj"); case DMU_OT_SPACE_MAP: return ("spacemap"); case DMU_OT_ERROR_LOG: return ("errlog"); default: return ("-"); } } struct errstr { int err; const char *str; }; static const struct errstr errstrtable[] = { { EIO, "io" }, { ECKSUM, "checksum" }, { EINVAL, "decompress" }, { EACCES, "decrypt" }, { ENXIO, "nxio" }, { ECHILD, "dtl" }, { EILSEQ, "corrupt" }, { ENOSYS, "noop" }, { 0, NULL }, }; static int str_to_err(const char *str) { for (int i = 0; errstrtable[i].str != NULL; i++) if (strcasecmp(errstrtable[i].str, str) == 0) return (errstrtable[i].err); return (-1); } static const char * err_to_str(int err) { for (int i = 0; errstrtable[i].str != NULL; i++) if (errstrtable[i].err == err) return (errstrtable[i].str); return ("[unknown]"); } /* * Print usage message. */ void usage(void) { (void) printf( "usage:\n" "\n" "\tzinject\n" "\n" "\t\tList all active injection records.\n" "\n" "\tzinject -c \n" "\n" "\t\tClear the particular record (if given a numeric ID), or\n" "\t\tall records if 'all' is specified.\n" "\n" "\tzinject -p pool\n" "\t\tInject a panic fault at the specified function. Only \n" "\t\tfunctions which call spa_vdev_config_exit(), or \n" "\t\tspa_vdev_exit() will trigger a panic.\n" "\n" "\tzinject -d device [-e errno] [-L ] [-F]\n" "\t\t[-T ] [-f frequency] pool\n\n" "\t\tInject a fault into a particular device or the device's\n" "\t\tlabel. Label injection can either be 'nvlist', 'uber',\n " "\t\t'pad1', or 'pad2'.\n" "\t\t'errno' can be 'nxio' (the default), 'io', 'dtl',\n" "\t\t'corrupt' (bit flip), or 'noop' (successfully do nothing).\n" "\t\t'frequency' is a value between 0.0001 and 100.0 that limits\n" "\t\tdevice error injection to a percentage of the IOs.\n" "\n" "\tzinject -d device -A -D pool\n" "\t\tPerform a specific action on a particular device.\n" "\n" "\tzinject -d device -D latency:lanes pool\n" "\n" "\t\tAdd an artificial delay to IO requests on a particular\n" "\t\tdevice, such that the requests take a minimum of 'latency'\n" "\t\tmilliseconds to complete. Each delay has an associated\n" "\t\tnumber of 'lanes' which defines the number of concurrent\n" "\t\tIO requests that can be processed.\n" "\n" "\t\tFor example, with a single lane delay of 10 ms (-D 10:1),\n" "\t\tthe device will only be able to service a single IO request\n" "\t\tat a time with each request taking 10 ms to complete. So,\n" "\t\tif only a single request is submitted every 10 ms, the\n" "\t\taverage latency will be 10 ms; but if more than one request\n" "\t\tis submitted every 10 ms, the average latency will be more\n" "\t\tthan 10 ms.\n" "\n" "\t\tSimilarly, if a delay of 10 ms is specified to have two\n" "\t\tlanes (-D 10:2), then the device will be able to service\n" "\t\ttwo requests at a time, each with a minimum latency of\n" "\t\t10 ms. So, if two requests are submitted every 10 ms, then\n" "\t\tthe average latency will be 10 ms; but if more than two\n" "\t\trequests are submitted every 10 ms, the average latency\n" "\t\twill be more than 10 ms.\n" "\n" "\t\tAlso note, these delays are additive. So two invocations\n" "\t\tof '-D 10:1', is roughly equivalent to a single invocation\n" "\t\tof '-D 10:2'. This also means, one can specify multiple\n" "\t\tlanes with differing target latencies. For example, an\n" "\t\tinvocation of '-D 10:1' followed by '-D 25:2' will\n" "\t\tcreate 3 lanes on the device; one lane with a latency\n" "\t\tof 10 ms and two lanes with a 25 ms latency.\n" "\n" + "\tzinject -P import|export -s pool\n" + "\t\tAdd an artificial delay to a future pool import or export,\n" + "\t\tsuch that the operation takes a minimum of supplied seconds\n" + "\t\tto complete.\n" + "\n" "\tzinject -I [-s | -g ] pool\n" "\t\tCause the pool to stop writing blocks yet not\n" "\t\treport errors for a duration. Simulates buggy hardware\n" "\t\tthat fails to honor cache flush requests.\n" "\t\tDefault duration is 30 seconds. The machine is panicked\n" "\t\tat the end of the duration.\n" "\n" "\tzinject -b objset:object:level:blkid pool\n" "\n" "\t\tInject an error into pool 'pool' with the numeric bookmark\n" "\t\tspecified by the remaining tuple. Each number is in\n" "\t\thexadecimal, and only one block can be specified.\n" "\n" "\tzinject [-q] <-t type> [-C dvas] [-e errno] [-l level]\n" "\t\t[-r range] [-a] [-m] [-u] [-f freq] \n" "\n" "\t\tInject an error into the object specified by the '-t' option\n" "\t\tand the object descriptor. The 'object' parameter is\n" "\t\tinterpreted depending on the '-t' option.\n" "\n" "\t\t-q\tQuiet mode. Only print out the handler number added.\n" "\t\t-e\tInject a specific error. Must be one of 'io',\n" "\t\t\t'checksum', 'decompress', or 'decrypt'. Default is 'io'.\n" "\t\t-C\tInject the given error only into specific DVAs. The\n" "\t\t\tDVAs should be specified as a list of 0-indexed DVAs\n" "\t\t\tseparated by commas (ex. '0,2').\n" "\t\t-l\tInject error at a particular block level. Default is " "0.\n" "\t\t-m\tAutomatically remount underlying filesystem.\n" "\t\t-r\tInject error over a particular logical range of an\n" "\t\t\tobject. Will be translated to the appropriate blkid\n" "\t\t\trange according to the object's properties.\n" "\t\t-a\tFlush the ARC cache. Can be specified without any\n" "\t\t\tassociated object.\n" "\t\t-u\tUnload the associated pool. Can be specified with only\n" "\t\t\ta pool object.\n" "\t\t-f\tOnly inject errors a fraction of the time. Expressed as\n" "\t\t\ta percentage between 0.0001 and 100.\n" "\n" "\t-t data\t\tInject an error into the plain file contents of a\n" "\t\t\tfile. The object must be specified as a complete path\n" "\t\t\tto a file on a ZFS filesystem.\n" "\n" "\t-t dnode\tInject an error into the metadnode in the block\n" "\t\t\tcorresponding to the dnode for a file or directory. The\n" "\t\t\t'-r' option is incompatible with this mode. The object\n" "\t\t\tis specified as a complete path to a file or directory\n" "\t\t\ton a ZFS filesystem.\n" "\n" "\t-t \tInject errors into the MOS for objects of the given\n" "\t\t\ttype. Valid types are: mos, mosdir, config, bpobj,\n" "\t\t\tspacemap, metaslab, errlog. The only valid is\n" "\t\t\tthe poolname.\n"); } static int iter_handlers(int (*func)(int, const char *, zinject_record_t *, void *), void *data) { zfs_cmd_t zc = {"\0"}; int ret; while (zfs_ioctl(g_zfs, ZFS_IOC_INJECT_LIST_NEXT, &zc) == 0) if ((ret = func((int)zc.zc_guid, zc.zc_name, &zc.zc_inject_record, data)) != 0) return (ret); if (errno != ENOENT) { (void) fprintf(stderr, "Unable to list handlers: %s\n", strerror(errno)); return (-1); } return (0); } static int print_data_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; - if (record->zi_guid != 0 || record->zi_func[0] != '\0') + if (record->zi_guid != 0 || record->zi_func[0] != '\0' || + record->zi_duration != 0) { return (0); + } if (*count == 0) { (void) printf("%3s %-15s %-6s %-6s %-8s %3s %-4s " "%-15s\n", "ID", "POOL", "OBJSET", "OBJECT", "TYPE", "LVL", "DVAs", "RANGE"); (void) printf("--- --------------- ------ " "------ -------- --- ---- ---------------\n"); } *count += 1; (void) printf("%3d %-15s %-6llu %-6llu %-8s %-3d 0x%02x ", id, pool, (u_longlong_t)record->zi_objset, (u_longlong_t)record->zi_object, type_to_name(record->zi_type), record->zi_level, record->zi_dvas); if (record->zi_start == 0 && record->zi_end == -1ULL) (void) printf("all\n"); else (void) printf("[%llu, %llu]\n", (u_longlong_t)record->zi_start, (u_longlong_t)record->zi_end); return (0); } static int print_device_handler(int id, const char *pool, zinject_record_t *record, void *data) { static const char *iotypestr[] = { "null", "read", "write", "free", "claim", "flush", "trim", "all", }; int *count = data; if (record->zi_guid == 0 || record->zi_func[0] != '\0') return (0); if (record->zi_cmd == ZINJECT_DELAY_IO) return (0); if (*count == 0) { (void) printf("%3s %-15s %-16s %-5s %-10s %-9s\n", "ID", "POOL", "GUID", "TYPE", "ERROR", "FREQ"); (void) printf( "--- --------------- ---------------- " "----- ---------- ---------\n"); } *count += 1; double freq = record->zi_freq == 0 ? 100.0f : (((double)record->zi_freq) / ZI_PERCENTAGE_MAX) * 100.0f; (void) printf("%3d %-15s %llx %-5s %-10s %8.4f%%\n", id, pool, (u_longlong_t)record->zi_guid, iotypestr[record->zi_iotype], err_to_str(record->zi_error), freq); return (0); } static int print_delay_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; if (record->zi_guid == 0 || record->zi_func[0] != '\0') return (0); if (record->zi_cmd != ZINJECT_DELAY_IO) return (0); if (*count == 0) { (void) printf("%3s %-15s %-15s %-15s %s\n", "ID", "POOL", "DELAY (ms)", "LANES", "GUID"); (void) printf("--- --------------- --------------- " "--------------- ----------------\n"); } *count += 1; (void) printf("%3d %-15s %-15llu %-15llu %llx\n", id, pool, (u_longlong_t)NSEC2MSEC(record->zi_timer), (u_longlong_t)record->zi_nlanes, (u_longlong_t)record->zi_guid); return (0); } static int print_panic_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; if (record->zi_func[0] == '\0') return (0); if (*count == 0) { (void) printf("%3s %-15s %s\n", "ID", "POOL", "FUNCTION"); (void) printf("--- --------------- ----------------\n"); } *count += 1; (void) printf("%3d %-15s %s\n", id, pool, record->zi_func); return (0); } +static int +print_pool_delay_handler(int id, const char *pool, zinject_record_t *record, + void *data) +{ + int *count = data; + + if (record->zi_cmd != ZINJECT_DELAY_IMPORT && + record->zi_cmd != ZINJECT_DELAY_EXPORT) { + return (0); + } + + if (*count == 0) { + (void) printf("%3s %-19s %-11s %s\n", + "ID", "POOL", "DELAY (sec)", "COMMAND"); + (void) printf("--- ------------------- -----------" + " -------\n"); + } + + *count += 1; + + (void) printf("%3d %-19s %-11llu %s\n", + id, pool, (u_longlong_t)record->zi_duration, + record->zi_cmd == ZINJECT_DELAY_IMPORT ? "import": "export"); + + return (0); +} + /* * Print all registered error handlers. Returns the number of handlers * registered. */ static int print_all_handlers(void) { int count = 0, total = 0; (void) iter_handlers(print_device_handler, &count); if (count > 0) { total += count; (void) printf("\n"); count = 0; } (void) iter_handlers(print_delay_handler, &count); if (count > 0) { total += count; (void) printf("\n"); count = 0; } (void) iter_handlers(print_data_handler, &count); if (count > 0) { total += count; (void) printf("\n"); count = 0; } + (void) iter_handlers(print_pool_delay_handler, &count); + if (count > 0) { + total += count; + (void) printf("\n"); + count = 0; + } + (void) iter_handlers(print_panic_handler, &count); return (count + total); } static int cancel_one_handler(int id, const char *pool, zinject_record_t *record, void *data) { (void) pool, (void) record, (void) data; zfs_cmd_t zc = {"\0"}; zc.zc_guid = (uint64_t)id; if (zfs_ioctl(g_zfs, ZFS_IOC_CLEAR_FAULT, &zc) != 0) { (void) fprintf(stderr, "failed to remove handler %d: %s\n", id, strerror(errno)); return (1); } return (0); } /* * Remove all fault injection handlers. */ static int cancel_all_handlers(void) { int ret = iter_handlers(cancel_one_handler, NULL); if (ret == 0) (void) printf("removed all registered handlers\n"); return (ret); } /* * Remove a specific fault injection handler. */ static int cancel_handler(int id) { zfs_cmd_t zc = {"\0"}; zc.zc_guid = (uint64_t)id; if (zfs_ioctl(g_zfs, ZFS_IOC_CLEAR_FAULT, &zc) != 0) { (void) fprintf(stderr, "failed to remove handler %d: %s\n", id, strerror(errno)); return (1); } (void) printf("removed handler %d\n", id); return (0); } /* * Register a new fault injection handler. */ static int register_handler(const char *pool, int flags, zinject_record_t *record, int quiet) { zfs_cmd_t zc = {"\0"}; (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); zc.zc_inject_record = *record; zc.zc_guid = flags; if (zfs_ioctl(g_zfs, ZFS_IOC_INJECT_FAULT, &zc) != 0) { - (void) fprintf(stderr, "failed to add handler: %s\n", - errno == EDOM ? "block level exceeds max level of object" : - strerror(errno)); + const char *errmsg = strerror(errno); + + switch (errno) { + case EDOM: + errmsg = "block level exceeds max level of object"; + break; + case EEXIST: + if (record->zi_cmd == ZINJECT_DELAY_IMPORT) + errmsg = "pool already imported"; + if (record->zi_cmd == ZINJECT_DELAY_EXPORT) + errmsg = "a handler already exists"; + break; + case ENOENT: + /* import delay injector running on older zfs module */ + if (record->zi_cmd == ZINJECT_DELAY_IMPORT) + errmsg = "import delay injector not supported"; + break; + default: + break; + } + (void) fprintf(stderr, "failed to add handler: %s\n", errmsg); return (1); } if (flags & ZINJECT_NULL) return (0); if (quiet) { (void) printf("%llu\n", (u_longlong_t)zc.zc_guid); } else { (void) printf("Added handler %llu with the following " "properties:\n", (u_longlong_t)zc.zc_guid); (void) printf(" pool: %s\n", pool); if (record->zi_guid) { (void) printf(" vdev: %llx\n", (u_longlong_t)record->zi_guid); } else if (record->zi_func[0] != '\0') { (void) printf(" panic function: %s\n", record->zi_func); } else if (record->zi_duration > 0) { (void) printf(" time: %lld seconds\n", (u_longlong_t)record->zi_duration); } else if (record->zi_duration < 0) { (void) printf(" txgs: %lld \n", (u_longlong_t)-record->zi_duration); + } else if (record->zi_timer > 0) { + (void) printf(" timer: %lld ms\n", + (u_longlong_t)NSEC2MSEC(record->zi_timer)); } else { (void) printf("objset: %llu\n", (u_longlong_t)record->zi_objset); (void) printf("object: %llu\n", (u_longlong_t)record->zi_object); (void) printf(" type: %llu\n", (u_longlong_t)record->zi_type); (void) printf(" level: %d\n", record->zi_level); if (record->zi_start == 0 && record->zi_end == -1ULL) (void) printf(" range: all\n"); else (void) printf(" range: [%llu, %llu)\n", (u_longlong_t)record->zi_start, (u_longlong_t)record->zi_end); (void) printf(" dvas: 0x%x\n", record->zi_dvas); } } return (0); } static int perform_action(const char *pool, zinject_record_t *record, int cmd) { zfs_cmd_t zc = {"\0"}; ASSERT(cmd == VDEV_STATE_DEGRADED || cmd == VDEV_STATE_FAULTED); (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); zc.zc_guid = record->zi_guid; zc.zc_cookie = cmd; if (zfs_ioctl(g_zfs, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) return (0); return (1); } static int parse_delay(char *str, uint64_t *delay, uint64_t *nlanes) { unsigned long scan_delay; unsigned long scan_nlanes; if (sscanf(str, "%lu:%lu", &scan_delay, &scan_nlanes) != 2) return (1); /* * We explicitly disallow a delay of zero here, because we key * off this value being non-zero in translate_device(), to * determine if the fault is a ZINJECT_DELAY_IO fault or not. */ if (scan_delay == 0) return (1); /* * The units for the CLI delay parameter is milliseconds, but * the data passed to the kernel is interpreted as nanoseconds. * Thus we scale the milliseconds to nanoseconds here, and this * nanosecond value is used to pass the delay to the kernel. */ *delay = MSEC2NSEC(scan_delay); *nlanes = scan_nlanes; return (0); } static int parse_frequency(const char *str, uint32_t *percent) { double val; char *post; val = strtod(str, &post); if (post == NULL || *post != '\0') return (EINVAL); /* valid range is [0.0001, 100.0] */ val /= 100.0f; if (val < 0.000001f || val > 1.0f) return (ERANGE); /* convert to an integer for use by kernel */ *percent = ((uint32_t)(val * ZI_PERCENTAGE_MAX)); return (0); } /* * This function converts a string specifier for DVAs into a bit mask. * The dva's provided by the user should be 0 indexed and separated by * a comma. For example: * "1" -> 0b0010 (0x2) * "0,1" -> 0b0011 (0x3) * "0,1,2" -> 0b0111 (0x7) */ static int parse_dvas(const char *str, uint32_t *dvas_out) { const char *c = str; uint32_t mask = 0; boolean_t need_delim = B_FALSE; /* max string length is 5 ("0,1,2") */ if (strlen(str) > 5 || strlen(str) == 0) return (EINVAL); while (*c != '\0') { switch (*c) { case '0': case '1': case '2': /* check for pipe between DVAs */ if (need_delim) return (EINVAL); /* check if this DVA has been set already */ if (mask & (1 << ((*c) - '0'))) return (EINVAL); mask |= (1 << ((*c) - '0')); need_delim = B_TRUE; break; case ',': need_delim = B_FALSE; break; default: /* check for invalid character */ return (EINVAL); } c++; } /* check for dangling delimiter */ if (!need_delim) return (EINVAL); *dvas_out = mask; return (0); } int main(int argc, char **argv) { int c; char *range = NULL; char *cancel = NULL; char *end; char *raw = NULL; char *device = NULL; int level = 0; int quiet = 0; int error = 0; int domount = 0; int io_type = ZIO_TYPES; int action = VDEV_STATE_UNKNOWN; err_type_t type = TYPE_INVAL; err_type_t label = TYPE_INVAL; zinject_record_t record = { 0 }; char pool[MAXNAMELEN] = ""; char dataset[MAXNAMELEN] = ""; zfs_handle_t *zhp = NULL; int nowrites = 0; int dur_txg = 0; int dur_secs = 0; int ret; int flags = 0; uint32_t dvas = 0; if ((g_zfs = libzfs_init()) == NULL) { (void) fprintf(stderr, "%s\n", libzfs_error_init(errno)); return (1); } libzfs_print_on_error(g_zfs, B_TRUE); if ((zfs_fd = open(ZFS_DEV, O_RDWR)) < 0) { (void) fprintf(stderr, "failed to open ZFS device\n"); libzfs_fini(g_zfs); return (1); } if (argc == 1) { /* * No arguments. Print the available handlers. If there are no * available handlers, direct the user to '-h' for help * information. */ if (print_all_handlers() == 0) { (void) printf("No handlers registered.\n"); (void) printf("Run 'zinject -h' for usage " "information.\n"); } libzfs_fini(g_zfs); return (0); } while ((c = getopt(argc, argv, - ":aA:b:C:d:D:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:")) != -1) { + ":aA:b:C:d:D:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:P:")) != -1) { switch (c) { case 'a': flags |= ZINJECT_FLUSH_ARC; break; case 'A': if (strcasecmp(optarg, "degrade") == 0) { action = VDEV_STATE_DEGRADED; } else if (strcasecmp(optarg, "fault") == 0) { action = VDEV_STATE_FAULTED; } else { (void) fprintf(stderr, "invalid action '%s': " "must be 'degrade' or 'fault'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'b': raw = optarg; break; case 'c': cancel = optarg; break; case 'C': ret = parse_dvas(optarg, &dvas); if (ret != 0) { (void) fprintf(stderr, "invalid DVA list '%s': " "DVAs should be 0 indexed and separated by " "commas.\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'd': device = optarg; break; case 'D': errno = 0; ret = parse_delay(optarg, &record.zi_timer, &record.zi_nlanes); if (ret != 0) { (void) fprintf(stderr, "invalid i/o delay " "value: '%s'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'e': error = str_to_err(optarg); if (error < 0) { (void) fprintf(stderr, "invalid error type " "'%s': must be one of: io decompress " "decrypt nxio dtl corrupt noop\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'f': ret = parse_frequency(optarg, &record.zi_freq); if (ret != 0) { (void) fprintf(stderr, "%sfrequency value must " "be in the range [0.0001, 100.0]\n", ret == EINVAL ? "invalid value: " : ret == ERANGE ? "out of range: " : ""); libzfs_fini(g_zfs); return (1); } break; case 'F': record.zi_failfast = B_TRUE; break; case 'g': dur_txg = 1; record.zi_duration = (int)strtol(optarg, &end, 10); if (record.zi_duration <= 0 || *end != '\0') { (void) fprintf(stderr, "invalid duration '%s': " "must be a positive integer\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } /* store duration of txgs as its negative */ record.zi_duration *= -1; break; case 'h': usage(); libzfs_fini(g_zfs); return (0); case 'I': /* default duration, if one hasn't yet been defined */ nowrites = 1; if (dur_secs == 0 && dur_txg == 0) record.zi_duration = 30; break; case 'l': level = (int)strtol(optarg, &end, 10); if (*end != '\0') { (void) fprintf(stderr, "invalid level '%s': " "must be an integer\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'm': domount = 1; break; case 'p': (void) strlcpy(record.zi_func, optarg, sizeof (record.zi_func)); record.zi_cmd = ZINJECT_PANIC; break; + case 'P': + if (strcasecmp(optarg, "import") == 0) { + record.zi_cmd = ZINJECT_DELAY_IMPORT; + } else if (strcasecmp(optarg, "export") == 0) { + record.zi_cmd = ZINJECT_DELAY_EXPORT; + } else { + (void) fprintf(stderr, "invalid command '%s': " + "must be 'import' or 'export'\n", optarg); + usage(); + libzfs_fini(g_zfs); + return (1); + } + break; case 'q': quiet = 1; break; case 'r': range = optarg; flags |= ZINJECT_CALC_RANGE; break; case 's': dur_secs = 1; record.zi_duration = (int)strtol(optarg, &end, 10); if (record.zi_duration <= 0 || *end != '\0') { (void) fprintf(stderr, "invalid duration '%s': " "must be a positive integer\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'T': if (strcasecmp(optarg, "read") == 0) { io_type = ZIO_TYPE_READ; } else if (strcasecmp(optarg, "write") == 0) { io_type = ZIO_TYPE_WRITE; } else if (strcasecmp(optarg, "free") == 0) { io_type = ZIO_TYPE_FREE; } else if (strcasecmp(optarg, "claim") == 0) { io_type = ZIO_TYPE_CLAIM; } else if (strcasecmp(optarg, "flush") == 0) { io_type = ZIO_TYPE_FLUSH; } else if (strcasecmp(optarg, "all") == 0) { io_type = ZIO_TYPES; } else { (void) fprintf(stderr, "invalid I/O type " "'%s': must be 'read', 'write', 'free', " "'claim', 'flush' or 'all'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 't': if ((type = name_to_type(optarg)) == TYPE_INVAL && !MOS_TYPE(type)) { (void) fprintf(stderr, "invalid type '%s'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'u': flags |= ZINJECT_UNLOAD_SPA; break; case 'L': if ((label = name_to_type(optarg)) == TYPE_INVAL && !LABEL_TYPE(type)) { (void) fprintf(stderr, "invalid label type " "'%s'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case ':': (void) fprintf(stderr, "option -%c requires an " "operand\n", optopt); usage(); libzfs_fini(g_zfs); return (1); case '?': (void) fprintf(stderr, "invalid option '%c'\n", optopt); usage(); libzfs_fini(g_zfs); return (2); } } argc -= optind; argv += optind; - if (record.zi_duration != 0) + if (record.zi_duration != 0 && record.zi_cmd == 0) record.zi_cmd = ZINJECT_IGNORED_WRITES; if (cancel != NULL) { /* * '-c' is invalid with any other options. */ if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || record.zi_freq > 0 || dvas != 0) { (void) fprintf(stderr, "cancel (-c) incompatible with " "any other options\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc != 0) { (void) fprintf(stderr, "extraneous argument to '-c'\n"); usage(); libzfs_fini(g_zfs); return (2); } if (strcmp(cancel, "all") == 0) { return (cancel_all_handlers()); } else { int id = (int)strtol(cancel, &end, 10); if (*end != '\0') { (void) fprintf(stderr, "invalid handle id '%s':" " must be an integer or 'all'\n", cancel); usage(); libzfs_fini(g_zfs); return (1); } return (cancel_handler(id)); } } if (device != NULL) { /* * Device (-d) injection uses a completely different mechanism * for doing injection, so handle it separately here. */ if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || dvas != 0) { (void) fprintf(stderr, "device (-d) incompatible with " "data error injection\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc != 1) { (void) fprintf(stderr, "device (-d) injection requires " "a single pool name\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; if (error == ECKSUM) { (void) fprintf(stderr, "device error type must be " "'io', 'nxio' or 'corrupt'\n"); libzfs_fini(g_zfs); return (1); } if (error == EILSEQ && (record.zi_freq == 0 || io_type != ZIO_TYPE_READ)) { (void) fprintf(stderr, "device corrupt errors require " "io type read and a frequency value\n"); libzfs_fini(g_zfs); return (1); } record.zi_iotype = io_type; if (translate_device(pool, device, label, &record) != 0) { libzfs_fini(g_zfs); return (1); } if (record.zi_nlanes) { switch (io_type) { case ZIO_TYPE_READ: case ZIO_TYPE_WRITE: case ZIO_TYPES: break; default: (void) fprintf(stderr, "I/O type for a delay " "must be 'read' or 'write'\n"); usage(); libzfs_fini(g_zfs); return (1); } } if (!error) error = ENXIO; if (action != VDEV_STATE_UNKNOWN) return (perform_action(pool, &record, action)); } else if (raw != NULL) { if (range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || record.zi_freq > 0 || dvas != 0) { (void) fprintf(stderr, "raw (-b) format with " "any other options\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc != 1) { (void) fprintf(stderr, "raw (-b) format expects a " "single pool name\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; if (error == ENXIO) { (void) fprintf(stderr, "data error type must be " "'checksum' or 'io'\n"); libzfs_fini(g_zfs); return (1); } record.zi_cmd = ZINJECT_DATA_FAULT; if (translate_raw(raw, &record) != 0) { libzfs_fini(g_zfs); return (1); } if (!error) error = EIO; } else if (record.zi_cmd == ZINJECT_PANIC) { if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || device != NULL || record.zi_freq > 0 || dvas != 0) { - (void) fprintf(stderr, "panic (-p) incompatible with " - "other options\n"); + (void) fprintf(stderr, "%s incompatible with other " + "options\n", "import|export delay (-P)"); usage(); libzfs_fini(g_zfs); return (2); } if (argc < 1 || argc > 2) { (void) fprintf(stderr, "panic (-p) injection requires " "a single pool name and an optional id\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); if (argv[1] != NULL) record.zi_type = atoi(argv[1]); dataset[0] = '\0'; + } else if (record.zi_cmd == ZINJECT_DELAY_IMPORT || + record.zi_cmd == ZINJECT_DELAY_EXPORT) { + if (raw != NULL || range != NULL || type != TYPE_INVAL || + level != 0 || device != NULL || record.zi_freq > 0 || + dvas != 0) { + (void) fprintf(stderr, "%s incompatible with other " + "options\n", "import|export delay (-P)"); + usage(); + libzfs_fini(g_zfs); + return (2); + } + + if (argc != 1 || record.zi_duration <= 0) { + (void) fprintf(stderr, "import|export delay (-P) " + "injection requires a duration (-s) and a single " + "pool name\n"); + usage(); + libzfs_fini(g_zfs); + return (2); + } + + (void) strlcpy(pool, argv[0], sizeof (pool)); } else if (record.zi_cmd == ZINJECT_IGNORED_WRITES) { if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_freq > 0 || dvas != 0) { (void) fprintf(stderr, "hardware failure (-I) " "incompatible with other options\n"); usage(); libzfs_fini(g_zfs); return (2); } if (nowrites == 0) { (void) fprintf(stderr, "-s or -g meaningless " "without -I (ignore writes)\n"); usage(); libzfs_fini(g_zfs); return (2); } else if (dur_secs && dur_txg) { (void) fprintf(stderr, "choose a duration either " "in seconds (-s) or a number of txgs (-g) " "but not both\n"); usage(); libzfs_fini(g_zfs); return (2); } else if (argc != 1) { (void) fprintf(stderr, "ignore writes (-I) " "injection requires a single pool name\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; } else if (type == TYPE_INVAL) { if (flags == 0) { (void) fprintf(stderr, "at least one of '-b', '-d', " "'-t', '-a', '-p', '-I' or '-u' " "must be specified\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc == 1 && (flags & ZINJECT_UNLOAD_SPA)) { (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; } else if (argc != 0) { (void) fprintf(stderr, "extraneous argument for " "'-f'\n"); usage(); libzfs_fini(g_zfs); return (2); } flags |= ZINJECT_NULL; } else { if (argc != 1) { (void) fprintf(stderr, "missing object\n"); usage(); libzfs_fini(g_zfs); return (2); } if (error == ENXIO || error == EILSEQ) { (void) fprintf(stderr, "data error type must be " "'checksum' or 'io'\n"); libzfs_fini(g_zfs); return (1); } if (dvas != 0) { if (error == EACCES || error == EINVAL) { (void) fprintf(stderr, "the '-C' option may " "not be used with logical data errors " "'decrypt' and 'decompress'\n"); libzfs_fini(g_zfs); return (1); } record.zi_dvas = dvas; } if (error == EACCES) { if (type != TYPE_DATA) { (void) fprintf(stderr, "decryption errors " "may only be injected for 'data' types\n"); libzfs_fini(g_zfs); return (1); } record.zi_cmd = ZINJECT_DECRYPT_FAULT; /* * Internally, ZFS actually uses ECKSUM for decryption * errors since EACCES is used to indicate the key was * not found. */ error = ECKSUM; } else { record.zi_cmd = ZINJECT_DATA_FAULT; } if (translate_record(type, argv[0], range, level, &record, pool, dataset) != 0) { libzfs_fini(g_zfs); return (1); } if (!error) error = EIO; } /* * If this is pool-wide metadata, unmount everything. The ioctl() will * unload the pool, so that we trigger spa-wide reopen of metadata next * time we access the pool. */ if (dataset[0] != '\0' && domount) { if ((zhp = zfs_open(g_zfs, dataset, ZFS_TYPE_DATASET)) == NULL) { libzfs_fini(g_zfs); return (1); } if (zfs_unmount(zhp, NULL, 0) != 0) { libzfs_fini(g_zfs); return (1); } } record.zi_error = error; ret = register_handler(pool, flags, &record, quiet); if (dataset[0] != '\0' && domount) ret = (zfs_mount(zhp, NULL, 0) != 0); libzfs_fini(g_zfs); return (ret); } diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_main.c b/sys/contrib/openzfs/cmd/zpool/zpool_main.c index d670cd1afeb1..636eb2a301cd 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_main.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_main.c @@ -1,11630 +1,11668 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2011, 2024 by Delphix. All rights reserved. * Copyright (c) 2012 by Frederik Wessels. All rights reserved. * Copyright (c) 2012 by Cyril Plisko. All rights reserved. * Copyright (c) 2013 by Prasad Joshi (sTec). All rights reserved. * Copyright 2016 Igor Kozhukhov . * Copyright (c) 2017 Datto Inc. * Copyright (c) 2017 Open-E, Inc. All Rights Reserved. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2019, loli10K * Copyright (c) 2021, Colm Buckley * Copyright (c) 2021, Klara Inc. * Copyright [2021] Hewlett Packard Enterprise Development LP */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "zpool_util.h" #include "zfs_comutil.h" #include "zfeature_common.h" #include "statcommon.h" libzfs_handle_t *g_zfs; static int zpool_do_create(int, char **); static int zpool_do_destroy(int, char **); static int zpool_do_add(int, char **); static int zpool_do_remove(int, char **); static int zpool_do_labelclear(int, char **); static int zpool_do_checkpoint(int, char **); static int zpool_do_list(int, char **); static int zpool_do_iostat(int, char **); static int zpool_do_status(int, char **); static int zpool_do_online(int, char **); static int zpool_do_offline(int, char **); static int zpool_do_clear(int, char **); static int zpool_do_reopen(int, char **); static int zpool_do_reguid(int, char **); static int zpool_do_attach(int, char **); static int zpool_do_detach(int, char **); static int zpool_do_replace(int, char **); static int zpool_do_split(int, char **); static int zpool_do_initialize(int, char **); static int zpool_do_scrub(int, char **); static int zpool_do_resilver(int, char **); static int zpool_do_trim(int, char **); static int zpool_do_import(int, char **); static int zpool_do_export(int, char **); static int zpool_do_upgrade(int, char **); static int zpool_do_history(int, char **); static int zpool_do_events(int, char **); static int zpool_do_get(int, char **); static int zpool_do_set(int, char **); static int zpool_do_sync(int, char **); static int zpool_do_version(int, char **); static int zpool_do_wait(int, char **); static int zpool_do_help(int argc, char **argv); static zpool_compat_status_t zpool_do_load_compat( const char *, boolean_t *); enum zpool_options { ZPOOL_OPTION_POWER = 1024, ZPOOL_OPTION_ALLOW_INUSE, ZPOOL_OPTION_ALLOW_REPLICATION_MISMATCH, ZPOOL_OPTION_ALLOW_ASHIFT_MISMATCH }; /* * These libumem hooks provide a reasonable set of defaults for the allocator's * debugging facilities. */ #ifdef DEBUG const char * _umem_debug_init(void) { return ("default,verbose"); /* $UMEM_DEBUG setting */ } const char * _umem_logging_init(void) { return ("fail,contents"); /* $UMEM_LOGGING setting */ } #endif typedef enum { HELP_ADD, HELP_ATTACH, HELP_CLEAR, HELP_CREATE, HELP_CHECKPOINT, HELP_DESTROY, HELP_DETACH, HELP_EXPORT, HELP_HISTORY, HELP_IMPORT, HELP_IOSTAT, HELP_LABELCLEAR, HELP_LIST, HELP_OFFLINE, HELP_ONLINE, HELP_REPLACE, HELP_REMOVE, HELP_INITIALIZE, HELP_SCRUB, HELP_RESILVER, HELP_TRIM, HELP_STATUS, HELP_UPGRADE, HELP_EVENTS, HELP_GET, HELP_SET, HELP_SPLIT, HELP_SYNC, HELP_REGUID, HELP_REOPEN, HELP_VERSION, HELP_WAIT } zpool_help_t; /* * Flags for stats to display with "zpool iostats" */ enum iostat_type { IOS_DEFAULT = 0, IOS_LATENCY = 1, IOS_QUEUES = 2, IOS_L_HISTO = 3, IOS_RQ_HISTO = 4, IOS_COUNT, /* always last element */ }; /* iostat_type entries as bitmasks */ #define IOS_DEFAULT_M (1ULL << IOS_DEFAULT) #define IOS_LATENCY_M (1ULL << IOS_LATENCY) #define IOS_QUEUES_M (1ULL << IOS_QUEUES) #define IOS_L_HISTO_M (1ULL << IOS_L_HISTO) #define IOS_RQ_HISTO_M (1ULL << IOS_RQ_HISTO) /* Mask of all the histo bits */ #define IOS_ANYHISTO_M (IOS_L_HISTO_M | IOS_RQ_HISTO_M) /* * Lookup table for iostat flags to nvlist names. Basically a list * of all the nvlists a flag requires. Also specifies the order in * which data gets printed in zpool iostat. */ static const char *vsx_type_to_nvlist[IOS_COUNT][15] = { [IOS_L_HISTO] = { ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO, ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO, ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO, NULL}, [IOS_LATENCY] = { ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO, ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO, NULL}, [IOS_QUEUES] = { ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE, NULL}, [IOS_RQ_HISTO] = { ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO, ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO, ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO, ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO, ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO, ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO, ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO, ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO, ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO, ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO, NULL}, }; /* * Given a cb->cb_flags with a histogram bit set, return the iostat_type. * Right now, only one histo bit is ever set at one time, so we can * just do a highbit64(a) */ #define IOS_HISTO_IDX(a) (highbit64(a & IOS_ANYHISTO_M) - 1) typedef struct zpool_command { const char *name; int (*func)(int, char **); zpool_help_t usage; } zpool_command_t; /* * Master command table. Each ZFS command has a name, associated function, and * usage message. The usage messages need to be internationalized, so we have * to have a function to return the usage message based on a command index. * * These commands are organized according to how they are displayed in the usage * message. An empty command (one with a NULL name) indicates an empty line in * the generic usage message. */ static zpool_command_t command_table[] = { { "version", zpool_do_version, HELP_VERSION }, { NULL }, { "create", zpool_do_create, HELP_CREATE }, { "destroy", zpool_do_destroy, HELP_DESTROY }, { NULL }, { "add", zpool_do_add, HELP_ADD }, { "remove", zpool_do_remove, HELP_REMOVE }, { NULL }, { "labelclear", zpool_do_labelclear, HELP_LABELCLEAR }, { NULL }, { "checkpoint", zpool_do_checkpoint, HELP_CHECKPOINT }, { NULL }, { "list", zpool_do_list, HELP_LIST }, { "iostat", zpool_do_iostat, HELP_IOSTAT }, { "status", zpool_do_status, HELP_STATUS }, { NULL }, { "online", zpool_do_online, HELP_ONLINE }, { "offline", zpool_do_offline, HELP_OFFLINE }, { "clear", zpool_do_clear, HELP_CLEAR }, { "reopen", zpool_do_reopen, HELP_REOPEN }, { NULL }, { "attach", zpool_do_attach, HELP_ATTACH }, { "detach", zpool_do_detach, HELP_DETACH }, { "replace", zpool_do_replace, HELP_REPLACE }, { "split", zpool_do_split, HELP_SPLIT }, { NULL }, { "initialize", zpool_do_initialize, HELP_INITIALIZE }, { "resilver", zpool_do_resilver, HELP_RESILVER }, { "scrub", zpool_do_scrub, HELP_SCRUB }, { "trim", zpool_do_trim, HELP_TRIM }, { NULL }, { "import", zpool_do_import, HELP_IMPORT }, { "export", zpool_do_export, HELP_EXPORT }, { "upgrade", zpool_do_upgrade, HELP_UPGRADE }, { "reguid", zpool_do_reguid, HELP_REGUID }, { NULL }, { "history", zpool_do_history, HELP_HISTORY }, { "events", zpool_do_events, HELP_EVENTS }, { NULL }, { "get", zpool_do_get, HELP_GET }, { "set", zpool_do_set, HELP_SET }, { "sync", zpool_do_sync, HELP_SYNC }, { NULL }, { "wait", zpool_do_wait, HELP_WAIT }, }; #define NCOMMAND (ARRAY_SIZE(command_table)) #define VDEV_ALLOC_CLASS_LOGS "logs" static zpool_command_t *current_command; static zfs_type_t current_prop_type = (ZFS_TYPE_POOL | ZFS_TYPE_VDEV); static char history_str[HIS_MAX_RECORD_LEN]; static boolean_t log_history = B_TRUE; static uint_t timestamp_fmt = NODATE; static const char * get_usage(zpool_help_t idx) { switch (idx) { case HELP_ADD: return (gettext("\tadd [-afgLnP] [-o property=value] " " ...\n")); case HELP_ATTACH: return (gettext("\tattach [-fsw] [-o property=value] " " \n")); case HELP_CLEAR: return (gettext("\tclear [[--power]|[-nF]] [device]\n")); case HELP_CREATE: return (gettext("\tcreate [-fnd] [-o property=value] ... \n" "\t [-O file-system-property=value] ... \n" "\t [-m mountpoint] [-R root] ...\n")); case HELP_CHECKPOINT: return (gettext("\tcheckpoint [-d [-w]] ...\n")); case HELP_DESTROY: return (gettext("\tdestroy [-f] \n")); case HELP_DETACH: return (gettext("\tdetach \n")); case HELP_EXPORT: return (gettext("\texport [-af] ...\n")); case HELP_HISTORY: return (gettext("\thistory [-il] [] ...\n")); case HELP_IMPORT: return (gettext("\timport [-d dir] [-D]\n" "\timport [-o mntopts] [-o property=value] ... \n" "\t [-d dir | -c cachefile] [-D] [-l] [-f] [-m] [-N] " "[-R root] [-F [-n]] -a\n" "\timport [-o mntopts] [-o property=value] ... \n" "\t [-d dir | -c cachefile] [-D] [-l] [-f] [-m] [-N] " "[-R root] [-F [-n]]\n" "\t [--rewind-to-checkpoint] [newpool]\n")); case HELP_IOSTAT: return (gettext("\tiostat [[[-c [script1,script2,...]" "[-lq]]|[-rw]] [-T d | u] [-ghHLpPvy]\n" "\t [[pool ...]|[pool vdev ...]|[vdev ...]]" " [[-n] interval [count]]\n")); case HELP_LABELCLEAR: return (gettext("\tlabelclear [-f] \n")); case HELP_LIST: return (gettext("\tlist [-gHLpPv] [-o property[,...]] " "[-T d|u] [pool] ... \n" "\t [interval [count]]\n")); case HELP_OFFLINE: return (gettext("\toffline [--power]|[[-f][-t]] " " ...\n")); case HELP_ONLINE: return (gettext("\tonline [--power][-e] " "...\n")); case HELP_REPLACE: return (gettext("\treplace [-fsw] [-o property=value] " " [new-device]\n")); case HELP_REMOVE: return (gettext("\tremove [-npsw] ...\n")); case HELP_REOPEN: return (gettext("\treopen [-n] \n")); case HELP_INITIALIZE: return (gettext("\tinitialize [-c | -s | -u] [-w] " "[ ...]\n")); case HELP_SCRUB: return (gettext("\tscrub [-s | -p] [-w] [-e] ...\n")); case HELP_RESILVER: return (gettext("\tresilver ...\n")); case HELP_TRIM: return (gettext("\ttrim [-dw] [-r ] [-c | -s] " "[ ...]\n")); case HELP_STATUS: return (gettext("\tstatus [--power] [-c [script1,script2,...]] " "[-DegiLpPstvx] [-T d|u] [pool] ...\n" "\t [interval [count]]\n")); case HELP_UPGRADE: return (gettext("\tupgrade\n" "\tupgrade -v\n" "\tupgrade [-V version] <-a | pool ...>\n")); case HELP_EVENTS: return (gettext("\tevents [-vHf [pool] | -c]\n")); case HELP_GET: return (gettext("\tget [-Hp] [-o \"all\" | field[,...]] " "<\"all\" | property[,...]> ...\n")); case HELP_SET: return (gettext("\tset \n" "\tset \n")); case HELP_SPLIT: return (gettext("\tsplit [-gLnPl] [-R altroot] [-o mntopts]\n" "\t [-o property=value] " "[ ...]\n")); case HELP_REGUID: return (gettext("\treguid \n")); case HELP_SYNC: return (gettext("\tsync [pool] ...\n")); case HELP_VERSION: return (gettext("\tversion\n")); case HELP_WAIT: return (gettext("\twait [-Hp] [-T d|u] [-t [,...]] " " [interval]\n")); default: __builtin_unreachable(); } } static void zpool_collect_leaves(zpool_handle_t *zhp, nvlist_t *nvroot, nvlist_t *res) { uint_t children = 0; nvlist_t **child; uint_t i; (void) nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, &children); if (children == 0) { char *path = zpool_vdev_name(g_zfs, zhp, nvroot, VDEV_NAME_PATH); if (strcmp(path, VDEV_TYPE_INDIRECT) != 0 && strcmp(path, VDEV_TYPE_HOLE) != 0) fnvlist_add_boolean(res, path); free(path); return; } for (i = 0; i < children; i++) { zpool_collect_leaves(zhp, child[i], res); } } /* * Callback routine that will print out a pool property value. */ static int print_pool_prop_cb(int prop, void *cb) { FILE *fp = cb; (void) fprintf(fp, "\t%-19s ", zpool_prop_to_name(prop)); if (zpool_prop_readonly(prop)) (void) fprintf(fp, " NO "); else (void) fprintf(fp, " YES "); if (zpool_prop_values(prop) == NULL) (void) fprintf(fp, "-\n"); else (void) fprintf(fp, "%s\n", zpool_prop_values(prop)); return (ZPROP_CONT); } /* * Callback routine that will print out a vdev property value. */ static int print_vdev_prop_cb(int prop, void *cb) { FILE *fp = cb; (void) fprintf(fp, "\t%-19s ", vdev_prop_to_name(prop)); if (vdev_prop_readonly(prop)) (void) fprintf(fp, " NO "); else (void) fprintf(fp, " YES "); if (vdev_prop_values(prop) == NULL) (void) fprintf(fp, "-\n"); else (void) fprintf(fp, "%s\n", vdev_prop_values(prop)); return (ZPROP_CONT); } /* * Given a leaf vdev name like 'L5' return its VDEV_CONFIG_PATH like * '/dev/disk/by-vdev/L5'. */ static const char * vdev_name_to_path(zpool_handle_t *zhp, char *vdev) { nvlist_t *vdev_nv = zpool_find_vdev(zhp, vdev, NULL, NULL, NULL); if (vdev_nv == NULL) { return (NULL); } return (fnvlist_lookup_string(vdev_nv, ZPOOL_CONFIG_PATH)); } static int zpool_power_on(zpool_handle_t *zhp, char *vdev) { return (zpool_power(zhp, vdev, B_TRUE)); } static int zpool_power_on_and_disk_wait(zpool_handle_t *zhp, char *vdev) { int rc; rc = zpool_power_on(zhp, vdev); if (rc != 0) return (rc); zpool_disk_wait(vdev_name_to_path(zhp, vdev)); return (0); } static int zpool_power_on_pool_and_wait_for_devices(zpool_handle_t *zhp) { nvlist_t *nv; const char *path = NULL; int rc; /* Power up all the devices first */ FOR_EACH_REAL_LEAF_VDEV(zhp, nv) { path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH); if (path != NULL) { rc = zpool_power_on(zhp, (char *)path); if (rc != 0) { return (rc); } } } /* * Wait for their devices to show up. Since we powered them on * at roughly the same time, they should all come online around * the same time. */ FOR_EACH_REAL_LEAF_VDEV(zhp, nv) { path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH); zpool_disk_wait(path); } return (0); } static int zpool_power_off(zpool_handle_t *zhp, char *vdev) { return (zpool_power(zhp, vdev, B_FALSE)); } /* * Display usage message. If we're inside a command, display only the usage for * that command. Otherwise, iterate over the entire command table and display * a complete usage message. */ static __attribute__((noreturn)) void usage(boolean_t requested) { FILE *fp = requested ? stdout : stderr; if (current_command == NULL) { int i; (void) fprintf(fp, gettext("usage: zpool command args ...\n")); (void) fprintf(fp, gettext("where 'command' is one of the following:\n\n")); for (i = 0; i < NCOMMAND; i++) { if (command_table[i].name == NULL) (void) fprintf(fp, "\n"); else (void) fprintf(fp, "%s", get_usage(command_table[i].usage)); } (void) fprintf(fp, gettext("\nFor further help on a command or topic, " "run: %s\n"), "zpool help []"); } else { (void) fprintf(fp, gettext("usage:\n")); (void) fprintf(fp, "%s", get_usage(current_command->usage)); } if (current_command != NULL && current_prop_type != (ZFS_TYPE_POOL | ZFS_TYPE_VDEV) && ((strcmp(current_command->name, "set") == 0) || (strcmp(current_command->name, "get") == 0) || (strcmp(current_command->name, "list") == 0))) { (void) fprintf(fp, "%s", gettext("\nthe following properties are supported:\n")); (void) fprintf(fp, "\n\t%-19s %s %s\n\n", "PROPERTY", "EDIT", "VALUES"); /* Iterate over all properties */ if (current_prop_type == ZFS_TYPE_POOL) { (void) zprop_iter(print_pool_prop_cb, fp, B_FALSE, B_TRUE, current_prop_type); (void) fprintf(fp, "\t%-19s ", "feature@..."); (void) fprintf(fp, "YES " "disabled | enabled | active\n"); (void) fprintf(fp, gettext("\nThe feature@ properties " "must be appended with a feature name.\n" "See zpool-features(7).\n")); } else if (current_prop_type == ZFS_TYPE_VDEV) { (void) zprop_iter(print_vdev_prop_cb, fp, B_FALSE, B_TRUE, current_prop_type); } } /* * See comments at end of main(). */ if (getenv("ZFS_ABORT") != NULL) { (void) printf("dumping core by request\n"); abort(); } exit(requested ? 0 : 2); } /* * zpool initialize [-c | -s | -u] [-w] [ ...] * Initialize all unused blocks in the specified vdevs, or all vdevs in the pool * if none specified. * * -c Cancel. Ends active initializing. * -s Suspend. Initializing can then be restarted with no flags. * -u Uninitialize. Clears initialization state. * -w Wait. Blocks until initializing has completed. */ int zpool_do_initialize(int argc, char **argv) { int c; char *poolname; zpool_handle_t *zhp; nvlist_t *vdevs; int err = 0; boolean_t wait = B_FALSE; struct option long_options[] = { {"cancel", no_argument, NULL, 'c'}, {"suspend", no_argument, NULL, 's'}, {"uninit", no_argument, NULL, 'u'}, {"wait", no_argument, NULL, 'w'}, {0, 0, 0, 0} }; pool_initialize_func_t cmd_type = POOL_INITIALIZE_START; while ((c = getopt_long(argc, argv, "csuw", long_options, NULL)) != -1) { switch (c) { case 'c': if (cmd_type != POOL_INITIALIZE_START && cmd_type != POOL_INITIALIZE_CANCEL) { (void) fprintf(stderr, gettext("-c cannot be " "combined with other options\n")); usage(B_FALSE); } cmd_type = POOL_INITIALIZE_CANCEL; break; case 's': if (cmd_type != POOL_INITIALIZE_START && cmd_type != POOL_INITIALIZE_SUSPEND) { (void) fprintf(stderr, gettext("-s cannot be " "combined with other options\n")); usage(B_FALSE); } cmd_type = POOL_INITIALIZE_SUSPEND; break; case 'u': if (cmd_type != POOL_INITIALIZE_START && cmd_type != POOL_INITIALIZE_UNINIT) { (void) fprintf(stderr, gettext("-u cannot be " "combined with other options\n")); usage(B_FALSE); } cmd_type = POOL_INITIALIZE_UNINIT; break; case 'w': wait = B_TRUE; break; case '?': if (optopt != 0) { (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); } else { (void) fprintf(stderr, gettext("invalid option '%s'\n"), argv[optind - 1]); } usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); return (-1); } if (wait && (cmd_type != POOL_INITIALIZE_START)) { (void) fprintf(stderr, gettext("-w cannot be used with -c, -s" "or -u\n")); usage(B_FALSE); } poolname = argv[0]; zhp = zpool_open(g_zfs, poolname); if (zhp == NULL) return (-1); vdevs = fnvlist_alloc(); if (argc == 1) { /* no individual leaf vdevs specified, so add them all */ nvlist_t *config = zpool_get_config(zhp, NULL); nvlist_t *nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); zpool_collect_leaves(zhp, nvroot, vdevs); } else { for (int i = 1; i < argc; i++) { fnvlist_add_boolean(vdevs, argv[i]); } } if (wait) err = zpool_initialize_wait(zhp, cmd_type, vdevs); else err = zpool_initialize(zhp, cmd_type, vdevs); fnvlist_free(vdevs); zpool_close(zhp); return (err); } /* * print a pool vdev config for dry runs */ static void print_vdev_tree(zpool_handle_t *zhp, const char *name, nvlist_t *nv, int indent, const char *match, int name_flags) { nvlist_t **child; uint_t c, children; char *vname; boolean_t printed = B_FALSE; if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { if (name != NULL) (void) printf("\t%*s%s\n", indent, "", name); return; } for (c = 0; c < children; c++) { uint64_t is_log = B_FALSE, is_hole = B_FALSE; const char *class = ""; (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, &is_hole); if (is_hole == B_TRUE) { continue; } (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &is_log); if (is_log) class = VDEV_ALLOC_BIAS_LOG; (void) nvlist_lookup_string(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS, &class); if (strcmp(match, class) != 0) continue; if (!printed && name != NULL) { (void) printf("\t%*s%s\n", indent, "", name); printed = B_TRUE; } vname = zpool_vdev_name(g_zfs, zhp, child[c], name_flags); print_vdev_tree(zhp, vname, child[c], indent + 2, "", name_flags); free(vname); } } /* * Print the list of l2cache devices for dry runs. */ static void print_cache_list(nvlist_t *nv, int indent) { nvlist_t **child; uint_t c, children; if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, &child, &children) == 0 && children > 0) { (void) printf("\t%*s%s\n", indent, "", "cache"); } else { return; } for (c = 0; c < children; c++) { char *vname; vname = zpool_vdev_name(g_zfs, NULL, child[c], 0); (void) printf("\t%*s%s\n", indent + 2, "", vname); free(vname); } } /* * Print the list of spares for dry runs. */ static void print_spare_list(nvlist_t *nv, int indent) { nvlist_t **child; uint_t c, children; if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child, &children) == 0 && children > 0) { (void) printf("\t%*s%s\n", indent, "", "spares"); } else { return; } for (c = 0; c < children; c++) { char *vname; vname = zpool_vdev_name(g_zfs, NULL, child[c], 0); (void) printf("\t%*s%s\n", indent + 2, "", vname); free(vname); } } static boolean_t prop_list_contains_feature(nvlist_t *proplist) { nvpair_t *nvp; for (nvp = nvlist_next_nvpair(proplist, NULL); NULL != nvp; nvp = nvlist_next_nvpair(proplist, nvp)) { if (zpool_prop_feature(nvpair_name(nvp))) return (B_TRUE); } return (B_FALSE); } /* * Add a property pair (name, string-value) into a property nvlist. */ static int add_prop_list(const char *propname, const char *propval, nvlist_t **props, boolean_t poolprop) { zpool_prop_t prop = ZPOOL_PROP_INVAL; nvlist_t *proplist; const char *normnm; const char *strval; if (*props == NULL && nvlist_alloc(props, NV_UNIQUE_NAME, 0) != 0) { (void) fprintf(stderr, gettext("internal error: out of memory\n")); return (1); } proplist = *props; if (poolprop) { const char *vname = zpool_prop_to_name(ZPOOL_PROP_VERSION); const char *cname = zpool_prop_to_name(ZPOOL_PROP_COMPATIBILITY); if ((prop = zpool_name_to_prop(propname)) == ZPOOL_PROP_INVAL && (!zpool_prop_feature(propname) && !zpool_prop_vdev(propname))) { (void) fprintf(stderr, gettext("property '%s' is " "not a valid pool or vdev property\n"), propname); return (2); } /* * feature@ properties and version should not be specified * at the same time. */ if ((prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname) && nvlist_exists(proplist, vname)) || (prop == ZPOOL_PROP_VERSION && prop_list_contains_feature(proplist))) { (void) fprintf(stderr, gettext("'feature@' and " "'version' properties cannot be specified " "together\n")); return (2); } /* * if version is specified, only "legacy" compatibility * may be requested */ if ((prop == ZPOOL_PROP_COMPATIBILITY && strcmp(propval, ZPOOL_COMPAT_LEGACY) != 0 && nvlist_exists(proplist, vname)) || (prop == ZPOOL_PROP_VERSION && nvlist_exists(proplist, cname) && strcmp(fnvlist_lookup_string(proplist, cname), ZPOOL_COMPAT_LEGACY) != 0)) { (void) fprintf(stderr, gettext("when 'version' is " "specified, the 'compatibility' feature may only " "be set to '" ZPOOL_COMPAT_LEGACY "'\n")); return (2); } if (zpool_prop_feature(propname) || zpool_prop_vdev(propname)) normnm = propname; else normnm = zpool_prop_to_name(prop); } else { zfs_prop_t fsprop = zfs_name_to_prop(propname); if (zfs_prop_valid_for_type(fsprop, ZFS_TYPE_FILESYSTEM, B_FALSE)) { normnm = zfs_prop_to_name(fsprop); } else if (zfs_prop_user(propname) || zfs_prop_userquota(propname)) { normnm = propname; } else { (void) fprintf(stderr, gettext("property '%s' is " "not a valid filesystem property\n"), propname); return (2); } } if (nvlist_lookup_string(proplist, normnm, &strval) == 0 && prop != ZPOOL_PROP_CACHEFILE) { (void) fprintf(stderr, gettext("property '%s' " "specified multiple times\n"), propname); return (2); } if (nvlist_add_string(proplist, normnm, propval) != 0) { (void) fprintf(stderr, gettext("internal " "error: out of memory\n")); return (1); } return (0); } /* * Set a default property pair (name, string-value) in a property nvlist */ static int add_prop_list_default(const char *propname, const char *propval, nvlist_t **props) { const char *pval; if (nvlist_lookup_string(*props, propname, &pval) == 0) return (0); return (add_prop_list(propname, propval, props, B_TRUE)); } /* * zpool add [-afgLnP] [-o property=value] ... * * -a Disable the ashift validation checks * -f Force addition of devices, even if they appear in use * -g Display guid for individual vdev name. * -L Follow links when resolving vdev path name. * -n Do not add the devices, but display the resulting layout if * they were to be added. * -o Set property=value. * -P Display full path for vdev name. * * Adds the given vdevs to 'pool'. As with create, the bulk of this work is * handled by make_root_vdev(), which constructs the nvlist needed to pass to * libzfs. */ int zpool_do_add(int argc, char **argv) { boolean_t check_replication = B_TRUE; boolean_t check_inuse = B_TRUE; boolean_t dryrun = B_FALSE; boolean_t check_ashift = B_TRUE; boolean_t force = B_FALSE; int name_flags = 0; int c; nvlist_t *nvroot; char *poolname; int ret; zpool_handle_t *zhp; nvlist_t *config; nvlist_t *props = NULL; char *propval; struct option long_options[] = { {"allow-in-use", no_argument, NULL, ZPOOL_OPTION_ALLOW_INUSE}, {"allow-replication-mismatch", no_argument, NULL, ZPOOL_OPTION_ALLOW_REPLICATION_MISMATCH}, {"allow-ashift-mismatch", no_argument, NULL, ZPOOL_OPTION_ALLOW_ASHIFT_MISMATCH}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, "fgLno:P", long_options, NULL)) != -1) { switch (c) { case 'f': force = B_TRUE; break; case 'g': name_flags |= VDEV_NAME_GUID; break; case 'L': name_flags |= VDEV_NAME_FOLLOW_LINKS; break; case 'n': dryrun = B_TRUE; break; case 'o': if ((propval = strchr(optarg, '=')) == NULL) { (void) fprintf(stderr, gettext("missing " "'=' for -o option\n")); usage(B_FALSE); } *propval = '\0'; propval++; if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) || (add_prop_list(optarg, propval, &props, B_TRUE))) usage(B_FALSE); break; case 'P': name_flags |= VDEV_NAME_PATH; break; case ZPOOL_OPTION_ALLOW_INUSE: check_inuse = B_FALSE; break; case ZPOOL_OPTION_ALLOW_REPLICATION_MISMATCH: check_replication = B_FALSE; break; case ZPOOL_OPTION_ALLOW_ASHIFT_MISMATCH: check_ashift = B_FALSE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing vdev specification\n")); usage(B_FALSE); } if (force) { if (!check_inuse || !check_replication || !check_ashift) { (void) fprintf(stderr, gettext("'-f' option is not " "allowed with '--allow-replication-mismatch', " "'--allow-ashift-mismatch', or " "'--allow-in-use'\n")); usage(B_FALSE); } check_inuse = B_FALSE; check_replication = B_FALSE; check_ashift = B_FALSE; } poolname = argv[0]; argc--; argv++; if ((zhp = zpool_open(g_zfs, poolname)) == NULL) return (1); if ((config = zpool_get_config(zhp, NULL)) == NULL) { (void) fprintf(stderr, gettext("pool '%s' is unavailable\n"), poolname); zpool_close(zhp); return (1); } /* unless manually specified use "ashift" pool property (if set) */ if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) { int intval; zprop_source_t src; char strval[ZPOOL_MAXPROPLEN]; intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src); if (src != ZPROP_SRC_DEFAULT) { (void) sprintf(strval, "%" PRId32, intval); verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval, &props, B_TRUE) == 0); } } /* pass off to make_root_vdev for processing */ nvroot = make_root_vdev(zhp, props, !check_inuse, check_replication, B_FALSE, dryrun, argc, argv); if (nvroot == NULL) { zpool_close(zhp); return (1); } if (dryrun) { nvlist_t *poolnvroot; nvlist_t **l2child, **sparechild; uint_t l2children, sparechildren, c; char *vname; boolean_t hadcache = B_FALSE, hadspare = B_FALSE; verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &poolnvroot) == 0); (void) printf(gettext("would update '%s' to the following " "configuration:\n\n"), zpool_get_name(zhp)); /* print original main pool and new tree */ print_vdev_tree(zhp, poolname, poolnvroot, 0, "", name_flags | VDEV_NAME_TYPE_ID); print_vdev_tree(zhp, NULL, nvroot, 0, "", name_flags); /* print other classes: 'dedup', 'special', and 'log' */ if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_DEDUP)) { print_vdev_tree(zhp, "dedup", poolnvroot, 0, VDEV_ALLOC_BIAS_DEDUP, name_flags); print_vdev_tree(zhp, NULL, nvroot, 0, VDEV_ALLOC_BIAS_DEDUP, name_flags); } else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_DEDUP)) { print_vdev_tree(zhp, "dedup", nvroot, 0, VDEV_ALLOC_BIAS_DEDUP, name_flags); } if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_SPECIAL)) { print_vdev_tree(zhp, "special", poolnvroot, 0, VDEV_ALLOC_BIAS_SPECIAL, name_flags); print_vdev_tree(zhp, NULL, nvroot, 0, VDEV_ALLOC_BIAS_SPECIAL, name_flags); } else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_SPECIAL)) { print_vdev_tree(zhp, "special", nvroot, 0, VDEV_ALLOC_BIAS_SPECIAL, name_flags); } if (num_logs(poolnvroot) > 0) { print_vdev_tree(zhp, "logs", poolnvroot, 0, VDEV_ALLOC_BIAS_LOG, name_flags); print_vdev_tree(zhp, NULL, nvroot, 0, VDEV_ALLOC_BIAS_LOG, name_flags); } else if (num_logs(nvroot) > 0) { print_vdev_tree(zhp, "logs", nvroot, 0, VDEV_ALLOC_BIAS_LOG, name_flags); } /* Do the same for the caches */ if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_L2CACHE, &l2child, &l2children) == 0 && l2children) { hadcache = B_TRUE; (void) printf(gettext("\tcache\n")); for (c = 0; c < l2children; c++) { vname = zpool_vdev_name(g_zfs, NULL, l2child[c], name_flags); (void) printf("\t %s\n", vname); free(vname); } } if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2child, &l2children) == 0 && l2children) { if (!hadcache) (void) printf(gettext("\tcache\n")); for (c = 0; c < l2children; c++) { vname = zpool_vdev_name(g_zfs, NULL, l2child[c], name_flags); (void) printf("\t %s\n", vname); free(vname); } } /* And finally the spares */ if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_SPARES, &sparechild, &sparechildren) == 0 && sparechildren > 0) { hadspare = B_TRUE; (void) printf(gettext("\tspares\n")); for (c = 0; c < sparechildren; c++) { vname = zpool_vdev_name(g_zfs, NULL, sparechild[c], name_flags); (void) printf("\t %s\n", vname); free(vname); } } if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &sparechild, &sparechildren) == 0 && sparechildren > 0) { if (!hadspare) (void) printf(gettext("\tspares\n")); for (c = 0; c < sparechildren; c++) { vname = zpool_vdev_name(g_zfs, NULL, sparechild[c], name_flags); (void) printf("\t %s\n", vname); free(vname); } } ret = 0; } else { ret = (zpool_add(zhp, nvroot, check_ashift) != 0); } nvlist_free(props); nvlist_free(nvroot); zpool_close(zhp); return (ret); } /* * zpool remove [-npsw] ... * * Removes the given vdev from the pool. */ int zpool_do_remove(int argc, char **argv) { char *poolname; int i, ret = 0; zpool_handle_t *zhp = NULL; boolean_t stop = B_FALSE; int c; boolean_t noop = B_FALSE; boolean_t parsable = B_FALSE; boolean_t wait = B_FALSE; /* check options */ while ((c = getopt(argc, argv, "npsw")) != -1) { switch (c) { case 'n': noop = B_TRUE; break; case 'p': parsable = B_TRUE; break; case 's': stop = B_TRUE; break; case 'w': wait = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); } poolname = argv[0]; if ((zhp = zpool_open(g_zfs, poolname)) == NULL) return (1); if (stop && noop) { zpool_close(zhp); (void) fprintf(stderr, gettext("stop request ignored\n")); return (0); } if (stop) { if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if (zpool_vdev_remove_cancel(zhp) != 0) ret = 1; if (wait) { (void) fprintf(stderr, gettext("invalid option " "combination: -w cannot be used with -s\n")); usage(B_FALSE); } } else { if (argc < 2) { (void) fprintf(stderr, gettext("missing device\n")); usage(B_FALSE); } for (i = 1; i < argc; i++) { if (noop) { uint64_t size; if (zpool_vdev_indirect_size(zhp, argv[i], &size) != 0) { ret = 1; break; } if (parsable) { (void) printf("%s %llu\n", argv[i], (unsigned long long)size); } else { char valstr[32]; zfs_nicenum(size, valstr, sizeof (valstr)); (void) printf("Memory that will be " "used after removing %s: %s\n", argv[i], valstr); } } else { if (zpool_vdev_remove(zhp, argv[i]) != 0) ret = 1; } } if (ret == 0 && wait) ret = zpool_wait(zhp, ZPOOL_WAIT_REMOVE); } zpool_close(zhp); return (ret); } /* * Return 1 if a vdev is active (being used in a pool) * Return 0 if a vdev is inactive (offlined or faulted, or not in active pool) * * This is useful for checking if a disk in an active pool is offlined or * faulted. */ static int vdev_is_active(char *vdev_path) { int fd; fd = open(vdev_path, O_EXCL); if (fd < 0) { return (1); /* cant open O_EXCL - disk is active */ } close(fd); return (0); /* disk is inactive in the pool */ } /* * zpool labelclear [-f] * * -f Force clearing the label for the vdevs which are members of * the exported or foreign pools. * * Verifies that the vdev is not active and zeros out the label information * on the device. */ int zpool_do_labelclear(int argc, char **argv) { char vdev[MAXPATHLEN]; char *name = NULL; int c, fd = -1, ret = 0; nvlist_t *config; pool_state_t state; boolean_t inuse = B_FALSE; boolean_t force = B_FALSE; /* check options */ while ((c = getopt(argc, argv, "f")) != -1) { switch (c) { case 'f': force = B_TRUE; break; default: (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* get vdev name */ if (argc < 1) { (void) fprintf(stderr, gettext("missing vdev name\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } (void) strlcpy(vdev, argv[0], sizeof (vdev)); /* * If we cannot open an absolute path, we quit. * Otherwise if the provided vdev name doesn't point to a file, * try prepending expected disk paths and partition numbers. */ if ((fd = open(vdev, O_RDWR)) < 0) { int error; if (vdev[0] == '/') { (void) fprintf(stderr, gettext("failed to open " "%s: %s\n"), vdev, strerror(errno)); return (1); } error = zfs_resolve_shortname(argv[0], vdev, MAXPATHLEN); if (error == 0 && zfs_dev_is_whole_disk(vdev)) { if (zfs_append_partition(vdev, MAXPATHLEN) == -1) error = ENOENT; } if (error || ((fd = open(vdev, O_RDWR)) < 0)) { if (errno == ENOENT) { (void) fprintf(stderr, gettext( "failed to find device %s, try " "specifying absolute path instead\n"), argv[0]); return (1); } (void) fprintf(stderr, gettext("failed to open %s:" " %s\n"), vdev, strerror(errno)); return (1); } } /* * Flush all dirty pages for the block device. This should not be * fatal when the device does not support BLKFLSBUF as would be the * case for a file vdev. */ if ((zfs_dev_flush(fd) != 0) && (errno != ENOTTY)) (void) fprintf(stderr, gettext("failed to invalidate " "cache for %s: %s\n"), vdev, strerror(errno)); if (zpool_read_label(fd, &config, NULL) != 0) { (void) fprintf(stderr, gettext("failed to read label from %s\n"), vdev); ret = 1; goto errout; } nvlist_free(config); ret = zpool_in_use(g_zfs, fd, &state, &name, &inuse); if (ret != 0) { (void) fprintf(stderr, gettext("failed to check state for %s\n"), vdev); ret = 1; goto errout; } if (!inuse) goto wipe_label; switch (state) { default: case POOL_STATE_ACTIVE: case POOL_STATE_SPARE: case POOL_STATE_L2CACHE: /* * We allow the user to call 'zpool offline -f' * on an offlined disk in an active pool. We can check if * the disk is online by calling vdev_is_active(). */ if (force && !vdev_is_active(vdev)) break; (void) fprintf(stderr, gettext( "%s is a member (%s) of pool \"%s\""), vdev, zpool_pool_state_to_name(state), name); if (force) { (void) fprintf(stderr, gettext( ". Offline the disk first to clear its label.")); } printf("\n"); ret = 1; goto errout; case POOL_STATE_EXPORTED: if (force) break; (void) fprintf(stderr, gettext( "use '-f' to override the following error:\n" "%s is a member of exported pool \"%s\"\n"), vdev, name); ret = 1; goto errout; case POOL_STATE_POTENTIALLY_ACTIVE: if (force) break; (void) fprintf(stderr, gettext( "use '-f' to override the following error:\n" "%s is a member of potentially active pool \"%s\"\n"), vdev, name); ret = 1; goto errout; case POOL_STATE_DESTROYED: /* inuse should never be set for a destroyed pool */ assert(0); break; } wipe_label: ret = zpool_clear_label(fd); if (ret != 0) { (void) fprintf(stderr, gettext("failed to clear label for %s\n"), vdev); } errout: free(name); (void) close(fd); return (ret); } /* * zpool create [-fnd] [-o property=value] ... * [-O file-system-property=value] ... * [-R root] [-m mountpoint] ... * * -f Force creation, even if devices appear in use * -n Do not create the pool, but display the resulting layout if it * were to be created. * -R Create a pool under an alternate root * -m Set default mountpoint for the root dataset. By default it's * '/' * -o Set property=value. * -o Set feature@feature=enabled|disabled. * -d Don't automatically enable all supported pool features * (individual features can be enabled with -o). * -O Set fsproperty=value in the pool's root file system * * Creates the named pool according to the given vdev specification. The * bulk of the vdev processing is done in make_root_vdev() in zpool_vdev.c. * Once we get the nvlist back from make_root_vdev(), we either print out the * contents (if '-n' was specified), or pass it to libzfs to do the creation. */ int zpool_do_create(int argc, char **argv) { boolean_t force = B_FALSE; boolean_t dryrun = B_FALSE; boolean_t enable_pool_features = B_TRUE; int c; nvlist_t *nvroot = NULL; char *poolname; char *tname = NULL; int ret = 1; char *altroot = NULL; char *compat = NULL; char *mountpoint = NULL; nvlist_t *fsprops = NULL; nvlist_t *props = NULL; char *propval; /* check options */ while ((c = getopt(argc, argv, ":fndR:m:o:O:t:")) != -1) { switch (c) { case 'f': force = B_TRUE; break; case 'n': dryrun = B_TRUE; break; case 'd': enable_pool_features = B_FALSE; break; case 'R': altroot = optarg; if (add_prop_list(zpool_prop_to_name( ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE)) goto errout; if (add_prop_list_default(zpool_prop_to_name( ZPOOL_PROP_CACHEFILE), "none", &props)) goto errout; break; case 'm': /* Equivalent to -O mountpoint=optarg */ mountpoint = optarg; break; case 'o': if ((propval = strchr(optarg, '=')) == NULL) { (void) fprintf(stderr, gettext("missing " "'=' for -o option\n")); goto errout; } *propval = '\0'; propval++; if (add_prop_list(optarg, propval, &props, B_TRUE)) goto errout; /* * If the user is creating a pool that doesn't support * feature flags, don't enable any features. */ if (zpool_name_to_prop(optarg) == ZPOOL_PROP_VERSION) { char *end; u_longlong_t ver; ver = strtoull(propval, &end, 10); if (*end == '\0' && ver < SPA_VERSION_FEATURES) { enable_pool_features = B_FALSE; } } if (zpool_name_to_prop(optarg) == ZPOOL_PROP_ALTROOT) altroot = propval; if (zpool_name_to_prop(optarg) == ZPOOL_PROP_COMPATIBILITY) compat = propval; break; case 'O': if ((propval = strchr(optarg, '=')) == NULL) { (void) fprintf(stderr, gettext("missing " "'=' for -O option\n")); goto errout; } *propval = '\0'; propval++; /* * Mountpoints are checked and then added later. * Uniquely among properties, they can be specified * more than once, to avoid conflict with -m. */ if (0 == strcmp(optarg, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT))) { mountpoint = propval; } else if (add_prop_list(optarg, propval, &fsprops, B_FALSE)) { goto errout; } break; case 't': /* * Sanity check temporary pool name. */ if (strchr(optarg, '/') != NULL) { (void) fprintf(stderr, gettext("cannot create " "'%s': invalid character '/' in temporary " "name\n"), optarg); (void) fprintf(stderr, gettext("use 'zfs " "create' to create a dataset\n")); goto errout; } if (add_prop_list(zpool_prop_to_name( ZPOOL_PROP_TNAME), optarg, &props, B_TRUE)) goto errout; if (add_prop_list_default(zpool_prop_to_name( ZPOOL_PROP_CACHEFILE), "none", &props)) goto errout; tname = optarg; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); goto badusage; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); goto badusage; } } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); goto badusage; } if (argc < 2) { (void) fprintf(stderr, gettext("missing vdev specification\n")); goto badusage; } poolname = argv[0]; /* * As a special case, check for use of '/' in the name, and direct the * user to use 'zfs create' instead. */ if (strchr(poolname, '/') != NULL) { (void) fprintf(stderr, gettext("cannot create '%s': invalid " "character '/' in pool name\n"), poolname); (void) fprintf(stderr, gettext("use 'zfs create' to " "create a dataset\n")); goto errout; } /* pass off to make_root_vdev for bulk processing */ nvroot = make_root_vdev(NULL, props, force, !force, B_FALSE, dryrun, argc - 1, argv + 1); if (nvroot == NULL) goto errout; /* make_root_vdev() allows 0 toplevel children if there are spares */ if (!zfs_allocatable_devs(nvroot)) { (void) fprintf(stderr, gettext("invalid vdev " "specification: at least one toplevel vdev must be " "specified\n")); goto errout; } if (altroot != NULL && altroot[0] != '/') { (void) fprintf(stderr, gettext("invalid alternate root '%s': " "must be an absolute path\n"), altroot); goto errout; } /* * Check the validity of the mountpoint and direct the user to use the * '-m' mountpoint option if it looks like its in use. */ if (mountpoint == NULL || (strcmp(mountpoint, ZFS_MOUNTPOINT_LEGACY) != 0 && strcmp(mountpoint, ZFS_MOUNTPOINT_NONE) != 0)) { char buf[MAXPATHLEN]; DIR *dirp; if (mountpoint && mountpoint[0] != '/') { (void) fprintf(stderr, gettext("invalid mountpoint " "'%s': must be an absolute path, 'legacy', or " "'none'\n"), mountpoint); goto errout; } if (mountpoint == NULL) { if (altroot != NULL) (void) snprintf(buf, sizeof (buf), "%s/%s", altroot, poolname); else (void) snprintf(buf, sizeof (buf), "/%s", poolname); } else { if (altroot != NULL) (void) snprintf(buf, sizeof (buf), "%s%s", altroot, mountpoint); else (void) snprintf(buf, sizeof (buf), "%s", mountpoint); } if ((dirp = opendir(buf)) == NULL && errno != ENOENT) { (void) fprintf(stderr, gettext("mountpoint '%s' : " "%s\n"), buf, strerror(errno)); (void) fprintf(stderr, gettext("use '-m' " "option to provide a different default\n")); goto errout; } else if (dirp) { int count = 0; while (count < 3 && readdir(dirp) != NULL) count++; (void) closedir(dirp); if (count > 2) { (void) fprintf(stderr, gettext("mountpoint " "'%s' exists and is not empty\n"), buf); (void) fprintf(stderr, gettext("use '-m' " "option to provide a " "different default\n")); goto errout; } } } /* * Now that the mountpoint's validity has been checked, ensure that * the property is set appropriately prior to creating the pool. */ if (mountpoint != NULL) { ret = add_prop_list(zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), mountpoint, &fsprops, B_FALSE); if (ret != 0) goto errout; } ret = 1; if (dryrun) { /* * For a dry run invocation, print out a basic message and run * through all the vdevs in the list and print out in an * appropriate hierarchy. */ (void) printf(gettext("would create '%s' with the " "following layout:\n\n"), poolname); print_vdev_tree(NULL, poolname, nvroot, 0, "", 0); print_vdev_tree(NULL, "dedup", nvroot, 0, VDEV_ALLOC_BIAS_DEDUP, 0); print_vdev_tree(NULL, "special", nvroot, 0, VDEV_ALLOC_BIAS_SPECIAL, 0); print_vdev_tree(NULL, "logs", nvroot, 0, VDEV_ALLOC_BIAS_LOG, 0); print_cache_list(nvroot, 0); print_spare_list(nvroot, 0); ret = 0; } else { /* * Load in feature set. * Note: if compatibility property not given, we'll have * NULL, which means 'all features'. */ boolean_t requested_features[SPA_FEATURES]; if (zpool_do_load_compat(compat, requested_features) != ZPOOL_COMPATIBILITY_OK) goto errout; /* * props contains list of features to enable. * For each feature: * - remove it if feature@name=disabled * - leave it there if feature@name=enabled * - add it if: * - enable_pool_features (ie: no '-d' or '-o version') * - it's supported by the kernel module * - it's in the requested feature set * - warn if it's enabled but not in compat */ for (spa_feature_t i = 0; i < SPA_FEATURES; i++) { char propname[MAXPATHLEN]; const char *propval; zfeature_info_t *feat = &spa_feature_table[i]; (void) snprintf(propname, sizeof (propname), "feature@%s", feat->fi_uname); if (!nvlist_lookup_string(props, propname, &propval)) { if (strcmp(propval, ZFS_FEATURE_DISABLED) == 0) { (void) nvlist_remove_all(props, propname); } else if (strcmp(propval, ZFS_FEATURE_ENABLED) == 0 && !requested_features[i]) { (void) fprintf(stderr, gettext( "Warning: feature \"%s\" enabled " "but is not in specified " "'compatibility' feature set.\n"), feat->fi_uname); } } else if ( enable_pool_features && feat->fi_zfs_mod_supported && requested_features[i]) { ret = add_prop_list(propname, ZFS_FEATURE_ENABLED, &props, B_TRUE); if (ret != 0) goto errout; } } ret = 1; if (zpool_create(g_zfs, poolname, nvroot, props, fsprops) == 0) { zfs_handle_t *pool = zfs_open(g_zfs, tname ? tname : poolname, ZFS_TYPE_FILESYSTEM); if (pool != NULL) { if (zfs_mount(pool, NULL, 0) == 0) { ret = zfs_share(pool, NULL); zfs_commit_shares(NULL); } zfs_close(pool); } } else if (libzfs_errno(g_zfs) == EZFS_INVALIDNAME) { (void) fprintf(stderr, gettext("pool name may have " "been omitted\n")); } } errout: nvlist_free(nvroot); nvlist_free(fsprops); nvlist_free(props); return (ret); badusage: nvlist_free(fsprops); nvlist_free(props); usage(B_FALSE); return (2); } /* * zpool destroy * * -f Forcefully unmount any datasets * * Destroy the given pool. Automatically unmounts any datasets in the pool. */ int zpool_do_destroy(int argc, char **argv) { boolean_t force = B_FALSE; int c; char *pool; zpool_handle_t *zhp; int ret; /* check options */ while ((c = getopt(argc, argv, "f")) != -1) { switch (c) { case 'f': force = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* check arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } pool = argv[0]; if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) { /* * As a special case, check for use of '/' in the name, and * direct the user to use 'zfs destroy' instead. */ if (strchr(pool, '/') != NULL) (void) fprintf(stderr, gettext("use 'zfs destroy' to " "destroy a dataset\n")); return (1); } if (zpool_disable_datasets(zhp, force) != 0) { (void) fprintf(stderr, gettext("could not destroy '%s': " "could not unmount datasets\n"), zpool_get_name(zhp)); zpool_close(zhp); return (1); } /* The history must be logged as part of the export */ log_history = B_FALSE; ret = (zpool_destroy(zhp, history_str) != 0); zpool_close(zhp); return (ret); } typedef struct export_cbdata { boolean_t force; boolean_t hardforce; } export_cbdata_t; /* * Export one pool */ static int zpool_export_one(zpool_handle_t *zhp, void *data) { export_cbdata_t *cb = data; if (zpool_disable_datasets(zhp, cb->force) != 0) return (1); /* The history must be logged as part of the export */ log_history = B_FALSE; if (cb->hardforce) { if (zpool_export_force(zhp, history_str) != 0) return (1); } else if (zpool_export(zhp, cb->force, history_str) != 0) { return (1); } return (0); } /* * zpool export [-f] ... * * -a Export all pools * -f Forcefully unmount datasets * * Export the given pools. By default, the command will attempt to cleanly * unmount any active datasets within the pool. If the '-f' flag is specified, * then the datasets will be forcefully unmounted. */ int zpool_do_export(int argc, char **argv) { export_cbdata_t cb; boolean_t do_all = B_FALSE; boolean_t force = B_FALSE; boolean_t hardforce = B_FALSE; int c, ret; /* check options */ while ((c = getopt(argc, argv, "afF")) != -1) { switch (c) { case 'a': do_all = B_TRUE; break; case 'f': force = B_TRUE; break; case 'F': hardforce = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } cb.force = force; cb.hardforce = hardforce; argc -= optind; argv += optind; if (do_all) { if (argc != 0) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } return (for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, zpool_export_one, &cb)); } /* check arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool argument\n")); usage(B_FALSE); } ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, zpool_export_one, &cb); return (ret); } /* * Given a vdev configuration, determine the maximum width needed for the device * name column. */ static int max_width(zpool_handle_t *zhp, nvlist_t *nv, int depth, int max, int name_flags) { static const char *const subtypes[] = {ZPOOL_CONFIG_SPARES, ZPOOL_CONFIG_L2CACHE, ZPOOL_CONFIG_CHILDREN}; char *name = zpool_vdev_name(g_zfs, zhp, nv, name_flags); max = MAX(strlen(name) + depth, max); free(name); nvlist_t **child; uint_t children; for (size_t i = 0; i < ARRAY_SIZE(subtypes); ++i) if (nvlist_lookup_nvlist_array(nv, subtypes[i], &child, &children) == 0) for (uint_t c = 0; c < children; ++c) max = MAX(max_width(zhp, child[c], depth + 2, max, name_flags), max); return (max); } typedef struct spare_cbdata { uint64_t cb_guid; zpool_handle_t *cb_zhp; } spare_cbdata_t; static boolean_t find_vdev(nvlist_t *nv, uint64_t search) { uint64_t guid; nvlist_t **child; uint_t c, children; if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 && search == guid) return (B_TRUE); if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) == 0) { for (c = 0; c < children; c++) if (find_vdev(child[c], search)) return (B_TRUE); } return (B_FALSE); } static int find_spare(zpool_handle_t *zhp, void *data) { spare_cbdata_t *cbp = data; nvlist_t *config, *nvroot; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); if (find_vdev(nvroot, cbp->cb_guid)) { cbp->cb_zhp = zhp; return (1); } zpool_close(zhp); return (0); } typedef struct status_cbdata { int cb_count; int cb_name_flags; int cb_namewidth; boolean_t cb_allpools; boolean_t cb_verbose; boolean_t cb_literal; boolean_t cb_explain; boolean_t cb_first; boolean_t cb_dedup_stats; boolean_t cb_print_unhealthy; boolean_t cb_print_status; boolean_t cb_print_slow_ios; boolean_t cb_print_vdev_init; boolean_t cb_print_vdev_trim; vdev_cmd_data_list_t *vcdl; boolean_t cb_print_power; } status_cbdata_t; /* Return 1 if string is NULL, empty, or whitespace; return 0 otherwise. */ static boolean_t is_blank_str(const char *str) { for (; str != NULL && *str != '\0'; ++str) if (!isblank(*str)) return (B_FALSE); return (B_TRUE); } /* Print command output lines for specific vdev in a specific pool */ static void zpool_print_cmd(vdev_cmd_data_list_t *vcdl, const char *pool, const char *path) { vdev_cmd_data_t *data; int i, j; const char *val; for (i = 0; i < vcdl->count; i++) { if ((strcmp(vcdl->data[i].path, path) != 0) || (strcmp(vcdl->data[i].pool, pool) != 0)) { /* Not the vdev we're looking for */ continue; } data = &vcdl->data[i]; /* Print out all the output values for this vdev */ for (j = 0; j < vcdl->uniq_cols_cnt; j++) { val = NULL; /* Does this vdev have values for this column? */ for (int k = 0; k < data->cols_cnt; k++) { if (strcmp(data->cols[k], vcdl->uniq_cols[j]) == 0) { /* yes it does, record the value */ val = data->lines[k]; break; } } /* * Mark empty values with dashes to make output * awk-able. */ if (val == NULL || is_blank_str(val)) val = "-"; printf("%*s", vcdl->uniq_cols_width[j], val); if (j < vcdl->uniq_cols_cnt - 1) fputs(" ", stdout); } /* Print out any values that aren't in a column at the end */ for (j = data->cols_cnt; j < data->lines_cnt; j++) { /* Did we have any columns? If so print a spacer. */ if (vcdl->uniq_cols_cnt > 0) fputs(" ", stdout); val = data->lines[j]; fputs(val ?: "", stdout); } break; } } /* * Print vdev initialization status for leaves */ static void print_status_initialize(vdev_stat_t *vs, boolean_t verbose) { if (verbose) { if ((vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE || vs->vs_initialize_state == VDEV_INITIALIZE_SUSPENDED || vs->vs_initialize_state == VDEV_INITIALIZE_COMPLETE) && !vs->vs_scan_removing) { char zbuf[1024]; char tbuf[256]; time_t t = vs->vs_initialize_action_time; int initialize_pct = 100; if (vs->vs_initialize_state != VDEV_INITIALIZE_COMPLETE) { initialize_pct = (vs->vs_initialize_bytes_done * 100 / (vs->vs_initialize_bytes_est + 1)); } (void) ctime_r(&t, tbuf); tbuf[24] = 0; switch (vs->vs_initialize_state) { case VDEV_INITIALIZE_SUSPENDED: (void) snprintf(zbuf, sizeof (zbuf), ", %s %s", gettext("suspended, started at"), tbuf); break; case VDEV_INITIALIZE_ACTIVE: (void) snprintf(zbuf, sizeof (zbuf), ", %s %s", gettext("started at"), tbuf); break; case VDEV_INITIALIZE_COMPLETE: (void) snprintf(zbuf, sizeof (zbuf), ", %s %s", gettext("completed at"), tbuf); break; } (void) printf(gettext(" (%d%% initialized%s)"), initialize_pct, zbuf); } else { (void) printf(gettext(" (uninitialized)")); } } else if (vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE) { (void) printf(gettext(" (initializing)")); } } /* * Print vdev TRIM status for leaves */ static void print_status_trim(vdev_stat_t *vs, boolean_t verbose) { if (verbose) { if ((vs->vs_trim_state == VDEV_TRIM_ACTIVE || vs->vs_trim_state == VDEV_TRIM_SUSPENDED || vs->vs_trim_state == VDEV_TRIM_COMPLETE) && !vs->vs_scan_removing) { char zbuf[1024]; char tbuf[256]; time_t t = vs->vs_trim_action_time; int trim_pct = 100; if (vs->vs_trim_state != VDEV_TRIM_COMPLETE) { trim_pct = (vs->vs_trim_bytes_done * 100 / (vs->vs_trim_bytes_est + 1)); } (void) ctime_r(&t, tbuf); tbuf[24] = 0; switch (vs->vs_trim_state) { case VDEV_TRIM_SUSPENDED: (void) snprintf(zbuf, sizeof (zbuf), ", %s %s", gettext("suspended, started at"), tbuf); break; case VDEV_TRIM_ACTIVE: (void) snprintf(zbuf, sizeof (zbuf), ", %s %s", gettext("started at"), tbuf); break; case VDEV_TRIM_COMPLETE: (void) snprintf(zbuf, sizeof (zbuf), ", %s %s", gettext("completed at"), tbuf); break; } (void) printf(gettext(" (%d%% trimmed%s)"), trim_pct, zbuf); } else if (vs->vs_trim_notsup) { (void) printf(gettext(" (trim unsupported)")); } else { (void) printf(gettext(" (untrimmed)")); } } else if (vs->vs_trim_state == VDEV_TRIM_ACTIVE) { (void) printf(gettext(" (trimming)")); } } /* * Return the color associated with a health string. This includes returning * NULL for no color change. */ static const char * health_str_to_color(const char *health) { if (strcmp(health, gettext("FAULTED")) == 0 || strcmp(health, gettext("SUSPENDED")) == 0 || strcmp(health, gettext("UNAVAIL")) == 0) { return (ANSI_RED); } if (strcmp(health, gettext("OFFLINE")) == 0 || strcmp(health, gettext("DEGRADED")) == 0 || strcmp(health, gettext("REMOVED")) == 0) { return (ANSI_YELLOW); } return (NULL); } /* * Called for each leaf vdev. Returns 0 if the vdev is healthy. * A vdev is unhealthy if any of the following are true: * 1) there are read, write, or checksum errors, * 2) its state is not ONLINE, or * 3) slow IO reporting was requested (-s) and there are slow IOs. */ static int vdev_health_check_cb(void *hdl_data, nvlist_t *nv, void *data) { status_cbdata_t *cb = data; vdev_stat_t *vs; uint_t vsc; (void) hdl_data; if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) != 0) return (1); if (vs->vs_checksum_errors || vs->vs_read_errors || vs->vs_write_errors || vs->vs_state != VDEV_STATE_HEALTHY) return (1); if (cb->cb_print_slow_ios && vs->vs_slow_ios) return (1); return (0); } /* * Print out configuration state as requested by status_callback. */ static void print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name, nvlist_t *nv, int depth, boolean_t isspare, vdev_rebuild_stat_t *vrs) { nvlist_t **child, *root; uint_t c, i, vsc, children; pool_scan_stat_t *ps = NULL; vdev_stat_t *vs; char rbuf[6], wbuf[6], cbuf[6]; char *vname; uint64_t notpresent; spare_cbdata_t spare_cb; const char *state; const char *type; const char *path = NULL; const char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL, *scolor = NULL; if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) children = 0; verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) == 0); verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); if (strcmp(type, VDEV_TYPE_INDIRECT) == 0) return; state = zpool_state_to_name(vs->vs_state, vs->vs_aux); if (isspare) { /* * For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for * online drives. */ if (vs->vs_aux == VDEV_AUX_SPARED) state = gettext("INUSE"); else if (vs->vs_state == VDEV_STATE_HEALTHY) state = gettext("AVAIL"); } /* * If '-e' is specified then top-level vdevs and their children * can be pruned if all of their leaves are healthy. */ if (cb->cb_print_unhealthy && depth > 0 && for_each_vdev_in_nvlist(nv, vdev_health_check_cb, cb) == 0) { return; } printf_color(health_str_to_color(state), "\t%*s%-*s %-8s", depth, "", cb->cb_namewidth - depth, name, state); if (!isspare) { if (vs->vs_read_errors) rcolor = ANSI_RED; if (vs->vs_write_errors) wcolor = ANSI_RED; if (vs->vs_checksum_errors) ccolor = ANSI_RED; if (vs->vs_slow_ios) scolor = ANSI_BLUE; if (cb->cb_literal) { fputc(' ', stdout); printf_color(rcolor, "%5llu", (u_longlong_t)vs->vs_read_errors); fputc(' ', stdout); printf_color(wcolor, "%5llu", (u_longlong_t)vs->vs_write_errors); fputc(' ', stdout); printf_color(ccolor, "%5llu", (u_longlong_t)vs->vs_checksum_errors); } else { zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf)); zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf)); zfs_nicenum(vs->vs_checksum_errors, cbuf, sizeof (cbuf)); fputc(' ', stdout); printf_color(rcolor, "%5s", rbuf); fputc(' ', stdout); printf_color(wcolor, "%5s", wbuf); fputc(' ', stdout); printf_color(ccolor, "%5s", cbuf); } if (cb->cb_print_slow_ios) { if (children == 0) { /* Only leafs vdevs have slow IOs */ zfs_nicenum(vs->vs_slow_ios, rbuf, sizeof (rbuf)); } else { snprintf(rbuf, sizeof (rbuf), "-"); } if (cb->cb_literal) printf_color(scolor, " %5llu", (u_longlong_t)vs->vs_slow_ios); else printf_color(scolor, " %5s", rbuf); } if (cb->cb_print_power) { if (children == 0) { /* Only leaf vdevs have physical slots */ switch (zpool_power_current_state(zhp, (char *) fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH))) { case 0: printf_color(ANSI_RED, " %5s", gettext("off")); break; case 1: printf(" %5s", gettext("on")); break; default: printf(" %5s", "-"); } } else { printf(" %5s", "-"); } } } if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, ¬present) == 0) { verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0); (void) printf(" %s %s", gettext("was"), path); } else if (vs->vs_aux != 0) { (void) printf(" "); color_start(ANSI_RED); switch (vs->vs_aux) { case VDEV_AUX_OPEN_FAILED: (void) printf(gettext("cannot open")); break; case VDEV_AUX_BAD_GUID_SUM: (void) printf(gettext("missing device")); break; case VDEV_AUX_NO_REPLICAS: (void) printf(gettext("insufficient replicas")); break; case VDEV_AUX_VERSION_NEWER: (void) printf(gettext("newer version")); break; case VDEV_AUX_UNSUP_FEAT: (void) printf(gettext("unsupported feature(s)")); break; case VDEV_AUX_ASHIFT_TOO_BIG: (void) printf(gettext("unsupported minimum blocksize")); break; case VDEV_AUX_SPARED: verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &spare_cb.cb_guid) == 0); if (zpool_iter(g_zfs, find_spare, &spare_cb) == 1) { if (strcmp(zpool_get_name(spare_cb.cb_zhp), zpool_get_name(zhp)) == 0) (void) printf(gettext("currently in " "use")); else (void) printf(gettext("in use by " "pool '%s'"), zpool_get_name(spare_cb.cb_zhp)); zpool_close(spare_cb.cb_zhp); } else { (void) printf(gettext("currently in use")); } break; case VDEV_AUX_ERR_EXCEEDED: if (vs->vs_read_errors + vs->vs_write_errors + vs->vs_checksum_errors == 0 && children == 0 && vs->vs_slow_ios > 0) { (void) printf(gettext("too many slow I/Os")); } else { (void) printf(gettext("too many errors")); } break; case VDEV_AUX_IO_FAILURE: (void) printf(gettext("experienced I/O failures")); break; case VDEV_AUX_BAD_LOG: (void) printf(gettext("bad intent log")); break; case VDEV_AUX_EXTERNAL: (void) printf(gettext("external device fault")); break; case VDEV_AUX_SPLIT_POOL: (void) printf(gettext("split into new pool")); break; case VDEV_AUX_ACTIVE: (void) printf(gettext("currently in use")); break; case VDEV_AUX_CHILDREN_OFFLINE: (void) printf(gettext("all children offline")); break; case VDEV_AUX_BAD_LABEL: (void) printf(gettext("invalid label")); break; default: (void) printf(gettext("corrupted data")); break; } color_end(); } else if (children == 0 && !isspare && getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") == NULL && VDEV_STAT_VALID(vs_physical_ashift, vsc) && vs->vs_configured_ashift < vs->vs_physical_ashift) { (void) printf( gettext(" block size: %dB configured, %dB native"), 1 << vs->vs_configured_ashift, 1 << vs->vs_physical_ashift); } if (vs->vs_scan_removing != 0) { (void) printf(gettext(" (removing)")); } else if (VDEV_STAT_VALID(vs_noalloc, vsc) && vs->vs_noalloc != 0) { (void) printf(gettext(" (non-allocating)")); } /* The root vdev has the scrub/resilver stats */ root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL), ZPOOL_CONFIG_VDEV_TREE); (void) nvlist_lookup_uint64_array(root, ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &c); /* * If you force fault a drive that's resilvering, its scan stats can * get frozen in time, giving the false impression that it's * being resilvered. That's why we check the state to see if the vdev * is healthy before reporting "resilvering" or "repairing". */ if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0 && vs->vs_state == VDEV_STATE_HEALTHY) { if (vs->vs_scan_processed != 0) { (void) printf(gettext(" (%s)"), (ps->pss_func == POOL_SCAN_RESILVER) ? "resilvering" : "repairing"); } else if (vs->vs_resilver_deferred) { (void) printf(gettext(" (awaiting resilver)")); } } /* The top-level vdevs have the rebuild stats */ if (vrs != NULL && vrs->vrs_state == VDEV_REBUILD_ACTIVE && children == 0 && vs->vs_state == VDEV_STATE_HEALTHY) { if (vs->vs_rebuild_processed != 0) { (void) printf(gettext(" (resilvering)")); } } if (cb->vcdl != NULL) { if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { printf(" "); zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path); } } /* Display vdev initialization and trim status for leaves. */ if (children == 0) { print_status_initialize(vs, cb->cb_print_vdev_init); print_status_trim(vs, cb->cb_print_vdev_trim); } (void) printf("\n"); for (c = 0; c < children; c++) { uint64_t islog = B_FALSE, ishole = B_FALSE; /* Don't print logs or holes here */ (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &islog); (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, &ishole); if (islog || ishole) continue; /* Only print normal classes here */ if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS)) continue; /* Provide vdev_rebuild_stats to children if available */ if (vrs == NULL) { (void) nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i); } vname = zpool_vdev_name(g_zfs, zhp, child[c], cb->cb_name_flags | VDEV_NAME_TYPE_ID); print_status_config(zhp, cb, vname, child[c], depth + 2, isspare, vrs); free(vname); } } /* * Print the configuration of an exported pool. Iterate over all vdevs in the * pool, printing out the name and status for each one. */ static void print_import_config(status_cbdata_t *cb, const char *name, nvlist_t *nv, int depth) { nvlist_t **child; uint_t c, children; vdev_stat_t *vs; const char *type; char *vname; verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); if (strcmp(type, VDEV_TYPE_MISSING) == 0 || strcmp(type, VDEV_TYPE_HOLE) == 0) return; verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &c) == 0); (void) printf("\t%*s%-*s", depth, "", cb->cb_namewidth - depth, name); (void) printf(" %s", zpool_state_to_name(vs->vs_state, vs->vs_aux)); if (vs->vs_aux != 0) { (void) printf(" "); switch (vs->vs_aux) { case VDEV_AUX_OPEN_FAILED: (void) printf(gettext("cannot open")); break; case VDEV_AUX_BAD_GUID_SUM: (void) printf(gettext("missing device")); break; case VDEV_AUX_NO_REPLICAS: (void) printf(gettext("insufficient replicas")); break; case VDEV_AUX_VERSION_NEWER: (void) printf(gettext("newer version")); break; case VDEV_AUX_UNSUP_FEAT: (void) printf(gettext("unsupported feature(s)")); break; case VDEV_AUX_ERR_EXCEEDED: (void) printf(gettext("too many errors")); break; case VDEV_AUX_ACTIVE: (void) printf(gettext("currently in use")); break; case VDEV_AUX_CHILDREN_OFFLINE: (void) printf(gettext("all children offline")); break; case VDEV_AUX_BAD_LABEL: (void) printf(gettext("invalid label")); break; default: (void) printf(gettext("corrupted data")); break; } } (void) printf("\n"); if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) return; for (c = 0; c < children; c++) { uint64_t is_log = B_FALSE; (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &is_log); if (is_log) continue; if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS)) continue; vname = zpool_vdev_name(g_zfs, NULL, child[c], cb->cb_name_flags | VDEV_NAME_TYPE_ID); print_import_config(cb, vname, child[c], depth + 2); free(vname); } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, &child, &children) == 0) { (void) printf(gettext("\tcache\n")); for (c = 0; c < children; c++) { vname = zpool_vdev_name(g_zfs, NULL, child[c], cb->cb_name_flags); (void) printf("\t %s\n", vname); free(vname); } } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child, &children) == 0) { (void) printf(gettext("\tspares\n")); for (c = 0; c < children; c++) { vname = zpool_vdev_name(g_zfs, NULL, child[c], cb->cb_name_flags); (void) printf("\t %s\n", vname); free(vname); } } } /* * Print specialized class vdevs. * * These are recorded as top level vdevs in the main pool child array * but with "is_log" set to 1 or an "alloc_bias" string. We use either * print_status_config() or print_import_config() to print the top level * class vdevs then any of their children (eg mirrored slogs) are printed * recursively - which works because only the top level vdev is marked. */ static void print_class_vdevs(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t *nv, const char *class) { uint_t c, children; nvlist_t **child; boolean_t printed = B_FALSE; assert(zhp != NULL || !cb->cb_verbose); if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) return; for (c = 0; c < children; c++) { uint64_t is_log = B_FALSE; const char *bias = NULL; const char *type = NULL; (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &is_log); if (is_log) { bias = (char *)VDEV_ALLOC_CLASS_LOGS; } else { (void) nvlist_lookup_string(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS, &bias); (void) nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type); } if (bias == NULL || strcmp(bias, class) != 0) continue; if (!is_log && strcmp(type, VDEV_TYPE_INDIRECT) == 0) continue; if (!printed) { (void) printf("\t%s\t\n", gettext(class)); printed = B_TRUE; } char *name = zpool_vdev_name(g_zfs, zhp, child[c], cb->cb_name_flags | VDEV_NAME_TYPE_ID); if (cb->cb_print_status) print_status_config(zhp, cb, name, child[c], 2, B_FALSE, NULL); else print_import_config(cb, name, child[c], 2); free(name); } } /* * Display the status for the given pool. */ static int show_import(nvlist_t *config, boolean_t report_error) { uint64_t pool_state; vdev_stat_t *vs; const char *name; uint64_t guid; uint64_t hostid = 0; const char *msgid; const char *hostname = "unknown"; nvlist_t *nvroot, *nvinfo; zpool_status_t reason; zpool_errata_t errata; const char *health; uint_t vsc; const char *comment; status_cbdata_t cb = { 0 }; verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, &name) == 0); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) == 0); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &pool_state) == 0); verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) == 0); health = zpool_state_to_name(vs->vs_state, vs->vs_aux); reason = zpool_import_status(config, &msgid, &errata); /* * If we're importing using a cachefile, then we won't report any * errors unless we are in the scan phase of the import. */ if (reason != ZPOOL_STATUS_OK && !report_error) return (reason); (void) printf(gettext(" pool: %s\n"), name); (void) printf(gettext(" id: %llu\n"), (u_longlong_t)guid); (void) printf(gettext(" state: %s"), health); if (pool_state == POOL_STATE_DESTROYED) (void) printf(gettext(" (DESTROYED)")); (void) printf("\n"); switch (reason) { case ZPOOL_STATUS_MISSING_DEV_R: case ZPOOL_STATUS_MISSING_DEV_NR: case ZPOOL_STATUS_BAD_GUID_SUM: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices are " "missing from the system.\n")); break; case ZPOOL_STATUS_CORRUPT_LABEL_R: case ZPOOL_STATUS_CORRUPT_LABEL_NR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices contains" " corrupted data.\n")); break; case ZPOOL_STATUS_CORRUPT_DATA: (void) printf( gettext(" status: The pool data is corrupted.\n")); break; case ZPOOL_STATUS_OFFLINE_DEV: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices " "are offlined.\n")); break; case ZPOOL_STATUS_CORRUPT_POOL: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool metadata is " "corrupted.\n")); break; case ZPOOL_STATUS_VERSION_OLDER: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool is formatted using " "a legacy on-disk version.\n")); break; case ZPOOL_STATUS_VERSION_NEWER: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool is formatted using " "an incompatible version.\n")); break; case ZPOOL_STATUS_FEAT_DISABLED: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("Some supported " "features are not enabled on the pool.\n\t" "(Note that they may be intentionally disabled " "if the\n\t'compatibility' property is set.)\n")); break; case ZPOOL_STATUS_COMPATIBILITY_ERR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("Error reading or parsing " "the file(s) indicated by the 'compatibility'\n" "property.\n")); break; case ZPOOL_STATUS_INCOMPATIBLE_FEAT: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more features " "are enabled on the pool despite not being\n" "requested by the 'compatibility' property.\n")); break; case ZPOOL_STATUS_UNSUP_FEAT_READ: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool uses the following " "feature(s) not supported on this system:\n")); color_start(ANSI_YELLOW); zpool_print_unsup_feat(config); color_end(); break; case ZPOOL_STATUS_UNSUP_FEAT_WRITE: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool can only be " "accessed in read-only mode on this system. It\n\tcannot be" " accessed in read-write mode because it uses the " "following\n\tfeature(s) not supported on this system:\n")); color_start(ANSI_YELLOW); zpool_print_unsup_feat(config); color_end(); break; case ZPOOL_STATUS_HOSTID_ACTIVE: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool is currently " "imported by another system.\n")); break; case ZPOOL_STATUS_HOSTID_REQUIRED: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool has the " "multihost property on. It cannot\n\tbe safely imported " "when the system hostid is not set.\n")); break; case ZPOOL_STATUS_HOSTID_MISMATCH: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool was last accessed " "by another system.\n")); break; case ZPOOL_STATUS_FAULTED_DEV_R: case ZPOOL_STATUS_FAULTED_DEV_NR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices are " "faulted.\n")); break; case ZPOOL_STATUS_BAD_LOG: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("An intent log record cannot " "be read.\n")); break; case ZPOOL_STATUS_RESILVERING: case ZPOOL_STATUS_REBUILDING: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices were " "being resilvered.\n")); break; case ZPOOL_STATUS_ERRATA: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"), errata); break; case ZPOOL_STATUS_NON_NATIVE_ASHIFT: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices are " "configured to use a non-native block size.\n" "\tExpect reduced performance.\n")); break; default: /* * No other status can be seen when importing pools. */ assert(reason == ZPOOL_STATUS_OK); } /* * Print out an action according to the overall state of the pool. */ if (vs->vs_state == VDEV_STATE_HEALTHY) { if (reason == ZPOOL_STATUS_VERSION_OLDER || reason == ZPOOL_STATUS_FEAT_DISABLED) { (void) printf(gettext(" action: The pool can be " "imported using its name or numeric identifier, " "though\n\tsome features will not be available " "without an explicit 'zpool upgrade'.\n")); } else if (reason == ZPOOL_STATUS_COMPATIBILITY_ERR) { (void) printf(gettext(" action: The pool can be " "imported using its name or numeric\n\tidentifier, " "though the file(s) indicated by its " "'compatibility'\n\tproperty cannot be parsed at " "this time.\n")); } else if (reason == ZPOOL_STATUS_HOSTID_MISMATCH) { (void) printf(gettext(" action: The pool can be " "imported using its name or numeric " "identifier and\n\tthe '-f' flag.\n")); } else if (reason == ZPOOL_STATUS_ERRATA) { switch (errata) { case ZPOOL_ERRATA_NONE: break; case ZPOOL_ERRATA_ZOL_2094_SCRUB: (void) printf(gettext(" action: The pool can " "be imported using its name or numeric " "identifier,\n\thowever there is a compat" "ibility issue which should be corrected" "\n\tby running 'zpool scrub'\n")); break; case ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY: (void) printf(gettext(" action: The pool can" "not be imported with this version of ZFS " "due to\n\tan active asynchronous destroy. " "Revert to an earlier version\n\tand " "allow the destroy to complete before " "updating.\n")); break; case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION: (void) printf(gettext(" action: Existing " "encrypted datasets contain an on-disk " "incompatibility, which\n\tneeds to be " "corrected. Backup these datasets to new " "encrypted datasets\n\tand destroy the " "old ones.\n")); break; case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION: (void) printf(gettext(" action: Existing " "encrypted snapshots and bookmarks contain " "an on-disk\n\tincompatibility. This may " "cause on-disk corruption if they are used" "\n\twith 'zfs recv'. To correct the " "issue, enable the bookmark_v2 feature.\n\t" "No additional action is needed if there " "are no encrypted snapshots or\n\t" "bookmarks. If preserving the encrypted " "snapshots and bookmarks is\n\trequired, " "use a non-raw send to backup and restore " "them. Alternately,\n\tthey may be removed" " to resolve the incompatibility.\n")); break; default: /* * All errata must contain an action message. */ assert(0); } } else { (void) printf(gettext(" action: The pool can be " "imported using its name or numeric " "identifier.\n")); } } else if (vs->vs_state == VDEV_STATE_DEGRADED) { (void) printf(gettext(" action: The pool can be imported " "despite missing or damaged devices. The\n\tfault " "tolerance of the pool may be compromised if imported.\n")); } else { switch (reason) { case ZPOOL_STATUS_VERSION_NEWER: (void) printf(gettext(" action: The pool cannot be " "imported. Access the pool on a system running " "newer\n\tsoftware, or recreate the pool from " "backup.\n")); break; case ZPOOL_STATUS_UNSUP_FEAT_READ: printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("The pool cannot be " "imported. Access the pool on a system that " "supports\n\tthe required feature(s), or recreate " "the pool from backup.\n")); break; case ZPOOL_STATUS_UNSUP_FEAT_WRITE: printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("The pool cannot be " "imported in read-write mode. Import the pool " "with\n" "\t\"-o readonly=on\", access the pool on a system " "that supports the\n\trequired feature(s), or " "recreate the pool from backup.\n")); break; case ZPOOL_STATUS_MISSING_DEV_R: case ZPOOL_STATUS_MISSING_DEV_NR: case ZPOOL_STATUS_BAD_GUID_SUM: (void) printf(gettext(" action: The pool cannot be " "imported. Attach the missing\n\tdevices and try " "again.\n")); break; case ZPOOL_STATUS_HOSTID_ACTIVE: VERIFY0(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo)); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME)) hostname = fnvlist_lookup_string(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID)) hostid = fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_HOSTID); (void) printf(gettext(" action: The pool must be " "exported from %s (hostid=%"PRIx64")\n\tbefore it " "can be safely imported.\n"), hostname, hostid); break; case ZPOOL_STATUS_HOSTID_REQUIRED: (void) printf(gettext(" action: Set a unique system " "hostid with the zgenhostid(8) command.\n")); break; default: (void) printf(gettext(" action: The pool cannot be " "imported due to damaged devices or data.\n")); } } /* Print the comment attached to the pool. */ if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0) (void) printf(gettext("comment: %s\n"), comment); /* * If the state is "closed" or "can't open", and the aux state * is "corrupt data": */ if (((vs->vs_state == VDEV_STATE_CLOSED) || (vs->vs_state == VDEV_STATE_CANT_OPEN)) && (vs->vs_aux == VDEV_AUX_CORRUPT_DATA)) { if (pool_state == POOL_STATE_DESTROYED) (void) printf(gettext("\tThe pool was destroyed, " "but can be imported using the '-Df' flags.\n")); else if (pool_state != POOL_STATE_EXPORTED) (void) printf(gettext("\tThe pool may be active on " "another system, but can be imported using\n\t" "the '-f' flag.\n")); } if (msgid != NULL) { (void) printf(gettext( " see: https://openzfs.github.io/openzfs-docs/msg/%s\n"), msgid); } (void) printf(gettext(" config:\n\n")); cb.cb_namewidth = max_width(NULL, nvroot, 0, strlen(name), VDEV_NAME_TYPE_ID); if (cb.cb_namewidth < 10) cb.cb_namewidth = 10; print_import_config(&cb, name, nvroot, 0); print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_DEDUP); print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_SPECIAL); print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_CLASS_LOGS); if (reason == ZPOOL_STATUS_BAD_GUID_SUM) { (void) printf(gettext("\n\tAdditional devices are known to " "be part of this pool, though their\n\texact " "configuration cannot be determined.\n")); } return (0); } static boolean_t zfs_force_import_required(nvlist_t *config) { uint64_t state; uint64_t hostid = 0; nvlist_t *nvinfo; state = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE); nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO); /* * The hostid on LOAD_INFO comes from the MOS label via * spa_tryimport(). If its not there then we're likely talking to an * older kernel, so use the top one, which will be from the label * discovered in zpool_find_import(), or if a cachefile is in use, the * local hostid. */ if (nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_HOSTID, &hostid) != 0) (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid); if (state != POOL_STATE_EXPORTED && hostid != get_system_hostid()) return (B_TRUE); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE)) { mmp_state_t mmp_state = fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_STATE); if (mmp_state != MMP_STATE_INACTIVE) return (B_TRUE); } return (B_FALSE); } /* * Perform the import for the given configuration. This passes the heavy * lifting off to zpool_import_props(), and then mounts the datasets contained * within the pool. */ static int do_import(nvlist_t *config, const char *newname, const char *mntopts, nvlist_t *props, int flags) { int ret = 0; int ms_status = 0; zpool_handle_t *zhp; const char *name; uint64_t version; name = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME); version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION); if (!SPA_VERSION_IS_SUPPORTED(version)) { (void) fprintf(stderr, gettext("cannot import '%s': pool " "is formatted using an unsupported ZFS version\n"), name); return (1); } else if (zfs_force_import_required(config) && !(flags & ZFS_IMPORT_ANY_HOST)) { mmp_state_t mmp_state = MMP_STATE_INACTIVE; nvlist_t *nvinfo; nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE)) mmp_state = fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_STATE); if (mmp_state == MMP_STATE_ACTIVE) { const char *hostname = ""; uint64_t hostid = 0; if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME)) hostname = fnvlist_lookup_string(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID)) hostid = fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_HOSTID); (void) fprintf(stderr, gettext("cannot import '%s': " "pool is imported on %s (hostid: " "0x%"PRIx64")\nExport the pool on the other " "system, then run 'zpool import'.\n"), name, hostname, hostid); } else if (mmp_state == MMP_STATE_NO_HOSTID) { (void) fprintf(stderr, gettext("Cannot import '%s': " "pool has the multihost property on and the\n" "system's hostid is not set. Set a unique hostid " "with the zgenhostid(8) command.\n"), name); } else { const char *hostname = ""; time_t timestamp = 0; uint64_t hostid = 0; if (nvlist_exists(nvinfo, ZPOOL_CONFIG_HOSTNAME)) hostname = fnvlist_lookup_string(nvinfo, ZPOOL_CONFIG_HOSTNAME); else if (nvlist_exists(config, ZPOOL_CONFIG_HOSTNAME)) hostname = fnvlist_lookup_string(config, ZPOOL_CONFIG_HOSTNAME); if (nvlist_exists(config, ZPOOL_CONFIG_TIMESTAMP)) timestamp = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_HOSTID)) hostid = fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_HOSTID); else if (nvlist_exists(config, ZPOOL_CONFIG_HOSTID)) hostid = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID); (void) fprintf(stderr, gettext("cannot import '%s': " "pool was previously in use from another system.\n" "Last accessed by %s (hostid=%"PRIx64") at %s" "The pool can be imported, use 'zpool import -f' " "to import the pool.\n"), name, hostname, hostid, ctime(×tamp)); } return (1); } if (zpool_import_props(g_zfs, config, newname, props, flags) != 0) return (1); if (newname != NULL) name = newname; if ((zhp = zpool_open_canfail(g_zfs, name)) == NULL) return (1); /* * Loading keys is best effort. We don't want to return immediately * if it fails but we do want to give the error to the caller. */ if (flags & ZFS_IMPORT_LOAD_KEYS && zfs_crypto_attempt_load_keys(g_zfs, name) != 0) ret = 1; if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL && !(flags & ZFS_IMPORT_ONLY)) { ms_status = zpool_enable_datasets(zhp, mntopts, 0); if (ms_status == EZFS_SHAREFAILED) { (void) fprintf(stderr, gettext("Import was " - "successful, but unable to share some datasets")); + "successful, but unable to share some datasets\n")); } else if (ms_status == EZFS_MOUNTFAILED) { (void) fprintf(stderr, gettext("Import was " - "successful, but unable to mount some datasets")); + "successful, but unable to mount some datasets\n")); } } zpool_close(zhp); return (ret); } +typedef struct import_parameters { + nvlist_t *ip_config; + const char *ip_mntopts; + nvlist_t *ip_props; + int ip_flags; + int *ip_err; +} import_parameters_t; + +static void +do_import_task(void *arg) +{ + import_parameters_t *ip = arg; + *ip->ip_err |= do_import(ip->ip_config, NULL, ip->ip_mntopts, + ip->ip_props, ip->ip_flags); + free(ip); +} + + static int import_pools(nvlist_t *pools, nvlist_t *props, char *mntopts, int flags, - char *orig_name, char *new_name, - boolean_t do_destroyed, boolean_t pool_specified, boolean_t do_all, - importargs_t *import) + char *orig_name, char *new_name, importargs_t *import) { nvlist_t *config = NULL; nvlist_t *found_config = NULL; uint64_t pool_state; + boolean_t pool_specified = (import->poolname != NULL || + import->guid != 0); + + + tpool_t *tp = NULL; + if (import->do_all) { + tp = tpool_create(1, 5 * sysconf(_SC_NPROCESSORS_ONLN), + 0, NULL); + } /* * At this point we have a list of import candidate configs. Even if * we were searching by pool name or guid, we still need to * post-process the list to deal with pool state and possible * duplicate names. */ int err = 0; nvpair_t *elem = NULL; boolean_t first = B_TRUE; while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) { verify(nvpair_value_nvlist(elem, &config) == 0); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &pool_state) == 0); - if (!do_destroyed && pool_state == POOL_STATE_DESTROYED) + if (!import->do_destroyed && + pool_state == POOL_STATE_DESTROYED) continue; - if (do_destroyed && pool_state != POOL_STATE_DESTROYED) + if (import->do_destroyed && + pool_state != POOL_STATE_DESTROYED) continue; verify(nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY, import->policy) == 0); if (!pool_specified) { if (first) first = B_FALSE; - else if (!do_all) + else if (!import->do_all) (void) fputc('\n', stdout); - if (do_all) { - err |= do_import(config, NULL, mntopts, - props, flags); + if (import->do_all) { + import_parameters_t *ip = safe_malloc( + sizeof (import_parameters_t)); + + ip->ip_config = config; + ip->ip_mntopts = mntopts; + ip->ip_props = props; + ip->ip_flags = flags; + ip->ip_err = &err; + + (void) tpool_dispatch(tp, do_import_task, + (void *)ip); } else { /* * If we're importing from cachefile, then * we don't want to report errors until we * are in the scan phase of the import. If * we get an error, then we return that error * to invoke the scan phase. */ if (import->cachefile && !import->scan) err = show_import(config, B_FALSE); else (void) show_import(config, B_TRUE); } } else if (import->poolname != NULL) { const char *name; /* * We are searching for a pool based on name. */ verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, &name) == 0); if (strcmp(name, import->poolname) == 0) { if (found_config != NULL) { (void) fprintf(stderr, gettext( "cannot import '%s': more than " "one matching pool\n"), import->poolname); (void) fprintf(stderr, gettext( "import by numeric ID instead\n")); err = B_TRUE; } found_config = config; } } else { uint64_t guid; /* * Search for a pool by guid. */ verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) == 0); if (guid == import->guid) found_config = config; } } + if (import->do_all) { + tpool_wait(tp); + tpool_destroy(tp); + } /* * If we were searching for a specific pool, verify that we found a * pool, and then do the import. */ if (pool_specified && err == 0) { if (found_config == NULL) { (void) fprintf(stderr, gettext("cannot import '%s': " "no such pool available\n"), orig_name); err = B_TRUE; } else { err |= do_import(found_config, new_name, mntopts, props, flags); } } /* * If we were just looking for pools, report an error if none were * found. */ if (!pool_specified && first) (void) fprintf(stderr, gettext("no pools available to import\n")); return (err); } typedef struct target_exists_args { const char *poolname; uint64_t poolguid; } target_exists_args_t; static int name_or_guid_exists(zpool_handle_t *zhp, void *data) { target_exists_args_t *args = data; nvlist_t *config = zpool_get_config(zhp, NULL); int found = 0; if (config == NULL) return (0); if (args->poolname != NULL) { const char *pool_name; verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0); if (strcmp(pool_name, args->poolname) == 0) found = 1; } else { uint64_t pool_guid; verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0); if (pool_guid == args->poolguid) found = 1; } zpool_close(zhp); return (found); } /* * zpool checkpoint * checkpoint --discard * * -d Discard the checkpoint from a checkpointed * --discard pool. * * -w Wait for discarding a checkpoint to complete. * --wait * * Checkpoints the specified pool, by taking a "snapshot" of its * current state. A pool can only have one checkpoint at a time. */ int zpool_do_checkpoint(int argc, char **argv) { boolean_t discard, wait; char *pool; zpool_handle_t *zhp; int c, err; struct option long_options[] = { {"discard", no_argument, NULL, 'd'}, {"wait", no_argument, NULL, 'w'}, {0, 0, 0, 0} }; discard = B_FALSE; wait = B_FALSE; while ((c = getopt_long(argc, argv, ":dw", long_options, NULL)) != -1) { switch (c) { case 'd': discard = B_TRUE; break; case 'w': wait = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (wait && !discard) { (void) fprintf(stderr, gettext("--wait only valid when " "--discard also specified\n")); usage(B_FALSE); } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing pool argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } pool = argv[0]; if ((zhp = zpool_open(g_zfs, pool)) == NULL) { /* As a special case, check for use of '/' in the name */ if (strchr(pool, '/') != NULL) (void) fprintf(stderr, gettext("'zpool checkpoint' " "doesn't work on datasets. To save the state " "of a dataset from a specific point in time " "please use 'zfs snapshot'\n")); return (1); } if (discard) { err = (zpool_discard_checkpoint(zhp) != 0); if (err == 0 && wait) err = zpool_wait(zhp, ZPOOL_WAIT_CKPT_DISCARD); } else { err = (zpool_checkpoint(zhp) != 0); } zpool_close(zhp); return (err); } #define CHECKPOINT_OPT 1024 /* * zpool import [-d dir] [-D] * import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l] * [-d dir | -c cachefile | -s] [-f] -a * import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l] * [-d dir | -c cachefile | -s] [-f] [-n] [-F] * [newpool] * * -c Read pool information from a cachefile instead of searching * devices. If importing from a cachefile config fails, then * fallback to searching for devices only in the directories that * exist in the cachefile. * * -d Scan in a specific directory, other than /dev/. More than * one directory can be specified using multiple '-d' options. * * -D Scan for previously destroyed pools or import all or only * specified destroyed pools. * * -R Temporarily import the pool, with all mountpoints relative to * the given root. The pool will remain exported when the machine * is rebooted. * * -V Import even in the presence of faulted vdevs. This is an * intentionally undocumented option for testing purposes, and * treats the pool configuration as complete, leaving any bad * vdevs in the FAULTED state. In other words, it does verbatim * import. * * -f Force import, even if it appears that the pool is active. * * -F Attempt rewind if necessary. * * -n See if rewind would work, but don't actually rewind. * * -N Import the pool but don't mount datasets. * * -T Specify a starting txg to use for import. This option is * intentionally undocumented option for testing purposes. * * -a Import all pools found. * * -l Load encryption keys while importing. * * -o Set property=value and/or temporary mount options (without '='). * * -s Scan using the default search path, the libblkid cache will * not be consulted. * * --rewind-to-checkpoint * Import the pool and revert back to the checkpoint. * * The import command scans for pools to import, and import pools based on pool * name and GUID. The pool can also be renamed as part of the import process. */ int zpool_do_import(int argc, char **argv) { char **searchdirs = NULL; char *env, *envdup = NULL; int nsearch = 0; int c; int err = 0; nvlist_t *pools = NULL; boolean_t do_all = B_FALSE; boolean_t do_destroyed = B_FALSE; char *mntopts = NULL; uint64_t searchguid = 0; char *searchname = NULL; char *propval; nvlist_t *policy = NULL; nvlist_t *props = NULL; int flags = ZFS_IMPORT_NORMAL; uint32_t rewind_policy = ZPOOL_NO_REWIND; boolean_t dryrun = B_FALSE; boolean_t do_rewind = B_FALSE; boolean_t xtreme_rewind = B_FALSE; boolean_t do_scan = B_FALSE; boolean_t pool_exists = B_FALSE; - boolean_t pool_specified = B_FALSE; uint64_t txg = -1ULL; char *cachefile = NULL; importargs_t idata = { 0 }; char *endptr; struct option long_options[] = { {"rewind-to-checkpoint", no_argument, NULL, CHECKPOINT_OPT}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, ":aCc:d:DEfFlmnNo:R:stT:VX", long_options, NULL)) != -1) { switch (c) { case 'a': do_all = B_TRUE; break; case 'c': cachefile = optarg; break; case 'd': searchdirs = safe_realloc(searchdirs, (nsearch + 1) * sizeof (char *)); searchdirs[nsearch++] = optarg; break; case 'D': do_destroyed = B_TRUE; break; case 'f': flags |= ZFS_IMPORT_ANY_HOST; break; case 'F': do_rewind = B_TRUE; break; case 'l': flags |= ZFS_IMPORT_LOAD_KEYS; break; case 'm': flags |= ZFS_IMPORT_MISSING_LOG; break; case 'n': dryrun = B_TRUE; break; case 'N': flags |= ZFS_IMPORT_ONLY; break; case 'o': if ((propval = strchr(optarg, '=')) != NULL) { *propval = '\0'; propval++; if (add_prop_list(optarg, propval, &props, B_TRUE)) goto error; } else { mntopts = optarg; } break; case 'R': if (add_prop_list(zpool_prop_to_name( ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE)) goto error; if (add_prop_list_default(zpool_prop_to_name( ZPOOL_PROP_CACHEFILE), "none", &props)) goto error; break; case 's': do_scan = B_TRUE; break; case 't': flags |= ZFS_IMPORT_TEMP_NAME; if (add_prop_list_default(zpool_prop_to_name( ZPOOL_PROP_CACHEFILE), "none", &props)) goto error; break; case 'T': errno = 0; txg = strtoull(optarg, &endptr, 0); if (errno != 0 || *endptr != '\0') { (void) fprintf(stderr, gettext("invalid txg value\n")); usage(B_FALSE); } rewind_policy = ZPOOL_DO_REWIND | ZPOOL_EXTREME_REWIND; break; case 'V': flags |= ZFS_IMPORT_VERBATIM; break; case 'X': xtreme_rewind = B_TRUE; break; case CHECKPOINT_OPT: flags |= ZFS_IMPORT_CHECKPOINT; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (cachefile && nsearch != 0) { (void) fprintf(stderr, gettext("-c is incompatible with -d\n")); usage(B_FALSE); } if (cachefile && do_scan) { (void) fprintf(stderr, gettext("-c is incompatible with -s\n")); usage(B_FALSE); } if ((flags & ZFS_IMPORT_LOAD_KEYS) && (flags & ZFS_IMPORT_ONLY)) { (void) fprintf(stderr, gettext("-l is incompatible with -N\n")); usage(B_FALSE); } if ((flags & ZFS_IMPORT_LOAD_KEYS) && !do_all && argc == 0) { (void) fprintf(stderr, gettext("-l is only meaningful during " "an import\n")); usage(B_FALSE); } if ((dryrun || xtreme_rewind) && !do_rewind) { (void) fprintf(stderr, gettext("-n or -X only meaningful with -F\n")); usage(B_FALSE); } if (dryrun) rewind_policy = ZPOOL_TRY_REWIND; else if (do_rewind) rewind_policy = ZPOOL_DO_REWIND; if (xtreme_rewind) rewind_policy |= ZPOOL_EXTREME_REWIND; /* In the future, we can capture further policy and include it here */ if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 || nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, txg) != 0 || nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY, rewind_policy) != 0) goto error; /* check argument count */ if (do_all) { if (argc != 0) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } } else { if (argc > 2) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } } /* * Check for the effective uid. We do this explicitly here because * otherwise any attempt to discover pools will silently fail. */ if (argc == 0 && geteuid() != 0) { (void) fprintf(stderr, gettext("cannot " "discover pools: permission denied\n")); free(searchdirs); nvlist_free(props); nvlist_free(policy); return (1); } /* * Depending on the arguments given, we do one of the following: * * Iterate through all pools and display information about * each one. * * -a Iterate through all pools and try to import each one. * * Find the pool that corresponds to the given GUID/pool * name and import that one. * * -D Above options applies only to destroyed pools. */ if (argc != 0) { char *endptr; errno = 0; searchguid = strtoull(argv[0], &endptr, 10); if (errno != 0 || *endptr != '\0') { searchname = argv[0]; searchguid = 0; } - pool_specified = B_TRUE; /* * User specified a name or guid. Ensure it's unique. */ target_exists_args_t search = {searchname, searchguid}; pool_exists = zpool_iter(g_zfs, name_or_guid_exists, &search); } /* * Check the environment for the preferred search path. */ if ((searchdirs == NULL) && (env = getenv("ZPOOL_IMPORT_PATH"))) { char *dir, *tmp = NULL; envdup = strdup(env); for (dir = strtok_r(envdup, ":", &tmp); dir != NULL; dir = strtok_r(NULL, ":", &tmp)) { searchdirs = safe_realloc(searchdirs, (nsearch + 1) * sizeof (char *)); searchdirs[nsearch++] = dir; } } idata.path = searchdirs; idata.paths = nsearch; idata.poolname = searchname; idata.guid = searchguid; idata.cachefile = cachefile; idata.scan = do_scan; idata.policy = policy; + idata.do_destroyed = do_destroyed; + idata.do_all = do_all; libpc_handle_t lpch = { .lpc_lib_handle = g_zfs, .lpc_ops = &libzfs_config_ops, .lpc_printerr = B_TRUE }; pools = zpool_search_import(&lpch, &idata); if (pools != NULL && pool_exists && (argc == 1 || strcmp(argv[0], argv[1]) == 0)) { (void) fprintf(stderr, gettext("cannot import '%s': " "a pool with that name already exists\n"), argv[0]); (void) fprintf(stderr, gettext("use the form '%s " " ' to give it a new name\n"), "zpool import"); err = 1; } else if (pools == NULL && pool_exists) { (void) fprintf(stderr, gettext("cannot import '%s': " "a pool with that name is already created/imported,\n"), argv[0]); (void) fprintf(stderr, gettext("and no additional pools " "with that name were found\n")); err = 1; } else if (pools == NULL) { if (argc != 0) { (void) fprintf(stderr, gettext("cannot import '%s': " "no such pool available\n"), argv[0]); } err = 1; } if (err == 1) { free(searchdirs); free(envdup); nvlist_free(policy); nvlist_free(pools); nvlist_free(props); return (1); } err = import_pools(pools, props, mntopts, flags, - argc >= 1 ? argv[0] : NULL, - argc >= 2 ? argv[1] : NULL, - do_destroyed, pool_specified, do_all, &idata); + argc >= 1 ? argv[0] : NULL, argc >= 2 ? argv[1] : NULL, &idata); /* * If we're using the cachefile and we failed to import, then * fallback to scanning the directory for pools that match * those in the cachefile. */ if (err != 0 && cachefile != NULL) { (void) printf(gettext("cachefile import failed, retrying\n")); /* * We use the scan flag to gather the directories that exist * in the cachefile. If we need to fallback to searching for * the pool config, we will only search devices in these * directories. */ idata.scan = B_TRUE; nvlist_free(pools); pools = zpool_search_import(&lpch, &idata); err = import_pools(pools, props, mntopts, flags, - argc >= 1 ? argv[0] : NULL, - argc >= 2 ? argv[1] : NULL, - do_destroyed, pool_specified, do_all, &idata); + argc >= 1 ? argv[0] : NULL, argc >= 2 ? argv[1] : NULL, + &idata); } error: nvlist_free(props); nvlist_free(pools); nvlist_free(policy); free(searchdirs); free(envdup); return (err ? 1 : 0); } /* * zpool sync [-f] [pool] ... * * -f (undocumented) force uberblock (and config including zpool cache file) * update. * * Sync the specified pool(s). * Without arguments "zpool sync" will sync all pools. * This command initiates TXG sync(s) and will return after the TXG(s) commit. * */ static int zpool_do_sync(int argc, char **argv) { int ret; boolean_t force = B_FALSE; /* check options */ while ((ret = getopt(argc, argv, "f")) != -1) { switch (ret) { case 'f': force = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* if argc == 0 we will execute zpool_sync_one on all pools */ ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL, B_FALSE, zpool_sync_one, &force); return (ret); } typedef struct iostat_cbdata { uint64_t cb_flags; int cb_namewidth; int cb_iteration; boolean_t cb_verbose; boolean_t cb_literal; boolean_t cb_scripted; zpool_list_t *cb_list; vdev_cmd_data_list_t *vcdl; vdev_cbdata_t cb_vdevs; } iostat_cbdata_t; /* iostat labels */ typedef struct name_and_columns { const char *name; /* Column name */ unsigned int columns; /* Center name to this number of columns */ } name_and_columns_t; #define IOSTAT_MAX_LABELS 15 /* Max number of labels on one line */ static const name_and_columns_t iostat_top_labels[][IOSTAT_MAX_LABELS] = { [IOS_DEFAULT] = {{"capacity", 2}, {"operations", 2}, {"bandwidth", 2}, {NULL}}, [IOS_LATENCY] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2}, {"asyncq_wait", 2}, {"scrub", 1}, {"trim", 1}, {"rebuild", 1}, {NULL}}, [IOS_QUEUES] = {{"syncq_read", 2}, {"syncq_write", 2}, {"asyncq_read", 2}, {"asyncq_write", 2}, {"scrubq_read", 2}, {"trimq_write", 2}, {"rebuildq_write", 2}, {NULL}}, [IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2}, {"asyncq_wait", 2}, {NULL}}, [IOS_RQ_HISTO] = {{"sync_read", 2}, {"sync_write", 2}, {"async_read", 2}, {"async_write", 2}, {"scrub", 2}, {"trim", 2}, {"rebuild", 2}, {NULL}}, }; /* Shorthand - if "columns" field not set, default to 1 column */ static const name_and_columns_t iostat_bottom_labels[][IOSTAT_MAX_LABELS] = { [IOS_DEFAULT] = {{"alloc"}, {"free"}, {"read"}, {"write"}, {"read"}, {"write"}, {NULL}}, [IOS_LATENCY] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"}, {"write"}, {"read"}, {"write"}, {"wait"}, {"wait"}, {"wait"}, {NULL}}, [IOS_QUEUES] = {{"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {NULL}}, [IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"}, {"write"}, {"read"}, {"write"}, {"scrub"}, {"trim"}, {"rebuild"}, {NULL}}, [IOS_RQ_HISTO] = {{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {NULL}}, }; static const char *histo_to_title[] = { [IOS_L_HISTO] = "latency", [IOS_RQ_HISTO] = "req_size", }; /* * Return the number of labels in a null-terminated name_and_columns_t * array. * */ static unsigned int label_array_len(const name_and_columns_t *labels) { int i = 0; while (labels[i].name) i++; return (i); } /* * Return the number of strings in a null-terminated string array. * For example: * * const char foo[] = {"bar", "baz", NULL} * * returns 2 */ static uint64_t str_array_len(const char *array[]) { uint64_t i = 0; while (array[i]) i++; return (i); } /* * Return a default column width for default/latency/queue columns. This does * not include histograms, which have their columns autosized. */ static unsigned int default_column_width(iostat_cbdata_t *cb, enum iostat_type type) { unsigned long column_width = 5; /* Normal niceprint */ static unsigned long widths[] = { /* * Choose some sane default column sizes for printing the * raw numbers. */ [IOS_DEFAULT] = 15, /* 1PB capacity */ [IOS_LATENCY] = 10, /* 1B ns = 10sec */ [IOS_QUEUES] = 6, /* 1M queue entries */ [IOS_L_HISTO] = 10, /* 1B ns = 10sec */ [IOS_RQ_HISTO] = 6, /* 1M queue entries */ }; if (cb->cb_literal) column_width = widths[type]; return (column_width); } /* * Print the column labels, i.e: * * capacity operations bandwidth * alloc free read write read write ... * * If force_column_width is set, use it for the column width. If not set, use * the default column width. */ static void print_iostat_labels(iostat_cbdata_t *cb, unsigned int force_column_width, const name_and_columns_t labels[][IOSTAT_MAX_LABELS]) { int i, idx, s; int text_start, rw_column_width, spaces_to_end; uint64_t flags = cb->cb_flags; uint64_t f; unsigned int column_width = force_column_width; /* For each bit set in flags */ for (f = flags; f; f &= ~(1ULL << idx)) { idx = lowbit64(f) - 1; if (!force_column_width) column_width = default_column_width(cb, idx); /* Print our top labels centered over "read write" label. */ for (i = 0; i < label_array_len(labels[idx]); i++) { const char *name = labels[idx][i].name; /* * We treat labels[][].columns == 0 as shorthand * for one column. It makes writing out the label * tables more concise. */ unsigned int columns = MAX(1, labels[idx][i].columns); unsigned int slen = strlen(name); rw_column_width = (column_width * columns) + (2 * (columns - 1)); text_start = (int)((rw_column_width) / columns - slen / columns); if (text_start < 0) text_start = 0; printf(" "); /* Two spaces between columns */ /* Space from beginning of column to label */ for (s = 0; s < text_start; s++) printf(" "); printf("%s", name); /* Print space after label to end of column */ spaces_to_end = rw_column_width - text_start - slen; if (spaces_to_end < 0) spaces_to_end = 0; for (s = 0; s < spaces_to_end; s++) printf(" "); } } } /* * print_cmd_columns - Print custom column titles from -c * * If the user specified the "zpool status|iostat -c" then print their custom * column titles in the header. For example, print_cmd_columns() would print * the " col1 col2" part of this: * * $ zpool iostat -vc 'echo col1=val1; echo col2=val2' * ... * capacity operations bandwidth * pool alloc free read write read write col1 col2 * ---------- ----- ----- ----- ----- ----- ----- ---- ---- * mypool 269K 1008M 0 0 107 946 * mirror 269K 1008M 0 0 107 946 * sdb - - 0 0 102 473 val1 val2 * sdc - - 0 0 5 473 val1 val2 * ---------- ----- ----- ----- ----- ----- ----- ---- ---- */ static void print_cmd_columns(vdev_cmd_data_list_t *vcdl, int use_dashes) { int i, j; vdev_cmd_data_t *data = &vcdl->data[0]; if (vcdl->count == 0 || data == NULL) return; /* * Each vdev cmd should have the same column names unless the user did * something weird with their cmd. Just take the column names from the * first vdev and assume it works for all of them. */ for (i = 0; i < vcdl->uniq_cols_cnt; i++) { printf(" "); if (use_dashes) { for (j = 0; j < vcdl->uniq_cols_width[i]; j++) printf("-"); } else { printf_color(ANSI_BOLD, "%*s", vcdl->uniq_cols_width[i], vcdl->uniq_cols[i]); } } } /* * Utility function to print out a line of dashes like: * * -------------------------------- ----- ----- ----- ----- ----- * * ...or a dashed named-row line like: * * logs - - - - - * * @cb: iostat data * * @force_column_width If non-zero, use the value as the column width. * Otherwise use the default column widths. * * @name: Print a dashed named-row line starting * with @name. Otherwise, print a regular * dashed line. */ static void print_iostat_dashes(iostat_cbdata_t *cb, unsigned int force_column_width, const char *name) { int i; unsigned int namewidth; uint64_t flags = cb->cb_flags; uint64_t f; int idx; const name_and_columns_t *labels; const char *title; if (cb->cb_flags & IOS_ANYHISTO_M) { title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]; } else if (cb->cb_vdevs.cb_names_count) { title = "vdev"; } else { title = "pool"; } namewidth = MAX(MAX(strlen(title), cb->cb_namewidth), name ? strlen(name) : 0); if (name) { printf("%-*s", namewidth, name); } else { for (i = 0; i < namewidth; i++) (void) printf("-"); } /* For each bit in flags */ for (f = flags; f; f &= ~(1ULL << idx)) { unsigned int column_width; idx = lowbit64(f) - 1; if (force_column_width) column_width = force_column_width; else column_width = default_column_width(cb, idx); labels = iostat_bottom_labels[idx]; for (i = 0; i < label_array_len(labels); i++) { if (name) printf(" %*s-", column_width - 1, " "); else printf(" %.*s", column_width, "--------------------"); } } } static void print_iostat_separator_impl(iostat_cbdata_t *cb, unsigned int force_column_width) { print_iostat_dashes(cb, force_column_width, NULL); } static void print_iostat_separator(iostat_cbdata_t *cb) { print_iostat_separator_impl(cb, 0); } static void print_iostat_header_impl(iostat_cbdata_t *cb, unsigned int force_column_width, const char *histo_vdev_name) { unsigned int namewidth; const char *title; color_start(ANSI_BOLD); if (cb->cb_flags & IOS_ANYHISTO_M) { title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]; } else if (cb->cb_vdevs.cb_names_count) { title = "vdev"; } else { title = "pool"; } namewidth = MAX(MAX(strlen(title), cb->cb_namewidth), histo_vdev_name ? strlen(histo_vdev_name) : 0); if (histo_vdev_name) printf("%-*s", namewidth, histo_vdev_name); else printf("%*s", namewidth, ""); print_iostat_labels(cb, force_column_width, iostat_top_labels); printf("\n"); printf("%-*s", namewidth, title); print_iostat_labels(cb, force_column_width, iostat_bottom_labels); if (cb->vcdl != NULL) print_cmd_columns(cb->vcdl, 0); printf("\n"); print_iostat_separator_impl(cb, force_column_width); if (cb->vcdl != NULL) print_cmd_columns(cb->vcdl, 1); color_end(); printf("\n"); } static void print_iostat_header(iostat_cbdata_t *cb) { print_iostat_header_impl(cb, 0, NULL); } /* * Prints a size string (i.e. 120M) with the suffix ("M") colored * by order of magnitude. Uses column_size to add padding. */ static void print_stat_color(const char *statbuf, unsigned int column_size) { fputs(" ", stdout); size_t len = strlen(statbuf); while (len < column_size) { fputc(' ', stdout); column_size--; } if (*statbuf == '0') { color_start(ANSI_GRAY); fputc('0', stdout); } else { for (; *statbuf; statbuf++) { if (*statbuf == 'K') color_start(ANSI_GREEN); else if (*statbuf == 'M') color_start(ANSI_YELLOW); else if (*statbuf == 'G') color_start(ANSI_RED); else if (*statbuf == 'T') color_start(ANSI_BOLD_BLUE); else if (*statbuf == 'P') color_start(ANSI_MAGENTA); else if (*statbuf == 'E') color_start(ANSI_CYAN); fputc(*statbuf, stdout); if (--column_size <= 0) break; } } color_end(); } /* * Display a single statistic. */ static void print_one_stat(uint64_t value, enum zfs_nicenum_format format, unsigned int column_size, boolean_t scripted) { char buf[64]; zfs_nicenum_format(value, buf, sizeof (buf), format); if (scripted) printf("\t%s", buf); else print_stat_color(buf, column_size); } /* * Calculate the default vdev stats * * Subtract oldvs from newvs, apply a scaling factor, and save the resulting * stats into calcvs. */ static void calc_default_iostats(vdev_stat_t *oldvs, vdev_stat_t *newvs, vdev_stat_t *calcvs) { int i; memcpy(calcvs, newvs, sizeof (*calcvs)); for (i = 0; i < ARRAY_SIZE(calcvs->vs_ops); i++) calcvs->vs_ops[i] = (newvs->vs_ops[i] - oldvs->vs_ops[i]); for (i = 0; i < ARRAY_SIZE(calcvs->vs_bytes); i++) calcvs->vs_bytes[i] = (newvs->vs_bytes[i] - oldvs->vs_bytes[i]); } /* * Internal representation of the extended iostats data. * * The extended iostat stats are exported in nvlists as either uint64_t arrays * or single uint64_t's. We make both look like arrays to make them easier * to process. In order to make single uint64_t's look like arrays, we set * __data to the stat data, and then set *data = &__data with count = 1. Then, * we can just use *data and count. */ struct stat_array { uint64_t *data; uint_t count; /* Number of entries in data[] */ uint64_t __data; /* Only used when data is a single uint64_t */ }; static uint64_t stat_histo_max(struct stat_array *nva, unsigned int len) { uint64_t max = 0; int i; for (i = 0; i < len; i++) max = MAX(max, array64_max(nva[i].data, nva[i].count)); return (max); } /* * Helper function to lookup a uint64_t array or uint64_t value and store its * data as a stat_array. If the nvpair is a single uint64_t value, then we make * it look like a one element array to make it easier to process. */ static int nvpair64_to_stat_array(nvlist_t *nvl, const char *name, struct stat_array *nva) { nvpair_t *tmp; int ret; verify(nvlist_lookup_nvpair(nvl, name, &tmp) == 0); switch (nvpair_type(tmp)) { case DATA_TYPE_UINT64_ARRAY: ret = nvpair_value_uint64_array(tmp, &nva->data, &nva->count); break; case DATA_TYPE_UINT64: ret = nvpair_value_uint64(tmp, &nva->__data); nva->data = &nva->__data; nva->count = 1; break; default: /* Not a uint64_t */ ret = EINVAL; break; } return (ret); } /* * Given a list of nvlist names, look up the extended stats in newnv and oldnv, * subtract them, and return the results in a newly allocated stat_array. * You must free the returned array after you are done with it with * free_calc_stats(). * * Additionally, you can set "oldnv" to NULL if you simply want the newnv * values. */ static struct stat_array * calc_and_alloc_stats_ex(const char **names, unsigned int len, nvlist_t *oldnv, nvlist_t *newnv) { nvlist_t *oldnvx = NULL, *newnvx; struct stat_array *oldnva, *newnva, *calcnva; int i, j; unsigned int alloc_size = (sizeof (struct stat_array)) * len; /* Extract our extended stats nvlist from the main list */ verify(nvlist_lookup_nvlist(newnv, ZPOOL_CONFIG_VDEV_STATS_EX, &newnvx) == 0); if (oldnv) { verify(nvlist_lookup_nvlist(oldnv, ZPOOL_CONFIG_VDEV_STATS_EX, &oldnvx) == 0); } newnva = safe_malloc(alloc_size); oldnva = safe_malloc(alloc_size); calcnva = safe_malloc(alloc_size); for (j = 0; j < len; j++) { verify(nvpair64_to_stat_array(newnvx, names[j], &newnva[j]) == 0); calcnva[j].count = newnva[j].count; alloc_size = calcnva[j].count * sizeof (calcnva[j].data[0]); calcnva[j].data = safe_malloc(alloc_size); memcpy(calcnva[j].data, newnva[j].data, alloc_size); if (oldnvx) { verify(nvpair64_to_stat_array(oldnvx, names[j], &oldnva[j]) == 0); for (i = 0; i < oldnva[j].count; i++) calcnva[j].data[i] -= oldnva[j].data[i]; } } free(newnva); free(oldnva); return (calcnva); } static void free_calc_stats(struct stat_array *nva, unsigned int len) { int i; for (i = 0; i < len; i++) free(nva[i].data); free(nva); } static void print_iostat_histo(struct stat_array *nva, unsigned int len, iostat_cbdata_t *cb, unsigned int column_width, unsigned int namewidth, double scale) { int i, j; char buf[6]; uint64_t val; enum zfs_nicenum_format format; unsigned int buckets; unsigned int start_bucket; if (cb->cb_literal) format = ZFS_NICENUM_RAW; else format = ZFS_NICENUM_1024; /* All these histos are the same size, so just use nva[0].count */ buckets = nva[0].count; if (cb->cb_flags & IOS_RQ_HISTO_M) { /* Start at 512 - req size should never be lower than this */ start_bucket = 9; } else { start_bucket = 0; } for (j = start_bucket; j < buckets; j++) { /* Print histogram bucket label */ if (cb->cb_flags & IOS_L_HISTO_M) { /* Ending range of this bucket */ val = (1UL << (j + 1)) - 1; zfs_nicetime(val, buf, sizeof (buf)); } else { /* Request size (starting range of bucket) */ val = (1UL << j); zfs_nicenum(val, buf, sizeof (buf)); } if (cb->cb_scripted) printf("%llu", (u_longlong_t)val); else printf("%-*s", namewidth, buf); /* Print the values on the line */ for (i = 0; i < len; i++) { print_one_stat(nva[i].data[j] * scale, format, column_width, cb->cb_scripted); } printf("\n"); } } static void print_solid_separator(unsigned int length) { while (length--) printf("-"); printf("\n"); } static void print_iostat_histos(iostat_cbdata_t *cb, nvlist_t *oldnv, nvlist_t *newnv, double scale, const char *name) { unsigned int column_width; unsigned int namewidth; unsigned int entire_width; enum iostat_type type; struct stat_array *nva; const char **names; unsigned int names_len; /* What type of histo are we? */ type = IOS_HISTO_IDX(cb->cb_flags); /* Get NULL-terminated array of nvlist names for our histo */ names = vsx_type_to_nvlist[type]; names_len = str_array_len(names); /* num of names */ nva = calc_and_alloc_stats_ex(names, names_len, oldnv, newnv); if (cb->cb_literal) { column_width = MAX(5, (unsigned int) log10(stat_histo_max(nva, names_len)) + 1); } else { column_width = 5; } namewidth = MAX(cb->cb_namewidth, strlen(histo_to_title[IOS_HISTO_IDX(cb->cb_flags)])); /* * Calculate the entire line width of what we're printing. The * +2 is for the two spaces between columns: */ /* read write */ /* ----- ----- */ /* |___| <---------- column_width */ /* */ /* |__________| <--- entire_width */ /* */ entire_width = namewidth + (column_width + 2) * label_array_len(iostat_bottom_labels[type]); if (cb->cb_scripted) printf("%s\n", name); else print_iostat_header_impl(cb, column_width, name); print_iostat_histo(nva, names_len, cb, column_width, namewidth, scale); free_calc_stats(nva, names_len); if (!cb->cb_scripted) print_solid_separator(entire_width); } /* * Calculate the average latency of a power-of-two latency histogram */ static uint64_t single_histo_average(uint64_t *histo, unsigned int buckets) { int i; uint64_t count = 0, total = 0; for (i = 0; i < buckets; i++) { /* * Our buckets are power-of-two latency ranges. Use the * midpoint latency of each bucket to calculate the average. * For example: * * Bucket Midpoint * 8ns-15ns: 12ns * 16ns-31ns: 24ns * ... */ if (histo[i] != 0) { total += histo[i] * (((1UL << i) + ((1UL << i)/2))); count += histo[i]; } } /* Prevent divide by zero */ return (count == 0 ? 0 : total / count); } static void print_iostat_queues(iostat_cbdata_t *cb, nvlist_t *newnv) { const char *names[] = { ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE, ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE, ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE, ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE, ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE, }; struct stat_array *nva; unsigned int column_width = default_column_width(cb, IOS_QUEUES); enum zfs_nicenum_format format; nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), NULL, newnv); if (cb->cb_literal) format = ZFS_NICENUM_RAW; else format = ZFS_NICENUM_1024; for (int i = 0; i < ARRAY_SIZE(names); i++) { uint64_t val = nva[i].data[0]; print_one_stat(val, format, column_width, cb->cb_scripted); } free_calc_stats(nva, ARRAY_SIZE(names)); } static void print_iostat_latency(iostat_cbdata_t *cb, nvlist_t *oldnv, nvlist_t *newnv) { int i; uint64_t val; const char *names[] = { ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO, ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO, ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO, ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO, ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO, }; struct stat_array *nva; unsigned int column_width = default_column_width(cb, IOS_LATENCY); enum zfs_nicenum_format format; nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), oldnv, newnv); if (cb->cb_literal) format = ZFS_NICENUM_RAWTIME; else format = ZFS_NICENUM_TIME; /* Print our avg latencies on the line */ for (i = 0; i < ARRAY_SIZE(names); i++) { /* Compute average latency for a latency histo */ val = single_histo_average(nva[i].data, nva[i].count); print_one_stat(val, format, column_width, cb->cb_scripted); } free_calc_stats(nva, ARRAY_SIZE(names)); } /* * Print default statistics (capacity/operations/bandwidth) */ static void print_iostat_default(vdev_stat_t *vs, iostat_cbdata_t *cb, double scale) { unsigned int column_width = default_column_width(cb, IOS_DEFAULT); enum zfs_nicenum_format format; char na; /* char to print for "not applicable" values */ if (cb->cb_literal) { format = ZFS_NICENUM_RAW; na = '0'; } else { format = ZFS_NICENUM_1024; na = '-'; } /* only toplevel vdevs have capacity stats */ if (vs->vs_space == 0) { if (cb->cb_scripted) printf("\t%c\t%c", na, na); else printf(" %*c %*c", column_width, na, column_width, na); } else { print_one_stat(vs->vs_alloc, format, column_width, cb->cb_scripted); print_one_stat(vs->vs_space - vs->vs_alloc, format, column_width, cb->cb_scripted); } print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_READ] * scale), format, column_width, cb->cb_scripted); print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_WRITE] * scale), format, column_width, cb->cb_scripted); print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_READ] * scale), format, column_width, cb->cb_scripted); print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_WRITE] * scale), format, column_width, cb->cb_scripted); } static const char *const class_name[] = { VDEV_ALLOC_BIAS_DEDUP, VDEV_ALLOC_BIAS_SPECIAL, VDEV_ALLOC_CLASS_LOGS }; /* * Print out all the statistics for the given vdev. This can either be the * toplevel configuration, or called recursively. If 'name' is NULL, then this * is a verbose output, and we don't want to display the toplevel pool stats. * * Returns the number of stat lines printed. */ static unsigned int print_vdev_stats(zpool_handle_t *zhp, const char *name, nvlist_t *oldnv, nvlist_t *newnv, iostat_cbdata_t *cb, int depth) { nvlist_t **oldchild, **newchild; uint_t c, children, oldchildren; vdev_stat_t *oldvs, *newvs, *calcvs; vdev_stat_t zerovs = { 0 }; char *vname; int i; int ret = 0; uint64_t tdelta; double scale; if (strcmp(name, VDEV_TYPE_INDIRECT) == 0) return (ret); calcvs = safe_malloc(sizeof (*calcvs)); if (oldnv != NULL) { verify(nvlist_lookup_uint64_array(oldnv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&oldvs, &c) == 0); } else { oldvs = &zerovs; } /* Do we only want to see a specific vdev? */ for (i = 0; i < cb->cb_vdevs.cb_names_count; i++) { /* Yes we do. Is this the vdev? */ if (strcmp(name, cb->cb_vdevs.cb_names[i]) == 0) { /* * This is our vdev. Since it is the only vdev we * will be displaying, make depth = 0 so that it * doesn't get indented. */ depth = 0; break; } } if (cb->cb_vdevs.cb_names_count && (i == cb->cb_vdevs.cb_names_count)) { /* Couldn't match the name */ goto children; } verify(nvlist_lookup_uint64_array(newnv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&newvs, &c) == 0); /* * Print the vdev name unless it's is a histogram. Histograms * display the vdev name in the header itself. */ if (!(cb->cb_flags & IOS_ANYHISTO_M)) { if (cb->cb_scripted) { printf("%s", name); } else { if (strlen(name) + depth > cb->cb_namewidth) (void) printf("%*s%s", depth, "", name); else (void) printf("%*s%s%*s", depth, "", name, (int)(cb->cb_namewidth - strlen(name) - depth), ""); } } /* Calculate our scaling factor */ tdelta = newvs->vs_timestamp - oldvs->vs_timestamp; if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_ANYHISTO_M)) { /* * If we specify printing histograms with no time interval, then * print the histogram numbers over the entire lifetime of the * vdev. */ scale = 1; } else { if (tdelta == 0) scale = 1.0; else scale = (double)NANOSEC / tdelta; } if (cb->cb_flags & IOS_DEFAULT_M) { calc_default_iostats(oldvs, newvs, calcvs); print_iostat_default(calcvs, cb, scale); } if (cb->cb_flags & IOS_LATENCY_M) print_iostat_latency(cb, oldnv, newnv); if (cb->cb_flags & IOS_QUEUES_M) print_iostat_queues(cb, newnv); if (cb->cb_flags & IOS_ANYHISTO_M) { printf("\n"); print_iostat_histos(cb, oldnv, newnv, scale, name); } if (cb->vcdl != NULL) { const char *path; if (nvlist_lookup_string(newnv, ZPOOL_CONFIG_PATH, &path) == 0) { printf(" "); zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path); } } if (!(cb->cb_flags & IOS_ANYHISTO_M)) printf("\n"); ret++; children: free(calcvs); if (!cb->cb_verbose) return (ret); if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_CHILDREN, &newchild, &children) != 0) return (ret); if (oldnv) { if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_CHILDREN, &oldchild, &oldchildren) != 0) return (ret); children = MIN(oldchildren, children); } /* * print normal top-level devices */ for (c = 0; c < children; c++) { uint64_t ishole = B_FALSE, islog = B_FALSE; (void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_HOLE, &ishole); (void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_LOG, &islog); if (ishole || islog) continue; if (nvlist_exists(newchild[c], ZPOOL_CONFIG_ALLOCATION_BIAS)) continue; vname = zpool_vdev_name(g_zfs, zhp, newchild[c], cb->cb_vdevs.cb_name_flags | VDEV_NAME_TYPE_ID); ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL, newchild[c], cb, depth + 2); free(vname); } /* * print all other top-level devices */ for (uint_t n = 0; n < ARRAY_SIZE(class_name); n++) { boolean_t printed = B_FALSE; for (c = 0; c < children; c++) { uint64_t islog = B_FALSE; const char *bias = NULL; const char *type = NULL; (void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_LOG, &islog); if (islog) { bias = VDEV_ALLOC_CLASS_LOGS; } else { (void) nvlist_lookup_string(newchild[c], ZPOOL_CONFIG_ALLOCATION_BIAS, &bias); (void) nvlist_lookup_string(newchild[c], ZPOOL_CONFIG_TYPE, &type); } if (bias == NULL || strcmp(bias, class_name[n]) != 0) continue; if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0) continue; if (!printed) { if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted && !cb->cb_vdevs.cb_names) { print_iostat_dashes(cb, 0, class_name[n]); } printf("\n"); printed = B_TRUE; } vname = zpool_vdev_name(g_zfs, zhp, newchild[c], cb->cb_vdevs.cb_name_flags | VDEV_NAME_TYPE_ID); ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL, newchild[c], cb, depth + 2); free(vname); } } /* * Include level 2 ARC devices in iostat output */ if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_L2CACHE, &newchild, &children) != 0) return (ret); if (oldnv) { if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_L2CACHE, &oldchild, &oldchildren) != 0) return (ret); children = MIN(oldchildren, children); } if (children > 0) { if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted && !cb->cb_vdevs.cb_names) { print_iostat_dashes(cb, 0, "cache"); } printf("\n"); for (c = 0; c < children; c++) { vname = zpool_vdev_name(g_zfs, zhp, newchild[c], cb->cb_vdevs.cb_name_flags); ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL, newchild[c], cb, depth + 2); free(vname); } } return (ret); } static int refresh_iostat(zpool_handle_t *zhp, void *data) { iostat_cbdata_t *cb = data; boolean_t missing; /* * If the pool has disappeared, remove it from the list and continue. */ if (zpool_refresh_stats(zhp, &missing) != 0) return (-1); if (missing) pool_list_remove(cb->cb_list, zhp); return (0); } /* * Callback to print out the iostats for the given pool. */ static int print_iostat(zpool_handle_t *zhp, void *data) { iostat_cbdata_t *cb = data; nvlist_t *oldconfig, *newconfig; nvlist_t *oldnvroot, *newnvroot; int ret; newconfig = zpool_get_config(zhp, &oldconfig); if (cb->cb_iteration == 1) oldconfig = NULL; verify(nvlist_lookup_nvlist(newconfig, ZPOOL_CONFIG_VDEV_TREE, &newnvroot) == 0); if (oldconfig == NULL) oldnvroot = NULL; else verify(nvlist_lookup_nvlist(oldconfig, ZPOOL_CONFIG_VDEV_TREE, &oldnvroot) == 0); ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot, cb, 0); if ((ret != 0) && !(cb->cb_flags & IOS_ANYHISTO_M) && !cb->cb_scripted && cb->cb_verbose && !cb->cb_vdevs.cb_names_count) { print_iostat_separator(cb); if (cb->vcdl != NULL) { print_cmd_columns(cb->vcdl, 1); } printf("\n"); } return (ret); } static int get_columns(void) { struct winsize ws; int columns = 80; int error; if (isatty(STDOUT_FILENO)) { error = ioctl(STDOUT_FILENO, TIOCGWINSZ, &ws); if (error == 0) columns = ws.ws_col; } else { columns = 999; } return (columns); } /* * Return the required length of the pool/vdev name column. The minimum * allowed width and output formatting flags must be provided. */ static int get_namewidth(zpool_handle_t *zhp, int min_width, int flags, boolean_t verbose) { nvlist_t *config, *nvroot; int width = min_width; if ((config = zpool_get_config(zhp, NULL)) != NULL) { verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); size_t poolname_len = strlen(zpool_get_name(zhp)); if (verbose == B_FALSE) { width = MAX(poolname_len, min_width); } else { width = MAX(poolname_len, max_width(zhp, nvroot, 0, min_width, flags)); } } return (width); } /* * Parse the input string, get the 'interval' and 'count' value if there is one. */ static void get_interval_count(int *argcp, char **argv, float *iv, unsigned long *cnt) { float interval = 0; unsigned long count = 0; int argc = *argcp; /* * Determine if the last argument is an integer or a pool name */ if (argc > 0 && zfs_isnumber(argv[argc - 1])) { char *end; errno = 0; interval = strtof(argv[argc - 1], &end); if (*end == '\0' && errno == 0) { if (interval == 0) { (void) fprintf(stderr, gettext( "interval cannot be zero\n")); usage(B_FALSE); } /* * Ignore the last parameter */ argc--; } else { /* * If this is not a valid number, just plow on. The * user will get a more informative error message later * on. */ interval = 0; } } /* * If the last argument is also an integer, then we have both a count * and an interval. */ if (argc > 0 && zfs_isnumber(argv[argc - 1])) { char *end; errno = 0; count = interval; interval = strtof(argv[argc - 1], &end); if (*end == '\0' && errno == 0) { if (interval == 0) { (void) fprintf(stderr, gettext( "interval cannot be zero\n")); usage(B_FALSE); } /* * Ignore the last parameter */ argc--; } else { interval = 0; } } *iv = interval; *cnt = count; *argcp = argc; } static void get_timestamp_arg(char c) { if (c == 'u') timestamp_fmt = UDATE; else if (c == 'd') timestamp_fmt = DDATE; else usage(B_FALSE); } /* * Return stat flags that are supported by all pools by both the module and * zpool iostat. "*data" should be initialized to all 0xFFs before running. * It will get ANDed down until only the flags that are supported on all pools * remain. */ static int get_stat_flags_cb(zpool_handle_t *zhp, void *data) { uint64_t *mask = data; nvlist_t *config, *nvroot, *nvx; uint64_t flags = 0; int i, j; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); /* Default stats are always supported, but for completeness.. */ if (nvlist_exists(nvroot, ZPOOL_CONFIG_VDEV_STATS)) flags |= IOS_DEFAULT_M; /* Get our extended stats nvlist from the main list */ if (nvlist_lookup_nvlist(nvroot, ZPOOL_CONFIG_VDEV_STATS_EX, &nvx) != 0) { /* * No extended stats; they're probably running an older * module. No big deal, we support that too. */ goto end; } /* For each extended stat, make sure all its nvpairs are supported */ for (j = 0; j < ARRAY_SIZE(vsx_type_to_nvlist); j++) { if (!vsx_type_to_nvlist[j][0]) continue; /* Start off by assuming the flag is supported, then check */ flags |= (1ULL << j); for (i = 0; vsx_type_to_nvlist[j][i]; i++) { if (!nvlist_exists(nvx, vsx_type_to_nvlist[j][i])) { /* flag isn't supported */ flags = flags & ~(1ULL << j); break; } } } end: *mask = *mask & flags; return (0); } /* * Return a bitmask of stats that are supported on all pools by both the module * and zpool iostat. */ static uint64_t get_stat_flags(zpool_list_t *list) { uint64_t mask = -1; /* * get_stat_flags_cb() will lop off bits from "mask" until only the * flags that are supported on all pools remain. */ pool_list_iter(list, B_FALSE, get_stat_flags_cb, &mask); return (mask); } /* * Return 1 if cb_data->cb_names[0] is this vdev's name, 0 otherwise. */ static int is_vdev_cb(void *zhp_data, nvlist_t *nv, void *cb_data) { uint64_t guid; vdev_cbdata_t *cb = cb_data; zpool_handle_t *zhp = zhp_data; if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0) return (0); return (guid == zpool_vdev_path_to_guid(zhp, cb->cb_names[0])); } /* * Returns 1 if cb_data->cb_names[0] is a vdev name, 0 otherwise. */ static int is_vdev(zpool_handle_t *zhp, void *cb_data) { return (for_each_vdev(zhp, is_vdev_cb, cb_data)); } /* * Check if vdevs are in a pool * * Return 1 if all argv[] strings are vdev names in pool "pool_name". Otherwise * return 0. If pool_name is NULL, then search all pools. */ static int are_vdevs_in_pool(int argc, char **argv, char *pool_name, vdev_cbdata_t *cb) { char **tmp_name; int ret = 0; int i; int pool_count = 0; if ((argc == 0) || !*argv) return (0); if (pool_name) pool_count = 1; /* Temporarily hijack cb_names for a second... */ tmp_name = cb->cb_names; /* Go though our list of prospective vdev names */ for (i = 0; i < argc; i++) { cb->cb_names = argv + i; /* Is this name a vdev in our pools? */ ret = for_each_pool(pool_count, &pool_name, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, is_vdev, cb); if (!ret) { /* No match */ break; } } cb->cb_names = tmp_name; return (ret); } static int is_pool_cb(zpool_handle_t *zhp, void *data) { char *name = data; if (strcmp(name, zpool_get_name(zhp)) == 0) return (1); return (0); } /* * Do we have a pool named *name? If so, return 1, otherwise 0. */ static int is_pool(char *name) { return (for_each_pool(0, NULL, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, is_pool_cb, name)); } /* Are all our argv[] strings pool names? If so return 1, 0 otherwise. */ static int are_all_pools(int argc, char **argv) { if ((argc == 0) || !*argv) return (0); while (--argc >= 0) if (!is_pool(argv[argc])) return (0); return (1); } /* * Helper function to print out vdev/pool names we can't resolve. Used for an * error message. */ static void error_list_unresolved_vdevs(int argc, char **argv, char *pool_name, vdev_cbdata_t *cb) { int i; char *name; char *str; for (i = 0; i < argc; i++) { name = argv[i]; if (is_pool(name)) str = gettext("pool"); else if (are_vdevs_in_pool(1, &name, pool_name, cb)) str = gettext("vdev in this pool"); else if (are_vdevs_in_pool(1, &name, NULL, cb)) str = gettext("vdev in another pool"); else str = gettext("unknown"); fprintf(stderr, "\t%s (%s)\n", name, str); } } /* * Same as get_interval_count(), but with additional checks to not misinterpret * guids as interval/count values. Assumes VDEV_NAME_GUID is set in * cb.cb_vdevs.cb_name_flags. */ static void get_interval_count_filter_guids(int *argc, char **argv, float *interval, unsigned long *count, iostat_cbdata_t *cb) { char **tmpargv = argv; int argc_for_interval = 0; /* Is the last arg an interval value? Or a guid? */ if (*argc >= 1 && !are_vdevs_in_pool(1, &argv[*argc - 1], NULL, &cb->cb_vdevs)) { /* * The last arg is not a guid, so it's probably an * interval value. */ argc_for_interval++; if (*argc >= 2 && !are_vdevs_in_pool(1, &argv[*argc - 2], NULL, &cb->cb_vdevs)) { /* * The 2nd to last arg is not a guid, so it's probably * an interval value. */ argc_for_interval++; } } /* Point to our list of possible intervals */ tmpargv = &argv[*argc - argc_for_interval]; *argc = *argc - argc_for_interval; get_interval_count(&argc_for_interval, tmpargv, interval, count); } /* * Terminal height, in rows. Returns -1 if stdout is not connected to a TTY or * if we were unable to determine its size. */ static int terminal_height(void) { struct winsize win; if (isatty(STDOUT_FILENO) == 0) return (-1); if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) != -1 && win.ws_row > 0) return (win.ws_row); return (-1); } /* * Run one of the zpool status/iostat -c scripts with the help (-h) option and * print the result. * * name: Short name of the script ('iostat'). * path: Full path to the script ('/usr/local/etc/zfs/zpool.d/iostat'); */ static void print_zpool_script_help(char *name, char *path) { char *argv[] = {path, (char *)"-h", NULL}; char **lines = NULL; int lines_cnt = 0; int rc; rc = libzfs_run_process_get_stdout_nopath(path, argv, NULL, &lines, &lines_cnt); if (rc != 0 || lines == NULL || lines_cnt <= 0) { if (lines != NULL) libzfs_free_str_array(lines, lines_cnt); return; } for (int i = 0; i < lines_cnt; i++) if (!is_blank_str(lines[i])) printf(" %-14s %s\n", name, lines[i]); libzfs_free_str_array(lines, lines_cnt); } /* * Go though the zpool status/iostat -c scripts in the user's path, run their * help option (-h), and print out the results. */ static void print_zpool_dir_scripts(char *dirpath) { DIR *dir; struct dirent *ent; char fullpath[MAXPATHLEN]; struct stat dir_stat; if ((dir = opendir(dirpath)) != NULL) { /* print all the files and directories within directory */ while ((ent = readdir(dir)) != NULL) { if (snprintf(fullpath, sizeof (fullpath), "%s/%s", dirpath, ent->d_name) >= sizeof (fullpath)) { (void) fprintf(stderr, gettext("internal error: " "ZPOOL_SCRIPTS_PATH too large.\n")); exit(1); } /* Print the scripts */ if (stat(fullpath, &dir_stat) == 0) if (dir_stat.st_mode & S_IXUSR && S_ISREG(dir_stat.st_mode)) print_zpool_script_help(ent->d_name, fullpath); } closedir(dir); } } /* * Print out help text for all zpool status/iostat -c scripts. */ static void print_zpool_script_list(const char *subcommand) { char *dir, *sp, *tmp; printf(gettext("Available 'zpool %s -c' commands:\n"), subcommand); sp = zpool_get_cmd_search_path(); if (sp == NULL) return; for (dir = strtok_r(sp, ":", &tmp); dir != NULL; dir = strtok_r(NULL, ":", &tmp)) print_zpool_dir_scripts(dir); free(sp); } /* * Set the minimum pool/vdev name column width. The width must be at least 10, * but may be as large as the column width - 42 so it still fits on one line. * NOTE: 42 is the width of the default capacity/operations/bandwidth output */ static int get_namewidth_iostat(zpool_handle_t *zhp, void *data) { iostat_cbdata_t *cb = data; int width, available_width; /* * get_namewidth() returns the maximum width of any name in that column * for any pool/vdev/device line that will be output. */ width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_vdevs.cb_name_flags | VDEV_NAME_TYPE_ID, cb->cb_verbose); /* * The width we are calculating is the width of the header and also the * padding width for names that are less than maximum width. The stats * take up 42 characters, so the width available for names is: */ available_width = get_columns() - 42; /* * If the maximum width fits on a screen, then great! Make everything * line up by justifying all lines to the same width. If that max * width is larger than what's available, the name plus stats won't fit * on one line, and justifying to that width would cause every line to * wrap on the screen. We only want lines with long names to wrap. * Limit the padding to what won't wrap. */ if (width > available_width) width = available_width; /* * And regardless of whatever the screen width is (get_columns can * return 0 if the width is not known or less than 42 for a narrow * terminal) have the width be a minimum of 10. */ if (width < 10) width = 10; /* Save the calculated width */ cb->cb_namewidth = width; return (0); } /* * zpool iostat [[-c [script1,script2,...]] [-lq]|[-rw]] [-ghHLpPvy] [-n name] * [-T d|u] [[ pool ...]|[pool vdev ...]|[vdev ...]] * [interval [count]] * * -c CMD For each vdev, run command CMD * -g Display guid for individual vdev name. * -L Follow links when resolving vdev path name. * -P Display full path for vdev name. * -v Display statistics for individual vdevs * -h Display help * -p Display values in parsable (exact) format. * -H Scripted mode. Don't display headers, and separate properties * by a single tab. * -l Display average latency * -q Display queue depths * -w Display latency histograms * -r Display request size histogram * -T Display a timestamp in date(1) or Unix format * -n Only print headers once * * This command can be tricky because we want to be able to deal with pool * creation/destruction as well as vdev configuration changes. The bulk of this * processing is handled by the pool_list_* routines in zpool_iter.c. We rely * on pool_list_update() to detect the addition of new pools. Configuration * changes are all handled within libzfs. */ int zpool_do_iostat(int argc, char **argv) { int c; int ret; int npools; float interval = 0; unsigned long count = 0; int winheight = 24; zpool_list_t *list; boolean_t verbose = B_FALSE; boolean_t latency = B_FALSE, l_histo = B_FALSE, rq_histo = B_FALSE; boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE; boolean_t omit_since_boot = B_FALSE; boolean_t guid = B_FALSE; boolean_t follow_links = B_FALSE; boolean_t full_name = B_FALSE; boolean_t headers_once = B_FALSE; iostat_cbdata_t cb = { 0 }; char *cmd = NULL; /* Used for printing error message */ const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q', [IOS_L_HISTO] = 'w', [IOS_RQ_HISTO] = 'r'}; uint64_t unsupported_flags; /* check options */ while ((c = getopt(argc, argv, "c:gLPT:vyhplqrwnH")) != -1) { switch (c) { case 'c': if (cmd != NULL) { fprintf(stderr, gettext("Can't set -c flag twice\n")); exit(1); } if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL && !libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) { fprintf(stderr, gettext( "Can't run -c, disabled by " "ZPOOL_SCRIPTS_ENABLED.\n")); exit(1); } if ((getuid() <= 0 || geteuid() <= 0) && !libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) { fprintf(stderr, gettext( "Can't run -c with root privileges " "unless ZPOOL_SCRIPTS_AS_ROOT is set.\n")); exit(1); } cmd = optarg; verbose = B_TRUE; break; case 'g': guid = B_TRUE; break; case 'L': follow_links = B_TRUE; break; case 'P': full_name = B_TRUE; break; case 'T': get_timestamp_arg(*optarg); break; case 'v': verbose = B_TRUE; break; case 'p': parsable = B_TRUE; break; case 'l': latency = B_TRUE; break; case 'q': queues = B_TRUE; break; case 'H': scripted = B_TRUE; break; case 'w': l_histo = B_TRUE; break; case 'r': rq_histo = B_TRUE; break; case 'y': omit_since_boot = B_TRUE; break; case 'n': headers_once = B_TRUE; break; case 'h': usage(B_FALSE); break; case '?': if (optopt == 'c') { print_zpool_script_list("iostat"); exit(0); } else { fprintf(stderr, gettext("invalid option '%c'\n"), optopt); } usage(B_FALSE); } } argc -= optind; argv += optind; cb.cb_literal = parsable; cb.cb_scripted = scripted; if (guid) cb.cb_vdevs.cb_name_flags |= VDEV_NAME_GUID; if (follow_links) cb.cb_vdevs.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS; if (full_name) cb.cb_vdevs.cb_name_flags |= VDEV_NAME_PATH; cb.cb_iteration = 0; cb.cb_namewidth = 0; cb.cb_verbose = verbose; /* Get our interval and count values (if any) */ if (guid) { get_interval_count_filter_guids(&argc, argv, &interval, &count, &cb); } else { get_interval_count(&argc, argv, &interval, &count); } if (argc == 0) { /* No args, so just print the defaults. */ } else if (are_all_pools(argc, argv)) { /* All the args are pool names */ } else if (are_vdevs_in_pool(argc, argv, NULL, &cb.cb_vdevs)) { /* All the args are vdevs */ cb.cb_vdevs.cb_names = argv; cb.cb_vdevs.cb_names_count = argc; argc = 0; /* No pools to process */ } else if (are_all_pools(1, argv)) { /* The first arg is a pool name */ if (are_vdevs_in_pool(argc - 1, argv + 1, argv[0], &cb.cb_vdevs)) { /* ...and the rest are vdev names */ cb.cb_vdevs.cb_names = argv + 1; cb.cb_vdevs.cb_names_count = argc - 1; argc = 1; /* One pool to process */ } else { fprintf(stderr, gettext("Expected either a list of ")); fprintf(stderr, gettext("pools, or list of vdevs in")); fprintf(stderr, " \"%s\", ", argv[0]); fprintf(stderr, gettext("but got:\n")); error_list_unresolved_vdevs(argc - 1, argv + 1, argv[0], &cb.cb_vdevs); fprintf(stderr, "\n"); usage(B_FALSE); return (1); } } else { /* * The args don't make sense. The first arg isn't a pool name, * nor are all the args vdevs. */ fprintf(stderr, gettext("Unable to parse pools/vdevs list.\n")); fprintf(stderr, "\n"); return (1); } if (cb.cb_vdevs.cb_names_count != 0) { /* * If user specified vdevs, it implies verbose. */ cb.cb_verbose = B_TRUE; } /* * Construct the list of all interesting pools. */ ret = 0; if ((list = pool_list_get(argc, argv, NULL, ZFS_TYPE_POOL, parsable, &ret)) == NULL) return (1); if (pool_list_count(list) == 0 && argc != 0) { pool_list_free(list); return (1); } if (pool_list_count(list) == 0 && interval == 0) { pool_list_free(list); (void) fprintf(stderr, gettext("no pools available\n")); return (1); } if ((l_histo || rq_histo) && (cmd != NULL || latency || queues)) { pool_list_free(list); (void) fprintf(stderr, gettext("[-r|-w] isn't allowed with [-c|-l|-q]\n")); usage(B_FALSE); return (1); } if (l_histo && rq_histo) { pool_list_free(list); (void) fprintf(stderr, gettext("Only one of [-r|-w] can be passed at a time\n")); usage(B_FALSE); return (1); } /* * Enter the main iostat loop. */ cb.cb_list = list; if (l_histo) { /* * Histograms tables look out of place when you try to display * them with the other stats, so make a rule that you can only * print histograms by themselves. */ cb.cb_flags = IOS_L_HISTO_M; } else if (rq_histo) { cb.cb_flags = IOS_RQ_HISTO_M; } else { cb.cb_flags = IOS_DEFAULT_M; if (latency) cb.cb_flags |= IOS_LATENCY_M; if (queues) cb.cb_flags |= IOS_QUEUES_M; } /* * See if the module supports all the stats we want to display. */ unsupported_flags = cb.cb_flags & ~get_stat_flags(list); if (unsupported_flags) { uint64_t f; int idx; fprintf(stderr, gettext("The loaded zfs module doesn't support:")); /* for each bit set in unsupported_flags */ for (f = unsupported_flags; f; f &= ~(1ULL << idx)) { idx = lowbit64(f) - 1; fprintf(stderr, " -%c", flag_to_arg[idx]); } fprintf(stderr, ". Try running a newer module.\n"); pool_list_free(list); return (1); } for (;;) { if ((npools = pool_list_count(list)) == 0) (void) fprintf(stderr, gettext("no pools available\n")); else { /* * If this is the first iteration and -y was supplied * we skip any printing. */ boolean_t skip = (omit_since_boot && cb.cb_iteration == 0); /* * Refresh all statistics. This is done as an * explicit step before calculating the maximum name * width, so that any * configuration changes are * properly accounted for. */ (void) pool_list_iter(list, B_FALSE, refresh_iostat, &cb); /* * Iterate over all pools to determine the maximum width * for the pool / device name column across all pools. */ cb.cb_namewidth = 0; (void) pool_list_iter(list, B_FALSE, get_namewidth_iostat, &cb); if (timestamp_fmt != NODATE) print_timestamp(timestamp_fmt); if (cmd != NULL && cb.cb_verbose && !(cb.cb_flags & IOS_ANYHISTO_M)) { cb.vcdl = all_pools_for_each_vdev_run(argc, argv, cmd, g_zfs, cb.cb_vdevs.cb_names, cb.cb_vdevs.cb_names_count, cb.cb_vdevs.cb_name_flags); } else { cb.vcdl = NULL; } /* * Check terminal size so we can print headers * even when terminal window has its height * changed. */ winheight = terminal_height(); /* * Are we connected to TTY? If not, headers_once * should be true, to avoid breaking scripts. */ if (winheight < 0) headers_once = B_TRUE; /* * If it's the first time and we're not skipping it, * or either skip or verbose mode, print the header. * * The histogram code explicitly prints its header on * every vdev, so skip this for histograms. */ if (((++cb.cb_iteration == 1 && !skip) || (skip != verbose) || (!headers_once && (cb.cb_iteration % winheight) == 0)) && (!(cb.cb_flags & IOS_ANYHISTO_M)) && !cb.cb_scripted) print_iostat_header(&cb); if (skip) { (void) fflush(stdout); (void) fsleep(interval); continue; } pool_list_iter(list, B_FALSE, print_iostat, &cb); /* * If there's more than one pool, and we're not in * verbose mode (which prints a separator for us), * then print a separator. * * In addition, if we're printing specific vdevs then * we also want an ending separator. */ if (((npools > 1 && !verbose && !(cb.cb_flags & IOS_ANYHISTO_M)) || (!(cb.cb_flags & IOS_ANYHISTO_M) && cb.cb_vdevs.cb_names_count)) && !cb.cb_scripted) { print_iostat_separator(&cb); if (cb.vcdl != NULL) print_cmd_columns(cb.vcdl, 1); printf("\n"); } if (cb.vcdl != NULL) free_vdev_cmd_data_list(cb.vcdl); } if (interval == 0) break; if (count != 0 && --count == 0) break; (void) fflush(stdout); (void) fsleep(interval); } pool_list_free(list); return (ret); } typedef struct list_cbdata { boolean_t cb_verbose; int cb_name_flags; int cb_namewidth; boolean_t cb_scripted; zprop_list_t *cb_proplist; boolean_t cb_literal; } list_cbdata_t; /* * Given a list of columns to display, output appropriate headers for each one. */ static void print_header(list_cbdata_t *cb) { zprop_list_t *pl = cb->cb_proplist; char headerbuf[ZPOOL_MAXPROPLEN]; const char *header; boolean_t first = B_TRUE; boolean_t right_justify; size_t width = 0; for (; pl != NULL; pl = pl->pl_next) { width = pl->pl_width; if (first && cb->cb_verbose) { /* * Reset the width to accommodate the verbose listing * of devices. */ width = cb->cb_namewidth; } if (!first) (void) fputs(" ", stdout); else first = B_FALSE; right_justify = B_FALSE; if (pl->pl_prop != ZPROP_USERPROP) { header = zpool_prop_column_name(pl->pl_prop); right_justify = zpool_prop_align_right(pl->pl_prop); } else { int i; for (i = 0; pl->pl_user_prop[i] != '\0'; i++) headerbuf[i] = toupper(pl->pl_user_prop[i]); headerbuf[i] = '\0'; header = headerbuf; } if (pl->pl_next == NULL && !right_justify) (void) fputs(header, stdout); else if (right_justify) (void) printf("%*s", (int)width, header); else (void) printf("%-*s", (int)width, header); } (void) fputc('\n', stdout); } /* * Given a pool and a list of properties, print out all the properties according * to the described layout. Used by zpool_do_list(). */ static void print_pool(zpool_handle_t *zhp, list_cbdata_t *cb) { zprop_list_t *pl = cb->cb_proplist; boolean_t first = B_TRUE; char property[ZPOOL_MAXPROPLEN]; const char *propstr; boolean_t right_justify; size_t width; for (; pl != NULL; pl = pl->pl_next) { width = pl->pl_width; if (first && cb->cb_verbose) { /* * Reset the width to accommodate the verbose listing * of devices. */ width = cb->cb_namewidth; } if (!first) { if (cb->cb_scripted) (void) fputc('\t', stdout); else (void) fputs(" ", stdout); } else { first = B_FALSE; } right_justify = B_FALSE; if (pl->pl_prop != ZPROP_USERPROP) { if (zpool_get_prop(zhp, pl->pl_prop, property, sizeof (property), NULL, cb->cb_literal) != 0) propstr = "-"; else propstr = property; right_justify = zpool_prop_align_right(pl->pl_prop); } else if ((zpool_prop_feature(pl->pl_user_prop) || zpool_prop_unsupported(pl->pl_user_prop)) && zpool_prop_get_feature(zhp, pl->pl_user_prop, property, sizeof (property)) == 0) { propstr = property; } else if (zfs_prop_user(pl->pl_user_prop) && zpool_get_userprop(zhp, pl->pl_user_prop, property, sizeof (property), NULL) == 0) { propstr = property; } else { propstr = "-"; } /* * If this is being called in scripted mode, or if this is the * last column and it is left-justified, don't include a width * format specifier. */ if (cb->cb_scripted || (pl->pl_next == NULL && !right_justify)) (void) fputs(propstr, stdout); else if (right_justify) (void) printf("%*s", (int)width, propstr); else (void) printf("%-*s", (int)width, propstr); } (void) fputc('\n', stdout); } static void print_one_column(zpool_prop_t prop, uint64_t value, const char *str, boolean_t scripted, boolean_t valid, enum zfs_nicenum_format format) { char propval[64]; boolean_t fixed; size_t width = zprop_width(prop, &fixed, ZFS_TYPE_POOL); switch (prop) { case ZPOOL_PROP_SIZE: case ZPOOL_PROP_EXPANDSZ: case ZPOOL_PROP_CHECKPOINT: case ZPOOL_PROP_DEDUPRATIO: if (value == 0) (void) strlcpy(propval, "-", sizeof (propval)); else zfs_nicenum_format(value, propval, sizeof (propval), format); break; case ZPOOL_PROP_FRAGMENTATION: if (value == ZFS_FRAG_INVALID) { (void) strlcpy(propval, "-", sizeof (propval)); } else if (format == ZFS_NICENUM_RAW) { (void) snprintf(propval, sizeof (propval), "%llu", (unsigned long long)value); } else { (void) snprintf(propval, sizeof (propval), "%llu%%", (unsigned long long)value); } break; case ZPOOL_PROP_CAPACITY: /* capacity value is in parts-per-10,000 (aka permyriad) */ if (format == ZFS_NICENUM_RAW) (void) snprintf(propval, sizeof (propval), "%llu", (unsigned long long)value / 100); else (void) snprintf(propval, sizeof (propval), value < 1000 ? "%1.2f%%" : value < 10000 ? "%2.1f%%" : "%3.0f%%", value / 100.0); break; case ZPOOL_PROP_HEALTH: width = 8; (void) strlcpy(propval, str, sizeof (propval)); break; default: zfs_nicenum_format(value, propval, sizeof (propval), format); } if (!valid) (void) strlcpy(propval, "-", sizeof (propval)); if (scripted) (void) printf("\t%s", propval); else (void) printf(" %*s", (int)width, propval); } /* * print static default line per vdev * not compatible with '-o' option */ static void print_list_stats(zpool_handle_t *zhp, const char *name, nvlist_t *nv, list_cbdata_t *cb, int depth, boolean_t isspare) { nvlist_t **child; vdev_stat_t *vs; uint_t c, children; char *vname; boolean_t scripted = cb->cb_scripted; uint64_t islog = B_FALSE; const char *dashes = "%-*s - - - - " "- - - - -\n"; verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &c) == 0); if (name != NULL) { boolean_t toplevel = (vs->vs_space != 0); uint64_t cap; enum zfs_nicenum_format format; const char *state; if (cb->cb_literal) format = ZFS_NICENUM_RAW; else format = ZFS_NICENUM_1024; if (strcmp(name, VDEV_TYPE_INDIRECT) == 0) return; if (scripted) (void) printf("\t%s", name); else if (strlen(name) + depth > cb->cb_namewidth) (void) printf("%*s%s", depth, "", name); else (void) printf("%*s%s%*s", depth, "", name, (int)(cb->cb_namewidth - strlen(name) - depth), ""); /* * Print the properties for the individual vdevs. Some * properties are only applicable to toplevel vdevs. The * 'toplevel' boolean value is passed to the print_one_column() * to indicate that the value is valid. */ if (VDEV_STAT_VALID(vs_pspace, c) && vs->vs_pspace) print_one_column(ZPOOL_PROP_SIZE, vs->vs_pspace, NULL, scripted, B_TRUE, format); else print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL, scripted, toplevel, format); print_one_column(ZPOOL_PROP_ALLOCATED, vs->vs_alloc, NULL, scripted, toplevel, format); print_one_column(ZPOOL_PROP_FREE, vs->vs_space - vs->vs_alloc, NULL, scripted, toplevel, format); print_one_column(ZPOOL_PROP_CHECKPOINT, vs->vs_checkpoint_space, NULL, scripted, toplevel, format); print_one_column(ZPOOL_PROP_EXPANDSZ, vs->vs_esize, NULL, scripted, B_TRUE, format); print_one_column(ZPOOL_PROP_FRAGMENTATION, vs->vs_fragmentation, NULL, scripted, (vs->vs_fragmentation != ZFS_FRAG_INVALID && toplevel), format); cap = (vs->vs_space == 0) ? 0 : (vs->vs_alloc * 10000 / vs->vs_space); print_one_column(ZPOOL_PROP_CAPACITY, cap, NULL, scripted, toplevel, format); print_one_column(ZPOOL_PROP_DEDUPRATIO, 0, NULL, scripted, toplevel, format); state = zpool_state_to_name(vs->vs_state, vs->vs_aux); if (isspare) { if (vs->vs_aux == VDEV_AUX_SPARED) state = "INUSE"; else if (vs->vs_state == VDEV_STATE_HEALTHY) state = "AVAIL"; } print_one_column(ZPOOL_PROP_HEALTH, 0, state, scripted, B_TRUE, format); (void) fputc('\n', stdout); } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) return; /* list the normal vdevs first */ for (c = 0; c < children; c++) { uint64_t ishole = B_FALSE; if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, &ishole) == 0 && ishole) continue; if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &islog) == 0 && islog) continue; if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS)) continue; vname = zpool_vdev_name(g_zfs, zhp, child[c], cb->cb_name_flags | VDEV_NAME_TYPE_ID); print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE); free(vname); } /* list the classes: 'logs', 'dedup', and 'special' */ for (uint_t n = 0; n < ARRAY_SIZE(class_name); n++) { boolean_t printed = B_FALSE; for (c = 0; c < children; c++) { const char *bias = NULL; const char *type = NULL; if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &islog) == 0 && islog) { bias = VDEV_ALLOC_CLASS_LOGS; } else { (void) nvlist_lookup_string(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS, &bias); (void) nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type); } if (bias == NULL || strcmp(bias, class_name[n]) != 0) continue; if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0) continue; if (!printed) { /* LINTED E_SEC_PRINTF_VAR_FMT */ (void) printf(dashes, cb->cb_namewidth, class_name[n]); printed = B_TRUE; } vname = zpool_vdev_name(g_zfs, zhp, child[c], cb->cb_name_flags | VDEV_NAME_TYPE_ID); print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE); free(vname); } } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, &child, &children) == 0 && children > 0) { /* LINTED E_SEC_PRINTF_VAR_FMT */ (void) printf(dashes, cb->cb_namewidth, "cache"); for (c = 0; c < children; c++) { vname = zpool_vdev_name(g_zfs, zhp, child[c], cb->cb_name_flags); print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE); free(vname); } } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child, &children) == 0 && children > 0) { /* LINTED E_SEC_PRINTF_VAR_FMT */ (void) printf(dashes, cb->cb_namewidth, "spare"); for (c = 0; c < children; c++) { vname = zpool_vdev_name(g_zfs, zhp, child[c], cb->cb_name_flags); print_list_stats(zhp, vname, child[c], cb, depth + 2, B_TRUE); free(vname); } } } /* * Generic callback function to list a pool. */ static int list_callback(zpool_handle_t *zhp, void *data) { list_cbdata_t *cbp = data; print_pool(zhp, cbp); if (cbp->cb_verbose) { nvlist_t *config, *nvroot; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); print_list_stats(zhp, NULL, nvroot, cbp, 0, B_FALSE); } return (0); } /* * Set the minimum pool/vdev name column width. The width must be at least 9, * but may be as large as needed. */ static int get_namewidth_list(zpool_handle_t *zhp, void *data) { list_cbdata_t *cb = data; int width; width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags | VDEV_NAME_TYPE_ID, cb->cb_verbose); if (width < 9) width = 9; cb->cb_namewidth = width; return (0); } /* * zpool list [-gHLpP] [-o prop[,prop]*] [-T d|u] [pool] ... [interval [count]] * * -g Display guid for individual vdev name. * -H Scripted mode. Don't display headers, and separate properties * by a single tab. * -L Follow links when resolving vdev path name. * -o List of properties to display. Defaults to * "name,size,allocated,free,expandsize,fragmentation,capacity," * "dedupratio,health,altroot" * -p Display values in parsable (exact) format. * -P Display full path for vdev name. * -T Display a timestamp in date(1) or Unix format * * List all pools in the system, whether or not they're healthy. Output space * statistics for each one, as well as health status summary. */ int zpool_do_list(int argc, char **argv) { int c; int ret = 0; list_cbdata_t cb = { 0 }; static char default_props[] = "name,size,allocated,free,checkpoint,expandsize,fragmentation," "capacity,dedupratio,health,altroot"; char *props = default_props; float interval = 0; unsigned long count = 0; zpool_list_t *list; boolean_t first = B_TRUE; current_prop_type = ZFS_TYPE_POOL; /* check options */ while ((c = getopt(argc, argv, ":gHLo:pPT:v")) != -1) { switch (c) { case 'g': cb.cb_name_flags |= VDEV_NAME_GUID; break; case 'H': cb.cb_scripted = B_TRUE; break; case 'L': cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS; break; case 'o': props = optarg; break; case 'P': cb.cb_name_flags |= VDEV_NAME_PATH; break; case 'p': cb.cb_literal = B_TRUE; break; case 'T': get_timestamp_arg(*optarg); break; case 'v': cb.cb_verbose = B_TRUE; cb.cb_namewidth = 8; /* 8 until precalc is avail */ break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; get_interval_count(&argc, argv, &interval, &count); if (zprop_get_list(g_zfs, props, &cb.cb_proplist, ZFS_TYPE_POOL) != 0) usage(B_FALSE); for (;;) { if ((list = pool_list_get(argc, argv, &cb.cb_proplist, ZFS_TYPE_POOL, cb.cb_literal, &ret)) == NULL) return (1); if (pool_list_count(list) == 0) break; cb.cb_namewidth = 0; (void) pool_list_iter(list, B_FALSE, get_namewidth_list, &cb); if (timestamp_fmt != NODATE) print_timestamp(timestamp_fmt); if (!cb.cb_scripted && (first || cb.cb_verbose)) { print_header(&cb); first = B_FALSE; } ret = pool_list_iter(list, B_TRUE, list_callback, &cb); if (interval == 0) break; if (count != 0 && --count == 0) break; pool_list_free(list); (void) fflush(stdout); (void) fsleep(interval); } if (argc == 0 && !cb.cb_scripted && pool_list_count(list) == 0) { (void) printf(gettext("no pools available\n")); ret = 0; } pool_list_free(list); zprop_free_list(cb.cb_proplist); return (ret); } static int zpool_do_attach_or_replace(int argc, char **argv, int replacing) { boolean_t force = B_FALSE; boolean_t rebuild = B_FALSE; boolean_t wait = B_FALSE; int c; nvlist_t *nvroot; char *poolname, *old_disk, *new_disk; zpool_handle_t *zhp; nvlist_t *props = NULL; char *propval; int ret; /* check options */ while ((c = getopt(argc, argv, "fo:sw")) != -1) { switch (c) { case 'f': force = B_TRUE; break; case 'o': if ((propval = strchr(optarg, '=')) == NULL) { (void) fprintf(stderr, gettext("missing " "'=' for -o option\n")); usage(B_FALSE); } *propval = '\0'; propval++; if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) || (add_prop_list(optarg, propval, &props, B_TRUE))) usage(B_FALSE); break; case 's': rebuild = B_TRUE; break; case 'w': wait = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); } poolname = argv[0]; if (argc < 2) { (void) fprintf(stderr, gettext("missing specification\n")); usage(B_FALSE); } old_disk = argv[1]; if (argc < 3) { if (!replacing) { (void) fprintf(stderr, gettext("missing specification\n")); usage(B_FALSE); } new_disk = old_disk; argc -= 1; argv += 1; } else { new_disk = argv[2]; argc -= 2; argv += 2; } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if ((zhp = zpool_open(g_zfs, poolname)) == NULL) { nvlist_free(props); return (1); } if (zpool_get_config(zhp, NULL) == NULL) { (void) fprintf(stderr, gettext("pool '%s' is unavailable\n"), poolname); zpool_close(zhp); nvlist_free(props); return (1); } /* unless manually specified use "ashift" pool property (if set) */ if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) { int intval; zprop_source_t src; char strval[ZPOOL_MAXPROPLEN]; intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src); if (src != ZPROP_SRC_DEFAULT) { (void) sprintf(strval, "%" PRId32, intval); verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval, &props, B_TRUE) == 0); } } nvroot = make_root_vdev(zhp, props, force, B_FALSE, replacing, B_FALSE, argc, argv); if (nvroot == NULL) { zpool_close(zhp); nvlist_free(props); return (1); } ret = zpool_vdev_attach(zhp, old_disk, new_disk, nvroot, replacing, rebuild); if (ret == 0 && wait) { zpool_wait_activity_t activity = ZPOOL_WAIT_RESILVER; char raidz_prefix[] = "raidz"; if (replacing) { activity = ZPOOL_WAIT_REPLACE; } else if (strncmp(old_disk, raidz_prefix, strlen(raidz_prefix)) == 0) { activity = ZPOOL_WAIT_RAIDZ_EXPAND; } ret = zpool_wait(zhp, activity); } nvlist_free(props); nvlist_free(nvroot); zpool_close(zhp); return (ret); } /* * zpool replace [-fsw] [-o property=value] * * -f Force attach, even if appears to be in use. * -s Use sequential instead of healing reconstruction for resilver. * -o Set property=value. * -w Wait for replacing to complete before returning * * Replace with . */ int zpool_do_replace(int argc, char **argv) { return (zpool_do_attach_or_replace(argc, argv, B_TRUE)); } /* * zpool attach [-fsw] [-o property=value] | * * -f Force attach, even if appears to be in use. * -s Use sequential instead of healing reconstruction for resilver. * -o Set property=value. * -w Wait for resilvering (mirror) or expansion (raidz) to complete * before returning. * * Attach to a or , where the vdev can be of type * mirror or raidz. If is not part of a mirror, then will * be transformed into a mirror of and . When a mirror * is involved, will begin life with a DTL of [0, now], and will * immediately begin to resilver itself. For the raidz case, a expansion will * commence and reflow the raidz data across all the disks including the * . */ int zpool_do_attach(int argc, char **argv) { return (zpool_do_attach_or_replace(argc, argv, B_FALSE)); } /* * zpool detach [-f] * * -f Force detach of , even if DTLs argue against it * (not supported yet) * * Detach a device from a mirror. The operation will be refused if * is the last device in the mirror, or if the DTLs indicate that this device * has the only valid copy of some data. */ int zpool_do_detach(int argc, char **argv) { int c; char *poolname, *path; zpool_handle_t *zhp; int ret; /* check options */ while ((c = getopt(argc, argv, "")) != -1) { switch (c) { case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing specification\n")); usage(B_FALSE); } poolname = argv[0]; path = argv[1]; if ((zhp = zpool_open(g_zfs, poolname)) == NULL) return (1); ret = zpool_vdev_detach(zhp, path); zpool_close(zhp); return (ret); } /* * zpool split [-gLnP] [-o prop=val] ... * [-o mntopt] ... * [-R altroot] [ ...] * * -g Display guid for individual vdev name. * -L Follow links when resolving vdev path name. * -n Do not split the pool, but display the resulting layout if * it were to be split. * -o Set property=value, or set mount options. * -P Display full path for vdev name. * -R Mount the split-off pool under an alternate root. * -l Load encryption keys while importing. * * Splits the named pool and gives it the new pool name. Devices to be split * off may be listed, provided that no more than one device is specified * per top-level vdev mirror. The newly split pool is left in an exported * state unless -R is specified. * * Restrictions: the top-level of the pool pool must only be made up of * mirrors; all devices in the pool must be healthy; no device may be * undergoing a resilvering operation. */ int zpool_do_split(int argc, char **argv) { char *srcpool, *newpool, *propval; char *mntopts = NULL; splitflags_t flags; int c, ret = 0; int ms_status = 0; boolean_t loadkeys = B_FALSE; zpool_handle_t *zhp; nvlist_t *config, *props = NULL; flags.dryrun = B_FALSE; flags.import = B_FALSE; flags.name_flags = 0; /* check options */ while ((c = getopt(argc, argv, ":gLR:lno:P")) != -1) { switch (c) { case 'g': flags.name_flags |= VDEV_NAME_GUID; break; case 'L': flags.name_flags |= VDEV_NAME_FOLLOW_LINKS; break; case 'R': flags.import = B_TRUE; if (add_prop_list( zpool_prop_to_name(ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE) != 0) { nvlist_free(props); usage(B_FALSE); } break; case 'l': loadkeys = B_TRUE; break; case 'n': flags.dryrun = B_TRUE; break; case 'o': if ((propval = strchr(optarg, '=')) != NULL) { *propval = '\0'; propval++; if (add_prop_list(optarg, propval, &props, B_TRUE) != 0) { nvlist_free(props); usage(B_FALSE); } } else { mntopts = optarg; } break; case 'P': flags.name_flags |= VDEV_NAME_PATH; break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); break; } } if (!flags.import && mntopts != NULL) { (void) fprintf(stderr, gettext("setting mntopts is only " "valid when importing the pool\n")); usage(B_FALSE); } if (!flags.import && loadkeys) { (void) fprintf(stderr, gettext("loading keys is only " "valid when importing the pool\n")); usage(B_FALSE); } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("Missing pool name\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("Missing new pool name\n")); usage(B_FALSE); } srcpool = argv[0]; newpool = argv[1]; argc -= 2; argv += 2; if ((zhp = zpool_open(g_zfs, srcpool)) == NULL) { nvlist_free(props); return (1); } config = split_mirror_vdev(zhp, newpool, props, flags, argc, argv); if (config == NULL) { ret = 1; } else { if (flags.dryrun) { (void) printf(gettext("would create '%s' with the " "following layout:\n\n"), newpool); print_vdev_tree(NULL, newpool, config, 0, "", flags.name_flags); print_vdev_tree(NULL, "dedup", config, 0, VDEV_ALLOC_BIAS_DEDUP, 0); print_vdev_tree(NULL, "special", config, 0, VDEV_ALLOC_BIAS_SPECIAL, 0); } } zpool_close(zhp); if (ret != 0 || flags.dryrun || !flags.import) { nvlist_free(config); nvlist_free(props); return (ret); } /* * The split was successful. Now we need to open the new * pool and import it. */ if ((zhp = zpool_open_canfail(g_zfs, newpool)) == NULL) { nvlist_free(config); nvlist_free(props); return (1); } if (loadkeys) { ret = zfs_crypto_attempt_load_keys(g_zfs, newpool); if (ret != 0) ret = 1; } if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL) { ms_status = zpool_enable_datasets(zhp, mntopts, 0); if (ms_status == EZFS_SHAREFAILED) { (void) fprintf(stderr, gettext("Split was successful, " "datasets are mounted but sharing of some datasets " "has failed\n")); } else if (ms_status == EZFS_MOUNTFAILED) { (void) fprintf(stderr, gettext("Split was successful" ", but some datasets could not be mounted\n")); (void) fprintf(stderr, gettext("Try doing '%s' with a " "different altroot\n"), "zpool import"); } } zpool_close(zhp); nvlist_free(config); nvlist_free(props); return (ret); } /* * zpool online [--power] ... * * --power: Power on the enclosure slot to the drive (if possible) */ int zpool_do_online(int argc, char **argv) { int c, i; char *poolname; zpool_handle_t *zhp; int ret = 0; vdev_state_t newstate; int flags = 0; boolean_t is_power_on = B_FALSE; struct option long_options[] = { {"power", no_argument, NULL, ZPOOL_OPTION_POWER}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, "e", long_options, NULL)) != -1) { switch (c) { case 'e': flags |= ZFS_ONLINE_EXPAND; break; case ZPOOL_OPTION_POWER: is_power_on = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (libzfs_envvar_is_set("ZPOOL_AUTO_POWER_ON_SLOT")) is_power_on = B_TRUE; argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing device name\n")); usage(B_FALSE); } poolname = argv[0]; if ((zhp = zpool_open(g_zfs, poolname)) == NULL) return (1); for (i = 1; i < argc; i++) { vdev_state_t oldstate; boolean_t avail_spare, l2cache; int rc; if (is_power_on) { rc = zpool_power_on_and_disk_wait(zhp, argv[i]); if (rc == ENOTSUP) { (void) fprintf(stderr, gettext("Power control not supported\n")); } if (rc != 0) return (rc); } nvlist_t *tgt = zpool_find_vdev(zhp, argv[i], &avail_spare, &l2cache, NULL); if (tgt == NULL) { ret = 1; continue; } uint_t vsc; oldstate = ((vdev_stat_t *)fnvlist_lookup_uint64_array(tgt, ZPOOL_CONFIG_VDEV_STATS, &vsc))->vs_state; if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) { if (newstate != VDEV_STATE_HEALTHY) { (void) printf(gettext("warning: device '%s' " "onlined, but remains in faulted state\n"), argv[i]); if (newstate == VDEV_STATE_FAULTED) (void) printf(gettext("use 'zpool " "clear' to restore a faulted " "device\n")); else (void) printf(gettext("use 'zpool " "replace' to replace devices " "that are no longer present\n")); if ((flags & ZFS_ONLINE_EXPAND)) { (void) printf(gettext("%s: failed " "to expand usable space on " "unhealthy device '%s'\n"), (oldstate >= VDEV_STATE_DEGRADED ? "error" : "warning"), argv[i]); if (oldstate >= VDEV_STATE_DEGRADED) { ret = 1; break; } } } } else { ret = 1; } } zpool_close(zhp); return (ret); } /* * zpool offline [-ft]|[--power] ... * * * -f Force the device into a faulted state. * * -t Only take the device off-line temporarily. The offline/faulted * state will not be persistent across reboots. * * --power Power off the enclosure slot to the drive (if possible) */ int zpool_do_offline(int argc, char **argv) { int c, i; char *poolname; zpool_handle_t *zhp; int ret = 0; boolean_t istmp = B_FALSE; boolean_t fault = B_FALSE; boolean_t is_power_off = B_FALSE; struct option long_options[] = { {"power", no_argument, NULL, ZPOOL_OPTION_POWER}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, "ft", long_options, NULL)) != -1) { switch (c) { case 'f': fault = B_TRUE; break; case 't': istmp = B_TRUE; break; case ZPOOL_OPTION_POWER: is_power_off = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (is_power_off && fault) { (void) fprintf(stderr, gettext("-0 and -f cannot be used together\n")); usage(B_FALSE); return (1); } if (is_power_off && istmp) { (void) fprintf(stderr, gettext("-0 and -t cannot be used together\n")); usage(B_FALSE); return (1); } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name\n")); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing device name\n")); usage(B_FALSE); } poolname = argv[0]; if ((zhp = zpool_open(g_zfs, poolname)) == NULL) return (1); for (i = 1; i < argc; i++) { uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]); if (is_power_off) { /* * Note: we have to power off first, then set REMOVED, * or else zpool_vdev_set_removed_state() returns * EAGAIN. */ ret = zpool_power_off(zhp, argv[i]); if (ret != 0) { (void) fprintf(stderr, "%s %s %d\n", gettext("unable to power off slot for"), argv[i], ret); } zpool_vdev_set_removed_state(zhp, guid, VDEV_AUX_NONE); } else if (fault) { vdev_aux_t aux; if (istmp == B_FALSE) { /* Force the fault to persist across imports */ aux = VDEV_AUX_EXTERNAL_PERSIST; } else { aux = VDEV_AUX_EXTERNAL; } if (guid == 0 || zpool_vdev_fault(zhp, guid, aux) != 0) ret = 1; } else { if (zpool_vdev_offline(zhp, argv[i], istmp) != 0) ret = 1; } } zpool_close(zhp); return (ret); } /* * zpool clear [-nF]|[--power] [device] * * Clear all errors associated with a pool or a particular device. */ int zpool_do_clear(int argc, char **argv) { int c; int ret = 0; boolean_t dryrun = B_FALSE; boolean_t do_rewind = B_FALSE; boolean_t xtreme_rewind = B_FALSE; boolean_t is_power_on = B_FALSE; uint32_t rewind_policy = ZPOOL_NO_REWIND; nvlist_t *policy = NULL; zpool_handle_t *zhp; char *pool, *device; struct option long_options[] = { {"power", no_argument, NULL, ZPOOL_OPTION_POWER}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, "FnX", long_options, NULL)) != -1) { switch (c) { case 'F': do_rewind = B_TRUE; break; case 'n': dryrun = B_TRUE; break; case 'X': xtreme_rewind = B_TRUE; break; case ZPOOL_OPTION_POWER: is_power_on = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (libzfs_envvar_is_set("ZPOOL_AUTO_POWER_ON_SLOT")) is_power_on = B_TRUE; argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name\n")); usage(B_FALSE); } if (argc > 2) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if ((dryrun || xtreme_rewind) && !do_rewind) { (void) fprintf(stderr, gettext("-n or -X only meaningful with -F\n")); usage(B_FALSE); } if (dryrun) rewind_policy = ZPOOL_TRY_REWIND; else if (do_rewind) rewind_policy = ZPOOL_DO_REWIND; if (xtreme_rewind) rewind_policy |= ZPOOL_EXTREME_REWIND; /* In future, further rewind policy choices can be passed along here */ if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 || nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY, rewind_policy) != 0) { return (1); } pool = argv[0]; device = argc == 2 ? argv[1] : NULL; if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) { nvlist_free(policy); return (1); } if (is_power_on) { if (device == NULL) { zpool_power_on_pool_and_wait_for_devices(zhp); } else { zpool_power_on_and_disk_wait(zhp, device); } } if (zpool_clear(zhp, device, policy) != 0) ret = 1; zpool_close(zhp); nvlist_free(policy); return (ret); } /* * zpool reguid */ int zpool_do_reguid(int argc, char **argv) { int c; char *poolname; zpool_handle_t *zhp; int ret = 0; /* check options */ while ((c = getopt(argc, argv, "")) != -1) { switch (c) { case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* get pool name and check number of arguments */ if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } poolname = argv[0]; if ((zhp = zpool_open(g_zfs, poolname)) == NULL) return (1); ret = zpool_reguid(zhp); zpool_close(zhp); return (ret); } /* * zpool reopen * * Reopen the pool so that the kernel can update the sizes of all vdevs. */ int zpool_do_reopen(int argc, char **argv) { int c; int ret = 0; boolean_t scrub_restart = B_TRUE; /* check options */ while ((c = getopt(argc, argv, "n")) != -1) { switch (c) { case 'n': scrub_restart = B_FALSE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; /* if argc == 0 we will execute zpool_reopen_one on all pools */ ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, zpool_reopen_one, &scrub_restart); return (ret); } typedef struct scrub_cbdata { int cb_type; pool_scrub_cmd_t cb_scrub_cmd; } scrub_cbdata_t; static boolean_t zpool_has_checkpoint(zpool_handle_t *zhp) { nvlist_t *config, *nvroot; config = zpool_get_config(zhp, NULL); if (config != NULL) { pool_checkpoint_stat_t *pcs = NULL; uint_t c; nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c); if (pcs == NULL || pcs->pcs_state == CS_NONE) return (B_FALSE); assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS || pcs->pcs_state == CS_CHECKPOINT_DISCARDING); return (B_TRUE); } return (B_FALSE); } static int scrub_callback(zpool_handle_t *zhp, void *data) { scrub_cbdata_t *cb = data; int err; /* * Ignore faulted pools. */ if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { (void) fprintf(stderr, gettext("cannot scan '%s': pool is " "currently unavailable\n"), zpool_get_name(zhp)); return (1); } err = zpool_scan(zhp, cb->cb_type, cb->cb_scrub_cmd); if (err == 0 && zpool_has_checkpoint(zhp) && cb->cb_type == POOL_SCAN_SCRUB) { (void) printf(gettext("warning: will not scrub state that " "belongs to the checkpoint of pool '%s'\n"), zpool_get_name(zhp)); } return (err != 0); } static int wait_callback(zpool_handle_t *zhp, void *data) { zpool_wait_activity_t *act = data; return (zpool_wait(zhp, *act)); } /* * zpool scrub [-s | -p] [-w] [-e] ... * * -e Only scrub blocks in the error log. * -s Stop. Stops any in-progress scrub. * -p Pause. Pause in-progress scrub. * -w Wait. Blocks until scrub has completed. */ int zpool_do_scrub(int argc, char **argv) { int c; scrub_cbdata_t cb; boolean_t wait = B_FALSE; int error; cb.cb_type = POOL_SCAN_SCRUB; cb.cb_scrub_cmd = POOL_SCRUB_NORMAL; boolean_t is_error_scrub = B_FALSE; boolean_t is_pause = B_FALSE; boolean_t is_stop = B_FALSE; /* check options */ while ((c = getopt(argc, argv, "spwe")) != -1) { switch (c) { case 'e': is_error_scrub = B_TRUE; break; case 's': is_stop = B_TRUE; break; case 'p': is_pause = B_TRUE; break; case 'w': wait = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } if (is_pause && is_stop) { (void) fprintf(stderr, gettext("invalid option " "combination :-s and -p are mutually exclusive\n")); usage(B_FALSE); } else { if (is_error_scrub) cb.cb_type = POOL_SCAN_ERRORSCRUB; if (is_pause) { cb.cb_scrub_cmd = POOL_SCRUB_PAUSE; } else if (is_stop) { cb.cb_type = POOL_SCAN_NONE; } else { cb.cb_scrub_cmd = POOL_SCRUB_NORMAL; } } if (wait && (cb.cb_type == POOL_SCAN_NONE || cb.cb_scrub_cmd == POOL_SCRUB_PAUSE)) { (void) fprintf(stderr, gettext("invalid option combination: " "-w cannot be used with -p or -s\n")); usage(B_FALSE); } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); } error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, scrub_callback, &cb); if (wait && !error) { zpool_wait_activity_t act = ZPOOL_WAIT_SCRUB; error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, wait_callback, &act); } return (error); } /* * zpool resilver ... * * Restarts any in-progress resilver */ int zpool_do_resilver(int argc, char **argv) { int c; scrub_cbdata_t cb; cb.cb_type = POOL_SCAN_RESILVER; cb.cb_scrub_cmd = POOL_SCRUB_NORMAL; /* check options */ while ((c = getopt(argc, argv, "")) != -1) { switch (c) { case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); } return (for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, scrub_callback, &cb)); } /* * zpool trim [-d] [-r ] [-c | -s] [ ...] * * -c Cancel. Ends any in-progress trim. * -d Secure trim. Requires kernel and device support. * -r Sets the TRIM rate in bytes (per second). Supports * adding a multiplier suffix such as 'k' or 'm'. * -s Suspend. TRIM can then be restarted with no flags. * -w Wait. Blocks until trimming has completed. */ int zpool_do_trim(int argc, char **argv) { struct option long_options[] = { {"cancel", no_argument, NULL, 'c'}, {"secure", no_argument, NULL, 'd'}, {"rate", required_argument, NULL, 'r'}, {"suspend", no_argument, NULL, 's'}, {"wait", no_argument, NULL, 'w'}, {0, 0, 0, 0} }; pool_trim_func_t cmd_type = POOL_TRIM_START; uint64_t rate = 0; boolean_t secure = B_FALSE; boolean_t wait = B_FALSE; int c; while ((c = getopt_long(argc, argv, "cdr:sw", long_options, NULL)) != -1) { switch (c) { case 'c': if (cmd_type != POOL_TRIM_START && cmd_type != POOL_TRIM_CANCEL) { (void) fprintf(stderr, gettext("-c cannot be " "combined with other options\n")); usage(B_FALSE); } cmd_type = POOL_TRIM_CANCEL; break; case 'd': if (cmd_type != POOL_TRIM_START) { (void) fprintf(stderr, gettext("-d cannot be " "combined with the -c or -s options\n")); usage(B_FALSE); } secure = B_TRUE; break; case 'r': if (cmd_type != POOL_TRIM_START) { (void) fprintf(stderr, gettext("-r cannot be " "combined with the -c or -s options\n")); usage(B_FALSE); } if (zfs_nicestrtonum(g_zfs, optarg, &rate) == -1) { (void) fprintf(stderr, "%s: %s\n", gettext("invalid value for rate"), libzfs_error_description(g_zfs)); usage(B_FALSE); } break; case 's': if (cmd_type != POOL_TRIM_START && cmd_type != POOL_TRIM_SUSPEND) { (void) fprintf(stderr, gettext("-s cannot be " "combined with other options\n")); usage(B_FALSE); } cmd_type = POOL_TRIM_SUSPEND; break; case 'w': wait = B_TRUE; break; case '?': if (optopt != 0) { (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); } else { (void) fprintf(stderr, gettext("invalid option '%s'\n"), argv[optind - 1]); } usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing pool name argument\n")); usage(B_FALSE); return (-1); } if (wait && (cmd_type != POOL_TRIM_START)) { (void) fprintf(stderr, gettext("-w cannot be used with -c or " "-s\n")); usage(B_FALSE); } char *poolname = argv[0]; zpool_handle_t *zhp = zpool_open(g_zfs, poolname); if (zhp == NULL) return (-1); trimflags_t trim_flags = { .secure = secure, .rate = rate, .wait = wait, }; nvlist_t *vdevs = fnvlist_alloc(); if (argc == 1) { /* no individual leaf vdevs specified, so add them all */ nvlist_t *config = zpool_get_config(zhp, NULL); nvlist_t *nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); zpool_collect_leaves(zhp, nvroot, vdevs); trim_flags.fullpool = B_TRUE; } else { trim_flags.fullpool = B_FALSE; for (int i = 1; i < argc; i++) { fnvlist_add_boolean(vdevs, argv[i]); } } int error = zpool_trim(zhp, cmd_type, vdevs, &trim_flags); fnvlist_free(vdevs); zpool_close(zhp); return (error); } /* * Converts a total number of seconds to a human readable string broken * down in to days/hours/minutes/seconds. */ static void secs_to_dhms(uint64_t total, char *buf) { uint64_t days = total / 60 / 60 / 24; uint64_t hours = (total / 60 / 60) % 24; uint64_t mins = (total / 60) % 60; uint64_t secs = (total % 60); if (days > 0) { (void) sprintf(buf, "%llu days %02llu:%02llu:%02llu", (u_longlong_t)days, (u_longlong_t)hours, (u_longlong_t)mins, (u_longlong_t)secs); } else { (void) sprintf(buf, "%02llu:%02llu:%02llu", (u_longlong_t)hours, (u_longlong_t)mins, (u_longlong_t)secs); } } /* * Print out detailed error scrub status. */ static void print_err_scrub_status(pool_scan_stat_t *ps) { time_t start, end, pause; uint64_t total_secs_left; uint64_t secs_left, mins_left, hours_left, days_left; uint64_t examined, to_be_examined; if (ps == NULL || ps->pss_error_scrub_func != POOL_SCAN_ERRORSCRUB) { return; } (void) printf(gettext(" scrub: ")); start = ps->pss_error_scrub_start; end = ps->pss_error_scrub_end; pause = ps->pss_pass_error_scrub_pause; examined = ps->pss_error_scrub_examined; to_be_examined = ps->pss_error_scrub_to_be_examined; assert(ps->pss_error_scrub_func == POOL_SCAN_ERRORSCRUB); if (ps->pss_error_scrub_state == DSS_FINISHED) { total_secs_left = end - start; days_left = total_secs_left / 60 / 60 / 24; hours_left = (total_secs_left / 60 / 60) % 24; mins_left = (total_secs_left / 60) % 60; secs_left = (total_secs_left % 60); (void) printf(gettext("scrubbed %llu error blocks in %llu days " "%02llu:%02llu:%02llu on %s"), (u_longlong_t)examined, (u_longlong_t)days_left, (u_longlong_t)hours_left, (u_longlong_t)mins_left, (u_longlong_t)secs_left, ctime(&end)); return; } else if (ps->pss_error_scrub_state == DSS_CANCELED) { (void) printf(gettext("error scrub canceled on %s"), ctime(&end)); return; } assert(ps->pss_error_scrub_state == DSS_ERRORSCRUBBING); /* Error scrub is in progress. */ if (pause == 0) { (void) printf(gettext("error scrub in progress since %s"), ctime(&start)); } else { (void) printf(gettext("error scrub paused since %s"), ctime(&pause)); (void) printf(gettext("\terror scrub started on %s"), ctime(&start)); } double fraction_done = (double)examined / (to_be_examined + examined); (void) printf(gettext("\t%.2f%% done, issued I/O for %llu error" " blocks"), 100 * fraction_done, (u_longlong_t)examined); (void) printf("\n"); } /* * Print out detailed scrub status. */ static void print_scan_scrub_resilver_status(pool_scan_stat_t *ps) { time_t start, end, pause; uint64_t pass_scanned, scanned, pass_issued, issued, total_s, total_i; uint64_t elapsed, scan_rate, issue_rate; double fraction_done; char processed_buf[7], scanned_buf[7], issued_buf[7], total_s_buf[7]; char total_i_buf[7], srate_buf[7], irate_buf[7], time_buf[32]; printf(" "); printf_color(ANSI_BOLD, gettext("scan:")); printf(" "); /* If there's never been a scan, there's not much to say. */ if (ps == NULL || ps->pss_func == POOL_SCAN_NONE || ps->pss_func >= POOL_SCAN_FUNCS) { (void) printf(gettext("none requested\n")); return; } start = ps->pss_start_time; end = ps->pss_end_time; pause = ps->pss_pass_scrub_pause; zfs_nicebytes(ps->pss_processed, processed_buf, sizeof (processed_buf)); int is_resilver = ps->pss_func == POOL_SCAN_RESILVER; int is_scrub = ps->pss_func == POOL_SCAN_SCRUB; assert(is_resilver || is_scrub); /* Scan is finished or canceled. */ if (ps->pss_state == DSS_FINISHED) { secs_to_dhms(end - start, time_buf); if (is_scrub) { (void) printf(gettext("scrub repaired %s " "in %s with %llu errors on %s"), processed_buf, time_buf, (u_longlong_t)ps->pss_errors, ctime(&end)); } else if (is_resilver) { (void) printf(gettext("resilvered %s " "in %s with %llu errors on %s"), processed_buf, time_buf, (u_longlong_t)ps->pss_errors, ctime(&end)); } return; } else if (ps->pss_state == DSS_CANCELED) { if (is_scrub) { (void) printf(gettext("scrub canceled on %s"), ctime(&end)); } else if (is_resilver) { (void) printf(gettext("resilver canceled on %s"), ctime(&end)); } return; } assert(ps->pss_state == DSS_SCANNING); /* Scan is in progress. Resilvers can't be paused. */ if (is_scrub) { if (pause == 0) { (void) printf(gettext("scrub in progress since %s"), ctime(&start)); } else { (void) printf(gettext("scrub paused since %s"), ctime(&pause)); (void) printf(gettext("\tscrub started on %s"), ctime(&start)); } } else if (is_resilver) { (void) printf(gettext("resilver in progress since %s"), ctime(&start)); } scanned = ps->pss_examined; pass_scanned = ps->pss_pass_exam; issued = ps->pss_issued; pass_issued = ps->pss_pass_issued; total_s = ps->pss_to_examine; total_i = ps->pss_to_examine - ps->pss_skipped; /* we are only done with a block once we have issued the IO for it */ fraction_done = (double)issued / total_i; /* elapsed time for this pass, rounding up to 1 if it's 0 */ elapsed = time(NULL) - ps->pss_pass_start; elapsed -= ps->pss_pass_scrub_spent_paused; elapsed = (elapsed != 0) ? elapsed : 1; scan_rate = pass_scanned / elapsed; issue_rate = pass_issued / elapsed; /* format all of the numbers we will be reporting */ zfs_nicebytes(scanned, scanned_buf, sizeof (scanned_buf)); zfs_nicebytes(issued, issued_buf, sizeof (issued_buf)); zfs_nicebytes(total_s, total_s_buf, sizeof (total_s_buf)); zfs_nicebytes(total_i, total_i_buf, sizeof (total_i_buf)); /* do not print estimated time if we have a paused scrub */ (void) printf(gettext("\t%s / %s scanned"), scanned_buf, total_s_buf); if (pause == 0 && scan_rate > 0) { zfs_nicebytes(scan_rate, srate_buf, sizeof (srate_buf)); (void) printf(gettext(" at %s/s"), srate_buf); } (void) printf(gettext(", %s / %s issued"), issued_buf, total_i_buf); if (pause == 0 && issue_rate > 0) { zfs_nicebytes(issue_rate, irate_buf, sizeof (irate_buf)); (void) printf(gettext(" at %s/s"), irate_buf); } (void) printf(gettext("\n")); if (is_resilver) { (void) printf(gettext("\t%s resilvered, %.2f%% done"), processed_buf, 100 * fraction_done); } else if (is_scrub) { (void) printf(gettext("\t%s repaired, %.2f%% done"), processed_buf, 100 * fraction_done); } if (pause == 0) { /* * Only provide an estimate iff: * 1) we haven't yet issued all we expected, and * 2) the issue rate exceeds 10 MB/s, and * 3) it's either: * a) a resilver which has started repairs, or * b) a scrub which has entered the issue phase. */ if (total_i >= issued && issue_rate >= 10 * 1024 * 1024 && ((is_resilver && ps->pss_processed > 0) || (is_scrub && issued > 0))) { secs_to_dhms((total_i - issued) / issue_rate, time_buf); (void) printf(gettext(", %s to go\n"), time_buf); } else { (void) printf(gettext(", no estimated " "completion time\n")); } } else { (void) printf(gettext("\n")); } } static void print_rebuild_status_impl(vdev_rebuild_stat_t *vrs, uint_t c, char *vdev_name) { if (vrs == NULL || vrs->vrs_state == VDEV_REBUILD_NONE) return; printf(" "); printf_color(ANSI_BOLD, gettext("scan:")); printf(" "); uint64_t bytes_scanned = vrs->vrs_bytes_scanned; uint64_t bytes_issued = vrs->vrs_bytes_issued; uint64_t bytes_rebuilt = vrs->vrs_bytes_rebuilt; uint64_t bytes_est_s = vrs->vrs_bytes_est; uint64_t bytes_est_i = vrs->vrs_bytes_est; if (c > offsetof(vdev_rebuild_stat_t, vrs_pass_bytes_skipped) / 8) bytes_est_i -= vrs->vrs_pass_bytes_skipped; uint64_t scan_rate = (vrs->vrs_pass_bytes_scanned / (vrs->vrs_pass_time_ms + 1)) * 1000; uint64_t issue_rate = (vrs->vrs_pass_bytes_issued / (vrs->vrs_pass_time_ms + 1)) * 1000; double scan_pct = MIN((double)bytes_scanned * 100 / (bytes_est_s + 1), 100); /* Format all of the numbers we will be reporting */ char bytes_scanned_buf[7], bytes_issued_buf[7]; char bytes_rebuilt_buf[7], bytes_est_s_buf[7], bytes_est_i_buf[7]; char scan_rate_buf[7], issue_rate_buf[7], time_buf[32]; zfs_nicebytes(bytes_scanned, bytes_scanned_buf, sizeof (bytes_scanned_buf)); zfs_nicebytes(bytes_issued, bytes_issued_buf, sizeof (bytes_issued_buf)); zfs_nicebytes(bytes_rebuilt, bytes_rebuilt_buf, sizeof (bytes_rebuilt_buf)); zfs_nicebytes(bytes_est_s, bytes_est_s_buf, sizeof (bytes_est_s_buf)); zfs_nicebytes(bytes_est_i, bytes_est_i_buf, sizeof (bytes_est_i_buf)); time_t start = vrs->vrs_start_time; time_t end = vrs->vrs_end_time; /* Rebuild is finished or canceled. */ if (vrs->vrs_state == VDEV_REBUILD_COMPLETE) { secs_to_dhms(vrs->vrs_scan_time_ms / 1000, time_buf); (void) printf(gettext("resilvered (%s) %s in %s " "with %llu errors on %s"), vdev_name, bytes_rebuilt_buf, time_buf, (u_longlong_t)vrs->vrs_errors, ctime(&end)); return; } else if (vrs->vrs_state == VDEV_REBUILD_CANCELED) { (void) printf(gettext("resilver (%s) canceled on %s"), vdev_name, ctime(&end)); return; } else if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) { (void) printf(gettext("resilver (%s) in progress since %s"), vdev_name, ctime(&start)); } assert(vrs->vrs_state == VDEV_REBUILD_ACTIVE); (void) printf(gettext("\t%s / %s scanned"), bytes_scanned_buf, bytes_est_s_buf); if (scan_rate > 0) { zfs_nicebytes(scan_rate, scan_rate_buf, sizeof (scan_rate_buf)); (void) printf(gettext(" at %s/s"), scan_rate_buf); } (void) printf(gettext(", %s / %s issued"), bytes_issued_buf, bytes_est_i_buf); if (issue_rate > 0) { zfs_nicebytes(issue_rate, issue_rate_buf, sizeof (issue_rate_buf)); (void) printf(gettext(" at %s/s"), issue_rate_buf); } (void) printf(gettext("\n")); (void) printf(gettext("\t%s resilvered, %.2f%% done"), bytes_rebuilt_buf, scan_pct); if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) { if (bytes_est_s >= bytes_scanned && scan_rate >= 10 * 1024 * 1024) { secs_to_dhms((bytes_est_s - bytes_scanned) / scan_rate, time_buf); (void) printf(gettext(", %s to go\n"), time_buf); } else { (void) printf(gettext(", no estimated " "completion time\n")); } } else { (void) printf(gettext("\n")); } } /* * Print rebuild status for top-level vdevs. */ static void print_rebuild_status(zpool_handle_t *zhp, nvlist_t *nvroot) { nvlist_t **child; uint_t children; if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) children = 0; for (uint_t c = 0; c < children; c++) { vdev_rebuild_stat_t *vrs; uint_t i; if (nvlist_lookup_uint64_array(child[c], ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) { char *name = zpool_vdev_name(g_zfs, zhp, child[c], VDEV_NAME_TYPE_ID); print_rebuild_status_impl(vrs, i, name); free(name); } } } /* * As we don't scrub checkpointed blocks, we want to warn the user that we * skipped scanning some blocks if a checkpoint exists or existed at any * time during the scan. If a sequential instead of healing reconstruction * was performed then the blocks were reconstructed. However, their checksums * have not been verified so we still print the warning. */ static void print_checkpoint_scan_warning(pool_scan_stat_t *ps, pool_checkpoint_stat_t *pcs) { if (ps == NULL || pcs == NULL) return; if (pcs->pcs_state == CS_NONE || pcs->pcs_state == CS_CHECKPOINT_DISCARDING) return; assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS); if (ps->pss_state == DSS_NONE) return; if ((ps->pss_state == DSS_FINISHED || ps->pss_state == DSS_CANCELED) && ps->pss_end_time < pcs->pcs_start_time) return; if (ps->pss_state == DSS_FINISHED || ps->pss_state == DSS_CANCELED) { (void) printf(gettext(" scan warning: skipped blocks " "that are only referenced by the checkpoint.\n")); } else { assert(ps->pss_state == DSS_SCANNING); (void) printf(gettext(" scan warning: skipping blocks " "that are only referenced by the checkpoint.\n")); } } /* * Returns B_TRUE if there is an active rebuild in progress. Otherwise, * B_FALSE is returned and 'rebuild_end_time' is set to the end time for * the last completed (or cancelled) rebuild. */ static boolean_t check_rebuilding(nvlist_t *nvroot, uint64_t *rebuild_end_time) { nvlist_t **child; uint_t children; boolean_t rebuilding = B_FALSE; uint64_t end_time = 0; if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) children = 0; for (uint_t c = 0; c < children; c++) { vdev_rebuild_stat_t *vrs; uint_t i; if (nvlist_lookup_uint64_array(child[c], ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) { if (vrs->vrs_end_time > end_time) end_time = vrs->vrs_end_time; if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) { rebuilding = B_TRUE; end_time = 0; break; } } } if (rebuild_end_time != NULL) *rebuild_end_time = end_time; return (rebuilding); } /* * Print the scan status. */ static void print_scan_status(zpool_handle_t *zhp, nvlist_t *nvroot) { uint64_t rebuild_end_time = 0, resilver_end_time = 0; boolean_t have_resilver = B_FALSE, have_scrub = B_FALSE; boolean_t have_errorscrub = B_FALSE; boolean_t active_resilver = B_FALSE; pool_checkpoint_stat_t *pcs = NULL; pool_scan_stat_t *ps = NULL; uint_t c; time_t scrub_start = 0, errorscrub_start = 0; if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &c) == 0) { if (ps->pss_func == POOL_SCAN_RESILVER) { resilver_end_time = ps->pss_end_time; active_resilver = (ps->pss_state == DSS_SCANNING); } have_resilver = (ps->pss_func == POOL_SCAN_RESILVER); have_scrub = (ps->pss_func == POOL_SCAN_SCRUB); scrub_start = ps->pss_start_time; if (c > offsetof(pool_scan_stat_t, pss_pass_error_scrub_pause) / 8) { have_errorscrub = (ps->pss_error_scrub_func == POOL_SCAN_ERRORSCRUB); errorscrub_start = ps->pss_error_scrub_start; } } boolean_t active_rebuild = check_rebuilding(nvroot, &rebuild_end_time); boolean_t have_rebuild = (active_rebuild || (rebuild_end_time > 0)); /* Always print the scrub status when available. */ if (have_scrub && scrub_start > errorscrub_start) print_scan_scrub_resilver_status(ps); else if (have_errorscrub && errorscrub_start >= scrub_start) print_err_scrub_status(ps); /* * When there is an active resilver or rebuild print its status. * Otherwise print the status of the last resilver or rebuild. */ if (active_resilver || (!active_rebuild && have_resilver && resilver_end_time && resilver_end_time > rebuild_end_time)) { print_scan_scrub_resilver_status(ps); } else if (active_rebuild || (!active_resilver && have_rebuild && rebuild_end_time && rebuild_end_time > resilver_end_time)) { print_rebuild_status(zhp, nvroot); } (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c); print_checkpoint_scan_warning(ps, pcs); } /* * Print out detailed removal status. */ static void print_removal_status(zpool_handle_t *zhp, pool_removal_stat_t *prs) { char copied_buf[7], examined_buf[7], total_buf[7], rate_buf[7]; time_t start, end; nvlist_t *config, *nvroot; nvlist_t **child; uint_t children; char *vdev_name; if (prs == NULL || prs->prs_state == DSS_NONE) return; /* * Determine name of vdev. */ config = zpool_get_config(zhp, NULL); nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, &children) == 0); assert(prs->prs_removing_vdev < children); vdev_name = zpool_vdev_name(g_zfs, zhp, child[prs->prs_removing_vdev], B_TRUE); printf_color(ANSI_BOLD, gettext("remove: ")); start = prs->prs_start_time; end = prs->prs_end_time; zfs_nicenum(prs->prs_copied, copied_buf, sizeof (copied_buf)); /* * Removal is finished or canceled. */ if (prs->prs_state == DSS_FINISHED) { uint64_t minutes_taken = (end - start) / 60; (void) printf(gettext("Removal of vdev %llu copied %s " "in %lluh%um, completed on %s"), (longlong_t)prs->prs_removing_vdev, copied_buf, (u_longlong_t)(minutes_taken / 60), (uint_t)(minutes_taken % 60), ctime((time_t *)&end)); } else if (prs->prs_state == DSS_CANCELED) { (void) printf(gettext("Removal of %s canceled on %s"), vdev_name, ctime(&end)); } else { uint64_t copied, total, elapsed, mins_left, hours_left; double fraction_done; uint_t rate; assert(prs->prs_state == DSS_SCANNING); /* * Removal is in progress. */ (void) printf(gettext( "Evacuation of %s in progress since %s"), vdev_name, ctime(&start)); copied = prs->prs_copied > 0 ? prs->prs_copied : 1; total = prs->prs_to_copy; fraction_done = (double)copied / total; /* elapsed time for this pass */ elapsed = time(NULL) - prs->prs_start_time; elapsed = elapsed > 0 ? elapsed : 1; rate = copied / elapsed; rate = rate > 0 ? rate : 1; mins_left = ((total - copied) / rate) / 60; hours_left = mins_left / 60; zfs_nicenum(copied, examined_buf, sizeof (examined_buf)); zfs_nicenum(total, total_buf, sizeof (total_buf)); zfs_nicenum(rate, rate_buf, sizeof (rate_buf)); /* * do not print estimated time if hours_left is more than * 30 days */ (void) printf(gettext( "\t%s copied out of %s at %s/s, %.2f%% done"), examined_buf, total_buf, rate_buf, 100 * fraction_done); if (hours_left < (30 * 24)) { (void) printf(gettext(", %lluh%um to go\n"), (u_longlong_t)hours_left, (uint_t)(mins_left % 60)); } else { (void) printf(gettext( ", (copy is slow, no estimated time)\n")); } } free(vdev_name); if (prs->prs_mapping_memory > 0) { char mem_buf[7]; zfs_nicenum(prs->prs_mapping_memory, mem_buf, sizeof (mem_buf)); (void) printf(gettext( "\t%s memory used for removed device mappings\n"), mem_buf); } } /* * Print out detailed raidz expansion status. */ static void print_raidz_expand_status(zpool_handle_t *zhp, pool_raidz_expand_stat_t *pres) { char copied_buf[7]; if (pres == NULL || pres->pres_state == DSS_NONE) return; /* * Determine name of vdev. */ nvlist_t *config = zpool_get_config(zhp, NULL); nvlist_t *nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); nvlist_t **child; uint_t children; verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, &children) == 0); assert(pres->pres_expanding_vdev < children); printf_color(ANSI_BOLD, gettext("expand: ")); time_t start = pres->pres_start_time; time_t end = pres->pres_end_time; char *vname = zpool_vdev_name(g_zfs, zhp, child[pres->pres_expanding_vdev], 0); zfs_nicenum(pres->pres_reflowed, copied_buf, sizeof (copied_buf)); /* * Expansion is finished or canceled. */ if (pres->pres_state == DSS_FINISHED) { char time_buf[32]; secs_to_dhms(end - start, time_buf); (void) printf(gettext("expanded %s-%u copied %s in %s, " "on %s"), vname, (int)pres->pres_expanding_vdev, copied_buf, time_buf, ctime((time_t *)&end)); } else { char examined_buf[7], total_buf[7], rate_buf[7]; uint64_t copied, total, elapsed, secs_left; double fraction_done; uint_t rate; assert(pres->pres_state == DSS_SCANNING); /* * Expansion is in progress. */ (void) printf(gettext( "expansion of %s-%u in progress since %s"), vname, (int)pres->pres_expanding_vdev, ctime(&start)); copied = pres->pres_reflowed > 0 ? pres->pres_reflowed : 1; total = pres->pres_to_reflow; fraction_done = (double)copied / total; /* elapsed time for this pass */ elapsed = time(NULL) - pres->pres_start_time; elapsed = elapsed > 0 ? elapsed : 1; rate = copied / elapsed; rate = rate > 0 ? rate : 1; secs_left = (total - copied) / rate; zfs_nicenum(copied, examined_buf, sizeof (examined_buf)); zfs_nicenum(total, total_buf, sizeof (total_buf)); zfs_nicenum(rate, rate_buf, sizeof (rate_buf)); /* * do not print estimated time if hours_left is more than * 30 days */ (void) printf(gettext("\t%s / %s copied at %s/s, %.2f%% done"), examined_buf, total_buf, rate_buf, 100 * fraction_done); if (pres->pres_waiting_for_resilver) { (void) printf(gettext(", paused for resilver or " "clear\n")); } else if (secs_left < (30 * 24 * 3600)) { char time_buf[32]; secs_to_dhms(secs_left, time_buf); (void) printf(gettext(", %s to go\n"), time_buf); } else { (void) printf(gettext( ", (copy is slow, no estimated time)\n")); } } free(vname); } static void print_checkpoint_status(pool_checkpoint_stat_t *pcs) { time_t start; char space_buf[7]; if (pcs == NULL || pcs->pcs_state == CS_NONE) return; (void) printf(gettext("checkpoint: ")); start = pcs->pcs_start_time; zfs_nicenum(pcs->pcs_space, space_buf, sizeof (space_buf)); if (pcs->pcs_state == CS_CHECKPOINT_EXISTS) { char *date = ctime(&start); /* * ctime() adds a newline at the end of the generated * string, thus the weird format specifier and the * strlen() call used to chop it off from the output. */ (void) printf(gettext("created %.*s, consumes %s\n"), (int)(strlen(date) - 1), date, space_buf); return; } assert(pcs->pcs_state == CS_CHECKPOINT_DISCARDING); (void) printf(gettext("discarding, %s remaining.\n"), space_buf); } static void print_error_log(zpool_handle_t *zhp) { nvlist_t *nverrlist = NULL; nvpair_t *elem; char *pathname; size_t len = MAXPATHLEN * 2; if (zpool_get_errlog(zhp, &nverrlist) != 0) return; (void) printf("errors: Permanent errors have been " "detected in the following files:\n\n"); pathname = safe_malloc(len); elem = NULL; while ((elem = nvlist_next_nvpair(nverrlist, elem)) != NULL) { nvlist_t *nv; uint64_t dsobj, obj; verify(nvpair_value_nvlist(elem, &nv) == 0); verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_DATASET, &dsobj) == 0); verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_OBJECT, &obj) == 0); zpool_obj_to_path(zhp, dsobj, obj, pathname, len); (void) printf("%7s %s\n", "", pathname); } free(pathname); nvlist_free(nverrlist); } static void print_spares(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t **spares, uint_t nspares) { uint_t i; char *name; if (nspares == 0) return; (void) printf(gettext("\tspares\n")); for (i = 0; i < nspares; i++) { name = zpool_vdev_name(g_zfs, zhp, spares[i], cb->cb_name_flags); print_status_config(zhp, cb, name, spares[i], 2, B_TRUE, NULL); free(name); } } static void print_l2cache(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t **l2cache, uint_t nl2cache) { uint_t i; char *name; if (nl2cache == 0) return; (void) printf(gettext("\tcache\n")); for (i = 0; i < nl2cache; i++) { name = zpool_vdev_name(g_zfs, zhp, l2cache[i], cb->cb_name_flags); print_status_config(zhp, cb, name, l2cache[i], 2, B_FALSE, NULL); free(name); } } static void print_dedup_stats(nvlist_t *config) { ddt_histogram_t *ddh; ddt_stat_t *dds; ddt_object_t *ddo; uint_t c; char dspace[6], mspace[6]; /* * If the pool was faulted then we may not have been able to * obtain the config. Otherwise, if we have anything in the dedup * table continue processing the stats. */ if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_OBJ_STATS, (uint64_t **)&ddo, &c) != 0) return; (void) printf("\n"); (void) printf(gettext(" dedup: ")); if (ddo->ddo_count == 0) { (void) printf(gettext("no DDT entries\n")); return; } zfs_nicebytes(ddo->ddo_dspace, dspace, sizeof (dspace)); zfs_nicebytes(ddo->ddo_mspace, mspace, sizeof (mspace)); (void) printf("DDT entries %llu, size %s on disk, %s in core\n", (u_longlong_t)ddo->ddo_count, dspace, mspace); verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_STATS, (uint64_t **)&dds, &c) == 0); verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_HISTOGRAM, (uint64_t **)&ddh, &c) == 0); zpool_dump_ddt(dds, ddh); } /* * Display a summary of pool status. Displays a summary such as: * * pool: tank * status: DEGRADED * reason: One or more devices ... * see: https://openzfs.github.io/openzfs-docs/msg/ZFS-xxxx-01 * config: * mirror DEGRADED * c1t0d0 OK * c2t0d0 UNAVAIL * * When given the '-v' option, we print out the complete config. If the '-e' * option is specified, then we print out error rate information as well. */ static int status_callback(zpool_handle_t *zhp, void *data) { status_cbdata_t *cbp = data; nvlist_t *config, *nvroot; const char *msgid; zpool_status_t reason; zpool_errata_t errata; const char *health; uint_t c; vdev_stat_t *vs; config = zpool_get_config(zhp, NULL); reason = zpool_get_status(zhp, &msgid, &errata); cbp->cb_count++; /* * If we were given 'zpool status -x', only report those pools with * problems. */ if (cbp->cb_explain && (reason == ZPOOL_STATUS_OK || reason == ZPOOL_STATUS_VERSION_OLDER || reason == ZPOOL_STATUS_FEAT_DISABLED || reason == ZPOOL_STATUS_COMPATIBILITY_ERR || reason == ZPOOL_STATUS_INCOMPATIBLE_FEAT)) { if (!cbp->cb_allpools) { (void) printf(gettext("pool '%s' is healthy\n"), zpool_get_name(zhp)); if (cbp->cb_first) cbp->cb_first = B_FALSE; } return (0); } if (cbp->cb_first) cbp->cb_first = B_FALSE; else (void) printf("\n"); nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &c) == 0); health = zpool_get_state_str(zhp); printf(" "); printf_color(ANSI_BOLD, gettext("pool:")); printf(" %s\n", zpool_get_name(zhp)); fputc(' ', stdout); printf_color(ANSI_BOLD, gettext("state: ")); printf_color(health_str_to_color(health), "%s", health); fputc('\n', stdout); switch (reason) { case ZPOOL_STATUS_MISSING_DEV_R: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices could " "not be opened. Sufficient replicas exist for\n\tthe pool " "to continue functioning in a degraded state.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Attach the missing device " "and online it using 'zpool online'.\n")); break; case ZPOOL_STATUS_MISSING_DEV_NR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices could " "not be opened. There are insufficient\n\treplicas for the" " pool to continue functioning.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Attach the missing device " "and online it using 'zpool online'.\n")); break; case ZPOOL_STATUS_CORRUPT_LABEL_R: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices could " "not be used because the label is missing or\n\tinvalid. " "Sufficient replicas exist for the pool to continue\n\t" "functioning in a degraded state.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Replace the device using " "'zpool replace'.\n")); break; case ZPOOL_STATUS_CORRUPT_LABEL_NR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices could " "not be used because the label is missing \n\tor invalid. " "There are insufficient replicas for the pool to " "continue\n\tfunctioning.\n")); zpool_explain_recover(zpool_get_handle(zhp), zpool_get_name(zhp), reason, config); break; case ZPOOL_STATUS_FAILING_DEV: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices has " "experienced an unrecoverable error. An\n\tattempt was " "made to correct the error. Applications are " "unaffected.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Determine if the " "device needs to be replaced, and clear the errors\n\tusing" " 'zpool clear' or replace the device with 'zpool " "replace'.\n")); break; case ZPOOL_STATUS_OFFLINE_DEV: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices has " "been taken offline by the administrator.\n\tSufficient " "replicas exist for the pool to continue functioning in " "a\n\tdegraded state.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Online the device " "using 'zpool online' or replace the device with\n\t'zpool " "replace'.\n")); break; case ZPOOL_STATUS_REMOVED_DEV: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices has " "been removed by the administrator.\n\tSufficient " "replicas exist for the pool to continue functioning in " "a\n\tdegraded state.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Online the device " "using zpool online' or replace the device with\n\t'zpool " "replace'.\n")); break; case ZPOOL_STATUS_RESILVERING: case ZPOOL_STATUS_REBUILDING: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices is " "currently being resilvered. The pool will\n\tcontinue " "to function, possibly in a degraded state.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Wait for the resilver to " "complete.\n")); break; case ZPOOL_STATUS_REBUILD_SCRUB: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices have " "been sequentially resilvered, scrubbing\n\tthe pool " "is recommended.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Use 'zpool scrub' to " "verify all data checksums.\n")); break; case ZPOOL_STATUS_CORRUPT_DATA: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices has " "experienced an error resulting in data\n\tcorruption. " "Applications may be affected.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Restore the file in question" " if possible. Otherwise restore the\n\tentire pool from " "backup.\n")); break; case ZPOOL_STATUS_CORRUPT_POOL: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool metadata is " "corrupted and the pool cannot be opened.\n")); zpool_explain_recover(zpool_get_handle(zhp), zpool_get_name(zhp), reason, config); break; case ZPOOL_STATUS_VERSION_OLDER: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool is formatted using " "a legacy on-disk format. The pool can\n\tstill be used, " "but some features are unavailable.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Upgrade the pool using " "'zpool upgrade'. Once this is done, the\n\tpool will no " "longer be accessible on software that does not support\n\t" "feature flags.\n")); break; case ZPOOL_STATUS_VERSION_NEWER: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool has been upgraded " "to a newer, incompatible on-disk version.\n\tThe pool " "cannot be accessed on this system.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Access the pool from a " "system running more recent software, or\n\trestore the " "pool from backup.\n")); break; case ZPOOL_STATUS_FEAT_DISABLED: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("Some supported and " "requested features are not enabled on the pool.\n\t" "The pool can still be used, but some features are " "unavailable.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Enable all features using " "'zpool upgrade'. Once this is done,\n\tthe pool may no " "longer be accessible by software that does not support\n\t" "the features. See zpool-features(7) for details.\n")); break; case ZPOOL_STATUS_COMPATIBILITY_ERR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("This pool has a " "compatibility list specified, but it could not be\n\t" "read/parsed at this time. The pool can still be used, " "but this\n\tshould be investigated.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Check the value of the " "'compatibility' property against the\n\t" "appropriate file in " ZPOOL_SYSCONF_COMPAT_D " or " ZPOOL_DATA_COMPAT_D ".\n")); break; case ZPOOL_STATUS_INCOMPATIBLE_FEAT: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more features " "are enabled on the pool despite not being\n\t" "requested by the 'compatibility' property.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Consider setting " "'compatibility' to an appropriate value, or\n\t" "adding needed features to the relevant file in\n\t" ZPOOL_SYSCONF_COMPAT_D " or " ZPOOL_DATA_COMPAT_D ".\n")); break; case ZPOOL_STATUS_UNSUP_FEAT_READ: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed " "on this system because it uses the\n\tfollowing feature(s)" " not supported on this system:\n")); zpool_print_unsup_feat(config); (void) printf("\n"); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Access the pool from a " "system that supports the required feature(s),\n\tor " "restore the pool from backup.\n")); break; case ZPOOL_STATUS_UNSUP_FEAT_WRITE: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool can only be " "accessed in read-only mode on this system. It\n\tcannot be" " accessed in read-write mode because it uses the " "following\n\tfeature(s) not supported on this system:\n")); zpool_print_unsup_feat(config); (void) printf("\n"); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed " "in read-write mode. Import the pool with\n" "\t\"-o readonly=on\", access the pool from a system that " "supports the\n\trequired feature(s), or restore the " "pool from backup.\n")); break; case ZPOOL_STATUS_FAULTED_DEV_R: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices are " "faulted in response to persistent errors.\n\tSufficient " "replicas exist for the pool to continue functioning " "in a\n\tdegraded state.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Replace the faulted device, " "or use 'zpool clear' to mark the device\n\trepaired.\n")); break; case ZPOOL_STATUS_FAULTED_DEV_NR: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices are " "faulted in response to persistent errors. There are " "insufficient replicas for the pool to\n\tcontinue " "functioning.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Destroy and re-create the " "pool from a backup source. Manually marking the device\n" "\trepaired using 'zpool clear' may allow some data " "to be recovered.\n")); break; case ZPOOL_STATUS_IO_FAILURE_MMP: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("The pool is suspended " "because multihost writes failed or were delayed;\n\t" "another system could import the pool undetected.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Make sure the pool's devices" " are connected, then reboot your system and\n\timport the " "pool.\n")); break; case ZPOOL_STATUS_IO_FAILURE_WAIT: case ZPOOL_STATUS_IO_FAILURE_CONTINUE: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("One or more devices are " "faulted in response to IO failures.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Make sure the affected " "devices are connected, then run 'zpool clear'.\n")); break; case ZPOOL_STATUS_BAD_LOG: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("An intent log record " "could not be read.\n" "\tWaiting for administrator intervention to fix the " "faulted pool.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Either restore the affected " "device(s) and run 'zpool online',\n" "\tor ignore the intent log records by running " "'zpool clear'.\n")); break; case ZPOOL_STATUS_NON_NATIVE_ASHIFT: (void) printf(gettext("status: One or more devices are " "configured to use a non-native block size.\n" "\tExpect reduced performance.\n")); (void) printf(gettext("action: Replace affected devices with " "devices that support the\n\tconfigured block size, or " "migrate data to a properly configured\n\tpool.\n")); break; case ZPOOL_STATUS_HOSTID_MISMATCH: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("Mismatch between pool hostid" " and system hostid on imported pool.\n\tThis pool was " "previously imported into a system with a different " "hostid,\n\tand then was verbatim imported into this " "system.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("Export this pool on all " "systems on which it is imported.\n" "\tThen import it to correct the mismatch.\n")); break; case ZPOOL_STATUS_ERRATA: printf_color(ANSI_BOLD, gettext("status: ")); printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"), errata); switch (errata) { case ZPOOL_ERRATA_NONE: break; case ZPOOL_ERRATA_ZOL_2094_SCRUB: printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("To correct the issue" " run 'zpool scrub'.\n")); break; case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION: (void) printf(gettext("\tExisting encrypted datasets " "contain an on-disk incompatibility\n\twhich " "needs to be corrected.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("To correct the issue" " backup existing encrypted datasets to new\n\t" "encrypted datasets and destroy the old ones. " "'zfs mount -o ro' can\n\tbe used to temporarily " "mount existing encrypted datasets readonly.\n")); break; case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION: (void) printf(gettext("\tExisting encrypted snapshots " "and bookmarks contain an on-disk\n\tincompat" "ibility. This may cause on-disk corruption if " "they are used\n\twith 'zfs recv'.\n")); printf_color(ANSI_BOLD, gettext("action: ")); printf_color(ANSI_YELLOW, gettext("To correct the" "issue, enable the bookmark_v2 feature. No " "additional\n\taction is needed if there are no " "encrypted snapshots or bookmarks.\n\tIf preserving" "the encrypted snapshots and bookmarks is required," " use\n\ta non-raw send to backup and restore them." " Alternately, they may be\n\tremoved to resolve " "the incompatibility.\n")); break; default: /* * All errata which allow the pool to be imported * must contain an action message. */ assert(0); } break; default: /* * The remaining errors can't actually be generated, yet. */ assert(reason == ZPOOL_STATUS_OK); } if (msgid != NULL) { printf(" "); printf_color(ANSI_BOLD, gettext("see:")); printf(gettext( " https://openzfs.github.io/openzfs-docs/msg/%s\n"), msgid); } if (config != NULL) { uint64_t nerr; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; print_scan_status(zhp, nvroot); pool_removal_stat_t *prs = NULL; (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c); print_removal_status(zhp, prs); pool_checkpoint_stat_t *pcs = NULL; (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c); print_checkpoint_status(pcs); pool_raidz_expand_stat_t *pres = NULL; (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_RAIDZ_EXPAND_STATS, (uint64_t **)&pres, &c); print_raidz_expand_status(zhp, pres); cbp->cb_namewidth = max_width(zhp, nvroot, 0, 0, cbp->cb_name_flags | VDEV_NAME_TYPE_ID); if (cbp->cb_namewidth < 10) cbp->cb_namewidth = 10; color_start(ANSI_BOLD); (void) printf(gettext("config:\n\n")); (void) printf(gettext("\t%-*s %-8s %5s %5s %5s"), cbp->cb_namewidth, "NAME", "STATE", "READ", "WRITE", "CKSUM"); color_end(); if (cbp->cb_print_slow_ios) { printf_color(ANSI_BOLD, " %5s", gettext("SLOW")); } if (cbp->cb_print_power) { printf_color(ANSI_BOLD, " %5s", gettext("POWER")); } if (cbp->vcdl != NULL) print_cmd_columns(cbp->vcdl, 0); printf("\n"); print_status_config(zhp, cbp, zpool_get_name(zhp), nvroot, 0, B_FALSE, NULL); print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_DEDUP); print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_SPECIAL); print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_CLASS_LOGS); if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0) print_l2cache(zhp, cbp, l2cache, nl2cache); if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) print_spares(zhp, cbp, spares, nspares); if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT, &nerr) == 0) { (void) printf("\n"); if (nerr == 0) { (void) printf(gettext( "errors: No known data errors\n")); } else if (!cbp->cb_verbose) { color_start(ANSI_RED); (void) printf(gettext("errors: %llu data " "errors, use '-v' for a list\n"), (u_longlong_t)nerr); color_end(); } else { print_error_log(zhp); } } if (cbp->cb_dedup_stats) print_dedup_stats(config); } else { (void) printf(gettext("config: The configuration cannot be " "determined.\n")); } return (0); } /* * zpool status [-c [script1,script2,...]] [-DegiLpPstvx] [--power] [-T d|u] ... * [pool] [interval [count]] * * -c CMD For each vdev, run command CMD * -D Display dedup status (undocumented) * -e Display only unhealthy vdevs * -g Display guid for individual vdev name. * -i Display vdev initialization status. * -L Follow links when resolving vdev path name. * -p Display values in parsable (exact) format. * -P Display full path for vdev name. * -s Display slow IOs column. * -t Display vdev TRIM status. * -T Display a timestamp in date(1) or Unix format * -v Display complete error logs * -x Display only pools with potential problems * --power Display vdev enclosure slot power status * * Describes the health status of all pools or some subset. */ int zpool_do_status(int argc, char **argv) { int c; int ret; float interval = 0; unsigned long count = 0; status_cbdata_t cb = { 0 }; char *cmd = NULL; struct option long_options[] = { {"power", no_argument, NULL, ZPOOL_OPTION_POWER}, {0, 0, 0, 0} }; /* check options */ while ((c = getopt_long(argc, argv, "c:DegiLpPstT:vx", long_options, NULL)) != -1) { switch (c) { case 'c': if (cmd != NULL) { fprintf(stderr, gettext("Can't set -c flag twice\n")); exit(1); } if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL && !libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) { fprintf(stderr, gettext( "Can't run -c, disabled by " "ZPOOL_SCRIPTS_ENABLED.\n")); exit(1); } if ((getuid() <= 0 || geteuid() <= 0) && !libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) { fprintf(stderr, gettext( "Can't run -c with root privileges " "unless ZPOOL_SCRIPTS_AS_ROOT is set.\n")); exit(1); } cmd = optarg; break; case 'D': cb.cb_dedup_stats = B_TRUE; break; case 'e': cb.cb_print_unhealthy = B_TRUE; break; case 'g': cb.cb_name_flags |= VDEV_NAME_GUID; break; case 'i': cb.cb_print_vdev_init = B_TRUE; break; case 'L': cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS; break; case 'p': cb.cb_literal = B_TRUE; break; case 'P': cb.cb_name_flags |= VDEV_NAME_PATH; break; case 's': cb.cb_print_slow_ios = B_TRUE; break; case 't': cb.cb_print_vdev_trim = B_TRUE; break; case 'T': get_timestamp_arg(*optarg); break; case 'v': cb.cb_verbose = B_TRUE; break; case 'x': cb.cb_explain = B_TRUE; break; case ZPOOL_OPTION_POWER: cb.cb_print_power = B_TRUE; break; case '?': if (optopt == 'c') { print_zpool_script_list("status"); exit(0); } else { fprintf(stderr, gettext("invalid option '%c'\n"), optopt); } usage(B_FALSE); } } argc -= optind; argv += optind; get_interval_count(&argc, argv, &interval, &count); if (argc == 0) cb.cb_allpools = B_TRUE; cb.cb_first = B_TRUE; cb.cb_print_status = B_TRUE; for (;;) { if (timestamp_fmt != NODATE) print_timestamp(timestamp_fmt); if (cmd != NULL) cb.vcdl = all_pools_for_each_vdev_run(argc, argv, cmd, NULL, NULL, 0, 0); ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL, cb.cb_literal, status_callback, &cb); if (cb.vcdl != NULL) free_vdev_cmd_data_list(cb.vcdl); if (argc == 0 && cb.cb_count == 0) (void) fprintf(stderr, gettext("no pools available\n")); else if (cb.cb_explain && cb.cb_first && cb.cb_allpools) (void) printf(gettext("all pools are healthy\n")); if (ret != 0) return (ret); if (interval == 0) break; if (count != 0 && --count == 0) break; (void) fflush(stdout); (void) fsleep(interval); } return (0); } typedef struct upgrade_cbdata { int cb_first; int cb_argc; uint64_t cb_version; char **cb_argv; } upgrade_cbdata_t; static int check_unsupp_fs(zfs_handle_t *zhp, void *unsupp_fs) { int zfs_version = (int)zfs_prop_get_int(zhp, ZFS_PROP_VERSION); int *count = (int *)unsupp_fs; if (zfs_version > ZPL_VERSION) { (void) printf(gettext("%s (v%d) is not supported by this " "implementation of ZFS.\n"), zfs_get_name(zhp), zfs_version); (*count)++; } zfs_iter_filesystems_v2(zhp, 0, check_unsupp_fs, unsupp_fs); zfs_close(zhp); return (0); } static int upgrade_version(zpool_handle_t *zhp, uint64_t version) { int ret; nvlist_t *config; uint64_t oldversion; int unsupp_fs = 0; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &oldversion) == 0); char compat[ZFS_MAXPROPLEN]; if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat, ZFS_MAXPROPLEN, NULL, B_FALSE) != 0) compat[0] = '\0'; assert(SPA_VERSION_IS_SUPPORTED(oldversion)); assert(oldversion < version); ret = zfs_iter_root(zpool_get_handle(zhp), check_unsupp_fs, &unsupp_fs); if (ret != 0) return (ret); if (unsupp_fs) { (void) fprintf(stderr, gettext("Upgrade not performed due " "to %d unsupported filesystems (max v%d).\n"), unsupp_fs, (int)ZPL_VERSION); return (1); } if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) { (void) fprintf(stderr, gettext("Upgrade not performed because " "'compatibility' property set to '" ZPOOL_COMPAT_LEGACY "'.\n")); return (1); } ret = zpool_upgrade(zhp, version); if (ret != 0) return (ret); if (version >= SPA_VERSION_FEATURES) { (void) printf(gettext("Successfully upgraded " "'%s' from version %llu to feature flags.\n"), zpool_get_name(zhp), (u_longlong_t)oldversion); } else { (void) printf(gettext("Successfully upgraded " "'%s' from version %llu to version %llu.\n"), zpool_get_name(zhp), (u_longlong_t)oldversion, (u_longlong_t)version); } return (0); } static int upgrade_enable_all(zpool_handle_t *zhp, int *countp) { int i, ret, count; boolean_t firstff = B_TRUE; nvlist_t *enabled = zpool_get_features(zhp); char compat[ZFS_MAXPROPLEN]; if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat, ZFS_MAXPROPLEN, NULL, B_FALSE) != 0) compat[0] = '\0'; boolean_t requested_features[SPA_FEATURES]; if (zpool_do_load_compat(compat, requested_features) != ZPOOL_COMPATIBILITY_OK) return (-1); count = 0; for (i = 0; i < SPA_FEATURES; i++) { const char *fname = spa_feature_table[i].fi_uname; const char *fguid = spa_feature_table[i].fi_guid; if (!spa_feature_table[i].fi_zfs_mod_supported) continue; if (!nvlist_exists(enabled, fguid) && requested_features[i]) { char *propname; verify(-1 != asprintf(&propname, "feature@%s", fname)); ret = zpool_set_prop(zhp, propname, ZFS_FEATURE_ENABLED); if (ret != 0) { free(propname); return (ret); } count++; if (firstff) { (void) printf(gettext("Enabled the " "following features on '%s':\n"), zpool_get_name(zhp)); firstff = B_FALSE; } (void) printf(gettext(" %s\n"), fname); free(propname); } } if (countp != NULL) *countp = count; return (0); } static int upgrade_cb(zpool_handle_t *zhp, void *arg) { upgrade_cbdata_t *cbp = arg; nvlist_t *config; uint64_t version; boolean_t modified_pool = B_FALSE; int ret; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) == 0); assert(SPA_VERSION_IS_SUPPORTED(version)); if (version < cbp->cb_version) { cbp->cb_first = B_FALSE; ret = upgrade_version(zhp, cbp->cb_version); if (ret != 0) return (ret); modified_pool = B_TRUE; /* * If they did "zpool upgrade -a", then we could * be doing ioctls to different pools. We need * to log this history once to each pool, and bypass * the normal history logging that happens in main(). */ (void) zpool_log_history(g_zfs, history_str); log_history = B_FALSE; } if (cbp->cb_version >= SPA_VERSION_FEATURES) { int count; ret = upgrade_enable_all(zhp, &count); if (ret != 0) return (ret); if (count > 0) { cbp->cb_first = B_FALSE; modified_pool = B_TRUE; } } if (modified_pool) { (void) printf("\n"); (void) after_zpool_upgrade(zhp); } return (0); } static int upgrade_list_older_cb(zpool_handle_t *zhp, void *arg) { upgrade_cbdata_t *cbp = arg; nvlist_t *config; uint64_t version; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) == 0); assert(SPA_VERSION_IS_SUPPORTED(version)); if (version < SPA_VERSION_FEATURES) { if (cbp->cb_first) { (void) printf(gettext("The following pools are " "formatted with legacy version numbers and can\n" "be upgraded to use feature flags. After " "being upgraded, these pools\nwill no " "longer be accessible by software that does not " "support feature\nflags.\n\n" "Note that setting a pool's 'compatibility' " "feature to '" ZPOOL_COMPAT_LEGACY "' will\n" "inhibit upgrades.\n\n")); (void) printf(gettext("VER POOL\n")); (void) printf(gettext("--- ------------\n")); cbp->cb_first = B_FALSE; } (void) printf("%2llu %s\n", (u_longlong_t)version, zpool_get_name(zhp)); } return (0); } static int upgrade_list_disabled_cb(zpool_handle_t *zhp, void *arg) { upgrade_cbdata_t *cbp = arg; nvlist_t *config; uint64_t version; config = zpool_get_config(zhp, NULL); verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) == 0); if (version >= SPA_VERSION_FEATURES) { int i; boolean_t poolfirst = B_TRUE; nvlist_t *enabled = zpool_get_features(zhp); for (i = 0; i < SPA_FEATURES; i++) { const char *fguid = spa_feature_table[i].fi_guid; const char *fname = spa_feature_table[i].fi_uname; if (!spa_feature_table[i].fi_zfs_mod_supported) continue; if (!nvlist_exists(enabled, fguid)) { if (cbp->cb_first) { (void) printf(gettext("\nSome " "supported features are not " "enabled on the following pools. " "Once a\nfeature is enabled the " "pool may become incompatible with " "software\nthat does not support " "the feature. See " "zpool-features(7) for " "details.\n\n" "Note that the pool " "'compatibility' feature can be " "used to inhibit\nfeature " "upgrades.\n\n")); (void) printf(gettext("POOL " "FEATURE\n")); (void) printf(gettext("------" "---------\n")); cbp->cb_first = B_FALSE; } if (poolfirst) { (void) printf(gettext("%s\n"), zpool_get_name(zhp)); poolfirst = B_FALSE; } (void) printf(gettext(" %s\n"), fname); } /* * If they did "zpool upgrade -a", then we could * be doing ioctls to different pools. We need * to log this history once to each pool, and bypass * the normal history logging that happens in main(). */ (void) zpool_log_history(g_zfs, history_str); log_history = B_FALSE; } } return (0); } static int upgrade_one(zpool_handle_t *zhp, void *data) { boolean_t modified_pool = B_FALSE; upgrade_cbdata_t *cbp = data; uint64_t cur_version; int ret; if (strcmp("log", zpool_get_name(zhp)) == 0) { (void) fprintf(stderr, gettext("'log' is now a reserved word\n" "Pool 'log' must be renamed using export and import" " to upgrade.\n")); return (1); } cur_version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); if (cur_version > cbp->cb_version) { (void) printf(gettext("Pool '%s' is already formatted " "using more current version '%llu'.\n\n"), zpool_get_name(zhp), (u_longlong_t)cur_version); return (0); } if (cbp->cb_version != SPA_VERSION && cur_version == cbp->cb_version) { (void) printf(gettext("Pool '%s' is already formatted " "using version %llu.\n\n"), zpool_get_name(zhp), (u_longlong_t)cbp->cb_version); return (0); } if (cur_version != cbp->cb_version) { modified_pool = B_TRUE; ret = upgrade_version(zhp, cbp->cb_version); if (ret != 0) return (ret); } if (cbp->cb_version >= SPA_VERSION_FEATURES) { int count = 0; ret = upgrade_enable_all(zhp, &count); if (ret != 0) return (ret); if (count != 0) { modified_pool = B_TRUE; } else if (cur_version == SPA_VERSION) { (void) printf(gettext("Pool '%s' already has all " "supported and requested features enabled.\n"), zpool_get_name(zhp)); } } if (modified_pool) { (void) printf("\n"); (void) after_zpool_upgrade(zhp); } return (0); } /* * zpool upgrade * zpool upgrade -v * zpool upgrade [-V version] <-a | pool ...> * * With no arguments, display downrev'd ZFS pool available for upgrade. * Individual pools can be upgraded by specifying the pool, and '-a' will * upgrade all pools. */ int zpool_do_upgrade(int argc, char **argv) { int c; upgrade_cbdata_t cb = { 0 }; int ret = 0; boolean_t showversions = B_FALSE; boolean_t upgradeall = B_FALSE; char *end; /* check options */ while ((c = getopt(argc, argv, ":avV:")) != -1) { switch (c) { case 'a': upgradeall = B_TRUE; break; case 'v': showversions = B_TRUE; break; case 'V': cb.cb_version = strtoll(optarg, &end, 10); if (*end != '\0' || !SPA_VERSION_IS_SUPPORTED(cb.cb_version)) { (void) fprintf(stderr, gettext("invalid version '%s'\n"), optarg); usage(B_FALSE); } break; case ':': (void) fprintf(stderr, gettext("missing argument for " "'%c' option\n"), optopt); usage(B_FALSE); break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } cb.cb_argc = argc; cb.cb_argv = argv; argc -= optind; argv += optind; if (cb.cb_version == 0) { cb.cb_version = SPA_VERSION; } else if (!upgradeall && argc == 0) { (void) fprintf(stderr, gettext("-V option is " "incompatible with other arguments\n")); usage(B_FALSE); } if (showversions) { if (upgradeall || argc != 0) { (void) fprintf(stderr, gettext("-v option is " "incompatible with other arguments\n")); usage(B_FALSE); } } else if (upgradeall) { if (argc != 0) { (void) fprintf(stderr, gettext("-a option should not " "be used along with a pool name\n")); usage(B_FALSE); } } (void) printf("%s", gettext("This system supports ZFS pool feature " "flags.\n\n")); if (showversions) { int i; (void) printf(gettext("The following features are " "supported:\n\n")); (void) printf(gettext("FEAT DESCRIPTION\n")); (void) printf("----------------------------------------------" "---------------\n"); for (i = 0; i < SPA_FEATURES; i++) { zfeature_info_t *fi = &spa_feature_table[i]; if (!fi->fi_zfs_mod_supported) continue; const char *ro = (fi->fi_flags & ZFEATURE_FLAG_READONLY_COMPAT) ? " (read-only compatible)" : ""; (void) printf("%-37s%s\n", fi->fi_uname, ro); (void) printf(" %s\n", fi->fi_desc); } (void) printf("\n"); (void) printf(gettext("The following legacy versions are also " "supported:\n\n")); (void) printf(gettext("VER DESCRIPTION\n")); (void) printf("--- -----------------------------------------" "---------------\n"); (void) printf(gettext(" 1 Initial ZFS version\n")); (void) printf(gettext(" 2 Ditto blocks " "(replicated metadata)\n")); (void) printf(gettext(" 3 Hot spares and double parity " "RAID-Z\n")); (void) printf(gettext(" 4 zpool history\n")); (void) printf(gettext(" 5 Compression using the gzip " "algorithm\n")); (void) printf(gettext(" 6 bootfs pool property\n")); (void) printf(gettext(" 7 Separate intent log devices\n")); (void) printf(gettext(" 8 Delegated administration\n")); (void) printf(gettext(" 9 refquota and refreservation " "properties\n")); (void) printf(gettext(" 10 Cache devices\n")); (void) printf(gettext(" 11 Improved scrub performance\n")); (void) printf(gettext(" 12 Snapshot properties\n")); (void) printf(gettext(" 13 snapused property\n")); (void) printf(gettext(" 14 passthrough-x aclinherit\n")); (void) printf(gettext(" 15 user/group space accounting\n")); (void) printf(gettext(" 16 stmf property support\n")); (void) printf(gettext(" 17 Triple-parity RAID-Z\n")); (void) printf(gettext(" 18 Snapshot user holds\n")); (void) printf(gettext(" 19 Log device removal\n")); (void) printf(gettext(" 20 Compression using zle " "(zero-length encoding)\n")); (void) printf(gettext(" 21 Deduplication\n")); (void) printf(gettext(" 22 Received properties\n")); (void) printf(gettext(" 23 Slim ZIL\n")); (void) printf(gettext(" 24 System attributes\n")); (void) printf(gettext(" 25 Improved scrub stats\n")); (void) printf(gettext(" 26 Improved snapshot deletion " "performance\n")); (void) printf(gettext(" 27 Improved snapshot creation " "performance\n")); (void) printf(gettext(" 28 Multiple vdev replacements\n")); (void) printf(gettext("\nFor more information on a particular " "version, including supported releases,\n")); (void) printf(gettext("see the ZFS Administration Guide.\n\n")); } else if (argc == 0 && upgradeall) { cb.cb_first = B_TRUE; ret = zpool_iter(g_zfs, upgrade_cb, &cb); if (ret == 0 && cb.cb_first) { if (cb.cb_version == SPA_VERSION) { (void) printf(gettext("All pools are already " "formatted using feature flags.\n\n")); (void) printf(gettext("Every feature flags " "pool already has all supported and " "requested features enabled.\n")); } else { (void) printf(gettext("All pools are already " "formatted with version %llu or higher.\n"), (u_longlong_t)cb.cb_version); } } } else if (argc == 0) { cb.cb_first = B_TRUE; ret = zpool_iter(g_zfs, upgrade_list_older_cb, &cb); assert(ret == 0); if (cb.cb_first) { (void) printf(gettext("All pools are formatted " "using feature flags.\n\n")); } else { (void) printf(gettext("\nUse 'zpool upgrade -v' " "for a list of available legacy versions.\n")); } cb.cb_first = B_TRUE; ret = zpool_iter(g_zfs, upgrade_list_disabled_cb, &cb); assert(ret == 0); if (cb.cb_first) { (void) printf(gettext("Every feature flags pool has " "all supported and requested features enabled.\n")); } else { (void) printf(gettext("\n")); } } else { ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL, B_FALSE, upgrade_one, &cb); } return (ret); } typedef struct hist_cbdata { boolean_t first; boolean_t longfmt; boolean_t internal; } hist_cbdata_t; static void print_history_records(nvlist_t *nvhis, hist_cbdata_t *cb) { nvlist_t **records; uint_t numrecords; int i; verify(nvlist_lookup_nvlist_array(nvhis, ZPOOL_HIST_RECORD, &records, &numrecords) == 0); for (i = 0; i < numrecords; i++) { nvlist_t *rec = records[i]; char tbuf[64] = ""; if (nvlist_exists(rec, ZPOOL_HIST_TIME)) { time_t tsec; struct tm t; tsec = fnvlist_lookup_uint64(records[i], ZPOOL_HIST_TIME); (void) localtime_r(&tsec, &t); (void) strftime(tbuf, sizeof (tbuf), "%F.%T", &t); } if (nvlist_exists(rec, ZPOOL_HIST_ELAPSED_NS)) { uint64_t elapsed_ns = fnvlist_lookup_int64(records[i], ZPOOL_HIST_ELAPSED_NS); (void) snprintf(tbuf + strlen(tbuf), sizeof (tbuf) - strlen(tbuf), " (%lldms)", (long long)elapsed_ns / 1000 / 1000); } if (nvlist_exists(rec, ZPOOL_HIST_CMD)) { (void) printf("%s %s", tbuf, fnvlist_lookup_string(rec, ZPOOL_HIST_CMD)); } else if (nvlist_exists(rec, ZPOOL_HIST_INT_EVENT)) { int ievent = fnvlist_lookup_uint64(rec, ZPOOL_HIST_INT_EVENT); if (!cb->internal) continue; if (ievent >= ZFS_NUM_LEGACY_HISTORY_EVENTS) { (void) printf("%s unrecognized record:\n", tbuf); dump_nvlist(rec, 4); continue; } (void) printf("%s [internal %s txg:%lld] %s", tbuf, zfs_history_event_names[ievent], (longlong_t)fnvlist_lookup_uint64( rec, ZPOOL_HIST_TXG), fnvlist_lookup_string(rec, ZPOOL_HIST_INT_STR)); } else if (nvlist_exists(rec, ZPOOL_HIST_INT_NAME)) { if (!cb->internal) continue; (void) printf("%s [txg:%lld] %s", tbuf, (longlong_t)fnvlist_lookup_uint64( rec, ZPOOL_HIST_TXG), fnvlist_lookup_string(rec, ZPOOL_HIST_INT_NAME)); if (nvlist_exists(rec, ZPOOL_HIST_DSNAME)) { (void) printf(" %s (%llu)", fnvlist_lookup_string(rec, ZPOOL_HIST_DSNAME), (u_longlong_t)fnvlist_lookup_uint64(rec, ZPOOL_HIST_DSID)); } (void) printf(" %s", fnvlist_lookup_string(rec, ZPOOL_HIST_INT_STR)); } else if (nvlist_exists(rec, ZPOOL_HIST_IOCTL)) { if (!cb->internal) continue; (void) printf("%s ioctl %s\n", tbuf, fnvlist_lookup_string(rec, ZPOOL_HIST_IOCTL)); if (nvlist_exists(rec, ZPOOL_HIST_INPUT_NVL)) { (void) printf(" input:\n"); dump_nvlist(fnvlist_lookup_nvlist(rec, ZPOOL_HIST_INPUT_NVL), 8); } if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_NVL)) { (void) printf(" output:\n"); dump_nvlist(fnvlist_lookup_nvlist(rec, ZPOOL_HIST_OUTPUT_NVL), 8); } if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_SIZE)) { (void) printf(" output nvlist omitted; " "original size: %lldKB\n", (longlong_t)fnvlist_lookup_int64(rec, ZPOOL_HIST_OUTPUT_SIZE) / 1024); } if (nvlist_exists(rec, ZPOOL_HIST_ERRNO)) { (void) printf(" errno: %lld\n", (longlong_t)fnvlist_lookup_int64(rec, ZPOOL_HIST_ERRNO)); } } else { if (!cb->internal) continue; (void) printf("%s unrecognized record:\n", tbuf); dump_nvlist(rec, 4); } if (!cb->longfmt) { (void) printf("\n"); continue; } (void) printf(" ["); if (nvlist_exists(rec, ZPOOL_HIST_WHO)) { uid_t who = fnvlist_lookup_uint64(rec, ZPOOL_HIST_WHO); struct passwd *pwd = getpwuid(who); (void) printf("user %d ", (int)who); if (pwd != NULL) (void) printf("(%s) ", pwd->pw_name); } if (nvlist_exists(rec, ZPOOL_HIST_HOST)) { (void) printf("on %s", fnvlist_lookup_string(rec, ZPOOL_HIST_HOST)); } if (nvlist_exists(rec, ZPOOL_HIST_ZONE)) { (void) printf(":%s", fnvlist_lookup_string(rec, ZPOOL_HIST_ZONE)); } (void) printf("]"); (void) printf("\n"); } } /* * Print out the command history for a specific pool. */ static int get_history_one(zpool_handle_t *zhp, void *data) { nvlist_t *nvhis; int ret; hist_cbdata_t *cb = (hist_cbdata_t *)data; uint64_t off = 0; boolean_t eof = B_FALSE; cb->first = B_FALSE; (void) printf(gettext("History for '%s':\n"), zpool_get_name(zhp)); while (!eof) { if ((ret = zpool_get_history(zhp, &nvhis, &off, &eof)) != 0) return (ret); print_history_records(nvhis, cb); nvlist_free(nvhis); } (void) printf("\n"); return (ret); } /* * zpool history * * Displays the history of commands that modified pools. */ int zpool_do_history(int argc, char **argv) { hist_cbdata_t cbdata = { 0 }; int ret; int c; cbdata.first = B_TRUE; /* check options */ while ((c = getopt(argc, argv, "li")) != -1) { switch (c) { case 'l': cbdata.longfmt = B_TRUE; break; case 'i': cbdata.internal = B_TRUE; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL, B_FALSE, get_history_one, &cbdata); if (argc == 0 && cbdata.first == B_TRUE) { (void) fprintf(stderr, gettext("no pools available\n")); return (0); } return (ret); } typedef struct ev_opts { int verbose; int scripted; int follow; int clear; char poolname[ZFS_MAX_DATASET_NAME_LEN]; } ev_opts_t; static void zpool_do_events_short(nvlist_t *nvl, ev_opts_t *opts) { char ctime_str[26], str[32]; const char *ptr; int64_t *tv; uint_t n; verify(nvlist_lookup_int64_array(nvl, FM_EREPORT_TIME, &tv, &n) == 0); memset(str, ' ', 32); (void) ctime_r((const time_t *)&tv[0], ctime_str); (void) memcpy(str, ctime_str+4, 6); /* 'Jun 30' */ (void) memcpy(str+7, ctime_str+20, 4); /* '1993' */ (void) memcpy(str+12, ctime_str+11, 8); /* '21:49:08' */ (void) sprintf(str+20, ".%09lld", (longlong_t)tv[1]); /* '.123456789' */ if (opts->scripted) (void) printf(gettext("%s\t"), str); else (void) printf(gettext("%s "), str); verify(nvlist_lookup_string(nvl, FM_CLASS, &ptr) == 0); (void) printf(gettext("%s\n"), ptr); } static void zpool_do_events_nvprint(nvlist_t *nvl, int depth) { nvpair_t *nvp; for (nvp = nvlist_next_nvpair(nvl, NULL); nvp != NULL; nvp = nvlist_next_nvpair(nvl, nvp)) { data_type_t type = nvpair_type(nvp); const char *name = nvpair_name(nvp); boolean_t b; uint8_t i8; uint16_t i16; uint32_t i32; uint64_t i64; const char *str; nvlist_t *cnv; printf(gettext("%*s%s = "), depth, "", name); switch (type) { case DATA_TYPE_BOOLEAN: printf(gettext("%s"), "1"); break; case DATA_TYPE_BOOLEAN_VALUE: (void) nvpair_value_boolean_value(nvp, &b); printf(gettext("%s"), b ? "1" : "0"); break; case DATA_TYPE_BYTE: (void) nvpair_value_byte(nvp, &i8); printf(gettext("0x%x"), i8); break; case DATA_TYPE_INT8: (void) nvpair_value_int8(nvp, (void *)&i8); printf(gettext("0x%x"), i8); break; case DATA_TYPE_UINT8: (void) nvpair_value_uint8(nvp, &i8); printf(gettext("0x%x"), i8); break; case DATA_TYPE_INT16: (void) nvpair_value_int16(nvp, (void *)&i16); printf(gettext("0x%x"), i16); break; case DATA_TYPE_UINT16: (void) nvpair_value_uint16(nvp, &i16); printf(gettext("0x%x"), i16); break; case DATA_TYPE_INT32: (void) nvpair_value_int32(nvp, (void *)&i32); printf(gettext("0x%x"), i32); break; case DATA_TYPE_UINT32: (void) nvpair_value_uint32(nvp, &i32); printf(gettext("0x%x"), i32); break; case DATA_TYPE_INT64: (void) nvpair_value_int64(nvp, (void *)&i64); printf(gettext("0x%llx"), (u_longlong_t)i64); break; case DATA_TYPE_UINT64: (void) nvpair_value_uint64(nvp, &i64); /* * translate vdev state values to readable * strings to aide zpool events consumers */ if (strcmp(name, FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE) == 0 || strcmp(name, FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE) == 0) { printf(gettext("\"%s\" (0x%llx)"), zpool_state_to_name(i64, VDEV_AUX_NONE), (u_longlong_t)i64); } else { printf(gettext("0x%llx"), (u_longlong_t)i64); } break; case DATA_TYPE_HRTIME: (void) nvpair_value_hrtime(nvp, (void *)&i64); printf(gettext("0x%llx"), (u_longlong_t)i64); break; case DATA_TYPE_STRING: (void) nvpair_value_string(nvp, &str); printf(gettext("\"%s\""), str ? str : ""); break; case DATA_TYPE_NVLIST: printf(gettext("(embedded nvlist)\n")); (void) nvpair_value_nvlist(nvp, &cnv); zpool_do_events_nvprint(cnv, depth + 8); printf(gettext("%*s(end %s)"), depth, "", name); break; case DATA_TYPE_NVLIST_ARRAY: { nvlist_t **val; uint_t i, nelem; (void) nvpair_value_nvlist_array(nvp, &val, &nelem); printf(gettext("(%d embedded nvlists)\n"), nelem); for (i = 0; i < nelem; i++) { printf(gettext("%*s%s[%d] = %s\n"), depth, "", name, i, "(embedded nvlist)"); zpool_do_events_nvprint(val[i], depth + 8); printf(gettext("%*s(end %s[%i])\n"), depth, "", name, i); } printf(gettext("%*s(end %s)\n"), depth, "", name); } break; case DATA_TYPE_INT8_ARRAY: { int8_t *val; uint_t i, nelem; (void) nvpair_value_int8_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%x "), val[i]); break; } case DATA_TYPE_UINT8_ARRAY: { uint8_t *val; uint_t i, nelem; (void) nvpair_value_uint8_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%x "), val[i]); break; } case DATA_TYPE_INT16_ARRAY: { int16_t *val; uint_t i, nelem; (void) nvpair_value_int16_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%x "), val[i]); break; } case DATA_TYPE_UINT16_ARRAY: { uint16_t *val; uint_t i, nelem; (void) nvpair_value_uint16_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%x "), val[i]); break; } case DATA_TYPE_INT32_ARRAY: { int32_t *val; uint_t i, nelem; (void) nvpair_value_int32_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%x "), val[i]); break; } case DATA_TYPE_UINT32_ARRAY: { uint32_t *val; uint_t i, nelem; (void) nvpair_value_uint32_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%x "), val[i]); break; } case DATA_TYPE_INT64_ARRAY: { int64_t *val; uint_t i, nelem; (void) nvpair_value_int64_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%llx "), (u_longlong_t)val[i]); break; } case DATA_TYPE_UINT64_ARRAY: { uint64_t *val; uint_t i, nelem; (void) nvpair_value_uint64_array(nvp, &val, &nelem); for (i = 0; i < nelem; i++) printf(gettext("0x%llx "), (u_longlong_t)val[i]); break; } case DATA_TYPE_STRING_ARRAY: { const char **str; uint_t i, nelem; (void) nvpair_value_string_array(nvp, &str, &nelem); for (i = 0; i < nelem; i++) printf(gettext("\"%s\" "), str[i] ? str[i] : ""); break; } case DATA_TYPE_BOOLEAN_ARRAY: case DATA_TYPE_BYTE_ARRAY: case DATA_TYPE_DOUBLE: case DATA_TYPE_DONTCARE: case DATA_TYPE_UNKNOWN: printf(gettext("")); break; } printf(gettext("\n")); } } static int zpool_do_events_next(ev_opts_t *opts) { nvlist_t *nvl; int zevent_fd, ret, dropped; const char *pool; zevent_fd = open(ZFS_DEV, O_RDWR); VERIFY(zevent_fd >= 0); if (!opts->scripted) (void) printf(gettext("%-30s %s\n"), "TIME", "CLASS"); while (1) { ret = zpool_events_next(g_zfs, &nvl, &dropped, (opts->follow ? ZEVENT_NONE : ZEVENT_NONBLOCK), zevent_fd); if (ret || nvl == NULL) break; if (dropped > 0) (void) printf(gettext("dropped %d events\n"), dropped); if (strlen(opts->poolname) > 0 && nvlist_lookup_string(nvl, FM_FMRI_ZFS_POOL, &pool) == 0 && strcmp(opts->poolname, pool) != 0) continue; zpool_do_events_short(nvl, opts); if (opts->verbose) { zpool_do_events_nvprint(nvl, 8); printf(gettext("\n")); } (void) fflush(stdout); nvlist_free(nvl); } VERIFY(0 == close(zevent_fd)); return (ret); } static int zpool_do_events_clear(void) { int count, ret; ret = zpool_events_clear(g_zfs, &count); if (!ret) (void) printf(gettext("cleared %d events\n"), count); return (ret); } /* * zpool events [-vHf [pool] | -c] * * Displays events logs by ZFS. */ int zpool_do_events(int argc, char **argv) { ev_opts_t opts = { 0 }; int ret; int c; /* check options */ while ((c = getopt(argc, argv, "vHfc")) != -1) { switch (c) { case 'v': opts.verbose = 1; break; case 'H': opts.scripted = 1; break; case 'f': opts.follow = 1; break; case 'c': opts.clear = 1; break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } else if (argc == 1) { (void) strlcpy(opts.poolname, argv[0], sizeof (opts.poolname)); if (!zfs_name_valid(opts.poolname, ZFS_TYPE_POOL)) { (void) fprintf(stderr, gettext("invalid pool name '%s'\n"), opts.poolname); usage(B_FALSE); } } if ((argc == 1 || opts.verbose || opts.scripted || opts.follow) && opts.clear) { (void) fprintf(stderr, gettext("invalid options combined with -c\n")); usage(B_FALSE); } if (opts.clear) ret = zpool_do_events_clear(); else ret = zpool_do_events_next(&opts); return (ret); } static int get_callback_vdev(zpool_handle_t *zhp, char *vdevname, void *data) { zprop_get_cbdata_t *cbp = (zprop_get_cbdata_t *)data; char value[ZFS_MAXPROPLEN]; zprop_source_t srctype; for (zprop_list_t *pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) { char *prop_name; /* * If the first property is pool name, it is a special * placeholder that we can skip. This will also skip * over the name property when 'all' is specified. */ if (pl->pl_prop == ZPOOL_PROP_NAME && pl == cbp->cb_proplist) continue; if (pl->pl_prop == ZPROP_INVAL) { prop_name = pl->pl_user_prop; } else { prop_name = (char *)vdev_prop_to_name(pl->pl_prop); } if (zpool_get_vdev_prop(zhp, vdevname, pl->pl_prop, prop_name, value, sizeof (value), &srctype, cbp->cb_literal) == 0) { zprop_print_one_property(vdevname, cbp, prop_name, value, srctype, NULL, NULL); } } return (0); } static int get_callback_vdev_cb(void *zhp_data, nvlist_t *nv, void *data) { zpool_handle_t *zhp = zhp_data; zprop_get_cbdata_t *cbp = (zprop_get_cbdata_t *)data; char *vdevname; const char *type; int ret; /* * zpool_vdev_name() transforms the root vdev name (i.e., root-0) to the * pool name for display purposes, which is not desired. Fallback to * zpool_vdev_name() when not dealing with the root vdev. */ type = fnvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE); if (zhp != NULL && strcmp(type, "root") == 0) vdevname = strdup("root-0"); else vdevname = zpool_vdev_name(g_zfs, zhp, nv, cbp->cb_vdevs.cb_name_flags); (void) vdev_expand_proplist(zhp, vdevname, &cbp->cb_proplist); ret = get_callback_vdev(zhp, vdevname, data); free(vdevname); return (ret); } static int get_callback(zpool_handle_t *zhp, void *data) { zprop_get_cbdata_t *cbp = (zprop_get_cbdata_t *)data; char value[ZFS_MAXPROPLEN]; zprop_source_t srctype; zprop_list_t *pl; int vid; if (cbp->cb_type == ZFS_TYPE_VDEV) { if (strcmp(cbp->cb_vdevs.cb_names[0], "all-vdevs") == 0) { for_each_vdev(zhp, get_callback_vdev_cb, data); } else { /* Adjust column widths for vdev properties */ for (vid = 0; vid < cbp->cb_vdevs.cb_names_count; vid++) { vdev_expand_proplist(zhp, cbp->cb_vdevs.cb_names[vid], &cbp->cb_proplist); } /* Display the properties */ for (vid = 0; vid < cbp->cb_vdevs.cb_names_count; vid++) { get_callback_vdev(zhp, cbp->cb_vdevs.cb_names[vid], data); } } } else { assert(cbp->cb_type == ZFS_TYPE_POOL); for (pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) { /* * Skip the special fake placeholder. This will also * skip over the name property when 'all' is specified. */ if (pl->pl_prop == ZPOOL_PROP_NAME && pl == cbp->cb_proplist) continue; if (pl->pl_prop == ZPROP_INVAL && zfs_prop_user(pl->pl_user_prop)) { srctype = ZPROP_SRC_LOCAL; if (zpool_get_userprop(zhp, pl->pl_user_prop, value, sizeof (value), &srctype) != 0) continue; zprop_print_one_property(zpool_get_name(zhp), cbp, pl->pl_user_prop, value, srctype, NULL, NULL); } else if (pl->pl_prop == ZPROP_INVAL && (zpool_prop_feature(pl->pl_user_prop) || zpool_prop_unsupported(pl->pl_user_prop))) { srctype = ZPROP_SRC_LOCAL; if (zpool_prop_get_feature(zhp, pl->pl_user_prop, value, sizeof (value)) == 0) { zprop_print_one_property( zpool_get_name(zhp), cbp, pl->pl_user_prop, value, srctype, NULL, NULL); } } else { if (zpool_get_prop(zhp, pl->pl_prop, value, sizeof (value), &srctype, cbp->cb_literal) != 0) continue; zprop_print_one_property(zpool_get_name(zhp), cbp, zpool_prop_to_name(pl->pl_prop), value, srctype, NULL, NULL); } } } return (0); } /* * zpool get [-Hp] [-o "all" | field[,...]] <"all" | property[,...]> ... * * -H Scripted mode. Don't display headers, and separate properties * by a single tab. * -o List of columns to display. Defaults to * "name,property,value,source". * -p Display values in parsable (exact) format. * * Get properties of pools in the system. Output space statistics * for each one as well as other attributes. */ int zpool_do_get(int argc, char **argv) { zprop_get_cbdata_t cb = { 0 }; zprop_list_t fake_name = { 0 }; int ret; int c, i; char *propstr = NULL; char *vdev = NULL; cb.cb_first = B_TRUE; /* * Set up default columns and sources. */ cb.cb_sources = ZPROP_SRC_ALL; cb.cb_columns[0] = GET_COL_NAME; cb.cb_columns[1] = GET_COL_PROPERTY; cb.cb_columns[2] = GET_COL_VALUE; cb.cb_columns[3] = GET_COL_SOURCE; cb.cb_type = ZFS_TYPE_POOL; cb.cb_vdevs.cb_name_flags |= VDEV_NAME_TYPE_ID; current_prop_type = cb.cb_type; /* check options */ while ((c = getopt(argc, argv, ":Hpo:")) != -1) { switch (c) { case 'p': cb.cb_literal = B_TRUE; break; case 'H': cb.cb_scripted = B_TRUE; break; case 'o': memset(&cb.cb_columns, 0, sizeof (cb.cb_columns)); i = 0; for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const col_opts[] = { "name", "property", "value", "source", "all" }; static const zfs_get_column_t col_cols[] = { GET_COL_NAME, GET_COL_PROPERTY, GET_COL_VALUE, GET_COL_SOURCE }; if (i == ZFS_GET_NCOLS - 1) { (void) fprintf(stderr, gettext("too " "many fields given to -o " "option\n")); usage(B_FALSE); } for (c = 0; c < ARRAY_SIZE(col_opts); ++c) if (strcmp(tok, col_opts[c]) == 0) goto found; (void) fprintf(stderr, gettext("invalid column name '%s'\n"), tok); usage(B_FALSE); found: if (c >= 4) { if (i > 0) { (void) fprintf(stderr, gettext("\"all\" conflicts " "with specific fields " "given to -o option\n")); usage(B_FALSE); } memcpy(cb.cb_columns, col_cols, sizeof (col_cols)); i = ZFS_GET_NCOLS - 1; } else cb.cb_columns[i++] = col_cols[c]; } break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; if (argc < 1) { (void) fprintf(stderr, gettext("missing property " "argument\n")); usage(B_FALSE); } /* Properties list is needed later by zprop_get_list() */ propstr = argv[0]; argc--; argv++; if (argc == 0) { /* No args, so just print the defaults. */ } else if (are_all_pools(argc, argv)) { /* All the args are pool names */ } else if (are_all_pools(1, argv)) { /* The first arg is a pool name */ if ((argc == 2 && strcmp(argv[1], "all-vdevs") == 0) || (argc == 2 && strcmp(argv[1], "root") == 0) || are_vdevs_in_pool(argc - 1, argv + 1, argv[0], &cb.cb_vdevs)) { if (strcmp(argv[1], "root") == 0) vdev = strdup("root-0"); else vdev = strdup(argv[1]); /* ... and the rest are vdev names */ cb.cb_vdevs.cb_names = &vdev; cb.cb_vdevs.cb_names_count = argc - 1; cb.cb_type = ZFS_TYPE_VDEV; argc = 1; /* One pool to process */ } else { fprintf(stderr, gettext("Expected a list of vdevs in" " \"%s\", but got:\n"), argv[0]); error_list_unresolved_vdevs(argc - 1, argv + 1, argv[0], &cb.cb_vdevs); fprintf(stderr, "\n"); usage(B_FALSE); return (1); } } else { /* * The first arg isn't the name of a valid pool. */ fprintf(stderr, gettext("Cannot get properties of %s: " "no such pool available.\n"), argv[0]); return (1); } if (zprop_get_list(g_zfs, propstr, &cb.cb_proplist, cb.cb_type) != 0) { /* Use correct list of valid properties (pool or vdev) */ current_prop_type = cb.cb_type; usage(B_FALSE); } if (cb.cb_proplist != NULL) { fake_name.pl_prop = ZPOOL_PROP_NAME; fake_name.pl_width = strlen(gettext("NAME")); fake_name.pl_next = cb.cb_proplist; cb.cb_proplist = &fake_name; } ret = for_each_pool(argc, argv, B_TRUE, &cb.cb_proplist, cb.cb_type, cb.cb_literal, get_callback, &cb); if (cb.cb_proplist == &fake_name) zprop_free_list(fake_name.pl_next); else zprop_free_list(cb.cb_proplist); if (vdev != NULL) free(vdev); return (ret); } typedef struct set_cbdata { char *cb_propname; char *cb_value; zfs_type_t cb_type; vdev_cbdata_t cb_vdevs; boolean_t cb_any_successful; } set_cbdata_t; static int set_pool_callback(zpool_handle_t *zhp, set_cbdata_t *cb) { int error; /* Check if we have out-of-bounds features */ if (strcmp(cb->cb_propname, ZPOOL_CONFIG_COMPATIBILITY) == 0) { boolean_t features[SPA_FEATURES]; if (zpool_do_load_compat(cb->cb_value, features) != ZPOOL_COMPATIBILITY_OK) return (-1); nvlist_t *enabled = zpool_get_features(zhp); spa_feature_t i; for (i = 0; i < SPA_FEATURES; i++) { const char *fguid = spa_feature_table[i].fi_guid; if (nvlist_exists(enabled, fguid) && !features[i]) break; } if (i < SPA_FEATURES) (void) fprintf(stderr, gettext("Warning: one or " "more features already enabled on pool '%s'\n" "are not present in this compatibility set.\n"), zpool_get_name(zhp)); } /* if we're setting a feature, check it's in compatibility set */ if (zpool_prop_feature(cb->cb_propname) && strcmp(cb->cb_value, ZFS_FEATURE_ENABLED) == 0) { char *fname = strchr(cb->cb_propname, '@') + 1; spa_feature_t f; if (zfeature_lookup_name(fname, &f) == 0) { char compat[ZFS_MAXPROPLEN]; if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat, ZFS_MAXPROPLEN, NULL, B_FALSE) != 0) compat[0] = '\0'; boolean_t features[SPA_FEATURES]; if (zpool_do_load_compat(compat, features) != ZPOOL_COMPATIBILITY_OK) { (void) fprintf(stderr, gettext("Error: " "cannot enable feature '%s' on pool '%s'\n" "because the pool's 'compatibility' " "property cannot be parsed.\n"), fname, zpool_get_name(zhp)); return (-1); } if (!features[f]) { (void) fprintf(stderr, gettext("Error: " "cannot enable feature '%s' on pool '%s'\n" "as it is not specified in this pool's " "current compatibility set.\n" "Consider setting 'compatibility' to a " "less restrictive set, or to 'off'.\n"), fname, zpool_get_name(zhp)); return (-1); } } } error = zpool_set_prop(zhp, cb->cb_propname, cb->cb_value); return (error); } static int set_callback(zpool_handle_t *zhp, void *data) { int error; set_cbdata_t *cb = (set_cbdata_t *)data; if (cb->cb_type == ZFS_TYPE_VDEV) { error = zpool_set_vdev_prop(zhp, *cb->cb_vdevs.cb_names, cb->cb_propname, cb->cb_value); } else { assert(cb->cb_type == ZFS_TYPE_POOL); error = set_pool_callback(zhp, cb); } cb->cb_any_successful = !error; return (error); } int zpool_do_set(int argc, char **argv) { set_cbdata_t cb = { 0 }; int error; char *vdev = NULL; current_prop_type = ZFS_TYPE_POOL; if (argc > 1 && argv[1][0] == '-') { (void) fprintf(stderr, gettext("invalid option '%c'\n"), argv[1][1]); usage(B_FALSE); } if (argc < 2) { (void) fprintf(stderr, gettext("missing property=value " "argument\n")); usage(B_FALSE); } if (argc < 3) { (void) fprintf(stderr, gettext("missing pool name\n")); usage(B_FALSE); } if (argc > 4) { (void) fprintf(stderr, gettext("too many pool names\n")); usage(B_FALSE); } cb.cb_propname = argv[1]; cb.cb_type = ZFS_TYPE_POOL; cb.cb_vdevs.cb_name_flags |= VDEV_NAME_TYPE_ID; cb.cb_value = strchr(cb.cb_propname, '='); if (cb.cb_value == NULL) { (void) fprintf(stderr, gettext("missing value in " "property=value argument\n")); usage(B_FALSE); } *(cb.cb_value) = '\0'; cb.cb_value++; argc -= 2; argv += 2; /* argv[0] is pool name */ if (!is_pool(argv[0])) { (void) fprintf(stderr, gettext("cannot open '%s': is not a pool\n"), argv[0]); return (EINVAL); } /* argv[1], when supplied, is vdev name */ if (argc == 2) { if (strcmp(argv[1], "root") == 0) vdev = strdup("root-0"); else vdev = strdup(argv[1]); if (!are_vdevs_in_pool(1, &vdev, argv[0], &cb.cb_vdevs)) { (void) fprintf(stderr, gettext( "cannot find '%s' in '%s': device not in pool\n"), vdev, argv[0]); free(vdev); return (EINVAL); } cb.cb_vdevs.cb_names = &vdev; cb.cb_vdevs.cb_names_count = 1; cb.cb_type = ZFS_TYPE_VDEV; } error = for_each_pool(1, argv, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE, set_callback, &cb); if (vdev != NULL) free(vdev); return (error); } /* Add up the total number of bytes left to initialize/trim across all vdevs */ static uint64_t vdev_activity_remaining(nvlist_t *nv, zpool_wait_activity_t activity) { uint64_t bytes_remaining; nvlist_t **child; uint_t c, children; vdev_stat_t *vs; assert(activity == ZPOOL_WAIT_INITIALIZE || activity == ZPOOL_WAIT_TRIM); verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &c) == 0); if (activity == ZPOOL_WAIT_INITIALIZE && vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE) bytes_remaining = vs->vs_initialize_bytes_est - vs->vs_initialize_bytes_done; else if (activity == ZPOOL_WAIT_TRIM && vs->vs_trim_state == VDEV_TRIM_ACTIVE) bytes_remaining = vs->vs_trim_bytes_est - vs->vs_trim_bytes_done; else bytes_remaining = 0; if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) children = 0; for (c = 0; c < children; c++) bytes_remaining += vdev_activity_remaining(child[c], activity); return (bytes_remaining); } /* Add up the total number of bytes left to rebuild across top-level vdevs */ static uint64_t vdev_activity_top_remaining(nvlist_t *nv) { uint64_t bytes_remaining = 0; nvlist_t **child; uint_t children; int error; if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) children = 0; for (uint_t c = 0; c < children; c++) { vdev_rebuild_stat_t *vrs; uint_t i; error = nvlist_lookup_uint64_array(child[c], ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i); if (error == 0) { if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) { bytes_remaining += (vrs->vrs_bytes_est - vrs->vrs_bytes_rebuilt); } } } return (bytes_remaining); } /* Whether any vdevs are 'spare' or 'replacing' vdevs */ static boolean_t vdev_any_spare_replacing(nvlist_t *nv) { nvlist_t **child; uint_t c, children; const char *vdev_type; (void) nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &vdev_type); if (strcmp(vdev_type, VDEV_TYPE_REPLACING) == 0 || strcmp(vdev_type, VDEV_TYPE_SPARE) == 0 || strcmp(vdev_type, VDEV_TYPE_DRAID_SPARE) == 0) { return (B_TRUE); } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) children = 0; for (c = 0; c < children; c++) { if (vdev_any_spare_replacing(child[c])) return (B_TRUE); } return (B_FALSE); } typedef struct wait_data { char *wd_poolname; boolean_t wd_scripted; boolean_t wd_exact; boolean_t wd_headers_once; boolean_t wd_should_exit; /* Which activities to wait for */ boolean_t wd_enabled[ZPOOL_WAIT_NUM_ACTIVITIES]; float wd_interval; pthread_cond_t wd_cv; pthread_mutex_t wd_mutex; } wait_data_t; /* * Print to stdout a single line, containing one column for each activity that * we are waiting for specifying how many bytes of work are left for that * activity. */ static void print_wait_status_row(wait_data_t *wd, zpool_handle_t *zhp, int row) { nvlist_t *config, *nvroot; uint_t c; int i; pool_checkpoint_stat_t *pcs = NULL; pool_scan_stat_t *pss = NULL; pool_removal_stat_t *prs = NULL; pool_raidz_expand_stat_t *pres = NULL; const char *const headers[] = {"DISCARD", "FREE", "INITIALIZE", "REPLACE", "REMOVE", "RESILVER", "SCRUB", "TRIM", "RAIDZ_EXPAND"}; int col_widths[ZPOOL_WAIT_NUM_ACTIVITIES]; /* Calculate the width of each column */ for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) { /* * Make sure we have enough space in the col for pretty-printed * numbers and for the column header, and then leave a couple * spaces between cols for readability. */ col_widths[i] = MAX(strlen(headers[i]), 6) + 2; } if (timestamp_fmt != NODATE) print_timestamp(timestamp_fmt); /* Print header if appropriate */ int term_height = terminal_height(); boolean_t reprint_header = (!wd->wd_headers_once && term_height > 0 && row % (term_height-1) == 0); if (!wd->wd_scripted && (row == 0 || reprint_header)) { for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) { if (wd->wd_enabled[i]) (void) printf("%*s", col_widths[i], headers[i]); } (void) fputc('\n', stdout); } /* Bytes of work remaining in each activity */ int64_t bytes_rem[ZPOOL_WAIT_NUM_ACTIVITIES] = {0}; bytes_rem[ZPOOL_WAIT_FREE] = zpool_get_prop_int(zhp, ZPOOL_PROP_FREEING, NULL); config = zpool_get_config(zhp, NULL); nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c); if (pcs != NULL && pcs->pcs_state == CS_CHECKPOINT_DISCARDING) bytes_rem[ZPOOL_WAIT_CKPT_DISCARD] = pcs->pcs_space; (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c); if (prs != NULL && prs->prs_state == DSS_SCANNING) bytes_rem[ZPOOL_WAIT_REMOVE] = prs->prs_to_copy - prs->prs_copied; (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&pss, &c); if (pss != NULL && pss->pss_state == DSS_SCANNING && pss->pss_pass_scrub_pause == 0) { int64_t rem = pss->pss_to_examine - pss->pss_issued; if (pss->pss_func == POOL_SCAN_SCRUB) bytes_rem[ZPOOL_WAIT_SCRUB] = rem; else bytes_rem[ZPOOL_WAIT_RESILVER] = rem; } else if (check_rebuilding(nvroot, NULL)) { bytes_rem[ZPOOL_WAIT_RESILVER] = vdev_activity_top_remaining(nvroot); } (void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_RAIDZ_EXPAND_STATS, (uint64_t **)&pres, &c); if (pres != NULL && pres->pres_state == DSS_SCANNING) { int64_t rem = pres->pres_to_reflow - pres->pres_reflowed; bytes_rem[ZPOOL_WAIT_RAIDZ_EXPAND] = rem; } bytes_rem[ZPOOL_WAIT_INITIALIZE] = vdev_activity_remaining(nvroot, ZPOOL_WAIT_INITIALIZE); bytes_rem[ZPOOL_WAIT_TRIM] = vdev_activity_remaining(nvroot, ZPOOL_WAIT_TRIM); /* * A replace finishes after resilvering finishes, so the amount of work * left for a replace is the same as for resilvering. * * It isn't quite correct to say that if we have any 'spare' or * 'replacing' vdevs and a resilver is happening, then a replace is in * progress, like we do here. When a hot spare is used, the faulted vdev * is not removed after the hot spare is resilvered, so parent 'spare' * vdev is not removed either. So we could have a 'spare' vdev, but be * resilvering for a different reason. However, we use it as a heuristic * because we don't have access to the DTLs, which could tell us whether * or not we have really finished resilvering a hot spare. */ if (vdev_any_spare_replacing(nvroot)) bytes_rem[ZPOOL_WAIT_REPLACE] = bytes_rem[ZPOOL_WAIT_RESILVER]; for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) { char buf[64]; if (!wd->wd_enabled[i]) continue; if (wd->wd_exact) { (void) snprintf(buf, sizeof (buf), "%" PRIi64, bytes_rem[i]); } else { zfs_nicenum(bytes_rem[i], buf, sizeof (buf)); } if (wd->wd_scripted) (void) printf(i == 0 ? "%s" : "\t%s", buf); else (void) printf(" %*s", col_widths[i] - 1, buf); } (void) printf("\n"); (void) fflush(stdout); } static void * wait_status_thread(void *arg) { wait_data_t *wd = (wait_data_t *)arg; zpool_handle_t *zhp; if ((zhp = zpool_open(g_zfs, wd->wd_poolname)) == NULL) return (void *)(1); for (int row = 0; ; row++) { boolean_t missing; struct timespec timeout; int ret = 0; (void) clock_gettime(CLOCK_REALTIME, &timeout); if (zpool_refresh_stats(zhp, &missing) != 0 || missing || zpool_props_refresh(zhp) != 0) { zpool_close(zhp); return (void *)(uintptr_t)(missing ? 0 : 1); } print_wait_status_row(wd, zhp, row); timeout.tv_sec += floor(wd->wd_interval); long nanos = timeout.tv_nsec + (wd->wd_interval - floor(wd->wd_interval)) * NANOSEC; if (nanos >= NANOSEC) { timeout.tv_sec++; timeout.tv_nsec = nanos - NANOSEC; } else { timeout.tv_nsec = nanos; } pthread_mutex_lock(&wd->wd_mutex); if (!wd->wd_should_exit) ret = pthread_cond_timedwait(&wd->wd_cv, &wd->wd_mutex, &timeout); pthread_mutex_unlock(&wd->wd_mutex); if (ret == 0) { break; /* signaled by main thread */ } else if (ret != ETIMEDOUT) { (void) fprintf(stderr, gettext("pthread_cond_timedwait " "failed: %s\n"), strerror(ret)); zpool_close(zhp); return (void *)(uintptr_t)(1); } } zpool_close(zhp); return (void *)(0); } int zpool_do_wait(int argc, char **argv) { boolean_t verbose = B_FALSE; int c, i; unsigned long count; pthread_t status_thr; int error = 0; zpool_handle_t *zhp; wait_data_t wd; wd.wd_scripted = B_FALSE; wd.wd_exact = B_FALSE; wd.wd_headers_once = B_FALSE; wd.wd_should_exit = B_FALSE; pthread_mutex_init(&wd.wd_mutex, NULL); pthread_cond_init(&wd.wd_cv, NULL); /* By default, wait for all types of activity. */ for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) wd.wd_enabled[i] = B_TRUE; while ((c = getopt(argc, argv, "HpT:t:")) != -1) { switch (c) { case 'H': wd.wd_scripted = B_TRUE; break; case 'n': wd.wd_headers_once = B_TRUE; break; case 'p': wd.wd_exact = B_TRUE; break; case 'T': get_timestamp_arg(*optarg); break; case 't': /* Reset activities array */ memset(&wd.wd_enabled, 0, sizeof (wd.wd_enabled)); for (char *tok; (tok = strsep(&optarg, ",")); ) { static const char *const col_opts[] = { "discard", "free", "initialize", "replace", "remove", "resilver", "scrub", "trim", "raidz_expand" }; for (i = 0; i < ARRAY_SIZE(col_opts); ++i) if (strcmp(tok, col_opts[i]) == 0) { wd.wd_enabled[i] = B_TRUE; goto found; } (void) fprintf(stderr, gettext("invalid activity '%s'\n"), tok); usage(B_FALSE); found:; } break; case '?': (void) fprintf(stderr, gettext("invalid option '%c'\n"), optopt); usage(B_FALSE); } } argc -= optind; argv += optind; get_interval_count(&argc, argv, &wd.wd_interval, &count); if (count != 0) { /* This subcmd only accepts an interval, not a count */ (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } if (wd.wd_interval != 0) verbose = B_TRUE; if (argc < 1) { (void) fprintf(stderr, gettext("missing 'pool' argument\n")); usage(B_FALSE); } if (argc > 1) { (void) fprintf(stderr, gettext("too many arguments\n")); usage(B_FALSE); } wd.wd_poolname = argv[0]; if ((zhp = zpool_open(g_zfs, wd.wd_poolname)) == NULL) return (1); if (verbose) { /* * We use a separate thread for printing status updates because * the main thread will call lzc_wait(), which blocks as long * as an activity is in progress, which can be a long time. */ if (pthread_create(&status_thr, NULL, wait_status_thread, &wd) != 0) { (void) fprintf(stderr, gettext("failed to create status" "thread: %s\n"), strerror(errno)); zpool_close(zhp); return (1); } } /* * Loop over all activities that we are supposed to wait for until none * of them are in progress. Note that this means we can end up waiting * for more activities to complete than just those that were in progress * when we began waiting; if an activity we are interested in begins * while we are waiting for another activity, we will wait for both to * complete before exiting. */ for (;;) { boolean_t missing = B_FALSE; boolean_t any_waited = B_FALSE; for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) { boolean_t waited; if (!wd.wd_enabled[i]) continue; error = zpool_wait_status(zhp, i, &missing, &waited); if (error != 0 || missing) break; any_waited = (any_waited || waited); } if (error != 0 || missing || !any_waited) break; } zpool_close(zhp); if (verbose) { uintptr_t status; pthread_mutex_lock(&wd.wd_mutex); wd.wd_should_exit = B_TRUE; pthread_cond_signal(&wd.wd_cv); pthread_mutex_unlock(&wd.wd_mutex); (void) pthread_join(status_thr, (void *)&status); if (status != 0) error = status; } pthread_mutex_destroy(&wd.wd_mutex); pthread_cond_destroy(&wd.wd_cv); return (error); } static int find_command_idx(const char *command, int *idx) { for (int i = 0; i < NCOMMAND; ++i) { if (command_table[i].name == NULL) continue; if (strcmp(command, command_table[i].name) == 0) { *idx = i; return (0); } } return (1); } /* * Display version message */ static int zpool_do_version(int argc, char **argv) { (void) argc, (void) argv; return (zfs_version_print() != 0); } /* Display documentation */ static int zpool_do_help(int argc, char **argv) { char page[MAXNAMELEN]; if (argc < 3 || strcmp(argv[2], "zpool") == 0) strcpy(page, "zpool"); else if (strcmp(argv[2], "concepts") == 0 || strcmp(argv[2], "props") == 0) snprintf(page, sizeof (page), "zpool%s", argv[2]); else snprintf(page, sizeof (page), "zpool-%s", argv[2]); execlp("man", "man", page, NULL); fprintf(stderr, "couldn't run man program: %s", strerror(errno)); return (-1); } /* * Do zpool_load_compat() and print error message on failure */ static zpool_compat_status_t zpool_do_load_compat(const char *compat, boolean_t *list) { char report[1024]; zpool_compat_status_t ret; ret = zpool_load_compat(compat, list, report, 1024); switch (ret) { case ZPOOL_COMPATIBILITY_OK: break; case ZPOOL_COMPATIBILITY_NOFILES: case ZPOOL_COMPATIBILITY_BADFILE: case ZPOOL_COMPATIBILITY_BADTOKEN: (void) fprintf(stderr, "Error: %s\n", report); break; case ZPOOL_COMPATIBILITY_WARNTOKEN: (void) fprintf(stderr, "Warning: %s\n", report); ret = ZPOOL_COMPATIBILITY_OK; break; } return (ret); } int main(int argc, char **argv) { int ret = 0; int i = 0; char *cmdname; char **newargv; (void) setlocale(LC_ALL, ""); (void) setlocale(LC_NUMERIC, "C"); (void) textdomain(TEXT_DOMAIN); srand(time(NULL)); opterr = 0; /* * Make sure the user has specified some command. */ if (argc < 2) { (void) fprintf(stderr, gettext("missing command\n")); usage(B_FALSE); } cmdname = argv[1]; /* * Special case '-?' */ if ((strcmp(cmdname, "-?") == 0) || strcmp(cmdname, "--help") == 0) usage(B_TRUE); /* * Special case '-V|--version' */ if ((strcmp(cmdname, "-V") == 0) || (strcmp(cmdname, "--version") == 0)) return (zpool_do_version(argc, argv)); /* * Special case 'help' */ if (strcmp(cmdname, "help") == 0) return (zpool_do_help(argc, argv)); if ((g_zfs = libzfs_init()) == NULL) { (void) fprintf(stderr, "%s\n", libzfs_error_init(errno)); return (1); } libzfs_print_on_error(g_zfs, B_TRUE); zfs_save_arguments(argc, argv, history_str, sizeof (history_str)); /* * Many commands modify input strings for string parsing reasons. * We create a copy to protect the original argv. */ newargv = safe_malloc((argc + 1) * sizeof (newargv[0])); for (i = 0; i < argc; i++) newargv[i] = strdup(argv[i]); newargv[argc] = NULL; /* * Run the appropriate command. */ if (find_command_idx(cmdname, &i) == 0) { current_command = &command_table[i]; ret = command_table[i].func(argc - 1, newargv + 1); } else if (strchr(cmdname, '=')) { verify(find_command_idx("set", &i) == 0); current_command = &command_table[i]; ret = command_table[i].func(argc, newargv); } else if (strcmp(cmdname, "freeze") == 0 && argc == 3) { /* * 'freeze' is a vile debugging abomination, so we treat * it as such. */ zfs_cmd_t zc = {"\0"}; (void) strlcpy(zc.zc_name, argv[2], sizeof (zc.zc_name)); ret = zfs_ioctl(g_zfs, ZFS_IOC_POOL_FREEZE, &zc); if (ret != 0) { (void) fprintf(stderr, gettext("failed to freeze pool: %d\n"), errno); ret = 1; } log_history = 0; } else { (void) fprintf(stderr, gettext("unrecognized " "command '%s'\n"), cmdname); usage(B_FALSE); ret = 1; } for (i = 0; i < argc; i++) free(newargv[i]); free(newargv); if (ret == 0 && log_history) (void) zpool_log_history(g_zfs, history_str); libzfs_fini(g_zfs); /* * The 'ZFS_ABORT' environment variable causes us to dump core on exit * for the purposes of running ::findleaks. */ if (getenv("ZFS_ABORT") != NULL) { (void) printf("dumping core by request\n"); abort(); } return (ret); } diff --git a/sys/contrib/openzfs/include/libzutil.h b/sys/contrib/openzfs/include/libzutil.h index d9a9a65753dd..e2108ceeaa44 100644 --- a/sys/contrib/openzfs/include/libzutil.h +++ b/sys/contrib/openzfs/include/libzutil.h @@ -1,284 +1,286 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2018 by Delphix. All rights reserved. + * Copyright (c) 2018, 2024 by Delphix. All rights reserved. */ #ifndef _LIBZUTIL_H #define _LIBZUTIL_H extern __attribute__((visibility("default"))) #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Default wait time in milliseconds for a device name to be created. */ #define DISK_LABEL_WAIT (30 * 1000) /* 30 seconds */ /* * Pool Config Operations * * These are specific to the library libzfs or libzpool instance. */ typedef nvlist_t *refresh_config_func_t(void *, nvlist_t *); typedef int pool_active_func_t(void *, const char *, uint64_t, boolean_t *); typedef const struct pool_config_ops { refresh_config_func_t *pco_refresh_config; pool_active_func_t *pco_pool_active; } pool_config_ops_t; /* * An instance of pool_config_ops_t is expected in the caller's binary. */ _LIBZUTIL_H pool_config_ops_t libzfs_config_ops; _LIBZUTIL_H pool_config_ops_t libzpool_config_ops; typedef enum lpc_error { LPC_SUCCESS = 0, /* no error -- success */ LPC_BADCACHE = 2000, /* out of memory */ LPC_BADPATH, /* must be an absolute path */ LPC_NOMEM, /* out of memory */ LPC_EACCESS, /* some devices require root privileges */ LPC_UNKNOWN } lpc_error_t; typedef struct importargs { char **path; /* a list of paths to search */ int paths; /* number of paths to search */ const char *poolname; /* name of a pool to find */ uint64_t guid; /* guid of a pool to find */ const char *cachefile; /* cachefile to use for import */ boolean_t can_be_active; /* can the pool be active? */ boolean_t scan; /* prefer scanning to libblkid cache */ nvlist_t *policy; /* load policy (max txg, rewind, etc.) */ + boolean_t do_destroyed; + boolean_t do_all; } importargs_t; typedef struct libpc_handle { int lpc_error; boolean_t lpc_printerr; boolean_t lpc_open_access_error; boolean_t lpc_desc_active; char lpc_desc[1024]; pool_config_ops_t *lpc_ops; void *lpc_lib_handle; } libpc_handle_t; _LIBZUTIL_H const char *libpc_error_description(libpc_handle_t *); _LIBZUTIL_H nvlist_t *zpool_search_import(libpc_handle_t *, importargs_t *); _LIBZUTIL_H int zpool_find_config(libpc_handle_t *, const char *, nvlist_t **, importargs_t *); _LIBZUTIL_H const char * const * zpool_default_search_paths(size_t *count); _LIBZUTIL_H int zpool_read_label(int, nvlist_t **, int *); _LIBZUTIL_H int zpool_label_disk_wait(const char *, int); _LIBZUTIL_H int zpool_disk_wait(const char *); struct udev_device; _LIBZUTIL_H int zfs_device_get_devid(struct udev_device *, char *, size_t); _LIBZUTIL_H int zfs_device_get_physical(struct udev_device *, char *, size_t); _LIBZUTIL_H void update_vdev_config_dev_strs(nvlist_t *); /* * Default device paths */ #define DISK_ROOT "/dev" #define UDISK_ROOT "/dev/disk" #define ZVOL_ROOT "/dev/zvol" _LIBZUTIL_H int zfs_append_partition(char *path, size_t max_len); _LIBZUTIL_H int zfs_resolve_shortname(const char *name, char *path, size_t pathlen); _LIBZUTIL_H char *zfs_strip_partition(const char *); _LIBZUTIL_H const char *zfs_strip_path(const char *); _LIBZUTIL_H int zfs_strcmp_pathname(const char *, const char *, int); _LIBZUTIL_H boolean_t zfs_dev_is_dm(const char *); _LIBZUTIL_H boolean_t zfs_dev_is_whole_disk(const char *); _LIBZUTIL_H int zfs_dev_flush(int); _LIBZUTIL_H char *zfs_get_underlying_path(const char *); _LIBZUTIL_H char *zfs_get_enclosure_sysfs_path(const char *); _LIBZUTIL_H boolean_t is_mpath_whole_disk(const char *); _LIBZUTIL_H boolean_t zfs_isnumber(const char *); /* * Formats for iostat numbers. Examples: "12K", "30ms", "4B", "2321234", "-". * * ZFS_NICENUM_1024: Print kilo, mega, tera, peta, exa.. * ZFS_NICENUM_BYTES: Print single bytes ("13B"), kilo, mega, tera... * ZFS_NICENUM_TIME: Print nanosecs, microsecs, millisecs, seconds... * ZFS_NICENUM_RAW: Print the raw number without any formatting * ZFS_NICENUM_RAWTIME: Same as RAW, but print dashes ('-') for zero. */ enum zfs_nicenum_format { ZFS_NICENUM_1024 = 0, ZFS_NICENUM_BYTES = 1, ZFS_NICENUM_TIME = 2, ZFS_NICENUM_RAW = 3, ZFS_NICENUM_RAWTIME = 4 }; /* * Convert a number to a human-readable form. */ _LIBZUTIL_H void zfs_nicebytes(uint64_t, char *, size_t); _LIBZUTIL_H void zfs_nicenum(uint64_t, char *, size_t); _LIBZUTIL_H void zfs_nicenum_format(uint64_t, char *, size_t, enum zfs_nicenum_format); _LIBZUTIL_H void zfs_nicetime(uint64_t, char *, size_t); _LIBZUTIL_H void zfs_niceraw(uint64_t, char *, size_t); #define nicenum(num, buf, size) zfs_nicenum(num, buf, size) _LIBZUTIL_H void zpool_dump_ddt(const ddt_stat_t *, const ddt_histogram_t *); _LIBZUTIL_H int zpool_history_unpack(char *, uint64_t, uint64_t *, nvlist_t ***, uint_t *); _LIBZUTIL_H void fsleep(float sec); _LIBZUTIL_H int zpool_getenv_int(const char *env, int default_val); struct zfs_cmd; /* * List of colors to use */ #define ANSI_BLACK "\033[0;30m" #define ANSI_RED "\033[0;31m" #define ANSI_GREEN "\033[0;32m" #define ANSI_YELLOW "\033[0;33m" #define ANSI_BLUE "\033[0;34m" #define ANSI_BOLD_BLUE "\033[1;34m" /* light blue */ #define ANSI_MAGENTA "\033[0;35m" #define ANSI_CYAN "\033[0;36m" #define ANSI_GRAY "\033[0;37m" #define ANSI_RESET "\033[0m" #define ANSI_BOLD "\033[1m" _LIBZUTIL_H int use_color(void); _LIBZUTIL_H void color_start(const char *color); _LIBZUTIL_H void color_end(void); _LIBZUTIL_H int printf_color(const char *color, const char *format, ...); _LIBZUTIL_H const char *zfs_basename(const char *path); _LIBZUTIL_H ssize_t zfs_dirnamelen(const char *path); #ifdef __linux__ extern char **environ; _LIBZUTIL_H void zfs_setproctitle_init(int argc, char *argv[], char *envp[]); _LIBZUTIL_H void zfs_setproctitle(const char *fmt, ...); #else #define zfs_setproctitle(fmt, ...) setproctitle(fmt, ##__VA_ARGS__) #define zfs_setproctitle_init(x, y, z) ((void)0) #endif /* * These functions are used by the ZFS libraries and cmd/zpool code, but are * not exported in the ABI. */ typedef int (*pool_vdev_iter_f)(void *, nvlist_t *, void *); int for_each_vdev_cb(void *zhp, nvlist_t *nv, pool_vdev_iter_f func, void *data); int for_each_vdev_macro_helper_func(void *zhp_data, nvlist_t *nv, void *data); int for_each_real_leaf_vdev_macro_helper_func(void *zhp_data, nvlist_t *nv, void *data); /* * Often you'll want to iterate over all the vdevs in the pool, but don't want * to use for_each_vdev() since it requires a callback function. * * Instead you can use FOR_EACH_VDEV(): * * zpool_handle_t *zhp // Assume this is initialized * nvlist_t *nv * ... * FOR_EACH_VDEV(zhp, nv) { * const char *path = NULL; * nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path); * printf("Looking at vdev %s\n", path); * } * * Note: FOR_EACH_VDEV runs in O(n^2) time where n = number of vdevs. However, * there's an upper limit of 256 vdevs per dRAID top-level vdevs (TLDs), 255 for * raidz2 TLDs, a real world limit of ~500 vdevs for mirrors, so this shouldn't * really be an issue. * * Here are some micro-benchmarks of a complete FOR_EACH_VDEV loop on a RAID0 * pool: * * 100 vdevs = 0.7ms * 500 vdevs = 17ms * 750 vdevs = 40ms * 1000 vdevs = 82ms * * The '__nv += 0' at the end of the for() loop gets around a "comma or * semicolon followed by non-blank" checkstyle error. Note on most compliers * the '__nv += 0' can just be replaced with 'NULL', but gcc on Centos 7 * will give a 'warning: statement with no effect' error if you do that. */ #define __FOR_EACH_VDEV(__zhp, __nv, __func) { \ __nv = zpool_get_config(__zhp, NULL); \ VERIFY0(nvlist_lookup_nvlist(__nv, ZPOOL_CONFIG_VDEV_TREE, &__nv)); \ } \ for (nvlist_t *__root_nv = __nv, *__state = (nvlist_t *)0; \ for_each_vdev_cb(&__state, __root_nv, __func, &__nv) == 1; \ __nv += 0) #define FOR_EACH_VDEV(__zhp, __nv) \ __FOR_EACH_VDEV(__zhp, __nv, for_each_vdev_macro_helper_func) /* * "real leaf" vdevs are leaf vdevs that are real devices (disks or files). * This excludes leaf vdevs like like draid spares. */ #define FOR_EACH_REAL_LEAF_VDEV(__zhp, __nv) \ __FOR_EACH_VDEV(__zhp, __nv, for_each_real_leaf_vdev_macro_helper_func) int for_each_vdev_in_nvlist(nvlist_t *nvroot, pool_vdev_iter_f func, void *data); void update_vdevs_config_dev_sysfs_path(nvlist_t *config); _LIBZUTIL_H void update_vdev_config_dev_sysfs_path(nvlist_t *nv, const char *path, const char *key); /* * Thread-safe strerror() for use in ZFS libraries */ static inline char *zfs_strerror(int errnum) { return (strerror_l(errnum, uselocale(0))); } #ifdef __cplusplus } #endif #endif /* _LIBZUTIL_H */ diff --git a/sys/contrib/openzfs/include/sys/spa.h b/sys/contrib/openzfs/include/sys/spa.h index b969f05afe48..ca15025ba33c 100644 --- a/sys/contrib/openzfs/include/sys/spa.h +++ b/sys/contrib/openzfs/include/sys/spa.h @@ -1,1250 +1,1252 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2024 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2017 Joyent, Inc. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2019, Allan Jude * Copyright (c) 2019, Klara Inc. * Copyright (c) 2019, Datto Inc. */ #ifndef _SYS_SPA_H #define _SYS_SPA_H #include #include #include #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Forward references that lots of things need. */ typedef struct spa spa_t; typedef struct vdev vdev_t; typedef struct metaslab metaslab_t; typedef struct metaslab_group metaslab_group_t; typedef struct metaslab_class metaslab_class_t; typedef struct zio zio_t; typedef struct zilog zilog_t; typedef struct spa_aux_vdev spa_aux_vdev_t; typedef struct zbookmark_phys zbookmark_phys_t; typedef struct zbookmark_err_phys zbookmark_err_phys_t; struct bpobj; struct bplist; struct dsl_pool; struct dsl_dataset; struct dsl_crypto_params; /* * Alignment Shift (ashift) is an immutable, internal top-level vdev property * which can only be set at vdev creation time. Physical writes are always done * according to it, which makes 2^ashift the smallest possible IO on a vdev. * * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB * (2^16 = 65,536). */ #define ASHIFT_MIN 9 #define ASHIFT_MAX 16 /* * Size of block to hold the configuration data (a packed nvlist) */ #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) /* * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. * The ASIZE encoding should be at least 64 times larger (6 more bits) * to support up to 4-way RAID-Z mirror mode with worst-case gang block * overhead, three DVAs per bp, plus one more bit in case we do anything * else that expands the ASIZE. */ #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ #define SPA_COMPRESSBITS 7 #define SPA_VDEVBITS 24 #define SPA_COMPRESSMASK ((1U << SPA_COMPRESSBITS) - 1) /* * All SPA data is represented by 128-bit data virtual addresses (DVAs). * The members of the dva_t should be considered opaque outside the SPA. */ typedef struct dva { uint64_t dva_word[2]; } dva_t; /* * Some checksums/hashes need a 256-bit initialization salt. This salt is kept * secret and is suitable for use in MAC algorithms as the key. */ typedef struct zio_cksum_salt { uint8_t zcs_bytes[32]; } zio_cksum_salt_t; /* * Each block is described by its DVAs, time of birth, checksum, etc. * The word-by-word, bit-by-bit layout of the blkptr is as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | pad | vdev1 | pad | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 1 |G| offset1 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 2 | pad | vdev2 | pad | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 3 |G| offset2 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 4 | pad | vdev3 | pad | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 5 |G| offset3 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 8 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 9 | physical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | fill count | * +-------+-------+-------+-------+-------+-------+-------+-------+ * c | checksum[0] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * d | checksum[1] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * e | checksum[2] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * f | checksum[3] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * vdev virtual device ID * offset offset into virtual device * LSIZE logical size * PSIZE physical size (after compression) * ASIZE allocated size (including RAID-Z parity and gang block headers) * cksum checksum function * comp compression function * G gang block indicator * B byteorder (endianness) * D dedup * X encryption * E blkptr_t contains embedded data (see below) * lvl level of indirection * type DMU object type * phys birth txg when dva[0] was written; zero if same as logical birth txg * note that typically all the dva's would be written in this * txg, but they could be different if they were moved by * device removal. * log. birth transaction group in which the block was logically born * fill count number of non-zero blocks under this bp * checksum[4] 256-bit checksum of the data this bp describes */ /* * The blkptr_t's of encrypted blocks also need to store the encryption * parameters so that the block can be decrypted. This layout is as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | vdev1 | pad | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 1 |G| offset1 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 2 | vdev2 | pad | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 3 |G| offset2 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 4 | salt | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 5 | IV1 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 8 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 9 | physical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | IV2 | fill count | * +-------+-------+-------+-------+-------+-------+-------+-------+ * c | checksum[0] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * d | checksum[1] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * e | MAC[0] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * f | MAC[1] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * salt Salt for generating encryption keys * IV1 First 64 bits of encryption IV * X Block requires encryption handling (set to 1) * E blkptr_t contains embedded data (set to 0, see below) * fill count number of non-zero blocks under this bp (truncated to 32 bits) * IV2 Last 32 bits of encryption IV * checksum[2] 128-bit checksum of the data this bp describes * MAC[2] 128-bit message authentication code for this data * * The X bit being set indicates that this block is one of 3 types. If this is * a level 0 block with an encrypted object type, the block is encrypted * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted * object type, this block is authenticated with an HMAC (see * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC * words to store a checksum-of-MACs from the level below (see * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED() * refers to both encrypted and authenticated blocks and BP_USES_CRYPT() * refers to any of these 3 kinds of blocks. * * The additional encryption parameters are the salt, IV, and MAC which are * explained in greater detail in the block comment at the top of zio_crypt.c. * The MAC occupies half of the checksum space since it serves a very similar * purpose: to prevent data corruption on disk. The only functional difference * is that the checksum is used to detect on-disk corruption whether or not the * encryption key is loaded and the MAC provides additional protection against * malicious disk tampering. We use the 3rd DVA to store the salt and first * 64 bits of the IV. As a result encrypted blocks can only have 2 copies * maximum instead of the normal 3. The last 32 bits of the IV are stored in * the upper bits of what is usually the fill count. Note that only blocks at * level 0 or -2 are ever encrypted, which allows us to guarantee that these * 32 bits are not trampled over by other code (see zio_crypt.c for details). * The salt and IV are not used for authenticated bps or bps with an indirect * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits * for the fill count. */ /* * "Embedded" blkptr_t's don't actually point to a block, instead they * have a data payload embedded in the blkptr_t itself. See the comment * in blkptr.c for more details. * * The blkptr_t is laid out as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | payload | * 1 | payload | * 2 | payload | * 3 | payload | * 4 | payload | * 5 | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | payload | * 8 | payload | * 9 | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | payload | * c | payload | * d | payload | * e | payload | * f | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * payload contains the embedded data * B (byteorder) byteorder (endianness) * D (dedup) padding (set to zero) * X encryption (set to zero) * E (embedded) set to one * lvl indirection level * type DMU object type * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) * comp compression function of payload * PSIZE size of payload after compression, in bytes * LSIZE logical size of payload, in bytes * note that 25 bits is enough to store the largest * "normal" BP's LSIZE (2^16 * 2^9) in bytes * log. birth transaction group in which the block was logically born * * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded * bp's they are stored in units of SPA_MINBLOCKSHIFT. * Generally, the generic BP_GET_*() macros can be used on embedded BP's. * The B, D, X, lvl, type, and comp fields are stored the same as with normal * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before * other macros, as they assert that they are only used on BP's of the correct * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use * the payload space for encryption parameters (see the comment above on * how encryption parameters are stored). */ #define BPE_GET_ETYPE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET((bp)->blk_prop, 40, 8)) #define BPE_SET_ETYPE(bp, t) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET((bp)->blk_prop, 40, 8, t); \ } while (0) #define BPE_GET_LSIZE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) #define BPE_SET_LSIZE(bp, x) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ } while (0) #define BPE_GET_PSIZE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) #define BPE_SET_PSIZE(bp, x) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ } while (0) typedef enum bp_embedded_type { BP_EMBEDDED_TYPE_DATA, BP_EMBEDDED_TYPE_RESERVED, /* Reserved for Delphix byteswap feature. */ BP_EMBEDDED_TYPE_REDACTED, NUM_BP_EMBEDDED_TYPES } bp_embedded_type_t; #define BPE_NUM_WORDS 14 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) #define BPE_IS_PAYLOADWORD(bp, wp) \ ((wp) != &(bp)->blk_prop && (wp) != (&(bp)->blk_birth_word[1])) #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */ /* * A block is a hole when it has either 1) never been written to, or * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads * without physically allocating disk space. Holes are represented in the * blkptr_t structure by zeroed blk_dva. Correct checking for holes is * done through the BP_IS_HOLE macro. For holes, the logical size, level, * DMU object type, and birth times are all also stored for holes that * were written to at some point (i.e. were punched after having been filled). */ typedef struct blkptr { dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ uint64_t blk_prop; /* size, compression, type, etc */ uint64_t blk_pad[2]; /* Extra space for the future */ uint64_t blk_birth_word[2]; uint64_t blk_fill; /* fill count */ zio_cksum_t blk_cksum; /* 256-bit checksum */ } blkptr_t; /* * Macros to get and set fields in a bp or DVA. */ /* * Note, for gang blocks, DVA_GET_ASIZE() is the total space allocated for * this gang DVA including its children BP's. The space allocated at this * DVA's vdev/offset is vdev_gang_header_asize(vdev). */ #define DVA_GET_ASIZE(dva) \ BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) #define DVA_SET_ASIZE(dva, x) \ BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ SPA_MINBLOCKSHIFT, 0, x) #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS) #define DVA_SET_VDEV(dva, x) \ BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x) #define DVA_GET_OFFSET(dva) \ BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) #define DVA_SET_OFFSET(dva, x) \ BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) #define BP_GET_LSIZE(bp) \ (BP_IS_EMBEDDED(bp) ? \ (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) #define BP_SET_LSIZE(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, \ 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ } while (0) #define BP_GET_PSIZE(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) #define BP_SET_PSIZE(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, \ 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ } while (0) #define BP_GET_COMPRESS(bp) \ BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS) #define BP_SET_COMPRESS(bp, x) \ BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x) #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) #define BP_GET_CHECKSUM(bp) \ (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ BF64_GET((bp)->blk_prop, 40, 8)) #define BP_SET_CHECKSUM(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET((bp)->blk_prop, 40, 8, x); \ } while (0) #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) /* encrypted, authenticated, and MAC cksum bps use the same bit */ #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1) #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x) #define BP_IS_ENCRYPTED(bp) \ (BP_USES_CRYPT(bp) && \ BP_GET_LEVEL(bp) <= 0 && \ DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) #define BP_IS_AUTHENTICATED(bp) \ (BP_USES_CRYPT(bp) && \ BP_GET_LEVEL(bp) <= 0 && \ !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \ (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0) #define BP_IS_PROTECTED(bp) \ (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp)) #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) #define BP_GET_FREE(bp) BF64_GET((bp)->blk_fill, 0, 1) #define BP_SET_FREE(bp, x) BF64_SET((bp)->blk_fill, 0, 1, x) #define BP_GET_LOGICAL_BIRTH(bp) (bp)->blk_birth_word[1] #define BP_SET_LOGICAL_BIRTH(bp, x) ((bp)->blk_birth_word[1] = (x)) #define BP_GET_PHYSICAL_BIRTH(bp) (bp)->blk_birth_word[0] #define BP_SET_PHYSICAL_BIRTH(bp, x) ((bp)->blk_birth_word[0] = (x)) #define BP_GET_BIRTH(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ BP_GET_PHYSICAL_BIRTH(bp) ? BP_GET_PHYSICAL_BIRTH(bp) : \ BP_GET_LOGICAL_BIRTH(bp)) #define BP_SET_BIRTH(bp, logical, physical) \ { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BP_SET_LOGICAL_BIRTH(bp, logical); \ BP_SET_PHYSICAL_BIRTH(bp, \ ((logical) == (physical) ? 0 : (physical))); \ } #define BP_GET_FILL(bp) \ ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \ ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill)) #define BP_SET_FILL(bp, fill) \ { \ if (BP_IS_ENCRYPTED(bp)) \ BF64_SET((bp)->blk_fill, 0, 32, fill); \ else \ (bp)->blk_fill = fill; \ } #define BP_GET_IV2(bp) \ (ASSERT(BP_IS_ENCRYPTED(bp)), \ BF64_GET((bp)->blk_fill, 32, 32)) #define BP_SET_IV2(bp, iv2) \ { \ ASSERT(BP_IS_ENCRYPTED(bp)); \ BF64_SET((bp)->blk_fill, 32, 32, iv2); \ } #define BP_IS_METADATA(bp) \ (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) #define BP_GET_ASIZE(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) #define BP_GET_UCSIZE(bp) \ (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) #define BP_GET_NDVAS(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) #define BP_COUNT_GANG(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ DVA_GET_GANG(&(bp)->blk_dva[1]) + \ (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))) #define DVA_EQUAL(dva1, dva2) \ ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ (dva1)->dva_word[0] == (dva2)->dva_word[0]) #define BP_EQUAL(bp1, bp2) \ (BP_GET_BIRTH(bp1) == BP_GET_BIRTH(bp2) && \ BP_GET_LOGICAL_BIRTH(bp1) == BP_GET_LOGICAL_BIRTH(bp2) && \ DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) #define BP_IS_GANG(bp) \ (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ (dva)->dva_word[1] == 0ULL) #define BP_IS_HOLE(bp) \ (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) #define BP_SET_REDACTED(bp) \ { \ BP_SET_EMBEDDED(bp, B_TRUE); \ BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \ } #define BP_IS_REDACTED(bp) \ (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED) /* BP_IS_RAIDZ(bp) assumes no block compression */ #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ BP_GET_PSIZE(bp)) #define BP_ZERO(bp) \ { \ (bp)->blk_dva[0].dva_word[0] = 0; \ (bp)->blk_dva[0].dva_word[1] = 0; \ (bp)->blk_dva[1].dva_word[0] = 0; \ (bp)->blk_dva[1].dva_word[1] = 0; \ (bp)->blk_dva[2].dva_word[0] = 0; \ (bp)->blk_dva[2].dva_word[1] = 0; \ (bp)->blk_prop = 0; \ (bp)->blk_pad[0] = 0; \ (bp)->blk_pad[1] = 0; \ (bp)->blk_birth_word[0] = 0; \ (bp)->blk_birth_word[1] = 0; \ (bp)->blk_fill = 0; \ ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ } #ifdef _ZFS_BIG_ENDIAN #define ZFS_HOST_BYTEORDER (0ULL) #else #define ZFS_HOST_BYTEORDER (1ULL) #endif #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) #define BP_SPRINTF_LEN 400 /* * This macro allows code sharing between zfs, libzpool, and mdb. * 'func' is either kmem_scnprintf() or mdb_snprintf(). * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. */ #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ { \ static const char *const copyname[] = \ { "zero", "single", "double", "triple" }; \ int len = 0; \ int copies = 0; \ const char *crypt_type; \ if (bp != NULL) { \ if (BP_IS_ENCRYPTED(bp)) { \ crypt_type = "encrypted"; \ /* LINTED E_SUSPICIOUS_COMPARISON */ \ } else if (BP_IS_AUTHENTICATED(bp)) { \ crypt_type = "authenticated"; \ } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \ crypt_type = "indirect-MAC"; \ } else { \ crypt_type = "unencrypted"; \ } \ } \ if (bp == NULL) { \ len += func(buf + len, size - len, ""); \ } else if (BP_IS_HOLE(bp)) { \ len += func(buf + len, size - len, \ "HOLE [L%llu %s] " \ "size=%llxL birth=%lluL", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ (u_longlong_t)BP_GET_LSIZE(bp), \ (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp)); \ } else if (BP_IS_EMBEDDED(bp)) { \ len = func(buf + len, size - len, \ "EMBEDDED [L%llu %s] et=%u %s " \ "size=%llxL/%llxP birth=%lluL", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ (int)BPE_GET_ETYPE(bp), \ compress, \ (u_longlong_t)BPE_GET_LSIZE(bp), \ (u_longlong_t)BPE_GET_PSIZE(bp), \ (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp)); \ } else if (BP_IS_REDACTED(bp)) { \ len += func(buf + len, size - len, \ "REDACTED [L%llu %s] size=%llxL birth=%lluL", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ (u_longlong_t)BP_GET_LSIZE(bp), \ (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp)); \ } else { \ for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ const dva_t *dva = &bp->blk_dva[d]; \ if (DVA_IS_VALID(dva)) \ copies++; \ len += func(buf + len, size - len, \ "DVA[%d]=<%llu:%llx:%llx>%c", d, \ (u_longlong_t)DVA_GET_VDEV(dva), \ (u_longlong_t)DVA_GET_OFFSET(dva), \ (u_longlong_t)DVA_GET_ASIZE(dva), \ ws); \ } \ ASSERT3S(copies, >, 0); \ if (BP_IS_ENCRYPTED(bp)) { \ len += func(buf + len, size - len, \ "salt=%llx iv=%llx:%llx%c", \ (u_longlong_t)bp->blk_dva[2].dva_word[0], \ (u_longlong_t)bp->blk_dva[2].dva_word[1], \ (u_longlong_t)BP_GET_IV2(bp), \ ws); \ } \ if (BP_IS_GANG(bp) && \ DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ copies--; \ len += func(buf + len, size - len, \ "[L%llu %s] %s %s %s %s %s %s %s%c" \ "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ "cksum=%016llx:%016llx:%016llx:%016llx", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ checksum, \ compress, \ crypt_type, \ BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ BP_IS_GANG(bp) ? "gang" : "contiguous", \ BP_GET_DEDUP(bp) ? "dedup" : "unique", \ copyname[copies], \ ws, \ (u_longlong_t)BP_GET_LSIZE(bp), \ (u_longlong_t)BP_GET_PSIZE(bp), \ (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp), \ (u_longlong_t)BP_GET_BIRTH(bp), \ (u_longlong_t)BP_GET_FILL(bp), \ ws, \ (u_longlong_t)bp->blk_cksum.zc_word[0], \ (u_longlong_t)bp->blk_cksum.zc_word[1], \ (u_longlong_t)bp->blk_cksum.zc_word[2], \ (u_longlong_t)bp->blk_cksum.zc_word[3]); \ } \ ASSERT(len < size); \ } #define BP_GET_BUFC_TYPE(bp) \ (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA) typedef enum spa_import_type { SPA_IMPORT_EXISTING, SPA_IMPORT_ASSEMBLE } spa_import_type_t; typedef enum spa_mode { SPA_MODE_UNINIT = 0, SPA_MODE_READ = 1, SPA_MODE_WRITE = 2, } spa_mode_t; /* * Send TRIM commands in-line during normal pool operation while deleting. * OFF: no * ON: yes */ typedef enum { SPA_AUTOTRIM_OFF = 0, /* default */ SPA_AUTOTRIM_ON, } spa_autotrim_t; /* * Reason TRIM command was issued, used internally for accounting purposes. */ typedef enum trim_type { TRIM_TYPE_MANUAL = 0, TRIM_TYPE_AUTO = 1, TRIM_TYPE_SIMPLE = 2 } trim_type_t; /* state manipulation functions */ extern int spa_open(const char *pool, spa_t **, const void *tag); extern int spa_open_rewind(const char *pool, spa_t **, const void *tag, nvlist_t *policy, nvlist_t **config); extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, size_t buflen); extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, nvlist_t *zplprops, struct dsl_crypto_params *dcp); extern int spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags); extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); extern int spa_destroy(const char *pool); extern int spa_checkpoint(const char *pool); extern int spa_checkpoint_discard(const char *pool); extern int spa_export(const char *pool, nvlist_t **oldconfig, boolean_t force, boolean_t hardforce); extern int spa_reset(const char *pool); extern void spa_async_request(spa_t *spa, int flag); extern void spa_async_unrequest(spa_t *spa, int flag); extern void spa_async_suspend(spa_t *spa); extern void spa_async_resume(spa_t *spa); extern int spa_async_tasks(spa_t *spa); extern spa_t *spa_inject_addref(char *pool); extern void spa_inject_delref(spa_t *spa); extern void spa_scan_stat_init(spa_t *spa); extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); extern int bpobj_enqueue_alloc_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx); extern int bpobj_enqueue_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx); #define SPA_ASYNC_CONFIG_UPDATE 0x01 #define SPA_ASYNC_REMOVE 0x02 #define SPA_ASYNC_PROBE 0x04 #define SPA_ASYNC_RESILVER_DONE 0x08 #define SPA_ASYNC_RESILVER 0x10 #define SPA_ASYNC_AUTOEXPAND 0x20 #define SPA_ASYNC_REMOVE_DONE 0x40 #define SPA_ASYNC_REMOVE_STOP 0x80 #define SPA_ASYNC_INITIALIZE_RESTART 0x100 #define SPA_ASYNC_TRIM_RESTART 0x200 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400 #define SPA_ASYNC_L2CACHE_REBUILD 0x800 #define SPA_ASYNC_L2CACHE_TRIM 0x1000 #define SPA_ASYNC_REBUILD_DONE 0x2000 #define SPA_ASYNC_DETACH_SPARE 0x4000 /* device manipulation */ extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot, boolean_t ashift_check); extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing, int rebuild); extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done); extern int spa_vdev_alloc(spa_t *spa, uint64_t guid); extern int spa_vdev_noalloc(spa_t *spa, uint64_t guid); extern boolean_t spa_vdev_remove_active(spa_t *spa); extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, nvlist_t *vdev_errlist); extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist); extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); extern int spa_vdev_split_mirror(spa_t *spa, const char *newname, nvlist_t *config, nvlist_t *props, boolean_t exp); /* spare state (which is global across all pools) */ extern void spa_spare_add(vdev_t *vd); extern void spa_spare_remove(vdev_t *vd); extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); extern void spa_spare_activate(vdev_t *vd); /* L2ARC state (which is global across all pools) */ extern void spa_l2cache_add(vdev_t *vd); extern void spa_l2cache_remove(vdev_t *vd); extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); extern void spa_l2cache_activate(vdev_t *vd); extern void spa_l2cache_drop(spa_t *spa); /* scanning */ extern int spa_scan(spa_t *spa, pool_scan_func_t func); extern int spa_scan_stop(spa_t *spa); extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag); /* spa syncing */ extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ extern void spa_sync_allpools(void); extern uint_t zfs_sync_pass_deferred_free; /* spa sync taskqueues */ taskq_t *spa_sync_tq_create(spa_t *spa, const char *name); void spa_sync_tq_destroy(spa_t *spa); void spa_select_allocator(zio_t *zio); /* spa namespace global mutex */ extern kmutex_t spa_namespace_lock; +extern avl_tree_t spa_namespace_avl; +extern kcondvar_t spa_namespace_cv; /* * SPA configuration functions in spa_config.c */ #define SPA_CONFIG_UPDATE_POOL 0 #define SPA_CONFIG_UPDATE_VDEVS 1 extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t, boolean_t); extern void spa_config_load(void); extern int spa_all_configs(uint64_t *generation, nvlist_t **pools); extern void spa_config_set(spa_t *spa, nvlist_t *config); extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats); extern void spa_config_update(spa_t *spa, int what); extern int spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id, int atype); /* * Miscellaneous SPA routines in spa_misc.c */ /* Namespace manipulation */ extern spa_t *spa_lookup(const char *name); extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); extern void spa_remove(spa_t *spa); extern spa_t *spa_next(spa_t *prev); /* Refcount functions */ extern void spa_open_ref(spa_t *spa, const void *tag); extern void spa_close(spa_t *spa, const void *tag); extern void spa_async_close(spa_t *spa, const void *tag); extern boolean_t spa_refcount_zero(spa_t *spa); #define SCL_NONE 0x00 #define SCL_CONFIG 0x01 #define SCL_STATE 0x02 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ #define SCL_ALLOC 0x08 #define SCL_ZIO 0x10 #define SCL_FREE 0x20 #define SCL_VDEV 0x40 #define SCL_LOCKS 7 #define SCL_ALL ((1 << SCL_LOCKS) - 1) #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) /* Historical pool statistics */ typedef struct spa_history_kstat { kmutex_t lock; uint64_t count; uint64_t size; kstat_t *kstat; void *priv; list_t list; } spa_history_kstat_t; typedef struct spa_history_list { uint64_t size; procfs_list_t procfs_list; } spa_history_list_t; typedef struct spa_stats { spa_history_list_t read_history; spa_history_list_t txg_history; spa_history_kstat_t tx_assign_histogram; spa_history_list_t mmp_history; spa_history_kstat_t state; /* pool state */ spa_history_kstat_t guid; /* pool guid */ spa_history_kstat_t iostats; } spa_stats_t; typedef enum txg_state { TXG_STATE_BIRTH = 0, TXG_STATE_OPEN = 1, TXG_STATE_QUIESCED = 2, TXG_STATE_WAIT_FOR_SYNC = 3, TXG_STATE_SYNCED = 4, TXG_STATE_COMMITTED = 5, } txg_state_t; typedef struct txg_stat { vdev_stat_t vs1; vdev_stat_t vs2; uint64_t txg; uint64_t ndirty; } txg_stat_t; /* Assorted pool IO kstats */ typedef struct spa_iostats { kstat_named_t trim_extents_written; kstat_named_t trim_bytes_written; kstat_named_t trim_extents_skipped; kstat_named_t trim_bytes_skipped; kstat_named_t trim_extents_failed; kstat_named_t trim_bytes_failed; kstat_named_t autotrim_extents_written; kstat_named_t autotrim_bytes_written; kstat_named_t autotrim_extents_skipped; kstat_named_t autotrim_bytes_skipped; kstat_named_t autotrim_extents_failed; kstat_named_t autotrim_bytes_failed; kstat_named_t simple_trim_extents_written; kstat_named_t simple_trim_bytes_written; kstat_named_t simple_trim_extents_skipped; kstat_named_t simple_trim_bytes_skipped; kstat_named_t simple_trim_extents_failed; kstat_named_t simple_trim_bytes_failed; } spa_iostats_t; extern void spa_stats_init(spa_t *spa); extern void spa_stats_destroy(spa_t *spa); extern void spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb, uint32_t aflags); extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time); extern int spa_txg_history_set(spa_t *spa, uint64_t txg, txg_state_t completed_state, hrtime_t completed_time); extern txg_stat_t *spa_txg_history_init_io(spa_t *, uint64_t, struct dsl_pool *); extern void spa_txg_history_fini_io(spa_t *, txg_stat_t *); extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs); extern int spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_kstat_id); extern int spa_mmp_history_set(spa_t *spa, uint64_t mmp_kstat_id, int io_error, hrtime_t duration); extern void spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp, uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_kstat_id, int error); extern void spa_iostats_trim_add(spa_t *spa, trim_type_t type, uint64_t extents_written, uint64_t bytes_written, uint64_t extents_skipped, uint64_t bytes_skipped, uint64_t extents_failed, uint64_t bytes_failed); extern void spa_import_progress_add(spa_t *spa); extern void spa_import_progress_remove(uint64_t spa_guid); extern int spa_import_progress_set_mmp_check(uint64_t pool_guid, uint64_t mmp_sec_remaining); extern int spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t max_txg); extern int spa_import_progress_set_state(uint64_t pool_guid, spa_load_state_t spa_load_state); extern void spa_import_progress_set_notes(spa_t *spa, const char *fmt, ...) __printflike(2, 3); extern void spa_import_progress_set_notes_nolog(spa_t *spa, const char *fmt, ...) __printflike(2, 3); /* Pool configuration locks */ extern int spa_config_tryenter(spa_t *spa, int locks, const void *tag, krw_t rw); extern void spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw); extern void spa_config_enter_mmp(spa_t *spa, int locks, const void *tag, krw_t rw); extern void spa_config_exit(spa_t *spa, int locks, const void *tag); extern int spa_config_held(spa_t *spa, int locks, krw_t rw); /* Pool vdev add/remove lock */ extern uint64_t spa_vdev_enter(spa_t *spa); extern uint64_t spa_vdev_detach_enter(spa_t *spa, uint64_t guid); extern uint64_t spa_vdev_config_enter(spa_t *spa); extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, const char *tag); extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); /* Pool vdev state change lock */ extern void spa_vdev_state_enter(spa_t *spa, int oplock); extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); /* Log state */ typedef enum spa_log_state { SPA_LOG_UNKNOWN = 0, /* unknown log state */ SPA_LOG_MISSING, /* missing log(s) */ SPA_LOG_CLEAR, /* clear the log(s) */ SPA_LOG_GOOD, /* log(s) are good */ } spa_log_state_t; extern spa_log_state_t spa_get_log_state(spa_t *spa); extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); extern int spa_reset_logs(spa_t *spa); /* Log claim callback */ extern void spa_claim_notify(zio_t *zio); extern void spa_deadman(void *); /* Accessor functions */ extern boolean_t spa_shutting_down(spa_t *spa); extern struct dsl_pool *spa_get_dsl(spa_t *spa); extern boolean_t spa_is_initializing(spa_t *spa); extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa); extern blkptr_t *spa_get_rootblkptr(spa_t *spa); extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); extern void spa_altroot(spa_t *, char *, size_t); extern uint32_t spa_sync_pass(spa_t *spa); extern char *spa_name(spa_t *spa); extern uint64_t spa_guid(spa_t *spa); extern uint64_t spa_load_guid(spa_t *spa); extern uint64_t spa_last_synced_txg(spa_t *spa); extern uint64_t spa_first_txg(spa_t *spa); extern uint64_t spa_syncing_txg(spa_t *spa); extern uint64_t spa_final_dirty_txg(spa_t *spa); extern uint64_t spa_version(spa_t *spa); extern pool_state_t spa_state(spa_t *spa); extern spa_load_state_t spa_load_state(spa_t *spa); extern uint64_t spa_freeze_txg(spa_t *spa); extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize); extern uint64_t spa_get_dspace(spa_t *spa); extern uint64_t spa_get_checkpoint_space(spa_t *spa); extern uint64_t spa_get_slop_space(spa_t *spa); extern void spa_update_dspace(spa_t *spa); extern uint64_t spa_version(spa_t *spa); extern boolean_t spa_deflate(spa_t *spa); extern metaslab_class_t *spa_normal_class(spa_t *spa); extern metaslab_class_t *spa_log_class(spa_t *spa); extern metaslab_class_t *spa_embedded_log_class(spa_t *spa); extern metaslab_class_t *spa_special_class(spa_t *spa); extern metaslab_class_t *spa_dedup_class(spa_t *spa); extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype, uint_t level, uint_t special_smallblk); extern void spa_evicting_os_register(spa_t *, objset_t *os); extern void spa_evicting_os_deregister(spa_t *, objset_t *os); extern void spa_evicting_os_wait(spa_t *spa); extern int spa_max_replication(spa_t *spa); extern int spa_prev_software_version(spa_t *spa); extern uint64_t spa_get_failmode(spa_t *spa); extern uint64_t spa_get_deadman_failmode(spa_t *spa); extern void spa_set_deadman_failmode(spa_t *spa, const char *failmode); extern boolean_t spa_suspended(spa_t *spa); extern uint64_t spa_bootfs(spa_t *spa); extern uint64_t spa_delegation(spa_t *spa); extern objset_t *spa_meta_objset(spa_t *spa); extern space_map_t *spa_syncing_log_sm(spa_t *spa); extern uint64_t spa_deadman_synctime(spa_t *spa); extern uint64_t spa_deadman_ziotime(spa_t *spa); extern uint64_t spa_dirty_data(spa_t *spa); extern spa_autotrim_t spa_get_autotrim(spa_t *spa); extern int spa_get_allocator(spa_t *spa); extern void spa_set_allocator(spa_t *spa, const char *allocator); /* Miscellaneous support routines */ extern void spa_load_failed(spa_t *spa, const char *fmt, ...) __attribute__((format(printf, 2, 3))); extern void spa_load_note(spa_t *spa, const char *fmt, ...) __attribute__((format(printf, 2, 3))); extern void spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx); extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); extern char *spa_strdup(const char *); extern void spa_strfree(char *); extern uint64_t spa_generate_guid(spa_t *spa); extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); extern void spa_freeze(spa_t *spa); extern int spa_change_guid(spa_t *spa); extern void spa_upgrade(spa_t *spa, uint64_t version); extern void spa_evict_all(void); extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t l2cache); extern boolean_t spa_has_l2cache(spa_t *, uint64_t guid); extern boolean_t spa_has_spare(spa_t *, uint64_t guid); extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); extern boolean_t spa_has_slogs(spa_t *spa); extern boolean_t spa_is_root(spa_t *spa); extern boolean_t spa_writeable(spa_t *spa); extern boolean_t spa_has_pending_synctask(spa_t *spa); extern int spa_maxblocksize(spa_t *spa); extern int spa_maxdnodesize(spa_t *spa); extern boolean_t spa_has_checkpoint(spa_t *spa); extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa); extern boolean_t spa_suspend_async_destroy(spa_t *spa); extern uint64_t spa_min_claim_txg(spa_t *spa); extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva, const blkptr_t *bp); typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size, void *arg); extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp, spa_remap_cb_t callback, void *arg); extern uint64_t spa_get_last_removal_txg(spa_t *spa); extern boolean_t spa_trust_config(spa_t *spa); extern uint64_t spa_missing_tvds_allowed(spa_t *spa); extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing); extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa); extern uint64_t spa_total_metaslabs(spa_t *spa); extern boolean_t spa_multihost(spa_t *spa); extern uint32_t spa_get_hostid(spa_t *spa); extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *); extern boolean_t spa_livelist_delete_check(spa_t *spa); extern spa_mode_t spa_mode(spa_t *spa); extern uint64_t zfs_strtonum(const char *str, char **nptr); extern char *spa_his_ievent_table[]; extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, char *his_buf); extern int spa_history_log(spa_t *spa, const char *his_buf); extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); extern void spa_history_log_version(spa_t *spa, const char *operation, dmu_tx_t *tx); extern void spa_history_log_internal(spa_t *spa, const char *operation, dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5); extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5); extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5); extern const char *spa_state_to_name(spa_t *spa); /* error handling */ struct zbookmark_phys; extern void spa_log_error(spa_t *spa, const zbookmark_phys_t *zb, const uint64_t birth); extern void spa_remove_error(spa_t *spa, zbookmark_phys_t *zb, uint64_t birth); extern int zfs_ereport_post(const char *clazz, spa_t *spa, vdev_t *vd, const zbookmark_phys_t *zb, zio_t *zio, uint64_t state); extern boolean_t zfs_ereport_is_valid(const char *clazz, spa_t *spa, vdev_t *vd, zio_t *zio); extern void zfs_ereport_taskq_fini(void); extern void zfs_ereport_clear(spa_t *spa, vdev_t *vd); extern nvlist_t *zfs_event_create(spa_t *spa, vdev_t *vd, const char *type, const char *name, nvlist_t *aux); extern void zfs_post_remove(spa_t *spa, vdev_t *vd); extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate); extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); extern uint64_t spa_approx_errlog_size(spa_t *spa); extern int spa_get_errlog(spa_t *spa, void *uaddr, uint64_t *count); extern uint64_t spa_get_last_errlog_size(spa_t *spa); extern void spa_errlog_rotate(spa_t *spa); extern void spa_errlog_drain(spa_t *spa); extern void spa_errlog_sync(spa_t *spa, uint64_t txg); extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); extern void spa_delete_dataset_errlog(spa_t *spa, uint64_t ds, dmu_tx_t *tx); extern void spa_swap_errlog(spa_t *spa, uint64_t new_head_ds, uint64_t old_head_ds, dmu_tx_t *tx); extern void sync_error_list(spa_t *spa, avl_tree_t *t, uint64_t *obj, dmu_tx_t *tx); extern void spa_upgrade_errlog(spa_t *spa, dmu_tx_t *tx); extern int find_top_affected_fs(spa_t *spa, uint64_t head_ds, zbookmark_err_phys_t *zep, uint64_t *top_affected_fs); extern int find_birth_txg(struct dsl_dataset *ds, zbookmark_err_phys_t *zep, uint64_t *birth_txg); extern void zep_to_zb(uint64_t dataset, zbookmark_err_phys_t *zep, zbookmark_phys_t *zb); extern void name_to_errphys(char *buf, zbookmark_err_phys_t *zep); /* vdev mirror */ extern void vdev_mirror_stat_init(void); extern void vdev_mirror_stat_fini(void); /* Initialization and termination */ extern void spa_init(spa_mode_t mode); extern void spa_fini(void); extern void spa_boot_init(void); /* properties */ extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); /* asynchronous event notification */ extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl, const char *name); extern void zfs_ereport_zvol_post(const char *subclass, const char *name, const char *device_name, const char *raw_name); /* waiting for pool activities to complete */ extern int spa_wait(const char *pool, zpool_wait_activity_t activity, boolean_t *waited); extern int spa_wait_tag(const char *name, zpool_wait_activity_t activity, uint64_t tag, boolean_t *waited); extern void spa_notify_waiters(spa_t *spa); extern void spa_wake_waiters(spa_t *spa); extern void spa_import_os(spa_t *spa); extern void spa_export_os(spa_t *spa); extern void spa_activate_os(spa_t *spa); extern void spa_deactivate_os(spa_t *spa); /* module param call functions */ int param_set_deadman_ziotime(ZFS_MODULE_PARAM_ARGS); int param_set_deadman_synctime(ZFS_MODULE_PARAM_ARGS); int param_set_slop_shift(ZFS_MODULE_PARAM_ARGS); int param_set_deadman_failmode(ZFS_MODULE_PARAM_ARGS); int param_set_active_allocator(ZFS_MODULE_PARAM_ARGS); #ifdef ZFS_DEBUG #define dprintf_bp(bp, fmt, ...) do { \ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ kmem_free(__blkbuf, BP_SPRINTF_LEN); \ } \ } while (0) #else #define dprintf_bp(bp, fmt, ...) #endif extern spa_mode_t spa_mode_global; extern int zfs_deadman_enabled; extern uint64_t zfs_deadman_synctime_ms; extern uint64_t zfs_deadman_ziotime_ms; extern uint64_t zfs_deadman_checktime_ms; extern kmem_cache_t *zio_buf_cache[]; extern kmem_cache_t *zio_data_buf_cache[]; #ifdef __cplusplus } #endif #endif /* _SYS_SPA_H */ diff --git a/sys/contrib/openzfs/include/sys/spa_impl.h b/sys/contrib/openzfs/include/sys/spa_impl.h index 0cd0c4720fbe..d7da085ab313 100644 --- a/sys/contrib/openzfs/include/sys/spa_impl.h +++ b/sys/contrib/openzfs/include/sys/spa_impl.h @@ -1,493 +1,494 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2011, 2019 by Delphix. All rights reserved. + * Copyright (c) 2011, 2024 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2016 Actifio, Inc. All rights reserved. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2019 Datto Inc. */ #ifndef _SYS_SPA_IMPL_H #define _SYS_SPA_IMPL_H #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif typedef struct spa_alloc { kmutex_t spaa_lock; avl_tree_t spaa_tree; } ____cacheline_aligned spa_alloc_t; typedef struct spa_error_entry { zbookmark_phys_t se_bookmark; char *se_name; avl_node_t se_avl; zbookmark_err_phys_t se_zep; /* not accounted in avl_find */ } spa_error_entry_t; typedef struct spa_history_phys { uint64_t sh_pool_create_len; /* ending offset of zpool create */ uint64_t sh_phys_max_off; /* physical EOF */ uint64_t sh_bof; /* logical BOF */ uint64_t sh_eof; /* logical EOF */ uint64_t sh_records_lost; /* num of records overwritten */ } spa_history_phys_t; /* * All members must be uint64_t, for byteswap purposes. */ typedef struct spa_removing_phys { uint64_t sr_state; /* dsl_scan_state_t */ /* * The vdev ID that we most recently attempted to remove, * or -1 if no removal has been attempted. */ uint64_t sr_removing_vdev; /* * The vdev ID that we most recently successfully removed, * or -1 if no devices have been removed. */ uint64_t sr_prev_indirect_vdev; uint64_t sr_start_time; uint64_t sr_end_time; /* * Note that we can not use the space map's or indirect mapping's * accounting as a substitute for these values, because we need to * count frees of not-yet-copied data as though it did the copy. * Otherwise, we could get into a situation where copied > to_copy, * or we complete before copied == to_copy. */ uint64_t sr_to_copy; /* bytes that need to be copied */ uint64_t sr_copied; /* bytes that have been copied or freed */ } spa_removing_phys_t; /* * This struct is stored as an entry in the DMU_POOL_DIRECTORY_OBJECT * (with key DMU_POOL_CONDENSING_INDIRECT). It is present if a condense * of an indirect vdev's mapping object is in progress. */ typedef struct spa_condensing_indirect_phys { /* * The vdev ID of the indirect vdev whose indirect mapping is * being condensed. */ uint64_t scip_vdev; /* * The vdev's old obsolete spacemap. This spacemap's contents are * being integrated into the new mapping. */ uint64_t scip_prev_obsolete_sm_object; /* * The new mapping object that is being created. */ uint64_t scip_next_mapping_object; } spa_condensing_indirect_phys_t; struct spa_aux_vdev { uint64_t sav_object; /* MOS object for device list */ nvlist_t *sav_config; /* cached device config */ vdev_t **sav_vdevs; /* devices */ int sav_count; /* number devices */ boolean_t sav_sync; /* sync the device list */ nvlist_t **sav_pending; /* pending device additions */ uint_t sav_npending; /* # pending devices */ }; typedef struct spa_config_lock { kmutex_t scl_lock; kthread_t *scl_writer; int scl_write_wanted; int scl_count; kcondvar_t scl_cv; } ____cacheline_aligned spa_config_lock_t; typedef struct spa_config_dirent { list_node_t scd_link; char *scd_path; } spa_config_dirent_t; typedef enum zio_taskq_type { ZIO_TASKQ_ISSUE = 0, ZIO_TASKQ_ISSUE_HIGH, ZIO_TASKQ_INTERRUPT, ZIO_TASKQ_INTERRUPT_HIGH, ZIO_TASKQ_TYPES } zio_taskq_type_t; /* * State machine for the zpool-poolname process. The states transitions * are done as follows: * * From To Routine * PROC_NONE -> PROC_CREATED spa_activate() * PROC_CREATED -> PROC_ACTIVE spa_thread() * PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate() * PROC_DEACTIVATE -> PROC_GONE spa_thread() * PROC_GONE -> PROC_NONE spa_deactivate() */ typedef enum spa_proc_state { SPA_PROC_NONE, /* spa_proc = &p0, no process created */ SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */ SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */ SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */ SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */ } spa_proc_state_t; typedef struct spa_taskqs { uint_t stqs_count; taskq_t **stqs_taskq; } spa_taskqs_t; /* one for each thread in the spa sync taskq */ typedef struct spa_syncthread_info { kthread_t *sti_thread; taskq_t *sti_wr_iss_tq; /* assigned wr_iss taskq */ } spa_syncthread_info_t; typedef enum spa_all_vdev_zap_action { AVZ_ACTION_NONE = 0, AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */ AVZ_ACTION_REBUILD, /* Populate the new AVZ, see spa_avz_rebuild */ AVZ_ACTION_INITIALIZE } spa_avz_action_t; typedef enum spa_config_source { SPA_CONFIG_SRC_NONE = 0, SPA_CONFIG_SRC_SCAN, /* scan of path (default: /dev/dsk) */ SPA_CONFIG_SRC_CACHEFILE, /* any cachefile */ SPA_CONFIG_SRC_TRYIMPORT, /* returned from call to tryimport */ SPA_CONFIG_SRC_SPLIT, /* new pool in a pool split */ SPA_CONFIG_SRC_MOS /* MOS, but not always from right txg */ } spa_config_source_t; struct spa { /* * Fields protected by spa_namespace_lock. */ char spa_name[ZFS_MAX_DATASET_NAME_LEN]; /* pool name */ char *spa_comment; /* comment */ avl_node_t spa_avl; /* node in spa_namespace_avl */ nvlist_t *spa_config; /* last synced config */ nvlist_t *spa_config_syncing; /* currently syncing config */ nvlist_t *spa_config_splitting; /* config for splitting */ nvlist_t *spa_load_info; /* info and errors from load */ uint64_t spa_config_txg; /* txg of last config change */ uint32_t spa_sync_pass; /* iterate-to-convergence */ pool_state_t spa_state; /* pool state */ int spa_inject_ref; /* injection references */ uint8_t spa_sync_on; /* sync threads are running */ spa_load_state_t spa_load_state; /* current load operation */ boolean_t spa_indirect_vdevs_loaded; /* mappings loaded? */ boolean_t spa_trust_config; /* do we trust vdev tree? */ boolean_t spa_is_splitting; /* in the middle of a split? */ spa_config_source_t spa_config_source; /* where config comes from? */ uint64_t spa_import_flags; /* import specific flags */ spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES]; dsl_pool_t *spa_dsl_pool; boolean_t spa_is_initializing; /* true while opening pool */ boolean_t spa_is_exporting; /* true while exporting pool */ + kthread_t *spa_load_thread; /* loading, no namespace lock */ metaslab_class_t *spa_normal_class; /* normal data class */ metaslab_class_t *spa_log_class; /* intent log data class */ metaslab_class_t *spa_embedded_log_class; /* log on normal vdevs */ metaslab_class_t *spa_special_class; /* special allocation class */ metaslab_class_t *spa_dedup_class; /* dedup allocation class */ uint64_t spa_first_txg; /* first txg after spa_open() */ uint64_t spa_final_txg; /* txg of export/destroy */ uint64_t spa_freeze_txg; /* freeze pool at this txg */ uint64_t spa_load_max_txg; /* best initial ub_txg */ uint64_t spa_claim_max_txg; /* highest claimed birth txg */ inode_timespec_t spa_loaded_ts; /* 1st successful open time */ objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */ kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */ list_t spa_evicting_os_list; /* Objsets being evicted. */ kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */ txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */ vdev_t *spa_root_vdev; /* top-level vdev container */ uint64_t spa_min_ashift; /* of vdevs in normal class */ uint64_t spa_max_ashift; /* of vdevs in normal class */ uint64_t spa_min_alloc; /* of vdevs in normal class */ uint64_t spa_gcd_alloc; /* of vdevs in normal class */ uint64_t spa_config_guid; /* config pool guid */ uint64_t spa_load_guid; /* spa_load initialized guid */ uint64_t spa_last_synced_guid; /* last synced guid */ list_t spa_config_dirty_list; /* vdevs with dirty config */ list_t spa_state_dirty_list; /* vdevs with dirty state */ /* * spa_allocs is an array, whose lengths is stored in spa_alloc_count. * There is one tree and one lock for each allocator, to help improve * allocation performance in write-heavy workloads. */ spa_alloc_t *spa_allocs; int spa_alloc_count; int spa_active_allocator; /* selectable allocator */ /* per-allocator sync thread taskqs */ taskq_t *spa_sync_tq; spa_syncthread_info_t *spa_syncthreads; spa_aux_vdev_t spa_spares; /* hot spares */ spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */ boolean_t spa_aux_sync_uber; /* need to sync aux uber */ nvlist_t *spa_label_features; /* Features for reading MOS */ uint64_t spa_config_object; /* MOS object for pool config */ uint64_t spa_config_generation; /* config generation number */ uint64_t spa_syncing_txg; /* txg currently syncing */ bpobj_t spa_deferred_bpobj; /* deferred-free bplist */ bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */ zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ /* checksum context templates */ kmutex_t spa_cksum_tmpls_lock; void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; uberblock_t spa_ubsync; /* last synced uberblock */ uberblock_t spa_uberblock; /* current uberblock */ boolean_t spa_extreme_rewind; /* rewind past deferred frees */ kmutex_t spa_scrub_lock; /* resilver/scrub lock */ uint64_t spa_scrub_inflight; /* in-flight scrub bytes */ /* in-flight verification bytes */ uint64_t spa_load_verify_bytes; kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */ uint8_t spa_scrub_active; /* active or suspended? */ uint8_t spa_scrub_type; /* type of scrub we're doing */ uint8_t spa_scrub_finished; /* indicator to rotate logs */ uint8_t spa_scrub_started; /* started since last boot */ uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */ uint64_t spa_scan_pass_start; /* start time per pass/reboot */ uint64_t spa_scan_pass_scrub_pause; /* scrub pause time */ uint64_t spa_scan_pass_scrub_spent_paused; /* total paused */ uint64_t spa_scan_pass_exam; /* examined bytes per pass */ uint64_t spa_scan_pass_issued; /* issued bytes per pass */ /* error scrub pause time in milliseconds */ uint64_t spa_scan_pass_errorscrub_pause; /* total error scrub paused time in milliseconds */ uint64_t spa_scan_pass_errorscrub_spent_paused; /* * We are in the middle of a resilver, and another resilver * is needed once this one completes. This is set iff any * vdev_resilver_deferred is set. */ boolean_t spa_resilver_deferred; kmutex_t spa_async_lock; /* protect async state */ kthread_t *spa_async_thread; /* thread doing async task */ int spa_async_suspended; /* async tasks suspended */ kcondvar_t spa_async_cv; /* wait for thread_exit() */ uint16_t spa_async_tasks; /* async task mask */ uint64_t spa_missing_tvds; /* unopenable tvds on load */ uint64_t spa_missing_tvds_allowed; /* allow loading spa? */ uint64_t spa_nonallocating_dspace; spa_removing_phys_t spa_removing_phys; spa_vdev_removal_t *spa_vdev_removal; spa_condensing_indirect_phys_t spa_condensing_indirect_phys; spa_condensing_indirect_t *spa_condensing_indirect; zthr_t *spa_condense_zthr; /* zthr doing condense. */ vdev_raidz_expand_t *spa_raidz_expand; zthr_t *spa_raidz_expand_zthr; uint64_t spa_checkpoint_txg; /* the txg of the checkpoint */ spa_checkpoint_info_t spa_checkpoint_info; /* checkpoint accounting */ zthr_t *spa_checkpoint_discard_zthr; space_map_t *spa_syncing_log_sm; /* current log space map */ avl_tree_t spa_sm_logs_by_txg; kmutex_t spa_flushed_ms_lock; /* for metaslabs_by_flushed */ avl_tree_t spa_metaslabs_by_flushed; spa_unflushed_stats_t spa_unflushed_stats; list_t spa_log_summary; uint64_t spa_log_flushall_txg; zthr_t *spa_livelist_delete_zthr; /* deleting livelists */ zthr_t *spa_livelist_condense_zthr; /* condensing livelists */ uint64_t spa_livelists_to_delete; /* set of livelists to free */ livelist_condense_entry_t spa_to_condense; /* next to condense */ char *spa_root; /* alternate root directory */ uint64_t spa_ena; /* spa-wide ereport ENA */ int spa_last_open_failed; /* error if last open failed */ uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */ uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */ uint64_t spa_load_txg; /* ub txg that loaded */ uint64_t spa_load_txg_ts; /* timestamp from that ub */ uint64_t spa_load_meta_errors; /* verify metadata err count */ uint64_t spa_load_data_errors; /* verify data err count */ uint64_t spa_verify_min_txg; /* start txg of verify scrub */ kmutex_t spa_errlog_lock; /* error log lock */ uint64_t spa_errlog_last; /* last error log object */ uint64_t spa_errlog_scrub; /* scrub error log object */ kmutex_t spa_errlist_lock; /* error list/ereport lock */ avl_tree_t spa_errlist_last; /* last error list */ avl_tree_t spa_errlist_scrub; /* scrub error list */ avl_tree_t spa_errlist_healed; /* list of healed blocks */ uint64_t spa_deflate; /* should we deflate? */ uint64_t spa_history; /* history object */ kmutex_t spa_history_lock; /* history lock */ vdev_t *spa_pending_vdev; /* pending vdev additions */ kmutex_t spa_props_lock; /* property lock */ uint64_t spa_pool_props_object; /* object for properties */ uint64_t spa_bootfs; /* default boot filesystem */ uint64_t spa_failmode; /* failure mode for the pool */ uint64_t spa_deadman_failmode; /* failure mode for deadman */ uint64_t spa_delegation; /* delegation on/off */ list_t spa_config_list; /* previous cache file(s) */ /* per-CPU array of root of async I/O: */ zio_t **spa_async_zio_root; zio_t *spa_suspend_zio_root; /* root of all suspended I/O */ zio_t *spa_txg_zio[TXG_SIZE]; /* spa_sync() waits for this */ kmutex_t spa_suspend_lock; /* protects suspend_zio_root */ kcondvar_t spa_suspend_cv; /* notification of resume */ zio_suspend_reason_t spa_suspended; /* pool is suspended */ uint8_t spa_claiming; /* pool is doing zil_claim() */ boolean_t spa_is_root; /* pool is root */ int spa_minref; /* num refs when first opened */ spa_mode_t spa_mode; /* SPA_MODE_{READ|WRITE} */ boolean_t spa_read_spacemaps; /* spacemaps available if ro */ spa_log_state_t spa_log_state; /* log state */ uint64_t spa_autoexpand; /* lun expansion on/off */ ddt_t *spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */ uint64_t spa_ddt_stat_object; /* DDT statistics */ uint64_t spa_dedup_dspace; /* Cache get_dedup_dspace() */ uint64_t spa_dedup_checksum; /* default dedup checksum */ uint64_t spa_dspace; /* dspace in normal class */ struct brt *spa_brt; /* in-core BRT */ kmutex_t spa_vdev_top_lock; /* dueling offline/remove */ kmutex_t spa_proc_lock; /* protects spa_proc* */ kcondvar_t spa_proc_cv; /* spa_proc_state transitions */ spa_proc_state_t spa_proc_state; /* see definition */ proc_t *spa_proc; /* "zpool-poolname" process */ uintptr_t spa_did; /* if procp != p0, did of t1 */ boolean_t spa_autoreplace; /* autoreplace set in open */ int spa_vdev_locks; /* locks grabbed */ uint64_t spa_creation_version; /* version at pool creation */ uint64_t spa_prev_software_version; /* See ub_software_version */ uint64_t spa_feat_for_write_obj; /* required to write to pool */ uint64_t spa_feat_for_read_obj; /* required to read from pool */ uint64_t spa_feat_desc_obj; /* Feature descriptions */ uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */ kmutex_t spa_feat_stats_lock; /* protects spa_feat_stats */ nvlist_t *spa_feat_stats; /* Cache of enabled features */ /* cache feature refcounts */ uint64_t spa_feat_refcount_cache[SPA_FEATURES]; taskqid_t spa_deadman_tqid; /* Task id */ uint64_t spa_deadman_calls; /* number of deadman calls */ hrtime_t spa_sync_starttime; /* starting time of spa_sync */ uint64_t spa_deadman_synctime; /* deadman sync expiration */ uint64_t spa_deadman_ziotime; /* deadman zio expiration */ uint64_t spa_all_vdev_zaps; /* ZAP of per-vd ZAP obj #s */ spa_avz_action_t spa_avz_action; /* destroy/rebuild AVZ? */ uint64_t spa_autotrim; /* automatic background trim? */ uint64_t spa_errata; /* errata issues detected */ spa_stats_t spa_stats; /* assorted spa statistics */ spa_keystore_t spa_keystore; /* loaded crypto keys */ /* arc_memory_throttle() parameters during low memory condition */ uint64_t spa_lowmem_page_load; /* memory load during txg */ uint64_t spa_lowmem_last_txg; /* txg window start */ hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */ taskq_t *spa_zvol_taskq; /* Taskq for minor management */ taskq_t *spa_metaslab_taskq; /* Taskq for metaslab preload */ taskq_t *spa_prefetch_taskq; /* Taskq for prefetch threads */ taskq_t *spa_upgrade_taskq; /* Taskq for upgrade jobs */ uint64_t spa_multihost; /* multihost aware (mmp) */ mmp_thread_t spa_mmp; /* multihost mmp thread */ list_t spa_leaf_list; /* list of leaf vdevs */ uint64_t spa_leaf_list_gen; /* track leaf_list changes */ uint32_t spa_hostid; /* cached system hostid */ /* synchronization for threads in spa_wait */ kmutex_t spa_activities_lock; kcondvar_t spa_activities_cv; kcondvar_t spa_waiters_cv; int spa_waiters; /* number of waiting threads */ boolean_t spa_waiters_cancel; /* waiters should return */ char *spa_compatibility; /* compatibility file(s) */ /* * spa_refcount & spa_config_lock must be the last elements * because zfs_refcount_t changes size based on compilation options. * In order for the MDB module to function correctly, the other * fields must remain in the same location. */ spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */ zfs_refcount_t spa_refcount; /* number of opens */ }; extern char *spa_config_path; extern const char *zfs_deadman_failmode; extern uint_t spa_slop_shift; extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q, task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent, zio_t *zio); extern void spa_taskq_dispatch_sync(spa_t *, zio_type_t t, zio_taskq_type_t q, task_func_t *func, void *arg, uint_t flags); extern void spa_load_spares(spa_t *spa); extern void spa_load_l2cache(spa_t *spa); extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name); extern void spa_event_post(sysevent_t *ev); extern int param_set_deadman_failmode_common(const char *val); extern void spa_set_deadman_synctime(hrtime_t ns); extern void spa_set_deadman_ziotime(hrtime_t ns); extern const char *spa_history_zone(void); extern const char *zfs_active_allocator; extern int param_set_active_allocator_common(const char *val); #ifdef __cplusplus } #endif #endif /* _SYS_SPA_IMPL_H */ diff --git a/sys/contrib/openzfs/include/sys/zfs_ioctl.h b/sys/contrib/openzfs/include/sys/zfs_ioctl.h index 26dfe97604de..525d40759fdd 100644 --- a/sys/contrib/openzfs/include/sys/zfs_ioctl.h +++ b/sys/contrib/openzfs/include/sys/zfs_ioctl.h @@ -1,586 +1,588 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2012, 2020 by Delphix. All rights reserved. + * Copyright (c) 2012, 2024 by Delphix. All rights reserved. * Copyright 2016 RackTop Systems. * Copyright (c) 2017, Intel Corporation. */ #ifndef _SYS_ZFS_IOCTL_H #define _SYS_ZFS_IOCTL_H #include #include #include #include #include #include #ifdef _KERNEL #include #endif /* _KERNEL */ #ifdef __cplusplus extern "C" { #endif /* * The structures in this file are passed between userland and the * kernel. Userland may be running a 32-bit process, while the kernel * is 64-bit. Therefore, these structures need to compile the same in * 32-bit and 64-bit. This means not using type "long", and adding * explicit padding so that the 32-bit structure will not be packed more * tightly than the 64-bit structure (which requires 64-bit alignment). */ /* * Property values for snapdir */ #define ZFS_SNAPDIR_HIDDEN 0 #define ZFS_SNAPDIR_VISIBLE 1 /* * Property values for snapdev */ #define ZFS_SNAPDEV_HIDDEN 0 #define ZFS_SNAPDEV_VISIBLE 1 /* * Property values for acltype */ #define ZFS_ACLTYPE_OFF 0 #define ZFS_ACLTYPE_POSIX 1 #define ZFS_ACLTYPE_NFSV4 2 /* * Field manipulation macros for the drr_versioninfo field of the * send stream header. */ /* * Header types for zfs send streams. */ typedef enum drr_headertype { DMU_SUBSTREAM = 0x1, DMU_COMPOUNDSTREAM = 0x2 } drr_headertype_t; #define DMU_GET_STREAM_HDRTYPE(vi) BF64_GET((vi), 0, 2) #define DMU_SET_STREAM_HDRTYPE(vi, x) BF64_SET((vi), 0, 2, x) #define DMU_GET_FEATUREFLAGS(vi) BF64_GET((vi), 2, 30) #define DMU_SET_FEATUREFLAGS(vi, x) BF64_SET((vi), 2, 30, x) /* * Feature flags for zfs send streams (flags in drr_versioninfo) */ #define DMU_BACKUP_FEATURE_DEDUP (1 << 0) #define DMU_BACKUP_FEATURE_DEDUPPROPS (1 << 1) #define DMU_BACKUP_FEATURE_SA_SPILL (1 << 2) /* flags #3 - #15 are reserved for incompatible closed-source implementations */ #define DMU_BACKUP_FEATURE_EMBED_DATA (1 << 16) #define DMU_BACKUP_FEATURE_LZ4 (1 << 17) /* flag #18 is reserved for a Delphix feature */ #define DMU_BACKUP_FEATURE_LARGE_BLOCKS (1 << 19) #define DMU_BACKUP_FEATURE_RESUMING (1 << 20) #define DMU_BACKUP_FEATURE_REDACTED (1 << 21) #define DMU_BACKUP_FEATURE_COMPRESSED (1 << 22) #define DMU_BACKUP_FEATURE_LARGE_DNODE (1 << 23) #define DMU_BACKUP_FEATURE_RAW (1 << 24) #define DMU_BACKUP_FEATURE_ZSTD (1 << 25) #define DMU_BACKUP_FEATURE_HOLDS (1 << 26) /* * The SWITCH_TO_LARGE_BLOCKS feature indicates that we can receive * incremental LARGE_BLOCKS streams (those with WRITE records of >128KB) even * if the previous send did not use LARGE_BLOCKS, and thus its large blocks * were split into multiple 128KB WRITE records. (See * flush_write_batch_impl() and receive_object()). Older software that does * not support this flag may encounter a bug when switching to large blocks, * which causes files to incorrectly be zeroed. * * This flag is currently not set on any send streams. In the future, we * intend for incremental send streams of snapshots that have large blocks to * use LARGE_BLOCKS by default, and these streams will also have the * SWITCH_TO_LARGE_BLOCKS feature set. This ensures that streams from the * default use of "zfs send" won't encounter the bug mentioned above. */ #define DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS (1 << 27) /* flag #28 is reserved for a Nutanix feature */ /* * flag #29 is the last unused bit. It is reserved to indicate a to-be-designed * extension to the stream format which will accomodate more feature flags. * If you need to add another feature flag, please reach out to the OpenZFS * community, e.g., on GitHub or Slack. */ /* * Mask of all supported backup features */ #define DMU_BACKUP_FEATURE_MASK (DMU_BACKUP_FEATURE_SA_SPILL | \ DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_LZ4 | \ DMU_BACKUP_FEATURE_RESUMING | DMU_BACKUP_FEATURE_LARGE_BLOCKS | \ DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_LARGE_DNODE | \ DMU_BACKUP_FEATURE_RAW | DMU_BACKUP_FEATURE_HOLDS | \ DMU_BACKUP_FEATURE_REDACTED | DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS | \ DMU_BACKUP_FEATURE_ZSTD) /* Are all features in the given flag word currently supported? */ #define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK)) typedef enum dmu_send_resume_token_version { ZFS_SEND_RESUME_TOKEN_VERSION = 1 } dmu_send_resume_token_version_t; /* * The drr_versioninfo field of the dmu_replay_record has the * following layout: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * | reserved | feature-flags |C|S| * +-------+-------+-------+-------+-------+-------+-------+-------+ * * The low order two bits indicate the header type: SUBSTREAM (0x1) * or COMPOUNDSTREAM (0x2). Using two bits for this is historical: * this field used to be a version number, where the two version types * were 1 and 2. Using two bits for this allows earlier versions of * the code to be able to recognize send streams that don't use any * of the features indicated by feature flags. */ #define DMU_BACKUP_MAGIC 0x2F5bacbacULL /* * Send stream flags. Bits 24-31 are reserved for vendor-specific * implementations and should not be used. */ #define DRR_FLAG_CLONE (1<<0) #define DRR_FLAG_CI_DATA (1<<1) /* * This send stream, if it is a full send, includes the FREE and FREEOBJECT * records that are created by the sending process. This means that the send * stream can be received as a clone, even though it is not an incremental. * This is not implemented as a feature flag, because the receiving side does * not need to have implemented it to receive this stream; it is fully backwards * compatible. We need a flag, though, because full send streams without it * cannot necessarily be received as a clone correctly. */ #define DRR_FLAG_FREERECORDS (1<<2) /* * When DRR_FLAG_SPILL_BLOCK is set it indicates the DRR_OBJECT_SPILL * and DRR_SPILL_UNMODIFIED flags are meaningful in the send stream. * * When DRR_FLAG_SPILL_BLOCK is set, DRR_OBJECT records will have * DRR_OBJECT_SPILL set if and only if they should have a spill block * (either an existing one, or a new one in the send stream). When clear * the object does not have a spill block and any existing spill block * should be freed. * * Similarly, when DRR_FLAG_SPILL_BLOCK is set, DRR_SPILL records will * have DRR_SPILL_UNMODIFIED set if and only if they were included for * backward compatibility purposes, and can be safely ignored by new versions * of zfs receive. Previous versions of ZFS which do not understand the * DRR_FLAG_SPILL_BLOCK will process this record and recreate any missing * spill blocks. */ #define DRR_FLAG_SPILL_BLOCK (1<<3) /* * flags in the drr_flags field in the DRR_WRITE, DRR_SPILL, DRR_OBJECT, * DRR_WRITE_BYREF, and DRR_OBJECT_RANGE blocks */ #define DRR_CHECKSUM_DEDUP (1<<0) /* not used for SPILL records */ #define DRR_RAW_BYTESWAP (1<<1) #define DRR_OBJECT_SPILL (1<<2) /* OBJECT record has a spill block */ #define DRR_SPILL_UNMODIFIED (1<<2) /* SPILL record for unmodified block */ #define DRR_IS_DEDUP_CAPABLE(flags) ((flags) & DRR_CHECKSUM_DEDUP) #define DRR_IS_RAW_BYTESWAPPED(flags) ((flags) & DRR_RAW_BYTESWAP) #define DRR_OBJECT_HAS_SPILL(flags) ((flags) & DRR_OBJECT_SPILL) #define DRR_SPILL_IS_UNMODIFIED(flags) ((flags) & DRR_SPILL_UNMODIFIED) /* deal with compressed drr_write replay records */ #define DRR_WRITE_COMPRESSED(drrw) ((drrw)->drr_compressiontype != 0) #define DRR_WRITE_PAYLOAD_SIZE(drrw) \ (DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \ (drrw)->drr_logical_size) #define DRR_SPILL_PAYLOAD_SIZE(drrs) \ ((drrs)->drr_compressed_size ? \ (drrs)->drr_compressed_size : (drrs)->drr_length) #define DRR_OBJECT_PAYLOAD_SIZE(drro) \ ((drro)->drr_raw_bonuslen != 0 ? \ (drro)->drr_raw_bonuslen : P2ROUNDUP((drro)->drr_bonuslen, 8)) /* * zfs ioctl command structure */ /* Header is used in C++ so can't forward declare untagged struct */ struct drr_begin { uint64_t drr_magic; uint64_t drr_versioninfo; /* was drr_version */ uint64_t drr_creation_time; dmu_objset_type_t drr_type; uint32_t drr_flags; uint64_t drr_toguid; uint64_t drr_fromguid; char drr_toname[MAXNAMELEN]; }; typedef struct dmu_replay_record { enum { DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS, DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF, DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_OBJECT_RANGE, DRR_REDACT, DRR_NUMTYPES } drr_type; uint32_t drr_payloadlen; union { struct drr_begin drr_begin; struct drr_end { zio_cksum_t drr_checksum; uint64_t drr_toguid; } drr_end; struct drr_object { uint64_t drr_object; dmu_object_type_t drr_type; dmu_object_type_t drr_bonustype; uint32_t drr_blksz; uint32_t drr_bonuslen; uint8_t drr_checksumtype; uint8_t drr_compress; uint8_t drr_dn_slots; uint8_t drr_flags; uint32_t drr_raw_bonuslen; uint64_t drr_toguid; /* only (possibly) nonzero for raw streams */ uint8_t drr_indblkshift; uint8_t drr_nlevels; uint8_t drr_nblkptr; uint8_t drr_pad[5]; uint64_t drr_maxblkid; /* bonus content follows */ } drr_object; struct drr_freeobjects { uint64_t drr_firstobj; uint64_t drr_numobjs; uint64_t drr_toguid; } drr_freeobjects; struct drr_write { uint64_t drr_object; dmu_object_type_t drr_type; uint32_t drr_pad; uint64_t drr_offset; uint64_t drr_logical_size; uint64_t drr_toguid; uint8_t drr_checksumtype; uint8_t drr_flags; uint8_t drr_compressiontype; uint8_t drr_pad2[5]; /* deduplication key */ ddt_key_t drr_key; /* only nonzero if drr_compressiontype is not 0 */ uint64_t drr_compressed_size; /* only nonzero for raw streams */ uint8_t drr_salt[ZIO_DATA_SALT_LEN]; uint8_t drr_iv[ZIO_DATA_IV_LEN]; uint8_t drr_mac[ZIO_DATA_MAC_LEN]; /* content follows */ } drr_write; struct drr_free { uint64_t drr_object; uint64_t drr_offset; uint64_t drr_length; uint64_t drr_toguid; } drr_free; struct drr_write_byref { /* where to put the data */ uint64_t drr_object; uint64_t drr_offset; uint64_t drr_length; uint64_t drr_toguid; /* where to find the prior copy of the data */ uint64_t drr_refguid; uint64_t drr_refobject; uint64_t drr_refoffset; /* properties of the data */ uint8_t drr_checksumtype; uint8_t drr_flags; uint8_t drr_pad2[6]; ddt_key_t drr_key; /* deduplication key */ } drr_write_byref; struct drr_spill { uint64_t drr_object; uint64_t drr_length; uint64_t drr_toguid; uint8_t drr_flags; uint8_t drr_compressiontype; uint8_t drr_pad[6]; /* only nonzero for raw streams */ uint64_t drr_compressed_size; uint8_t drr_salt[ZIO_DATA_SALT_LEN]; uint8_t drr_iv[ZIO_DATA_IV_LEN]; uint8_t drr_mac[ZIO_DATA_MAC_LEN]; dmu_object_type_t drr_type; /* spill data follows */ } drr_spill; struct drr_write_embedded { uint64_t drr_object; uint64_t drr_offset; /* logical length, should equal blocksize */ uint64_t drr_length; uint64_t drr_toguid; uint8_t drr_compression; uint8_t drr_etype; uint8_t drr_pad[6]; uint32_t drr_lsize; /* uncompressed size of payload */ uint32_t drr_psize; /* compr. (real) size of payload */ /* (possibly compressed) content follows */ } drr_write_embedded; struct drr_object_range { uint64_t drr_firstobj; uint64_t drr_numslots; uint64_t drr_toguid; uint8_t drr_salt[ZIO_DATA_SALT_LEN]; uint8_t drr_iv[ZIO_DATA_IV_LEN]; uint8_t drr_mac[ZIO_DATA_MAC_LEN]; uint8_t drr_flags; uint8_t drr_pad[3]; } drr_object_range; struct drr_redact { uint64_t drr_object; uint64_t drr_offset; uint64_t drr_length; uint64_t drr_toguid; } drr_redact; /* * Note: drr_checksum is overlaid with all record types * except DRR_BEGIN. Therefore its (non-pad) members * must not overlap with members from the other structs. * We accomplish this by putting its members at the very * end of the struct. */ struct drr_checksum { uint64_t drr_pad[34]; /* * fletcher-4 checksum of everything preceding the * checksum. */ zio_cksum_t drr_checksum; } drr_checksum; } drr_u; } dmu_replay_record_t; /* diff record range types */ typedef enum diff_type { DDR_NONE = 0x1, DDR_INUSE = 0x2, DDR_FREE = 0x4 } diff_type_t; /* * The diff reports back ranges of free or in-use objects. */ typedef struct dmu_diff_record { uint64_t ddr_type; uint64_t ddr_first; uint64_t ddr_last; } dmu_diff_record_t; typedef struct zinject_record { uint64_t zi_objset; uint64_t zi_object; uint64_t zi_start; uint64_t zi_end; uint64_t zi_guid; uint32_t zi_level; uint32_t zi_error; uint64_t zi_type; uint32_t zi_freq; uint32_t zi_failfast; char zi_func[MAXNAMELEN]; uint32_t zi_iotype; int32_t zi_duration; uint64_t zi_timer; uint64_t zi_nlanes; uint32_t zi_cmd; uint32_t zi_dvas; } zinject_record_t; #define ZINJECT_NULL 0x1 #define ZINJECT_FLUSH_ARC 0x2 #define ZINJECT_UNLOAD_SPA 0x4 #define ZINJECT_CALC_RANGE 0x8 #define ZEVENT_NONE 0x0 #define ZEVENT_NONBLOCK 0x1 #define ZEVENT_SIZE 1024 #define ZEVENT_SEEK_START 0 #define ZEVENT_SEEK_END UINT64_MAX /* scaled frequency ranges */ #define ZI_PERCENTAGE_MIN 4294UL #define ZI_PERCENTAGE_MAX UINT32_MAX #define ZI_NO_DVA (-1) typedef enum zinject_type { ZINJECT_UNINITIALIZED, ZINJECT_DATA_FAULT, ZINJECT_DEVICE_FAULT, ZINJECT_LABEL_FAULT, ZINJECT_IGNORED_WRITES, ZINJECT_PANIC, ZINJECT_DELAY_IO, ZINJECT_DECRYPT_FAULT, + ZINJECT_DELAY_IMPORT, + ZINJECT_DELAY_EXPORT, } zinject_type_t; typedef struct zfs_share { uint64_t z_exportdata; uint64_t z_sharedata; uint64_t z_sharetype; /* 0 = share, 1 = unshare */ uint64_t z_sharemax; /* max length of share string */ } zfs_share_t; /* * ZFS file systems may behave the usual, POSIX-compliant way, where * name lookups are case-sensitive. They may also be set up so that * all the name lookups are case-insensitive, or so that only some * lookups, the ones that set an FIGNORECASE flag, are case-insensitive. */ typedef enum zfs_case { ZFS_CASE_SENSITIVE, ZFS_CASE_INSENSITIVE, ZFS_CASE_MIXED } zfs_case_t; /* * Note: this struct must have the same layout in 32-bit and 64-bit, so * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit * kernel. Therefore, we add padding to it so that no "hidden" padding * is automatically added on 64-bit (but not on 32-bit). */ typedef struct zfs_cmd { char zc_name[MAXPATHLEN]; /* name of pool or dataset */ uint64_t zc_nvlist_src; /* really (char *) */ uint64_t zc_nvlist_src_size; uint64_t zc_nvlist_dst; /* really (char *) */ uint64_t zc_nvlist_dst_size; boolean_t zc_nvlist_dst_filled; /* put an nvlist in dst? */ int zc_pad2; /* * The following members are for legacy ioctls which haven't been * converted to the new method. */ uint64_t zc_history; /* really (char *) */ char zc_value[MAXPATHLEN * 2]; char zc_string[MAXNAMELEN]; uint64_t zc_guid; uint64_t zc_nvlist_conf; /* really (char *) */ uint64_t zc_nvlist_conf_size; uint64_t zc_cookie; uint64_t zc_objset_type; uint64_t zc_perm_action; uint64_t zc_history_len; uint64_t zc_history_offset; uint64_t zc_obj; uint64_t zc_iflags; /* internal to zfs(7fs) */ zfs_share_t zc_share; dmu_objset_stats_t zc_objset_stats; struct drr_begin zc_begin_record; zinject_record_t zc_inject_record; uint32_t zc_defer_destroy; uint32_t zc_flags; uint64_t zc_action_handle; int zc_cleanup_fd; uint8_t zc_simple; uint8_t zc_pad[3]; /* alignment */ uint64_t zc_sendobj; uint64_t zc_fromobj; uint64_t zc_createtxg; zfs_stat_t zc_stat; uint64_t zc_zoneid; } zfs_cmd_t; typedef struct zfs_useracct { char zu_domain[256]; uid_t zu_rid; uint32_t zu_pad; uint64_t zu_space; } zfs_useracct_t; #define ZFSDEV_MAX_MINOR (1 << 16) #define ZPOOL_EXPORT_AFTER_SPLIT 0x1 #ifdef _KERNEL struct objset; struct zfsvfs; typedef struct zfs_creat { nvlist_t *zct_zplprops; nvlist_t *zct_props; } zfs_creat_t; extern int zfs_secpolicy_snapshot_perms(const char *, cred_t *); extern int zfs_secpolicy_rename_perms(const char *, const char *, cred_t *); extern int zfs_secpolicy_destroy_perms(const char *, cred_t *); extern void zfs_unmount_snap(const char *); extern void zfs_destroy_unmount_origin(const char *); extern int getzfsvfs_impl(struct objset *, struct zfsvfs **); extern int getzfsvfs(const char *, struct zfsvfs **); enum zfsdev_state_type { ZST_ONEXIT, ZST_ZEVENT, ZST_ALL, }; /* * The zfsdev_state_t structure is managed as a singly-linked list * from which items are never deleted. This allows for lock-free * reading of the list so long as assignments to the zs_next and * reads from zs_minor are performed atomically. Empty items are * indicated by storing -1 into zs_minor. */ typedef struct zfsdev_state { struct zfsdev_state *zs_next; /* next zfsdev_state_t link */ minor_t zs_minor; /* made up minor number */ void *zs_onexit; /* onexit data */ void *zs_zevent; /* zevent data */ } zfsdev_state_t; extern void *zfsdev_get_state(minor_t minor, enum zfsdev_state_type which); extern int zfsdev_getminor(zfs_file_t *fp, minor_t *minorp); extern uint_t zfs_allow_log_key; #endif /* _KERNEL */ #ifdef __cplusplus } #endif #endif /* _SYS_ZFS_IOCTL_H */ diff --git a/sys/contrib/openzfs/include/sys/zio.h b/sys/contrib/openzfs/include/sys/zio.h index 545b9cf0c3c5..4037b429982b 100644 --- a/sys/contrib/openzfs/include/sys/zio.h +++ b/sys/contrib/openzfs/include/sys/zio.h @@ -1,726 +1,728 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. - * Copyright (c) 2012, 2020 by Delphix. All rights reserved. + * Copyright (c) 2012, 2024 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright 2016 Toomas Soome * Copyright (c) 2019, Allan Jude * Copyright (c) 2019, 2023, 2024, Klara Inc. * Copyright (c) 2019-2020, Michael Niewöhner */ #ifndef _ZIO_H #define _ZIO_H #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Embedded checksum */ #define ZEC_MAGIC 0x210da7ab10c7a11ULL typedef struct zio_eck { uint64_t zec_magic; /* for validation, endianness */ zio_cksum_t zec_cksum; /* 256-bit checksum */ } zio_eck_t; /* * Gang block headers are self-checksumming and contain an array * of block pointers. */ #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t)) / sizeof (blkptr_t)) #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t) - \ (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ sizeof (uint64_t)) typedef struct zio_gbh { blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; uint64_t zg_filler[SPA_GBH_FILLER]; zio_eck_t zg_tail; } zio_gbh_phys_t; enum zio_checksum { ZIO_CHECKSUM_INHERIT = 0, ZIO_CHECKSUM_ON, ZIO_CHECKSUM_OFF, ZIO_CHECKSUM_LABEL, ZIO_CHECKSUM_GANG_HEADER, ZIO_CHECKSUM_ZILOG, ZIO_CHECKSUM_FLETCHER_2, ZIO_CHECKSUM_FLETCHER_4, ZIO_CHECKSUM_SHA256, ZIO_CHECKSUM_ZILOG2, ZIO_CHECKSUM_NOPARITY, ZIO_CHECKSUM_SHA512, ZIO_CHECKSUM_SKEIN, ZIO_CHECKSUM_EDONR, ZIO_CHECKSUM_BLAKE3, ZIO_CHECKSUM_FUNCTIONS }; /* * The number of "legacy" compression functions which can be set on individual * objects. */ #define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON #define ZIO_CHECKSUM_MASK 0xffULL #define ZIO_CHECKSUM_VERIFY (1U << 8) #define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256 /* macros defining encryption lengths */ #define ZIO_OBJSET_MAC_LEN 32 #define ZIO_DATA_IV_LEN 12 #define ZIO_DATA_SALT_LEN 8 #define ZIO_DATA_MAC_LEN 16 /* * The number of "legacy" compression functions which can be set on individual * objects. */ #define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4 /* * The meaning of "compress = on" selected by the compression features enabled * on a given pool. */ #define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB #define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_ON #define BOOTFS_COMPRESS_VALID(compress) \ ((compress) == ZIO_COMPRESS_LZJB || \ (compress) == ZIO_COMPRESS_LZ4 || \ (compress) == ZIO_COMPRESS_GZIP_1 || \ (compress) == ZIO_COMPRESS_GZIP_2 || \ (compress) == ZIO_COMPRESS_GZIP_3 || \ (compress) == ZIO_COMPRESS_GZIP_4 || \ (compress) == ZIO_COMPRESS_GZIP_5 || \ (compress) == ZIO_COMPRESS_GZIP_6 || \ (compress) == ZIO_COMPRESS_GZIP_7 || \ (compress) == ZIO_COMPRESS_GZIP_8 || \ (compress) == ZIO_COMPRESS_GZIP_9 || \ (compress) == ZIO_COMPRESS_ZLE || \ (compress) == ZIO_COMPRESS_ZSTD || \ (compress) == ZIO_COMPRESS_ON || \ (compress) == ZIO_COMPRESS_OFF) #define ZIO_COMPRESS_ALGO(x) (x & SPA_COMPRESSMASK) #define ZIO_COMPRESS_LEVEL(x) ((x & ~SPA_COMPRESSMASK) >> SPA_COMPRESSBITS) #define ZIO_COMPRESS_RAW(type, level) (type | ((level) << SPA_COMPRESSBITS)) #define ZIO_COMPLEVEL_ZSTD(level) \ ZIO_COMPRESS_RAW(ZIO_COMPRESS_ZSTD, level) #define ZIO_FAILURE_MODE_WAIT 0 #define ZIO_FAILURE_MODE_CONTINUE 1 #define ZIO_FAILURE_MODE_PANIC 2 typedef enum zio_suspend_reason { ZIO_SUSPEND_NONE = 0, ZIO_SUSPEND_IOERR, ZIO_SUSPEND_MMP, } zio_suspend_reason_t; /* * This was originally an enum type. However, those are 32-bit and there is no * way to make a 64-bit enum type. Since we ran out of bits for flags, we were * forced to upgrade it to a uint64_t. */ typedef uint64_t zio_flag_t; /* * Flags inherited by gang, ddt, and vdev children, * and that must be equal for two zios to aggregate */ #define ZIO_FLAG_DONT_AGGREGATE (1ULL << 0) #define ZIO_FLAG_IO_REPAIR (1ULL << 1) #define ZIO_FLAG_SELF_HEAL (1ULL << 2) #define ZIO_FLAG_RESILVER (1ULL << 3) #define ZIO_FLAG_SCRUB (1ULL << 4) #define ZIO_FLAG_SCAN_THREAD (1ULL << 5) #define ZIO_FLAG_PHYSICAL (1ULL << 6) #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1) /* * Flags inherited by ddt, gang, and vdev children. */ #define ZIO_FLAG_CANFAIL (1ULL << 7) /* must be first for INHERIT */ #define ZIO_FLAG_SPECULATIVE (1ULL << 8) #define ZIO_FLAG_CONFIG_WRITER (1ULL << 9) #define ZIO_FLAG_DONT_RETRY (1ULL << 10) #define ZIO_FLAG_NODATA (1ULL << 12) #define ZIO_FLAG_INDUCE_DAMAGE (1ULL << 13) #define ZIO_FLAG_IO_ALLOCATING (1ULL << 14) #define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1) #define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1) /* * Flags inherited by vdev children. */ #define ZIO_FLAG_IO_RETRY (1ULL << 15) /* must be first for INHERIT */ #define ZIO_FLAG_PROBE (1ULL << 16) #define ZIO_FLAG_TRYHARD (1ULL << 17) #define ZIO_FLAG_OPTIONAL (1ULL << 18) #define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1) /* * Flags not inherited by any children. */ #define ZIO_FLAG_DONT_QUEUE (1ULL << 19) /* must be first for INHERIT */ #define ZIO_FLAG_DONT_PROPAGATE (1ULL << 20) #define ZIO_FLAG_IO_BYPASS (1ULL << 21) #define ZIO_FLAG_IO_REWRITE (1ULL << 22) #define ZIO_FLAG_RAW_COMPRESS (1ULL << 23) #define ZIO_FLAG_RAW_ENCRYPT (1ULL << 24) #define ZIO_FLAG_GANG_CHILD (1ULL << 25) #define ZIO_FLAG_DDT_CHILD (1ULL << 26) #define ZIO_FLAG_GODFATHER (1ULL << 27) #define ZIO_FLAG_NOPWRITE (1ULL << 28) #define ZIO_FLAG_REEXECUTED (1ULL << 29) #define ZIO_FLAG_DELEGATED (1ULL << 30) #define ZIO_ALLOCATOR_NONE (-1) #define ZIO_HAS_ALLOCATOR(zio) ((zio)->io_allocator != ZIO_ALLOCATOR_NONE) #define ZIO_FLAG_MUSTSUCCEED 0 #define ZIO_FLAG_RAW (ZIO_FLAG_RAW_COMPRESS | ZIO_FLAG_RAW_ENCRYPT) #define ZIO_DDT_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \ ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL) #define ZIO_GANG_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \ ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL) #define ZIO_VDEV_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \ ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL) #define ZIO_CHILD_BIT(x) (1U << (x)) #define ZIO_CHILD_BIT_IS_SET(val, x) ((val) & (1U << (x))) enum zio_child { ZIO_CHILD_VDEV = 0, ZIO_CHILD_GANG, ZIO_CHILD_DDT, ZIO_CHILD_LOGICAL, ZIO_CHILD_TYPES }; #define ZIO_CHILD_VDEV_BIT ZIO_CHILD_BIT(ZIO_CHILD_VDEV) #define ZIO_CHILD_GANG_BIT ZIO_CHILD_BIT(ZIO_CHILD_GANG) #define ZIO_CHILD_DDT_BIT ZIO_CHILD_BIT(ZIO_CHILD_DDT) #define ZIO_CHILD_LOGICAL_BIT ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL) #define ZIO_CHILD_ALL_BITS \ (ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT | \ ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT) enum zio_wait_type { ZIO_WAIT_READY = 0, ZIO_WAIT_DONE, ZIO_WAIT_TYPES }; typedef void zio_done_func_t(zio_t *zio); extern int zio_exclude_metadata; extern int zio_dva_throttle_enabled; extern const char *const zio_type_name[ZIO_TYPES]; /* * A bookmark is a four-tuple that uniquely * identifies any block in the pool. By convention, the meta-objset (MOS) * is objset 0, and the meta-dnode is object 0. This covers all blocks * except root blocks and ZIL blocks, which are defined as follows: * * Root blocks (objset_phys_t) are object 0, level -1: . * ZIL blocks are bookmarked . * dmu_sync()ed ZIL data blocks are bookmarked . * dnode visit bookmarks are . * * Note: this structure is called a bookmark because its original purpose * was to remember where to resume a pool-wide traverse. * * Note: this structure is passed between userland and the kernel, and is * stored on disk (by virtue of being incorporated into other on-disk * structures, e.g. dsl_scan_phys_t). * * If the head_errlog feature is enabled a different on-disk format for error * logs is used. This introduces the use of an error bookmark, a four-tuple * that uniquely identifies any error block * in the pool. The birth transaction group is used to track whether the block * has been overwritten by newer data or added to a snapshot since its marking * as an error. */ struct zbookmark_phys { uint64_t zb_objset; uint64_t zb_object; int64_t zb_level; uint64_t zb_blkid; }; struct zbookmark_err_phys { uint64_t zb_object; int64_t zb_level; uint64_t zb_blkid; uint64_t zb_birth; }; #define SET_BOOKMARK(zb, objset, object, level, blkid) \ { \ (zb)->zb_objset = objset; \ (zb)->zb_object = object; \ (zb)->zb_level = level; \ (zb)->zb_blkid = blkid; \ } #define ZB_DESTROYED_OBJSET (-1ULL) #define ZB_ROOT_OBJECT (0ULL) #define ZB_ROOT_LEVEL (-1LL) #define ZB_ROOT_BLKID (0ULL) #define ZB_ZIL_OBJECT (0ULL) #define ZB_ZIL_LEVEL (-2LL) #define ZB_DNODE_LEVEL (-3LL) #define ZB_DNODE_BLKID (0ULL) #define ZB_IS_ZERO(zb) \ ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \ (zb)->zb_level == 0 && (zb)->zb_blkid == 0) #define ZB_IS_ROOT(zb) \ ((zb)->zb_object == ZB_ROOT_OBJECT && \ (zb)->zb_level == ZB_ROOT_LEVEL && \ (zb)->zb_blkid == ZB_ROOT_BLKID) typedef struct zio_prop { enum zio_checksum zp_checksum; enum zio_compress zp_compress; uint8_t zp_complevel; uint8_t zp_level; uint8_t zp_copies; dmu_object_type_t zp_type; boolean_t zp_dedup; boolean_t zp_dedup_verify; boolean_t zp_nopwrite; boolean_t zp_brtwrite; boolean_t zp_encrypt; boolean_t zp_byteorder; uint8_t zp_salt[ZIO_DATA_SALT_LEN]; uint8_t zp_iv[ZIO_DATA_IV_LEN]; uint8_t zp_mac[ZIO_DATA_MAC_LEN]; uint32_t zp_zpl_smallblk; } zio_prop_t; typedef struct zio_cksum_report zio_cksum_report_t; typedef void zio_cksum_finish_f(zio_cksum_report_t *rep, const abd_t *good_data); typedef void zio_cksum_free_f(void *cbdata, size_t size); struct zio_bad_cksum; /* defined in zio_checksum.h */ struct dnode_phys; struct abd; struct zio_cksum_report { struct zio_cksum_report *zcr_next; nvlist_t *zcr_ereport; nvlist_t *zcr_detector; void *zcr_cbdata; size_t zcr_cbinfo; /* passed to zcr_free() */ uint64_t zcr_sector; uint64_t zcr_align; uint64_t zcr_length; zio_cksum_finish_f *zcr_finish; zio_cksum_free_f *zcr_free; /* internal use only */ struct zio_bad_cksum *zcr_ckinfo; /* information from failure */ }; typedef struct zio_vsd_ops { zio_done_func_t *vsd_free; } zio_vsd_ops_t; typedef struct zio_gang_node { zio_gbh_phys_t *gn_gbh; struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS]; } zio_gang_node_t; typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp, zio_gang_node_t *gn, struct abd *data, uint64_t offset); typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size); typedef struct zio_transform { struct abd *zt_orig_abd; uint64_t zt_orig_size; uint64_t zt_bufsize; zio_transform_func_t *zt_transform; struct zio_transform *zt_next; } zio_transform_t; typedef zio_t *zio_pipe_stage_t(zio_t *zio); /* * The io_reexecute flags are distinct from io_flags because the child must * be able to propagate them to the parent. The normal io_flags are local * to the zio, not protected by any lock, and not modifiable by children; * the reexecute flags are protected by io_lock, modifiable by children, * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set. */ #define ZIO_REEXECUTE_NOW 0x01 #define ZIO_REEXECUTE_SUSPEND 0x02 /* * The io_trim flags are used to specify the type of TRIM to perform. They * only apply to ZIO_TYPE_TRIM zios are distinct from io_flags. */ enum trim_flag { ZIO_TRIM_SECURE = 1U << 0, }; typedef struct zio_alloc_list { list_t zal_list; uint64_t zal_size; } zio_alloc_list_t; typedef struct zio_link { zio_t *zl_parent; zio_t *zl_child; list_node_t zl_parent_node; list_node_t zl_child_node; } zio_link_t; enum zio_qstate { ZIO_QS_NONE = 0, ZIO_QS_QUEUED, ZIO_QS_ACTIVE, }; struct zio { /* Core information about this I/O */ zbookmark_phys_t io_bookmark; zio_prop_t io_prop; zio_type_t io_type; enum zio_child io_child_type; enum trim_flag io_trim_flags; zio_priority_t io_priority; uint8_t io_reexecute; uint8_t io_state[ZIO_WAIT_TYPES]; uint64_t io_txg; spa_t *io_spa; blkptr_t *io_bp; blkptr_t *io_bp_override; blkptr_t io_bp_copy; list_t io_parent_list; list_t io_child_list; zio_t *io_logical; zio_transform_t *io_transform_stack; /* Callback info */ zio_done_func_t *io_ready; zio_done_func_t *io_children_ready; zio_done_func_t *io_done; void *io_private; int64_t io_prev_space_delta; /* DMU private */ blkptr_t io_bp_orig; /* io_lsize != io_orig_size iff this is a raw write */ uint64_t io_lsize; /* Data represented by this I/O */ struct abd *io_abd; struct abd *io_orig_abd; uint64_t io_size; uint64_t io_orig_size; /* Stuff for the vdev stack */ vdev_t *io_vd; void *io_vsd; const zio_vsd_ops_t *io_vsd_ops; metaslab_class_t *io_metaslab_class; /* dva throttle class */ enum zio_qstate io_queue_state; /* vdev queue state */ union { list_node_t l; avl_node_t a; } io_queue_node ____cacheline_aligned; /* allocator and vdev queues */ avl_node_t io_offset_node; /* vdev offset queues */ uint64_t io_offset; hrtime_t io_timestamp; /* submitted at */ hrtime_t io_queued_timestamp; hrtime_t io_target_timestamp; hrtime_t io_delta; /* vdev queue service delta */ hrtime_t io_delay; /* Device access time (disk or */ /* file). */ zio_alloc_list_t io_alloc_list; /* Internal pipeline state */ zio_flag_t io_flags; enum zio_stage io_stage; enum zio_stage io_pipeline; zio_flag_t io_orig_flags; enum zio_stage io_orig_stage; enum zio_stage io_orig_pipeline; enum zio_stage io_pipeline_trace; int io_error; int io_child_error[ZIO_CHILD_TYPES]; uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES]; uint64_t *io_stall; zio_t *io_gang_leader; zio_gang_node_t *io_gang_tree; void *io_executor; void *io_waiter; void *io_bio; kmutex_t io_lock; kcondvar_t io_cv; int io_allocator; /* FMA state */ zio_cksum_report_t *io_cksum_report; uint64_t io_ena; /* Taskq dispatching state */ taskq_ent_t io_tqent; /* write issue taskq selection, based upon sync thread */ taskq_t *io_wr_iss_tq; }; enum blk_verify_flag { BLK_VERIFY_ONLY, BLK_VERIFY_LOG, BLK_VERIFY_HALT }; enum blk_config_flag { BLK_CONFIG_HELD, // SCL_VDEV held for writer BLK_CONFIG_NEEDED, // SCL_VDEV should be obtained for reader BLK_CONFIG_SKIP, // skip checks which require SCL_VDEV }; extern int zio_bookmark_compare(const void *, const void *); extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *priv, zio_flag_t flags); extern zio_t *zio_root(spa_t *spa, zio_done_func_t *done, void *priv, zio_flag_t flags); extern void zio_destroy(zio_t *zio); extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, struct abd *data, uint64_t lsize, zio_done_func_t *done, void *priv, zio_priority_t priority, zio_flag_t flags, const zbookmark_phys_t *zb); extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *children_ready, zio_done_func_t *done, void *priv, zio_priority_t priority, zio_flag_t flags, const zbookmark_phys_t *zb); extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, struct abd *data, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, zio_flag_t flags, zbookmark_phys_t *zb); extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite, boolean_t brtwrite); extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp); extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *priv, zio_flag_t flags); extern zio_t *zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, zio_flag_t flags, enum trim_flag trim_flags); extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, struct abd *data, int checksum, zio_done_func_t *done, void *priv, zio_priority_t priority, zio_flag_t flags, boolean_t labels); extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, struct abd *data, int checksum, zio_done_func_t *done, void *priv, zio_priority_t priority, zio_flag_t flags, boolean_t labels); extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_flag_t flags); extern int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp, uint64_t size, boolean_t *slog); extern void zio_flush(zio_t *zio, vdev_t *vd); extern void zio_shrink(zio_t *zio, uint64_t size); extern int zio_wait(zio_t *zio); extern void zio_nowait(zio_t *zio); extern void zio_execute(void *zio); extern void zio_interrupt(void *zio); extern void zio_delay_init(zio_t *zio); extern void zio_delay_interrupt(zio_t *zio); extern void zio_deadman(zio_t *zio, const char *tag); extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **); extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **); extern zio_t *zio_unique_parent(zio_t *cio); extern void zio_add_child(zio_t *pio, zio_t *cio); extern void zio_add_child_first(zio_t *pio, zio_t *cio); extern void *zio_buf_alloc(size_t size); extern void zio_buf_free(void *buf, size_t size); extern void *zio_data_buf_alloc(size_t size); extern void zio_data_buf_free(void *buf, size_t size); extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size, uint64_t bufsize, zio_transform_func_t *transform); extern void zio_pop_transforms(zio_t *zio); extern void zio_resubmit_stage_async(void *); extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd, uint64_t offset, struct abd *data, uint64_t size, int type, zio_priority_t priority, zio_flag_t flags, zio_done_func_t *done, void *priv); extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, struct abd *data, uint64_t size, zio_type_t type, zio_priority_t priority, zio_flag_t flags, zio_done_func_t *done, void *priv); extern void zio_vdev_io_bypass(zio_t *zio); extern void zio_vdev_io_reissue(zio_t *zio); extern void zio_vdev_io_redone(zio_t *zio); extern void zio_change_priority(zio_t *pio, zio_priority_t priority); extern void zio_checksum_verified(zio_t *zio); extern int zio_worst_error(int e1, int e2); extern enum zio_checksum zio_checksum_select(enum zio_checksum child, enum zio_checksum parent); extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child, enum zio_checksum parent); extern enum zio_compress zio_compress_select(spa_t *spa, enum zio_compress child, enum zio_compress parent); extern uint8_t zio_complevel_select(spa_t *spa, enum zio_compress compress, uint8_t child, uint8_t parent); extern void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t); extern int zio_resume(spa_t *spa); extern void zio_resume_wait(spa_t *spa); extern boolean_t zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp, enum blk_config_flag blk_config, enum blk_verify_flag blk_verify); /* * Initial setup and teardown. */ extern void zio_init(void); extern void zio_fini(void); /* * Fault injection */ struct zinject_record; extern uint32_t zio_injection_enabled; extern int zio_inject_fault(char *name, int flags, int *id, struct zinject_record *record); extern int zio_inject_list_next(int *id, char *name, size_t buflen, struct zinject_record *record); extern int zio_clear_fault(int id); extern void zio_handle_panic_injection(spa_t *spa, const char *tag, uint64_t type); extern int zio_handle_decrypt_injection(spa_t *spa, const zbookmark_phys_t *zb, uint64_t type, int error); extern int zio_handle_fault_injection(zio_t *zio, int error); extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error); extern int zio_handle_device_injections(vdev_t *vd, zio_t *zio, int err1, int err2); extern int zio_handle_label_injection(zio_t *zio, int error); extern void zio_handle_ignored_writes(zio_t *zio); extern hrtime_t zio_handle_io_delay(zio_t *zio); +extern void zio_handle_import_delay(spa_t *spa, hrtime_t elapsed); +extern void zio_handle_export_delay(spa_t *spa, hrtime_t elapsed); /* * Checksum ereport functions */ extern int zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset, uint64_t length, struct zio_bad_cksum *info); extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report, const abd_t *good_data, const abd_t *bad_data, boolean_t drop_if_identical); extern void zfs_ereport_free_checksum(zio_cksum_report_t *report); /* If we have the good data in hand, this function can be used */ extern int zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd, const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset, uint64_t length, const abd_t *good_data, const abd_t *bad_data, struct zio_bad_cksum *info); void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr); extern void zfs_ereport_snapshot_post(const char *subclass, spa_t *spa, const char *name); /* Called from spa_sync(), but primarily an injection handler */ extern void spa_handle_ignored_writes(spa_t *spa); /* zbookmark_phys functions */ boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block); boolean_t zbookmark_subtree_tbd(const struct dnode_phys *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block); int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2); #ifdef __cplusplus } #endif #endif /* _ZIO_H */ diff --git a/sys/contrib/openzfs/man/man8/zfs-set.8 b/sys/contrib/openzfs/man/man8/zfs-set.8 index c01bcc643e5d..8cc19caf3f00 100644 --- a/sys/contrib/openzfs/man/man8/zfs-set.8 +++ b/sys/contrib/openzfs/man/man8/zfs-set.8 @@ -1,316 +1,325 @@ .\" .\" CDDL HEADER START .\" .\" The contents of this file are subject to the terms of the .\" Common Development and Distribution License (the "License"). .\" You may not use this file except in compliance with the License. .\" .\" You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE .\" or https://opensource.org/licenses/CDDL-1.0. .\" See the License for the specific language governing permissions .\" and limitations under the License. .\" .\" When distributing Covered Code, include this CDDL HEADER in each .\" file and include the License file at usr/src/OPENSOLARIS.LICENSE. .\" If applicable, add the following below this CDDL HEADER, with the .\" fields enclosed by brackets "[]" replaced with your own identifying .\" information: Portions Copyright [yyyy] [name of copyright owner] .\" .\" CDDL HEADER END .\" .\" Copyright (c) 2009 Sun Microsystems, Inc. All Rights Reserved. .\" Copyright 2011 Joshua M. Clulow .\" Copyright (c) 2011, 2019 by Delphix. All rights reserved. .\" Copyright (c) 2013 by Saso Kiselkov. All rights reserved. .\" Copyright (c) 2014, Joyent, Inc. All rights reserved. .\" Copyright (c) 2014 by Adam Stevko. All rights reserved. .\" Copyright (c) 2014 Integros [integros.com] .\" Copyright 2019 Richard Laager. All rights reserved. .\" Copyright 2018 Nexenta Systems, Inc. .\" Copyright 2019 Joyent, Inc. .\" -.Dd March 16, 2022 +.Dd April 20, 2024 .Dt ZFS-SET 8 .Os . .Sh NAME .Nm zfs-set .Nd set properties on ZFS datasets .Sh SYNOPSIS .Nm zfs .Cm set .Op Fl u .Ar property Ns = Ns Ar value Oo Ar property Ns = Ns Ar value Oc Ns … .Ar filesystem Ns | Ns Ar volume Ns | Ns Ar snapshot Ns … .Nm zfs .Cm get .Op Fl r Ns | Ns Fl d Ar depth .Op Fl Hp .Oo Fl o Ar field Ns Oo , Ns Ar field Oc Ns … Oc .Oo Fl s Ar source Ns Oo , Ns Ar source Oc Ns … Oc .Oo Fl t Ar type Ns Oo , Ns Ar type Oc Ns … Oc .Cm all Ns | Ns Ar property Ns Oo , Ns Ar property Oc Ns … .Oo Ar filesystem Ns | Ns Ar volume Ns | Ns Ar snapshot Ns | Ns Ar bookmark Oc Ns … .Nm zfs .Cm inherit .Op Fl rS .Ar property Ar filesystem Ns | Ns Ar volume Ns | Ns Ar snapshot Ns … . .Sh DESCRIPTION .Bl -tag -width "" .It Xo .Nm zfs .Cm set .Op Fl u .Ar property Ns = Ns Ar value Oo Ar property Ns = Ns Ar value Oc Ns … .Ar filesystem Ns | Ns Ar volume Ns | Ns Ar snapshot Ns … .Xc Only some properties can be edited. See .Xr zfsprops 7 for more information on what properties can be set and acceptable values. Numeric values can be specified as exact values, or in a human-readable form with a suffix of .Sy B , K , M , G , T , P , E , Z .Po for bytes, kilobytes, megabytes, gigabytes, terabytes, petabytes, exabytes, or zettabytes, respectively .Pc . User properties can be set on snapshots. For more information, see the .Em User Properties section of .Xr zfsprops 7 . .Bl -tag -width "-u" .It Fl u Update mountpoint, sharenfs, sharesmb property but do not mount or share the dataset. .El .It Xo .Nm zfs .Cm get .Op Fl r Ns | Ns Fl d Ar depth .Op Fl Hp .Oo Fl o Ar field Ns Oo , Ns Ar field Oc Ns … Oc .Oo Fl s Ar source Ns Oo , Ns Ar source Oc Ns … Oc .Oo Fl t Ar type Ns Oo , Ns Ar type Oc Ns … Oc .Cm all Ns | Ns Ar property Ns Oo , Ns Ar property Oc Ns … .Oo Ar filesystem Ns | Ns Ar volume Ns | Ns Ar snapshot Ns | Ns Ar bookmark Oc Ns … .Xc Displays properties for the given datasets. If no datasets are specified, then the command displays properties for all datasets on the system. For each property, the following columns are displayed: .Bl -tag -compact -offset 4n -width "property" .It Sy name Dataset name .It Sy property Property name .It Sy value Property value .It Sy source Property source .Sy local , default , inherited , temporary , received , No or Sy - Pq none . .El .Pp All columns are displayed by default, though this can be controlled by using the .Fl o option. This command takes a comma-separated list of properties as described in the .Sx Native Properties and .Sx User Properties sections of .Xr zfsprops 7 . .Pp The value .Sy all can be used to display all properties that apply to the given dataset's type .Pq Sy filesystem , volume , snapshot , No or Sy bookmark . .Bl -tag -width "-s source" .It Fl H Display output in a form more easily parsed by scripts. Any headers are omitted, and fields are explicitly separated by a single tab instead of an arbitrary amount of space. .It Fl d Ar depth Recursively display any children of the dataset, limiting the recursion to .Ar depth . A depth of .Sy 1 will display only the dataset and its direct children. .It Fl o Ar field A comma-separated list of columns to display, defaults to .Sy name , Ns Sy property , Ns Sy value , Ns Sy source . .It Fl p Display numbers in parsable .Pq exact values. .It Fl r Recursively display properties for any children. .It Fl s Ar source A comma-separated list of sources to display. Those properties coming from a source other than those in this list are ignored. Each source must be one of the following: .Sy local , default , inherited , temporary , received , No or Sy none . The default value is all sources. .It Fl t Ar type A comma-separated list of types to display, where .Ar type is one of .Sy filesystem , snapshot , volume , bookmark , No or Sy all . +.Sy fs , +.Sy snap , +or +.Sy vol +can be used as aliases for +.Sy filesystem , +.Sy snapshot , +or +.Sy volume . .El .It Xo .Nm zfs .Cm inherit .Op Fl rS .Ar property Ar filesystem Ns | Ns Ar volume Ns | Ns Ar snapshot Ns … .Xc Clears the specified property, causing it to be inherited from an ancestor, restored to default if no ancestor has the property set, or with the .Fl S option reverted to the received value if one exists. See .Xr zfsprops 7 for a listing of default values, and details on which properties can be inherited. .Bl -tag -width "-r" .It Fl r Recursively inherit the given property for all children. .It Fl S Revert the property to the received value, if one exists; otherwise, for non-inheritable properties, to the default; otherwise, operate as if the .Fl S option was not specified. .El .El . .Sh EXAMPLES .\" These are, respectively, examples 1, 4, 6, 7, 11, 14, 16 from zfs.8 .\" Make sure to update them bidirectionally .Ss Example 1 : No Creating a ZFS File System Hierarchy The following commands create a file system named .Ar pool/home and a file system named .Ar pool/home/bob . The mount point .Pa /export/home is set for the parent file system, and is automatically inherited by the child file system. .Dl # Nm zfs Cm create Ar pool/home .Dl # Nm zfs Cm set Sy mountpoint Ns = Ns Ar /export/home pool/home .Dl # Nm zfs Cm create Ar pool/home/bob . .Ss Example 2 : No Disabling and Enabling File System Compression The following command disables the .Sy compression property for all file systems under .Ar pool/home . The next command explicitly enables .Sy compression for .Ar pool/home/anne . .Dl # Nm zfs Cm set Sy compression Ns = Ns Sy off Ar pool/home .Dl # Nm zfs Cm set Sy compression Ns = Ns Sy on Ar pool/home/anne . .Ss Example 3 : No Setting a Quota on a ZFS File System The following command sets a quota of 50 Gbytes for .Ar pool/home/bob : .Dl # Nm zfs Cm set Sy quota Ns = Ns Ar 50G pool/home/bob . .Ss Example 4 : No Listing ZFS Properties The following command lists all properties for .Ar pool/home/bob : .Bd -literal -compact -offset Ds .No # Nm zfs Cm get Sy all Ar pool/home/bob NAME PROPERTY VALUE SOURCE pool/home/bob type filesystem - pool/home/bob creation Tue Jul 21 15:53 2009 - pool/home/bob used 21K - pool/home/bob available 20.0G - pool/home/bob referenced 21K - pool/home/bob compressratio 1.00x - pool/home/bob mounted yes - pool/home/bob quota 20G local pool/home/bob reservation none default pool/home/bob recordsize 128K default pool/home/bob mountpoint /pool/home/bob default pool/home/bob sharenfs off default pool/home/bob checksum on default pool/home/bob compression on local pool/home/bob atime on default pool/home/bob devices on default pool/home/bob exec on default pool/home/bob setuid on default pool/home/bob readonly off default pool/home/bob zoned off default pool/home/bob snapdir hidden default pool/home/bob acltype off default pool/home/bob aclmode discard default pool/home/bob aclinherit restricted default pool/home/bob canmount on default pool/home/bob xattr on default pool/home/bob copies 1 default pool/home/bob version 4 - pool/home/bob utf8only off - pool/home/bob normalization none - pool/home/bob casesensitivity sensitive - pool/home/bob vscan off default pool/home/bob nbmand off default pool/home/bob sharesmb off default pool/home/bob refquota none default pool/home/bob refreservation none default pool/home/bob primarycache all default pool/home/bob secondarycache all default pool/home/bob usedbysnapshots 0 - pool/home/bob usedbydataset 21K - pool/home/bob usedbychildren 0 - pool/home/bob usedbyrefreservation 0 - .Ed .Pp The following command gets a single property value: .Bd -literal -compact -offset Ds .No # Nm zfs Cm get Fl H o Sy value compression Ar pool/home/bob on .Ed .Pp The following command lists all properties with local settings for .Ar pool/home/bob : .Bd -literal -compact -offset Ds .No # Nm zfs Cm get Fl r s Sy local Fl o Sy name , Ns Sy property , Ns Sy value all Ar pool/home/bob NAME PROPERTY VALUE pool/home/bob quota 20G pool/home/bob compression on .Ed . .Ss Example 5 : No Inheriting ZFS Properties The following command causes .Ar pool/home/bob No and Ar pool/home/anne to inherit the .Sy checksum property from their parent. .Dl # Nm zfs Cm inherit Sy checksum Ar pool/home/bob pool/home/anne . .Ss Example 6 : No Setting User Properties The following example sets the user-defined .Ar com.example : Ns Ar department property for a dataset: .Dl # Nm zfs Cm set Ar com.example : Ns Ar department Ns = Ns Ar 12345 tank/accounting . .Ss Example 7 : No Setting sharenfs Property Options on a ZFS File System The following commands show how to set .Sy sharenfs property options to enable read-write access for a set of IP addresses and to enable root access for system .Qq neo on the .Ar tank/home file system: .Dl # Nm zfs Cm set Sy sharenfs Ns = Ns ' Ns Ar rw Ns =@123.123.0.0/16:[::1],root= Ns Ar neo Ns ' tank/home .Pp If you are using DNS for host name resolution, specify the fully-qualified hostname. . .Sh SEE ALSO .Xr zfsprops 7 , .Xr zfs-list 8 diff --git a/sys/contrib/openzfs/man/man8/zinject.8 b/sys/contrib/openzfs/man/man8/zinject.8 index f67b5e378dc3..ad9e7a42bfac 100644 --- a/sys/contrib/openzfs/man/man8/zinject.8 +++ b/sys/contrib/openzfs/man/man8/zinject.8 @@ -1,301 +1,309 @@ .\" .\" CDDL HEADER START .\" .\" The contents of this file are subject to the terms of the .\" Common Development and Distribution License (the "License"). .\" You may not use this file except in compliance with the License. .\" .\" You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE .\" or https://opensource.org/licenses/CDDL-1.0. .\" See the License for the specific language governing permissions .\" and limitations under the License. .\" .\" When distributing Covered Code, include this CDDL HEADER in each .\" file and include the License file at usr/src/OPENSOLARIS.LICENSE. .\" If applicable, add the following below this CDDL HEADER, with the .\" fields enclosed by brackets "[]" replaced with your own identifying .\" information: Portions Copyright [yyyy] [name of copyright owner] .\" .\" CDDL HEADER END .\" .\" Copyright 2013 Darik Horn . All rights reserved. .\" Copyright (c) 2024, Klara Inc. .\" .\" lint-ok: WARNING: sections out of conventional order: Sh SYNOPSIS .\" .Dd April 4, 2024 .Dt ZINJECT 8 .Os . .Sh NAME .Nm zinject .Nd ZFS Fault Injector .Sh DESCRIPTION .Nm creates artificial problems in a ZFS pool by simulating data corruption or device failures. This program is dangerous. . .Sh SYNOPSIS .Bl -tag -width Ds .It Xo .Nm zinject .Xc List injection records. . .It Xo .Nm zinject .Fl b Ar objset : Ns Ar object : Ns Ar level : Ns Ar start : Ns Ar end .Op Fl f Ar frequency .Fl amu .Op pool .Xc Force an error into the pool at a bookmark. . .It Xo .Nm zinject .Fl c Ar id Ns | Ns Sy all .Xc Cancel injection records. . .It Xo .Nm zinject .Fl d Ar vdev .Fl A Sy degrade Ns | Ns Sy fault .Ar pool .Xc Force a vdev into the DEGRADED or FAULTED state. . .It Xo .Nm zinject .Fl d Ar vdev .Fl D Ar latency : Ns Ar lanes .Op Fl T Ar read|write .Ar pool .Xc Add an artificial delay to I/O requests on a particular device, such that the requests take a minimum of .Ar latency milliseconds to complete. Each delay has an associated number of .Ar lanes which defines the number of concurrent I/O requests that can be processed. .Pp For example, with a single lane delay of 10 ms .No (\& Ns Fl D Ar 10 : Ns Ar 1 ) , the device will only be able to service a single I/O request at a time with each request taking 10 ms to complete. So, if only a single request is submitted every 10 ms, the average latency will be 10 ms; but if more than one request is submitted every 10 ms, the average latency will be more than 10 ms. .Pp Similarly, if a delay of 10 ms is specified to have two lanes .No (\& Ns Fl D Ar 10 : Ns Ar 2 ) , then the device will be able to service two requests at a time, each with a minimum latency of 10 ms. So, if two requests are submitted every 10 ms, then the average latency will be 10 ms; but if more than two requests are submitted every 10 ms, the average latency will be more than 10 ms. .Pp Also note, these delays are additive. So two invocations of .Fl D Ar 10 : Ns Ar 1 are roughly equivalent to a single invocation of .Fl D Ar 10 : Ns Ar 2 . This also means, that one can specify multiple lanes with differing target latencies. For example, an invocation of .Fl D Ar 10 : Ns Ar 1 followed by .Fl D Ar 25 : Ns Ar 2 will create 3 lanes on the device: one lane with a latency of 10 ms and two lanes with a 25 ms latency. . .It Xo .Nm zinject .Fl d Ar vdev .Op Fl e Ar device_error .Op Fl L Ar label_error .Op Fl T Ar failure .Op Fl f Ar frequency .Op Fl F .Ar pool .Xc Force a vdev error. . .It Xo .Nm zinject +.Fl i Ar seconds +.Ar pool +.Xc +Add an artificial delay during the future import of a pool. +This injector is automatically cleared after the import is finished. +. +.It Xo +.Nm zinject .Fl I .Op Fl s Ar seconds Ns | Ns Fl g Ar txgs .Ar pool .Xc Simulate a hardware failure that fails to honor a cache flush. . .It Xo .Nm zinject .Fl p Ar function .Ar pool .Xc Panic inside the specified function. . .It Xo .Nm zinject .Fl t Sy data .Fl C Ar dvas .Op Fl e Ar device_error .Op Fl f Ar frequency .Op Fl l Ar level .Op Fl r Ar range .Op Fl amq .Ar path .Xc Force an error into the contents of a file. . .It Xo .Nm zinject .Fl t Sy dnode .Fl C Ar dvas .Op Fl e Ar device_error .Op Fl f Ar frequency .Op Fl l Ar level .Op Fl amq .Ar path .Xc Force an error into the metadnode for a file or directory. . .It Xo .Nm zinject .Fl t Ar mos_type .Fl C Ar dvas .Op Fl e Ar device_error .Op Fl f Ar frequency .Op Fl l Ar level .Op Fl r Ar range .Op Fl amqu .Ar pool .Xc Force an error into the MOS of a pool. .El .Sh OPTIONS .Bl -tag -width "-C dvas" .It Fl a Flush the ARC before injection. .It Fl b Ar objset : Ns Ar object : Ns Ar level : Ns Ar start : Ns Ar end Force an error into the pool at this bookmark tuple. Each number is in hexadecimal, and only one block can be specified. .It Fl C Ar dvas Inject the given error only into specific DVAs. The mask should be specified as a list of 0-indexed DVAs separated by commas .No (e.g . Ar 0,2 Ns No ). This option is not applicable to logical data errors such as .Sy decompress and .Sy decrypt . .It Fl d Ar vdev A vdev specified by path or GUID. .It Fl e Ar device_error Specify .Bl -tag -compact -width "decompress" .It Sy checksum for an ECKSUM error, .It Sy decompress for a data decompression error, .It Sy decrypt for a data decryption error, .It Sy corrupt to flip a bit in the data after a read, .It Sy dtl for an ECHILD error, .It Sy io for an EIO error where reopening the device will succeed, .It Sy nxio for an ENXIO error where reopening the device will fail, or .It Sy noop to drop the IO without executing it, and return success. .El .Pp For EIO and ENXIO, the "failed" reads or writes still occur. The probe simply sets the error value reported by the I/O pipeline so it appears the read or write failed. Decryption errors only currently work with file data. .It Fl f Ar frequency Only inject errors a fraction of the time. Expressed as a real number percentage between .Sy 0.0001 and .Sy 100 . .It Fl F Fail faster. Do fewer checks. .It Fl f Ar txgs Run for this many transaction groups before reporting failure. .It Fl h Print the usage message. .It Fl l Ar level Inject an error at a particular block level. The default is .Sy 0 . .It Fl L Ar label_error Set the label error region to one of .Sy nvlist , .Sy pad1 , .Sy pad2 , or .Sy uber . .It Fl m Automatically remount the underlying filesystem. .It Fl q Quiet mode. Only print the handler number added. .It Fl r Ar range Inject an error over a particular logical range of an object, which will be translated to the appropriate blkid range according to the object's properties. .It Fl s Ar seconds Run for this many seconds before reporting failure. .It Fl T Ar failure Set the failure type to one of .Sy all , .Sy ioctl , .Sy claim , .Sy free , .Sy read , or .Sy write . .It Fl t Ar mos_type Set this to .Bl -tag -compact -width "spacemap" .It Sy mos for any data in the MOS, .It Sy mosdir for an object directory, .It Sy config for the pool configuration, .It Sy bpobj for the block pointer list, .It Sy spacemap for the space map, .It Sy metaslab for the metaslab, or .It Sy errlog for the persistent error log. .El .It Fl u Unload the pool after injection. .El . .Sh ENVIRONMENT VARIABLES .Bl -tag -width "ZF" .It Ev ZFS_HOSTID Run .Nm in debug mode. .El . .Sh SEE ALSO .Xr zfs 8 , .Xr zpool 8 diff --git a/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c b/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c index d3255dcbc0f7..cee7410c8833 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c @@ -1,1273 +1,1301 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2014 by Chunwei Chen. All rights reserved. * Copyright (c) 2019 by Delphix. All rights reserved. * Copyright (c) 2023, 2024, Klara Inc. */ /* * See abd.c for a general overview of the arc buffered data (ABD). * * Linear buffers act exactly like normal buffers and are always mapped into the * kernel's virtual memory space, while scattered ABD data chunks are allocated * as physical pages and then mapped in only while they are actually being * accessed through one of the abd_* library functions. Using scattered ABDs * provides several benefits: * * (1) They avoid use of kmem_*, preventing performance problems where running * kmem_reap on very large memory systems never finishes and causes * constant TLB shootdowns. * * (2) Fragmentation is less of an issue since when we are at the limit of * allocatable space, we won't have to search around for a long free * hole in the VA space for large ARC allocations. Each chunk is mapped in * individually, so even if we are using HIGHMEM (see next point) we * wouldn't need to worry about finding a contiguous address range. * * (3) If we are not using HIGHMEM, then all physical memory is always * mapped into the kernel's address space, so we also avoid the map / * unmap costs on each ABD access. * * If we are not using HIGHMEM, scattered buffers which have only one chunk * can be treated as linear buffers, because they are contiguous in the * kernel's virtual address space. See abd_alloc_chunks() for details. */ #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #include #endif #ifdef _KERNEL #if defined(MAX_ORDER) #define ABD_MAX_ORDER (MAX_ORDER) #elif defined(MAX_PAGE_ORDER) #define ABD_MAX_ORDER (MAX_PAGE_ORDER) #endif #else #define ABD_MAX_ORDER (1) #endif typedef struct abd_stats { kstat_named_t abdstat_struct_size; kstat_named_t abdstat_linear_cnt; kstat_named_t abdstat_linear_data_size; kstat_named_t abdstat_scatter_cnt; kstat_named_t abdstat_scatter_data_size; kstat_named_t abdstat_scatter_chunk_waste; kstat_named_t abdstat_scatter_orders[ABD_MAX_ORDER]; kstat_named_t abdstat_scatter_page_multi_chunk; kstat_named_t abdstat_scatter_page_multi_zone; kstat_named_t abdstat_scatter_page_alloc_retry; kstat_named_t abdstat_scatter_sg_table_retry; } abd_stats_t; static abd_stats_t abd_stats = { /* Amount of memory occupied by all of the abd_t struct allocations */ { "struct_size", KSTAT_DATA_UINT64 }, /* * The number of linear ABDs which are currently allocated, excluding * ABDs which don't own their data (for instance the ones which were * allocated through abd_get_offset() and abd_get_from_buf()). If an * ABD takes ownership of its buf then it will become tracked. */ { "linear_cnt", KSTAT_DATA_UINT64 }, /* Amount of data stored in all linear ABDs tracked by linear_cnt */ { "linear_data_size", KSTAT_DATA_UINT64 }, /* * The number of scatter ABDs which are currently allocated, excluding * ABDs which don't own their data (for instance the ones which were * allocated through abd_get_offset()). */ { "scatter_cnt", KSTAT_DATA_UINT64 }, /* Amount of data stored in all scatter ABDs tracked by scatter_cnt */ { "scatter_data_size", KSTAT_DATA_UINT64 }, /* * The amount of space wasted at the end of the last chunk across all * scatter ABDs tracked by scatter_cnt. */ { "scatter_chunk_waste", KSTAT_DATA_UINT64 }, /* * The number of compound allocations of a given order. These * allocations are spread over all currently allocated ABDs, and * act as a measure of memory fragmentation. */ { { "scatter_order_N", KSTAT_DATA_UINT64 } }, /* * The number of scatter ABDs which contain multiple chunks. * ABDs are preferentially allocated from the minimum number of * contiguous multi-page chunks, a single chunk is optimal. */ { "scatter_page_multi_chunk", KSTAT_DATA_UINT64 }, /* * The number of scatter ABDs which are split across memory zones. * ABDs are preferentially allocated using pages from a single zone. */ { "scatter_page_multi_zone", KSTAT_DATA_UINT64 }, /* * The total number of retries encountered when attempting to * allocate the pages to populate the scatter ABD. */ { "scatter_page_alloc_retry", KSTAT_DATA_UINT64 }, /* * The total number of retries encountered when attempting to * allocate the sg table for an ABD. */ { "scatter_sg_table_retry", KSTAT_DATA_UINT64 }, }; static struct { wmsum_t abdstat_struct_size; wmsum_t abdstat_linear_cnt; wmsum_t abdstat_linear_data_size; wmsum_t abdstat_scatter_cnt; wmsum_t abdstat_scatter_data_size; wmsum_t abdstat_scatter_chunk_waste; wmsum_t abdstat_scatter_orders[ABD_MAX_ORDER]; wmsum_t abdstat_scatter_page_multi_chunk; wmsum_t abdstat_scatter_page_multi_zone; wmsum_t abdstat_scatter_page_alloc_retry; wmsum_t abdstat_scatter_sg_table_retry; } abd_sums; #define abd_for_each_sg(abd, sg, n, i) \ for_each_sg(ABD_SCATTER(abd).abd_sgl, sg, n, i) /* * zfs_abd_scatter_min_size is the minimum allocation size to use scatter * ABD's. Smaller allocations will use linear ABD's which uses * zio_[data_]buf_alloc(). * * Scatter ABD's use at least one page each, so sub-page allocations waste * some space when allocated as scatter (e.g. 2KB scatter allocation wastes * half of each page). Using linear ABD's for small allocations means that * they will be put on slabs which contain many allocations. This can * improve memory efficiency, but it also makes it much harder for ARC * evictions to actually free pages, because all the buffers on one slab need * to be freed in order for the slab (and underlying pages) to be freed. * Typically, 512B and 1KB kmem caches have 16 buffers per slab, so it's * possible for them to actually waste more memory than scatter (one page per * buf = wasting 3/4 or 7/8th; one buf per slab = wasting 15/16th). * * Spill blocks are typically 512B and are heavily used on systems running * selinux with the default dnode size and the `xattr=sa` property set. * * By default we use linear allocations for 512B and 1KB, and scatter * allocations for larger (1.5KB and up). */ static int zfs_abd_scatter_min_size = 512 * 3; /* * We use a scattered SPA_MAXBLOCKSIZE sized ABD whose pages are * just a single zero'd page. This allows us to conserve memory by * only using a single zero page for the scatterlist. */ abd_t *abd_zero_scatter = NULL; struct page; /* * _KERNEL - Will point to ZERO_PAGE if it is available or it will be * an allocated zero'd PAGESIZE buffer. * Userspace - Will be an allocated zero'ed PAGESIZE buffer. * * abd_zero_page is assigned to each of the pages of abd_zero_scatter. */ static struct page *abd_zero_page = NULL; static kmem_cache_t *abd_cache = NULL; static kstat_t *abd_ksp; static uint_t abd_chunkcnt_for_bytes(size_t size) { return (P2ROUNDUP(size, PAGESIZE) / PAGESIZE); } abd_t * abd_alloc_struct_impl(size_t size) { /* * In Linux we do not use the size passed in during ABD * allocation, so we just ignore it. */ (void) size; abd_t *abd = kmem_cache_alloc(abd_cache, KM_PUSHPAGE); ASSERT3P(abd, !=, NULL); ABDSTAT_INCR(abdstat_struct_size, sizeof (abd_t)); return (abd); } void abd_free_struct_impl(abd_t *abd) { kmem_cache_free(abd_cache, abd); ABDSTAT_INCR(abdstat_struct_size, -(int)sizeof (abd_t)); } #ifdef _KERNEL static unsigned zfs_abd_scatter_max_order = ABD_MAX_ORDER - 1; /* * Mark zfs data pages so they can be excluded from kernel crash dumps */ #ifdef _LP64 #define ABD_FILE_CACHE_PAGE 0x2F5ABDF11ECAC4E static inline void abd_mark_zfs_page(struct page *page) { get_page(page); SetPagePrivate(page); set_page_private(page, ABD_FILE_CACHE_PAGE); } static inline void abd_unmark_zfs_page(struct page *page) { set_page_private(page, 0UL); ClearPagePrivate(page); put_page(page); } #else #define abd_mark_zfs_page(page) #define abd_unmark_zfs_page(page) #endif /* _LP64 */ #ifndef CONFIG_HIGHMEM #ifndef __GFP_RECLAIM #define __GFP_RECLAIM __GFP_WAIT #endif /* * The goal is to minimize fragmentation by preferentially populating ABDs * with higher order compound pages from a single zone. Allocation size is * progressively decreased until it can be satisfied without performing * reclaim or compaction. When necessary this function will degenerate to * allocating individual pages and allowing reclaim to satisfy allocations. */ void abd_alloc_chunks(abd_t *abd, size_t size) { struct list_head pages; struct sg_table table; struct scatterlist *sg; struct page *page, *tmp_page = NULL; gfp_t gfp = __GFP_NOWARN | GFP_NOIO; gfp_t gfp_comp = (gfp | __GFP_NORETRY | __GFP_COMP) & ~__GFP_RECLAIM; unsigned int max_order = MIN(zfs_abd_scatter_max_order, ABD_MAX_ORDER - 1); unsigned int nr_pages = abd_chunkcnt_for_bytes(size); unsigned int chunks = 0, zones = 0; size_t remaining_size; int nid = NUMA_NO_NODE; unsigned int alloc_pages = 0; INIT_LIST_HEAD(&pages); ASSERT3U(alloc_pages, <, nr_pages); while (alloc_pages < nr_pages) { unsigned int chunk_pages; unsigned int order; order = MIN(highbit64(nr_pages - alloc_pages) - 1, max_order); chunk_pages = (1U << order); page = alloc_pages_node(nid, order ? gfp_comp : gfp, order); if (page == NULL) { if (order == 0) { ABDSTAT_BUMP(abdstat_scatter_page_alloc_retry); schedule_timeout_interruptible(1); } else { max_order = MAX(0, order - 1); } continue; } list_add_tail(&page->lru, &pages); if ((nid != NUMA_NO_NODE) && (page_to_nid(page) != nid)) zones++; nid = page_to_nid(page); ABDSTAT_BUMP(abdstat_scatter_orders[order]); chunks++; alloc_pages += chunk_pages; } ASSERT3S(alloc_pages, ==, nr_pages); while (sg_alloc_table(&table, chunks, gfp)) { ABDSTAT_BUMP(abdstat_scatter_sg_table_retry); schedule_timeout_interruptible(1); } sg = table.sgl; remaining_size = size; list_for_each_entry_safe(page, tmp_page, &pages, lru) { size_t sg_size = MIN(PAGESIZE << compound_order(page), remaining_size); sg_set_page(sg, page, sg_size, 0); abd_mark_zfs_page(page); remaining_size -= sg_size; sg = sg_next(sg); list_del(&page->lru); } /* * These conditions ensure that a possible transformation to a linear * ABD would be valid. */ ASSERT(!PageHighMem(sg_page(table.sgl))); ASSERT0(ABD_SCATTER(abd).abd_offset); if (table.nents == 1) { /* * Since there is only one entry, this ABD can be represented * as a linear buffer. All single-page (4K) ABD's can be * represented this way. Some multi-page ABD's can also be * represented this way, if we were able to allocate a single * "chunk" (higher-order "page" which represents a power-of-2 * series of physically-contiguous pages). This is often the * case for 2-page (8K) ABD's. * * Representing a single-entry scatter ABD as a linear ABD * has the performance advantage of avoiding the copy (and * allocation) in abd_borrow_buf_copy / abd_return_buf_copy. * A performance increase of around 5% has been observed for * ARC-cached reads (of small blocks which can take advantage * of this). * * Note that this optimization is only possible because the * pages are always mapped into the kernel's address space. * This is not the case for highmem pages, so the * optimization can not be made there. */ abd->abd_flags |= ABD_FLAG_LINEAR; abd->abd_flags |= ABD_FLAG_LINEAR_PAGE; abd->abd_u.abd_linear.abd_sgl = table.sgl; ABD_LINEAR_BUF(abd) = page_address(sg_page(table.sgl)); } else if (table.nents > 1) { ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk); abd->abd_flags |= ABD_FLAG_MULTI_CHUNK; if (zones) { ABDSTAT_BUMP(abdstat_scatter_page_multi_zone); abd->abd_flags |= ABD_FLAG_MULTI_ZONE; } ABD_SCATTER(abd).abd_sgl = table.sgl; ABD_SCATTER(abd).abd_nents = table.nents; } } #else /* * Allocate N individual pages to construct a scatter ABD. This function * makes no attempt to request contiguous pages and requires the minimal * number of kernel interfaces. It's designed for maximum compatibility. */ void abd_alloc_chunks(abd_t *abd, size_t size) { struct scatterlist *sg = NULL; struct sg_table table; struct page *page; gfp_t gfp = __GFP_NOWARN | GFP_NOIO; int nr_pages = abd_chunkcnt_for_bytes(size); int i = 0; while (sg_alloc_table(&table, nr_pages, gfp)) { ABDSTAT_BUMP(abdstat_scatter_sg_table_retry); schedule_timeout_interruptible(1); } ASSERT3U(table.nents, ==, nr_pages); ABD_SCATTER(abd).abd_sgl = table.sgl; ABD_SCATTER(abd).abd_nents = nr_pages; abd_for_each_sg(abd, sg, nr_pages, i) { while ((page = __page_cache_alloc(gfp)) == NULL) { ABDSTAT_BUMP(abdstat_scatter_page_alloc_retry); schedule_timeout_interruptible(1); } ABDSTAT_BUMP(abdstat_scatter_orders[0]); sg_set_page(sg, page, PAGESIZE, 0); abd_mark_zfs_page(page); } if (nr_pages > 1) { ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk); abd->abd_flags |= ABD_FLAG_MULTI_CHUNK; } } #endif /* !CONFIG_HIGHMEM */ /* * This must be called if any of the sg_table allocation functions * are called. */ static void abd_free_sg_table(abd_t *abd) { struct sg_table table; table.sgl = ABD_SCATTER(abd).abd_sgl; table.nents = table.orig_nents = ABD_SCATTER(abd).abd_nents; sg_free_table(&table); } void abd_free_chunks(abd_t *abd) { struct scatterlist *sg = NULL; struct page *page; int nr_pages = ABD_SCATTER(abd).abd_nents; int order, i = 0; if (abd->abd_flags & ABD_FLAG_MULTI_ZONE) ABDSTAT_BUMPDOWN(abdstat_scatter_page_multi_zone); if (abd->abd_flags & ABD_FLAG_MULTI_CHUNK) ABDSTAT_BUMPDOWN(abdstat_scatter_page_multi_chunk); abd_for_each_sg(abd, sg, nr_pages, i) { page = sg_page(sg); abd_unmark_zfs_page(page); order = compound_order(page); __free_pages(page, order); ASSERT3U(sg->length, <=, PAGE_SIZE << order); ABDSTAT_BUMPDOWN(abdstat_scatter_orders[order]); } abd_free_sg_table(abd); } /* * Allocate scatter ABD of size SPA_MAXBLOCKSIZE, where each page in * the scatterlist will be set to the zero'd out buffer abd_zero_page. */ static void abd_alloc_zero_scatter(void) { struct scatterlist *sg = NULL; struct sg_table table; gfp_t gfp = __GFP_NOWARN | GFP_NOIO; int nr_pages = abd_chunkcnt_for_bytes(SPA_MAXBLOCKSIZE); int i = 0; #if defined(HAVE_ZERO_PAGE_GPL_ONLY) gfp_t gfp_zero_page = gfp | __GFP_ZERO; while ((abd_zero_page = __page_cache_alloc(gfp_zero_page)) == NULL) { ABDSTAT_BUMP(abdstat_scatter_page_alloc_retry); schedule_timeout_interruptible(1); } abd_mark_zfs_page(abd_zero_page); #else abd_zero_page = ZERO_PAGE(0); #endif /* HAVE_ZERO_PAGE_GPL_ONLY */ while (sg_alloc_table(&table, nr_pages, gfp)) { ABDSTAT_BUMP(abdstat_scatter_sg_table_retry); schedule_timeout_interruptible(1); } ASSERT3U(table.nents, ==, nr_pages); abd_zero_scatter = abd_alloc_struct(SPA_MAXBLOCKSIZE); abd_zero_scatter->abd_flags |= ABD_FLAG_OWNER; ABD_SCATTER(abd_zero_scatter).abd_offset = 0; ABD_SCATTER(abd_zero_scatter).abd_sgl = table.sgl; ABD_SCATTER(abd_zero_scatter).abd_nents = nr_pages; abd_zero_scatter->abd_size = SPA_MAXBLOCKSIZE; abd_zero_scatter->abd_flags |= ABD_FLAG_MULTI_CHUNK | ABD_FLAG_ZEROS; abd_for_each_sg(abd_zero_scatter, sg, nr_pages, i) { sg_set_page(sg, abd_zero_page, PAGESIZE, 0); } ABDSTAT_BUMP(abdstat_scatter_cnt); ABDSTAT_INCR(abdstat_scatter_data_size, PAGESIZE); ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk); } #else /* _KERNEL */ #ifndef PAGE_SHIFT #define PAGE_SHIFT (highbit64(PAGESIZE)-1) #endif #define zfs_kmap_atomic(chunk) ((void *)chunk) #define zfs_kunmap_atomic(addr) do { (void)(addr); } while (0) #define local_irq_save(flags) do { (void)(flags); } while (0) #define local_irq_restore(flags) do { (void)(flags); } while (0) #define nth_page(pg, i) \ ((struct page *)((void *)(pg) + (i) * PAGESIZE)) struct scatterlist { struct page *page; int length; int end; }; static void sg_init_table(struct scatterlist *sg, int nr) { memset(sg, 0, nr * sizeof (struct scatterlist)); sg[nr - 1].end = 1; } /* * This must be called if any of the sg_table allocation functions * are called. */ static void abd_free_sg_table(abd_t *abd) { int nents = ABD_SCATTER(abd).abd_nents; vmem_free(ABD_SCATTER(abd).abd_sgl, nents * sizeof (struct scatterlist)); } #define for_each_sg(sgl, sg, nr, i) \ for ((i) = 0, (sg) = (sgl); (i) < (nr); (i)++, (sg) = sg_next(sg)) static inline void sg_set_page(struct scatterlist *sg, struct page *page, unsigned int len, unsigned int offset) { /* currently we don't use offset */ ASSERT(offset == 0); sg->page = page; sg->length = len; } static inline struct page * sg_page(struct scatterlist *sg) { return (sg->page); } static inline struct scatterlist * sg_next(struct scatterlist *sg) { if (sg->end) return (NULL); return (sg + 1); } void abd_alloc_chunks(abd_t *abd, size_t size) { unsigned nr_pages = abd_chunkcnt_for_bytes(size); struct scatterlist *sg; int i; ABD_SCATTER(abd).abd_sgl = vmem_alloc(nr_pages * sizeof (struct scatterlist), KM_SLEEP); sg_init_table(ABD_SCATTER(abd).abd_sgl, nr_pages); abd_for_each_sg(abd, sg, nr_pages, i) { struct page *p = umem_alloc_aligned(PAGESIZE, 64, KM_SLEEP); sg_set_page(sg, p, PAGESIZE, 0); } ABD_SCATTER(abd).abd_nents = nr_pages; } void abd_free_chunks(abd_t *abd) { int i, n = ABD_SCATTER(abd).abd_nents; struct scatterlist *sg; abd_for_each_sg(abd, sg, n, i) { struct page *p = nth_page(sg_page(sg), 0); umem_free_aligned(p, PAGESIZE); } abd_free_sg_table(abd); } static void abd_alloc_zero_scatter(void) { unsigned nr_pages = abd_chunkcnt_for_bytes(SPA_MAXBLOCKSIZE); struct scatterlist *sg; int i; abd_zero_page = umem_alloc_aligned(PAGESIZE, 64, KM_SLEEP); memset(abd_zero_page, 0, PAGESIZE); abd_zero_scatter = abd_alloc_struct(SPA_MAXBLOCKSIZE); abd_zero_scatter->abd_flags |= ABD_FLAG_OWNER; abd_zero_scatter->abd_flags |= ABD_FLAG_MULTI_CHUNK | ABD_FLAG_ZEROS; ABD_SCATTER(abd_zero_scatter).abd_offset = 0; ABD_SCATTER(abd_zero_scatter).abd_nents = nr_pages; abd_zero_scatter->abd_size = SPA_MAXBLOCKSIZE; ABD_SCATTER(abd_zero_scatter).abd_sgl = vmem_alloc(nr_pages * sizeof (struct scatterlist), KM_SLEEP); sg_init_table(ABD_SCATTER(abd_zero_scatter).abd_sgl, nr_pages); abd_for_each_sg(abd_zero_scatter, sg, nr_pages, i) { sg_set_page(sg, abd_zero_page, PAGESIZE, 0); } ABDSTAT_BUMP(abdstat_scatter_cnt); ABDSTAT_INCR(abdstat_scatter_data_size, PAGESIZE); ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk); } #endif /* _KERNEL */ boolean_t abd_size_alloc_linear(size_t size) { return (!zfs_abd_scatter_enabled || size < zfs_abd_scatter_min_size); } void abd_update_scatter_stats(abd_t *abd, abd_stats_op_t op) { ASSERT(op == ABDSTAT_INCR || op == ABDSTAT_DECR); int waste = P2ROUNDUP(abd->abd_size, PAGESIZE) - abd->abd_size; if (op == ABDSTAT_INCR) { ABDSTAT_BUMP(abdstat_scatter_cnt); ABDSTAT_INCR(abdstat_scatter_data_size, abd->abd_size); ABDSTAT_INCR(abdstat_scatter_chunk_waste, waste); arc_space_consume(waste, ARC_SPACE_ABD_CHUNK_WASTE); } else { ABDSTAT_BUMPDOWN(abdstat_scatter_cnt); ABDSTAT_INCR(abdstat_scatter_data_size, -(int)abd->abd_size); ABDSTAT_INCR(abdstat_scatter_chunk_waste, -waste); arc_space_return(waste, ARC_SPACE_ABD_CHUNK_WASTE); } } void abd_update_linear_stats(abd_t *abd, abd_stats_op_t op) { ASSERT(op == ABDSTAT_INCR || op == ABDSTAT_DECR); if (op == ABDSTAT_INCR) { ABDSTAT_BUMP(abdstat_linear_cnt); ABDSTAT_INCR(abdstat_linear_data_size, abd->abd_size); } else { ABDSTAT_BUMPDOWN(abdstat_linear_cnt); ABDSTAT_INCR(abdstat_linear_data_size, -(int)abd->abd_size); } } void abd_verify_scatter(abd_t *abd) { size_t n; int i = 0; struct scatterlist *sg = NULL; ASSERT3U(ABD_SCATTER(abd).abd_nents, >, 0); ASSERT3U(ABD_SCATTER(abd).abd_offset, <, ABD_SCATTER(abd).abd_sgl->length); n = ABD_SCATTER(abd).abd_nents; abd_for_each_sg(abd, sg, n, i) { ASSERT3P(sg_page(sg), !=, NULL); } } static void abd_free_zero_scatter(void) { ABDSTAT_BUMPDOWN(abdstat_scatter_cnt); ABDSTAT_INCR(abdstat_scatter_data_size, -(int)PAGESIZE); ABDSTAT_BUMPDOWN(abdstat_scatter_page_multi_chunk); abd_free_sg_table(abd_zero_scatter); abd_free_struct(abd_zero_scatter); abd_zero_scatter = NULL; ASSERT3P(abd_zero_page, !=, NULL); #if defined(_KERNEL) #if defined(HAVE_ZERO_PAGE_GPL_ONLY) abd_unmark_zfs_page(abd_zero_page); __free_page(abd_zero_page); #endif /* HAVE_ZERO_PAGE_GPL_ONLY */ #else umem_free_aligned(abd_zero_page, PAGESIZE); #endif /* _KERNEL */ } static int abd_kstats_update(kstat_t *ksp, int rw) { abd_stats_t *as = ksp->ks_data; if (rw == KSTAT_WRITE) return (EACCES); as->abdstat_struct_size.value.ui64 = wmsum_value(&abd_sums.abdstat_struct_size); as->abdstat_linear_cnt.value.ui64 = wmsum_value(&abd_sums.abdstat_linear_cnt); as->abdstat_linear_data_size.value.ui64 = wmsum_value(&abd_sums.abdstat_linear_data_size); as->abdstat_scatter_cnt.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_cnt); as->abdstat_scatter_data_size.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_data_size); as->abdstat_scatter_chunk_waste.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_chunk_waste); for (int i = 0; i < ABD_MAX_ORDER; i++) { as->abdstat_scatter_orders[i].value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_orders[i]); } as->abdstat_scatter_page_multi_chunk.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_page_multi_chunk); as->abdstat_scatter_page_multi_zone.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_page_multi_zone); as->abdstat_scatter_page_alloc_retry.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_page_alloc_retry); as->abdstat_scatter_sg_table_retry.value.ui64 = wmsum_value(&abd_sums.abdstat_scatter_sg_table_retry); return (0); } void abd_init(void) { int i; abd_cache = kmem_cache_create("abd_t", sizeof (abd_t), 0, NULL, NULL, NULL, NULL, NULL, 0); wmsum_init(&abd_sums.abdstat_struct_size, 0); wmsum_init(&abd_sums.abdstat_linear_cnt, 0); wmsum_init(&abd_sums.abdstat_linear_data_size, 0); wmsum_init(&abd_sums.abdstat_scatter_cnt, 0); wmsum_init(&abd_sums.abdstat_scatter_data_size, 0); wmsum_init(&abd_sums.abdstat_scatter_chunk_waste, 0); for (i = 0; i < ABD_MAX_ORDER; i++) wmsum_init(&abd_sums.abdstat_scatter_orders[i], 0); wmsum_init(&abd_sums.abdstat_scatter_page_multi_chunk, 0); wmsum_init(&abd_sums.abdstat_scatter_page_multi_zone, 0); wmsum_init(&abd_sums.abdstat_scatter_page_alloc_retry, 0); wmsum_init(&abd_sums.abdstat_scatter_sg_table_retry, 0); abd_ksp = kstat_create("zfs", 0, "abdstats", "misc", KSTAT_TYPE_NAMED, sizeof (abd_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (abd_ksp != NULL) { for (i = 0; i < ABD_MAX_ORDER; i++) { snprintf(abd_stats.abdstat_scatter_orders[i].name, KSTAT_STRLEN, "scatter_order_%d", i); abd_stats.abdstat_scatter_orders[i].data_type = KSTAT_DATA_UINT64; } abd_ksp->ks_data = &abd_stats; abd_ksp->ks_update = abd_kstats_update; kstat_install(abd_ksp); } abd_alloc_zero_scatter(); } void abd_fini(void) { abd_free_zero_scatter(); if (abd_ksp != NULL) { kstat_delete(abd_ksp); abd_ksp = NULL; } wmsum_fini(&abd_sums.abdstat_struct_size); wmsum_fini(&abd_sums.abdstat_linear_cnt); wmsum_fini(&abd_sums.abdstat_linear_data_size); wmsum_fini(&abd_sums.abdstat_scatter_cnt); wmsum_fini(&abd_sums.abdstat_scatter_data_size); wmsum_fini(&abd_sums.abdstat_scatter_chunk_waste); for (int i = 0; i < ABD_MAX_ORDER; i++) wmsum_fini(&abd_sums.abdstat_scatter_orders[i]); wmsum_fini(&abd_sums.abdstat_scatter_page_multi_chunk); wmsum_fini(&abd_sums.abdstat_scatter_page_multi_zone); wmsum_fini(&abd_sums.abdstat_scatter_page_alloc_retry); wmsum_fini(&abd_sums.abdstat_scatter_sg_table_retry); if (abd_cache) { kmem_cache_destroy(abd_cache); abd_cache = NULL; } } void abd_free_linear_page(abd_t *abd) { /* Transform it back into a scatter ABD for freeing */ struct scatterlist *sg = abd->abd_u.abd_linear.abd_sgl; abd->abd_flags &= ~ABD_FLAG_LINEAR; abd->abd_flags &= ~ABD_FLAG_LINEAR_PAGE; ABD_SCATTER(abd).abd_nents = 1; ABD_SCATTER(abd).abd_offset = 0; ABD_SCATTER(abd).abd_sgl = sg; abd_free_chunks(abd); abd_update_scatter_stats(abd, ABDSTAT_DECR); } /* * If we're going to use this ABD for doing I/O using the block layer, the * consumer of the ABD data doesn't care if it's scattered or not, and we don't * plan to store this ABD in memory for a long period of time, we should * allocate the ABD type that requires the least data copying to do the I/O. * * On Linux the optimal thing to do would be to use abd_get_offset() and * construct a new ABD which shares the original pages thereby eliminating * the copy. But for the moment a new linear ABD is allocated until this * performance optimization can be implemented. */ abd_t * abd_alloc_for_io(size_t size, boolean_t is_metadata) { return (abd_alloc(size, is_metadata)); } abd_t * abd_get_offset_scatter(abd_t *abd, abd_t *sabd, size_t off, size_t size) { (void) size; int i = 0; struct scatterlist *sg = NULL; abd_verify(sabd); ASSERT3U(off, <=, sabd->abd_size); size_t new_offset = ABD_SCATTER(sabd).abd_offset + off; if (abd == NULL) abd = abd_alloc_struct(0); /* * Even if this buf is filesystem metadata, we only track that * if we own the underlying data buffer, which is not true in * this case. Therefore, we don't ever use ABD_FLAG_META here. */ abd_for_each_sg(sabd, sg, ABD_SCATTER(sabd).abd_nents, i) { if (new_offset < sg->length) break; new_offset -= sg->length; } ABD_SCATTER(abd).abd_sgl = sg; ABD_SCATTER(abd).abd_offset = new_offset; ABD_SCATTER(abd).abd_nents = ABD_SCATTER(sabd).abd_nents - i; return (abd); } /* * Initialize the abd_iter. */ void abd_iter_init(struct abd_iter *aiter, abd_t *abd) { ASSERT(!abd_is_gang(abd)); abd_verify(abd); memset(aiter, 0, sizeof (struct abd_iter)); aiter->iter_abd = abd; if (!abd_is_linear(abd)) { aiter->iter_offset = ABD_SCATTER(abd).abd_offset; aiter->iter_sg = ABD_SCATTER(abd).abd_sgl; } } /* * This is just a helper function to see if we have exhausted the * abd_iter and reached the end. */ boolean_t abd_iter_at_end(struct abd_iter *aiter) { ASSERT3U(aiter->iter_pos, <=, aiter->iter_abd->abd_size); return (aiter->iter_pos == aiter->iter_abd->abd_size); } /* * Advance the iterator by a certain amount. Cannot be called when a chunk is * in use. This can be safely called when the aiter has already exhausted, in * which case this does nothing. */ void abd_iter_advance(struct abd_iter *aiter, size_t amount) { /* * Ensure that last chunk is not in use. abd_iterate_*() must clear * this state (directly or abd_iter_unmap()) before advancing. */ ASSERT3P(aiter->iter_mapaddr, ==, NULL); ASSERT0(aiter->iter_mapsize); ASSERT3P(aiter->iter_page, ==, NULL); ASSERT0(aiter->iter_page_doff); ASSERT0(aiter->iter_page_dsize); /* There's nothing left to advance to, so do nothing */ if (abd_iter_at_end(aiter)) return; aiter->iter_pos += amount; aiter->iter_offset += amount; if (!abd_is_linear(aiter->iter_abd)) { while (aiter->iter_offset >= aiter->iter_sg->length) { aiter->iter_offset -= aiter->iter_sg->length; aiter->iter_sg = sg_next(aiter->iter_sg); if (aiter->iter_sg == NULL) { ASSERT0(aiter->iter_offset); break; } } } } /* * Map the current chunk into aiter. This can be safely called when the aiter * has already exhausted, in which case this does nothing. */ void abd_iter_map(struct abd_iter *aiter) { void *paddr; size_t offset = 0; ASSERT3P(aiter->iter_mapaddr, ==, NULL); ASSERT0(aiter->iter_mapsize); /* There's nothing left to iterate over, so do nothing */ if (abd_iter_at_end(aiter)) return; if (abd_is_linear(aiter->iter_abd)) { ASSERT3U(aiter->iter_pos, ==, aiter->iter_offset); offset = aiter->iter_offset; aiter->iter_mapsize = aiter->iter_abd->abd_size - offset; paddr = ABD_LINEAR_BUF(aiter->iter_abd); } else { offset = aiter->iter_offset; aiter->iter_mapsize = MIN(aiter->iter_sg->length - offset, aiter->iter_abd->abd_size - aiter->iter_pos); paddr = zfs_kmap_atomic(sg_page(aiter->iter_sg)); } aiter->iter_mapaddr = (char *)paddr + offset; } /* * Unmap the current chunk from aiter. This can be safely called when the aiter * has already exhausted, in which case this does nothing. */ void abd_iter_unmap(struct abd_iter *aiter) { /* There's nothing left to unmap, so do nothing */ if (abd_iter_at_end(aiter)) return; if (!abd_is_linear(aiter->iter_abd)) { /* LINTED E_FUNC_SET_NOT_USED */ zfs_kunmap_atomic(aiter->iter_mapaddr - aiter->iter_offset); } ASSERT3P(aiter->iter_mapaddr, !=, NULL); ASSERT3U(aiter->iter_mapsize, >, 0); aiter->iter_mapaddr = NULL; aiter->iter_mapsize = 0; } void abd_cache_reap_now(void) { } #if defined(_KERNEL) + /* - * Yield the next page struct and data offset and size within it, without + * This is abd_iter_page(), the function underneath abd_iterate_page_func(). + * It yields the next page struct and data offset and size within it, without * mapping it into the address space. */ + +/* + * "Compound pages" are a group of pages that can be referenced from a single + * struct page *. Its organised as a "head" page, followed by a series of + * "tail" pages. + * + * In OpenZFS, compound pages are allocated using the __GFP_COMP flag, which we + * get from scatter ABDs and SPL vmalloc slabs (ie >16K allocations). So a + * great many of the IO buffers we get are going to be of this type. + * + * The tail pages are just regular PAGESIZE pages, and can be safely used + * as-is. However, the head page has length covering itself and all the tail + * pages. If the ABD chunk spans multiple pages, then we can use the head page + * and a >PAGESIZE length, which is far more efficient. + * + * Before kernel 4.5 however, compound page heads were refcounted separately + * from tail pages, such that moving back to the head page would require us to + * take a reference to it and releasing it once we're completely finished with + * it. In practice, that means when our caller is done with the ABD, which we + * have no insight into from here. Rather than contort this API to track head + * page references on such ancient kernels, we disable this special compound + * page handling on 4.5, instead just using treating each page within it as a + * regular PAGESIZE page (which it is). This is slightly less efficient, but + * makes everything far simpler. + * + * The below test sets/clears ABD_ITER_COMPOUND_PAGES to enable/disable the + * special handling, and also defines the ABD_ITER_PAGE_SIZE(page) macro to + * understand compound pages, or not, as required. + */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 5, 0) +#define ABD_ITER_COMPOUND_PAGES 1 +#define ABD_ITER_PAGE_SIZE(page) \ + (PageCompound(page) ? page_size(page) : PAGESIZE) +#else +#undef ABD_ITER_COMPOUND_PAGES +#define ABD_ITER_PAGE_SIZE(page) (PAGESIZE) +#endif + void abd_iter_page(struct abd_iter *aiter) { if (abd_iter_at_end(aiter)) { aiter->iter_page = NULL; aiter->iter_page_doff = 0; aiter->iter_page_dsize = 0; return; } struct page *page; size_t doff, dsize; + /* + * Find the page, and the start of the data within it. This is computed + * differently for linear and scatter ABDs; linear is referenced by + * virtual memory location, while scatter is referenced by page + * pointer. + */ if (abd_is_linear(aiter->iter_abd)) { ASSERT3U(aiter->iter_pos, ==, aiter->iter_offset); /* memory address at iter_pos */ void *paddr = ABD_LINEAR_BUF(aiter->iter_abd) + aiter->iter_pos; /* struct page for address */ page = is_vmalloc_addr(paddr) ? vmalloc_to_page(paddr) : virt_to_page(paddr); /* offset of address within the page */ doff = offset_in_page(paddr); - - /* total data remaining in abd from this position */ - dsize = aiter->iter_abd->abd_size - aiter->iter_offset; } else { ASSERT(!abd_is_gang(aiter->iter_abd)); /* current scatter page */ - page = sg_page(aiter->iter_sg); + page = nth_page(sg_page(aiter->iter_sg), + aiter->iter_offset >> PAGE_SHIFT); /* position within page */ - doff = aiter->iter_offset; - - /* remaining data in scatterlist */ - dsize = MIN(aiter->iter_sg->length - aiter->iter_offset, - aiter->iter_abd->abd_size - aiter->iter_pos); + doff = aiter->iter_offset & (PAGESIZE - 1); } - ASSERT(page); -#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 5, 0) +#ifdef ABD_ITER_COMPOUND_PAGES if (PageTail(page)) { /* - * This page is part of a "compound page", which is a group of - * pages that can be referenced from a single struct page *. - * Its organised as a "head" page, followed by a series of - * "tail" pages. - * - * In OpenZFS, compound pages are allocated using the - * __GFP_COMP flag, which we get from scatter ABDs and SPL - * vmalloc slabs (ie >16K allocations). So a great many of the - * IO buffers we get are going to be of this type. - * - * The tail pages are just regular PAGE_SIZE pages, and can be - * safely used as-is. However, the head page has length - * covering itself and all the tail pages. If this ABD chunk - * spans multiple pages, then we can use the head page and a - * >PAGE_SIZE length, which is far more efficient. - * - * To do this, we need to adjust the offset to be counted from - * the head page. struct page for compound pages are stored - * contiguously, so we can just adjust by a simple offset. - * - * Before kernel 4.5, compound page heads were refcounted - * separately, such that moving back to the head page would - * require us to take a reference to it and releasing it once - * we're completely finished with it. In practice, that means - * when our caller is done with the ABD, which we have no - * insight into from here. Rather than contort this API to - * track head page references on such ancient kernels, we just - * compile this block out and use the tail pages directly. This - * is slightly less efficient, but makes everything far - * simpler. + * If this is a compound tail page, move back to the head, and + * adjust the offset to match. This may let us yield a much + * larger amount of data from a single logical page, and so + * leave our caller with fewer pages to process. */ struct page *head = compound_head(page); doff += ((page - head) * PAGESIZE); page = head; } #endif - /* final page and position within it */ + ASSERT(page); + + /* + * Compute the maximum amount of data we can take from this page. This + * is the smaller of: + * - the remaining space in the page + * - the remaining space in this scatterlist entry (which may not cover + * the entire page) + * - the remaining space in the abd (which may not cover the entire + * scatterlist entry) + */ + dsize = MIN(ABD_ITER_PAGE_SIZE(page) - doff, + aiter->iter_abd->abd_size - aiter->iter_pos); + if (!abd_is_linear(aiter->iter_abd)) + dsize = MIN(dsize, aiter->iter_sg->length - aiter->iter_offset); + ASSERT3U(dsize, >, 0); + + /* final iterator outputs */ aiter->iter_page = page; aiter->iter_page_doff = doff; - - /* amount of data in the chunk, up to the end of the page */ - aiter->iter_page_dsize = MIN(dsize, page_size(page) - doff); + aiter->iter_page_dsize = dsize; } /* * Note: ABD BIO functions only needed to support vdev_classic. See comments in * vdev_disk.c. */ /* * bio_nr_pages for ABD. * @off is the offset in @abd */ unsigned long abd_nr_pages_off(abd_t *abd, unsigned int size, size_t off) { unsigned long pos; if (abd_is_gang(abd)) { unsigned long count = 0; for (abd_t *cabd = abd_gang_get_offset(abd, &off); cabd != NULL && size != 0; cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) { ASSERT3U(off, <, cabd->abd_size); int mysize = MIN(size, cabd->abd_size - off); count += abd_nr_pages_off(cabd, mysize, off); size -= mysize; off = 0; } return (count); } if (abd_is_linear(abd)) pos = (unsigned long)abd_to_buf(abd) + off; else pos = ABD_SCATTER(abd).abd_offset + off; return (((pos + size + PAGESIZE - 1) >> PAGE_SHIFT) - (pos >> PAGE_SHIFT)); } static unsigned int bio_map(struct bio *bio, void *buf_ptr, unsigned int bio_size) { unsigned int offset, size, i; struct page *page; offset = offset_in_page(buf_ptr); for (i = 0; i < bio->bi_max_vecs; i++) { size = PAGE_SIZE - offset; if (bio_size <= 0) break; if (size > bio_size) size = bio_size; if (is_vmalloc_addr(buf_ptr)) page = vmalloc_to_page(buf_ptr); else page = virt_to_page(buf_ptr); /* * Some network related block device uses tcp_sendpage, which * doesn't behave well when using 0-count page, this is a * safety net to catch them. */ ASSERT3S(page_count(page), >, 0); if (bio_add_page(bio, page, size, offset) != size) break; buf_ptr += size; bio_size -= size; offset = 0; } return (bio_size); } /* * bio_map for gang ABD. */ static unsigned int abd_gang_bio_map_off(struct bio *bio, abd_t *abd, unsigned int io_size, size_t off) { ASSERT(abd_is_gang(abd)); for (abd_t *cabd = abd_gang_get_offset(abd, &off); cabd != NULL; cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) { ASSERT3U(off, <, cabd->abd_size); int size = MIN(io_size, cabd->abd_size - off); int remainder = abd_bio_map_off(bio, cabd, size, off); io_size -= (size - remainder); if (io_size == 0 || remainder > 0) return (io_size); off = 0; } ASSERT0(io_size); return (io_size); } /* * bio_map for ABD. * @off is the offset in @abd * Remaining IO size is returned */ unsigned int abd_bio_map_off(struct bio *bio, abd_t *abd, unsigned int io_size, size_t off) { struct abd_iter aiter; ASSERT3U(io_size, <=, abd->abd_size - off); if (abd_is_linear(abd)) return (bio_map(bio, ((char *)abd_to_buf(abd)) + off, io_size)); ASSERT(!abd_is_linear(abd)); if (abd_is_gang(abd)) return (abd_gang_bio_map_off(bio, abd, io_size, off)); abd_iter_init(&aiter, abd); abd_iter_advance(&aiter, off); for (int i = 0; i < bio->bi_max_vecs; i++) { struct page *pg; size_t len, sgoff, pgoff; struct scatterlist *sg; if (io_size <= 0) break; sg = aiter.iter_sg; sgoff = aiter.iter_offset; pgoff = sgoff & (PAGESIZE - 1); len = MIN(io_size, PAGESIZE - pgoff); ASSERT(len > 0); pg = nth_page(sg_page(sg), sgoff >> PAGE_SHIFT); if (bio_add_page(bio, pg, len, pgoff) != len) break; io_size -= len; abd_iter_advance(&aiter, len); } return (io_size); } /* Tunable Parameters */ module_param(zfs_abd_scatter_enabled, int, 0644); MODULE_PARM_DESC(zfs_abd_scatter_enabled, "Toggle whether ABD allocations must be linear."); module_param(zfs_abd_scatter_min_size, int, 0644); MODULE_PARM_DESC(zfs_abd_scatter_min_size, "Minimum size of scatter allocations."); /* CSTYLED */ module_param(zfs_abd_scatter_max_order, uint, 0644); MODULE_PARM_DESC(zfs_abd_scatter_max_order, "Maximum order allocation used for a scatter ABD."); #endif /* _KERNEL */ diff --git a/sys/contrib/openzfs/module/zfs/arc.c b/sys/contrib/openzfs/module/zfs/arc.c index 6954051b1d19..51039af9bcc0 100644 --- a/sys/contrib/openzfs/module/zfs/arc.c +++ b/sys/contrib/openzfs/module/zfs/arc.c @@ -1,10770 +1,10751 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2018, Joyent, Inc. * Copyright (c) 2011, 2020, Delphix. All rights reserved. * Copyright (c) 2014, Saso Kiselkov. All rights reserved. * Copyright (c) 2017, Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2019, loli10K . All rights reserved. * Copyright (c) 2020, George Amanakis. All rights reserved. * Copyright (c) 2019, Klara Inc. * Copyright (c) 2019, Allan Jude * Copyright (c) 2020, The FreeBSD Foundation [1] * * [1] Portions of this software were developed by Allan Jude * under sponsorship from the FreeBSD Foundation. */ /* * DVA-based Adjustable Replacement Cache * * While much of the theory of operation used here is * based on the self-tuning, low overhead replacement cache * presented by Megiddo and Modha at FAST 2003, there are some * significant differences: * * 1. The Megiddo and Modha model assumes any page is evictable. * Pages in its cache cannot be "locked" into memory. This makes * the eviction algorithm simple: evict the last page in the list. * This also make the performance characteristics easy to reason * about. Our cache is not so simple. At any given moment, some * subset of the blocks in the cache are un-evictable because we * have handed out a reference to them. Blocks are only evictable * when there are no external references active. This makes * eviction far more problematic: we choose to evict the evictable * blocks that are the "lowest" in the list. * * There are times when it is not possible to evict the requested * space. In these circumstances we are unable to adjust the cache * size. To prevent the cache growing unbounded at these times we * implement a "cache throttle" that slows the flow of new data * into the cache until we can make space available. * * 2. The Megiddo and Modha model assumes a fixed cache size. * Pages are evicted when the cache is full and there is a cache * miss. Our model has a variable sized cache. It grows with * high use, but also tries to react to memory pressure from the * operating system: decreasing its size when system memory is * tight. * * 3. The Megiddo and Modha model assumes a fixed page size. All * elements of the cache are therefore exactly the same size. So * when adjusting the cache size following a cache miss, its simply * a matter of choosing a single page to evict. In our model, we * have variable sized cache blocks (ranging from 512 bytes to * 128K bytes). We therefore choose a set of blocks to evict to make * space for a cache miss that approximates as closely as possible * the space used by the new block. * * See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache" * by N. Megiddo & D. Modha, FAST 2003 */ /* * The locking model: * * A new reference to a cache buffer can be obtained in two * ways: 1) via a hash table lookup using the DVA as a key, * or 2) via one of the ARC lists. The arc_read() interface * uses method 1, while the internal ARC algorithms for * adjusting the cache use method 2. We therefore provide two * types of locks: 1) the hash table lock array, and 2) the * ARC list locks. * * Buffers do not have their own mutexes, rather they rely on the * hash table mutexes for the bulk of their protection (i.e. most * fields in the arc_buf_hdr_t are protected by these mutexes). * * buf_hash_find() returns the appropriate mutex (held) when it * locates the requested buffer in the hash table. It returns * NULL for the mutex if the buffer was not in the table. * * buf_hash_remove() expects the appropriate hash mutex to be * already held before it is invoked. * * Each ARC state also has a mutex which is used to protect the * buffer list associated with the state. When attempting to * obtain a hash table lock while holding an ARC list lock you * must use: mutex_tryenter() to avoid deadlock. Also note that * the active state mutex must be held before the ghost state mutex. * * It as also possible to register a callback which is run when the * metadata limit is reached and no buffers can be safely evicted. In * this case the arc user should drop a reference on some arc buffers so * they can be reclaimed. For example, when using the ZPL each dentry * holds a references on a znode. These dentries must be pruned before * the arc buffer holding the znode can be safely evicted. * * Note that the majority of the performance stats are manipulated * with atomic operations. * * The L2ARC uses the l2ad_mtx on each vdev for the following: * * - L2ARC buflist creation * - L2ARC buflist eviction * - L2ARC write completion, which walks L2ARC buflists * - ARC header destruction, as it removes from L2ARC buflists * - ARC header release, as it removes from L2ARC buflists */ /* * ARC operation: * * Every block that is in the ARC is tracked by an arc_buf_hdr_t structure. * This structure can point either to a block that is still in the cache or to * one that is only accessible in an L2 ARC device, or it can provide * information about a block that was recently evicted. If a block is * only accessible in the L2ARC, then the arc_buf_hdr_t only has enough * information to retrieve it from the L2ARC device. This information is * stored in the l2arc_buf_hdr_t sub-structure of the arc_buf_hdr_t. A block * that is in this state cannot access the data directly. * * Blocks that are actively being referenced or have not been evicted * are cached in the L1ARC. The L1ARC (l1arc_buf_hdr_t) is a structure within * the arc_buf_hdr_t that will point to the data block in memory. A block can * only be read by a consumer if it has an l1arc_buf_hdr_t. The L1ARC * caches data in two ways -- in a list of ARC buffers (arc_buf_t) and * also in the arc_buf_hdr_t's private physical data block pointer (b_pabd). * * The L1ARC's data pointer may or may not be uncompressed. The ARC has the * ability to store the physical data (b_pabd) associated with the DVA of the * arc_buf_hdr_t. Since the b_pabd is a copy of the on-disk physical block, * it will match its on-disk compression characteristics. This behavior can be * disabled by setting 'zfs_compressed_arc_enabled' to B_FALSE. When the * compressed ARC functionality is disabled, the b_pabd will point to an * uncompressed version of the on-disk data. * * Data in the L1ARC is not accessed by consumers of the ARC directly. Each * arc_buf_hdr_t can have multiple ARC buffers (arc_buf_t) which reference it. * Each ARC buffer (arc_buf_t) is being actively accessed by a specific ARC * consumer. The ARC will provide references to this data and will keep it * cached until it is no longer in use. The ARC caches only the L1ARC's physical * data block and will evict any arc_buf_t that is no longer referenced. The * amount of memory consumed by the arc_buf_ts' data buffers can be seen via the * "overhead_size" kstat. * * Depending on the consumer, an arc_buf_t can be requested in uncompressed or * compressed form. The typical case is that consumers will want uncompressed * data, and when that happens a new data buffer is allocated where the data is * decompressed for them to use. Currently the only consumer who wants * compressed arc_buf_t's is "zfs send", when it streams data exactly as it * exists on disk. When this happens, the arc_buf_t's data buffer is shared * with the arc_buf_hdr_t. * * Here is a diagram showing an arc_buf_hdr_t referenced by two arc_buf_t's. The * first one is owned by a compressed send consumer (and therefore references * the same compressed data buffer as the arc_buf_hdr_t) and the second could be * used by any other consumer (and has its own uncompressed copy of the data * buffer). * * arc_buf_hdr_t * +-----------+ * | fields | * | common to | * | L1- and | * | L2ARC | * +-----------+ * | l2arc_buf_hdr_t * | | * +-----------+ * | l1arc_buf_hdr_t * | | arc_buf_t * | b_buf +------------>+-----------+ arc_buf_t * | b_pabd +-+ |b_next +---->+-----------+ * +-----------+ | |-----------| |b_next +-->NULL * | |b_comp = T | +-----------+ * | |b_data +-+ |b_comp = F | * | +-----------+ | |b_data +-+ * +->+------+ | +-----------+ | * compressed | | | | * data | |<--------------+ | uncompressed * +------+ compressed, | data * shared +-->+------+ * data | | * | | * +------+ * * When a consumer reads a block, the ARC must first look to see if the * arc_buf_hdr_t is cached. If the hdr is cached then the ARC allocates a new * arc_buf_t and either copies uncompressed data into a new data buffer from an * existing uncompressed arc_buf_t, decompresses the hdr's b_pabd buffer into a * new data buffer, or shares the hdr's b_pabd buffer, depending on whether the * hdr is compressed and the desired compression characteristics of the * arc_buf_t consumer. If the arc_buf_t ends up sharing data with the * arc_buf_hdr_t and both of them are uncompressed then the arc_buf_t must be * the last buffer in the hdr's b_buf list, however a shared compressed buf can * be anywhere in the hdr's list. * * The diagram below shows an example of an uncompressed ARC hdr that is * sharing its data with an arc_buf_t (note that the shared uncompressed buf is * the last element in the buf list): * * arc_buf_hdr_t * +-----------+ * | | * | | * | | * +-----------+ * l2arc_buf_hdr_t| | * | | * +-----------+ * l1arc_buf_hdr_t| | * | | arc_buf_t (shared) * | b_buf +------------>+---------+ arc_buf_t * | | |b_next +---->+---------+ * | b_pabd +-+ |---------| |b_next +-->NULL * +-----------+ | | | +---------+ * | |b_data +-+ | | * | +---------+ | |b_data +-+ * +->+------+ | +---------+ | * | | | | * uncompressed | | | | * data +------+ | | * ^ +->+------+ | * | uncompressed | | | * | data | | | * | +------+ | * +---------------------------------+ * * Writing to the ARC requires that the ARC first discard the hdr's b_pabd * since the physical block is about to be rewritten. The new data contents * will be contained in the arc_buf_t. As the I/O pipeline performs the write, * it may compress the data before writing it to disk. The ARC will be called * with the transformed data and will memcpy the transformed on-disk block into * a newly allocated b_pabd. Writes are always done into buffers which have * either been loaned (and hence are new and don't have other readers) or * buffers which have been released (and hence have their own hdr, if there * were originally other readers of the buf's original hdr). This ensures that * the ARC only needs to update a single buf and its hdr after a write occurs. * * When the L2ARC is in use, it will also take advantage of the b_pabd. The * L2ARC will always write the contents of b_pabd to the L2ARC. This means * that when compressed ARC is enabled that the L2ARC blocks are identical * to the on-disk block in the main data pool. This provides a significant * advantage since the ARC can leverage the bp's checksum when reading from the * L2ARC to determine if the contents are valid. However, if the compressed * ARC is disabled, then the L2ARC's block must be transformed to look * like the physical block in the main data pool before comparing the * checksum and determining its validity. * * The L1ARC has a slightly different system for storing encrypted data. * Raw (encrypted + possibly compressed) data has a few subtle differences from * data that is just compressed. The biggest difference is that it is not * possible to decrypt encrypted data (or vice-versa) if the keys aren't loaded. * The other difference is that encryption cannot be treated as a suggestion. * If a caller would prefer compressed data, but they actually wind up with * uncompressed data the worst thing that could happen is there might be a * performance hit. If the caller requests encrypted data, however, we must be * sure they actually get it or else secret information could be leaked. Raw * data is stored in hdr->b_crypt_hdr.b_rabd. An encrypted header, therefore, * may have both an encrypted version and a decrypted version of its data at * once. When a caller needs a raw arc_buf_t, it is allocated and the data is * copied out of this header. To avoid complications with b_pabd, raw buffers * cannot be shared. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef _KERNEL /* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */ boolean_t arc_watch = B_FALSE; #endif /* * This thread's job is to keep enough free memory in the system, by * calling arc_kmem_reap_soon() plus arc_reduce_target_size(), which improves * arc_available_memory(). */ static zthr_t *arc_reap_zthr; /* * This thread's job is to keep arc_size under arc_c, by calling * arc_evict(), which improves arc_is_overflowing(). */ static zthr_t *arc_evict_zthr; static arc_buf_hdr_t **arc_state_evict_markers; static int arc_state_evict_marker_count; static kmutex_t arc_evict_lock; static boolean_t arc_evict_needed = B_FALSE; static clock_t arc_last_uncached_flush; /* * Count of bytes evicted since boot. */ static uint64_t arc_evict_count; /* * List of arc_evict_waiter_t's, representing threads waiting for the * arc_evict_count to reach specific values. */ static list_t arc_evict_waiters; /* * When arc_is_overflowing(), arc_get_data_impl() waits for this percent of * the requested amount of data to be evicted. For example, by default for * every 2KB that's evicted, 1KB of it may be "reused" by a new allocation. * Since this is above 100%, it ensures that progress is made towards getting * arc_size under arc_c. Since this is finite, it ensures that allocations * can still happen, even during the potentially long time that arc_size is * more than arc_c. */ static uint_t zfs_arc_eviction_pct = 200; /* * The number of headers to evict in arc_evict_state_impl() before * dropping the sublist lock and evicting from another sublist. A lower * value means we're more likely to evict the "correct" header (i.e. the * oldest header in the arc state), but comes with higher overhead * (i.e. more invocations of arc_evict_state_impl()). */ static uint_t zfs_arc_evict_batch_limit = 10; /* number of seconds before growing cache again */ uint_t arc_grow_retry = 5; /* * Minimum time between calls to arc_kmem_reap_soon(). */ static const int arc_kmem_cache_reap_retry_ms = 1000; /* shift of arc_c for calculating overflow limit in arc_get_data_impl */ static int zfs_arc_overflow_shift = 8; /* log2(fraction of arc to reclaim) */ uint_t arc_shrink_shift = 7; /* percent of pagecache to reclaim arc to */ #ifdef _KERNEL uint_t zfs_arc_pc_percent = 0; #endif /* * log2(fraction of ARC which must be free to allow growing). * I.e. If there is less than arc_c >> arc_no_grow_shift free memory, * when reading a new block into the ARC, we will evict an equal-sized block * from the ARC. * * This must be less than arc_shrink_shift, so that when we shrink the ARC, * we will still not allow it to grow. */ uint_t arc_no_grow_shift = 5; /* * minimum lifespan of a prefetch block in clock ticks * (initialized in arc_init()) */ static uint_t arc_min_prefetch_ms; static uint_t arc_min_prescient_prefetch_ms; /* * If this percent of memory is free, don't throttle. */ uint_t arc_lotsfree_percent = 10; /* * The arc has filled available memory and has now warmed up. */ boolean_t arc_warm; /* * These tunables are for performance analysis. */ uint64_t zfs_arc_max = 0; uint64_t zfs_arc_min = 0; static uint64_t zfs_arc_dnode_limit = 0; static uint_t zfs_arc_dnode_reduce_percent = 10; static uint_t zfs_arc_grow_retry = 0; static uint_t zfs_arc_shrink_shift = 0; uint_t zfs_arc_average_blocksize = 8 * 1024; /* 8KB */ /* * ARC dirty data constraints for arc_tempreserve_space() throttle: * * total dirty data limit * * anon block dirty limit * * each pool's anon allowance */ static const unsigned long zfs_arc_dirty_limit_percent = 50; static const unsigned long zfs_arc_anon_limit_percent = 25; static const unsigned long zfs_arc_pool_dirty_percent = 20; /* * Enable or disable compressed arc buffers. */ int zfs_compressed_arc_enabled = B_TRUE; /* * Balance between metadata and data on ghost hits. Values above 100 * increase metadata caching by proportionally reducing effect of ghost * data hits on target data/metadata rate. */ static uint_t zfs_arc_meta_balance = 500; /* * Percentage that can be consumed by dnodes of ARC meta buffers. */ static uint_t zfs_arc_dnode_limit_percent = 10; /* * These tunables are Linux-specific */ static uint64_t zfs_arc_sys_free = 0; static uint_t zfs_arc_min_prefetch_ms = 0; static uint_t zfs_arc_min_prescient_prefetch_ms = 0; static uint_t zfs_arc_lotsfree_percent = 10; /* * Number of arc_prune threads */ static int zfs_arc_prune_task_threads = 1; /* The 7 states: */ arc_state_t ARC_anon; arc_state_t ARC_mru; arc_state_t ARC_mru_ghost; arc_state_t ARC_mfu; arc_state_t ARC_mfu_ghost; arc_state_t ARC_l2c_only; arc_state_t ARC_uncached; arc_stats_t arc_stats = { { "hits", KSTAT_DATA_UINT64 }, { "iohits", KSTAT_DATA_UINT64 }, { "misses", KSTAT_DATA_UINT64 }, { "demand_data_hits", KSTAT_DATA_UINT64 }, { "demand_data_iohits", KSTAT_DATA_UINT64 }, { "demand_data_misses", KSTAT_DATA_UINT64 }, { "demand_metadata_hits", KSTAT_DATA_UINT64 }, { "demand_metadata_iohits", KSTAT_DATA_UINT64 }, { "demand_metadata_misses", KSTAT_DATA_UINT64 }, { "prefetch_data_hits", KSTAT_DATA_UINT64 }, { "prefetch_data_iohits", KSTAT_DATA_UINT64 }, { "prefetch_data_misses", KSTAT_DATA_UINT64 }, { "prefetch_metadata_hits", KSTAT_DATA_UINT64 }, { "prefetch_metadata_iohits", KSTAT_DATA_UINT64 }, { "prefetch_metadata_misses", KSTAT_DATA_UINT64 }, { "mru_hits", KSTAT_DATA_UINT64 }, { "mru_ghost_hits", KSTAT_DATA_UINT64 }, { "mfu_hits", KSTAT_DATA_UINT64 }, { "mfu_ghost_hits", KSTAT_DATA_UINT64 }, { "uncached_hits", KSTAT_DATA_UINT64 }, { "deleted", KSTAT_DATA_UINT64 }, { "mutex_miss", KSTAT_DATA_UINT64 }, { "access_skip", KSTAT_DATA_UINT64 }, { "evict_skip", KSTAT_DATA_UINT64 }, { "evict_not_enough", KSTAT_DATA_UINT64 }, { "evict_l2_cached", KSTAT_DATA_UINT64 }, { "evict_l2_eligible", KSTAT_DATA_UINT64 }, { "evict_l2_eligible_mfu", KSTAT_DATA_UINT64 }, { "evict_l2_eligible_mru", KSTAT_DATA_UINT64 }, { "evict_l2_ineligible", KSTAT_DATA_UINT64 }, { "evict_l2_skip", KSTAT_DATA_UINT64 }, { "hash_elements", KSTAT_DATA_UINT64 }, { "hash_elements_max", KSTAT_DATA_UINT64 }, { "hash_collisions", KSTAT_DATA_UINT64 }, { "hash_chains", KSTAT_DATA_UINT64 }, { "hash_chain_max", KSTAT_DATA_UINT64 }, { "meta", KSTAT_DATA_UINT64 }, { "pd", KSTAT_DATA_UINT64 }, { "pm", KSTAT_DATA_UINT64 }, { "c", KSTAT_DATA_UINT64 }, { "c_min", KSTAT_DATA_UINT64 }, { "c_max", KSTAT_DATA_UINT64 }, { "size", KSTAT_DATA_UINT64 }, { "compressed_size", KSTAT_DATA_UINT64 }, { "uncompressed_size", KSTAT_DATA_UINT64 }, { "overhead_size", KSTAT_DATA_UINT64 }, { "hdr_size", KSTAT_DATA_UINT64 }, { "data_size", KSTAT_DATA_UINT64 }, { "metadata_size", KSTAT_DATA_UINT64 }, { "dbuf_size", KSTAT_DATA_UINT64 }, { "dnode_size", KSTAT_DATA_UINT64 }, { "bonus_size", KSTAT_DATA_UINT64 }, #if defined(COMPAT_FREEBSD11) { "other_size", KSTAT_DATA_UINT64 }, #endif { "anon_size", KSTAT_DATA_UINT64 }, { "anon_data", KSTAT_DATA_UINT64 }, { "anon_metadata", KSTAT_DATA_UINT64 }, { "anon_evictable_data", KSTAT_DATA_UINT64 }, { "anon_evictable_metadata", KSTAT_DATA_UINT64 }, { "mru_size", KSTAT_DATA_UINT64 }, { "mru_data", KSTAT_DATA_UINT64 }, { "mru_metadata", KSTAT_DATA_UINT64 }, { "mru_evictable_data", KSTAT_DATA_UINT64 }, { "mru_evictable_metadata", KSTAT_DATA_UINT64 }, { "mru_ghost_size", KSTAT_DATA_UINT64 }, { "mru_ghost_data", KSTAT_DATA_UINT64 }, { "mru_ghost_metadata", KSTAT_DATA_UINT64 }, { "mru_ghost_evictable_data", KSTAT_DATA_UINT64 }, { "mru_ghost_evictable_metadata", KSTAT_DATA_UINT64 }, { "mfu_size", KSTAT_DATA_UINT64 }, { "mfu_data", KSTAT_DATA_UINT64 }, { "mfu_metadata", KSTAT_DATA_UINT64 }, { "mfu_evictable_data", KSTAT_DATA_UINT64 }, { "mfu_evictable_metadata", KSTAT_DATA_UINT64 }, { "mfu_ghost_size", KSTAT_DATA_UINT64 }, { "mfu_ghost_data", KSTAT_DATA_UINT64 }, { "mfu_ghost_metadata", KSTAT_DATA_UINT64 }, { "mfu_ghost_evictable_data", KSTAT_DATA_UINT64 }, { "mfu_ghost_evictable_metadata", KSTAT_DATA_UINT64 }, { "uncached_size", KSTAT_DATA_UINT64 }, { "uncached_data", KSTAT_DATA_UINT64 }, { "uncached_metadata", KSTAT_DATA_UINT64 }, { "uncached_evictable_data", KSTAT_DATA_UINT64 }, { "uncached_evictable_metadata", KSTAT_DATA_UINT64 }, { "l2_hits", KSTAT_DATA_UINT64 }, { "l2_misses", KSTAT_DATA_UINT64 }, { "l2_prefetch_asize", KSTAT_DATA_UINT64 }, { "l2_mru_asize", KSTAT_DATA_UINT64 }, { "l2_mfu_asize", KSTAT_DATA_UINT64 }, { "l2_bufc_data_asize", KSTAT_DATA_UINT64 }, { "l2_bufc_metadata_asize", KSTAT_DATA_UINT64 }, { "l2_feeds", KSTAT_DATA_UINT64 }, { "l2_rw_clash", KSTAT_DATA_UINT64 }, { "l2_read_bytes", KSTAT_DATA_UINT64 }, { "l2_write_bytes", KSTAT_DATA_UINT64 }, { "l2_writes_sent", KSTAT_DATA_UINT64 }, { "l2_writes_done", KSTAT_DATA_UINT64 }, { "l2_writes_error", KSTAT_DATA_UINT64 }, { "l2_writes_lock_retry", KSTAT_DATA_UINT64 }, { "l2_evict_lock_retry", KSTAT_DATA_UINT64 }, { "l2_evict_reading", KSTAT_DATA_UINT64 }, { "l2_evict_l1cached", KSTAT_DATA_UINT64 }, { "l2_free_on_write", KSTAT_DATA_UINT64 }, { "l2_abort_lowmem", KSTAT_DATA_UINT64 }, { "l2_cksum_bad", KSTAT_DATA_UINT64 }, { "l2_io_error", KSTAT_DATA_UINT64 }, { "l2_size", KSTAT_DATA_UINT64 }, { "l2_asize", KSTAT_DATA_UINT64 }, { "l2_hdr_size", KSTAT_DATA_UINT64 }, { "l2_log_blk_writes", KSTAT_DATA_UINT64 }, { "l2_log_blk_avg_asize", KSTAT_DATA_UINT64 }, { "l2_log_blk_asize", KSTAT_DATA_UINT64 }, { "l2_log_blk_count", KSTAT_DATA_UINT64 }, { "l2_data_to_meta_ratio", KSTAT_DATA_UINT64 }, { "l2_rebuild_success", KSTAT_DATA_UINT64 }, { "l2_rebuild_unsupported", KSTAT_DATA_UINT64 }, { "l2_rebuild_io_errors", KSTAT_DATA_UINT64 }, { "l2_rebuild_dh_errors", KSTAT_DATA_UINT64 }, { "l2_rebuild_cksum_lb_errors", KSTAT_DATA_UINT64 }, { "l2_rebuild_lowmem", KSTAT_DATA_UINT64 }, { "l2_rebuild_size", KSTAT_DATA_UINT64 }, { "l2_rebuild_asize", KSTAT_DATA_UINT64 }, { "l2_rebuild_bufs", KSTAT_DATA_UINT64 }, { "l2_rebuild_bufs_precached", KSTAT_DATA_UINT64 }, { "l2_rebuild_log_blks", KSTAT_DATA_UINT64 }, { "memory_throttle_count", KSTAT_DATA_UINT64 }, { "memory_direct_count", KSTAT_DATA_UINT64 }, { "memory_indirect_count", KSTAT_DATA_UINT64 }, { "memory_all_bytes", KSTAT_DATA_UINT64 }, { "memory_free_bytes", KSTAT_DATA_UINT64 }, { "memory_available_bytes", KSTAT_DATA_INT64 }, { "arc_no_grow", KSTAT_DATA_UINT64 }, { "arc_tempreserve", KSTAT_DATA_UINT64 }, { "arc_loaned_bytes", KSTAT_DATA_UINT64 }, { "arc_prune", KSTAT_DATA_UINT64 }, { "arc_meta_used", KSTAT_DATA_UINT64 }, { "arc_dnode_limit", KSTAT_DATA_UINT64 }, { "async_upgrade_sync", KSTAT_DATA_UINT64 }, { "predictive_prefetch", KSTAT_DATA_UINT64 }, { "demand_hit_predictive_prefetch", KSTAT_DATA_UINT64 }, { "demand_iohit_predictive_prefetch", KSTAT_DATA_UINT64 }, { "prescient_prefetch", KSTAT_DATA_UINT64 }, { "demand_hit_prescient_prefetch", KSTAT_DATA_UINT64 }, { "demand_iohit_prescient_prefetch", KSTAT_DATA_UINT64 }, { "arc_need_free", KSTAT_DATA_UINT64 }, { "arc_sys_free", KSTAT_DATA_UINT64 }, { "arc_raw_size", KSTAT_DATA_UINT64 }, { "cached_only_in_progress", KSTAT_DATA_UINT64 }, { "abd_chunk_waste_size", KSTAT_DATA_UINT64 }, }; arc_sums_t arc_sums; #define ARCSTAT_MAX(stat, val) { \ uint64_t m; \ while ((val) > (m = arc_stats.stat.value.ui64) && \ (m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \ continue; \ } /* * We define a macro to allow ARC hits/misses to be easily broken down by * two separate conditions, giving a total of four different subtypes for * each of hits and misses (so eight statistics total). */ #define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \ if (cond1) { \ if (cond2) { \ ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \ } else { \ ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \ } \ } else { \ if (cond2) { \ ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \ } else { \ ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\ } \ } /* * This macro allows us to use kstats as floating averages. Each time we * update this kstat, we first factor it and the update value by * ARCSTAT_AVG_FACTOR to shrink the new value's contribution to the overall * average. This macro assumes that integer loads and stores are atomic, but * is not safe for multiple writers updating the kstat in parallel (only the * last writer's update will remain). */ #define ARCSTAT_F_AVG_FACTOR 3 #define ARCSTAT_F_AVG(stat, value) \ do { \ uint64_t x = ARCSTAT(stat); \ x = x - x / ARCSTAT_F_AVG_FACTOR + \ (value) / ARCSTAT_F_AVG_FACTOR; \ ARCSTAT(stat) = x; \ } while (0) static kstat_t *arc_ksp; /* * There are several ARC variables that are critical to export as kstats -- * but we don't want to have to grovel around in the kstat whenever we wish to * manipulate them. For these variables, we therefore define them to be in * terms of the statistic variable. This assures that we are not introducing * the possibility of inconsistency by having shadow copies of the variables, * while still allowing the code to be readable. */ #define arc_tempreserve ARCSTAT(arcstat_tempreserve) #define arc_loaned_bytes ARCSTAT(arcstat_loaned_bytes) #define arc_dnode_limit ARCSTAT(arcstat_dnode_limit) /* max size for dnodes */ #define arc_need_free ARCSTAT(arcstat_need_free) /* waiting to be evicted */ hrtime_t arc_growtime; list_t arc_prune_list; kmutex_t arc_prune_mtx; taskq_t *arc_prune_taskq; #define GHOST_STATE(state) \ ((state) == arc_mru_ghost || (state) == arc_mfu_ghost || \ (state) == arc_l2c_only) #define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_FLAG_IN_HASH_TABLE) #define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) #define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR) #define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH) #define HDR_PRESCIENT_PREFETCH(hdr) \ ((hdr)->b_flags & ARC_FLAG_PRESCIENT_PREFETCH) #define HDR_COMPRESSION_ENABLED(hdr) \ ((hdr)->b_flags & ARC_FLAG_COMPRESSED_ARC) #define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE) #define HDR_UNCACHED(hdr) ((hdr)->b_flags & ARC_FLAG_UNCACHED) #define HDR_L2_READING(hdr) \ (((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \ ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)) #define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING) #define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED) #define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD) #define HDR_PROTECTED(hdr) ((hdr)->b_flags & ARC_FLAG_PROTECTED) #define HDR_NOAUTH(hdr) ((hdr)->b_flags & ARC_FLAG_NOAUTH) #define HDR_SHARED_DATA(hdr) ((hdr)->b_flags & ARC_FLAG_SHARED_DATA) #define HDR_ISTYPE_METADATA(hdr) \ ((hdr)->b_flags & ARC_FLAG_BUFC_METADATA) #define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr)) #define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR) #define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR) #define HDR_HAS_RABD(hdr) \ (HDR_HAS_L1HDR(hdr) && HDR_PROTECTED(hdr) && \ (hdr)->b_crypt_hdr.b_rabd != NULL) #define HDR_ENCRYPTED(hdr) \ (HDR_PROTECTED(hdr) && DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot)) #define HDR_AUTHENTICATED(hdr) \ (HDR_PROTECTED(hdr) && !DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot)) /* For storing compression mode in b_flags */ #define HDR_COMPRESS_OFFSET (highbit64(ARC_FLAG_COMPRESS_0) - 1) #define HDR_GET_COMPRESS(hdr) ((enum zio_compress)BF32_GET((hdr)->b_flags, \ HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS)) #define HDR_SET_COMPRESS(hdr, cmp) BF32_SET((hdr)->b_flags, \ HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS, (cmp)); #define ARC_BUF_LAST(buf) ((buf)->b_next == NULL) #define ARC_BUF_SHARED(buf) ((buf)->b_flags & ARC_BUF_FLAG_SHARED) #define ARC_BUF_COMPRESSED(buf) ((buf)->b_flags & ARC_BUF_FLAG_COMPRESSED) #define ARC_BUF_ENCRYPTED(buf) ((buf)->b_flags & ARC_BUF_FLAG_ENCRYPTED) /* * Other sizes */ #define HDR_FULL_SIZE ((int64_t)sizeof (arc_buf_hdr_t)) #define HDR_L2ONLY_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_l1hdr)) /* * Hash table routines */ #define BUF_LOCKS 2048 typedef struct buf_hash_table { uint64_t ht_mask; arc_buf_hdr_t **ht_table; kmutex_t ht_locks[BUF_LOCKS] ____cacheline_aligned; } buf_hash_table_t; static buf_hash_table_t buf_hash_table; #define BUF_HASH_INDEX(spa, dva, birth) \ (buf_hash(spa, dva, birth) & buf_hash_table.ht_mask) #define BUF_HASH_LOCK(idx) (&buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)]) #define HDR_LOCK(hdr) \ (BUF_HASH_LOCK(BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth))) uint64_t zfs_crc64_table[256]; /* * Level 2 ARC */ #define L2ARC_WRITE_SIZE (32 * 1024 * 1024) /* initial write max */ #define L2ARC_HEADROOM 8 /* num of writes */ /* * If we discover during ARC scan any buffers to be compressed, we boost * our headroom for the next scanning cycle by this percentage multiple. */ #define L2ARC_HEADROOM_BOOST 200 #define L2ARC_FEED_SECS 1 /* caching interval secs */ #define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */ /* * We can feed L2ARC from two states of ARC buffers, mru and mfu, * and each of the state has two types: data and metadata. */ #define L2ARC_FEED_TYPES 4 /* L2ARC Performance Tunables */ uint64_t l2arc_write_max = L2ARC_WRITE_SIZE; /* def max write size */ uint64_t l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra warmup write */ uint64_t l2arc_headroom = L2ARC_HEADROOM; /* # of dev writes */ uint64_t l2arc_headroom_boost = L2ARC_HEADROOM_BOOST; uint64_t l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */ uint64_t l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval msecs */ int l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */ int l2arc_feed_again = B_TRUE; /* turbo warmup */ int l2arc_norw = B_FALSE; /* no reads during writes */ static uint_t l2arc_meta_percent = 33; /* limit on headers size */ /* * L2ARC Internals */ static list_t L2ARC_dev_list; /* device list */ static list_t *l2arc_dev_list; /* device list pointer */ static kmutex_t l2arc_dev_mtx; /* device list mutex */ static l2arc_dev_t *l2arc_dev_last; /* last device used */ static list_t L2ARC_free_on_write; /* free after write buf list */ static list_t *l2arc_free_on_write; /* free after write list ptr */ static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */ static uint64_t l2arc_ndev; /* number of devices */ typedef struct l2arc_read_callback { arc_buf_hdr_t *l2rcb_hdr; /* read header */ blkptr_t l2rcb_bp; /* original blkptr */ zbookmark_phys_t l2rcb_zb; /* original bookmark */ int l2rcb_flags; /* original flags */ abd_t *l2rcb_abd; /* temporary buffer */ } l2arc_read_callback_t; typedef struct l2arc_data_free { /* protected by l2arc_free_on_write_mtx */ abd_t *l2df_abd; size_t l2df_size; arc_buf_contents_t l2df_type; list_node_t l2df_list_node; } l2arc_data_free_t; typedef enum arc_fill_flags { ARC_FILL_LOCKED = 1 << 0, /* hdr lock is held */ ARC_FILL_COMPRESSED = 1 << 1, /* fill with compressed data */ ARC_FILL_ENCRYPTED = 1 << 2, /* fill with encrypted data */ ARC_FILL_NOAUTH = 1 << 3, /* don't attempt to authenticate */ ARC_FILL_IN_PLACE = 1 << 4 /* fill in place (special case) */ } arc_fill_flags_t; typedef enum arc_ovf_level { ARC_OVF_NONE, /* ARC within target size. */ ARC_OVF_SOME, /* ARC is slightly overflowed. */ ARC_OVF_SEVERE /* ARC is severely overflowed. */ } arc_ovf_level_t; static kmutex_t l2arc_feed_thr_lock; static kcondvar_t l2arc_feed_thr_cv; static uint8_t l2arc_thread_exit; static kmutex_t l2arc_rebuild_thr_lock; static kcondvar_t l2arc_rebuild_thr_cv; enum arc_hdr_alloc_flags { ARC_HDR_ALLOC_RDATA = 0x1, ARC_HDR_USE_RESERVE = 0x4, ARC_HDR_ALLOC_LINEAR = 0x8, }; static abd_t *arc_get_data_abd(arc_buf_hdr_t *, uint64_t, const void *, int); static void *arc_get_data_buf(arc_buf_hdr_t *, uint64_t, const void *); static void arc_get_data_impl(arc_buf_hdr_t *, uint64_t, const void *, int); static void arc_free_data_abd(arc_buf_hdr_t *, abd_t *, uint64_t, const void *); static void arc_free_data_buf(arc_buf_hdr_t *, void *, uint64_t, const void *); static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size, const void *tag); static void arc_hdr_free_abd(arc_buf_hdr_t *, boolean_t); static void arc_hdr_alloc_abd(arc_buf_hdr_t *, int); static void arc_hdr_destroy(arc_buf_hdr_t *); static void arc_access(arc_buf_hdr_t *, arc_flags_t, boolean_t); static void arc_buf_watch(arc_buf_t *); static void arc_change_state(arc_state_t *, arc_buf_hdr_t *); static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *); static uint32_t arc_bufc_to_flags(arc_buf_contents_t); static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags); static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags); static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *); static void l2arc_read_done(zio_t *); static void l2arc_do_free_on_write(void); static void l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr, boolean_t state_only); static void arc_prune_async(uint64_t adjust); #define l2arc_hdr_arcstats_increment(hdr) \ l2arc_hdr_arcstats_update((hdr), B_TRUE, B_FALSE) #define l2arc_hdr_arcstats_decrement(hdr) \ l2arc_hdr_arcstats_update((hdr), B_FALSE, B_FALSE) #define l2arc_hdr_arcstats_increment_state(hdr) \ l2arc_hdr_arcstats_update((hdr), B_TRUE, B_TRUE) #define l2arc_hdr_arcstats_decrement_state(hdr) \ l2arc_hdr_arcstats_update((hdr), B_FALSE, B_TRUE) /* * l2arc_exclude_special : A zfs module parameter that controls whether buffers * present on special vdevs are eligibile for caching in L2ARC. If * set to 1, exclude dbufs on special vdevs from being cached to * L2ARC. */ int l2arc_exclude_special = 0; /* * l2arc_mfuonly : A ZFS module parameter that controls whether only MFU * metadata and data are cached from ARC into L2ARC. */ static int l2arc_mfuonly = 0; /* * L2ARC TRIM * l2arc_trim_ahead : A ZFS module parameter that controls how much ahead of * the current write size (l2arc_write_max) we should TRIM if we * have filled the device. It is defined as a percentage of the * write size. If set to 100 we trim twice the space required to * accommodate upcoming writes. A minimum of 64MB will be trimmed. * It also enables TRIM of the whole L2ARC device upon creation or * addition to an existing pool or if the header of the device is * invalid upon importing a pool or onlining a cache device. The * default is 0, which disables TRIM on L2ARC altogether as it can * put significant stress on the underlying storage devices. This * will vary depending of how well the specific device handles * these commands. */ static uint64_t l2arc_trim_ahead = 0; /* * Performance tuning of L2ARC persistence: * * l2arc_rebuild_enabled : A ZFS module parameter that controls whether adding * an L2ARC device (either at pool import or later) will attempt * to rebuild L2ARC buffer contents. * l2arc_rebuild_blocks_min_l2size : A ZFS module parameter that controls * whether log blocks are written to the L2ARC device. If the L2ARC * device is less than 1GB, the amount of data l2arc_evict() * evicts is significant compared to the amount of restored L2ARC * data. In this case do not write log blocks in L2ARC in order * not to waste space. */ static int l2arc_rebuild_enabled = B_TRUE; static uint64_t l2arc_rebuild_blocks_min_l2size = 1024 * 1024 * 1024; /* L2ARC persistence rebuild control routines. */ void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen); static __attribute__((noreturn)) void l2arc_dev_rebuild_thread(void *arg); static int l2arc_rebuild(l2arc_dev_t *dev); /* L2ARC persistence read I/O routines. */ static int l2arc_dev_hdr_read(l2arc_dev_t *dev); static int l2arc_log_blk_read(l2arc_dev_t *dev, const l2arc_log_blkptr_t *this_lp, const l2arc_log_blkptr_t *next_lp, l2arc_log_blk_phys_t *this_lb, l2arc_log_blk_phys_t *next_lb, zio_t *this_io, zio_t **next_io); static zio_t *l2arc_log_blk_fetch(vdev_t *vd, const l2arc_log_blkptr_t *lp, l2arc_log_blk_phys_t *lb); static void l2arc_log_blk_fetch_abort(zio_t *zio); /* L2ARC persistence block restoration routines. */ static void l2arc_log_blk_restore(l2arc_dev_t *dev, const l2arc_log_blk_phys_t *lb, uint64_t lb_asize); static void l2arc_hdr_restore(const l2arc_log_ent_phys_t *le, l2arc_dev_t *dev); /* L2ARC persistence write I/O routines. */ static uint64_t l2arc_log_blk_commit(l2arc_dev_t *dev, zio_t *pio, l2arc_write_callback_t *cb); /* L2ARC persistence auxiliary routines. */ boolean_t l2arc_log_blkptr_valid(l2arc_dev_t *dev, const l2arc_log_blkptr_t *lbp); static boolean_t l2arc_log_blk_insert(l2arc_dev_t *dev, const arc_buf_hdr_t *ab); boolean_t l2arc_range_check_overlap(uint64_t bottom, uint64_t top, uint64_t check); static void l2arc_blk_fetch_done(zio_t *zio); static inline uint64_t l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev); /* * We use Cityhash for this. It's fast, and has good hash properties without * requiring any large static buffers. */ static uint64_t buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth) { return (cityhash4(spa, dva->dva_word[0], dva->dva_word[1], birth)); } #define HDR_EMPTY(hdr) \ ((hdr)->b_dva.dva_word[0] == 0 && \ (hdr)->b_dva.dva_word[1] == 0) #define HDR_EMPTY_OR_LOCKED(hdr) \ (HDR_EMPTY(hdr) || MUTEX_HELD(HDR_LOCK(hdr))) #define HDR_EQUAL(spa, dva, birth, hdr) \ ((hdr)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \ ((hdr)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \ ((hdr)->b_birth == birth) && ((hdr)->b_spa == spa) static void buf_discard_identity(arc_buf_hdr_t *hdr) { hdr->b_dva.dva_word[0] = 0; hdr->b_dva.dva_word[1] = 0; hdr->b_birth = 0; } static arc_buf_hdr_t * buf_hash_find(uint64_t spa, const blkptr_t *bp, kmutex_t **lockp) { const dva_t *dva = BP_IDENTITY(bp); uint64_t birth = BP_GET_BIRTH(bp); uint64_t idx = BUF_HASH_INDEX(spa, dva, birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *hdr; mutex_enter(hash_lock); for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL; hdr = hdr->b_hash_next) { if (HDR_EQUAL(spa, dva, birth, hdr)) { *lockp = hash_lock; return (hdr); } } mutex_exit(hash_lock); *lockp = NULL; return (NULL); } /* * Insert an entry into the hash table. If there is already an element * equal to elem in the hash table, then the already existing element * will be returned and the new element will not be inserted. * Otherwise returns NULL. * If lockp == NULL, the caller is assumed to already hold the hash lock. */ static arc_buf_hdr_t * buf_hash_insert(arc_buf_hdr_t *hdr, kmutex_t **lockp) { uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *fhdr; uint32_t i; ASSERT(!DVA_IS_EMPTY(&hdr->b_dva)); ASSERT(hdr->b_birth != 0); ASSERT(!HDR_IN_HASH_TABLE(hdr)); if (lockp != NULL) { *lockp = hash_lock; mutex_enter(hash_lock); } else { ASSERT(MUTEX_HELD(hash_lock)); } for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL; fhdr = fhdr->b_hash_next, i++) { if (HDR_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr)) return (fhdr); } hdr->b_hash_next = buf_hash_table.ht_table[idx]; buf_hash_table.ht_table[idx] = hdr; arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE); /* collect some hash table performance data */ if (i > 0) { ARCSTAT_BUMP(arcstat_hash_collisions); if (i == 1) ARCSTAT_BUMP(arcstat_hash_chains); ARCSTAT_MAX(arcstat_hash_chain_max, i); } uint64_t he = atomic_inc_64_nv( &arc_stats.arcstat_hash_elements.value.ui64); ARCSTAT_MAX(arcstat_hash_elements_max, he); return (NULL); } static void buf_hash_remove(arc_buf_hdr_t *hdr) { arc_buf_hdr_t *fhdr, **hdrp; uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth); ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx))); ASSERT(HDR_IN_HASH_TABLE(hdr)); hdrp = &buf_hash_table.ht_table[idx]; while ((fhdr = *hdrp) != hdr) { ASSERT3P(fhdr, !=, NULL); hdrp = &fhdr->b_hash_next; } *hdrp = hdr->b_hash_next; hdr->b_hash_next = NULL; arc_hdr_clear_flags(hdr, ARC_FLAG_IN_HASH_TABLE); /* collect some hash table performance data */ atomic_dec_64(&arc_stats.arcstat_hash_elements.value.ui64); if (buf_hash_table.ht_table[idx] && buf_hash_table.ht_table[idx]->b_hash_next == NULL) ARCSTAT_BUMPDOWN(arcstat_hash_chains); } /* * Global data structures and functions for the buf kmem cache. */ static kmem_cache_t *hdr_full_cache; static kmem_cache_t *hdr_l2only_cache; static kmem_cache_t *buf_cache; static void buf_fini(void) { #if defined(_KERNEL) /* * Large allocations which do not require contiguous pages * should be using vmem_free() in the linux kernel\ */ vmem_free(buf_hash_table.ht_table, (buf_hash_table.ht_mask + 1) * sizeof (void *)); #else kmem_free(buf_hash_table.ht_table, (buf_hash_table.ht_mask + 1) * sizeof (void *)); #endif for (int i = 0; i < BUF_LOCKS; i++) mutex_destroy(BUF_HASH_LOCK(i)); kmem_cache_destroy(hdr_full_cache); kmem_cache_destroy(hdr_l2only_cache); kmem_cache_destroy(buf_cache); } /* * Constructor callback - called when the cache is empty * and a new buf is requested. */ static int hdr_full_cons(void *vbuf, void *unused, int kmflag) { (void) unused, (void) kmflag; arc_buf_hdr_t *hdr = vbuf; memset(hdr, 0, HDR_FULL_SIZE); hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; zfs_refcount_create(&hdr->b_l1hdr.b_refcnt); #ifdef ZFS_DEBUG mutex_init(&hdr->b_l1hdr.b_freeze_lock, NULL, MUTEX_DEFAULT, NULL); #endif multilist_link_init(&hdr->b_l1hdr.b_arc_node); list_link_init(&hdr->b_l2hdr.b_l2node); arc_space_consume(HDR_FULL_SIZE, ARC_SPACE_HDRS); return (0); } static int hdr_l2only_cons(void *vbuf, void *unused, int kmflag) { (void) unused, (void) kmflag; arc_buf_hdr_t *hdr = vbuf; memset(hdr, 0, HDR_L2ONLY_SIZE); arc_space_consume(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS); return (0); } static int buf_cons(void *vbuf, void *unused, int kmflag) { (void) unused, (void) kmflag; arc_buf_t *buf = vbuf; memset(buf, 0, sizeof (arc_buf_t)); arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS); return (0); } /* * Destructor callback - called when a cached buf is * no longer required. */ static void hdr_full_dest(void *vbuf, void *unused) { (void) unused; arc_buf_hdr_t *hdr = vbuf; ASSERT(HDR_EMPTY(hdr)); zfs_refcount_destroy(&hdr->b_l1hdr.b_refcnt); #ifdef ZFS_DEBUG mutex_destroy(&hdr->b_l1hdr.b_freeze_lock); #endif ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); arc_space_return(HDR_FULL_SIZE, ARC_SPACE_HDRS); } static void hdr_l2only_dest(void *vbuf, void *unused) { (void) unused; arc_buf_hdr_t *hdr = vbuf; ASSERT(HDR_EMPTY(hdr)); arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS); } static void buf_dest(void *vbuf, void *unused) { (void) unused; (void) vbuf; arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS); } static void buf_init(void) { uint64_t *ct = NULL; uint64_t hsize = 1ULL << 12; int i, j; /* * The hash table is big enough to fill all of physical memory * with an average block size of zfs_arc_average_blocksize (default 8K). * By default, the table will take up * totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers). */ while (hsize * zfs_arc_average_blocksize < arc_all_memory()) hsize <<= 1; retry: buf_hash_table.ht_mask = hsize - 1; #if defined(_KERNEL) /* * Large allocations which do not require contiguous pages * should be using vmem_alloc() in the linux kernel */ buf_hash_table.ht_table = vmem_zalloc(hsize * sizeof (void*), KM_SLEEP); #else buf_hash_table.ht_table = kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP); #endif if (buf_hash_table.ht_table == NULL) { ASSERT(hsize > (1ULL << 8)); hsize >>= 1; goto retry; } hdr_full_cache = kmem_cache_create("arc_buf_hdr_t_full", HDR_FULL_SIZE, 0, hdr_full_cons, hdr_full_dest, NULL, NULL, NULL, 0); hdr_l2only_cache = kmem_cache_create("arc_buf_hdr_t_l2only", HDR_L2ONLY_SIZE, 0, hdr_l2only_cons, hdr_l2only_dest, NULL, NULL, NULL, 0); buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t), 0, buf_cons, buf_dest, NULL, NULL, NULL, 0); for (i = 0; i < 256; i++) for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--) *ct = (*ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY); for (i = 0; i < BUF_LOCKS; i++) mutex_init(BUF_HASH_LOCK(i), NULL, MUTEX_DEFAULT, NULL); } #define ARC_MINTIME (hz>>4) /* 62 ms */ /* * This is the size that the buf occupies in memory. If the buf is compressed, * it will correspond to the compressed size. You should use this method of * getting the buf size unless you explicitly need the logical size. */ uint64_t arc_buf_size(arc_buf_t *buf) { return (ARC_BUF_COMPRESSED(buf) ? HDR_GET_PSIZE(buf->b_hdr) : HDR_GET_LSIZE(buf->b_hdr)); } uint64_t arc_buf_lsize(arc_buf_t *buf) { return (HDR_GET_LSIZE(buf->b_hdr)); } /* * This function will return B_TRUE if the buffer is encrypted in memory. * This buffer can be decrypted by calling arc_untransform(). */ boolean_t arc_is_encrypted(arc_buf_t *buf) { return (ARC_BUF_ENCRYPTED(buf) != 0); } /* * Returns B_TRUE if the buffer represents data that has not had its MAC * verified yet. */ boolean_t arc_is_unauthenticated(arc_buf_t *buf) { return (HDR_NOAUTH(buf->b_hdr) != 0); } void arc_get_raw_params(arc_buf_t *buf, boolean_t *byteorder, uint8_t *salt, uint8_t *iv, uint8_t *mac) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(HDR_PROTECTED(hdr)); memcpy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN); memcpy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN); memcpy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN); *byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ? ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER; } /* * Indicates how this buffer is compressed in memory. If it is not compressed * the value will be ZIO_COMPRESS_OFF. It can be made normally readable with * arc_untransform() as long as it is also unencrypted. */ enum zio_compress arc_get_compression(arc_buf_t *buf) { return (ARC_BUF_COMPRESSED(buf) ? HDR_GET_COMPRESS(buf->b_hdr) : ZIO_COMPRESS_OFF); } /* * Return the compression algorithm used to store this data in the ARC. If ARC * compression is enabled or this is an encrypted block, this will be the same * as what's used to store it on-disk. Otherwise, this will be ZIO_COMPRESS_OFF. */ static inline enum zio_compress arc_hdr_get_compress(arc_buf_hdr_t *hdr) { return (HDR_COMPRESSION_ENABLED(hdr) ? HDR_GET_COMPRESS(hdr) : ZIO_COMPRESS_OFF); } uint8_t arc_get_complevel(arc_buf_t *buf) { return (buf->b_hdr->b_complevel); } static inline boolean_t arc_buf_is_shared(arc_buf_t *buf) { boolean_t shared = (buf->b_data != NULL && buf->b_hdr->b_l1hdr.b_pabd != NULL && abd_is_linear(buf->b_hdr->b_l1hdr.b_pabd) && buf->b_data == abd_to_buf(buf->b_hdr->b_l1hdr.b_pabd)); IMPLY(shared, HDR_SHARED_DATA(buf->b_hdr)); EQUIV(shared, ARC_BUF_SHARED(buf)); IMPLY(shared, ARC_BUF_COMPRESSED(buf) || ARC_BUF_LAST(buf)); /* * It would be nice to assert arc_can_share() too, but the "hdr isn't * already being shared" requirement prevents us from doing that. */ return (shared); } /* * Free the checksum associated with this header. If there is no checksum, this * is a no-op. */ static inline void arc_cksum_free(arc_buf_hdr_t *hdr) { #ifdef ZFS_DEBUG ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&hdr->b_l1hdr.b_freeze_lock); if (hdr->b_l1hdr.b_freeze_cksum != NULL) { kmem_free(hdr->b_l1hdr.b_freeze_cksum, sizeof (zio_cksum_t)); hdr->b_l1hdr.b_freeze_cksum = NULL; } mutex_exit(&hdr->b_l1hdr.b_freeze_lock); #endif } /* * Return true iff at least one of the bufs on hdr is not compressed. * Encrypted buffers count as compressed. */ static boolean_t arc_hdr_has_uncompressed_buf(arc_buf_hdr_t *hdr) { ASSERT(hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY_OR_LOCKED(hdr)); for (arc_buf_t *b = hdr->b_l1hdr.b_buf; b != NULL; b = b->b_next) { if (!ARC_BUF_COMPRESSED(b)) { return (B_TRUE); } } return (B_FALSE); } /* * If we've turned on the ZFS_DEBUG_MODIFY flag, verify that the buf's data * matches the checksum that is stored in the hdr. If there is no checksum, * or if the buf is compressed, this is a no-op. */ static void arc_cksum_verify(arc_buf_t *buf) { #ifdef ZFS_DEBUG arc_buf_hdr_t *hdr = buf->b_hdr; zio_cksum_t zc; if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; if (ARC_BUF_COMPRESSED(buf)) return; ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&hdr->b_l1hdr.b_freeze_lock); if (hdr->b_l1hdr.b_freeze_cksum == NULL || HDR_IO_ERROR(hdr)) { mutex_exit(&hdr->b_l1hdr.b_freeze_lock); return; } fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, &zc); if (!ZIO_CHECKSUM_EQUAL(*hdr->b_l1hdr.b_freeze_cksum, zc)) panic("buffer modified while frozen!"); mutex_exit(&hdr->b_l1hdr.b_freeze_lock); #endif } /* * This function makes the assumption that data stored in the L2ARC * will be transformed exactly as it is in the main pool. Because of * this we can verify the checksum against the reading process's bp. */ static boolean_t arc_cksum_is_equal(arc_buf_hdr_t *hdr, zio_t *zio) { ASSERT(!BP_IS_EMBEDDED(zio->io_bp)); VERIFY3U(BP_GET_PSIZE(zio->io_bp), ==, HDR_GET_PSIZE(hdr)); /* * Block pointers always store the checksum for the logical data. * If the block pointer has the gang bit set, then the checksum * it represents is for the reconstituted data and not for an * individual gang member. The zio pipeline, however, must be able to * determine the checksum of each of the gang constituents so it * treats the checksum comparison differently than what we need * for l2arc blocks. This prevents us from using the * zio_checksum_error() interface directly. Instead we must call the * zio_checksum_error_impl() so that we can ensure the checksum is * generated using the correct checksum algorithm and accounts for the * logical I/O size and not just a gang fragment. */ return (zio_checksum_error_impl(zio->io_spa, zio->io_bp, BP_GET_CHECKSUM(zio->io_bp), zio->io_abd, zio->io_size, zio->io_offset, NULL) == 0); } /* * Given a buf full of data, if ZFS_DEBUG_MODIFY is enabled this computes a * checksum and attaches it to the buf's hdr so that we can ensure that the buf * isn't modified later on. If buf is compressed or there is already a checksum * on the hdr, this is a no-op (we only checksum uncompressed bufs). */ static void arc_cksum_compute(arc_buf_t *buf) { if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; #ifdef ZFS_DEBUG arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&hdr->b_l1hdr.b_freeze_lock); if (hdr->b_l1hdr.b_freeze_cksum != NULL || ARC_BUF_COMPRESSED(buf)) { mutex_exit(&hdr->b_l1hdr.b_freeze_lock); return; } ASSERT(!ARC_BUF_ENCRYPTED(buf)); ASSERT(!ARC_BUF_COMPRESSED(buf)); hdr->b_l1hdr.b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t), KM_SLEEP); fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, hdr->b_l1hdr.b_freeze_cksum); mutex_exit(&hdr->b_l1hdr.b_freeze_lock); #endif arc_buf_watch(buf); } #ifndef _KERNEL void arc_buf_sigsegv(int sig, siginfo_t *si, void *unused) { (void) sig, (void) unused; panic("Got SIGSEGV at address: 0x%lx\n", (long)si->si_addr); } #endif static void arc_buf_unwatch(arc_buf_t *buf) { #ifndef _KERNEL if (arc_watch) { ASSERT0(mprotect(buf->b_data, arc_buf_size(buf), PROT_READ | PROT_WRITE)); } #else (void) buf; #endif } static void arc_buf_watch(arc_buf_t *buf) { #ifndef _KERNEL if (arc_watch) ASSERT0(mprotect(buf->b_data, arc_buf_size(buf), PROT_READ)); #else (void) buf; #endif } static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *hdr) { arc_buf_contents_t type; if (HDR_ISTYPE_METADATA(hdr)) { type = ARC_BUFC_METADATA; } else { type = ARC_BUFC_DATA; } VERIFY3U(hdr->b_type, ==, type); return (type); } boolean_t arc_is_metadata(arc_buf_t *buf) { return (HDR_ISTYPE_METADATA(buf->b_hdr) != 0); } static uint32_t arc_bufc_to_flags(arc_buf_contents_t type) { switch (type) { case ARC_BUFC_DATA: /* metadata field is 0 if buffer contains normal data */ return (0); case ARC_BUFC_METADATA: return (ARC_FLAG_BUFC_METADATA); default: break; } panic("undefined ARC buffer type!"); return ((uint32_t)-1); } void arc_buf_thaw(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); arc_cksum_verify(buf); /* * Compressed buffers do not manipulate the b_freeze_cksum. */ if (ARC_BUF_COMPRESSED(buf)) return; ASSERT(HDR_HAS_L1HDR(hdr)); arc_cksum_free(hdr); arc_buf_unwatch(buf); } void arc_buf_freeze(arc_buf_t *buf) { if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; if (ARC_BUF_COMPRESSED(buf)) return; ASSERT(HDR_HAS_L1HDR(buf->b_hdr)); arc_cksum_compute(buf); } /* * The arc_buf_hdr_t's b_flags should never be modified directly. Instead, * the following functions should be used to ensure that the flags are * updated in a thread-safe way. When manipulating the flags either * the hash_lock must be held or the hdr must be undiscoverable. This * ensures that we're not racing with any other threads when updating * the flags. */ static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags) { ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); hdr->b_flags |= flags; } static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags) { ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); hdr->b_flags &= ~flags; } /* * Setting the compression bits in the arc_buf_hdr_t's b_flags is * done in a special way since we have to clear and set bits * at the same time. Consumers that wish to set the compression bits * must use this function to ensure that the flags are updated in * thread-safe manner. */ static void arc_hdr_set_compress(arc_buf_hdr_t *hdr, enum zio_compress cmp) { ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); /* * Holes and embedded blocks will always have a psize = 0 so * we ignore the compression of the blkptr and set the * want to uncompress them. Mark them as uncompressed. */ if (!zfs_compressed_arc_enabled || HDR_GET_PSIZE(hdr) == 0) { arc_hdr_clear_flags(hdr, ARC_FLAG_COMPRESSED_ARC); ASSERT(!HDR_COMPRESSION_ENABLED(hdr)); } else { arc_hdr_set_flags(hdr, ARC_FLAG_COMPRESSED_ARC); ASSERT(HDR_COMPRESSION_ENABLED(hdr)); } HDR_SET_COMPRESS(hdr, cmp); ASSERT3U(HDR_GET_COMPRESS(hdr), ==, cmp); } /* * Looks for another buf on the same hdr which has the data decompressed, copies * from it, and returns true. If no such buf exists, returns false. */ static boolean_t arc_buf_try_copy_decompressed_data(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; boolean_t copied = B_FALSE; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3P(buf->b_data, !=, NULL); ASSERT(!ARC_BUF_COMPRESSED(buf)); for (arc_buf_t *from = hdr->b_l1hdr.b_buf; from != NULL; from = from->b_next) { /* can't use our own data buffer */ if (from == buf) { continue; } if (!ARC_BUF_COMPRESSED(from)) { memcpy(buf->b_data, from->b_data, arc_buf_size(buf)); copied = B_TRUE; break; } } #ifdef ZFS_DEBUG /* * There were no decompressed bufs, so there should not be a * checksum on the hdr either. */ if (zfs_flags & ZFS_DEBUG_MODIFY) EQUIV(!copied, hdr->b_l1hdr.b_freeze_cksum == NULL); #endif return (copied); } /* * Allocates an ARC buf header that's in an evicted & L2-cached state. * This is used during l2arc reconstruction to make empty ARC buffers * which circumvent the regular disk->arc->l2arc path and instead come * into being in the reverse order, i.e. l2arc->arc. */ static arc_buf_hdr_t * arc_buf_alloc_l2only(size_t size, arc_buf_contents_t type, l2arc_dev_t *dev, dva_t dva, uint64_t daddr, int32_t psize, uint64_t birth, enum zio_compress compress, uint8_t complevel, boolean_t protected, boolean_t prefetch, arc_state_type_t arcs_state) { arc_buf_hdr_t *hdr; ASSERT(size != 0); hdr = kmem_cache_alloc(hdr_l2only_cache, KM_SLEEP); hdr->b_birth = birth; hdr->b_type = type; hdr->b_flags = 0; arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L2HDR); HDR_SET_LSIZE(hdr, size); HDR_SET_PSIZE(hdr, psize); arc_hdr_set_compress(hdr, compress); hdr->b_complevel = complevel; if (protected) arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED); if (prefetch) arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH); hdr->b_spa = spa_load_guid(dev->l2ad_vdev->vdev_spa); hdr->b_dva = dva; hdr->b_l2hdr.b_dev = dev; hdr->b_l2hdr.b_daddr = daddr; hdr->b_l2hdr.b_arcs_state = arcs_state; return (hdr); } /* * Return the size of the block, b_pabd, that is stored in the arc_buf_hdr_t. */ static uint64_t arc_hdr_size(arc_buf_hdr_t *hdr) { uint64_t size; if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF && HDR_GET_PSIZE(hdr) > 0) { size = HDR_GET_PSIZE(hdr); } else { ASSERT3U(HDR_GET_LSIZE(hdr), !=, 0); size = HDR_GET_LSIZE(hdr); } return (size); } static int arc_hdr_authenticate(arc_buf_hdr_t *hdr, spa_t *spa, uint64_t dsobj) { int ret; uint64_t csize; uint64_t lsize = HDR_GET_LSIZE(hdr); uint64_t psize = HDR_GET_PSIZE(hdr); void *tmpbuf = NULL; abd_t *abd = hdr->b_l1hdr.b_pabd; ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); ASSERT(HDR_AUTHENTICATED(hdr)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); /* * The MAC is calculated on the compressed data that is stored on disk. * However, if compressed arc is disabled we will only have the * decompressed data available to us now. Compress it into a temporary * abd so we can verify the MAC. The performance overhead of this will * be relatively low, since most objects in an encrypted objset will * be encrypted (instead of authenticated) anyway. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { csize = zio_compress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, &tmpbuf, lsize, hdr->b_complevel); ASSERT3P(tmpbuf, !=, NULL); ASSERT3U(csize, <=, psize); abd = abd_get_from_buf(tmpbuf, lsize); abd_take_ownership_of_buf(abd, B_TRUE); abd_zero_off(abd, csize, psize - csize); } /* * Authentication is best effort. We authenticate whenever the key is * available. If we succeed we clear ARC_FLAG_NOAUTH. */ if (hdr->b_crypt_hdr.b_ot == DMU_OT_OBJSET) { ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF); ASSERT3U(lsize, ==, psize); ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, abd, psize, hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); } else { ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, abd, psize, hdr->b_crypt_hdr.b_mac); } if (ret == 0) arc_hdr_clear_flags(hdr, ARC_FLAG_NOAUTH); else if (ret != ENOENT) goto error; if (tmpbuf != NULL) abd_free(abd); return (0); error: if (tmpbuf != NULL) abd_free(abd); return (ret); } /* * This function will take a header that only has raw encrypted data in * b_crypt_hdr.b_rabd and decrypt it into a new buffer which is stored in * b_l1hdr.b_pabd. If designated in the header flags, this function will * also decompress the data. */ static int arc_hdr_decrypt(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb) { int ret; abd_t *cabd = NULL; void *tmp = NULL; boolean_t no_crypt = B_FALSE; boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); ASSERT(HDR_ENCRYPTED(hdr)); arc_hdr_alloc_abd(hdr, 0); ret = spa_do_crypt_abd(B_FALSE, spa, zb, hdr->b_crypt_hdr.b_ot, B_FALSE, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv, hdr->b_crypt_hdr.b_mac, HDR_GET_PSIZE(hdr), hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd, &no_crypt); if (ret != 0) goto error; if (no_crypt) { abd_copy(hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd, HDR_GET_PSIZE(hdr)); } /* * If this header has disabled arc compression but the b_pabd is * compressed after decrypting it, we need to decompress the newly * decrypted data. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { /* * We want to make sure that we are correctly honoring the * zfs_abd_scatter_enabled setting, so we allocate an abd here * and then loan a buffer from it, rather than allocating a * linear buffer and wrapping it in an abd later. */ cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr, 0); tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr)); ret = zio_decompress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr), &hdr->b_complevel); if (ret != 0) { abd_return_buf(cabd, tmp, arc_hdr_size(hdr)); goto error; } abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr)); arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, arc_hdr_size(hdr), hdr); hdr->b_l1hdr.b_pabd = cabd; } return (0); error: arc_hdr_free_abd(hdr, B_FALSE); if (cabd != NULL) arc_free_data_buf(hdr, cabd, arc_hdr_size(hdr), hdr); return (ret); } /* * This function is called during arc_buf_fill() to prepare the header's * abd plaintext pointer for use. This involves authenticated protected * data and decrypting encrypted data into the plaintext abd. */ static int arc_fill_hdr_crypt(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, spa_t *spa, const zbookmark_phys_t *zb, boolean_t noauth) { int ret; ASSERT(HDR_PROTECTED(hdr)); if (hash_lock != NULL) mutex_enter(hash_lock); if (HDR_NOAUTH(hdr) && !noauth) { /* * The caller requested authenticated data but our data has * not been authenticated yet. Verify the MAC now if we can. */ ret = arc_hdr_authenticate(hdr, spa, zb->zb_objset); if (ret != 0) goto error; } else if (HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd == NULL) { /* * If we only have the encrypted version of the data, but the * unencrypted version was requested we take this opportunity * to store the decrypted version in the header for future use. */ ret = arc_hdr_decrypt(hdr, spa, zb); if (ret != 0) goto error; } ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); if (hash_lock != NULL) mutex_exit(hash_lock); return (0); error: if (hash_lock != NULL) mutex_exit(hash_lock); return (ret); } /* * This function is used by the dbuf code to decrypt bonus buffers in place. * The dbuf code itself doesn't have any locking for decrypting a shared dnode * block, so we use the hash lock here to protect against concurrent calls to * arc_buf_fill(). */ static void arc_buf_untransform_in_place(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(HDR_ENCRYPTED(hdr)); ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE); ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); ASSERT3PF(hdr->b_l1hdr.b_pabd, !=, NULL, "hdr %px buf %px", hdr, buf); zio_crypt_copy_dnode_bonus(hdr->b_l1hdr.b_pabd, buf->b_data, arc_buf_size(buf)); buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED; buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; } /* * Given a buf that has a data buffer attached to it, this function will * efficiently fill the buf with data of the specified compression setting from * the hdr and update the hdr's b_freeze_cksum if necessary. If the buf and hdr * are already sharing a data buf, no copy is performed. * * If the buf is marked as compressed but uncompressed data was requested, this * will allocate a new data buffer for the buf, remove that flag, and fill the * buf with uncompressed data. You can't request a compressed buf on a hdr with * uncompressed data, and (since we haven't added support for it yet) if you * want compressed data your buf must already be marked as compressed and have * the correct-sized data buffer. */ static int arc_buf_fill(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb, arc_fill_flags_t flags) { int error = 0; arc_buf_hdr_t *hdr = buf->b_hdr; boolean_t hdr_compressed = (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF); boolean_t compressed = (flags & ARC_FILL_COMPRESSED) != 0; boolean_t encrypted = (flags & ARC_FILL_ENCRYPTED) != 0; dmu_object_byteswap_t bswap = hdr->b_l1hdr.b_byteswap; kmutex_t *hash_lock = (flags & ARC_FILL_LOCKED) ? NULL : HDR_LOCK(hdr); ASSERT3P(buf->b_data, !=, NULL); IMPLY(compressed, hdr_compressed || ARC_BUF_ENCRYPTED(buf)); IMPLY(compressed, ARC_BUF_COMPRESSED(buf)); IMPLY(encrypted, HDR_ENCRYPTED(hdr)); IMPLY(encrypted, ARC_BUF_ENCRYPTED(buf)); IMPLY(encrypted, ARC_BUF_COMPRESSED(buf)); IMPLY(encrypted, !arc_buf_is_shared(buf)); /* * If the caller wanted encrypted data we just need to copy it from * b_rabd and potentially byteswap it. We won't be able to do any * further transforms on it. */ if (encrypted) { ASSERT(HDR_HAS_RABD(hdr)); abd_copy_to_buf(buf->b_data, hdr->b_crypt_hdr.b_rabd, HDR_GET_PSIZE(hdr)); goto byteswap; } /* * Adjust encrypted and authenticated headers to accommodate * the request if needed. Dnode blocks (ARC_FILL_IN_PLACE) are * allowed to fail decryption due to keys not being loaded * without being marked as an IO error. */ if (HDR_PROTECTED(hdr)) { error = arc_fill_hdr_crypt(hdr, hash_lock, spa, zb, !!(flags & ARC_FILL_NOAUTH)); if (error == EACCES && (flags & ARC_FILL_IN_PLACE) != 0) { return (error); } else if (error != 0) { if (hash_lock != NULL) mutex_enter(hash_lock); arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR); if (hash_lock != NULL) mutex_exit(hash_lock); return (error); } } /* * There is a special case here for dnode blocks which are * decrypting their bonus buffers. These blocks may request to * be decrypted in-place. This is necessary because there may * be many dnodes pointing into this buffer and there is * currently no method to synchronize replacing the backing * b_data buffer and updating all of the pointers. Here we use * the hash lock to ensure there are no races. If the need * arises for other types to be decrypted in-place, they must * add handling here as well. */ if ((flags & ARC_FILL_IN_PLACE) != 0) { ASSERT(!hdr_compressed); ASSERT(!compressed); ASSERT(!encrypted); if (HDR_ENCRYPTED(hdr) && ARC_BUF_ENCRYPTED(buf)) { ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE); if (hash_lock != NULL) mutex_enter(hash_lock); arc_buf_untransform_in_place(buf); if (hash_lock != NULL) mutex_exit(hash_lock); /* Compute the hdr's checksum if necessary */ arc_cksum_compute(buf); } return (0); } if (hdr_compressed == compressed) { if (ARC_BUF_SHARED(buf)) { ASSERT(arc_buf_is_shared(buf)); } else { abd_copy_to_buf(buf->b_data, hdr->b_l1hdr.b_pabd, arc_buf_size(buf)); } } else { ASSERT(hdr_compressed); ASSERT(!compressed); /* * If the buf is sharing its data with the hdr, unlink it and * allocate a new data buffer for the buf. */ if (ARC_BUF_SHARED(buf)) { ASSERTF(ARC_BUF_COMPRESSED(buf), "buf %p was uncompressed", buf); /* We need to give the buf its own b_data */ buf->b_flags &= ~ARC_BUF_FLAG_SHARED; buf->b_data = arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf); arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA); /* Previously overhead was 0; just add new overhead */ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr)); } else if (ARC_BUF_COMPRESSED(buf)) { ASSERT(!arc_buf_is_shared(buf)); /* We need to reallocate the buf's b_data */ arc_free_data_buf(hdr, buf->b_data, HDR_GET_PSIZE(hdr), buf); buf->b_data = arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf); /* We increased the size of b_data; update overhead */ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr) - HDR_GET_PSIZE(hdr)); } /* * Regardless of the buf's previous compression settings, it * should not be compressed at the end of this function. */ buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; /* * Try copying the data from another buf which already has a * decompressed version. If that's not possible, it's time to * bite the bullet and decompress the data from the hdr. */ if (arc_buf_try_copy_decompressed_data(buf)) { /* Skip byteswapping and checksumming (already done) */ return (0); } else { error = zio_decompress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, buf->b_data, HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr), &hdr->b_complevel); /* * Absent hardware errors or software bugs, this should * be impossible, but log it anyway so we can debug it. */ if (error != 0) { zfs_dbgmsg( "hdr %px, compress %d, psize %d, lsize %d", hdr, arc_hdr_get_compress(hdr), HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr)); if (hash_lock != NULL) mutex_enter(hash_lock); arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR); if (hash_lock != NULL) mutex_exit(hash_lock); return (SET_ERROR(EIO)); } } } byteswap: /* Byteswap the buf's data if necessary */ if (bswap != DMU_BSWAP_NUMFUNCS) { ASSERT(!HDR_SHARED_DATA(hdr)); ASSERT3U(bswap, <, DMU_BSWAP_NUMFUNCS); dmu_ot_byteswap[bswap].ob_func(buf->b_data, HDR_GET_LSIZE(hdr)); } /* Compute the hdr's checksum if necessary */ arc_cksum_compute(buf); return (0); } /* * If this function is being called to decrypt an encrypted buffer or verify an * authenticated one, the key must be loaded and a mapping must be made * available in the keystore via spa_keystore_create_mapping() or one of its * callers. */ int arc_untransform(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb, boolean_t in_place) { int ret; arc_fill_flags_t flags = 0; if (in_place) flags |= ARC_FILL_IN_PLACE; ret = arc_buf_fill(buf, spa, zb, flags); if (ret == ECKSUM) { /* * Convert authentication and decryption errors to EIO * (and generate an ereport) before leaving the ARC. */ ret = SET_ERROR(EIO); spa_log_error(spa, zb, buf->b_hdr->b_birth); (void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, zb, NULL, 0); } return (ret); } /* * Increment the amount of evictable space in the arc_state_t's refcount. * We account for the space used by the hdr and the arc buf individually * so that we can add and remove them from the refcount individually. */ static void arc_evictable_space_increment(arc_buf_hdr_t *hdr, arc_state_t *state) { arc_buf_contents_t type = arc_buf_type(hdr); ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(state)) { ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); (void) zfs_refcount_add_many(&state->arcs_esize[type], HDR_GET_LSIZE(hdr), hdr); return; } if (hdr->b_l1hdr.b_pabd != NULL) { (void) zfs_refcount_add_many(&state->arcs_esize[type], arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) zfs_refcount_add_many(&state->arcs_esize[type], HDR_GET_PSIZE(hdr), hdr); } for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { if (ARC_BUF_SHARED(buf)) continue; (void) zfs_refcount_add_many(&state->arcs_esize[type], arc_buf_size(buf), buf); } } /* * Decrement the amount of evictable space in the arc_state_t's refcount. * We account for the space used by the hdr and the arc buf individually * so that we can add and remove them from the refcount individually. */ static void arc_evictable_space_decrement(arc_buf_hdr_t *hdr, arc_state_t *state) { arc_buf_contents_t type = arc_buf_type(hdr); ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(state)) { ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); (void) zfs_refcount_remove_many(&state->arcs_esize[type], HDR_GET_LSIZE(hdr), hdr); return; } if (hdr->b_l1hdr.b_pabd != NULL) { (void) zfs_refcount_remove_many(&state->arcs_esize[type], arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) zfs_refcount_remove_many(&state->arcs_esize[type], HDR_GET_PSIZE(hdr), hdr); } for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { if (ARC_BUF_SHARED(buf)) continue; (void) zfs_refcount_remove_many(&state->arcs_esize[type], arc_buf_size(buf), buf); } } /* * Add a reference to this hdr indicating that someone is actively * referencing that memory. When the refcount transitions from 0 to 1, * we remove it from the respective arc_state_t list to indicate that * it is not evictable. */ static void add_reference(arc_buf_hdr_t *hdr, const void *tag) { arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT(HDR_HAS_L1HDR(hdr)); if (!HDR_EMPTY(hdr) && !MUTEX_HELD(HDR_LOCK(hdr))) { ASSERT(state == arc_anon); ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); } if ((zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) && state != arc_anon && state != arc_l2c_only) { /* We don't use the L2-only state list. */ multilist_remove(&state->arcs_list[arc_buf_type(hdr)], hdr); arc_evictable_space_decrement(hdr, state); } } /* * Remove a reference from this hdr. When the reference transitions from * 1 to 0 and we're not anonymous, then we add this hdr to the arc_state_t's * list making it eligible for eviction. */ static int remove_reference(arc_buf_hdr_t *hdr, const void *tag) { int cnt; arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(state == arc_anon || MUTEX_HELD(HDR_LOCK(hdr))); ASSERT(!GHOST_STATE(state)); /* arc_l2c_only counts as a ghost. */ if ((cnt = zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) != 0) return (cnt); if (state == arc_anon) { arc_hdr_destroy(hdr); return (0); } if (state == arc_uncached && !HDR_PREFETCH(hdr)) { arc_change_state(arc_anon, hdr); arc_hdr_destroy(hdr); return (0); } multilist_insert(&state->arcs_list[arc_buf_type(hdr)], hdr); arc_evictable_space_increment(hdr, state); return (0); } /* * Returns detailed information about a specific arc buffer. When the * state_index argument is set the function will calculate the arc header * list position for its arc state. Since this requires a linear traversal * callers are strongly encourage not to do this. However, it can be helpful * for targeted analysis so the functionality is provided. */ void arc_buf_info(arc_buf_t *ab, arc_buf_info_t *abi, int state_index) { (void) state_index; arc_buf_hdr_t *hdr = ab->b_hdr; l1arc_buf_hdr_t *l1hdr = NULL; l2arc_buf_hdr_t *l2hdr = NULL; arc_state_t *state = NULL; memset(abi, 0, sizeof (arc_buf_info_t)); if (hdr == NULL) return; abi->abi_flags = hdr->b_flags; if (HDR_HAS_L1HDR(hdr)) { l1hdr = &hdr->b_l1hdr; state = l1hdr->b_state; } if (HDR_HAS_L2HDR(hdr)) l2hdr = &hdr->b_l2hdr; if (l1hdr) { abi->abi_bufcnt = 0; for (arc_buf_t *buf = l1hdr->b_buf; buf; buf = buf->b_next) abi->abi_bufcnt++; abi->abi_access = l1hdr->b_arc_access; abi->abi_mru_hits = l1hdr->b_mru_hits; abi->abi_mru_ghost_hits = l1hdr->b_mru_ghost_hits; abi->abi_mfu_hits = l1hdr->b_mfu_hits; abi->abi_mfu_ghost_hits = l1hdr->b_mfu_ghost_hits; abi->abi_holds = zfs_refcount_count(&l1hdr->b_refcnt); } if (l2hdr) { abi->abi_l2arc_dattr = l2hdr->b_daddr; abi->abi_l2arc_hits = l2hdr->b_hits; } abi->abi_state_type = state ? state->arcs_state : ARC_STATE_ANON; abi->abi_state_contents = arc_buf_type(hdr); abi->abi_size = arc_hdr_size(hdr); } /* * Move the supplied buffer to the indicated state. The hash lock * for the buffer must be held by the caller. */ static void arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *hdr) { arc_state_t *old_state; int64_t refcnt; boolean_t update_old, update_new; arc_buf_contents_t type = arc_buf_type(hdr); /* * We almost always have an L1 hdr here, since we call arc_hdr_realloc() * in arc_read() when bringing a buffer out of the L2ARC. However, the * L1 hdr doesn't always exist when we change state to arc_anon before * destroying a header, in which case reallocating to add the L1 hdr is * pointless. */ if (HDR_HAS_L1HDR(hdr)) { old_state = hdr->b_l1hdr.b_state; refcnt = zfs_refcount_count(&hdr->b_l1hdr.b_refcnt); update_old = (hdr->b_l1hdr.b_buf != NULL || hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); IMPLY(GHOST_STATE(old_state), hdr->b_l1hdr.b_buf == NULL); IMPLY(GHOST_STATE(new_state), hdr->b_l1hdr.b_buf == NULL); IMPLY(old_state == arc_anon, hdr->b_l1hdr.b_buf == NULL || ARC_BUF_LAST(hdr->b_l1hdr.b_buf)); } else { old_state = arc_l2c_only; refcnt = 0; update_old = B_FALSE; } update_new = update_old; if (GHOST_STATE(old_state)) update_old = B_TRUE; if (GHOST_STATE(new_state)) update_new = B_TRUE; ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); ASSERT3P(new_state, !=, old_state); /* * If this buffer is evictable, transfer it from the * old state list to the new state list. */ if (refcnt == 0) { if (old_state != arc_anon && old_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); /* remove_reference() saves on insert. */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { multilist_remove(&old_state->arcs_list[type], hdr); arc_evictable_space_decrement(hdr, old_state); } } if (new_state != arc_anon && new_state != arc_l2c_only) { /* * An L1 header always exists here, since if we're * moving to some L1-cached state (i.e. not l2c_only or * anonymous), we realloc the header to add an L1hdr * beforehand. */ ASSERT(HDR_HAS_L1HDR(hdr)); multilist_insert(&new_state->arcs_list[type], hdr); arc_evictable_space_increment(hdr, new_state); } } ASSERT(!HDR_EMPTY(hdr)); if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr)) buf_hash_remove(hdr); /* adjust state sizes (ignore arc_l2c_only) */ if (update_new && new_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(new_state)) { /* * When moving a header to a ghost state, we first * remove all arc buffers. Thus, we'll have no arc * buffer to use for the reference. As a result, we * use the arc header pointer for the reference. */ (void) zfs_refcount_add_many( &new_state->arcs_size[type], HDR_GET_LSIZE(hdr), hdr); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); } else { /* * Each individual buffer holds a unique reference, * thus we must remove each of these references one * at a time. */ for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { /* * When the arc_buf_t is sharing the data * block with the hdr, the owner of the * reference belongs to the hdr. Only * add to the refcount if the arc_buf_t is * not shared. */ if (ARC_BUF_SHARED(buf)) continue; (void) zfs_refcount_add_many( &new_state->arcs_size[type], arc_buf_size(buf), buf); } if (hdr->b_l1hdr.b_pabd != NULL) { (void) zfs_refcount_add_many( &new_state->arcs_size[type], arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) zfs_refcount_add_many( &new_state->arcs_size[type], HDR_GET_PSIZE(hdr), hdr); } } } if (update_old && old_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(old_state)) { ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); /* * When moving a header off of a ghost state, * the header will not contain any arc buffers. * We use the arc header pointer for the reference * which is exactly what we did when we put the * header on the ghost state. */ (void) zfs_refcount_remove_many( &old_state->arcs_size[type], HDR_GET_LSIZE(hdr), hdr); } else { /* * Each individual buffer holds a unique reference, * thus we must remove each of these references one * at a time. */ for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { /* * When the arc_buf_t is sharing the data * block with the hdr, the owner of the * reference belongs to the hdr. Only * add to the refcount if the arc_buf_t is * not shared. */ if (ARC_BUF_SHARED(buf)) continue; (void) zfs_refcount_remove_many( &old_state->arcs_size[type], arc_buf_size(buf), buf); } ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); if (hdr->b_l1hdr.b_pabd != NULL) { (void) zfs_refcount_remove_many( &old_state->arcs_size[type], arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) zfs_refcount_remove_many( &old_state->arcs_size[type], HDR_GET_PSIZE(hdr), hdr); } } } if (HDR_HAS_L1HDR(hdr)) { hdr->b_l1hdr.b_state = new_state; if (HDR_HAS_L2HDR(hdr) && new_state != arc_l2c_only) { l2arc_hdr_arcstats_decrement_state(hdr); hdr->b_l2hdr.b_arcs_state = new_state->arcs_state; l2arc_hdr_arcstats_increment_state(hdr); } } } void arc_space_consume(uint64_t space, arc_space_type_t type) { ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES); switch (type) { default: break; case ARC_SPACE_DATA: ARCSTAT_INCR(arcstat_data_size, space); break; case ARC_SPACE_META: ARCSTAT_INCR(arcstat_metadata_size, space); break; case ARC_SPACE_BONUS: ARCSTAT_INCR(arcstat_bonus_size, space); break; case ARC_SPACE_DNODE: ARCSTAT_INCR(arcstat_dnode_size, space); break; case ARC_SPACE_DBUF: ARCSTAT_INCR(arcstat_dbuf_size, space); break; case ARC_SPACE_HDRS: ARCSTAT_INCR(arcstat_hdr_size, space); break; case ARC_SPACE_L2HDRS: aggsum_add(&arc_sums.arcstat_l2_hdr_size, space); break; case ARC_SPACE_ABD_CHUNK_WASTE: /* * Note: this includes space wasted by all scatter ABD's, not * just those allocated by the ARC. But the vast majority of * scatter ABD's come from the ARC, because other users are * very short-lived. */ ARCSTAT_INCR(arcstat_abd_chunk_waste_size, space); break; } if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE) ARCSTAT_INCR(arcstat_meta_used, space); aggsum_add(&arc_sums.arcstat_size, space); } void arc_space_return(uint64_t space, arc_space_type_t type) { ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES); switch (type) { default: break; case ARC_SPACE_DATA: ARCSTAT_INCR(arcstat_data_size, -space); break; case ARC_SPACE_META: ARCSTAT_INCR(arcstat_metadata_size, -space); break; case ARC_SPACE_BONUS: ARCSTAT_INCR(arcstat_bonus_size, -space); break; case ARC_SPACE_DNODE: ARCSTAT_INCR(arcstat_dnode_size, -space); break; case ARC_SPACE_DBUF: ARCSTAT_INCR(arcstat_dbuf_size, -space); break; case ARC_SPACE_HDRS: ARCSTAT_INCR(arcstat_hdr_size, -space); break; case ARC_SPACE_L2HDRS: aggsum_add(&arc_sums.arcstat_l2_hdr_size, -space); break; case ARC_SPACE_ABD_CHUNK_WASTE: ARCSTAT_INCR(arcstat_abd_chunk_waste_size, -space); break; } if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE) ARCSTAT_INCR(arcstat_meta_used, -space); ASSERT(aggsum_compare(&arc_sums.arcstat_size, space) >= 0); aggsum_add(&arc_sums.arcstat_size, -space); } /* * Given a hdr and a buf, returns whether that buf can share its b_data buffer * with the hdr's b_pabd. */ static boolean_t arc_can_share(arc_buf_hdr_t *hdr, arc_buf_t *buf) { /* * The criteria for sharing a hdr's data are: * 1. the buffer is not encrypted * 2. the hdr's compression matches the buf's compression * 3. the hdr doesn't need to be byteswapped * 4. the hdr isn't already being shared * 5. the buf is either compressed or it is the last buf in the hdr list * * Criterion #5 maintains the invariant that shared uncompressed * bufs must be the final buf in the hdr's b_buf list. Reading this, you * might ask, "if a compressed buf is allocated first, won't that be the * last thing in the list?", but in that case it's impossible to create * a shared uncompressed buf anyway (because the hdr must be compressed * to have the compressed buf). You might also think that #3 is * sufficient to make this guarantee, however it's possible * (specifically in the rare L2ARC write race mentioned in * arc_buf_alloc_impl()) there will be an existing uncompressed buf that * is shareable, but wasn't at the time of its allocation. Rather than * allow a new shared uncompressed buf to be created and then shuffle * the list around to make it the last element, this simply disallows * sharing if the new buf isn't the first to be added. */ ASSERT3P(buf->b_hdr, ==, hdr); boolean_t hdr_compressed = arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF; boolean_t buf_compressed = ARC_BUF_COMPRESSED(buf) != 0; return (!ARC_BUF_ENCRYPTED(buf) && buf_compressed == hdr_compressed && hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS && !HDR_SHARED_DATA(hdr) && (ARC_BUF_LAST(buf) || ARC_BUF_COMPRESSED(buf))); } /* * Allocate a buf for this hdr. If you care about the data that's in the hdr, * or if you want a compressed buffer, pass those flags in. Returns 0 if the * copy was made successfully, or an error code otherwise. */ static int arc_buf_alloc_impl(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb, const void *tag, boolean_t encrypted, boolean_t compressed, boolean_t noauth, boolean_t fill, arc_buf_t **ret) { arc_buf_t *buf; arc_fill_flags_t flags = ARC_FILL_LOCKED; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3U(HDR_GET_LSIZE(hdr), >, 0); VERIFY(hdr->b_type == ARC_BUFC_DATA || hdr->b_type == ARC_BUFC_METADATA); ASSERT3P(ret, !=, NULL); ASSERT3P(*ret, ==, NULL); IMPLY(encrypted, compressed); buf = *ret = kmem_cache_alloc(buf_cache, KM_PUSHPAGE); buf->b_hdr = hdr; buf->b_data = NULL; buf->b_next = hdr->b_l1hdr.b_buf; buf->b_flags = 0; add_reference(hdr, tag); /* * We're about to change the hdr's b_flags. We must either * hold the hash_lock or be undiscoverable. */ ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); /* * Only honor requests for compressed bufs if the hdr is actually * compressed. This must be overridden if the buffer is encrypted since * encrypted buffers cannot be decompressed. */ if (encrypted) { buf->b_flags |= ARC_BUF_FLAG_COMPRESSED; buf->b_flags |= ARC_BUF_FLAG_ENCRYPTED; flags |= ARC_FILL_COMPRESSED | ARC_FILL_ENCRYPTED; } else if (compressed && arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) { buf->b_flags |= ARC_BUF_FLAG_COMPRESSED; flags |= ARC_FILL_COMPRESSED; } if (noauth) { ASSERT0(encrypted); flags |= ARC_FILL_NOAUTH; } /* * If the hdr's data can be shared then we share the data buffer and * set the appropriate bit in the hdr's b_flags to indicate the hdr is * sharing it's b_pabd with the arc_buf_t. Otherwise, we allocate a new * buffer to store the buf's data. * * There are two additional restrictions here because we're sharing * hdr -> buf instead of the usual buf -> hdr. First, the hdr can't be * actively involved in an L2ARC write, because if this buf is used by * an arc_write() then the hdr's data buffer will be released when the * write completes, even though the L2ARC write might still be using it. * Second, the hdr's ABD must be linear so that the buf's user doesn't * need to be ABD-aware. It must be allocated via * zio_[data_]buf_alloc(), not as a page, because we need to be able * to abd_release_ownership_of_buf(), which isn't allowed on "linear * page" buffers because the ABD code needs to handle freeing them * specially. */ boolean_t can_share = arc_can_share(hdr, buf) && !HDR_L2_WRITING(hdr) && hdr->b_l1hdr.b_pabd != NULL && abd_is_linear(hdr->b_l1hdr.b_pabd) && !abd_is_linear_page(hdr->b_l1hdr.b_pabd); /* Set up b_data and sharing */ if (can_share) { buf->b_data = abd_to_buf(hdr->b_l1hdr.b_pabd); buf->b_flags |= ARC_BUF_FLAG_SHARED; arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA); } else { buf->b_data = arc_get_data_buf(hdr, arc_buf_size(buf), buf); ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf)); } VERIFY3P(buf->b_data, !=, NULL); hdr->b_l1hdr.b_buf = buf; /* * If the user wants the data from the hdr, we need to either copy or * decompress the data. */ if (fill) { ASSERT3P(zb, !=, NULL); return (arc_buf_fill(buf, spa, zb, flags)); } return (0); } static const char *arc_onloan_tag = "onloan"; static inline void arc_loaned_bytes_update(int64_t delta) { atomic_add_64(&arc_loaned_bytes, delta); /* assert that it did not wrap around */ ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0); } /* * Loan out an anonymous arc buffer. Loaned buffers are not counted as in * flight data by arc_tempreserve_space() until they are "returned". Loaned * buffers must be returned to the arc before they can be used by the DMU or * freed. */ arc_buf_t * arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size) { arc_buf_t *buf = arc_alloc_buf(spa, arc_onloan_tag, is_metadata ? ARC_BUFC_METADATA : ARC_BUFC_DATA, size); arc_loaned_bytes_update(arc_buf_size(buf)); return (buf); } arc_buf_t * arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize, enum zio_compress compression_type, uint8_t complevel) { arc_buf_t *buf = arc_alloc_compressed_buf(spa, arc_onloan_tag, psize, lsize, compression_type, complevel); arc_loaned_bytes_update(arc_buf_size(buf)); return (buf); } arc_buf_t * arc_loan_raw_buf(spa_t *spa, uint64_t dsobj, boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize, enum zio_compress compression_type, uint8_t complevel) { arc_buf_t *buf = arc_alloc_raw_buf(spa, arc_onloan_tag, dsobj, byteorder, salt, iv, mac, ot, psize, lsize, compression_type, complevel); atomic_add_64(&arc_loaned_bytes, psize); return (buf); } /* * Return a loaned arc buffer to the arc. */ void arc_return_buf(arc_buf_t *buf, const void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(buf->b_data, !=, NULL); ASSERT(HDR_HAS_L1HDR(hdr)); (void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag); (void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag); arc_loaned_bytes_update(-arc_buf_size(buf)); } /* Detach an arc_buf from a dbuf (tag) */ void arc_loan_inuse_buf(arc_buf_t *buf, const void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(buf->b_data, !=, NULL); ASSERT(HDR_HAS_L1HDR(hdr)); (void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag); (void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag); arc_loaned_bytes_update(arc_buf_size(buf)); } static void l2arc_free_abd_on_write(abd_t *abd, size_t size, arc_buf_contents_t type) { l2arc_data_free_t *df = kmem_alloc(sizeof (*df), KM_SLEEP); df->l2df_abd = abd; df->l2df_size = size; df->l2df_type = type; mutex_enter(&l2arc_free_on_write_mtx); list_insert_head(l2arc_free_on_write, df); mutex_exit(&l2arc_free_on_write_mtx); } static void arc_hdr_free_on_write(arc_buf_hdr_t *hdr, boolean_t free_rdata) { arc_state_t *state = hdr->b_l1hdr.b_state; arc_buf_contents_t type = arc_buf_type(hdr); uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr); /* protected by hash lock, if in the hash table */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT(state != arc_anon && state != arc_l2c_only); (void) zfs_refcount_remove_many(&state->arcs_esize[type], size, hdr); } (void) zfs_refcount_remove_many(&state->arcs_size[type], size, hdr); if (type == ARC_BUFC_METADATA) { arc_space_return(size, ARC_SPACE_META); } else { ASSERT(type == ARC_BUFC_DATA); arc_space_return(size, ARC_SPACE_DATA); } if (free_rdata) { l2arc_free_abd_on_write(hdr->b_crypt_hdr.b_rabd, size, type); } else { l2arc_free_abd_on_write(hdr->b_l1hdr.b_pabd, size, type); } } /* * Share the arc_buf_t's data with the hdr. Whenever we are sharing the * data buffer, we transfer the refcount ownership to the hdr and update * the appropriate kstats. */ static void arc_share_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(arc_can_share(hdr, buf)); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!ARC_BUF_ENCRYPTED(buf)); ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); /* * Start sharing the data buffer. We transfer the * refcount ownership to the hdr since it always owns * the refcount whenever an arc_buf_t is shared. */ zfs_refcount_transfer_ownership_many( &hdr->b_l1hdr.b_state->arcs_size[arc_buf_type(hdr)], arc_hdr_size(hdr), buf, hdr); hdr->b_l1hdr.b_pabd = abd_get_from_buf(buf->b_data, arc_buf_size(buf)); abd_take_ownership_of_buf(hdr->b_l1hdr.b_pabd, HDR_ISTYPE_METADATA(hdr)); arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA); buf->b_flags |= ARC_BUF_FLAG_SHARED; /* * Since we've transferred ownership to the hdr we need * to increment its compressed and uncompressed kstats and * decrement the overhead size. */ ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr)); ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr)); ARCSTAT_INCR(arcstat_overhead_size, -arc_buf_size(buf)); } static void arc_unshare_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(arc_buf_is_shared(buf)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); /* * We are no longer sharing this buffer so we need * to transfer its ownership to the rightful owner. */ zfs_refcount_transfer_ownership_many( &hdr->b_l1hdr.b_state->arcs_size[arc_buf_type(hdr)], arc_hdr_size(hdr), hdr, buf); arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA); abd_release_ownership_of_buf(hdr->b_l1hdr.b_pabd); abd_free(hdr->b_l1hdr.b_pabd); hdr->b_l1hdr.b_pabd = NULL; buf->b_flags &= ~ARC_BUF_FLAG_SHARED; /* * Since the buffer is no longer shared between * the arc buf and the hdr, count it as overhead. */ ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr)); ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr)); ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf)); } /* * Remove an arc_buf_t from the hdr's buf list and return the last * arc_buf_t on the list. If no buffers remain on the list then return * NULL. */ static arc_buf_t * arc_buf_remove(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); arc_buf_t **bufp = &hdr->b_l1hdr.b_buf; arc_buf_t *lastbuf = NULL; /* * Remove the buf from the hdr list and locate the last * remaining buffer on the list. */ while (*bufp != NULL) { if (*bufp == buf) *bufp = buf->b_next; /* * If we've removed a buffer in the middle of * the list then update the lastbuf and update * bufp. */ if (*bufp != NULL) { lastbuf = *bufp; bufp = &(*bufp)->b_next; } } buf->b_next = NULL; ASSERT3P(lastbuf, !=, buf); IMPLY(lastbuf != NULL, ARC_BUF_LAST(lastbuf)); return (lastbuf); } /* * Free up buf->b_data and pull the arc_buf_t off of the arc_buf_hdr_t's * list and free it. */ static void arc_buf_destroy_impl(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; /* * Free up the data associated with the buf but only if we're not * sharing this with the hdr. If we are sharing it with the hdr, the * hdr is responsible for doing the free. */ if (buf->b_data != NULL) { /* * We're about to change the hdr's b_flags. We must either * hold the hash_lock or be undiscoverable. */ ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); arc_cksum_verify(buf); arc_buf_unwatch(buf); if (ARC_BUF_SHARED(buf)) { arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA); } else { ASSERT(!arc_buf_is_shared(buf)); uint64_t size = arc_buf_size(buf); arc_free_data_buf(hdr, buf->b_data, size, buf); ARCSTAT_INCR(arcstat_overhead_size, -size); } buf->b_data = NULL; /* * If we have no more encrypted buffers and we've already * gotten a copy of the decrypted data we can free b_rabd * to save some space. */ if (ARC_BUF_ENCRYPTED(buf) && HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd != NULL && !HDR_IO_IN_PROGRESS(hdr)) { arc_buf_t *b; for (b = hdr->b_l1hdr.b_buf; b; b = b->b_next) { if (b != buf && ARC_BUF_ENCRYPTED(b)) break; } if (b == NULL) arc_hdr_free_abd(hdr, B_TRUE); } } arc_buf_t *lastbuf = arc_buf_remove(hdr, buf); if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) { /* * If the current arc_buf_t is sharing its data buffer with the * hdr, then reassign the hdr's b_pabd to share it with the new * buffer at the end of the list. The shared buffer is always * the last one on the hdr's buffer list. * * There is an equivalent case for compressed bufs, but since * they aren't guaranteed to be the last buf in the list and * that is an exceedingly rare case, we just allow that space be * wasted temporarily. We must also be careful not to share * encrypted buffers, since they cannot be shared. */ if (lastbuf != NULL && !ARC_BUF_ENCRYPTED(lastbuf)) { /* Only one buf can be shared at once */ ASSERT(!arc_buf_is_shared(lastbuf)); /* hdr is uncompressed so can't have compressed buf */ ASSERT(!ARC_BUF_COMPRESSED(lastbuf)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); arc_hdr_free_abd(hdr, B_FALSE); /* * We must setup a new shared block between the * last buffer and the hdr. The data would have * been allocated by the arc buf so we need to transfer * ownership to the hdr since it's now being shared. */ arc_share_buf(hdr, lastbuf); } } else if (HDR_SHARED_DATA(hdr)) { /* * Uncompressed shared buffers are always at the end * of the list. Compressed buffers don't have the * same requirements. This makes it hard to * simply assert that the lastbuf is shared so * we rely on the hdr's compression flags to determine * if we have a compressed, shared buffer. */ ASSERT3P(lastbuf, !=, NULL); ASSERT(arc_buf_is_shared(lastbuf) || arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF); } /* * Free the checksum if we're removing the last uncompressed buf from * this hdr. */ if (!arc_hdr_has_uncompressed_buf(hdr)) { arc_cksum_free(hdr); } /* clean up the buf */ buf->b_hdr = NULL; kmem_cache_free(buf_cache, buf); } static void arc_hdr_alloc_abd(arc_buf_hdr_t *hdr, int alloc_flags) { uint64_t size; boolean_t alloc_rdata = ((alloc_flags & ARC_HDR_ALLOC_RDATA) != 0); ASSERT3U(HDR_GET_LSIZE(hdr), >, 0); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!HDR_SHARED_DATA(hdr) || alloc_rdata); IMPLY(alloc_rdata, HDR_PROTECTED(hdr)); if (alloc_rdata) { size = HDR_GET_PSIZE(hdr); ASSERT3P(hdr->b_crypt_hdr.b_rabd, ==, NULL); hdr->b_crypt_hdr.b_rabd = arc_get_data_abd(hdr, size, hdr, alloc_flags); ASSERT3P(hdr->b_crypt_hdr.b_rabd, !=, NULL); ARCSTAT_INCR(arcstat_raw_size, size); } else { size = arc_hdr_size(hdr); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); hdr->b_l1hdr.b_pabd = arc_get_data_abd(hdr, size, hdr, alloc_flags); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); } ARCSTAT_INCR(arcstat_compressed_size, size); ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr)); } static void arc_hdr_free_abd(arc_buf_hdr_t *hdr, boolean_t free_rdata) { uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); IMPLY(free_rdata, HDR_HAS_RABD(hdr)); /* * If the hdr is currently being written to the l2arc then * we defer freeing the data by adding it to the l2arc_free_on_write * list. The l2arc will free the data once it's finished * writing it to the l2arc device. */ if (HDR_L2_WRITING(hdr)) { arc_hdr_free_on_write(hdr, free_rdata); ARCSTAT_BUMP(arcstat_l2_free_on_write); } else if (free_rdata) { arc_free_data_abd(hdr, hdr->b_crypt_hdr.b_rabd, size, hdr); } else { arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, size, hdr); } if (free_rdata) { hdr->b_crypt_hdr.b_rabd = NULL; ARCSTAT_INCR(arcstat_raw_size, -size); } else { hdr->b_l1hdr.b_pabd = NULL; } if (hdr->b_l1hdr.b_pabd == NULL && !HDR_HAS_RABD(hdr)) hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; ARCSTAT_INCR(arcstat_compressed_size, -size); ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr)); } /* * Allocate empty anonymous ARC header. The header will get its identity * assigned and buffers attached later as part of read or write operations. * * In case of read arc_read() assigns header its identify (b_dva + b_birth), * inserts it into ARC hash to become globally visible and allocates physical * (b_pabd) or raw (b_rabd) ABD buffer to read into from disk. On disk read * completion arc_read_done() allocates ARC buffer(s) as needed, potentially * sharing one of them with the physical ABD buffer. * * In case of write arc_alloc_buf() allocates ARC buffer to be filled with * data. Then after compression and/or encryption arc_write_ready() allocates * and fills (or potentially shares) physical (b_pabd) or raw (b_rabd) ABD * buffer. On disk write completion arc_write_done() assigns the header its * new identity (b_dva + b_birth) and inserts into ARC hash. * * In case of partial overwrite the old data is read first as described. Then * arc_release() either allocates new anonymous ARC header and moves the ARC * buffer to it, or reuses the old ARC header by discarding its identity and * removing it from ARC hash. After buffer modification normal write process * follows as described. */ static arc_buf_hdr_t * arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize, boolean_t protected, enum zio_compress compression_type, uint8_t complevel, arc_buf_contents_t type) { arc_buf_hdr_t *hdr; VERIFY(type == ARC_BUFC_DATA || type == ARC_BUFC_METADATA); hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE); ASSERT(HDR_EMPTY(hdr)); #ifdef ZFS_DEBUG ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); #endif HDR_SET_PSIZE(hdr, psize); HDR_SET_LSIZE(hdr, lsize); hdr->b_spa = spa; hdr->b_type = type; hdr->b_flags = 0; arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L1HDR); arc_hdr_set_compress(hdr, compression_type); hdr->b_complevel = complevel; if (protected) arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED); hdr->b_l1hdr.b_state = arc_anon; hdr->b_l1hdr.b_arc_access = 0; hdr->b_l1hdr.b_mru_hits = 0; hdr->b_l1hdr.b_mru_ghost_hits = 0; hdr->b_l1hdr.b_mfu_hits = 0; hdr->b_l1hdr.b_mfu_ghost_hits = 0; hdr->b_l1hdr.b_buf = NULL; ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); return (hdr); } /* * Transition between the two allocation states for the arc_buf_hdr struct. * The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without * (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller * version is used when a cache buffer is only in the L2ARC in order to reduce * memory usage. */ static arc_buf_hdr_t * arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new) { ASSERT(HDR_HAS_L2HDR(hdr)); arc_buf_hdr_t *nhdr; l2arc_dev_t *dev = hdr->b_l2hdr.b_dev; ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) || (old == hdr_l2only_cache && new == hdr_full_cache)); nhdr = kmem_cache_alloc(new, KM_PUSHPAGE); ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); buf_hash_remove(hdr); memcpy(nhdr, hdr, HDR_L2ONLY_SIZE); if (new == hdr_full_cache) { arc_hdr_set_flags(nhdr, ARC_FLAG_HAS_L1HDR); /* * arc_access and arc_change_state need to be aware that a * header has just come out of L2ARC, so we set its state to * l2c_only even though it's about to change. */ nhdr->b_l1hdr.b_state = arc_l2c_only; /* Verify previous threads set to NULL before freeing */ ASSERT3P(nhdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); } else { ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); #ifdef ZFS_DEBUG ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); #endif /* * If we've reached here, We must have been called from * arc_evict_hdr(), as such we should have already been * removed from any ghost list we were previously on * (which protects us from racing with arc_evict_state), * thus no locking is needed during this check. */ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); /* * A buffer must not be moved into the arc_l2c_only * state if it's not finished being written out to the * l2arc device. Otherwise, the b_l1hdr.b_pabd field * might try to be accessed, even though it was removed. */ VERIFY(!HDR_L2_WRITING(hdr)); VERIFY3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR); } /* * The header has been reallocated so we need to re-insert it into any * lists it was on. */ (void) buf_hash_insert(nhdr, NULL); ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node)); mutex_enter(&dev->l2ad_mtx); /* * We must place the realloc'ed header back into the list at * the same spot. Otherwise, if it's placed earlier in the list, * l2arc_write_buffers() could find it during the function's * write phase, and try to write it out to the l2arc. */ list_insert_after(&dev->l2ad_buflist, hdr, nhdr); list_remove(&dev->l2ad_buflist, hdr); mutex_exit(&dev->l2ad_mtx); /* * Since we're using the pointer address as the tag when * incrementing and decrementing the l2ad_alloc refcount, we * must remove the old pointer (that we're about to destroy) and * add the new pointer to the refcount. Otherwise we'd remove * the wrong pointer address when calling arc_hdr_destroy() later. */ (void) zfs_refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); (void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(nhdr), nhdr); buf_discard_identity(hdr); kmem_cache_free(old, hdr); return (nhdr); } /* * This function is used by the send / receive code to convert a newly * allocated arc_buf_t to one that is suitable for a raw encrypted write. It * is also used to allow the root objset block to be updated without altering * its embedded MACs. Both block types will always be uncompressed so we do not * have to worry about compression type or psize. */ void arc_convert_to_raw(arc_buf_t *buf, uint64_t dsobj, boolean_t byteorder, dmu_object_type_t ot, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(ot == DMU_OT_DNODE || ot == DMU_OT_OBJSET); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); buf->b_flags |= (ARC_BUF_FLAG_COMPRESSED | ARC_BUF_FLAG_ENCRYPTED); arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED); hdr->b_crypt_hdr.b_dsobj = dsobj; hdr->b_crypt_hdr.b_ot = ot; hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ? DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot); if (!arc_hdr_has_uncompressed_buf(hdr)) arc_cksum_free(hdr); if (salt != NULL) memcpy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN); if (iv != NULL) memcpy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN); if (mac != NULL) memcpy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN); } /* * Allocate a new arc_buf_hdr_t and arc_buf_t and return the buf to the caller. * The buf is returned thawed since we expect the consumer to modify it. */ arc_buf_t * arc_alloc_buf(spa_t *spa, const void *tag, arc_buf_contents_t type, int32_t size) { arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size, B_FALSE, ZIO_COMPRESS_OFF, 0, type); arc_buf_t *buf = NULL; VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_FALSE, B_FALSE, B_FALSE, &buf)); arc_buf_thaw(buf); return (buf); } /* * Allocate a compressed buf in the same manner as arc_alloc_buf. Don't use this * for bufs containing metadata. */ arc_buf_t * arc_alloc_compressed_buf(spa_t *spa, const void *tag, uint64_t psize, uint64_t lsize, enum zio_compress compression_type, uint8_t complevel) { ASSERT3U(lsize, >, 0); ASSERT3U(lsize, >=, psize); ASSERT3U(compression_type, >, ZIO_COMPRESS_OFF); ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS); arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_FALSE, compression_type, complevel, ARC_BUFC_DATA); arc_buf_t *buf = NULL; VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_TRUE, B_FALSE, B_FALSE, &buf)); arc_buf_thaw(buf); /* * To ensure that the hdr has the correct data in it if we call * arc_untransform() on this buf before it's been written to disk, * it's easiest if we just set up sharing between the buf and the hdr. */ arc_share_buf(hdr, buf); return (buf); } arc_buf_t * arc_alloc_raw_buf(spa_t *spa, const void *tag, uint64_t dsobj, boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize, enum zio_compress compression_type, uint8_t complevel) { arc_buf_hdr_t *hdr; arc_buf_t *buf; arc_buf_contents_t type = DMU_OT_IS_METADATA(ot) ? ARC_BUFC_METADATA : ARC_BUFC_DATA; ASSERT3U(lsize, >, 0); ASSERT3U(lsize, >=, psize); ASSERT3U(compression_type, >=, ZIO_COMPRESS_OFF); ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS); hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_TRUE, compression_type, complevel, type); hdr->b_crypt_hdr.b_dsobj = dsobj; hdr->b_crypt_hdr.b_ot = ot; hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ? DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot); memcpy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN); memcpy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN); memcpy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN); /* * This buffer will be considered encrypted even if the ot is not an * encrypted type. It will become authenticated instead in * arc_write_ready(). */ buf = NULL; VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_TRUE, B_TRUE, B_FALSE, B_FALSE, &buf)); arc_buf_thaw(buf); return (buf); } static void l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr, boolean_t state_only) { l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr; l2arc_dev_t *dev = l2hdr->b_dev; uint64_t lsize = HDR_GET_LSIZE(hdr); uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize); arc_buf_contents_t type = hdr->b_type; int64_t lsize_s; int64_t psize_s; int64_t asize_s; if (incr) { lsize_s = lsize; psize_s = psize; asize_s = asize; } else { lsize_s = -lsize; psize_s = -psize; asize_s = -asize; } /* If the buffer is a prefetch, count it as such. */ if (HDR_PREFETCH(hdr)) { ARCSTAT_INCR(arcstat_l2_prefetch_asize, asize_s); } else { /* * We use the value stored in the L2 header upon initial * caching in L2ARC. This value will be updated in case * an MRU/MRU_ghost buffer transitions to MFU but the L2ARC * metadata (log entry) cannot currently be updated. Having * the ARC state in the L2 header solves the problem of a * possibly absent L1 header (apparent in buffers restored * from persistent L2ARC). */ switch (hdr->b_l2hdr.b_arcs_state) { case ARC_STATE_MRU_GHOST: case ARC_STATE_MRU: ARCSTAT_INCR(arcstat_l2_mru_asize, asize_s); break; case ARC_STATE_MFU_GHOST: case ARC_STATE_MFU: ARCSTAT_INCR(arcstat_l2_mfu_asize, asize_s); break; default: break; } } if (state_only) return; ARCSTAT_INCR(arcstat_l2_psize, psize_s); ARCSTAT_INCR(arcstat_l2_lsize, lsize_s); switch (type) { case ARC_BUFC_DATA: ARCSTAT_INCR(arcstat_l2_bufc_data_asize, asize_s); break; case ARC_BUFC_METADATA: ARCSTAT_INCR(arcstat_l2_bufc_metadata_asize, asize_s); break; default: break; } } static void arc_hdr_l2hdr_destroy(arc_buf_hdr_t *hdr) { l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr; l2arc_dev_t *dev = l2hdr->b_dev; uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize); ASSERT(MUTEX_HELD(&dev->l2ad_mtx)); ASSERT(HDR_HAS_L2HDR(hdr)); list_remove(&dev->l2ad_buflist, hdr); l2arc_hdr_arcstats_decrement(hdr); vdev_space_update(dev->l2ad_vdev, -asize, 0, 0); (void) zfs_refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR); } static void arc_hdr_destroy(arc_buf_hdr_t *hdr) { if (HDR_HAS_L1HDR(hdr)) { ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); } ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(!HDR_IN_HASH_TABLE(hdr)); if (HDR_HAS_L2HDR(hdr)) { l2arc_dev_t *dev = hdr->b_l2hdr.b_dev; boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx); if (!buflist_held) mutex_enter(&dev->l2ad_mtx); /* * Even though we checked this conditional above, we * need to check this again now that we have the * l2ad_mtx. This is because we could be racing with * another thread calling l2arc_evict() which might have * destroyed this header's L2 portion as we were waiting * to acquire the l2ad_mtx. If that happens, we don't * want to re-destroy the header's L2 portion. */ if (HDR_HAS_L2HDR(hdr)) { if (!HDR_EMPTY(hdr)) buf_discard_identity(hdr); arc_hdr_l2hdr_destroy(hdr); } if (!buflist_held) mutex_exit(&dev->l2ad_mtx); } /* * The header's identify can only be safely discarded once it is no * longer discoverable. This requires removing it from the hash table * and the l2arc header list. After this point the hash lock can not * be used to protect the header. */ if (!HDR_EMPTY(hdr)) buf_discard_identity(hdr); if (HDR_HAS_L1HDR(hdr)) { arc_cksum_free(hdr); while (hdr->b_l1hdr.b_buf != NULL) arc_buf_destroy_impl(hdr->b_l1hdr.b_buf); if (hdr->b_l1hdr.b_pabd != NULL) arc_hdr_free_abd(hdr, B_FALSE); if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); } ASSERT3P(hdr->b_hash_next, ==, NULL); if (HDR_HAS_L1HDR(hdr)) { ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); #ifdef ZFS_DEBUG ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); #endif kmem_cache_free(hdr_full_cache, hdr); } else { kmem_cache_free(hdr_l2only_cache, hdr); } } void arc_buf_destroy(arc_buf_t *buf, const void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; if (hdr->b_l1hdr.b_state == arc_anon) { ASSERT3P(hdr->b_l1hdr.b_buf, ==, buf); ASSERT(ARC_BUF_LAST(buf)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); VERIFY0(remove_reference(hdr, tag)); return; } kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); ASSERT3P(hdr, ==, buf->b_hdr); ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL); ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT3P(hdr->b_l1hdr.b_state, !=, arc_anon); ASSERT3P(buf->b_data, !=, NULL); arc_buf_destroy_impl(buf); (void) remove_reference(hdr, tag); mutex_exit(hash_lock); } /* * Evict the arc_buf_hdr that is provided as a parameter. The resultant * state of the header is dependent on its state prior to entering this * function. The following transitions are possible: * * - arc_mru -> arc_mru_ghost * - arc_mfu -> arc_mfu_ghost * - arc_mru_ghost -> arc_l2c_only * - arc_mru_ghost -> deleted * - arc_mfu_ghost -> arc_l2c_only * - arc_mfu_ghost -> deleted * - arc_uncached -> deleted * * Return total size of evicted data buffers for eviction progress tracking. * When evicting from ghost states return logical buffer size to make eviction * progress at the same (or at least comparable) rate as from non-ghost states. * * Return *real_evicted for actual ARC size reduction to wake up threads * waiting for it. For non-ghost states it includes size of evicted data * buffers (the headers are not freed there). For ghost states it includes * only the evicted headers size. */ static int64_t arc_evict_hdr(arc_buf_hdr_t *hdr, uint64_t *real_evicted) { arc_state_t *evicted_state, *state; int64_t bytes_evicted = 0; uint_t min_lifetime = HDR_PRESCIENT_PREFETCH(hdr) ? arc_min_prescient_prefetch_ms : arc_min_prefetch_ms; ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt)); *real_evicted = 0; state = hdr->b_l1hdr.b_state; if (GHOST_STATE(state)) { /* * l2arc_write_buffers() relies on a header's L1 portion * (i.e. its b_pabd field) during it's write phase. * Thus, we cannot push a header onto the arc_l2c_only * state (removing its L1 piece) until the header is * done being written to the l2arc. */ if (HDR_HAS_L2HDR(hdr) && HDR_L2_WRITING(hdr)) { ARCSTAT_BUMP(arcstat_evict_l2_skip); return (bytes_evicted); } ARCSTAT_BUMP(arcstat_deleted); bytes_evicted += HDR_GET_LSIZE(hdr); DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr); if (HDR_HAS_L2HDR(hdr)) { ASSERT(hdr->b_l1hdr.b_pabd == NULL); ASSERT(!HDR_HAS_RABD(hdr)); /* * This buffer is cached on the 2nd Level ARC; * don't destroy the header. */ arc_change_state(arc_l2c_only, hdr); /* * dropping from L1+L2 cached to L2-only, * realloc to remove the L1 header. */ (void) arc_hdr_realloc(hdr, hdr_full_cache, hdr_l2only_cache); *real_evicted += HDR_FULL_SIZE - HDR_L2ONLY_SIZE; } else { arc_change_state(arc_anon, hdr); arc_hdr_destroy(hdr); *real_evicted += HDR_FULL_SIZE; } return (bytes_evicted); } ASSERT(state == arc_mru || state == arc_mfu || state == arc_uncached); evicted_state = (state == arc_uncached) ? arc_anon : ((state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost); /* prefetch buffers have a minimum lifespan */ if ((hdr->b_flags & (ARC_FLAG_PREFETCH | ARC_FLAG_INDIRECT)) && ddi_get_lbolt() - hdr->b_l1hdr.b_arc_access < MSEC_TO_TICK(min_lifetime)) { ARCSTAT_BUMP(arcstat_evict_skip); return (bytes_evicted); } if (HDR_HAS_L2HDR(hdr)) { ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr)); } else { if (l2arc_write_eligible(hdr->b_spa, hdr)) { ARCSTAT_INCR(arcstat_evict_l2_eligible, HDR_GET_LSIZE(hdr)); switch (state->arcs_state) { case ARC_STATE_MRU: ARCSTAT_INCR( arcstat_evict_l2_eligible_mru, HDR_GET_LSIZE(hdr)); break; case ARC_STATE_MFU: ARCSTAT_INCR( arcstat_evict_l2_eligible_mfu, HDR_GET_LSIZE(hdr)); break; default: break; } } else { ARCSTAT_INCR(arcstat_evict_l2_ineligible, HDR_GET_LSIZE(hdr)); } } bytes_evicted += arc_hdr_size(hdr); *real_evicted += arc_hdr_size(hdr); /* * If this hdr is being evicted and has a compressed buffer then we * discard it here before we change states. This ensures that the * accounting is updated correctly in arc_free_data_impl(). */ if (hdr->b_l1hdr.b_pabd != NULL) arc_hdr_free_abd(hdr, B_FALSE); if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); arc_change_state(evicted_state, hdr); DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr); if (evicted_state == arc_anon) { arc_hdr_destroy(hdr); *real_evicted += HDR_FULL_SIZE; } else { ASSERT(HDR_IN_HASH_TABLE(hdr)); } return (bytes_evicted); } static void arc_set_need_free(void) { ASSERT(MUTEX_HELD(&arc_evict_lock)); int64_t remaining = arc_free_memory() - arc_sys_free / 2; arc_evict_waiter_t *aw = list_tail(&arc_evict_waiters); if (aw == NULL) { arc_need_free = MAX(-remaining, 0); } else { arc_need_free = MAX(-remaining, (int64_t)(aw->aew_count - arc_evict_count)); } } static uint64_t arc_evict_state_impl(multilist_t *ml, int idx, arc_buf_hdr_t *marker, uint64_t spa, uint64_t bytes) { multilist_sublist_t *mls; uint64_t bytes_evicted = 0, real_evicted = 0; arc_buf_hdr_t *hdr; kmutex_t *hash_lock; uint_t evict_count = zfs_arc_evict_batch_limit; ASSERT3P(marker, !=, NULL); mls = multilist_sublist_lock_idx(ml, idx); for (hdr = multilist_sublist_prev(mls, marker); likely(hdr != NULL); hdr = multilist_sublist_prev(mls, marker)) { if ((evict_count == 0) || (bytes_evicted >= bytes)) break; /* * To keep our iteration location, move the marker * forward. Since we're not holding hdr's hash lock, we * must be very careful and not remove 'hdr' from the * sublist. Otherwise, other consumers might mistake the * 'hdr' as not being on a sublist when they call the * multilist_link_active() function (they all rely on * the hash lock protecting concurrent insertions and * removals). multilist_sublist_move_forward() was * specifically implemented to ensure this is the case * (only 'marker' will be removed and re-inserted). */ multilist_sublist_move_forward(mls, marker); /* * The only case where the b_spa field should ever be * zero, is the marker headers inserted by * arc_evict_state(). It's possible for multiple threads * to be calling arc_evict_state() concurrently (e.g. * dsl_pool_close() and zio_inject_fault()), so we must * skip any markers we see from these other threads. */ if (hdr->b_spa == 0) continue; /* we're only interested in evicting buffers of a certain spa */ if (spa != 0 && hdr->b_spa != spa) { ARCSTAT_BUMP(arcstat_evict_skip); continue; } hash_lock = HDR_LOCK(hdr); /* * We aren't calling this function from any code path * that would already be holding a hash lock, so we're * asserting on this assumption to be defensive in case * this ever changes. Without this check, it would be * possible to incorrectly increment arcstat_mutex_miss * below (e.g. if the code changed such that we called * this function with a hash lock held). */ ASSERT(!MUTEX_HELD(hash_lock)); if (mutex_tryenter(hash_lock)) { uint64_t revicted; uint64_t evicted = arc_evict_hdr(hdr, &revicted); mutex_exit(hash_lock); bytes_evicted += evicted; real_evicted += revicted; /* * If evicted is zero, arc_evict_hdr() must have * decided to skip this header, don't increment * evict_count in this case. */ if (evicted != 0) evict_count--; } else { ARCSTAT_BUMP(arcstat_mutex_miss); } } multilist_sublist_unlock(mls); /* * Increment the count of evicted bytes, and wake up any threads that * are waiting for the count to reach this value. Since the list is * ordered by ascending aew_count, we pop off the beginning of the * list until we reach the end, or a waiter that's past the current * "count". Doing this outside the loop reduces the number of times * we need to acquire the global arc_evict_lock. * * Only wake when there's sufficient free memory in the system * (specifically, arc_sys_free/2, which by default is a bit more than * 1/64th of RAM). See the comments in arc_wait_for_eviction(). */ mutex_enter(&arc_evict_lock); arc_evict_count += real_evicted; if (arc_free_memory() > arc_sys_free / 2) { arc_evict_waiter_t *aw; while ((aw = list_head(&arc_evict_waiters)) != NULL && aw->aew_count <= arc_evict_count) { list_remove(&arc_evict_waiters, aw); cv_broadcast(&aw->aew_cv); } } arc_set_need_free(); mutex_exit(&arc_evict_lock); /* * If the ARC size is reduced from arc_c_max to arc_c_min (especially * if the average cached block is small), eviction can be on-CPU for * many seconds. To ensure that other threads that may be bound to * this CPU are able to make progress, make a voluntary preemption * call here. */ kpreempt(KPREEMPT_SYNC); return (bytes_evicted); } static arc_buf_hdr_t * arc_state_alloc_marker(void) { arc_buf_hdr_t *marker = kmem_cache_alloc(hdr_full_cache, KM_SLEEP); /* * A b_spa of 0 is used to indicate that this header is * a marker. This fact is used in arc_evict_state_impl(). */ marker->b_spa = 0; return (marker); } static void arc_state_free_marker(arc_buf_hdr_t *marker) { kmem_cache_free(hdr_full_cache, marker); } /* * Allocate an array of buffer headers used as placeholders during arc state * eviction. */ static arc_buf_hdr_t ** arc_state_alloc_markers(int count) { arc_buf_hdr_t **markers; markers = kmem_zalloc(sizeof (*markers) * count, KM_SLEEP); for (int i = 0; i < count; i++) markers[i] = arc_state_alloc_marker(); return (markers); } static void arc_state_free_markers(arc_buf_hdr_t **markers, int count) { for (int i = 0; i < count; i++) arc_state_free_marker(markers[i]); kmem_free(markers, sizeof (*markers) * count); } /* * Evict buffers from the given arc state, until we've removed the * specified number of bytes. Move the removed buffers to the * appropriate evict state. * * This function makes a "best effort". It skips over any buffers * it can't get a hash_lock on, and so, may not catch all candidates. * It may also return without evicting as much space as requested. * * If bytes is specified using the special value ARC_EVICT_ALL, this * will evict all available (i.e. unlocked and evictable) buffers from * the given arc state; which is used by arc_flush(). */ static uint64_t arc_evict_state(arc_state_t *state, arc_buf_contents_t type, uint64_t spa, uint64_t bytes) { uint64_t total_evicted = 0; multilist_t *ml = &state->arcs_list[type]; int num_sublists; arc_buf_hdr_t **markers; num_sublists = multilist_get_num_sublists(ml); /* * If we've tried to evict from each sublist, made some * progress, but still have not hit the target number of bytes * to evict, we want to keep trying. The markers allow us to * pick up where we left off for each individual sublist, rather * than starting from the tail each time. */ if (zthr_iscurthread(arc_evict_zthr)) { markers = arc_state_evict_markers; ASSERT3S(num_sublists, <=, arc_state_evict_marker_count); } else { markers = arc_state_alloc_markers(num_sublists); } for (int i = 0; i < num_sublists; i++) { multilist_sublist_t *mls; mls = multilist_sublist_lock_idx(ml, i); multilist_sublist_insert_tail(mls, markers[i]); multilist_sublist_unlock(mls); } /* * While we haven't hit our target number of bytes to evict, or * we're evicting all available buffers. */ while (total_evicted < bytes) { int sublist_idx = multilist_get_random_index(ml); uint64_t scan_evicted = 0; /* * Start eviction using a randomly selected sublist, * this is to try and evenly balance eviction across all * sublists. Always starting at the same sublist * (e.g. index 0) would cause evictions to favor certain * sublists over others. */ for (int i = 0; i < num_sublists; i++) { uint64_t bytes_remaining; uint64_t bytes_evicted; if (total_evicted < bytes) bytes_remaining = bytes - total_evicted; else break; bytes_evicted = arc_evict_state_impl(ml, sublist_idx, markers[sublist_idx], spa, bytes_remaining); scan_evicted += bytes_evicted; total_evicted += bytes_evicted; /* we've reached the end, wrap to the beginning */ if (++sublist_idx >= num_sublists) sublist_idx = 0; } /* * If we didn't evict anything during this scan, we have * no reason to believe we'll evict more during another * scan, so break the loop. */ if (scan_evicted == 0) { /* This isn't possible, let's make that obvious */ ASSERT3S(bytes, !=, 0); /* * When bytes is ARC_EVICT_ALL, the only way to * break the loop is when scan_evicted is zero. * In that case, we actually have evicted enough, * so we don't want to increment the kstat. */ if (bytes != ARC_EVICT_ALL) { ASSERT3S(total_evicted, <, bytes); ARCSTAT_BUMP(arcstat_evict_not_enough); } break; } } for (int i = 0; i < num_sublists; i++) { multilist_sublist_t *mls = multilist_sublist_lock_idx(ml, i); multilist_sublist_remove(mls, markers[i]); multilist_sublist_unlock(mls); } if (markers != arc_state_evict_markers) arc_state_free_markers(markers, num_sublists); return (total_evicted); } /* * Flush all "evictable" data of the given type from the arc state * specified. This will not evict any "active" buffers (i.e. referenced). * * When 'retry' is set to B_FALSE, the function will make a single pass * over the state and evict any buffers that it can. Since it doesn't * continually retry the eviction, it might end up leaving some buffers * in the ARC due to lock misses. * * When 'retry' is set to B_TRUE, the function will continually retry the * eviction until *all* evictable buffers have been removed from the * state. As a result, if concurrent insertions into the state are * allowed (e.g. if the ARC isn't shutting down), this function might * wind up in an infinite loop, continually trying to evict buffers. */ static uint64_t arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type, boolean_t retry) { uint64_t evicted = 0; while (zfs_refcount_count(&state->arcs_esize[type]) != 0) { evicted += arc_evict_state(state, type, spa, ARC_EVICT_ALL); if (!retry) break; } return (evicted); } /* * Evict the specified number of bytes from the state specified. This * function prevents us from trying to evict more from a state's list * than is "evictable", and to skip evicting altogether when passed a * negative value for "bytes". In contrast, arc_evict_state() will * evict everything it can, when passed a negative value for "bytes". */ static uint64_t arc_evict_impl(arc_state_t *state, arc_buf_contents_t type, int64_t bytes) { uint64_t delta; if (bytes > 0 && zfs_refcount_count(&state->arcs_esize[type]) > 0) { delta = MIN(zfs_refcount_count(&state->arcs_esize[type]), bytes); return (arc_evict_state(state, type, 0, delta)); } return (0); } /* * Adjust specified fraction, taking into account initial ghost state(s) size, * ghost hit bytes towards increasing the fraction, ghost hit bytes towards * decreasing it, plus a balance factor, controlling the decrease rate, used * to balance metadata vs data. */ static uint64_t arc_evict_adj(uint64_t frac, uint64_t total, uint64_t up, uint64_t down, uint_t balance) { if (total < 8 || up + down == 0) return (frac); /* * We should not have more ghost hits than ghost size, but they * may get close. Restrict maximum adjustment in that case. */ if (up + down >= total / 4) { uint64_t scale = (up + down) / (total / 8); up /= scale; down /= scale; } /* Get maximal dynamic range by choosing optimal shifts. */ int s = highbit64(total); s = MIN(64 - s, 32); uint64_t ofrac = (1ULL << 32) - frac; if (frac >= 4 * ofrac) up /= frac / (2 * ofrac + 1); up = (up << s) / (total >> (32 - s)); if (ofrac >= 4 * frac) down /= ofrac / (2 * frac + 1); down = (down << s) / (total >> (32 - s)); down = down * 100 / balance; return (frac + up - down); } /* * Evict buffers from the cache, such that arcstat_size is capped by arc_c. */ static uint64_t arc_evict(void) { uint64_t asize, bytes, total_evicted = 0; int64_t e, mrud, mrum, mfud, mfum, w; static uint64_t ogrd, ogrm, ogfd, ogfm; static uint64_t gsrd, gsrm, gsfd, gsfm; uint64_t ngrd, ngrm, ngfd, ngfm; /* Get current size of ARC states we can evict from. */ mrud = zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_DATA]) + zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_DATA]); mrum = zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_METADATA]) + zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_METADATA]); mfud = zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_DATA]); mfum = zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_METADATA]); uint64_t d = mrud + mfud; uint64_t m = mrum + mfum; uint64_t t = d + m; /* Get ARC ghost hits since last eviction. */ ngrd = wmsum_value(&arc_mru_ghost->arcs_hits[ARC_BUFC_DATA]); uint64_t grd = ngrd - ogrd; ogrd = ngrd; ngrm = wmsum_value(&arc_mru_ghost->arcs_hits[ARC_BUFC_METADATA]); uint64_t grm = ngrm - ogrm; ogrm = ngrm; ngfd = wmsum_value(&arc_mfu_ghost->arcs_hits[ARC_BUFC_DATA]); uint64_t gfd = ngfd - ogfd; ogfd = ngfd; ngfm = wmsum_value(&arc_mfu_ghost->arcs_hits[ARC_BUFC_METADATA]); uint64_t gfm = ngfm - ogfm; ogfm = ngfm; /* Adjust ARC states balance based on ghost hits. */ arc_meta = arc_evict_adj(arc_meta, gsrd + gsrm + gsfd + gsfm, grm + gfm, grd + gfd, zfs_arc_meta_balance); arc_pd = arc_evict_adj(arc_pd, gsrd + gsfd, grd, gfd, 100); arc_pm = arc_evict_adj(arc_pm, gsrm + gsfm, grm, gfm, 100); asize = aggsum_value(&arc_sums.arcstat_size); int64_t wt = t - (asize - arc_c); /* * Try to reduce pinned dnodes if more than 3/4 of wanted metadata * target is not evictable or if they go over arc_dnode_limit. */ int64_t prune = 0; int64_t dn = wmsum_value(&arc_sums.arcstat_dnode_size); w = wt * (int64_t)(arc_meta >> 16) >> 16; if (zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_METADATA]) + zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_METADATA]) - zfs_refcount_count(&arc_mru->arcs_esize[ARC_BUFC_METADATA]) - zfs_refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]) > w * 3 / 4) { prune = dn / sizeof (dnode_t) * zfs_arc_dnode_reduce_percent / 100; } else if (dn > arc_dnode_limit) { prune = (dn - arc_dnode_limit) / sizeof (dnode_t) * zfs_arc_dnode_reduce_percent / 100; } if (prune > 0) arc_prune_async(prune); /* Evict MRU metadata. */ w = wt * (int64_t)(arc_meta * arc_pm >> 48) >> 16; e = MIN((int64_t)(asize - arc_c), (int64_t)(mrum - w)); bytes = arc_evict_impl(arc_mru, ARC_BUFC_METADATA, e); total_evicted += bytes; mrum -= bytes; asize -= bytes; /* Evict MFU metadata. */ w = wt * (int64_t)(arc_meta >> 16) >> 16; e = MIN((int64_t)(asize - arc_c), (int64_t)(m - w)); bytes = arc_evict_impl(arc_mfu, ARC_BUFC_METADATA, e); total_evicted += bytes; mfum -= bytes; asize -= bytes; /* Evict MRU data. */ wt -= m - total_evicted; w = wt * (int64_t)(arc_pd >> 16) >> 16; e = MIN((int64_t)(asize - arc_c), (int64_t)(mrud - w)); bytes = arc_evict_impl(arc_mru, ARC_BUFC_DATA, e); total_evicted += bytes; mrud -= bytes; asize -= bytes; /* Evict MFU data. */ e = asize - arc_c; bytes = arc_evict_impl(arc_mfu, ARC_BUFC_DATA, e); mfud -= bytes; total_evicted += bytes; /* * Evict ghost lists * * Size of each state's ghost list represents how much that state * may grow by shrinking the other states. Would it need to shrink * other states to zero (that is unlikely), its ghost size would be * equal to sum of other three state sizes. But excessive ghost * size may result in false ghost hits (too far back), that may * never result in real cache hits if several states are competing. * So choose some arbitraty point of 1/2 of other state sizes. */ gsrd = (mrum + mfud + mfum) / 2; e = zfs_refcount_count(&arc_mru_ghost->arcs_size[ARC_BUFC_DATA]) - gsrd; (void) arc_evict_impl(arc_mru_ghost, ARC_BUFC_DATA, e); gsrm = (mrud + mfud + mfum) / 2; e = zfs_refcount_count(&arc_mru_ghost->arcs_size[ARC_BUFC_METADATA]) - gsrm; (void) arc_evict_impl(arc_mru_ghost, ARC_BUFC_METADATA, e); gsfd = (mrud + mrum + mfum) / 2; e = zfs_refcount_count(&arc_mfu_ghost->arcs_size[ARC_BUFC_DATA]) - gsfd; (void) arc_evict_impl(arc_mfu_ghost, ARC_BUFC_DATA, e); gsfm = (mrud + mrum + mfud) / 2; e = zfs_refcount_count(&arc_mfu_ghost->arcs_size[ARC_BUFC_METADATA]) - gsfm; (void) arc_evict_impl(arc_mfu_ghost, ARC_BUFC_METADATA, e); return (total_evicted); } void arc_flush(spa_t *spa, boolean_t retry) { uint64_t guid = 0; /* * If retry is B_TRUE, a spa must not be specified since we have * no good way to determine if all of a spa's buffers have been * evicted from an arc state. */ ASSERT(!retry || spa == NULL); if (spa != NULL) guid = spa_load_guid(spa); (void) arc_flush_state(arc_mru, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mru, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mfu, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mfu, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_uncached, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_uncached, guid, ARC_BUFC_METADATA, retry); } void arc_reduce_target_size(int64_t to_free) { uint64_t c = arc_c; if (c <= arc_c_min) return; /* * All callers want the ARC to actually evict (at least) this much * memory. Therefore we reduce from the lower of the current size and * the target size. This way, even if arc_c is much higher than * arc_size (as can be the case after many calls to arc_freed(), we will * immediately have arc_c < arc_size and therefore the arc_evict_zthr * will evict. */ uint64_t asize = aggsum_value(&arc_sums.arcstat_size); if (asize < c) to_free += c - asize; arc_c = MAX((int64_t)c - to_free, (int64_t)arc_c_min); /* See comment in arc_evict_cb_check() on why lock+flag */ mutex_enter(&arc_evict_lock); arc_evict_needed = B_TRUE; mutex_exit(&arc_evict_lock); zthr_wakeup(arc_evict_zthr); } /* * Determine if the system is under memory pressure and is asking * to reclaim memory. A return value of B_TRUE indicates that the system * is under memory pressure and that the arc should adjust accordingly. */ boolean_t arc_reclaim_needed(void) { return (arc_available_memory() < 0); } void arc_kmem_reap_soon(void) { size_t i; kmem_cache_t *prev_cache = NULL; kmem_cache_t *prev_data_cache = NULL; #ifdef _KERNEL #if defined(_ILP32) /* * Reclaim unused memory from all kmem caches. */ kmem_reap(); #endif #endif for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) { #if defined(_ILP32) /* reach upper limit of cache size on 32-bit */ if (zio_buf_cache[i] == NULL) break; #endif if (zio_buf_cache[i] != prev_cache) { prev_cache = zio_buf_cache[i]; kmem_cache_reap_now(zio_buf_cache[i]); } if (zio_data_buf_cache[i] != prev_data_cache) { prev_data_cache = zio_data_buf_cache[i]; kmem_cache_reap_now(zio_data_buf_cache[i]); } } kmem_cache_reap_now(buf_cache); kmem_cache_reap_now(hdr_full_cache); kmem_cache_reap_now(hdr_l2only_cache); kmem_cache_reap_now(zfs_btree_leaf_cache); abd_cache_reap_now(); } static boolean_t arc_evict_cb_check(void *arg, zthr_t *zthr) { (void) arg, (void) zthr; #ifdef ZFS_DEBUG /* * This is necessary in order to keep the kstat information * up to date for tools that display kstat data such as the * mdb ::arc dcmd and the Linux crash utility. These tools * typically do not call kstat's update function, but simply * dump out stats from the most recent update. Without * this call, these commands may show stale stats for the * anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even * with this call, the data might be out of date if the * evict thread hasn't been woken recently; but that should * suffice. The arc_state_t structures can be queried * directly if more accurate information is needed. */ if (arc_ksp != NULL) arc_ksp->ks_update(arc_ksp, KSTAT_READ); #endif /* * We have to rely on arc_wait_for_eviction() to tell us when to * evict, rather than checking if we are overflowing here, so that we * are sure to not leave arc_wait_for_eviction() waiting on aew_cv. * If we have become "not overflowing" since arc_wait_for_eviction() * checked, we need to wake it up. We could broadcast the CV here, * but arc_wait_for_eviction() may have not yet gone to sleep. We * would need to use a mutex to ensure that this function doesn't * broadcast until arc_wait_for_eviction() has gone to sleep (e.g. * the arc_evict_lock). However, the lock ordering of such a lock * would necessarily be incorrect with respect to the zthr_lock, * which is held before this function is called, and is held by * arc_wait_for_eviction() when it calls zthr_wakeup(). */ if (arc_evict_needed) return (B_TRUE); /* * If we have buffers in uncached state, evict them periodically. */ return ((zfs_refcount_count(&arc_uncached->arcs_esize[ARC_BUFC_DATA]) + zfs_refcount_count(&arc_uncached->arcs_esize[ARC_BUFC_METADATA]) && ddi_get_lbolt() - arc_last_uncached_flush > MSEC_TO_TICK(arc_min_prefetch_ms / 2))); } /* * Keep arc_size under arc_c by running arc_evict which evicts data * from the ARC. */ static void arc_evict_cb(void *arg, zthr_t *zthr) { (void) arg; uint64_t evicted = 0; fstrans_cookie_t cookie = spl_fstrans_mark(); /* Always try to evict from uncached state. */ arc_last_uncached_flush = ddi_get_lbolt(); evicted += arc_flush_state(arc_uncached, 0, ARC_BUFC_DATA, B_FALSE); evicted += arc_flush_state(arc_uncached, 0, ARC_BUFC_METADATA, B_FALSE); /* Evict from other states only if told to. */ if (arc_evict_needed) evicted += arc_evict(); /* * If evicted is zero, we couldn't evict anything * via arc_evict(). This could be due to hash lock * collisions, but more likely due to the majority of * arc buffers being unevictable. Therefore, even if * arc_size is above arc_c, another pass is unlikely to * be helpful and could potentially cause us to enter an * infinite loop. Additionally, zthr_iscancelled() is * checked here so that if the arc is shutting down, the * broadcast will wake any remaining arc evict waiters. * * Note we cancel using zthr instead of arc_evict_zthr * because the latter may not yet be initializd when the * callback is first invoked. */ mutex_enter(&arc_evict_lock); arc_evict_needed = !zthr_iscancelled(zthr) && evicted > 0 && aggsum_compare(&arc_sums.arcstat_size, arc_c) > 0; if (!arc_evict_needed) { /* * We're either no longer overflowing, or we * can't evict anything more, so we should wake * arc_get_data_impl() sooner. */ arc_evict_waiter_t *aw; while ((aw = list_remove_head(&arc_evict_waiters)) != NULL) { cv_broadcast(&aw->aew_cv); } arc_set_need_free(); } mutex_exit(&arc_evict_lock); spl_fstrans_unmark(cookie); } static boolean_t arc_reap_cb_check(void *arg, zthr_t *zthr) { (void) arg, (void) zthr; int64_t free_memory = arc_available_memory(); static int reap_cb_check_counter = 0; /* * If a kmem reap is already active, don't schedule more. We must * check for this because kmem_cache_reap_soon() won't actually * block on the cache being reaped (this is to prevent callers from * becoming implicitly blocked by a system-wide kmem reap -- which, * on a system with many, many full magazines, can take minutes). */ if (!kmem_cache_reap_active() && free_memory < 0) { arc_no_grow = B_TRUE; arc_warm = B_TRUE; /* * Wait at least zfs_grow_retry (default 5) seconds * before considering growing. */ arc_growtime = gethrtime() + SEC2NSEC(arc_grow_retry); return (B_TRUE); } else if (free_memory < arc_c >> arc_no_grow_shift) { arc_no_grow = B_TRUE; } else if (gethrtime() >= arc_growtime) { arc_no_grow = B_FALSE; } /* * Called unconditionally every 60 seconds to reclaim unused * zstd compression and decompression context. This is done * here to avoid the need for an independent thread. */ if (!((reap_cb_check_counter++) % 60)) zfs_zstd_cache_reap_now(); return (B_FALSE); } /* * Keep enough free memory in the system by reaping the ARC's kmem * caches. To cause more slabs to be reapable, we may reduce the * target size of the cache (arc_c), causing the arc_evict_cb() * to free more buffers. */ static void arc_reap_cb(void *arg, zthr_t *zthr) { (void) arg, (void) zthr; int64_t free_memory; fstrans_cookie_t cookie = spl_fstrans_mark(); /* * Kick off asynchronous kmem_reap()'s of all our caches. */ arc_kmem_reap_soon(); /* * Wait at least arc_kmem_cache_reap_retry_ms between * arc_kmem_reap_soon() calls. Without this check it is possible to * end up in a situation where we spend lots of time reaping * caches, while we're near arc_c_min. Waiting here also gives the * subsequent free memory check a chance of finding that the * asynchronous reap has already freed enough memory, and we don't * need to call arc_reduce_target_size(). */ delay((hz * arc_kmem_cache_reap_retry_ms + 999) / 1000); /* * Reduce the target size as needed to maintain the amount of free * memory in the system at a fraction of the arc_size (1/128th by * default). If oversubscribed (free_memory < 0) then reduce the * target arc_size by the deficit amount plus the fractional * amount. If free memory is positive but less than the fractional * amount, reduce by what is needed to hit the fractional amount. */ free_memory = arc_available_memory(); int64_t can_free = arc_c - arc_c_min; if (can_free > 0) { int64_t to_free = (can_free >> arc_shrink_shift) - free_memory; if (to_free > 0) arc_reduce_target_size(to_free); } spl_fstrans_unmark(cookie); } #ifdef _KERNEL /* * Determine the amount of memory eligible for eviction contained in the * ARC. All clean data reported by the ghost lists can always be safely * evicted. Due to arc_c_min, the same does not hold for all clean data * contained by the regular mru and mfu lists. * * In the case of the regular mru and mfu lists, we need to report as * much clean data as possible, such that evicting that same reported * data will not bring arc_size below arc_c_min. Thus, in certain * circumstances, the total amount of clean data in the mru and mfu * lists might not actually be evictable. * * The following two distinct cases are accounted for: * * 1. The sum of the amount of dirty data contained by both the mru and * mfu lists, plus the ARC's other accounting (e.g. the anon list), * is greater than or equal to arc_c_min. * (i.e. amount of dirty data >= arc_c_min) * * This is the easy case; all clean data contained by the mru and mfu * lists is evictable. Evicting all clean data can only drop arc_size * to the amount of dirty data, which is greater than arc_c_min. * * 2. The sum of the amount of dirty data contained by both the mru and * mfu lists, plus the ARC's other accounting (e.g. the anon list), * is less than arc_c_min. * (i.e. arc_c_min > amount of dirty data) * * 2.1. arc_size is greater than or equal arc_c_min. * (i.e. arc_size >= arc_c_min > amount of dirty data) * * In this case, not all clean data from the regular mru and mfu * lists is actually evictable; we must leave enough clean data * to keep arc_size above arc_c_min. Thus, the maximum amount of * evictable data from the two lists combined, is exactly the * difference between arc_size and arc_c_min. * * 2.2. arc_size is less than arc_c_min * (i.e. arc_c_min > arc_size > amount of dirty data) * * In this case, none of the data contained in the mru and mfu * lists is evictable, even if it's clean. Since arc_size is * already below arc_c_min, evicting any more would only * increase this negative difference. */ #endif /* _KERNEL */ /* * Adapt arc info given the number of bytes we are trying to add and * the state that we are coming from. This function is only called * when we are adding new content to the cache. */ static void arc_adapt(uint64_t bytes) { /* * Wake reap thread if we do not have any available memory */ if (arc_reclaim_needed()) { zthr_wakeup(arc_reap_zthr); return; } if (arc_no_grow) return; if (arc_c >= arc_c_max) return; /* * If we're within (2 * maxblocksize) bytes of the target * cache size, increment the target cache size */ if (aggsum_upper_bound(&arc_sums.arcstat_size) + 2 * SPA_MAXBLOCKSIZE >= arc_c) { uint64_t dc = MAX(bytes, SPA_OLD_MAXBLOCKSIZE); if (atomic_add_64_nv(&arc_c, dc) > arc_c_max) arc_c = arc_c_max; } } /* * Check if arc_size has grown past our upper threshold, determined by * zfs_arc_overflow_shift. */ static arc_ovf_level_t arc_is_overflowing(boolean_t use_reserve) { /* Always allow at least one block of overflow */ int64_t overflow = MAX(SPA_MAXBLOCKSIZE, arc_c >> zfs_arc_overflow_shift); /* * We just compare the lower bound here for performance reasons. Our * primary goals are to make sure that the arc never grows without * bound, and that it can reach its maximum size. This check * accomplishes both goals. The maximum amount we could run over by is * 2 * aggsum_borrow_multiplier * NUM_CPUS * the average size of a block * in the ARC. In practice, that's in the tens of MB, which is low * enough to be safe. */ int64_t over = aggsum_lower_bound(&arc_sums.arcstat_size) - arc_c - overflow / 2; if (!use_reserve) overflow /= 2; return (over < 0 ? ARC_OVF_NONE : over < overflow ? ARC_OVF_SOME : ARC_OVF_SEVERE); } static abd_t * arc_get_data_abd(arc_buf_hdr_t *hdr, uint64_t size, const void *tag, int alloc_flags) { arc_buf_contents_t type = arc_buf_type(hdr); arc_get_data_impl(hdr, size, tag, alloc_flags); if (alloc_flags & ARC_HDR_ALLOC_LINEAR) return (abd_alloc_linear(size, type == ARC_BUFC_METADATA)); else return (abd_alloc(size, type == ARC_BUFC_METADATA)); } static void * arc_get_data_buf(arc_buf_hdr_t *hdr, uint64_t size, const void *tag) { arc_buf_contents_t type = arc_buf_type(hdr); arc_get_data_impl(hdr, size, tag, 0); if (type == ARC_BUFC_METADATA) { return (zio_buf_alloc(size)); } else { ASSERT(type == ARC_BUFC_DATA); return (zio_data_buf_alloc(size)); } } /* * Wait for the specified amount of data (in bytes) to be evicted from the * ARC, and for there to be sufficient free memory in the system. Waiting for * eviction ensures that the memory used by the ARC decreases. Waiting for * free memory ensures that the system won't run out of free pages, regardless * of ARC behavior and settings. See arc_lowmem_init(). */ void arc_wait_for_eviction(uint64_t amount, boolean_t use_reserve) { switch (arc_is_overflowing(use_reserve)) { case ARC_OVF_NONE: return; case ARC_OVF_SOME: /* * This is a bit racy without taking arc_evict_lock, but the * worst that can happen is we either call zthr_wakeup() extra * time due to race with other thread here, or the set flag * get cleared by arc_evict_cb(), which is unlikely due to * big hysteresis, but also not important since at this level * of overflow the eviction is purely advisory. Same time * taking the global lock here every time without waiting for * the actual eviction creates a significant lock contention. */ if (!arc_evict_needed) { arc_evict_needed = B_TRUE; zthr_wakeup(arc_evict_zthr); } return; case ARC_OVF_SEVERE: default: { arc_evict_waiter_t aw; list_link_init(&aw.aew_node); cv_init(&aw.aew_cv, NULL, CV_DEFAULT, NULL); uint64_t last_count = 0; mutex_enter(&arc_evict_lock); if (!list_is_empty(&arc_evict_waiters)) { arc_evict_waiter_t *last = list_tail(&arc_evict_waiters); last_count = last->aew_count; } else if (!arc_evict_needed) { arc_evict_needed = B_TRUE; zthr_wakeup(arc_evict_zthr); } /* * Note, the last waiter's count may be less than * arc_evict_count if we are low on memory in which * case arc_evict_state_impl() may have deferred * wakeups (but still incremented arc_evict_count). */ aw.aew_count = MAX(last_count, arc_evict_count) + amount; list_insert_tail(&arc_evict_waiters, &aw); arc_set_need_free(); DTRACE_PROBE3(arc__wait__for__eviction, uint64_t, amount, uint64_t, arc_evict_count, uint64_t, aw.aew_count); /* * We will be woken up either when arc_evict_count reaches * aew_count, or when the ARC is no longer overflowing and * eviction completes. * In case of "false" wakeup, we will still be on the list. */ do { cv_wait(&aw.aew_cv, &arc_evict_lock); } while (list_link_active(&aw.aew_node)); mutex_exit(&arc_evict_lock); cv_destroy(&aw.aew_cv); } } } /* * Allocate a block and return it to the caller. If we are hitting the * hard limit for the cache size, we must sleep, waiting for the eviction * thread to catch up. If we're past the target size but below the hard * limit, we'll only signal the reclaim thread and continue on. */ static void arc_get_data_impl(arc_buf_hdr_t *hdr, uint64_t size, const void *tag, int alloc_flags) { arc_adapt(size); /* * If arc_size is currently overflowing, we must be adding data * faster than we are evicting. To ensure we don't compound the * problem by adding more data and forcing arc_size to grow even * further past it's target size, we wait for the eviction thread to * make some progress. We also wait for there to be sufficient free * memory in the system, as measured by arc_free_memory(). * * Specifically, we wait for zfs_arc_eviction_pct percent of the * requested size to be evicted. This should be more than 100%, to * ensure that that progress is also made towards getting arc_size * under arc_c. See the comment above zfs_arc_eviction_pct. */ arc_wait_for_eviction(size * zfs_arc_eviction_pct / 100, alloc_flags & ARC_HDR_USE_RESERVE); arc_buf_contents_t type = arc_buf_type(hdr); if (type == ARC_BUFC_METADATA) { arc_space_consume(size, ARC_SPACE_META); } else { arc_space_consume(size, ARC_SPACE_DATA); } /* * Update the state size. Note that ghost states have a * "ghost size" and so don't need to be updated. */ arc_state_t *state = hdr->b_l1hdr.b_state; if (!GHOST_STATE(state)) { (void) zfs_refcount_add_many(&state->arcs_size[type], size, tag); /* * If this is reached via arc_read, the link is * protected by the hash lock. If reached via * arc_buf_alloc, the header should not be accessed by * any other thread. And, if reached via arc_read_done, * the hash lock will protect it if it's found in the * hash table; otherwise no other thread should be * trying to [add|remove]_reference it. */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); (void) zfs_refcount_add_many(&state->arcs_esize[type], size, tag); } } } static void arc_free_data_abd(arc_buf_hdr_t *hdr, abd_t *abd, uint64_t size, const void *tag) { arc_free_data_impl(hdr, size, tag); abd_free(abd); } static void arc_free_data_buf(arc_buf_hdr_t *hdr, void *buf, uint64_t size, const void *tag) { arc_buf_contents_t type = arc_buf_type(hdr); arc_free_data_impl(hdr, size, tag); if (type == ARC_BUFC_METADATA) { zio_buf_free(buf, size); } else { ASSERT(type == ARC_BUFC_DATA); zio_data_buf_free(buf, size); } } /* * Free the arc data buffer. */ static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size, const void *tag) { arc_state_t *state = hdr->b_l1hdr.b_state; arc_buf_contents_t type = arc_buf_type(hdr); /* protected by hash lock, if in the hash table */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT(state != arc_anon && state != arc_l2c_only); (void) zfs_refcount_remove_many(&state->arcs_esize[type], size, tag); } (void) zfs_refcount_remove_many(&state->arcs_size[type], size, tag); VERIFY3U(hdr->b_type, ==, type); if (type == ARC_BUFC_METADATA) { arc_space_return(size, ARC_SPACE_META); } else { ASSERT(type == ARC_BUFC_DATA); arc_space_return(size, ARC_SPACE_DATA); } } /* * This routine is called whenever a buffer is accessed. */ static void arc_access(arc_buf_hdr_t *hdr, arc_flags_t arc_flags, boolean_t hit) { ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); ASSERT(HDR_HAS_L1HDR(hdr)); /* * Update buffer prefetch status. */ boolean_t was_prefetch = HDR_PREFETCH(hdr); boolean_t now_prefetch = arc_flags & ARC_FLAG_PREFETCH; if (was_prefetch != now_prefetch) { if (was_prefetch) { ARCSTAT_CONDSTAT(hit, demand_hit, demand_iohit, HDR_PRESCIENT_PREFETCH(hdr), prescient, predictive, prefetch); } if (HDR_HAS_L2HDR(hdr)) l2arc_hdr_arcstats_decrement_state(hdr); if (was_prefetch) { arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH | ARC_FLAG_PRESCIENT_PREFETCH); } else { arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH); } if (HDR_HAS_L2HDR(hdr)) l2arc_hdr_arcstats_increment_state(hdr); } if (now_prefetch) { if (arc_flags & ARC_FLAG_PRESCIENT_PREFETCH) { arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH); ARCSTAT_BUMP(arcstat_prescient_prefetch); } else { ARCSTAT_BUMP(arcstat_predictive_prefetch); } } if (arc_flags & ARC_FLAG_L2CACHE) arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE); clock_t now = ddi_get_lbolt(); if (hdr->b_l1hdr.b_state == arc_anon) { arc_state_t *new_state; /* * This buffer is not in the cache, and does not appear in * our "ghost" lists. Add it to the MRU or uncached state. */ ASSERT0(hdr->b_l1hdr.b_arc_access); hdr->b_l1hdr.b_arc_access = now; if (HDR_UNCACHED(hdr)) { new_state = arc_uncached; DTRACE_PROBE1(new_state__uncached, arc_buf_hdr_t *, hdr); } else { new_state = arc_mru; DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); } arc_change_state(new_state, hdr); } else if (hdr->b_l1hdr.b_state == arc_mru) { /* * This buffer has been accessed once recently and either * its read is still in progress or it is in the cache. */ if (HDR_IO_IN_PROGRESS(hdr)) { hdr->b_l1hdr.b_arc_access = now; return; } hdr->b_l1hdr.b_mru_hits++; ARCSTAT_BUMP(arcstat_mru_hits); /* * If the previous access was a prefetch, then it already * handled possible promotion, so nothing more to do for now. */ if (was_prefetch) { hdr->b_l1hdr.b_arc_access = now; return; } /* * If more than ARC_MINTIME have passed from the previous * hit, promote the buffer to the MFU state. */ if (ddi_time_after(now, hdr->b_l1hdr.b_arc_access + ARC_MINTIME)) { hdr->b_l1hdr.b_arc_access = now; DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(arc_mfu, hdr); } } else if (hdr->b_l1hdr.b_state == arc_mru_ghost) { arc_state_t *new_state; /* * This buffer has been accessed once recently, but was * evicted from the cache. Would we have bigger MRU, it * would be an MRU hit, so handle it the same way, except * we don't need to check the previous access time. */ hdr->b_l1hdr.b_mru_ghost_hits++; ARCSTAT_BUMP(arcstat_mru_ghost_hits); hdr->b_l1hdr.b_arc_access = now; wmsum_add(&arc_mru_ghost->arcs_hits[arc_buf_type(hdr)], arc_hdr_size(hdr)); if (was_prefetch) { new_state = arc_mru; DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); } else { new_state = arc_mfu; DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); } arc_change_state(new_state, hdr); } else if (hdr->b_l1hdr.b_state == arc_mfu) { /* * This buffer has been accessed more than once and either * still in the cache or being restored from one of ghosts. */ if (!HDR_IO_IN_PROGRESS(hdr)) { hdr->b_l1hdr.b_mfu_hits++; ARCSTAT_BUMP(arcstat_mfu_hits); } hdr->b_l1hdr.b_arc_access = now; } else if (hdr->b_l1hdr.b_state == arc_mfu_ghost) { /* * This buffer has been accessed more than once recently, but * has been evicted from the cache. Would we have bigger MFU * it would stay in cache, so move it back to MFU state. */ hdr->b_l1hdr.b_mfu_ghost_hits++; ARCSTAT_BUMP(arcstat_mfu_ghost_hits); hdr->b_l1hdr.b_arc_access = now; wmsum_add(&arc_mfu_ghost->arcs_hits[arc_buf_type(hdr)], arc_hdr_size(hdr)); DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(arc_mfu, hdr); } else if (hdr->b_l1hdr.b_state == arc_uncached) { /* * This buffer is uncacheable, but we got a hit. Probably * a demand read after prefetch. Nothing more to do here. */ if (!HDR_IO_IN_PROGRESS(hdr)) ARCSTAT_BUMP(arcstat_uncached_hits); hdr->b_l1hdr.b_arc_access = now; } else if (hdr->b_l1hdr.b_state == arc_l2c_only) { /* * This buffer is on the 2nd Level ARC and was not accessed * for a long time, so treat it as new and put into MRU. */ hdr->b_l1hdr.b_arc_access = now; DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); arc_change_state(arc_mru, hdr); } else { cmn_err(CE_PANIC, "invalid arc state 0x%p", hdr->b_l1hdr.b_state); } } /* * This routine is called by dbuf_hold() to update the arc_access() state * which otherwise would be skipped for entries in the dbuf cache. */ void arc_buf_access(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; /* * Avoid taking the hash_lock when possible as an optimization. * The header must be checked again under the hash_lock in order * to handle the case where it is concurrently being released. */ if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) return; kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) { mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_access_skip); return; } ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu || hdr->b_l1hdr.b_state == arc_uncached); DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr); arc_access(hdr, 0, B_TRUE); mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_hits); ARCSTAT_CONDSTAT(B_TRUE /* demand */, demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits); } /* a generic arc_read_done_func_t which you can use */ void arc_bcopy_func(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *arg) { (void) zio, (void) zb, (void) bp; if (buf == NULL) return; memcpy(arg, buf->b_data, arc_buf_size(buf)); arc_buf_destroy(buf, arg); } /* a generic arc_read_done_func_t */ void arc_getbuf_func(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *arg) { (void) zb, (void) bp; arc_buf_t **bufp = arg; if (buf == NULL) { ASSERT(zio == NULL || zio->io_error != 0); *bufp = NULL; } else { ASSERT(zio == NULL || zio->io_error == 0); *bufp = buf; ASSERT(buf->b_data != NULL); } } static void arc_hdr_verify(arc_buf_hdr_t *hdr, blkptr_t *bp) { if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) { ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0); ASSERT3U(arc_hdr_get_compress(hdr), ==, ZIO_COMPRESS_OFF); } else { if (HDR_COMPRESSION_ENABLED(hdr)) { ASSERT3U(arc_hdr_get_compress(hdr), ==, BP_GET_COMPRESS(bp)); } ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp)); ASSERT3U(HDR_GET_PSIZE(hdr), ==, BP_GET_PSIZE(bp)); ASSERT3U(!!HDR_PROTECTED(hdr), ==, BP_IS_PROTECTED(bp)); } } static void arc_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; arc_buf_hdr_t *hdr = zio->io_private; kmutex_t *hash_lock = NULL; arc_callback_t *callback_list; arc_callback_t *acb; /* * The hdr was inserted into hash-table and removed from lists * prior to starting I/O. We should find this header, since * it's in the hash table, and it should be legit since it's * not possible to evict it during the I/O. The only possible * reason for it not to be found is if we were freed during the * read. */ if (HDR_IN_HASH_TABLE(hdr)) { arc_buf_hdr_t *found; ASSERT3U(hdr->b_birth, ==, BP_GET_BIRTH(zio->io_bp)); ASSERT3U(hdr->b_dva.dva_word[0], ==, BP_IDENTITY(zio->io_bp)->dva_word[0]); ASSERT3U(hdr->b_dva.dva_word[1], ==, BP_IDENTITY(zio->io_bp)->dva_word[1]); found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock); ASSERT((found == hdr && DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) || (found == hdr && HDR_L2_READING(hdr))); ASSERT3P(hash_lock, !=, NULL); } if (BP_IS_PROTECTED(bp)) { hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp); hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset; zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv); if (zio->io_error == 0) { if (BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG) { void *tmpbuf; tmpbuf = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t)); zio_crypt_decode_mac_zil(tmpbuf, hdr->b_crypt_hdr.b_mac); abd_return_buf(zio->io_abd, tmpbuf, sizeof (zil_chain_t)); } else { zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac); } } } if (zio->io_error == 0) { /* byteswap if necessary */ if (BP_SHOULD_BYTESWAP(zio->io_bp)) { if (BP_GET_LEVEL(zio->io_bp) > 0) { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64; } else { hdr->b_l1hdr.b_byteswap = DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp)); } } else { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; } if (!HDR_L2_READING(hdr)) { hdr->b_complevel = zio->io_prop.zp_complevel; } } arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED); if (l2arc_noprefetch && HDR_PREFETCH(hdr)) arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE); callback_list = hdr->b_l1hdr.b_acb; ASSERT3P(callback_list, !=, NULL); hdr->b_l1hdr.b_acb = NULL; /* * If a read request has a callback (i.e. acb_done is not NULL), then we * make a buf containing the data according to the parameters which were * passed in. The implementation of arc_buf_alloc_impl() ensures that we * aren't needlessly decompressing the data multiple times. */ int callback_cnt = 0; for (acb = callback_list; acb != NULL; acb = acb->acb_next) { /* We need the last one to call below in original order. */ callback_list = acb; if (!acb->acb_done || acb->acb_nobuf) continue; callback_cnt++; if (zio->io_error != 0) continue; int error = arc_buf_alloc_impl(hdr, zio->io_spa, &acb->acb_zb, acb->acb_private, acb->acb_encrypted, acb->acb_compressed, acb->acb_noauth, B_TRUE, &acb->acb_buf); /* * Assert non-speculative zios didn't fail because an * encryption key wasn't loaded */ ASSERT((zio->io_flags & ZIO_FLAG_SPECULATIVE) || error != EACCES); /* * If we failed to decrypt, report an error now (as the zio * layer would have done if it had done the transforms). */ if (error == ECKSUM) { ASSERT(BP_IS_PROTECTED(bp)); error = SET_ERROR(EIO); if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(zio->io_spa, &acb->acb_zb, BP_GET_LOGICAL_BIRTH(zio->io_bp)); (void) zfs_ereport_post( FM_EREPORT_ZFS_AUTHENTICATION, zio->io_spa, NULL, &acb->acb_zb, zio, 0); } } if (error != 0) { /* * Decompression or decryption failed. Set * io_error so that when we call acb_done * (below), we will indicate that the read * failed. Note that in the unusual case * where one callback is compressed and another * uncompressed, we will mark all of them * as failed, even though the uncompressed * one can't actually fail. In this case, * the hdr will not be anonymous, because * if there are multiple callbacks, it's * because multiple threads found the same * arc buf in the hash table. */ zio->io_error = error; } } /* * If there are multiple callbacks, we must have the hash lock, * because the only way for multiple threads to find this hdr is * in the hash table. This ensures that if there are multiple * callbacks, the hdr is not anonymous. If it were anonymous, * we couldn't use arc_buf_destroy() in the error case below. */ ASSERT(callback_cnt < 2 || hash_lock != NULL); if (zio->io_error == 0) { arc_hdr_verify(hdr, zio->io_bp); } else { arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR); if (hdr->b_l1hdr.b_state != arc_anon) arc_change_state(arc_anon, hdr); if (HDR_IN_HASH_TABLE(hdr)) buf_hash_remove(hdr); } arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); (void) remove_reference(hdr, hdr); if (hash_lock != NULL) mutex_exit(hash_lock); /* execute each callback and free its structure */ while ((acb = callback_list) != NULL) { if (acb->acb_done != NULL) { if (zio->io_error != 0 && acb->acb_buf != NULL) { /* * If arc_buf_alloc_impl() fails during * decompression, the buf will still be * allocated, and needs to be freed here. */ arc_buf_destroy(acb->acb_buf, acb->acb_private); acb->acb_buf = NULL; } acb->acb_done(zio, &zio->io_bookmark, zio->io_bp, acb->acb_buf, acb->acb_private); } if (acb->acb_zio_dummy != NULL) { acb->acb_zio_dummy->io_error = zio->io_error; zio_nowait(acb->acb_zio_dummy); } callback_list = acb->acb_prev; if (acb->acb_wait) { mutex_enter(&acb->acb_wait_lock); acb->acb_wait_error = zio->io_error; acb->acb_wait = B_FALSE; cv_signal(&acb->acb_wait_cv); mutex_exit(&acb->acb_wait_lock); /* acb will be freed by the waiting thread. */ } else { kmem_free(acb, sizeof (arc_callback_t)); } } } /* * "Read" the block at the specified DVA (in bp) via the * cache. If the block is found in the cache, invoke the provided * callback immediately and return. Note that the `zio' parameter * in the callback will be NULL in this case, since no IO was * required. If the block is not in the cache pass the read request * on to the spa with a substitute callback function, so that the * requested block will be added to the cache. * * If a read request arrives for a block that has a read in-progress, * either wait for the in-progress read to complete (and return the * results); or, if this is a read with a "done" func, add a record * to the read to invoke the "done" func when the read completes, * and return; or just return. * * arc_read_done() will invoke all the requested "done" functions * for readers of this block. */ int arc_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, arc_read_done_func_t *done, void *private, zio_priority_t priority, int zio_flags, arc_flags_t *arc_flags, const zbookmark_phys_t *zb) { arc_buf_hdr_t *hdr = NULL; kmutex_t *hash_lock = NULL; zio_t *rzio; uint64_t guid = spa_load_guid(spa); boolean_t compressed_read = (zio_flags & ZIO_FLAG_RAW_COMPRESS) != 0; boolean_t encrypted_read = BP_IS_ENCRYPTED(bp) && (zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0; boolean_t noauth_read = BP_IS_AUTHENTICATED(bp) && (zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0; boolean_t embedded_bp = !!BP_IS_EMBEDDED(bp); boolean_t no_buf = *arc_flags & ARC_FLAG_NO_BUF; arc_buf_t *buf = NULL; int rc = 0; ASSERT(!embedded_bp || BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA); ASSERT(!BP_IS_HOLE(bp)); ASSERT(!BP_IS_REDACTED(bp)); /* * Normally SPL_FSTRANS will already be set since kernel threads which * expect to call the DMU interfaces will set it when created. System * calls are similarly handled by setting/cleaning the bit in the * registered callback (module/os/.../zfs/zpl_*). * * External consumers such as Lustre which call the exported DMU * interfaces may not have set SPL_FSTRANS. To avoid a deadlock * on the hash_lock always set and clear the bit. */ fstrans_cookie_t cookie = spl_fstrans_mark(); top: /* * Verify the block pointer contents are reasonable. This should * always be the case since the blkptr is protected by a checksum. * However, if there is damage it's desirable to detect this early * and treat it as a checksum error. This allows an alternate blkptr * to be tried when one is available (e.g. ditto blocks). */ if (!zfs_blkptr_verify(spa, bp, (zio_flags & ZIO_FLAG_CONFIG_WRITER) ? BLK_CONFIG_HELD : BLK_CONFIG_NEEDED, BLK_VERIFY_LOG)) { rc = SET_ERROR(ECKSUM); goto done; } if (!embedded_bp) { /* * Embedded BP's have no DVA and require no I/O to "read". * Create an anonymous arc buf to back it. */ hdr = buf_hash_find(guid, bp, &hash_lock); } /* * Determine if we have an L1 cache hit or a cache miss. For simplicity * we maintain encrypted data separately from compressed / uncompressed * data. If the user is requesting raw encrypted data and we don't have * that in the header we will read from disk to guarantee that we can * get it even if the encryption keys aren't loaded. */ if (hdr != NULL && HDR_HAS_L1HDR(hdr) && (HDR_HAS_RABD(hdr) || (hdr->b_l1hdr.b_pabd != NULL && !encrypted_read))) { boolean_t is_data = !HDR_ISTYPE_METADATA(hdr); if (HDR_IO_IN_PROGRESS(hdr)) { if (*arc_flags & ARC_FLAG_CACHED_ONLY) { mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_cached_only_in_progress); rc = SET_ERROR(ENOENT); goto done; } zio_t *head_zio = hdr->b_l1hdr.b_acb->acb_zio_head; ASSERT3P(head_zio, !=, NULL); if ((hdr->b_flags & ARC_FLAG_PRIO_ASYNC_READ) && priority == ZIO_PRIORITY_SYNC_READ) { /* * This is a sync read that needs to wait for * an in-flight async read. Request that the * zio have its priority upgraded. */ zio_change_priority(head_zio, priority); DTRACE_PROBE1(arc__async__upgrade__sync, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_async_upgrade_sync); } DTRACE_PROBE1(arc__iohit, arc_buf_hdr_t *, hdr); arc_access(hdr, *arc_flags, B_FALSE); /* * If there are multiple threads reading the same block * and that block is not yet in the ARC, then only one * thread will do the physical I/O and all other * threads will wait until that I/O completes. * Synchronous reads use the acb_wait_cv whereas nowait * reads register a callback. Both are signalled/called * in arc_read_done. * * Errors of the physical I/O may need to be propagated. * Synchronous read errors are returned here from * arc_read_done via acb_wait_error. Nowait reads * attach the acb_zio_dummy zio to pio and * arc_read_done propagates the physical I/O's io_error * to acb_zio_dummy, and thereby to pio. */ arc_callback_t *acb = NULL; if (done || pio || *arc_flags & ARC_FLAG_WAIT) { acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP); acb->acb_done = done; acb->acb_private = private; acb->acb_compressed = compressed_read; acb->acb_encrypted = encrypted_read; acb->acb_noauth = noauth_read; acb->acb_nobuf = no_buf; if (*arc_flags & ARC_FLAG_WAIT) { acb->acb_wait = B_TRUE; mutex_init(&acb->acb_wait_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&acb->acb_wait_cv, NULL, CV_DEFAULT, NULL); } acb->acb_zb = *zb; if (pio != NULL) { acb->acb_zio_dummy = zio_null(pio, spa, NULL, NULL, NULL, zio_flags); } acb->acb_zio_head = head_zio; acb->acb_next = hdr->b_l1hdr.b_acb; hdr->b_l1hdr.b_acb->acb_prev = acb; hdr->b_l1hdr.b_acb = acb; } mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_iohits); ARCSTAT_CONDSTAT(!(*arc_flags & ARC_FLAG_PREFETCH), demand, prefetch, is_data, data, metadata, iohits); if (*arc_flags & ARC_FLAG_WAIT) { mutex_enter(&acb->acb_wait_lock); while (acb->acb_wait) { cv_wait(&acb->acb_wait_cv, &acb->acb_wait_lock); } rc = acb->acb_wait_error; mutex_exit(&acb->acb_wait_lock); mutex_destroy(&acb->acb_wait_lock); cv_destroy(&acb->acb_wait_cv); kmem_free(acb, sizeof (arc_callback_t)); } goto out; } ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu || hdr->b_l1hdr.b_state == arc_uncached); DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr); arc_access(hdr, *arc_flags, B_TRUE); if (done && !no_buf) { ASSERT(!embedded_bp || !BP_IS_HOLE(bp)); /* Get a buf with the desired data in it. */ rc = arc_buf_alloc_impl(hdr, spa, zb, private, encrypted_read, compressed_read, noauth_read, B_TRUE, &buf); if (rc == ECKSUM) { /* * Convert authentication and decryption errors * to EIO (and generate an ereport if needed) * before leaving the ARC. */ rc = SET_ERROR(EIO); if ((zio_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(spa, zb, hdr->b_birth); (void) zfs_ereport_post( FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, zb, NULL, 0); } } if (rc != 0) { arc_buf_destroy_impl(buf); buf = NULL; (void) remove_reference(hdr, private); } /* assert any errors weren't due to unloaded keys */ ASSERT((zio_flags & ZIO_FLAG_SPECULATIVE) || rc != EACCES); } mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_hits); ARCSTAT_CONDSTAT(!(*arc_flags & ARC_FLAG_PREFETCH), demand, prefetch, is_data, data, metadata, hits); *arc_flags |= ARC_FLAG_CACHED; goto done; } else { uint64_t lsize = BP_GET_LSIZE(bp); uint64_t psize = BP_GET_PSIZE(bp); arc_callback_t *acb; vdev_t *vd = NULL; uint64_t addr = 0; boolean_t devw = B_FALSE; uint64_t size; abd_t *hdr_abd; int alloc_flags = encrypted_read ? ARC_HDR_ALLOC_RDATA : 0; arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp); if (*arc_flags & ARC_FLAG_CACHED_ONLY) { if (hash_lock != NULL) mutex_exit(hash_lock); rc = SET_ERROR(ENOENT); goto done; } if (hdr == NULL) { /* * This block is not in the cache or it has * embedded data. */ arc_buf_hdr_t *exists = NULL; hdr = arc_hdr_alloc(guid, psize, lsize, BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), 0, type); if (!embedded_bp) { hdr->b_dva = *BP_IDENTITY(bp); hdr->b_birth = BP_GET_BIRTH(bp); exists = buf_hash_insert(hdr, &hash_lock); } if (exists != NULL) { /* somebody beat us to the hash insert */ mutex_exit(hash_lock); buf_discard_identity(hdr); arc_hdr_destroy(hdr); goto top; /* restart the IO request */ } } else { /* * This block is in the ghost cache or encrypted data * was requested and we didn't have it. If it was * L2-only (and thus didn't have an L1 hdr), * we realloc the header to add an L1 hdr. */ if (!HDR_HAS_L1HDR(hdr)) { hdr = arc_hdr_realloc(hdr, hdr_l2only_cache, hdr_full_cache); } if (GHOST_STATE(hdr->b_l1hdr.b_state)) { ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT0(zfs_refcount_count( &hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); #ifdef ZFS_DEBUG ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); #endif } else if (HDR_IO_IN_PROGRESS(hdr)) { /* * If this header already had an IO in progress * and we are performing another IO to fetch * encrypted data we must wait until the first * IO completes so as not to confuse * arc_read_done(). This should be very rare * and so the performance impact shouldn't * matter. */ arc_callback_t *acb = kmem_zalloc( sizeof (arc_callback_t), KM_SLEEP); acb->acb_wait = B_TRUE; mutex_init(&acb->acb_wait_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&acb->acb_wait_cv, NULL, CV_DEFAULT, NULL); acb->acb_zio_head = hdr->b_l1hdr.b_acb->acb_zio_head; acb->acb_next = hdr->b_l1hdr.b_acb; hdr->b_l1hdr.b_acb->acb_prev = acb; hdr->b_l1hdr.b_acb = acb; mutex_exit(hash_lock); mutex_enter(&acb->acb_wait_lock); while (acb->acb_wait) { cv_wait(&acb->acb_wait_cv, &acb->acb_wait_lock); } mutex_exit(&acb->acb_wait_lock); mutex_destroy(&acb->acb_wait_lock); cv_destroy(&acb->acb_wait_cv); kmem_free(acb, sizeof (arc_callback_t)); goto top; } } if (*arc_flags & ARC_FLAG_UNCACHED) { arc_hdr_set_flags(hdr, ARC_FLAG_UNCACHED); if (!encrypted_read) alloc_flags |= ARC_HDR_ALLOC_LINEAR; } /* * Take additional reference for IO_IN_PROGRESS. It stops * arc_access() from putting this header without any buffers * and so other references but obviously nonevictable onto * the evictable list of MRU or MFU state. */ add_reference(hdr, hdr); if (!embedded_bp) arc_access(hdr, *arc_flags, B_FALSE); arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); arc_hdr_alloc_abd(hdr, alloc_flags); if (encrypted_read) { ASSERT(HDR_HAS_RABD(hdr)); size = HDR_GET_PSIZE(hdr); hdr_abd = hdr->b_crypt_hdr.b_rabd; zio_flags |= ZIO_FLAG_RAW; } else { ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); size = arc_hdr_size(hdr); hdr_abd = hdr->b_l1hdr.b_pabd; if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) { zio_flags |= ZIO_FLAG_RAW_COMPRESS; } /* * For authenticated bp's, we do not ask the ZIO layer * to authenticate them since this will cause the entire * IO to fail if the key isn't loaded. Instead, we * defer authentication until arc_buf_fill(), which will * verify the data when the key is available. */ if (BP_IS_AUTHENTICATED(bp)) zio_flags |= ZIO_FLAG_RAW_ENCRYPT; } if (BP_IS_AUTHENTICATED(bp)) arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH); if (BP_GET_LEVEL(bp) > 0) arc_hdr_set_flags(hdr, ARC_FLAG_INDIRECT); ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state)); acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP); acb->acb_done = done; acb->acb_private = private; acb->acb_compressed = compressed_read; acb->acb_encrypted = encrypted_read; acb->acb_noauth = noauth_read; acb->acb_zb = *zb; ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); hdr->b_l1hdr.b_acb = acb; if (HDR_HAS_L2HDR(hdr) && (vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) { devw = hdr->b_l2hdr.b_dev->l2ad_writing; addr = hdr->b_l2hdr.b_daddr; /* * Lock out L2ARC device removal. */ if (vdev_is_dead(vd) || !spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER)) vd = NULL; } /* * We count both async reads and scrub IOs as asynchronous so * that both can be upgraded in the event of a cache hit while * the read IO is still in-flight. */ if (priority == ZIO_PRIORITY_ASYNC_READ || priority == ZIO_PRIORITY_SCRUB) arc_hdr_set_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ); else arc_hdr_clear_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ); /* * At this point, we have a level 1 cache miss or a blkptr * with embedded data. Try again in L2ARC if possible. */ ASSERT3U(HDR_GET_LSIZE(hdr), ==, lsize); /* * Skip ARC stat bump for block pointers with embedded * data. The data are read from the blkptr itself via * decode_embedded_bp_compressed(). */ if (!embedded_bp) { DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr, blkptr_t *, bp, uint64_t, lsize, zbookmark_phys_t *, zb); ARCSTAT_BUMP(arcstat_misses); ARCSTAT_CONDSTAT(!(*arc_flags & ARC_FLAG_PREFETCH), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, misses); zfs_racct_read(size, 1); } /* Check if the spa even has l2 configured */ const boolean_t spa_has_l2 = l2arc_ndev != 0 && spa->spa_l2cache.sav_count > 0; if (vd != NULL && spa_has_l2 && !(l2arc_norw && devw)) { /* * Read from the L2ARC if the following are true: * 1. The L2ARC vdev was previously cached. * 2. This buffer still has L2ARC metadata. * 3. This buffer isn't currently writing to the L2ARC. * 4. The L2ARC entry wasn't evicted, which may * also have invalidated the vdev. */ if (HDR_HAS_L2HDR(hdr) && !HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr)) { l2arc_read_callback_t *cb; abd_t *abd; uint64_t asize; DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_hits); hdr->b_l2hdr.b_hits++; cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP); cb->l2rcb_hdr = hdr; cb->l2rcb_bp = *bp; cb->l2rcb_zb = *zb; cb->l2rcb_flags = zio_flags; /* * When Compressed ARC is disabled, but the * L2ARC block is compressed, arc_hdr_size() * will have returned LSIZE rather than PSIZE. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr) && HDR_GET_PSIZE(hdr) != 0) { size = HDR_GET_PSIZE(hdr); } asize = vdev_psize_to_asize(vd, size); if (asize != size) { abd = abd_alloc_for_io(asize, HDR_ISTYPE_METADATA(hdr)); cb->l2rcb_abd = abd; } else { abd = hdr_abd; } ASSERT(addr >= VDEV_LABEL_START_SIZE && addr + asize <= vd->vdev_psize - VDEV_LABEL_END_SIZE); /* * l2arc read. The SCL_L2ARC lock will be * released by l2arc_read_done(). * Issue a null zio if the underlying buffer * was squashed to zero size by compression. */ ASSERT3U(arc_hdr_get_compress(hdr), !=, ZIO_COMPRESS_EMPTY); rzio = zio_read_phys(pio, vd, addr, asize, abd, ZIO_CHECKSUM_OFF, l2arc_read_done, cb, priority, zio_flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, B_FALSE); acb->acb_zio_head = rzio; if (hash_lock != NULL) mutex_exit(hash_lock); DTRACE_PROBE2(l2arc__read, vdev_t *, vd, zio_t *, rzio); ARCSTAT_INCR(arcstat_l2_read_bytes, HDR_GET_PSIZE(hdr)); if (*arc_flags & ARC_FLAG_NOWAIT) { zio_nowait(rzio); goto out; } ASSERT(*arc_flags & ARC_FLAG_WAIT); if (zio_wait(rzio) == 0) goto out; /* l2arc read error; goto zio_read() */ if (hash_lock != NULL) mutex_enter(hash_lock); } else { DTRACE_PROBE1(l2arc__miss, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_misses); if (HDR_L2_WRITING(hdr)) ARCSTAT_BUMP(arcstat_l2_rw_clash); spa_config_exit(spa, SCL_L2ARC, vd); } } else { if (vd != NULL) spa_config_exit(spa, SCL_L2ARC, vd); /* * Only a spa with l2 should contribute to l2 * miss stats. (Including the case of having a * faulted cache device - that's also a miss.) */ if (spa_has_l2) { /* * Skip ARC stat bump for block pointers with * embedded data. The data are read from the * blkptr itself via * decode_embedded_bp_compressed(). */ if (!embedded_bp) { DTRACE_PROBE1(l2arc__miss, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_misses); } } } rzio = zio_read(pio, spa, bp, hdr_abd, size, arc_read_done, hdr, priority, zio_flags, zb); acb->acb_zio_head = rzio; if (hash_lock != NULL) mutex_exit(hash_lock); if (*arc_flags & ARC_FLAG_WAIT) { rc = zio_wait(rzio); goto out; } ASSERT(*arc_flags & ARC_FLAG_NOWAIT); zio_nowait(rzio); } out: /* embedded bps don't actually go to disk */ if (!embedded_bp) spa_read_history_add(spa, zb, *arc_flags); spl_fstrans_unmark(cookie); return (rc); done: if (done) done(NULL, zb, bp, buf, private); if (pio && rc != 0) { zio_t *zio = zio_null(pio, spa, NULL, NULL, NULL, zio_flags); zio->io_error = rc; zio_nowait(zio); } goto out; } arc_prune_t * arc_add_prune_callback(arc_prune_func_t *func, void *private) { arc_prune_t *p; p = kmem_alloc(sizeof (*p), KM_SLEEP); p->p_pfunc = func; p->p_private = private; list_link_init(&p->p_node); zfs_refcount_create(&p->p_refcnt); mutex_enter(&arc_prune_mtx); zfs_refcount_add(&p->p_refcnt, &arc_prune_list); list_insert_head(&arc_prune_list, p); mutex_exit(&arc_prune_mtx); return (p); } void arc_remove_prune_callback(arc_prune_t *p) { boolean_t wait = B_FALSE; mutex_enter(&arc_prune_mtx); list_remove(&arc_prune_list, p); if (zfs_refcount_remove(&p->p_refcnt, &arc_prune_list) > 0) wait = B_TRUE; mutex_exit(&arc_prune_mtx); /* wait for arc_prune_task to finish */ if (wait) taskq_wait_outstanding(arc_prune_taskq, 0); ASSERT0(zfs_refcount_count(&p->p_refcnt)); zfs_refcount_destroy(&p->p_refcnt); kmem_free(p, sizeof (*p)); } /* * Helper function for arc_prune_async() it is responsible for safely * handling the execution of a registered arc_prune_func_t. */ static void arc_prune_task(void *ptr) { arc_prune_t *ap = (arc_prune_t *)ptr; arc_prune_func_t *func = ap->p_pfunc; if (func != NULL) func(ap->p_adjust, ap->p_private); (void) zfs_refcount_remove(&ap->p_refcnt, func); } /* * Notify registered consumers they must drop holds on a portion of the ARC * buffers they reference. This provides a mechanism to ensure the ARC can * honor the metadata limit and reclaim otherwise pinned ARC buffers. * * This operation is performed asynchronously so it may be safely called * in the context of the arc_reclaim_thread(). A reference is taken here * for each registered arc_prune_t and the arc_prune_task() is responsible * for releasing it once the registered arc_prune_func_t has completed. */ static void arc_prune_async(uint64_t adjust) { arc_prune_t *ap; mutex_enter(&arc_prune_mtx); for (ap = list_head(&arc_prune_list); ap != NULL; ap = list_next(&arc_prune_list, ap)) { if (zfs_refcount_count(&ap->p_refcnt) >= 2) continue; zfs_refcount_add(&ap->p_refcnt, ap->p_pfunc); ap->p_adjust = adjust; if (taskq_dispatch(arc_prune_taskq, arc_prune_task, ap, TQ_SLEEP) == TASKQID_INVALID) { (void) zfs_refcount_remove(&ap->p_refcnt, ap->p_pfunc); continue; } ARCSTAT_BUMP(arcstat_prune); } mutex_exit(&arc_prune_mtx); } /* * Notify the arc that a block was freed, and thus will never be used again. */ void arc_freed(spa_t *spa, const blkptr_t *bp) { arc_buf_hdr_t *hdr; kmutex_t *hash_lock; uint64_t guid = spa_load_guid(spa); ASSERT(!BP_IS_EMBEDDED(bp)); hdr = buf_hash_find(guid, bp, &hash_lock); if (hdr == NULL) return; /* * We might be trying to free a block that is still doing I/O * (i.e. prefetch) or has some other reference (i.e. a dedup-ed, * dmu_sync-ed block). A block may also have a reference if it is * part of a dedup-ed, dmu_synced write. The dmu_sync() function would * have written the new block to its final resting place on disk but * without the dedup flag set. This would have left the hdr in the MRU * state and discoverable. When the txg finally syncs it detects that * the block was overridden in open context and issues an override I/O. * Since this is a dedup block, the override I/O will determine if the * block is already in the DDT. If so, then it will replace the io_bp * with the bp from the DDT and allow the I/O to finish. When the I/O * reaches the done callback, dbuf_write_override_done, it will * check to see if the io_bp and io_bp_override are identical. * If they are not, then it indicates that the bp was replaced with * the bp in the DDT and the override bp is freed. This allows * us to arrive here with a reference on a block that is being * freed. So if we have an I/O in progress, or a reference to * this hdr, then we don't destroy the hdr. */ if (!HDR_HAS_L1HDR(hdr) || zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) { arc_change_state(arc_anon, hdr); arc_hdr_destroy(hdr); mutex_exit(hash_lock); } else { mutex_exit(hash_lock); } } /* * Release this buffer from the cache, making it an anonymous buffer. This * must be done after a read and prior to modifying the buffer contents. * If the buffer has more than one reference, we must make * a new hdr for the buffer. */ void arc_release(arc_buf_t *buf, const void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; /* * It would be nice to assert that if its DMU metadata (level > * 0 || it's the dnode file), then it must be syncing context. * But we don't know that information at this level. */ ASSERT(HDR_HAS_L1HDR(hdr)); /* * We don't grab the hash lock prior to this check, because if * the buffer's header is in the arc_anon state, it won't be * linked into the hash table. */ if (hdr->b_l1hdr.b_state == arc_anon) { ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(!HDR_IN_HASH_TABLE(hdr)); ASSERT(!HDR_HAS_L2HDR(hdr)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, buf); ASSERT(ARC_BUF_LAST(buf)); ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1); ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); hdr->b_l1hdr.b_arc_access = 0; /* * If the buf is being overridden then it may already * have a hdr that is not empty. */ buf_discard_identity(hdr); arc_buf_thaw(buf); return; } kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); /* * This assignment is only valid as long as the hash_lock is * held, we must be careful not to reference state or the * b_state field after dropping the lock. */ arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT3P(state, !=, arc_anon); /* this buffer is not on any list */ ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), >, 0); if (HDR_HAS_L2HDR(hdr)) { mutex_enter(&hdr->b_l2hdr.b_dev->l2ad_mtx); /* * We have to recheck this conditional again now that * we're holding the l2ad_mtx to prevent a race with * another thread which might be concurrently calling * l2arc_evict(). In that case, l2arc_evict() might have * destroyed the header's L2 portion as we were waiting * to acquire the l2ad_mtx. */ if (HDR_HAS_L2HDR(hdr)) arc_hdr_l2hdr_destroy(hdr); mutex_exit(&hdr->b_l2hdr.b_dev->l2ad_mtx); } /* * Do we have more than one buf? */ if (hdr->b_l1hdr.b_buf != buf || !ARC_BUF_LAST(buf)) { arc_buf_hdr_t *nhdr; uint64_t spa = hdr->b_spa; uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t lsize = HDR_GET_LSIZE(hdr); boolean_t protected = HDR_PROTECTED(hdr); enum zio_compress compress = arc_hdr_get_compress(hdr); arc_buf_contents_t type = arc_buf_type(hdr); VERIFY3U(hdr->b_type, ==, type); ASSERT(hdr->b_l1hdr.b_buf != buf || buf->b_next != NULL); VERIFY3S(remove_reference(hdr, tag), >, 0); if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) { ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf); ASSERT(ARC_BUF_LAST(buf)); } /* * Pull the data off of this hdr and attach it to * a new anonymous hdr. Also find the last buffer * in the hdr's buffer list. */ arc_buf_t *lastbuf = arc_buf_remove(hdr, buf); ASSERT3P(lastbuf, !=, NULL); /* * If the current arc_buf_t and the hdr are sharing their data * buffer, then we must stop sharing that block. */ if (ARC_BUF_SHARED(buf)) { ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf); ASSERT(!arc_buf_is_shared(lastbuf)); /* * First, sever the block sharing relationship between * buf and the arc_buf_hdr_t. */ arc_unshare_buf(hdr, buf); /* * Now we need to recreate the hdr's b_pabd. Since we * have lastbuf handy, we try to share with it, but if * we can't then we allocate a new b_pabd and copy the * data from buf into it. */ if (arc_can_share(hdr, lastbuf)) { arc_share_buf(hdr, lastbuf); } else { arc_hdr_alloc_abd(hdr, 0); abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data, psize); } VERIFY3P(lastbuf->b_data, !=, NULL); } else if (HDR_SHARED_DATA(hdr)) { /* * Uncompressed shared buffers are always at the end * of the list. Compressed buffers don't have the * same requirements. This makes it hard to * simply assert that the lastbuf is shared so * we rely on the hdr's compression flags to determine * if we have a compressed, shared buffer. */ ASSERT(arc_buf_is_shared(lastbuf) || arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF); ASSERT(!arc_buf_is_shared(buf)); } ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); ASSERT3P(state, !=, arc_l2c_only); (void) zfs_refcount_remove_many(&state->arcs_size[type], arc_buf_size(buf), buf); if (zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) { ASSERT3P(state, !=, arc_l2c_only); (void) zfs_refcount_remove_many( &state->arcs_esize[type], arc_buf_size(buf), buf); } arc_cksum_verify(buf); arc_buf_unwatch(buf); /* if this is the last uncompressed buf free the checksum */ if (!arc_hdr_has_uncompressed_buf(hdr)) arc_cksum_free(hdr); mutex_exit(hash_lock); nhdr = arc_hdr_alloc(spa, psize, lsize, protected, compress, hdr->b_complevel, type); ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL); ASSERT0(zfs_refcount_count(&nhdr->b_l1hdr.b_refcnt)); VERIFY3U(nhdr->b_type, ==, type); ASSERT(!HDR_SHARED_DATA(nhdr)); nhdr->b_l1hdr.b_buf = buf; (void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, tag); buf->b_hdr = nhdr; (void) zfs_refcount_add_many(&arc_anon->arcs_size[type], arc_buf_size(buf), buf); } else { ASSERT(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 1); /* protected by hash lock, or hdr is on arc_anon */ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); hdr->b_l1hdr.b_mru_hits = 0; hdr->b_l1hdr.b_mru_ghost_hits = 0; hdr->b_l1hdr.b_mfu_hits = 0; hdr->b_l1hdr.b_mfu_ghost_hits = 0; arc_change_state(arc_anon, hdr); hdr->b_l1hdr.b_arc_access = 0; mutex_exit(hash_lock); buf_discard_identity(hdr); arc_buf_thaw(buf); } } int arc_released(arc_buf_t *buf) { return (buf->b_data != NULL && buf->b_hdr->b_l1hdr.b_state == arc_anon); } #ifdef ZFS_DEBUG int arc_referenced(arc_buf_t *buf) { return (zfs_refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt)); } #endif static void arc_write_ready(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; blkptr_t *bp = zio->io_bp; uint64_t psize = BP_IS_HOLE(bp) ? 0 : BP_GET_PSIZE(bp); fstrans_cookie_t cookie = spl_fstrans_mark(); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!zfs_refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL); /* * If we're reexecuting this zio because the pool suspended, then * cleanup any state that was previously set the first time the * callback was invoked. */ if (zio->io_flags & ZIO_FLAG_REEXECUTED) { arc_cksum_free(hdr); arc_buf_unwatch(buf); if (hdr->b_l1hdr.b_pabd != NULL) { if (ARC_BUF_SHARED(buf)) { arc_unshare_buf(hdr, buf); } else { ASSERT(!arc_buf_is_shared(buf)); arc_hdr_free_abd(hdr, B_FALSE); } } if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); } ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); ASSERT(!HDR_SHARED_DATA(hdr)); ASSERT(!arc_buf_is_shared(buf)); callback->awcb_ready(zio, buf, callback->awcb_private); if (HDR_IO_IN_PROGRESS(hdr)) { ASSERT(zio->io_flags & ZIO_FLAG_REEXECUTED); } else { arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); add_reference(hdr, hdr); /* For IO_IN_PROGRESS. */ } if (BP_IS_PROTECTED(bp)) { /* ZIL blocks are written through zio_rewrite */ ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG); if (BP_SHOULD_BYTESWAP(bp)) { if (BP_GET_LEVEL(bp) > 0) { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64; } else { hdr->b_l1hdr.b_byteswap = DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); } } else { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; } arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED); hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp); hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset; zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv); zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac); } else { arc_hdr_clear_flags(hdr, ARC_FLAG_PROTECTED); } /* * If this block was written for raw encryption but the zio layer * ended up only authenticating it, adjust the buffer flags now. */ if (BP_IS_AUTHENTICATED(bp) && ARC_BUF_ENCRYPTED(buf)) { arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH); buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED; if (BP_GET_COMPRESS(bp) == ZIO_COMPRESS_OFF) buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; } else if (BP_IS_HOLE(bp) && ARC_BUF_ENCRYPTED(buf)) { buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED; buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; } /* this must be done after the buffer flags are adjusted */ arc_cksum_compute(buf); enum zio_compress compress; if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) { compress = ZIO_COMPRESS_OFF; } else { ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp)); compress = BP_GET_COMPRESS(bp); } HDR_SET_PSIZE(hdr, psize); arc_hdr_set_compress(hdr, compress); hdr->b_complevel = zio->io_prop.zp_complevel; if (zio->io_error != 0 || psize == 0) goto out; /* * Fill the hdr with data. If the buffer is encrypted we have no choice * but to copy the data into b_radb. If the hdr is compressed, the data * we want is available from the zio, otherwise we can take it from * the buf. * * We might be able to share the buf's data with the hdr here. However, * doing so would cause the ARC to be full of linear ABDs if we write a * lot of shareable data. As a compromise, we check whether scattered * ABDs are allowed, and assume that if they are then the user wants * the ARC to be primarily filled with them regardless of the data being * written. Therefore, if they're allowed then we allocate one and copy * the data into it; otherwise, we share the data directly if we can. */ if (ARC_BUF_ENCRYPTED(buf)) { ASSERT3U(psize, >, 0); ASSERT(ARC_BUF_COMPRESSED(buf)); arc_hdr_alloc_abd(hdr, ARC_HDR_ALLOC_RDATA | ARC_HDR_USE_RESERVE); abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize); } else if (!(HDR_UNCACHED(hdr) || abd_size_alloc_linear(arc_buf_size(buf))) || !arc_can_share(hdr, buf)) { /* * Ideally, we would always copy the io_abd into b_pabd, but the * user may have disabled compressed ARC, thus we must check the * hdr's compression setting rather than the io_bp's. */ if (BP_IS_ENCRYPTED(bp)) { ASSERT3U(psize, >, 0); arc_hdr_alloc_abd(hdr, ARC_HDR_ALLOC_RDATA | ARC_HDR_USE_RESERVE); abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize); } else if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF && !ARC_BUF_COMPRESSED(buf)) { ASSERT3U(psize, >, 0); arc_hdr_alloc_abd(hdr, ARC_HDR_USE_RESERVE); abd_copy(hdr->b_l1hdr.b_pabd, zio->io_abd, psize); } else { ASSERT3U(zio->io_orig_size, ==, arc_hdr_size(hdr)); arc_hdr_alloc_abd(hdr, ARC_HDR_USE_RESERVE); abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data, arc_buf_size(buf)); } } else { ASSERT3P(buf->b_data, ==, abd_to_buf(zio->io_orig_abd)); ASSERT3U(zio->io_orig_size, ==, arc_buf_size(buf)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, buf); ASSERT(ARC_BUF_LAST(buf)); arc_share_buf(hdr, buf); } out: arc_hdr_verify(hdr, bp); spl_fstrans_unmark(cookie); } static void arc_write_children_ready(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; callback->awcb_children_ready(zio, buf, callback->awcb_private); } static void arc_write_done(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); if (zio->io_error == 0) { arc_hdr_verify(hdr, zio->io_bp); if (BP_IS_HOLE(zio->io_bp) || BP_IS_EMBEDDED(zio->io_bp)) { buf_discard_identity(hdr); } else { hdr->b_dva = *BP_IDENTITY(zio->io_bp); hdr->b_birth = BP_GET_BIRTH(zio->io_bp); } } else { ASSERT(HDR_EMPTY(hdr)); } /* * If the block to be written was all-zero or compressed enough to be * embedded in the BP, no write was performed so there will be no * dva/birth/checksum. The buffer must therefore remain anonymous * (and uncached). */ if (!HDR_EMPTY(hdr)) { arc_buf_hdr_t *exists; kmutex_t *hash_lock; ASSERT3U(zio->io_error, ==, 0); arc_cksum_verify(buf); exists = buf_hash_insert(hdr, &hash_lock); if (exists != NULL) { /* * This can only happen if we overwrite for * sync-to-convergence, because we remove * buffers from the hash table when we arc_free(). */ if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp)) panic("bad overwrite, hdr=%p exists=%p", (void *)hdr, (void *)exists); ASSERT(zfs_refcount_is_zero( &exists->b_l1hdr.b_refcnt)); arc_change_state(arc_anon, exists); arc_hdr_destroy(exists); mutex_exit(hash_lock); exists = buf_hash_insert(hdr, &hash_lock); ASSERT3P(exists, ==, NULL); } else if (zio->io_flags & ZIO_FLAG_NOPWRITE) { /* nopwrite */ ASSERT(zio->io_prop.zp_nopwrite); if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp)) panic("bad nopwrite, hdr=%p exists=%p", (void *)hdr, (void *)exists); } else { /* Dedup */ ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL); ASSERT(ARC_BUF_LAST(hdr->b_l1hdr.b_buf)); ASSERT(hdr->b_l1hdr.b_state == arc_anon); ASSERT(BP_GET_DEDUP(zio->io_bp)); ASSERT(BP_GET_LEVEL(zio->io_bp) == 0); } } arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); VERIFY3S(remove_reference(hdr, hdr), >, 0); /* if it's not anon, we are doing a scrub */ if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon) arc_access(hdr, 0, B_FALSE); mutex_exit(hash_lock); } else { arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); VERIFY3S(remove_reference(hdr, hdr), >, 0); } callback->awcb_done(zio, buf, callback->awcb_private); abd_free(zio->io_abd); kmem_free(callback, sizeof (arc_write_callback_t)); } zio_t * arc_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, arc_buf_t *buf, boolean_t uncached, boolean_t l2arc, const zio_prop_t *zp, arc_write_done_func_t *ready, arc_write_done_func_t *children_ready, arc_write_done_func_t *done, void *private, zio_priority_t priority, int zio_flags, const zbookmark_phys_t *zb) { arc_buf_hdr_t *hdr = buf->b_hdr; arc_write_callback_t *callback; zio_t *zio; zio_prop_t localprop = *zp; ASSERT3P(ready, !=, NULL); ASSERT3P(done, !=, NULL); ASSERT(!HDR_IO_ERROR(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL); if (uncached) arc_hdr_set_flags(hdr, ARC_FLAG_UNCACHED); else if (l2arc) arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE); if (ARC_BUF_ENCRYPTED(buf)) { ASSERT(ARC_BUF_COMPRESSED(buf)); localprop.zp_encrypt = B_TRUE; localprop.zp_compress = HDR_GET_COMPRESS(hdr); localprop.zp_complevel = hdr->b_complevel; localprop.zp_byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ? ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER; memcpy(localprop.zp_salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN); memcpy(localprop.zp_iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN); memcpy(localprop.zp_mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN); if (DMU_OT_IS_ENCRYPTED(localprop.zp_type)) { localprop.zp_nopwrite = B_FALSE; localprop.zp_copies = MIN(localprop.zp_copies, SPA_DVAS_PER_BP - 1); } zio_flags |= ZIO_FLAG_RAW; } else if (ARC_BUF_COMPRESSED(buf)) { ASSERT3U(HDR_GET_LSIZE(hdr), !=, arc_buf_size(buf)); localprop.zp_compress = HDR_GET_COMPRESS(hdr); localprop.zp_complevel = hdr->b_complevel; zio_flags |= ZIO_FLAG_RAW_COMPRESS; } callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP); callback->awcb_ready = ready; callback->awcb_children_ready = children_ready; callback->awcb_done = done; callback->awcb_private = private; callback->awcb_buf = buf; /* * The hdr's b_pabd is now stale, free it now. A new data block * will be allocated when the zio pipeline calls arc_write_ready(). */ if (hdr->b_l1hdr.b_pabd != NULL) { /* * If the buf is currently sharing the data block with * the hdr then we need to break that relationship here. * The hdr will remain with a NULL data pointer and the * buf will take sole ownership of the block. */ if (ARC_BUF_SHARED(buf)) { arc_unshare_buf(hdr, buf); } else { ASSERT(!arc_buf_is_shared(buf)); arc_hdr_free_abd(hdr, B_FALSE); } VERIFY3P(buf->b_data, !=, NULL); } if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); if (!(zio_flags & ZIO_FLAG_RAW)) arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF); ASSERT(!arc_buf_is_shared(buf)); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); zio = zio_write(pio, spa, txg, bp, abd_get_from_buf(buf->b_data, HDR_GET_LSIZE(hdr)), HDR_GET_LSIZE(hdr), arc_buf_size(buf), &localprop, arc_write_ready, (children_ready != NULL) ? arc_write_children_ready : NULL, arc_write_done, callback, priority, zio_flags, zb); return (zio); } void arc_tempreserve_clear(uint64_t reserve) { atomic_add_64(&arc_tempreserve, -reserve); ASSERT((int64_t)arc_tempreserve >= 0); } int arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg) { int error; uint64_t anon_size; if (!arc_no_grow && reserve > arc_c/4 && reserve * 4 > (2ULL << SPA_MAXBLOCKSHIFT)) arc_c = MIN(arc_c_max, reserve * 4); /* * Throttle when the calculated memory footprint for the TXG * exceeds the target ARC size. */ if (reserve > arc_c) { DMU_TX_STAT_BUMP(dmu_tx_memory_reserve); return (SET_ERROR(ERESTART)); } /* * Don't count loaned bufs as in flight dirty data to prevent long * network delays from blocking transactions that are ready to be * assigned to a txg. */ /* assert that it has not wrapped around */ ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0); anon_size = MAX((int64_t) (zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_DATA]) + zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_METADATA]) - arc_loaned_bytes), 0); /* * Writes will, almost always, require additional memory allocations * in order to compress/encrypt/etc the data. We therefore need to * make sure that there is sufficient available memory for this. */ error = arc_memory_throttle(spa, reserve, txg); if (error != 0) return (error); /* * Throttle writes when the amount of dirty data in the cache * gets too large. We try to keep the cache less than half full * of dirty blocks so that our sync times don't grow too large. * * In the case of one pool being built on another pool, we want * to make sure we don't end up throttling the lower (backing) * pool when the upper pool is the majority contributor to dirty * data. To insure we make forward progress during throttling, we * also check the current pool's net dirty data and only throttle * if it exceeds zfs_arc_pool_dirty_percent of the anonymous dirty * data in the cache. * * Note: if two requests come in concurrently, we might let them * both succeed, when one of them should fail. Not a huge deal. */ uint64_t total_dirty = reserve + arc_tempreserve + anon_size; uint64_t spa_dirty_anon = spa_dirty_data(spa); uint64_t rarc_c = arc_warm ? arc_c : arc_c_max; if (total_dirty > rarc_c * zfs_arc_dirty_limit_percent / 100 && anon_size > rarc_c * zfs_arc_anon_limit_percent / 100 && spa_dirty_anon > anon_size * zfs_arc_pool_dirty_percent / 100) { #ifdef ZFS_DEBUG uint64_t meta_esize = zfs_refcount_count( &arc_anon->arcs_esize[ARC_BUFC_METADATA]); uint64_t data_esize = zfs_refcount_count(&arc_anon->arcs_esize[ARC_BUFC_DATA]); dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK " "anon_data=%lluK tempreserve=%lluK rarc_c=%lluK\n", (u_longlong_t)arc_tempreserve >> 10, (u_longlong_t)meta_esize >> 10, (u_longlong_t)data_esize >> 10, (u_longlong_t)reserve >> 10, (u_longlong_t)rarc_c >> 10); #endif DMU_TX_STAT_BUMP(dmu_tx_dirty_throttle); return (SET_ERROR(ERESTART)); } atomic_add_64(&arc_tempreserve, reserve); return (0); } static void arc_kstat_update_state(arc_state_t *state, kstat_named_t *size, kstat_named_t *data, kstat_named_t *metadata, kstat_named_t *evict_data, kstat_named_t *evict_metadata) { data->value.ui64 = zfs_refcount_count(&state->arcs_size[ARC_BUFC_DATA]); metadata->value.ui64 = zfs_refcount_count(&state->arcs_size[ARC_BUFC_METADATA]); size->value.ui64 = data->value.ui64 + metadata->value.ui64; evict_data->value.ui64 = zfs_refcount_count(&state->arcs_esize[ARC_BUFC_DATA]); evict_metadata->value.ui64 = zfs_refcount_count(&state->arcs_esize[ARC_BUFC_METADATA]); } static int arc_kstat_update(kstat_t *ksp, int rw) { arc_stats_t *as = ksp->ks_data; if (rw == KSTAT_WRITE) return (SET_ERROR(EACCES)); as->arcstat_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_hits); as->arcstat_iohits.value.ui64 = wmsum_value(&arc_sums.arcstat_iohits); as->arcstat_misses.value.ui64 = wmsum_value(&arc_sums.arcstat_misses); as->arcstat_demand_data_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_data_hits); as->arcstat_demand_data_iohits.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_data_iohits); as->arcstat_demand_data_misses.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_data_misses); as->arcstat_demand_metadata_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_metadata_hits); as->arcstat_demand_metadata_iohits.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_metadata_iohits); as->arcstat_demand_metadata_misses.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_metadata_misses); as->arcstat_prefetch_data_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_prefetch_data_hits); as->arcstat_prefetch_data_iohits.value.ui64 = wmsum_value(&arc_sums.arcstat_prefetch_data_iohits); as->arcstat_prefetch_data_misses.value.ui64 = wmsum_value(&arc_sums.arcstat_prefetch_data_misses); as->arcstat_prefetch_metadata_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_prefetch_metadata_hits); as->arcstat_prefetch_metadata_iohits.value.ui64 = wmsum_value(&arc_sums.arcstat_prefetch_metadata_iohits); as->arcstat_prefetch_metadata_misses.value.ui64 = wmsum_value(&arc_sums.arcstat_prefetch_metadata_misses); as->arcstat_mru_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_mru_hits); as->arcstat_mru_ghost_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_mru_ghost_hits); as->arcstat_mfu_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_mfu_hits); as->arcstat_mfu_ghost_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_mfu_ghost_hits); as->arcstat_uncached_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_uncached_hits); as->arcstat_deleted.value.ui64 = wmsum_value(&arc_sums.arcstat_deleted); as->arcstat_mutex_miss.value.ui64 = wmsum_value(&arc_sums.arcstat_mutex_miss); as->arcstat_access_skip.value.ui64 = wmsum_value(&arc_sums.arcstat_access_skip); as->arcstat_evict_skip.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_skip); as->arcstat_evict_not_enough.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_not_enough); as->arcstat_evict_l2_cached.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_l2_cached); as->arcstat_evict_l2_eligible.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_l2_eligible); as->arcstat_evict_l2_eligible_mfu.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mfu); as->arcstat_evict_l2_eligible_mru.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mru); as->arcstat_evict_l2_ineligible.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_l2_ineligible); as->arcstat_evict_l2_skip.value.ui64 = wmsum_value(&arc_sums.arcstat_evict_l2_skip); as->arcstat_hash_collisions.value.ui64 = wmsum_value(&arc_sums.arcstat_hash_collisions); as->arcstat_hash_chains.value.ui64 = wmsum_value(&arc_sums.arcstat_hash_chains); as->arcstat_size.value.ui64 = aggsum_value(&arc_sums.arcstat_size); as->arcstat_compressed_size.value.ui64 = wmsum_value(&arc_sums.arcstat_compressed_size); as->arcstat_uncompressed_size.value.ui64 = wmsum_value(&arc_sums.arcstat_uncompressed_size); as->arcstat_overhead_size.value.ui64 = wmsum_value(&arc_sums.arcstat_overhead_size); as->arcstat_hdr_size.value.ui64 = wmsum_value(&arc_sums.arcstat_hdr_size); as->arcstat_data_size.value.ui64 = wmsum_value(&arc_sums.arcstat_data_size); as->arcstat_metadata_size.value.ui64 = wmsum_value(&arc_sums.arcstat_metadata_size); as->arcstat_dbuf_size.value.ui64 = wmsum_value(&arc_sums.arcstat_dbuf_size); #if defined(COMPAT_FREEBSD11) as->arcstat_other_size.value.ui64 = wmsum_value(&arc_sums.arcstat_bonus_size) + wmsum_value(&arc_sums.arcstat_dnode_size) + wmsum_value(&arc_sums.arcstat_dbuf_size); #endif arc_kstat_update_state(arc_anon, &as->arcstat_anon_size, &as->arcstat_anon_data, &as->arcstat_anon_metadata, &as->arcstat_anon_evictable_data, &as->arcstat_anon_evictable_metadata); arc_kstat_update_state(arc_mru, &as->arcstat_mru_size, &as->arcstat_mru_data, &as->arcstat_mru_metadata, &as->arcstat_mru_evictable_data, &as->arcstat_mru_evictable_metadata); arc_kstat_update_state(arc_mru_ghost, &as->arcstat_mru_ghost_size, &as->arcstat_mru_ghost_data, &as->arcstat_mru_ghost_metadata, &as->arcstat_mru_ghost_evictable_data, &as->arcstat_mru_ghost_evictable_metadata); arc_kstat_update_state(arc_mfu, &as->arcstat_mfu_size, &as->arcstat_mfu_data, &as->arcstat_mfu_metadata, &as->arcstat_mfu_evictable_data, &as->arcstat_mfu_evictable_metadata); arc_kstat_update_state(arc_mfu_ghost, &as->arcstat_mfu_ghost_size, &as->arcstat_mfu_ghost_data, &as->arcstat_mfu_ghost_metadata, &as->arcstat_mfu_ghost_evictable_data, &as->arcstat_mfu_ghost_evictable_metadata); arc_kstat_update_state(arc_uncached, &as->arcstat_uncached_size, &as->arcstat_uncached_data, &as->arcstat_uncached_metadata, &as->arcstat_uncached_evictable_data, &as->arcstat_uncached_evictable_metadata); as->arcstat_dnode_size.value.ui64 = wmsum_value(&arc_sums.arcstat_dnode_size); as->arcstat_bonus_size.value.ui64 = wmsum_value(&arc_sums.arcstat_bonus_size); as->arcstat_l2_hits.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_hits); as->arcstat_l2_misses.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_misses); as->arcstat_l2_prefetch_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_prefetch_asize); as->arcstat_l2_mru_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_mru_asize); as->arcstat_l2_mfu_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_mfu_asize); as->arcstat_l2_bufc_data_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_bufc_data_asize); as->arcstat_l2_bufc_metadata_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_bufc_metadata_asize); as->arcstat_l2_feeds.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_feeds); as->arcstat_l2_rw_clash.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rw_clash); as->arcstat_l2_read_bytes.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_read_bytes); as->arcstat_l2_write_bytes.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_write_bytes); as->arcstat_l2_writes_sent.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_writes_sent); as->arcstat_l2_writes_done.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_writes_done); as->arcstat_l2_writes_error.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_writes_error); as->arcstat_l2_writes_lock_retry.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_writes_lock_retry); as->arcstat_l2_evict_lock_retry.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_evict_lock_retry); as->arcstat_l2_evict_reading.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_evict_reading); as->arcstat_l2_evict_l1cached.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_evict_l1cached); as->arcstat_l2_free_on_write.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_free_on_write); as->arcstat_l2_abort_lowmem.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_abort_lowmem); as->arcstat_l2_cksum_bad.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_cksum_bad); as->arcstat_l2_io_error.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_io_error); as->arcstat_l2_lsize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_lsize); as->arcstat_l2_psize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_psize); as->arcstat_l2_hdr_size.value.ui64 = aggsum_value(&arc_sums.arcstat_l2_hdr_size); as->arcstat_l2_log_blk_writes.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_log_blk_writes); as->arcstat_l2_log_blk_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_log_blk_asize); as->arcstat_l2_log_blk_count.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_log_blk_count); as->arcstat_l2_rebuild_success.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_success); as->arcstat_l2_rebuild_abort_unsupported.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_unsupported); as->arcstat_l2_rebuild_abort_io_errors.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_io_errors); as->arcstat_l2_rebuild_abort_dh_errors.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_dh_errors); as->arcstat_l2_rebuild_abort_cksum_lb_errors.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors); as->arcstat_l2_rebuild_abort_lowmem.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_lowmem); as->arcstat_l2_rebuild_size.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_size); as->arcstat_l2_rebuild_asize.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_asize); as->arcstat_l2_rebuild_bufs.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs); as->arcstat_l2_rebuild_bufs_precached.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs_precached); as->arcstat_l2_rebuild_log_blks.value.ui64 = wmsum_value(&arc_sums.arcstat_l2_rebuild_log_blks); as->arcstat_memory_throttle_count.value.ui64 = wmsum_value(&arc_sums.arcstat_memory_throttle_count); as->arcstat_memory_direct_count.value.ui64 = wmsum_value(&arc_sums.arcstat_memory_direct_count); as->arcstat_memory_indirect_count.value.ui64 = wmsum_value(&arc_sums.arcstat_memory_indirect_count); as->arcstat_memory_all_bytes.value.ui64 = arc_all_memory(); as->arcstat_memory_free_bytes.value.ui64 = arc_free_memory(); as->arcstat_memory_available_bytes.value.i64 = arc_available_memory(); as->arcstat_prune.value.ui64 = wmsum_value(&arc_sums.arcstat_prune); as->arcstat_meta_used.value.ui64 = wmsum_value(&arc_sums.arcstat_meta_used); as->arcstat_async_upgrade_sync.value.ui64 = wmsum_value(&arc_sums.arcstat_async_upgrade_sync); as->arcstat_predictive_prefetch.value.ui64 = wmsum_value(&arc_sums.arcstat_predictive_prefetch); as->arcstat_demand_hit_predictive_prefetch.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_hit_predictive_prefetch); as->arcstat_demand_iohit_predictive_prefetch.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_iohit_predictive_prefetch); as->arcstat_prescient_prefetch.value.ui64 = wmsum_value(&arc_sums.arcstat_prescient_prefetch); as->arcstat_demand_hit_prescient_prefetch.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_hit_prescient_prefetch); as->arcstat_demand_iohit_prescient_prefetch.value.ui64 = wmsum_value(&arc_sums.arcstat_demand_iohit_prescient_prefetch); as->arcstat_raw_size.value.ui64 = wmsum_value(&arc_sums.arcstat_raw_size); as->arcstat_cached_only_in_progress.value.ui64 = wmsum_value(&arc_sums.arcstat_cached_only_in_progress); as->arcstat_abd_chunk_waste_size.value.ui64 = wmsum_value(&arc_sums.arcstat_abd_chunk_waste_size); return (0); } /* * This function *must* return indices evenly distributed between all * sublists of the multilist. This is needed due to how the ARC eviction * code is laid out; arc_evict_state() assumes ARC buffers are evenly * distributed between all sublists and uses this assumption when * deciding which sublist to evict from and how much to evict from it. */ static unsigned int arc_state_multilist_index_func(multilist_t *ml, void *obj) { arc_buf_hdr_t *hdr = obj; /* * We rely on b_dva to generate evenly distributed index * numbers using buf_hash below. So, as an added precaution, * let's make sure we never add empty buffers to the arc lists. */ ASSERT(!HDR_EMPTY(hdr)); /* * The assumption here, is the hash value for a given * arc_buf_hdr_t will remain constant throughout its lifetime * (i.e. its b_spa, b_dva, and b_birth fields don't change). * Thus, we don't need to store the header's sublist index * on insertion, as this index can be recalculated on removal. * * Also, the low order bits of the hash value are thought to be * distributed evenly. Otherwise, in the case that the multilist * has a power of two number of sublists, each sublists' usage * would not be evenly distributed. In this context full 64bit * division would be a waste of time, so limit it to 32 bits. */ return ((unsigned int)buf_hash(hdr->b_spa, &hdr->b_dva, hdr->b_birth) % multilist_get_num_sublists(ml)); } static unsigned int arc_state_l2c_multilist_index_func(multilist_t *ml, void *obj) { panic("Header %p insert into arc_l2c_only %p", obj, ml); } #define WARN_IF_TUNING_IGNORED(tuning, value, do_warn) do { \ if ((do_warn) && (tuning) && ((tuning) != (value))) { \ cmn_err(CE_WARN, \ "ignoring tunable %s (using %llu instead)", \ (#tuning), (u_longlong_t)(value)); \ } \ } while (0) /* * Called during module initialization and periodically thereafter to * apply reasonable changes to the exposed performance tunings. Can also be * called explicitly by param_set_arc_*() functions when ARC tunables are * updated manually. Non-zero zfs_* values which differ from the currently set * values will be applied. */ void arc_tuning_update(boolean_t verbose) { uint64_t allmem = arc_all_memory(); /* Valid range: 32M - */ if ((zfs_arc_min) && (zfs_arc_min != arc_c_min) && (zfs_arc_min >= 2ULL << SPA_MAXBLOCKSHIFT) && (zfs_arc_min <= arc_c_max)) { arc_c_min = zfs_arc_min; arc_c = MAX(arc_c, arc_c_min); } WARN_IF_TUNING_IGNORED(zfs_arc_min, arc_c_min, verbose); /* Valid range: 64M - */ if ((zfs_arc_max) && (zfs_arc_max != arc_c_max) && (zfs_arc_max >= MIN_ARC_MAX) && (zfs_arc_max < allmem) && (zfs_arc_max > arc_c_min)) { arc_c_max = zfs_arc_max; arc_c = MIN(arc_c, arc_c_max); if (arc_dnode_limit > arc_c_max) arc_dnode_limit = arc_c_max; } WARN_IF_TUNING_IGNORED(zfs_arc_max, arc_c_max, verbose); /* Valid range: 0 - */ arc_dnode_limit = zfs_arc_dnode_limit ? zfs_arc_dnode_limit : MIN(zfs_arc_dnode_limit_percent, 100) * arc_c_max / 100; WARN_IF_TUNING_IGNORED(zfs_arc_dnode_limit, arc_dnode_limit, verbose); /* Valid range: 1 - N */ if (zfs_arc_grow_retry) arc_grow_retry = zfs_arc_grow_retry; /* Valid range: 1 - N */ if (zfs_arc_shrink_shift) { arc_shrink_shift = zfs_arc_shrink_shift; arc_no_grow_shift = MIN(arc_no_grow_shift, arc_shrink_shift -1); } /* Valid range: 1 - N ms */ if (zfs_arc_min_prefetch_ms) arc_min_prefetch_ms = zfs_arc_min_prefetch_ms; /* Valid range: 1 - N ms */ if (zfs_arc_min_prescient_prefetch_ms) { arc_min_prescient_prefetch_ms = zfs_arc_min_prescient_prefetch_ms; } /* Valid range: 0 - 100 */ if (zfs_arc_lotsfree_percent <= 100) arc_lotsfree_percent = zfs_arc_lotsfree_percent; WARN_IF_TUNING_IGNORED(zfs_arc_lotsfree_percent, arc_lotsfree_percent, verbose); /* Valid range: 0 - */ if ((zfs_arc_sys_free) && (zfs_arc_sys_free != arc_sys_free)) arc_sys_free = MIN(zfs_arc_sys_free, allmem); WARN_IF_TUNING_IGNORED(zfs_arc_sys_free, arc_sys_free, verbose); } static void arc_state_multilist_init(multilist_t *ml, multilist_sublist_index_func_t *index_func, int *maxcountp) { multilist_create(ml, sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), index_func); *maxcountp = MAX(*maxcountp, multilist_get_num_sublists(ml)); } static void arc_state_init(void) { int num_sublists = 0; arc_state_multilist_init(&arc_mru->arcs_list[ARC_BUFC_METADATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mru->arcs_list[ARC_BUFC_DATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mfu->arcs_list[ARC_BUFC_METADATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mfu->arcs_list[ARC_BUFC_DATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_uncached->arcs_list[ARC_BUFC_METADATA], arc_state_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_uncached->arcs_list[ARC_BUFC_DATA], arc_state_multilist_index_func, &num_sublists); /* * L2 headers should never be on the L2 state list since they don't * have L1 headers allocated. Special index function asserts that. */ arc_state_multilist_init(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA], arc_state_l2c_multilist_index_func, &num_sublists); arc_state_multilist_init(&arc_l2c_only->arcs_list[ARC_BUFC_DATA], arc_state_l2c_multilist_index_func, &num_sublists); /* * Keep track of the number of markers needed to reclaim buffers from * any ARC state. The markers will be pre-allocated so as to minimize * the number of memory allocations performed by the eviction thread. */ arc_state_evict_marker_count = num_sublists; zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_uncached->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_uncached->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_create(&arc_anon->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_anon->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mru->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mru->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mru_ghost->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mru_ghost->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mfu->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mfu->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_mfu_ghost->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_mfu_ghost->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_l2c_only->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_l2c_only->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_create(&arc_uncached->arcs_size[ARC_BUFC_DATA]); zfs_refcount_create(&arc_uncached->arcs_size[ARC_BUFC_METADATA]); wmsum_init(&arc_mru_ghost->arcs_hits[ARC_BUFC_DATA], 0); wmsum_init(&arc_mru_ghost->arcs_hits[ARC_BUFC_METADATA], 0); wmsum_init(&arc_mfu_ghost->arcs_hits[ARC_BUFC_DATA], 0); wmsum_init(&arc_mfu_ghost->arcs_hits[ARC_BUFC_METADATA], 0); wmsum_init(&arc_sums.arcstat_hits, 0); wmsum_init(&arc_sums.arcstat_iohits, 0); wmsum_init(&arc_sums.arcstat_misses, 0); wmsum_init(&arc_sums.arcstat_demand_data_hits, 0); wmsum_init(&arc_sums.arcstat_demand_data_iohits, 0); wmsum_init(&arc_sums.arcstat_demand_data_misses, 0); wmsum_init(&arc_sums.arcstat_demand_metadata_hits, 0); wmsum_init(&arc_sums.arcstat_demand_metadata_iohits, 0); wmsum_init(&arc_sums.arcstat_demand_metadata_misses, 0); wmsum_init(&arc_sums.arcstat_prefetch_data_hits, 0); wmsum_init(&arc_sums.arcstat_prefetch_data_iohits, 0); wmsum_init(&arc_sums.arcstat_prefetch_data_misses, 0); wmsum_init(&arc_sums.arcstat_prefetch_metadata_hits, 0); wmsum_init(&arc_sums.arcstat_prefetch_metadata_iohits, 0); wmsum_init(&arc_sums.arcstat_prefetch_metadata_misses, 0); wmsum_init(&arc_sums.arcstat_mru_hits, 0); wmsum_init(&arc_sums.arcstat_mru_ghost_hits, 0); wmsum_init(&arc_sums.arcstat_mfu_hits, 0); wmsum_init(&arc_sums.arcstat_mfu_ghost_hits, 0); wmsum_init(&arc_sums.arcstat_uncached_hits, 0); wmsum_init(&arc_sums.arcstat_deleted, 0); wmsum_init(&arc_sums.arcstat_mutex_miss, 0); wmsum_init(&arc_sums.arcstat_access_skip, 0); wmsum_init(&arc_sums.arcstat_evict_skip, 0); wmsum_init(&arc_sums.arcstat_evict_not_enough, 0); wmsum_init(&arc_sums.arcstat_evict_l2_cached, 0); wmsum_init(&arc_sums.arcstat_evict_l2_eligible, 0); wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mfu, 0); wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mru, 0); wmsum_init(&arc_sums.arcstat_evict_l2_ineligible, 0); wmsum_init(&arc_sums.arcstat_evict_l2_skip, 0); wmsum_init(&arc_sums.arcstat_hash_collisions, 0); wmsum_init(&arc_sums.arcstat_hash_chains, 0); aggsum_init(&arc_sums.arcstat_size, 0); wmsum_init(&arc_sums.arcstat_compressed_size, 0); wmsum_init(&arc_sums.arcstat_uncompressed_size, 0); wmsum_init(&arc_sums.arcstat_overhead_size, 0); wmsum_init(&arc_sums.arcstat_hdr_size, 0); wmsum_init(&arc_sums.arcstat_data_size, 0); wmsum_init(&arc_sums.arcstat_metadata_size, 0); wmsum_init(&arc_sums.arcstat_dbuf_size, 0); wmsum_init(&arc_sums.arcstat_dnode_size, 0); wmsum_init(&arc_sums.arcstat_bonus_size, 0); wmsum_init(&arc_sums.arcstat_l2_hits, 0); wmsum_init(&arc_sums.arcstat_l2_misses, 0); wmsum_init(&arc_sums.arcstat_l2_prefetch_asize, 0); wmsum_init(&arc_sums.arcstat_l2_mru_asize, 0); wmsum_init(&arc_sums.arcstat_l2_mfu_asize, 0); wmsum_init(&arc_sums.arcstat_l2_bufc_data_asize, 0); wmsum_init(&arc_sums.arcstat_l2_bufc_metadata_asize, 0); wmsum_init(&arc_sums.arcstat_l2_feeds, 0); wmsum_init(&arc_sums.arcstat_l2_rw_clash, 0); wmsum_init(&arc_sums.arcstat_l2_read_bytes, 0); wmsum_init(&arc_sums.arcstat_l2_write_bytes, 0); wmsum_init(&arc_sums.arcstat_l2_writes_sent, 0); wmsum_init(&arc_sums.arcstat_l2_writes_done, 0); wmsum_init(&arc_sums.arcstat_l2_writes_error, 0); wmsum_init(&arc_sums.arcstat_l2_writes_lock_retry, 0); wmsum_init(&arc_sums.arcstat_l2_evict_lock_retry, 0); wmsum_init(&arc_sums.arcstat_l2_evict_reading, 0); wmsum_init(&arc_sums.arcstat_l2_evict_l1cached, 0); wmsum_init(&arc_sums.arcstat_l2_free_on_write, 0); wmsum_init(&arc_sums.arcstat_l2_abort_lowmem, 0); wmsum_init(&arc_sums.arcstat_l2_cksum_bad, 0); wmsum_init(&arc_sums.arcstat_l2_io_error, 0); wmsum_init(&arc_sums.arcstat_l2_lsize, 0); wmsum_init(&arc_sums.arcstat_l2_psize, 0); aggsum_init(&arc_sums.arcstat_l2_hdr_size, 0); wmsum_init(&arc_sums.arcstat_l2_log_blk_writes, 0); wmsum_init(&arc_sums.arcstat_l2_log_blk_asize, 0); wmsum_init(&arc_sums.arcstat_l2_log_blk_count, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_success, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_unsupported, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_io_errors, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_dh_errors, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_lowmem, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_size, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_asize, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs_precached, 0); wmsum_init(&arc_sums.arcstat_l2_rebuild_log_blks, 0); wmsum_init(&arc_sums.arcstat_memory_throttle_count, 0); wmsum_init(&arc_sums.arcstat_memory_direct_count, 0); wmsum_init(&arc_sums.arcstat_memory_indirect_count, 0); wmsum_init(&arc_sums.arcstat_prune, 0); wmsum_init(&arc_sums.arcstat_meta_used, 0); wmsum_init(&arc_sums.arcstat_async_upgrade_sync, 0); wmsum_init(&arc_sums.arcstat_predictive_prefetch, 0); wmsum_init(&arc_sums.arcstat_demand_hit_predictive_prefetch, 0); wmsum_init(&arc_sums.arcstat_demand_iohit_predictive_prefetch, 0); wmsum_init(&arc_sums.arcstat_prescient_prefetch, 0); wmsum_init(&arc_sums.arcstat_demand_hit_prescient_prefetch, 0); wmsum_init(&arc_sums.arcstat_demand_iohit_prescient_prefetch, 0); wmsum_init(&arc_sums.arcstat_raw_size, 0); wmsum_init(&arc_sums.arcstat_cached_only_in_progress, 0); wmsum_init(&arc_sums.arcstat_abd_chunk_waste_size, 0); arc_anon->arcs_state = ARC_STATE_ANON; arc_mru->arcs_state = ARC_STATE_MRU; arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST; arc_mfu->arcs_state = ARC_STATE_MFU; arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST; arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY; arc_uncached->arcs_state = ARC_STATE_UNCACHED; } static void arc_state_fini(void) { zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_uncached->arcs_esize[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_uncached->arcs_esize[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_anon->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_anon->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mru->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mru->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mru_ghost->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mru_ghost->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mfu->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mfu->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_mfu_ghost->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_mfu_ghost->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_l2c_only->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_l2c_only->arcs_size[ARC_BUFC_METADATA]); zfs_refcount_destroy(&arc_uncached->arcs_size[ARC_BUFC_DATA]); zfs_refcount_destroy(&arc_uncached->arcs_size[ARC_BUFC_METADATA]); multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_uncached->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_uncached->arcs_list[ARC_BUFC_DATA]); wmsum_fini(&arc_mru_ghost->arcs_hits[ARC_BUFC_DATA]); wmsum_fini(&arc_mru_ghost->arcs_hits[ARC_BUFC_METADATA]); wmsum_fini(&arc_mfu_ghost->arcs_hits[ARC_BUFC_DATA]); wmsum_fini(&arc_mfu_ghost->arcs_hits[ARC_BUFC_METADATA]); wmsum_fini(&arc_sums.arcstat_hits); wmsum_fini(&arc_sums.arcstat_iohits); wmsum_fini(&arc_sums.arcstat_misses); wmsum_fini(&arc_sums.arcstat_demand_data_hits); wmsum_fini(&arc_sums.arcstat_demand_data_iohits); wmsum_fini(&arc_sums.arcstat_demand_data_misses); wmsum_fini(&arc_sums.arcstat_demand_metadata_hits); wmsum_fini(&arc_sums.arcstat_demand_metadata_iohits); wmsum_fini(&arc_sums.arcstat_demand_metadata_misses); wmsum_fini(&arc_sums.arcstat_prefetch_data_hits); wmsum_fini(&arc_sums.arcstat_prefetch_data_iohits); wmsum_fini(&arc_sums.arcstat_prefetch_data_misses); wmsum_fini(&arc_sums.arcstat_prefetch_metadata_hits); wmsum_fini(&arc_sums.arcstat_prefetch_metadata_iohits); wmsum_fini(&arc_sums.arcstat_prefetch_metadata_misses); wmsum_fini(&arc_sums.arcstat_mru_hits); wmsum_fini(&arc_sums.arcstat_mru_ghost_hits); wmsum_fini(&arc_sums.arcstat_mfu_hits); wmsum_fini(&arc_sums.arcstat_mfu_ghost_hits); wmsum_fini(&arc_sums.arcstat_uncached_hits); wmsum_fini(&arc_sums.arcstat_deleted); wmsum_fini(&arc_sums.arcstat_mutex_miss); wmsum_fini(&arc_sums.arcstat_access_skip); wmsum_fini(&arc_sums.arcstat_evict_skip); wmsum_fini(&arc_sums.arcstat_evict_not_enough); wmsum_fini(&arc_sums.arcstat_evict_l2_cached); wmsum_fini(&arc_sums.arcstat_evict_l2_eligible); wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mfu); wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mru); wmsum_fini(&arc_sums.arcstat_evict_l2_ineligible); wmsum_fini(&arc_sums.arcstat_evict_l2_skip); wmsum_fini(&arc_sums.arcstat_hash_collisions); wmsum_fini(&arc_sums.arcstat_hash_chains); aggsum_fini(&arc_sums.arcstat_size); wmsum_fini(&arc_sums.arcstat_compressed_size); wmsum_fini(&arc_sums.arcstat_uncompressed_size); wmsum_fini(&arc_sums.arcstat_overhead_size); wmsum_fini(&arc_sums.arcstat_hdr_size); wmsum_fini(&arc_sums.arcstat_data_size); wmsum_fini(&arc_sums.arcstat_metadata_size); wmsum_fini(&arc_sums.arcstat_dbuf_size); wmsum_fini(&arc_sums.arcstat_dnode_size); wmsum_fini(&arc_sums.arcstat_bonus_size); wmsum_fini(&arc_sums.arcstat_l2_hits); wmsum_fini(&arc_sums.arcstat_l2_misses); wmsum_fini(&arc_sums.arcstat_l2_prefetch_asize); wmsum_fini(&arc_sums.arcstat_l2_mru_asize); wmsum_fini(&arc_sums.arcstat_l2_mfu_asize); wmsum_fini(&arc_sums.arcstat_l2_bufc_data_asize); wmsum_fini(&arc_sums.arcstat_l2_bufc_metadata_asize); wmsum_fini(&arc_sums.arcstat_l2_feeds); wmsum_fini(&arc_sums.arcstat_l2_rw_clash); wmsum_fini(&arc_sums.arcstat_l2_read_bytes); wmsum_fini(&arc_sums.arcstat_l2_write_bytes); wmsum_fini(&arc_sums.arcstat_l2_writes_sent); wmsum_fini(&arc_sums.arcstat_l2_writes_done); wmsum_fini(&arc_sums.arcstat_l2_writes_error); wmsum_fini(&arc_sums.arcstat_l2_writes_lock_retry); wmsum_fini(&arc_sums.arcstat_l2_evict_lock_retry); wmsum_fini(&arc_sums.arcstat_l2_evict_reading); wmsum_fini(&arc_sums.arcstat_l2_evict_l1cached); wmsum_fini(&arc_sums.arcstat_l2_free_on_write); wmsum_fini(&arc_sums.arcstat_l2_abort_lowmem); wmsum_fini(&arc_sums.arcstat_l2_cksum_bad); wmsum_fini(&arc_sums.arcstat_l2_io_error); wmsum_fini(&arc_sums.arcstat_l2_lsize); wmsum_fini(&arc_sums.arcstat_l2_psize); aggsum_fini(&arc_sums.arcstat_l2_hdr_size); wmsum_fini(&arc_sums.arcstat_l2_log_blk_writes); wmsum_fini(&arc_sums.arcstat_l2_log_blk_asize); wmsum_fini(&arc_sums.arcstat_l2_log_blk_count); wmsum_fini(&arc_sums.arcstat_l2_rebuild_success); wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_unsupported); wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_io_errors); wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_dh_errors); wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors); wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_lowmem); wmsum_fini(&arc_sums.arcstat_l2_rebuild_size); wmsum_fini(&arc_sums.arcstat_l2_rebuild_asize); wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs); wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs_precached); wmsum_fini(&arc_sums.arcstat_l2_rebuild_log_blks); wmsum_fini(&arc_sums.arcstat_memory_throttle_count); wmsum_fini(&arc_sums.arcstat_memory_direct_count); wmsum_fini(&arc_sums.arcstat_memory_indirect_count); wmsum_fini(&arc_sums.arcstat_prune); wmsum_fini(&arc_sums.arcstat_meta_used); wmsum_fini(&arc_sums.arcstat_async_upgrade_sync); wmsum_fini(&arc_sums.arcstat_predictive_prefetch); wmsum_fini(&arc_sums.arcstat_demand_hit_predictive_prefetch); wmsum_fini(&arc_sums.arcstat_demand_iohit_predictive_prefetch); wmsum_fini(&arc_sums.arcstat_prescient_prefetch); wmsum_fini(&arc_sums.arcstat_demand_hit_prescient_prefetch); wmsum_fini(&arc_sums.arcstat_demand_iohit_prescient_prefetch); wmsum_fini(&arc_sums.arcstat_raw_size); wmsum_fini(&arc_sums.arcstat_cached_only_in_progress); wmsum_fini(&arc_sums.arcstat_abd_chunk_waste_size); } uint64_t arc_target_bytes(void) { return (arc_c); } void arc_set_limits(uint64_t allmem) { /* Set min cache to 1/32 of all memory, or 32MB, whichever is more. */ arc_c_min = MAX(allmem / 32, 2ULL << SPA_MAXBLOCKSHIFT); /* How to set default max varies by platform. */ arc_c_max = arc_default_max(arc_c_min, allmem); } void arc_init(void) { uint64_t percent, allmem = arc_all_memory(); mutex_init(&arc_evict_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&arc_evict_waiters, sizeof (arc_evict_waiter_t), offsetof(arc_evict_waiter_t, aew_node)); arc_min_prefetch_ms = 1000; arc_min_prescient_prefetch_ms = 6000; #if defined(_KERNEL) arc_lowmem_init(); #endif arc_set_limits(allmem); #ifdef _KERNEL /* * If zfs_arc_max is non-zero at init, meaning it was set in the kernel * environment before the module was loaded, don't block setting the * maximum because it is less than arc_c_min, instead, reset arc_c_min * to a lower value. * zfs_arc_min will be handled by arc_tuning_update(). */ if (zfs_arc_max != 0 && zfs_arc_max >= MIN_ARC_MAX && zfs_arc_max < allmem) { arc_c_max = zfs_arc_max; if (arc_c_min >= arc_c_max) { arc_c_min = MAX(zfs_arc_max / 2, 2ULL << SPA_MAXBLOCKSHIFT); } } #else /* * In userland, there's only the memory pressure that we artificially * create (see arc_available_memory()). Don't let arc_c get too * small, because it can cause transactions to be larger than * arc_c, causing arc_tempreserve_space() to fail. */ arc_c_min = MAX(arc_c_max / 2, 2ULL << SPA_MAXBLOCKSHIFT); #endif arc_c = arc_c_min; /* * 32-bit fixed point fractions of metadata from total ARC size, * MRU data from all data and MRU metadata from all metadata. */ arc_meta = (1ULL << 32) / 4; /* Metadata is 25% of arc_c. */ arc_pd = (1ULL << 32) / 2; /* Data MRU is 50% of data. */ arc_pm = (1ULL << 32) / 2; /* Metadata MRU is 50% of metadata. */ percent = MIN(zfs_arc_dnode_limit_percent, 100); arc_dnode_limit = arc_c_max * percent / 100; /* Apply user specified tunings */ arc_tuning_update(B_TRUE); /* if kmem_flags are set, lets try to use less memory */ if (kmem_debugging()) arc_c = arc_c / 2; if (arc_c < arc_c_min) arc_c = arc_c_min; arc_register_hotplug(); arc_state_init(); buf_init(); list_create(&arc_prune_list, sizeof (arc_prune_t), offsetof(arc_prune_t, p_node)); mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL); arc_prune_taskq = taskq_create("arc_prune", zfs_arc_prune_task_threads, defclsyspri, 100, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC); arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED, sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (arc_ksp != NULL) { arc_ksp->ks_data = &arc_stats; arc_ksp->ks_update = arc_kstat_update; kstat_install(arc_ksp); } arc_state_evict_markers = arc_state_alloc_markers(arc_state_evict_marker_count); arc_evict_zthr = zthr_create_timer("arc_evict", arc_evict_cb_check, arc_evict_cb, NULL, SEC2NSEC(1), defclsyspri); arc_reap_zthr = zthr_create_timer("arc_reap", arc_reap_cb_check, arc_reap_cb, NULL, SEC2NSEC(1), minclsyspri); arc_warm = B_FALSE; /* * Calculate maximum amount of dirty data per pool. * * If it has been set by a module parameter, take that. * Otherwise, use a percentage of physical memory defined by * zfs_dirty_data_max_percent (default 10%) with a cap at * zfs_dirty_data_max_max (default 4G or 25% of physical memory). */ #ifdef __LP64__ if (zfs_dirty_data_max_max == 0) zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024, allmem * zfs_dirty_data_max_max_percent / 100); #else if (zfs_dirty_data_max_max == 0) zfs_dirty_data_max_max = MIN(1ULL * 1024 * 1024 * 1024, allmem * zfs_dirty_data_max_max_percent / 100); #endif if (zfs_dirty_data_max == 0) { zfs_dirty_data_max = allmem * zfs_dirty_data_max_percent / 100; zfs_dirty_data_max = MIN(zfs_dirty_data_max, zfs_dirty_data_max_max); } if (zfs_wrlog_data_max == 0) { /* * dp_wrlog_total is reduced for each txg at the end of * spa_sync(). However, dp_dirty_total is reduced every time * a block is written out. Thus under normal operation, * dp_wrlog_total could grow 2 times as big as * zfs_dirty_data_max. */ zfs_wrlog_data_max = zfs_dirty_data_max * 2; } } void arc_fini(void) { arc_prune_t *p; #ifdef _KERNEL arc_lowmem_fini(); #endif /* _KERNEL */ /* Use B_TRUE to ensure *all* buffers are evicted */ arc_flush(NULL, B_TRUE); if (arc_ksp != NULL) { kstat_delete(arc_ksp); arc_ksp = NULL; } taskq_wait(arc_prune_taskq); taskq_destroy(arc_prune_taskq); mutex_enter(&arc_prune_mtx); while ((p = list_remove_head(&arc_prune_list)) != NULL) { (void) zfs_refcount_remove(&p->p_refcnt, &arc_prune_list); zfs_refcount_destroy(&p->p_refcnt); kmem_free(p, sizeof (*p)); } mutex_exit(&arc_prune_mtx); list_destroy(&arc_prune_list); mutex_destroy(&arc_prune_mtx); (void) zthr_cancel(arc_evict_zthr); (void) zthr_cancel(arc_reap_zthr); arc_state_free_markers(arc_state_evict_markers, arc_state_evict_marker_count); mutex_destroy(&arc_evict_lock); list_destroy(&arc_evict_waiters); /* * Free any buffers that were tagged for destruction. This needs * to occur before arc_state_fini() runs and destroys the aggsum * values which are updated when freeing scatter ABDs. */ l2arc_do_free_on_write(); /* * buf_fini() must proceed arc_state_fini() because buf_fin() may * trigger the release of kmem magazines, which can callback to * arc_space_return() which accesses aggsums freed in act_state_fini(). */ buf_fini(); arc_state_fini(); arc_unregister_hotplug(); /* * We destroy the zthrs after all the ARC state has been * torn down to avoid the case of them receiving any * wakeup() signals after they are destroyed. */ zthr_destroy(arc_evict_zthr); zthr_destroy(arc_reap_zthr); ASSERT0(arc_loaned_bytes); } /* * Level 2 ARC * * The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk. * It uses dedicated storage devices to hold cached data, which are populated * using large infrequent writes. The main role of this cache is to boost * the performance of random read workloads. The intended L2ARC devices * include short-stroked disks, solid state disks, and other media with * substantially faster read latency than disk. * * +-----------------------+ * | ARC | * +-----------------------+ * | ^ ^ * | | | * l2arc_feed_thread() arc_read() * | | | * | l2arc read | * V | | * +---------------+ | * | L2ARC | | * +---------------+ | * | ^ | * l2arc_write() | | * | | | * V | | * +-------+ +-------+ * | vdev | | vdev | * | cache | | cache | * +-------+ +-------+ * +=========+ .-----. * : L2ARC : |-_____-| * : devices : | Disks | * +=========+ `-_____-' * * Read requests are satisfied from the following sources, in order: * * 1) ARC * 2) vdev cache of L2ARC devices * 3) L2ARC devices * 4) vdev cache of disks * 5) disks * * Some L2ARC device types exhibit extremely slow write performance. * To accommodate for this there are some significant differences between * the L2ARC and traditional cache design: * * 1. There is no eviction path from the ARC to the L2ARC. Evictions from * the ARC behave as usual, freeing buffers and placing headers on ghost * lists. The ARC does not send buffers to the L2ARC during eviction as * this would add inflated write latencies for all ARC memory pressure. * * 2. The L2ARC attempts to cache data from the ARC before it is evicted. * It does this by periodically scanning buffers from the eviction-end of * the MFU and MRU ARC lists, copying them to the L2ARC devices if they are * not already there. It scans until a headroom of buffers is satisfied, * which itself is a buffer for ARC eviction. If a compressible buffer is * found during scanning and selected for writing to an L2ARC device, we * temporarily boost scanning headroom during the next scan cycle to make * sure we adapt to compression effects (which might significantly reduce * the data volume we write to L2ARC). The thread that does this is * l2arc_feed_thread(), illustrated below; example sizes are included to * provide a better sense of ratio than this diagram: * * head --> tail * +---------------------+----------+ * ARC_mfu |:::::#:::::::::::::::|o#o###o###|-->. # already on L2ARC * +---------------------+----------+ | o L2ARC eligible * ARC_mru |:#:::::::::::::::::::|#o#ooo####|-->| : ARC buffer * +---------------------+----------+ | * 15.9 Gbytes ^ 32 Mbytes | * headroom | * l2arc_feed_thread() * | * l2arc write hand <--[oooo]--' * | 8 Mbyte * | write max * V * +==============================+ * L2ARC dev |####|#|###|###| |####| ... | * +==============================+ * 32 Gbytes * * 3. If an ARC buffer is copied to the L2ARC but then hit instead of * evicted, then the L2ARC has cached a buffer much sooner than it probably * needed to, potentially wasting L2ARC device bandwidth and storage. It is * safe to say that this is an uncommon case, since buffers at the end of * the ARC lists have moved there due to inactivity. * * 4. If the ARC evicts faster than the L2ARC can maintain a headroom, * then the L2ARC simply misses copying some buffers. This serves as a * pressure valve to prevent heavy read workloads from both stalling the ARC * with waits and clogging the L2ARC with writes. This also helps prevent * the potential for the L2ARC to churn if it attempts to cache content too * quickly, such as during backups of the entire pool. * * 5. After system boot and before the ARC has filled main memory, there are * no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru * lists can remain mostly static. Instead of searching from tail of these * lists as pictured, the l2arc_feed_thread() will search from the list heads * for eligible buffers, greatly increasing its chance of finding them. * * The L2ARC device write speed is also boosted during this time so that * the L2ARC warms up faster. Since there have been no ARC evictions yet, * there are no L2ARC reads, and no fear of degrading read performance * through increased writes. * * 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that * the vdev queue can aggregate them into larger and fewer writes. Each * device is written to in a rotor fashion, sweeping writes through * available space then repeating. * * 7. The L2ARC does not store dirty content. It never needs to flush * write buffers back to disk based storage. * * 8. If an ARC buffer is written (and dirtied) which also exists in the * L2ARC, the now stale L2ARC buffer is immediately dropped. * * The performance of the L2ARC can be tweaked by a number of tunables, which * may be necessary for different workloads: * * l2arc_write_max max write bytes per interval * l2arc_write_boost extra write bytes during device warmup * l2arc_noprefetch skip caching prefetched buffers * l2arc_headroom number of max device writes to precache * l2arc_headroom_boost when we find compressed buffers during ARC * scanning, we multiply headroom by this * percentage factor for the next scan cycle, * since more compressed buffers are likely to * be present * l2arc_feed_secs seconds between L2ARC writing * * Tunables may be removed or added as future performance improvements are * integrated, and also may become zpool properties. * * There are three key functions that control how the L2ARC warms up: * * l2arc_write_eligible() check if a buffer is eligible to cache * l2arc_write_size() calculate how much to write * l2arc_write_interval() calculate sleep delay between writes * * These three functions determine what to write, how much, and how quickly * to send writes. * * L2ARC persistence: * * When writing buffers to L2ARC, we periodically add some metadata to * make sure we can pick them up after reboot, thus dramatically reducing * the impact that any downtime has on the performance of storage systems * with large caches. * * The implementation works fairly simply by integrating the following two * modifications: * * *) When writing to the L2ARC, we occasionally write a "l2arc log block", * which is an additional piece of metadata which describes what's been * written. This allows us to rebuild the arc_buf_hdr_t structures of the * main ARC buffers. There are 2 linked-lists of log blocks headed by * dh_start_lbps[2]. We alternate which chain we append to, so they are * time-wise and offset-wise interleaved, but that is an optimization rather * than for correctness. The log block also includes a pointer to the * previous block in its chain. * * *) We reserve SPA_MINBLOCKSIZE of space at the start of each L2ARC device * for our header bookkeeping purposes. This contains a device header, * which contains our top-level reference structures. We update it each * time we write a new log block, so that we're able to locate it in the * L2ARC device. If this write results in an inconsistent device header * (e.g. due to power failure), we detect this by verifying the header's * checksum and simply fail to reconstruct the L2ARC after reboot. * * Implementation diagram: * * +=== L2ARC device (not to scale) ======================================+ * | ___two newest log block pointers__.__________ | * | / \dh_start_lbps[1] | * | / \ \dh_start_lbps[0]| * |.___/__. V V | * ||L2 dev|....|lb |bufs |lb |bufs |lb |bufs |lb |bufs |lb |---(empty)---| * || hdr| ^ /^ /^ / / | * |+------+ ...--\-------/ \-----/--\------/ / | * | \--------------/ \--------------/ | * +======================================================================+ * * As can be seen on the diagram, rather than using a simple linked list, * we use a pair of linked lists with alternating elements. This is a * performance enhancement due to the fact that we only find out the * address of the next log block access once the current block has been * completely read in. Obviously, this hurts performance, because we'd be * keeping the device's I/O queue at only a 1 operation deep, thus * incurring a large amount of I/O round-trip latency. Having two lists * allows us to fetch two log blocks ahead of where we are currently * rebuilding L2ARC buffers. * * On-device data structures: * * L2ARC device header: l2arc_dev_hdr_phys_t * L2ARC log block: l2arc_log_blk_phys_t * * L2ARC reconstruction: * * When writing data, we simply write in the standard rotary fashion, * evicting buffers as we go and simply writing new data over them (writing * a new log block every now and then). This obviously means that once we * loop around the end of the device, we will start cutting into an already * committed log block (and its referenced data buffers), like so: * * current write head__ __old tail * \ / * V V * <--|bufs |lb |bufs |lb | |bufs |lb |bufs |lb |--> * ^ ^^^^^^^^^___________________________________ * | \ * <> may overwrite this blk and/or its bufs --' * * When importing the pool, we detect this situation and use it to stop * our scanning process (see l2arc_rebuild). * * There is one significant caveat to consider when rebuilding ARC contents * from an L2ARC device: what about invalidated buffers? Given the above * construction, we cannot update blocks which we've already written to amend * them to remove buffers which were invalidated. Thus, during reconstruction, * we might be populating the cache with buffers for data that's not on the * main pool anymore, or may have been overwritten! * * As it turns out, this isn't a problem. Every arc_read request includes * both the DVA and, crucially, the birth TXG of the BP the caller is * looking for. So even if the cache were populated by completely rotten * blocks for data that had been long deleted and/or overwritten, we'll * never actually return bad data from the cache, since the DVA with the * birth TXG uniquely identify a block in space and time - once created, * a block is immutable on disk. The worst thing we have done is wasted * some time and memory at l2arc rebuild to reconstruct outdated ARC * entries that will get dropped from the l2arc as it is being updated * with new blocks. * * L2ARC buffers that have been evicted by l2arc_evict() ahead of the write * hand are not restored. This is done by saving the offset (in bytes) * l2arc_evict() has evicted to in the L2ARC device header and taking it * into account when restoring buffers. */ static boolean_t l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *hdr) { /* * A buffer is *not* eligible for the L2ARC if it: * 1. belongs to a different spa. * 2. is already cached on the L2ARC. * 3. has an I/O in progress (it may be an incomplete read). * 4. is flagged not eligible (zfs property). */ if (hdr->b_spa != spa_guid || HDR_HAS_L2HDR(hdr) || HDR_IO_IN_PROGRESS(hdr) || !HDR_L2CACHE(hdr)) return (B_FALSE); return (B_TRUE); } static uint64_t l2arc_write_size(l2arc_dev_t *dev) { uint64_t size; /* * Make sure our globals have meaningful values in case the user * altered them. */ size = l2arc_write_max; if (size == 0) { cmn_err(CE_NOTE, "l2arc_write_max must be greater than zero, " "resetting it to the default (%d)", L2ARC_WRITE_SIZE); size = l2arc_write_max = L2ARC_WRITE_SIZE; } if (arc_warm == B_FALSE) size += l2arc_write_boost; /* We need to add in the worst case scenario of log block overhead. */ size += l2arc_log_blk_overhead(size, dev); if (dev->l2ad_vdev->vdev_has_trim && l2arc_trim_ahead > 0) { /* * Trim ahead of the write size 64MB or (l2arc_trim_ahead/100) * times the writesize, whichever is greater. */ size += MAX(64 * 1024 * 1024, (size * l2arc_trim_ahead) / 100); } /* * Make sure the write size does not exceed the size of the cache * device. This is important in l2arc_evict(), otherwise infinite * iteration can occur. */ size = MIN(size, (dev->l2ad_end - dev->l2ad_start) / 4); size = P2ROUNDUP(size, 1ULL << dev->l2ad_vdev->vdev_ashift); return (size); } static clock_t l2arc_write_interval(clock_t began, uint64_t wanted, uint64_t wrote) { clock_t interval, next, now; /* * If the ARC lists are busy, increase our write rate; if the * lists are stale, idle back. This is achieved by checking * how much we previously wrote - if it was more than half of * what we wanted, schedule the next write much sooner. */ if (l2arc_feed_again && wrote > (wanted / 2)) interval = (hz * l2arc_feed_min_ms) / 1000; else interval = hz * l2arc_feed_secs; now = ddi_get_lbolt(); next = MAX(now, MIN(now + interval, began + interval)); return (next); } /* * Cycle through L2ARC devices. This is how L2ARC load balances. * If a device is returned, this also returns holding the spa config lock. */ static l2arc_dev_t * l2arc_dev_get_next(void) { l2arc_dev_t *first, *next = NULL; /* * Lock out the removal of spas (spa_namespace_lock), then removal * of cache devices (l2arc_dev_mtx). Once a device has been selected, * both locks will be dropped and a spa config lock held instead. */ mutex_enter(&spa_namespace_lock); mutex_enter(&l2arc_dev_mtx); /* if there are no vdevs, there is nothing to do */ if (l2arc_ndev == 0) goto out; first = NULL; next = l2arc_dev_last; do { /* loop around the list looking for a non-faulted vdev */ if (next == NULL) { next = list_head(l2arc_dev_list); } else { next = list_next(l2arc_dev_list, next); if (next == NULL) next = list_head(l2arc_dev_list); } /* if we have come back to the start, bail out */ if (first == NULL) first = next; else if (next == first) break; ASSERT3P(next, !=, NULL); } while (vdev_is_dead(next->l2ad_vdev) || next->l2ad_rebuild || next->l2ad_trim_all); /* if we were unable to find any usable vdevs, return NULL */ if (vdev_is_dead(next->l2ad_vdev) || next->l2ad_rebuild || next->l2ad_trim_all) next = NULL; l2arc_dev_last = next; out: mutex_exit(&l2arc_dev_mtx); /* * Grab the config lock to prevent the 'next' device from being * removed while we are writing to it. */ if (next != NULL) spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER); mutex_exit(&spa_namespace_lock); return (next); } /* * Free buffers that were tagged for destruction. */ static void l2arc_do_free_on_write(void) { l2arc_data_free_t *df; mutex_enter(&l2arc_free_on_write_mtx); while ((df = list_remove_head(l2arc_free_on_write)) != NULL) { ASSERT3P(df->l2df_abd, !=, NULL); abd_free(df->l2df_abd); kmem_free(df, sizeof (l2arc_data_free_t)); } mutex_exit(&l2arc_free_on_write_mtx); } /* * A write to a cache device has completed. Update all headers to allow * reads from these buffers to begin. */ static void l2arc_write_done(zio_t *zio) { l2arc_write_callback_t *cb; l2arc_lb_abd_buf_t *abd_buf; l2arc_lb_ptr_buf_t *lb_ptr_buf; l2arc_dev_t *dev; l2arc_dev_hdr_phys_t *l2dhdr; list_t *buflist; arc_buf_hdr_t *head, *hdr, *hdr_prev; kmutex_t *hash_lock; int64_t bytes_dropped = 0; cb = zio->io_private; ASSERT3P(cb, !=, NULL); dev = cb->l2wcb_dev; l2dhdr = dev->l2ad_dev_hdr; ASSERT3P(dev, !=, NULL); head = cb->l2wcb_head; ASSERT3P(head, !=, NULL); buflist = &dev->l2ad_buflist; ASSERT3P(buflist, !=, NULL); DTRACE_PROBE2(l2arc__iodone, zio_t *, zio, l2arc_write_callback_t *, cb); /* * All writes completed, or an error was hit. */ top: mutex_enter(&dev->l2ad_mtx); for (hdr = list_prev(buflist, head); hdr; hdr = hdr_prev) { hdr_prev = list_prev(buflist, hdr); hash_lock = HDR_LOCK(hdr); /* * We cannot use mutex_enter or else we can deadlock * with l2arc_write_buffers (due to swapping the order * the hash lock and l2ad_mtx are taken). */ if (!mutex_tryenter(hash_lock)) { /* * Missed the hash lock. We must retry so we * don't leave the ARC_FLAG_L2_WRITING bit set. */ ARCSTAT_BUMP(arcstat_l2_writes_lock_retry); /* * We don't want to rescan the headers we've * already marked as having been written out, so * we reinsert the head node so we can pick up * where we left off. */ list_remove(buflist, head); list_insert_after(buflist, hdr, head); mutex_exit(&dev->l2ad_mtx); /* * We wait for the hash lock to become available * to try and prevent busy waiting, and increase * the chance we'll be able to acquire the lock * the next time around. */ mutex_enter(hash_lock); mutex_exit(hash_lock); goto top; } /* * We could not have been moved into the arc_l2c_only * state while in-flight due to our ARC_FLAG_L2_WRITING * bit being set. Let's just ensure that's being enforced. */ ASSERT(HDR_HAS_L1HDR(hdr)); /* * Skipped - drop L2ARC entry and mark the header as no * longer L2 eligibile. */ if (zio->io_error != 0) { /* * Error - drop L2ARC entry. */ list_remove(buflist, hdr); arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR); uint64_t psize = HDR_GET_PSIZE(hdr); l2arc_hdr_arcstats_decrement(hdr); bytes_dropped += vdev_psize_to_asize(dev->l2ad_vdev, psize); (void) zfs_refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); } /* * Allow ARC to begin reads and ghost list evictions to * this L2ARC entry. */ arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING); mutex_exit(hash_lock); } /* * Free the allocated abd buffers for writing the log blocks. * If the zio failed reclaim the allocated space and remove the * pointers to these log blocks from the log block pointer list * of the L2ARC device. */ while ((abd_buf = list_remove_tail(&cb->l2wcb_abd_list)) != NULL) { abd_free(abd_buf->abd); zio_buf_free(abd_buf, sizeof (*abd_buf)); if (zio->io_error != 0) { lb_ptr_buf = list_remove_head(&dev->l2ad_lbptr_list); /* * L2BLK_GET_PSIZE returns aligned size for log * blocks. */ uint64_t asize = L2BLK_GET_PSIZE((lb_ptr_buf->lb_ptr)->lbp_prop); bytes_dropped += asize; ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize); ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count); zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf); (void) zfs_refcount_remove(&dev->l2ad_lb_count, lb_ptr_buf); kmem_free(lb_ptr_buf->lb_ptr, sizeof (l2arc_log_blkptr_t)); kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t)); } } list_destroy(&cb->l2wcb_abd_list); if (zio->io_error != 0) { ARCSTAT_BUMP(arcstat_l2_writes_error); /* * Restore the lbps array in the header to its previous state. * If the list of log block pointers is empty, zero out the * log block pointers in the device header. */ lb_ptr_buf = list_head(&dev->l2ad_lbptr_list); for (int i = 0; i < 2; i++) { if (lb_ptr_buf == NULL) { /* * If the list is empty zero out the device * header. Otherwise zero out the second log * block pointer in the header. */ if (i == 0) { memset(l2dhdr, 0, dev->l2ad_dev_hdr_asize); } else { memset(&l2dhdr->dh_start_lbps[i], 0, sizeof (l2arc_log_blkptr_t)); } break; } memcpy(&l2dhdr->dh_start_lbps[i], lb_ptr_buf->lb_ptr, sizeof (l2arc_log_blkptr_t)); lb_ptr_buf = list_next(&dev->l2ad_lbptr_list, lb_ptr_buf); } } ARCSTAT_BUMP(arcstat_l2_writes_done); list_remove(buflist, head); ASSERT(!HDR_HAS_L1HDR(head)); kmem_cache_free(hdr_l2only_cache, head); mutex_exit(&dev->l2ad_mtx); ASSERT(dev->l2ad_vdev != NULL); vdev_space_update(dev->l2ad_vdev, -bytes_dropped, 0, 0); l2arc_do_free_on_write(); kmem_free(cb, sizeof (l2arc_write_callback_t)); } static int l2arc_untransform(zio_t *zio, l2arc_read_callback_t *cb) { int ret; spa_t *spa = zio->io_spa; arc_buf_hdr_t *hdr = cb->l2rcb_hdr; blkptr_t *bp = zio->io_bp; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; boolean_t no_crypt = B_FALSE; /* * ZIL data is never be written to the L2ARC, so we don't need * special handling for its unique MAC storage. */ ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG); ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); /* * If the data was encrypted, decrypt it now. Note that * we must check the bp here and not the hdr, since the * hdr does not have its encryption parameters updated * until arc_read_done(). */ if (BP_IS_ENCRYPTED(bp)) { abd_t *eabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr, ARC_HDR_USE_RESERVE); zio_crypt_decode_params_bp(bp, salt, iv); zio_crypt_decode_mac_bp(bp, mac); ret = spa_do_crypt_abd(B_FALSE, spa, &cb->l2rcb_zb, BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, HDR_GET_PSIZE(hdr), eabd, hdr->b_l1hdr.b_pabd, &no_crypt); if (ret != 0) { arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr); goto error; } /* * If we actually performed decryption, replace b_pabd * with the decrypted data. Otherwise we can just throw * our decryption buffer away. */ if (!no_crypt) { arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, arc_hdr_size(hdr), hdr); hdr->b_l1hdr.b_pabd = eabd; zio->io_abd = eabd; } else { arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr); } } /* * If the L2ARC block was compressed, but ARC compression * is disabled we decompress the data into a new buffer and * replace the existing data. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { abd_t *cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr, ARC_HDR_USE_RESERVE); void *tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr)); ret = zio_decompress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr), &hdr->b_complevel); if (ret != 0) { abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr)); arc_free_data_abd(hdr, cabd, arc_hdr_size(hdr), hdr); goto error; } abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr)); arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, arc_hdr_size(hdr), hdr); hdr->b_l1hdr.b_pabd = cabd; zio->io_abd = cabd; zio->io_size = HDR_GET_LSIZE(hdr); } return (0); error: return (ret); } /* * A read to a cache device completed. Validate buffer contents before * handing over to the regular ARC routines. */ static void l2arc_read_done(zio_t *zio) { int tfm_error = 0; l2arc_read_callback_t *cb = zio->io_private; arc_buf_hdr_t *hdr; kmutex_t *hash_lock; boolean_t valid_cksum; boolean_t using_rdata = (BP_IS_ENCRYPTED(&cb->l2rcb_bp) && (cb->l2rcb_flags & ZIO_FLAG_RAW_ENCRYPT)); ASSERT3P(zio->io_vd, !=, NULL); ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE); spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd); ASSERT3P(cb, !=, NULL); hdr = cb->l2rcb_hdr; ASSERT3P(hdr, !=, NULL); hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); /* * If the data was read into a temporary buffer, * move it and free the buffer. */ if (cb->l2rcb_abd != NULL) { ASSERT3U(arc_hdr_size(hdr), <, zio->io_size); if (zio->io_error == 0) { if (using_rdata) { abd_copy(hdr->b_crypt_hdr.b_rabd, cb->l2rcb_abd, arc_hdr_size(hdr)); } else { abd_copy(hdr->b_l1hdr.b_pabd, cb->l2rcb_abd, arc_hdr_size(hdr)); } } /* * The following must be done regardless of whether * there was an error: * - free the temporary buffer * - point zio to the real ARC buffer * - set zio size accordingly * These are required because zio is either re-used for * an I/O of the block in the case of the error * or the zio is passed to arc_read_done() and it * needs real data. */ abd_free(cb->l2rcb_abd); zio->io_size = zio->io_orig_size = arc_hdr_size(hdr); if (using_rdata) { ASSERT(HDR_HAS_RABD(hdr)); zio->io_abd = zio->io_orig_abd = hdr->b_crypt_hdr.b_rabd; } else { ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); zio->io_abd = zio->io_orig_abd = hdr->b_l1hdr.b_pabd; } } ASSERT3P(zio->io_abd, !=, NULL); /* * Check this survived the L2ARC journey. */ ASSERT(zio->io_abd == hdr->b_l1hdr.b_pabd || (HDR_HAS_RABD(hdr) && zio->io_abd == hdr->b_crypt_hdr.b_rabd)); zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */ zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */ zio->io_prop.zp_complevel = hdr->b_complevel; valid_cksum = arc_cksum_is_equal(hdr, zio); /* * b_rabd will always match the data as it exists on disk if it is * being used. Therefore if we are reading into b_rabd we do not * attempt to untransform the data. */ if (valid_cksum && !using_rdata) tfm_error = l2arc_untransform(zio, cb); if (valid_cksum && tfm_error == 0 && zio->io_error == 0 && !HDR_L2_EVICTED(hdr)) { mutex_exit(hash_lock); zio->io_private = hdr; arc_read_done(zio); } else { /* * Buffer didn't survive caching. Increment stats and * reissue to the original storage device. */ if (zio->io_error != 0) { ARCSTAT_BUMP(arcstat_l2_io_error); } else { zio->io_error = SET_ERROR(EIO); } if (!valid_cksum || tfm_error != 0) ARCSTAT_BUMP(arcstat_l2_cksum_bad); /* * If there's no waiter, issue an async i/o to the primary * storage now. If there *is* a waiter, the caller must * issue the i/o in a context where it's OK to block. */ if (zio->io_waiter == NULL) { zio_t *pio = zio_unique_parent(zio); void *abd = (using_rdata) ? hdr->b_crypt_hdr.b_rabd : hdr->b_l1hdr.b_pabd; ASSERT(!pio || pio->io_child_type == ZIO_CHILD_LOGICAL); zio = zio_read(pio, zio->io_spa, zio->io_bp, abd, zio->io_size, arc_read_done, hdr, zio->io_priority, cb->l2rcb_flags, &cb->l2rcb_zb); /* * Original ZIO will be freed, so we need to update * ARC header with the new ZIO pointer to be used * by zio_change_priority() in arc_read(). */ for (struct arc_callback *acb = hdr->b_l1hdr.b_acb; acb != NULL; acb = acb->acb_next) acb->acb_zio_head = zio; mutex_exit(hash_lock); zio_nowait(zio); } else { mutex_exit(hash_lock); } } kmem_free(cb, sizeof (l2arc_read_callback_t)); } /* * This is the list priority from which the L2ARC will search for pages to * cache. This is used within loops (0..3) to cycle through lists in the * desired order. This order can have a significant effect on cache * performance. * * Currently the metadata lists are hit first, MFU then MRU, followed by * the data lists. This function returns a locked list, and also returns * the lock pointer. */ static multilist_sublist_t * l2arc_sublist_lock(int list_num) { multilist_t *ml = NULL; unsigned int idx; ASSERT(list_num >= 0 && list_num < L2ARC_FEED_TYPES); switch (list_num) { case 0: ml = &arc_mfu->arcs_list[ARC_BUFC_METADATA]; break; case 1: ml = &arc_mru->arcs_list[ARC_BUFC_METADATA]; break; case 2: ml = &arc_mfu->arcs_list[ARC_BUFC_DATA]; break; case 3: ml = &arc_mru->arcs_list[ARC_BUFC_DATA]; break; default: return (NULL); } /* * Return a randomly-selected sublist. This is acceptable * because the caller feeds only a little bit of data for each * call (8MB). Subsequent calls will result in different * sublists being selected. */ idx = multilist_get_random_index(ml); return (multilist_sublist_lock_idx(ml, idx)); } /* * Calculates the maximum overhead of L2ARC metadata log blocks for a given * L2ARC write size. l2arc_evict and l2arc_write_size need to include this * overhead in processing to make sure there is enough headroom available * when writing buffers. */ static inline uint64_t l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev) { if (dev->l2ad_log_entries == 0) { return (0); } else { uint64_t log_entries = write_sz >> SPA_MINBLOCKSHIFT; uint64_t log_blocks = (log_entries + dev->l2ad_log_entries - 1) / dev->l2ad_log_entries; return (vdev_psize_to_asize(dev->l2ad_vdev, sizeof (l2arc_log_blk_phys_t)) * log_blocks); } } /* * Evict buffers from the device write hand to the distance specified in * bytes. This distance may span populated buffers, it may span nothing. * This is clearing a region on the L2ARC device ready for writing. * If the 'all' boolean is set, every buffer is evicted. */ static void l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all) { list_t *buflist; arc_buf_hdr_t *hdr, *hdr_prev; kmutex_t *hash_lock; uint64_t taddr; l2arc_lb_ptr_buf_t *lb_ptr_buf, *lb_ptr_buf_prev; vdev_t *vd = dev->l2ad_vdev; boolean_t rerun; buflist = &dev->l2ad_buflist; top: rerun = B_FALSE; if (dev->l2ad_hand + distance > dev->l2ad_end) { /* * When there is no space to accommodate upcoming writes, * evict to the end. Then bump the write and evict hands * to the start and iterate. This iteration does not * happen indefinitely as we make sure in * l2arc_write_size() that when the write hand is reset, * the write size does not exceed the end of the device. */ rerun = B_TRUE; taddr = dev->l2ad_end; } else { taddr = dev->l2ad_hand + distance; } DTRACE_PROBE4(l2arc__evict, l2arc_dev_t *, dev, list_t *, buflist, uint64_t, taddr, boolean_t, all); if (!all) { /* * This check has to be placed after deciding whether to * iterate (rerun). */ if (dev->l2ad_first) { /* * This is the first sweep through the device. There is * nothing to evict. We have already trimmmed the * whole device. */ goto out; } else { /* * Trim the space to be evicted. */ if (vd->vdev_has_trim && dev->l2ad_evict < taddr && l2arc_trim_ahead > 0) { /* * We have to drop the spa_config lock because * vdev_trim_range() will acquire it. * l2ad_evict already accounts for the label * size. To prevent vdev_trim_ranges() from * adding it again, we subtract it from * l2ad_evict. */ spa_config_exit(dev->l2ad_spa, SCL_L2ARC, dev); vdev_trim_simple(vd, dev->l2ad_evict - VDEV_LABEL_START_SIZE, taddr - dev->l2ad_evict); spa_config_enter(dev->l2ad_spa, SCL_L2ARC, dev, RW_READER); } /* * When rebuilding L2ARC we retrieve the evict hand * from the header of the device. Of note, l2arc_evict() * does not actually delete buffers from the cache * device, but trimming may do so depending on the * hardware implementation. Thus keeping track of the * evict hand is useful. */ dev->l2ad_evict = MAX(dev->l2ad_evict, taddr); } } retry: mutex_enter(&dev->l2ad_mtx); /* * We have to account for evicted log blocks. Run vdev_space_update() * on log blocks whose offset (in bytes) is before the evicted offset * (in bytes) by searching in the list of pointers to log blocks * present in the L2ARC device. */ for (lb_ptr_buf = list_tail(&dev->l2ad_lbptr_list); lb_ptr_buf; lb_ptr_buf = lb_ptr_buf_prev) { lb_ptr_buf_prev = list_prev(&dev->l2ad_lbptr_list, lb_ptr_buf); /* L2BLK_GET_PSIZE returns aligned size for log blocks */ uint64_t asize = L2BLK_GET_PSIZE( (lb_ptr_buf->lb_ptr)->lbp_prop); /* * We don't worry about log blocks left behind (ie * lbp_payload_start < l2ad_hand) because l2arc_write_buffers() * will never write more than l2arc_evict() evicts. */ if (!all && l2arc_log_blkptr_valid(dev, lb_ptr_buf->lb_ptr)) { break; } else { vdev_space_update(vd, -asize, 0, 0); ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize); ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count); zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf); (void) zfs_refcount_remove(&dev->l2ad_lb_count, lb_ptr_buf); list_remove(&dev->l2ad_lbptr_list, lb_ptr_buf); kmem_free(lb_ptr_buf->lb_ptr, sizeof (l2arc_log_blkptr_t)); kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t)); } } for (hdr = list_tail(buflist); hdr; hdr = hdr_prev) { hdr_prev = list_prev(buflist, hdr); ASSERT(!HDR_EMPTY(hdr)); hash_lock = HDR_LOCK(hdr); /* * We cannot use mutex_enter or else we can deadlock * with l2arc_write_buffers (due to swapping the order * the hash lock and l2ad_mtx are taken). */ if (!mutex_tryenter(hash_lock)) { /* * Missed the hash lock. Retry. */ ARCSTAT_BUMP(arcstat_l2_evict_lock_retry); mutex_exit(&dev->l2ad_mtx); mutex_enter(hash_lock); mutex_exit(hash_lock); goto retry; } /* * A header can't be on this list if it doesn't have L2 header. */ ASSERT(HDR_HAS_L2HDR(hdr)); /* Ensure this header has finished being written. */ ASSERT(!HDR_L2_WRITING(hdr)); ASSERT(!HDR_L2_WRITE_HEAD(hdr)); if (!all && (hdr->b_l2hdr.b_daddr >= dev->l2ad_evict || hdr->b_l2hdr.b_daddr < dev->l2ad_hand)) { /* * We've evicted to the target address, * or the end of the device. */ mutex_exit(hash_lock); break; } if (!HDR_HAS_L1HDR(hdr)) { ASSERT(!HDR_L2_READING(hdr)); /* * This doesn't exist in the ARC. Destroy. * arc_hdr_destroy() will call list_remove() * and decrement arcstat_l2_lsize. */ arc_change_state(arc_anon, hdr); arc_hdr_destroy(hdr); } else { ASSERT(hdr->b_l1hdr.b_state != arc_l2c_only); ARCSTAT_BUMP(arcstat_l2_evict_l1cached); /* * Invalidate issued or about to be issued * reads, since we may be about to write * over this location. */ if (HDR_L2_READING(hdr)) { ARCSTAT_BUMP(arcstat_l2_evict_reading); arc_hdr_set_flags(hdr, ARC_FLAG_L2_EVICTED); } arc_hdr_l2hdr_destroy(hdr); } mutex_exit(hash_lock); } mutex_exit(&dev->l2ad_mtx); out: /* * We need to check if we evict all buffers, otherwise we may iterate * unnecessarily. */ if (!all && rerun) { /* * Bump device hand to the device start if it is approaching the * end. l2arc_evict() has already evicted ahead for this case. */ dev->l2ad_hand = dev->l2ad_start; dev->l2ad_evict = dev->l2ad_start; dev->l2ad_first = B_FALSE; goto top; } if (!all) { /* * In case of cache device removal (all) the following * assertions may be violated without functional consequences * as the device is about to be removed. */ ASSERT3U(dev->l2ad_hand + distance, <, dev->l2ad_end); if (!dev->l2ad_first) ASSERT3U(dev->l2ad_hand, <=, dev->l2ad_evict); } } /* * Handle any abd transforms that might be required for writing to the L2ARC. * If successful, this function will always return an abd with the data * transformed as it is on disk in a new abd of asize bytes. */ static int l2arc_apply_transforms(spa_t *spa, arc_buf_hdr_t *hdr, uint64_t asize, abd_t **abd_out) { int ret; - void *tmp = NULL; abd_t *cabd = NULL, *eabd = NULL, *to_write = hdr->b_l1hdr.b_pabd; enum zio_compress compress = HDR_GET_COMPRESS(hdr); uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t size = arc_hdr_size(hdr); boolean_t ismd = HDR_ISTYPE_METADATA(hdr); boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); dsl_crypto_key_t *dck = NULL; uint8_t mac[ZIO_DATA_MAC_LEN] = { 0 }; boolean_t no_crypt = B_FALSE; ASSERT((HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) || HDR_ENCRYPTED(hdr) || HDR_SHARED_DATA(hdr) || psize != asize); ASSERT3U(psize, <=, asize); /* * If this data simply needs its own buffer, we simply allocate it * and copy the data. This may be done to eliminate a dependency on a * shared buffer or to reallocate the buffer to match asize. */ - if (HDR_HAS_RABD(hdr) && asize != psize) { - ASSERT3U(asize, >=, psize); + if (HDR_HAS_RABD(hdr)) { + ASSERT3U(asize, >, psize); to_write = abd_alloc_for_io(asize, ismd); abd_copy(to_write, hdr->b_crypt_hdr.b_rabd, psize); - if (psize != asize) - abd_zero_off(to_write, psize, asize - psize); + abd_zero_off(to_write, psize, asize - psize); goto out; } if ((compress == ZIO_COMPRESS_OFF || HDR_COMPRESSION_ENABLED(hdr)) && !HDR_ENCRYPTED(hdr)) { ASSERT3U(size, ==, psize); to_write = abd_alloc_for_io(asize, ismd); abd_copy(to_write, hdr->b_l1hdr.b_pabd, size); - if (size != asize) + if (asize > size) abd_zero_off(to_write, size, asize - size); goto out; } if (compress != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { - /* - * In some cases, we can wind up with size > asize, so - * we need to opt for the larger allocation option here. - * - * (We also need abd_return_buf_copy in all cases because - * it's an ASSERT() to modify the buffer before returning it - * with arc_return_buf(), and all the compressors - * write things before deciding to fail compression in nearly - * every case.) - */ - uint64_t bufsize = MAX(size, asize); - cabd = abd_alloc_for_io(bufsize, ismd); - tmp = abd_borrow_buf(cabd, bufsize); - - psize = zio_compress_data(compress, to_write, &tmp, size, - hdr->b_complevel); - - if (psize >= asize) { - psize = HDR_GET_PSIZE(hdr); - abd_return_buf_copy(cabd, tmp, bufsize); - HDR_SET_COMPRESS(hdr, ZIO_COMPRESS_OFF); - to_write = cabd; - abd_copy(to_write, hdr->b_l1hdr.b_pabd, psize); - if (psize != asize) - abd_zero_off(to_write, psize, asize - psize); - goto encrypt; + size_t bufsize = MAX(size, asize); + void *buf = zio_buf_alloc(bufsize); + uint64_t csize = zio_compress_data(compress, to_write, &buf, + size, hdr->b_complevel); + if (csize > psize) { + /* + * We can't re-compress the block into the original + * psize. Even if it fits into asize, it does not + * matter, since checksum will never match on read. + */ + zio_buf_free(buf, bufsize); + return (SET_ERROR(EIO)); } - ASSERT3U(psize, <=, HDR_GET_PSIZE(hdr)); - if (psize < asize) - memset((char *)tmp + psize, 0, bufsize - psize); - psize = HDR_GET_PSIZE(hdr); - abd_return_buf_copy(cabd, tmp, bufsize); - to_write = cabd; + if (asize > csize) + memset((char *)buf + csize, 0, asize - csize); + to_write = cabd = abd_get_from_buf(buf, bufsize); + abd_take_ownership_of_buf(cabd, B_TRUE); } -encrypt: if (HDR_ENCRYPTED(hdr)) { eabd = abd_alloc_for_io(asize, ismd); /* * If the dataset was disowned before the buffer * made it to this point, the key to re-encrypt * it won't be available. In this case we simply * won't write the buffer to the L2ARC. */ ret = spa_keystore_lookup_key(spa, hdr->b_crypt_hdr.b_dsobj, FTAG, &dck); if (ret != 0) goto error; ret = zio_do_crypt_abd(B_TRUE, &dck->dck_key, hdr->b_crypt_hdr.b_ot, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv, mac, psize, to_write, eabd, &no_crypt); if (ret != 0) goto error; if (no_crypt) abd_copy(eabd, to_write, psize); if (psize != asize) abd_zero_off(eabd, psize, asize - psize); /* assert that the MAC we got here matches the one we saved */ ASSERT0(memcmp(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN)); spa_keystore_dsl_key_rele(spa, dck, FTAG); if (to_write == cabd) abd_free(cabd); to_write = eabd; } out: ASSERT3P(to_write, !=, hdr->b_l1hdr.b_pabd); *abd_out = to_write; return (0); error: if (dck != NULL) spa_keystore_dsl_key_rele(spa, dck, FTAG); if (cabd != NULL) abd_free(cabd); if (eabd != NULL) abd_free(eabd); *abd_out = NULL; return (ret); } static void l2arc_blk_fetch_done(zio_t *zio) { l2arc_read_callback_t *cb; cb = zio->io_private; if (cb->l2rcb_abd != NULL) abd_free(cb->l2rcb_abd); kmem_free(cb, sizeof (l2arc_read_callback_t)); } /* * Find and write ARC buffers to the L2ARC device. * * An ARC_FLAG_L2_WRITING flag is set so that the L2ARC buffers are not valid * for reading until they have completed writing. * The headroom_boost is an in-out parameter used to maintain headroom boost * state between calls to this function. * * Returns the number of bytes actually written (which may be smaller than * the delta by which the device hand has changed due to alignment and the * writing of log blocks). */ static uint64_t l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz) { arc_buf_hdr_t *hdr, *head, *marker; uint64_t write_asize, write_psize, headroom; boolean_t full, from_head = !arc_warm; l2arc_write_callback_t *cb = NULL; zio_t *pio, *wzio; uint64_t guid = spa_load_guid(spa); l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr; ASSERT3P(dev->l2ad_vdev, !=, NULL); pio = NULL; write_asize = write_psize = 0; full = B_FALSE; head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE); arc_hdr_set_flags(head, ARC_FLAG_L2_WRITE_HEAD | ARC_FLAG_HAS_L2HDR); marker = arc_state_alloc_marker(); /* * Copy buffers for L2ARC writing. */ for (int pass = 0; pass < L2ARC_FEED_TYPES; pass++) { /* * If pass == 1 or 3, we cache MRU metadata and data * respectively. */ if (l2arc_mfuonly) { if (pass == 1 || pass == 3) continue; } uint64_t passed_sz = 0; headroom = target_sz * l2arc_headroom; if (zfs_compressed_arc_enabled) headroom = (headroom * l2arc_headroom_boost) / 100; /* * Until the ARC is warm and starts to evict, read from the * head of the ARC lists rather than the tail. */ multilist_sublist_t *mls = l2arc_sublist_lock(pass); ASSERT3P(mls, !=, NULL); if (from_head) hdr = multilist_sublist_head(mls); else hdr = multilist_sublist_tail(mls); while (hdr != NULL) { kmutex_t *hash_lock; abd_t *to_write = NULL; hash_lock = HDR_LOCK(hdr); if (!mutex_tryenter(hash_lock)) { skip: /* Skip this buffer rather than waiting. */ if (from_head) hdr = multilist_sublist_next(mls, hdr); else hdr = multilist_sublist_prev(mls, hdr); continue; } passed_sz += HDR_GET_LSIZE(hdr); if (l2arc_headroom != 0 && passed_sz > headroom) { /* * Searched too far. */ mutex_exit(hash_lock); break; } if (!l2arc_write_eligible(guid, hdr)) { mutex_exit(hash_lock); goto skip; } ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3U(HDR_GET_PSIZE(hdr), >, 0); ASSERT3U(arc_hdr_size(hdr), >, 0); ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize); /* * If the allocated size of this buffer plus the max * size for the pending log block exceeds the evicted * target size, terminate writing buffers for this run. */ if (write_asize + asize + sizeof (l2arc_log_blk_phys_t) > target_sz) { full = B_TRUE; mutex_exit(hash_lock); break; } /* * We should not sleep with sublist lock held or it * may block ARC eviction. Insert a marker to save * the position and drop the lock. */ if (from_head) { multilist_sublist_insert_after(mls, hdr, marker); } else { multilist_sublist_insert_before(mls, hdr, marker); } multilist_sublist_unlock(mls); /* * If this header has b_rabd, we can use this since it * must always match the data exactly as it exists on * disk. Otherwise, the L2ARC can normally use the * hdr's data, but if we're sharing data between the * hdr and one of its bufs, L2ARC needs its own copy of * the data so that the ZIO below can't race with the * buf consumer. To ensure that this copy will be * available for the lifetime of the ZIO and be cleaned * up afterwards, we add it to the l2arc_free_on_write * queue. If we need to apply any transforms to the * data (compression, encryption) we will also need the * extra buffer. */ if (HDR_HAS_RABD(hdr) && psize == asize) { to_write = hdr->b_crypt_hdr.b_rabd; } else if ((HDR_COMPRESSION_ENABLED(hdr) || HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF) && !HDR_ENCRYPTED(hdr) && !HDR_SHARED_DATA(hdr) && psize == asize) { to_write = hdr->b_l1hdr.b_pabd; } else { int ret; arc_buf_contents_t type = arc_buf_type(hdr); ret = l2arc_apply_transforms(spa, hdr, asize, &to_write); if (ret != 0) { arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE); mutex_exit(hash_lock); goto next; } l2arc_free_abd_on_write(to_write, asize, type); } hdr->b_l2hdr.b_dev = dev; hdr->b_l2hdr.b_daddr = dev->l2ad_hand; hdr->b_l2hdr.b_hits = 0; hdr->b_l2hdr.b_arcs_state = hdr->b_l1hdr.b_state->arcs_state; mutex_enter(&dev->l2ad_mtx); if (pio == NULL) { /* * Insert a dummy header on the buflist so * l2arc_write_done() can find where the * write buffers begin without searching. */ list_insert_head(&dev->l2ad_buflist, head); } list_insert_head(&dev->l2ad_buflist, hdr); mutex_exit(&dev->l2ad_mtx); arc_hdr_set_flags(hdr, ARC_FLAG_HAS_L2HDR | ARC_FLAG_L2_WRITING); (void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); l2arc_hdr_arcstats_increment(hdr); boolean_t commit = l2arc_log_blk_insert(dev, hdr); mutex_exit(hash_lock); if (pio == NULL) { cb = kmem_alloc( sizeof (l2arc_write_callback_t), KM_SLEEP); cb->l2wcb_dev = dev; cb->l2wcb_head = head; list_create(&cb->l2wcb_abd_list, sizeof (l2arc_lb_abd_buf_t), offsetof(l2arc_lb_abd_buf_t, node)); pio = zio_root(spa, l2arc_write_done, cb, ZIO_FLAG_CANFAIL); } wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand, asize, to_write, ZIO_CHECKSUM_OFF, NULL, hdr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE); DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev, zio_t *, wzio); zio_nowait(wzio); write_psize += psize; write_asize += asize; dev->l2ad_hand += asize; vdev_space_update(dev->l2ad_vdev, asize, 0, 0); if (commit) { /* l2ad_hand will be adjusted inside. */ write_asize += l2arc_log_blk_commit(dev, pio, cb); } next: multilist_sublist_lock(mls); if (from_head) hdr = multilist_sublist_next(mls, marker); else hdr = multilist_sublist_prev(mls, marker); multilist_sublist_remove(mls, marker); } multilist_sublist_unlock(mls); if (full == B_TRUE) break; } arc_state_free_marker(marker); /* No buffers selected for writing? */ if (pio == NULL) { ASSERT0(write_psize); ASSERT(!HDR_HAS_L1HDR(head)); kmem_cache_free(hdr_l2only_cache, head); /* * Although we did not write any buffers l2ad_evict may * have advanced. */ if (dev->l2ad_evict != l2dhdr->dh_evict) l2arc_dev_hdr_update(dev); return (0); } if (!dev->l2ad_first) ASSERT3U(dev->l2ad_hand, <=, dev->l2ad_evict); ASSERT3U(write_asize, <=, target_sz); ARCSTAT_BUMP(arcstat_l2_writes_sent); ARCSTAT_INCR(arcstat_l2_write_bytes, write_psize); dev->l2ad_writing = B_TRUE; (void) zio_wait(pio); dev->l2ad_writing = B_FALSE; /* * Update the device header after the zio completes as * l2arc_write_done() may have updated the memory holding the log block * pointers in the device header. */ l2arc_dev_hdr_update(dev); return (write_asize); } static boolean_t l2arc_hdr_limit_reached(void) { int64_t s = aggsum_upper_bound(&arc_sums.arcstat_l2_hdr_size); return (arc_reclaim_needed() || (s > (arc_warm ? arc_c : arc_c_max) * l2arc_meta_percent / 100)); } /* * This thread feeds the L2ARC at regular intervals. This is the beating * heart of the L2ARC. */ static __attribute__((noreturn)) void l2arc_feed_thread(void *unused) { (void) unused; callb_cpr_t cpr; l2arc_dev_t *dev; spa_t *spa; uint64_t size, wrote; clock_t begin, next = ddi_get_lbolt(); fstrans_cookie_t cookie; CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG); mutex_enter(&l2arc_feed_thr_lock); cookie = spl_fstrans_mark(); while (l2arc_thread_exit == 0) { CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait_idle(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock, next); CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock); next = ddi_get_lbolt() + hz; /* * Quick check for L2ARC devices. */ mutex_enter(&l2arc_dev_mtx); if (l2arc_ndev == 0) { mutex_exit(&l2arc_dev_mtx); continue; } mutex_exit(&l2arc_dev_mtx); begin = ddi_get_lbolt(); /* * This selects the next l2arc device to write to, and in * doing so the next spa to feed from: dev->l2ad_spa. This * will return NULL if there are now no l2arc devices or if * they are all faulted. * * If a device is returned, its spa's config lock is also * held to prevent device removal. l2arc_dev_get_next() * will grab and release l2arc_dev_mtx. */ if ((dev = l2arc_dev_get_next()) == NULL) continue; spa = dev->l2ad_spa; ASSERT3P(spa, !=, NULL); /* * If the pool is read-only then force the feed thread to * sleep a little longer. */ if (!spa_writeable(spa)) { next = ddi_get_lbolt() + 5 * l2arc_feed_secs * hz; spa_config_exit(spa, SCL_L2ARC, dev); continue; } /* * Avoid contributing to memory pressure. */ if (l2arc_hdr_limit_reached()) { ARCSTAT_BUMP(arcstat_l2_abort_lowmem); spa_config_exit(spa, SCL_L2ARC, dev); continue; } ARCSTAT_BUMP(arcstat_l2_feeds); size = l2arc_write_size(dev); /* * Evict L2ARC buffers that will be overwritten. */ l2arc_evict(dev, size, B_FALSE); /* * Write ARC buffers. */ wrote = l2arc_write_buffers(spa, dev, size); /* * Calculate interval between writes. */ next = l2arc_write_interval(begin, size, wrote); spa_config_exit(spa, SCL_L2ARC, dev); } spl_fstrans_unmark(cookie); l2arc_thread_exit = 0; cv_broadcast(&l2arc_feed_thr_cv); CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */ thread_exit(); } boolean_t l2arc_vdev_present(vdev_t *vd) { return (l2arc_vdev_get(vd) != NULL); } /* * Returns the l2arc_dev_t associated with a particular vdev_t or NULL if * the vdev_t isn't an L2ARC device. */ l2arc_dev_t * l2arc_vdev_get(vdev_t *vd) { l2arc_dev_t *dev; mutex_enter(&l2arc_dev_mtx); for (dev = list_head(l2arc_dev_list); dev != NULL; dev = list_next(l2arc_dev_list, dev)) { if (dev->l2ad_vdev == vd) break; } mutex_exit(&l2arc_dev_mtx); return (dev); } static void l2arc_rebuild_dev(l2arc_dev_t *dev, boolean_t reopen) { l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr; uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize; spa_t *spa = dev->l2ad_spa; /* * The L2ARC has to hold at least the payload of one log block for * them to be restored (persistent L2ARC). The payload of a log block * depends on the amount of its log entries. We always write log blocks * with 1022 entries. How many of them are committed or restored depends * on the size of the L2ARC device. Thus the maximum payload of * one log block is 1022 * SPA_MAXBLOCKSIZE = 16GB. If the L2ARC device * is less than that, we reduce the amount of committed and restored * log entries per block so as to enable persistence. */ if (dev->l2ad_end < l2arc_rebuild_blocks_min_l2size) { dev->l2ad_log_entries = 0; } else { dev->l2ad_log_entries = MIN((dev->l2ad_end - dev->l2ad_start) >> SPA_MAXBLOCKSHIFT, L2ARC_LOG_BLK_MAX_ENTRIES); } /* * Read the device header, if an error is returned do not rebuild L2ARC. */ if (l2arc_dev_hdr_read(dev) == 0 && dev->l2ad_log_entries > 0) { /* * If we are onlining a cache device (vdev_reopen) that was * still present (l2arc_vdev_present()) and rebuild is enabled, * we should evict all ARC buffers and pointers to log blocks * and reclaim their space before restoring its contents to * L2ARC. */ if (reopen) { if (!l2arc_rebuild_enabled) { return; } else { l2arc_evict(dev, 0, B_TRUE); /* start a new log block */ dev->l2ad_log_ent_idx = 0; dev->l2ad_log_blk_payload_asize = 0; dev->l2ad_log_blk_payload_start = 0; } } /* * Just mark the device as pending for a rebuild. We won't * be starting a rebuild in line here as it would block pool * import. Instead spa_load_impl will hand that off to an * async task which will call l2arc_spa_rebuild_start. */ dev->l2ad_rebuild = B_TRUE; } else if (spa_writeable(spa)) { /* * In this case TRIM the whole device if l2arc_trim_ahead > 0, * otherwise create a new header. We zero out the memory holding * the header to reset dh_start_lbps. If we TRIM the whole * device the new header will be written by * vdev_trim_l2arc_thread() at the end of the TRIM to update the * trim_state in the header too. When reading the header, if * trim_state is not VDEV_TRIM_COMPLETE and l2arc_trim_ahead > 0 * we opt to TRIM the whole device again. */ if (l2arc_trim_ahead > 0) { dev->l2ad_trim_all = B_TRUE; } else { memset(l2dhdr, 0, l2dhdr_asize); l2arc_dev_hdr_update(dev); } } } /* * Add a vdev for use by the L2ARC. By this point the spa has already * validated the vdev and opened it. */ void l2arc_add_vdev(spa_t *spa, vdev_t *vd) { l2arc_dev_t *adddev; uint64_t l2dhdr_asize; ASSERT(!l2arc_vdev_present(vd)); /* * Create a new l2arc device entry. */ adddev = vmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP); adddev->l2ad_spa = spa; adddev->l2ad_vdev = vd; /* leave extra size for an l2arc device header */ l2dhdr_asize = adddev->l2ad_dev_hdr_asize = MAX(sizeof (*adddev->l2ad_dev_hdr), 1 << vd->vdev_ashift); adddev->l2ad_start = VDEV_LABEL_START_SIZE + l2dhdr_asize; adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd); ASSERT3U(adddev->l2ad_start, <, adddev->l2ad_end); adddev->l2ad_hand = adddev->l2ad_start; adddev->l2ad_evict = adddev->l2ad_start; adddev->l2ad_first = B_TRUE; adddev->l2ad_writing = B_FALSE; adddev->l2ad_trim_all = B_FALSE; list_link_init(&adddev->l2ad_node); adddev->l2ad_dev_hdr = kmem_zalloc(l2dhdr_asize, KM_SLEEP); mutex_init(&adddev->l2ad_mtx, NULL, MUTEX_DEFAULT, NULL); /* * This is a list of all ARC buffers that are still valid on the * device. */ list_create(&adddev->l2ad_buflist, sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l2hdr.b_l2node)); /* * This is a list of pointers to log blocks that are still present * on the device. */ list_create(&adddev->l2ad_lbptr_list, sizeof (l2arc_lb_ptr_buf_t), offsetof(l2arc_lb_ptr_buf_t, node)); vdev_space_update(vd, 0, 0, adddev->l2ad_end - adddev->l2ad_hand); zfs_refcount_create(&adddev->l2ad_alloc); zfs_refcount_create(&adddev->l2ad_lb_asize); zfs_refcount_create(&adddev->l2ad_lb_count); /* * Decide if dev is eligible for L2ARC rebuild or whole device * trimming. This has to happen before the device is added in the * cache device list and l2arc_dev_mtx is released. Otherwise * l2arc_feed_thread() might already start writing on the * device. */ l2arc_rebuild_dev(adddev, B_FALSE); /* * Add device to global list */ mutex_enter(&l2arc_dev_mtx); list_insert_head(l2arc_dev_list, adddev); atomic_inc_64(&l2arc_ndev); mutex_exit(&l2arc_dev_mtx); } /* * Decide if a vdev is eligible for L2ARC rebuild, called from vdev_reopen() * in case of onlining a cache device. */ void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen) { l2arc_dev_t *dev = NULL; dev = l2arc_vdev_get(vd); ASSERT3P(dev, !=, NULL); /* * In contrast to l2arc_add_vdev() we do not have to worry about * l2arc_feed_thread() invalidating previous content when onlining a * cache device. The device parameters (l2ad*) are not cleared when * offlining the device and writing new buffers will not invalidate * all previous content. In worst case only buffers that have not had * their log block written to the device will be lost. * When onlining the cache device (ie offline->online without exporting * the pool in between) this happens: * vdev_reopen() -> vdev_open() -> l2arc_rebuild_vdev() * | | * vdev_is_dead() = B_FALSE l2ad_rebuild = B_TRUE * During the time where vdev_is_dead = B_FALSE and until l2ad_rebuild * is set to B_TRUE we might write additional buffers to the device. */ l2arc_rebuild_dev(dev, reopen); } /* * Remove a vdev from the L2ARC. */ void l2arc_remove_vdev(vdev_t *vd) { l2arc_dev_t *remdev = NULL; /* * Find the device by vdev */ remdev = l2arc_vdev_get(vd); ASSERT3P(remdev, !=, NULL); /* * Cancel any ongoing or scheduled rebuild. */ mutex_enter(&l2arc_rebuild_thr_lock); if (remdev->l2ad_rebuild_began == B_TRUE) { remdev->l2ad_rebuild_cancel = B_TRUE; while (remdev->l2ad_rebuild == B_TRUE) cv_wait(&l2arc_rebuild_thr_cv, &l2arc_rebuild_thr_lock); } mutex_exit(&l2arc_rebuild_thr_lock); /* * Remove device from global list */ mutex_enter(&l2arc_dev_mtx); list_remove(l2arc_dev_list, remdev); l2arc_dev_last = NULL; /* may have been invalidated */ atomic_dec_64(&l2arc_ndev); mutex_exit(&l2arc_dev_mtx); /* * Clear all buflists and ARC references. L2ARC device flush. */ l2arc_evict(remdev, 0, B_TRUE); list_destroy(&remdev->l2ad_buflist); ASSERT(list_is_empty(&remdev->l2ad_lbptr_list)); list_destroy(&remdev->l2ad_lbptr_list); mutex_destroy(&remdev->l2ad_mtx); zfs_refcount_destroy(&remdev->l2ad_alloc); zfs_refcount_destroy(&remdev->l2ad_lb_asize); zfs_refcount_destroy(&remdev->l2ad_lb_count); kmem_free(remdev->l2ad_dev_hdr, remdev->l2ad_dev_hdr_asize); vmem_free(remdev, sizeof (l2arc_dev_t)); } void l2arc_init(void) { l2arc_thread_exit = 0; l2arc_ndev = 0; mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL); mutex_init(&l2arc_rebuild_thr_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&l2arc_rebuild_thr_cv, NULL, CV_DEFAULT, NULL); mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL); mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL); l2arc_dev_list = &L2ARC_dev_list; l2arc_free_on_write = &L2ARC_free_on_write; list_create(l2arc_dev_list, sizeof (l2arc_dev_t), offsetof(l2arc_dev_t, l2ad_node)); list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t), offsetof(l2arc_data_free_t, l2df_list_node)); } void l2arc_fini(void) { mutex_destroy(&l2arc_feed_thr_lock); cv_destroy(&l2arc_feed_thr_cv); mutex_destroy(&l2arc_rebuild_thr_lock); cv_destroy(&l2arc_rebuild_thr_cv); mutex_destroy(&l2arc_dev_mtx); mutex_destroy(&l2arc_free_on_write_mtx); list_destroy(l2arc_dev_list); list_destroy(l2arc_free_on_write); } void l2arc_start(void) { if (!(spa_mode_global & SPA_MODE_WRITE)) return; (void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0, TS_RUN, defclsyspri); } void l2arc_stop(void) { if (!(spa_mode_global & SPA_MODE_WRITE)) return; mutex_enter(&l2arc_feed_thr_lock); cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */ l2arc_thread_exit = 1; while (l2arc_thread_exit != 0) cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock); mutex_exit(&l2arc_feed_thr_lock); } /* * Punches out rebuild threads for the L2ARC devices in a spa. This should * be called after pool import from the spa async thread, since starting * these threads directly from spa_import() will make them part of the * "zpool import" context and delay process exit (and thus pool import). */ void l2arc_spa_rebuild_start(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); /* * Locate the spa's l2arc devices and kick off rebuild threads. */ for (int i = 0; i < spa->spa_l2cache.sav_count; i++) { l2arc_dev_t *dev = l2arc_vdev_get(spa->spa_l2cache.sav_vdevs[i]); if (dev == NULL) { /* Don't attempt a rebuild if the vdev is UNAVAIL */ continue; } mutex_enter(&l2arc_rebuild_thr_lock); if (dev->l2ad_rebuild && !dev->l2ad_rebuild_cancel) { dev->l2ad_rebuild_began = B_TRUE; (void) thread_create(NULL, 0, l2arc_dev_rebuild_thread, dev, 0, &p0, TS_RUN, minclsyspri); } mutex_exit(&l2arc_rebuild_thr_lock); } } /* * Main entry point for L2ARC rebuilding. */ static __attribute__((noreturn)) void l2arc_dev_rebuild_thread(void *arg) { l2arc_dev_t *dev = arg; VERIFY(!dev->l2ad_rebuild_cancel); VERIFY(dev->l2ad_rebuild); (void) l2arc_rebuild(dev); mutex_enter(&l2arc_rebuild_thr_lock); dev->l2ad_rebuild_began = B_FALSE; dev->l2ad_rebuild = B_FALSE; mutex_exit(&l2arc_rebuild_thr_lock); thread_exit(); } /* * This function implements the actual L2ARC metadata rebuild. It: * starts reading the log block chain and restores each block's contents * to memory (reconstructing arc_buf_hdr_t's). * * Operation stops under any of the following conditions: * * 1) We reach the end of the log block chain. * 2) We encounter *any* error condition (cksum errors, io errors) */ static int l2arc_rebuild(l2arc_dev_t *dev) { vdev_t *vd = dev->l2ad_vdev; spa_t *spa = vd->vdev_spa; int err = 0; l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr; l2arc_log_blk_phys_t *this_lb, *next_lb; zio_t *this_io = NULL, *next_io = NULL; l2arc_log_blkptr_t lbps[2]; l2arc_lb_ptr_buf_t *lb_ptr_buf; boolean_t lock_held; this_lb = vmem_zalloc(sizeof (*this_lb), KM_SLEEP); next_lb = vmem_zalloc(sizeof (*next_lb), KM_SLEEP); /* * We prevent device removal while issuing reads to the device, * then during the rebuilding phases we drop this lock again so * that a spa_unload or device remove can be initiated - this is * safe, because the spa will signal us to stop before removing * our device and wait for us to stop. */ spa_config_enter(spa, SCL_L2ARC, vd, RW_READER); lock_held = B_TRUE; /* * Retrieve the persistent L2ARC device state. * L2BLK_GET_PSIZE returns aligned size for log blocks. */ dev->l2ad_evict = MAX(l2dhdr->dh_evict, dev->l2ad_start); dev->l2ad_hand = MAX(l2dhdr->dh_start_lbps[0].lbp_daddr + L2BLK_GET_PSIZE((&l2dhdr->dh_start_lbps[0])->lbp_prop), dev->l2ad_start); dev->l2ad_first = !!(l2dhdr->dh_flags & L2ARC_DEV_HDR_EVICT_FIRST); vd->vdev_trim_action_time = l2dhdr->dh_trim_action_time; vd->vdev_trim_state = l2dhdr->dh_trim_state; /* * In case the zfs module parameter l2arc_rebuild_enabled is false * we do not start the rebuild process. */ if (!l2arc_rebuild_enabled) goto out; /* Prepare the rebuild process */ memcpy(lbps, l2dhdr->dh_start_lbps, sizeof (lbps)); /* Start the rebuild process */ for (;;) { if (!l2arc_log_blkptr_valid(dev, &lbps[0])) break; if ((err = l2arc_log_blk_read(dev, &lbps[0], &lbps[1], this_lb, next_lb, this_io, &next_io)) != 0) goto out; /* * Our memory pressure valve. If the system is running low * on memory, rather than swamping memory with new ARC buf * hdrs, we opt not to rebuild the L2ARC. At this point, * however, we have already set up our L2ARC dev to chain in * new metadata log blocks, so the user may choose to offline/ * online the L2ARC dev at a later time (or re-import the pool) * to reconstruct it (when there's less memory pressure). */ if (l2arc_hdr_limit_reached()) { ARCSTAT_BUMP(arcstat_l2_rebuild_abort_lowmem); cmn_err(CE_NOTE, "System running low on memory, " "aborting L2ARC rebuild."); err = SET_ERROR(ENOMEM); goto out; } spa_config_exit(spa, SCL_L2ARC, vd); lock_held = B_FALSE; /* * Now that we know that the next_lb checks out alright, we * can start reconstruction from this log block. * L2BLK_GET_PSIZE returns aligned size for log blocks. */ uint64_t asize = L2BLK_GET_PSIZE((&lbps[0])->lbp_prop); l2arc_log_blk_restore(dev, this_lb, asize); /* * log block restored, include its pointer in the list of * pointers to log blocks present in the L2ARC device. */ lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP); lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t), KM_SLEEP); memcpy(lb_ptr_buf->lb_ptr, &lbps[0], sizeof (l2arc_log_blkptr_t)); mutex_enter(&dev->l2ad_mtx); list_insert_tail(&dev->l2ad_lbptr_list, lb_ptr_buf); ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize); ARCSTAT_BUMP(arcstat_l2_log_blk_count); zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf); zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf); mutex_exit(&dev->l2ad_mtx); vdev_space_update(vd, asize, 0, 0); /* * Protection against loops of log blocks: * * l2ad_hand l2ad_evict * V V * l2ad_start |=======================================| l2ad_end * -----|||----|||---|||----||| * (3) (2) (1) (0) * ---|||---|||----|||---||| * (7) (6) (5) (4) * * In this situation the pointer of log block (4) passes * l2arc_log_blkptr_valid() but the log block should not be * restored as it is overwritten by the payload of log block * (0). Only log blocks (0)-(3) should be restored. We check * whether l2ad_evict lies in between the payload starting * offset of the next log block (lbps[1].lbp_payload_start) * and the payload starting offset of the present log block * (lbps[0].lbp_payload_start). If true and this isn't the * first pass, we are looping from the beginning and we should * stop. */ if (l2arc_range_check_overlap(lbps[1].lbp_payload_start, lbps[0].lbp_payload_start, dev->l2ad_evict) && !dev->l2ad_first) goto out; kpreempt(KPREEMPT_SYNC); for (;;) { mutex_enter(&l2arc_rebuild_thr_lock); if (dev->l2ad_rebuild_cancel) { dev->l2ad_rebuild = B_FALSE; cv_signal(&l2arc_rebuild_thr_cv); mutex_exit(&l2arc_rebuild_thr_lock); err = SET_ERROR(ECANCELED); goto out; } mutex_exit(&l2arc_rebuild_thr_lock); if (spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER)) { lock_held = B_TRUE; break; } /* * L2ARC config lock held by somebody in writer, * possibly due to them trying to remove us. They'll * likely to want us to shut down, so after a little * delay, we check l2ad_rebuild_cancel and retry * the lock again. */ delay(1); } /* * Continue with the next log block. */ lbps[0] = lbps[1]; lbps[1] = this_lb->lb_prev_lbp; PTR_SWAP(this_lb, next_lb); this_io = next_io; next_io = NULL; } if (this_io != NULL) l2arc_log_blk_fetch_abort(this_io); out: if (next_io != NULL) l2arc_log_blk_fetch_abort(next_io); vmem_free(this_lb, sizeof (*this_lb)); vmem_free(next_lb, sizeof (*next_lb)); if (!l2arc_rebuild_enabled) { spa_history_log_internal(spa, "L2ARC rebuild", NULL, "disabled"); } else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) > 0) { ARCSTAT_BUMP(arcstat_l2_rebuild_success); spa_history_log_internal(spa, "L2ARC rebuild", NULL, "successful, restored %llu blocks", (u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count)); } else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) == 0) { /* * No error but also nothing restored, meaning the lbps array * in the device header points to invalid/non-present log * blocks. Reset the header. */ spa_history_log_internal(spa, "L2ARC rebuild", NULL, "no valid log blocks"); memset(l2dhdr, 0, dev->l2ad_dev_hdr_asize); l2arc_dev_hdr_update(dev); } else if (err == ECANCELED) { /* * In case the rebuild was canceled do not log to spa history * log as the pool may be in the process of being removed. */ zfs_dbgmsg("L2ARC rebuild aborted, restored %llu blocks", (u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count)); } else if (err != 0) { spa_history_log_internal(spa, "L2ARC rebuild", NULL, "aborted, restored %llu blocks", (u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count)); } if (lock_held) spa_config_exit(spa, SCL_L2ARC, vd); return (err); } /* * Attempts to read the device header on the provided L2ARC device and writes * it to `hdr'. On success, this function returns 0, otherwise the appropriate * error code is returned. */ static int l2arc_dev_hdr_read(l2arc_dev_t *dev) { int err; uint64_t guid; l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr; const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize; abd_t *abd; guid = spa_guid(dev->l2ad_vdev->vdev_spa); abd = abd_get_from_buf(l2dhdr, l2dhdr_asize); err = zio_wait(zio_read_phys(NULL, dev->l2ad_vdev, VDEV_LABEL_START_SIZE, l2dhdr_asize, abd, ZIO_CHECKSUM_LABEL, NULL, NULL, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_SPECULATIVE, B_FALSE)); abd_free(abd); if (err != 0) { ARCSTAT_BUMP(arcstat_l2_rebuild_abort_dh_errors); zfs_dbgmsg("L2ARC IO error (%d) while reading device header, " "vdev guid: %llu", err, (u_longlong_t)dev->l2ad_vdev->vdev_guid); return (err); } if (l2dhdr->dh_magic == BSWAP_64(L2ARC_DEV_HDR_MAGIC)) byteswap_uint64_array(l2dhdr, sizeof (*l2dhdr)); if (l2dhdr->dh_magic != L2ARC_DEV_HDR_MAGIC || l2dhdr->dh_spa_guid != guid || l2dhdr->dh_vdev_guid != dev->l2ad_vdev->vdev_guid || l2dhdr->dh_version != L2ARC_PERSISTENT_VERSION || l2dhdr->dh_log_entries != dev->l2ad_log_entries || l2dhdr->dh_end != dev->l2ad_end || !l2arc_range_check_overlap(dev->l2ad_start, dev->l2ad_end, l2dhdr->dh_evict) || (l2dhdr->dh_trim_state != VDEV_TRIM_COMPLETE && l2arc_trim_ahead > 0)) { /* * Attempt to rebuild a device containing no actual dev hdr * or containing a header from some other pool or from another * version of persistent L2ARC. */ ARCSTAT_BUMP(arcstat_l2_rebuild_abort_unsupported); return (SET_ERROR(ENOTSUP)); } return (0); } /* * Reads L2ARC log blocks from storage and validates their contents. * * This function implements a simple fetcher to make sure that while * we're processing one buffer the L2ARC is already fetching the next * one in the chain. * * The arguments this_lp and next_lp point to the current and next log block * address in the block chain. Similarly, this_lb and next_lb hold the * l2arc_log_blk_phys_t's of the current and next L2ARC blk. * * The `this_io' and `next_io' arguments are used for block fetching. * When issuing the first blk IO during rebuild, you should pass NULL for * `this_io'. This function will then issue a sync IO to read the block and * also issue an async IO to fetch the next block in the block chain. The * fetched IO is returned in `next_io'. On subsequent calls to this * function, pass the value returned in `next_io' from the previous call * as `this_io' and a fresh `next_io' pointer to hold the next fetch IO. * Prior to the call, you should initialize your `next_io' pointer to be * NULL. If no fetch IO was issued, the pointer is left set at NULL. * * On success, this function returns 0, otherwise it returns an appropriate * error code. On error the fetching IO is aborted and cleared before * returning from this function. Therefore, if we return `success', the * caller can assume that we have taken care of cleanup of fetch IOs. */ static int l2arc_log_blk_read(l2arc_dev_t *dev, const l2arc_log_blkptr_t *this_lbp, const l2arc_log_blkptr_t *next_lbp, l2arc_log_blk_phys_t *this_lb, l2arc_log_blk_phys_t *next_lb, zio_t *this_io, zio_t **next_io) { int err = 0; zio_cksum_t cksum; abd_t *abd = NULL; uint64_t asize; ASSERT(this_lbp != NULL && next_lbp != NULL); ASSERT(this_lb != NULL && next_lb != NULL); ASSERT(next_io != NULL && *next_io == NULL); ASSERT(l2arc_log_blkptr_valid(dev, this_lbp)); /* * Check to see if we have issued the IO for this log block in a * previous run. If not, this is the first call, so issue it now. */ if (this_io == NULL) { this_io = l2arc_log_blk_fetch(dev->l2ad_vdev, this_lbp, this_lb); } /* * Peek to see if we can start issuing the next IO immediately. */ if (l2arc_log_blkptr_valid(dev, next_lbp)) { /* * Start issuing IO for the next log block early - this * should help keep the L2ARC device busy while we * decompress and restore this log block. */ *next_io = l2arc_log_blk_fetch(dev->l2ad_vdev, next_lbp, next_lb); } /* Wait for the IO to read this log block to complete */ if ((err = zio_wait(this_io)) != 0) { ARCSTAT_BUMP(arcstat_l2_rebuild_abort_io_errors); zfs_dbgmsg("L2ARC IO error (%d) while reading log block, " "offset: %llu, vdev guid: %llu", err, (u_longlong_t)this_lbp->lbp_daddr, (u_longlong_t)dev->l2ad_vdev->vdev_guid); goto cleanup; } /* * Make sure the buffer checks out. * L2BLK_GET_PSIZE returns aligned size for log blocks. */ asize = L2BLK_GET_PSIZE((this_lbp)->lbp_prop); fletcher_4_native(this_lb, asize, NULL, &cksum); if (!ZIO_CHECKSUM_EQUAL(cksum, this_lbp->lbp_cksum)) { ARCSTAT_BUMP(arcstat_l2_rebuild_abort_cksum_lb_errors); zfs_dbgmsg("L2ARC log block cksum failed, offset: %llu, " "vdev guid: %llu, l2ad_hand: %llu, l2ad_evict: %llu", (u_longlong_t)this_lbp->lbp_daddr, (u_longlong_t)dev->l2ad_vdev->vdev_guid, (u_longlong_t)dev->l2ad_hand, (u_longlong_t)dev->l2ad_evict); err = SET_ERROR(ECKSUM); goto cleanup; } /* Now we can take our time decoding this buffer */ switch (L2BLK_GET_COMPRESS((this_lbp)->lbp_prop)) { case ZIO_COMPRESS_OFF: break; case ZIO_COMPRESS_LZ4: abd = abd_alloc_for_io(asize, B_TRUE); abd_copy_from_buf_off(abd, this_lb, 0, asize); if ((err = zio_decompress_data( L2BLK_GET_COMPRESS((this_lbp)->lbp_prop), abd, this_lb, asize, sizeof (*this_lb), NULL)) != 0) { err = SET_ERROR(EINVAL); goto cleanup; } break; default: err = SET_ERROR(EINVAL); goto cleanup; } if (this_lb->lb_magic == BSWAP_64(L2ARC_LOG_BLK_MAGIC)) byteswap_uint64_array(this_lb, sizeof (*this_lb)); if (this_lb->lb_magic != L2ARC_LOG_BLK_MAGIC) { err = SET_ERROR(EINVAL); goto cleanup; } cleanup: /* Abort an in-flight fetch I/O in case of error */ if (err != 0 && *next_io != NULL) { l2arc_log_blk_fetch_abort(*next_io); *next_io = NULL; } if (abd != NULL) abd_free(abd); return (err); } /* * Restores the payload of a log block to ARC. This creates empty ARC hdr * entries which only contain an l2arc hdr, essentially restoring the * buffers to their L2ARC evicted state. This function also updates space * usage on the L2ARC vdev to make sure it tracks restored buffers. */ static void l2arc_log_blk_restore(l2arc_dev_t *dev, const l2arc_log_blk_phys_t *lb, uint64_t lb_asize) { uint64_t size = 0, asize = 0; uint64_t log_entries = dev->l2ad_log_entries; /* * Usually arc_adapt() is called only for data, not headers, but * since we may allocate significant amount of memory here, let ARC * grow its arc_c. */ arc_adapt(log_entries * HDR_L2ONLY_SIZE); for (int i = log_entries - 1; i >= 0; i--) { /* * Restore goes in the reverse temporal direction to preserve * correct temporal ordering of buffers in the l2ad_buflist. * l2arc_hdr_restore also does a list_insert_tail instead of * list_insert_head on the l2ad_buflist: * * LIST l2ad_buflist LIST * HEAD <------ (time) ------ TAIL * direction +-----+-----+-----+-----+-----+ direction * of l2arc <== | buf | buf | buf | buf | buf | ===> of rebuild * fill +-----+-----+-----+-----+-----+ * ^ ^ * | | * | | * l2arc_feed_thread l2arc_rebuild * will place new bufs here restores bufs here * * During l2arc_rebuild() the device is not used by * l2arc_feed_thread() as dev->l2ad_rebuild is set to true. */ size += L2BLK_GET_LSIZE((&lb->lb_entries[i])->le_prop); asize += vdev_psize_to_asize(dev->l2ad_vdev, L2BLK_GET_PSIZE((&lb->lb_entries[i])->le_prop)); l2arc_hdr_restore(&lb->lb_entries[i], dev); } /* * Record rebuild stats: * size Logical size of restored buffers in the L2ARC * asize Aligned size of restored buffers in the L2ARC */ ARCSTAT_INCR(arcstat_l2_rebuild_size, size); ARCSTAT_INCR(arcstat_l2_rebuild_asize, asize); ARCSTAT_INCR(arcstat_l2_rebuild_bufs, log_entries); ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, lb_asize); ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio, asize / lb_asize); ARCSTAT_BUMP(arcstat_l2_rebuild_log_blks); } /* * Restores a single ARC buf hdr from a log entry. The ARC buffer is put * into a state indicating that it has been evicted to L2ARC. */ static void l2arc_hdr_restore(const l2arc_log_ent_phys_t *le, l2arc_dev_t *dev) { arc_buf_hdr_t *hdr, *exists; kmutex_t *hash_lock; arc_buf_contents_t type = L2BLK_GET_TYPE((le)->le_prop); uint64_t asize; /* * Do all the allocation before grabbing any locks, this lets us * sleep if memory is full and we don't have to deal with failed * allocations. */ hdr = arc_buf_alloc_l2only(L2BLK_GET_LSIZE((le)->le_prop), type, dev, le->le_dva, le->le_daddr, L2BLK_GET_PSIZE((le)->le_prop), le->le_birth, L2BLK_GET_COMPRESS((le)->le_prop), le->le_complevel, L2BLK_GET_PROTECTED((le)->le_prop), L2BLK_GET_PREFETCH((le)->le_prop), L2BLK_GET_STATE((le)->le_prop)); asize = vdev_psize_to_asize(dev->l2ad_vdev, L2BLK_GET_PSIZE((le)->le_prop)); /* * vdev_space_update() has to be called before arc_hdr_destroy() to * avoid underflow since the latter also calls vdev_space_update(). */ l2arc_hdr_arcstats_increment(hdr); vdev_space_update(dev->l2ad_vdev, asize, 0, 0); mutex_enter(&dev->l2ad_mtx); list_insert_tail(&dev->l2ad_buflist, hdr); (void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); mutex_exit(&dev->l2ad_mtx); exists = buf_hash_insert(hdr, &hash_lock); if (exists) { /* Buffer was already cached, no need to restore it. */ arc_hdr_destroy(hdr); /* * If the buffer is already cached, check whether it has * L2ARC metadata. If not, enter them and update the flag. * This is important is case of onlining a cache device, since * we previously evicted all L2ARC metadata from ARC. */ if (!HDR_HAS_L2HDR(exists)) { arc_hdr_set_flags(exists, ARC_FLAG_HAS_L2HDR); exists->b_l2hdr.b_dev = dev; exists->b_l2hdr.b_daddr = le->le_daddr; exists->b_l2hdr.b_arcs_state = L2BLK_GET_STATE((le)->le_prop); mutex_enter(&dev->l2ad_mtx); list_insert_tail(&dev->l2ad_buflist, exists); (void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(exists), exists); mutex_exit(&dev->l2ad_mtx); l2arc_hdr_arcstats_increment(exists); vdev_space_update(dev->l2ad_vdev, asize, 0, 0); } ARCSTAT_BUMP(arcstat_l2_rebuild_bufs_precached); } mutex_exit(hash_lock); } /* * Starts an asynchronous read IO to read a log block. This is used in log * block reconstruction to start reading the next block before we are done * decoding and reconstructing the current block, to keep the l2arc device * nice and hot with read IO to process. * The returned zio will contain a newly allocated memory buffers for the IO * data which should then be freed by the caller once the zio is no longer * needed (i.e. due to it having completed). If you wish to abort this * zio, you should do so using l2arc_log_blk_fetch_abort, which takes * care of disposing of the allocated buffers correctly. */ static zio_t * l2arc_log_blk_fetch(vdev_t *vd, const l2arc_log_blkptr_t *lbp, l2arc_log_blk_phys_t *lb) { uint32_t asize; zio_t *pio; l2arc_read_callback_t *cb; /* L2BLK_GET_PSIZE returns aligned size for log blocks */ asize = L2BLK_GET_PSIZE((lbp)->lbp_prop); ASSERT(asize <= sizeof (l2arc_log_blk_phys_t)); cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP); cb->l2rcb_abd = abd_get_from_buf(lb, asize); pio = zio_root(vd->vdev_spa, l2arc_blk_fetch_done, cb, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY); (void) zio_nowait(zio_read_phys(pio, vd, lbp->lbp_daddr, asize, cb->l2rcb_abd, ZIO_CHECKSUM_OFF, NULL, NULL, ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, B_FALSE)); return (pio); } /* * Aborts a zio returned from l2arc_log_blk_fetch and frees the data * buffers allocated for it. */ static void l2arc_log_blk_fetch_abort(zio_t *zio) { (void) zio_wait(zio); } /* * Creates a zio to update the device header on an l2arc device. */ void l2arc_dev_hdr_update(l2arc_dev_t *dev) { l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr; const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize; abd_t *abd; int err; VERIFY(spa_config_held(dev->l2ad_spa, SCL_STATE_ALL, RW_READER)); l2dhdr->dh_magic = L2ARC_DEV_HDR_MAGIC; l2dhdr->dh_version = L2ARC_PERSISTENT_VERSION; l2dhdr->dh_spa_guid = spa_guid(dev->l2ad_vdev->vdev_spa); l2dhdr->dh_vdev_guid = dev->l2ad_vdev->vdev_guid; l2dhdr->dh_log_entries = dev->l2ad_log_entries; l2dhdr->dh_evict = dev->l2ad_evict; l2dhdr->dh_start = dev->l2ad_start; l2dhdr->dh_end = dev->l2ad_end; l2dhdr->dh_lb_asize = zfs_refcount_count(&dev->l2ad_lb_asize); l2dhdr->dh_lb_count = zfs_refcount_count(&dev->l2ad_lb_count); l2dhdr->dh_flags = 0; l2dhdr->dh_trim_action_time = dev->l2ad_vdev->vdev_trim_action_time; l2dhdr->dh_trim_state = dev->l2ad_vdev->vdev_trim_state; if (dev->l2ad_first) l2dhdr->dh_flags |= L2ARC_DEV_HDR_EVICT_FIRST; abd = abd_get_from_buf(l2dhdr, l2dhdr_asize); err = zio_wait(zio_write_phys(NULL, dev->l2ad_vdev, VDEV_LABEL_START_SIZE, l2dhdr_asize, abd, ZIO_CHECKSUM_LABEL, NULL, NULL, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE)); abd_free(abd); if (err != 0) { zfs_dbgmsg("L2ARC IO error (%d) while writing device header, " "vdev guid: %llu", err, (u_longlong_t)dev->l2ad_vdev->vdev_guid); } } /* * Commits a log block to the L2ARC device. This routine is invoked from * l2arc_write_buffers when the log block fills up. * This function allocates some memory to temporarily hold the serialized * buffer to be written. This is then released in l2arc_write_done. */ static uint64_t l2arc_log_blk_commit(l2arc_dev_t *dev, zio_t *pio, l2arc_write_callback_t *cb) { l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk; l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr; uint64_t psize, asize; zio_t *wzio; l2arc_lb_abd_buf_t *abd_buf; uint8_t *tmpbuf = NULL; l2arc_lb_ptr_buf_t *lb_ptr_buf; VERIFY3S(dev->l2ad_log_ent_idx, ==, dev->l2ad_log_entries); abd_buf = zio_buf_alloc(sizeof (*abd_buf)); abd_buf->abd = abd_get_from_buf(lb, sizeof (*lb)); lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP); lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t), KM_SLEEP); /* link the buffer into the block chain */ lb->lb_prev_lbp = l2dhdr->dh_start_lbps[1]; lb->lb_magic = L2ARC_LOG_BLK_MAGIC; /* * l2arc_log_blk_commit() may be called multiple times during a single * l2arc_write_buffers() call. Save the allocated abd buffers in a list * so we can free them in l2arc_write_done() later on. */ list_insert_tail(&cb->l2wcb_abd_list, abd_buf); /* try to compress the buffer */ psize = zio_compress_data(ZIO_COMPRESS_LZ4, abd_buf->abd, (void **) &tmpbuf, sizeof (*lb), 0); /* a log block is never entirely zero */ ASSERT(psize != 0); asize = vdev_psize_to_asize(dev->l2ad_vdev, psize); ASSERT(asize <= sizeof (*lb)); /* * Update the start log block pointer in the device header to point * to the log block we're about to write. */ l2dhdr->dh_start_lbps[1] = l2dhdr->dh_start_lbps[0]; l2dhdr->dh_start_lbps[0].lbp_daddr = dev->l2ad_hand; l2dhdr->dh_start_lbps[0].lbp_payload_asize = dev->l2ad_log_blk_payload_asize; l2dhdr->dh_start_lbps[0].lbp_payload_start = dev->l2ad_log_blk_payload_start; L2BLK_SET_LSIZE( (&l2dhdr->dh_start_lbps[0])->lbp_prop, sizeof (*lb)); L2BLK_SET_PSIZE( (&l2dhdr->dh_start_lbps[0])->lbp_prop, asize); L2BLK_SET_CHECKSUM( (&l2dhdr->dh_start_lbps[0])->lbp_prop, ZIO_CHECKSUM_FLETCHER_4); if (asize < sizeof (*lb)) { /* compression succeeded */ memset(tmpbuf + psize, 0, asize - psize); L2BLK_SET_COMPRESS( (&l2dhdr->dh_start_lbps[0])->lbp_prop, ZIO_COMPRESS_LZ4); } else { /* compression failed */ memcpy(tmpbuf, lb, sizeof (*lb)); L2BLK_SET_COMPRESS( (&l2dhdr->dh_start_lbps[0])->lbp_prop, ZIO_COMPRESS_OFF); } /* checksum what we're about to write */ fletcher_4_native(tmpbuf, asize, NULL, &l2dhdr->dh_start_lbps[0].lbp_cksum); abd_free(abd_buf->abd); /* perform the write itself */ abd_buf->abd = abd_get_from_buf(tmpbuf, sizeof (*lb)); abd_take_ownership_of_buf(abd_buf->abd, B_TRUE); wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand, asize, abd_buf->abd, ZIO_CHECKSUM_OFF, NULL, NULL, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE); DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev, zio_t *, wzio); (void) zio_nowait(wzio); dev->l2ad_hand += asize; /* * Include the committed log block's pointer in the list of pointers * to log blocks present in the L2ARC device. */ memcpy(lb_ptr_buf->lb_ptr, &l2dhdr->dh_start_lbps[0], sizeof (l2arc_log_blkptr_t)); mutex_enter(&dev->l2ad_mtx); list_insert_head(&dev->l2ad_lbptr_list, lb_ptr_buf); ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize); ARCSTAT_BUMP(arcstat_l2_log_blk_count); zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf); zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf); mutex_exit(&dev->l2ad_mtx); vdev_space_update(dev->l2ad_vdev, asize, 0, 0); /* bump the kstats */ ARCSTAT_INCR(arcstat_l2_write_bytes, asize); ARCSTAT_BUMP(arcstat_l2_log_blk_writes); ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, asize); ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio, dev->l2ad_log_blk_payload_asize / asize); /* start a new log block */ dev->l2ad_log_ent_idx = 0; dev->l2ad_log_blk_payload_asize = 0; dev->l2ad_log_blk_payload_start = 0; return (asize); } /* * Validates an L2ARC log block address to make sure that it can be read * from the provided L2ARC device. */ boolean_t l2arc_log_blkptr_valid(l2arc_dev_t *dev, const l2arc_log_blkptr_t *lbp) { /* L2BLK_GET_PSIZE returns aligned size for log blocks */ uint64_t asize = L2BLK_GET_PSIZE((lbp)->lbp_prop); uint64_t end = lbp->lbp_daddr + asize - 1; uint64_t start = lbp->lbp_payload_start; boolean_t evicted = B_FALSE; /* * A log block is valid if all of the following conditions are true: * - it fits entirely (including its payload) between l2ad_start and * l2ad_end * - it has a valid size * - neither the log block itself nor part of its payload was evicted * by l2arc_evict(): * * l2ad_hand l2ad_evict * | | lbp_daddr * | start | | end * | | | | | * V V V V V * l2ad_start ============================================ l2ad_end * --------------------------|||| * ^ ^ * | log block * payload */ evicted = l2arc_range_check_overlap(start, end, dev->l2ad_hand) || l2arc_range_check_overlap(start, end, dev->l2ad_evict) || l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, start) || l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, end); return (start >= dev->l2ad_start && end <= dev->l2ad_end && asize > 0 && asize <= sizeof (l2arc_log_blk_phys_t) && (!evicted || dev->l2ad_first)); } /* * Inserts ARC buffer header `hdr' into the current L2ARC log block on * the device. The buffer being inserted must be present in L2ARC. * Returns B_TRUE if the L2ARC log block is full and needs to be committed * to L2ARC, or B_FALSE if it still has room for more ARC buffers. */ static boolean_t l2arc_log_blk_insert(l2arc_dev_t *dev, const arc_buf_hdr_t *hdr) { l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk; l2arc_log_ent_phys_t *le; if (dev->l2ad_log_entries == 0) return (B_FALSE); int index = dev->l2ad_log_ent_idx++; ASSERT3S(index, <, dev->l2ad_log_entries); ASSERT(HDR_HAS_L2HDR(hdr)); le = &lb->lb_entries[index]; memset(le, 0, sizeof (*le)); le->le_dva = hdr->b_dva; le->le_birth = hdr->b_birth; le->le_daddr = hdr->b_l2hdr.b_daddr; if (index == 0) dev->l2ad_log_blk_payload_start = le->le_daddr; L2BLK_SET_LSIZE((le)->le_prop, HDR_GET_LSIZE(hdr)); L2BLK_SET_PSIZE((le)->le_prop, HDR_GET_PSIZE(hdr)); L2BLK_SET_COMPRESS((le)->le_prop, HDR_GET_COMPRESS(hdr)); le->le_complevel = hdr->b_complevel; L2BLK_SET_TYPE((le)->le_prop, hdr->b_type); L2BLK_SET_PROTECTED((le)->le_prop, !!(HDR_PROTECTED(hdr))); L2BLK_SET_PREFETCH((le)->le_prop, !!(HDR_PREFETCH(hdr))); L2BLK_SET_STATE((le)->le_prop, hdr->b_l2hdr.b_arcs_state); dev->l2ad_log_blk_payload_asize += vdev_psize_to_asize(dev->l2ad_vdev, HDR_GET_PSIZE(hdr)); return (dev->l2ad_log_ent_idx == dev->l2ad_log_entries); } /* * Checks whether a given L2ARC device address sits in a time-sequential * range. The trick here is that the L2ARC is a rotary buffer, so we can't * just do a range comparison, we need to handle the situation in which the * range wraps around the end of the L2ARC device. Arguments: * bottom -- Lower end of the range to check (written to earlier). * top -- Upper end of the range to check (written to later). * check -- The address for which we want to determine if it sits in * between the top and bottom. * * The 3-way conditional below represents the following cases: * * bottom < top : Sequentially ordered case: * --------+-------------------+ * | (overlap here?) | * L2ARC dev V V * |---------------============--------------| * * bottom > top: Looped-around case: * --------+------------------+ * | (overlap here?) | * L2ARC dev V V * |===============---------------===========| * ^ ^ * | (or here?) | * +---------------+--------- * * top == bottom : Just a single address comparison. */ boolean_t l2arc_range_check_overlap(uint64_t bottom, uint64_t top, uint64_t check) { if (bottom < top) return (bottom <= check && check <= top); else if (bottom > top) return (check <= top || bottom <= check); else return (check == top); } EXPORT_SYMBOL(arc_buf_size); EXPORT_SYMBOL(arc_write); EXPORT_SYMBOL(arc_read); EXPORT_SYMBOL(arc_buf_info); EXPORT_SYMBOL(arc_getbuf_func); EXPORT_SYMBOL(arc_add_prune_callback); EXPORT_SYMBOL(arc_remove_prune_callback); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min, param_set_arc_min, spl_param_get_u64, ZMOD_RW, "Minimum ARC size in bytes"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, max, param_set_arc_max, spl_param_get_u64, ZMOD_RW, "Maximum ARC size in bytes"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_balance, UINT, ZMOD_RW, "Balance between metadata and data on ghost hits."); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, grow_retry, param_set_arc_int, param_get_uint, ZMOD_RW, "Seconds before growing ARC size"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, shrink_shift, param_set_arc_int, param_get_uint, ZMOD_RW, "log2(fraction of ARC to reclaim)"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, pc_percent, UINT, ZMOD_RW, "Percent of pagecache to reclaim ARC to"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, average_blocksize, UINT, ZMOD_RD, "Target average block size"); ZFS_MODULE_PARAM(zfs, zfs_, compressed_arc_enabled, INT, ZMOD_RW, "Disable compressed ARC buffers"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prefetch_ms, param_set_arc_int, param_get_uint, ZMOD_RW, "Min life of prefetch block in ms"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prescient_prefetch_ms, param_set_arc_int, param_get_uint, ZMOD_RW, "Min life of prescient prefetched block in ms"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_max, U64, ZMOD_RW, "Max write bytes per interval"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_boost, U64, ZMOD_RW, "Extra write bytes during device warmup"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom, U64, ZMOD_RW, "Number of max device writes to precache"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom_boost, U64, ZMOD_RW, "Compressed l2arc_headroom multiplier"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, trim_ahead, U64, ZMOD_RW, "TRIM ahead L2ARC write size multiplier"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_secs, U64, ZMOD_RW, "Seconds between L2ARC writing"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_min_ms, U64, ZMOD_RW, "Min feed interval in milliseconds"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, noprefetch, INT, ZMOD_RW, "Skip caching prefetched buffers"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_again, INT, ZMOD_RW, "Turbo L2ARC warmup"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, norw, INT, ZMOD_RW, "No reads during writes"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, meta_percent, UINT, ZMOD_RW, "Percent of ARC size allowed for L2ARC-only headers"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_enabled, INT, ZMOD_RW, "Rebuild the L2ARC when importing a pool"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_blocks_min_l2size, U64, ZMOD_RW, "Min size in bytes to write rebuild log blocks in L2ARC"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, mfuonly, INT, ZMOD_RW, "Cache only MFU data from ARC into L2ARC"); ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, exclude_special, INT, ZMOD_RW, "Exclude dbufs on special vdevs from being cached to L2ARC if set."); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, lotsfree_percent, param_set_arc_int, param_get_uint, ZMOD_RW, "System free memory I/O throttle in bytes"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, sys_free, param_set_arc_u64, spl_param_get_u64, ZMOD_RW, "System free memory target size in bytes"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit, param_set_arc_u64, spl_param_get_u64, ZMOD_RW, "Minimum bytes of dnodes in ARC"); ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit_percent, param_set_arc_int, param_get_uint, ZMOD_RW, "Percent of ARC meta buffers for dnodes"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, dnode_reduce_percent, UINT, ZMOD_RW, "Percentage of excess dnodes to try to unpin"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, eviction_pct, UINT, ZMOD_RW, "When full, ARC allocation waits for eviction of this % of alloc size"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, evict_batch_limit, UINT, ZMOD_RW, "The number of headers to evict per sublist before moving to the next"); ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, prune_task_threads, INT, ZMOD_RW, "Number of arc_prune threads"); diff --git a/sys/contrib/openzfs/module/zfs/dbuf.c b/sys/contrib/openzfs/module/zfs/dbuf.c index 5f3643f573f7..bb913f556374 100644 --- a/sys/contrib/openzfs/module/zfs/dbuf.c +++ b/sys/contrib/openzfs/module/zfs/dbuf.c @@ -1,5260 +1,5254 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, 2020 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright (c) 2019, Klara Inc. * Copyright (c) 2019, Allan Jude * Copyright (c) 2021, 2022 by Pawel Jakub Dawidek */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static kstat_t *dbuf_ksp; typedef struct dbuf_stats { /* * Various statistics about the size of the dbuf cache. */ kstat_named_t cache_count; kstat_named_t cache_size_bytes; kstat_named_t cache_size_bytes_max; /* * Statistics regarding the bounds on the dbuf cache size. */ kstat_named_t cache_target_bytes; kstat_named_t cache_lowater_bytes; kstat_named_t cache_hiwater_bytes; /* * Total number of dbuf cache evictions that have occurred. */ kstat_named_t cache_total_evicts; /* * The distribution of dbuf levels in the dbuf cache and * the total size of all dbufs at each level. */ kstat_named_t cache_levels[DN_MAX_LEVELS]; kstat_named_t cache_levels_bytes[DN_MAX_LEVELS]; /* * Statistics about the dbuf hash table. */ kstat_named_t hash_hits; kstat_named_t hash_misses; kstat_named_t hash_collisions; kstat_named_t hash_elements; kstat_named_t hash_elements_max; /* * Number of sublists containing more than one dbuf in the dbuf * hash table. Keep track of the longest hash chain. */ kstat_named_t hash_chains; kstat_named_t hash_chain_max; /* * Number of times a dbuf_create() discovers that a dbuf was * already created and in the dbuf hash table. */ kstat_named_t hash_insert_race; /* * Number of entries in the hash table dbuf and mutex arrays. */ kstat_named_t hash_table_count; kstat_named_t hash_mutex_count; /* * Statistics about the size of the metadata dbuf cache. */ kstat_named_t metadata_cache_count; kstat_named_t metadata_cache_size_bytes; kstat_named_t metadata_cache_size_bytes_max; /* * For diagnostic purposes, this is incremented whenever we can't add * something to the metadata cache because it's full, and instead put * the data in the regular dbuf cache. */ kstat_named_t metadata_cache_overflow; } dbuf_stats_t; dbuf_stats_t dbuf_stats = { { "cache_count", KSTAT_DATA_UINT64 }, { "cache_size_bytes", KSTAT_DATA_UINT64 }, { "cache_size_bytes_max", KSTAT_DATA_UINT64 }, { "cache_target_bytes", KSTAT_DATA_UINT64 }, { "cache_lowater_bytes", KSTAT_DATA_UINT64 }, { "cache_hiwater_bytes", KSTAT_DATA_UINT64 }, { "cache_total_evicts", KSTAT_DATA_UINT64 }, { { "cache_levels_N", KSTAT_DATA_UINT64 } }, { { "cache_levels_bytes_N", KSTAT_DATA_UINT64 } }, { "hash_hits", KSTAT_DATA_UINT64 }, { "hash_misses", KSTAT_DATA_UINT64 }, { "hash_collisions", KSTAT_DATA_UINT64 }, { "hash_elements", KSTAT_DATA_UINT64 }, { "hash_elements_max", KSTAT_DATA_UINT64 }, { "hash_chains", KSTAT_DATA_UINT64 }, { "hash_chain_max", KSTAT_DATA_UINT64 }, { "hash_insert_race", KSTAT_DATA_UINT64 }, { "hash_table_count", KSTAT_DATA_UINT64 }, { "hash_mutex_count", KSTAT_DATA_UINT64 }, { "metadata_cache_count", KSTAT_DATA_UINT64 }, { "metadata_cache_size_bytes", KSTAT_DATA_UINT64 }, { "metadata_cache_size_bytes_max", KSTAT_DATA_UINT64 }, { "metadata_cache_overflow", KSTAT_DATA_UINT64 } }; struct { wmsum_t cache_count; wmsum_t cache_total_evicts; wmsum_t cache_levels[DN_MAX_LEVELS]; wmsum_t cache_levels_bytes[DN_MAX_LEVELS]; wmsum_t hash_hits; wmsum_t hash_misses; wmsum_t hash_collisions; wmsum_t hash_chains; wmsum_t hash_insert_race; wmsum_t metadata_cache_count; wmsum_t metadata_cache_overflow; } dbuf_sums; #define DBUF_STAT_INCR(stat, val) \ - wmsum_add(&dbuf_sums.stat, val); + wmsum_add(&dbuf_sums.stat, val) #define DBUF_STAT_DECR(stat, val) \ - DBUF_STAT_INCR(stat, -(val)); + DBUF_STAT_INCR(stat, -(val)) #define DBUF_STAT_BUMP(stat) \ - DBUF_STAT_INCR(stat, 1); + DBUF_STAT_INCR(stat, 1) #define DBUF_STAT_BUMPDOWN(stat) \ - DBUF_STAT_INCR(stat, -1); + DBUF_STAT_INCR(stat, -1) #define DBUF_STAT_MAX(stat, v) { \ uint64_t _m; \ while ((v) > (_m = dbuf_stats.stat.value.ui64) && \ (_m != atomic_cas_64(&dbuf_stats.stat.value.ui64, _m, (v))))\ continue; \ } static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx); static void dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr); -static int dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags); /* * Global data structures and functions for the dbuf cache. */ static kmem_cache_t *dbuf_kmem_cache; static taskq_t *dbu_evict_taskq; static kthread_t *dbuf_cache_evict_thread; static kmutex_t dbuf_evict_lock; static kcondvar_t dbuf_evict_cv; static boolean_t dbuf_evict_thread_exit; /* * There are two dbuf caches; each dbuf can only be in one of them at a time. * * 1. Cache of metadata dbufs, to help make read-heavy administrative commands * from /sbin/zfs run faster. The "metadata cache" specifically stores dbufs * that represent the metadata that describes filesystems/snapshots/ * bookmarks/properties/etc. We only evict from this cache when we export a * pool, to short-circuit as much I/O as possible for all administrative * commands that need the metadata. There is no eviction policy for this * cache, because we try to only include types in it which would occupy a * very small amount of space per object but create a large impact on the * performance of these commands. Instead, after it reaches a maximum size * (which should only happen on very small memory systems with a very large * number of filesystem objects), we stop taking new dbufs into the * metadata cache, instead putting them in the normal dbuf cache. * * 2. LRU cache of dbufs. The dbuf cache maintains a list of dbufs that * are not currently held but have been recently released. These dbufs * are not eligible for arc eviction until they are aged out of the cache. * Dbufs that are aged out of the cache will be immediately destroyed and * become eligible for arc eviction. * * Dbufs are added to these caches once the last hold is released. If a dbuf is * later accessed and still exists in the dbuf cache, then it will be removed * from the cache and later re-added to the head of the cache. * * If a given dbuf meets the requirements for the metadata cache, it will go * there, otherwise it will be considered for the generic LRU dbuf cache. The * caches and the refcounts tracking their sizes are stored in an array indexed * by those caches' matching enum values (from dbuf_cached_state_t). */ typedef struct dbuf_cache { multilist_t cache; zfs_refcount_t size ____cacheline_aligned; } dbuf_cache_t; dbuf_cache_t dbuf_caches[DB_CACHE_MAX]; /* Size limits for the caches */ static uint64_t dbuf_cache_max_bytes = UINT64_MAX; static uint64_t dbuf_metadata_cache_max_bytes = UINT64_MAX; /* Set the default sizes of the caches to log2 fraction of arc size */ static uint_t dbuf_cache_shift = 5; static uint_t dbuf_metadata_cache_shift = 6; /* Set the dbuf hash mutex count as log2 shift (dynamic by default) */ static uint_t dbuf_mutex_cache_shift = 0; static unsigned long dbuf_cache_target_bytes(void); static unsigned long dbuf_metadata_cache_target_bytes(void); /* * The LRU dbuf cache uses a three-stage eviction policy: * - A low water marker designates when the dbuf eviction thread * should stop evicting from the dbuf cache. * - When we reach the maximum size (aka mid water mark), we * signal the eviction thread to run. * - The high water mark indicates when the eviction thread * is unable to keep up with the incoming load and eviction must * happen in the context of the calling thread. * * The dbuf cache: * (max size) * low water mid water hi water * +----------------------------------------+----------+----------+ * | | | | * | | | | * | | | | * | | | | * +----------------------------------------+----------+----------+ * stop signal evict * evicting eviction directly * thread * * The high and low water marks indicate the operating range for the eviction * thread. The low water mark is, by default, 90% of the total size of the * cache and the high water mark is at 110% (both of these percentages can be * changed by setting dbuf_cache_lowater_pct and dbuf_cache_hiwater_pct, * respectively). The eviction thread will try to ensure that the cache remains * within this range by waking up every second and checking if the cache is * above the low water mark. The thread can also be woken up by callers adding * elements into the cache if the cache is larger than the mid water (i.e max * cache size). Once the eviction thread is woken up and eviction is required, * it will continue evicting buffers until it's able to reduce the cache size * to the low water mark. If the cache size continues to grow and hits the high * water mark, then callers adding elements to the cache will begin to evict * directly from the cache until the cache is no longer above the high water * mark. */ /* * The percentage above and below the maximum cache size. */ static uint_t dbuf_cache_hiwater_pct = 10; static uint_t dbuf_cache_lowater_pct = 10; static int dbuf_cons(void *vdb, void *unused, int kmflag) { (void) unused, (void) kmflag; dmu_buf_impl_t *db = vdb; memset(db, 0, sizeof (dmu_buf_impl_t)); mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL); rw_init(&db->db_rwlock, NULL, RW_DEFAULT, NULL); cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL); multilist_link_init(&db->db_cache_link); zfs_refcount_create(&db->db_holds); return (0); } static void dbuf_dest(void *vdb, void *unused) { (void) unused; dmu_buf_impl_t *db = vdb; mutex_destroy(&db->db_mtx); rw_destroy(&db->db_rwlock); cv_destroy(&db->db_changed); ASSERT(!multilist_link_active(&db->db_cache_link)); zfs_refcount_destroy(&db->db_holds); } /* * dbuf hash table routines */ static dbuf_hash_table_t dbuf_hash_table; /* * We use Cityhash for this. It's fast, and has good hash properties without * requiring any large static buffers. */ static uint64_t dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid) { return (cityhash4((uintptr_t)os, obj, (uint64_t)lvl, blkid)); } #define DTRACE_SET_STATE(db, why) \ DTRACE_PROBE2(dbuf__state_change, dmu_buf_impl_t *, db, \ const char *, why) #define DBUF_EQUAL(dbuf, os, obj, level, blkid) \ ((dbuf)->db.db_object == (obj) && \ (dbuf)->db_objset == (os) && \ (dbuf)->db_level == (level) && \ (dbuf)->db_blkid == (blkid)) dmu_buf_impl_t * dbuf_find(objset_t *os, uint64_t obj, uint8_t level, uint64_t blkid, uint64_t *hash_out) { dbuf_hash_table_t *h = &dbuf_hash_table; uint64_t hv; uint64_t idx; dmu_buf_impl_t *db; hv = dbuf_hash(os, obj, level, blkid); idx = hv & h->hash_table_mask; mutex_enter(DBUF_HASH_MUTEX(h, idx)); for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) { if (DBUF_EQUAL(db, os, obj, level, blkid)) { mutex_enter(&db->db_mtx); if (db->db_state != DB_EVICTING) { mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (db); } mutex_exit(&db->db_mtx); } } mutex_exit(DBUF_HASH_MUTEX(h, idx)); if (hash_out != NULL) *hash_out = hv; return (NULL); } static dmu_buf_impl_t * dbuf_find_bonus(objset_t *os, uint64_t object) { dnode_t *dn; dmu_buf_impl_t *db = NULL; if (dnode_hold(os, object, FTAG, &dn) == 0) { rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_bonus != NULL) { db = dn->dn_bonus; mutex_enter(&db->db_mtx); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); } return (db); } /* * Insert an entry into the hash table. If there is already an element * equal to elem in the hash table, then the already existing element * will be returned and the new element will not be inserted. * Otherwise returns NULL. */ static dmu_buf_impl_t * dbuf_hash_insert(dmu_buf_impl_t *db) { dbuf_hash_table_t *h = &dbuf_hash_table; objset_t *os = db->db_objset; uint64_t obj = db->db.db_object; int level = db->db_level; uint64_t blkid, idx; dmu_buf_impl_t *dbf; uint32_t i; blkid = db->db_blkid; ASSERT3U(dbuf_hash(os, obj, level, blkid), ==, db->db_hash); idx = db->db_hash & h->hash_table_mask; mutex_enter(DBUF_HASH_MUTEX(h, idx)); for (dbf = h->hash_table[idx], i = 0; dbf != NULL; dbf = dbf->db_hash_next, i++) { if (DBUF_EQUAL(dbf, os, obj, level, blkid)) { mutex_enter(&dbf->db_mtx); if (dbf->db_state != DB_EVICTING) { mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (dbf); } mutex_exit(&dbf->db_mtx); } } if (i > 0) { DBUF_STAT_BUMP(hash_collisions); if (i == 1) DBUF_STAT_BUMP(hash_chains); DBUF_STAT_MAX(hash_chain_max, i); } mutex_enter(&db->db_mtx); db->db_hash_next = h->hash_table[idx]; h->hash_table[idx] = db; mutex_exit(DBUF_HASH_MUTEX(h, idx)); uint64_t he = atomic_inc_64_nv(&dbuf_stats.hash_elements.value.ui64); DBUF_STAT_MAX(hash_elements_max, he); return (NULL); } /* * This returns whether this dbuf should be stored in the metadata cache, which * is based on whether it's from one of the dnode types that store data related * to traversing dataset hierarchies. */ static boolean_t dbuf_include_in_metadata_cache(dmu_buf_impl_t *db) { DB_DNODE_ENTER(db); dmu_object_type_t type = DB_DNODE(db)->dn_type; DB_DNODE_EXIT(db); /* Check if this dbuf is one of the types we care about */ if (DMU_OT_IS_METADATA_CACHED(type)) { /* If we hit this, then we set something up wrong in dmu_ot */ ASSERT(DMU_OT_IS_METADATA(type)); /* * Sanity check for small-memory systems: don't allocate too * much memory for this purpose. */ if (zfs_refcount_count( &dbuf_caches[DB_DBUF_METADATA_CACHE].size) > dbuf_metadata_cache_target_bytes()) { DBUF_STAT_BUMP(metadata_cache_overflow); return (B_FALSE); } return (B_TRUE); } return (B_FALSE); } /* * Remove an entry from the hash table. It must be in the EVICTING state. */ static void dbuf_hash_remove(dmu_buf_impl_t *db) { dbuf_hash_table_t *h = &dbuf_hash_table; uint64_t idx; dmu_buf_impl_t *dbf, **dbp; ASSERT3U(dbuf_hash(db->db_objset, db->db.db_object, db->db_level, db->db_blkid), ==, db->db_hash); idx = db->db_hash & h->hash_table_mask; /* * We mustn't hold db_mtx to maintain lock ordering: * DBUF_HASH_MUTEX > db_mtx. */ ASSERT(zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db_state == DB_EVICTING); ASSERT(!MUTEX_HELD(&db->db_mtx)); mutex_enter(DBUF_HASH_MUTEX(h, idx)); dbp = &h->hash_table[idx]; while ((dbf = *dbp) != db) { dbp = &dbf->db_hash_next; ASSERT(dbf != NULL); } *dbp = db->db_hash_next; db->db_hash_next = NULL; if (h->hash_table[idx] && h->hash_table[idx]->db_hash_next == NULL) DBUF_STAT_BUMPDOWN(hash_chains); mutex_exit(DBUF_HASH_MUTEX(h, idx)); atomic_dec_64(&dbuf_stats.hash_elements.value.ui64); } typedef enum { DBVU_EVICTING, DBVU_NOT_EVICTING } dbvu_verify_type_t; static void dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type) { #ifdef ZFS_DEBUG int64_t holds; if (db->db_user == NULL) return; /* Only data blocks support the attachment of user data. */ ASSERT(db->db_level == 0); /* Clients must resolve a dbuf before attaching user data. */ ASSERT(db->db.db_data != NULL); ASSERT3U(db->db_state, ==, DB_CACHED); holds = zfs_refcount_count(&db->db_holds); if (verify_type == DBVU_EVICTING) { /* * Immediate eviction occurs when holds == dirtycnt. * For normal eviction buffers, holds is zero on * eviction, except when dbuf_fix_old_data() calls * dbuf_clear_data(). However, the hold count can grow * during eviction even though db_mtx is held (see * dmu_bonus_hold() for an example), so we can only * test the generic invariant that holds >= dirtycnt. */ ASSERT3U(holds, >=, db->db_dirtycnt); } else { if (db->db_user_immediate_evict == TRUE) ASSERT3U(holds, >=, db->db_dirtycnt); else ASSERT3U(holds, >, 0); } #endif } static void dbuf_evict_user(dmu_buf_impl_t *db) { dmu_buf_user_t *dbu = db->db_user; ASSERT(MUTEX_HELD(&db->db_mtx)); if (dbu == NULL) return; dbuf_verify_user(db, DBVU_EVICTING); db->db_user = NULL; #ifdef ZFS_DEBUG if (dbu->dbu_clear_on_evict_dbufp != NULL) *dbu->dbu_clear_on_evict_dbufp = NULL; #endif if (db->db_caching_status != DB_NO_CACHE) { /* * This is a cached dbuf, so the size of the user data is * included in its cached amount. We adjust it here because the * user data has already been detached from the dbuf, and the * sync functions are not supposed to touch it (the dbuf might * not exist anymore by the time the sync functions run. */ uint64_t size = dbu->dbu_size; (void) zfs_refcount_remove_many( &dbuf_caches[db->db_caching_status].size, size, db); if (db->db_caching_status == DB_DBUF_CACHE) DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size); } /* * There are two eviction callbacks - one that we call synchronously * and one that we invoke via a taskq. The async one is useful for * avoiding lock order reversals and limiting stack depth. * * Note that if we have a sync callback but no async callback, * it's likely that the sync callback will free the structure * containing the dbu. In that case we need to take care to not * dereference dbu after calling the sync evict func. */ boolean_t has_async = (dbu->dbu_evict_func_async != NULL); if (dbu->dbu_evict_func_sync != NULL) dbu->dbu_evict_func_sync(dbu); if (has_async) { taskq_dispatch_ent(dbu_evict_taskq, dbu->dbu_evict_func_async, dbu, 0, &dbu->dbu_tqent); } } boolean_t dbuf_is_metadata(dmu_buf_impl_t *db) { /* * Consider indirect blocks and spill blocks to be meta data. */ if (db->db_level > 0 || db->db_blkid == DMU_SPILL_BLKID) { return (B_TRUE); } else { boolean_t is_metadata; DB_DNODE_ENTER(db); is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type); DB_DNODE_EXIT(db); return (is_metadata); } } /* * We want to exclude buffers that are on a special allocation class from * L2ARC. */ boolean_t dbuf_is_l2cacheable(dmu_buf_impl_t *db) { if (db->db_objset->os_secondary_cache == ZFS_CACHE_ALL || (db->db_objset->os_secondary_cache == ZFS_CACHE_METADATA && dbuf_is_metadata(db))) { if (l2arc_exclude_special == 0) return (B_TRUE); blkptr_t *bp = db->db_blkptr; if (bp == NULL || BP_IS_HOLE(bp)) return (B_FALSE); uint64_t vdev = DVA_GET_VDEV(bp->blk_dva); vdev_t *rvd = db->db_objset->os_spa->spa_root_vdev; vdev_t *vd = NULL; if (vdev < rvd->vdev_children) vd = rvd->vdev_child[vdev]; if (vd == NULL) return (B_TRUE); if (vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL && vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) return (B_TRUE); } return (B_FALSE); } static inline boolean_t dnode_level_is_l2cacheable(blkptr_t *bp, dnode_t *dn, int64_t level) { if (dn->dn_objset->os_secondary_cache == ZFS_CACHE_ALL || (dn->dn_objset->os_secondary_cache == ZFS_CACHE_METADATA && (level > 0 || DMU_OT_IS_METADATA(dn->dn_handle->dnh_dnode->dn_type)))) { if (l2arc_exclude_special == 0) return (B_TRUE); if (bp == NULL || BP_IS_HOLE(bp)) return (B_FALSE); uint64_t vdev = DVA_GET_VDEV(bp->blk_dva); vdev_t *rvd = dn->dn_objset->os_spa->spa_root_vdev; vdev_t *vd = NULL; if (vdev < rvd->vdev_children) vd = rvd->vdev_child[vdev]; if (vd == NULL) return (B_TRUE); if (vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL && vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) return (B_TRUE); } return (B_FALSE); } /* * This function *must* return indices evenly distributed between all * sublists of the multilist. This is needed due to how the dbuf eviction * code is laid out; dbuf_evict_thread() assumes dbufs are evenly * distributed between all sublists and uses this assumption when * deciding which sublist to evict from and how much to evict from it. */ static unsigned int dbuf_cache_multilist_index_func(multilist_t *ml, void *obj) { dmu_buf_impl_t *db = obj; /* * The assumption here, is the hash value for a given * dmu_buf_impl_t will remain constant throughout it's lifetime * (i.e. it's objset, object, level and blkid fields don't change). * Thus, we don't need to store the dbuf's sublist index * on insertion, as this index can be recalculated on removal. * * Also, the low order bits of the hash value are thought to be * distributed evenly. Otherwise, in the case that the multilist * has a power of two number of sublists, each sublists' usage * would not be evenly distributed. In this context full 64bit * division would be a waste of time, so limit it to 32 bits. */ return ((unsigned int)dbuf_hash(db->db_objset, db->db.db_object, db->db_level, db->db_blkid) % multilist_get_num_sublists(ml)); } /* * The target size of the dbuf cache can grow with the ARC target, * unless limited by the tunable dbuf_cache_max_bytes. */ static inline unsigned long dbuf_cache_target_bytes(void) { return (MIN(dbuf_cache_max_bytes, arc_target_bytes() >> dbuf_cache_shift)); } /* * The target size of the dbuf metadata cache can grow with the ARC target, * unless limited by the tunable dbuf_metadata_cache_max_bytes. */ static inline unsigned long dbuf_metadata_cache_target_bytes(void) { return (MIN(dbuf_metadata_cache_max_bytes, arc_target_bytes() >> dbuf_metadata_cache_shift)); } static inline uint64_t dbuf_cache_hiwater_bytes(void) { uint64_t dbuf_cache_target = dbuf_cache_target_bytes(); return (dbuf_cache_target + (dbuf_cache_target * dbuf_cache_hiwater_pct) / 100); } static inline uint64_t dbuf_cache_lowater_bytes(void) { uint64_t dbuf_cache_target = dbuf_cache_target_bytes(); return (dbuf_cache_target - (dbuf_cache_target * dbuf_cache_lowater_pct) / 100); } static inline boolean_t dbuf_cache_above_lowater(void) { return (zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) > dbuf_cache_lowater_bytes()); } /* * Evict the oldest eligible dbuf from the dbuf cache. */ static void dbuf_evict_one(void) { int idx = multilist_get_random_index(&dbuf_caches[DB_DBUF_CACHE].cache); multilist_sublist_t *mls = multilist_sublist_lock_idx( &dbuf_caches[DB_DBUF_CACHE].cache, idx); ASSERT(!MUTEX_HELD(&dbuf_evict_lock)); dmu_buf_impl_t *db = multilist_sublist_tail(mls); while (db != NULL && mutex_tryenter(&db->db_mtx) == 0) { db = multilist_sublist_prev(mls, db); } DTRACE_PROBE2(dbuf__evict__one, dmu_buf_impl_t *, db, multilist_sublist_t *, mls); if (db != NULL) { multilist_sublist_remove(mls, db); multilist_sublist_unlock(mls); uint64_t size = db->db.db_size + dmu_buf_user_size(&db->db); (void) zfs_refcount_remove_many( &dbuf_caches[DB_DBUF_CACHE].size, size, db); DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size); ASSERT3U(db->db_caching_status, ==, DB_DBUF_CACHE); db->db_caching_status = DB_NO_CACHE; dbuf_destroy(db); DBUF_STAT_BUMP(cache_total_evicts); } else { multilist_sublist_unlock(mls); } } /* * The dbuf evict thread is responsible for aging out dbufs from the * cache. Once the cache has reached it's maximum size, dbufs are removed * and destroyed. The eviction thread will continue running until the size * of the dbuf cache is at or below the maximum size. Once the dbuf is aged * out of the cache it is destroyed and becomes eligible for arc eviction. */ static __attribute__((noreturn)) void dbuf_evict_thread(void *unused) { (void) unused; callb_cpr_t cpr; CALLB_CPR_INIT(&cpr, &dbuf_evict_lock, callb_generic_cpr, FTAG); mutex_enter(&dbuf_evict_lock); while (!dbuf_evict_thread_exit) { while (!dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) { CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait_idle_hires(&dbuf_evict_cv, &dbuf_evict_lock, SEC2NSEC(1), MSEC2NSEC(1), 0); CALLB_CPR_SAFE_END(&cpr, &dbuf_evict_lock); } mutex_exit(&dbuf_evict_lock); /* * Keep evicting as long as we're above the low water mark * for the cache. We do this without holding the locks to * minimize lock contention. */ while (dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) { dbuf_evict_one(); } mutex_enter(&dbuf_evict_lock); } dbuf_evict_thread_exit = B_FALSE; cv_broadcast(&dbuf_evict_cv); CALLB_CPR_EXIT(&cpr); /* drops dbuf_evict_lock */ thread_exit(); } /* * Wake up the dbuf eviction thread if the dbuf cache is at its max size. * If the dbuf cache is at its high water mark, then evict a dbuf from the * dbuf cache using the caller's context. */ static void dbuf_evict_notify(uint64_t size) { /* * We check if we should evict without holding the dbuf_evict_lock, * because it's OK to occasionally make the wrong decision here, * and grabbing the lock results in massive lock contention. */ if (size > dbuf_cache_target_bytes()) { if (size > dbuf_cache_hiwater_bytes()) dbuf_evict_one(); cv_signal(&dbuf_evict_cv); } } static int dbuf_kstat_update(kstat_t *ksp, int rw) { dbuf_stats_t *ds = ksp->ks_data; dbuf_hash_table_t *h = &dbuf_hash_table; if (rw == KSTAT_WRITE) return (SET_ERROR(EACCES)); ds->cache_count.value.ui64 = wmsum_value(&dbuf_sums.cache_count); ds->cache_size_bytes.value.ui64 = zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size); ds->cache_target_bytes.value.ui64 = dbuf_cache_target_bytes(); ds->cache_hiwater_bytes.value.ui64 = dbuf_cache_hiwater_bytes(); ds->cache_lowater_bytes.value.ui64 = dbuf_cache_lowater_bytes(); ds->cache_total_evicts.value.ui64 = wmsum_value(&dbuf_sums.cache_total_evicts); for (int i = 0; i < DN_MAX_LEVELS; i++) { ds->cache_levels[i].value.ui64 = wmsum_value(&dbuf_sums.cache_levels[i]); ds->cache_levels_bytes[i].value.ui64 = wmsum_value(&dbuf_sums.cache_levels_bytes[i]); } ds->hash_hits.value.ui64 = wmsum_value(&dbuf_sums.hash_hits); ds->hash_misses.value.ui64 = wmsum_value(&dbuf_sums.hash_misses); ds->hash_collisions.value.ui64 = wmsum_value(&dbuf_sums.hash_collisions); ds->hash_chains.value.ui64 = wmsum_value(&dbuf_sums.hash_chains); ds->hash_insert_race.value.ui64 = wmsum_value(&dbuf_sums.hash_insert_race); ds->hash_table_count.value.ui64 = h->hash_table_mask + 1; ds->hash_mutex_count.value.ui64 = h->hash_mutex_mask + 1; ds->metadata_cache_count.value.ui64 = wmsum_value(&dbuf_sums.metadata_cache_count); ds->metadata_cache_size_bytes.value.ui64 = zfs_refcount_count( &dbuf_caches[DB_DBUF_METADATA_CACHE].size); ds->metadata_cache_overflow.value.ui64 = wmsum_value(&dbuf_sums.metadata_cache_overflow); return (0); } void dbuf_init(void) { uint64_t hmsize, hsize = 1ULL << 16; dbuf_hash_table_t *h = &dbuf_hash_table; /* * The hash table is big enough to fill one eighth of physical memory * with an average block size of zfs_arc_average_blocksize (default 8K). * By default, the table will take up * totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers). */ while (hsize * zfs_arc_average_blocksize < arc_all_memory() / 8) hsize <<= 1; h->hash_table = NULL; while (h->hash_table == NULL) { h->hash_table_mask = hsize - 1; h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP); if (h->hash_table == NULL) hsize >>= 1; ASSERT3U(hsize, >=, 1ULL << 10); } /* * The hash table buckets are protected by an array of mutexes where * each mutex is reponsible for protecting 128 buckets. A minimum * array size of 8192 is targeted to avoid contention. */ if (dbuf_mutex_cache_shift == 0) hmsize = MAX(hsize >> 7, 1ULL << 13); else hmsize = 1ULL << MIN(dbuf_mutex_cache_shift, 24); h->hash_mutexes = NULL; while (h->hash_mutexes == NULL) { h->hash_mutex_mask = hmsize - 1; h->hash_mutexes = vmem_zalloc(hmsize * sizeof (kmutex_t), KM_SLEEP); if (h->hash_mutexes == NULL) hmsize >>= 1; } dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t", sizeof (dmu_buf_impl_t), 0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0); for (int i = 0; i < hmsize; i++) mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL); dbuf_stats_init(h); /* * All entries are queued via taskq_dispatch_ent(), so min/maxalloc * configuration is not required. */ dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0); for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) { multilist_create(&dbuf_caches[dcs].cache, sizeof (dmu_buf_impl_t), offsetof(dmu_buf_impl_t, db_cache_link), dbuf_cache_multilist_index_func); zfs_refcount_create(&dbuf_caches[dcs].size); } dbuf_evict_thread_exit = B_FALSE; mutex_init(&dbuf_evict_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&dbuf_evict_cv, NULL, CV_DEFAULT, NULL); dbuf_cache_evict_thread = thread_create(NULL, 0, dbuf_evict_thread, NULL, 0, &p0, TS_RUN, minclsyspri); wmsum_init(&dbuf_sums.cache_count, 0); wmsum_init(&dbuf_sums.cache_total_evicts, 0); for (int i = 0; i < DN_MAX_LEVELS; i++) { wmsum_init(&dbuf_sums.cache_levels[i], 0); wmsum_init(&dbuf_sums.cache_levels_bytes[i], 0); } wmsum_init(&dbuf_sums.hash_hits, 0); wmsum_init(&dbuf_sums.hash_misses, 0); wmsum_init(&dbuf_sums.hash_collisions, 0); wmsum_init(&dbuf_sums.hash_chains, 0); wmsum_init(&dbuf_sums.hash_insert_race, 0); wmsum_init(&dbuf_sums.metadata_cache_count, 0); wmsum_init(&dbuf_sums.metadata_cache_overflow, 0); dbuf_ksp = kstat_create("zfs", 0, "dbufstats", "misc", KSTAT_TYPE_NAMED, sizeof (dbuf_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (dbuf_ksp != NULL) { for (int i = 0; i < DN_MAX_LEVELS; i++) { snprintf(dbuf_stats.cache_levels[i].name, KSTAT_STRLEN, "cache_level_%d", i); dbuf_stats.cache_levels[i].data_type = KSTAT_DATA_UINT64; snprintf(dbuf_stats.cache_levels_bytes[i].name, KSTAT_STRLEN, "cache_level_%d_bytes", i); dbuf_stats.cache_levels_bytes[i].data_type = KSTAT_DATA_UINT64; } dbuf_ksp->ks_data = &dbuf_stats; dbuf_ksp->ks_update = dbuf_kstat_update; kstat_install(dbuf_ksp); } } void dbuf_fini(void) { dbuf_hash_table_t *h = &dbuf_hash_table; dbuf_stats_destroy(); for (int i = 0; i < (h->hash_mutex_mask + 1); i++) mutex_destroy(&h->hash_mutexes[i]); vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); vmem_free(h->hash_mutexes, (h->hash_mutex_mask + 1) * sizeof (kmutex_t)); kmem_cache_destroy(dbuf_kmem_cache); taskq_destroy(dbu_evict_taskq); mutex_enter(&dbuf_evict_lock); dbuf_evict_thread_exit = B_TRUE; while (dbuf_evict_thread_exit) { cv_signal(&dbuf_evict_cv); cv_wait(&dbuf_evict_cv, &dbuf_evict_lock); } mutex_exit(&dbuf_evict_lock); mutex_destroy(&dbuf_evict_lock); cv_destroy(&dbuf_evict_cv); for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) { zfs_refcount_destroy(&dbuf_caches[dcs].size); multilist_destroy(&dbuf_caches[dcs].cache); } if (dbuf_ksp != NULL) { kstat_delete(dbuf_ksp); dbuf_ksp = NULL; } wmsum_fini(&dbuf_sums.cache_count); wmsum_fini(&dbuf_sums.cache_total_evicts); for (int i = 0; i < DN_MAX_LEVELS; i++) { wmsum_fini(&dbuf_sums.cache_levels[i]); wmsum_fini(&dbuf_sums.cache_levels_bytes[i]); } wmsum_fini(&dbuf_sums.hash_hits); wmsum_fini(&dbuf_sums.hash_misses); wmsum_fini(&dbuf_sums.hash_collisions); wmsum_fini(&dbuf_sums.hash_chains); wmsum_fini(&dbuf_sums.hash_insert_race); wmsum_fini(&dbuf_sums.metadata_cache_count); wmsum_fini(&dbuf_sums.metadata_cache_overflow); } /* * Other stuff. */ #ifdef ZFS_DEBUG static void dbuf_verify(dmu_buf_impl_t *db) { dnode_t *dn; dbuf_dirty_record_t *dr; uint32_t txg_prev; ASSERT(MUTEX_HELD(&db->db_mtx)); if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY)) return; ASSERT(db->db_objset != NULL); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn == NULL) { ASSERT(db->db_parent == NULL); ASSERT(db->db_blkptr == NULL); } else { ASSERT3U(db->db.db_object, ==, dn->dn_object); ASSERT3P(db->db_objset, ==, dn->dn_objset); ASSERT3U(db->db_level, <, dn->dn_nlevels); ASSERT(db->db_blkid == DMU_BONUS_BLKID || db->db_blkid == DMU_SPILL_BLKID || !avl_is_empty(&dn->dn_dbufs)); } if (db->db_blkid == DMU_BONUS_BLKID) { ASSERT(dn != NULL); ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen); ASSERT3U(db->db.db_offset, ==, DMU_BONUS_BLKID); } else if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn != NULL); ASSERT0(db->db.db_offset); } else { ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size); } if ((dr = list_head(&db->db_dirty_records)) != NULL) { ASSERT(dr->dr_dbuf == db); txg_prev = dr->dr_txg; for (dr = list_next(&db->db_dirty_records, dr); dr != NULL; dr = list_next(&db->db_dirty_records, dr)) { ASSERT(dr->dr_dbuf == db); ASSERT(txg_prev > dr->dr_txg); txg_prev = dr->dr_txg; } } /* * We can't assert that db_size matches dn_datablksz because it * can be momentarily different when another thread is doing * dnode_set_blksz(). */ if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) { dr = db->db_data_pending; /* * It should only be modified in syncing context, so * make sure we only have one copy of the data. */ ASSERT(dr == NULL || dr->dt.dl.dr_data == db->db_buf); } /* verify db->db_blkptr */ if (db->db_blkptr) { if (db->db_parent == dn->dn_dbuf) { /* db is pointed to by the dnode */ /* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */ if (DMU_OBJECT_IS_SPECIAL(db->db.db_object)) ASSERT(db->db_parent == NULL); else ASSERT(db->db_parent != NULL); if (db->db_blkid != DMU_SPILL_BLKID) ASSERT3P(db->db_blkptr, ==, &dn->dn_phys->dn_blkptr[db->db_blkid]); } else { /* db is pointed to by an indirect block */ int epb __maybe_unused = db->db_parent->db.db_size >> SPA_BLKPTRSHIFT; ASSERT3U(db->db_parent->db_level, ==, db->db_level+1); ASSERT3U(db->db_parent->db.db_object, ==, db->db.db_object); /* * dnode_grow_indblksz() can make this fail if we don't * have the parent's rwlock. XXX indblksz no longer * grows. safe to do this now? */ if (RW_LOCK_HELD(&db->db_parent->db_rwlock)) { ASSERT3P(db->db_blkptr, ==, ((blkptr_t *)db->db_parent->db.db_data + db->db_blkid % epb)); } } } if ((db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr)) && (db->db_buf == NULL || db->db_buf->b_data) && db->db.db_data && db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_FILL && (dn == NULL || !dn->dn_free_txg)) { /* * If the blkptr isn't set but they have nonzero data, * it had better be dirty, otherwise we'll lose that * data when we evict this buffer. * * There is an exception to this rule for indirect blocks; in * this case, if the indirect block is a hole, we fill in a few * fields on each of the child blocks (importantly, birth time) * to prevent hole birth times from being lost when you * partially fill in a hole. */ if (db->db_dirtycnt == 0) { if (db->db_level == 0) { uint64_t *buf = db->db.db_data; int i; for (i = 0; i < db->db.db_size >> 3; i++) { ASSERT(buf[i] == 0); } } else { blkptr_t *bps = db->db.db_data; ASSERT3U(1 << DB_DNODE(db)->dn_indblkshift, ==, db->db.db_size); /* * We want to verify that all the blkptrs in the * indirect block are holes, but we may have * automatically set up a few fields for them. * We iterate through each blkptr and verify * they only have those fields set. */ for (int i = 0; i < db->db.db_size / sizeof (blkptr_t); i++) { blkptr_t *bp = &bps[i]; ASSERT(ZIO_CHECKSUM_IS_ZERO( &bp->blk_cksum)); ASSERT( DVA_IS_EMPTY(&bp->blk_dva[0]) && DVA_IS_EMPTY(&bp->blk_dva[1]) && DVA_IS_EMPTY(&bp->blk_dva[2])); ASSERT0(bp->blk_fill); ASSERT0(bp->blk_pad[0]); ASSERT0(bp->blk_pad[1]); ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT(BP_IS_HOLE(bp)); ASSERT0(BP_GET_PHYSICAL_BIRTH(bp)); } } } } DB_DNODE_EXIT(db); } #endif static void dbuf_clear_data(dmu_buf_impl_t *db) { ASSERT(MUTEX_HELD(&db->db_mtx)); dbuf_evict_user(db); ASSERT3P(db->db_buf, ==, NULL); db->db.db_data = NULL; if (db->db_state != DB_NOFILL) { db->db_state = DB_UNCACHED; DTRACE_SET_STATE(db, "clear data"); } } static void dbuf_set_data(dmu_buf_impl_t *db, arc_buf_t *buf) { ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(buf != NULL); db->db_buf = buf; ASSERT(buf->b_data != NULL); db->db.db_data = buf->b_data; } static arc_buf_t * dbuf_alloc_arcbuf(dmu_buf_impl_t *db) { spa_t *spa = db->db_objset->os_spa; return (arc_alloc_buf(spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size)); } /* * Loan out an arc_buf for read. Return the loaned arc_buf. */ arc_buf_t * dbuf_loan_arcbuf(dmu_buf_impl_t *db) { arc_buf_t *abuf; ASSERT(db->db_blkid != DMU_BONUS_BLKID); mutex_enter(&db->db_mtx); if (arc_released(db->db_buf) || zfs_refcount_count(&db->db_holds) > 1) { int blksz = db->db.db_size; spa_t *spa = db->db_objset->os_spa; mutex_exit(&db->db_mtx); abuf = arc_loan_buf(spa, B_FALSE, blksz); memcpy(abuf->b_data, db->db.db_data, blksz); } else { abuf = db->db_buf; arc_loan_inuse_buf(abuf, db); db->db_buf = NULL; dbuf_clear_data(db); mutex_exit(&db->db_mtx); } return (abuf); } /* * Calculate which level n block references the data at the level 0 offset * provided. */ uint64_t dbuf_whichblock(const dnode_t *dn, const int64_t level, const uint64_t offset) { if (dn->dn_datablkshift != 0 && dn->dn_indblkshift != 0) { /* * The level n blkid is equal to the level 0 blkid divided by * the number of level 0s in a level n block. * * The level 0 blkid is offset >> datablkshift = * offset / 2^datablkshift. * * The number of level 0s in a level n is the number of block * pointers in an indirect block, raised to the power of level. * This is 2^(indblkshift - SPA_BLKPTRSHIFT)^level = * 2^(level*(indblkshift - SPA_BLKPTRSHIFT)). * * Thus, the level n blkid is: offset / * ((2^datablkshift)*(2^(level*(indblkshift-SPA_BLKPTRSHIFT)))) * = offset / 2^(datablkshift + level * * (indblkshift - SPA_BLKPTRSHIFT)) * = offset >> (datablkshift + level * * (indblkshift - SPA_BLKPTRSHIFT)) */ const unsigned exp = dn->dn_datablkshift + level * (dn->dn_indblkshift - SPA_BLKPTRSHIFT); if (exp >= 8 * sizeof (offset)) { /* This only happens on the highest indirection level */ ASSERT3U(level, ==, dn->dn_nlevels - 1); return (0); } ASSERT3U(exp, <, 8 * sizeof (offset)); return (offset >> exp); } else { ASSERT3U(offset, <, dn->dn_datablksz); return (0); } } /* * This function is used to lock the parent of the provided dbuf. This should be * used when modifying or reading db_blkptr. */ db_lock_type_t dmu_buf_lock_parent(dmu_buf_impl_t *db, krw_t rw, const void *tag) { enum db_lock_type ret = DLT_NONE; if (db->db_parent != NULL) { rw_enter(&db->db_parent->db_rwlock, rw); ret = DLT_PARENT; } else if (dmu_objset_ds(db->db_objset) != NULL) { rrw_enter(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, rw, tag); ret = DLT_OBJSET; } /* * We only return a DLT_NONE lock when it's the top-most indirect block * of the meta-dnode of the MOS. */ return (ret); } /* * We need to pass the lock type in because it's possible that the block will * move from being the topmost indirect block in a dnode (and thus, have no * parent) to not the top-most via an indirection increase. This would cause a * panic if we didn't pass the lock type in. */ void dmu_buf_unlock_parent(dmu_buf_impl_t *db, db_lock_type_t type, const void *tag) { if (type == DLT_PARENT) rw_exit(&db->db_parent->db_rwlock); else if (type == DLT_OBJSET) rrw_exit(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, tag); } static void dbuf_read_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *vdb) { (void) zb, (void) bp; dmu_buf_impl_t *db = vdb; mutex_enter(&db->db_mtx); ASSERT3U(db->db_state, ==, DB_READ); /* * All reads are synchronous, so we must have a hold on the dbuf */ ASSERT(zfs_refcount_count(&db->db_holds) > 0); ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); if (buf == NULL) { /* i/o error */ ASSERT(zio == NULL || zio->io_error != 0); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT3P(db->db_buf, ==, NULL); db->db_state = DB_UNCACHED; DTRACE_SET_STATE(db, "i/o error"); } else if (db->db_level == 0 && db->db_freed_in_flight) { /* freed in flight */ ASSERT(zio == NULL || zio->io_error == 0); arc_release(buf, db); memset(buf->b_data, 0, db->db.db_size); arc_buf_freeze(buf); db->db_freed_in_flight = FALSE; dbuf_set_data(db, buf); db->db_state = DB_CACHED; DTRACE_SET_STATE(db, "freed in flight"); } else { /* success */ ASSERT(zio == NULL || zio->io_error == 0); dbuf_set_data(db, buf); db->db_state = DB_CACHED; DTRACE_SET_STATE(db, "successful read"); } cv_broadcast(&db->db_changed); dbuf_rele_and_unlock(db, NULL, B_FALSE); } /* * Shortcut for performing reads on bonus dbufs. Returns * an error if we fail to verify the dnode associated with * a decrypted block. Otherwise success. */ static int -dbuf_read_bonus(dmu_buf_impl_t *db, dnode_t *dn, uint32_t flags) +dbuf_read_bonus(dmu_buf_impl_t *db, dnode_t *dn) { - int bonuslen, max_bonuslen, err; - - err = dbuf_read_verify_dnode_crypt(db, flags); - if (err) - return (err); + int bonuslen, max_bonuslen; bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen); max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(DB_DNODE_HELD(db)); ASSERT3U(bonuslen, <=, db->db.db_size); db->db.db_data = kmem_alloc(max_bonuslen, KM_SLEEP); arc_space_consume(max_bonuslen, ARC_SPACE_BONUS); if (bonuslen < max_bonuslen) memset(db->db.db_data, 0, max_bonuslen); if (bonuslen) memcpy(db->db.db_data, DN_BONUS(dn->dn_phys), bonuslen); db->db_state = DB_CACHED; DTRACE_SET_STATE(db, "bonus buffer filled"); return (0); } static void dbuf_handle_indirect_hole(dmu_buf_impl_t *db, dnode_t *dn, blkptr_t *dbbp) { blkptr_t *bps = db->db.db_data; uint32_t indbs = 1ULL << dn->dn_indblkshift; int n_bps = indbs >> SPA_BLKPTRSHIFT; for (int i = 0; i < n_bps; i++) { blkptr_t *bp = &bps[i]; ASSERT3U(BP_GET_LSIZE(dbbp), ==, indbs); BP_SET_LSIZE(bp, BP_GET_LEVEL(dbbp) == 1 ? dn->dn_datablksz : BP_GET_LSIZE(dbbp)); BP_SET_TYPE(bp, BP_GET_TYPE(dbbp)); BP_SET_LEVEL(bp, BP_GET_LEVEL(dbbp) - 1); BP_SET_BIRTH(bp, BP_GET_LOGICAL_BIRTH(dbbp), 0); } } /* * Handle reads on dbufs that are holes, if necessary. This function * requires that the dbuf's mutex is held. Returns success (0) if action * was taken, ENOENT if no action was taken. */ static int dbuf_read_hole(dmu_buf_impl_t *db, dnode_t *dn, blkptr_t *bp) { ASSERT(MUTEX_HELD(&db->db_mtx)); int is_hole = bp == NULL || BP_IS_HOLE(bp); /* * For level 0 blocks only, if the above check fails: * Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync() * processes the delete record and clears the bp while we are waiting * for the dn_mtx (resulting in a "no" from block_freed). */ if (!is_hole && db->db_level == 0) is_hole = dnode_block_freed(dn, db->db_blkid) || BP_IS_HOLE(bp); if (is_hole) { dbuf_set_data(db, dbuf_alloc_arcbuf(db)); memset(db->db.db_data, 0, db->db.db_size); if (bp != NULL && db->db_level > 0 && BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) != 0) { dbuf_handle_indirect_hole(db, dn, bp); } db->db_state = DB_CACHED; DTRACE_SET_STATE(db, "hole read satisfied"); return (0); } return (ENOENT); } /* * This function ensures that, when doing a decrypting read of a block, * we make sure we have decrypted the dnode associated with it. We must do * this so that we ensure we are fully authenticating the checksum-of-MACs * tree from the root of the objset down to this block. Indirect blocks are * always verified against their secure checksum-of-MACs assuming that the * dnode containing them is correct. Now that we are doing a decrypting read, * we can be sure that the key is loaded and verify that assumption. This is * especially important considering that we always read encrypted dnode * blocks as raw data (without verifying their MACs) to start, and * decrypt / authenticate them when we need to read an encrypted bonus buffer. */ static int -dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags) +dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, dnode_t *dn, uint32_t flags) { - int err = 0; objset_t *os = db->db_objset; - arc_buf_t *dnode_abuf; - dnode_t *dn; + dmu_buf_impl_t *dndb; + arc_buf_t *dnbuf; zbookmark_phys_t zb; - - ASSERT(MUTEX_HELD(&db->db_mtx)); + int err; if ((flags & DB_RF_NO_DECRYPT) != 0 || - !os->os_encrypted || os->os_raw_receive) + !os->os_encrypted || os->os_raw_receive || + (dndb = dn->dn_dbuf) == NULL) return (0); - DB_DNODE_ENTER(db); - dn = DB_DNODE(db); - dnode_abuf = (dn->dn_dbuf != NULL) ? dn->dn_dbuf->db_buf : NULL; - - if (dnode_abuf == NULL || !arc_is_encrypted(dnode_abuf)) { - DB_DNODE_EXIT(db); + dnbuf = dndb->db_buf; + if (!arc_is_encrypted(dnbuf)) return (0); - } + + mutex_enter(&dndb->db_mtx); + + /* + * Since dnode buffer is modified by sync process, there can be only + * one copy of it. It means we can not modify (decrypt) it while it + * is being written. I don't see how this may happen now, since + * encrypted dnode writes by receive should be completed before any + * plain-text reads due to txg wait, but better be safe than sorry. + */ + while (1) { + if (!arc_is_encrypted(dnbuf)) { + mutex_exit(&dndb->db_mtx); + return (0); + } + dbuf_dirty_record_t *dr = dndb->db_data_pending; + if (dr == NULL || dr->dt.dl.dr_data != dnbuf) + break; + cv_wait(&dndb->db_changed, &dndb->db_mtx); + }; SET_BOOKMARK(&zb, dmu_objset_id(os), - DMU_META_DNODE_OBJECT, 0, dn->dn_dbuf->db_blkid); - err = arc_untransform(dnode_abuf, os->os_spa, &zb, B_TRUE); + DMU_META_DNODE_OBJECT, 0, dndb->db_blkid); + err = arc_untransform(dnbuf, os->os_spa, &zb, B_TRUE); /* * An error code of EACCES tells us that the key is still not * available. This is ok if we are only reading authenticated * (and therefore non-encrypted) blocks. */ if (err == EACCES && ((db->db_blkid != DMU_BONUS_BLKID && !DMU_OT_IS_ENCRYPTED(dn->dn_type)) || (db->db_blkid == DMU_BONUS_BLKID && !DMU_OT_IS_ENCRYPTED(dn->dn_bonustype)))) err = 0; - DB_DNODE_EXIT(db); + mutex_exit(&dndb->db_mtx); return (err); } /* * Drops db_mtx and the parent lock specified by dblt and tag before * returning. */ static int dbuf_read_impl(dmu_buf_impl_t *db, dnode_t *dn, zio_t *zio, uint32_t flags, db_lock_type_t dblt, const void *tag) { zbookmark_phys_t zb; uint32_t aflags = ARC_FLAG_NOWAIT; int err, zio_flags; blkptr_t bp, *bpp = NULL; ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); ASSERT(db->db_buf == NULL); ASSERT(db->db_parent == NULL || RW_LOCK_HELD(&db->db_parent->db_rwlock)); if (db->db_blkid == DMU_BONUS_BLKID) { - err = dbuf_read_bonus(db, dn, flags); + err = dbuf_read_bonus(db, dn); goto early_unlock; } /* * If we have a pending block clone, we don't want to read the * underlying block, but the content of the block being cloned, * pointed by the dirty record, so we have the most recent data. * If there is no dirty record, then we hit a race in a sync * process when the dirty record is already removed, while the * dbuf is not yet destroyed. Such case is equivalent to uncached. */ if (db->db_state == DB_NOFILL) { dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records); if (dr != NULL) { if (!dr->dt.dl.dr_brtwrite) { err = EIO; goto early_unlock; } bp = dr->dt.dl.dr_overridden_by; bpp = &bp; } } if (bpp == NULL && db->db_blkptr != NULL) { bp = *db->db_blkptr; bpp = &bp; } err = dbuf_read_hole(db, dn, bpp); if (err == 0) goto early_unlock; ASSERT(bpp != NULL); /* * Any attempt to read a redacted block should result in an error. This * will never happen under normal conditions, but can be useful for * debugging purposes. */ if (BP_IS_REDACTED(bpp)) { ASSERT(dsl_dataset_feature_is_active( db->db_objset->os_dsl_dataset, SPA_FEATURE_REDACTED_DATASETS)); err = SET_ERROR(EIO); goto early_unlock; } SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); /* * All bps of an encrypted os should have the encryption bit set. * If this is not true it indicates tampering and we report an error. */ if (db->db_objset->os_encrypted && !BP_USES_CRYPT(bpp)) { spa_log_error(db->db_objset->os_spa, &zb, BP_GET_LOGICAL_BIRTH(bpp)); err = SET_ERROR(EIO); goto early_unlock; } - err = dbuf_read_verify_dnode_crypt(db, flags); - if (err != 0) - goto early_unlock; - db->db_state = DB_READ; DTRACE_SET_STATE(db, "read issued"); mutex_exit(&db->db_mtx); if (!DBUF_IS_CACHEABLE(db)) aflags |= ARC_FLAG_UNCACHED; else if (dbuf_is_l2cacheable(db)) aflags |= ARC_FLAG_L2CACHE; dbuf_add_ref(db, NULL); zio_flags = (flags & DB_RF_CANFAIL) ? ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED; if ((flags & DB_RF_NO_DECRYPT) && BP_IS_PROTECTED(db->db_blkptr)) zio_flags |= ZIO_FLAG_RAW; /* * The zio layer will copy the provided blkptr later, but we have our * own copy so that we can release the parent's rwlock. We have to * do that so that if dbuf_read_done is called synchronously (on * an l1 cache hit) we don't acquire the db_mtx while holding the * parent's rwlock, which would be a lock ordering violation. */ dmu_buf_unlock_parent(db, dblt, tag); return (arc_read(zio, db->db_objset->os_spa, bpp, dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb)); early_unlock: mutex_exit(&db->db_mtx); dmu_buf_unlock_parent(db, dblt, tag); return (err); } /* * This is our just-in-time copy function. It makes a copy of buffers that * have been modified in a previous transaction group before we access them in * the current active group. * * This function is used in three places: when we are dirtying a buffer for the * first time in a txg, when we are freeing a range in a dnode that includes * this buffer, and when we are accessing a buffer which was received compressed * and later referenced in a WRITE_BYREF record. * * Note that when we are called from dbuf_free_range() we do not put a hold on * the buffer, we just traverse the active dbuf list for the dnode. */ static void dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg) { dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db.db_data != NULL); ASSERT(db->db_level == 0); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT); if (dr == NULL || (dr->dt.dl.dr_data != ((db->db_blkid == DMU_BONUS_BLKID) ? db->db.db_data : db->db_buf))) return; /* * If the last dirty record for this dbuf has not yet synced * and its referencing the dbuf data, either: * reset the reference to point to a new copy, * or (if there a no active holders) * just null out the current db_data pointer. */ ASSERT3U(dr->dr_txg, >=, txg - 2); if (db->db_blkid == DMU_BONUS_BLKID) { dnode_t *dn = DB_DNODE(db); int bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); dr->dt.dl.dr_data = kmem_alloc(bonuslen, KM_SLEEP); arc_space_consume(bonuslen, ARC_SPACE_BONUS); memcpy(dr->dt.dl.dr_data, db->db.db_data, bonuslen); } else if (zfs_refcount_count(&db->db_holds) > db->db_dirtycnt) { dnode_t *dn = DB_DNODE(db); int size = arc_buf_size(db->db_buf); arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); spa_t *spa = db->db_objset->os_spa; enum zio_compress compress_type = arc_get_compression(db->db_buf); uint8_t complevel = arc_get_complevel(db->db_buf); if (arc_is_encrypted(db->db_buf)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(db->db_buf, &byteorder, salt, iv, mac); dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db, dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac, dn->dn_type, size, arc_buf_lsize(db->db_buf), compress_type, complevel); } else if (compress_type != ZIO_COMPRESS_OFF) { ASSERT3U(type, ==, ARC_BUFC_DATA); dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db, size, arc_buf_lsize(db->db_buf), compress_type, complevel); } else { dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size); } memcpy(dr->dt.dl.dr_data->b_data, db->db.db_data, size); } else { db->db_buf = NULL; dbuf_clear_data(db); } } int dbuf_read(dmu_buf_impl_t *db, zio_t *pio, uint32_t flags) { - int err = 0; - boolean_t prefetch; dnode_t *dn; + boolean_t miss = B_TRUE, need_wait = B_FALSE, prefetch; + int err; - /* - * We don't have to hold the mutex to check db_state because it - * can't be freed while we have a hold on the buffer. - */ ASSERT(!zfs_refcount_is_zero(&db->db_holds)); DB_DNODE_ENTER(db); dn = DB_DNODE(db); + /* + * Ensure that this block's dnode has been decrypted if the caller + * has requested decrypted data. + */ + err = dbuf_read_verify_dnode_crypt(db, dn, flags); + if (err != 0) + goto done; + prefetch = db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && (flags & DB_RF_NOPREFETCH) == 0; mutex_enter(&db->db_mtx); if (flags & DB_RF_PARTIAL_FIRST) db->db_partial_read = B_TRUE; else if (!(flags & DB_RF_PARTIAL_MORE)) db->db_partial_read = B_FALSE; - if (db->db_state == DB_CACHED) { + miss = (db->db_state != DB_CACHED); + + if (db->db_state == DB_READ || db->db_state == DB_FILL) { /* - * Ensure that this block's dnode has been decrypted if - * the caller has requested decrypted data. + * Another reader came in while the dbuf was in flight between + * UNCACHED and CACHED. Either a writer will finish filling + * the buffer, sending the dbuf to CACHED, or the first reader's + * request will reach the read_done callback and send the dbuf + * to CACHED. Otherwise, a failure occurred and the dbuf will + * be sent to UNCACHED. */ - err = dbuf_read_verify_dnode_crypt(db, flags); + if (flags & DB_RF_NEVERWAIT) { + mutex_exit(&db->db_mtx); + DB_DNODE_EXIT(db); + goto done; + } + do { + ASSERT(db->db_state == DB_READ || + (flags & DB_RF_HAVESTRUCT) == 0); + DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *, db, + zio_t *, pio); + cv_wait(&db->db_changed, &db->db_mtx); + } while (db->db_state == DB_READ || db->db_state == DB_FILL); + if (db->db_state == DB_UNCACHED) { + err = SET_ERROR(EIO); + mutex_exit(&db->db_mtx); + DB_DNODE_EXIT(db); + goto done; + } + } + if (db->db_state == DB_CACHED) { /* * If the arc buf is compressed or encrypted and the caller * requested uncompressed data, we need to untransform it * before returning. We also call arc_untransform() on any * unauthenticated blocks, which will verify their MAC if * the key is now available. */ - if (err == 0 && db->db_buf != NULL && - (flags & DB_RF_NO_DECRYPT) == 0 && + if ((flags & DB_RF_NO_DECRYPT) == 0 && db->db_buf != NULL && (arc_is_encrypted(db->db_buf) || arc_is_unauthenticated(db->db_buf) || arc_get_compression(db->db_buf) != ZIO_COMPRESS_OFF)) { spa_t *spa = dn->dn_objset->os_spa; zbookmark_phys_t zb; SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); dbuf_fix_old_data(db, spa_syncing_txg(spa)); err = arc_untransform(db->db_buf, spa, &zb, B_FALSE); dbuf_set_data(db, db->db_buf); } mutex_exit(&db->db_mtx); - if (err == 0 && prefetch) { - dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE, - B_FALSE, flags & DB_RF_HAVESTRUCT); - } - DB_DNODE_EXIT(db); - DBUF_STAT_BUMP(hash_hits); - } else if (db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL) { - boolean_t need_wait = B_FALSE; - + } else { + ASSERT(db->db_state == DB_UNCACHED || + db->db_state == DB_NOFILL); db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG); - if (pio == NULL && (db->db_state == DB_NOFILL || (db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr)))) { spa_t *spa = dn->dn_objset->os_spa; pio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); need_wait = B_TRUE; } err = dbuf_read_impl(db, dn, pio, flags, dblt, FTAG); - /* - * dbuf_read_impl has dropped db_mtx and our parent's rwlock - * for us - */ - if (!err && prefetch) { - dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE, - db->db_state != DB_CACHED, - flags & DB_RF_HAVESTRUCT); - } - - DB_DNODE_EXIT(db); - DBUF_STAT_BUMP(hash_misses); + /* dbuf_read_impl drops db_mtx and parent's rwlock. */ + miss = (db->db_state != DB_CACHED); + } - /* - * If we created a zio_root we must execute it to avoid - * leaking it, even if it isn't attached to any work due - * to an error in dbuf_read_impl(). - */ - if (need_wait) { - if (err == 0) - err = zio_wait(pio); - else - (void) zio_wait(pio); - pio = NULL; - } - } else { - /* - * Another reader came in while the dbuf was in flight - * between UNCACHED and CACHED. Either a writer will finish - * writing the buffer (sending the dbuf to CACHED) or the - * first reader's request will reach the read_done callback - * and send the dbuf to CACHED. Otherwise, a failure - * occurred and the dbuf went to UNCACHED. - */ - mutex_exit(&db->db_mtx); - if (prefetch) { - dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE, - B_TRUE, flags & DB_RF_HAVESTRUCT); - } - DB_DNODE_EXIT(db); - DBUF_STAT_BUMP(hash_misses); + if (err == 0 && prefetch) { + dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE, miss, + flags & DB_RF_HAVESTRUCT); + } + DB_DNODE_EXIT(db); - /* Skip the wait per the caller's request. */ - if ((flags & DB_RF_NEVERWAIT) == 0) { - mutex_enter(&db->db_mtx); - while (db->db_state == DB_READ || - db->db_state == DB_FILL) { - ASSERT(db->db_state == DB_READ || - (flags & DB_RF_HAVESTRUCT) == 0); - DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *, - db, zio_t *, pio); - cv_wait(&db->db_changed, &db->db_mtx); - } - if (db->db_state == DB_UNCACHED) - err = SET_ERROR(EIO); - mutex_exit(&db->db_mtx); - } + /* + * If we created a zio we must execute it to avoid leaking it, even if + * it isn't attached to any work due to an error in dbuf_read_impl(). + */ + if (need_wait) { + if (err == 0) + err = zio_wait(pio); + else + (void) zio_wait(pio); + pio = NULL; } +done: + if (miss) + DBUF_STAT_BUMP(hash_misses); + else + DBUF_STAT_BUMP(hash_hits); if (pio && err != 0) { zio_t *zio = zio_null(pio, pio->io_spa, NULL, NULL, NULL, ZIO_FLAG_CANFAIL); zio->io_error = err; zio_nowait(zio); } return (err); } static void dbuf_noread(dmu_buf_impl_t *db) { ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); if (db->db_state == DB_UNCACHED) { ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); dbuf_set_data(db, dbuf_alloc_arcbuf(db)); db->db_state = DB_FILL; DTRACE_SET_STATE(db, "assigning filled buffer"); } else if (db->db_state == DB_NOFILL) { dbuf_clear_data(db); } else { ASSERT3U(db->db_state, ==, DB_CACHED); } mutex_exit(&db->db_mtx); } void dbuf_unoverride(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; blkptr_t *bp = &dr->dt.dl.dr_overridden_by; uint64_t txg = dr->dr_txg; ASSERT(MUTEX_HELD(&db->db_mtx)); /* * This assert is valid because dmu_sync() expects to be called by * a zilog's get_data while holding a range lock. This call only * comes from dbuf_dirty() callers who must also hold a range lock. */ ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC); ASSERT(db->db_level == 0); if (db->db_blkid == DMU_BONUS_BLKID || dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN) return; ASSERT(db->db_data_pending != dr); /* free this block */ if (!BP_IS_HOLE(bp) && !dr->dt.dl.dr_nopwrite) zio_free(db->db_objset->os_spa, txg, bp); if (dr->dt.dl.dr_brtwrite) { ASSERT0P(dr->dt.dl.dr_data); dr->dt.dl.dr_data = db->db_buf; } dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; dr->dt.dl.dr_nopwrite = B_FALSE; dr->dt.dl.dr_brtwrite = B_FALSE; dr->dt.dl.dr_has_raw_params = B_FALSE; /* * Release the already-written buffer, so we leave it in * a consistent dirty state. Note that all callers are * modifying the buffer, so they will immediately do * another (redundant) arc_release(). Therefore, leave * the buf thawed to save the effort of freezing & * immediately re-thawing it. */ if (dr->dt.dl.dr_data) arc_release(dr->dt.dl.dr_data, db); } /* * Evict (if its unreferenced) or clear (if its referenced) any level-0 * data blocks in the free range, so that any future readers will find * empty blocks. */ void dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid, dmu_tx_t *tx) { dmu_buf_impl_t *db_search; dmu_buf_impl_t *db, *db_next; uint64_t txg = tx->tx_txg; avl_index_t where; dbuf_dirty_record_t *dr; if (end_blkid > dn->dn_maxblkid && !(start_blkid == DMU_SPILL_BLKID || end_blkid == DMU_SPILL_BLKID)) end_blkid = dn->dn_maxblkid; dprintf_dnode(dn, "start=%llu end=%llu\n", (u_longlong_t)start_blkid, (u_longlong_t)end_blkid); db_search = kmem_alloc(sizeof (dmu_buf_impl_t), KM_SLEEP); db_search->db_level = 0; db_search->db_blkid = start_blkid; db_search->db_state = DB_SEARCH; mutex_enter(&dn->dn_dbufs_mtx); db = avl_find(&dn->dn_dbufs, db_search, &where); ASSERT3P(db, ==, NULL); db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER); for (; db != NULL; db = db_next) { db_next = AVL_NEXT(&dn->dn_dbufs, db); ASSERT(db->db_blkid != DMU_BONUS_BLKID); if (db->db_level != 0 || db->db_blkid > end_blkid) { break; } ASSERT3U(db->db_blkid, >=, start_blkid); /* found a level 0 buffer in the range */ mutex_enter(&db->db_mtx); if (dbuf_undirty(db, tx)) { /* mutex has been dropped and dbuf destroyed */ continue; } if (db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL || db->db_state == DB_EVICTING) { ASSERT(db->db.db_data == NULL); mutex_exit(&db->db_mtx); continue; } if (db->db_state == DB_READ || db->db_state == DB_FILL) { /* will be handled in dbuf_read_done or dbuf_rele */ db->db_freed_in_flight = TRUE; mutex_exit(&db->db_mtx); continue; } if (zfs_refcount_count(&db->db_holds) == 0) { ASSERT(db->db_buf); dbuf_destroy(db); continue; } /* The dbuf is referenced */ dr = list_head(&db->db_dirty_records); if (dr != NULL) { if (dr->dr_txg == txg) { /* * This buffer is "in-use", re-adjust the file * size to reflect that this buffer may * contain new data when we sync. */ if (db->db_blkid != DMU_SPILL_BLKID && db->db_blkid > dn->dn_maxblkid) dn->dn_maxblkid = db->db_blkid; dbuf_unoverride(dr); } else { /* * This dbuf is not dirty in the open context. * Either uncache it (if its not referenced in * the open context) or reset its contents to * empty. */ dbuf_fix_old_data(db, txg); } } /* clear the contents if its cached */ if (db->db_state == DB_CACHED) { ASSERT(db->db.db_data != NULL); arc_release(db->db_buf, db); rw_enter(&db->db_rwlock, RW_WRITER); memset(db->db.db_data, 0, db->db.db_size); rw_exit(&db->db_rwlock); arc_buf_freeze(db->db_buf); } mutex_exit(&db->db_mtx); } mutex_exit(&dn->dn_dbufs_mtx); kmem_free(db_search, sizeof (dmu_buf_impl_t)); } void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx) { arc_buf_t *buf, *old_buf; dbuf_dirty_record_t *dr; int osize = db->db.db_size; arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); dnode_t *dn; ASSERT(db->db_blkid != DMU_BONUS_BLKID); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* * XXX we should be doing a dbuf_read, checking the return * value and returning that up to our callers */ dmu_buf_will_dirty(&db->db, tx); /* create the data buffer for the new block */ buf = arc_alloc_buf(dn->dn_objset->os_spa, db, type, size); /* copy old block data to the new block */ old_buf = db->db_buf; memcpy(buf->b_data, old_buf->b_data, MIN(osize, size)); /* zero the remainder */ if (size > osize) memset((uint8_t *)buf->b_data + osize, 0, size - osize); mutex_enter(&db->db_mtx); dbuf_set_data(db, buf); arc_buf_destroy(old_buf, db); db->db.db_size = size; dr = list_head(&db->db_dirty_records); /* dirty record added by dmu_buf_will_dirty() */ VERIFY(dr != NULL); if (db->db_level == 0) dr->dt.dl.dr_data = buf; ASSERT3U(dr->dr_txg, ==, tx->tx_txg); ASSERT3U(dr->dr_accounted, ==, osize); dr->dr_accounted = size; mutex_exit(&db->db_mtx); dmu_objset_willuse_space(dn->dn_objset, size - osize, tx); DB_DNODE_EXIT(db); } void dbuf_release_bp(dmu_buf_impl_t *db) { objset_t *os __maybe_unused = db->db_objset; ASSERT(dsl_pool_sync_context(dmu_objset_pool(os))); ASSERT(arc_released(os->os_phys_buf) || list_link_active(&os->os_dsl_dataset->ds_synced_link)); ASSERT(db->db_parent == NULL || arc_released(db->db_parent->db_buf)); (void) arc_release(db->db_buf, db); } /* * We already have a dirty record for this TXG, and we are being * dirtied again. */ static void dbuf_redirty(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; ASSERT(MUTEX_HELD(&db->db_mtx)); if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID) { /* * If this buffer has already been written out, * we now need to reset its state. */ dbuf_unoverride(dr); if (db->db.db_object != DMU_META_DNODE_OBJECT && db->db_state != DB_NOFILL) { /* Already released on initial dirty, so just thaw. */ ASSERT(arc_released(db->db_buf)); arc_buf_thaw(db->db_buf); } } } dbuf_dirty_record_t * dbuf_dirty_lightweight(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx) { rw_enter(&dn->dn_struct_rwlock, RW_READER); IMPLY(dn->dn_objset->os_raw_receive, dn->dn_maxblkid >= blkid); dnode_new_blkid(dn, blkid, tx, B_TRUE, B_FALSE); ASSERT(dn->dn_maxblkid >= blkid); dbuf_dirty_record_t *dr = kmem_zalloc(sizeof (*dr), KM_SLEEP); list_link_init(&dr->dr_dirty_node); list_link_init(&dr->dr_dbuf_node); dr->dr_dnode = dn; dr->dr_txg = tx->tx_txg; dr->dt.dll.dr_blkid = blkid; dr->dr_accounted = dn->dn_datablksz; /* * There should not be any dbuf for the block that we're dirtying. * Otherwise the buffer contents could be inconsistent between the * dbuf and the lightweight dirty record. */ ASSERT3P(NULL, ==, dbuf_find(dn->dn_objset, dn->dn_object, 0, blkid, NULL)); mutex_enter(&dn->dn_mtx); int txgoff = tx->tx_txg & TXG_MASK; if (dn->dn_free_ranges[txgoff] != NULL) { range_tree_clear(dn->dn_free_ranges[txgoff], blkid, 1); } if (dn->dn_nlevels == 1) { ASSERT3U(blkid, <, dn->dn_nblkptr); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); rw_exit(&dn->dn_struct_rwlock); dnode_setdirty(dn, tx); } else { mutex_exit(&dn->dn_mtx); int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; dmu_buf_impl_t *parent_db = dbuf_hold_level(dn, 1, blkid >> epbs, FTAG); rw_exit(&dn->dn_struct_rwlock); if (parent_db == NULL) { kmem_free(dr, sizeof (*dr)); return (NULL); } int err = dbuf_read(parent_db, NULL, (DB_RF_NOPREFETCH | DB_RF_CANFAIL)); if (err != 0) { dbuf_rele(parent_db, FTAG); kmem_free(dr, sizeof (*dr)); return (NULL); } dbuf_dirty_record_t *parent_dr = dbuf_dirty(parent_db, tx); dbuf_rele(parent_db, FTAG); mutex_enter(&parent_dr->dt.di.dr_mtx); ASSERT3U(parent_dr->dr_txg, ==, tx->tx_txg); list_insert_tail(&parent_dr->dt.di.dr_children, dr); mutex_exit(&parent_dr->dt.di.dr_mtx); dr->dr_parent = parent_dr; } dmu_objset_willuse_space(dn->dn_objset, dr->dr_accounted, tx); return (dr); } dbuf_dirty_record_t * dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx) { dnode_t *dn; objset_t *os; dbuf_dirty_record_t *dr, *dr_next, *dr_head; int txgoff = tx->tx_txg & TXG_MASK; boolean_t drop_struct_rwlock = B_FALSE; ASSERT(tx->tx_txg != 0); ASSERT(!zfs_refcount_is_zero(&db->db_holds)); DMU_TX_DIRTY_BUF(tx, db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* * Shouldn't dirty a regular buffer in syncing context. Private * objects may be dirtied in syncing context, but only if they * were already pre-dirtied in open context. */ #ifdef ZFS_DEBUG if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); } ASSERT(!dmu_tx_is_syncing(tx) || BP_IS_HOLE(dn->dn_objset->os_rootbp) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) || dn->dn_objset->os_dsl_dataset == NULL); if (dn->dn_objset->os_dsl_dataset != NULL) rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG); #endif /* * We make this assert for private objects as well, but after we * check if we're already dirty. They are allowed to re-dirty * in syncing context. */ ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx == (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN)); mutex_enter(&db->db_mtx); /* * XXX make this true for indirects too? The problem is that * transactions created with dmu_tx_create_assigned() from * syncing context don't bother holding ahead. */ ASSERT(db->db_level != 0 || db->db_state == DB_CACHED || db->db_state == DB_FILL || db->db_state == DB_NOFILL); mutex_enter(&dn->dn_mtx); dnode_set_dirtyctx(dn, tx, db); if (tx->tx_txg > dn->dn_dirty_txg) dn->dn_dirty_txg = tx->tx_txg; mutex_exit(&dn->dn_mtx); if (db->db_blkid == DMU_SPILL_BLKID) dn->dn_have_spill = B_TRUE; /* * If this buffer is already dirty, we're done. */ dr_head = list_head(&db->db_dirty_records); ASSERT(dr_head == NULL || dr_head->dr_txg <= tx->tx_txg || db->db.db_object == DMU_META_DNODE_OBJECT); dr_next = dbuf_find_dirty_lte(db, tx->tx_txg); if (dr_next && dr_next->dr_txg == tx->tx_txg) { DB_DNODE_EXIT(db); dbuf_redirty(dr_next); mutex_exit(&db->db_mtx); return (dr_next); } /* * Only valid if not already dirty. */ ASSERT(dn->dn_object == 0 || dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx == (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN)); ASSERT3U(dn->dn_nlevels, >, db->db_level); /* * We should only be dirtying in syncing context if it's the * mos or we're initializing the os or it's a special object. * However, we are allowed to dirty in syncing context provided * we already dirtied it in open context. Hence we must make * this assertion only if we're not already dirty. */ os = dn->dn_objset; VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(os->os_spa)); #ifdef ZFS_DEBUG if (dn->dn_objset->os_dsl_dataset != NULL) rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); ASSERT(!dmu_tx_is_syncing(tx) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) || os->os_dsl_dataset == NULL || BP_IS_HOLE(os->os_rootbp)); if (dn->dn_objset->os_dsl_dataset != NULL) rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); #endif ASSERT(db->db.db_size != 0); dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size); if (db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_NOFILL) { dmu_objset_willuse_space(os, db->db.db_size, tx); } /* * If this buffer is dirty in an old transaction group we need * to make a copy of it so that the changes we make in this * transaction group won't leak out when we sync the older txg. */ dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP); list_link_init(&dr->dr_dirty_node); list_link_init(&dr->dr_dbuf_node); dr->dr_dnode = dn; if (db->db_level == 0) { void *data_old = db->db_buf; if (db->db_state != DB_NOFILL) { if (db->db_blkid == DMU_BONUS_BLKID) { dbuf_fix_old_data(db, tx->tx_txg); data_old = db->db.db_data; } else if (db->db.db_object != DMU_META_DNODE_OBJECT) { /* * Release the data buffer from the cache so * that we can modify it without impacting * possible other users of this cached data * block. Note that indirect blocks and * private objects are not released until the * syncing state (since they are only modified * then). */ arc_release(db->db_buf, db); dbuf_fix_old_data(db, tx->tx_txg); data_old = db->db_buf; } ASSERT(data_old != NULL); } dr->dt.dl.dr_data = data_old; } else { mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_NOLOCKDEP, NULL); list_create(&dr->dt.di.dr_children, sizeof (dbuf_dirty_record_t), offsetof(dbuf_dirty_record_t, dr_dirty_node)); } if (db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_NOFILL) { dr->dr_accounted = db->db.db_size; } dr->dr_dbuf = db; dr->dr_txg = tx->tx_txg; list_insert_before(&db->db_dirty_records, dr_next, dr); /* * We could have been freed_in_flight between the dbuf_noread * and dbuf_dirty. We win, as though the dbuf_noread() had * happened after the free. */ if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && db->db_blkid != DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (dn->dn_free_ranges[txgoff] != NULL) { range_tree_clear(dn->dn_free_ranges[txgoff], db->db_blkid, 1); } mutex_exit(&dn->dn_mtx); db->db_freed_in_flight = FALSE; } /* * This buffer is now part of this txg */ dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg); db->db_dirtycnt += 1; ASSERT3U(db->db_dirtycnt, <=, 3); mutex_exit(&db->db_mtx); if (db->db_blkid == DMU_BONUS_BLKID || db->db_blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); dnode_setdirty(dn, tx); DB_DNODE_EXIT(db); return (dr); } if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) { rw_enter(&dn->dn_struct_rwlock, RW_READER); drop_struct_rwlock = B_TRUE; } /* * If we are overwriting a dedup BP, then unless it is snapshotted, * when we get to syncing context we will need to decrement its * refcount in the DDT. Prefetch the relevant DDT block so that * syncing context won't have to wait for the i/o. */ if (db->db_blkptr != NULL) { db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG); ddt_prefetch(os->os_spa, db->db_blkptr); dmu_buf_unlock_parent(db, dblt, FTAG); } /* * We need to hold the dn_struct_rwlock to make this assertion, * because it protects dn_phys / dn_next_nlevels from changing. */ ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) || dn->dn_phys->dn_nlevels > db->db_level || dn->dn_next_nlevels[txgoff] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level); if (db->db_level == 0) { ASSERT(!db->db_objset->os_raw_receive || dn->dn_maxblkid >= db->db_blkid); dnode_new_blkid(dn, db->db_blkid, tx, drop_struct_rwlock, B_FALSE); ASSERT(dn->dn_maxblkid >= db->db_blkid); } if (db->db_level+1 < dn->dn_nlevels) { dmu_buf_impl_t *parent = db->db_parent; dbuf_dirty_record_t *di; int parent_held = FALSE; if (db->db_parent == NULL || db->db_parent == dn->dn_dbuf) { int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; parent = dbuf_hold_level(dn, db->db_level + 1, db->db_blkid >> epbs, FTAG); ASSERT(parent != NULL); parent_held = TRUE; } if (drop_struct_rwlock) rw_exit(&dn->dn_struct_rwlock); ASSERT3U(db->db_level + 1, ==, parent->db_level); di = dbuf_dirty(parent, tx); if (parent_held) dbuf_rele(parent, FTAG); mutex_enter(&db->db_mtx); /* * Since we've dropped the mutex, it's possible that * dbuf_undirty() might have changed this out from under us. */ if (list_head(&db->db_dirty_records) == dr || dn->dn_object == DMU_META_DNODE_OBJECT) { mutex_enter(&di->dt.di.dr_mtx); ASSERT3U(di->dr_txg, ==, tx->tx_txg); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&di->dt.di.dr_children, dr); mutex_exit(&di->dt.di.dr_mtx); dr->dr_parent = di; } mutex_exit(&db->db_mtx); } else { ASSERT(db->db_level + 1 == dn->dn_nlevels); ASSERT(db->db_blkid < dn->dn_nblkptr); ASSERT(db->db_parent == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); if (drop_struct_rwlock) rw_exit(&dn->dn_struct_rwlock); } dnode_setdirty(dn, tx); DB_DNODE_EXIT(db); return (dr); } static void dbuf_undirty_bonus(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; if (dr->dt.dl.dr_data != db->db.db_data) { struct dnode *dn = dr->dr_dnode; int max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); kmem_free(dr->dt.dl.dr_data, max_bonuslen); arc_space_return(max_bonuslen, ARC_SPACE_BONUS); } db->db_data_pending = NULL; ASSERT(list_next(&db->db_dirty_records, dr) == NULL); list_remove(&db->db_dirty_records, dr); if (dr->dr_dbuf->db_level != 0) { mutex_destroy(&dr->dt.di.dr_mtx); list_destroy(&dr->dt.di.dr_children); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); ASSERT3U(db->db_dirtycnt, >, 0); db->db_dirtycnt -= 1; } /* * Undirty a buffer in the transaction group referenced by the given * transaction. Return whether this evicted the dbuf. */ boolean_t dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx) { uint64_t txg = tx->tx_txg; boolean_t brtwrite; ASSERT(txg != 0); /* * Due to our use of dn_nlevels below, this can only be called * in open context, unless we are operating on the MOS. * From syncing context, dn_nlevels may be different from the * dn_nlevels used when dbuf was dirtied. */ ASSERT(db->db_objset == dmu_objset_pool(db->db_objset)->dp_meta_objset || txg != spa_syncing_txg(dmu_objset_spa(db->db_objset))); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT0(db->db_level); ASSERT(MUTEX_HELD(&db->db_mtx)); /* * If this buffer is not dirty, we're done. */ dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, txg); if (dr == NULL) return (B_FALSE); ASSERT(dr->dr_dbuf == db); brtwrite = dr->dt.dl.dr_brtwrite; if (brtwrite) { /* * We are freeing a block that we cloned in the same * transaction group. */ brt_pending_remove(dmu_objset_spa(db->db_objset), &dr->dt.dl.dr_overridden_by, tx); } dnode_t *dn = dr->dr_dnode; dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size); ASSERT(db->db.db_size != 0); dsl_pool_undirty_space(dmu_objset_pool(dn->dn_objset), dr->dr_accounted, txg); list_remove(&db->db_dirty_records, dr); /* * Note that there are three places in dbuf_dirty() * where this dirty record may be put on a list. * Make sure to do a list_remove corresponding to * every one of those list_insert calls. */ if (dr->dr_parent) { mutex_enter(&dr->dr_parent->dt.di.dr_mtx); list_remove(&dr->dr_parent->dt.di.dr_children, dr); mutex_exit(&dr->dr_parent->dt.di.dr_mtx); } else if (db->db_blkid == DMU_SPILL_BLKID || db->db_level + 1 == dn->dn_nlevels) { ASSERT(db->db_blkptr == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr); mutex_exit(&dn->dn_mtx); } if (db->db_state != DB_NOFILL && !brtwrite) { dbuf_unoverride(dr); ASSERT(db->db_buf != NULL); ASSERT(dr->dt.dl.dr_data != NULL); if (dr->dt.dl.dr_data != db->db_buf) arc_buf_destroy(dr->dt.dl.dr_data, db); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; if (zfs_refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) { ASSERT(db->db_state == DB_NOFILL || brtwrite || arc_released(db->db_buf)); dbuf_destroy(db); return (B_TRUE); } return (B_FALSE); } static void dmu_buf_will_dirty_impl(dmu_buf_t *db_fake, int flags, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; boolean_t undirty = B_FALSE; ASSERT(tx->tx_txg != 0); ASSERT(!zfs_refcount_is_zero(&db->db_holds)); /* * Quick check for dirtiness to improve performance for some workloads * (e.g. file deletion with indirect blocks cached). */ mutex_enter(&db->db_mtx); if (db->db_state == DB_CACHED || db->db_state == DB_NOFILL) { /* * It's possible that the dbuf is already dirty but not cached, * because there are some calls to dbuf_dirty() that don't * go through dmu_buf_will_dirty(). */ dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, tx->tx_txg); if (dr != NULL) { if (db->db_level == 0 && dr->dt.dl.dr_brtwrite) { /* * Block cloning: If we are dirtying a cloned * level 0 block, we cannot simply redirty it, * because this dr has no associated data. * We will go through a full undirtying below, * before dirtying it again. */ undirty = B_TRUE; } else { /* This dbuf is already dirty and cached. */ dbuf_redirty(dr); mutex_exit(&db->db_mtx); return; } } } mutex_exit(&db->db_mtx); DB_DNODE_ENTER(db); if (RW_WRITE_HELD(&DB_DNODE(db)->dn_struct_rwlock)) flags |= DB_RF_HAVESTRUCT; DB_DNODE_EXIT(db); /* * Block cloning: Do the dbuf_read() before undirtying the dbuf, as we * want to make sure dbuf_read() will read the pending cloned block and * not the uderlying block that is being replaced. dbuf_undirty() will * do dbuf_unoverride(), so we will end up with cloned block content, * without overridden BP. */ (void) dbuf_read(db, NULL, flags); if (undirty) { mutex_enter(&db->db_mtx); VERIFY(!dbuf_undirty(db, tx)); mutex_exit(&db->db_mtx); } (void) dbuf_dirty(db, tx); } void dmu_buf_will_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_will_dirty_impl(db_fake, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH, tx); } boolean_t dmu_buf_is_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_dirty_record_t *dr; mutex_enter(&db->db_mtx); dr = dbuf_find_dirty_eq(db, tx->tx_txg); mutex_exit(&db->db_mtx); return (dr != NULL); } void dmu_buf_will_clone(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; /* * Block cloning: We are going to clone into this block, so undirty * modifications done to this block so far in this txg. This includes * writes and clones into this block. */ mutex_enter(&db->db_mtx); DBUF_VERIFY(db); VERIFY(!dbuf_undirty(db, tx)); ASSERT0P(dbuf_find_dirty_eq(db, tx->tx_txg)); if (db->db_buf != NULL) { arc_buf_destroy(db->db_buf, db); db->db_buf = NULL; dbuf_clear_data(db); } db->db_state = DB_NOFILL; DTRACE_SET_STATE(db, "allocating NOFILL buffer for clone"); DBUF_VERIFY(db); mutex_exit(&db->db_mtx); dbuf_noread(db); (void) dbuf_dirty(db, tx); } void dmu_buf_will_not_fill(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; mutex_enter(&db->db_mtx); db->db_state = DB_NOFILL; DTRACE_SET_STATE(db, "allocating NOFILL buffer"); mutex_exit(&db->db_mtx); dbuf_noread(db); (void) dbuf_dirty(db, tx); } void dmu_buf_will_fill(dmu_buf_t *db_fake, dmu_tx_t *tx, boolean_t canfail) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(tx->tx_txg != 0); ASSERT(db->db_level == 0); ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx)); mutex_enter(&db->db_mtx); if (db->db_state == DB_NOFILL) { /* * Block cloning: We will be completely overwriting a block * cloned in this transaction group, so let's undirty the * pending clone and mark the block as uncached. This will be * as if the clone was never done. But if the fill can fail * we should have a way to return back to the cloned data. */ if (canfail && dbuf_find_dirty_eq(db, tx->tx_txg) != NULL) { mutex_exit(&db->db_mtx); dmu_buf_will_dirty(db_fake, tx); return; } VERIFY(!dbuf_undirty(db, tx)); db->db_state = DB_UNCACHED; } mutex_exit(&db->db_mtx); dbuf_noread(db); (void) dbuf_dirty(db, tx); } /* * This function is effectively the same as dmu_buf_will_dirty(), but * indicates the caller expects raw encrypted data in the db, and provides * the crypt params (byteorder, salt, iv, mac) which should be stored in the * blkptr_t when this dbuf is written. This is only used for blocks of * dnodes, during raw receive. */ void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_dirty_record_t *dr; /* * dr_has_raw_params is only processed for blocks of dnodes * (see dbuf_sync_dnode_leaf_crypt()). */ ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); ASSERT3U(db->db_level, ==, 0); ASSERT(db->db_objset->os_raw_receive); dmu_buf_will_dirty_impl(db_fake, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_NO_DECRYPT, tx); dr = dbuf_find_dirty_eq(db, tx->tx_txg); ASSERT3P(dr, !=, NULL); dr->dt.dl.dr_has_raw_params = B_TRUE; dr->dt.dl.dr_byteorder = byteorder; memcpy(dr->dt.dl.dr_salt, salt, ZIO_DATA_SALT_LEN); memcpy(dr->dt.dl.dr_iv, iv, ZIO_DATA_IV_LEN); memcpy(dr->dt.dl.dr_mac, mac, ZIO_DATA_MAC_LEN); } static void dbuf_override_impl(dmu_buf_impl_t *db, const blkptr_t *bp, dmu_tx_t *tx) { struct dirty_leaf *dl; dbuf_dirty_record_t *dr; dr = list_head(&db->db_dirty_records); ASSERT3P(dr, !=, NULL); ASSERT3U(dr->dr_txg, ==, tx->tx_txg); dl = &dr->dt.dl; dl->dr_overridden_by = *bp; dl->dr_override_state = DR_OVERRIDDEN; BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by, dr->dr_txg); } boolean_t dmu_buf_fill_done(dmu_buf_t *dbuf, dmu_tx_t *tx, boolean_t failed) { (void) tx; dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf; mutex_enter(&db->db_mtx); DBUF_VERIFY(db); if (db->db_state == DB_FILL) { if (db->db_level == 0 && db->db_freed_in_flight) { ASSERT(db->db_blkid != DMU_BONUS_BLKID); /* we were freed while filling */ /* XXX dbuf_undirty? */ memset(db->db.db_data, 0, db->db.db_size); db->db_freed_in_flight = FALSE; db->db_state = DB_CACHED; DTRACE_SET_STATE(db, "fill done handling freed in flight"); failed = B_FALSE; } else if (failed) { VERIFY(!dbuf_undirty(db, tx)); db->db_buf = NULL; dbuf_clear_data(db); DTRACE_SET_STATE(db, "fill failed"); } else { db->db_state = DB_CACHED; DTRACE_SET_STATE(db, "fill done"); } cv_broadcast(&db->db_changed); } else { db->db_state = DB_CACHED; failed = B_FALSE; } mutex_exit(&db->db_mtx); return (failed); } void dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data, bp_embedded_type_t etype, enum zio_compress comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf; struct dirty_leaf *dl; dmu_object_type_t type; dbuf_dirty_record_t *dr; if (etype == BP_EMBEDDED_TYPE_DATA) { ASSERT(spa_feature_is_active(dmu_objset_spa(db->db_objset), SPA_FEATURE_EMBEDDED_DATA)); } DB_DNODE_ENTER(db); type = DB_DNODE(db)->dn_type; DB_DNODE_EXIT(db); ASSERT0(db->db_level); ASSERT(db->db_blkid != DMU_BONUS_BLKID); dmu_buf_will_not_fill(dbuf, tx); dr = list_head(&db->db_dirty_records); ASSERT3P(dr, !=, NULL); ASSERT3U(dr->dr_txg, ==, tx->tx_txg); dl = &dr->dt.dl; encode_embedded_bp_compressed(&dl->dr_overridden_by, data, comp, uncompressed_size, compressed_size); BPE_SET_ETYPE(&dl->dr_overridden_by, etype); BP_SET_TYPE(&dl->dr_overridden_by, type); BP_SET_LEVEL(&dl->dr_overridden_by, 0); BP_SET_BYTEORDER(&dl->dr_overridden_by, byteorder); dl->dr_override_state = DR_OVERRIDDEN; BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by, dr->dr_txg); } void dmu_buf_redact(dmu_buf_t *dbuf, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf; dmu_object_type_t type; ASSERT(dsl_dataset_feature_is_active(db->db_objset->os_dsl_dataset, SPA_FEATURE_REDACTED_DATASETS)); DB_DNODE_ENTER(db); type = DB_DNODE(db)->dn_type; DB_DNODE_EXIT(db); ASSERT0(db->db_level); dmu_buf_will_not_fill(dbuf, tx); blkptr_t bp = { { { {0} } } }; BP_SET_TYPE(&bp, type); BP_SET_LEVEL(&bp, 0); BP_SET_BIRTH(&bp, tx->tx_txg, 0); BP_SET_REDACTED(&bp); BPE_SET_LSIZE(&bp, dbuf->db_size); dbuf_override_impl(db, &bp, tx); } /* * Directly assign a provided arc buf to a given dbuf if it's not referenced * by anybody except our caller. Otherwise copy arcbuf's contents to dbuf. */ void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx) { ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(db->db_level == 0); ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf)); ASSERT(buf != NULL); ASSERT3U(arc_buf_lsize(buf), ==, db->db.db_size); ASSERT(tx->tx_txg != 0); arc_return_buf(buf, db); ASSERT(arc_released(buf)); mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); ASSERT(db->db_state == DB_CACHED || db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); if (db->db_state == DB_CACHED && zfs_refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) { /* * In practice, we will never have a case where we have an * encrypted arc buffer while additional holds exist on the * dbuf. We don't handle this here so we simply assert that * fact instead. */ ASSERT(!arc_is_encrypted(buf)); mutex_exit(&db->db_mtx); (void) dbuf_dirty(db, tx); memcpy(db->db.db_data, buf->b_data, db->db.db_size); arc_buf_destroy(buf, db); return; } if (db->db_state == DB_CACHED) { dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records); ASSERT(db->db_buf != NULL); if (dr != NULL && dr->dr_txg == tx->tx_txg) { ASSERT(dr->dt.dl.dr_data == db->db_buf); if (!arc_released(db->db_buf)) { ASSERT(dr->dt.dl.dr_override_state == DR_OVERRIDDEN); arc_release(db->db_buf, db); } dr->dt.dl.dr_data = buf; arc_buf_destroy(db->db_buf, db); } else if (dr == NULL || dr->dt.dl.dr_data != db->db_buf) { arc_release(db->db_buf, db); arc_buf_destroy(db->db_buf, db); } db->db_buf = NULL; } else if (db->db_state == DB_NOFILL) { /* * We will be completely replacing the cloned block. In case * it was cloned in this transaction group, let's undirty the * pending clone and mark the block as uncached. This will be * as if the clone was never done. */ VERIFY(!dbuf_undirty(db, tx)); db->db_state = DB_UNCACHED; } ASSERT(db->db_buf == NULL); dbuf_set_data(db, buf); db->db_state = DB_FILL; DTRACE_SET_STATE(db, "filling assigned arcbuf"); mutex_exit(&db->db_mtx); (void) dbuf_dirty(db, tx); dmu_buf_fill_done(&db->db, tx, B_FALSE); } void dbuf_destroy(dmu_buf_impl_t *db) { dnode_t *dn; dmu_buf_impl_t *parent = db->db_parent; dmu_buf_impl_t *dndb; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(zfs_refcount_is_zero(&db->db_holds)); if (db->db_buf != NULL) { arc_buf_destroy(db->db_buf, db); db->db_buf = NULL; } if (db->db_blkid == DMU_BONUS_BLKID) { int slots = DB_DNODE(db)->dn_num_slots; int bonuslen = DN_SLOTS_TO_BONUSLEN(slots); if (db->db.db_data != NULL) { kmem_free(db->db.db_data, bonuslen); arc_space_return(bonuslen, ARC_SPACE_BONUS); db->db_state = DB_UNCACHED; DTRACE_SET_STATE(db, "buffer cleared"); } } dbuf_clear_data(db); if (multilist_link_active(&db->db_cache_link)) { ASSERT(db->db_caching_status == DB_DBUF_CACHE || db->db_caching_status == DB_DBUF_METADATA_CACHE); multilist_remove(&dbuf_caches[db->db_caching_status].cache, db); ASSERT0(dmu_buf_user_size(&db->db)); (void) zfs_refcount_remove_many( &dbuf_caches[db->db_caching_status].size, db->db.db_size, db); if (db->db_caching_status == DB_DBUF_METADATA_CACHE) { DBUF_STAT_BUMPDOWN(metadata_cache_count); } else { DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], db->db.db_size); } db->db_caching_status = DB_NO_CACHE; } ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); ASSERT(db->db_data_pending == NULL); ASSERT(list_is_empty(&db->db_dirty_records)); db->db_state = DB_EVICTING; DTRACE_SET_STATE(db, "buffer eviction started"); db->db_blkptr = NULL; /* * Now that db_state is DB_EVICTING, nobody else can find this via * the hash table. We can now drop db_mtx, which allows us to * acquire the dn_dbufs_mtx. */ mutex_exit(&db->db_mtx); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dndb = dn->dn_dbuf; if (db->db_blkid != DMU_BONUS_BLKID) { boolean_t needlock = !MUTEX_HELD(&dn->dn_dbufs_mtx); if (needlock) mutex_enter_nested(&dn->dn_dbufs_mtx, NESTED_SINGLE); avl_remove(&dn->dn_dbufs, db); membar_producer(); DB_DNODE_EXIT(db); if (needlock) mutex_exit(&dn->dn_dbufs_mtx); /* * Decrementing the dbuf count means that the hold corresponding * to the removed dbuf is no longer discounted in dnode_move(), * so the dnode cannot be moved until after we release the hold. * The membar_producer() ensures visibility of the decremented * value in dnode_move(), since DB_DNODE_EXIT doesn't actually * release any lock. */ mutex_enter(&dn->dn_mtx); dnode_rele_and_unlock(dn, db, B_TRUE); db->db_dnode_handle = NULL; dbuf_hash_remove(db); } else { DB_DNODE_EXIT(db); } ASSERT(zfs_refcount_is_zero(&db->db_holds)); db->db_parent = NULL; ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); ASSERT(db->db_hash_next == NULL); ASSERT(db->db_blkptr == NULL); ASSERT(db->db_data_pending == NULL); ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); ASSERT(!multilist_link_active(&db->db_cache_link)); /* * If this dbuf is referenced from an indirect dbuf, * decrement the ref count on the indirect dbuf. */ if (parent && parent != dndb) { mutex_enter(&parent->db_mtx); dbuf_rele_and_unlock(parent, db, B_TRUE); } kmem_cache_free(dbuf_kmem_cache, db); arc_space_return(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); } /* * Note: While bpp will always be updated if the function returns success, * parentp will not be updated if the dnode does not have dn_dbuf filled in; * this happens when the dnode is the meta-dnode, or {user|group|project}used * object. */ __attribute__((always_inline)) static inline int dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse, dmu_buf_impl_t **parentp, blkptr_t **bpp) { *parentp = NULL; *bpp = NULL; ASSERT(blkid != DMU_BONUS_BLKID); if (blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (dn->dn_have_spill && (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) *bpp = DN_SPILL_BLKPTR(dn->dn_phys); else *bpp = NULL; dbuf_add_ref(dn->dn_dbuf, NULL); *parentp = dn->dn_dbuf; mutex_exit(&dn->dn_mtx); return (0); } int nlevels = (dn->dn_phys->dn_nlevels == 0) ? 1 : dn->dn_phys->dn_nlevels; int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(level * epbs, <, 64); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); /* * This assertion shouldn't trip as long as the max indirect block size * is less than 1M. The reason for this is that up to that point, * the number of levels required to address an entire object with blocks * of size SPA_MINBLOCKSIZE satisfies nlevels * epbs + 1 <= 64. In * other words, if N * epbs + 1 > 64, then if (N-1) * epbs + 1 > 55 * (i.e. we can address the entire object), objects will all use at most * N-1 levels and the assertion won't overflow. However, once epbs is * 13, 4 * 13 + 1 = 53, but 5 * 13 + 1 = 66. Then, 4 levels will not be * enough to address an entire object, so objects will have 5 levels, * but then this assertion will overflow. * * All this is to say that if we ever increase DN_MAX_INDBLKSHIFT, we * need to redo this logic to handle overflows. */ ASSERT(level >= nlevels || ((nlevels - level - 1) * epbs) + highbit64(dn->dn_phys->dn_nblkptr) <= 64); if (level >= nlevels || blkid >= ((uint64_t)dn->dn_phys->dn_nblkptr << ((nlevels - level - 1) * epbs)) || (fail_sparse && blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) { /* the buffer has no parent yet */ return (SET_ERROR(ENOENT)); } else if (level < nlevels-1) { /* this block is referenced from an indirect block */ int err; err = dbuf_hold_impl(dn, level + 1, blkid >> epbs, fail_sparse, FALSE, NULL, parentp); if (err) return (err); err = dbuf_read(*parentp, NULL, (DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH | DB_RF_CANFAIL)); if (err) { dbuf_rele(*parentp, NULL); *parentp = NULL; return (err); } rw_enter(&(*parentp)->db_rwlock, RW_READER); *bpp = ((blkptr_t *)(*parentp)->db.db_data) + (blkid & ((1ULL << epbs) - 1)); if (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs))) ASSERT(BP_IS_HOLE(*bpp)); rw_exit(&(*parentp)->db_rwlock); return (0); } else { /* the block is referenced from the dnode */ ASSERT3U(level, ==, nlevels-1); ASSERT(dn->dn_phys->dn_nblkptr == 0 || blkid < dn->dn_phys->dn_nblkptr); if (dn->dn_dbuf) { dbuf_add_ref(dn->dn_dbuf, NULL); *parentp = dn->dn_dbuf; } *bpp = &dn->dn_phys->dn_blkptr[blkid]; return (0); } } static dmu_buf_impl_t * dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid, dmu_buf_impl_t *parent, blkptr_t *blkptr, uint64_t hash) { objset_t *os = dn->dn_objset; dmu_buf_impl_t *db, *odb; ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_type != DMU_OT_NONE); db = kmem_cache_alloc(dbuf_kmem_cache, KM_SLEEP); list_create(&db->db_dirty_records, sizeof (dbuf_dirty_record_t), offsetof(dbuf_dirty_record_t, dr_dbuf_node)); db->db_objset = os; db->db.db_object = dn->dn_object; db->db_level = level; db->db_blkid = blkid; db->db_dirtycnt = 0; db->db_dnode_handle = dn->dn_handle; db->db_parent = parent; db->db_blkptr = blkptr; db->db_hash = hash; db->db_user = NULL; db->db_user_immediate_evict = FALSE; db->db_freed_in_flight = FALSE; db->db_pending_evict = FALSE; if (blkid == DMU_BONUS_BLKID) { ASSERT3P(parent, ==, dn->dn_dbuf); db->db.db_size = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots) - (dn->dn_nblkptr-1) * sizeof (blkptr_t); ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen); db->db.db_offset = DMU_BONUS_BLKID; db->db_state = DB_UNCACHED; DTRACE_SET_STATE(db, "bonus buffer created"); db->db_caching_status = DB_NO_CACHE; /* the bonus dbuf is not placed in the hash table */ arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); return (db); } else if (blkid == DMU_SPILL_BLKID) { db->db.db_size = (blkptr != NULL) ? BP_GET_LSIZE(blkptr) : SPA_MINBLOCKSIZE; db->db.db_offset = 0; } else { int blocksize = db->db_level ? 1 << dn->dn_indblkshift : dn->dn_datablksz; db->db.db_size = blocksize; db->db.db_offset = db->db_blkid * blocksize; } /* * Hold the dn_dbufs_mtx while we get the new dbuf * in the hash table *and* added to the dbufs list. * This prevents a possible deadlock with someone * trying to look up this dbuf before it's added to the * dn_dbufs list. */ mutex_enter(&dn->dn_dbufs_mtx); db->db_state = DB_EVICTING; /* not worth logging this state change */ if ((odb = dbuf_hash_insert(db)) != NULL) { /* someone else inserted it first */ mutex_exit(&dn->dn_dbufs_mtx); kmem_cache_free(dbuf_kmem_cache, db); DBUF_STAT_BUMP(hash_insert_race); return (odb); } avl_add(&dn->dn_dbufs, db); db->db_state = DB_UNCACHED; DTRACE_SET_STATE(db, "regular buffer created"); db->db_caching_status = DB_NO_CACHE; mutex_exit(&dn->dn_dbufs_mtx); arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); if (parent && parent != dn->dn_dbuf) dbuf_add_ref(parent, db); ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || zfs_refcount_count(&dn->dn_holds) > 0); (void) zfs_refcount_add(&dn->dn_holds, db); dprintf_dbuf(db, "db=%p\n", db); return (db); } /* * This function returns a block pointer and information about the object, * given a dnode and a block. This is a publicly accessible version of * dbuf_findbp that only returns some information, rather than the * dbuf. Note that the dnode passed in must be held, and the dn_struct_rwlock * should be locked as (at least) a reader. */ int dbuf_dnode_findbp(dnode_t *dn, uint64_t level, uint64_t blkid, blkptr_t *bp, uint16_t *datablkszsec, uint8_t *indblkshift) { dmu_buf_impl_t *dbp = NULL; blkptr_t *bp2; int err = 0; ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); err = dbuf_findbp(dn, level, blkid, B_FALSE, &dbp, &bp2); if (err == 0) { ASSERT3P(bp2, !=, NULL); *bp = *bp2; if (dbp != NULL) dbuf_rele(dbp, NULL); if (datablkszsec != NULL) *datablkszsec = dn->dn_phys->dn_datablkszsec; if (indblkshift != NULL) *indblkshift = dn->dn_phys->dn_indblkshift; } return (err); } typedef struct dbuf_prefetch_arg { spa_t *dpa_spa; /* The spa to issue the prefetch in. */ zbookmark_phys_t dpa_zb; /* The target block to prefetch. */ int dpa_epbs; /* Entries (blkptr_t's) Per Block Shift. */ int dpa_curlevel; /* The current level that we're reading */ dnode_t *dpa_dnode; /* The dnode associated with the prefetch */ zio_priority_t dpa_prio; /* The priority I/Os should be issued at. */ zio_t *dpa_zio; /* The parent zio_t for all prefetches. */ arc_flags_t dpa_aflags; /* Flags to pass to the final prefetch. */ dbuf_prefetch_fn dpa_cb; /* prefetch completion callback */ void *dpa_arg; /* prefetch completion arg */ } dbuf_prefetch_arg_t; static void dbuf_prefetch_fini(dbuf_prefetch_arg_t *dpa, boolean_t io_done) { if (dpa->dpa_cb != NULL) { dpa->dpa_cb(dpa->dpa_arg, dpa->dpa_zb.zb_level, dpa->dpa_zb.zb_blkid, io_done); } kmem_free(dpa, sizeof (*dpa)); } static void dbuf_issue_final_prefetch_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *iobp, arc_buf_t *abuf, void *private) { (void) zio, (void) zb, (void) iobp; dbuf_prefetch_arg_t *dpa = private; if (abuf != NULL) arc_buf_destroy(abuf, private); dbuf_prefetch_fini(dpa, B_TRUE); } /* * Actually issue the prefetch read for the block given. */ static void dbuf_issue_final_prefetch(dbuf_prefetch_arg_t *dpa, blkptr_t *bp) { ASSERT(!BP_IS_REDACTED(bp) || dsl_dataset_feature_is_active( dpa->dpa_dnode->dn_objset->os_dsl_dataset, SPA_FEATURE_REDACTED_DATASETS)); if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp) || BP_IS_REDACTED(bp)) return (dbuf_prefetch_fini(dpa, B_FALSE)); int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE; arc_flags_t aflags = dpa->dpa_aflags | ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH | ARC_FLAG_NO_BUF; /* dnodes are always read as raw and then converted later */ if (BP_GET_TYPE(bp) == DMU_OT_DNODE && BP_IS_PROTECTED(bp) && dpa->dpa_curlevel == 0) zio_flags |= ZIO_FLAG_RAW; ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp)); ASSERT3U(dpa->dpa_curlevel, ==, dpa->dpa_zb.zb_level); ASSERT(dpa->dpa_zio != NULL); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp, dbuf_issue_final_prefetch_done, dpa, dpa->dpa_prio, zio_flags, &aflags, &dpa->dpa_zb); } /* * Called when an indirect block above our prefetch target is read in. This * will either read in the next indirect block down the tree or issue the actual * prefetch if the next block down is our target. */ static void dbuf_prefetch_indirect_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *iobp, arc_buf_t *abuf, void *private) { (void) zb, (void) iobp; dbuf_prefetch_arg_t *dpa = private; ASSERT3S(dpa->dpa_zb.zb_level, <, dpa->dpa_curlevel); ASSERT3S(dpa->dpa_curlevel, >, 0); if (abuf == NULL) { ASSERT(zio == NULL || zio->io_error != 0); dbuf_prefetch_fini(dpa, B_TRUE); return; } ASSERT(zio == NULL || zio->io_error == 0); /* * The dpa_dnode is only valid if we are called with a NULL * zio. This indicates that the arc_read() returned without * first calling zio_read() to issue a physical read. Once * a physical read is made the dpa_dnode must be invalidated * as the locks guarding it may have been dropped. If the * dpa_dnode is still valid, then we want to add it to the dbuf * cache. To do so, we must hold the dbuf associated with the block * we just prefetched, read its contents so that we associate it * with an arc_buf_t, and then release it. */ if (zio != NULL) { ASSERT3S(BP_GET_LEVEL(zio->io_bp), ==, dpa->dpa_curlevel); if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) { ASSERT3U(BP_GET_PSIZE(zio->io_bp), ==, zio->io_size); } else { ASSERT3U(BP_GET_LSIZE(zio->io_bp), ==, zio->io_size); } ASSERT3P(zio->io_spa, ==, dpa->dpa_spa); dpa->dpa_dnode = NULL; } else if (dpa->dpa_dnode != NULL) { uint64_t curblkid = dpa->dpa_zb.zb_blkid >> (dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level)); dmu_buf_impl_t *db = dbuf_hold_level(dpa->dpa_dnode, dpa->dpa_curlevel, curblkid, FTAG); if (db == NULL) { arc_buf_destroy(abuf, private); dbuf_prefetch_fini(dpa, B_TRUE); return; } (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_HAVESTRUCT); dbuf_rele(db, FTAG); } dpa->dpa_curlevel--; uint64_t nextblkid = dpa->dpa_zb.zb_blkid >> (dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level)); blkptr_t *bp = ((blkptr_t *)abuf->b_data) + P2PHASE(nextblkid, 1ULL << dpa->dpa_epbs); ASSERT(!BP_IS_REDACTED(bp) || (dpa->dpa_dnode && dsl_dataset_feature_is_active( dpa->dpa_dnode->dn_objset->os_dsl_dataset, SPA_FEATURE_REDACTED_DATASETS))); if (BP_IS_HOLE(bp) || BP_IS_REDACTED(bp)) { arc_buf_destroy(abuf, private); dbuf_prefetch_fini(dpa, B_TRUE); return; } else if (dpa->dpa_curlevel == dpa->dpa_zb.zb_level) { ASSERT3U(nextblkid, ==, dpa->dpa_zb.zb_blkid); dbuf_issue_final_prefetch(dpa, bp); } else { arc_flags_t iter_aflags = ARC_FLAG_NOWAIT; zbookmark_phys_t zb; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (dpa->dpa_aflags & ARC_FLAG_L2CACHE) iter_aflags |= ARC_FLAG_L2CACHE; ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp)); SET_BOOKMARK(&zb, dpa->dpa_zb.zb_objset, dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp, dbuf_prefetch_indirect_done, dpa, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &iter_aflags, &zb); } arc_buf_destroy(abuf, private); } /* * Issue prefetch reads for the given block on the given level. If the indirect * blocks above that block are not in memory, we will read them in * asynchronously. As a result, this call never blocks waiting for a read to * complete. Note that the prefetch might fail if the dataset is encrypted and * the encryption key is unmapped before the IO completes. */ int dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio, arc_flags_t aflags, dbuf_prefetch_fn cb, void *arg) { blkptr_t bp; int epbs, nlevels, curlevel; uint64_t curblkid; ASSERT(blkid != DMU_BONUS_BLKID); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); if (blkid > dn->dn_maxblkid) goto no_issue; if (level == 0 && dnode_block_freed(dn, blkid)) goto no_issue; /* * This dnode hasn't been written to disk yet, so there's nothing to * prefetch. */ nlevels = dn->dn_phys->dn_nlevels; if (level >= nlevels || dn->dn_phys->dn_nblkptr == 0) goto no_issue; epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; if (dn->dn_phys->dn_maxblkid < blkid << (epbs * level)) goto no_issue; dmu_buf_impl_t *db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid, NULL); if (db != NULL) { mutex_exit(&db->db_mtx); /* * This dbuf already exists. It is either CACHED, or * (we assume) about to be read or filled. */ goto no_issue; } /* * Find the closest ancestor (indirect block) of the target block * that is present in the cache. In this indirect block, we will * find the bp that is at curlevel, curblkid. */ curlevel = level; curblkid = blkid; while (curlevel < nlevels - 1) { int parent_level = curlevel + 1; uint64_t parent_blkid = curblkid >> epbs; dmu_buf_impl_t *db; if (dbuf_hold_impl(dn, parent_level, parent_blkid, FALSE, TRUE, FTAG, &db) == 0) { blkptr_t *bpp = db->db_buf->b_data; bp = bpp[P2PHASE(curblkid, 1 << epbs)]; dbuf_rele(db, FTAG); break; } curlevel = parent_level; curblkid = parent_blkid; } if (curlevel == nlevels - 1) { /* No cached indirect blocks found. */ ASSERT3U(curblkid, <, dn->dn_phys->dn_nblkptr); bp = dn->dn_phys->dn_blkptr[curblkid]; } ASSERT(!BP_IS_REDACTED(&bp) || dsl_dataset_feature_is_active(dn->dn_objset->os_dsl_dataset, SPA_FEATURE_REDACTED_DATASETS)); if (BP_IS_HOLE(&bp) || BP_IS_REDACTED(&bp)) goto no_issue; ASSERT3U(curlevel, ==, BP_GET_LEVEL(&bp)); zio_t *pio = zio_root(dmu_objset_spa(dn->dn_objset), NULL, NULL, ZIO_FLAG_CANFAIL); dbuf_prefetch_arg_t *dpa = kmem_zalloc(sizeof (*dpa), KM_SLEEP); dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset; SET_BOOKMARK(&dpa->dpa_zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, dn->dn_object, level, blkid); dpa->dpa_curlevel = curlevel; dpa->dpa_prio = prio; dpa->dpa_aflags = aflags; dpa->dpa_spa = dn->dn_objset->os_spa; dpa->dpa_dnode = dn; dpa->dpa_epbs = epbs; dpa->dpa_zio = pio; dpa->dpa_cb = cb; dpa->dpa_arg = arg; if (!DNODE_LEVEL_IS_CACHEABLE(dn, level)) dpa->dpa_aflags |= ARC_FLAG_UNCACHED; else if (dnode_level_is_l2cacheable(&bp, dn, level)) dpa->dpa_aflags |= ARC_FLAG_L2CACHE; /* * If we have the indirect just above us, no need to do the asynchronous * prefetch chain; we'll just run the last step ourselves. If we're at * a higher level, though, we want to issue the prefetches for all the * indirect blocks asynchronously, so we can go on with whatever we were * doing. */ if (curlevel == level) { ASSERT3U(curblkid, ==, blkid); dbuf_issue_final_prefetch(dpa, &bp); } else { arc_flags_t iter_aflags = ARC_FLAG_NOWAIT; zbookmark_phys_t zb; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (dnode_level_is_l2cacheable(&bp, dn, level)) iter_aflags |= ARC_FLAG_L2CACHE; SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, dn->dn_object, curlevel, curblkid); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, &bp, dbuf_prefetch_indirect_done, dpa, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &iter_aflags, &zb); } /* * We use pio here instead of dpa_zio since it's possible that * dpa may have already been freed. */ zio_nowait(pio); return (1); no_issue: if (cb != NULL) cb(arg, level, blkid, B_FALSE); return (0); } int dbuf_prefetch(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio, arc_flags_t aflags) { return (dbuf_prefetch_impl(dn, level, blkid, prio, aflags, NULL, NULL)); } /* * Helper function for dbuf_hold_impl() to copy a buffer. Handles * the case of encrypted, compressed and uncompressed buffers by * allocating the new buffer, respectively, with arc_alloc_raw_buf(), * arc_alloc_compressed_buf() or arc_alloc_buf().* * * NOTE: Declared noinline to avoid stack bloat in dbuf_hold_impl(). */ noinline static void dbuf_hold_copy(dnode_t *dn, dmu_buf_impl_t *db) { dbuf_dirty_record_t *dr = db->db_data_pending; arc_buf_t *data = dr->dt.dl.dr_data; enum zio_compress compress_type = arc_get_compression(data); uint8_t complevel = arc_get_complevel(data); if (arc_is_encrypted(data)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(data, &byteorder, salt, iv, mac); dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db, dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac, dn->dn_type, arc_buf_size(data), arc_buf_lsize(data), compress_type, complevel)); } else if (compress_type != ZIO_COMPRESS_OFF) { dbuf_set_data(db, arc_alloc_compressed_buf( dn->dn_objset->os_spa, db, arc_buf_size(data), arc_buf_lsize(data), compress_type, complevel)); } else { dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size)); } rw_enter(&db->db_rwlock, RW_WRITER); memcpy(db->db.db_data, data->b_data, arc_buf_size(data)); rw_exit(&db->db_rwlock); } /* * Returns with db_holds incremented, and db_mtx not held. * Note: dn_struct_rwlock must be held. */ int dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, const void *tag, dmu_buf_impl_t **dbp) { dmu_buf_impl_t *db, *parent = NULL; uint64_t hv; /* If the pool has been created, verify the tx_sync_lock is not held */ spa_t *spa = dn->dn_objset->os_spa; dsl_pool_t *dp = spa->spa_dsl_pool; if (dp != NULL) { ASSERT(!MUTEX_HELD(&dp->dp_tx.tx_sync_lock)); } ASSERT(blkid != DMU_BONUS_BLKID); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT3U(dn->dn_nlevels, >, level); *dbp = NULL; /* dbuf_find() returns with db_mtx held */ db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid, &hv); if (db == NULL) { blkptr_t *bp = NULL; int err; if (fail_uncached) return (SET_ERROR(ENOENT)); ASSERT3P(parent, ==, NULL); err = dbuf_findbp(dn, level, blkid, fail_sparse, &parent, &bp); if (fail_sparse) { if (err == 0 && bp && BP_IS_HOLE(bp)) err = SET_ERROR(ENOENT); if (err) { if (parent) dbuf_rele(parent, NULL); return (err); } } if (err && err != ENOENT) return (err); db = dbuf_create(dn, level, blkid, parent, bp, hv); } if (fail_uncached && db->db_state != DB_CACHED) { mutex_exit(&db->db_mtx); return (SET_ERROR(ENOENT)); } if (db->db_buf != NULL) { arc_buf_access(db->db_buf); ASSERT3P(db->db.db_data, ==, db->db_buf->b_data); } ASSERT(db->db_buf == NULL || arc_referenced(db->db_buf)); /* * If this buffer is currently syncing out, and we are * still referencing it from db_data, we need to make a copy * of it in case we decide we want to dirty it again in this txg. */ if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && dn->dn_object != DMU_META_DNODE_OBJECT && db->db_state == DB_CACHED && db->db_data_pending) { dbuf_dirty_record_t *dr = db->db_data_pending; if (dr->dt.dl.dr_data == db->db_buf) { ASSERT3P(db->db_buf, !=, NULL); dbuf_hold_copy(dn, db); } } if (multilist_link_active(&db->db_cache_link)) { ASSERT(zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db_caching_status == DB_DBUF_CACHE || db->db_caching_status == DB_DBUF_METADATA_CACHE); multilist_remove(&dbuf_caches[db->db_caching_status].cache, db); uint64_t size = db->db.db_size + dmu_buf_user_size(&db->db); (void) zfs_refcount_remove_many( &dbuf_caches[db->db_caching_status].size, size, db); if (db->db_caching_status == DB_DBUF_METADATA_CACHE) { DBUF_STAT_BUMPDOWN(metadata_cache_count); } else { DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size); } db->db_caching_status = DB_NO_CACHE; } (void) zfs_refcount_add(&db->db_holds, tag); DBUF_VERIFY(db); mutex_exit(&db->db_mtx); /* NOTE: we can't rele the parent until after we drop the db_mtx */ if (parent) dbuf_rele(parent, NULL); ASSERT3P(DB_DNODE(db), ==, dn); ASSERT3U(db->db_blkid, ==, blkid); ASSERT3U(db->db_level, ==, level); *dbp = db; return (0); } dmu_buf_impl_t * dbuf_hold(dnode_t *dn, uint64_t blkid, const void *tag) { return (dbuf_hold_level(dn, 0, blkid, tag)); } dmu_buf_impl_t * dbuf_hold_level(dnode_t *dn, int level, uint64_t blkid, const void *tag) { dmu_buf_impl_t *db; int err = dbuf_hold_impl(dn, level, blkid, FALSE, FALSE, tag, &db); return (err ? NULL : db); } void dbuf_create_bonus(dnode_t *dn) { ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_bonus == NULL); dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL, dbuf_hash(dn->dn_objset, dn->dn_object, 0, DMU_BONUS_BLKID)); } int dbuf_spill_set_blksz(dmu_buf_t *db_fake, uint64_t blksz, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; if (db->db_blkid != DMU_SPILL_BLKID) return (SET_ERROR(ENOTSUP)); if (blksz == 0) blksz = SPA_MINBLOCKSIZE; ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset))); blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE); dbuf_new_size(db, blksz, tx); return (0); } void dbuf_rm_spill(dnode_t *dn, dmu_tx_t *tx) { dbuf_free_range(dn, DMU_SPILL_BLKID, DMU_SPILL_BLKID, tx); } #pragma weak dmu_buf_add_ref = dbuf_add_ref void dbuf_add_ref(dmu_buf_impl_t *db, const void *tag) { int64_t holds = zfs_refcount_add(&db->db_holds, tag); VERIFY3S(holds, >, 1); } #pragma weak dmu_buf_try_add_ref = dbuf_try_add_ref boolean_t dbuf_try_add_ref(dmu_buf_t *db_fake, objset_t *os, uint64_t obj, uint64_t blkid, const void *tag) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dmu_buf_impl_t *found_db; boolean_t result = B_FALSE; if (blkid == DMU_BONUS_BLKID) found_db = dbuf_find_bonus(os, obj); else found_db = dbuf_find(os, obj, 0, blkid, NULL); if (found_db != NULL) { if (db == found_db && dbuf_refcount(db) > db->db_dirtycnt) { (void) zfs_refcount_add(&db->db_holds, tag); result = B_TRUE; } mutex_exit(&found_db->db_mtx); } return (result); } /* * If you call dbuf_rele() you had better not be referencing the dnode handle * unless you have some other direct or indirect hold on the dnode. (An indirect * hold is a hold on one of the dnode's dbufs, including the bonus buffer.) * Without that, the dbuf_rele() could lead to a dnode_rele() followed by the * dnode's parent dbuf evicting its dnode handles. */ void dbuf_rele(dmu_buf_impl_t *db, const void *tag) { mutex_enter(&db->db_mtx); dbuf_rele_and_unlock(db, tag, B_FALSE); } void dmu_buf_rele(dmu_buf_t *db, const void *tag) { dbuf_rele((dmu_buf_impl_t *)db, tag); } /* * dbuf_rele() for an already-locked dbuf. This is necessary to allow * db_dirtycnt and db_holds to be updated atomically. The 'evicting' * argument should be set if we are already in the dbuf-evicting code * path, in which case we don't want to recursively evict. This allows us to * avoid deeply nested stacks that would have a call flow similar to this: * * dbuf_rele()-->dbuf_rele_and_unlock()-->dbuf_evict_notify() * ^ | * | | * +-----dbuf_destroy()<--dbuf_evict_one()<--------+ * */ void dbuf_rele_and_unlock(dmu_buf_impl_t *db, const void *tag, boolean_t evicting) { int64_t holds; uint64_t size; ASSERT(MUTEX_HELD(&db->db_mtx)); DBUF_VERIFY(db); /* * Remove the reference to the dbuf before removing its hold on the * dnode so we can guarantee in dnode_move() that a referenced bonus * buffer has a corresponding dnode hold. */ holds = zfs_refcount_remove(&db->db_holds, tag); ASSERT(holds >= 0); /* * We can't freeze indirects if there is a possibility that they * may be modified in the current syncing context. */ if (db->db_buf != NULL && holds == (db->db_level == 0 ? db->db_dirtycnt : 0)) { arc_buf_freeze(db->db_buf); } if (holds == db->db_dirtycnt && db->db_level == 0 && db->db_user_immediate_evict) dbuf_evict_user(db); if (holds == 0) { if (db->db_blkid == DMU_BONUS_BLKID) { dnode_t *dn; boolean_t evict_dbuf = db->db_pending_evict; /* * If the dnode moves here, we cannot cross this * barrier until the move completes. */ DB_DNODE_ENTER(db); dn = DB_DNODE(db); atomic_dec_32(&dn->dn_dbufs_count); /* * Decrementing the dbuf count means that the bonus * buffer's dnode hold is no longer discounted in * dnode_move(). The dnode cannot move until after * the dnode_rele() below. */ DB_DNODE_EXIT(db); /* * Do not reference db after its lock is dropped. * Another thread may evict it. */ mutex_exit(&db->db_mtx); if (evict_dbuf) dnode_evict_bonus(dn); dnode_rele(dn, db); } else if (db->db_buf == NULL) { /* * This is a special case: we never associated this * dbuf with any data allocated from the ARC. */ ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); dbuf_destroy(db); } else if (arc_released(db->db_buf)) { /* * This dbuf has anonymous data associated with it. */ dbuf_destroy(db); } else if (!(DBUF_IS_CACHEABLE(db) || db->db_partial_read) || db->db_pending_evict) { dbuf_destroy(db); } else if (!multilist_link_active(&db->db_cache_link)) { ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); dbuf_cached_state_t dcs = dbuf_include_in_metadata_cache(db) ? DB_DBUF_METADATA_CACHE : DB_DBUF_CACHE; db->db_caching_status = dcs; multilist_insert(&dbuf_caches[dcs].cache, db); uint64_t db_size = db->db.db_size + dmu_buf_user_size(&db->db); size = zfs_refcount_add_many( &dbuf_caches[dcs].size, db_size, db); uint8_t db_level = db->db_level; mutex_exit(&db->db_mtx); if (dcs == DB_DBUF_METADATA_CACHE) { DBUF_STAT_BUMP(metadata_cache_count); DBUF_STAT_MAX(metadata_cache_size_bytes_max, size); } else { DBUF_STAT_BUMP(cache_count); DBUF_STAT_MAX(cache_size_bytes_max, size); DBUF_STAT_BUMP(cache_levels[db_level]); DBUF_STAT_INCR(cache_levels_bytes[db_level], db_size); } if (dcs == DB_DBUF_CACHE && !evicting) dbuf_evict_notify(size); } } else { mutex_exit(&db->db_mtx); } } #pragma weak dmu_buf_refcount = dbuf_refcount uint64_t dbuf_refcount(dmu_buf_impl_t *db) { return (zfs_refcount_count(&db->db_holds)); } uint64_t dmu_buf_user_refcount(dmu_buf_t *db_fake) { uint64_t holds; dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; mutex_enter(&db->db_mtx); ASSERT3U(zfs_refcount_count(&db->db_holds), >=, db->db_dirtycnt); holds = zfs_refcount_count(&db->db_holds) - db->db_dirtycnt; mutex_exit(&db->db_mtx); return (holds); } void * dmu_buf_replace_user(dmu_buf_t *db_fake, dmu_buf_user_t *old_user, dmu_buf_user_t *new_user) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; mutex_enter(&db->db_mtx); dbuf_verify_user(db, DBVU_NOT_EVICTING); if (db->db_user == old_user) db->db_user = new_user; else old_user = db->db_user; dbuf_verify_user(db, DBVU_NOT_EVICTING); mutex_exit(&db->db_mtx); return (old_user); } void * dmu_buf_set_user(dmu_buf_t *db_fake, dmu_buf_user_t *user) { return (dmu_buf_replace_user(db_fake, NULL, user)); } void * dmu_buf_set_user_ie(dmu_buf_t *db_fake, dmu_buf_user_t *user) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; db->db_user_immediate_evict = TRUE; return (dmu_buf_set_user(db_fake, user)); } void * dmu_buf_remove_user(dmu_buf_t *db_fake, dmu_buf_user_t *user) { return (dmu_buf_replace_user(db_fake, user, NULL)); } void * dmu_buf_get_user(dmu_buf_t *db_fake) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_verify_user(db, DBVU_NOT_EVICTING); return (db->db_user); } uint64_t dmu_buf_user_size(dmu_buf_t *db_fake) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; if (db->db_user == NULL) return (0); return (atomic_load_64(&db->db_user->dbu_size)); } void dmu_buf_add_user_size(dmu_buf_t *db_fake, uint64_t nadd) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); ASSERT3P(db->db_user, !=, NULL); ASSERT3U(atomic_load_64(&db->db_user->dbu_size), <, UINT64_MAX - nadd); atomic_add_64(&db->db_user->dbu_size, nadd); } void dmu_buf_sub_user_size(dmu_buf_t *db_fake, uint64_t nsub) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); ASSERT3P(db->db_user, !=, NULL); ASSERT3U(atomic_load_64(&db->db_user->dbu_size), >=, nsub); atomic_sub_64(&db->db_user->dbu_size, nsub); } void dmu_buf_user_evict_wait(void) { taskq_wait(dbu_evict_taskq); } blkptr_t * dmu_buf_get_blkptr(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; return (dbi->db_blkptr); } objset_t * dmu_buf_get_objset(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; return (dbi->db_objset); } static void dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db) { /* ASSERT(dmu_tx_is_syncing(tx) */ ASSERT(MUTEX_HELD(&db->db_mtx)); if (db->db_blkptr != NULL) return; if (db->db_blkid == DMU_SPILL_BLKID) { db->db_blkptr = DN_SPILL_BLKPTR(dn->dn_phys); BP_ZERO(db->db_blkptr); return; } if (db->db_level == dn->dn_phys->dn_nlevels-1) { /* * This buffer was allocated at a time when there was * no available blkptrs from the dnode, or it was * inappropriate to hook it in (i.e., nlevels mismatch). */ ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr); ASSERT(db->db_parent == NULL); db->db_parent = dn->dn_dbuf; db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid]; DBUF_VERIFY(db); } else { dmu_buf_impl_t *parent = db->db_parent; int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT(dn->dn_phys->dn_nlevels > 1); if (parent == NULL) { mutex_exit(&db->db_mtx); rw_enter(&dn->dn_struct_rwlock, RW_READER); parent = dbuf_hold_level(dn, db->db_level + 1, db->db_blkid >> epbs, db); rw_exit(&dn->dn_struct_rwlock); mutex_enter(&db->db_mtx); db->db_parent = parent; } db->db_blkptr = (blkptr_t *)parent->db.db_data + (db->db_blkid & ((1ULL << epbs) - 1)); DBUF_VERIFY(db); } } static void dbuf_sync_bonus(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; void *data = dr->dt.dl.dr_data; ASSERT0(db->db_level); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db_blkid == DMU_BONUS_BLKID); ASSERT(data != NULL); dnode_t *dn = dr->dr_dnode; ASSERT3U(DN_MAX_BONUS_LEN(dn->dn_phys), <=, DN_SLOTS_TO_BONUSLEN(dn->dn_phys->dn_extra_slots + 1)); memcpy(DN_BONUS(dn->dn_phys), data, DN_MAX_BONUS_LEN(dn->dn_phys)); dbuf_sync_leaf_verify_bonus_dnode(dr); dbuf_undirty_bonus(dr); dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE); } /* * When syncing out a blocks of dnodes, adjust the block to deal with * encryption. Normally, we make sure the block is decrypted before writing * it. If we have crypt params, then we are writing a raw (encrypted) block, * from a raw receive. In this case, set the ARC buf's crypt params so * that the BP will be filled with the correct byteorder, salt, iv, and mac. */ static void dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr) { int err; dmu_buf_impl_t *db = dr->dr_dbuf; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); ASSERT3U(db->db_level, ==, 0); if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) { zbookmark_phys_t zb; /* * Unfortunately, there is currently no mechanism for * syncing context to handle decryption errors. An error * here is only possible if an attacker maliciously * changed a dnode block and updated the associated * checksums going up the block tree. */ SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); err = arc_untransform(db->db_buf, db->db_objset->os_spa, &zb, B_TRUE); if (err) panic("Invalid dnode block MAC"); } else if (dr->dt.dl.dr_has_raw_params) { (void) arc_release(dr->dt.dl.dr_data, db); arc_convert_to_raw(dr->dt.dl.dr_data, dmu_objset_id(db->db_objset), dr->dt.dl.dr_byteorder, DMU_OT_DNODE, dr->dt.dl.dr_salt, dr->dt.dl.dr_iv, dr->dt.dl.dr_mac); } } /* * dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it * is critical the we not allow the compiler to inline this function in to * dbuf_sync_list() thereby drastically bloating the stack usage. */ noinline static void dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn = dr->dr_dnode; ASSERT(dmu_tx_is_syncing(tx)); dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr); mutex_enter(&db->db_mtx); ASSERT(db->db_level > 0); DBUF_VERIFY(db); /* Read the block if it hasn't been read yet. */ if (db->db_buf == NULL) { mutex_exit(&db->db_mtx); (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED); mutex_enter(&db->db_mtx); } ASSERT3U(db->db_state, ==, DB_CACHED); ASSERT(db->db_buf != NULL); /* Indirect block size must match what the dnode thinks it is. */ ASSERT3U(db->db.db_size, ==, 1<dn_phys->dn_indblkshift); dbuf_check_blkptr(dn, db); /* Provide the pending dirty record to child dbufs */ db->db_data_pending = dr; mutex_exit(&db->db_mtx); dbuf_write(dr, db->db_buf, tx); zio_t *zio = dr->dr_zio; mutex_enter(&dr->dt.di.dr_mtx); dbuf_sync_list(&dr->dt.di.dr_children, db->db_level - 1, tx); ASSERT(list_head(&dr->dt.di.dr_children) == NULL); mutex_exit(&dr->dt.di.dr_mtx); zio_nowait(zio); } /* * Verify that the size of the data in our bonus buffer does not exceed * its recorded size. * * The purpose of this verification is to catch any cases in development * where the size of a phys structure (i.e space_map_phys_t) grows and, * due to incorrect feature management, older pools expect to read more * data even though they didn't actually write it to begin with. * * For a example, this would catch an error in the feature logic where we * open an older pool and we expect to write the space map histogram of * a space map with size SPACE_MAP_SIZE_V0. */ static void dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr) { #ifdef ZFS_DEBUG dnode_t *dn = dr->dr_dnode; /* * Encrypted bonus buffers can have data past their bonuslen. * Skip the verification of these blocks. */ if (DMU_OT_IS_ENCRYPTED(dn->dn_bonustype)) return; uint16_t bonuslen = dn->dn_phys->dn_bonuslen; uint16_t maxbonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); ASSERT3U(bonuslen, <=, maxbonuslen); arc_buf_t *datap = dr->dt.dl.dr_data; char *datap_end = ((char *)datap) + bonuslen; char *datap_max = ((char *)datap) + maxbonuslen; /* ensure that everything is zero after our data */ for (; datap_end < datap_max; datap_end++) ASSERT(*datap_end == 0); #endif } static blkptr_t * dbuf_lightweight_bp(dbuf_dirty_record_t *dr) { /* This must be a lightweight dirty record. */ ASSERT3P(dr->dr_dbuf, ==, NULL); dnode_t *dn = dr->dr_dnode; if (dn->dn_phys->dn_nlevels == 1) { VERIFY3U(dr->dt.dll.dr_blkid, <, dn->dn_phys->dn_nblkptr); return (&dn->dn_phys->dn_blkptr[dr->dt.dll.dr_blkid]); } else { dmu_buf_impl_t *parent_db = dr->dr_parent->dr_dbuf; int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; VERIFY3U(parent_db->db_level, ==, 1); VERIFY3P(parent_db->db_dnode_handle->dnh_dnode, ==, dn); VERIFY3U(dr->dt.dll.dr_blkid >> epbs, ==, parent_db->db_blkid); blkptr_t *bp = parent_db->db.db_data; return (&bp[dr->dt.dll.dr_blkid & ((1 << epbs) - 1)]); } } static void dbuf_lightweight_ready(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; blkptr_t *bp = zio->io_bp; if (zio->io_error != 0) return; dnode_t *dn = dr->dr_dnode; blkptr_t *bp_orig = dbuf_lightweight_bp(dr); spa_t *spa = dmu_objset_spa(dn->dn_objset); int64_t delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig); dnode_diduse_space(dn, delta); uint64_t blkid = dr->dt.dll.dr_blkid; mutex_enter(&dn->dn_mtx); if (blkid > dn->dn_phys->dn_maxblkid) { ASSERT0(dn->dn_objset->os_raw_receive); dn->dn_phys->dn_maxblkid = blkid; } mutex_exit(&dn->dn_mtx); if (!BP_IS_EMBEDDED(bp)) { uint64_t fill = BP_IS_HOLE(bp) ? 0 : 1; BP_SET_FILL(bp, fill); } dmu_buf_impl_t *parent_db; EQUIV(dr->dr_parent == NULL, dn->dn_phys->dn_nlevels == 1); if (dr->dr_parent == NULL) { parent_db = dn->dn_dbuf; } else { parent_db = dr->dr_parent->dr_dbuf; } rw_enter(&parent_db->db_rwlock, RW_WRITER); *bp_orig = *bp; rw_exit(&parent_db->db_rwlock); } static void dbuf_lightweight_done(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; VERIFY0(zio->io_error); objset_t *os = dr->dr_dnode->dn_objset; dmu_tx_t *tx = os->os_synctx; if (zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)) { ASSERT(BP_EQUAL(zio->io_bp, &zio->io_bp_orig)); } else { dsl_dataset_t *ds = os->os_dsl_dataset; (void) dsl_dataset_block_kill(ds, &zio->io_bp_orig, tx, B_TRUE); dsl_dataset_block_born(ds, zio->io_bp, tx); } dsl_pool_undirty_space(dmu_objset_pool(os), dr->dr_accounted, zio->io_txg); abd_free(dr->dt.dll.dr_abd); kmem_free(dr, sizeof (*dr)); } noinline static void dbuf_sync_lightweight(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { dnode_t *dn = dr->dr_dnode; zio_t *pio; if (dn->dn_phys->dn_nlevels == 1) { pio = dn->dn_zio; } else { pio = dr->dr_parent->dr_zio; } zbookmark_phys_t zb = { .zb_objset = dmu_objset_id(dn->dn_objset), .zb_object = dn->dn_object, .zb_level = 0, .zb_blkid = dr->dt.dll.dr_blkid, }; /* * See comment in dbuf_write(). This is so that zio->io_bp_orig * will have the old BP in dbuf_lightweight_done(). */ dr->dr_bp_copy = *dbuf_lightweight_bp(dr); dr->dr_zio = zio_write(pio, dmu_objset_spa(dn->dn_objset), dmu_tx_get_txg(tx), &dr->dr_bp_copy, dr->dt.dll.dr_abd, dn->dn_datablksz, abd_get_size(dr->dt.dll.dr_abd), &dr->dt.dll.dr_props, dbuf_lightweight_ready, NULL, dbuf_lightweight_done, dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED | dr->dt.dll.dr_flags, &zb); zio_nowait(dr->dr_zio); } /* * dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is * critical the we not allow the compiler to inline this function in to * dbuf_sync_list() thereby drastically bloating the stack usage. */ noinline static void dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { arc_buf_t **datap = &dr->dt.dl.dr_data; dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn = dr->dr_dnode; objset_t *os; uint64_t txg = tx->tx_txg; ASSERT(dmu_tx_is_syncing(tx)); dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr); mutex_enter(&db->db_mtx); /* * To be synced, we must be dirtied. But we * might have been freed after the dirty. */ if (db->db_state == DB_UNCACHED) { /* This buffer has been freed since it was dirtied */ ASSERT(db->db.db_data == NULL); } else if (db->db_state == DB_FILL) { /* This buffer was freed and is now being re-filled */ ASSERT(db->db.db_data != dr->dt.dl.dr_data); } else if (db->db_state == DB_READ) { /* * This buffer has a clone we need to write, and an in-flight * read on the BP we're about to clone. Its safe to issue the * write here because the read has already been issued and the * contents won't change. */ ASSERT(dr->dt.dl.dr_brtwrite && dr->dt.dl.dr_override_state == DR_OVERRIDDEN); } else { ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL); } DBUF_VERIFY(db); if (db->db_blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) { /* * In the previous transaction group, the bonus buffer * was entirely used to store the attributes for the * dnode which overrode the dn_spill field. However, * when adding more attributes to the file a spill * block was required to hold the extra attributes. * * Make sure to clear the garbage left in the dn_spill * field from the previous attributes in the bonus * buffer. Otherwise, after writing out the spill * block to the new allocated dva, it will free * the old block pointed to by the invalid dn_spill. */ db->db_blkptr = NULL; } dn->dn_phys->dn_flags |= DNODE_FLAG_SPILL_BLKPTR; mutex_exit(&dn->dn_mtx); } /* * If this is a bonus buffer, simply copy the bonus data into the * dnode. It will be written out when the dnode is synced (and it * will be synced, since it must have been dirty for dbuf_sync to * be called). */ if (db->db_blkid == DMU_BONUS_BLKID) { ASSERT(dr->dr_dbuf == db); dbuf_sync_bonus(dr, tx); return; } os = dn->dn_objset; /* * This function may have dropped the db_mtx lock allowing a dmu_sync * operation to sneak in. As a result, we need to ensure that we * don't check the dr_override_state until we have returned from * dbuf_check_blkptr. */ dbuf_check_blkptr(dn, db); /* * If this buffer is in the middle of an immediate write, * wait for the synchronous IO to complete. */ while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) { ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); cv_wait(&db->db_changed, &db->db_mtx); } /* * If this is a dnode block, ensure it is appropriately encrypted * or decrypted, depending on what we are writing to it this txg. */ if (os->os_encrypted && dn->dn_object == DMU_META_DNODE_OBJECT) dbuf_prepare_encrypted_dnode_leaf(dr); if (*datap != NULL && *datap == db->db_buf && dn->dn_object != DMU_META_DNODE_OBJECT && zfs_refcount_count(&db->db_holds) > 1 && dr->dt.dl.dr_override_state != DR_OVERRIDDEN) { /* * If this buffer is currently "in use" (i.e., there * are active holds and db_data still references it), * then make a copy before we start the write so that * any modifications from the open txg will not leak * into this write. * * NOTE: this copy does not need to be made for * objects only modified in the syncing context (e.g. * DNONE_DNODE blocks). */ int psize = arc_buf_size(*datap); int lsize = arc_buf_lsize(*datap); arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); enum zio_compress compress_type = arc_get_compression(*datap); uint8_t complevel = arc_get_complevel(*datap); if (arc_is_encrypted(*datap)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(*datap, &byteorder, salt, iv, mac); *datap = arc_alloc_raw_buf(os->os_spa, db, dmu_objset_id(os), byteorder, salt, iv, mac, dn->dn_type, psize, lsize, compress_type, complevel); } else if (compress_type != ZIO_COMPRESS_OFF) { ASSERT3U(type, ==, ARC_BUFC_DATA); *datap = arc_alloc_compressed_buf(os->os_spa, db, psize, lsize, compress_type, complevel); } else { *datap = arc_alloc_buf(os->os_spa, db, type, psize); } memcpy((*datap)->b_data, db->db.db_data, psize); } db->db_data_pending = dr; mutex_exit(&db->db_mtx); dbuf_write(dr, *datap, tx); ASSERT(!list_link_active(&dr->dr_dirty_node)); if (dn->dn_object == DMU_META_DNODE_OBJECT) { list_insert_tail(&dn->dn_dirty_records[txg & TXG_MASK], dr); } else { zio_nowait(dr->dr_zio); } } /* * Syncs out a range of dirty records for indirect or leaf dbufs. May be * called recursively from dbuf_sync_indirect(). */ void dbuf_sync_list(list_t *list, int level, dmu_tx_t *tx) { dbuf_dirty_record_t *dr; while ((dr = list_head(list))) { if (dr->dr_zio != NULL) { /* * If we find an already initialized zio then we * are processing the meta-dnode, and we have finished. * The dbufs for all dnodes are put back on the list * during processing, so that we can zio_wait() * these IOs after initiating all child IOs. */ ASSERT3U(dr->dr_dbuf->db.db_object, ==, DMU_META_DNODE_OBJECT); break; } list_remove(list, dr); if (dr->dr_dbuf == NULL) { dbuf_sync_lightweight(dr, tx); } else { if (dr->dr_dbuf->db_blkid != DMU_BONUS_BLKID && dr->dr_dbuf->db_blkid != DMU_SPILL_BLKID) { VERIFY3U(dr->dr_dbuf->db_level, ==, level); } if (dr->dr_dbuf->db_level > 0) dbuf_sync_indirect(dr, tx); else dbuf_sync_leaf(dr, tx); } } } static void dbuf_write_ready(zio_t *zio, arc_buf_t *buf, void *vdb) { (void) buf; dmu_buf_impl_t *db = vdb; dnode_t *dn; blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; spa_t *spa = zio->io_spa; int64_t delta; uint64_t fill = 0; int i; ASSERT3P(db->db_blkptr, !=, NULL); ASSERT3P(&db->db_data_pending->dr_bp_copy, ==, bp); DB_DNODE_ENTER(db); dn = DB_DNODE(db); delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig); dnode_diduse_space(dn, delta - zio->io_prev_space_delta); zio->io_prev_space_delta = delta; if (BP_GET_LOGICAL_BIRTH(bp) != 0) { ASSERT((db->db_blkid != DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_type) || (db->db_blkid == DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_bonustype) || BP_IS_EMBEDDED(bp)); ASSERT(BP_GET_LEVEL(bp) == db->db_level); } mutex_enter(&db->db_mtx); #ifdef ZFS_DEBUG if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR); ASSERT(!(BP_IS_HOLE(bp)) && db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys)); } #endif if (db->db_level == 0) { mutex_enter(&dn->dn_mtx); if (db->db_blkid > dn->dn_phys->dn_maxblkid && db->db_blkid != DMU_SPILL_BLKID) { ASSERT0(db->db_objset->os_raw_receive); dn->dn_phys->dn_maxblkid = db->db_blkid; } mutex_exit(&dn->dn_mtx); if (dn->dn_type == DMU_OT_DNODE) { i = 0; while (i < db->db.db_size) { dnode_phys_t *dnp = (void *)(((char *)db->db.db_data) + i); i += DNODE_MIN_SIZE; if (dnp->dn_type != DMU_OT_NONE) { fill++; for (int j = 0; j < dnp->dn_nblkptr; j++) { (void) zfs_blkptr_verify(spa, &dnp->dn_blkptr[j], BLK_CONFIG_SKIP, BLK_VERIFY_HALT); } if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) { (void) zfs_blkptr_verify(spa, DN_SPILL_BLKPTR(dnp), BLK_CONFIG_SKIP, BLK_VERIFY_HALT); } i += dnp->dn_extra_slots * DNODE_MIN_SIZE; } } } else { if (BP_IS_HOLE(bp)) { fill = 0; } else { fill = 1; } } } else { blkptr_t *ibp = db->db.db_data; ASSERT3U(db->db.db_size, ==, 1<dn_phys->dn_indblkshift); for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) { if (BP_IS_HOLE(ibp)) continue; (void) zfs_blkptr_verify(spa, ibp, BLK_CONFIG_SKIP, BLK_VERIFY_HALT); fill += BP_GET_FILL(ibp); } } DB_DNODE_EXIT(db); if (!BP_IS_EMBEDDED(bp)) BP_SET_FILL(bp, fill); mutex_exit(&db->db_mtx); db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_WRITER, FTAG); *db->db_blkptr = *bp; dmu_buf_unlock_parent(db, dblt, FTAG); } /* * This function gets called just prior to running through the compression * stage of the zio pipeline. If we're an indirect block comprised of only * holes, then we want this indirect to be compressed away to a hole. In * order to do that we must zero out any information about the holes that * this indirect points to prior to before we try to compress it. */ static void dbuf_write_children_ready(zio_t *zio, arc_buf_t *buf, void *vdb) { (void) zio, (void) buf; dmu_buf_impl_t *db = vdb; dnode_t *dn; blkptr_t *bp; unsigned int epbs, i; ASSERT3U(db->db_level, >, 0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(epbs, <, 31); /* Determine if all our children are holes */ for (i = 0, bp = db->db.db_data; i < 1ULL << epbs; i++, bp++) { if (!BP_IS_HOLE(bp)) break; } /* * If all the children are holes, then zero them all out so that * we may get compressed away. */ if (i == 1ULL << epbs) { /* * We only found holes. Grab the rwlock to prevent * anybody from reading the blocks we're about to * zero out. */ rw_enter(&db->db_rwlock, RW_WRITER); memset(db->db.db_data, 0, db->db.db_size); rw_exit(&db->db_rwlock); } DB_DNODE_EXIT(db); } static void dbuf_write_done(zio_t *zio, arc_buf_t *buf, void *vdb) { (void) buf; dmu_buf_impl_t *db = vdb; blkptr_t *bp_orig = &zio->io_bp_orig; blkptr_t *bp = db->db_blkptr; objset_t *os = db->db_objset; dmu_tx_t *tx = os->os_synctx; ASSERT0(zio->io_error); ASSERT(db->db_blkptr == bp); /* * For nopwrites and rewrites we ensure that the bp matches our * original and bypass all the accounting. */ if (zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)) { ASSERT(BP_EQUAL(bp, bp_orig)); } else { dsl_dataset_t *ds = os->os_dsl_dataset; (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); dsl_dataset_block_born(ds, bp, tx); } mutex_enter(&db->db_mtx); DBUF_VERIFY(db); dbuf_dirty_record_t *dr = db->db_data_pending; dnode_t *dn = dr->dr_dnode; ASSERT(!list_link_active(&dr->dr_dirty_node)); ASSERT(dr->dr_dbuf == db); ASSERT(list_next(&db->db_dirty_records, dr) == NULL); list_remove(&db->db_dirty_records, dr); #ifdef ZFS_DEBUG if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR); ASSERT(!(BP_IS_HOLE(db->db_blkptr)) && db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys)); } #endif if (db->db_level == 0) { ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN); if (dr->dt.dl.dr_data != NULL && dr->dt.dl.dr_data != db->db_buf) { arc_buf_destroy(dr->dt.dl.dr_data, db); } } else { ASSERT(list_head(&dr->dt.di.dr_children) == NULL); ASSERT3U(db->db.db_size, ==, 1 << dn->dn_phys->dn_indblkshift); if (!BP_IS_HOLE(db->db_blkptr)) { int epbs __maybe_unused = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(db->db_blkid, <=, dn->dn_phys->dn_maxblkid >> (db->db_level * epbs)); ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, db->db.db_size); } mutex_destroy(&dr->dt.di.dr_mtx); list_destroy(&dr->dt.di.dr_children); } cv_broadcast(&db->db_changed); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; db->db_data_pending = NULL; dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE); dsl_pool_undirty_space(dmu_objset_pool(os), dr->dr_accounted, zio->io_txg); kmem_free(dr, sizeof (dbuf_dirty_record_t)); } static void dbuf_write_nofill_ready(zio_t *zio) { dbuf_write_ready(zio, NULL, zio->io_private); } static void dbuf_write_nofill_done(zio_t *zio) { dbuf_write_done(zio, NULL, zio->io_private); } static void dbuf_write_override_ready(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; dmu_buf_impl_t *db = dr->dr_dbuf; dbuf_write_ready(zio, NULL, db); } static void dbuf_write_override_done(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; dmu_buf_impl_t *db = dr->dr_dbuf; blkptr_t *obp = &dr->dt.dl.dr_overridden_by; mutex_enter(&db->db_mtx); if (!BP_EQUAL(zio->io_bp, obp)) { if (!BP_IS_HOLE(obp)) dsl_free(spa_get_dsl(zio->io_spa), zio->io_txg, obp); arc_release(dr->dt.dl.dr_data, db); } mutex_exit(&db->db_mtx); dbuf_write_done(zio, NULL, db); if (zio->io_abd != NULL) abd_free(zio->io_abd); } typedef struct dbuf_remap_impl_callback_arg { objset_t *drica_os; uint64_t drica_blk_birth; dmu_tx_t *drica_tx; } dbuf_remap_impl_callback_arg_t; static void dbuf_remap_impl_callback(uint64_t vdev, uint64_t offset, uint64_t size, void *arg) { dbuf_remap_impl_callback_arg_t *drica = arg; objset_t *os = drica->drica_os; spa_t *spa = dmu_objset_spa(os); dmu_tx_t *tx = drica->drica_tx; ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); if (os == spa_meta_objset(spa)) { spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx); } else { dsl_dataset_block_remapped(dmu_objset_ds(os), vdev, offset, size, drica->drica_blk_birth, tx); } } static void dbuf_remap_impl(dnode_t *dn, blkptr_t *bp, krwlock_t *rw, dmu_tx_t *tx) { blkptr_t bp_copy = *bp; spa_t *spa = dmu_objset_spa(dn->dn_objset); dbuf_remap_impl_callback_arg_t drica; ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); drica.drica_os = dn->dn_objset; drica.drica_blk_birth = BP_GET_LOGICAL_BIRTH(bp); drica.drica_tx = tx; if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback, &drica)) { /* * If the blkptr being remapped is tracked by a livelist, * then we need to make sure the livelist reflects the update. * First, cancel out the old blkptr by appending a 'FREE' * entry. Next, add an 'ALLOC' to track the new version. This * way we avoid trying to free an inaccurate blkptr at delete. * Note that embedded blkptrs are not tracked in livelists. */ if (dn->dn_objset != spa_meta_objset(spa)) { dsl_dataset_t *ds = dmu_objset_ds(dn->dn_objset); if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) && BP_GET_LOGICAL_BIRTH(bp) > ds->ds_dir->dd_origin_txg) { ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT(dsl_dir_is_clone(ds->ds_dir)); ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_LIVELIST)); bplist_append(&ds->ds_dir->dd_pending_frees, bp); bplist_append(&ds->ds_dir->dd_pending_allocs, &bp_copy); } } /* * The db_rwlock prevents dbuf_read_impl() from * dereferencing the BP while we are changing it. To * avoid lock contention, only grab it when we are actually * changing the BP. */ if (rw != NULL) rw_enter(rw, RW_WRITER); *bp = bp_copy; if (rw != NULL) rw_exit(rw); } } /* * Remap any existing BP's to concrete vdevs, if possible. */ static void dbuf_remap(dnode_t *dn, dmu_buf_impl_t *db, dmu_tx_t *tx) { spa_t *spa = dmu_objset_spa(db->db_objset); ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); if (!spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)) return; if (db->db_level > 0) { blkptr_t *bp = db->db.db_data; for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) { dbuf_remap_impl(dn, &bp[i], &db->db_rwlock, tx); } } else if (db->db.db_object == DMU_META_DNODE_OBJECT) { dnode_phys_t *dnp = db->db.db_data; ASSERT3U(db->db_dnode_handle->dnh_dnode->dn_type, ==, DMU_OT_DNODE); for (int i = 0; i < db->db.db_size >> DNODE_SHIFT; i += dnp[i].dn_extra_slots + 1) { for (int j = 0; j < dnp[i].dn_nblkptr; j++) { krwlock_t *lock = (dn->dn_dbuf == NULL ? NULL : &dn->dn_dbuf->db_rwlock); dbuf_remap_impl(dn, &dnp[i].dn_blkptr[j], lock, tx); } } } } /* * Populate dr->dr_zio with a zio to commit a dirty buffer to disk. * Caller is responsible for issuing the zio_[no]wait(dr->dr_zio). */ static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn = dr->dr_dnode; objset_t *os; dmu_buf_impl_t *parent = db->db_parent; uint64_t txg = tx->tx_txg; zbookmark_phys_t zb; zio_prop_t zp; zio_t *pio; /* parent I/O */ int wp_flag = 0; ASSERT(dmu_tx_is_syncing(tx)); os = dn->dn_objset; if (db->db_level > 0 || dn->dn_type == DMU_OT_DNODE) { /* * Private object buffers are released here rather than in * dbuf_dirty() since they are only modified in the syncing * context and we don't want the overhead of making multiple * copies of the data. */ if (BP_IS_HOLE(db->db_blkptr)) arc_buf_thaw(data); else dbuf_release_bp(db); dbuf_remap(dn, db, tx); } if (parent != dn->dn_dbuf) { /* Our parent is an indirect block. */ /* We have a dirty parent that has been scheduled for write. */ ASSERT(parent && parent->db_data_pending); /* Our parent's buffer is one level closer to the dnode. */ ASSERT(db->db_level == parent->db_level-1); /* * We're about to modify our parent's db_data by modifying * our block pointer, so the parent must be released. */ ASSERT(arc_released(parent->db_buf)); pio = parent->db_data_pending->dr_zio; } else { /* Our parent is the dnode itself. */ ASSERT((db->db_level == dn->dn_phys->dn_nlevels-1 && db->db_blkid != DMU_SPILL_BLKID) || (db->db_blkid == DMU_SPILL_BLKID && db->db_level == 0)); if (db->db_blkid != DMU_SPILL_BLKID) ASSERT3P(db->db_blkptr, ==, &dn->dn_phys->dn_blkptr[db->db_blkid]); pio = dn->dn_zio; } ASSERT(db->db_level == 0 || data == db->db_buf); ASSERT3U(BP_GET_LOGICAL_BIRTH(db->db_blkptr), <=, txg); ASSERT(pio); SET_BOOKMARK(&zb, os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : DMU_META_OBJSET, db->db.db_object, db->db_level, db->db_blkid); if (db->db_blkid == DMU_SPILL_BLKID) wp_flag = WP_SPILL; wp_flag |= (data == NULL) ? WP_NOFILL : 0; dmu_write_policy(os, dn, db->db_level, wp_flag, &zp); /* * We copy the blkptr now (rather than when we instantiate the dirty * record), because its value can change between open context and * syncing context. We do not need to hold dn_struct_rwlock to read * db_blkptr because we are in syncing context. */ dr->dr_bp_copy = *db->db_blkptr; if (db->db_level == 0 && dr->dt.dl.dr_override_state == DR_OVERRIDDEN) { /* * The BP for this block has been provided by open context * (by dmu_sync() or dmu_buf_write_embedded()). */ abd_t *contents = (data != NULL) ? abd_get_from_buf(data->b_data, arc_buf_size(data)) : NULL; dr->dr_zio = zio_write(pio, os->os_spa, txg, &dr->dr_bp_copy, contents, db->db.db_size, db->db.db_size, &zp, dbuf_write_override_ready, NULL, dbuf_write_override_done, dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); mutex_enter(&db->db_mtx); dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; zio_write_override(dr->dr_zio, &dr->dt.dl.dr_overridden_by, dr->dt.dl.dr_copies, dr->dt.dl.dr_nopwrite, dr->dt.dl.dr_brtwrite); mutex_exit(&db->db_mtx); } else if (data == NULL) { ASSERT(zp.zp_checksum == ZIO_CHECKSUM_OFF || zp.zp_checksum == ZIO_CHECKSUM_NOPARITY); dr->dr_zio = zio_write(pio, os->os_spa, txg, &dr->dr_bp_copy, NULL, db->db.db_size, db->db.db_size, &zp, dbuf_write_nofill_ready, NULL, dbuf_write_nofill_done, db, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED | ZIO_FLAG_NODATA, &zb); } else { ASSERT(arc_released(data)); /* * For indirect blocks, we want to setup the children * ready callback so that we can properly handle an indirect * block that only contains holes. */ arc_write_done_func_t *children_ready_cb = NULL; if (db->db_level != 0) children_ready_cb = dbuf_write_children_ready; dr->dr_zio = arc_write(pio, os->os_spa, txg, &dr->dr_bp_copy, data, !DBUF_IS_CACHEABLE(db), dbuf_is_l2cacheable(db), &zp, dbuf_write_ready, children_ready_cb, dbuf_write_done, db, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); } } EXPORT_SYMBOL(dbuf_find); EXPORT_SYMBOL(dbuf_is_metadata); EXPORT_SYMBOL(dbuf_destroy); EXPORT_SYMBOL(dbuf_loan_arcbuf); EXPORT_SYMBOL(dbuf_whichblock); EXPORT_SYMBOL(dbuf_read); EXPORT_SYMBOL(dbuf_unoverride); EXPORT_SYMBOL(dbuf_free_range); EXPORT_SYMBOL(dbuf_new_size); EXPORT_SYMBOL(dbuf_release_bp); EXPORT_SYMBOL(dbuf_dirty); EXPORT_SYMBOL(dmu_buf_set_crypt_params); EXPORT_SYMBOL(dmu_buf_will_dirty); EXPORT_SYMBOL(dmu_buf_is_dirty); EXPORT_SYMBOL(dmu_buf_will_clone); EXPORT_SYMBOL(dmu_buf_will_not_fill); EXPORT_SYMBOL(dmu_buf_will_fill); EXPORT_SYMBOL(dmu_buf_fill_done); EXPORT_SYMBOL(dmu_buf_rele); EXPORT_SYMBOL(dbuf_assign_arcbuf); EXPORT_SYMBOL(dbuf_prefetch); EXPORT_SYMBOL(dbuf_hold_impl); EXPORT_SYMBOL(dbuf_hold); EXPORT_SYMBOL(dbuf_hold_level); EXPORT_SYMBOL(dbuf_create_bonus); EXPORT_SYMBOL(dbuf_spill_set_blksz); EXPORT_SYMBOL(dbuf_rm_spill); EXPORT_SYMBOL(dbuf_add_ref); EXPORT_SYMBOL(dbuf_rele); EXPORT_SYMBOL(dbuf_rele_and_unlock); EXPORT_SYMBOL(dbuf_refcount); EXPORT_SYMBOL(dbuf_sync_list); EXPORT_SYMBOL(dmu_buf_set_user); EXPORT_SYMBOL(dmu_buf_set_user_ie); EXPORT_SYMBOL(dmu_buf_get_user); EXPORT_SYMBOL(dmu_buf_get_blkptr); ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, max_bytes, U64, ZMOD_RW, "Maximum size in bytes of the dbuf cache."); ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, hiwater_pct, UINT, ZMOD_RW, "Percentage over dbuf_cache_max_bytes for direct dbuf eviction."); ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, lowater_pct, UINT, ZMOD_RW, "Percentage below dbuf_cache_max_bytes when dbuf eviction stops."); ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_max_bytes, U64, ZMOD_RW, "Maximum size in bytes of dbuf metadata cache."); ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, cache_shift, UINT, ZMOD_RW, "Set size of dbuf cache to log2 fraction of arc size."); ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_shift, UINT, ZMOD_RW, "Set size of dbuf metadata cache to log2 fraction of arc size."); ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, mutex_cache_shift, UINT, ZMOD_RD, "Set size of dbuf cache mutex array as log2 shift."); diff --git a/sys/contrib/openzfs/module/zfs/spa.c b/sys/contrib/openzfs/module/zfs/spa.c index f67d980ae4c6..96daf51b696a 100644 --- a/sys/contrib/openzfs/module/zfs/spa.c +++ b/sys/contrib/openzfs/module/zfs/spa.c @@ -1,10835 +1,10857 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2024 by Delphix. All rights reserved. * Copyright (c) 2018, Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2016 Toomas Soome * Copyright (c) 2016 Actifio, Inc. All rights reserved. * Copyright 2018 Joyent, Inc. * Copyright (c) 2017, 2019, Datto Inc. All rights reserved. * Copyright 2017 Joyent, Inc. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2021, Colm Buckley * Copyright (c) 2023 Hewlett Packard Enterprise Development LP. */ /* * SPA: Storage Pool Allocator * * This file contains all the routines used when modifying on-disk SPA state. * This includes opening, importing, destroying, exporting a pool, and syncing a * pool. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #include #include #endif /* _KERNEL */ #include "zfs_prop.h" #include "zfs_comutil.h" #include /* * spa_thread() existed on Illumos as a parent thread for the various worker * threads that actually run the pool, as a way to both reference the entire * pool work as a single object, and to share properties like scheduling * options. It has not yet been adapted to Linux or FreeBSD. This define is * used to mark related parts of the code to make things easier for the reader, * and to compile this code out. It can be removed when someone implements it, * moves it to some Illumos-specific place, or removes it entirely. */ #undef HAVE_SPA_THREAD /* * The "System Duty Cycle" scheduling class is an Illumos feature to help * prevent CPU-intensive kernel threads from affecting latency on interactive * threads. It doesn't exist on Linux or FreeBSD, so the supporting code is * gated behind a define. On Illumos SDC depends on spa_thread(), but * spa_thread() also has other uses, so this is a separate define. */ #undef HAVE_SYSDC /* * The interval, in seconds, at which failed configuration cache file writes * should be retried. */ int zfs_ccw_retry_interval = 300; typedef enum zti_modes { ZTI_MODE_FIXED, /* value is # of threads (min 1) */ ZTI_MODE_SCALE, /* Taskqs scale with CPUs. */ ZTI_MODE_SYNC, /* sync thread assigned */ ZTI_MODE_NULL, /* don't create a taskq */ ZTI_NMODES } zti_modes_t; #define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) } #define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 } #define ZTI_SCALE { ZTI_MODE_SCALE, 0, 1 } #define ZTI_SYNC { ZTI_MODE_SYNC, 0, 1 } #define ZTI_NULL { ZTI_MODE_NULL, 0, 0 } #define ZTI_N(n) ZTI_P(n, 1) #define ZTI_ONE ZTI_N(1) typedef struct zio_taskq_info { zti_modes_t zti_mode; uint_t zti_value; uint_t zti_count; } zio_taskq_info_t; static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = { "iss", "iss_h", "int", "int_h" }; /* * This table defines the taskq settings for each ZFS I/O type. When * initializing a pool, we use this table to create an appropriately sized * taskq. Some operations are low volume and therefore have a small, static * number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE * macros. Other operations process a large amount of data; the ZTI_SCALE * macro causes us to create a taskq oriented for throughput. Some operations * are so high frequency and short-lived that the taskq itself can become a * point of lock contention. The ZTI_P(#, #) macro indicates that we need an * additional degree of parallelism specified by the number of threads per- * taskq and the number of taskqs; when dispatching an event in this case, the * particular taskq is chosen at random. ZTI_SCALE uses a number of taskqs * that scales with the number of CPUs. * * The different taskq priorities are to handle the different contexts (issue * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that * need to be handled with minimum delay. */ static zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = { /* ISSUE ISSUE_HIGH INTR INTR_HIGH */ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */ { ZTI_N(8), ZTI_NULL, ZTI_SCALE, ZTI_NULL }, /* READ */ { ZTI_SYNC, ZTI_N(5), ZTI_SCALE, ZTI_N(5) }, /* WRITE */ { ZTI_SCALE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FLUSH */ { ZTI_N(4), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* TRIM */ }; static void spa_sync_version(void *arg, dmu_tx_t *tx); static void spa_sync_props(void *arg, dmu_tx_t *tx); static boolean_t spa_has_active_shared_spare(spa_t *spa); static int spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport); static void spa_vdev_resilver_done(spa_t *spa); /* * Percentage of all CPUs that can be used by the metaslab preload taskq. */ static uint_t metaslab_preload_pct = 50; static uint_t zio_taskq_batch_pct = 80; /* 1 thread per cpu in pset */ static uint_t zio_taskq_batch_tpq; /* threads per taskq */ #ifdef HAVE_SYSDC static const boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */ static const uint_t zio_taskq_basedc = 80; /* base duty cycle */ #endif #ifdef HAVE_SPA_THREAD static const boolean_t spa_create_process = B_TRUE; /* no process => no sysdc */ #endif static uint_t zio_taskq_wr_iss_ncpus = 0; /* * Report any spa_load_verify errors found, but do not fail spa_load. * This is used by zdb to analyze non-idle pools. */ boolean_t spa_load_verify_dryrun = B_FALSE; /* * Allow read spacemaps in case of readonly import (spa_mode == SPA_MODE_READ). * This is used by zdb for spacemaps verification. */ boolean_t spa_mode_readable_spacemaps = B_FALSE; /* * This (illegal) pool name is used when temporarily importing a spa_t in order * to get the vdev stats associated with the imported devices. */ #define TRYIMPORT_NAME "$import" /* * For debugging purposes: print out vdev tree during pool import. */ static int spa_load_print_vdev_tree = B_FALSE; /* * A non-zero value for zfs_max_missing_tvds means that we allow importing * pools with missing top-level vdevs. This is strictly intended for advanced * pool recovery cases since missing data is almost inevitable. Pools with * missing devices can only be imported read-only for safety reasons, and their * fail-mode will be automatically set to "continue". * * With 1 missing vdev we should be able to import the pool and mount all * datasets. User data that was not modified after the missing device has been * added should be recoverable. This means that snapshots created prior to the * addition of that device should be completely intact. * * With 2 missing vdevs, some datasets may fail to mount since there are * dataset statistics that are stored as regular metadata. Some data might be * recoverable if those vdevs were added recently. * * With 3 or more missing vdevs, the pool is severely damaged and MOS entries * may be missing entirely. Chances of data recovery are very low. Note that * there are also risks of performing an inadvertent rewind as we might be * missing all the vdevs with the latest uberblocks. */ uint64_t zfs_max_missing_tvds = 0; /* * The parameters below are similar to zfs_max_missing_tvds but are only * intended for a preliminary open of the pool with an untrusted config which * might be incomplete or out-dated. * * We are more tolerant for pools opened from a cachefile since we could have * an out-dated cachefile where a device removal was not registered. * We could have set the limit arbitrarily high but in the case where devices * are really missing we would want to return the proper error codes; we chose * SPA_DVAS_PER_BP - 1 so that some copies of the MOS would still be available * and we get a chance to retrieve the trusted config. */ uint64_t zfs_max_missing_tvds_cachefile = SPA_DVAS_PER_BP - 1; /* * In the case where config was assembled by scanning device paths (/dev/dsks * by default) we are less tolerant since all the existing devices should have * been detected and we want spa_load to return the right error codes. */ uint64_t zfs_max_missing_tvds_scan = 0; /* * Debugging aid that pauses spa_sync() towards the end. */ static const boolean_t zfs_pause_spa_sync = B_FALSE; /* * Variables to indicate the livelist condense zthr func should wait at certain * points for the livelist to be removed - used to test condense/destroy races */ static int zfs_livelist_condense_zthr_pause = 0; static int zfs_livelist_condense_sync_pause = 0; /* * Variables to track whether or not condense cancellation has been * triggered in testing. */ static int zfs_livelist_condense_sync_cancel = 0; static int zfs_livelist_condense_zthr_cancel = 0; /* * Variable to track whether or not extra ALLOC blkptrs were added to a * livelist entry while it was being condensed (caused by the way we track * remapped blkptrs in dbuf_remap_impl) */ static int zfs_livelist_condense_new_alloc = 0; /* * ========================================================================== * SPA properties routines * ========================================================================== */ /* * Add a (source=src, propname=propval) list to an nvlist. */ static void spa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, const char *strval, uint64_t intval, zprop_source_t src) { const char *propname = zpool_prop_to_name(prop); nvlist_t *propval; propval = fnvlist_alloc(); fnvlist_add_uint64(propval, ZPROP_SOURCE, src); if (strval != NULL) fnvlist_add_string(propval, ZPROP_VALUE, strval); else fnvlist_add_uint64(propval, ZPROP_VALUE, intval); fnvlist_add_nvlist(nvl, propname, propval); nvlist_free(propval); } /* * Add a user property (source=src, propname=propval) to an nvlist. */ static void spa_prop_add_user(nvlist_t *nvl, const char *propname, char *strval, zprop_source_t src) { nvlist_t *propval; VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0); VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0); VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0); nvlist_free(propval); } /* * Get property values from the spa configuration. */ static void spa_prop_get_config(spa_t *spa, nvlist_t **nvp) { vdev_t *rvd = spa->spa_root_vdev; dsl_pool_t *pool = spa->spa_dsl_pool; uint64_t size, alloc, cap, version; const zprop_source_t src = ZPROP_SRC_NONE; spa_config_dirent_t *dp; metaslab_class_t *mc = spa_normal_class(spa); ASSERT(MUTEX_HELD(&spa->spa_props_lock)); if (rvd != NULL) { alloc = metaslab_class_get_alloc(mc); alloc += metaslab_class_get_alloc(spa_special_class(spa)); alloc += metaslab_class_get_alloc(spa_dedup_class(spa)); alloc += metaslab_class_get_alloc(spa_embedded_log_class(spa)); size = metaslab_class_get_space(mc); size += metaslab_class_get_space(spa_special_class(spa)); size += metaslab_class_get_space(spa_dedup_class(spa)); size += metaslab_class_get_space(spa_embedded_log_class(spa)); spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src); spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src); spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src); spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL, size - alloc, src); spa_prop_add_list(*nvp, ZPOOL_PROP_CHECKPOINT, NULL, spa->spa_checkpoint_info.sci_dspace, src); spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL, metaslab_class_fragmentation(mc), src); spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL, metaslab_class_expandable_space(mc), src); spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL, (spa_mode(spa) == SPA_MODE_READ), src); cap = (size == 0) ? 0 : (alloc * 100 / size); spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src); spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL, ddt_get_pool_dedup_ratio(spa), src); spa_prop_add_list(*nvp, ZPOOL_PROP_BCLONEUSED, NULL, brt_get_used(spa), src); spa_prop_add_list(*nvp, ZPOOL_PROP_BCLONESAVED, NULL, brt_get_saved(spa), src); spa_prop_add_list(*nvp, ZPOOL_PROP_BCLONERATIO, NULL, brt_get_ratio(spa), src); spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL, rvd->vdev_state, src); version = spa_version(spa); if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) { spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, ZPROP_SRC_DEFAULT); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, ZPROP_SRC_LOCAL); } spa_prop_add_list(*nvp, ZPOOL_PROP_LOAD_GUID, NULL, spa_load_guid(spa), src); } if (pool != NULL) { /* * The $FREE directory was introduced in SPA_VERSION_DEADLISTS, * when opening pools before this version freedir will be NULL. */ if (pool->dp_free_dir != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL, dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes, src); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL, 0, src); } if (pool->dp_leak_dir != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL, dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes, src); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL, 0, src); } } spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src); if (spa->spa_comment != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment, 0, ZPROP_SRC_LOCAL); } if (spa->spa_compatibility != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_COMPATIBILITY, spa->spa_compatibility, 0, ZPROP_SRC_LOCAL); } if (spa->spa_root != NULL) spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root, 0, ZPROP_SRC_LOCAL); if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) { spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL, MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL, SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE); } if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) { spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL, DNODE_MAX_SIZE, ZPROP_SRC_NONE); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL, DNODE_MIN_SIZE, ZPROP_SRC_NONE); } if ((dp = list_head(&spa->spa_config_list)) != NULL) { if (dp->scd_path == NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, "none", 0, ZPROP_SRC_LOCAL); } else if (strcmp(dp->scd_path, spa_config_path) != 0) { spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, dp->scd_path, 0, ZPROP_SRC_LOCAL); } } } /* * Get zpool property values. */ int spa_prop_get(spa_t *spa, nvlist_t **nvp) { objset_t *mos = spa->spa_meta_objset; zap_cursor_t zc; zap_attribute_t za; dsl_pool_t *dp; int err; err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP); if (err) return (err); dp = spa_get_dsl(spa); dsl_pool_config_enter(dp, FTAG); mutex_enter(&spa->spa_props_lock); /* * Get properties from the spa config. */ spa_prop_get_config(spa, nvp); /* If no pool property object, no more prop to get. */ if (mos == NULL || spa->spa_pool_props_object == 0) goto out; /* * Get properties from the MOS pool property object. */ for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object); (err = zap_cursor_retrieve(&zc, &za)) == 0; zap_cursor_advance(&zc)) { uint64_t intval = 0; char *strval = NULL; zprop_source_t src = ZPROP_SRC_DEFAULT; zpool_prop_t prop; if ((prop = zpool_name_to_prop(za.za_name)) == ZPOOL_PROP_INVAL && !zfs_prop_user(za.za_name)) continue; switch (za.za_integer_length) { case 8: /* integer property */ if (za.za_first_integer != zpool_prop_default_numeric(prop)) src = ZPROP_SRC_LOCAL; if (prop == ZPOOL_PROP_BOOTFS) { dsl_dataset_t *ds = NULL; err = dsl_dataset_hold_obj(dp, za.za_first_integer, FTAG, &ds); if (err != 0) break; strval = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); dsl_dataset_name(ds, strval); dsl_dataset_rele(ds, FTAG); } else { strval = NULL; intval = za.za_first_integer; } spa_prop_add_list(*nvp, prop, strval, intval, src); if (strval != NULL) kmem_free(strval, ZFS_MAX_DATASET_NAME_LEN); break; case 1: /* string property */ strval = kmem_alloc(za.za_num_integers, KM_SLEEP); err = zap_lookup(mos, spa->spa_pool_props_object, za.za_name, 1, za.za_num_integers, strval); if (err) { kmem_free(strval, za.za_num_integers); break; } if (prop != ZPOOL_PROP_INVAL) { spa_prop_add_list(*nvp, prop, strval, 0, src); } else { src = ZPROP_SRC_LOCAL; spa_prop_add_user(*nvp, za.za_name, strval, src); } kmem_free(strval, za.za_num_integers); break; default: break; } } zap_cursor_fini(&zc); out: mutex_exit(&spa->spa_props_lock); dsl_pool_config_exit(dp, FTAG); if (err && err != ENOENT) { nvlist_free(*nvp); *nvp = NULL; return (err); } return (0); } /* * Validate the given pool properties nvlist and modify the list * for the property values to be set. */ static int spa_prop_validate(spa_t *spa, nvlist_t *props) { nvpair_t *elem; int error = 0, reset_bootfs = 0; uint64_t objnum = 0; boolean_t has_feature = B_FALSE; elem = NULL; while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { uint64_t intval; const char *strval, *slash, *check, *fname; const char *propname = nvpair_name(elem); zpool_prop_t prop = zpool_name_to_prop(propname); switch (prop) { case ZPOOL_PROP_INVAL: /* * Sanitize the input. */ if (zfs_prop_user(propname)) { if (strlen(propname) >= ZAP_MAXNAMELEN) { error = SET_ERROR(ENAMETOOLONG); break; } if (strlen(fnvpair_value_string(elem)) >= ZAP_MAXVALUELEN) { error = SET_ERROR(E2BIG); break; } } else if (zpool_prop_feature(propname)) { if (nvpair_type(elem) != DATA_TYPE_UINT64) { error = SET_ERROR(EINVAL); break; } if (nvpair_value_uint64(elem, &intval) != 0) { error = SET_ERROR(EINVAL); break; } if (intval != 0) { error = SET_ERROR(EINVAL); break; } fname = strchr(propname, '@') + 1; if (zfeature_lookup_name(fname, NULL) != 0) { error = SET_ERROR(EINVAL); break; } has_feature = B_TRUE; } else { error = SET_ERROR(EINVAL); break; } break; case ZPOOL_PROP_VERSION: error = nvpair_value_uint64(elem, &intval); if (!error && (intval < spa_version(spa) || intval > SPA_VERSION_BEFORE_FEATURES || has_feature)) error = SET_ERROR(EINVAL); break; case ZPOOL_PROP_DELEGATION: case ZPOOL_PROP_AUTOREPLACE: case ZPOOL_PROP_LISTSNAPS: case ZPOOL_PROP_AUTOEXPAND: case ZPOOL_PROP_AUTOTRIM: error = nvpair_value_uint64(elem, &intval); if (!error && intval > 1) error = SET_ERROR(EINVAL); break; case ZPOOL_PROP_MULTIHOST: error = nvpair_value_uint64(elem, &intval); if (!error && intval > 1) error = SET_ERROR(EINVAL); if (!error) { uint32_t hostid = zone_get_hostid(NULL); if (hostid) spa->spa_hostid = hostid; else error = SET_ERROR(ENOTSUP); } break; case ZPOOL_PROP_BOOTFS: /* * If the pool version is less than SPA_VERSION_BOOTFS, * or the pool is still being created (version == 0), * the bootfs property cannot be set. */ if (spa_version(spa) < SPA_VERSION_BOOTFS) { error = SET_ERROR(ENOTSUP); break; } /* * Make sure the vdev config is bootable */ if (!vdev_is_bootable(spa->spa_root_vdev)) { error = SET_ERROR(ENOTSUP); break; } reset_bootfs = 1; error = nvpair_value_string(elem, &strval); if (!error) { objset_t *os; if (strval == NULL || strval[0] == '\0') { objnum = zpool_prop_default_numeric( ZPOOL_PROP_BOOTFS); break; } error = dmu_objset_hold(strval, FTAG, &os); if (error != 0) break; /* Must be ZPL. */ if (dmu_objset_type(os) != DMU_OST_ZFS) { error = SET_ERROR(ENOTSUP); } else { objnum = dmu_objset_id(os); } dmu_objset_rele(os, FTAG); } break; case ZPOOL_PROP_FAILUREMODE: error = nvpair_value_uint64(elem, &intval); if (!error && intval > ZIO_FAILURE_MODE_PANIC) error = SET_ERROR(EINVAL); /* * This is a special case which only occurs when * the pool has completely failed. This allows * the user to change the in-core failmode property * without syncing it out to disk (I/Os might * currently be blocked). We do this by returning * EIO to the caller (spa_prop_set) to trick it * into thinking we encountered a property validation * error. */ if (!error && spa_suspended(spa)) { spa->spa_failmode = intval; error = SET_ERROR(EIO); } break; case ZPOOL_PROP_CACHEFILE: if ((error = nvpair_value_string(elem, &strval)) != 0) break; if (strval[0] == '\0') break; if (strcmp(strval, "none") == 0) break; if (strval[0] != '/') { error = SET_ERROR(EINVAL); break; } slash = strrchr(strval, '/'); ASSERT(slash != NULL); if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || strcmp(slash, "/..") == 0) error = SET_ERROR(EINVAL); break; case ZPOOL_PROP_COMMENT: if ((error = nvpair_value_string(elem, &strval)) != 0) break; for (check = strval; *check != '\0'; check++) { if (!isprint(*check)) { error = SET_ERROR(EINVAL); break; } } if (strlen(strval) > ZPROP_MAX_COMMENT) error = SET_ERROR(E2BIG); break; default: break; } if (error) break; } (void) nvlist_remove_all(props, zpool_prop_to_name(ZPOOL_PROP_DEDUPDITTO)); if (!error && reset_bootfs) { error = nvlist_remove(props, zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING); if (!error) { error = nvlist_add_uint64(props, zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum); } } return (error); } void spa_configfile_set(spa_t *spa, nvlist_t *nvp, boolean_t need_sync) { const char *cachefile; spa_config_dirent_t *dp; if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), &cachefile) != 0) return; dp = kmem_alloc(sizeof (spa_config_dirent_t), KM_SLEEP); if (cachefile[0] == '\0') dp->scd_path = spa_strdup(spa_config_path); else if (strcmp(cachefile, "none") == 0) dp->scd_path = NULL; else dp->scd_path = spa_strdup(cachefile); list_insert_head(&spa->spa_config_list, dp); if (need_sync) spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); } int spa_prop_set(spa_t *spa, nvlist_t *nvp) { int error; nvpair_t *elem = NULL; boolean_t need_sync = B_FALSE; if ((error = spa_prop_validate(spa, nvp)) != 0) return (error); while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) { zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem)); if (prop == ZPOOL_PROP_CACHEFILE || prop == ZPOOL_PROP_ALTROOT || prop == ZPOOL_PROP_READONLY) continue; if (prop == ZPOOL_PROP_INVAL && zfs_prop_user(nvpair_name(elem))) { need_sync = B_TRUE; break; } if (prop == ZPOOL_PROP_VERSION || prop == ZPOOL_PROP_INVAL) { uint64_t ver = 0; if (prop == ZPOOL_PROP_VERSION) { VERIFY(nvpair_value_uint64(elem, &ver) == 0); } else { ASSERT(zpool_prop_feature(nvpair_name(elem))); ver = SPA_VERSION_FEATURES; need_sync = B_TRUE; } /* Save time if the version is already set. */ if (ver == spa_version(spa)) continue; /* * In addition to the pool directory object, we might * create the pool properties object, the features for * read object, the features for write object, or the * feature descriptions object. */ error = dsl_sync_task(spa->spa_name, NULL, spa_sync_version, &ver, 6, ZFS_SPACE_CHECK_RESERVED); if (error) return (error); continue; } need_sync = B_TRUE; break; } if (need_sync) { return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props, nvp, 6, ZFS_SPACE_CHECK_RESERVED)); } return (0); } /* * If the bootfs property value is dsobj, clear it. */ void spa_prop_clear_bootfs(spa_t *spa, uint64_t dsobj, dmu_tx_t *tx) { if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) { VERIFY(zap_remove(spa->spa_meta_objset, spa->spa_pool_props_object, zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0); spa->spa_bootfs = 0; } } static int spa_change_guid_check(void *arg, dmu_tx_t *tx) { uint64_t *newguid __maybe_unused = arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *rvd = spa->spa_root_vdev; uint64_t vdev_state; if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) { int error = (spa_has_checkpoint(spa)) ? ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT; return (SET_ERROR(error)); } spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); vdev_state = rvd->vdev_state; spa_config_exit(spa, SCL_STATE, FTAG); if (vdev_state != VDEV_STATE_HEALTHY) return (SET_ERROR(ENXIO)); ASSERT3U(spa_guid(spa), !=, *newguid); return (0); } static void spa_change_guid_sync(void *arg, dmu_tx_t *tx) { uint64_t *newguid = arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; uint64_t oldguid; vdev_t *rvd = spa->spa_root_vdev; oldguid = spa_guid(spa); spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); rvd->vdev_guid = *newguid; rvd->vdev_guid_sum += (*newguid - oldguid); vdev_config_dirty(rvd); spa_config_exit(spa, SCL_STATE, FTAG); spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu", (u_longlong_t)oldguid, (u_longlong_t)*newguid); } /* * Change the GUID for the pool. This is done so that we can later * re-import a pool built from a clone of our own vdevs. We will modify * the root vdev's guid, our own pool guid, and then mark all of our * vdevs dirty. Note that we must make sure that all our vdevs are * online when we do this, or else any vdevs that weren't present * would be orphaned from our pool. We are also going to issue a * sysevent to update any watchers. */ int spa_change_guid(spa_t *spa) { int error; uint64_t guid; mutex_enter(&spa->spa_vdev_top_lock); mutex_enter(&spa_namespace_lock); guid = spa_generate_guid(NULL); error = dsl_sync_task(spa->spa_name, spa_change_guid_check, spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED); if (error == 0) { /* * Clear the kobj flag from all the vdevs to allow * vdev_cache_process_kobj_evt() to post events to all the * vdevs since GUID is updated. */ vdev_clear_kobj_evt(spa->spa_root_vdev); for (int i = 0; i < spa->spa_l2cache.sav_count; i++) vdev_clear_kobj_evt(spa->spa_l2cache.sav_vdevs[i]); spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE); spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_REGUID); } mutex_exit(&spa_namespace_lock); mutex_exit(&spa->spa_vdev_top_lock); return (error); } /* * ========================================================================== * SPA state manipulation (open/create/destroy/import/export) * ========================================================================== */ static int spa_error_entry_compare(const void *a, const void *b) { const spa_error_entry_t *sa = (const spa_error_entry_t *)a; const spa_error_entry_t *sb = (const spa_error_entry_t *)b; int ret; ret = memcmp(&sa->se_bookmark, &sb->se_bookmark, sizeof (zbookmark_phys_t)); return (TREE_ISIGN(ret)); } /* * Utility function which retrieves copies of the current logs and * re-initializes them in the process. */ void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub) { ASSERT(MUTEX_HELD(&spa->spa_errlist_lock)); memcpy(last, &spa->spa_errlist_last, sizeof (avl_tree_t)); memcpy(scrub, &spa->spa_errlist_scrub, sizeof (avl_tree_t)); avl_create(&spa->spa_errlist_scrub, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); avl_create(&spa->spa_errlist_last, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); } static void spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q) { const zio_taskq_info_t *ztip = &zio_taskqs[t][q]; enum zti_modes mode = ztip->zti_mode; uint_t value = ztip->zti_value; uint_t count = ztip->zti_count; spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; uint_t cpus, flags = TASKQ_DYNAMIC; switch (mode) { case ZTI_MODE_FIXED: ASSERT3U(value, >, 0); break; case ZTI_MODE_SYNC: /* * Create one wr_iss taskq for every 'zio_taskq_wr_iss_ncpus', * not to exceed the number of spa allocators. */ if (zio_taskq_wr_iss_ncpus == 0) { count = MAX(boot_ncpus / spa->spa_alloc_count, 1); } else { count = MAX(1, boot_ncpus / MAX(1, zio_taskq_wr_iss_ncpus)); } count = MAX(count, (zio_taskq_batch_pct + 99) / 100); count = MIN(count, spa->spa_alloc_count); /* * zio_taskq_batch_pct is unbounded and may exceed 100%, but no * single taskq may have more threads than 100% of online cpus. */ value = (zio_taskq_batch_pct + count / 2) / count; value = MIN(value, 100); flags |= TASKQ_THREADS_CPU_PCT; break; case ZTI_MODE_SCALE: flags |= TASKQ_THREADS_CPU_PCT; /* * We want more taskqs to reduce lock contention, but we want * less for better request ordering and CPU utilization. */ cpus = MAX(1, boot_ncpus * zio_taskq_batch_pct / 100); if (zio_taskq_batch_tpq > 0) { count = MAX(1, (cpus + zio_taskq_batch_tpq / 2) / zio_taskq_batch_tpq); } else { /* * Prefer 6 threads per taskq, but no more taskqs * than threads in them on large systems. For 80%: * * taskq taskq total * cpus taskqs percent threads threads * ------- ------- ------- ------- ------- * 1 1 80% 1 1 * 2 1 80% 1 1 * 4 1 80% 3 3 * 8 2 40% 3 6 * 16 3 27% 4 12 * 32 5 16% 5 25 * 64 7 11% 7 49 * 128 10 8% 10 100 * 256 14 6% 15 210 */ count = 1 + cpus / 6; while (count * count > cpus) count--; } /* Limit each taskq within 100% to not trigger assertion. */ count = MAX(count, (zio_taskq_batch_pct + 99) / 100); value = (zio_taskq_batch_pct + count / 2) / count; break; case ZTI_MODE_NULL: tqs->stqs_count = 0; tqs->stqs_taskq = NULL; return; default: panic("unrecognized mode for %s_%s taskq (%u:%u) in " "spa_taskqs_init()", zio_type_name[t], zio_taskq_types[q], mode, value); break; } ASSERT3U(count, >, 0); tqs->stqs_count = count; tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP); for (uint_t i = 0; i < count; i++) { taskq_t *tq; char name[32]; if (count > 1) (void) snprintf(name, sizeof (name), "%s_%s_%u", zio_type_name[t], zio_taskq_types[q], i); else (void) snprintf(name, sizeof (name), "%s_%s", zio_type_name[t], zio_taskq_types[q]); #ifdef HAVE_SYSDC if (zio_taskq_sysdc && spa->spa_proc != &p0) { (void) zio_taskq_basedc; tq = taskq_create_sysdc(name, value, 50, INT_MAX, spa->spa_proc, zio_taskq_basedc, flags); } else { #endif pri_t pri = maxclsyspri; /* * The write issue taskq can be extremely CPU * intensive. Run it at slightly less important * priority than the other taskqs. * * Under Linux and FreeBSD this means incrementing * the priority value as opposed to platforms like * illumos where it should be decremented. * * On FreeBSD, if priorities divided by four (RQ_PPQ) * are equal then a difference between them is * insignificant. */ if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE) { #if defined(__linux__) pri++; #elif defined(__FreeBSD__) pri += 4; #else #error "unknown OS" #endif } tq = taskq_create_proc(name, value, pri, 50, INT_MAX, spa->spa_proc, flags); #ifdef HAVE_SYSDC } #endif tqs->stqs_taskq[i] = tq; } } static void spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; if (tqs->stqs_taskq == NULL) { ASSERT3U(tqs->stqs_count, ==, 0); return; } for (uint_t i = 0; i < tqs->stqs_count; i++) { ASSERT3P(tqs->stqs_taskq[i], !=, NULL); taskq_destroy(tqs->stqs_taskq[i]); } kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *)); tqs->stqs_taskq = NULL; } #ifdef _KERNEL /* * The READ and WRITE rows of zio_taskqs are configurable at module load time * by setting zio_taskq_read or zio_taskq_write. * * Example (the defaults for READ and WRITE) * zio_taskq_read='fixed,1,8 null scale null' * zio_taskq_write='sync fixed,1,5 scale fixed,1,5' * * Each sets the entire row at a time. * * 'fixed' is parameterised: fixed,Q,T where Q is number of taskqs, T is number * of threads per taskq. * * 'null' can only be set on the high-priority queues (queue selection for * high-priority queues will fall back to the regular queue if the high-pri * is NULL. */ static const char *const modes[ZTI_NMODES] = { "fixed", "scale", "sync", "null" }; /* Parse the incoming config string. Modifies cfg */ static int spa_taskq_param_set(zio_type_t t, char *cfg) { int err = 0; zio_taskq_info_t row[ZIO_TASKQ_TYPES] = {{0}}; char *next = cfg, *tok, *c; /* * Parse out each element from the string and fill `row`. The entire * row has to be set at once, so any errors are flagged by just * breaking out of this loop early. */ uint_t q; for (q = 0; q < ZIO_TASKQ_TYPES; q++) { /* `next` is the start of the config */ if (next == NULL) break; /* Eat up leading space */ while (isspace(*next)) next++; if (*next == '\0') break; /* Mode ends at space or end of string */ tok = next; next = strchr(tok, ' '); if (next != NULL) *next++ = '\0'; /* Parameters start after a comma */ c = strchr(tok, ','); if (c != NULL) *c++ = '\0'; /* Match mode string */ uint_t mode; for (mode = 0; mode < ZTI_NMODES; mode++) if (strcmp(tok, modes[mode]) == 0) break; if (mode == ZTI_NMODES) break; /* Invalid canary */ row[q].zti_mode = ZTI_NMODES; /* Per-mode setup */ switch (mode) { /* * FIXED is parameterised: number of queues, and number of * threads per queue. */ case ZTI_MODE_FIXED: { /* No parameters? */ if (c == NULL || *c == '\0') break; /* Find next parameter */ tok = c; c = strchr(tok, ','); if (c == NULL) break; /* Take digits and convert */ unsigned long long nq; if (!(isdigit(*tok))) break; err = ddi_strtoull(tok, &tok, 10, &nq); /* Must succeed and also end at the next param sep */ if (err != 0 || tok != c) break; /* Move past the comma */ tok++; /* Need another number */ if (!(isdigit(*tok))) break; /* Remember start to make sure we moved */ c = tok; /* Take digits */ unsigned long long ntpq; err = ddi_strtoull(tok, &tok, 10, &ntpq); /* Must succeed, and moved forward */ if (err != 0 || tok == c || *tok != '\0') break; /* * sanity; zero queues/threads make no sense, and * 16K is almost certainly more than anyone will ever * need and avoids silly numbers like UINT32_MAX */ if (nq == 0 || nq >= 16384 || ntpq == 0 || ntpq >= 16384) break; const zio_taskq_info_t zti = ZTI_P(ntpq, nq); row[q] = zti; break; } case ZTI_MODE_SCALE: { const zio_taskq_info_t zti = ZTI_SCALE; row[q] = zti; break; } case ZTI_MODE_SYNC: { const zio_taskq_info_t zti = ZTI_SYNC; row[q] = zti; break; } case ZTI_MODE_NULL: { /* * Can only null the high-priority queues; the general- * purpose ones have to exist. */ if (q != ZIO_TASKQ_ISSUE_HIGH && q != ZIO_TASKQ_INTERRUPT_HIGH) break; const zio_taskq_info_t zti = ZTI_NULL; row[q] = zti; break; } default: break; } /* Ensure we set a mode */ if (row[q].zti_mode == ZTI_NMODES) break; } /* Didn't get a full row, fail */ if (q < ZIO_TASKQ_TYPES) return (SET_ERROR(EINVAL)); /* Eat trailing space */ if (next != NULL) while (isspace(*next)) next++; /* If there's anything left over then fail */ if (next != NULL && *next != '\0') return (SET_ERROR(EINVAL)); /* Success! Copy it into the real config */ for (q = 0; q < ZIO_TASKQ_TYPES; q++) zio_taskqs[t][q] = row[q]; return (0); } static int spa_taskq_param_get(zio_type_t t, char *buf, boolean_t add_newline) { int pos = 0; /* Build paramater string from live config */ const char *sep = ""; for (uint_t q = 0; q < ZIO_TASKQ_TYPES; q++) { const zio_taskq_info_t *zti = &zio_taskqs[t][q]; if (zti->zti_mode == ZTI_MODE_FIXED) pos += sprintf(&buf[pos], "%s%s,%u,%u", sep, modes[zti->zti_mode], zti->zti_count, zti->zti_value); else pos += sprintf(&buf[pos], "%s%s", sep, modes[zti->zti_mode]); sep = " "; } if (add_newline) buf[pos++] = '\n'; buf[pos] = '\0'; return (pos); } #ifdef __linux__ static int spa_taskq_read_param_set(const char *val, zfs_kernel_param_t *kp) { char *cfg = kmem_strdup(val); int err = spa_taskq_param_set(ZIO_TYPE_READ, cfg); kmem_free(cfg, strlen(val)+1); return (-err); } static int spa_taskq_read_param_get(char *buf, zfs_kernel_param_t *kp) { return (spa_taskq_param_get(ZIO_TYPE_READ, buf, TRUE)); } static int spa_taskq_write_param_set(const char *val, zfs_kernel_param_t *kp) { char *cfg = kmem_strdup(val); int err = spa_taskq_param_set(ZIO_TYPE_WRITE, cfg); kmem_free(cfg, strlen(val)+1); return (-err); } static int spa_taskq_write_param_get(char *buf, zfs_kernel_param_t *kp) { return (spa_taskq_param_get(ZIO_TYPE_WRITE, buf, TRUE)); } #else /* * On FreeBSD load-time parameters can be set up before malloc() is available, * so we have to do all the parsing work on the stack. */ #define SPA_TASKQ_PARAM_MAX (128) static int spa_taskq_read_param(ZFS_MODULE_PARAM_ARGS) { char buf[SPA_TASKQ_PARAM_MAX]; int err; (void) spa_taskq_param_get(ZIO_TYPE_READ, buf, FALSE); err = sysctl_handle_string(oidp, buf, sizeof (buf), req); if (err || req->newptr == NULL) return (err); return (spa_taskq_param_set(ZIO_TYPE_READ, buf)); } static int spa_taskq_write_param(ZFS_MODULE_PARAM_ARGS) { char buf[SPA_TASKQ_PARAM_MAX]; int err; (void) spa_taskq_param_get(ZIO_TYPE_WRITE, buf, FALSE); err = sysctl_handle_string(oidp, buf, sizeof (buf), req); if (err || req->newptr == NULL) return (err); return (spa_taskq_param_set(ZIO_TYPE_WRITE, buf)); } #endif #endif /* _KERNEL */ /* * Dispatch a task to the appropriate taskq for the ZFS I/O type and priority. * Note that a type may have multiple discrete taskqs to avoid lock contention * on the taskq itself. */ static taskq_t * spa_taskq_dispatch_select(spa_t *spa, zio_type_t t, zio_taskq_type_t q, zio_t *zio) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; taskq_t *tq; ASSERT3P(tqs->stqs_taskq, !=, NULL); ASSERT3U(tqs->stqs_count, !=, 0); if ((t == ZIO_TYPE_WRITE) && (q == ZIO_TASKQ_ISSUE) && (zio != NULL) && (zio->io_wr_iss_tq != NULL)) { /* dispatch to assigned write issue taskq */ tq = zio->io_wr_iss_tq; return (tq); } if (tqs->stqs_count == 1) { tq = tqs->stqs_taskq[0]; } else { tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count]; } return (tq); } void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q, task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent, zio_t *zio) { taskq_t *tq = spa_taskq_dispatch_select(spa, t, q, zio); taskq_dispatch_ent(tq, func, arg, flags, ent); } /* * Same as spa_taskq_dispatch_ent() but block on the task until completion. */ void spa_taskq_dispatch_sync(spa_t *spa, zio_type_t t, zio_taskq_type_t q, task_func_t *func, void *arg, uint_t flags) { taskq_t *tq = spa_taskq_dispatch_select(spa, t, q, NULL); taskqid_t id = taskq_dispatch(tq, func, arg, flags); if (id) taskq_wait_id(tq, id); } static void spa_create_zio_taskqs(spa_t *spa) { for (int t = 0; t < ZIO_TYPES; t++) { for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { spa_taskqs_init(spa, t, q); } } } #if defined(_KERNEL) && defined(HAVE_SPA_THREAD) static void spa_thread(void *arg) { psetid_t zio_taskq_psrset_bind = PS_NONE; callb_cpr_t cprinfo; spa_t *spa = arg; user_t *pu = PTOU(curproc); CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr, spa->spa_name); ASSERT(curproc != &p0); (void) snprintf(pu->u_psargs, sizeof (pu->u_psargs), "zpool-%s", spa->spa_name); (void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm)); /* bind this thread to the requested psrset */ if (zio_taskq_psrset_bind != PS_NONE) { pool_lock(); mutex_enter(&cpu_lock); mutex_enter(&pidlock); mutex_enter(&curproc->p_lock); if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind, 0, NULL, NULL) == 0) { curthread->t_bind_pset = zio_taskq_psrset_bind; } else { cmn_err(CE_WARN, "Couldn't bind process for zfs pool \"%s\" to " "pset %d\n", spa->spa_name, zio_taskq_psrset_bind); } mutex_exit(&curproc->p_lock); mutex_exit(&pidlock); mutex_exit(&cpu_lock); pool_unlock(); } #ifdef HAVE_SYSDC if (zio_taskq_sysdc) { sysdc_thread_enter(curthread, 100, 0); } #endif spa->spa_proc = curproc; spa->spa_did = curthread->t_did; spa_create_zio_taskqs(spa); mutex_enter(&spa->spa_proc_lock); ASSERT(spa->spa_proc_state == SPA_PROC_CREATED); spa->spa_proc_state = SPA_PROC_ACTIVE; cv_broadcast(&spa->spa_proc_cv); CALLB_CPR_SAFE_BEGIN(&cprinfo); while (spa->spa_proc_state == SPA_PROC_ACTIVE) cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock); ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE); spa->spa_proc_state = SPA_PROC_GONE; spa->spa_proc = &p0; cv_broadcast(&spa->spa_proc_cv); CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */ mutex_enter(&curproc->p_lock); lwp_exit(); } #endif extern metaslab_ops_t *metaslab_allocator(spa_t *spa); /* * Activate an uninitialized pool. */ static void spa_activate(spa_t *spa, spa_mode_t mode) { metaslab_ops_t *msp = metaslab_allocator(spa); ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); spa->spa_state = POOL_STATE_ACTIVE; spa->spa_mode = mode; spa->spa_read_spacemaps = spa_mode_readable_spacemaps; spa->spa_normal_class = metaslab_class_create(spa, msp); spa->spa_log_class = metaslab_class_create(spa, msp); spa->spa_embedded_log_class = metaslab_class_create(spa, msp); spa->spa_special_class = metaslab_class_create(spa, msp); spa->spa_dedup_class = metaslab_class_create(spa, msp); /* Try to create a covering process */ mutex_enter(&spa->spa_proc_lock); ASSERT(spa->spa_proc_state == SPA_PROC_NONE); ASSERT(spa->spa_proc == &p0); spa->spa_did = 0; #ifdef HAVE_SPA_THREAD /* Only create a process if we're going to be around a while. */ if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) { if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri, NULL, 0) == 0) { spa->spa_proc_state = SPA_PROC_CREATED; while (spa->spa_proc_state == SPA_PROC_CREATED) { cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); } ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); ASSERT(spa->spa_proc != &p0); ASSERT(spa->spa_did != 0); } else { #ifdef _KERNEL cmn_err(CE_WARN, "Couldn't create process for zfs pool \"%s\"\n", spa->spa_name); #endif } } #endif /* HAVE_SPA_THREAD */ mutex_exit(&spa->spa_proc_lock); /* If we didn't create a process, we need to create our taskqs. */ if (spa->spa_proc == &p0) { spa_create_zio_taskqs(spa); } for (size_t i = 0; i < TXG_SIZE; i++) { spa->spa_txg_zio[i] = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); } list_create(&spa->spa_config_dirty_list, sizeof (vdev_t), offsetof(vdev_t, vdev_config_dirty_node)); list_create(&spa->spa_evicting_os_list, sizeof (objset_t), offsetof(objset_t, os_evicting_node)); list_create(&spa->spa_state_dirty_list, sizeof (vdev_t), offsetof(vdev_t, vdev_state_dirty_node)); txg_list_create(&spa->spa_vdev_txg_list, spa, offsetof(struct vdev, vdev_txg_node)); avl_create(&spa->spa_errlist_scrub, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); avl_create(&spa->spa_errlist_last, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); avl_create(&spa->spa_errlist_healed, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); spa_activate_os(spa); spa_keystore_init(&spa->spa_keystore); /* * This taskq is used to perform zvol-minor-related tasks * asynchronously. This has several advantages, including easy * resolution of various deadlocks. * * The taskq must be single threaded to ensure tasks are always * processed in the order in which they were dispatched. * * A taskq per pool allows one to keep the pools independent. * This way if one pool is suspended, it will not impact another. * * The preferred location to dispatch a zvol minor task is a sync * task. In this context, there is easy access to the spa_t and minimal * error handling is required because the sync task must succeed. */ spa->spa_zvol_taskq = taskq_create("z_zvol", 1, defclsyspri, 1, INT_MAX, 0); /* * The taskq to preload metaslabs. */ spa->spa_metaslab_taskq = taskq_create("z_metaslab", metaslab_preload_pct, maxclsyspri, 1, INT_MAX, TASKQ_DYNAMIC | TASKQ_THREADS_CPU_PCT); /* * Taskq dedicated to prefetcher threads: this is used to prevent the * pool traverse code from monopolizing the global (and limited) * system_taskq by inappropriately scheduling long running tasks on it. */ spa->spa_prefetch_taskq = taskq_create("z_prefetch", 100, defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC | TASKQ_THREADS_CPU_PCT); /* * The taskq to upgrade datasets in this pool. Currently used by * feature SPA_FEATURE_USEROBJ_ACCOUNTING/SPA_FEATURE_PROJECT_QUOTA. */ spa->spa_upgrade_taskq = taskq_create("z_upgrade", 100, defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC | TASKQ_THREADS_CPU_PCT); } /* * Opposite of spa_activate(). */ static void spa_deactivate(spa_t *spa) { ASSERT(spa->spa_sync_on == B_FALSE); ASSERT(spa->spa_dsl_pool == NULL); ASSERT(spa->spa_root_vdev == NULL); ASSERT(spa->spa_async_zio_root == NULL); ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED); spa_evicting_os_wait(spa); if (spa->spa_zvol_taskq) { taskq_destroy(spa->spa_zvol_taskq); spa->spa_zvol_taskq = NULL; } if (spa->spa_metaslab_taskq) { taskq_destroy(spa->spa_metaslab_taskq); spa->spa_metaslab_taskq = NULL; } if (spa->spa_prefetch_taskq) { taskq_destroy(spa->spa_prefetch_taskq); spa->spa_prefetch_taskq = NULL; } if (spa->spa_upgrade_taskq) { taskq_destroy(spa->spa_upgrade_taskq); spa->spa_upgrade_taskq = NULL; } txg_list_destroy(&spa->spa_vdev_txg_list); list_destroy(&spa->spa_config_dirty_list); list_destroy(&spa->spa_evicting_os_list); list_destroy(&spa->spa_state_dirty_list); taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid); for (int t = 0; t < ZIO_TYPES; t++) { for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { spa_taskqs_fini(spa, t, q); } } for (size_t i = 0; i < TXG_SIZE; i++) { ASSERT3P(spa->spa_txg_zio[i], !=, NULL); VERIFY0(zio_wait(spa->spa_txg_zio[i])); spa->spa_txg_zio[i] = NULL; } metaslab_class_destroy(spa->spa_normal_class); spa->spa_normal_class = NULL; metaslab_class_destroy(spa->spa_log_class); spa->spa_log_class = NULL; metaslab_class_destroy(spa->spa_embedded_log_class); spa->spa_embedded_log_class = NULL; metaslab_class_destroy(spa->spa_special_class); spa->spa_special_class = NULL; metaslab_class_destroy(spa->spa_dedup_class); spa->spa_dedup_class = NULL; /* * If this was part of an import or the open otherwise failed, we may * still have errors left in the queues. Empty them just in case. */ spa_errlog_drain(spa); avl_destroy(&spa->spa_errlist_scrub); avl_destroy(&spa->spa_errlist_last); avl_destroy(&spa->spa_errlist_healed); spa_keystore_fini(&spa->spa_keystore); spa->spa_state = POOL_STATE_UNINITIALIZED; mutex_enter(&spa->spa_proc_lock); if (spa->spa_proc_state != SPA_PROC_NONE) { ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); spa->spa_proc_state = SPA_PROC_DEACTIVATE; cv_broadcast(&spa->spa_proc_cv); while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) { ASSERT(spa->spa_proc != &p0); cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); } ASSERT(spa->spa_proc_state == SPA_PROC_GONE); spa->spa_proc_state = SPA_PROC_NONE; } ASSERT(spa->spa_proc == &p0); mutex_exit(&spa->spa_proc_lock); /* * We want to make sure spa_thread() has actually exited the ZFS * module, so that the module can't be unloaded out from underneath * it. */ if (spa->spa_did != 0) { thread_join(spa->spa_did); spa->spa_did = 0; } spa_deactivate_os(spa); } /* * Verify a pool configuration, and construct the vdev tree appropriately. This * will create all the necessary vdevs in the appropriate layout, with each vdev * in the CLOSED state. This will prep the pool before open/creation/import. * All vdev validation is done by the vdev_alloc() routine. */ int spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id, int atype) { nvlist_t **child; uint_t children; int error; if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0) return (error); if ((*vdp)->vdev_ops->vdev_op_leaf) return (0); error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children); if (error == ENOENT) return (0); if (error) { vdev_free(*vdp); *vdp = NULL; return (SET_ERROR(EINVAL)); } for (int c = 0; c < children; c++) { vdev_t *vd; if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c, atype)) != 0) { vdev_free(*vdp); *vdp = NULL; return (error); } } ASSERT(*vdp != NULL); return (0); } static boolean_t spa_should_flush_logs_on_unload(spa_t *spa) { if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) return (B_FALSE); if (!spa_writeable(spa)) return (B_FALSE); if (!spa->spa_sync_on) return (B_FALSE); if (spa_state(spa) != POOL_STATE_EXPORTED) return (B_FALSE); if (zfs_keep_log_spacemaps_at_export) return (B_FALSE); return (B_TRUE); } /* * Opens a transaction that will set the flag that will instruct * spa_sync to attempt to flush all the metaslabs for that txg. */ static void spa_unload_log_sm_flush_all(spa_t *spa) { dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); ASSERT3U(spa->spa_log_flushall_txg, ==, 0); spa->spa_log_flushall_txg = dmu_tx_get_txg(tx); dmu_tx_commit(tx); txg_wait_synced(spa_get_dsl(spa), spa->spa_log_flushall_txg); } static void spa_unload_log_sm_metadata(spa_t *spa) { void *cookie = NULL; spa_log_sm_t *sls; log_summary_entry_t *e; while ((sls = avl_destroy_nodes(&spa->spa_sm_logs_by_txg, &cookie)) != NULL) { VERIFY0(sls->sls_mscount); kmem_free(sls, sizeof (spa_log_sm_t)); } while ((e = list_remove_head(&spa->spa_log_summary)) != NULL) { VERIFY0(e->lse_mscount); kmem_free(e, sizeof (log_summary_entry_t)); } spa->spa_unflushed_stats.sus_nblocks = 0; spa->spa_unflushed_stats.sus_memused = 0; spa->spa_unflushed_stats.sus_blocklimit = 0; } static void spa_destroy_aux_threads(spa_t *spa) { if (spa->spa_condense_zthr != NULL) { zthr_destroy(spa->spa_condense_zthr); spa->spa_condense_zthr = NULL; } if (spa->spa_checkpoint_discard_zthr != NULL) { zthr_destroy(spa->spa_checkpoint_discard_zthr); spa->spa_checkpoint_discard_zthr = NULL; } if (spa->spa_livelist_delete_zthr != NULL) { zthr_destroy(spa->spa_livelist_delete_zthr); spa->spa_livelist_delete_zthr = NULL; } if (spa->spa_livelist_condense_zthr != NULL) { zthr_destroy(spa->spa_livelist_condense_zthr); spa->spa_livelist_condense_zthr = NULL; } if (spa->spa_raidz_expand_zthr != NULL) { zthr_destroy(spa->spa_raidz_expand_zthr); spa->spa_raidz_expand_zthr = NULL; } } /* * Opposite of spa_load(). */ static void spa_unload(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa_state(spa) != POOL_STATE_UNINITIALIZED); spa_import_progress_remove(spa_guid(spa)); spa_load_note(spa, "UNLOADING"); spa_wake_waiters(spa); /* * If we have set the spa_final_txg, we have already performed the * tasks below in spa_export_common(). We should not redo it here since * we delay the final TXGs beyond what spa_final_txg is set at. */ if (spa->spa_final_txg == UINT64_MAX) { /* * If the log space map feature is enabled and the pool is * getting exported (but not destroyed), we want to spend some * time flushing as many metaslabs as we can in an attempt to * destroy log space maps and save import time. */ if (spa_should_flush_logs_on_unload(spa)) spa_unload_log_sm_flush_all(spa); /* * Stop async tasks. */ spa_async_suspend(spa); if (spa->spa_root_vdev) { vdev_t *root_vdev = spa->spa_root_vdev; vdev_initialize_stop_all(root_vdev, VDEV_INITIALIZE_ACTIVE); vdev_trim_stop_all(root_vdev, VDEV_TRIM_ACTIVE); vdev_autotrim_stop_all(spa); vdev_rebuild_stop_all(spa); } } /* * Stop syncing. */ if (spa->spa_sync_on) { txg_sync_stop(spa->spa_dsl_pool); spa->spa_sync_on = B_FALSE; } /* * This ensures that there is no async metaslab prefetching * while we attempt to unload the spa. */ taskq_wait(spa->spa_metaslab_taskq); if (spa->spa_mmp.mmp_thread) mmp_thread_stop(spa); /* * Wait for any outstanding async I/O to complete. */ if (spa->spa_async_zio_root != NULL) { for (int i = 0; i < max_ncpus; i++) (void) zio_wait(spa->spa_async_zio_root[i]); kmem_free(spa->spa_async_zio_root, max_ncpus * sizeof (void *)); spa->spa_async_zio_root = NULL; } if (spa->spa_vdev_removal != NULL) { spa_vdev_removal_destroy(spa->spa_vdev_removal); spa->spa_vdev_removal = NULL; } spa_destroy_aux_threads(spa); spa_condense_fini(spa); bpobj_close(&spa->spa_deferred_bpobj); spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); /* * Close all vdevs. */ if (spa->spa_root_vdev) vdev_free(spa->spa_root_vdev); ASSERT(spa->spa_root_vdev == NULL); /* * Close the dsl pool. */ if (spa->spa_dsl_pool) { dsl_pool_close(spa->spa_dsl_pool); spa->spa_dsl_pool = NULL; spa->spa_meta_objset = NULL; } ddt_unload(spa); brt_unload(spa); spa_unload_log_sm_metadata(spa); /* * Drop and purge level 2 cache */ spa_l2cache_drop(spa); if (spa->spa_spares.sav_vdevs) { for (int i = 0; i < spa->spa_spares.sav_count; i++) vdev_free(spa->spa_spares.sav_vdevs[i]); kmem_free(spa->spa_spares.sav_vdevs, spa->spa_spares.sav_count * sizeof (void *)); spa->spa_spares.sav_vdevs = NULL; } if (spa->spa_spares.sav_config) { nvlist_free(spa->spa_spares.sav_config); spa->spa_spares.sav_config = NULL; } spa->spa_spares.sav_count = 0; if (spa->spa_l2cache.sav_vdevs) { for (int i = 0; i < spa->spa_l2cache.sav_count; i++) { vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]); vdev_free(spa->spa_l2cache.sav_vdevs[i]); } kmem_free(spa->spa_l2cache.sav_vdevs, spa->spa_l2cache.sav_count * sizeof (void *)); spa->spa_l2cache.sav_vdevs = NULL; } if (spa->spa_l2cache.sav_config) { nvlist_free(spa->spa_l2cache.sav_config); spa->spa_l2cache.sav_config = NULL; } spa->spa_l2cache.sav_count = 0; spa->spa_async_suspended = 0; spa->spa_indirect_vdevs_loaded = B_FALSE; if (spa->spa_comment != NULL) { spa_strfree(spa->spa_comment); spa->spa_comment = NULL; } if (spa->spa_compatibility != NULL) { spa_strfree(spa->spa_compatibility); spa->spa_compatibility = NULL; } spa->spa_raidz_expand = NULL; spa_config_exit(spa, SCL_ALL, spa); } /* * Load (or re-load) the current list of vdevs describing the active spares for * this pool. When this is called, we have some form of basic information in * 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and * then re-generate a more complete list including status information. */ void spa_load_spares(spa_t *spa) { nvlist_t **spares; uint_t nspares; int i; vdev_t *vd, *tvd; #ifndef _KERNEL /* * zdb opens both the current state of the pool and the * checkpointed state (if present), with a different spa_t. * * As spare vdevs are shared among open pools, we skip loading * them when we load the checkpointed state of the pool. */ if (!spa_writeable(spa)) return; #endif ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); /* * First, close and free any existing spare vdevs. */ if (spa->spa_spares.sav_vdevs) { for (i = 0; i < spa->spa_spares.sav_count; i++) { vd = spa->spa_spares.sav_vdevs[i]; /* Undo the call to spa_activate() below */ if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, B_FALSE)) != NULL && tvd->vdev_isspare) spa_spare_remove(tvd); vdev_close(vd); vdev_free(vd); } kmem_free(spa->spa_spares.sav_vdevs, spa->spa_spares.sav_count * sizeof (void *)); } if (spa->spa_spares.sav_config == NULL) nspares = 0; else VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, &spares, &nspares)); spa->spa_spares.sav_count = (int)nspares; spa->spa_spares.sav_vdevs = NULL; if (nspares == 0) return; /* * Construct the array of vdevs, opening them to get status in the * process. For each spare, there is potentially two different vdev_t * structures associated with it: one in the list of spares (used only * for basic validation purposes) and one in the active vdev * configuration (if it's spared in). During this phase we open and * validate each vdev on the spare list. If the vdev also exists in the * active configuration, then we also mark this vdev as an active spare. */ spa->spa_spares.sav_vdevs = kmem_zalloc(nspares * sizeof (void *), KM_SLEEP); for (i = 0; i < spa->spa_spares.sav_count; i++) { VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0, VDEV_ALLOC_SPARE) == 0); ASSERT(vd != NULL); spa->spa_spares.sav_vdevs[i] = vd; if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, B_FALSE)) != NULL) { if (!tvd->vdev_isspare) spa_spare_add(tvd); /* * We only mark the spare active if we were successfully * able to load the vdev. Otherwise, importing a pool * with a bad active spare would result in strange * behavior, because multiple pool would think the spare * is actively in use. * * There is a vulnerability here to an equally bizarre * circumstance, where a dead active spare is later * brought back to life (onlined or otherwise). Given * the rarity of this scenario, and the extra complexity * it adds, we ignore the possibility. */ if (!vdev_is_dead(tvd)) spa_spare_activate(tvd); } vd->vdev_top = vd; vd->vdev_aux = &spa->spa_spares; if (vdev_open(vd) != 0) continue; if (vdev_validate_aux(vd) == 0) spa_spare_add(vd); } /* * Recompute the stashed list of spares, with status information * this time. */ fnvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES); spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *), KM_SLEEP); for (i = 0; i < spa->spa_spares.sav_count; i++) spares[i] = vdev_config_generate(spa, spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE); fnvlist_add_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares, spa->spa_spares.sav_count); for (i = 0; i < spa->spa_spares.sav_count; i++) nvlist_free(spares[i]); kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *)); } /* * Load (or re-load) the current list of vdevs describing the active l2cache for * this pool. When this is called, we have some form of basic information in * 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and * then re-generate a more complete list including status information. * Devices which are already active have their details maintained, and are * not re-opened. */ void spa_load_l2cache(spa_t *spa) { nvlist_t **l2cache = NULL; uint_t nl2cache; int i, j, oldnvdevs; uint64_t guid; vdev_t *vd, **oldvdevs, **newvdevs; spa_aux_vdev_t *sav = &spa->spa_l2cache; #ifndef _KERNEL /* * zdb opens both the current state of the pool and the * checkpointed state (if present), with a different spa_t. * * As L2 caches are part of the ARC which is shared among open * pools, we skip loading them when we load the checkpointed * state of the pool. */ if (!spa_writeable(spa)) return; #endif ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); oldvdevs = sav->sav_vdevs; oldnvdevs = sav->sav_count; sav->sav_vdevs = NULL; sav->sav_count = 0; if (sav->sav_config == NULL) { nl2cache = 0; newvdevs = NULL; goto out; } VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache)); newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP); /* * Process new nvlist of vdevs. */ for (i = 0; i < nl2cache; i++) { guid = fnvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID); newvdevs[i] = NULL; for (j = 0; j < oldnvdevs; j++) { vd = oldvdevs[j]; if (vd != NULL && guid == vd->vdev_guid) { /* * Retain previous vdev for add/remove ops. */ newvdevs[i] = vd; oldvdevs[j] = NULL; break; } } if (newvdevs[i] == NULL) { /* * Create new vdev */ VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0, VDEV_ALLOC_L2CACHE) == 0); ASSERT(vd != NULL); newvdevs[i] = vd; /* * Commit this vdev as an l2cache device, * even if it fails to open. */ spa_l2cache_add(vd); vd->vdev_top = vd; vd->vdev_aux = sav; spa_l2cache_activate(vd); if (vdev_open(vd) != 0) continue; (void) vdev_validate_aux(vd); if (!vdev_is_dead(vd)) l2arc_add_vdev(spa, vd); /* * Upon cache device addition to a pool or pool * creation with a cache device or if the header * of the device is invalid we issue an async * TRIM command for the whole device which will * execute if l2arc_trim_ahead > 0. */ spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM); } } sav->sav_vdevs = newvdevs; sav->sav_count = (int)nl2cache; /* * Recompute the stashed list of l2cache devices, with status * information this time. */ fnvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE); if (sav->sav_count > 0) l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP); for (i = 0; i < sav->sav_count; i++) l2cache[i] = vdev_config_generate(spa, sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE); fnvlist_add_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache, sav->sav_count); out: /* * Purge vdevs that were dropped */ if (oldvdevs) { for (i = 0; i < oldnvdevs; i++) { uint64_t pool; vd = oldvdevs[i]; if (vd != NULL) { ASSERT(vd->vdev_isl2cache); if (spa_l2cache_exists(vd->vdev_guid, &pool) && pool != 0ULL && l2arc_vdev_present(vd)) l2arc_remove_vdev(vd); vdev_clear_stats(vd); vdev_free(vd); } } kmem_free(oldvdevs, oldnvdevs * sizeof (void *)); } for (i = 0; i < sav->sav_count; i++) nvlist_free(l2cache[i]); if (sav->sav_count) kmem_free(l2cache, sav->sav_count * sizeof (void *)); } static int load_nvlist(spa_t *spa, uint64_t obj, nvlist_t **value) { dmu_buf_t *db; char *packed = NULL; size_t nvsize = 0; int error; *value = NULL; error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db); if (error) return (error); nvsize = *(uint64_t *)db->db_data; dmu_buf_rele(db, FTAG); packed = vmem_alloc(nvsize, KM_SLEEP); error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed, DMU_READ_PREFETCH); if (error == 0) error = nvlist_unpack(packed, nvsize, value, 0); vmem_free(packed, nvsize); return (error); } /* * Concrete top-level vdevs that are not missing and are not logs. At every * spa_sync we write new uberblocks to at least SPA_SYNC_MIN_VDEVS core tvds. */ static uint64_t spa_healthy_core_tvds(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; uint64_t tvds = 0; for (uint64_t i = 0; i < rvd->vdev_children; i++) { vdev_t *vd = rvd->vdev_child[i]; if (vd->vdev_islog) continue; if (vdev_is_concrete(vd) && !vdev_is_dead(vd)) tvds++; } return (tvds); } /* * Checks to see if the given vdev could not be opened, in which case we post a * sysevent to notify the autoreplace code that the device has been removed. */ static void spa_check_removed(vdev_t *vd) { for (uint64_t c = 0; c < vd->vdev_children; c++) spa_check_removed(vd->vdev_child[c]); if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) && vdev_is_concrete(vd)) { zfs_post_autoreplace(vd->vdev_spa, vd); spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_CHECK); } } static int spa_check_for_missing_logs(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; /* * If we're doing a normal import, then build up any additional * diagnostic information about missing log devices. * We'll pass this up to the user for further processing. */ if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) { nvlist_t **child, *nv; uint64_t idx = 0; child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t *), KM_SLEEP); nv = fnvlist_alloc(); for (uint64_t c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; /* * We consider a device as missing only if it failed * to open (i.e. offline or faulted is not considered * as missing). */ if (tvd->vdev_islog && tvd->vdev_state == VDEV_STATE_CANT_OPEN) { child[idx++] = vdev_config_generate(spa, tvd, B_FALSE, VDEV_CONFIG_MISSING); } } if (idx > 0) { fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, (const nvlist_t * const *)child, idx); fnvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_MISSING_DEVICES, nv); for (uint64_t i = 0; i < idx; i++) nvlist_free(child[i]); } nvlist_free(nv); kmem_free(child, rvd->vdev_children * sizeof (char **)); if (idx > 0) { spa_load_failed(spa, "some log devices are missing"); vdev_dbgmsg_print_tree(rvd, 2); return (SET_ERROR(ENXIO)); } } else { for (uint64_t c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; if (tvd->vdev_islog && tvd->vdev_state == VDEV_STATE_CANT_OPEN) { spa_set_log_state(spa, SPA_LOG_CLEAR); spa_load_note(spa, "some log devices are " "missing, ZIL is dropped."); vdev_dbgmsg_print_tree(rvd, 2); break; } } } return (0); } /* * Check for missing log devices */ static boolean_t spa_check_logs(spa_t *spa) { boolean_t rv = B_FALSE; dsl_pool_t *dp = spa_get_dsl(spa); switch (spa->spa_log_state) { default: break; case SPA_LOG_MISSING: /* need to recheck in case slog has been restored */ case SPA_LOG_UNKNOWN: rv = (dmu_objset_find_dp(dp, dp->dp_root_dir_obj, zil_check_log_chain, NULL, DS_FIND_CHILDREN) != 0); if (rv) spa_set_log_state(spa, SPA_LOG_MISSING); break; } return (rv); } /* * Passivate any log vdevs (note, does not apply to embedded log metaslabs). */ static boolean_t spa_passivate_log(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; boolean_t slog_found = B_FALSE; ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; if (tvd->vdev_islog) { ASSERT3P(tvd->vdev_log_mg, ==, NULL); metaslab_group_passivate(tvd->vdev_mg); slog_found = B_TRUE; } } return (slog_found); } /* * Activate any log vdevs (note, does not apply to embedded log metaslabs). */ static void spa_activate_log(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; if (tvd->vdev_islog) { ASSERT3P(tvd->vdev_log_mg, ==, NULL); metaslab_group_activate(tvd->vdev_mg); } } } int spa_reset_logs(spa_t *spa) { int error; error = dmu_objset_find(spa_name(spa), zil_reset, NULL, DS_FIND_CHILDREN); if (error == 0) { /* * We successfully offlined the log device, sync out the * current txg so that the "stubby" block can be removed * by zil_sync(). */ txg_wait_synced(spa->spa_dsl_pool, 0); } return (error); } static void spa_aux_check_removed(spa_aux_vdev_t *sav) { for (int i = 0; i < sav->sav_count; i++) spa_check_removed(sav->sav_vdevs[i]); } void spa_claim_notify(zio_t *zio) { spa_t *spa = zio->io_spa; if (zio->io_error) return; mutex_enter(&spa->spa_props_lock); /* any mutex will do */ if (spa->spa_claim_max_txg < BP_GET_LOGICAL_BIRTH(zio->io_bp)) spa->spa_claim_max_txg = BP_GET_LOGICAL_BIRTH(zio->io_bp); mutex_exit(&spa->spa_props_lock); } typedef struct spa_load_error { boolean_t sle_verify_data; uint64_t sle_meta_count; uint64_t sle_data_count; } spa_load_error_t; static void spa_load_verify_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; spa_load_error_t *sle = zio->io_private; dmu_object_type_t type = BP_GET_TYPE(bp); int error = zio->io_error; spa_t *spa = zio->io_spa; abd_free(zio->io_abd); if (error) { if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) && type != DMU_OT_INTENT_LOG) atomic_inc_64(&sle->sle_meta_count); else atomic_inc_64(&sle->sle_data_count); } mutex_enter(&spa->spa_scrub_lock); spa->spa_load_verify_bytes -= BP_GET_PSIZE(bp); cv_broadcast(&spa->spa_scrub_io_cv); mutex_exit(&spa->spa_scrub_lock); } /* * Maximum number of inflight bytes is the log2 fraction of the arc size. * By default, we set it to 1/16th of the arc. */ static uint_t spa_load_verify_shift = 4; static int spa_load_verify_metadata = B_TRUE; static int spa_load_verify_data = B_TRUE; static int spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg) { zio_t *rio = arg; spa_load_error_t *sle = rio->io_private; (void) zilog, (void) dnp; /* * Note: normally this routine will not be called if * spa_load_verify_metadata is not set. However, it may be useful * to manually set the flag after the traversal has begun. */ if (!spa_load_verify_metadata) return (0); /* * Sanity check the block pointer in order to detect obvious damage * before using the contents in subsequent checks or in zio_read(). * When damaged consider it to be a metadata error since we cannot * trust the BP_GET_TYPE and BP_GET_LEVEL values. */ if (!zfs_blkptr_verify(spa, bp, BLK_CONFIG_NEEDED, BLK_VERIFY_LOG)) { atomic_inc_64(&sle->sle_meta_count); return (0); } if (zb->zb_level == ZB_DNODE_LEVEL || BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp) || BP_IS_REDACTED(bp)) return (0); if (!BP_IS_METADATA(bp) && (!spa_load_verify_data || !sle->sle_verify_data)) return (0); uint64_t maxinflight_bytes = arc_target_bytes() >> spa_load_verify_shift; size_t size = BP_GET_PSIZE(bp); mutex_enter(&spa->spa_scrub_lock); while (spa->spa_load_verify_bytes >= maxinflight_bytes) cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock); spa->spa_load_verify_bytes += size; mutex_exit(&spa->spa_scrub_lock); zio_nowait(zio_read(rio, spa, bp, abd_alloc_for_io(size, B_FALSE), size, spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB, ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb)); return (0); } static int verify_dataset_name_len(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg) { (void) dp, (void) arg; if (dsl_dataset_namelen(ds) >= ZFS_MAX_DATASET_NAME_LEN) return (SET_ERROR(ENAMETOOLONG)); return (0); } static int spa_load_verify(spa_t *spa) { zio_t *rio; spa_load_error_t sle = { 0 }; zpool_load_policy_t policy; boolean_t verify_ok = B_FALSE; int error = 0; zpool_get_load_policy(spa->spa_config, &policy); if (policy.zlp_rewind & ZPOOL_NEVER_REWIND || policy.zlp_maxmeta == UINT64_MAX) return (0); dsl_pool_config_enter(spa->spa_dsl_pool, FTAG); error = dmu_objset_find_dp(spa->spa_dsl_pool, spa->spa_dsl_pool->dp_root_dir_obj, verify_dataset_name_len, NULL, DS_FIND_CHILDREN); dsl_pool_config_exit(spa->spa_dsl_pool, FTAG); if (error != 0) return (error); /* * Verify data only if we are rewinding or error limit was set. * Otherwise nothing except dbgmsg care about it to waste time. */ sle.sle_verify_data = (policy.zlp_rewind & ZPOOL_REWIND_MASK) || (policy.zlp_maxdata < UINT64_MAX); rio = zio_root(spa, NULL, &sle, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE); if (spa_load_verify_metadata) { if (spa->spa_extreme_rewind) { spa_load_note(spa, "performing a complete scan of the " "pool since extreme rewind is on. This may take " "a very long time.\n (spa_load_verify_data=%u, " "spa_load_verify_metadata=%u)", spa_load_verify_data, spa_load_verify_metadata); } error = traverse_pool(spa, spa->spa_verify_min_txg, TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA | TRAVERSE_NO_DECRYPT, spa_load_verify_cb, rio); } (void) zio_wait(rio); ASSERT0(spa->spa_load_verify_bytes); spa->spa_load_meta_errors = sle.sle_meta_count; spa->spa_load_data_errors = sle.sle_data_count; if (sle.sle_meta_count != 0 || sle.sle_data_count != 0) { spa_load_note(spa, "spa_load_verify found %llu metadata errors " "and %llu data errors", (u_longlong_t)sle.sle_meta_count, (u_longlong_t)sle.sle_data_count); } if (spa_load_verify_dryrun || (!error && sle.sle_meta_count <= policy.zlp_maxmeta && sle.sle_data_count <= policy.zlp_maxdata)) { int64_t loss = 0; verify_ok = B_TRUE; spa->spa_load_txg = spa->spa_uberblock.ub_txg; spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp; loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts; fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_TIME, spa->spa_load_txg_ts); fnvlist_add_int64(spa->spa_load_info, ZPOOL_CONFIG_REWIND_TIME, loss); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_META_ERRORS, sle.sle_meta_count); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count); } else { spa->spa_load_max_txg = spa->spa_uberblock.ub_txg; } if (spa_load_verify_dryrun) return (0); if (error) { if (error != ENXIO && error != EIO) error = SET_ERROR(EIO); return (error); } return (verify_ok ? 0 : EIO); } /* * Find a value in the pool props object. */ static void spa_prop_find(spa_t *spa, zpool_prop_t prop, uint64_t *val) { (void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object, zpool_prop_to_name(prop), sizeof (uint64_t), 1, val); } /* * Find a value in the pool directory object. */ static int spa_dir_prop(spa_t *spa, const char *name, uint64_t *val, boolean_t log_enoent) { int error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, name, sizeof (uint64_t), 1, val); if (error != 0 && (error != ENOENT || log_enoent)) { spa_load_failed(spa, "couldn't get '%s' value in MOS directory " "[error=%d]", name, error); } return (error); } static int spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err) { vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux); return (SET_ERROR(err)); } boolean_t spa_livelist_delete_check(spa_t *spa) { return (spa->spa_livelists_to_delete != 0); } static boolean_t spa_livelist_delete_cb_check(void *arg, zthr_t *z) { (void) z; spa_t *spa = arg; return (spa_livelist_delete_check(spa)); } static int delete_blkptr_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { spa_t *spa = arg; zio_free(spa, tx->tx_txg, bp); dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD, -bp_get_dsize_sync(spa, bp), -BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx); return (0); } static int dsl_get_next_livelist_obj(objset_t *os, uint64_t zap_obj, uint64_t *llp) { int err; zap_cursor_t zc; zap_attribute_t za; zap_cursor_init(&zc, os, zap_obj); err = zap_cursor_retrieve(&zc, &za); zap_cursor_fini(&zc); if (err == 0) *llp = za.za_first_integer; return (err); } /* * Components of livelist deletion that must be performed in syncing * context: freeing block pointers and updating the pool-wide data * structures to indicate how much work is left to do */ typedef struct sublist_delete_arg { spa_t *spa; dsl_deadlist_t *ll; uint64_t key; bplist_t *to_free; } sublist_delete_arg_t; static void sublist_delete_sync(void *arg, dmu_tx_t *tx) { sublist_delete_arg_t *sda = arg; spa_t *spa = sda->spa; dsl_deadlist_t *ll = sda->ll; uint64_t key = sda->key; bplist_t *to_free = sda->to_free; bplist_iterate(to_free, delete_blkptr_cb, spa, tx); dsl_deadlist_remove_entry(ll, key, tx); } typedef struct livelist_delete_arg { spa_t *spa; uint64_t ll_obj; uint64_t zap_obj; } livelist_delete_arg_t; static void livelist_delete_sync(void *arg, dmu_tx_t *tx) { livelist_delete_arg_t *lda = arg; spa_t *spa = lda->spa; uint64_t ll_obj = lda->ll_obj; uint64_t zap_obj = lda->zap_obj; objset_t *mos = spa->spa_meta_objset; uint64_t count; /* free the livelist and decrement the feature count */ VERIFY0(zap_remove_int(mos, zap_obj, ll_obj, tx)); dsl_deadlist_free(mos, ll_obj, tx); spa_feature_decr(spa, SPA_FEATURE_LIVELIST, tx); VERIFY0(zap_count(mos, zap_obj, &count)); if (count == 0) { /* no more livelists to delete */ VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DELETED_CLONES, tx)); VERIFY0(zap_destroy(mos, zap_obj, tx)); spa->spa_livelists_to_delete = 0; spa_notify_waiters(spa); } } /* * Load in the value for the livelist to be removed and open it. Then, * load its first sublist and determine which block pointers should actually * be freed. Then, call a synctask which performs the actual frees and updates * the pool-wide livelist data. */ static void spa_livelist_delete_cb(void *arg, zthr_t *z) { spa_t *spa = arg; uint64_t ll_obj = 0, count; objset_t *mos = spa->spa_meta_objset; uint64_t zap_obj = spa->spa_livelists_to_delete; /* * Determine the next livelist to delete. This function should only * be called if there is at least one deleted clone. */ VERIFY0(dsl_get_next_livelist_obj(mos, zap_obj, &ll_obj)); VERIFY0(zap_count(mos, ll_obj, &count)); if (count > 0) { dsl_deadlist_t *ll; dsl_deadlist_entry_t *dle; bplist_t to_free; ll = kmem_zalloc(sizeof (dsl_deadlist_t), KM_SLEEP); dsl_deadlist_open(ll, mos, ll_obj); dle = dsl_deadlist_first(ll); ASSERT3P(dle, !=, NULL); bplist_create(&to_free); int err = dsl_process_sub_livelist(&dle->dle_bpobj, &to_free, z, NULL); if (err == 0) { sublist_delete_arg_t sync_arg = { .spa = spa, .ll = ll, .key = dle->dle_mintxg, .to_free = &to_free }; zfs_dbgmsg("deleting sublist (id %llu) from" " livelist %llu, %lld remaining", (u_longlong_t)dle->dle_bpobj.bpo_object, (u_longlong_t)ll_obj, (longlong_t)count - 1); VERIFY0(dsl_sync_task(spa_name(spa), NULL, sublist_delete_sync, &sync_arg, 0, ZFS_SPACE_CHECK_DESTROY)); } else { VERIFY3U(err, ==, EINTR); } bplist_clear(&to_free); bplist_destroy(&to_free); dsl_deadlist_close(ll); kmem_free(ll, sizeof (dsl_deadlist_t)); } else { livelist_delete_arg_t sync_arg = { .spa = spa, .ll_obj = ll_obj, .zap_obj = zap_obj }; zfs_dbgmsg("deletion of livelist %llu completed", (u_longlong_t)ll_obj); VERIFY0(dsl_sync_task(spa_name(spa), NULL, livelist_delete_sync, &sync_arg, 0, ZFS_SPACE_CHECK_DESTROY)); } } static void spa_start_livelist_destroy_thread(spa_t *spa) { ASSERT3P(spa->spa_livelist_delete_zthr, ==, NULL); spa->spa_livelist_delete_zthr = zthr_create("z_livelist_destroy", spa_livelist_delete_cb_check, spa_livelist_delete_cb, spa, minclsyspri); } typedef struct livelist_new_arg { bplist_t *allocs; bplist_t *frees; } livelist_new_arg_t; static int livelist_track_new_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx) { ASSERT(tx == NULL); livelist_new_arg_t *lna = arg; if (bp_freed) { bplist_append(lna->frees, bp); } else { bplist_append(lna->allocs, bp); zfs_livelist_condense_new_alloc++; } return (0); } typedef struct livelist_condense_arg { spa_t *spa; bplist_t to_keep; uint64_t first_size; uint64_t next_size; } livelist_condense_arg_t; static void spa_livelist_condense_sync(void *arg, dmu_tx_t *tx) { livelist_condense_arg_t *lca = arg; spa_t *spa = lca->spa; bplist_t new_frees; dsl_dataset_t *ds = spa->spa_to_condense.ds; /* Have we been cancelled? */ if (spa->spa_to_condense.cancelled) { zfs_livelist_condense_sync_cancel++; goto out; } dsl_deadlist_entry_t *first = spa->spa_to_condense.first; dsl_deadlist_entry_t *next = spa->spa_to_condense.next; dsl_deadlist_t *ll = &ds->ds_dir->dd_livelist; /* * It's possible that the livelist was changed while the zthr was * running. Therefore, we need to check for new blkptrs in the two * entries being condensed and continue to track them in the livelist. * Because of the way we handle remapped blkptrs (see dbuf_remap_impl), * it's possible that the newly added blkptrs are FREEs or ALLOCs so * we need to sort them into two different bplists. */ uint64_t first_obj = first->dle_bpobj.bpo_object; uint64_t next_obj = next->dle_bpobj.bpo_object; uint64_t cur_first_size = first->dle_bpobj.bpo_phys->bpo_num_blkptrs; uint64_t cur_next_size = next->dle_bpobj.bpo_phys->bpo_num_blkptrs; bplist_create(&new_frees); livelist_new_arg_t new_bps = { .allocs = &lca->to_keep, .frees = &new_frees, }; if (cur_first_size > lca->first_size) { VERIFY0(livelist_bpobj_iterate_from_nofree(&first->dle_bpobj, livelist_track_new_cb, &new_bps, lca->first_size)); } if (cur_next_size > lca->next_size) { VERIFY0(livelist_bpobj_iterate_from_nofree(&next->dle_bpobj, livelist_track_new_cb, &new_bps, lca->next_size)); } dsl_deadlist_clear_entry(first, ll, tx); ASSERT(bpobj_is_empty(&first->dle_bpobj)); dsl_deadlist_remove_entry(ll, next->dle_mintxg, tx); bplist_iterate(&lca->to_keep, dsl_deadlist_insert_alloc_cb, ll, tx); bplist_iterate(&new_frees, dsl_deadlist_insert_free_cb, ll, tx); bplist_destroy(&new_frees); char dsname[ZFS_MAX_DATASET_NAME_LEN]; dsl_dataset_name(ds, dsname); zfs_dbgmsg("txg %llu condensing livelist of %s (id %llu), bpobj %llu " "(%llu blkptrs) and bpobj %llu (%llu blkptrs) -> bpobj %llu " "(%llu blkptrs)", (u_longlong_t)tx->tx_txg, dsname, (u_longlong_t)ds->ds_object, (u_longlong_t)first_obj, (u_longlong_t)cur_first_size, (u_longlong_t)next_obj, (u_longlong_t)cur_next_size, (u_longlong_t)first->dle_bpobj.bpo_object, (u_longlong_t)first->dle_bpobj.bpo_phys->bpo_num_blkptrs); out: dmu_buf_rele(ds->ds_dbuf, spa); spa->spa_to_condense.ds = NULL; bplist_clear(&lca->to_keep); bplist_destroy(&lca->to_keep); kmem_free(lca, sizeof (livelist_condense_arg_t)); spa->spa_to_condense.syncing = B_FALSE; } static void spa_livelist_condense_cb(void *arg, zthr_t *t) { while (zfs_livelist_condense_zthr_pause && !(zthr_has_waiters(t) || zthr_iscancelled(t))) delay(1); spa_t *spa = arg; dsl_deadlist_entry_t *first = spa->spa_to_condense.first; dsl_deadlist_entry_t *next = spa->spa_to_condense.next; uint64_t first_size, next_size; livelist_condense_arg_t *lca = kmem_alloc(sizeof (livelist_condense_arg_t), KM_SLEEP); bplist_create(&lca->to_keep); /* * Process the livelists (matching FREEs and ALLOCs) in open context * so we have minimal work in syncing context to condense. * * We save bpobj sizes (first_size and next_size) to use later in * syncing context to determine if entries were added to these sublists * while in open context. This is possible because the clone is still * active and open for normal writes and we want to make sure the new, * unprocessed blockpointers are inserted into the livelist normally. * * Note that dsl_process_sub_livelist() both stores the size number of * blockpointers and iterates over them while the bpobj's lock held, so * the sizes returned to us are consistent which what was actually * processed. */ int err = dsl_process_sub_livelist(&first->dle_bpobj, &lca->to_keep, t, &first_size); if (err == 0) err = dsl_process_sub_livelist(&next->dle_bpobj, &lca->to_keep, t, &next_size); if (err == 0) { while (zfs_livelist_condense_sync_pause && !(zthr_has_waiters(t) || zthr_iscancelled(t))) delay(1); dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); dmu_tx_mark_netfree(tx); dmu_tx_hold_space(tx, 1); err = dmu_tx_assign(tx, TXG_NOWAIT | TXG_NOTHROTTLE); if (err == 0) { /* * Prevent the condense zthr restarting before * the synctask completes. */ spa->spa_to_condense.syncing = B_TRUE; lca->spa = spa; lca->first_size = first_size; lca->next_size = next_size; dsl_sync_task_nowait(spa_get_dsl(spa), spa_livelist_condense_sync, lca, tx); dmu_tx_commit(tx); return; } } /* * Condensing can not continue: either it was externally stopped or * we were unable to assign to a tx because the pool has run out of * space. In the second case, we'll just end up trying to condense * again in a later txg. */ ASSERT(err != 0); bplist_clear(&lca->to_keep); bplist_destroy(&lca->to_keep); kmem_free(lca, sizeof (livelist_condense_arg_t)); dmu_buf_rele(spa->spa_to_condense.ds->ds_dbuf, spa); spa->spa_to_condense.ds = NULL; if (err == EINTR) zfs_livelist_condense_zthr_cancel++; } /* * Check that there is something to condense but that a condense is not * already in progress and that condensing has not been cancelled. */ static boolean_t spa_livelist_condense_cb_check(void *arg, zthr_t *z) { (void) z; spa_t *spa = arg; if ((spa->spa_to_condense.ds != NULL) && (spa->spa_to_condense.syncing == B_FALSE) && (spa->spa_to_condense.cancelled == B_FALSE)) { return (B_TRUE); } return (B_FALSE); } static void spa_start_livelist_condensing_thread(spa_t *spa) { spa->spa_to_condense.ds = NULL; spa->spa_to_condense.first = NULL; spa->spa_to_condense.next = NULL; spa->spa_to_condense.syncing = B_FALSE; spa->spa_to_condense.cancelled = B_FALSE; ASSERT3P(spa->spa_livelist_condense_zthr, ==, NULL); spa->spa_livelist_condense_zthr = zthr_create("z_livelist_condense", spa_livelist_condense_cb_check, spa_livelist_condense_cb, spa, minclsyspri); } static void spa_spawn_aux_threads(spa_t *spa) { ASSERT(spa_writeable(spa)); - ASSERT(MUTEX_HELD(&spa_namespace_lock)); - spa_start_raidz_expansion_thread(spa); spa_start_indirect_condensing_thread(spa); spa_start_livelist_destroy_thread(spa); spa_start_livelist_condensing_thread(spa); ASSERT3P(spa->spa_checkpoint_discard_zthr, ==, NULL); spa->spa_checkpoint_discard_zthr = zthr_create("z_checkpoint_discard", spa_checkpoint_discard_thread_check, spa_checkpoint_discard_thread, spa, minclsyspri); } /* * Fix up config after a partly-completed split. This is done with the * ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off * pool have that entry in their config, but only the splitting one contains * a list of all the guids of the vdevs that are being split off. * * This function determines what to do with that list: either rejoin * all the disks to the pool, or complete the splitting process. To attempt * the rejoin, each disk that is offlined is marked online again, and * we do a reopen() call. If the vdev label for every disk that was * marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL) * then we call vdev_split() on each disk, and complete the split. * * Otherwise we leave the config alone, with all the vdevs in place in * the original pool. */ static void spa_try_repair(spa_t *spa, nvlist_t *config) { uint_t extracted; uint64_t *glist; uint_t i, gcount; nvlist_t *nvl; vdev_t **vd; boolean_t attempt_reopen; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0) return; /* check that the config is complete */ if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, &glist, &gcount) != 0) return; vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP); /* attempt to online all the vdevs & validate */ attempt_reopen = B_TRUE; for (i = 0; i < gcount; i++) { if (glist[i] == 0) /* vdev is hole */ continue; vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE); if (vd[i] == NULL) { /* * Don't bother attempting to reopen the disks; * just do the split. */ attempt_reopen = B_FALSE; } else { /* attempt to re-online it */ vd[i]->vdev_offline = B_FALSE; } } if (attempt_reopen) { vdev_reopen(spa->spa_root_vdev); /* check each device to see what state it's in */ for (extracted = 0, i = 0; i < gcount; i++) { if (vd[i] != NULL && vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL) break; ++extracted; } } /* * If every disk has been moved to the new pool, or if we never * even attempted to look at them, then we split them off for * good. */ if (!attempt_reopen || gcount == extracted) { for (i = 0; i < gcount; i++) if (vd[i] != NULL) vdev_split(vd[i]); vdev_reopen(spa->spa_root_vdev); } kmem_free(vd, gcount * sizeof (vdev_t *)); } static int spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type) { const char *ereport = FM_EREPORT_ZFS_POOL; int error; spa->spa_load_state = state; (void) spa_import_progress_set_state(spa_guid(spa), spa_load_state(spa)); spa_import_progress_set_notes(spa, "spa_load()"); gethrestime(&spa->spa_loaded_ts); error = spa_load_impl(spa, type, &ereport); /* * Don't count references from objsets that are already closed * and are making their way through the eviction process. */ spa_evicting_os_wait(spa); spa->spa_minref = zfs_refcount_count(&spa->spa_refcount); if (error) { if (error != EEXIST) { spa->spa_loaded_ts.tv_sec = 0; spa->spa_loaded_ts.tv_nsec = 0; } if (error != EBADF) { (void) zfs_ereport_post(ereport, spa, NULL, NULL, NULL, 0); } } spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE; spa->spa_ena = 0; (void) spa_import_progress_set_state(spa_guid(spa), spa_load_state(spa)); return (error); } #ifdef ZFS_DEBUG /* * Count the number of per-vdev ZAPs associated with all of the vdevs in the * vdev tree rooted in the given vd, and ensure that each ZAP is present in the * spa's per-vdev ZAP list. */ static uint64_t vdev_count_verify_zaps(vdev_t *vd) { spa_t *spa = vd->vdev_spa; uint64_t total = 0; if (spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_AVZ_V2) && vd->vdev_root_zap != 0) { total++; ASSERT0(zap_lookup_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps, vd->vdev_root_zap)); } if (vd->vdev_top_zap != 0) { total++; ASSERT0(zap_lookup_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps, vd->vdev_top_zap)); } if (vd->vdev_leaf_zap != 0) { total++; ASSERT0(zap_lookup_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps, vd->vdev_leaf_zap)); } for (uint64_t i = 0; i < vd->vdev_children; i++) { total += vdev_count_verify_zaps(vd->vdev_child[i]); } return (total); } #else #define vdev_count_verify_zaps(vd) ((void) sizeof (vd), 0) #endif /* * Determine whether the activity check is required. */ static boolean_t spa_activity_check_required(spa_t *spa, uberblock_t *ub, nvlist_t *label, nvlist_t *config) { uint64_t state = 0; uint64_t hostid = 0; uint64_t tryconfig_txg = 0; uint64_t tryconfig_timestamp = 0; uint16_t tryconfig_mmp_seq = 0; nvlist_t *nvinfo; if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) { nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO); (void) nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG, &tryconfig_txg); (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP, &tryconfig_timestamp); (void) nvlist_lookup_uint16(nvinfo, ZPOOL_CONFIG_MMP_SEQ, &tryconfig_mmp_seq); } (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state); /* * Disable the MMP activity check - This is used by zdb which * is intended to be used on potentially active pools. */ if (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) return (B_FALSE); /* * Skip the activity check when the MMP feature is disabled. */ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay == 0) return (B_FALSE); /* * If the tryconfig_ values are nonzero, they are the results of an * earlier tryimport. If they all match the uberblock we just found, * then the pool has not changed and we return false so we do not test * a second time. */ if (tryconfig_txg && tryconfig_txg == ub->ub_txg && tryconfig_timestamp && tryconfig_timestamp == ub->ub_timestamp && tryconfig_mmp_seq && tryconfig_mmp_seq == (MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0)) return (B_FALSE); /* * Allow the activity check to be skipped when importing the pool * on the same host which last imported it. Since the hostid from * configuration may be stale use the one read from the label. */ if (nvlist_exists(label, ZPOOL_CONFIG_HOSTID)) hostid = fnvlist_lookup_uint64(label, ZPOOL_CONFIG_HOSTID); if (hostid == spa_get_hostid(spa)) return (B_FALSE); /* * Skip the activity test when the pool was cleanly exported. */ if (state != POOL_STATE_ACTIVE) return (B_FALSE); return (B_TRUE); } /* * Nanoseconds the activity check must watch for changes on-disk. */ static uint64_t spa_activity_check_duration(spa_t *spa, uberblock_t *ub) { uint64_t import_intervals = MAX(zfs_multihost_import_intervals, 1); uint64_t multihost_interval = MSEC2NSEC( MMP_INTERVAL_OK(zfs_multihost_interval)); uint64_t import_delay = MAX(NANOSEC, import_intervals * multihost_interval); /* * Local tunables determine a minimum duration except for the case * where we know when the remote host will suspend the pool if MMP * writes do not land. * * See Big Theory comment at the top of mmp.c for the reasoning behind * these cases and times. */ ASSERT(MMP_IMPORT_SAFETY_FACTOR >= 100); if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) && MMP_FAIL_INT(ub) > 0) { /* MMP on remote host will suspend pool after failed writes */ import_delay = MMP_FAIL_INT(ub) * MSEC2NSEC(MMP_INTERVAL(ub)) * MMP_IMPORT_SAFETY_FACTOR / 100; zfs_dbgmsg("fail_intvals>0 import_delay=%llu ub_mmp " "mmp_fails=%llu ub_mmp mmp_interval=%llu " "import_intervals=%llu", (u_longlong_t)import_delay, (u_longlong_t)MMP_FAIL_INT(ub), (u_longlong_t)MMP_INTERVAL(ub), (u_longlong_t)import_intervals); } else if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) && MMP_FAIL_INT(ub) == 0) { /* MMP on remote host will never suspend pool */ import_delay = MAX(import_delay, (MSEC2NSEC(MMP_INTERVAL(ub)) + ub->ub_mmp_delay) * import_intervals); zfs_dbgmsg("fail_intvals=0 import_delay=%llu ub_mmp " "mmp_interval=%llu ub_mmp_delay=%llu " "import_intervals=%llu", (u_longlong_t)import_delay, (u_longlong_t)MMP_INTERVAL(ub), (u_longlong_t)ub->ub_mmp_delay, (u_longlong_t)import_intervals); } else if (MMP_VALID(ub)) { /* * zfs-0.7 compatibility case */ import_delay = MAX(import_delay, (multihost_interval + ub->ub_mmp_delay) * import_intervals); zfs_dbgmsg("import_delay=%llu ub_mmp_delay=%llu " "import_intervals=%llu leaves=%u", (u_longlong_t)import_delay, (u_longlong_t)ub->ub_mmp_delay, (u_longlong_t)import_intervals, vdev_count_leaves(spa)); } else { /* Using local tunings is the only reasonable option */ zfs_dbgmsg("pool last imported on non-MMP aware " "host using import_delay=%llu multihost_interval=%llu " "import_intervals=%llu", (u_longlong_t)import_delay, (u_longlong_t)multihost_interval, (u_longlong_t)import_intervals); } return (import_delay); } /* * Perform the import activity check. If the user canceled the import or * we detected activity then fail. */ static int spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config) { uint64_t txg = ub->ub_txg; uint64_t timestamp = ub->ub_timestamp; uint64_t mmp_config = ub->ub_mmp_config; uint16_t mmp_seq = MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0; uint64_t import_delay; hrtime_t import_expire, now; nvlist_t *mmp_label = NULL; vdev_t *rvd = spa->spa_root_vdev; kcondvar_t cv; kmutex_t mtx; int error = 0; cv_init(&cv, NULL, CV_DEFAULT, NULL); mutex_init(&mtx, NULL, MUTEX_DEFAULT, NULL); mutex_enter(&mtx); /* * If ZPOOL_CONFIG_MMP_TXG is present an activity check was performed * during the earlier tryimport. If the txg recorded there is 0 then * the pool is known to be active on another host. * * Otherwise, the pool might be in use on another host. Check for * changes in the uberblocks on disk if necessary. */ if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) { nvlist_t *nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO); if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_TXG) && fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG) == 0) { vdev_uberblock_load(rvd, ub, &mmp_label); error = SET_ERROR(EREMOTEIO); goto out; } } import_delay = spa_activity_check_duration(spa, ub); /* Add a small random factor in case of simultaneous imports (0-25%) */ import_delay += import_delay * random_in_range(250) / 1000; import_expire = gethrtime() + import_delay; spa_import_progress_set_notes(spa, "Checking MMP activity, waiting " "%llu ms", (u_longlong_t)NSEC2MSEC(import_delay)); int interations = 0; while ((now = gethrtime()) < import_expire) { if (interations++ % 30 == 0) { spa_import_progress_set_notes(spa, "Checking MMP " "activity, %llu ms remaining", (u_longlong_t)NSEC2MSEC(import_expire - now)); } (void) spa_import_progress_set_mmp_check(spa_guid(spa), NSEC2SEC(import_expire - gethrtime())); vdev_uberblock_load(rvd, ub, &mmp_label); if (txg != ub->ub_txg || timestamp != ub->ub_timestamp || mmp_seq != (MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0)) { zfs_dbgmsg("multihost activity detected " "txg %llu ub_txg %llu " "timestamp %llu ub_timestamp %llu " "mmp_config %#llx ub_mmp_config %#llx", (u_longlong_t)txg, (u_longlong_t)ub->ub_txg, (u_longlong_t)timestamp, (u_longlong_t)ub->ub_timestamp, (u_longlong_t)mmp_config, (u_longlong_t)ub->ub_mmp_config); error = SET_ERROR(EREMOTEIO); break; } if (mmp_label) { nvlist_free(mmp_label); mmp_label = NULL; } error = cv_timedwait_sig(&cv, &mtx, ddi_get_lbolt() + hz); if (error != -1) { error = SET_ERROR(EINTR); break; } error = 0; } out: mutex_exit(&mtx); mutex_destroy(&mtx); cv_destroy(&cv); /* * If the pool is determined to be active store the status in the * spa->spa_load_info nvlist. If the remote hostname or hostid are * available from configuration read from disk store them as well. * This allows 'zpool import' to generate a more useful message. * * ZPOOL_CONFIG_MMP_STATE - observed pool status (mandatory) * ZPOOL_CONFIG_MMP_HOSTNAME - hostname from the active pool * ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool */ if (error == EREMOTEIO) { const char *hostname = ""; uint64_t hostid = 0; if (mmp_label) { if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) { hostname = fnvlist_lookup_string(mmp_label, ZPOOL_CONFIG_HOSTNAME); fnvlist_add_string(spa->spa_load_info, ZPOOL_CONFIG_MMP_HOSTNAME, hostname); } if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) { hostid = fnvlist_lookup_uint64(mmp_label, ZPOOL_CONFIG_HOSTID); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_HOSTID, hostid); } } fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_STATE, MMP_STATE_ACTIVE); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_TXG, 0); error = spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO); } if (mmp_label) nvlist_free(mmp_label); return (error); } static int spa_verify_host(spa_t *spa, nvlist_t *mos_config) { uint64_t hostid; const char *hostname; uint64_t myhostid = 0; if (!spa_is_root(spa) && nvlist_lookup_uint64(mos_config, ZPOOL_CONFIG_HOSTID, &hostid) == 0) { hostname = fnvlist_lookup_string(mos_config, ZPOOL_CONFIG_HOSTNAME); myhostid = zone_get_hostid(NULL); if (hostid != 0 && myhostid != 0 && hostid != myhostid) { cmn_err(CE_WARN, "pool '%s' could not be " "loaded as it was last accessed by " "another system (host: %s hostid: 0x%llx). " "See: https://openzfs.github.io/openzfs-docs/msg/" "ZFS-8000-EY", spa_name(spa), hostname, (u_longlong_t)hostid); spa_load_failed(spa, "hostid verification failed: pool " "last accessed by host: %s (hostid: 0x%llx)", hostname, (u_longlong_t)hostid); return (SET_ERROR(EBADF)); } } return (0); } static int spa_ld_parse_config(spa_t *spa, spa_import_type_t type) { int error = 0; nvlist_t *nvtree, *nvl, *config = spa->spa_config; int parse; vdev_t *rvd; uint64_t pool_guid; const char *comment; const char *compatibility; /* * Versioning wasn't explicitly added to the label until later, so if * it's not present treat it as the initial version. */ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &spa->spa_ubsync.ub_version) != 0) spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL; if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) { spa_load_failed(spa, "invalid config provided: '%s' missing", ZPOOL_CONFIG_POOL_GUID); return (SET_ERROR(EINVAL)); } /* * If we are doing an import, ensure that the pool is not already * imported by checking if its pool guid already exists in the * spa namespace. * * The only case that we allow an already imported pool to be * imported again, is when the pool is checkpointed and we want to * look at its checkpointed state from userland tools like zdb. */ #ifdef _KERNEL if ((spa->spa_load_state == SPA_LOAD_IMPORT || spa->spa_load_state == SPA_LOAD_TRYIMPORT) && spa_guid_exists(pool_guid, 0)) { #else if ((spa->spa_load_state == SPA_LOAD_IMPORT || spa->spa_load_state == SPA_LOAD_TRYIMPORT) && spa_guid_exists(pool_guid, 0) && !spa_importing_readonly_checkpoint(spa)) { #endif spa_load_failed(spa, "a pool with guid %llu is already open", (u_longlong_t)pool_guid); return (SET_ERROR(EEXIST)); } spa->spa_config_guid = pool_guid; nvlist_free(spa->spa_load_info); spa->spa_load_info = fnvlist_alloc(); ASSERT(spa->spa_comment == NULL); if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0) spa->spa_comment = spa_strdup(comment); ASSERT(spa->spa_compatibility == NULL); if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMPATIBILITY, &compatibility) == 0) spa->spa_compatibility = spa_strdup(compatibility); (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &spa->spa_config_txg); if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) == 0) spa->spa_config_splitting = fnvlist_dup(nvl); if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtree)) { spa_load_failed(spa, "invalid config provided: '%s' missing", ZPOOL_CONFIG_VDEV_TREE); return (SET_ERROR(EINVAL)); } /* * Create "The Godfather" zio to hold all async IOs */ spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *), KM_SLEEP); for (int i = 0; i < max_ncpus; i++) { spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); } /* * Parse the configuration into a vdev tree. We explicitly set the * value that will be returned by spa_version() since parsing the * configuration requires knowing the version number. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); parse = (type == SPA_IMPORT_EXISTING ? VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT); error = spa_config_parse(spa, &rvd, nvtree, NULL, 0, parse); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) { spa_load_failed(spa, "unable to parse config [error=%d]", error); return (error); } ASSERT(spa->spa_root_vdev == rvd); ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT); ASSERT3U(spa->spa_max_ashift, <=, SPA_MAXBLOCKSHIFT); if (type != SPA_IMPORT_ASSEMBLE) { ASSERT(spa_guid(spa) == pool_guid); } return (0); } /* * Recursively open all vdevs in the vdev tree. This function is called twice: * first with the untrusted config, then with the trusted config. */ static int spa_ld_open_vdevs(spa_t *spa) { int error = 0; /* * spa_missing_tvds_allowed defines how many top-level vdevs can be * missing/unopenable for the root vdev to be still considered openable. */ if (spa->spa_trust_config) { spa->spa_missing_tvds_allowed = zfs_max_missing_tvds; } else if (spa->spa_config_source == SPA_CONFIG_SRC_CACHEFILE) { spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_cachefile; } else if (spa->spa_config_source == SPA_CONFIG_SRC_SCAN) { spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_scan; } else { spa->spa_missing_tvds_allowed = 0; } spa->spa_missing_tvds_allowed = MAX(zfs_max_missing_tvds, spa->spa_missing_tvds_allowed); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = vdev_open(spa->spa_root_vdev); spa_config_exit(spa, SCL_ALL, FTAG); if (spa->spa_missing_tvds != 0) { spa_load_note(spa, "vdev tree has %lld missing top-level " "vdevs.", (u_longlong_t)spa->spa_missing_tvds); if (spa->spa_trust_config && (spa->spa_mode & SPA_MODE_WRITE)) { /* * Although theoretically we could allow users to open * incomplete pools in RW mode, we'd need to add a lot * of extra logic (e.g. adjust pool space to account * for missing vdevs). * This limitation also prevents users from accidentally * opening the pool in RW mode during data recovery and * damaging it further. */ spa_load_note(spa, "pools with missing top-level " "vdevs can only be opened in read-only mode."); error = SET_ERROR(ENXIO); } else { spa_load_note(spa, "current settings allow for maximum " "%lld missing top-level vdevs at this stage.", (u_longlong_t)spa->spa_missing_tvds_allowed); } } if (error != 0) { spa_load_failed(spa, "unable to open vdev tree [error=%d]", error); } if (spa->spa_missing_tvds != 0 || error != 0) vdev_dbgmsg_print_tree(spa->spa_root_vdev, 2); return (error); } /* * We need to validate the vdev labels against the configuration that * we have in hand. This function is called twice: first with an untrusted * config, then with a trusted config. The validation is more strict when the * config is trusted. */ static int spa_ld_validate_vdevs(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = vdev_validate(rvd); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) { spa_load_failed(spa, "vdev_validate failed [error=%d]", error); return (error); } if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) { spa_load_failed(spa, "cannot open vdev tree after invalidating " "some vdevs"); vdev_dbgmsg_print_tree(rvd, 2); return (SET_ERROR(ENXIO)); } return (0); } static void spa_ld_select_uberblock_done(spa_t *spa, uberblock_t *ub) { spa->spa_state = POOL_STATE_ACTIVE; spa->spa_ubsync = spa->spa_uberblock; spa->spa_verify_min_txg = spa->spa_extreme_rewind ? TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1; spa->spa_first_txg = spa->spa_last_ubsync_txg ? spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1; spa->spa_claim_max_txg = spa->spa_first_txg; spa->spa_prev_software_version = ub->ub_software_version; } static int spa_ld_select_uberblock(spa_t *spa, spa_import_type_t type) { vdev_t *rvd = spa->spa_root_vdev; nvlist_t *label; uberblock_t *ub = &spa->spa_uberblock; boolean_t activity_check = B_FALSE; /* * If we are opening the checkpointed state of the pool by * rewinding to it, at this point we will have written the * checkpointed uberblock to the vdev labels, so searching * the labels will find the right uberblock. However, if * we are opening the checkpointed state read-only, we have * not modified the labels. Therefore, we must ignore the * labels and continue using the spa_uberblock that was set * by spa_ld_checkpoint_rewind. * * Note that it would be fine to ignore the labels when * rewinding (opening writeable) as well. However, if we * crash just after writing the labels, we will end up * searching the labels. Doing so in the common case means * that this code path gets exercised normally, rather than * just in the edge case. */ if (ub->ub_checkpoint_txg != 0 && spa_importing_readonly_checkpoint(spa)) { spa_ld_select_uberblock_done(spa, ub); return (0); } /* * Find the best uberblock. */ vdev_uberblock_load(rvd, ub, &label); /* * If we weren't able to find a single valid uberblock, return failure. */ if (ub->ub_txg == 0) { nvlist_free(label); spa_load_failed(spa, "no valid uberblock found"); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); } if (spa->spa_load_max_txg != UINT64_MAX) { (void) spa_import_progress_set_max_txg(spa_guid(spa), (u_longlong_t)spa->spa_load_max_txg); } spa_load_note(spa, "using uberblock with txg=%llu", (u_longlong_t)ub->ub_txg); if (ub->ub_raidz_reflow_info != 0) { spa_load_note(spa, "uberblock raidz_reflow_info: " "state=%u offset=%llu", (int)RRSS_GET_STATE(ub), (u_longlong_t)RRSS_GET_OFFSET(ub)); } /* * For pools which have the multihost property on determine if the * pool is truly inactive and can be safely imported. Prevent * hosts which don't have a hostid set from importing the pool. */ activity_check = spa_activity_check_required(spa, ub, label, spa->spa_config); if (activity_check) { if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay && spa_get_hostid(spa) == 0) { nvlist_free(label); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID); return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO)); } int error = spa_activity_check(spa, ub, spa->spa_config); if (error) { nvlist_free(label); return (error); } fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_STATE, MMP_STATE_INACTIVE); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_TXG, ub->ub_txg); fnvlist_add_uint16(spa->spa_load_info, ZPOOL_CONFIG_MMP_SEQ, (MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0)); } /* * If the pool has an unsupported version we can't open it. */ if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) { nvlist_free(label); spa_load_failed(spa, "version %llu is not supported", (u_longlong_t)ub->ub_version); return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP)); } if (ub->ub_version >= SPA_VERSION_FEATURES) { nvlist_t *features; /* * If we weren't able to find what's necessary for reading the * MOS in the label, return failure. */ if (label == NULL) { spa_load_failed(spa, "label config unavailable"); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); } if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_FEATURES_FOR_READ, &features) != 0) { nvlist_free(label); spa_load_failed(spa, "invalid label: '%s' missing", ZPOOL_CONFIG_FEATURES_FOR_READ); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); } /* * Update our in-core representation with the definitive values * from the label. */ nvlist_free(spa->spa_label_features); spa->spa_label_features = fnvlist_dup(features); } nvlist_free(label); /* * Look through entries in the label nvlist's features_for_read. If * there is a feature listed there which we don't understand then we * cannot open a pool. */ if (ub->ub_version >= SPA_VERSION_FEATURES) { nvlist_t *unsup_feat; unsup_feat = fnvlist_alloc(); for (nvpair_t *nvp = nvlist_next_nvpair(spa->spa_label_features, NULL); nvp != NULL; nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) { if (!zfeature_is_supported(nvpair_name(nvp))) { fnvlist_add_string(unsup_feat, nvpair_name(nvp), ""); } } if (!nvlist_empty(unsup_feat)) { fnvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat); nvlist_free(unsup_feat); spa_load_failed(spa, "some features are unsupported"); return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP)); } nvlist_free(unsup_feat); } if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_try_repair(spa, spa->spa_config); spa_config_exit(spa, SCL_ALL, FTAG); nvlist_free(spa->spa_config_splitting); spa->spa_config_splitting = NULL; } /* * Initialize internal SPA structures. */ spa_ld_select_uberblock_done(spa, ub); return (0); } static int spa_ld_open_rootbp(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool); if (error != 0) { spa_load_failed(spa, "unable to open rootbp in dsl_pool_init " "[error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset; return (0); } static int spa_ld_trusted_config(spa_t *spa, spa_import_type_t type, boolean_t reloading) { vdev_t *mrvd, *rvd = spa->spa_root_vdev; nvlist_t *nv, *mos_config, *policy; int error = 0, copy_error; uint64_t healthy_tvds, healthy_tvds_mos; uint64_t mos_config_txg; if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object, B_TRUE) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * If we're assembling a pool from a split, the config provided is * already trusted so there is nothing to do. */ if (type == SPA_IMPORT_ASSEMBLE) return (0); healthy_tvds = spa_healthy_core_tvds(spa); if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0) { spa_load_failed(spa, "unable to retrieve MOS config"); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } /* * If we are doing an open, pool owner wasn't verified yet, thus do * the verification here. */ if (spa->spa_load_state == SPA_LOAD_OPEN) { error = spa_verify_host(spa, mos_config); if (error != 0) { nvlist_free(mos_config); return (error); } } nv = fnvlist_lookup_nvlist(mos_config, ZPOOL_CONFIG_VDEV_TREE); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * Build a new vdev tree from the trusted config */ error = spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD); if (error != 0) { nvlist_free(mos_config); spa_config_exit(spa, SCL_ALL, FTAG); spa_load_failed(spa, "spa_config_parse failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error)); } /* * Vdev paths in the MOS may be obsolete. If the untrusted config was * obtained by scanning /dev/dsk, then it will have the right vdev * paths. We update the trusted MOS config with this information. * We first try to copy the paths with vdev_copy_path_strict, which * succeeds only when both configs have exactly the same vdev tree. * If that fails, we fall back to a more flexible method that has a * best effort policy. */ copy_error = vdev_copy_path_strict(rvd, mrvd); if (copy_error != 0 || spa_load_print_vdev_tree) { spa_load_note(spa, "provided vdev tree:"); vdev_dbgmsg_print_tree(rvd, 2); spa_load_note(spa, "MOS vdev tree:"); vdev_dbgmsg_print_tree(mrvd, 2); } if (copy_error != 0) { spa_load_note(spa, "vdev_copy_path_strict failed, falling " "back to vdev_copy_path_relaxed"); vdev_copy_path_relaxed(rvd, mrvd); } vdev_close(rvd); vdev_free(rvd); spa->spa_root_vdev = mrvd; rvd = mrvd; spa_config_exit(spa, SCL_ALL, FTAG); /* * If 'zpool import' used a cached config, then the on-disk hostid and * hostname may be different to the cached config in ways that should * prevent import. Userspace can't discover this without a scan, but * we know, so we add these values to LOAD_INFO so the caller can know * the difference. * * Note that we have to do this before the config is regenerated, * because the new config will have the hostid and hostname for this * host, in readiness for import. */ if (nvlist_exists(mos_config, ZPOOL_CONFIG_HOSTID)) fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_HOSTID, fnvlist_lookup_uint64(mos_config, ZPOOL_CONFIG_HOSTID)); if (nvlist_exists(mos_config, ZPOOL_CONFIG_HOSTNAME)) fnvlist_add_string(spa->spa_load_info, ZPOOL_CONFIG_HOSTNAME, fnvlist_lookup_string(mos_config, ZPOOL_CONFIG_HOSTNAME)); /* * We will use spa_config if we decide to reload the spa or if spa_load * fails and we rewind. We must thus regenerate the config using the * MOS information with the updated paths. ZPOOL_LOAD_POLICY is used to * pass settings on how to load the pool and is not stored in the MOS. * We copy it over to our new, trusted config. */ mos_config_txg = fnvlist_lookup_uint64(mos_config, ZPOOL_CONFIG_POOL_TXG); nvlist_free(mos_config); mos_config = spa_config_generate(spa, NULL, mos_config_txg, B_FALSE); if (nvlist_lookup_nvlist(spa->spa_config, ZPOOL_LOAD_POLICY, &policy) == 0) fnvlist_add_nvlist(mos_config, ZPOOL_LOAD_POLICY, policy); spa_config_set(spa, mos_config); spa->spa_config_source = SPA_CONFIG_SRC_MOS; /* * Now that we got the config from the MOS, we should be more strict * in checking blkptrs and can make assumptions about the consistency * of the vdev tree. spa_trust_config must be set to true before opening * vdevs in order for them to be writeable. */ spa->spa_trust_config = B_TRUE; /* * Open and validate the new vdev tree */ error = spa_ld_open_vdevs(spa); if (error != 0) return (error); error = spa_ld_validate_vdevs(spa); if (error != 0) return (error); if (copy_error != 0 || spa_load_print_vdev_tree) { spa_load_note(spa, "final vdev tree:"); vdev_dbgmsg_print_tree(rvd, 2); } if (spa->spa_load_state != SPA_LOAD_TRYIMPORT && !spa->spa_extreme_rewind && zfs_max_missing_tvds == 0) { /* * Sanity check to make sure that we are indeed loading the * latest uberblock. If we missed SPA_SYNC_MIN_VDEVS tvds * in the config provided and they happened to be the only ones * to have the latest uberblock, we could involuntarily perform * an extreme rewind. */ healthy_tvds_mos = spa_healthy_core_tvds(spa); if (healthy_tvds_mos - healthy_tvds >= SPA_SYNC_MIN_VDEVS) { spa_load_note(spa, "config provided misses too many " "top-level vdevs compared to MOS (%lld vs %lld). ", (u_longlong_t)healthy_tvds, (u_longlong_t)healthy_tvds_mos); spa_load_note(spa, "vdev tree:"); vdev_dbgmsg_print_tree(rvd, 2); if (reloading) { spa_load_failed(spa, "config was already " "provided from MOS. Aborting."); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } spa_load_note(spa, "spa must be reloaded using MOS " "config"); return (SET_ERROR(EAGAIN)); } } error = spa_check_for_missing_logs(spa); if (error != 0) return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO)); if (rvd->vdev_guid_sum != spa->spa_uberblock.ub_guid_sum) { spa_load_failed(spa, "uberblock guid sum doesn't match MOS " "guid sum (%llu != %llu)", (u_longlong_t)spa->spa_uberblock.ub_guid_sum, (u_longlong_t)rvd->vdev_guid_sum); return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO)); } return (0); } static int spa_ld_open_indirect_vdev_metadata(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; /* * Everything that we read before spa_remove_init() must be stored * on concreted vdevs. Therefore we do this as early as possible. */ error = spa_remove_init(spa); if (error != 0) { spa_load_failed(spa, "spa_remove_init failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } /* * Retrieve information needed to condense indirect vdev mappings. */ error = spa_condense_init(spa); if (error != 0) { spa_load_failed(spa, "spa_condense_init failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error)); } return (0); } static int spa_ld_check_features(spa_t *spa, boolean_t *missing_feat_writep) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; if (spa_version(spa) >= SPA_VERSION_FEATURES) { boolean_t missing_feat_read = B_FALSE; nvlist_t *unsup_feat, *enabled_feat; if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ, &spa->spa_feat_for_read_obj, B_TRUE) != 0) { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE, &spa->spa_feat_for_write_obj, B_TRUE) != 0) { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS, &spa->spa_feat_desc_obj, B_TRUE) != 0) { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } enabled_feat = fnvlist_alloc(); unsup_feat = fnvlist_alloc(); if (!spa_features_check(spa, B_FALSE, unsup_feat, enabled_feat)) missing_feat_read = B_TRUE; if (spa_writeable(spa) || spa->spa_load_state == SPA_LOAD_TRYIMPORT) { if (!spa_features_check(spa, B_TRUE, unsup_feat, enabled_feat)) { *missing_feat_writep = B_TRUE; } } fnvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat); if (!nvlist_empty(unsup_feat)) { fnvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat); } fnvlist_free(enabled_feat); fnvlist_free(unsup_feat); if (!missing_feat_read) { fnvlist_add_boolean(spa->spa_load_info, ZPOOL_CONFIG_CAN_RDONLY); } /* * If the state is SPA_LOAD_TRYIMPORT, our objective is * twofold: to determine whether the pool is available for * import in read-write mode and (if it is not) whether the * pool is available for import in read-only mode. If the pool * is available for import in read-write mode, it is displayed * as available in userland; if it is not available for import * in read-only mode, it is displayed as unavailable in * userland. If the pool is available for import in read-only * mode but not read-write mode, it is displayed as unavailable * in userland with a special note that the pool is actually * available for open in read-only mode. * * As a result, if the state is SPA_LOAD_TRYIMPORT and we are * missing a feature for write, we must first determine whether * the pool can be opened read-only before returning to * userland in order to know whether to display the * abovementioned note. */ if (missing_feat_read || (*missing_feat_writep && spa_writeable(spa))) { spa_load_failed(spa, "pool uses unsupported features"); return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP)); } /* * Load refcounts for ZFS features from disk into an in-memory * cache during SPA initialization. */ for (spa_feature_t i = 0; i < SPA_FEATURES; i++) { uint64_t refcount; error = feature_get_refcount_from_disk(spa, &spa_feature_table[i], &refcount); if (error == 0) { spa->spa_feat_refcount_cache[i] = refcount; } else if (error == ENOTSUP) { spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; } else { spa_load_failed(spa, "error getting refcount " "for feature %s [error=%d]", spa_feature_table[i].fi_guid, error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } } } if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) { if (spa_dir_prop(spa, DMU_POOL_FEATURE_ENABLED_TXG, &spa->spa_feat_enabled_txg_obj, B_TRUE) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } /* * Encryption was added before bookmark_v2, even though bookmark_v2 * is now a dependency. If this pool has encryption enabled without * bookmark_v2, trigger an errata message. */ if (spa_feature_is_enabled(spa, SPA_FEATURE_ENCRYPTION) && !spa_feature_is_enabled(spa, SPA_FEATURE_BOOKMARK_V2)) { spa->spa_errata = ZPOOL_ERRATA_ZOL_8308_ENCRYPTION; } return (0); } static int spa_ld_load_special_directories(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; spa->spa_is_initializing = B_TRUE; error = dsl_pool_open(spa->spa_dsl_pool); spa->spa_is_initializing = B_FALSE; if (error != 0) { spa_load_failed(spa, "dsl_pool_open failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } return (0); } static int spa_ld_get_props(spa_t *spa) { int error = 0; uint64_t obj; vdev_t *rvd = spa->spa_root_vdev; /* Grab the checksum salt from the MOS. */ error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1, sizeof (spa->spa_cksum_salt.zcs_bytes), spa->spa_cksum_salt.zcs_bytes); if (error == ENOENT) { /* Generate a new salt for subsequent use */ (void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes, sizeof (spa->spa_cksum_salt.zcs_bytes)); } else if (error != 0) { spa_load_failed(spa, "unable to retrieve checksum salt from " "MOS [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj, B_TRUE) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj); if (error != 0) { spa_load_failed(spa, "error opening deferred-frees bpobj " "[error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } /* * Load the bit that tells us to use the new accounting function * (raid-z deflation). If we have an older pool, this will not * be present. */ error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION, &spa->spa_creation_version, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the persistent error log. If we have an older pool, this will * not be present. */ error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB, &spa->spa_errlog_scrub, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the livelist deletion field. If a livelist is queued for * deletion, indicate that in the spa */ error = spa_dir_prop(spa, DMU_POOL_DELETED_CLONES, &spa->spa_livelists_to_delete, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the history object. If we have an older pool, this * will not be present. */ error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the per-vdev ZAP map. If we have an older pool, this will not * be present; in this case, defer its creation to a later time to * avoid dirtying the MOS this early / out of sync context. See * spa_sync_config_object. */ /* The sentinel is only available in the MOS config. */ nvlist_t *mos_config; if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0) { spa_load_failed(spa, "unable to retrieve MOS config"); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP, &spa->spa_all_vdev_zaps, B_FALSE); if (error == ENOENT) { VERIFY(!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)); spa->spa_avz_action = AVZ_ACTION_INITIALIZE; ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev)); } else if (error != 0) { nvlist_free(mos_config); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) { /* * An older version of ZFS overwrote the sentinel value, so * we have orphaned per-vdev ZAPs in the MOS. Defer their * destruction to later; see spa_sync_config_object. */ spa->spa_avz_action = AVZ_ACTION_DESTROY; /* * We're assuming that no vdevs have had their ZAPs created * before this. Better be sure of it. */ ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev)); } nvlist_free(mos_config); spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object, B_FALSE); if (error && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (error == 0) { uint64_t autoreplace = 0; spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs); spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace); spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation); spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode); spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand); spa_prop_find(spa, ZPOOL_PROP_MULTIHOST, &spa->spa_multihost); spa_prop_find(spa, ZPOOL_PROP_AUTOTRIM, &spa->spa_autotrim); spa->spa_autoreplace = (autoreplace != 0); } /* * If we are importing a pool with missing top-level vdevs, * we enforce that the pool doesn't panic or get suspended on * error since the likelihood of missing data is extremely high. */ if (spa->spa_missing_tvds > 0 && spa->spa_failmode != ZIO_FAILURE_MODE_CONTINUE && spa->spa_load_state != SPA_LOAD_TRYIMPORT) { spa_load_note(spa, "forcing failmode to 'continue' " "as some top level vdevs are missing"); spa->spa_failmode = ZIO_FAILURE_MODE_CONTINUE; } return (0); } static int spa_ld_open_aux_vdevs(spa_t *spa, spa_import_type_t type) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; /* * If we're assembling the pool from the split-off vdevs of * an existing pool, we don't want to attach the spares & cache * devices. */ /* * Load any hot spares for this pool. */ error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { ASSERT(spa_version(spa) >= SPA_VERSION_SPARES); if (load_nvlist(spa, spa->spa_spares.sav_object, &spa->spa_spares.sav_config) != 0) { spa_load_failed(spa, "error loading spares nvlist"); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); } else if (error == 0) { spa->spa_spares.sav_sync = B_TRUE; } /* * Load any level 2 ARC devices for this pool. */ error = spa_dir_prop(spa, DMU_POOL_L2CACHE, &spa->spa_l2cache.sav_object, B_FALSE); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE); if (load_nvlist(spa, spa->spa_l2cache.sav_object, &spa->spa_l2cache.sav_config) != 0) { spa_load_failed(spa, "error loading l2cache nvlist"); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); } else if (error == 0) { spa->spa_l2cache.sav_sync = B_TRUE; } return (0); } static int spa_ld_load_vdev_metadata(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; /* * If the 'multihost' property is set, then never allow a pool to * be imported when the system hostid is zero. The exception to * this rule is zdb which is always allowed to access pools. */ if (spa_multihost(spa) && spa_get_hostid(spa) == 0 && (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) == 0) { fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID); return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO)); } /* * If the 'autoreplace' property is set, then post a resource notifying * the ZFS DE that it should not issue any faults for unopenable * devices. We also iterate over the vdevs, and post a sysevent for any * unopenable vdevs so that the normal autoreplace handler can take * over. */ if (spa->spa_autoreplace && spa->spa_load_state != SPA_LOAD_TRYIMPORT) { spa_check_removed(spa->spa_root_vdev); /* * For the import case, this is done in spa_import(), because * at this point we're using the spare definitions from * the MOS config, not necessarily from the userland config. */ if (spa->spa_load_state != SPA_LOAD_IMPORT) { spa_aux_check_removed(&spa->spa_spares); spa_aux_check_removed(&spa->spa_l2cache); } } /* * Load the vdev metadata such as metaslabs, DTLs, spacemap object, etc. */ error = vdev_load(rvd); if (error != 0) { spa_load_failed(spa, "vdev_load failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error)); } error = spa_ld_log_spacemaps(spa); if (error != 0) { spa_load_failed(spa, "spa_ld_log_spacemaps failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error)); } /* * Propagate the leaf DTLs we just loaded all the way up the vdev tree. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); vdev_dtl_reassess(rvd, 0, 0, B_FALSE, B_FALSE); spa_config_exit(spa, SCL_ALL, FTAG); return (0); } static int spa_ld_load_dedup_tables(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; error = ddt_load(spa); if (error != 0) { spa_load_failed(spa, "ddt_load failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } return (0); } static int spa_ld_load_brt(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; error = brt_load(spa); if (error != 0) { spa_load_failed(spa, "brt_load failed [error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } return (0); } static int spa_ld_verify_logs(spa_t *spa, spa_import_type_t type, const char **ereport) { vdev_t *rvd = spa->spa_root_vdev; if (type != SPA_IMPORT_ASSEMBLE && spa_writeable(spa)) { boolean_t missing = spa_check_logs(spa); if (missing) { if (spa->spa_missing_tvds != 0) { spa_load_note(spa, "spa_check_logs failed " "so dropping the logs"); } else { *ereport = FM_EREPORT_ZFS_LOG_REPLAY; spa_load_failed(spa, "spa_check_logs failed"); return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO)); } } } return (0); } static int spa_ld_verify_pool_data(spa_t *spa) { int error = 0; vdev_t *rvd = spa->spa_root_vdev; /* * We've successfully opened the pool, verify that we're ready * to start pushing transactions. */ if (spa->spa_load_state != SPA_LOAD_TRYIMPORT) { error = spa_load_verify(spa); if (error != 0) { spa_load_failed(spa, "spa_load_verify failed " "[error=%d]", error); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error)); } } return (0); } static void spa_ld_claim_log_blocks(spa_t *spa) { dmu_tx_t *tx; dsl_pool_t *dp = spa_get_dsl(spa); /* * Claim log blocks that haven't been committed yet. * This must all happen in a single txg. * Note: spa_claim_max_txg is updated by spa_claim_notify(), * invoked from zil_claim_log_block()'s i/o done callback. * Price of rollback is that we abandon the log. */ spa->spa_claiming = B_TRUE; tx = dmu_tx_create_assigned(dp, spa_first_txg(spa)); (void) dmu_objset_find_dp(dp, dp->dp_root_dir_obj, zil_claim, tx, DS_FIND_CHILDREN); dmu_tx_commit(tx); spa->spa_claiming = B_FALSE; spa_set_log_state(spa, SPA_LOG_GOOD); } static void spa_ld_check_for_config_update(spa_t *spa, uint64_t config_cache_txg, boolean_t update_config_cache) { vdev_t *rvd = spa->spa_root_vdev; int need_update = B_FALSE; /* * If the config cache is stale, or we have uninitialized * metaslabs (see spa_vdev_add()), then update the config. * * If this is a verbatim import, trust the current * in-core spa_config and update the disk labels. */ if (update_config_cache || config_cache_txg != spa->spa_config_txg || spa->spa_load_state == SPA_LOAD_IMPORT || spa->spa_load_state == SPA_LOAD_RECOVER || (spa->spa_import_flags & ZFS_IMPORT_VERBATIM)) need_update = B_TRUE; for (int c = 0; c < rvd->vdev_children; c++) if (rvd->vdev_child[c]->vdev_ms_array == 0) need_update = B_TRUE; /* * Update the config cache asynchronously in case we're the * root pool, in which case the config cache isn't writable yet. */ if (need_update) spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); } static void spa_ld_prepare_for_reload(spa_t *spa) { spa_mode_t mode = spa->spa_mode; int async_suspended = spa->spa_async_suspended; spa_unload(spa); spa_deactivate(spa); spa_activate(spa, mode); /* * We save the value of spa_async_suspended as it gets reset to 0 by * spa_unload(). We want to restore it back to the original value before * returning as we might be calling spa_async_resume() later. */ spa->spa_async_suspended = async_suspended; } static int spa_ld_read_checkpoint_txg(spa_t *spa) { uberblock_t checkpoint; int error = 0; ASSERT0(spa->spa_checkpoint_txg); - ASSERT(MUTEX_HELD(&spa_namespace_lock)); + ASSERT(MUTEX_HELD(&spa_namespace_lock) || + spa->spa_load_thread == curthread); error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t), sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint); if (error == ENOENT) return (0); if (error != 0) return (error); ASSERT3U(checkpoint.ub_txg, !=, 0); ASSERT3U(checkpoint.ub_checkpoint_txg, !=, 0); ASSERT3U(checkpoint.ub_timestamp, !=, 0); spa->spa_checkpoint_txg = checkpoint.ub_txg; spa->spa_checkpoint_info.sci_timestamp = checkpoint.ub_timestamp; return (0); } static int spa_ld_mos_init(spa_t *spa, spa_import_type_t type) { int error = 0; ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE); /* * Never trust the config that is provided unless we are assembling * a pool following a split. * This means don't trust blkptrs and the vdev tree in general. This * also effectively puts the spa in read-only mode since * spa_writeable() checks for spa_trust_config to be true. * We will later load a trusted config from the MOS. */ if (type != SPA_IMPORT_ASSEMBLE) spa->spa_trust_config = B_FALSE; /* * Parse the config provided to create a vdev tree. */ error = spa_ld_parse_config(spa, type); if (error != 0) return (error); spa_import_progress_add(spa); /* * Now that we have the vdev tree, try to open each vdev. This involves * opening the underlying physical device, retrieving its geometry and * probing the vdev with a dummy I/O. The state of each vdev will be set * based on the success of those operations. After this we'll be ready * to read from the vdevs. */ error = spa_ld_open_vdevs(spa); if (error != 0) return (error); /* * Read the label of each vdev and make sure that the GUIDs stored * there match the GUIDs in the config provided. * If we're assembling a new pool that's been split off from an * existing pool, the labels haven't yet been updated so we skip * validation for now. */ if (type != SPA_IMPORT_ASSEMBLE) { error = spa_ld_validate_vdevs(spa); if (error != 0) return (error); } /* * Read all vdev labels to find the best uberblock (i.e. latest, * unless spa_load_max_txg is set) and store it in spa_uberblock. We * get the list of features required to read blkptrs in the MOS from * the vdev label with the best uberblock and verify that our version * of zfs supports them all. */ error = spa_ld_select_uberblock(spa, type); if (error != 0) return (error); /* * Pass that uberblock to the dsl_pool layer which will open the root * blkptr. This blkptr points to the latest version of the MOS and will * allow us to read its contents. */ error = spa_ld_open_rootbp(spa); if (error != 0) return (error); return (0); } static int spa_ld_checkpoint_rewind(spa_t *spa) { uberblock_t checkpoint; int error = 0; ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT); error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t), sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint); if (error != 0) { spa_load_failed(spa, "unable to retrieve checkpointed " "uberblock from the MOS config [error=%d]", error); if (error == ENOENT) error = ZFS_ERR_NO_CHECKPOINT; return (error); } ASSERT3U(checkpoint.ub_txg, <, spa->spa_uberblock.ub_txg); ASSERT3U(checkpoint.ub_txg, ==, checkpoint.ub_checkpoint_txg); /* * We need to update the txg and timestamp of the checkpointed * uberblock to be higher than the latest one. This ensures that * the checkpointed uberblock is selected if we were to close and * reopen the pool right after we've written it in the vdev labels. * (also see block comment in vdev_uberblock_compare) */ checkpoint.ub_txg = spa->spa_uberblock.ub_txg + 1; checkpoint.ub_timestamp = gethrestime_sec(); /* * Set current uberblock to be the checkpointed uberblock. */ spa->spa_uberblock = checkpoint; /* * If we are doing a normal rewind, then the pool is open for * writing and we sync the "updated" checkpointed uberblock to * disk. Once this is done, we've basically rewound the whole * pool and there is no way back. * * There are cases when we don't want to attempt and sync the * checkpointed uberblock to disk because we are opening a * pool as read-only. Specifically, verifying the checkpointed * state with zdb, and importing the checkpointed state to get * a "preview" of its content. */ if (spa_writeable(spa)) { vdev_t *rvd = spa->spa_root_vdev; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL }; int svdcount = 0; int children = rvd->vdev_children; int c0 = random_in_range(children); for (int c = 0; c < children; c++) { vdev_t *vd = rvd->vdev_child[(c0 + c) % children]; /* Stop when revisiting the first vdev */ if (c > 0 && svd[0] == vd) break; if (vd->vdev_ms_array == 0 || vd->vdev_islog || !vdev_is_concrete(vd)) continue; svd[svdcount++] = vd; if (svdcount == SPA_SYNC_MIN_VDEVS) break; } error = vdev_config_sync(svd, svdcount, spa->spa_first_txg); if (error == 0) spa->spa_last_synced_guid = rvd->vdev_guid; spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) { spa_load_failed(spa, "failed to write checkpointed " "uberblock to the vdev labels [error=%d]", error); return (error); } } return (0); } static int spa_ld_mos_with_trusted_config(spa_t *spa, spa_import_type_t type, boolean_t *update_config_cache) { int error; /* * Parse the config for pool, open and validate vdevs, * select an uberblock, and use that uberblock to open * the MOS. */ error = spa_ld_mos_init(spa, type); if (error != 0) return (error); /* * Retrieve the trusted config stored in the MOS and use it to create * a new, exact version of the vdev tree, then reopen all vdevs. */ error = spa_ld_trusted_config(spa, type, B_FALSE); if (error == EAGAIN) { if (update_config_cache != NULL) *update_config_cache = B_TRUE; /* * Redo the loading process with the trusted config if it is * too different from the untrusted config. */ spa_ld_prepare_for_reload(spa); spa_load_note(spa, "RELOADING"); error = spa_ld_mos_init(spa, type); if (error != 0) return (error); error = spa_ld_trusted_config(spa, type, B_TRUE); if (error != 0) return (error); } else if (error != 0) { return (error); } return (0); } /* * Load an existing storage pool, using the config provided. This config * describes which vdevs are part of the pool and is later validated against * partial configs present in each vdev's label and an entire copy of the * config stored in the MOS. */ static int spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport) { int error = 0; boolean_t missing_feat_write = B_FALSE; boolean_t checkpoint_rewind = (spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT); boolean_t update_config_cache = B_FALSE; + hrtime_t load_start = gethrtime(); ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE); spa_load_note(spa, "LOADING"); error = spa_ld_mos_with_trusted_config(spa, type, &update_config_cache); if (error != 0) return (error); /* * If we are rewinding to the checkpoint then we need to repeat * everything we've done so far in this function but this time * selecting the checkpointed uberblock and using that to open * the MOS. */ if (checkpoint_rewind) { /* * If we are rewinding to the checkpoint update config cache * anyway. */ update_config_cache = B_TRUE; /* * Extract the checkpointed uberblock from the current MOS * and use this as the pool's uberblock from now on. If the * pool is imported as writeable we also write the checkpoint * uberblock to the labels, making the rewind permanent. */ error = spa_ld_checkpoint_rewind(spa); if (error != 0) return (error); /* * Redo the loading process again with the * checkpointed uberblock. */ spa_ld_prepare_for_reload(spa); spa_load_note(spa, "LOADING checkpointed uberblock"); error = spa_ld_mos_with_trusted_config(spa, type, NULL); if (error != 0) return (error); } + /* + * Drop the namespace lock for the rest of the function. + */ + spa->spa_load_thread = curthread; + mutex_exit(&spa_namespace_lock); + /* * Retrieve the checkpoint txg if the pool has a checkpoint. */ spa_import_progress_set_notes(spa, "Loading checkpoint txg"); error = spa_ld_read_checkpoint_txg(spa); if (error != 0) - return (error); + goto fail; /* * Retrieve the mapping of indirect vdevs. Those vdevs were removed * from the pool and their contents were re-mapped to other vdevs. Note * that everything that we read before this step must have been * rewritten on concrete vdevs after the last device removal was * initiated. Otherwise we could be reading from indirect vdevs before * we have loaded their mappings. */ spa_import_progress_set_notes(spa, "Loading indirect vdev metadata"); error = spa_ld_open_indirect_vdev_metadata(spa); if (error != 0) - return (error); + goto fail; /* * Retrieve the full list of active features from the MOS and check if * they are all supported. */ spa_import_progress_set_notes(spa, "Checking feature flags"); error = spa_ld_check_features(spa, &missing_feat_write); if (error != 0) - return (error); + goto fail; /* * Load several special directories from the MOS needed by the dsl_pool * layer. */ spa_import_progress_set_notes(spa, "Loading special MOS directories"); error = spa_ld_load_special_directories(spa); if (error != 0) - return (error); + goto fail; /* * Retrieve pool properties from the MOS. */ spa_import_progress_set_notes(spa, "Loading properties"); error = spa_ld_get_props(spa); if (error != 0) - return (error); + goto fail; /* * Retrieve the list of auxiliary devices - cache devices and spares - * and open them. */ spa_import_progress_set_notes(spa, "Loading AUX vdevs"); error = spa_ld_open_aux_vdevs(spa, type); if (error != 0) - return (error); + goto fail; /* * Load the metadata for all vdevs. Also check if unopenable devices * should be autoreplaced. */ spa_import_progress_set_notes(spa, "Loading vdev metadata"); error = spa_ld_load_vdev_metadata(spa); if (error != 0) - return (error); + goto fail; spa_import_progress_set_notes(spa, "Loading dedup tables"); error = spa_ld_load_dedup_tables(spa); if (error != 0) - return (error); + goto fail; spa_import_progress_set_notes(spa, "Loading BRT"); error = spa_ld_load_brt(spa); if (error != 0) - return (error); + goto fail; /* * Verify the logs now to make sure we don't have any unexpected errors * when we claim log blocks later. */ spa_import_progress_set_notes(spa, "Verifying Log Devices"); error = spa_ld_verify_logs(spa, type, ereport); if (error != 0) - return (error); + goto fail; if (missing_feat_write) { ASSERT(spa->spa_load_state == SPA_LOAD_TRYIMPORT); /* * At this point, we know that we can open the pool in * read-only mode but not read-write mode. We now have enough * information and can return to userland. */ - return (spa_vdev_err(spa->spa_root_vdev, VDEV_AUX_UNSUP_FEAT, - ENOTSUP)); + error = spa_vdev_err(spa->spa_root_vdev, VDEV_AUX_UNSUP_FEAT, + ENOTSUP); + goto fail; } /* * Traverse the last txgs to make sure the pool was left off in a safe * state. When performing an extreme rewind, we verify the whole pool, * which can take a very long time. */ spa_import_progress_set_notes(spa, "Verifying pool data"); error = spa_ld_verify_pool_data(spa); if (error != 0) - return (error); + goto fail; /* * Calculate the deflated space for the pool. This must be done before * we write anything to the pool because we'd need to update the space * accounting using the deflated sizes. */ spa_import_progress_set_notes(spa, "Calculating deflated space"); spa_update_dspace(spa); /* * We have now retrieved all the information we needed to open the * pool. If we are importing the pool in read-write mode, a few * additional steps must be performed to finish the import. */ spa_import_progress_set_notes(spa, "Starting import"); if (spa_writeable(spa) && (spa->spa_load_state == SPA_LOAD_RECOVER || spa->spa_load_max_txg == UINT64_MAX)) { uint64_t config_cache_txg = spa->spa_config_txg; ASSERT(spa->spa_load_state != SPA_LOAD_TRYIMPORT); /* * Before we do any zio_write's, complete the raidz expansion * scratch space copying, if necessary. */ if (RRSS_GET_STATE(&spa->spa_uberblock) == RRSS_SCRATCH_VALID) vdev_raidz_reflow_copy_scratch(spa); /* * In case of a checkpoint rewind, log the original txg * of the checkpointed uberblock. */ if (checkpoint_rewind) { spa_history_log_internal(spa, "checkpoint rewind", NULL, "rewound state to txg=%llu", (u_longlong_t)spa->spa_uberblock.ub_checkpoint_txg); } spa_import_progress_set_notes(spa, "Claiming ZIL blocks"); /* * Traverse the ZIL and claim all blocks. */ spa_ld_claim_log_blocks(spa); /* * Kick-off the syncing thread. */ spa->spa_sync_on = B_TRUE; txg_sync_start(spa->spa_dsl_pool); mmp_thread_start(spa); /* * Wait for all claims to sync. We sync up to the highest * claimed log block birth time so that claimed log blocks * don't appear to be from the future. spa_claim_max_txg * will have been set for us by ZIL traversal operations * performed above. */ spa_import_progress_set_notes(spa, "Syncing ZIL claims"); txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg); /* * Check if we need to request an update of the config. On the * next sync, we would update the config stored in vdev labels * and the cachefile (by default /etc/zfs/zpool.cache). */ spa_import_progress_set_notes(spa, "Updating configs"); spa_ld_check_for_config_update(spa, config_cache_txg, update_config_cache); /* * Check if a rebuild was in progress and if so resume it. * Then check all DTLs to see if anything needs resilvering. * The resilver will be deferred if a rebuild was started. */ spa_import_progress_set_notes(spa, "Starting resilvers"); if (vdev_rebuild_active(spa->spa_root_vdev)) { vdev_rebuild_restart(spa); } else if (!dsl_scan_resilvering(spa->spa_dsl_pool) && vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) { spa_async_request(spa, SPA_ASYNC_RESILVER); } /* * Log the fact that we booted up (so that we can detect if * we rebooted in the middle of an operation). */ spa_history_log_version(spa, "open", NULL); spa_import_progress_set_notes(spa, "Restarting device removals"); spa_restart_removal(spa); spa_spawn_aux_threads(spa); /* * Delete any inconsistent datasets. * * Note: * Since we may be issuing deletes for clones here, * we make sure to do so after we've spawned all the * auxiliary threads above (from which the livelist * deletion zthr is part of). */ spa_import_progress_set_notes(spa, "Cleaning up inconsistent objsets"); (void) dmu_objset_find(spa_name(spa), dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN); /* * Clean up any stale temporary dataset userrefs. */ spa_import_progress_set_notes(spa, "Cleaning up temporary userrefs"); dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa_import_progress_set_notes(spa, "Restarting initialize"); vdev_initialize_restart(spa->spa_root_vdev); spa_import_progress_set_notes(spa, "Restarting TRIM"); vdev_trim_restart(spa->spa_root_vdev); vdev_autotrim_restart(spa); spa_config_exit(spa, SCL_CONFIG, FTAG); spa_import_progress_set_notes(spa, "Finished importing"); } + zio_handle_import_delay(spa, gethrtime() - load_start); spa_import_progress_remove(spa_guid(spa)); spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD); spa_load_note(spa, "LOADED"); +fail: + mutex_enter(&spa_namespace_lock); + spa->spa_load_thread = NULL; + cv_broadcast(&spa_namespace_cv); + + return (error); - return (0); } static int spa_load_retry(spa_t *spa, spa_load_state_t state) { spa_mode_t mode = spa->spa_mode; spa_unload(spa); spa_deactivate(spa); spa->spa_load_max_txg = spa->spa_uberblock.ub_txg - 1; spa_activate(spa, mode); spa_async_suspend(spa); spa_load_note(spa, "spa_load_retry: rewind, max txg: %llu", (u_longlong_t)spa->spa_load_max_txg); return (spa_load(spa, state, SPA_IMPORT_EXISTING)); } /* * If spa_load() fails this function will try loading prior txg's. If * 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool * will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this * function will not rewind the pool and will return the same error as * spa_load(). */ static int spa_load_best(spa_t *spa, spa_load_state_t state, uint64_t max_request, int rewind_flags) { nvlist_t *loadinfo = NULL; nvlist_t *config = NULL; int load_error, rewind_error; uint64_t safe_rewind_txg; uint64_t min_txg; if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) { spa->spa_load_max_txg = spa->spa_load_txg; spa_set_log_state(spa, SPA_LOG_CLEAR); } else { spa->spa_load_max_txg = max_request; if (max_request != UINT64_MAX) spa->spa_extreme_rewind = B_TRUE; } load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING); if (load_error == 0) return (0); if (load_error == ZFS_ERR_NO_CHECKPOINT) { /* * When attempting checkpoint-rewind on a pool with no * checkpoint, we should not attempt to load uberblocks * from previous txgs when spa_load fails. */ ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT); spa_import_progress_remove(spa_guid(spa)); return (load_error); } if (spa->spa_root_vdev != NULL) config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg; spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp; if (rewind_flags & ZPOOL_NEVER_REWIND) { nvlist_free(config); spa_import_progress_remove(spa_guid(spa)); return (load_error); } if (state == SPA_LOAD_RECOVER) { /* Price of rolling back is discarding txgs, including log */ spa_set_log_state(spa, SPA_LOG_CLEAR); } else { /* * If we aren't rolling back save the load info from our first * import attempt so that we can restore it after attempting * to rewind. */ loadinfo = spa->spa_load_info; spa->spa_load_info = fnvlist_alloc(); } spa->spa_load_max_txg = spa->spa_last_ubsync_txg; safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE; min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ? TXG_INITIAL : safe_rewind_txg; /* * Continue as long as we're finding errors, we're still within * the acceptable rewind range, and we're still finding uberblocks */ while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg && spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) { if (spa->spa_load_max_txg < safe_rewind_txg) spa->spa_extreme_rewind = B_TRUE; rewind_error = spa_load_retry(spa, state); } spa->spa_extreme_rewind = B_FALSE; spa->spa_load_max_txg = UINT64_MAX; if (config && (rewind_error || state != SPA_LOAD_RECOVER)) spa_config_set(spa, config); else nvlist_free(config); if (state == SPA_LOAD_RECOVER) { ASSERT3P(loadinfo, ==, NULL); spa_import_progress_remove(spa_guid(spa)); return (rewind_error); } else { /* Store the rewind info as part of the initial load info */ fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO, spa->spa_load_info); /* Restore the initial load info */ fnvlist_free(spa->spa_load_info); spa->spa_load_info = loadinfo; spa_import_progress_remove(spa_guid(spa)); return (load_error); } } /* * Pool Open/Import * * The import case is identical to an open except that the configuration is sent * down from userland, instead of grabbed from the configuration cache. For the * case of an open, the pool configuration will exist in the * POOL_STATE_UNINITIALIZED state. * * The stats information (gen/count/ustats) is used to gather vdev statistics at * the same time open the pool, without having to keep around the spa_t in some * ambiguous state. */ static int spa_open_common(const char *pool, spa_t **spapp, const void *tag, nvlist_t *nvpolicy, nvlist_t **config) { spa_t *spa; spa_load_state_t state = SPA_LOAD_OPEN; int error; int locked = B_FALSE; int firstopen = B_FALSE; *spapp = NULL; /* * As disgusting as this is, we need to support recursive calls to this * function because dsl_dir_open() is called during spa_load(), and ends * up calling spa_open() again. The real fix is to figure out how to * avoid dsl_dir_open() calling this in the first place. */ if (MUTEX_NOT_HELD(&spa_namespace_lock)) { mutex_enter(&spa_namespace_lock); locked = B_TRUE; } if ((spa = spa_lookup(pool)) == NULL) { if (locked) mutex_exit(&spa_namespace_lock); return (SET_ERROR(ENOENT)); } if (spa->spa_state == POOL_STATE_UNINITIALIZED) { zpool_load_policy_t policy; firstopen = B_TRUE; zpool_get_load_policy(nvpolicy ? nvpolicy : spa->spa_config, &policy); if (policy.zlp_rewind & ZPOOL_DO_REWIND) state = SPA_LOAD_RECOVER; spa_activate(spa, spa_mode_global); if (state != SPA_LOAD_RECOVER) spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE; zfs_dbgmsg("spa_open_common: opening %s", pool); error = spa_load_best(spa, state, policy.zlp_txg, policy.zlp_rewind); if (error == EBADF) { /* * If vdev_validate() returns failure (indicated by * EBADF), it indicates that one of the vdevs indicates * that the pool has been exported or destroyed. If * this is the case, the config cache is out of sync and * we should remove the pool from the namespace. */ spa_unload(spa); spa_deactivate(spa); spa_write_cachefile(spa, B_TRUE, B_TRUE, B_FALSE); spa_remove(spa); if (locked) mutex_exit(&spa_namespace_lock); return (SET_ERROR(ENOENT)); } if (error) { /* * We can't open the pool, but we still have useful * information: the state of each vdev after the * attempted vdev_open(). Return this to the user. */ if (config != NULL && spa->spa_config) { *config = fnvlist_dup(spa->spa_config); fnvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info); } spa_unload(spa); spa_deactivate(spa); spa->spa_last_open_failed = error; if (locked) mutex_exit(&spa_namespace_lock); *spapp = NULL; return (error); } } spa_open_ref(spa, tag); if (config != NULL) *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); /* * If we've recovered the pool, pass back any information we * gathered while doing the load. */ if (state == SPA_LOAD_RECOVER && config != NULL) { fnvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info); } if (locked) { spa->spa_last_open_failed = 0; spa->spa_last_ubsync_txg = 0; spa->spa_load_txg = 0; mutex_exit(&spa_namespace_lock); } if (firstopen) zvol_create_minors_recursive(spa_name(spa)); *spapp = spa; return (0); } int spa_open_rewind(const char *name, spa_t **spapp, const void *tag, nvlist_t *policy, nvlist_t **config) { return (spa_open_common(name, spapp, tag, policy, config)); } int spa_open(const char *name, spa_t **spapp, const void *tag) { return (spa_open_common(name, spapp, tag, NULL, NULL)); } /* * Lookup the given spa_t, incrementing the inject count in the process, * preventing it from being exported or destroyed. */ spa_t * spa_inject_addref(char *name) { spa_t *spa; mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(name)) == NULL) { mutex_exit(&spa_namespace_lock); return (NULL); } spa->spa_inject_ref++; mutex_exit(&spa_namespace_lock); return (spa); } void spa_inject_delref(spa_t *spa) { mutex_enter(&spa_namespace_lock); spa->spa_inject_ref--; mutex_exit(&spa_namespace_lock); } /* * Add spares device information to the nvlist. */ static void spa_add_spares(spa_t *spa, nvlist_t *config) { nvlist_t **spares; uint_t i, nspares; nvlist_t *nvroot; uint64_t guid; vdev_stat_t *vs; uint_t vsc; uint64_t pool; ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); if (spa->spa_spares.sav_count == 0) return; nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, &spares, &nspares)); if (nspares != 0) { fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares, nspares); VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares)); /* * Go through and find any spares which have since been * repurposed as an active spare. If this is the case, update * their status appropriately. */ for (i = 0; i < nspares; i++) { guid = fnvlist_lookup_uint64(spares[i], ZPOOL_CONFIG_GUID); VERIFY0(nvlist_lookup_uint64_array(spares[i], ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)); if (spa_spare_exists(guid, &pool, NULL) && pool != 0ULL) { vs->vs_state = VDEV_STATE_CANT_OPEN; vs->vs_aux = VDEV_AUX_SPARED; } else { vs->vs_state = spa->spa_spares.sav_vdevs[i]->vdev_state; } } } } /* * Add l2cache device information to the nvlist, including vdev stats. */ static void spa_add_l2cache(spa_t *spa, nvlist_t *config) { nvlist_t **l2cache; uint_t i, j, nl2cache; nvlist_t *nvroot; uint64_t guid; vdev_t *vd; vdev_stat_t *vs; uint_t vsc; ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); if (spa->spa_l2cache.sav_count == 0) return; nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); VERIFY0(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache)); if (nl2cache != 0) { fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache, nl2cache); VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache)); /* * Update level 2 cache device stats. */ for (i = 0; i < nl2cache; i++) { guid = fnvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID); vd = NULL; for (j = 0; j < spa->spa_l2cache.sav_count; j++) { if (guid == spa->spa_l2cache.sav_vdevs[j]->vdev_guid) { vd = spa->spa_l2cache.sav_vdevs[j]; break; } } ASSERT(vd != NULL); VERIFY0(nvlist_lookup_uint64_array(l2cache[i], ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)); vdev_get_stats(vd, vs); vdev_config_generate_stats(vd, l2cache[i]); } } } static void spa_feature_stats_from_disk(spa_t *spa, nvlist_t *features) { zap_cursor_t zc; zap_attribute_t za; if (spa->spa_feat_for_read_obj != 0) { for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_feat_for_read_obj); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { ASSERT(za.za_integer_length == sizeof (uint64_t) && za.za_num_integers == 1); VERIFY0(nvlist_add_uint64(features, za.za_name, za.za_first_integer)); } zap_cursor_fini(&zc); } if (spa->spa_feat_for_write_obj != 0) { for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_feat_for_write_obj); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { ASSERT(za.za_integer_length == sizeof (uint64_t) && za.za_num_integers == 1); VERIFY0(nvlist_add_uint64(features, za.za_name, za.za_first_integer)); } zap_cursor_fini(&zc); } } static void spa_feature_stats_from_cache(spa_t *spa, nvlist_t *features) { int i; for (i = 0; i < SPA_FEATURES; i++) { zfeature_info_t feature = spa_feature_table[i]; uint64_t refcount; if (feature_get_refcount(spa, &feature, &refcount) != 0) continue; VERIFY0(nvlist_add_uint64(features, feature.fi_guid, refcount)); } } /* * Store a list of pool features and their reference counts in the * config. * * The first time this is called on a spa, allocate a new nvlist, fetch * the pool features and reference counts from disk, then save the list * in the spa. In subsequent calls on the same spa use the saved nvlist * and refresh its values from the cached reference counts. This * ensures we don't block here on I/O on a suspended pool so 'zpool * clear' can resume the pool. */ static void spa_add_feature_stats(spa_t *spa, nvlist_t *config) { nvlist_t *features; ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); mutex_enter(&spa->spa_feat_stats_lock); features = spa->spa_feat_stats; if (features != NULL) { spa_feature_stats_from_cache(spa, features); } else { VERIFY0(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP)); spa->spa_feat_stats = features; spa_feature_stats_from_disk(spa, features); } VERIFY0(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS, features)); mutex_exit(&spa->spa_feat_stats_lock); } int spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen) { int error; spa_t *spa; *config = NULL; error = spa_open_common(name, &spa, FTAG, NULL, config); if (spa != NULL) { /* * This still leaves a window of inconsistency where the spares * or l2cache devices could change and the config would be * self-inconsistent. */ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); if (*config != NULL) { uint64_t loadtimes[2]; loadtimes[0] = spa->spa_loaded_ts.tv_sec; loadtimes[1] = spa->spa_loaded_ts.tv_nsec; fnvlist_add_uint64_array(*config, ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2); fnvlist_add_uint64(*config, ZPOOL_CONFIG_ERRCOUNT, spa_approx_errlog_size(spa)); if (spa_suspended(spa)) { fnvlist_add_uint64(*config, ZPOOL_CONFIG_SUSPENDED, spa->spa_failmode); fnvlist_add_uint64(*config, ZPOOL_CONFIG_SUSPENDED_REASON, spa->spa_suspended); } spa_add_spares(spa, *config); spa_add_l2cache(spa, *config); spa_add_feature_stats(spa, *config); } } /* * We want to get the alternate root even for faulted pools, so we cheat * and call spa_lookup() directly. */ if (altroot) { if (spa == NULL) { mutex_enter(&spa_namespace_lock); spa = spa_lookup(name); if (spa) spa_altroot(spa, altroot, buflen); else altroot[0] = '\0'; spa = NULL; mutex_exit(&spa_namespace_lock); } else { spa_altroot(spa, altroot, buflen); } } if (spa != NULL) { spa_config_exit(spa, SCL_CONFIG, FTAG); spa_close(spa, FTAG); } return (error); } /* * Validate that the auxiliary device array is well formed. We must have an * array of nvlists, each which describes a valid leaf vdev. If this is an * import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be * specified, as long as they are well-formed. */ static int spa_validate_aux_devs(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode, spa_aux_vdev_t *sav, const char *config, uint64_t version, vdev_labeltype_t label) { nvlist_t **dev; uint_t i, ndev; vdev_t *vd; int error; ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); /* * It's acceptable to have no devs specified. */ if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0) return (0); if (ndev == 0) return (SET_ERROR(EINVAL)); /* * Make sure the pool is formatted with a version that supports this * device type. */ if (spa_version(spa) < version) return (SET_ERROR(ENOTSUP)); /* * Set the pending device list so we correctly handle device in-use * checking. */ sav->sav_pending = dev; sav->sav_npending = ndev; for (i = 0; i < ndev; i++) { if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0, mode)) != 0) goto out; if (!vd->vdev_ops->vdev_op_leaf) { vdev_free(vd); error = SET_ERROR(EINVAL); goto out; } vd->vdev_top = vd; if ((error = vdev_open(vd)) == 0 && (error = vdev_label_init(vd, crtxg, label)) == 0) { fnvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID, vd->vdev_guid); } vdev_free(vd); if (error && (mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE)) goto out; else error = 0; } out: sav->sav_pending = NULL; sav->sav_npending = 0; return (error); } static int spa_validate_aux(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode) { int error; ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode, &spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES, VDEV_LABEL_SPARE)) != 0) { return (error); } return (spa_validate_aux_devs(spa, nvroot, crtxg, mode, &spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE, VDEV_LABEL_L2CACHE)); } static void spa_set_aux_vdevs(spa_aux_vdev_t *sav, nvlist_t **devs, int ndevs, const char *config) { int i; if (sav->sav_config != NULL) { nvlist_t **olddevs; uint_t oldndevs; nvlist_t **newdevs; /* * Generate new dev list by concatenating with the * current dev list. */ VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config, config, &olddevs, &oldndevs)); newdevs = kmem_alloc(sizeof (void *) * (ndevs + oldndevs), KM_SLEEP); for (i = 0; i < oldndevs; i++) newdevs[i] = fnvlist_dup(olddevs[i]); for (i = 0; i < ndevs; i++) newdevs[i + oldndevs] = fnvlist_dup(devs[i]); fnvlist_remove(sav->sav_config, config); fnvlist_add_nvlist_array(sav->sav_config, config, (const nvlist_t * const *)newdevs, ndevs + oldndevs); for (i = 0; i < oldndevs + ndevs; i++) nvlist_free(newdevs[i]); kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *)); } else { /* * Generate a new dev list. */ sav->sav_config = fnvlist_alloc(); fnvlist_add_nvlist_array(sav->sav_config, config, (const nvlist_t * const *)devs, ndevs); } } /* * Stop and drop level 2 ARC devices */ void spa_l2cache_drop(spa_t *spa) { vdev_t *vd; int i; spa_aux_vdev_t *sav = &spa->spa_l2cache; for (i = 0; i < sav->sav_count; i++) { uint64_t pool; vd = sav->sav_vdevs[i]; ASSERT(vd != NULL); if (spa_l2cache_exists(vd->vdev_guid, &pool) && pool != 0ULL && l2arc_vdev_present(vd)) l2arc_remove_vdev(vd); } } /* * Verify encryption parameters for spa creation. If we are encrypting, we must * have the encryption feature flag enabled. */ static int spa_create_check_encryption_params(dsl_crypto_params_t *dcp, boolean_t has_encryption) { if (dcp->cp_crypt != ZIO_CRYPT_OFF && dcp->cp_crypt != ZIO_CRYPT_INHERIT && !has_encryption) return (SET_ERROR(ENOTSUP)); return (dmu_objset_create_crypt_check(NULL, dcp, NULL)); } /* * Pool Creation */ int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, nvlist_t *zplprops, dsl_crypto_params_t *dcp) { spa_t *spa; const char *altroot = NULL; vdev_t *rvd; dsl_pool_t *dp; dmu_tx_t *tx; int error = 0; uint64_t txg = TXG_INITIAL; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; uint64_t version, obj, ndraid = 0; boolean_t has_features; boolean_t has_encryption; boolean_t has_allocclass; spa_feature_t feat; const char *feat_name; const char *poolname; nvlist_t *nvl; if (props == NULL || nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_TNAME), &poolname) != 0) poolname = (char *)pool; /* * If this pool already exists, return failure. */ mutex_enter(&spa_namespace_lock); if (spa_lookup(poolname) != NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(EEXIST)); } /* * Allocate a new spa_t structure. */ nvl = fnvlist_alloc(); fnvlist_add_string(nvl, ZPOOL_CONFIG_POOL_NAME, pool); (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); spa = spa_add(poolname, nvl, altroot); fnvlist_free(nvl); spa_activate(spa, spa_mode_global); if (props && (error = spa_prop_validate(spa, props))) { spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } /* * Temporary pool names should never be written to disk. */ if (poolname != pool) spa->spa_import_flags |= ZFS_IMPORT_TEMP_NAME; has_features = B_FALSE; has_encryption = B_FALSE; has_allocclass = B_FALSE; for (nvpair_t *elem = nvlist_next_nvpair(props, NULL); elem != NULL; elem = nvlist_next_nvpair(props, elem)) { if (zpool_prop_feature(nvpair_name(elem))) { has_features = B_TRUE; feat_name = strchr(nvpair_name(elem), '@') + 1; VERIFY0(zfeature_lookup_name(feat_name, &feat)); if (feat == SPA_FEATURE_ENCRYPTION) has_encryption = B_TRUE; if (feat == SPA_FEATURE_ALLOCATION_CLASSES) has_allocclass = B_TRUE; } } /* verify encryption params, if they were provided */ if (dcp != NULL) { error = spa_create_check_encryption_params(dcp, has_encryption); if (error != 0) { spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } } if (!has_allocclass && zfs_special_devs(nvroot, NULL)) { spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (ENOTSUP); } if (has_features || nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) { version = SPA_VERSION; } ASSERT(SPA_VERSION_IS_SUPPORTED(version)); spa->spa_first_txg = txg; spa->spa_uberblock.ub_txg = txg - 1; spa->spa_uberblock.ub_version = version; spa->spa_ubsync = spa->spa_uberblock; spa->spa_load_state = SPA_LOAD_CREATE; spa->spa_removing_phys.sr_state = DSS_NONE; spa->spa_removing_phys.sr_removing_vdev = -1; spa->spa_removing_phys.sr_prev_indirect_vdev = -1; spa->spa_indirect_vdevs_loaded = B_TRUE; /* * Create "The Godfather" zio to hold all async IOs */ spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *), KM_SLEEP); for (int i = 0; i < max_ncpus; i++) { spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); } /* * Create the root vdev. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD); ASSERT(error != 0 || rvd != NULL); ASSERT(error != 0 || spa->spa_root_vdev == rvd); if (error == 0 && !zfs_allocatable_devs(nvroot)) error = SET_ERROR(EINVAL); if (error == 0 && (error = vdev_create(rvd, txg, B_FALSE)) == 0 && (error = vdev_draid_spare_create(nvroot, rvd, &ndraid, 0)) == 0 && (error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) == 0) { /* * instantiate the metaslab groups (this will dirty the vdevs) * we can no longer error exit past this point */ for (int c = 0; error == 0 && c < rvd->vdev_children; c++) { vdev_t *vd = rvd->vdev_child[c]; vdev_metaslab_set_size(vd); vdev_expand(vd, txg); } } spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) { spa_unload(spa); spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } /* * Get the list of spares, if specified. */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { spa->spa_spares.sav_config = fnvlist_alloc(); fnvlist_add_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares, nspares); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_spares.sav_sync = B_TRUE; } /* * Get the list of level 2 cache devices, if specified. */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0) { VERIFY0(nvlist_alloc(&spa->spa_l2cache.sav_config, NV_UNIQUE_NAME, KM_SLEEP)); fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache, nl2cache); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_l2cache.sav_sync = B_TRUE; } spa->spa_is_initializing = B_TRUE; spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, dcp, txg); spa->spa_is_initializing = B_FALSE; /* * Create DDTs (dedup tables). */ ddt_create(spa); /* * Create BRT table and BRT table object. */ brt_create(spa); spa_update_dspace(spa); tx = dmu_tx_create_assigned(dp, txg); /* * Create the pool's history object. */ if (version >= SPA_VERSION_ZPOOL_HISTORY && !spa->spa_history) spa_history_create_obj(spa, tx); spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_CREATE); spa_history_log_version(spa, "create", tx); /* * Create the pool config object. */ spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset, DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE, DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx); if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG, sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) { cmn_err(CE_PANIC, "failed to add pool config"); } if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION, sizeof (uint64_t), 1, &version, tx) != 0) { cmn_err(CE_PANIC, "failed to add pool version"); } /* Newly created pools with the right version are always deflated. */ if (version >= SPA_VERSION_RAIDZ_DEFLATE) { spa->spa_deflate = TRUE; if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) { cmn_err(CE_PANIC, "failed to add deflate"); } } /* * Create the deferred-free bpobj. Turn off compression * because sync-to-convergence takes longer if the blocksize * keeps changing. */ obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx); dmu_object_set_compress(spa->spa_meta_objset, obj, ZIO_COMPRESS_OFF, tx); if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ, sizeof (uint64_t), 1, &obj, tx) != 0) { cmn_err(CE_PANIC, "failed to add bpobj"); } VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj)); /* * Generate some random noise for salted checksums to operate on. */ (void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes, sizeof (spa->spa_cksum_salt.zcs_bytes)); /* * Set pool properties. */ spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS); spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE); spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND); spa->spa_multihost = zpool_prop_default_numeric(ZPOOL_PROP_MULTIHOST); spa->spa_autotrim = zpool_prop_default_numeric(ZPOOL_PROP_AUTOTRIM); if (props != NULL) { spa_configfile_set(spa, props, B_FALSE); spa_sync_props(props, tx); } for (int i = 0; i < ndraid; i++) spa_feature_incr(spa, SPA_FEATURE_DRAID, tx); dmu_tx_commit(tx); spa->spa_sync_on = B_TRUE; txg_sync_start(dp); mmp_thread_start(spa); txg_wait_synced(dp, txg); spa_spawn_aux_threads(spa); spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE); /* * Don't count references from objsets that are already closed * and are making their way through the eviction process. */ spa_evicting_os_wait(spa); spa->spa_minref = zfs_refcount_count(&spa->spa_refcount); spa->spa_load_state = SPA_LOAD_NONE; spa_import_os(spa); mutex_exit(&spa_namespace_lock); return (0); } /* * Import a non-root pool into the system. */ int spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags) { spa_t *spa; const char *altroot = NULL; spa_load_state_t state = SPA_LOAD_IMPORT; zpool_load_policy_t policy; spa_mode_t mode = spa_mode_global; uint64_t readonly = B_FALSE; int error; nvlist_t *nvroot; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; /* * If a pool with this name exists, return failure. */ mutex_enter(&spa_namespace_lock); if (spa_lookup(pool) != NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(EEXIST)); } /* * Create and initialize the spa structure. */ (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); (void) nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly); if (readonly) mode = SPA_MODE_READ; spa = spa_add(pool, config, altroot); spa->spa_import_flags = flags; /* * Verbatim import - Take a pool and insert it into the namespace * as if it had been loaded at boot. */ if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) { if (props != NULL) spa_configfile_set(spa, props, B_FALSE); spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE); spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT); zfs_dbgmsg("spa_import: verbatim import of %s", pool); mutex_exit(&spa_namespace_lock); return (0); } spa_activate(spa, mode); /* * Don't start async tasks until we know everything is healthy. */ spa_async_suspend(spa); zpool_get_load_policy(config, &policy); if (policy.zlp_rewind & ZPOOL_DO_REWIND) state = SPA_LOAD_RECOVER; spa->spa_config_source = SPA_CONFIG_SRC_TRYIMPORT; if (state != SPA_LOAD_RECOVER) { spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; zfs_dbgmsg("spa_import: importing %s", pool); } else { zfs_dbgmsg("spa_import: importing %s, max_txg=%lld " "(RECOVERY MODE)", pool, (longlong_t)policy.zlp_txg); } error = spa_load_best(spa, state, policy.zlp_txg, policy.zlp_rewind); /* * Propagate anything learned while loading the pool and pass it * back to caller (i.e. rewind info, missing devices, etc). */ fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * Toss any existing sparelist, as it doesn't have any validity * anymore, and conflicts with spa_has_spare(). */ if (spa->spa_spares.sav_config) { nvlist_free(spa->spa_spares.sav_config); spa->spa_spares.sav_config = NULL; spa_load_spares(spa); } if (spa->spa_l2cache.sav_config) { nvlist_free(spa->spa_l2cache.sav_config); spa->spa_l2cache.sav_config = NULL; spa_load_l2cache(spa); } nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE); spa_config_exit(spa, SCL_ALL, FTAG); if (props != NULL) spa_configfile_set(spa, props, B_FALSE); if (error != 0 || (props && spa_writeable(spa) && (error = spa_prop_set(spa, props)))) { spa_unload(spa); spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } spa_async_resume(spa); /* * Override any spares and level 2 cache devices as specified by * the user, as these may have correct device names/devids, etc. */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { if (spa->spa_spares.sav_config) fnvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES); else spa->spa_spares.sav_config = fnvlist_alloc(); fnvlist_add_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares, nspares); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_spares.sav_sync = B_TRUE; } if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0) { if (spa->spa_l2cache.sav_config) fnvlist_remove(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE); else spa->spa_l2cache.sav_config = fnvlist_alloc(); fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache, nl2cache); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_l2cache.sav_sync = B_TRUE; } /* * Check for any removed devices. */ if (spa->spa_autoreplace) { spa_aux_check_removed(&spa->spa_spares); spa_aux_check_removed(&spa->spa_l2cache); } if (spa_writeable(spa)) { /* * Update the config cache to include the newly-imported pool. */ spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); } /* * It's possible that the pool was expanded while it was exported. * We kick off an async task to handle this for us. */ spa_async_request(spa, SPA_ASYNC_AUTOEXPAND); spa_history_log_version(spa, "import", NULL); spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT); mutex_exit(&spa_namespace_lock); zvol_create_minors_recursive(pool); spa_import_os(spa); return (0); } nvlist_t * spa_tryimport(nvlist_t *tryconfig) { nvlist_t *config = NULL; const char *poolname, *cachefile; spa_t *spa; uint64_t state; int error; zpool_load_policy_t policy; if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname)) return (NULL); if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state)) return (NULL); /* * Create and initialize the spa structure. */ + char *name = kmem_alloc(MAXPATHLEN, KM_SLEEP); + (void) snprintf(name, MAXPATHLEN, "%s-%llx-%s", + TRYIMPORT_NAME, (u_longlong_t)curthread, poolname); + mutex_enter(&spa_namespace_lock); - spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL); + spa = spa_add(name, tryconfig, NULL); spa_activate(spa, SPA_MODE_READ); + kmem_free(name, MAXPATHLEN); /* * Rewind pool if a max txg was provided. */ zpool_get_load_policy(spa->spa_config, &policy); if (policy.zlp_txg != UINT64_MAX) { spa->spa_load_max_txg = policy.zlp_txg; spa->spa_extreme_rewind = B_TRUE; zfs_dbgmsg("spa_tryimport: importing %s, max_txg=%lld", poolname, (longlong_t)policy.zlp_txg); } else { zfs_dbgmsg("spa_tryimport: importing %s", poolname); } if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_CACHEFILE, &cachefile) == 0) { zfs_dbgmsg("spa_tryimport: using cachefile '%s'", cachefile); spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE; } else { spa->spa_config_source = SPA_CONFIG_SRC_SCAN; } /* * spa_import() relies on a pool config fetched by spa_try_import() * for spare/cache devices. Import flags are not passed to * spa_tryimport(), which makes it return early due to a missing log * device and missing retrieving the cache device and spare eventually. * Passing ZFS_IMPORT_MISSING_LOG to spa_tryimport() makes it fetch * the correct configuration regardless of the missing log device. */ spa->spa_import_flags |= ZFS_IMPORT_MISSING_LOG; error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING); /* * If 'tryconfig' was at least parsable, return the current config. */ if (spa->spa_root_vdev != NULL) { config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, poolname); fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, state); fnvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP, spa->spa_uberblock.ub_timestamp); fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info); fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA, spa->spa_errata); /* * If the bootfs property exists on this pool then we * copy it out so that external consumers can tell which * pools are bootable. */ if ((!error || error == EEXIST) && spa->spa_bootfs) { char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP); /* * We have to play games with the name since the * pool was opened as TRYIMPORT_NAME. */ if (dsl_dsobj_to_dsname(spa_name(spa), spa->spa_bootfs, tmpname) == 0) { char *cp; char *dsname; dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP); cp = strchr(tmpname, '/'); if (cp == NULL) { (void) strlcpy(dsname, tmpname, MAXPATHLEN); } else { (void) snprintf(dsname, MAXPATHLEN, "%s/%s", poolname, ++cp); } fnvlist_add_string(config, ZPOOL_CONFIG_BOOTFS, dsname); kmem_free(dsname, MAXPATHLEN); } kmem_free(tmpname, MAXPATHLEN); } /* * Add the list of hot spares and level 2 cache devices. */ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa_add_spares(spa, config); spa_add_l2cache(spa, config); spa_config_exit(spa, SCL_CONFIG, FTAG); } spa_unload(spa); spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (config); } /* * Pool export/destroy * * The act of destroying or exporting a pool is very simple. We make sure there * is no more pending I/O and any references to the pool are gone. Then, we * update the pool state and sync all the labels to disk, removing the * configuration from the cache afterwards. If the 'hardforce' flag is set, then * we don't sync the labels or remove the configuration cache. */ static int spa_export_common(const char *pool, int new_state, nvlist_t **oldconfig, boolean_t force, boolean_t hardforce) { int error; spa_t *spa; + hrtime_t export_start = gethrtime(); if (oldconfig) *oldconfig = NULL; if (!(spa_mode_global & SPA_MODE_WRITE)) return (SET_ERROR(EROFS)); mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(pool)) == NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(ENOENT)); } if (spa->spa_is_exporting) { /* the pool is being exported by another thread */ mutex_exit(&spa_namespace_lock); return (SET_ERROR(ZFS_ERR_EXPORT_IN_PROGRESS)); } spa->spa_is_exporting = B_TRUE; /* * Put a hold on the pool, drop the namespace lock, stop async tasks, * reacquire the namespace lock, and see if we can export. */ spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); spa_async_suspend(spa); if (spa->spa_zvol_taskq) { zvol_remove_minors(spa, spa_name(spa), B_TRUE); taskq_wait(spa->spa_zvol_taskq); } mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); if (spa->spa_state == POOL_STATE_UNINITIALIZED) goto export_spa; /* * The pool will be in core if it's openable, in which case we can * modify its state. Objsets may be open only because they're dirty, * so we have to force it to sync before checking spa_refcnt. */ if (spa->spa_sync_on) { txg_wait_synced(spa->spa_dsl_pool, 0); spa_evicting_os_wait(spa); } /* * A pool cannot be exported or destroyed if there are active * references. If we are resetting a pool, allow references by * fault injection handlers. */ if (!spa_refcount_zero(spa) || (spa->spa_inject_ref != 0)) { error = SET_ERROR(EBUSY); goto fail; } if (spa->spa_sync_on) { vdev_t *rvd = spa->spa_root_vdev; /* * A pool cannot be exported if it has an active shared spare. * This is to prevent other pools stealing the active spare * from an exported pool. At user's own will, such pool can * be forcedly exported. */ if (!force && new_state == POOL_STATE_EXPORTED && spa_has_active_shared_spare(spa)) { error = SET_ERROR(EXDEV); goto fail; } /* * We're about to export or destroy this pool. Make sure * we stop all initialization and trim activity here before * we set the spa_final_txg. This will ensure that all * dirty data resulting from the initialization is * committed to disk before we unload the pool. */ vdev_initialize_stop_all(rvd, VDEV_INITIALIZE_ACTIVE); vdev_trim_stop_all(rvd, VDEV_TRIM_ACTIVE); vdev_autotrim_stop_all(spa); vdev_rebuild_stop_all(spa); /* * We want this to be reflected on every label, * so mark them all dirty. spa_unload() will do the * final sync that pushes these changes out. */ if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa->spa_state = new_state; vdev_config_dirty(rvd); spa_config_exit(spa, SCL_ALL, FTAG); } /* * If the log space map feature is enabled and the pool is * getting exported (but not destroyed), we want to spend some * time flushing as many metaslabs as we can in an attempt to * destroy log space maps and save import time. This has to be * done before we set the spa_final_txg, otherwise * spa_sync() -> spa_flush_metaslabs() may dirty the final TXGs. * spa_should_flush_logs_on_unload() should be called after * spa_state has been set to the new_state. */ if (spa_should_flush_logs_on_unload(spa)) spa_unload_log_sm_flush_all(spa); if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa->spa_final_txg = spa_last_synced_txg(spa) + TXG_DEFER_SIZE + 1; spa_config_exit(spa, SCL_ALL, FTAG); } } export_spa: spa_export_os(spa); if (new_state == POOL_STATE_DESTROYED) spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_DESTROY); else if (new_state == POOL_STATE_EXPORTED) spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_EXPORT); if (spa->spa_state != POOL_STATE_UNINITIALIZED) { spa_unload(spa); spa_deactivate(spa); } if (oldconfig && spa->spa_config) *oldconfig = fnvlist_dup(spa->spa_config); if (new_state != POOL_STATE_UNINITIALIZED) { if (!hardforce) spa_write_cachefile(spa, B_TRUE, B_TRUE, B_FALSE); spa_remove(spa); } else { /* * If spa_remove() is not called for this spa_t and * there is any possibility that it can be reused, * we make sure to reset the exporting flag. */ spa->spa_is_exporting = B_FALSE; } + if (new_state == POOL_STATE_EXPORTED) + zio_handle_export_delay(spa, gethrtime() - export_start); + mutex_exit(&spa_namespace_lock); return (0); fail: spa->spa_is_exporting = B_FALSE; spa_async_resume(spa); mutex_exit(&spa_namespace_lock); return (error); } /* * Destroy a storage pool. */ int spa_destroy(const char *pool) { return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL, B_FALSE, B_FALSE)); } /* * Export a storage pool. */ int spa_export(const char *pool, nvlist_t **oldconfig, boolean_t force, boolean_t hardforce) { return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig, force, hardforce)); } /* * Similar to spa_export(), this unloads the spa_t without actually removing it * from the namespace in any way. */ int spa_reset(const char *pool) { return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL, B_FALSE, B_FALSE)); } /* * ========================================================================== * Device manipulation * ========================================================================== */ /* * This is called as a synctask to increment the draid feature flag */ static void spa_draid_feature_incr(void *arg, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; int draid = (int)(uintptr_t)arg; for (int c = 0; c < draid; c++) spa_feature_incr(spa, SPA_FEATURE_DRAID, tx); } /* * Add a device to a storage pool. */ int spa_vdev_add(spa_t *spa, nvlist_t *nvroot, boolean_t check_ashift) { uint64_t txg, ndraid = 0; int error; vdev_t *rvd = spa->spa_root_vdev; vdev_t *vd, *tvd; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0, VDEV_ALLOC_ADD)) != 0) return (spa_vdev_exit(spa, NULL, txg, error)); spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) != 0) nspares = 0; if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) != 0) nl2cache = 0; if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0) return (spa_vdev_exit(spa, vd, txg, EINVAL)); if (vd->vdev_children != 0 && (error = vdev_create(vd, txg, B_FALSE)) != 0) { return (spa_vdev_exit(spa, vd, txg, error)); } /* * The virtual dRAID spares must be added after vdev tree is created * and the vdev guids are generated. The guid of their associated * dRAID is stored in the config and used when opening the spare. */ if ((error = vdev_draid_spare_create(nvroot, vd, &ndraid, rvd->vdev_children)) == 0) { if (ndraid > 0 && nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) != 0) nspares = 0; } else { return (spa_vdev_exit(spa, vd, txg, error)); } /* * We must validate the spares and l2cache devices after checking the * children. Otherwise, vdev_inuse() will blindly overwrite the spare. */ if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0) return (spa_vdev_exit(spa, vd, txg, error)); /* * If we are in the middle of a device removal, we can only add * devices which match the existing devices in the pool. * If we are in the middle of a removal, or have some indirect * vdevs, we can not add raidz or dRAID top levels. */ if (spa->spa_vdev_removal != NULL || spa->spa_removing_phys.sr_prev_indirect_vdev != -1) { for (int c = 0; c < vd->vdev_children; c++) { tvd = vd->vdev_child[c]; if (spa->spa_vdev_removal != NULL && tvd->vdev_ashift != spa->spa_max_ashift) { return (spa_vdev_exit(spa, vd, txg, EINVAL)); } /* Fail if top level vdev is raidz or a dRAID */ if (vdev_get_nparity(tvd) != 0) return (spa_vdev_exit(spa, vd, txg, EINVAL)); /* * Need the top level mirror to be * a mirror of leaf vdevs only */ if (tvd->vdev_ops == &vdev_mirror_ops) { for (uint64_t cid = 0; cid < tvd->vdev_children; cid++) { vdev_t *cvd = tvd->vdev_child[cid]; if (!cvd->vdev_ops->vdev_op_leaf) { return (spa_vdev_exit(spa, vd, txg, EINVAL)); } } } } } if (check_ashift && spa->spa_max_ashift == spa->spa_min_ashift) { for (int c = 0; c < vd->vdev_children; c++) { tvd = vd->vdev_child[c]; if (tvd->vdev_ashift != spa->spa_max_ashift) { return (spa_vdev_exit(spa, vd, txg, ZFS_ERR_ASHIFT_MISMATCH)); } } } for (int c = 0; c < vd->vdev_children; c++) { tvd = vd->vdev_child[c]; vdev_remove_child(vd, tvd); tvd->vdev_id = rvd->vdev_children; vdev_add_child(rvd, tvd); vdev_config_dirty(tvd); } if (nspares != 0) { spa_set_aux_vdevs(&spa->spa_spares, spares, nspares, ZPOOL_CONFIG_SPARES); spa_load_spares(spa); spa->spa_spares.sav_sync = B_TRUE; } if (nl2cache != 0) { spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache, ZPOOL_CONFIG_L2CACHE); spa_load_l2cache(spa); spa->spa_l2cache.sav_sync = B_TRUE; } /* * We can't increment a feature while holding spa_vdev so we * have to do it in a synctask. */ if (ndraid != 0) { dmu_tx_t *tx; tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg); dsl_sync_task_nowait(spa->spa_dsl_pool, spa_draid_feature_incr, (void *)(uintptr_t)ndraid, tx); dmu_tx_commit(tx); } /* * We have to be careful when adding new vdevs to an existing pool. * If other threads start allocating from these vdevs before we * sync the config cache, and we lose power, then upon reboot we may * fail to open the pool because there are DVAs that the config cache * can't translate. Therefore, we first add the vdevs without * initializing metaslabs; sync the config cache (via spa_vdev_exit()); * and then let spa_config_update() initialize the new metaslabs. * * spa_load() checks for added-but-not-initialized vdevs, so that * if we lose power at any point in this sequence, the remaining * steps will be completed the next time we load the pool. */ (void) spa_vdev_exit(spa, vd, txg, 0); mutex_enter(&spa_namespace_lock); spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); spa_event_notify(spa, NULL, NULL, ESC_ZFS_VDEV_ADD); mutex_exit(&spa_namespace_lock); return (0); } /* * Attach a device to a vdev specified by its guid. The vdev type can be * a mirror, a raidz, or a leaf device that is also a top-level (e.g. a * single device). When the vdev is a single device, a mirror vdev will be * automatically inserted. * * If 'replacing' is specified, the new device is intended to replace the * existing device; in this case the two devices are made into their own * mirror using the 'replacing' vdev, which is functionally identical to * the mirror vdev (it actually reuses all the same ops) but has a few * extra rules: you can't attach to it after it's been created, and upon * completion of resilvering, the first disk (the one being replaced) * is automatically detached. * * If 'rebuild' is specified, then sequential reconstruction (a.ka. rebuild) * should be performed instead of traditional healing reconstruction. From * an administrators perspective these are both resilver operations. */ int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing, int rebuild) { uint64_t txg, dtl_max_txg; vdev_t *rvd = spa->spa_root_vdev; vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd; vdev_ops_t *pvops; char *oldvdpath, *newvdpath; int newvd_isspare = B_FALSE; int error; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); oldvd = spa_lookup_by_guid(spa, guid, B_FALSE); ASSERT(MUTEX_HELD(&spa_namespace_lock)); if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) { error = (spa_has_checkpoint(spa)) ? ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT; return (spa_vdev_exit(spa, NULL, txg, error)); } if (rebuild) { if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD)) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); if (dsl_scan_resilvering(spa_get_dsl(spa)) || dsl_scan_resilver_scheduled(spa_get_dsl(spa))) { return (spa_vdev_exit(spa, NULL, txg, ZFS_ERR_RESILVER_IN_PROGRESS)); } } else { if (vdev_rebuild_active(rvd)) return (spa_vdev_exit(spa, NULL, txg, ZFS_ERR_REBUILD_IN_PROGRESS)); } if (spa->spa_vdev_removal != NULL) { return (spa_vdev_exit(spa, NULL, txg, ZFS_ERR_DEVRM_IN_PROGRESS)); } if (oldvd == NULL) return (spa_vdev_exit(spa, NULL, txg, ENODEV)); boolean_t raidz = oldvd->vdev_ops == &vdev_raidz_ops; if (raidz) { if (!spa_feature_is_enabled(spa, SPA_FEATURE_RAIDZ_EXPANSION)) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); /* * Can't expand a raidz while prior expand is in progress. */ if (spa->spa_raidz_expand != NULL) { return (spa_vdev_exit(spa, NULL, txg, ZFS_ERR_RAIDZ_EXPAND_IN_PROGRESS)); } } else if (!oldvd->vdev_ops->vdev_op_leaf) { return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); } if (raidz) pvd = oldvd; else pvd = oldvd->vdev_parent; if (spa_config_parse(spa, &newrootvd, nvroot, NULL, 0, VDEV_ALLOC_ATTACH) != 0) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); if (newrootvd->vdev_children != 1) return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); newvd = newrootvd->vdev_child[0]; if (!newvd->vdev_ops->vdev_op_leaf) return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); if ((error = vdev_create(newrootvd, txg, replacing)) != 0) return (spa_vdev_exit(spa, newrootvd, txg, error)); /* * log, dedup and special vdevs should not be replaced by spares. */ if ((oldvd->vdev_top->vdev_alloc_bias != VDEV_BIAS_NONE || oldvd->vdev_top->vdev_islog) && newvd->vdev_isspare) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } /* * A dRAID spare can only replace a child of its parent dRAID vdev. */ if (newvd->vdev_ops == &vdev_draid_spare_ops && oldvd->vdev_top != vdev_draid_spare_get_parent(newvd)) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } if (rebuild) { /* * For rebuilds, the top vdev must support reconstruction * using only space maps. This means the only allowable * vdevs types are the root vdev, a mirror, or dRAID. */ tvd = pvd; if (pvd->vdev_top != NULL) tvd = pvd->vdev_top; if (tvd->vdev_ops != &vdev_mirror_ops && tvd->vdev_ops != &vdev_root_ops && tvd->vdev_ops != &vdev_draid_ops) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } } if (!replacing) { /* * For attach, the only allowable parent is a mirror or * the root vdev. A raidz vdev can be attached to, but * you cannot attach to a raidz child. */ if (pvd->vdev_ops != &vdev_mirror_ops && pvd->vdev_ops != &vdev_root_ops && !raidz) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); pvops = &vdev_mirror_ops; } else { /* * Active hot spares can only be replaced by inactive hot * spares. */ if (pvd->vdev_ops == &vdev_spare_ops && oldvd->vdev_isspare && !spa_has_spare(spa, newvd->vdev_guid)) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); /* * If the source is a hot spare, and the parent isn't already a * spare, then we want to create a new hot spare. Otherwise, we * want to create a replacing vdev. The user is not allowed to * attach to a spared vdev child unless the 'isspare' state is * the same (spare replaces spare, non-spare replaces * non-spare). */ if (pvd->vdev_ops == &vdev_replacing_ops && spa_version(spa) < SPA_VERSION_MULTI_REPLACE) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } else if (pvd->vdev_ops == &vdev_spare_ops && newvd->vdev_isspare != oldvd->vdev_isspare) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } if (newvd->vdev_isspare) pvops = &vdev_spare_ops; else pvops = &vdev_replacing_ops; } /* * Make sure the new device is big enough. */ vdev_t *min_vdev = raidz ? oldvd->vdev_child[0] : oldvd; if (newvd->vdev_asize < vdev_get_min_asize(min_vdev)) return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW)); /* * The new device cannot have a higher alignment requirement * than the top-level vdev. */ if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); /* * RAIDZ-expansion-specific checks. */ if (raidz) { if (vdev_raidz_attach_check(newvd) != 0) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); /* * Fail early if a child is not healthy or being replaced */ for (int i = 0; i < oldvd->vdev_children; i++) { if (vdev_is_dead(oldvd->vdev_child[i]) || !oldvd->vdev_child[i]->vdev_ops->vdev_op_leaf) { return (spa_vdev_exit(spa, newrootvd, txg, ENXIO)); } /* Also fail if reserved boot area is in-use */ if (vdev_check_boot_reserve(spa, oldvd->vdev_child[i]) != 0) { return (spa_vdev_exit(spa, newrootvd, txg, EADDRINUSE)); } } } if (raidz) { /* * Note: oldvdpath is freed by spa_strfree(), but * kmem_asprintf() is freed by kmem_strfree(), so we have to * move it to a spa_strdup-ed string. */ char *tmp = kmem_asprintf("raidz%u-%u", (uint_t)vdev_get_nparity(oldvd), (uint_t)oldvd->vdev_id); oldvdpath = spa_strdup(tmp); kmem_strfree(tmp); } else { oldvdpath = spa_strdup(oldvd->vdev_path); } newvdpath = spa_strdup(newvd->vdev_path); /* * If this is an in-place replacement, update oldvd's path and devid * to make it distinguishable from newvd, and unopenable from now on. */ if (strcmp(oldvdpath, newvdpath) == 0) { spa_strfree(oldvd->vdev_path); oldvd->vdev_path = kmem_alloc(strlen(newvdpath) + 5, KM_SLEEP); (void) sprintf(oldvd->vdev_path, "%s/old", newvdpath); if (oldvd->vdev_devid != NULL) { spa_strfree(oldvd->vdev_devid); oldvd->vdev_devid = NULL; } spa_strfree(oldvdpath); oldvdpath = spa_strdup(oldvd->vdev_path); } /* * If the parent is not a mirror, or if we're replacing, insert the new * mirror/replacing/spare vdev above oldvd. */ if (!raidz && pvd->vdev_ops != pvops) { pvd = vdev_add_parent(oldvd, pvops); ASSERT(pvd->vdev_ops == pvops); ASSERT(oldvd->vdev_parent == pvd); } ASSERT(pvd->vdev_top->vdev_parent == rvd); /* * Extract the new device from its root and add it to pvd. */ vdev_remove_child(newrootvd, newvd); newvd->vdev_id = pvd->vdev_children; newvd->vdev_crtxg = oldvd->vdev_crtxg; vdev_add_child(pvd, newvd); /* * Reevaluate the parent vdev state. */ vdev_propagate_state(pvd); tvd = newvd->vdev_top; ASSERT(pvd->vdev_top == tvd); ASSERT(tvd->vdev_parent == rvd); vdev_config_dirty(tvd); /* * Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account * for any dmu_sync-ed blocks. It will propagate upward when * spa_vdev_exit() calls vdev_dtl_reassess(). */ dtl_max_txg = txg + TXG_CONCURRENT_STATES; if (raidz) { /* * Wait for the youngest allocations and frees to sync, * and then wait for the deferral of those frees to finish. */ spa_vdev_config_exit(spa, NULL, txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG); vdev_initialize_stop_all(tvd, VDEV_INITIALIZE_ACTIVE); vdev_trim_stop_all(tvd, VDEV_TRIM_ACTIVE); vdev_autotrim_stop_wait(tvd); dtl_max_txg = spa_vdev_config_enter(spa); tvd->vdev_rz_expanding = B_TRUE; vdev_dirty_leaves(tvd, VDD_DTL, dtl_max_txg); vdev_config_dirty(tvd); dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool, dtl_max_txg); dsl_sync_task_nowait(spa->spa_dsl_pool, vdev_raidz_attach_sync, newvd, tx); dmu_tx_commit(tx); } else { vdev_dtl_dirty(newvd, DTL_MISSING, TXG_INITIAL, dtl_max_txg - TXG_INITIAL); if (newvd->vdev_isspare) { spa_spare_activate(newvd); spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_SPARE); } newvd_isspare = newvd->vdev_isspare; /* * Mark newvd's DTL dirty in this txg. */ vdev_dirty(tvd, VDD_DTL, newvd, txg); /* * Schedule the resilver or rebuild to restart in the future. * We do this to ensure that dmu_sync-ed blocks have been * stitched into the respective datasets. */ if (rebuild) { newvd->vdev_rebuild_txg = txg; vdev_rebuild(tvd); } else { newvd->vdev_resilver_txg = txg; if (dsl_scan_resilvering(spa_get_dsl(spa)) && spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER)) { vdev_defer_resilver(newvd); } else { dsl_scan_restart_resilver(spa->spa_dsl_pool, dtl_max_txg); } } } if (spa->spa_bootfs) spa_event_notify(spa, newvd, NULL, ESC_ZFS_BOOTFS_VDEV_ATTACH); spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_ATTACH); /* * Commit the config */ (void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0); spa_history_log_internal(spa, "vdev attach", NULL, "%s vdev=%s %s vdev=%s", replacing && newvd_isspare ? "spare in" : replacing ? "replace" : "attach", newvdpath, replacing ? "for" : "to", oldvdpath); spa_strfree(oldvdpath); spa_strfree(newvdpath); return (0); } /* * Detach a device from a mirror or replacing vdev. * * If 'replace_done' is specified, only detach if the parent * is a replacing or a spare vdev. */ int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) { uint64_t txg; int error; vdev_t *rvd __maybe_unused = spa->spa_root_vdev; vdev_t *vd, *pvd, *cvd, *tvd; boolean_t unspare = B_FALSE; uint64_t unspare_guid = 0; char *vdpath; ASSERT(spa_writeable(spa)); txg = spa_vdev_detach_enter(spa, guid); vd = spa_lookup_by_guid(spa, guid, B_FALSE); /* * Besides being called directly from the userland through the * ioctl interface, spa_vdev_detach() can be potentially called * at the end of spa_vdev_resilver_done(). * * In the regular case, when we have a checkpoint this shouldn't * happen as we never empty the DTLs of a vdev during the scrub * [see comment in dsl_scan_done()]. Thus spa_vdev_resilvering_done() * should never get here when we have a checkpoint. * * That said, even in a case when we checkpoint the pool exactly * as spa_vdev_resilver_done() calls this function everything * should be fine as the resilver will return right away. */ ASSERT(MUTEX_HELD(&spa_namespace_lock)); if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) { error = (spa_has_checkpoint(spa)) ? ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT; return (spa_vdev_exit(spa, NULL, txg, error)); } if (vd == NULL) return (spa_vdev_exit(spa, NULL, txg, ENODEV)); if (!vd->vdev_ops->vdev_op_leaf) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); pvd = vd->vdev_parent; /* * If the parent/child relationship is not as expected, don't do it. * Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing * vdev that's replacing B with C. The user's intent in replacing * is to go from M(A,B) to M(A,C). If the user decides to cancel * the replace by detaching C, the expected behavior is to end up * M(A,B). But suppose that right after deciding to detach C, * the replacement of B completes. We would have M(A,C), and then * ask to detach C, which would leave us with just A -- not what * the user wanted. To prevent this, we make sure that the * parent/child relationship hasn't changed -- in this example, * that C's parent is still the replacing vdev R. */ if (pvd->vdev_guid != pguid && pguid != 0) return (spa_vdev_exit(spa, NULL, txg, EBUSY)); /* * Only 'replacing' or 'spare' vdevs can be replaced. */ if (replace_done && pvd->vdev_ops != &vdev_replacing_ops && pvd->vdev_ops != &vdev_spare_ops) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); ASSERT(pvd->vdev_ops != &vdev_spare_ops || spa_version(spa) >= SPA_VERSION_SPARES); /* * Only mirror, replacing, and spare vdevs support detach. */ if (pvd->vdev_ops != &vdev_replacing_ops && pvd->vdev_ops != &vdev_mirror_ops && pvd->vdev_ops != &vdev_spare_ops) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); /* * If this device has the only valid copy of some data, * we cannot safely detach it. */ if (vdev_dtl_required(vd)) return (spa_vdev_exit(spa, NULL, txg, EBUSY)); ASSERT(pvd->vdev_children >= 2); /* * If we are detaching the second disk from a replacing vdev, then * check to see if we changed the original vdev's path to have "/old" * at the end in spa_vdev_attach(). If so, undo that change now. */ if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 && vd->vdev_path != NULL) { size_t len = strlen(vd->vdev_path); for (int c = 0; c < pvd->vdev_children; c++) { cvd = pvd->vdev_child[c]; if (cvd == vd || cvd->vdev_path == NULL) continue; if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 && strcmp(cvd->vdev_path + len, "/old") == 0) { spa_strfree(cvd->vdev_path); cvd->vdev_path = spa_strdup(vd->vdev_path); break; } } } /* * If we are detaching the original disk from a normal spare, then it * implies that the spare should become a real disk, and be removed * from the active spare list for the pool. dRAID spares on the * other hand are coupled to the pool and thus should never be removed * from the spares list. */ if (pvd->vdev_ops == &vdev_spare_ops && vd->vdev_id == 0) { vdev_t *last_cvd = pvd->vdev_child[pvd->vdev_children - 1]; if (last_cvd->vdev_isspare && last_cvd->vdev_ops != &vdev_draid_spare_ops) { unspare = B_TRUE; } } /* * Erase the disk labels so the disk can be used for other things. * This must be done after all other error cases are handled, * but before we disembowel vd (so we can still do I/O to it). * But if we can't do it, don't treat the error as fatal -- * it may be that the unwritability of the disk is the reason * it's being detached! */ (void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); /* * Remove vd from its parent and compact the parent's children. */ vdev_remove_child(pvd, vd); vdev_compact_children(pvd); /* * Remember one of the remaining children so we can get tvd below. */ cvd = pvd->vdev_child[pvd->vdev_children - 1]; /* * If we need to remove the remaining child from the list of hot spares, * do it now, marking the vdev as no longer a spare in the process. * We must do this before vdev_remove_parent(), because that can * change the GUID if it creates a new toplevel GUID. For a similar * reason, we must remove the spare now, in the same txg as the detach; * otherwise someone could attach a new sibling, change the GUID, and * the subsequent attempt to spa_vdev_remove(unspare_guid) would fail. */ if (unspare) { ASSERT(cvd->vdev_isspare); spa_spare_remove(cvd); unspare_guid = cvd->vdev_guid; (void) spa_vdev_remove(spa, unspare_guid, B_TRUE); cvd->vdev_unspare = B_TRUE; } /* * If the parent mirror/replacing vdev only has one child, * the parent is no longer needed. Remove it from the tree. */ if (pvd->vdev_children == 1) { if (pvd->vdev_ops == &vdev_spare_ops) cvd->vdev_unspare = B_FALSE; vdev_remove_parent(cvd); } /* * We don't set tvd until now because the parent we just removed * may have been the previous top-level vdev. */ tvd = cvd->vdev_top; ASSERT(tvd->vdev_parent == rvd); /* * Reevaluate the parent vdev state. */ vdev_propagate_state(cvd); /* * If the 'autoexpand' property is set on the pool then automatically * try to expand the size of the pool. For example if the device we * just detached was smaller than the others, it may be possible to * add metaslabs (i.e. grow the pool). We need to reopen the vdev * first so that we can obtain the updated sizes of the leaf vdevs. */ if (spa->spa_autoexpand) { vdev_reopen(tvd); vdev_expand(tvd, txg); } vdev_config_dirty(tvd); /* * Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that * vd->vdev_detached is set and free vd's DTL object in syncing context. * But first make sure we're not on any *other* txg's DTL list, to * prevent vd from being accessed after it's freed. */ vdpath = spa_strdup(vd->vdev_path ? vd->vdev_path : "none"); for (int t = 0; t < TXG_SIZE; t++) (void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t); vd->vdev_detached = B_TRUE; vdev_dirty(tvd, VDD_DTL, vd, txg); spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE); spa_notify_waiters(spa); /* hang on to the spa before we release the lock */ spa_open_ref(spa, FTAG); error = spa_vdev_exit(spa, vd, txg, 0); spa_history_log_internal(spa, "detach", NULL, "vdev=%s", vdpath); spa_strfree(vdpath); /* * If this was the removal of the original device in a hot spare vdev, * then we want to go through and remove the device from the hot spare * list of every other pool. */ if (unspare) { spa_t *altspa = NULL; mutex_enter(&spa_namespace_lock); while ((altspa = spa_next(altspa)) != NULL) { if (altspa->spa_state != POOL_STATE_ACTIVE || altspa == spa) continue; spa_open_ref(altspa, FTAG); mutex_exit(&spa_namespace_lock); (void) spa_vdev_remove(altspa, unspare_guid, B_TRUE); mutex_enter(&spa_namespace_lock); spa_close(altspa, FTAG); } mutex_exit(&spa_namespace_lock); /* search the rest of the vdevs for spares to remove */ spa_vdev_resilver_done(spa); } /* all done with the spa; OK to release */ mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); mutex_exit(&spa_namespace_lock); return (error); } static int spa_vdev_initialize_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type, list_t *vd_list) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); /* Look up vdev and ensure it's a leaf. */ vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE); if (vd == NULL || vd->vdev_detached) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(ENODEV)); } else if (!vd->vdev_ops->vdev_op_leaf || !vdev_is_concrete(vd)) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(EINVAL)); } else if (!vdev_writeable(vd)) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(EROFS)); } mutex_enter(&vd->vdev_initialize_lock); spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); /* * When we activate an initialize action we check to see * if the vdev_initialize_thread is NULL. We do this instead * of using the vdev_initialize_state since there might be * a previous initialization process which has completed but * the thread is not exited. */ if (cmd_type == POOL_INITIALIZE_START && (vd->vdev_initialize_thread != NULL || vd->vdev_top->vdev_removing || vd->vdev_top->vdev_rz_expanding)) { mutex_exit(&vd->vdev_initialize_lock); return (SET_ERROR(EBUSY)); } else if (cmd_type == POOL_INITIALIZE_CANCEL && (vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE && vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED)) { mutex_exit(&vd->vdev_initialize_lock); return (SET_ERROR(ESRCH)); } else if (cmd_type == POOL_INITIALIZE_SUSPEND && vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE) { mutex_exit(&vd->vdev_initialize_lock); return (SET_ERROR(ESRCH)); } else if (cmd_type == POOL_INITIALIZE_UNINIT && vd->vdev_initialize_thread != NULL) { mutex_exit(&vd->vdev_initialize_lock); return (SET_ERROR(EBUSY)); } switch (cmd_type) { case POOL_INITIALIZE_START: vdev_initialize(vd); break; case POOL_INITIALIZE_CANCEL: vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, vd_list); break; case POOL_INITIALIZE_SUSPEND: vdev_initialize_stop(vd, VDEV_INITIALIZE_SUSPENDED, vd_list); break; case POOL_INITIALIZE_UNINIT: vdev_uninitialize(vd); break; default: panic("invalid cmd_type %llu", (unsigned long long)cmd_type); } mutex_exit(&vd->vdev_initialize_lock); return (0); } int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, nvlist_t *vdev_errlist) { int total_errors = 0; list_t vd_list; list_create(&vd_list, sizeof (vdev_t), offsetof(vdev_t, vdev_initialize_node)); /* * We hold the namespace lock through the whole function * to prevent any changes to the pool while we're starting or * stopping initialization. The config and state locks are held so that * we can properly assess the vdev state before we commit to * the initializing operation. */ mutex_enter(&spa_namespace_lock); for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL); pair != NULL; pair = nvlist_next_nvpair(nv, pair)) { uint64_t vdev_guid = fnvpair_value_uint64(pair); int error = spa_vdev_initialize_impl(spa, vdev_guid, cmd_type, &vd_list); if (error != 0) { char guid_as_str[MAXNAMELEN]; (void) snprintf(guid_as_str, sizeof (guid_as_str), "%llu", (unsigned long long)vdev_guid); fnvlist_add_int64(vdev_errlist, guid_as_str, error); total_errors++; } } /* Wait for all initialize threads to stop. */ vdev_initialize_stop_wait(spa, &vd_list); /* Sync out the initializing state */ txg_wait_synced(spa->spa_dsl_pool, 0); mutex_exit(&spa_namespace_lock); list_destroy(&vd_list); return (total_errors); } static int spa_vdev_trim_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type, uint64_t rate, boolean_t partial, boolean_t secure, list_t *vd_list) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); /* Look up vdev and ensure it's a leaf. */ vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE); if (vd == NULL || vd->vdev_detached) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(ENODEV)); } else if (!vd->vdev_ops->vdev_op_leaf || !vdev_is_concrete(vd)) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(EINVAL)); } else if (!vdev_writeable(vd)) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(EROFS)); } else if (!vd->vdev_has_trim) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(EOPNOTSUPP)); } else if (secure && !vd->vdev_has_securetrim) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (SET_ERROR(EOPNOTSUPP)); } mutex_enter(&vd->vdev_trim_lock); spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); /* * When we activate a TRIM action we check to see if the * vdev_trim_thread is NULL. We do this instead of using the * vdev_trim_state since there might be a previous TRIM process * which has completed but the thread is not exited. */ if (cmd_type == POOL_TRIM_START && (vd->vdev_trim_thread != NULL || vd->vdev_top->vdev_removing || vd->vdev_top->vdev_rz_expanding)) { mutex_exit(&vd->vdev_trim_lock); return (SET_ERROR(EBUSY)); } else if (cmd_type == POOL_TRIM_CANCEL && (vd->vdev_trim_state != VDEV_TRIM_ACTIVE && vd->vdev_trim_state != VDEV_TRIM_SUSPENDED)) { mutex_exit(&vd->vdev_trim_lock); return (SET_ERROR(ESRCH)); } else if (cmd_type == POOL_TRIM_SUSPEND && vd->vdev_trim_state != VDEV_TRIM_ACTIVE) { mutex_exit(&vd->vdev_trim_lock); return (SET_ERROR(ESRCH)); } switch (cmd_type) { case POOL_TRIM_START: vdev_trim(vd, rate, partial, secure); break; case POOL_TRIM_CANCEL: vdev_trim_stop(vd, VDEV_TRIM_CANCELED, vd_list); break; case POOL_TRIM_SUSPEND: vdev_trim_stop(vd, VDEV_TRIM_SUSPENDED, vd_list); break; default: panic("invalid cmd_type %llu", (unsigned long long)cmd_type); } mutex_exit(&vd->vdev_trim_lock); return (0); } /* * Initiates a manual TRIM for the requested vdevs. This kicks off individual * TRIM threads for each child vdev. These threads pass over all of the free * space in the vdev's metaslabs and issues TRIM commands for that space. */ int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist) { int total_errors = 0; list_t vd_list; list_create(&vd_list, sizeof (vdev_t), offsetof(vdev_t, vdev_trim_node)); /* * We hold the namespace lock through the whole function * to prevent any changes to the pool while we're starting or * stopping TRIM. The config and state locks are held so that * we can properly assess the vdev state before we commit to * the TRIM operation. */ mutex_enter(&spa_namespace_lock); for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL); pair != NULL; pair = nvlist_next_nvpair(nv, pair)) { uint64_t vdev_guid = fnvpair_value_uint64(pair); int error = spa_vdev_trim_impl(spa, vdev_guid, cmd_type, rate, partial, secure, &vd_list); if (error != 0) { char guid_as_str[MAXNAMELEN]; (void) snprintf(guid_as_str, sizeof (guid_as_str), "%llu", (unsigned long long)vdev_guid); fnvlist_add_int64(vdev_errlist, guid_as_str, error); total_errors++; } } /* Wait for all TRIM threads to stop. */ vdev_trim_stop_wait(spa, &vd_list); /* Sync out the TRIM state */ txg_wait_synced(spa->spa_dsl_pool, 0); mutex_exit(&spa_namespace_lock); list_destroy(&vd_list); return (total_errors); } /* * Split a set of devices from their mirrors, and create a new pool from them. */ int spa_vdev_split_mirror(spa_t *spa, const char *newname, nvlist_t *config, nvlist_t *props, boolean_t exp) { int error = 0; uint64_t txg, *glist; spa_t *newspa; uint_t c, children, lastlog; nvlist_t **child, *nvl, *tmp; dmu_tx_t *tx; const char *altroot = NULL; vdev_t *rvd, **vml = NULL; /* vdev modify list */ boolean_t activate_slog; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); ASSERT(MUTEX_HELD(&spa_namespace_lock)); if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) { error = (spa_has_checkpoint(spa)) ? ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT; return (spa_vdev_exit(spa, NULL, txg, error)); } /* clear the log and flush everything up to now */ activate_slog = spa_passivate_log(spa); (void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); error = spa_reset_logs(spa); txg = spa_vdev_config_enter(spa); if (activate_slog) spa_activate_log(spa); if (error != 0) return (spa_vdev_exit(spa, NULL, txg, error)); /* check new spa name before going any further */ if (spa_lookup(newname) != NULL) return (spa_vdev_exit(spa, NULL, txg, EEXIST)); /* * scan through all the children to ensure they're all mirrors */ if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 || nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); /* first, check to ensure we've got the right child count */ rvd = spa->spa_root_vdev; lastlog = 0; for (c = 0; c < rvd->vdev_children; c++) { vdev_t *vd = rvd->vdev_child[c]; /* don't count the holes & logs as children */ if (vd->vdev_islog || (vd->vdev_ops != &vdev_indirect_ops && !vdev_is_concrete(vd))) { if (lastlog == 0) lastlog = c; continue; } lastlog = 0; } if (children != (lastlog != 0 ? lastlog : rvd->vdev_children)) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); /* next, ensure no spare or cache devices are part of the split */ if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 || nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP); glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP); /* then, loop over each vdev and validate it */ for (c = 0; c < children; c++) { uint64_t is_hole = 0; (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, &is_hole); if (is_hole != 0) { if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole || spa->spa_root_vdev->vdev_child[c]->vdev_islog) { continue; } else { error = SET_ERROR(EINVAL); break; } } /* deal with indirect vdevs */ if (spa->spa_root_vdev->vdev_child[c]->vdev_ops == &vdev_indirect_ops) continue; /* which disk is going to be split? */ if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID, &glist[c]) != 0) { error = SET_ERROR(EINVAL); break; } /* look it up in the spa */ vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE); if (vml[c] == NULL) { error = SET_ERROR(ENODEV); break; } /* make sure there's nothing stopping the split */ if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops || vml[c]->vdev_islog || !vdev_is_concrete(vml[c]) || vml[c]->vdev_isspare || vml[c]->vdev_isl2cache || !vdev_writeable(vml[c]) || vml[c]->vdev_children != 0 || vml[c]->vdev_state != VDEV_STATE_HEALTHY || c != spa->spa_root_vdev->vdev_child[c]->vdev_id) { error = SET_ERROR(EINVAL); break; } if (vdev_dtl_required(vml[c]) || vdev_resilver_needed(vml[c], NULL, NULL)) { error = SET_ERROR(EBUSY); break; } /* we need certain info from the top level */ fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY, vml[c]->vdev_top->vdev_ms_array); fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT, vml[c]->vdev_top->vdev_ms_shift); fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE, vml[c]->vdev_top->vdev_asize); fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT, vml[c]->vdev_top->vdev_ashift); /* transfer per-vdev ZAPs */ ASSERT3U(vml[c]->vdev_leaf_zap, !=, 0); VERIFY0(nvlist_add_uint64(child[c], ZPOOL_CONFIG_VDEV_LEAF_ZAP, vml[c]->vdev_leaf_zap)); ASSERT3U(vml[c]->vdev_top->vdev_top_zap, !=, 0); VERIFY0(nvlist_add_uint64(child[c], ZPOOL_CONFIG_VDEV_TOP_ZAP, vml[c]->vdev_parent->vdev_top_zap)); } if (error != 0) { kmem_free(vml, children * sizeof (vdev_t *)); kmem_free(glist, children * sizeof (uint64_t)); return (spa_vdev_exit(spa, NULL, txg, error)); } /* stop writers from using the disks */ for (c = 0; c < children; c++) { if (vml[c] != NULL) vml[c]->vdev_offline = B_TRUE; } vdev_reopen(spa->spa_root_vdev); /* * Temporarily record the splitting vdevs in the spa config. This * will disappear once the config is regenerated. */ nvl = fnvlist_alloc(); fnvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, glist, children); kmem_free(glist, children * sizeof (uint64_t)); mutex_enter(&spa->spa_props_lock); fnvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, nvl); mutex_exit(&spa->spa_props_lock); spa->spa_config_splitting = nvl; vdev_config_dirty(spa->spa_root_vdev); /* configure and create the new pool */ fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname); fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE); fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa)); fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, spa->spa_config_txg); fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, spa_generate_guid(NULL)); VERIFY0(nvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)); (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); /* add the new pool to the namespace */ newspa = spa_add(newname, config, altroot); newspa->spa_avz_action = AVZ_ACTION_REBUILD; newspa->spa_config_txg = spa->spa_config_txg; spa_set_log_state(newspa, SPA_LOG_CLEAR); /* release the spa config lock, retaining the namespace lock */ spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); if (zio_injection_enabled) zio_handle_panic_injection(spa, FTAG, 1); spa_activate(newspa, spa_mode_global); spa_async_suspend(newspa); /* * Temporarily stop the initializing and TRIM activity. We set the * state to ACTIVE so that we know to resume initializing or TRIM * once the split has completed. */ list_t vd_initialize_list; list_create(&vd_initialize_list, sizeof (vdev_t), offsetof(vdev_t, vdev_initialize_node)); list_t vd_trim_list; list_create(&vd_trim_list, sizeof (vdev_t), offsetof(vdev_t, vdev_trim_node)); for (c = 0; c < children; c++) { if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) { mutex_enter(&vml[c]->vdev_initialize_lock); vdev_initialize_stop(vml[c], VDEV_INITIALIZE_ACTIVE, &vd_initialize_list); mutex_exit(&vml[c]->vdev_initialize_lock); mutex_enter(&vml[c]->vdev_trim_lock); vdev_trim_stop(vml[c], VDEV_TRIM_ACTIVE, &vd_trim_list); mutex_exit(&vml[c]->vdev_trim_lock); } } vdev_initialize_stop_wait(spa, &vd_initialize_list); vdev_trim_stop_wait(spa, &vd_trim_list); list_destroy(&vd_initialize_list); list_destroy(&vd_trim_list); newspa->spa_config_source = SPA_CONFIG_SRC_SPLIT; newspa->spa_is_splitting = B_TRUE; /* create the new pool from the disks of the original pool */ error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE); if (error) goto out; /* if that worked, generate a real config for the new pool */ if (newspa->spa_root_vdev != NULL) { newspa->spa_config_splitting = fnvlist_alloc(); fnvlist_add_uint64(newspa->spa_config_splitting, ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa)); spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL, B_TRUE)); } /* set the props */ if (props != NULL) { spa_configfile_set(newspa, props, B_FALSE); error = spa_prop_set(newspa, props); if (error) goto out; } /* flush everything */ txg = spa_vdev_config_enter(newspa); vdev_config_dirty(newspa->spa_root_vdev); (void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG); if (zio_injection_enabled) zio_handle_panic_injection(spa, FTAG, 2); spa_async_resume(newspa); /* finally, update the original pool's config */ txg = spa_vdev_config_enter(spa); tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) dmu_tx_abort(tx); for (c = 0; c < children; c++) { if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) { vdev_t *tvd = vml[c]->vdev_top; /* * Need to be sure the detachable VDEV is not * on any *other* txg's DTL list to prevent it * from being accessed after it's freed. */ for (int t = 0; t < TXG_SIZE; t++) { (void) txg_list_remove_this( &tvd->vdev_dtl_list, vml[c], t); } vdev_split(vml[c]); if (error == 0) spa_history_log_internal(spa, "detach", tx, "vdev=%s", vml[c]->vdev_path); vdev_free(vml[c]); } } spa->spa_avz_action = AVZ_ACTION_REBUILD; vdev_config_dirty(spa->spa_root_vdev); spa->spa_config_splitting = NULL; nvlist_free(nvl); if (error == 0) dmu_tx_commit(tx); (void) spa_vdev_exit(spa, NULL, txg, 0); if (zio_injection_enabled) zio_handle_panic_injection(spa, FTAG, 3); /* split is complete; log a history record */ spa_history_log_internal(newspa, "split", NULL, "from pool %s", spa_name(spa)); newspa->spa_is_splitting = B_FALSE; kmem_free(vml, children * sizeof (vdev_t *)); /* if we're not going to mount the filesystems in userland, export */ if (exp) error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL, B_FALSE, B_FALSE); return (error); out: spa_unload(newspa); spa_deactivate(newspa); spa_remove(newspa); txg = spa_vdev_config_enter(spa); /* re-online all offlined disks */ for (c = 0; c < children; c++) { if (vml[c] != NULL) vml[c]->vdev_offline = B_FALSE; } /* restart initializing or trimming disks as necessary */ spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART); spa_async_request(spa, SPA_ASYNC_TRIM_RESTART); spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART); vdev_reopen(spa->spa_root_vdev); nvlist_free(spa->spa_config_splitting); spa->spa_config_splitting = NULL; (void) spa_vdev_exit(spa, NULL, txg, error); kmem_free(vml, children * sizeof (vdev_t *)); return (error); } /* * Find any device that's done replacing, or a vdev marked 'unspare' that's * currently spared, so we can detach it. */ static vdev_t * spa_vdev_resilver_done_hunt(vdev_t *vd) { vdev_t *newvd, *oldvd; for (int c = 0; c < vd->vdev_children; c++) { oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]); if (oldvd != NULL) return (oldvd); } /* * Check for a completed replacement. We always consider the first * vdev in the list to be the oldest vdev, and the last one to be * the newest (see spa_vdev_attach() for how that works). In * the case where the newest vdev is faulted, we will not automatically * remove it after a resilver completes. This is OK as it will require * user intervention to determine which disk the admin wishes to keep. */ if (vd->vdev_ops == &vdev_replacing_ops) { ASSERT(vd->vdev_children > 1); newvd = vd->vdev_child[vd->vdev_children - 1]; oldvd = vd->vdev_child[0]; if (vdev_dtl_empty(newvd, DTL_MISSING) && vdev_dtl_empty(newvd, DTL_OUTAGE) && !vdev_dtl_required(oldvd)) return (oldvd); } /* * Check for a completed resilver with the 'unspare' flag set. * Also potentially update faulted state. */ if (vd->vdev_ops == &vdev_spare_ops) { vdev_t *first = vd->vdev_child[0]; vdev_t *last = vd->vdev_child[vd->vdev_children - 1]; if (last->vdev_unspare) { oldvd = first; newvd = last; } else if (first->vdev_unspare) { oldvd = last; newvd = first; } else { oldvd = NULL; } if (oldvd != NULL && vdev_dtl_empty(newvd, DTL_MISSING) && vdev_dtl_empty(newvd, DTL_OUTAGE) && !vdev_dtl_required(oldvd)) return (oldvd); vdev_propagate_state(vd); /* * If there are more than two spares attached to a disk, * and those spares are not required, then we want to * attempt to free them up now so that they can be used * by other pools. Once we're back down to a single * disk+spare, we stop removing them. */ if (vd->vdev_children > 2) { newvd = vd->vdev_child[1]; if (newvd->vdev_isspare && last->vdev_isspare && vdev_dtl_empty(last, DTL_MISSING) && vdev_dtl_empty(last, DTL_OUTAGE) && !vdev_dtl_required(newvd)) return (newvd); } } return (NULL); } static void spa_vdev_resilver_done(spa_t *spa) { vdev_t *vd, *pvd, *ppvd; uint64_t guid, sguid, pguid, ppguid; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) { pvd = vd->vdev_parent; ppvd = pvd->vdev_parent; guid = vd->vdev_guid; pguid = pvd->vdev_guid; ppguid = ppvd->vdev_guid; sguid = 0; /* * If we have just finished replacing a hot spared device, then * we need to detach the parent's first child (the original hot * spare) as well. */ if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 && ppvd->vdev_children == 2) { ASSERT(pvd->vdev_ops == &vdev_replacing_ops); sguid = ppvd->vdev_child[1]->vdev_guid; } ASSERT(vd->vdev_resilver_txg == 0 || !vdev_dtl_required(vd)); spa_config_exit(spa, SCL_ALL, FTAG); if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0) return; if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0) return; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); } spa_config_exit(spa, SCL_ALL, FTAG); /* * If a detach was not performed above replace waiters will not have * been notified. In which case we must do so now. */ spa_notify_waiters(spa); } /* * Update the stored path or FRU for this vdev. */ static int spa_vdev_set_common(spa_t *spa, uint64_t guid, const char *value, boolean_t ispath) { vdev_t *vd; boolean_t sync = B_FALSE; ASSERT(spa_writeable(spa)); spa_vdev_state_enter(spa, SCL_ALL); if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL) return (spa_vdev_state_exit(spa, NULL, ENOENT)); if (!vd->vdev_ops->vdev_op_leaf) return (spa_vdev_state_exit(spa, NULL, ENOTSUP)); if (ispath) { if (strcmp(value, vd->vdev_path) != 0) { spa_strfree(vd->vdev_path); vd->vdev_path = spa_strdup(value); sync = B_TRUE; } } else { if (vd->vdev_fru == NULL) { vd->vdev_fru = spa_strdup(value); sync = B_TRUE; } else if (strcmp(value, vd->vdev_fru) != 0) { spa_strfree(vd->vdev_fru); vd->vdev_fru = spa_strdup(value); sync = B_TRUE; } } return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0)); } int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath) { return (spa_vdev_set_common(spa, guid, newpath, B_TRUE)); } int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru) { return (spa_vdev_set_common(spa, guid, newfru, B_FALSE)); } /* * ========================================================================== * SPA Scanning * ========================================================================== */ int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd) { ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); if (dsl_scan_resilvering(spa->spa_dsl_pool)) return (SET_ERROR(EBUSY)); return (dsl_scrub_set_pause_resume(spa->spa_dsl_pool, cmd)); } int spa_scan_stop(spa_t *spa) { ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); if (dsl_scan_resilvering(spa->spa_dsl_pool)) return (SET_ERROR(EBUSY)); return (dsl_scan_cancel(spa->spa_dsl_pool)); } int spa_scan(spa_t *spa, pool_scan_func_t func) { ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE) return (SET_ERROR(ENOTSUP)); if (func == POOL_SCAN_RESILVER && !spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER)) return (SET_ERROR(ENOTSUP)); /* * If a resilver was requested, but there is no DTL on a * writeable leaf device, we have nothing to do. */ if (func == POOL_SCAN_RESILVER && !vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) { spa_async_request(spa, SPA_ASYNC_RESILVER_DONE); return (0); } if (func == POOL_SCAN_ERRORSCRUB && !spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) return (SET_ERROR(ENOTSUP)); return (dsl_scan(spa->spa_dsl_pool, func)); } /* * ========================================================================== * SPA async task processing * ========================================================================== */ static void spa_async_remove(spa_t *spa, vdev_t *vd) { if (vd->vdev_remove_wanted) { vd->vdev_remove_wanted = B_FALSE; vd->vdev_delayed_close = B_FALSE; vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE); /* * We want to clear the stats, but we don't want to do a full * vdev_clear() as that will cause us to throw away * degraded/faulted state as well as attempt to reopen the * device, all of which is a waste. */ vd->vdev_stat.vs_read_errors = 0; vd->vdev_stat.vs_write_errors = 0; vd->vdev_stat.vs_checksum_errors = 0; vdev_state_dirty(vd->vdev_top); /* Tell userspace that the vdev is gone. */ zfs_post_remove(spa, vd); } for (int c = 0; c < vd->vdev_children; c++) spa_async_remove(spa, vd->vdev_child[c]); } static void spa_async_probe(spa_t *spa, vdev_t *vd) { if (vd->vdev_probe_wanted) { vd->vdev_probe_wanted = B_FALSE; vdev_reopen(vd); /* vdev_open() does the actual probe */ } for (int c = 0; c < vd->vdev_children; c++) spa_async_probe(spa, vd->vdev_child[c]); } static void spa_async_autoexpand(spa_t *spa, vdev_t *vd) { if (!spa->spa_autoexpand) return; for (int c = 0; c < vd->vdev_children; c++) { vdev_t *cvd = vd->vdev_child[c]; spa_async_autoexpand(spa, cvd); } if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL) return; spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND); } static __attribute__((noreturn)) void spa_async_thread(void *arg) { spa_t *spa = (spa_t *)arg; dsl_pool_t *dp = spa->spa_dsl_pool; int tasks; ASSERT(spa->spa_sync_on); mutex_enter(&spa->spa_async_lock); tasks = spa->spa_async_tasks; spa->spa_async_tasks = 0; mutex_exit(&spa->spa_async_lock); /* * See if the config needs to be updated. */ if (tasks & SPA_ASYNC_CONFIG_UPDATE) { uint64_t old_space, new_space; mutex_enter(&spa_namespace_lock); old_space = metaslab_class_get_space(spa_normal_class(spa)); old_space += metaslab_class_get_space(spa_special_class(spa)); old_space += metaslab_class_get_space(spa_dedup_class(spa)); old_space += metaslab_class_get_space( spa_embedded_log_class(spa)); spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); new_space = metaslab_class_get_space(spa_normal_class(spa)); new_space += metaslab_class_get_space(spa_special_class(spa)); new_space += metaslab_class_get_space(spa_dedup_class(spa)); new_space += metaslab_class_get_space( spa_embedded_log_class(spa)); mutex_exit(&spa_namespace_lock); /* * If the pool grew as a result of the config update, * then log an internal history event. */ if (new_space != old_space) { spa_history_log_internal(spa, "vdev online", NULL, "pool '%s' size: %llu(+%llu)", spa_name(spa), (u_longlong_t)new_space, (u_longlong_t)(new_space - old_space)); } } /* * See if any devices need to be marked REMOVED. */ if (tasks & SPA_ASYNC_REMOVE) { spa_vdev_state_enter(spa, SCL_NONE); spa_async_remove(spa, spa->spa_root_vdev); for (int i = 0; i < spa->spa_l2cache.sav_count; i++) spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]); for (int i = 0; i < spa->spa_spares.sav_count; i++) spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]); (void) spa_vdev_state_exit(spa, NULL, 0); } if ((tasks & SPA_ASYNC_AUTOEXPAND) && !spa_suspended(spa)) { spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa_async_autoexpand(spa, spa->spa_root_vdev); spa_config_exit(spa, SCL_CONFIG, FTAG); } /* * See if any devices need to be probed. */ if (tasks & SPA_ASYNC_PROBE) { spa_vdev_state_enter(spa, SCL_NONE); spa_async_probe(spa, spa->spa_root_vdev); (void) spa_vdev_state_exit(spa, NULL, 0); } /* * If any devices are done replacing, detach them. */ if (tasks & SPA_ASYNC_RESILVER_DONE || tasks & SPA_ASYNC_REBUILD_DONE || tasks & SPA_ASYNC_DETACH_SPARE) { spa_vdev_resilver_done(spa); } /* * Kick off a resilver. */ if (tasks & SPA_ASYNC_RESILVER && !vdev_rebuild_active(spa->spa_root_vdev) && (!dsl_scan_resilvering(dp) || !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER))) dsl_scan_restart_resilver(dp, 0); if (tasks & SPA_ASYNC_INITIALIZE_RESTART) { mutex_enter(&spa_namespace_lock); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vdev_initialize_restart(spa->spa_root_vdev); spa_config_exit(spa, SCL_CONFIG, FTAG); mutex_exit(&spa_namespace_lock); } if (tasks & SPA_ASYNC_TRIM_RESTART) { mutex_enter(&spa_namespace_lock); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vdev_trim_restart(spa->spa_root_vdev); spa_config_exit(spa, SCL_CONFIG, FTAG); mutex_exit(&spa_namespace_lock); } if (tasks & SPA_ASYNC_AUTOTRIM_RESTART) { mutex_enter(&spa_namespace_lock); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vdev_autotrim_restart(spa); spa_config_exit(spa, SCL_CONFIG, FTAG); mutex_exit(&spa_namespace_lock); } /* * Kick off L2 cache whole device TRIM. */ if (tasks & SPA_ASYNC_L2CACHE_TRIM) { mutex_enter(&spa_namespace_lock); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vdev_trim_l2arc(spa); spa_config_exit(spa, SCL_CONFIG, FTAG); mutex_exit(&spa_namespace_lock); } /* * Kick off L2 cache rebuilding. */ if (tasks & SPA_ASYNC_L2CACHE_REBUILD) { mutex_enter(&spa_namespace_lock); spa_config_enter(spa, SCL_L2ARC, FTAG, RW_READER); l2arc_spa_rebuild_start(spa); spa_config_exit(spa, SCL_L2ARC, FTAG); mutex_exit(&spa_namespace_lock); } /* * Let the world know that we're done. */ mutex_enter(&spa->spa_async_lock); spa->spa_async_thread = NULL; cv_broadcast(&spa->spa_async_cv); mutex_exit(&spa->spa_async_lock); thread_exit(); } void spa_async_suspend(spa_t *spa) { mutex_enter(&spa->spa_async_lock); spa->spa_async_suspended++; while (spa->spa_async_thread != NULL) cv_wait(&spa->spa_async_cv, &spa->spa_async_lock); mutex_exit(&spa->spa_async_lock); spa_vdev_remove_suspend(spa); zthr_t *condense_thread = spa->spa_condense_zthr; if (condense_thread != NULL) zthr_cancel(condense_thread); zthr_t *raidz_expand_thread = spa->spa_raidz_expand_zthr; if (raidz_expand_thread != NULL) zthr_cancel(raidz_expand_thread); zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr; if (discard_thread != NULL) zthr_cancel(discard_thread); zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr; if (ll_delete_thread != NULL) zthr_cancel(ll_delete_thread); zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr; if (ll_condense_thread != NULL) zthr_cancel(ll_condense_thread); } void spa_async_resume(spa_t *spa) { mutex_enter(&spa->spa_async_lock); ASSERT(spa->spa_async_suspended != 0); spa->spa_async_suspended--; mutex_exit(&spa->spa_async_lock); spa_restart_removal(spa); zthr_t *condense_thread = spa->spa_condense_zthr; if (condense_thread != NULL) zthr_resume(condense_thread); zthr_t *raidz_expand_thread = spa->spa_raidz_expand_zthr; if (raidz_expand_thread != NULL) zthr_resume(raidz_expand_thread); zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr; if (discard_thread != NULL) zthr_resume(discard_thread); zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr; if (ll_delete_thread != NULL) zthr_resume(ll_delete_thread); zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr; if (ll_condense_thread != NULL) zthr_resume(ll_condense_thread); } static boolean_t spa_async_tasks_pending(spa_t *spa) { uint_t non_config_tasks; uint_t config_task; boolean_t config_task_suspended; non_config_tasks = spa->spa_async_tasks & ~SPA_ASYNC_CONFIG_UPDATE; config_task = spa->spa_async_tasks & SPA_ASYNC_CONFIG_UPDATE; if (spa->spa_ccw_fail_time == 0) { config_task_suspended = B_FALSE; } else { config_task_suspended = (gethrtime() - spa->spa_ccw_fail_time) < ((hrtime_t)zfs_ccw_retry_interval * NANOSEC); } return (non_config_tasks || (config_task && !config_task_suspended)); } static void spa_async_dispatch(spa_t *spa) { mutex_enter(&spa->spa_async_lock); if (spa_async_tasks_pending(spa) && !spa->spa_async_suspended && spa->spa_async_thread == NULL) spa->spa_async_thread = thread_create(NULL, 0, spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri); mutex_exit(&spa->spa_async_lock); } void spa_async_request(spa_t *spa, int task) { zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task); mutex_enter(&spa->spa_async_lock); spa->spa_async_tasks |= task; mutex_exit(&spa->spa_async_lock); } int spa_async_tasks(spa_t *spa) { return (spa->spa_async_tasks); } /* * ========================================================================== * SPA syncing routines * ========================================================================== */ static int bpobj_enqueue_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx) { bpobj_t *bpo = arg; bpobj_enqueue(bpo, bp, bp_freed, tx); return (0); } int bpobj_enqueue_alloc_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { return (bpobj_enqueue_cb(arg, bp, B_FALSE, tx)); } int bpobj_enqueue_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { return (bpobj_enqueue_cb(arg, bp, B_TRUE, tx)); } static int spa_free_sync_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { zio_t *pio = arg; zio_nowait(zio_free_sync(pio, pio->io_spa, dmu_tx_get_txg(tx), bp, pio->io_flags)); return (0); } static int bpobj_spa_free_sync_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx) { ASSERT(!bp_freed); return (spa_free_sync_cb(arg, bp, tx)); } /* * Note: this simple function is not inlined to make it easier to dtrace the * amount of time spent syncing frees. */ static void spa_sync_frees(spa_t *spa, bplist_t *bpl, dmu_tx_t *tx) { zio_t *zio = zio_root(spa, NULL, NULL, 0); bplist_iterate(bpl, spa_free_sync_cb, zio, tx); VERIFY(zio_wait(zio) == 0); } /* * Note: this simple function is not inlined to make it easier to dtrace the * amount of time spent syncing deferred frees. */ static void spa_sync_deferred_frees(spa_t *spa, dmu_tx_t *tx) { if (spa_sync_pass(spa) != 1) return; /* * Note: * If the log space map feature is active, we stop deferring * frees to the next TXG and therefore running this function * would be considered a no-op as spa_deferred_bpobj should * not have any entries. * * That said we run this function anyway (instead of returning * immediately) for the edge-case scenario where we just * activated the log space map feature in this TXG but we have * deferred frees from the previous TXG. */ zio_t *zio = zio_root(spa, NULL, NULL, 0); VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj, bpobj_spa_free_sync_cb, zio, tx), ==, 0); VERIFY0(zio_wait(zio)); } static void spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx) { char *packed = NULL; size_t bufsize; size_t nvsize = 0; dmu_buf_t *db; VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0); /* * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration * information. This avoids the dmu_buf_will_dirty() path and * saves us a pre-read to get data we don't actually care about. */ bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE); packed = vmem_alloc(bufsize, KM_SLEEP); VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR, KM_SLEEP) == 0); memset(packed + nvsize, 0, bufsize - nvsize); dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx); vmem_free(packed, bufsize); VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); dmu_buf_will_dirty(db, tx); *(uint64_t *)db->db_data = nvsize; dmu_buf_rele(db, FTAG); } static void spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx, const char *config, const char *entry) { nvlist_t *nvroot; nvlist_t **list; int i; if (!sav->sav_sync) return; /* * Update the MOS nvlist describing the list of available devices. * spa_validate_aux() will have already made sure this nvlist is * valid and the vdevs are labeled appropriately. */ if (sav->sav_object == 0) { sav->sav_object = dmu_object_alloc(spa->spa_meta_objset, DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx); VERIFY(zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1, &sav->sav_object, tx) == 0); } nvroot = fnvlist_alloc(); if (sav->sav_count == 0) { fnvlist_add_nvlist_array(nvroot, config, (const nvlist_t * const *)NULL, 0); } else { list = kmem_alloc(sav->sav_count*sizeof (void *), KM_SLEEP); for (i = 0; i < sav->sav_count; i++) list[i] = vdev_config_generate(spa, sav->sav_vdevs[i], B_FALSE, VDEV_CONFIG_L2CACHE); fnvlist_add_nvlist_array(nvroot, config, (const nvlist_t * const *)list, sav->sav_count); for (i = 0; i < sav->sav_count; i++) nvlist_free(list[i]); kmem_free(list, sav->sav_count * sizeof (void *)); } spa_sync_nvlist(spa, sav->sav_object, nvroot, tx); nvlist_free(nvroot); sav->sav_sync = B_FALSE; } /* * Rebuild spa's all-vdev ZAP from the vdev ZAPs indicated in each vdev_t. * The all-vdev ZAP must be empty. */ static void spa_avz_build(vdev_t *vd, uint64_t avz, dmu_tx_t *tx) { spa_t *spa = vd->vdev_spa; if (vd->vdev_root_zap != 0 && spa_feature_is_active(spa, SPA_FEATURE_AVZ_V2)) { VERIFY0(zap_add_int(spa->spa_meta_objset, avz, vd->vdev_root_zap, tx)); } if (vd->vdev_top_zap != 0) { VERIFY0(zap_add_int(spa->spa_meta_objset, avz, vd->vdev_top_zap, tx)); } if (vd->vdev_leaf_zap != 0) { VERIFY0(zap_add_int(spa->spa_meta_objset, avz, vd->vdev_leaf_zap, tx)); } for (uint64_t i = 0; i < vd->vdev_children; i++) { spa_avz_build(vd->vdev_child[i], avz, tx); } } static void spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) { nvlist_t *config; /* * If the pool is being imported from a pre-per-vdev-ZAP version of ZFS, * its config may not be dirty but we still need to build per-vdev ZAPs. * Similarly, if the pool is being assembled (e.g. after a split), we * need to rebuild the AVZ although the config may not be dirty. */ if (list_is_empty(&spa->spa_config_dirty_list) && spa->spa_avz_action == AVZ_ACTION_NONE) return; spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE || spa->spa_avz_action == AVZ_ACTION_INITIALIZE || spa->spa_all_vdev_zaps != 0); if (spa->spa_avz_action == AVZ_ACTION_REBUILD) { /* Make and build the new AVZ */ uint64_t new_avz = zap_create(spa->spa_meta_objset, DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx); spa_avz_build(spa->spa_root_vdev, new_avz, tx); /* Diff old AVZ with new one */ zap_cursor_t zc; zap_attribute_t za; for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_all_vdev_zaps); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { uint64_t vdzap = za.za_first_integer; if (zap_lookup_int(spa->spa_meta_objset, new_avz, vdzap) == ENOENT) { /* * ZAP is listed in old AVZ but not in new one; * destroy it */ VERIFY0(zap_destroy(spa->spa_meta_objset, vdzap, tx)); } } zap_cursor_fini(&zc); /* Destroy the old AVZ */ VERIFY0(zap_destroy(spa->spa_meta_objset, spa->spa_all_vdev_zaps, tx)); /* Replace the old AVZ in the dir obj with the new one */ VERIFY0(zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, sizeof (new_avz), 1, &new_avz, tx)); spa->spa_all_vdev_zaps = new_avz; } else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) { zap_cursor_t zc; zap_attribute_t za; /* Walk through the AVZ and destroy all listed ZAPs */ for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_all_vdev_zaps); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { uint64_t zap = za.za_first_integer; VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx)); } zap_cursor_fini(&zc); /* Destroy and unlink the AVZ itself */ VERIFY0(zap_destroy(spa->spa_meta_objset, spa->spa_all_vdev_zaps, tx)); VERIFY0(zap_remove(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx)); spa->spa_all_vdev_zaps = 0; } if (spa->spa_all_vdev_zaps == 0) { spa->spa_all_vdev_zaps = zap_create_link(spa->spa_meta_objset, DMU_OTN_ZAP_METADATA, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx); } spa->spa_avz_action = AVZ_ACTION_NONE; /* Create ZAPs for vdevs that don't have them. */ vdev_construct_zaps(spa->spa_root_vdev, tx); config = spa_config_generate(spa, spa->spa_root_vdev, dmu_tx_get_txg(tx), B_FALSE); /* * If we're upgrading the spa version then make sure that * the config object gets updated with the correct version. */ if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version) fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa->spa_uberblock.ub_version); spa_config_exit(spa, SCL_STATE, FTAG); nvlist_free(spa->spa_config_syncing); spa->spa_config_syncing = config; spa_sync_nvlist(spa, spa->spa_config_object, config, tx); } static void spa_sync_version(void *arg, dmu_tx_t *tx) { uint64_t *versionp = arg; uint64_t version = *versionp; spa_t *spa = dmu_tx_pool(tx)->dp_spa; /* * Setting the version is special cased when first creating the pool. */ ASSERT(tx->tx_txg != TXG_INITIAL); ASSERT(SPA_VERSION_IS_SUPPORTED(version)); ASSERT(version >= spa_version(spa)); spa->spa_uberblock.ub_version = version; vdev_config_dirty(spa->spa_root_vdev); spa_history_log_internal(spa, "set", tx, "version=%lld", (longlong_t)version); } /* * Set zpool properties. */ static void spa_sync_props(void *arg, dmu_tx_t *tx) { nvlist_t *nvp = arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; objset_t *mos = spa->spa_meta_objset; nvpair_t *elem = NULL; mutex_enter(&spa->spa_props_lock); while ((elem = nvlist_next_nvpair(nvp, elem))) { uint64_t intval; const char *strval, *fname; zpool_prop_t prop; const char *propname; const char *elemname = nvpair_name(elem); zprop_type_t proptype; spa_feature_t fid; switch (prop = zpool_name_to_prop(elemname)) { case ZPOOL_PROP_VERSION: intval = fnvpair_value_uint64(elem); /* * The version is synced separately before other * properties and should be correct by now. */ ASSERT3U(spa_version(spa), >=, intval); break; case ZPOOL_PROP_ALTROOT: /* * 'altroot' is a non-persistent property. It should * have been set temporarily at creation or import time. */ ASSERT(spa->spa_root != NULL); break; case ZPOOL_PROP_READONLY: case ZPOOL_PROP_CACHEFILE: /* * 'readonly' and 'cachefile' are also non-persistent * properties. */ break; case ZPOOL_PROP_COMMENT: strval = fnvpair_value_string(elem); if (spa->spa_comment != NULL) spa_strfree(spa->spa_comment); spa->spa_comment = spa_strdup(strval); /* * We need to dirty the configuration on all the vdevs * so that their labels get updated. We also need to * update the cache file to keep it in sync with the * MOS version. It's unnecessary to do this for pool * creation since the vdev's configuration has already * been dirtied. */ if (tx->tx_txg != TXG_INITIAL) { vdev_config_dirty(spa->spa_root_vdev); spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); } spa_history_log_internal(spa, "set", tx, "%s=%s", elemname, strval); break; case ZPOOL_PROP_COMPATIBILITY: strval = fnvpair_value_string(elem); if (spa->spa_compatibility != NULL) spa_strfree(spa->spa_compatibility); spa->spa_compatibility = spa_strdup(strval); /* * Dirty the configuration on vdevs as above. */ if (tx->tx_txg != TXG_INITIAL) { vdev_config_dirty(spa->spa_root_vdev); spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); } spa_history_log_internal(spa, "set", tx, "%s=%s", nvpair_name(elem), strval); break; case ZPOOL_PROP_INVAL: if (zpool_prop_feature(elemname)) { fname = strchr(elemname, '@') + 1; VERIFY0(zfeature_lookup_name(fname, &fid)); spa_feature_enable(spa, fid, tx); spa_history_log_internal(spa, "set", tx, "%s=enabled", elemname); break; } else if (!zfs_prop_user(elemname)) { ASSERT(zpool_prop_feature(elemname)); break; } zfs_fallthrough; default: /* * Set pool property values in the poolprops mos object. */ if (spa->spa_pool_props_object == 0) { spa->spa_pool_props_object = zap_create_link(mos, DMU_OT_POOL_PROPS, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS, tx); } /* normalize the property name */ if (prop == ZPOOL_PROP_INVAL) { propname = elemname; proptype = PROP_TYPE_STRING; } else { propname = zpool_prop_to_name(prop); proptype = zpool_prop_get_type(prop); } if (nvpair_type(elem) == DATA_TYPE_STRING) { ASSERT(proptype == PROP_TYPE_STRING); strval = fnvpair_value_string(elem); VERIFY0(zap_update(mos, spa->spa_pool_props_object, propname, 1, strlen(strval) + 1, strval, tx)); spa_history_log_internal(spa, "set", tx, "%s=%s", elemname, strval); } else if (nvpair_type(elem) == DATA_TYPE_UINT64) { intval = fnvpair_value_uint64(elem); if (proptype == PROP_TYPE_INDEX) { const char *unused; VERIFY0(zpool_prop_index_to_string( prop, intval, &unused)); } VERIFY0(zap_update(mos, spa->spa_pool_props_object, propname, 8, 1, &intval, tx)); spa_history_log_internal(spa, "set", tx, "%s=%lld", elemname, (longlong_t)intval); switch (prop) { case ZPOOL_PROP_DELEGATION: spa->spa_delegation = intval; break; case ZPOOL_PROP_BOOTFS: spa->spa_bootfs = intval; break; case ZPOOL_PROP_FAILUREMODE: spa->spa_failmode = intval; break; case ZPOOL_PROP_AUTOTRIM: spa->spa_autotrim = intval; spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART); break; case ZPOOL_PROP_AUTOEXPAND: spa->spa_autoexpand = intval; if (tx->tx_txg != TXG_INITIAL) spa_async_request(spa, SPA_ASYNC_AUTOEXPAND); break; case ZPOOL_PROP_MULTIHOST: spa->spa_multihost = intval; break; default: break; } } else { ASSERT(0); /* not allowed */ } } } mutex_exit(&spa->spa_props_lock); } /* * Perform one-time upgrade on-disk changes. spa_version() does not * reflect the new version this txg, so there must be no changes this * txg to anything that the upgrade code depends on after it executes. * Therefore this must be called after dsl_pool_sync() does the sync * tasks. */ static void spa_sync_upgrades(spa_t *spa, dmu_tx_t *tx) { if (spa_sync_pass(spa) != 1) return; dsl_pool_t *dp = spa->spa_dsl_pool; rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG); if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN && spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) { dsl_pool_create_origin(dp, tx); /* Keeping the origin open increases spa_minref */ spa->spa_minref += 3; } if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES && spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) { dsl_pool_upgrade_clones(dp, tx); } if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES && spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) { dsl_pool_upgrade_dir_clones(dp, tx); /* Keeping the freedir open increases spa_minref */ spa->spa_minref += 3; } if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES && spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) { spa_feature_create_zap_objects(spa, tx); } /* * LZ4_COMPRESS feature's behaviour was changed to activate_on_enable * when possibility to use lz4 compression for metadata was added * Old pools that have this feature enabled must be upgraded to have * this feature active */ if (spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) { boolean_t lz4_en = spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS); boolean_t lz4_ac = spa_feature_is_active(spa, SPA_FEATURE_LZ4_COMPRESS); if (lz4_en && !lz4_ac) spa_feature_incr(spa, SPA_FEATURE_LZ4_COMPRESS, tx); } /* * If we haven't written the salt, do so now. Note that the * feature may not be activated yet, but that's fine since * the presence of this ZAP entry is backwards compatible. */ if (zap_contains(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT) == ENOENT) { VERIFY0(zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1, sizeof (spa->spa_cksum_salt.zcs_bytes), spa->spa_cksum_salt.zcs_bytes, tx)); } rrw_exit(&dp->dp_config_rwlock, FTAG); } static void vdev_indirect_state_sync_verify(vdev_t *vd) { vdev_indirect_mapping_t *vim __maybe_unused = vd->vdev_indirect_mapping; vdev_indirect_births_t *vib __maybe_unused = vd->vdev_indirect_births; if (vd->vdev_ops == &vdev_indirect_ops) { ASSERT(vim != NULL); ASSERT(vib != NULL); } uint64_t obsolete_sm_object = 0; ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_object)); if (obsolete_sm_object != 0) { ASSERT(vd->vdev_obsolete_sm != NULL); ASSERT(vd->vdev_removing || vd->vdev_ops == &vdev_indirect_ops); ASSERT(vdev_indirect_mapping_num_entries(vim) > 0); ASSERT(vdev_indirect_mapping_bytes_mapped(vim) > 0); ASSERT3U(obsolete_sm_object, ==, space_map_object(vd->vdev_obsolete_sm)); ASSERT3U(vdev_indirect_mapping_bytes_mapped(vim), >=, space_map_allocated(vd->vdev_obsolete_sm)); } ASSERT(vd->vdev_obsolete_segments != NULL); /* * Since frees / remaps to an indirect vdev can only * happen in syncing context, the obsolete segments * tree must be empty when we start syncing. */ ASSERT0(range_tree_space(vd->vdev_obsolete_segments)); } /* * Set the top-level vdev's max queue depth. Evaluate each top-level's * async write queue depth in case it changed. The max queue depth will * not change in the middle of syncing out this txg. */ static void spa_sync_adjust_vdev_max_queue_depth(spa_t *spa) { ASSERT(spa_writeable(spa)); vdev_t *rvd = spa->spa_root_vdev; uint32_t max_queue_depth = zfs_vdev_async_write_max_active * zfs_vdev_queue_depth_pct / 100; metaslab_class_t *normal = spa_normal_class(spa); metaslab_class_t *special = spa_special_class(spa); metaslab_class_t *dedup = spa_dedup_class(spa); uint64_t slots_per_allocator = 0; for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; metaslab_group_t *mg = tvd->vdev_mg; if (mg == NULL || !metaslab_group_initialized(mg)) continue; metaslab_class_t *mc = mg->mg_class; if (mc != normal && mc != special && mc != dedup) continue; /* * It is safe to do a lock-free check here because only async * allocations look at mg_max_alloc_queue_depth, and async * allocations all happen from spa_sync(). */ for (int i = 0; i < mg->mg_allocators; i++) { ASSERT0(zfs_refcount_count( &(mg->mg_allocator[i].mga_alloc_queue_depth))); } mg->mg_max_alloc_queue_depth = max_queue_depth; for (int i = 0; i < mg->mg_allocators; i++) { mg->mg_allocator[i].mga_cur_max_alloc_queue_depth = zfs_vdev_def_queue_depth; } slots_per_allocator += zfs_vdev_def_queue_depth; } for (int i = 0; i < spa->spa_alloc_count; i++) { ASSERT0(zfs_refcount_count(&normal->mc_allocator[i]. mca_alloc_slots)); ASSERT0(zfs_refcount_count(&special->mc_allocator[i]. mca_alloc_slots)); ASSERT0(zfs_refcount_count(&dedup->mc_allocator[i]. mca_alloc_slots)); normal->mc_allocator[i].mca_alloc_max_slots = slots_per_allocator; special->mc_allocator[i].mca_alloc_max_slots = slots_per_allocator; dedup->mc_allocator[i].mca_alloc_max_slots = slots_per_allocator; } normal->mc_alloc_throttle_enabled = zio_dva_throttle_enabled; special->mc_alloc_throttle_enabled = zio_dva_throttle_enabled; dedup->mc_alloc_throttle_enabled = zio_dva_throttle_enabled; } static void spa_sync_condense_indirect(spa_t *spa, dmu_tx_t *tx) { ASSERT(spa_writeable(spa)); vdev_t *rvd = spa->spa_root_vdev; for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *vd = rvd->vdev_child[c]; vdev_indirect_state_sync_verify(vd); if (vdev_indirect_should_condense(vd)) { spa_condense_indirect_start_sync(vd, tx); break; } } } static void spa_sync_iterate_to_convergence(spa_t *spa, dmu_tx_t *tx) { objset_t *mos = spa->spa_meta_objset; dsl_pool_t *dp = spa->spa_dsl_pool; uint64_t txg = tx->tx_txg; bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK]; do { int pass = ++spa->spa_sync_pass; spa_sync_config_object(spa, tx); spa_sync_aux_dev(spa, &spa->spa_spares, tx, ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES); spa_sync_aux_dev(spa, &spa->spa_l2cache, tx, ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE); spa_errlog_sync(spa, txg); dsl_pool_sync(dp, txg); if (pass < zfs_sync_pass_deferred_free || spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) { /* * If the log space map feature is active we don't * care about deferred frees and the deferred bpobj * as the log space map should effectively have the * same results (i.e. appending only to one object). */ spa_sync_frees(spa, free_bpl, tx); } else { /* * We can not defer frees in pass 1, because * we sync the deferred frees later in pass 1. */ ASSERT3U(pass, >, 1); bplist_iterate(free_bpl, bpobj_enqueue_alloc_cb, &spa->spa_deferred_bpobj, tx); } brt_sync(spa, txg); ddt_sync(spa, txg); dsl_scan_sync(dp, tx); dsl_errorscrub_sync(dp, tx); svr_sync(spa, tx); spa_sync_upgrades(spa, tx); spa_flush_metaslabs(spa, tx); vdev_t *vd = NULL; while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)) != NULL) vdev_sync(vd, txg); if (pass == 1) { /* * dsl_pool_sync() -> dp_sync_tasks may have dirtied * the config. If that happens, this txg should not * be a no-op. So we must sync the config to the MOS * before checking for no-op. * * Note that when the config is dirty, it will * be written to the MOS (i.e. the MOS will be * dirtied) every time we call spa_sync_config_object() * in this txg. Therefore we can't call this after * dsl_pool_sync() every pass, because it would * prevent us from converging, since we'd dirty * the MOS every pass. * * Sync tasks can only be processed in pass 1, so * there's no need to do this in later passes. */ spa_sync_config_object(spa, tx); } /* * Note: We need to check if the MOS is dirty because we could * have marked the MOS dirty without updating the uberblock * (e.g. if we have sync tasks but no dirty user data). We need * to check the uberblock's rootbp because it is updated if we * have synced out dirty data (though in this case the MOS will * most likely also be dirty due to second order effects, we * don't want to rely on that here). */ if (pass == 1 && BP_GET_LOGICAL_BIRTH(&spa->spa_uberblock.ub_rootbp) < txg && !dmu_objset_is_dirty(mos, txg)) { /* * Nothing changed on the first pass, therefore this * TXG is a no-op. Avoid syncing deferred frees, so * that we can keep this TXG as a no-op. */ ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg)); ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg)); ASSERT(txg_list_empty(&dp->dp_sync_tasks, txg)); ASSERT(txg_list_empty(&dp->dp_early_sync_tasks, txg)); break; } spa_sync_deferred_frees(spa, tx); } while (dmu_objset_is_dirty(mos, txg)); } /* * Rewrite the vdev configuration (which includes the uberblock) to * commit the transaction group. * * If there are no dirty vdevs, we sync the uberblock to a few random * top-level vdevs that are known to be visible in the config cache * (see spa_vdev_add() for a complete description). If there *are* dirty * vdevs, sync the uberblock to all vdevs. */ static void spa_sync_rewrite_vdev_config(spa_t *spa, dmu_tx_t *tx) { vdev_t *rvd = spa->spa_root_vdev; uint64_t txg = tx->tx_txg; for (;;) { int error = 0; /* * We hold SCL_STATE to prevent vdev open/close/etc. * while we're attempting to write the vdev labels. */ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); if (list_is_empty(&spa->spa_config_dirty_list)) { vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL }; int svdcount = 0; int children = rvd->vdev_children; int c0 = random_in_range(children); for (int c = 0; c < children; c++) { vdev_t *vd = rvd->vdev_child[(c0 + c) % children]; /* Stop when revisiting the first vdev */ if (c > 0 && svd[0] == vd) break; if (vd->vdev_ms_array == 0 || vd->vdev_islog || !vdev_is_concrete(vd)) continue; svd[svdcount++] = vd; if (svdcount == SPA_SYNC_MIN_VDEVS) break; } error = vdev_config_sync(svd, svdcount, txg); } else { error = vdev_config_sync(rvd->vdev_child, rvd->vdev_children, txg); } if (error == 0) spa->spa_last_synced_guid = rvd->vdev_guid; spa_config_exit(spa, SCL_STATE, FTAG); if (error == 0) break; zio_suspend(spa, NULL, ZIO_SUSPEND_IOERR); zio_resume_wait(spa); } } /* * Sync the specified transaction group. New blocks may be dirtied as * part of the process, so we iterate until it converges. */ void spa_sync(spa_t *spa, uint64_t txg) { vdev_t *vd = NULL; VERIFY(spa_writeable(spa)); /* * Wait for i/os issued in open context that need to complete * before this txg syncs. */ (void) zio_wait(spa->spa_txg_zio[txg & TXG_MASK]); spa->spa_txg_zio[txg & TXG_MASK] = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); /* * Now that there can be no more cloning in this transaction group, * but we are still before issuing frees, we can process pending BRT * updates. */ brt_pending_apply(spa, txg); /* * Lock out configuration changes. */ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa->spa_syncing_txg = txg; spa->spa_sync_pass = 0; for (int i = 0; i < spa->spa_alloc_count; i++) { mutex_enter(&spa->spa_allocs[i].spaa_lock); VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree)); mutex_exit(&spa->spa_allocs[i].spaa_lock); } /* * If there are any pending vdev state changes, convert them * into config changes that go out with this transaction group. */ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) { /* Avoid holding the write lock unless actually necessary */ if (vd->vdev_aux == NULL) { vdev_state_clean(vd); vdev_config_dirty(vd); continue; } /* * We need the write lock here because, for aux vdevs, * calling vdev_config_dirty() modifies sav_config. * This is ugly and will become unnecessary when we * eliminate the aux vdev wart by integrating all vdevs * into the root vdev tree. */ spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_WRITER); while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) { vdev_state_clean(vd); vdev_config_dirty(vd); } spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); } spa_config_exit(spa, SCL_STATE, FTAG); dsl_pool_t *dp = spa->spa_dsl_pool; dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg); spa->spa_sync_starttime = gethrtime(); taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid); spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq, spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() + NSEC_TO_TICK(spa->spa_deadman_synctime)); /* * If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg, * set spa_deflate if we have no raid-z vdevs. */ if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE && spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) { vdev_t *rvd = spa->spa_root_vdev; int i; for (i = 0; i < rvd->vdev_children; i++) { vd = rvd->vdev_child[i]; if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE) break; } if (i == rvd->vdev_children) { spa->spa_deflate = TRUE; VERIFY0(zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, sizeof (uint64_t), 1, &spa->spa_deflate, tx)); } } spa_sync_adjust_vdev_max_queue_depth(spa); spa_sync_condense_indirect(spa, tx); spa_sync_iterate_to_convergence(spa, tx); #ifdef ZFS_DEBUG if (!list_is_empty(&spa->spa_config_dirty_list)) { /* * Make sure that the number of ZAPs for all the vdevs matches * the number of ZAPs in the per-vdev ZAP list. This only gets * called if the config is dirty; otherwise there may be * outstanding AVZ operations that weren't completed in * spa_sync_config_object. */ uint64_t all_vdev_zap_entry_count; ASSERT0(zap_count(spa->spa_meta_objset, spa->spa_all_vdev_zaps, &all_vdev_zap_entry_count)); ASSERT3U(vdev_count_verify_zaps(spa->spa_root_vdev), ==, all_vdev_zap_entry_count); } #endif if (spa->spa_vdev_removal != NULL) { ASSERT0(spa->spa_vdev_removal->svr_bytes_done[txg & TXG_MASK]); } spa_sync_rewrite_vdev_config(spa, tx); dmu_tx_commit(tx); taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid); spa->spa_deadman_tqid = 0; /* * Clear the dirty config list. */ while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL) vdev_config_clean(vd); /* * Now that the new config has synced transactionally, * let it become visible to the config cache. */ if (spa->spa_config_syncing != NULL) { spa_config_set(spa, spa->spa_config_syncing); spa->spa_config_txg = txg; spa->spa_config_syncing = NULL; } dsl_pool_sync_done(dp, txg); for (int i = 0; i < spa->spa_alloc_count; i++) { mutex_enter(&spa->spa_allocs[i].spaa_lock); VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree)); mutex_exit(&spa->spa_allocs[i].spaa_lock); } /* * Update usable space statistics. */ while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg))) != NULL) vdev_sync_done(vd, txg); metaslab_class_evict_old(spa->spa_normal_class, txg); metaslab_class_evict_old(spa->spa_log_class, txg); spa_sync_close_syncing_log_sm(spa); spa_update_dspace(spa); if (spa_get_autotrim(spa) == SPA_AUTOTRIM_ON) vdev_autotrim_kick(spa); /* * It had better be the case that we didn't dirty anything * since vdev_config_sync(). */ ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg)); ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg)); ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg)); while (zfs_pause_spa_sync) delay(1); spa->spa_sync_pass = 0; /* * Update the last synced uberblock here. We want to do this at * the end of spa_sync() so that consumers of spa_last_synced_txg() * will be guaranteed that all the processing associated with * that txg has been completed. */ spa->spa_ubsync = spa->spa_uberblock; spa_config_exit(spa, SCL_CONFIG, FTAG); spa_handle_ignored_writes(spa); /* * If any async tasks have been requested, kick them off. */ spa_async_dispatch(spa); } /* * Sync all pools. We don't want to hold the namespace lock across these * operations, so we take a reference on the spa_t and drop the lock during the * sync. */ void spa_sync_allpools(void) { spa_t *spa = NULL; mutex_enter(&spa_namespace_lock); while ((spa = spa_next(spa)) != NULL) { if (spa_state(spa) != POOL_STATE_ACTIVE || !spa_writeable(spa) || spa_suspended(spa)) continue; spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); txg_wait_synced(spa_get_dsl(spa), 0); mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); } mutex_exit(&spa_namespace_lock); } taskq_t * spa_sync_tq_create(spa_t *spa, const char *name) { kthread_t **kthreads; ASSERT(spa->spa_sync_tq == NULL); ASSERT3S(spa->spa_alloc_count, <=, boot_ncpus); /* * - do not allow more allocators than cpus. * - there may be more cpus than allocators. * - do not allow more sync taskq threads than allocators or cpus. */ int nthreads = spa->spa_alloc_count; spa->spa_syncthreads = kmem_zalloc(sizeof (spa_syncthread_info_t) * nthreads, KM_SLEEP); spa->spa_sync_tq = taskq_create_synced(name, nthreads, minclsyspri, nthreads, INT_MAX, TASKQ_PREPOPULATE, &kthreads); VERIFY(spa->spa_sync_tq != NULL); VERIFY(kthreads != NULL); spa_taskqs_t *tqs = &spa->spa_zio_taskq[ZIO_TYPE_WRITE][ZIO_TASKQ_ISSUE]; spa_syncthread_info_t *ti = spa->spa_syncthreads; for (int i = 0, w = 0; i < nthreads; i++, w++, ti++) { ti->sti_thread = kthreads[i]; if (w == tqs->stqs_count) { w = 0; } ti->sti_wr_iss_tq = tqs->stqs_taskq[w]; } kmem_free(kthreads, sizeof (*kthreads) * nthreads); return (spa->spa_sync_tq); } void spa_sync_tq_destroy(spa_t *spa) { ASSERT(spa->spa_sync_tq != NULL); taskq_wait(spa->spa_sync_tq); taskq_destroy(spa->spa_sync_tq); kmem_free(spa->spa_syncthreads, sizeof (spa_syncthread_info_t) * spa->spa_alloc_count); spa->spa_sync_tq = NULL; } void spa_select_allocator(zio_t *zio) { zbookmark_phys_t *bm = &zio->io_bookmark; spa_t *spa = zio->io_spa; ASSERT(zio->io_type == ZIO_TYPE_WRITE); /* * A gang block (for example) may have inherited its parent's * allocator, in which case there is nothing further to do here. */ if (ZIO_HAS_ALLOCATOR(zio)) return; ASSERT(spa != NULL); ASSERT(bm != NULL); /* * First try to use an allocator assigned to the syncthread, and set * the corresponding write issue taskq for the allocator. * Note, we must have an open pool to do this. */ if (spa->spa_sync_tq != NULL) { spa_syncthread_info_t *ti = spa->spa_syncthreads; for (int i = 0; i < spa->spa_alloc_count; i++, ti++) { if (ti->sti_thread == curthread) { zio->io_allocator = i; zio->io_wr_iss_tq = ti->sti_wr_iss_tq; return; } } } /* * We want to try to use as many allocators as possible to help improve * performance, but we also want logically adjacent IOs to be physically * adjacent to improve sequential read performance. We chunk each object * into 2^20 block regions, and then hash based on the objset, object, * level, and region to accomplish both of these goals. */ uint64_t hv = cityhash4(bm->zb_objset, bm->zb_object, bm->zb_level, bm->zb_blkid >> 20); zio->io_allocator = (uint_t)hv % spa->spa_alloc_count; zio->io_wr_iss_tq = NULL; } /* * ========================================================================== * Miscellaneous routines * ========================================================================== */ /* * Remove all pools in the system. */ void spa_evict_all(void) { spa_t *spa; /* * Remove all cached state. All pools should be closed now, * so every spa in the AVL tree should be unreferenced. */ mutex_enter(&spa_namespace_lock); while ((spa = spa_next(NULL)) != NULL) { /* * Stop async tasks. The async thread may need to detach * a device that's been replaced, which requires grabbing * spa_namespace_lock, so we must drop it here. */ spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); spa_async_suspend(spa); mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); if (spa->spa_state != POOL_STATE_UNINITIALIZED) { spa_unload(spa); spa_deactivate(spa); } spa_remove(spa); } mutex_exit(&spa_namespace_lock); } vdev_t * spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t aux) { vdev_t *vd; int i; if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL) return (vd); if (aux) { for (i = 0; i < spa->spa_l2cache.sav_count; i++) { vd = spa->spa_l2cache.sav_vdevs[i]; if (vd->vdev_guid == guid) return (vd); } for (i = 0; i < spa->spa_spares.sav_count; i++) { vd = spa->spa_spares.sav_vdevs[i]; if (vd->vdev_guid == guid) return (vd); } } return (NULL); } void spa_upgrade(spa_t *spa, uint64_t version) { ASSERT(spa_writeable(spa)); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * This should only be called for a non-faulted pool, and since a * future version would result in an unopenable pool, this shouldn't be * possible. */ ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version)); ASSERT3U(version, >=, spa->spa_uberblock.ub_version); spa->spa_uberblock.ub_version = version; vdev_config_dirty(spa->spa_root_vdev); spa_config_exit(spa, SCL_ALL, FTAG); txg_wait_synced(spa_get_dsl(spa), 0); } static boolean_t spa_has_aux_vdev(spa_t *spa, uint64_t guid, spa_aux_vdev_t *sav) { (void) spa; int i; uint64_t vdev_guid; for (i = 0; i < sav->sav_count; i++) if (sav->sav_vdevs[i]->vdev_guid == guid) return (B_TRUE); for (i = 0; i < sav->sav_npending; i++) { if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID, &vdev_guid) == 0 && vdev_guid == guid) return (B_TRUE); } return (B_FALSE); } boolean_t spa_has_l2cache(spa_t *spa, uint64_t guid) { return (spa_has_aux_vdev(spa, guid, &spa->spa_l2cache)); } boolean_t spa_has_spare(spa_t *spa, uint64_t guid) { return (spa_has_aux_vdev(spa, guid, &spa->spa_spares)); } /* * Check if a pool has an active shared spare device. * Note: reference count of an active spare is 2, as a spare and as a replace */ static boolean_t spa_has_active_shared_spare(spa_t *spa) { int i, refcnt; uint64_t pool; spa_aux_vdev_t *sav = &spa->spa_spares; for (i = 0; i < sav->sav_count; i++) { if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool, &refcnt) && pool != 0ULL && pool == spa_guid(spa) && refcnt > 2) return (B_TRUE); } return (B_FALSE); } uint64_t spa_total_metaslabs(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; uint64_t m = 0; for (uint64_t c = 0; c < rvd->vdev_children; c++) { vdev_t *vd = rvd->vdev_child[c]; if (!vdev_is_concrete(vd)) continue; m += vd->vdev_ms_count; } return (m); } /* * Notify any waiting threads that some activity has switched from being in- * progress to not-in-progress so that the thread can wake up and determine * whether it is finished waiting. */ void spa_notify_waiters(spa_t *spa) { /* * Acquiring spa_activities_lock here prevents the cv_broadcast from * happening between the waiting thread's check and cv_wait. */ mutex_enter(&spa->spa_activities_lock); cv_broadcast(&spa->spa_activities_cv); mutex_exit(&spa->spa_activities_lock); } /* * Notify any waiting threads that the pool is exporting, and then block until * they are finished using the spa_t. */ void spa_wake_waiters(spa_t *spa) { mutex_enter(&spa->spa_activities_lock); spa->spa_waiters_cancel = B_TRUE; cv_broadcast(&spa->spa_activities_cv); while (spa->spa_waiters != 0) cv_wait(&spa->spa_waiters_cv, &spa->spa_activities_lock); spa->spa_waiters_cancel = B_FALSE; mutex_exit(&spa->spa_activities_lock); } /* Whether the vdev or any of its descendants are being initialized/trimmed. */ static boolean_t spa_vdev_activity_in_progress_impl(vdev_t *vd, zpool_wait_activity_t activity) { spa_t *spa = vd->vdev_spa; ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER)); ASSERT(MUTEX_HELD(&spa->spa_activities_lock)); ASSERT(activity == ZPOOL_WAIT_INITIALIZE || activity == ZPOOL_WAIT_TRIM); kmutex_t *lock = activity == ZPOOL_WAIT_INITIALIZE ? &vd->vdev_initialize_lock : &vd->vdev_trim_lock; mutex_exit(&spa->spa_activities_lock); mutex_enter(lock); mutex_enter(&spa->spa_activities_lock); boolean_t in_progress = (activity == ZPOOL_WAIT_INITIALIZE) ? (vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) : (vd->vdev_trim_state == VDEV_TRIM_ACTIVE); mutex_exit(lock); if (in_progress) return (B_TRUE); for (int i = 0; i < vd->vdev_children; i++) { if (spa_vdev_activity_in_progress_impl(vd->vdev_child[i], activity)) return (B_TRUE); } return (B_FALSE); } /* * If use_guid is true, this checks whether the vdev specified by guid is * being initialized/trimmed. Otherwise, it checks whether any vdev in the pool * is being initialized/trimmed. The caller must hold the config lock and * spa_activities_lock. */ static int spa_vdev_activity_in_progress(spa_t *spa, boolean_t use_guid, uint64_t guid, zpool_wait_activity_t activity, boolean_t *in_progress) { mutex_exit(&spa->spa_activities_lock); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); mutex_enter(&spa->spa_activities_lock); vdev_t *vd; if (use_guid) { vd = spa_lookup_by_guid(spa, guid, B_FALSE); if (vd == NULL || !vd->vdev_ops->vdev_op_leaf) { spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (EINVAL); } } else { vd = spa->spa_root_vdev; } *in_progress = spa_vdev_activity_in_progress_impl(vd, activity); spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); return (0); } /* * Locking for waiting threads * --------------------------- * * Waiting threads need a way to check whether a given activity is in progress, * and then, if it is, wait for it to complete. Each activity will have some * in-memory representation of the relevant on-disk state which can be used to * determine whether or not the activity is in progress. The in-memory state and * the locking used to protect it will be different for each activity, and may * not be suitable for use with a cvar (e.g., some state is protected by the * config lock). To allow waiting threads to wait without any races, another * lock, spa_activities_lock, is used. * * When the state is checked, both the activity-specific lock (if there is one) * and spa_activities_lock are held. In some cases, the activity-specific lock * is acquired explicitly (e.g. the config lock). In others, the locking is * internal to some check (e.g. bpobj_is_empty). After checking, the waiting * thread releases the activity-specific lock and, if the activity is in * progress, then cv_waits using spa_activities_lock. * * The waiting thread is woken when another thread, one completing some * activity, updates the state of the activity and then calls * spa_notify_waiters, which will cv_broadcast. This 'completing' thread only * needs to hold its activity-specific lock when updating the state, and this * lock can (but doesn't have to) be dropped before calling spa_notify_waiters. * * Because spa_notify_waiters acquires spa_activities_lock before broadcasting, * and because it is held when the waiting thread checks the state of the * activity, it can never be the case that the completing thread both updates * the activity state and cv_broadcasts in between the waiting thread's check * and cv_wait. Thus, a waiting thread can never miss a wakeup. * * In order to prevent deadlock, when the waiting thread does its check, in some * cases it will temporarily drop spa_activities_lock in order to acquire the * activity-specific lock. The order in which spa_activities_lock and the * activity specific lock are acquired in the waiting thread is determined by * the order in which they are acquired in the completing thread; if the * completing thread calls spa_notify_waiters with the activity-specific lock * held, then the waiting thread must also acquire the activity-specific lock * first. */ static int spa_activity_in_progress(spa_t *spa, zpool_wait_activity_t activity, boolean_t use_tag, uint64_t tag, boolean_t *in_progress) { int error = 0; ASSERT(MUTEX_HELD(&spa->spa_activities_lock)); switch (activity) { case ZPOOL_WAIT_CKPT_DISCARD: *in_progress = (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT) && zap_contains(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT) == ENOENT); break; case ZPOOL_WAIT_FREE: *in_progress = ((spa_version(spa) >= SPA_VERSION_DEADLISTS && !bpobj_is_empty(&spa->spa_dsl_pool->dp_free_bpobj)) || spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY) || spa_livelist_delete_check(spa)); break; case ZPOOL_WAIT_INITIALIZE: case ZPOOL_WAIT_TRIM: error = spa_vdev_activity_in_progress(spa, use_tag, tag, activity, in_progress); break; case ZPOOL_WAIT_REPLACE: mutex_exit(&spa->spa_activities_lock); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); mutex_enter(&spa->spa_activities_lock); *in_progress = vdev_replace_in_progress(spa->spa_root_vdev); spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); break; case ZPOOL_WAIT_REMOVE: *in_progress = (spa->spa_removing_phys.sr_state == DSS_SCANNING); break; case ZPOOL_WAIT_RESILVER: *in_progress = vdev_rebuild_active(spa->spa_root_vdev); if (*in_progress) break; zfs_fallthrough; case ZPOOL_WAIT_SCRUB: { boolean_t scanning, paused, is_scrub; dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan; is_scrub = (scn->scn_phys.scn_func == POOL_SCAN_SCRUB); scanning = (scn->scn_phys.scn_state == DSS_SCANNING); paused = dsl_scan_is_paused_scrub(scn); *in_progress = (scanning && !paused && is_scrub == (activity == ZPOOL_WAIT_SCRUB)); break; } case ZPOOL_WAIT_RAIDZ_EXPAND: { vdev_raidz_expand_t *vre = spa->spa_raidz_expand; *in_progress = (vre != NULL && vre->vre_state == DSS_SCANNING); break; } default: panic("unrecognized value for activity %d", activity); } return (error); } static int spa_wait_common(const char *pool, zpool_wait_activity_t activity, boolean_t use_tag, uint64_t tag, boolean_t *waited) { /* * The tag is used to distinguish between instances of an activity. * 'initialize' and 'trim' are the only activities that we use this for. * The other activities can only have a single instance in progress in a * pool at one time, making the tag unnecessary. * * There can be multiple devices being replaced at once, but since they * all finish once resilvering finishes, we don't bother keeping track * of them individually, we just wait for them all to finish. */ if (use_tag && activity != ZPOOL_WAIT_INITIALIZE && activity != ZPOOL_WAIT_TRIM) return (EINVAL); if (activity < 0 || activity >= ZPOOL_WAIT_NUM_ACTIVITIES) return (EINVAL); spa_t *spa; int error = spa_open(pool, &spa, FTAG); if (error != 0) return (error); /* * Increment the spa's waiter count so that we can call spa_close and * still ensure that the spa_t doesn't get freed before this thread is * finished with it when the pool is exported. We want to call spa_close * before we start waiting because otherwise the additional ref would * prevent the pool from being exported or destroyed throughout the * potentially long wait. */ mutex_enter(&spa->spa_activities_lock); spa->spa_waiters++; spa_close(spa, FTAG); *waited = B_FALSE; for (;;) { boolean_t in_progress; error = spa_activity_in_progress(spa, activity, use_tag, tag, &in_progress); if (error || !in_progress || spa->spa_waiters_cancel) break; *waited = B_TRUE; if (cv_wait_sig(&spa->spa_activities_cv, &spa->spa_activities_lock) == 0) { error = EINTR; break; } } spa->spa_waiters--; cv_signal(&spa->spa_waiters_cv); mutex_exit(&spa->spa_activities_lock); return (error); } /* * Wait for a particular instance of the specified activity to complete, where * the instance is identified by 'tag' */ int spa_wait_tag(const char *pool, zpool_wait_activity_t activity, uint64_t tag, boolean_t *waited) { return (spa_wait_common(pool, activity, B_TRUE, tag, waited)); } /* * Wait for all instances of the specified activity complete */ int spa_wait(const char *pool, zpool_wait_activity_t activity, boolean_t *waited) { return (spa_wait_common(pool, activity, B_FALSE, 0, waited)); } sysevent_t * spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name) { sysevent_t *ev = NULL; #ifdef _KERNEL nvlist_t *resource; resource = zfs_event_create(spa, vd, FM_SYSEVENT_CLASS, name, hist_nvl); if (resource) { ev = kmem_alloc(sizeof (sysevent_t), KM_SLEEP); ev->resource = resource; } #else (void) spa, (void) vd, (void) hist_nvl, (void) name; #endif return (ev); } void spa_event_post(sysevent_t *ev) { #ifdef _KERNEL if (ev) { zfs_zevent_post(ev->resource, NULL, zfs_zevent_post_cb); kmem_free(ev, sizeof (*ev)); } #else (void) ev; #endif } /* * Post a zevent corresponding to the given sysevent. The 'name' must be one * of the event definitions in sys/sysevent/eventdefs.h. The payload will be * filled in from the spa and (optionally) the vdev. This doesn't do anything * in the userland libzpool, as we don't want consumers to misinterpret ztest * or zdb as real changes. */ void spa_event_notify(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name) { spa_event_post(spa_event_create(spa, vd, hist_nvl, name)); } /* state manipulation functions */ EXPORT_SYMBOL(spa_open); EXPORT_SYMBOL(spa_open_rewind); EXPORT_SYMBOL(spa_get_stats); EXPORT_SYMBOL(spa_create); EXPORT_SYMBOL(spa_import); EXPORT_SYMBOL(spa_tryimport); EXPORT_SYMBOL(spa_destroy); EXPORT_SYMBOL(spa_export); EXPORT_SYMBOL(spa_reset); EXPORT_SYMBOL(spa_async_request); EXPORT_SYMBOL(spa_async_suspend); EXPORT_SYMBOL(spa_async_resume); EXPORT_SYMBOL(spa_inject_addref); EXPORT_SYMBOL(spa_inject_delref); EXPORT_SYMBOL(spa_scan_stat_init); EXPORT_SYMBOL(spa_scan_get_stats); /* device manipulation */ EXPORT_SYMBOL(spa_vdev_add); EXPORT_SYMBOL(spa_vdev_attach); EXPORT_SYMBOL(spa_vdev_detach); EXPORT_SYMBOL(spa_vdev_setpath); EXPORT_SYMBOL(spa_vdev_setfru); EXPORT_SYMBOL(spa_vdev_split_mirror); /* spare statech is global across all pools) */ EXPORT_SYMBOL(spa_spare_add); EXPORT_SYMBOL(spa_spare_remove); EXPORT_SYMBOL(spa_spare_exists); EXPORT_SYMBOL(spa_spare_activate); /* L2ARC statech is global across all pools) */ EXPORT_SYMBOL(spa_l2cache_add); EXPORT_SYMBOL(spa_l2cache_remove); EXPORT_SYMBOL(spa_l2cache_exists); EXPORT_SYMBOL(spa_l2cache_activate); EXPORT_SYMBOL(spa_l2cache_drop); /* scanning */ EXPORT_SYMBOL(spa_scan); EXPORT_SYMBOL(spa_scan_stop); /* spa syncing */ EXPORT_SYMBOL(spa_sync); /* only for DMU use */ EXPORT_SYMBOL(spa_sync_allpools); /* properties */ EXPORT_SYMBOL(spa_prop_set); EXPORT_SYMBOL(spa_prop_get); EXPORT_SYMBOL(spa_prop_clear_bootfs); /* asynchronous event notification */ EXPORT_SYMBOL(spa_event_notify); ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, preload_pct, UINT, ZMOD_RW, "Percentage of CPUs to run a metaslab preload taskq"); /* BEGIN CSTYLED */ ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_shift, UINT, ZMOD_RW, "log2 fraction of arc that can be used by inflight I/Os when " "verifying pool during import"); /* END CSTYLED */ ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_metadata, INT, ZMOD_RW, "Set to traverse metadata on pool import"); ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_data, INT, ZMOD_RW, "Set to traverse data on pool import"); ZFS_MODULE_PARAM(zfs_spa, spa_, load_print_vdev_tree, INT, ZMOD_RW, "Print vdev tree to zfs_dbgmsg during pool import"); ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_pct, UINT, ZMOD_RD, "Percentage of CPUs to run an IO worker thread"); ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_tpq, UINT, ZMOD_RD, "Number of threads per IO worker taskqueue"); /* BEGIN CSTYLED */ ZFS_MODULE_PARAM(zfs, zfs_, max_missing_tvds, U64, ZMOD_RW, "Allow importing pool with up to this number of missing top-level " "vdevs (in read-only mode)"); /* END CSTYLED */ ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_pause, INT, ZMOD_RW, "Set the livelist condense zthr to pause"); ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_pause, INT, ZMOD_RW, "Set the livelist condense synctask to pause"); /* BEGIN CSTYLED */ ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_cancel, INT, ZMOD_RW, "Whether livelist condensing was canceled in the synctask"); ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_cancel, INT, ZMOD_RW, "Whether livelist condensing was canceled in the zthr function"); ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, new_alloc, INT, ZMOD_RW, "Whether extra ALLOC blkptrs were added to a livelist entry while it " "was being condensed"); #ifdef _KERNEL ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs_zio, zio_, taskq_read, spa_taskq_read_param_set, spa_taskq_read_param_get, ZMOD_RD, "Configure IO queues for read IO"); ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs_zio, zio_, taskq_write, spa_taskq_write_param_set, spa_taskq_write_param_get, ZMOD_RD, "Configure IO queues for write IO"); #endif /* END CSTYLED */ ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_wr_iss_ncpus, UINT, ZMOD_RW, "Number of CPUs to run write issue taskqs"); diff --git a/sys/contrib/openzfs/module/zfs/spa_misc.c b/sys/contrib/openzfs/module/zfs/spa_misc.c index 68b907614196..5fb7847b5d8b 100644 --- a/sys/contrib/openzfs/module/zfs/spa_misc.c +++ b/sys/contrib/openzfs/module/zfs/spa_misc.c @@ -1,3088 +1,3100 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2011, 2019 by Delphix. All rights reserved. + * Copyright (c) 2011, 2024 by Delphix. All rights reserved. * Copyright 2015 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2017 Datto Inc. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2019, loli10K . All rights reserved. * Copyright (c) 2023, Klara Inc. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "zfs_prop.h" #include #include #include #include /* * SPA locking * * There are three basic locks for managing spa_t structures: * * spa_namespace_lock (global mutex) * * This lock must be acquired to do any of the following: * * - Lookup a spa_t by name * - Add or remove a spa_t from the namespace * - Increase spa_refcount from non-zero * - Check if spa_refcount is zero * - Rename a spa_t * - add/remove/attach/detach devices - * - Held for the duration of create/destroy/import/export + * - Held for the duration of create/destroy/export + * - Held at the start and end of import * * It does not need to handle recursion. A create or destroy may * reference objects (files or zvols) in other pools, but by * definition they must have an existing reference, and will never need * to lookup a spa_t by name. * * spa_refcount (per-spa zfs_refcount_t protected by mutex) * * This reference count keep track of any active users of the spa_t. The * spa_t cannot be destroyed or freed while this is non-zero. Internally, * the refcount is never really 'zero' - opening a pool implicitly keeps * some references in the DMU. Internally we check against spa_minref, but * present the image of a zero/non-zero value to consumers. * * spa_config_lock[] (per-spa array of rwlocks) * * This protects the spa_t from config changes, and must be held in * the following circumstances: * * - RW_READER to perform I/O to the spa * - RW_WRITER to change the vdev config * * The locking order is fairly straightforward: * * spa_namespace_lock -> spa_refcount * * The namespace lock must be acquired to increase the refcount from 0 * or to check if it is zero. * * spa_refcount -> spa_config_lock[] * * There must be at least one valid reference on the spa_t to acquire * the config lock. * * spa_namespace_lock -> spa_config_lock[] * * The namespace lock must always be taken before the config lock. * * * The spa_namespace_lock can be acquired directly and is globally visible. * * The namespace is manipulated using the following functions, all of which * require the spa_namespace_lock to be held. * * spa_lookup() Lookup a spa_t by name. * * spa_add() Create a new spa_t in the namespace. * * spa_remove() Remove a spa_t from the namespace. This also * frees up any memory associated with the spa_t. * * spa_next() Returns the next spa_t in the system, or the * first if NULL is passed. * * spa_evict_all() Shutdown and remove all spa_t structures in * the system. * * spa_guid_exists() Determine whether a pool/device guid exists. * * The spa_refcount is manipulated using the following functions: * * spa_open_ref() Adds a reference to the given spa_t. Must be * called with spa_namespace_lock held if the * refcount is currently zero. * * spa_close() Remove a reference from the spa_t. This will * not free the spa_t or remove it from the * namespace. No locking is required. * * spa_refcount_zero() Returns true if the refcount is currently * zero. Must be called with spa_namespace_lock * held. * * The spa_config_lock[] is an array of rwlocks, ordered as follows: * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). * * To read the configuration, it suffices to hold one of these locks as reader. * To modify the configuration, you must hold all locks as writer. To modify * vdev state without altering the vdev tree's topology (e.g. online/offline), * you must hold SCL_STATE and SCL_ZIO as writer. * * We use these distinct config locks to avoid recursive lock entry. * For example, spa_sync() (which holds SCL_CONFIG as reader) induces * block allocations (SCL_ALLOC), which may require reading space maps * from disk (dmu_read() -> zio_read() -> SCL_ZIO). * * The spa config locks cannot be normal rwlocks because we need the * ability to hand off ownership. For example, SCL_ZIO is acquired * by the issuing thread and later released by an interrupt thread. * They do, however, obey the usual write-wanted semantics to prevent * writer (i.e. system administrator) starvation. * * The lock acquisition rules are as follows: * * SCL_CONFIG * Protects changes to the vdev tree topology, such as vdev * add/remove/attach/detach. Protects the dirty config list * (spa_config_dirty_list) and the set of spares and l2arc devices. * * SCL_STATE * Protects changes to pool state and vdev state, such as vdev * online/offline/fault/degrade/clear. Protects the dirty state list * (spa_state_dirty_list) and global pool state (spa_state). * * SCL_ALLOC * Protects changes to metaslab groups and classes. * Held as reader by metaslab_alloc() and metaslab_claim(). * * SCL_ZIO * Held by bp-level zios (those which have no io_vd upon entry) * to prevent changes to the vdev tree. The bp-level zio implicitly * protects all of its vdev child zios, which do not hold SCL_ZIO. * * SCL_FREE * Protects changes to metaslab groups and classes. * Held as reader by metaslab_free(). SCL_FREE is distinct from * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free * blocks in zio_done() while another i/o that holds either * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. * * SCL_VDEV * Held as reader to prevent changes to the vdev tree during trivial * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the * other locks, and lower than all of them, to ensure that it's safe * to acquire regardless of caller context. * * In addition, the following rules apply: * * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. * The lock ordering is SCL_CONFIG > spa_props_lock. * * (b) I/O operations on leaf vdevs. For any zio operation that takes * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), * or zio_write_phys() -- the caller must ensure that the config cannot * cannot change in the interim, and that the vdev cannot be reopened. * SCL_STATE as reader suffices for both. * * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). * * spa_vdev_enter() Acquire the namespace lock and the config lock * for writing. * * spa_vdev_exit() Release the config lock, wait for all I/O * to complete, sync the updated configs to the * cache, and release the namespace lock. * * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual * locking is, always, based on spa_namespace_lock and spa_config_lock[]. */ -static avl_tree_t spa_namespace_avl; +avl_tree_t spa_namespace_avl; kmutex_t spa_namespace_lock; -static kcondvar_t spa_namespace_cv; +kcondvar_t spa_namespace_cv; static const int spa_max_replication_override = SPA_DVAS_PER_BP; static kmutex_t spa_spare_lock; static avl_tree_t spa_spare_avl; static kmutex_t spa_l2cache_lock; static avl_tree_t spa_l2cache_avl; spa_mode_t spa_mode_global = SPA_MODE_UNINIT; #ifdef ZFS_DEBUG /* * Everything except dprintf, set_error, spa, and indirect_remap is on * by default in debug builds. */ int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SET_ERROR | ZFS_DEBUG_INDIRECT_REMAP); #else int zfs_flags = 0; #endif /* * zfs_recover can be set to nonzero to attempt to recover from * otherwise-fatal errors, typically caused by on-disk corruption. When * set, calls to zfs_panic_recover() will turn into warning messages. * This should only be used as a last resort, as it typically results * in leaked space, or worse. */ int zfs_recover = B_FALSE; /* * If destroy encounters an EIO while reading metadata (e.g. indirect * blocks), space referenced by the missing metadata can not be freed. * Normally this causes the background destroy to become "stalled", as * it is unable to make forward progress. While in this stalled state, * all remaining space to free from the error-encountering filesystem is * "temporarily leaked". Set this flag to cause it to ignore the EIO, * permanently leak the space from indirect blocks that can not be read, * and continue to free everything else that it can. * * The default, "stalling" behavior is useful if the storage partially * fails (i.e. some but not all i/os fail), and then later recovers. In * this case, we will be able to continue pool operations while it is * partially failed, and when it recovers, we can continue to free the * space, with no leaks. However, note that this case is actually * fairly rare. * * Typically pools either (a) fail completely (but perhaps temporarily, * e.g. a top-level vdev going offline), or (b) have localized, * permanent errors (e.g. disk returns the wrong data due to bit flip or * firmware bug). In case (a), this setting does not matter because the * pool will be suspended and the sync thread will not be able to make * forward progress regardless. In case (b), because the error is * permanent, the best we can do is leak the minimum amount of space, * which is what setting this flag will do. Therefore, it is reasonable * for this flag to normally be set, but we chose the more conservative * approach of not setting it, so that there is no possibility of * leaking space in the "partial temporary" failure case. */ int zfs_free_leak_on_eio = B_FALSE; /* * Expiration time in milliseconds. This value has two meanings. First it is * used to determine when the spa_deadman() logic should fire. By default the * spa_deadman() will fire if spa_sync() has not completed in 600 seconds. * Secondly, the value determines if an I/O is considered "hung". Any I/O that * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting * in one of three behaviors controlled by zfs_deadman_failmode. */ uint64_t zfs_deadman_synctime_ms = 600000UL; /* 10 min. */ /* * This value controls the maximum amount of time zio_wait() will block for an * outstanding IO. By default this is 300 seconds at which point the "hung" * behavior will be applied as described for zfs_deadman_synctime_ms. */ uint64_t zfs_deadman_ziotime_ms = 300000UL; /* 5 min. */ /* * Check time in milliseconds. This defines the frequency at which we check * for hung I/O. */ uint64_t zfs_deadman_checktime_ms = 60000UL; /* 1 min. */ /* * By default the deadman is enabled. */ int zfs_deadman_enabled = B_TRUE; /* * Controls the behavior of the deadman when it detects a "hung" I/O. * Valid values are zfs_deadman_failmode=. * * wait - Wait for the "hung" I/O (default) * continue - Attempt to recover from a "hung" I/O * panic - Panic the system */ const char *zfs_deadman_failmode = "wait"; /* * The worst case is single-sector max-parity RAID-Z blocks, in which * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) * times the size; so just assume that. Add to this the fact that * we can have up to 3 DVAs per bp, and one more factor of 2 because * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, * the worst case is: * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 */ uint_t spa_asize_inflation = 24; /* * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in * the pool to be consumed (bounded by spa_max_slop). This ensures that we * don't run the pool completely out of space, due to unaccounted changes (e.g. * to the MOS). It also limits the worst-case time to allocate space. If we * have less than this amount of free space, most ZPL operations (e.g. write, * create) will return ENOSPC. The ZIL metaslabs (spa_embedded_log_class) are * also part of this 3.2% of space which can't be consumed by normal writes; * the slop space "proper" (spa_get_slop_space()) is decreased by the embedded * log space. * * Certain operations (e.g. file removal, most administrative actions) can * use half the slop space. They will only return ENOSPC if less than half * the slop space is free. Typically, once the pool has less than the slop * space free, the user will use these operations to free up space in the pool. * These are the operations that call dsl_pool_adjustedsize() with the netfree * argument set to TRUE. * * Operations that are almost guaranteed to free up space in the absence of * a pool checkpoint can use up to three quarters of the slop space * (e.g zfs destroy). * * A very restricted set of operations are always permitted, regardless of * the amount of free space. These are the operations that call * dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net * increase in the amount of space used, it is possible to run the pool * completely out of space, causing it to be permanently read-only. * * Note that on very small pools, the slop space will be larger than * 3.2%, in an effort to have it be at least spa_min_slop (128MB), * but we never allow it to be more than half the pool size. * * Further, on very large pools, the slop space will be smaller than * 3.2%, to avoid reserving much more space than we actually need; bounded * by spa_max_slop (128GB). * * See also the comments in zfs_space_check_t. */ uint_t spa_slop_shift = 5; static const uint64_t spa_min_slop = 128ULL * 1024 * 1024; static const uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024; /* * Number of allocators to use, per spa instance */ static int spa_num_allocators = 4; /* * Spa active allocator. * Valid values are zfs_active_allocator=. */ const char *zfs_active_allocator = "dynamic"; void spa_load_failed(spa_t *spa, const char *fmt, ...) { va_list adx; char buf[256]; va_start(adx, fmt); (void) vsnprintf(buf, sizeof (buf), fmt, adx); va_end(adx); zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name, spa->spa_trust_config ? "trusted" : "untrusted", buf); } void spa_load_note(spa_t *spa, const char *fmt, ...) { va_list adx; char buf[256]; va_start(adx, fmt); (void) vsnprintf(buf, sizeof (buf), fmt, adx); va_end(adx); zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name, spa->spa_trust_config ? "trusted" : "untrusted", buf); spa_import_progress_set_notes_nolog(spa, "%s", buf); } /* * By default dedup and user data indirects land in the special class */ static int zfs_ddt_data_is_special = B_TRUE; static int zfs_user_indirect_is_special = B_TRUE; /* * The percentage of special class final space reserved for metadata only. * Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only * let metadata into the class. */ static uint_t zfs_special_class_metadata_reserve_pct = 25; /* * ========================================================================== * SPA config locking * ========================================================================== */ static void spa_config_lock_init(spa_t *spa) { for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); scl->scl_writer = NULL; scl->scl_write_wanted = 0; scl->scl_count = 0; } } static void spa_config_lock_destroy(spa_t *spa) { for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; mutex_destroy(&scl->scl_lock); cv_destroy(&scl->scl_cv); ASSERT(scl->scl_writer == NULL); ASSERT(scl->scl_write_wanted == 0); ASSERT(scl->scl_count == 0); } } int spa_config_tryenter(spa_t *spa, int locks, const void *tag, krw_t rw) { for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (!(locks & (1 << i))) continue; mutex_enter(&scl->scl_lock); if (rw == RW_READER) { if (scl->scl_writer || scl->scl_write_wanted) { mutex_exit(&scl->scl_lock); spa_config_exit(spa, locks & ((1 << i) - 1), tag); return (0); } } else { ASSERT(scl->scl_writer != curthread); if (scl->scl_count != 0) { mutex_exit(&scl->scl_lock); spa_config_exit(spa, locks & ((1 << i) - 1), tag); return (0); } scl->scl_writer = curthread; } scl->scl_count++; mutex_exit(&scl->scl_lock); } return (1); } static void spa_config_enter_impl(spa_t *spa, int locks, const void *tag, krw_t rw, int mmp_flag) { (void) tag; int wlocks_held = 0; ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (scl->scl_writer == curthread) wlocks_held |= (1 << i); if (!(locks & (1 << i))) continue; mutex_enter(&scl->scl_lock); if (rw == RW_READER) { while (scl->scl_writer || (!mmp_flag && scl->scl_write_wanted)) { cv_wait(&scl->scl_cv, &scl->scl_lock); } } else { ASSERT(scl->scl_writer != curthread); while (scl->scl_count != 0) { scl->scl_write_wanted++; cv_wait(&scl->scl_cv, &scl->scl_lock); scl->scl_write_wanted--; } scl->scl_writer = curthread; } scl->scl_count++; mutex_exit(&scl->scl_lock); } ASSERT3U(wlocks_held, <=, locks); } void spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw) { spa_config_enter_impl(spa, locks, tag, rw, 0); } /* * The spa_config_enter_mmp() allows the mmp thread to cut in front of * outstanding write lock requests. This is needed since the mmp updates are * time sensitive and failure to service them promptly will result in a * suspended pool. This pool suspension has been seen in practice when there is * a single disk in a pool that is responding slowly and presumably about to * fail. */ void spa_config_enter_mmp(spa_t *spa, int locks, const void *tag, krw_t rw) { spa_config_enter_impl(spa, locks, tag, rw, 1); } void spa_config_exit(spa_t *spa, int locks, const void *tag) { (void) tag; for (int i = SCL_LOCKS - 1; i >= 0; i--) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (!(locks & (1 << i))) continue; mutex_enter(&scl->scl_lock); ASSERT(scl->scl_count > 0); if (--scl->scl_count == 0) { ASSERT(scl->scl_writer == NULL || scl->scl_writer == curthread); scl->scl_writer = NULL; /* OK in either case */ cv_broadcast(&scl->scl_cv); } mutex_exit(&scl->scl_lock); } } int spa_config_held(spa_t *spa, int locks, krw_t rw) { int locks_held = 0; for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (!(locks & (1 << i))) continue; if ((rw == RW_READER && scl->scl_count != 0) || (rw == RW_WRITER && scl->scl_writer == curthread)) locks_held |= 1 << i; } return (locks_held); } /* * ========================================================================== * SPA namespace functions * ========================================================================== */ /* * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. * Returns NULL if no matching spa_t is found. */ spa_t * spa_lookup(const char *name) { static spa_t search; /* spa_t is large; don't allocate on stack */ spa_t *spa; avl_index_t where; char *cp; ASSERT(MUTEX_HELD(&spa_namespace_lock)); +retry: (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); /* * If it's a full dataset name, figure out the pool name and * just use that. */ cp = strpbrk(search.spa_name, "/@#"); if (cp != NULL) *cp = '\0'; spa = avl_find(&spa_namespace_avl, &search, &where); + if (spa == NULL) + return (NULL); + + if (spa->spa_load_thread != NULL && + spa->spa_load_thread != curthread) { + cv_wait(&spa_namespace_cv, &spa_namespace_lock); + goto retry; + } return (spa); } /* * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. * If the zfs_deadman_enabled flag is set then it inspects all vdev queues * looking for potentially hung I/Os. */ void spa_deadman(void *arg) { spa_t *spa = arg; /* Disable the deadman if the pool is suspended. */ if (spa_suspended(spa)) return; zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", (gethrtime() - spa->spa_sync_starttime) / NANOSEC, (u_longlong_t)++spa->spa_deadman_calls); if (zfs_deadman_enabled) vdev_deadman(spa->spa_root_vdev, FTAG); spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq, spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() + MSEC_TO_TICK(zfs_deadman_checktime_ms)); } static int spa_log_sm_sort_by_txg(const void *va, const void *vb) { const spa_log_sm_t *a = va; const spa_log_sm_t *b = vb; return (TREE_CMP(a->sls_txg, b->sls_txg)); } /* * Create an uninitialized spa_t with the given name. Requires * spa_namespace_lock. The caller must ensure that the spa_t doesn't already * exist by calling spa_lookup() first. */ spa_t * spa_add(const char *name, nvlist_t *config, const char *altroot) { spa_t *spa; spa_config_dirent_t *dp; ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_activities_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_waiters_cv, NULL, CV_DEFAULT, NULL); for (int t = 0; t < TXG_SIZE; t++) bplist_create(&spa->spa_free_bplist[t]); (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); spa->spa_state = POOL_STATE_UNINITIALIZED; spa->spa_freeze_txg = UINT64_MAX; spa->spa_final_txg = UINT64_MAX; spa->spa_load_max_txg = UINT64_MAX; spa->spa_proc = &p0; spa->spa_proc_state = SPA_PROC_NONE; spa->spa_trust_config = B_TRUE; spa->spa_hostid = zone_get_hostid(NULL); spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); spa->spa_deadman_ziotime = MSEC2NSEC(zfs_deadman_ziotime_ms); spa_set_deadman_failmode(spa, zfs_deadman_failmode); spa_set_allocator(spa, zfs_active_allocator); zfs_refcount_create(&spa->spa_refcount); spa_config_lock_init(spa); spa_stats_init(spa); + ASSERT(MUTEX_HELD(&spa_namespace_lock)); avl_add(&spa_namespace_avl, spa); /* * Set the alternate root, if there is one. */ if (altroot) spa->spa_root = spa_strdup(altroot); /* Do not allow more allocators than CPUs. */ spa->spa_alloc_count = MIN(MAX(spa_num_allocators, 1), boot_ncpus); spa->spa_allocs = kmem_zalloc(spa->spa_alloc_count * sizeof (spa_alloc_t), KM_SLEEP); for (int i = 0; i < spa->spa_alloc_count; i++) { mutex_init(&spa->spa_allocs[i].spaa_lock, NULL, MUTEX_DEFAULT, NULL); avl_create(&spa->spa_allocs[i].spaa_tree, zio_bookmark_compare, sizeof (zio_t), offsetof(zio_t, io_queue_node.a)); } avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed, sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node)); avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg, sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node)); list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t), offsetof(log_summary_entry_t, lse_node)); /* * Every pool starts with the default cachefile */ list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), offsetof(spa_config_dirent_t, scd_link)); dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); list_insert_head(&spa->spa_config_list, dp); VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, KM_SLEEP) == 0); if (config != NULL) { nvlist_t *features; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, &features) == 0) { VERIFY(nvlist_dup(features, &spa->spa_label_features, 0) == 0); } VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); } if (spa->spa_label_features == NULL) { VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, KM_SLEEP) == 0); } spa->spa_min_ashift = INT_MAX; spa->spa_max_ashift = 0; spa->spa_min_alloc = INT_MAX; spa->spa_gcd_alloc = INT_MAX; /* Reset cached value */ spa->spa_dedup_dspace = ~0ULL; /* * As a pool is being created, treat all features as disabled by * setting SPA_FEATURE_DISABLED for all entries in the feature * refcount cache. */ for (int i = 0; i < SPA_FEATURES; i++) { spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; } list_create(&spa->spa_leaf_list, sizeof (vdev_t), offsetof(vdev_t, vdev_leaf_node)); return (spa); } /* * Removes a spa_t from the namespace, freeing up any memory used. Requires * spa_namespace_lock. This is called only after the spa_t has been closed and * deactivated. */ void spa_remove(spa_t *spa) { spa_config_dirent_t *dp; ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED); ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0); ASSERT0(spa->spa_waiters); nvlist_free(spa->spa_config_splitting); avl_remove(&spa_namespace_avl, spa); - cv_broadcast(&spa_namespace_cv); if (spa->spa_root) spa_strfree(spa->spa_root); while ((dp = list_remove_head(&spa->spa_config_list)) != NULL) { if (dp->scd_path != NULL) spa_strfree(dp->scd_path); kmem_free(dp, sizeof (spa_config_dirent_t)); } for (int i = 0; i < spa->spa_alloc_count; i++) { avl_destroy(&spa->spa_allocs[i].spaa_tree); mutex_destroy(&spa->spa_allocs[i].spaa_lock); } kmem_free(spa->spa_allocs, spa->spa_alloc_count * sizeof (spa_alloc_t)); avl_destroy(&spa->spa_metaslabs_by_flushed); avl_destroy(&spa->spa_sm_logs_by_txg); list_destroy(&spa->spa_log_summary); list_destroy(&spa->spa_config_list); list_destroy(&spa->spa_leaf_list); nvlist_free(spa->spa_label_features); nvlist_free(spa->spa_load_info); nvlist_free(spa->spa_feat_stats); spa_config_set(spa, NULL); zfs_refcount_destroy(&spa->spa_refcount); spa_stats_destroy(spa); spa_config_lock_destroy(spa); for (int t = 0; t < TXG_SIZE; t++) bplist_destroy(&spa->spa_free_bplist[t]); zio_checksum_templates_free(spa); cv_destroy(&spa->spa_async_cv); cv_destroy(&spa->spa_evicting_os_cv); cv_destroy(&spa->spa_proc_cv); cv_destroy(&spa->spa_scrub_io_cv); cv_destroy(&spa->spa_suspend_cv); cv_destroy(&spa->spa_activities_cv); cv_destroy(&spa->spa_waiters_cv); mutex_destroy(&spa->spa_flushed_ms_lock); mutex_destroy(&spa->spa_async_lock); mutex_destroy(&spa->spa_errlist_lock); mutex_destroy(&spa->spa_errlog_lock); mutex_destroy(&spa->spa_evicting_os_lock); mutex_destroy(&spa->spa_history_lock); mutex_destroy(&spa->spa_proc_lock); mutex_destroy(&spa->spa_props_lock); mutex_destroy(&spa->spa_cksum_tmpls_lock); mutex_destroy(&spa->spa_scrub_lock); mutex_destroy(&spa->spa_suspend_lock); mutex_destroy(&spa->spa_vdev_top_lock); mutex_destroy(&spa->spa_feat_stats_lock); mutex_destroy(&spa->spa_activities_lock); kmem_free(spa, sizeof (spa_t)); } /* * Given a pool, return the next pool in the namespace, or NULL if there is * none. If 'prev' is NULL, return the first pool. */ spa_t * spa_next(spa_t *prev) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); if (prev) return (AVL_NEXT(&spa_namespace_avl, prev)); else return (avl_first(&spa_namespace_avl)); } /* * ========================================================================== * SPA refcount functions * ========================================================================== */ /* * Add a reference to the given spa_t. Must have at least one reference, or * have the namespace lock held. */ void spa_open_ref(spa_t *spa, const void *tag) { ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref || - MUTEX_HELD(&spa_namespace_lock)); + MUTEX_HELD(&spa_namespace_lock) || + spa->spa_load_thread == curthread); (void) zfs_refcount_add(&spa->spa_refcount, tag); } /* * Remove a reference to the given spa_t. Must have at least one reference, or * have the namespace lock held. */ void spa_close(spa_t *spa, const void *tag) { ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref || - MUTEX_HELD(&spa_namespace_lock)); + MUTEX_HELD(&spa_namespace_lock) || + spa->spa_load_thread == curthread); (void) zfs_refcount_remove(&spa->spa_refcount, tag); } /* * Remove a reference to the given spa_t held by a dsl dir that is * being asynchronously released. Async releases occur from a taskq * performing eviction of dsl datasets and dirs. The namespace lock * isn't held and the hold by the object being evicted may contribute to * spa_minref (e.g. dataset or directory released during pool export), * so the asserts in spa_close() do not apply. */ void spa_async_close(spa_t *spa, const void *tag) { (void) zfs_refcount_remove(&spa->spa_refcount, tag); } /* * Check to see if the spa refcount is zero. Must be called with * spa_namespace_lock held. We really compare against spa_minref, which is the * number of references acquired when opening a pool */ boolean_t spa_refcount_zero(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref); } /* * ========================================================================== * SPA spare and l2cache tracking * ========================================================================== */ /* * Hot spares and cache devices are tracked using the same code below, * for 'auxiliary' devices. */ typedef struct spa_aux { uint64_t aux_guid; uint64_t aux_pool; avl_node_t aux_avl; int aux_count; } spa_aux_t; static inline int spa_aux_compare(const void *a, const void *b) { const spa_aux_t *sa = (const spa_aux_t *)a; const spa_aux_t *sb = (const spa_aux_t *)b; return (TREE_CMP(sa->aux_guid, sb->aux_guid)); } static void spa_aux_add(vdev_t *vd, avl_tree_t *avl) { avl_index_t where; spa_aux_t search; spa_aux_t *aux; search.aux_guid = vd->vdev_guid; if ((aux = avl_find(avl, &search, &where)) != NULL) { aux->aux_count++; } else { aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); aux->aux_guid = vd->vdev_guid; aux->aux_count = 1; avl_insert(avl, aux, where); } } static void spa_aux_remove(vdev_t *vd, avl_tree_t *avl) { spa_aux_t search; spa_aux_t *aux; avl_index_t where; search.aux_guid = vd->vdev_guid; aux = avl_find(avl, &search, &where); ASSERT(aux != NULL); if (--aux->aux_count == 0) { avl_remove(avl, aux); kmem_free(aux, sizeof (spa_aux_t)); } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { aux->aux_pool = 0ULL; } } static boolean_t spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) { spa_aux_t search, *found; search.aux_guid = guid; found = avl_find(avl, &search, NULL); if (pool) { if (found) *pool = found->aux_pool; else *pool = 0ULL; } if (refcnt) { if (found) *refcnt = found->aux_count; else *refcnt = 0; } return (found != NULL); } static void spa_aux_activate(vdev_t *vd, avl_tree_t *avl) { spa_aux_t search, *found; avl_index_t where; search.aux_guid = vd->vdev_guid; found = avl_find(avl, &search, &where); ASSERT(found != NULL); ASSERT(found->aux_pool == 0ULL); found->aux_pool = spa_guid(vd->vdev_spa); } /* * Spares are tracked globally due to the following constraints: * * - A spare may be part of multiple pools. * - A spare may be added to a pool even if it's actively in use within * another pool. * - A spare in use in any pool can only be the source of a replacement if * the target is a spare in the same pool. * * We keep track of all spares on the system through the use of a reference * counted AVL tree. When a vdev is added as a spare, or used as a replacement * spare, then we bump the reference count in the AVL tree. In addition, we set * the 'vdev_isspare' member to indicate that the device is a spare (active or * inactive). When a spare is made active (used to replace a device in the * pool), we also keep track of which pool its been made a part of. * * The 'spa_spare_lock' protects the AVL tree. These functions are normally * called under the spa_namespace lock as part of vdev reconfiguration. The * separate spare lock exists for the status query path, which does not need to * be completely consistent with respect to other vdev configuration changes. */ static int spa_spare_compare(const void *a, const void *b) { return (spa_aux_compare(a, b)); } void spa_spare_add(vdev_t *vd) { mutex_enter(&spa_spare_lock); ASSERT(!vd->vdev_isspare); spa_aux_add(vd, &spa_spare_avl); vd->vdev_isspare = B_TRUE; mutex_exit(&spa_spare_lock); } void spa_spare_remove(vdev_t *vd) { mutex_enter(&spa_spare_lock); ASSERT(vd->vdev_isspare); spa_aux_remove(vd, &spa_spare_avl); vd->vdev_isspare = B_FALSE; mutex_exit(&spa_spare_lock); } boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) { boolean_t found; mutex_enter(&spa_spare_lock); found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); mutex_exit(&spa_spare_lock); return (found); } void spa_spare_activate(vdev_t *vd) { mutex_enter(&spa_spare_lock); ASSERT(vd->vdev_isspare); spa_aux_activate(vd, &spa_spare_avl); mutex_exit(&spa_spare_lock); } /* * Level 2 ARC devices are tracked globally for the same reasons as spares. * Cache devices currently only support one pool per cache device, and so * for these devices the aux reference count is currently unused beyond 1. */ static int spa_l2cache_compare(const void *a, const void *b) { return (spa_aux_compare(a, b)); } void spa_l2cache_add(vdev_t *vd) { mutex_enter(&spa_l2cache_lock); ASSERT(!vd->vdev_isl2cache); spa_aux_add(vd, &spa_l2cache_avl); vd->vdev_isl2cache = B_TRUE; mutex_exit(&spa_l2cache_lock); } void spa_l2cache_remove(vdev_t *vd) { mutex_enter(&spa_l2cache_lock); ASSERT(vd->vdev_isl2cache); spa_aux_remove(vd, &spa_l2cache_avl); vd->vdev_isl2cache = B_FALSE; mutex_exit(&spa_l2cache_lock); } boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool) { boolean_t found; mutex_enter(&spa_l2cache_lock); found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); mutex_exit(&spa_l2cache_lock); return (found); } void spa_l2cache_activate(vdev_t *vd) { mutex_enter(&spa_l2cache_lock); ASSERT(vd->vdev_isl2cache); spa_aux_activate(vd, &spa_l2cache_avl); mutex_exit(&spa_l2cache_lock); } /* * ========================================================================== * SPA vdev locking * ========================================================================== */ /* * Lock the given spa_t for the purpose of adding or removing a vdev. * Grabs the global spa_namespace_lock plus the spa config lock for writing. * It returns the next transaction group for the spa_t. */ uint64_t spa_vdev_enter(spa_t *spa) { mutex_enter(&spa->spa_vdev_top_lock); mutex_enter(&spa_namespace_lock); vdev_autotrim_stop_all(spa); return (spa_vdev_config_enter(spa)); } /* * The same as spa_vdev_enter() above but additionally takes the guid of * the vdev being detached. When there is a rebuild in process it will be * suspended while the vdev tree is modified then resumed by spa_vdev_exit(). * The rebuild is canceled if only a single child remains after the detach. */ uint64_t spa_vdev_detach_enter(spa_t *spa, uint64_t guid) { mutex_enter(&spa->spa_vdev_top_lock); mutex_enter(&spa_namespace_lock); vdev_autotrim_stop_all(spa); if (guid != 0) { vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE); if (vd) { vdev_rebuild_stop_wait(vd->vdev_top); } } return (spa_vdev_config_enter(spa)); } /* * Internal implementation for spa_vdev_enter(). Used when a vdev * operation requires multiple syncs (i.e. removing a device) while * keeping the spa_namespace_lock held. */ uint64_t spa_vdev_config_enter(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); return (spa_last_synced_txg(spa) + 1); } /* * Used in combination with spa_vdev_config_enter() to allow the syncing * of multiple transactions without releasing the spa_namespace_lock. */ void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, const char *tag) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); int config_changed = B_FALSE; ASSERT(txg > spa_last_synced_txg(spa)); spa->spa_pending_vdev = NULL; /* * Reassess the DTLs. */ vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE, B_FALSE); if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { config_changed = B_TRUE; spa->spa_config_generation++; } /* * Verify the metaslab classes. */ ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); ASSERT(metaslab_class_validate(spa_embedded_log_class(spa)) == 0); ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0); ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0); spa_config_exit(spa, SCL_ALL, spa); /* * Panic the system if the specified tag requires it. This * is useful for ensuring that configurations are updated * transactionally. */ if (zio_injection_enabled) zio_handle_panic_injection(spa, tag, 0); /* * Note: this txg_wait_synced() is important because it ensures * that there won't be more than one config change per txg. * This allows us to use the txg as the generation number. */ if (error == 0) txg_wait_synced(spa->spa_dsl_pool, txg); if (vd != NULL) { ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); if (vd->vdev_ops->vdev_op_leaf) { mutex_enter(&vd->vdev_initialize_lock); vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, NULL); mutex_exit(&vd->vdev_initialize_lock); mutex_enter(&vd->vdev_trim_lock); vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL); mutex_exit(&vd->vdev_trim_lock); } /* * The vdev may be both a leaf and top-level device. */ vdev_autotrim_stop_wait(vd); spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER); vdev_free(vd); spa_config_exit(spa, SCL_STATE_ALL, spa); } /* * If the config changed, update the config cache. */ if (config_changed) spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE); } /* * Unlock the spa_t after adding or removing a vdev. Besides undoing the * locking of spa_vdev_enter(), we also want make sure the transactions have * synced to disk, and then update the global configuration cache with the new * information. */ int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) { vdev_autotrim_restart(spa); vdev_rebuild_restart(spa); spa_vdev_config_exit(spa, vd, txg, error, FTAG); mutex_exit(&spa_namespace_lock); mutex_exit(&spa->spa_vdev_top_lock); return (error); } /* * Lock the given spa_t for the purpose of changing vdev state. */ void spa_vdev_state_enter(spa_t *spa, int oplocks) { int locks = SCL_STATE_ALL | oplocks; /* * Root pools may need to read of the underlying devfs filesystem * when opening up a vdev. Unfortunately if we're holding the * SCL_ZIO lock it will result in a deadlock when we try to issue * the read from the root filesystem. Instead we "prefetch" * the associated vnodes that we need prior to opening the * underlying devices and cache them so that we can prevent * any I/O when we are doing the actual open. */ if (spa_is_root(spa)) { int low = locks & ~(SCL_ZIO - 1); int high = locks & ~low; spa_config_enter(spa, high, spa, RW_WRITER); vdev_hold(spa->spa_root_vdev); spa_config_enter(spa, low, spa, RW_WRITER); } else { spa_config_enter(spa, locks, spa, RW_WRITER); } spa->spa_vdev_locks = locks; } int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) { boolean_t config_changed = B_FALSE; vdev_t *vdev_top; if (vd == NULL || vd == spa->spa_root_vdev) { vdev_top = spa->spa_root_vdev; } else { vdev_top = vd->vdev_top; } if (vd != NULL || error == 0) vdev_dtl_reassess(vdev_top, 0, 0, B_FALSE, B_FALSE); if (vd != NULL) { if (vd != spa->spa_root_vdev) vdev_state_dirty(vdev_top); config_changed = B_TRUE; spa->spa_config_generation++; } if (spa_is_root(spa)) vdev_rele(spa->spa_root_vdev); ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); spa_config_exit(spa, spa->spa_vdev_locks, spa); /* * If anything changed, wait for it to sync. This ensures that, * from the system administrator's perspective, zpool(8) commands * are synchronous. This is important for things like zpool offline: * when the command completes, you expect no further I/O from ZFS. */ if (vd != NULL) txg_wait_synced(spa->spa_dsl_pool, 0); /* * If the config changed, update the config cache. */ if (config_changed) { mutex_enter(&spa_namespace_lock); spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE); mutex_exit(&spa_namespace_lock); } return (error); } /* * ========================================================================== * Miscellaneous functions * ========================================================================== */ void spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) { if (!nvlist_exists(spa->spa_label_features, feature)) { fnvlist_add_boolean(spa->spa_label_features, feature); /* * When we are creating the pool (tx_txg==TXG_INITIAL), we can't * dirty the vdev config because lock SCL_CONFIG is not held. * Thankfully, in this case we don't need to dirty the config * because it will be written out anyway when we finish * creating the pool. */ if (tx->tx_txg != TXG_INITIAL) vdev_config_dirty(spa->spa_root_vdev); } } void spa_deactivate_mos_feature(spa_t *spa, const char *feature) { if (nvlist_remove_all(spa->spa_label_features, feature) == 0) vdev_config_dirty(spa->spa_root_vdev); } /* * Return the spa_t associated with given pool_guid, if it exists. If * device_guid is non-zero, determine whether the pool exists *and* contains * a device with the specified device_guid. */ spa_t * spa_by_guid(uint64_t pool_guid, uint64_t device_guid) { spa_t *spa; avl_tree_t *t = &spa_namespace_avl; ASSERT(MUTEX_HELD(&spa_namespace_lock)); for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { if (spa->spa_state == POOL_STATE_UNINITIALIZED) continue; if (spa->spa_root_vdev == NULL) continue; if (spa_guid(spa) == pool_guid) { if (device_guid == 0) break; if (vdev_lookup_by_guid(spa->spa_root_vdev, device_guid) != NULL) break; /* * Check any devices we may be in the process of adding. */ if (spa->spa_pending_vdev) { if (vdev_lookup_by_guid(spa->spa_pending_vdev, device_guid) != NULL) break; } } } return (spa); } /* * Determine whether a pool with the given pool_guid exists. */ boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) { return (spa_by_guid(pool_guid, device_guid) != NULL); } char * spa_strdup(const char *s) { size_t len; char *new; len = strlen(s); new = kmem_alloc(len + 1, KM_SLEEP); memcpy(new, s, len + 1); return (new); } void spa_strfree(char *s) { kmem_free(s, strlen(s) + 1); } uint64_t spa_generate_guid(spa_t *spa) { uint64_t guid; if (spa != NULL) { do { (void) random_get_pseudo_bytes((void *)&guid, sizeof (guid)); } while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)); } else { do { (void) random_get_pseudo_bytes((void *)&guid, sizeof (guid)); } while (guid == 0 || spa_guid_exists(guid, 0)); } return (guid); } void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) { char type[256]; const char *checksum = NULL; const char *compress = NULL; if (bp != NULL) { if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { dmu_object_byteswap_t bswap = DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); (void) snprintf(type, sizeof (type), "bswap %s %s", DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? "metadata" : "data", dmu_ot_byteswap[bswap].ob_name); } else { (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, sizeof (type)); } if (!BP_IS_EMBEDDED(bp)) { checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; } compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; } SNPRINTF_BLKPTR(kmem_scnprintf, ' ', buf, buflen, bp, type, checksum, compress); } void spa_freeze(spa_t *spa) { uint64_t freeze_txg = 0; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); if (spa->spa_freeze_txg == UINT64_MAX) { freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; spa->spa_freeze_txg = freeze_txg; } spa_config_exit(spa, SCL_ALL, FTAG); if (freeze_txg != 0) txg_wait_synced(spa_get_dsl(spa), freeze_txg); } void zfs_panic_recover(const char *fmt, ...) { va_list adx; va_start(adx, fmt); vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); va_end(adx); } /* * This is a stripped-down version of strtoull, suitable only for converting * lowercase hexadecimal numbers that don't overflow. */ uint64_t zfs_strtonum(const char *str, char **nptr) { uint64_t val = 0; char c; int digit; while ((c = *str) != '\0') { if (c >= '0' && c <= '9') digit = c - '0'; else if (c >= 'a' && c <= 'f') digit = 10 + c - 'a'; else break; val *= 16; val += digit; str++; } if (nptr) *nptr = (char *)str; return (val); } void spa_activate_allocation_classes(spa_t *spa, dmu_tx_t *tx) { /* * We bump the feature refcount for each special vdev added to the pool */ ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)); spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx); } /* * ========================================================================== * Accessor functions * ========================================================================== */ boolean_t spa_shutting_down(spa_t *spa) { return (spa->spa_async_suspended); } dsl_pool_t * spa_get_dsl(spa_t *spa) { return (spa->spa_dsl_pool); } boolean_t spa_is_initializing(spa_t *spa) { return (spa->spa_is_initializing); } boolean_t spa_indirect_vdevs_loaded(spa_t *spa) { return (spa->spa_indirect_vdevs_loaded); } blkptr_t * spa_get_rootblkptr(spa_t *spa) { return (&spa->spa_ubsync.ub_rootbp); } void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) { spa->spa_uberblock.ub_rootbp = *bp; } void spa_altroot(spa_t *spa, char *buf, size_t buflen) { if (spa->spa_root == NULL) buf[0] = '\0'; else (void) strlcpy(buf, spa->spa_root, buflen); } uint32_t spa_sync_pass(spa_t *spa) { return (spa->spa_sync_pass); } char * spa_name(spa_t *spa) { return (spa->spa_name); } uint64_t spa_guid(spa_t *spa) { dsl_pool_t *dp = spa_get_dsl(spa); uint64_t guid; /* * If we fail to parse the config during spa_load(), we can go through * the error path (which posts an ereport) and end up here with no root * vdev. We stash the original pool guid in 'spa_config_guid' to handle * this case. */ if (spa->spa_root_vdev == NULL) return (spa->spa_config_guid); guid = spa->spa_last_synced_guid != 0 ? spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; /* * Return the most recently synced out guid unless we're * in syncing context. */ if (dp && dsl_pool_sync_context(dp)) return (spa->spa_root_vdev->vdev_guid); else return (guid); } uint64_t spa_load_guid(spa_t *spa) { /* * This is a GUID that exists solely as a reference for the * purposes of the arc. It is generated at load time, and * is never written to persistent storage. */ return (spa->spa_load_guid); } uint64_t spa_last_synced_txg(spa_t *spa) { return (spa->spa_ubsync.ub_txg); } uint64_t spa_first_txg(spa_t *spa) { return (spa->spa_first_txg); } uint64_t spa_syncing_txg(spa_t *spa) { return (spa->spa_syncing_txg); } /* * Return the last txg where data can be dirtied. The final txgs * will be used to just clear out any deferred frees that remain. */ uint64_t spa_final_dirty_txg(spa_t *spa) { return (spa->spa_final_txg - TXG_DEFER_SIZE); } pool_state_t spa_state(spa_t *spa) { return (spa->spa_state); } spa_load_state_t spa_load_state(spa_t *spa) { return (spa->spa_load_state); } uint64_t spa_freeze_txg(spa_t *spa) { return (spa->spa_freeze_txg); } /* * Return the inflated asize for a logical write in bytes. This is used by the * DMU to calculate the space a logical write will require on disk. * If lsize is smaller than the largest physical block size allocatable on this * pool we use its value instead, since the write will end up using the whole * block anyway. */ uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize) { if (lsize == 0) return (0); /* No inflation needed */ return (MAX(lsize, 1 << spa->spa_max_ashift) * spa_asize_inflation); } /* * Return the amount of slop space in bytes. It is typically 1/32 of the pool * (3.2%), minus the embedded log space. On very small pools, it may be * slightly larger than this. On very large pools, it will be capped to * the value of spa_max_slop. The embedded log space is not included in * spa_dspace. By subtracting it, the usable space (per "zfs list") is a * constant 97% of the total space, regardless of metaslab size (assuming the * default spa_slop_shift=5 and a non-tiny pool). * * See the comment above spa_slop_shift for more details. */ uint64_t spa_get_slop_space(spa_t *spa) { uint64_t space = 0; uint64_t slop = 0; /* * Make sure spa_dedup_dspace has been set. */ if (spa->spa_dedup_dspace == ~0ULL) spa_update_dspace(spa); /* * spa_get_dspace() includes the space only logically "used" by * deduplicated data, so since it's not useful to reserve more * space with more deduplicated data, we subtract that out here. */ space = spa_get_dspace(spa) - spa->spa_dedup_dspace - brt_get_dspace(spa); slop = MIN(space >> spa_slop_shift, spa_max_slop); /* * Subtract the embedded log space, but no more than half the (3.2%) * unusable space. Note, the "no more than half" is only relevant if * zfs_embedded_slog_min_ms >> spa_slop_shift < 2, which is not true by * default. */ uint64_t embedded_log = metaslab_class_get_dspace(spa_embedded_log_class(spa)); slop -= MIN(embedded_log, slop >> 1); /* * Slop space should be at least spa_min_slop, but no more than half * the entire pool. */ slop = MAX(slop, MIN(space >> 1, spa_min_slop)); return (slop); } uint64_t spa_get_dspace(spa_t *spa) { return (spa->spa_dspace); } uint64_t spa_get_checkpoint_space(spa_t *spa) { return (spa->spa_checkpoint_info.sci_dspace); } void spa_update_dspace(spa_t *spa) { spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + ddt_get_dedup_dspace(spa) + brt_get_dspace(spa); if (spa->spa_nonallocating_dspace > 0) { /* * Subtract the space provided by all non-allocating vdevs that * contribute to dspace. If a file is overwritten, its old * blocks are freed and new blocks are allocated. If there are * no snapshots of the file, the available space should remain * the same. The old blocks could be freed from the * non-allocating vdev, but the new blocks must be allocated on * other (allocating) vdevs. By reserving the entire size of * the non-allocating vdevs (including allocated space), we * ensure that there will be enough space on the allocating * vdevs for this file overwrite to succeed. * * Note that the DMU/DSL doesn't actually know or care * how much space is allocated (it does its own tracking * of how much space has been logically used). So it * doesn't matter that the data we are moving may be * allocated twice (on the old device and the new device). */ ASSERT3U(spa->spa_dspace, >=, spa->spa_nonallocating_dspace); spa->spa_dspace -= spa->spa_nonallocating_dspace; } } /* * Return the failure mode that has been set to this pool. The default * behavior will be to block all I/Os when a complete failure occurs. */ uint64_t spa_get_failmode(spa_t *spa) { return (spa->spa_failmode); } boolean_t spa_suspended(spa_t *spa) { return (spa->spa_suspended != ZIO_SUSPEND_NONE); } uint64_t spa_version(spa_t *spa) { return (spa->spa_ubsync.ub_version); } boolean_t spa_deflate(spa_t *spa) { return (spa->spa_deflate); } metaslab_class_t * spa_normal_class(spa_t *spa) { return (spa->spa_normal_class); } metaslab_class_t * spa_log_class(spa_t *spa) { return (spa->spa_log_class); } metaslab_class_t * spa_embedded_log_class(spa_t *spa) { return (spa->spa_embedded_log_class); } metaslab_class_t * spa_special_class(spa_t *spa) { return (spa->spa_special_class); } metaslab_class_t * spa_dedup_class(spa_t *spa) { return (spa->spa_dedup_class); } /* * Locate an appropriate allocation class */ metaslab_class_t * spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype, uint_t level, uint_t special_smallblk) { /* * ZIL allocations determine their class in zio_alloc_zil(). */ ASSERT(objtype != DMU_OT_INTENT_LOG); boolean_t has_special_class = spa->spa_special_class->mc_groups != 0; if (DMU_OT_IS_DDT(objtype)) { if (spa->spa_dedup_class->mc_groups != 0) return (spa_dedup_class(spa)); else if (has_special_class && zfs_ddt_data_is_special) return (spa_special_class(spa)); else return (spa_normal_class(spa)); } /* Indirect blocks for user data can land in special if allowed */ if (level > 0 && (DMU_OT_IS_FILE(objtype) || objtype == DMU_OT_ZVOL)) { if (has_special_class && zfs_user_indirect_is_special) return (spa_special_class(spa)); else return (spa_normal_class(spa)); } if (DMU_OT_IS_METADATA(objtype) || level > 0) { if (has_special_class) return (spa_special_class(spa)); else return (spa_normal_class(spa)); } /* * Allow small file blocks in special class in some cases (like * for the dRAID vdev feature). But always leave a reserve of * zfs_special_class_metadata_reserve_pct exclusively for metadata. */ if (DMU_OT_IS_FILE(objtype) && has_special_class && size <= special_smallblk) { metaslab_class_t *special = spa_special_class(spa); uint64_t alloc = metaslab_class_get_alloc(special); uint64_t space = metaslab_class_get_space(special); uint64_t limit = (space * (100 - zfs_special_class_metadata_reserve_pct)) / 100; if (alloc < limit) return (special); } return (spa_normal_class(spa)); } void spa_evicting_os_register(spa_t *spa, objset_t *os) { mutex_enter(&spa->spa_evicting_os_lock); list_insert_head(&spa->spa_evicting_os_list, os); mutex_exit(&spa->spa_evicting_os_lock); } void spa_evicting_os_deregister(spa_t *spa, objset_t *os) { mutex_enter(&spa->spa_evicting_os_lock); list_remove(&spa->spa_evicting_os_list, os); cv_broadcast(&spa->spa_evicting_os_cv); mutex_exit(&spa->spa_evicting_os_lock); } void spa_evicting_os_wait(spa_t *spa) { mutex_enter(&spa->spa_evicting_os_lock); while (!list_is_empty(&spa->spa_evicting_os_list)) cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); mutex_exit(&spa->spa_evicting_os_lock); dmu_buf_user_evict_wait(); } int spa_max_replication(spa_t *spa) { /* * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to * handle BPs with more than one DVA allocated. Set our max * replication level accordingly. */ if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) return (1); return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); } int spa_prev_software_version(spa_t *spa) { return (spa->spa_prev_software_version); } uint64_t spa_deadman_synctime(spa_t *spa) { return (spa->spa_deadman_synctime); } spa_autotrim_t spa_get_autotrim(spa_t *spa) { return (spa->spa_autotrim); } uint64_t spa_deadman_ziotime(spa_t *spa) { return (spa->spa_deadman_ziotime); } uint64_t spa_get_deadman_failmode(spa_t *spa) { return (spa->spa_deadman_failmode); } void spa_set_deadman_failmode(spa_t *spa, const char *failmode) { if (strcmp(failmode, "wait") == 0) spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT; else if (strcmp(failmode, "continue") == 0) spa->spa_deadman_failmode = ZIO_FAILURE_MODE_CONTINUE; else if (strcmp(failmode, "panic") == 0) spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; else spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT; } void spa_set_deadman_ziotime(hrtime_t ns) { spa_t *spa = NULL; if (spa_mode_global != SPA_MODE_UNINIT) { mutex_enter(&spa_namespace_lock); while ((spa = spa_next(spa)) != NULL) spa->spa_deadman_ziotime = ns; mutex_exit(&spa_namespace_lock); } } void spa_set_deadman_synctime(hrtime_t ns) { spa_t *spa = NULL; if (spa_mode_global != SPA_MODE_UNINIT) { mutex_enter(&spa_namespace_lock); while ((spa = spa_next(spa)) != NULL) spa->spa_deadman_synctime = ns; mutex_exit(&spa_namespace_lock); } } uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva) { uint64_t asize = DVA_GET_ASIZE(dva); uint64_t dsize = asize; ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); if (asize != 0 && spa->spa_deflate) { vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); if (vd != NULL) dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; } return (dsize); } uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) { uint64_t dsize = 0; for (int d = 0; d < BP_GET_NDVAS(bp); d++) dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); return (dsize); } uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp) { uint64_t dsize = 0; spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); for (int d = 0; d < BP_GET_NDVAS(bp); d++) dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); spa_config_exit(spa, SCL_VDEV, FTAG); return (dsize); } uint64_t spa_dirty_data(spa_t *spa) { return (spa->spa_dsl_pool->dp_dirty_total); } /* * ========================================================================== * SPA Import Progress Routines * ========================================================================== */ typedef struct spa_import_progress { uint64_t pool_guid; /* unique id for updates */ char *pool_name; spa_load_state_t spa_load_state; char *spa_load_notes; uint64_t mmp_sec_remaining; /* MMP activity check */ uint64_t spa_load_max_txg; /* rewind txg */ procfs_list_node_t smh_node; } spa_import_progress_t; spa_history_list_t *spa_import_progress_list = NULL; static int spa_import_progress_show_header(struct seq_file *f) { seq_printf(f, "%-20s %-14s %-14s %-12s %-16s %s\n", "pool_guid", "load_state", "multihost_secs", "max_txg", "pool_name", "notes"); return (0); } static int spa_import_progress_show(struct seq_file *f, void *data) { spa_import_progress_t *sip = (spa_import_progress_t *)data; seq_printf(f, "%-20llu %-14llu %-14llu %-12llu %-16s %s\n", (u_longlong_t)sip->pool_guid, (u_longlong_t)sip->spa_load_state, (u_longlong_t)sip->mmp_sec_remaining, (u_longlong_t)sip->spa_load_max_txg, (sip->pool_name ? sip->pool_name : "-"), (sip->spa_load_notes ? sip->spa_load_notes : "-")); return (0); } /* Remove oldest elements from list until there are no more than 'size' left */ static void spa_import_progress_truncate(spa_history_list_t *shl, unsigned int size) { spa_import_progress_t *sip; while (shl->size > size) { sip = list_remove_head(&shl->procfs_list.pl_list); if (sip->pool_name) spa_strfree(sip->pool_name); if (sip->spa_load_notes) kmem_strfree(sip->spa_load_notes); kmem_free(sip, sizeof (spa_import_progress_t)); shl->size--; } IMPLY(size == 0, list_is_empty(&shl->procfs_list.pl_list)); } static void spa_import_progress_init(void) { spa_import_progress_list = kmem_zalloc(sizeof (spa_history_list_t), KM_SLEEP); spa_import_progress_list->size = 0; spa_import_progress_list->procfs_list.pl_private = spa_import_progress_list; procfs_list_install("zfs", NULL, "import_progress", 0644, &spa_import_progress_list->procfs_list, spa_import_progress_show, spa_import_progress_show_header, NULL, offsetof(spa_import_progress_t, smh_node)); } static void spa_import_progress_destroy(void) { spa_history_list_t *shl = spa_import_progress_list; procfs_list_uninstall(&shl->procfs_list); spa_import_progress_truncate(shl, 0); procfs_list_destroy(&shl->procfs_list); kmem_free(shl, sizeof (spa_history_list_t)); } int spa_import_progress_set_state(uint64_t pool_guid, spa_load_state_t load_state) { spa_history_list_t *shl = spa_import_progress_list; spa_import_progress_t *sip; int error = ENOENT; if (shl->size == 0) return (0); mutex_enter(&shl->procfs_list.pl_lock); for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; sip = list_prev(&shl->procfs_list.pl_list, sip)) { if (sip->pool_guid == pool_guid) { sip->spa_load_state = load_state; if (sip->spa_load_notes != NULL) { kmem_strfree(sip->spa_load_notes); sip->spa_load_notes = NULL; } error = 0; break; } } mutex_exit(&shl->procfs_list.pl_lock); return (error); } static void spa_import_progress_set_notes_impl(spa_t *spa, boolean_t log_dbgmsg, const char *fmt, va_list adx) { spa_history_list_t *shl = spa_import_progress_list; spa_import_progress_t *sip; uint64_t pool_guid = spa_guid(spa); if (shl->size == 0) return; char *notes = kmem_vasprintf(fmt, adx); mutex_enter(&shl->procfs_list.pl_lock); for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; sip = list_prev(&shl->procfs_list.pl_list, sip)) { if (sip->pool_guid == pool_guid) { if (sip->spa_load_notes != NULL) { kmem_strfree(sip->spa_load_notes); sip->spa_load_notes = NULL; } sip->spa_load_notes = notes; if (log_dbgmsg) zfs_dbgmsg("'%s' %s", sip->pool_name, notes); notes = NULL; break; } } mutex_exit(&shl->procfs_list.pl_lock); if (notes != NULL) kmem_strfree(notes); } void spa_import_progress_set_notes(spa_t *spa, const char *fmt, ...) { va_list adx; va_start(adx, fmt); spa_import_progress_set_notes_impl(spa, B_TRUE, fmt, adx); va_end(adx); } void spa_import_progress_set_notes_nolog(spa_t *spa, const char *fmt, ...) { va_list adx; va_start(adx, fmt); spa_import_progress_set_notes_impl(spa, B_FALSE, fmt, adx); va_end(adx); } int spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t load_max_txg) { spa_history_list_t *shl = spa_import_progress_list; spa_import_progress_t *sip; int error = ENOENT; if (shl->size == 0) return (0); mutex_enter(&shl->procfs_list.pl_lock); for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; sip = list_prev(&shl->procfs_list.pl_list, sip)) { if (sip->pool_guid == pool_guid) { sip->spa_load_max_txg = load_max_txg; error = 0; break; } } mutex_exit(&shl->procfs_list.pl_lock); return (error); } int spa_import_progress_set_mmp_check(uint64_t pool_guid, uint64_t mmp_sec_remaining) { spa_history_list_t *shl = spa_import_progress_list; spa_import_progress_t *sip; int error = ENOENT; if (shl->size == 0) return (0); mutex_enter(&shl->procfs_list.pl_lock); for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; sip = list_prev(&shl->procfs_list.pl_list, sip)) { if (sip->pool_guid == pool_guid) { sip->mmp_sec_remaining = mmp_sec_remaining; error = 0; break; } } mutex_exit(&shl->procfs_list.pl_lock); return (error); } /* * A new import is in progress, add an entry. */ void spa_import_progress_add(spa_t *spa) { spa_history_list_t *shl = spa_import_progress_list; spa_import_progress_t *sip; const char *poolname = NULL; sip = kmem_zalloc(sizeof (spa_import_progress_t), KM_SLEEP); sip->pool_guid = spa_guid(spa); (void) nvlist_lookup_string(spa->spa_config, ZPOOL_CONFIG_POOL_NAME, &poolname); if (poolname == NULL) poolname = spa_name(spa); sip->pool_name = spa_strdup(poolname); sip->spa_load_state = spa_load_state(spa); sip->spa_load_notes = NULL; mutex_enter(&shl->procfs_list.pl_lock); procfs_list_add(&shl->procfs_list, sip); shl->size++; mutex_exit(&shl->procfs_list.pl_lock); } void spa_import_progress_remove(uint64_t pool_guid) { spa_history_list_t *shl = spa_import_progress_list; spa_import_progress_t *sip; mutex_enter(&shl->procfs_list.pl_lock); for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; sip = list_prev(&shl->procfs_list.pl_list, sip)) { if (sip->pool_guid == pool_guid) { if (sip->pool_name) spa_strfree(sip->pool_name); if (sip->spa_load_notes) spa_strfree(sip->spa_load_notes); list_remove(&shl->procfs_list.pl_list, sip); shl->size--; kmem_free(sip, sizeof (spa_import_progress_t)); break; } } mutex_exit(&shl->procfs_list.pl_lock); } /* * ========================================================================== * Initialization and Termination * ========================================================================== */ static int spa_name_compare(const void *a1, const void *a2) { const spa_t *s1 = a1; const spa_t *s2 = a2; int s; s = strcmp(s1->spa_name, s2->spa_name); return (TREE_ISIGN(s)); } void spa_boot_init(void) { spa_config_load(); } void spa_init(spa_mode_t mode) { mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), offsetof(spa_t, spa_avl)); avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), offsetof(spa_aux_t, aux_avl)); avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), offsetof(spa_aux_t, aux_avl)); spa_mode_global = mode; #ifndef _KERNEL if (spa_mode_global != SPA_MODE_READ && dprintf_find_string("watch")) { struct sigaction sa; sa.sa_flags = SA_SIGINFO; sigemptyset(&sa.sa_mask); sa.sa_sigaction = arc_buf_sigsegv; if (sigaction(SIGSEGV, &sa, NULL) == -1) { perror("could not enable watchpoints: " "sigaction(SIGSEGV, ...) = "); } else { arc_watch = B_TRUE; } } #endif fm_init(); zfs_refcount_init(); unique_init(); zfs_btree_init(); metaslab_stat_init(); brt_init(); ddt_init(); zio_init(); dmu_init(); zil_init(); vdev_mirror_stat_init(); vdev_raidz_math_init(); vdev_file_init(); zfs_prop_init(); chksum_init(); zpool_prop_init(); zpool_feature_init(); spa_config_load(); vdev_prop_init(); l2arc_start(); scan_init(); qat_init(); spa_import_progress_init(); } void spa_fini(void) { l2arc_stop(); spa_evict_all(); vdev_file_fini(); vdev_mirror_stat_fini(); vdev_raidz_math_fini(); chksum_fini(); zil_fini(); dmu_fini(); zio_fini(); ddt_fini(); brt_fini(); metaslab_stat_fini(); zfs_btree_fini(); unique_fini(); zfs_refcount_fini(); fm_fini(); scan_fini(); qat_fini(); spa_import_progress_destroy(); avl_destroy(&spa_namespace_avl); avl_destroy(&spa_spare_avl); avl_destroy(&spa_l2cache_avl); cv_destroy(&spa_namespace_cv); mutex_destroy(&spa_namespace_lock); mutex_destroy(&spa_spare_lock); mutex_destroy(&spa_l2cache_lock); } /* * Return whether this pool has a dedicated slog device. No locking needed. * It's not a problem if the wrong answer is returned as it's only for * performance and not correctness. */ boolean_t spa_has_slogs(spa_t *spa) { return (spa->spa_log_class->mc_groups != 0); } spa_log_state_t spa_get_log_state(spa_t *spa) { return (spa->spa_log_state); } void spa_set_log_state(spa_t *spa, spa_log_state_t state) { spa->spa_log_state = state; } boolean_t spa_is_root(spa_t *spa) { return (spa->spa_is_root); } boolean_t spa_writeable(spa_t *spa) { return (!!(spa->spa_mode & SPA_MODE_WRITE) && spa->spa_trust_config); } /* * Returns true if there is a pending sync task in any of the current * syncing txg, the current quiescing txg, or the current open txg. */ boolean_t spa_has_pending_synctask(spa_t *spa) { return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) || !txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks)); } spa_mode_t spa_mode(spa_t *spa) { return (spa->spa_mode); } uint64_t spa_bootfs(spa_t *spa) { return (spa->spa_bootfs); } uint64_t spa_delegation(spa_t *spa) { return (spa->spa_delegation); } objset_t * spa_meta_objset(spa_t *spa) { return (spa->spa_meta_objset); } enum zio_checksum spa_dedup_checksum(spa_t *spa) { return (spa->spa_dedup_checksum); } /* * Reset pool scan stat per scan pass (or reboot). */ void spa_scan_stat_init(spa_t *spa) { /* data not stored on disk */ spa->spa_scan_pass_start = gethrestime_sec(); if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan)) spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start; else spa->spa_scan_pass_scrub_pause = 0; if (dsl_errorscrub_is_paused(spa->spa_dsl_pool->dp_scan)) spa->spa_scan_pass_errorscrub_pause = spa->spa_scan_pass_start; else spa->spa_scan_pass_errorscrub_pause = 0; spa->spa_scan_pass_scrub_spent_paused = 0; spa->spa_scan_pass_exam = 0; spa->spa_scan_pass_issued = 0; // error scrub stats spa->spa_scan_pass_errorscrub_spent_paused = 0; } /* * Get scan stats for zpool status reports */ int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) { dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; if (scn == NULL || (scn->scn_phys.scn_func == POOL_SCAN_NONE && scn->errorscrub_phys.dep_func == POOL_SCAN_NONE)) return (SET_ERROR(ENOENT)); memset(ps, 0, sizeof (pool_scan_stat_t)); /* data stored on disk */ ps->pss_func = scn->scn_phys.scn_func; ps->pss_state = scn->scn_phys.scn_state; ps->pss_start_time = scn->scn_phys.scn_start_time; ps->pss_end_time = scn->scn_phys.scn_end_time; ps->pss_to_examine = scn->scn_phys.scn_to_examine; ps->pss_examined = scn->scn_phys.scn_examined; ps->pss_skipped = scn->scn_phys.scn_skipped; ps->pss_processed = scn->scn_phys.scn_processed; ps->pss_errors = scn->scn_phys.scn_errors; /* data not stored on disk */ ps->pss_pass_exam = spa->spa_scan_pass_exam; ps->pss_pass_start = spa->spa_scan_pass_start; ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause; ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused; ps->pss_pass_issued = spa->spa_scan_pass_issued; ps->pss_issued = scn->scn_issued_before_pass + spa->spa_scan_pass_issued; /* error scrub data stored on disk */ ps->pss_error_scrub_func = scn->errorscrub_phys.dep_func; ps->pss_error_scrub_state = scn->errorscrub_phys.dep_state; ps->pss_error_scrub_start = scn->errorscrub_phys.dep_start_time; ps->pss_error_scrub_end = scn->errorscrub_phys.dep_end_time; ps->pss_error_scrub_examined = scn->errorscrub_phys.dep_examined; ps->pss_error_scrub_to_be_examined = scn->errorscrub_phys.dep_to_examine; /* error scrub data not stored on disk */ ps->pss_pass_error_scrub_pause = spa->spa_scan_pass_errorscrub_pause; return (0); } int spa_maxblocksize(spa_t *spa) { if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) return (SPA_MAXBLOCKSIZE); else return (SPA_OLD_MAXBLOCKSIZE); } /* * Returns the txg that the last device removal completed. No indirect mappings * have been added since this txg. */ uint64_t spa_get_last_removal_txg(spa_t *spa) { uint64_t vdevid; uint64_t ret = -1ULL; spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); /* * sr_prev_indirect_vdev is only modified while holding all the * config locks, so it is sufficient to hold SCL_VDEV as reader when * examining it. */ vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev; while (vdevid != -1ULL) { vdev_t *vd = vdev_lookup_top(spa, vdevid); vdev_indirect_births_t *vib = vd->vdev_indirect_births; ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops); /* * If the removal did not remap any data, we don't care. */ if (vdev_indirect_births_count(vib) != 0) { ret = vdev_indirect_births_last_entry_txg(vib); break; } vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev; } spa_config_exit(spa, SCL_VDEV, FTAG); IMPLY(ret != -1ULL, spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); return (ret); } int spa_maxdnodesize(spa_t *spa) { if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) return (DNODE_MAX_SIZE); else return (DNODE_MIN_SIZE); } boolean_t spa_multihost(spa_t *spa) { return (spa->spa_multihost ? B_TRUE : B_FALSE); } uint32_t spa_get_hostid(spa_t *spa) { return (spa->spa_hostid); } boolean_t spa_trust_config(spa_t *spa) { return (spa->spa_trust_config); } uint64_t spa_missing_tvds_allowed(spa_t *spa) { return (spa->spa_missing_tvds_allowed); } space_map_t * spa_syncing_log_sm(spa_t *spa) { return (spa->spa_syncing_log_sm); } void spa_set_missing_tvds(spa_t *spa, uint64_t missing) { spa->spa_missing_tvds = missing; } /* * Return the pool state string ("ONLINE", "DEGRADED", "SUSPENDED", etc). */ const char * spa_state_to_name(spa_t *spa) { ASSERT3P(spa, !=, NULL); /* * it is possible for the spa to exist, without root vdev * as the spa transitions during import/export */ vdev_t *rvd = spa->spa_root_vdev; if (rvd == NULL) { return ("TRANSITIONING"); } vdev_state_t state = rvd->vdev_state; vdev_aux_t aux = rvd->vdev_stat.vs_aux; if (spa_suspended(spa)) return ("SUSPENDED"); switch (state) { case VDEV_STATE_CLOSED: case VDEV_STATE_OFFLINE: return ("OFFLINE"); case VDEV_STATE_REMOVED: return ("REMOVED"); case VDEV_STATE_CANT_OPEN: if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) return ("FAULTED"); else if (aux == VDEV_AUX_SPLIT_POOL) return ("SPLIT"); else return ("UNAVAIL"); case VDEV_STATE_FAULTED: return ("FAULTED"); case VDEV_STATE_DEGRADED: return ("DEGRADED"); case VDEV_STATE_HEALTHY: return ("ONLINE"); default: break; } return ("UNKNOWN"); } boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; for (uint64_t c = 0; c < rvd->vdev_children; c++) { if (!vdev_is_spacemap_addressable(rvd->vdev_child[c])) return (B_FALSE); } return (B_TRUE); } boolean_t spa_has_checkpoint(spa_t *spa) { return (spa->spa_checkpoint_txg != 0); } boolean_t spa_importing_readonly_checkpoint(spa_t *spa) { return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) && spa->spa_mode == SPA_MODE_READ); } uint64_t spa_min_claim_txg(spa_t *spa) { uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg; if (checkpoint_txg != 0) return (checkpoint_txg + 1); return (spa->spa_first_txg); } /* * If there is a checkpoint, async destroys may consume more space from * the pool instead of freeing it. In an attempt to save the pool from * getting suspended when it is about to run out of space, we stop * processing async destroys. */ boolean_t spa_suspend_async_destroy(spa_t *spa) { dsl_pool_t *dp = spa_get_dsl(spa); uint64_t unreserved = dsl_pool_unreserved_space(dp, ZFS_SPACE_CHECK_EXTRA_RESERVED); uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes; uint64_t avail = (unreserved > used) ? (unreserved - used) : 0; if (spa_has_checkpoint(spa) && avail == 0) return (B_TRUE); return (B_FALSE); } #if defined(_KERNEL) int param_set_deadman_failmode_common(const char *val) { spa_t *spa = NULL; char *p; if (val == NULL) return (SET_ERROR(EINVAL)); if ((p = strchr(val, '\n')) != NULL) *p = '\0'; if (strcmp(val, "wait") != 0 && strcmp(val, "continue") != 0 && strcmp(val, "panic")) return (SET_ERROR(EINVAL)); if (spa_mode_global != SPA_MODE_UNINIT) { mutex_enter(&spa_namespace_lock); while ((spa = spa_next(spa)) != NULL) spa_set_deadman_failmode(spa, val); mutex_exit(&spa_namespace_lock); } return (0); } #endif /* Namespace manipulation */ EXPORT_SYMBOL(spa_lookup); EXPORT_SYMBOL(spa_add); EXPORT_SYMBOL(spa_remove); EXPORT_SYMBOL(spa_next); /* Refcount functions */ EXPORT_SYMBOL(spa_open_ref); EXPORT_SYMBOL(spa_close); EXPORT_SYMBOL(spa_refcount_zero); /* Pool configuration lock */ EXPORT_SYMBOL(spa_config_tryenter); EXPORT_SYMBOL(spa_config_enter); EXPORT_SYMBOL(spa_config_exit); EXPORT_SYMBOL(spa_config_held); /* Pool vdev add/remove lock */ EXPORT_SYMBOL(spa_vdev_enter); EXPORT_SYMBOL(spa_vdev_exit); /* Pool vdev state change lock */ EXPORT_SYMBOL(spa_vdev_state_enter); EXPORT_SYMBOL(spa_vdev_state_exit); /* Accessor functions */ EXPORT_SYMBOL(spa_shutting_down); EXPORT_SYMBOL(spa_get_dsl); EXPORT_SYMBOL(spa_get_rootblkptr); EXPORT_SYMBOL(spa_set_rootblkptr); EXPORT_SYMBOL(spa_altroot); EXPORT_SYMBOL(spa_sync_pass); EXPORT_SYMBOL(spa_name); EXPORT_SYMBOL(spa_guid); EXPORT_SYMBOL(spa_last_synced_txg); EXPORT_SYMBOL(spa_first_txg); EXPORT_SYMBOL(spa_syncing_txg); EXPORT_SYMBOL(spa_version); EXPORT_SYMBOL(spa_state); EXPORT_SYMBOL(spa_load_state); EXPORT_SYMBOL(spa_freeze_txg); EXPORT_SYMBOL(spa_get_dspace); EXPORT_SYMBOL(spa_update_dspace); EXPORT_SYMBOL(spa_deflate); EXPORT_SYMBOL(spa_normal_class); EXPORT_SYMBOL(spa_log_class); EXPORT_SYMBOL(spa_special_class); EXPORT_SYMBOL(spa_preferred_class); EXPORT_SYMBOL(spa_max_replication); EXPORT_SYMBOL(spa_prev_software_version); EXPORT_SYMBOL(spa_get_failmode); EXPORT_SYMBOL(spa_suspended); EXPORT_SYMBOL(spa_bootfs); EXPORT_SYMBOL(spa_delegation); EXPORT_SYMBOL(spa_meta_objset); EXPORT_SYMBOL(spa_maxblocksize); EXPORT_SYMBOL(spa_maxdnodesize); /* Miscellaneous support routines */ EXPORT_SYMBOL(spa_guid_exists); EXPORT_SYMBOL(spa_strdup); EXPORT_SYMBOL(spa_strfree); EXPORT_SYMBOL(spa_generate_guid); EXPORT_SYMBOL(snprintf_blkptr); EXPORT_SYMBOL(spa_freeze); EXPORT_SYMBOL(spa_upgrade); EXPORT_SYMBOL(spa_evict_all); EXPORT_SYMBOL(spa_lookup_by_guid); EXPORT_SYMBOL(spa_has_spare); EXPORT_SYMBOL(dva_get_dsize_sync); EXPORT_SYMBOL(bp_get_dsize_sync); EXPORT_SYMBOL(bp_get_dsize); EXPORT_SYMBOL(spa_has_slogs); EXPORT_SYMBOL(spa_is_root); EXPORT_SYMBOL(spa_writeable); EXPORT_SYMBOL(spa_mode); EXPORT_SYMBOL(spa_namespace_lock); EXPORT_SYMBOL(spa_trust_config); EXPORT_SYMBOL(spa_missing_tvds_allowed); EXPORT_SYMBOL(spa_set_missing_tvds); EXPORT_SYMBOL(spa_state_to_name); EXPORT_SYMBOL(spa_importing_readonly_checkpoint); EXPORT_SYMBOL(spa_min_claim_txg); EXPORT_SYMBOL(spa_suspend_async_destroy); EXPORT_SYMBOL(spa_has_checkpoint); EXPORT_SYMBOL(spa_top_vdevs_spacemap_addressable); ZFS_MODULE_PARAM(zfs, zfs_, flags, UINT, ZMOD_RW, "Set additional debugging flags"); ZFS_MODULE_PARAM(zfs, zfs_, recover, INT, ZMOD_RW, "Set to attempt to recover from fatal errors"); ZFS_MODULE_PARAM(zfs, zfs_, free_leak_on_eio, INT, ZMOD_RW, "Set to ignore IO errors during free and permanently leak the space"); ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, checktime_ms, U64, ZMOD_RW, "Dead I/O check interval in milliseconds"); ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, enabled, INT, ZMOD_RW, "Enable deadman timer"); ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, UINT, ZMOD_RW, "SPA size estimate multiplication factor"); ZFS_MODULE_PARAM(zfs, zfs_, ddt_data_is_special, INT, ZMOD_RW, "Place DDT data into the special class"); ZFS_MODULE_PARAM(zfs, zfs_, user_indirect_is_special, INT, ZMOD_RW, "Place user data indirect blocks into the special class"); /* BEGIN CSTYLED */ ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, failmode, param_set_deadman_failmode, param_get_charp, ZMOD_RW, "Failmode for deadman timer"); ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, synctime_ms, param_set_deadman_synctime, spl_param_get_u64, ZMOD_RW, "Pool sync expiration time in milliseconds"); ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, ziotime_ms, param_set_deadman_ziotime, spl_param_get_u64, ZMOD_RW, "IO expiration time in milliseconds"); ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, UINT, ZMOD_RW, "Small file blocks in special vdevs depends on this much " "free space available"); /* END CSTYLED */ ZFS_MODULE_PARAM_CALL(zfs_spa, spa_, slop_shift, param_set_slop_shift, param_get_uint, ZMOD_RW, "Reserved free space in pool"); ZFS_MODULE_PARAM(zfs, spa_, num_allocators, INT, ZMOD_RW, "Number of allocators per spa, capped by ncpus"); diff --git a/sys/contrib/openzfs/module/zfs/vdev_initialize.c b/sys/contrib/openzfs/module/zfs/vdev_initialize.c index 5aaef1a69986..c5e16af16692 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_initialize.c +++ b/sys/contrib/openzfs/module/zfs/vdev_initialize.c @@ -1,828 +1,829 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* - * Copyright (c) 2016, 2019 by Delphix. All rights reserved. + * Copyright (c) 2016, 2024 by Delphix. All rights reserved. */ #include #include #include #include #include #include #include #include #include /* * Value that is written to disk during initialization. */ static uint64_t zfs_initialize_value = 0xdeadbeefdeadbeeeULL; /* maximum number of I/Os outstanding per leaf vdev */ static const int zfs_initialize_limit = 1; /* size of initializing writes; default 1MiB, see zfs_remove_max_segment */ static uint64_t zfs_initialize_chunk_size = 1024 * 1024; static boolean_t vdev_initialize_should_stop(vdev_t *vd) { return (vd->vdev_initialize_exit_wanted || !vdev_writeable(vd) || vd->vdev_detached || vd->vdev_top->vdev_removing || vd->vdev_top->vdev_rz_expanding); } static void vdev_initialize_zap_update_sync(void *arg, dmu_tx_t *tx) { /* * We pass in the guid instead of the vdev_t since the vdev may * have been freed prior to the sync task being processed. This * happens when a vdev is detached as we call spa_config_vdev_exit(), * stop the initializing thread, schedule the sync task, and free * the vdev. Later when the scheduled sync task is invoked, it would * find that the vdev has been freed. */ uint64_t guid = *(uint64_t *)arg; uint64_t txg = dmu_tx_get_txg(tx); kmem_free(arg, sizeof (uint64_t)); vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE); if (vd == NULL || vd->vdev_top->vdev_removing || !vdev_is_concrete(vd) || vd->vdev_top->vdev_rz_expanding) return; uint64_t last_offset = vd->vdev_initialize_offset[txg & TXG_MASK]; vd->vdev_initialize_offset[txg & TXG_MASK] = 0; VERIFY(vd->vdev_leaf_zap != 0); objset_t *mos = vd->vdev_spa->spa_meta_objset; if (last_offset > 0) { vd->vdev_initialize_last_offset = last_offset; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET, sizeof (last_offset), 1, &last_offset, tx)); } if (vd->vdev_initialize_action_time > 0) { uint64_t val = (uint64_t)vd->vdev_initialize_action_time; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, sizeof (val), 1, &val, tx)); } uint64_t initialize_state = vd->vdev_initialize_state; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_STATE, sizeof (initialize_state), 1, &initialize_state, tx)); } static void vdev_initialize_zap_remove_sync(void *arg, dmu_tx_t *tx) { uint64_t guid = *(uint64_t *)arg; kmem_free(arg, sizeof (uint64_t)); vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE); if (vd == NULL || vd->vdev_top->vdev_removing || !vdev_is_concrete(vd)) return; ASSERT3S(vd->vdev_initialize_state, ==, VDEV_INITIALIZE_NONE); ASSERT3U(vd->vdev_leaf_zap, !=, 0); vd->vdev_initialize_last_offset = 0; vd->vdev_initialize_action_time = 0; objset_t *mos = vd->vdev_spa->spa_meta_objset; int error; error = zap_remove(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET, tx); VERIFY(error == 0 || error == ENOENT); error = zap_remove(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_STATE, tx); VERIFY(error == 0 || error == ENOENT); error = zap_remove(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, tx); VERIFY(error == 0 || error == ENOENT); } static void vdev_initialize_change_state(vdev_t *vd, vdev_initializing_state_t new_state) { ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); spa_t *spa = vd->vdev_spa; if (new_state == vd->vdev_initialize_state) return; /* * Copy the vd's guid, this will be freed by the sync task. */ uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); *guid = vd->vdev_guid; /* * If we're suspending, then preserving the original start time. */ if (vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED) { vd->vdev_initialize_action_time = gethrestime_sec(); } vdev_initializing_state_t old_state = vd->vdev_initialize_state; vd->vdev_initialize_state = new_state; dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); if (new_state != VDEV_INITIALIZE_NONE) { dsl_sync_task_nowait(spa_get_dsl(spa), vdev_initialize_zap_update_sync, guid, tx); } else { dsl_sync_task_nowait(spa_get_dsl(spa), vdev_initialize_zap_remove_sync, guid, tx); } switch (new_state) { case VDEV_INITIALIZE_ACTIVE: spa_history_log_internal(spa, "initialize", tx, "vdev=%s activated", vd->vdev_path); break; case VDEV_INITIALIZE_SUSPENDED: spa_history_log_internal(spa, "initialize", tx, "vdev=%s suspended", vd->vdev_path); break; case VDEV_INITIALIZE_CANCELED: if (old_state == VDEV_INITIALIZE_ACTIVE || old_state == VDEV_INITIALIZE_SUSPENDED) spa_history_log_internal(spa, "initialize", tx, "vdev=%s canceled", vd->vdev_path); break; case VDEV_INITIALIZE_COMPLETE: spa_history_log_internal(spa, "initialize", tx, "vdev=%s complete", vd->vdev_path); break; case VDEV_INITIALIZE_NONE: spa_history_log_internal(spa, "uninitialize", tx, "vdev=%s", vd->vdev_path); break; default: panic("invalid state %llu", (unsigned long long)new_state); } dmu_tx_commit(tx); if (new_state != VDEV_INITIALIZE_ACTIVE) spa_notify_waiters(spa); } static void vdev_initialize_cb(zio_t *zio) { vdev_t *vd = zio->io_vd; mutex_enter(&vd->vdev_initialize_io_lock); if (zio->io_error == ENXIO && !vdev_writeable(vd)) { /* * The I/O failed because the vdev was unavailable; roll the * last offset back. (This works because spa_sync waits on * spa_txg_zio before it runs sync tasks.) */ uint64_t *off = &vd->vdev_initialize_offset[zio->io_txg & TXG_MASK]; *off = MIN(*off, zio->io_offset); } else { /* * Since initializing is best-effort, we ignore I/O errors and * rely on vdev_probe to determine if the errors are more * critical. */ if (zio->io_error != 0) vd->vdev_stat.vs_initialize_errors++; vd->vdev_initialize_bytes_done += zio->io_orig_size; } ASSERT3U(vd->vdev_initialize_inflight, >, 0); vd->vdev_initialize_inflight--; cv_broadcast(&vd->vdev_initialize_io_cv); mutex_exit(&vd->vdev_initialize_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); } /* Takes care of physical writing and limiting # of concurrent ZIOs. */ static int vdev_initialize_write(vdev_t *vd, uint64_t start, uint64_t size, abd_t *data) { spa_t *spa = vd->vdev_spa; /* Limit inflight initializing I/Os */ mutex_enter(&vd->vdev_initialize_io_lock); while (vd->vdev_initialize_inflight >= zfs_initialize_limit) { cv_wait(&vd->vdev_initialize_io_cv, &vd->vdev_initialize_io_lock); } vd->vdev_initialize_inflight++; mutex_exit(&vd->vdev_initialize_io_lock); dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); uint64_t txg = dmu_tx_get_txg(tx); spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER); mutex_enter(&vd->vdev_initialize_lock); if (vd->vdev_initialize_offset[txg & TXG_MASK] == 0) { uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); *guid = vd->vdev_guid; /* This is the first write of this txg. */ dsl_sync_task_nowait(spa_get_dsl(spa), vdev_initialize_zap_update_sync, guid, tx); } /* * We know the vdev struct will still be around since all * consumers of vdev_free must stop the initialization first. */ if (vdev_initialize_should_stop(vd)) { mutex_enter(&vd->vdev_initialize_io_lock); ASSERT3U(vd->vdev_initialize_inflight, >, 0); vd->vdev_initialize_inflight--; mutex_exit(&vd->vdev_initialize_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); mutex_exit(&vd->vdev_initialize_lock); dmu_tx_commit(tx); return (SET_ERROR(EINTR)); } mutex_exit(&vd->vdev_initialize_lock); vd->vdev_initialize_offset[txg & TXG_MASK] = start + size; zio_nowait(zio_write_phys(spa->spa_txg_zio[txg & TXG_MASK], vd, start, size, data, ZIO_CHECKSUM_OFF, vdev_initialize_cb, NULL, ZIO_PRIORITY_INITIALIZING, ZIO_FLAG_CANFAIL, B_FALSE)); /* vdev_initialize_cb releases SCL_STATE_ALL */ dmu_tx_commit(tx); return (0); } /* * Callback to fill each ABD chunk with zfs_initialize_value. len must be * divisible by sizeof (uint64_t), and buf must be 8-byte aligned. The ABD * allocation will guarantee these for us. */ static int vdev_initialize_block_fill(void *buf, size_t len, void *unused) { (void) unused; ASSERT0(len % sizeof (uint64_t)); for (uint64_t i = 0; i < len; i += sizeof (uint64_t)) { *(uint64_t *)((char *)(buf) + i) = zfs_initialize_value; } return (0); } static abd_t * vdev_initialize_block_alloc(void) { /* Allocate ABD for filler data */ abd_t *data = abd_alloc_for_io(zfs_initialize_chunk_size, B_FALSE); ASSERT0(zfs_initialize_chunk_size % sizeof (uint64_t)); (void) abd_iterate_func(data, 0, zfs_initialize_chunk_size, vdev_initialize_block_fill, NULL); return (data); } static void vdev_initialize_block_free(abd_t *data) { abd_free(data); } static int vdev_initialize_ranges(vdev_t *vd, abd_t *data) { range_tree_t *rt = vd->vdev_initialize_tree; zfs_btree_t *bt = &rt->rt_root; zfs_btree_index_t where; for (range_seg_t *rs = zfs_btree_first(bt, &where); rs != NULL; rs = zfs_btree_next(bt, &where, &where)) { uint64_t size = rs_get_end(rs, rt) - rs_get_start(rs, rt); /* Split range into legally-sized physical chunks */ uint64_t writes_required = ((size - 1) / zfs_initialize_chunk_size) + 1; for (uint64_t w = 0; w < writes_required; w++) { int error; error = vdev_initialize_write(vd, VDEV_LABEL_START_SIZE + rs_get_start(rs, rt) + (w * zfs_initialize_chunk_size), MIN(size - (w * zfs_initialize_chunk_size), zfs_initialize_chunk_size), data); if (error != 0) return (error); } } return (0); } static void vdev_initialize_xlate_last_rs_end(void *arg, range_seg64_t *physical_rs) { uint64_t *last_rs_end = (uint64_t *)arg; if (physical_rs->rs_end > *last_rs_end) *last_rs_end = physical_rs->rs_end; } static void vdev_initialize_xlate_progress(void *arg, range_seg64_t *physical_rs) { vdev_t *vd = (vdev_t *)arg; uint64_t size = physical_rs->rs_end - physical_rs->rs_start; vd->vdev_initialize_bytes_est += size; if (vd->vdev_initialize_last_offset > physical_rs->rs_end) { vd->vdev_initialize_bytes_done += size; } else if (vd->vdev_initialize_last_offset > physical_rs->rs_start && vd->vdev_initialize_last_offset < physical_rs->rs_end) { vd->vdev_initialize_bytes_done += vd->vdev_initialize_last_offset - physical_rs->rs_start; } } static void vdev_initialize_calculate_progress(vdev_t *vd) { ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) || spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER)); ASSERT(vd->vdev_leaf_zap != 0); vd->vdev_initialize_bytes_est = 0; vd->vdev_initialize_bytes_done = 0; for (uint64_t i = 0; i < vd->vdev_top->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_top->vdev_ms[i]; mutex_enter(&msp->ms_lock); uint64_t ms_free = (msp->ms_size - metaslab_allocated_space(msp)) / vdev_get_ndisks(vd->vdev_top); /* * Convert the metaslab range to a physical range * on our vdev. We use this to determine if we are * in the middle of this metaslab range. */ range_seg64_t logical_rs, physical_rs, remain_rs; logical_rs.rs_start = msp->ms_start; logical_rs.rs_end = msp->ms_start + msp->ms_size; /* Metaslab space after this offset has not been initialized */ vdev_xlate(vd, &logical_rs, &physical_rs, &remain_rs); if (vd->vdev_initialize_last_offset <= physical_rs.rs_start) { vd->vdev_initialize_bytes_est += ms_free; mutex_exit(&msp->ms_lock); continue; } /* Metaslab space before this offset has been initialized */ uint64_t last_rs_end = physical_rs.rs_end; if (!vdev_xlate_is_empty(&remain_rs)) { vdev_xlate_walk(vd, &remain_rs, vdev_initialize_xlate_last_rs_end, &last_rs_end); } if (vd->vdev_initialize_last_offset > last_rs_end) { vd->vdev_initialize_bytes_done += ms_free; vd->vdev_initialize_bytes_est += ms_free; mutex_exit(&msp->ms_lock); continue; } /* * If we get here, we're in the middle of initializing this * metaslab. Load it and walk the free tree for more accurate * progress estimation. */ VERIFY0(metaslab_load(msp)); zfs_btree_index_t where; range_tree_t *rt = msp->ms_allocatable; for (range_seg_t *rs = zfs_btree_first(&rt->rt_root, &where); rs; rs = zfs_btree_next(&rt->rt_root, &where, &where)) { logical_rs.rs_start = rs_get_start(rs, rt); logical_rs.rs_end = rs_get_end(rs, rt); vdev_xlate_walk(vd, &logical_rs, vdev_initialize_xlate_progress, vd); } mutex_exit(&msp->ms_lock); } } static int vdev_initialize_load(vdev_t *vd) { int err = 0; ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) || spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER)); ASSERT(vd->vdev_leaf_zap != 0); if (vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE || vd->vdev_initialize_state == VDEV_INITIALIZE_SUSPENDED) { err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET, sizeof (vd->vdev_initialize_last_offset), 1, &vd->vdev_initialize_last_offset); if (err == ENOENT) { vd->vdev_initialize_last_offset = 0; err = 0; } } vdev_initialize_calculate_progress(vd); return (err); } static void vdev_initialize_xlate_range_add(void *arg, range_seg64_t *physical_rs) { vdev_t *vd = arg; /* Only add segments that we have not visited yet */ if (physical_rs->rs_end <= vd->vdev_initialize_last_offset) return; /* Pick up where we left off mid-range. */ if (vd->vdev_initialize_last_offset > physical_rs->rs_start) { zfs_dbgmsg("range write: vd %s changed (%llu, %llu) to " "(%llu, %llu)", vd->vdev_path, (u_longlong_t)physical_rs->rs_start, (u_longlong_t)physical_rs->rs_end, (u_longlong_t)vd->vdev_initialize_last_offset, (u_longlong_t)physical_rs->rs_end); ASSERT3U(physical_rs->rs_end, >, vd->vdev_initialize_last_offset); physical_rs->rs_start = vd->vdev_initialize_last_offset; } ASSERT3U(physical_rs->rs_end, >, physical_rs->rs_start); range_tree_add(vd->vdev_initialize_tree, physical_rs->rs_start, physical_rs->rs_end - physical_rs->rs_start); } /* * Convert the logical range into a physical range and add it to our * avl tree. */ static void vdev_initialize_range_add(void *arg, uint64_t start, uint64_t size) { vdev_t *vd = arg; range_seg64_t logical_rs; logical_rs.rs_start = start; logical_rs.rs_end = start + size; ASSERT(vd->vdev_ops->vdev_op_leaf); vdev_xlate_walk(vd, &logical_rs, vdev_initialize_xlate_range_add, arg); } static __attribute__((noreturn)) void vdev_initialize_thread(void *arg) { vdev_t *vd = arg; spa_t *spa = vd->vdev_spa; int error = 0; uint64_t ms_count = 0; ASSERT(vdev_is_concrete(vd)); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vd->vdev_initialize_last_offset = 0; VERIFY0(vdev_initialize_load(vd)); abd_t *deadbeef = vdev_initialize_block_alloc(); vd->vdev_initialize_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); for (uint64_t i = 0; !vd->vdev_detached && i < vd->vdev_top->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_top->vdev_ms[i]; boolean_t unload_when_done = B_FALSE; /* * If we've expanded the top-level vdev or it's our * first pass, calculate our progress. */ if (vd->vdev_top->vdev_ms_count != ms_count) { vdev_initialize_calculate_progress(vd); ms_count = vd->vdev_top->vdev_ms_count; } spa_config_exit(spa, SCL_CONFIG, FTAG); metaslab_disable(msp); mutex_enter(&msp->ms_lock); if (!msp->ms_loaded && !msp->ms_loading) unload_when_done = B_TRUE; VERIFY0(metaslab_load(msp)); range_tree_walk(msp->ms_allocatable, vdev_initialize_range_add, vd); mutex_exit(&msp->ms_lock); error = vdev_initialize_ranges(vd, deadbeef); metaslab_enable(msp, B_TRUE, unload_when_done); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); range_tree_vacate(vd->vdev_initialize_tree, NULL, NULL); if (error != 0) break; } spa_config_exit(spa, SCL_CONFIG, FTAG); mutex_enter(&vd->vdev_initialize_io_lock); while (vd->vdev_initialize_inflight > 0) { cv_wait(&vd->vdev_initialize_io_cv, &vd->vdev_initialize_io_lock); } mutex_exit(&vd->vdev_initialize_io_lock); range_tree_destroy(vd->vdev_initialize_tree); vdev_initialize_block_free(deadbeef); vd->vdev_initialize_tree = NULL; mutex_enter(&vd->vdev_initialize_lock); if (!vd->vdev_initialize_exit_wanted) { if (vdev_writeable(vd)) { vdev_initialize_change_state(vd, VDEV_INITIALIZE_COMPLETE); } else if (vd->vdev_faulted) { vdev_initialize_change_state(vd, VDEV_INITIALIZE_CANCELED); } } ASSERT(vd->vdev_initialize_thread != NULL || vd->vdev_initialize_inflight == 0); /* * Drop the vdev_initialize_lock while we sync out the * txg since it's possible that a device might be trying to * come online and must check to see if it needs to restart an * initialization. That thread will be holding the spa_config_lock * which would prevent the txg_wait_synced from completing. */ mutex_exit(&vd->vdev_initialize_lock); txg_wait_synced(spa_get_dsl(spa), 0); mutex_enter(&vd->vdev_initialize_lock); vd->vdev_initialize_thread = NULL; cv_broadcast(&vd->vdev_initialize_cv); mutex_exit(&vd->vdev_initialize_lock); thread_exit(); } /* * Initiates a device. Caller must hold vdev_initialize_lock. * Device must be a leaf and not already be initializing. */ void vdev_initialize(vdev_t *vd) { ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); ASSERT3P(vd->vdev_initialize_thread, ==, NULL); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_initialize_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); ASSERT(!vd->vdev_top->vdev_rz_expanding); vdev_initialize_change_state(vd, VDEV_INITIALIZE_ACTIVE); vd->vdev_initialize_thread = thread_create(NULL, 0, vdev_initialize_thread, vd, 0, &p0, TS_RUN, maxclsyspri); } /* * Uninitializes a device. Caller must hold vdev_initialize_lock. * Device must be a leaf and not already be initializing. */ void vdev_uninitialize(vdev_t *vd) { ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); ASSERT3P(vd->vdev_initialize_thread, ==, NULL); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_initialize_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); vdev_initialize_change_state(vd, VDEV_INITIALIZE_NONE); } /* * Wait for the initialize thread to be terminated (cancelled or stopped). */ static void vdev_initialize_stop_wait_impl(vdev_t *vd) { ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); while (vd->vdev_initialize_thread != NULL) cv_wait(&vd->vdev_initialize_cv, &vd->vdev_initialize_lock); ASSERT3P(vd->vdev_initialize_thread, ==, NULL); vd->vdev_initialize_exit_wanted = B_FALSE; } /* * Wait for vdev initialize threads which were either to cleanly exit. */ void vdev_initialize_stop_wait(spa_t *spa, list_t *vd_list) { (void) spa; vdev_t *vd; ASSERT(MUTEX_HELD(&spa_namespace_lock)); while ((vd = list_remove_head(vd_list)) != NULL) { mutex_enter(&vd->vdev_initialize_lock); vdev_initialize_stop_wait_impl(vd); mutex_exit(&vd->vdev_initialize_lock); } } /* * Stop initializing a device, with the resultant initializing state being * tgt_state. For blocking behavior pass NULL for vd_list. Otherwise, when * a list_t is provided the stopping vdev is inserted in to the list. Callers * are then required to call vdev_initialize_stop_wait() to block for all the * initialization threads to exit. The caller must hold vdev_initialize_lock * and must not be writing to the spa config, as the initializing thread may * try to enter the config as a reader before exiting. */ void vdev_initialize_stop(vdev_t *vd, vdev_initializing_state_t tgt_state, list_t *vd_list) { ASSERT(!spa_config_held(vd->vdev_spa, SCL_CONFIG|SCL_STATE, RW_WRITER)); ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); /* * Allow cancel requests to proceed even if the initialize thread * has stopped. */ if (vd->vdev_initialize_thread == NULL && tgt_state != VDEV_INITIALIZE_CANCELED) { return; } vdev_initialize_change_state(vd, tgt_state); vd->vdev_initialize_exit_wanted = B_TRUE; if (vd_list == NULL) { vdev_initialize_stop_wait_impl(vd); } else { ASSERT(MUTEX_HELD(&spa_namespace_lock)); list_insert_tail(vd_list, vd); } } static void vdev_initialize_stop_all_impl(vdev_t *vd, vdev_initializing_state_t tgt_state, list_t *vd_list) { if (vd->vdev_ops->vdev_op_leaf && vdev_is_concrete(vd)) { mutex_enter(&vd->vdev_initialize_lock); vdev_initialize_stop(vd, tgt_state, vd_list); mutex_exit(&vd->vdev_initialize_lock); return; } for (uint64_t i = 0; i < vd->vdev_children; i++) { vdev_initialize_stop_all_impl(vd->vdev_child[i], tgt_state, vd_list); } } /* * Convenience function to stop initializing of a vdev tree and set all * initialize thread pointers to NULL. */ void vdev_initialize_stop_all(vdev_t *vd, vdev_initializing_state_t tgt_state) { spa_t *spa = vd->vdev_spa; list_t vd_list; ASSERT(MUTEX_HELD(&spa_namespace_lock)); list_create(&vd_list, sizeof (vdev_t), offsetof(vdev_t, vdev_initialize_node)); vdev_initialize_stop_all_impl(vd, tgt_state, &vd_list); vdev_initialize_stop_wait(spa, &vd_list); if (vd->vdev_spa->spa_sync_on) { /* Make sure that our state has been synced to disk */ txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0); } list_destroy(&vd_list); } void vdev_initialize_restart(vdev_t *vd) { - ASSERT(MUTEX_HELD(&spa_namespace_lock)); + ASSERT(MUTEX_HELD(&spa_namespace_lock) || + vd->vdev_spa->spa_load_thread == curthread); ASSERT(!spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER)); if (vd->vdev_leaf_zap != 0) { mutex_enter(&vd->vdev_initialize_lock); uint64_t initialize_state = VDEV_INITIALIZE_NONE; int err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_STATE, sizeof (initialize_state), 1, &initialize_state); ASSERT(err == 0 || err == ENOENT); vd->vdev_initialize_state = initialize_state; uint64_t timestamp = 0; err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, sizeof (timestamp), 1, ×tamp); ASSERT(err == 0 || err == ENOENT); vd->vdev_initialize_action_time = timestamp; if ((vd->vdev_initialize_state == VDEV_INITIALIZE_SUSPENDED || vd->vdev_offline) && !vd->vdev_top->vdev_rz_expanding) { /* load progress for reporting, but don't resume */ VERIFY0(vdev_initialize_load(vd)); } else if (vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE && vdev_writeable(vd) && !vd->vdev_top->vdev_removing && !vd->vdev_top->vdev_rz_expanding && vd->vdev_initialize_thread == NULL) { vdev_initialize(vd); } mutex_exit(&vd->vdev_initialize_lock); } for (uint64_t i = 0; i < vd->vdev_children; i++) { vdev_initialize_restart(vd->vdev_child[i]); } } EXPORT_SYMBOL(vdev_initialize); EXPORT_SYMBOL(vdev_uninitialize); EXPORT_SYMBOL(vdev_initialize_stop); EXPORT_SYMBOL(vdev_initialize_stop_all); EXPORT_SYMBOL(vdev_initialize_stop_wait); EXPORT_SYMBOL(vdev_initialize_restart); ZFS_MODULE_PARAM(zfs, zfs_, initialize_value, U64, ZMOD_RW, "Value written during zpool initialize"); ZFS_MODULE_PARAM(zfs, zfs_, initialize_chunk_size, U64, ZMOD_RW, "Size in bytes of writes by zpool initialize"); diff --git a/sys/contrib/openzfs/module/zfs/vdev_rebuild.c b/sys/contrib/openzfs/module/zfs/vdev_rebuild.c index 6503390f7973..00ebd4c9fca4 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_rebuild.c +++ b/sys/contrib/openzfs/module/zfs/vdev_rebuild.c @@ -1,1173 +1,1175 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * * Copyright (c) 2018, Intel Corporation. * Copyright (c) 2020 by Lawrence Livermore National Security, LLC. * Copyright (c) 2022 Hewlett Packard Enterprise Development LP. + * Copyright (c) 2024 by Delphix. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include /* * This file contains the sequential reconstruction implementation for * resilvering. This form of resilvering is internally referred to as device * rebuild to avoid conflating it with the traditional healing reconstruction * performed by the dsl scan code. * * When replacing a device, or scrubbing the pool, ZFS has historically used * a process called resilvering which is a form of healing reconstruction. * This approach has the advantage that as blocks are read from disk their * checksums can be immediately verified and the data repaired. Unfortunately, * it also results in a random IO pattern to the disk even when extra care * is taken to sequentialize the IO as much as possible. This substantially * increases the time required to resilver the pool and restore redundancy. * * For mirrored devices it's possible to implement an alternate sequential * reconstruction strategy when resilvering. Sequential reconstruction * behaves like a traditional RAID rebuild and reconstructs a device in LBA * order without verifying the checksum. After this phase completes a second * scrub phase is started to verify all of the checksums. This two phase * process will take longer than the healing reconstruction described above. * However, it has that advantage that after the reconstruction first phase * completes redundancy has been restored. At this point the pool can incur * another device failure without risking data loss. * * There are a few noteworthy limitations and other advantages of resilvering * using sequential reconstruction vs healing reconstruction. * * Limitations: * * - Sequential reconstruction is not possible on RAIDZ due to its * variable stripe width. Note dRAID uses a fixed stripe width which * avoids this issue, but comes at the expense of some usable capacity. * * - Block checksums are not verified during sequential reconstruction. * Similar to traditional RAID the parity/mirror data is reconstructed * but cannot be immediately double checked. For this reason when the * last active resilver completes the pool is automatically scrubbed * by default. * * - Deferred resilvers using sequential reconstruction are not currently * supported. When adding another vdev to an active top-level resilver * it must be restarted. * * Advantages: * * - Sequential reconstruction is performed in LBA order which may be faster * than healing reconstruction particularly when using HDDs (or * especially with SMR devices). Only allocated capacity is resilvered. * * - Sequential reconstruction is not constrained by ZFS block boundaries. * This allows it to issue larger IOs to disk which span multiple blocks * allowing all of these logical blocks to be repaired with a single IO. * * - Unlike a healing resilver or scrub which are pool wide operations, * sequential reconstruction is handled by the top-level vdevs. This * allows for it to be started or canceled on a top-level vdev without * impacting any other top-level vdevs in the pool. * * - Data only referenced by a pool checkpoint will be repaired because * that space is reflected in the space maps. This differs for a * healing resilver or scrub which will not repair that data. */ /* * Size of rebuild reads; defaults to 1MiB per data disk and is capped at * SPA_MAXBLOCKSIZE. */ static uint64_t zfs_rebuild_max_segment = 1024 * 1024; /* * Maximum number of parallelly executed bytes per leaf vdev caused by a * sequential resilver. We attempt to strike a balance here between keeping * the vdev queues full of I/Os at all times and not overflowing the queues * to cause long latency, which would cause long txg sync times. * * A large default value can be safely used here because the default target * segment size is also large (zfs_rebuild_max_segment=1M). This helps keep * the queue depth short. * * 64MB was observed to deliver the best performance and set as the default. * Testing was performed with a 106-drive dRAID HDD pool (draid2:11d:106c) * and a rebuild rate of 1.2GB/s was measured to the distribute spare. * Smaller values were unable to fully saturate the available pool I/O. */ static uint64_t zfs_rebuild_vdev_limit = 64 << 20; /* * Automatically start a pool scrub when the last active sequential resilver * completes in order to verify the checksums of all blocks which have been * resilvered. This option is enabled by default and is strongly recommended. */ static int zfs_rebuild_scrub_enabled = 1; /* * For vdev_rebuild_initiate_sync() and vdev_rebuild_reset_sync(). */ static __attribute__((noreturn)) void vdev_rebuild_thread(void *arg); static void vdev_rebuild_reset_sync(void *arg, dmu_tx_t *tx); /* * Clear the per-vdev rebuild bytes value for a vdev tree. */ static void clear_rebuild_bytes(vdev_t *vd) { vdev_stat_t *vs = &vd->vdev_stat; for (uint64_t i = 0; i < vd->vdev_children; i++) clear_rebuild_bytes(vd->vdev_child[i]); mutex_enter(&vd->vdev_stat_lock); vs->vs_rebuild_processed = 0; mutex_exit(&vd->vdev_stat_lock); } /* * Determines whether a vdev_rebuild_thread() should be stopped. */ static boolean_t vdev_rebuild_should_stop(vdev_t *vd) { return (!vdev_writeable(vd) || vd->vdev_removing || vd->vdev_rebuild_exit_wanted || vd->vdev_rebuild_cancel_wanted || vd->vdev_rebuild_reset_wanted); } /* * Determine if the rebuild should be canceled. This may happen when all * vdevs with MISSING DTLs are detached. */ static boolean_t vdev_rebuild_should_cancel(vdev_t *vd) { vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; if (!vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg)) return (B_TRUE); return (B_FALSE); } /* * The sync task for updating the on-disk state of a rebuild. This is * scheduled by vdev_rebuild_range(). */ static void vdev_rebuild_update_sync(void *arg, dmu_tx_t *tx) { int vdev_id = (uintptr_t)arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *vd = vdev_lookup_top(spa, vdev_id); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; uint64_t txg = dmu_tx_get_txg(tx); mutex_enter(&vd->vdev_rebuild_lock); if (vr->vr_scan_offset[txg & TXG_MASK] > 0) { vrp->vrp_last_offset = vr->vr_scan_offset[txg & TXG_MASK]; vr->vr_scan_offset[txg & TXG_MASK] = 0; } vrp->vrp_scan_time_ms = vr->vr_prev_scan_time_ms + NSEC2MSEC(gethrtime() - vr->vr_pass_start_time); VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp, tx)); mutex_exit(&vd->vdev_rebuild_lock); } /* * Initialize the on-disk state for a new rebuild, start the rebuild thread. */ static void vdev_rebuild_initiate_sync(void *arg, dmu_tx_t *tx) { int vdev_id = (uintptr_t)arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *vd = vdev_lookup_top(spa, vdev_id); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; ASSERT(vd->vdev_rebuilding); spa_feature_incr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx); mutex_enter(&vd->vdev_rebuild_lock); memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES); vrp->vrp_rebuild_state = VDEV_REBUILD_ACTIVE; vrp->vrp_min_txg = 0; vrp->vrp_max_txg = dmu_tx_get_txg(tx); vrp->vrp_start_time = gethrestime_sec(); vrp->vrp_scan_time_ms = 0; vr->vr_prev_scan_time_ms = 0; /* * Rebuilds are currently only used when replacing a device, in which * case there must be DTL_MISSING entries. In the future, we could * allow rebuilds to be used in a way similar to a scrub. This would * be useful because it would allow us to rebuild the space used by * pool checkpoints. */ VERIFY(vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg)); VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp, tx)); spa_history_log_internal(spa, "rebuild", tx, "vdev_id=%llu vdev_guid=%llu started", (u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid); ASSERT3P(vd->vdev_rebuild_thread, ==, NULL); vd->vdev_rebuild_thread = thread_create(NULL, 0, vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri); mutex_exit(&vd->vdev_rebuild_lock); } static void vdev_rebuild_log_notify(spa_t *spa, vdev_t *vd, const char *name) { nvlist_t *aux = fnvlist_alloc(); fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE, "sequential"); spa_event_notify(spa, vd, aux, name); nvlist_free(aux); } /* * Called to request that a new rebuild be started. The feature will remain * active for the duration of the rebuild, then revert to the enabled state. */ static void vdev_rebuild_initiate(vdev_t *vd) { spa_t *spa = vd->vdev_spa; ASSERT(vd->vdev_top == vd); ASSERT(MUTEX_HELD(&vd->vdev_rebuild_lock)); ASSERT(!vd->vdev_rebuilding); dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); vd->vdev_rebuilding = B_TRUE; dsl_sync_task_nowait(spa_get_dsl(spa), vdev_rebuild_initiate_sync, (void *)(uintptr_t)vd->vdev_id, tx); dmu_tx_commit(tx); vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_START); } /* * Update the on-disk state to completed when a rebuild finishes. */ static void vdev_rebuild_complete_sync(void *arg, dmu_tx_t *tx) { int vdev_id = (uintptr_t)arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *vd = vdev_lookup_top(spa, vdev_id); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; mutex_enter(&vd->vdev_rebuild_lock); /* * Handle a second device failure if it occurs after all rebuild I/O * has completed but before this sync task has been executed. */ if (vd->vdev_rebuild_reset_wanted) { mutex_exit(&vd->vdev_rebuild_lock); vdev_rebuild_reset_sync(arg, tx); return; } vrp->vrp_rebuild_state = VDEV_REBUILD_COMPLETE; vrp->vrp_end_time = gethrestime_sec(); VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp, tx)); vdev_dtl_reassess(vd, tx->tx_txg, vrp->vrp_max_txg, B_TRUE, B_TRUE); spa_feature_decr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx); spa_history_log_internal(spa, "rebuild", tx, "vdev_id=%llu vdev_guid=%llu complete", (u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid); vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_FINISH); /* Handles detaching of spares */ spa_async_request(spa, SPA_ASYNC_REBUILD_DONE); vd->vdev_rebuilding = B_FALSE; mutex_exit(&vd->vdev_rebuild_lock); /* * While we're in syncing context take the opportunity to * setup the scrub when there are no more active rebuilds. */ pool_scan_func_t func = POOL_SCAN_SCRUB; if (dsl_scan_setup_check(&func, tx) == 0 && zfs_rebuild_scrub_enabled) { dsl_scan_setup_sync(&func, tx); } cv_broadcast(&vd->vdev_rebuild_cv); /* Clear recent error events (i.e. duplicate events tracking) */ zfs_ereport_clear(spa, NULL); } /* * Update the on-disk state to canceled when a rebuild finishes. */ static void vdev_rebuild_cancel_sync(void *arg, dmu_tx_t *tx) { int vdev_id = (uintptr_t)arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *vd = vdev_lookup_top(spa, vdev_id); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; mutex_enter(&vd->vdev_rebuild_lock); vrp->vrp_rebuild_state = VDEV_REBUILD_CANCELED; vrp->vrp_end_time = gethrestime_sec(); VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp, tx)); spa_feature_decr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx); spa_history_log_internal(spa, "rebuild", tx, "vdev_id=%llu vdev_guid=%llu canceled", (u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid); vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_FINISH); vd->vdev_rebuild_cancel_wanted = B_FALSE; vd->vdev_rebuilding = B_FALSE; mutex_exit(&vd->vdev_rebuild_lock); spa_notify_waiters(spa); cv_broadcast(&vd->vdev_rebuild_cv); } /* * Resets the progress of a running rebuild. This will occur when a new * vdev is added to rebuild. */ static void vdev_rebuild_reset_sync(void *arg, dmu_tx_t *tx) { int vdev_id = (uintptr_t)arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *vd = vdev_lookup_top(spa, vdev_id); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; mutex_enter(&vd->vdev_rebuild_lock); ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE); ASSERT3P(vd->vdev_rebuild_thread, ==, NULL); vrp->vrp_last_offset = 0; vrp->vrp_min_txg = 0; vrp->vrp_max_txg = dmu_tx_get_txg(tx); vrp->vrp_bytes_scanned = 0; vrp->vrp_bytes_issued = 0; vrp->vrp_bytes_rebuilt = 0; vrp->vrp_bytes_est = 0; vrp->vrp_scan_time_ms = 0; vr->vr_prev_scan_time_ms = 0; /* See vdev_rebuild_initiate_sync comment */ VERIFY(vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg)); VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp, tx)); spa_history_log_internal(spa, "rebuild", tx, "vdev_id=%llu vdev_guid=%llu reset", (u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid); vd->vdev_rebuild_reset_wanted = B_FALSE; ASSERT(vd->vdev_rebuilding); vd->vdev_rebuild_thread = thread_create(NULL, 0, vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri); mutex_exit(&vd->vdev_rebuild_lock); } /* * Clear the last rebuild status. */ void vdev_rebuild_clear_sync(void *arg, dmu_tx_t *tx) { int vdev_id = (uintptr_t)arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *vd = vdev_lookup_top(spa, vdev_id); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; objset_t *mos = spa_meta_objset(spa); mutex_enter(&vd->vdev_rebuild_lock); if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD) || vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE) { mutex_exit(&vd->vdev_rebuild_lock); return; } clear_rebuild_bytes(vd); memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES); if (vd->vdev_top_zap != 0 && zap_contains(mos, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS) == 0) { VERIFY0(zap_update(mos, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp, tx)); } mutex_exit(&vd->vdev_rebuild_lock); } /* * The zio_done_func_t callback for each rebuild I/O issued. It's responsible * for updating the rebuild stats and limiting the number of in flight I/Os. */ static void vdev_rebuild_cb(zio_t *zio) { vdev_rebuild_t *vr = zio->io_private; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; vdev_t *vd = vr->vr_top_vdev; mutex_enter(&vr->vr_io_lock); if (zio->io_error == ENXIO && !vdev_writeable(vd)) { /* * The I/O failed because the top-level vdev was unavailable. * Attempt to roll back to the last completed offset, in order * resume from the correct location if the pool is resumed. * (This works because spa_sync waits on spa_txg_zio before * it runs sync tasks.) */ uint64_t *off = &vr->vr_scan_offset[zio->io_txg & TXG_MASK]; *off = MIN(*off, zio->io_offset); } else if (zio->io_error) { vrp->vrp_errors++; } abd_free(zio->io_abd); ASSERT3U(vr->vr_bytes_inflight, >, 0); vr->vr_bytes_inflight -= zio->io_size; cv_broadcast(&vr->vr_io_cv); mutex_exit(&vr->vr_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); } /* * Initialize a block pointer that can be used to read the given segment * for sequential rebuild. */ static void vdev_rebuild_blkptr_init(blkptr_t *bp, vdev_t *vd, uint64_t start, uint64_t asize) { ASSERT(vd->vdev_ops == &vdev_draid_ops || vd->vdev_ops == &vdev_mirror_ops || vd->vdev_ops == &vdev_replacing_ops || vd->vdev_ops == &vdev_spare_ops); uint64_t psize = vd->vdev_ops == &vdev_draid_ops ? vdev_draid_asize_to_psize(vd, asize) : asize; BP_ZERO(bp); DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id); DVA_SET_OFFSET(&bp->blk_dva[0], start); DVA_SET_GANG(&bp->blk_dva[0], 0); DVA_SET_ASIZE(&bp->blk_dva[0], asize); BP_SET_BIRTH(bp, TXG_INITIAL, TXG_INITIAL); BP_SET_LSIZE(bp, psize); BP_SET_PSIZE(bp, psize); BP_SET_COMPRESS(bp, ZIO_COMPRESS_OFF); BP_SET_CHECKSUM(bp, ZIO_CHECKSUM_OFF); BP_SET_TYPE(bp, DMU_OT_NONE); BP_SET_LEVEL(bp, 0); BP_SET_DEDUP(bp, 0); BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER); } /* * Issues a rebuild I/O and takes care of rate limiting the number of queued * rebuild I/Os. The provided start and size must be properly aligned for the * top-level vdev type being rebuilt. */ static int vdev_rebuild_range(vdev_rebuild_t *vr, uint64_t start, uint64_t size) { uint64_t ms_id __maybe_unused = vr->vr_scan_msp->ms_id; vdev_t *vd = vr->vr_top_vdev; spa_t *spa = vd->vdev_spa; blkptr_t blk; ASSERT3U(ms_id, ==, start >> vd->vdev_ms_shift); ASSERT3U(ms_id, ==, (start + size - 1) >> vd->vdev_ms_shift); vr->vr_pass_bytes_scanned += size; vr->vr_rebuild_phys.vrp_bytes_scanned += size; /* * Rebuild the data in this range by constructing a special block * pointer. It has no relation to any existing blocks in the pool. * However, by disabling checksum verification and issuing a scrub IO * we can reconstruct and repair any children with missing data. */ vdev_rebuild_blkptr_init(&blk, vd, start, size); uint64_t psize = BP_GET_PSIZE(&blk); if (!vdev_dtl_need_resilver(vd, &blk.blk_dva[0], psize, TXG_UNKNOWN)) { vr->vr_pass_bytes_skipped += size; return (0); } mutex_enter(&vr->vr_io_lock); /* Limit in flight rebuild I/Os */ while (vr->vr_bytes_inflight >= vr->vr_bytes_inflight_max) cv_wait(&vr->vr_io_cv, &vr->vr_io_lock); vr->vr_bytes_inflight += psize; mutex_exit(&vr->vr_io_lock); dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); uint64_t txg = dmu_tx_get_txg(tx); spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER); mutex_enter(&vd->vdev_rebuild_lock); /* This is the first I/O for this txg. */ if (vr->vr_scan_offset[txg & TXG_MASK] == 0) { vr->vr_scan_offset[txg & TXG_MASK] = start; dsl_sync_task_nowait(spa_get_dsl(spa), vdev_rebuild_update_sync, (void *)(uintptr_t)vd->vdev_id, tx); } /* When exiting write out our progress. */ if (vdev_rebuild_should_stop(vd)) { mutex_enter(&vr->vr_io_lock); vr->vr_bytes_inflight -= psize; mutex_exit(&vr->vr_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); mutex_exit(&vd->vdev_rebuild_lock); dmu_tx_commit(tx); return (SET_ERROR(EINTR)); } mutex_exit(&vd->vdev_rebuild_lock); dmu_tx_commit(tx); vr->vr_scan_offset[txg & TXG_MASK] = start + size; vr->vr_pass_bytes_issued += size; vr->vr_rebuild_phys.vrp_bytes_issued += size; zio_nowait(zio_read(spa->spa_txg_zio[txg & TXG_MASK], spa, &blk, abd_alloc(psize, B_FALSE), psize, vdev_rebuild_cb, vr, ZIO_PRIORITY_REBUILD, ZIO_FLAG_RAW | ZIO_FLAG_CANFAIL | ZIO_FLAG_RESILVER, NULL)); return (0); } /* * Issues rebuild I/Os for all ranges in the provided vr->vr_tree range tree. */ static int vdev_rebuild_ranges(vdev_rebuild_t *vr) { vdev_t *vd = vr->vr_top_vdev; zfs_btree_t *t = &vr->vr_scan_tree->rt_root; zfs_btree_index_t idx; int error; for (range_seg_t *rs = zfs_btree_first(t, &idx); rs != NULL; rs = zfs_btree_next(t, &idx, &idx)) { uint64_t start = rs_get_start(rs, vr->vr_scan_tree); uint64_t size = rs_get_end(rs, vr->vr_scan_tree) - start; /* * zfs_scan_suspend_progress can be set to disable rebuild * progress for testing. See comment in dsl_scan_sync(). */ while (zfs_scan_suspend_progress && !vdev_rebuild_should_stop(vd)) { delay(hz); } while (size > 0) { uint64_t chunk_size; /* * Split range into legally-sized logical chunks * given the constraints of the top-level vdev * being rebuilt (dRAID or mirror). */ ASSERT3P(vd->vdev_ops, !=, NULL); chunk_size = vd->vdev_ops->vdev_op_rebuild_asize(vd, start, size, zfs_rebuild_max_segment); error = vdev_rebuild_range(vr, start, chunk_size); if (error != 0) return (error); size -= chunk_size; start += chunk_size; } } return (0); } /* * Calculates the estimated capacity which remains to be scanned. Since * we traverse the pool in metaslab order only allocated capacity beyond * the vrp_last_offset need be considered. All lower offsets must have * already been rebuilt and are thus already included in vrp_bytes_scanned. */ static void vdev_rebuild_update_bytes_est(vdev_t *vd, uint64_t ms_id) { vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; uint64_t bytes_est = vrp->vrp_bytes_scanned; if (vrp->vrp_last_offset < vd->vdev_ms[ms_id]->ms_start) return; for (uint64_t i = ms_id; i < vd->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_ms[i]; mutex_enter(&msp->ms_lock); bytes_est += metaslab_allocated_space(msp); mutex_exit(&msp->ms_lock); } vrp->vrp_bytes_est = bytes_est; } /* * Load from disk the top-level vdev's rebuild information. */ int vdev_rebuild_load(vdev_t *vd) { vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; spa_t *spa = vd->vdev_spa; int err = 0; mutex_enter(&vd->vdev_rebuild_lock); vd->vdev_rebuilding = B_FALSE; if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD)) { memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES); mutex_exit(&vd->vdev_rebuild_lock); return (SET_ERROR(ENOTSUP)); } ASSERT(vd->vdev_top == vd); err = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t), REBUILD_PHYS_ENTRIES, vrp); /* * A missing or damaged VDEV_TOP_ZAP_VDEV_REBUILD_PHYS should * not prevent a pool from being imported. Clear the rebuild * status allowing a new resilver/rebuild to be started. */ if (err == ENOENT || err == EOVERFLOW || err == ECKSUM) { memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES); } else if (err) { mutex_exit(&vd->vdev_rebuild_lock); return (err); } vr->vr_prev_scan_time_ms = vrp->vrp_scan_time_ms; vr->vr_top_vdev = vd; mutex_exit(&vd->vdev_rebuild_lock); return (0); } /* * Each scan thread is responsible for rebuilding a top-level vdev. The * rebuild progress in tracked on-disk in VDEV_TOP_ZAP_VDEV_REBUILD_PHYS. */ static __attribute__((noreturn)) void vdev_rebuild_thread(void *arg) { vdev_t *vd = arg; spa_t *spa = vd->vdev_spa; vdev_t *rvd = spa->spa_root_vdev; int error = 0; /* * If there's a scrub in process request that it be stopped. This * is not required for a correct rebuild, but we do want rebuilds to * emulate the resilver behavior as much as possible. */ dsl_pool_t *dsl = spa_get_dsl(spa); if (dsl_scan_scrubbing(dsl)) dsl_scan_cancel(dsl); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); mutex_enter(&vd->vdev_rebuild_lock); ASSERT3P(vd->vdev_top, ==, vd); ASSERT3P(vd->vdev_rebuild_thread, !=, NULL); ASSERT(vd->vdev_rebuilding); ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REBUILD)); ASSERT3B(vd->vdev_rebuild_cancel_wanted, ==, B_FALSE); vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; vr->vr_top_vdev = vd; vr->vr_scan_msp = NULL; vr->vr_scan_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); mutex_init(&vr->vr_io_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&vr->vr_io_cv, NULL, CV_DEFAULT, NULL); vr->vr_pass_start_time = gethrtime(); vr->vr_pass_bytes_scanned = 0; vr->vr_pass_bytes_issued = 0; vr->vr_pass_bytes_skipped = 0; uint64_t update_est_time = gethrtime(); vdev_rebuild_update_bytes_est(vd, 0); clear_rebuild_bytes(vr->vr_top_vdev); mutex_exit(&vd->vdev_rebuild_lock); /* * Systematically walk the metaslabs and issue rebuild I/Os for * all ranges in the allocated space map. */ for (uint64_t i = 0; i < vd->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_ms[i]; vr->vr_scan_msp = msp; /* * Calculate the max number of in-flight bytes for top-level * vdev scanning operations (minimum 1MB, maximum 1/2 of * arc_c_max shared by all top-level vdevs). Limits for the * issuing phase are done per top-level vdev and are handled * separately. */ uint64_t limit = (arc_c_max / 2) / MAX(rvd->vdev_children, 1); vr->vr_bytes_inflight_max = MIN(limit, MAX(1ULL << 20, zfs_rebuild_vdev_limit * vd->vdev_children)); /* * Removal of vdevs from the vdev tree may eliminate the need * for the rebuild, in which case it should be canceled. The * vdev_rebuild_cancel_wanted flag is set until the sync task * completes. This may be after the rebuild thread exits. */ if (vdev_rebuild_should_cancel(vd)) { vd->vdev_rebuild_cancel_wanted = B_TRUE; error = EINTR; break; } ASSERT0(range_tree_space(vr->vr_scan_tree)); /* Disable any new allocations to this metaslab */ spa_config_exit(spa, SCL_CONFIG, FTAG); metaslab_disable(msp); mutex_enter(&msp->ms_sync_lock); mutex_enter(&msp->ms_lock); /* * If there are outstanding allocations wait for them to be * synced. This is needed to ensure all allocated ranges are * on disk and therefore will be rebuilt. */ for (int j = 0; j < TXG_SIZE; j++) { if (range_tree_space(msp->ms_allocating[j])) { mutex_exit(&msp->ms_lock); mutex_exit(&msp->ms_sync_lock); txg_wait_synced(dsl, 0); mutex_enter(&msp->ms_sync_lock); mutex_enter(&msp->ms_lock); break; } } /* * When a metaslab has been allocated from read its allocated * ranges from the space map object into the vr_scan_tree. * Then add inflight / unflushed ranges and remove inflight / * unflushed frees. This is the minimum range to be rebuilt. */ if (msp->ms_sm != NULL) { VERIFY0(space_map_load(msp->ms_sm, vr->vr_scan_tree, SM_ALLOC)); for (int i = 0; i < TXG_SIZE; i++) { ASSERT0(range_tree_space( msp->ms_allocating[i])); } range_tree_walk(msp->ms_unflushed_allocs, range_tree_add, vr->vr_scan_tree); range_tree_walk(msp->ms_unflushed_frees, range_tree_remove, vr->vr_scan_tree); /* * Remove ranges which have already been rebuilt based * on the last offset. This can happen when restarting * a scan after exporting and re-importing the pool. */ range_tree_clear(vr->vr_scan_tree, 0, vrp->vrp_last_offset); } mutex_exit(&msp->ms_lock); mutex_exit(&msp->ms_sync_lock); /* * To provide an accurate estimate re-calculate the estimated * size every 5 minutes to account for recent allocations and * frees made to space maps which have not yet been rebuilt. */ if (gethrtime() > update_est_time + SEC2NSEC(300)) { update_est_time = gethrtime(); vdev_rebuild_update_bytes_est(vd, i); } /* * Walk the allocated space map and issue the rebuild I/O. */ error = vdev_rebuild_ranges(vr); range_tree_vacate(vr->vr_scan_tree, NULL, NULL); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); metaslab_enable(msp, B_FALSE, B_FALSE); if (error != 0) break; } range_tree_destroy(vr->vr_scan_tree); spa_config_exit(spa, SCL_CONFIG, FTAG); /* Wait for any remaining rebuild I/O to complete */ mutex_enter(&vr->vr_io_lock); while (vr->vr_bytes_inflight > 0) cv_wait(&vr->vr_io_cv, &vr->vr_io_lock); mutex_exit(&vr->vr_io_lock); mutex_destroy(&vr->vr_io_lock); cv_destroy(&vr->vr_io_cv); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); dsl_pool_t *dp = spa_get_dsl(spa); dmu_tx_t *tx = dmu_tx_create_dd(dp->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); mutex_enter(&vd->vdev_rebuild_lock); if (error == 0) { /* * After a successful rebuild clear the DTLs of all ranges * which were missing when the rebuild was started. These * ranges must have been rebuilt as a consequence of rebuilding * all allocated space. Note that unlike a scrub or resilver * the rebuild operation will reconstruct data only referenced * by a pool checkpoint. See the dsl_scan_done() comments. */ dsl_sync_task_nowait(dp, vdev_rebuild_complete_sync, (void *)(uintptr_t)vd->vdev_id, tx); } else if (vd->vdev_rebuild_cancel_wanted) { /* * The rebuild operation was canceled. This will occur when * a device participating in the rebuild is detached. */ dsl_sync_task_nowait(dp, vdev_rebuild_cancel_sync, (void *)(uintptr_t)vd->vdev_id, tx); } else if (vd->vdev_rebuild_reset_wanted) { /* * Reset the running rebuild without canceling and restarting * it. This will occur when a new device is attached and must * participate in the rebuild. */ dsl_sync_task_nowait(dp, vdev_rebuild_reset_sync, (void *)(uintptr_t)vd->vdev_id, tx); } else { /* * The rebuild operation should be suspended. This may occur * when detaching a child vdev or when exporting the pool. The * rebuild is left in the active state so it will be resumed. */ ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE); vd->vdev_rebuilding = B_FALSE; } dmu_tx_commit(tx); vd->vdev_rebuild_thread = NULL; mutex_exit(&vd->vdev_rebuild_lock); spa_config_exit(spa, SCL_CONFIG, FTAG); cv_broadcast(&vd->vdev_rebuild_cv); thread_exit(); } /* * Returns B_TRUE if any top-level vdev are rebuilding. */ boolean_t vdev_rebuild_active(vdev_t *vd) { spa_t *spa = vd->vdev_spa; boolean_t ret = B_FALSE; if (vd == spa->spa_root_vdev) { for (uint64_t i = 0; i < vd->vdev_children; i++) { ret = vdev_rebuild_active(vd->vdev_child[i]); if (ret) return (ret); } } else if (vd->vdev_top_zap != 0) { vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; mutex_enter(&vd->vdev_rebuild_lock); ret = (vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE); mutex_exit(&vd->vdev_rebuild_lock); } return (ret); } /* * Start a rebuild operation. The rebuild may be restarted when the * top-level vdev is currently actively rebuilding. */ void vdev_rebuild(vdev_t *vd) { vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp __maybe_unused = &vr->vr_rebuild_phys; ASSERT(vd->vdev_top == vd); ASSERT(vdev_is_concrete(vd)); ASSERT(!vd->vdev_removing); ASSERT(spa_feature_is_enabled(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD)); mutex_enter(&vd->vdev_rebuild_lock); if (vd->vdev_rebuilding) { ASSERT3U(vrp->vrp_rebuild_state, ==, VDEV_REBUILD_ACTIVE); /* * Signal a running rebuild operation that it should restart * from the beginning because a new device was attached. The * vdev_rebuild_reset_wanted flag is set until the sync task * completes. This may be after the rebuild thread exits. */ if (!vd->vdev_rebuild_reset_wanted) vd->vdev_rebuild_reset_wanted = B_TRUE; } else { vdev_rebuild_initiate(vd); } mutex_exit(&vd->vdev_rebuild_lock); } static void vdev_rebuild_restart_impl(vdev_t *vd) { spa_t *spa = vd->vdev_spa; if (vd == spa->spa_root_vdev) { for (uint64_t i = 0; i < vd->vdev_children; i++) vdev_rebuild_restart_impl(vd->vdev_child[i]); } else if (vd->vdev_top_zap != 0) { vdev_rebuild_t *vr = &vd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; mutex_enter(&vd->vdev_rebuild_lock); if (vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE && vdev_writeable(vd) && !vd->vdev_rebuilding) { ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REBUILD)); vd->vdev_rebuilding = B_TRUE; vd->vdev_rebuild_thread = thread_create(NULL, 0, vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri); } mutex_exit(&vd->vdev_rebuild_lock); } } /* * Conditionally restart all of the vdev_rebuild_thread's for a pool. The * feature flag must be active and the rebuild in the active state. This * cannot be used to start a new rebuild. */ void vdev_rebuild_restart(spa_t *spa) { - ASSERT(MUTEX_HELD(&spa_namespace_lock)); + ASSERT(MUTEX_HELD(&spa_namespace_lock) || + spa->spa_load_thread == curthread); vdev_rebuild_restart_impl(spa->spa_root_vdev); } /* * Stop and wait for all of the vdev_rebuild_thread's associated with the * vdev tree provide to be terminated (canceled or stopped). */ void vdev_rebuild_stop_wait(vdev_t *vd) { spa_t *spa = vd->vdev_spa; ASSERT(MUTEX_HELD(&spa_namespace_lock)); if (vd == spa->spa_root_vdev) { for (uint64_t i = 0; i < vd->vdev_children; i++) vdev_rebuild_stop_wait(vd->vdev_child[i]); } else if (vd->vdev_top_zap != 0) { ASSERT(vd == vd->vdev_top); mutex_enter(&vd->vdev_rebuild_lock); if (vd->vdev_rebuild_thread != NULL) { vd->vdev_rebuild_exit_wanted = B_TRUE; while (vd->vdev_rebuilding) { cv_wait(&vd->vdev_rebuild_cv, &vd->vdev_rebuild_lock); } vd->vdev_rebuild_exit_wanted = B_FALSE; } mutex_exit(&vd->vdev_rebuild_lock); } } /* * Stop all rebuild operations but leave them in the active state so they * will be resumed when importing the pool. */ void vdev_rebuild_stop_all(spa_t *spa) { vdev_rebuild_stop_wait(spa->spa_root_vdev); } /* * Rebuild statistics reported per top-level vdev. */ int vdev_rebuild_get_stats(vdev_t *tvd, vdev_rebuild_stat_t *vrs) { spa_t *spa = tvd->vdev_spa; if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD)) return (SET_ERROR(ENOTSUP)); if (tvd != tvd->vdev_top || tvd->vdev_top_zap == 0) return (SET_ERROR(EINVAL)); int error = zap_contains(spa_meta_objset(spa), tvd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS); if (error == ENOENT) { memset(vrs, 0, sizeof (vdev_rebuild_stat_t)); vrs->vrs_state = VDEV_REBUILD_NONE; error = 0; } else if (error == 0) { vdev_rebuild_t *vr = &tvd->vdev_rebuild_config; vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; mutex_enter(&tvd->vdev_rebuild_lock); vrs->vrs_state = vrp->vrp_rebuild_state; vrs->vrs_start_time = vrp->vrp_start_time; vrs->vrs_end_time = vrp->vrp_end_time; vrs->vrs_scan_time_ms = vrp->vrp_scan_time_ms; vrs->vrs_bytes_scanned = vrp->vrp_bytes_scanned; vrs->vrs_bytes_issued = vrp->vrp_bytes_issued; vrs->vrs_bytes_rebuilt = vrp->vrp_bytes_rebuilt; vrs->vrs_bytes_est = vrp->vrp_bytes_est; vrs->vrs_errors = vrp->vrp_errors; vrs->vrs_pass_time_ms = NSEC2MSEC(gethrtime() - vr->vr_pass_start_time); vrs->vrs_pass_bytes_scanned = vr->vr_pass_bytes_scanned; vrs->vrs_pass_bytes_issued = vr->vr_pass_bytes_issued; vrs->vrs_pass_bytes_skipped = vr->vr_pass_bytes_skipped; mutex_exit(&tvd->vdev_rebuild_lock); } return (error); } ZFS_MODULE_PARAM(zfs, zfs_, rebuild_max_segment, U64, ZMOD_RW, "Max segment size in bytes of rebuild reads"); ZFS_MODULE_PARAM(zfs, zfs_, rebuild_vdev_limit, U64, ZMOD_RW, "Max bytes in flight per leaf vdev for sequential resilvers"); ZFS_MODULE_PARAM(zfs, zfs_, rebuild_scrub_enabled, INT, ZMOD_RW, "Automatically scrub after sequential resilver completes"); diff --git a/sys/contrib/openzfs/module/zfs/vdev_trim.c b/sys/contrib/openzfs/module/zfs/vdev_trim.c index 7e3c5f684703..9753d5a1ea04 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_trim.c +++ b/sys/contrib/openzfs/module/zfs/vdev_trim.c @@ -1,1784 +1,1785 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* - * Copyright (c) 2016 by Delphix. All rights reserved. + * Copyright (c) 2016, 2024 by Delphix. All rights reserved. * Copyright (c) 2019 by Lawrence Livermore National Security, LLC. * Copyright (c) 2021 Hewlett Packard Enterprise Development LP * Copyright 2023 RackTop Systems, Inc. */ #include #include #include #include #include #include #include #include #include #include /* * TRIM is a feature which is used to notify a SSD that some previously * written space is no longer allocated by the pool. This is useful because * writes to a SSD must be performed to blocks which have first been erased. * Ensuring the SSD always has a supply of erased blocks for new writes * helps prevent the performance from deteriorating. * * There are two supported TRIM methods; manual and automatic. * * Manual TRIM: * * A manual TRIM is initiated by running the 'zpool trim' command. A single * 'vdev_trim' thread is created for each leaf vdev, and it is responsible for * managing that vdev TRIM process. This involves iterating over all the * metaslabs, calculating the unallocated space ranges, and then issuing the * required TRIM I/Os. * * While a metaslab is being actively trimmed it is not eligible to perform * new allocations. After traversing all of the metaslabs the thread is * terminated. Finally, both the requested options and current progress of * the TRIM are regularly written to the pool. This allows the TRIM to be * suspended and resumed as needed. * * Automatic TRIM: * * An automatic TRIM is enabled by setting the 'autotrim' pool property * to 'on'. When enabled, a `vdev_autotrim' thread is created for each * top-level (not leaf) vdev in the pool. These threads perform the same * core TRIM process as a manual TRIM, but with a few key differences. * * 1) Automatic TRIM happens continuously in the background and operates * solely on recently freed blocks (ms_trim not ms_allocatable). * * 2) Each thread is associated with a top-level (not leaf) vdev. This has * the benefit of simplifying the threading model, it makes it easier * to coordinate administrative commands, and it ensures only a single * metaslab is disabled at a time. Unlike manual TRIM, this means each * 'vdev_autotrim' thread is responsible for issuing TRIM I/Os for its * children. * * 3) There is no automatic TRIM progress information stored on disk, nor * is it reported by 'zpool status'. * * While the automatic TRIM process is highly effective it is more likely * than a manual TRIM to encounter tiny ranges. Ranges less than or equal to * 'zfs_trim_extent_bytes_min' (32k) are considered too small to efficiently * TRIM and are skipped. This means small amounts of freed space may not * be automatically trimmed. * * Furthermore, devices with attached hot spares and devices being actively * replaced are skipped. This is done to avoid adding additional stress to * a potentially unhealthy device and to minimize the required rebuild time. * * For this reason it may be beneficial to occasionally manually TRIM a pool * even when automatic TRIM is enabled. */ /* * Maximum size of TRIM I/O, ranges will be chunked in to 128MiB lengths. */ static unsigned int zfs_trim_extent_bytes_max = 128 * 1024 * 1024; /* * Minimum size of TRIM I/O, extents smaller than 32Kib will be skipped. */ static unsigned int zfs_trim_extent_bytes_min = 32 * 1024; /* * Skip uninitialized metaslabs during the TRIM process. This option is * useful for pools constructed from large thinly-provisioned devices where * TRIM operations are slow. As a pool ages an increasing fraction of * the pools metaslabs will be initialized progressively degrading the * usefulness of this option. This setting is stored when starting a * manual TRIM and will persist for the duration of the requested TRIM. */ unsigned int zfs_trim_metaslab_skip = 0; /* * Maximum number of queued TRIM I/Os per leaf vdev. The number of * concurrent TRIM I/Os issued to the device is controlled by the * zfs_vdev_trim_min_active and zfs_vdev_trim_max_active module options. */ static unsigned int zfs_trim_queue_limit = 10; /* * The minimum number of transaction groups between automatic trims of a * metaslab. This setting represents a trade-off between issuing more * efficient TRIM operations, by allowing them to be aggregated longer, * and issuing them promptly so the trimmed space is available. Note * that this value is a minimum; metaslabs can be trimmed less frequently * when there are a large number of ranges which need to be trimmed. * * Increasing this value will allow frees to be aggregated for a longer * time. This can result is larger TRIM operations, and increased memory * usage in order to track the ranges to be trimmed. Decreasing this value * has the opposite effect. The default value of 32 was determined though * testing to be a reasonable compromise. */ static unsigned int zfs_trim_txg_batch = 32; /* * The trim_args are a control structure which describe how a leaf vdev * should be trimmed. The core elements are the vdev, the metaslab being * trimmed and a range tree containing the extents to TRIM. All provided * ranges must be within the metaslab. */ typedef struct trim_args { /* * These fields are set by the caller of vdev_trim_ranges(). */ vdev_t *trim_vdev; /* Leaf vdev to TRIM */ metaslab_t *trim_msp; /* Disabled metaslab */ range_tree_t *trim_tree; /* TRIM ranges (in metaslab) */ trim_type_t trim_type; /* Manual or auto TRIM */ uint64_t trim_extent_bytes_max; /* Maximum TRIM I/O size */ uint64_t trim_extent_bytes_min; /* Minimum TRIM I/O size */ enum trim_flag trim_flags; /* TRIM flags (secure) */ /* * These fields are updated by vdev_trim_ranges(). */ hrtime_t trim_start_time; /* Start time */ uint64_t trim_bytes_done; /* Bytes trimmed */ } trim_args_t; /* * Determines whether a vdev_trim_thread() should be stopped. */ static boolean_t vdev_trim_should_stop(vdev_t *vd) { return (vd->vdev_trim_exit_wanted || !vdev_writeable(vd) || vd->vdev_detached || vd->vdev_top->vdev_removing || vd->vdev_top->vdev_rz_expanding); } /* * Determines whether a vdev_autotrim_thread() should be stopped. */ static boolean_t vdev_autotrim_should_stop(vdev_t *tvd) { return (tvd->vdev_autotrim_exit_wanted || !vdev_writeable(tvd) || tvd->vdev_removing || tvd->vdev_rz_expanding || spa_get_autotrim(tvd->vdev_spa) == SPA_AUTOTRIM_OFF); } /* * Wait for given number of kicks, return true if the wait is aborted due to * vdev_autotrim_exit_wanted. */ static boolean_t vdev_autotrim_wait_kick(vdev_t *vd, int num_of_kick) { mutex_enter(&vd->vdev_autotrim_lock); for (int i = 0; i < num_of_kick; i++) { if (vd->vdev_autotrim_exit_wanted) break; cv_wait_idle(&vd->vdev_autotrim_kick_cv, &vd->vdev_autotrim_lock); } boolean_t exit_wanted = vd->vdev_autotrim_exit_wanted; mutex_exit(&vd->vdev_autotrim_lock); return (exit_wanted); } /* * The sync task for updating the on-disk state of a manual TRIM. This * is scheduled by vdev_trim_change_state(). */ static void vdev_trim_zap_update_sync(void *arg, dmu_tx_t *tx) { /* * We pass in the guid instead of the vdev_t since the vdev may * have been freed prior to the sync task being processed. This * happens when a vdev is detached as we call spa_config_vdev_exit(), * stop the trimming thread, schedule the sync task, and free * the vdev. Later when the scheduled sync task is invoked, it would * find that the vdev has been freed. */ uint64_t guid = *(uint64_t *)arg; uint64_t txg = dmu_tx_get_txg(tx); kmem_free(arg, sizeof (uint64_t)); vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE); if (vd == NULL || vd->vdev_top->vdev_removing || !vdev_is_concrete(vd) || vd->vdev_top->vdev_rz_expanding) return; uint64_t last_offset = vd->vdev_trim_offset[txg & TXG_MASK]; vd->vdev_trim_offset[txg & TXG_MASK] = 0; VERIFY3U(vd->vdev_leaf_zap, !=, 0); objset_t *mos = vd->vdev_spa->spa_meta_objset; if (last_offset > 0 || vd->vdev_trim_last_offset == UINT64_MAX) { if (vd->vdev_trim_last_offset == UINT64_MAX) last_offset = 0; vd->vdev_trim_last_offset = last_offset; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_LAST_OFFSET, sizeof (last_offset), 1, &last_offset, tx)); } if (vd->vdev_trim_action_time > 0) { uint64_t val = (uint64_t)vd->vdev_trim_action_time; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_ACTION_TIME, sizeof (val), 1, &val, tx)); } if (vd->vdev_trim_rate > 0) { uint64_t rate = (uint64_t)vd->vdev_trim_rate; if (rate == UINT64_MAX) rate = 0; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_RATE, sizeof (rate), 1, &rate, tx)); } uint64_t partial = vd->vdev_trim_partial; if (partial == UINT64_MAX) partial = 0; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_PARTIAL, sizeof (partial), 1, &partial, tx)); uint64_t secure = vd->vdev_trim_secure; if (secure == UINT64_MAX) secure = 0; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_SECURE, sizeof (secure), 1, &secure, tx)); uint64_t trim_state = vd->vdev_trim_state; VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_STATE, sizeof (trim_state), 1, &trim_state, tx)); } /* * Update the on-disk state of a manual TRIM. This is called to request * that a TRIM be started/suspended/canceled, or to change one of the * TRIM options (partial, secure, rate). */ static void vdev_trim_change_state(vdev_t *vd, vdev_trim_state_t new_state, uint64_t rate, boolean_t partial, boolean_t secure) { ASSERT(MUTEX_HELD(&vd->vdev_trim_lock)); spa_t *spa = vd->vdev_spa; if (new_state == vd->vdev_trim_state) return; /* * Copy the vd's guid, this will be freed by the sync task. */ uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); *guid = vd->vdev_guid; /* * If we're suspending, then preserve the original start time. */ if (vd->vdev_trim_state != VDEV_TRIM_SUSPENDED) { vd->vdev_trim_action_time = gethrestime_sec(); } /* * If we're activating, then preserve the requested rate and trim * method. Setting the last offset and rate to UINT64_MAX is used * as a sentinel to indicate they should be reset to default values. */ if (new_state == VDEV_TRIM_ACTIVE) { if (vd->vdev_trim_state == VDEV_TRIM_COMPLETE || vd->vdev_trim_state == VDEV_TRIM_CANCELED) { vd->vdev_trim_last_offset = UINT64_MAX; vd->vdev_trim_rate = UINT64_MAX; vd->vdev_trim_partial = UINT64_MAX; vd->vdev_trim_secure = UINT64_MAX; } if (rate != 0) vd->vdev_trim_rate = rate; if (partial != 0) vd->vdev_trim_partial = partial; if (secure != 0) vd->vdev_trim_secure = secure; } vdev_trim_state_t old_state = vd->vdev_trim_state; boolean_t resumed = (old_state == VDEV_TRIM_SUSPENDED); vd->vdev_trim_state = new_state; dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); dsl_sync_task_nowait(spa_get_dsl(spa), vdev_trim_zap_update_sync, guid, tx); switch (new_state) { case VDEV_TRIM_ACTIVE: spa_event_notify(spa, vd, NULL, resumed ? ESC_ZFS_TRIM_RESUME : ESC_ZFS_TRIM_START); spa_history_log_internal(spa, "trim", tx, "vdev=%s activated", vd->vdev_path); break; case VDEV_TRIM_SUSPENDED: spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_SUSPEND); spa_history_log_internal(spa, "trim", tx, "vdev=%s suspended", vd->vdev_path); break; case VDEV_TRIM_CANCELED: if (old_state == VDEV_TRIM_ACTIVE || old_state == VDEV_TRIM_SUSPENDED) { spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_CANCEL); spa_history_log_internal(spa, "trim", tx, "vdev=%s canceled", vd->vdev_path); } break; case VDEV_TRIM_COMPLETE: spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_FINISH); spa_history_log_internal(spa, "trim", tx, "vdev=%s complete", vd->vdev_path); break; default: panic("invalid state %llu", (unsigned long long)new_state); } dmu_tx_commit(tx); if (new_state != VDEV_TRIM_ACTIVE) spa_notify_waiters(spa); } /* * The zio_done_func_t done callback for each manual TRIM issued. It is * responsible for updating the TRIM stats, reissuing failed TRIM I/Os, * and limiting the number of in flight TRIM I/Os. */ static void vdev_trim_cb(zio_t *zio) { vdev_t *vd = zio->io_vd; mutex_enter(&vd->vdev_trim_io_lock); if (zio->io_error == ENXIO && !vdev_writeable(vd)) { /* * The I/O failed because the vdev was unavailable; roll the * last offset back. (This works because spa_sync waits on * spa_txg_zio before it runs sync tasks.) */ uint64_t *offset = &vd->vdev_trim_offset[zio->io_txg & TXG_MASK]; *offset = MIN(*offset, zio->io_offset); } else { if (zio->io_error != 0) { vd->vdev_stat.vs_trim_errors++; spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_MANUAL, 0, 0, 0, 0, 1, zio->io_orig_size); } else { spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_MANUAL, 1, zio->io_orig_size, 0, 0, 0, 0); } vd->vdev_trim_bytes_done += zio->io_orig_size; } ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_MANUAL], >, 0); vd->vdev_trim_inflight[TRIM_TYPE_MANUAL]--; cv_broadcast(&vd->vdev_trim_io_cv); mutex_exit(&vd->vdev_trim_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); } /* * The zio_done_func_t done callback for each automatic TRIM issued. It * is responsible for updating the TRIM stats and limiting the number of * in flight TRIM I/Os. Automatic TRIM I/Os are best effort and are * never reissued on failure. */ static void vdev_autotrim_cb(zio_t *zio) { vdev_t *vd = zio->io_vd; mutex_enter(&vd->vdev_trim_io_lock); if (zio->io_error != 0) { vd->vdev_stat.vs_trim_errors++; spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_AUTO, 0, 0, 0, 0, 1, zio->io_orig_size); } else { spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_AUTO, 1, zio->io_orig_size, 0, 0, 0, 0); } ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_AUTO], >, 0); vd->vdev_trim_inflight[TRIM_TYPE_AUTO]--; cv_broadcast(&vd->vdev_trim_io_cv); mutex_exit(&vd->vdev_trim_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); } /* * The zio_done_func_t done callback for each TRIM issued via * vdev_trim_simple(). It is responsible for updating the TRIM stats and * limiting the number of in flight TRIM I/Os. Simple TRIM I/Os are best * effort and are never reissued on failure. */ static void vdev_trim_simple_cb(zio_t *zio) { vdev_t *vd = zio->io_vd; mutex_enter(&vd->vdev_trim_io_lock); if (zio->io_error != 0) { vd->vdev_stat.vs_trim_errors++; spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_SIMPLE, 0, 0, 0, 0, 1, zio->io_orig_size); } else { spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_SIMPLE, 1, zio->io_orig_size, 0, 0, 0, 0); } ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE], >, 0); vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE]--; cv_broadcast(&vd->vdev_trim_io_cv); mutex_exit(&vd->vdev_trim_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); } /* * Returns the average trim rate in bytes/sec for the ta->trim_vdev. */ static uint64_t vdev_trim_calculate_rate(trim_args_t *ta) { return (ta->trim_bytes_done * 1000 / (NSEC2MSEC(gethrtime() - ta->trim_start_time) + 1)); } /* * Issues a physical TRIM and takes care of rate limiting (bytes/sec) * and number of concurrent TRIM I/Os. */ static int vdev_trim_range(trim_args_t *ta, uint64_t start, uint64_t size) { vdev_t *vd = ta->trim_vdev; spa_t *spa = vd->vdev_spa; void *cb; mutex_enter(&vd->vdev_trim_io_lock); /* * Limit manual TRIM I/Os to the requested rate. This does not * apply to automatic TRIM since no per vdev rate can be specified. */ if (ta->trim_type == TRIM_TYPE_MANUAL) { while (vd->vdev_trim_rate != 0 && !vdev_trim_should_stop(vd) && vdev_trim_calculate_rate(ta) > vd->vdev_trim_rate) { cv_timedwait_idle(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock, ddi_get_lbolt() + MSEC_TO_TICK(10)); } } ta->trim_bytes_done += size; /* Limit in flight trimming I/Os */ while (vd->vdev_trim_inflight[0] + vd->vdev_trim_inflight[1] + vd->vdev_trim_inflight[2] >= zfs_trim_queue_limit) { cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock); } vd->vdev_trim_inflight[ta->trim_type]++; mutex_exit(&vd->vdev_trim_io_lock); dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); uint64_t txg = dmu_tx_get_txg(tx); spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER); mutex_enter(&vd->vdev_trim_lock); if (ta->trim_type == TRIM_TYPE_MANUAL && vd->vdev_trim_offset[txg & TXG_MASK] == 0) { uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); *guid = vd->vdev_guid; /* This is the first write of this txg. */ dsl_sync_task_nowait(spa_get_dsl(spa), vdev_trim_zap_update_sync, guid, tx); } /* * We know the vdev_t will still be around since all consumers of * vdev_free must stop the trimming first. */ if ((ta->trim_type == TRIM_TYPE_MANUAL && vdev_trim_should_stop(vd)) || (ta->trim_type == TRIM_TYPE_AUTO && vdev_autotrim_should_stop(vd->vdev_top))) { mutex_enter(&vd->vdev_trim_io_lock); vd->vdev_trim_inflight[ta->trim_type]--; mutex_exit(&vd->vdev_trim_io_lock); spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd); mutex_exit(&vd->vdev_trim_lock); dmu_tx_commit(tx); return (SET_ERROR(EINTR)); } mutex_exit(&vd->vdev_trim_lock); if (ta->trim_type == TRIM_TYPE_MANUAL) vd->vdev_trim_offset[txg & TXG_MASK] = start + size; if (ta->trim_type == TRIM_TYPE_MANUAL) { cb = vdev_trim_cb; } else if (ta->trim_type == TRIM_TYPE_AUTO) { cb = vdev_autotrim_cb; } else { cb = vdev_trim_simple_cb; } zio_nowait(zio_trim(spa->spa_txg_zio[txg & TXG_MASK], vd, start, size, cb, NULL, ZIO_PRIORITY_TRIM, ZIO_FLAG_CANFAIL, ta->trim_flags)); /* vdev_trim_cb and vdev_autotrim_cb release SCL_STATE_ALL */ dmu_tx_commit(tx); return (0); } /* * Issues TRIM I/Os for all ranges in the provided ta->trim_tree range tree. * Additional parameters describing how the TRIM should be performed must * be set in the trim_args structure. See the trim_args definition for * additional information. */ static int vdev_trim_ranges(trim_args_t *ta) { vdev_t *vd = ta->trim_vdev; zfs_btree_t *t = &ta->trim_tree->rt_root; zfs_btree_index_t idx; uint64_t extent_bytes_max = ta->trim_extent_bytes_max; uint64_t extent_bytes_min = ta->trim_extent_bytes_min; spa_t *spa = vd->vdev_spa; int error = 0; ta->trim_start_time = gethrtime(); ta->trim_bytes_done = 0; for (range_seg_t *rs = zfs_btree_first(t, &idx); rs != NULL; rs = zfs_btree_next(t, &idx, &idx)) { uint64_t size = rs_get_end(rs, ta->trim_tree) - rs_get_start(rs, ta->trim_tree); if (extent_bytes_min && size < extent_bytes_min) { spa_iostats_trim_add(spa, ta->trim_type, 0, 0, 1, size, 0, 0); continue; } /* Split range into legally-sized physical chunks */ uint64_t writes_required = ((size - 1) / extent_bytes_max) + 1; for (uint64_t w = 0; w < writes_required; w++) { error = vdev_trim_range(ta, VDEV_LABEL_START_SIZE + rs_get_start(rs, ta->trim_tree) + (w *extent_bytes_max), MIN(size - (w * extent_bytes_max), extent_bytes_max)); if (error != 0) { goto done; } } } done: /* * Make sure all TRIMs for this metaslab have completed before * returning. TRIM zios have lower priority over regular or syncing * zios, so all TRIM zios for this metaslab must complete before the * metaslab is re-enabled. Otherwise it's possible write zios to * this metaslab could cut ahead of still queued TRIM zios for this * metaslab causing corruption if the ranges overlap. */ mutex_enter(&vd->vdev_trim_io_lock); while (vd->vdev_trim_inflight[0] > 0) { cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock); } mutex_exit(&vd->vdev_trim_io_lock); return (error); } static void vdev_trim_xlate_last_rs_end(void *arg, range_seg64_t *physical_rs) { uint64_t *last_rs_end = (uint64_t *)arg; if (physical_rs->rs_end > *last_rs_end) *last_rs_end = physical_rs->rs_end; } static void vdev_trim_xlate_progress(void *arg, range_seg64_t *physical_rs) { vdev_t *vd = (vdev_t *)arg; uint64_t size = physical_rs->rs_end - physical_rs->rs_start; vd->vdev_trim_bytes_est += size; if (vd->vdev_trim_last_offset >= physical_rs->rs_end) { vd->vdev_trim_bytes_done += size; } else if (vd->vdev_trim_last_offset > physical_rs->rs_start && vd->vdev_trim_last_offset <= physical_rs->rs_end) { vd->vdev_trim_bytes_done += vd->vdev_trim_last_offset - physical_rs->rs_start; } } /* * Calculates the completion percentage of a manual TRIM. */ static void vdev_trim_calculate_progress(vdev_t *vd) { ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) || spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER)); ASSERT(vd->vdev_leaf_zap != 0); vd->vdev_trim_bytes_est = 0; vd->vdev_trim_bytes_done = 0; for (uint64_t i = 0; i < vd->vdev_top->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_top->vdev_ms[i]; mutex_enter(&msp->ms_lock); uint64_t ms_free = (msp->ms_size - metaslab_allocated_space(msp)) / vdev_get_ndisks(vd->vdev_top); /* * Convert the metaslab range to a physical range * on our vdev. We use this to determine if we are * in the middle of this metaslab range. */ range_seg64_t logical_rs, physical_rs, remain_rs; logical_rs.rs_start = msp->ms_start; logical_rs.rs_end = msp->ms_start + msp->ms_size; /* Metaslab space after this offset has not been trimmed. */ vdev_xlate(vd, &logical_rs, &physical_rs, &remain_rs); if (vd->vdev_trim_last_offset <= physical_rs.rs_start) { vd->vdev_trim_bytes_est += ms_free; mutex_exit(&msp->ms_lock); continue; } /* Metaslab space before this offset has been trimmed */ uint64_t last_rs_end = physical_rs.rs_end; if (!vdev_xlate_is_empty(&remain_rs)) { vdev_xlate_walk(vd, &remain_rs, vdev_trim_xlate_last_rs_end, &last_rs_end); } if (vd->vdev_trim_last_offset > last_rs_end) { vd->vdev_trim_bytes_done += ms_free; vd->vdev_trim_bytes_est += ms_free; mutex_exit(&msp->ms_lock); continue; } /* * If we get here, we're in the middle of trimming this * metaslab. Load it and walk the free tree for more * accurate progress estimation. */ VERIFY0(metaslab_load(msp)); range_tree_t *rt = msp->ms_allocatable; zfs_btree_t *bt = &rt->rt_root; zfs_btree_index_t idx; for (range_seg_t *rs = zfs_btree_first(bt, &idx); rs != NULL; rs = zfs_btree_next(bt, &idx, &idx)) { logical_rs.rs_start = rs_get_start(rs, rt); logical_rs.rs_end = rs_get_end(rs, rt); vdev_xlate_walk(vd, &logical_rs, vdev_trim_xlate_progress, vd); } mutex_exit(&msp->ms_lock); } } /* * Load from disk the vdev's manual TRIM information. This includes the * state, progress, and options provided when initiating the manual TRIM. */ static int vdev_trim_load(vdev_t *vd) { int err = 0; ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) || spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER)); ASSERT(vd->vdev_leaf_zap != 0); if (vd->vdev_trim_state == VDEV_TRIM_ACTIVE || vd->vdev_trim_state == VDEV_TRIM_SUSPENDED) { err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_LAST_OFFSET, sizeof (vd->vdev_trim_last_offset), 1, &vd->vdev_trim_last_offset); if (err == ENOENT) { vd->vdev_trim_last_offset = 0; err = 0; } if (err == 0) { err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_RATE, sizeof (vd->vdev_trim_rate), 1, &vd->vdev_trim_rate); if (err == ENOENT) { vd->vdev_trim_rate = 0; err = 0; } } if (err == 0) { err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_PARTIAL, sizeof (vd->vdev_trim_partial), 1, &vd->vdev_trim_partial); if (err == ENOENT) { vd->vdev_trim_partial = 0; err = 0; } } if (err == 0) { err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_SECURE, sizeof (vd->vdev_trim_secure), 1, &vd->vdev_trim_secure); if (err == ENOENT) { vd->vdev_trim_secure = 0; err = 0; } } } vdev_trim_calculate_progress(vd); return (err); } static void vdev_trim_xlate_range_add(void *arg, range_seg64_t *physical_rs) { trim_args_t *ta = arg; vdev_t *vd = ta->trim_vdev; /* * Only a manual trim will be traversing the vdev sequentially. * For an auto trim all valid ranges should be added. */ if (ta->trim_type == TRIM_TYPE_MANUAL) { /* Only add segments that we have not visited yet */ if (physical_rs->rs_end <= vd->vdev_trim_last_offset) return; /* Pick up where we left off mid-range. */ if (vd->vdev_trim_last_offset > physical_rs->rs_start) { ASSERT3U(physical_rs->rs_end, >, vd->vdev_trim_last_offset); physical_rs->rs_start = vd->vdev_trim_last_offset; } } ASSERT3U(physical_rs->rs_end, >, physical_rs->rs_start); range_tree_add(ta->trim_tree, physical_rs->rs_start, physical_rs->rs_end - physical_rs->rs_start); } /* * Convert the logical range into physical ranges and add them to the * range tree passed in the trim_args_t. */ static void vdev_trim_range_add(void *arg, uint64_t start, uint64_t size) { trim_args_t *ta = arg; vdev_t *vd = ta->trim_vdev; range_seg64_t logical_rs; logical_rs.rs_start = start; logical_rs.rs_end = start + size; /* * Every range to be trimmed must be part of ms_allocatable. * When ZFS_DEBUG_TRIM is set load the metaslab to verify this * is always the case. */ if (zfs_flags & ZFS_DEBUG_TRIM) { metaslab_t *msp = ta->trim_msp; VERIFY0(metaslab_load(msp)); VERIFY3B(msp->ms_loaded, ==, B_TRUE); VERIFY(range_tree_contains(msp->ms_allocatable, start, size)); } ASSERT(vd->vdev_ops->vdev_op_leaf); vdev_xlate_walk(vd, &logical_rs, vdev_trim_xlate_range_add, arg); } /* * Each manual TRIM thread is responsible for trimming the unallocated * space for each leaf vdev. This is accomplished by sequentially iterating * over its top-level metaslabs and issuing TRIM I/O for the space described * by its ms_allocatable. While a metaslab is undergoing trimming it is * not eligible for new allocations. */ static __attribute__((noreturn)) void vdev_trim_thread(void *arg) { vdev_t *vd = arg; spa_t *spa = vd->vdev_spa; trim_args_t ta; int error = 0; /* * The VDEV_LEAF_ZAP_TRIM_* entries may have been updated by * vdev_trim(). Wait for the updated values to be reflected * in the zap in order to start with the requested settings. */ txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0); ASSERT(vdev_is_concrete(vd)); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vd->vdev_trim_last_offset = 0; vd->vdev_trim_rate = 0; vd->vdev_trim_partial = 0; vd->vdev_trim_secure = 0; VERIFY0(vdev_trim_load(vd)); ta.trim_vdev = vd; ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max; ta.trim_extent_bytes_min = zfs_trim_extent_bytes_min; ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); ta.trim_type = TRIM_TYPE_MANUAL; ta.trim_flags = 0; /* * When a secure TRIM has been requested infer that the intent * is that everything must be trimmed. Override the default * minimum TRIM size to prevent ranges from being skipped. */ if (vd->vdev_trim_secure) { ta.trim_flags |= ZIO_TRIM_SECURE; ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE; } uint64_t ms_count = 0; for (uint64_t i = 0; !vd->vdev_detached && i < vd->vdev_top->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_top->vdev_ms[i]; /* * If we've expanded the top-level vdev or it's our * first pass, calculate our progress. */ if (vd->vdev_top->vdev_ms_count != ms_count) { vdev_trim_calculate_progress(vd); ms_count = vd->vdev_top->vdev_ms_count; } spa_config_exit(spa, SCL_CONFIG, FTAG); metaslab_disable(msp); mutex_enter(&msp->ms_lock); VERIFY0(metaslab_load(msp)); /* * If a partial TRIM was requested skip metaslabs which have * never been initialized and thus have never been written. */ if (msp->ms_sm == NULL && vd->vdev_trim_partial) { mutex_exit(&msp->ms_lock); metaslab_enable(msp, B_FALSE, B_FALSE); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vdev_trim_calculate_progress(vd); continue; } ta.trim_msp = msp; range_tree_walk(msp->ms_allocatable, vdev_trim_range_add, &ta); range_tree_vacate(msp->ms_trim, NULL, NULL); mutex_exit(&msp->ms_lock); error = vdev_trim_ranges(&ta); metaslab_enable(msp, B_TRUE, B_FALSE); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); range_tree_vacate(ta.trim_tree, NULL, NULL); if (error != 0) break; } spa_config_exit(spa, SCL_CONFIG, FTAG); range_tree_destroy(ta.trim_tree); mutex_enter(&vd->vdev_trim_lock); if (!vd->vdev_trim_exit_wanted) { if (vdev_writeable(vd)) { vdev_trim_change_state(vd, VDEV_TRIM_COMPLETE, vd->vdev_trim_rate, vd->vdev_trim_partial, vd->vdev_trim_secure); } else if (vd->vdev_faulted) { vdev_trim_change_state(vd, VDEV_TRIM_CANCELED, vd->vdev_trim_rate, vd->vdev_trim_partial, vd->vdev_trim_secure); } } ASSERT(vd->vdev_trim_thread != NULL || vd->vdev_trim_inflight[0] == 0); /* * Drop the vdev_trim_lock while we sync out the txg since it's * possible that a device might be trying to come online and must * check to see if it needs to restart a trim. That thread will be * holding the spa_config_lock which would prevent the txg_wait_synced * from completing. */ mutex_exit(&vd->vdev_trim_lock); txg_wait_synced(spa_get_dsl(spa), 0); mutex_enter(&vd->vdev_trim_lock); vd->vdev_trim_thread = NULL; cv_broadcast(&vd->vdev_trim_cv); mutex_exit(&vd->vdev_trim_lock); thread_exit(); } /* * Initiates a manual TRIM for the vdev_t. Callers must hold vdev_trim_lock, * the vdev_t must be a leaf and cannot already be manually trimming. */ void vdev_trim(vdev_t *vd, uint64_t rate, boolean_t partial, boolean_t secure) { ASSERT(MUTEX_HELD(&vd->vdev_trim_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); ASSERT3P(vd->vdev_trim_thread, ==, NULL); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_trim_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); ASSERT(!vd->vdev_rz_expanding); vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, rate, partial, secure); vd->vdev_trim_thread = thread_create(NULL, 0, vdev_trim_thread, vd, 0, &p0, TS_RUN, maxclsyspri); } /* * Wait for the trimming thread to be terminated (canceled or stopped). */ static void vdev_trim_stop_wait_impl(vdev_t *vd) { ASSERT(MUTEX_HELD(&vd->vdev_trim_lock)); while (vd->vdev_trim_thread != NULL) cv_wait(&vd->vdev_trim_cv, &vd->vdev_trim_lock); ASSERT3P(vd->vdev_trim_thread, ==, NULL); vd->vdev_trim_exit_wanted = B_FALSE; } /* * Wait for vdev trim threads which were listed to cleanly exit. */ void vdev_trim_stop_wait(spa_t *spa, list_t *vd_list) { (void) spa; vdev_t *vd; ASSERT(MUTEX_HELD(&spa_namespace_lock)); while ((vd = list_remove_head(vd_list)) != NULL) { mutex_enter(&vd->vdev_trim_lock); vdev_trim_stop_wait_impl(vd); mutex_exit(&vd->vdev_trim_lock); } } /* * Stop trimming a device, with the resultant trimming state being tgt_state. * For blocking behavior pass NULL for vd_list. Otherwise, when a list_t is * provided the stopping vdev is inserted in to the list. Callers are then * required to call vdev_trim_stop_wait() to block for all the trim threads * to exit. The caller must hold vdev_trim_lock and must not be writing to * the spa config, as the trimming thread may try to enter the config as a * reader before exiting. */ void vdev_trim_stop(vdev_t *vd, vdev_trim_state_t tgt_state, list_t *vd_list) { ASSERT(!spa_config_held(vd->vdev_spa, SCL_CONFIG|SCL_STATE, RW_WRITER)); ASSERT(MUTEX_HELD(&vd->vdev_trim_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); /* * Allow cancel requests to proceed even if the trim thread has * stopped. */ if (vd->vdev_trim_thread == NULL && tgt_state != VDEV_TRIM_CANCELED) return; vdev_trim_change_state(vd, tgt_state, 0, 0, 0); vd->vdev_trim_exit_wanted = B_TRUE; if (vd_list == NULL) { vdev_trim_stop_wait_impl(vd); } else { ASSERT(MUTEX_HELD(&spa_namespace_lock)); list_insert_tail(vd_list, vd); } } /* * Requests that all listed vdevs stop trimming. */ static void vdev_trim_stop_all_impl(vdev_t *vd, vdev_trim_state_t tgt_state, list_t *vd_list) { if (vd->vdev_ops->vdev_op_leaf && vdev_is_concrete(vd)) { mutex_enter(&vd->vdev_trim_lock); vdev_trim_stop(vd, tgt_state, vd_list); mutex_exit(&vd->vdev_trim_lock); return; } for (uint64_t i = 0; i < vd->vdev_children; i++) { vdev_trim_stop_all_impl(vd->vdev_child[i], tgt_state, vd_list); } } /* * Convenience function to stop trimming of a vdev tree and set all trim * thread pointers to NULL. */ void vdev_trim_stop_all(vdev_t *vd, vdev_trim_state_t tgt_state) { spa_t *spa = vd->vdev_spa; list_t vd_list; vdev_t *vd_l2cache; ASSERT(MUTEX_HELD(&spa_namespace_lock)); list_create(&vd_list, sizeof (vdev_t), offsetof(vdev_t, vdev_trim_node)); vdev_trim_stop_all_impl(vd, tgt_state, &vd_list); /* * Iterate over cache devices and request stop trimming the * whole device in case we export the pool or remove the cache * device prematurely. */ for (int i = 0; i < spa->spa_l2cache.sav_count; i++) { vd_l2cache = spa->spa_l2cache.sav_vdevs[i]; vdev_trim_stop_all_impl(vd_l2cache, tgt_state, &vd_list); } vdev_trim_stop_wait(spa, &vd_list); if (vd->vdev_spa->spa_sync_on) { /* Make sure that our state has been synced to disk */ txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0); } list_destroy(&vd_list); } /* * Conditionally restarts a manual TRIM given its on-disk state. */ void vdev_trim_restart(vdev_t *vd) { - ASSERT(MUTEX_HELD(&spa_namespace_lock)); + ASSERT(MUTEX_HELD(&spa_namespace_lock) || + vd->vdev_spa->spa_load_thread == curthread); ASSERT(!spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER)); if (vd->vdev_leaf_zap != 0) { mutex_enter(&vd->vdev_trim_lock); uint64_t trim_state = VDEV_TRIM_NONE; int err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_STATE, sizeof (trim_state), 1, &trim_state); ASSERT(err == 0 || err == ENOENT); vd->vdev_trim_state = trim_state; uint64_t timestamp = 0; err = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_ACTION_TIME, sizeof (timestamp), 1, ×tamp); ASSERT(err == 0 || err == ENOENT); vd->vdev_trim_action_time = timestamp; if ((vd->vdev_trim_state == VDEV_TRIM_SUSPENDED || vd->vdev_offline) && !vd->vdev_top->vdev_rz_expanding) { /* load progress for reporting, but don't resume */ VERIFY0(vdev_trim_load(vd)); } else if (vd->vdev_trim_state == VDEV_TRIM_ACTIVE && vdev_writeable(vd) && !vd->vdev_top->vdev_removing && !vd->vdev_top->vdev_rz_expanding && vd->vdev_trim_thread == NULL) { VERIFY0(vdev_trim_load(vd)); vdev_trim(vd, vd->vdev_trim_rate, vd->vdev_trim_partial, vd->vdev_trim_secure); } mutex_exit(&vd->vdev_trim_lock); } for (uint64_t i = 0; i < vd->vdev_children; i++) { vdev_trim_restart(vd->vdev_child[i]); } } /* * Used by the automatic TRIM when ZFS_DEBUG_TRIM is set to verify that * every TRIM range is contained within ms_allocatable. */ static void vdev_trim_range_verify(void *arg, uint64_t start, uint64_t size) { trim_args_t *ta = arg; metaslab_t *msp = ta->trim_msp; VERIFY3B(msp->ms_loaded, ==, B_TRUE); VERIFY3U(msp->ms_disabled, >, 0); VERIFY(range_tree_contains(msp->ms_allocatable, start, size)); } /* * Each automatic TRIM thread is responsible for managing the trimming of a * top-level vdev in the pool. No automatic TRIM state is maintained on-disk. * * N.B. This behavior is different from a manual TRIM where a thread * is created for each leaf vdev, instead of each top-level vdev. */ static __attribute__((noreturn)) void vdev_autotrim_thread(void *arg) { vdev_t *vd = arg; spa_t *spa = vd->vdev_spa; int shift = 0; mutex_enter(&vd->vdev_autotrim_lock); ASSERT3P(vd->vdev_top, ==, vd); ASSERT3P(vd->vdev_autotrim_thread, !=, NULL); mutex_exit(&vd->vdev_autotrim_lock); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); while (!vdev_autotrim_should_stop(vd)) { int txgs_per_trim = MAX(zfs_trim_txg_batch, 1); uint64_t extent_bytes_max = zfs_trim_extent_bytes_max; uint64_t extent_bytes_min = zfs_trim_extent_bytes_min; /* * All of the metaslabs are divided in to groups of size * num_metaslabs / zfs_trim_txg_batch. Each of these groups * is composed of metaslabs which are spread evenly over the * device. * * For example, when zfs_trim_txg_batch = 32 (default) then * group 0 will contain metaslabs 0, 32, 64, ...; * group 1 will contain metaslabs 1, 33, 65, ...; * group 2 will contain metaslabs 2, 34, 66, ...; and so on. * * On each pass through the while() loop one of these groups * is selected. This is accomplished by using a shift value * to select the starting metaslab, then striding over the * metaslabs using the zfs_trim_txg_batch size. This is * done to accomplish two things. * * 1) By dividing the metaslabs in to groups, and making sure * that each group takes a minimum of one txg to process. * Then zfs_trim_txg_batch controls the minimum number of * txgs which must occur before a metaslab is revisited. * * 2) Selecting non-consecutive metaslabs distributes the * TRIM commands for a group evenly over the entire device. * This can be advantageous for certain types of devices. */ for (uint64_t i = shift % txgs_per_trim; i < vd->vdev_ms_count; i += txgs_per_trim) { metaslab_t *msp = vd->vdev_ms[i]; range_tree_t *trim_tree; boolean_t issued_trim = B_FALSE; boolean_t wait_aborted = B_FALSE; spa_config_exit(spa, SCL_CONFIG, FTAG); metaslab_disable(msp); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); mutex_enter(&msp->ms_lock); /* * Skip the metaslab when it has never been allocated * or when there are no recent frees to trim. */ if (msp->ms_sm == NULL || range_tree_is_empty(msp->ms_trim)) { mutex_exit(&msp->ms_lock); metaslab_enable(msp, B_FALSE, B_FALSE); continue; } /* * Skip the metaslab when it has already been disabled. * This may happen when a manual TRIM or initialize * operation is running concurrently. In the case * of a manual TRIM, the ms_trim tree will have been * vacated. Only ranges added after the manual TRIM * disabled the metaslab will be included in the tree. * These will be processed when the automatic TRIM * next revisits this metaslab. */ if (msp->ms_disabled > 1) { mutex_exit(&msp->ms_lock); metaslab_enable(msp, B_FALSE, B_FALSE); continue; } /* * Allocate an empty range tree which is swapped in * for the existing ms_trim tree while it is processed. */ trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); range_tree_swap(&msp->ms_trim, &trim_tree); ASSERT(range_tree_is_empty(msp->ms_trim)); /* * There are two cases when constructing the per-vdev * trim trees for a metaslab. If the top-level vdev * has no children then it is also a leaf and should * be trimmed. Otherwise our children are the leaves * and a trim tree should be constructed for each. */ trim_args_t *tap; uint64_t children = vd->vdev_children; if (children == 0) { children = 1; tap = kmem_zalloc(sizeof (trim_args_t) * children, KM_SLEEP); tap[0].trim_vdev = vd; } else { tap = kmem_zalloc(sizeof (trim_args_t) * children, KM_SLEEP); for (uint64_t c = 0; c < children; c++) { tap[c].trim_vdev = vd->vdev_child[c]; } } for (uint64_t c = 0; c < children; c++) { trim_args_t *ta = &tap[c]; vdev_t *cvd = ta->trim_vdev; ta->trim_msp = msp; ta->trim_extent_bytes_max = extent_bytes_max; ta->trim_extent_bytes_min = extent_bytes_min; ta->trim_type = TRIM_TYPE_AUTO; ta->trim_flags = 0; if (cvd->vdev_detached || !vdev_writeable(cvd) || !cvd->vdev_has_trim || cvd->vdev_trim_thread != NULL) { continue; } /* * When a device has an attached hot spare, or * is being replaced it will not be trimmed. * This is done to avoid adding additional * stress to a potentially unhealthy device, * and to minimize the required rebuild time. */ if (!cvd->vdev_ops->vdev_op_leaf) continue; ta->trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); range_tree_walk(trim_tree, vdev_trim_range_add, ta); } mutex_exit(&msp->ms_lock); spa_config_exit(spa, SCL_CONFIG, FTAG); /* * Issue the TRIM I/Os for all ranges covered by the * TRIM trees. These ranges are safe to TRIM because * no new allocations will be performed until the call * to metaslab_enabled() below. */ for (uint64_t c = 0; c < children; c++) { trim_args_t *ta = &tap[c]; /* * Always yield to a manual TRIM if one has * been started for the child vdev. */ if (ta->trim_tree == NULL || ta->trim_vdev->vdev_trim_thread != NULL) { continue; } /* * After this point metaslab_enable() must be * called with the sync flag set. This is done * here because vdev_trim_ranges() is allowed * to be interrupted (EINTR) before issuing all * of the required TRIM I/Os. */ issued_trim = B_TRUE; int error = vdev_trim_ranges(ta); if (error) break; } /* * Verify every range which was trimmed is still * contained within the ms_allocatable tree. */ if (zfs_flags & ZFS_DEBUG_TRIM) { mutex_enter(&msp->ms_lock); VERIFY0(metaslab_load(msp)); VERIFY3P(tap[0].trim_msp, ==, msp); range_tree_walk(trim_tree, vdev_trim_range_verify, &tap[0]); mutex_exit(&msp->ms_lock); } range_tree_vacate(trim_tree, NULL, NULL); range_tree_destroy(trim_tree); /* * Wait for couples of kicks, to ensure the trim io is * synced. If the wait is aborted due to * vdev_autotrim_exit_wanted, we need to signal * metaslab_enable() to wait for sync. */ if (issued_trim) { wait_aborted = vdev_autotrim_wait_kick(vd, TXG_CONCURRENT_STATES + TXG_DEFER_SIZE); } metaslab_enable(msp, wait_aborted, B_FALSE); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); for (uint64_t c = 0; c < children; c++) { trim_args_t *ta = &tap[c]; if (ta->trim_tree == NULL) continue; range_tree_vacate(ta->trim_tree, NULL, NULL); range_tree_destroy(ta->trim_tree); } kmem_free(tap, sizeof (trim_args_t) * children); if (vdev_autotrim_should_stop(vd)) break; } spa_config_exit(spa, SCL_CONFIG, FTAG); vdev_autotrim_wait_kick(vd, 1); shift++; spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); } for (uint64_t c = 0; c < vd->vdev_children; c++) { vdev_t *cvd = vd->vdev_child[c]; mutex_enter(&cvd->vdev_trim_io_lock); while (cvd->vdev_trim_inflight[1] > 0) { cv_wait(&cvd->vdev_trim_io_cv, &cvd->vdev_trim_io_lock); } mutex_exit(&cvd->vdev_trim_io_lock); } spa_config_exit(spa, SCL_CONFIG, FTAG); /* * When exiting because the autotrim property was set to off, then * abandon any unprocessed ms_trim ranges to reclaim the memory. */ if (spa_get_autotrim(spa) == SPA_AUTOTRIM_OFF) { for (uint64_t i = 0; i < vd->vdev_ms_count; i++) { metaslab_t *msp = vd->vdev_ms[i]; mutex_enter(&msp->ms_lock); range_tree_vacate(msp->ms_trim, NULL, NULL); mutex_exit(&msp->ms_lock); } } mutex_enter(&vd->vdev_autotrim_lock); ASSERT(vd->vdev_autotrim_thread != NULL); vd->vdev_autotrim_thread = NULL; cv_broadcast(&vd->vdev_autotrim_cv); mutex_exit(&vd->vdev_autotrim_lock); thread_exit(); } /* * Starts an autotrim thread, if needed, for each top-level vdev which can be * trimmed. A top-level vdev which has been evacuated will never be trimmed. */ void vdev_autotrim(spa_t *spa) { vdev_t *root_vd = spa->spa_root_vdev; for (uint64_t i = 0; i < root_vd->vdev_children; i++) { vdev_t *tvd = root_vd->vdev_child[i]; mutex_enter(&tvd->vdev_autotrim_lock); if (vdev_writeable(tvd) && !tvd->vdev_removing && tvd->vdev_autotrim_thread == NULL && !tvd->vdev_rz_expanding) { ASSERT3P(tvd->vdev_top, ==, tvd); tvd->vdev_autotrim_thread = thread_create(NULL, 0, vdev_autotrim_thread, tvd, 0, &p0, TS_RUN, maxclsyspri); ASSERT(tvd->vdev_autotrim_thread != NULL); } mutex_exit(&tvd->vdev_autotrim_lock); } } /* * Wait for the vdev_autotrim_thread associated with the passed top-level * vdev to be terminated (canceled or stopped). */ void vdev_autotrim_stop_wait(vdev_t *tvd) { mutex_enter(&tvd->vdev_autotrim_lock); if (tvd->vdev_autotrim_thread != NULL) { tvd->vdev_autotrim_exit_wanted = B_TRUE; cv_broadcast(&tvd->vdev_autotrim_kick_cv); cv_wait(&tvd->vdev_autotrim_cv, &tvd->vdev_autotrim_lock); ASSERT3P(tvd->vdev_autotrim_thread, ==, NULL); tvd->vdev_autotrim_exit_wanted = B_FALSE; } mutex_exit(&tvd->vdev_autotrim_lock); } void vdev_autotrim_kick(spa_t *spa) { ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); vdev_t *root_vd = spa->spa_root_vdev; vdev_t *tvd; for (uint64_t i = 0; i < root_vd->vdev_children; i++) { tvd = root_vd->vdev_child[i]; mutex_enter(&tvd->vdev_autotrim_lock); if (tvd->vdev_autotrim_thread != NULL) cv_broadcast(&tvd->vdev_autotrim_kick_cv); mutex_exit(&tvd->vdev_autotrim_lock); } } /* * Wait for all of the vdev_autotrim_thread associated with the pool to * be terminated (canceled or stopped). */ void vdev_autotrim_stop_all(spa_t *spa) { vdev_t *root_vd = spa->spa_root_vdev; for (uint64_t i = 0; i < root_vd->vdev_children; i++) vdev_autotrim_stop_wait(root_vd->vdev_child[i]); } /* * Conditionally restart all of the vdev_autotrim_thread's for the pool. */ void vdev_autotrim_restart(spa_t *spa) { - ASSERT(MUTEX_HELD(&spa_namespace_lock)); - + ASSERT(MUTEX_HELD(&spa_namespace_lock) || + spa->spa_load_thread == curthread); if (spa->spa_autotrim) vdev_autotrim(spa); } static __attribute__((noreturn)) void vdev_trim_l2arc_thread(void *arg) { vdev_t *vd = arg; spa_t *spa = vd->vdev_spa; l2arc_dev_t *dev = l2arc_vdev_get(vd); trim_args_t ta = {0}; range_seg64_t physical_rs; ASSERT(vdev_is_concrete(vd)); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); vd->vdev_trim_last_offset = 0; vd->vdev_trim_rate = 0; vd->vdev_trim_partial = 0; vd->vdev_trim_secure = 0; ta.trim_vdev = vd; ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); ta.trim_type = TRIM_TYPE_MANUAL; ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max; ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE; ta.trim_flags = 0; physical_rs.rs_start = vd->vdev_trim_bytes_done = 0; physical_rs.rs_end = vd->vdev_trim_bytes_est = vdev_get_min_asize(vd); range_tree_add(ta.trim_tree, physical_rs.rs_start, physical_rs.rs_end - physical_rs.rs_start); mutex_enter(&vd->vdev_trim_lock); vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, 0, 0, 0); mutex_exit(&vd->vdev_trim_lock); (void) vdev_trim_ranges(&ta); spa_config_exit(spa, SCL_CONFIG, FTAG); mutex_enter(&vd->vdev_trim_io_lock); while (vd->vdev_trim_inflight[TRIM_TYPE_MANUAL] > 0) { cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock); } mutex_exit(&vd->vdev_trim_io_lock); range_tree_vacate(ta.trim_tree, NULL, NULL); range_tree_destroy(ta.trim_tree); mutex_enter(&vd->vdev_trim_lock); if (!vd->vdev_trim_exit_wanted && vdev_writeable(vd)) { vdev_trim_change_state(vd, VDEV_TRIM_COMPLETE, vd->vdev_trim_rate, vd->vdev_trim_partial, vd->vdev_trim_secure); } ASSERT(vd->vdev_trim_thread != NULL || vd->vdev_trim_inflight[TRIM_TYPE_MANUAL] == 0); /* * Drop the vdev_trim_lock while we sync out the txg since it's * possible that a device might be trying to come online and * must check to see if it needs to restart a trim. That thread * will be holding the spa_config_lock which would prevent the * txg_wait_synced from completing. Same strategy as in * vdev_trim_thread(). */ mutex_exit(&vd->vdev_trim_lock); txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0); mutex_enter(&vd->vdev_trim_lock); /* * Update the header of the cache device here, before * broadcasting vdev_trim_cv which may lead to the removal * of the device. The same applies for setting l2ad_trim_all to * false. */ spa_config_enter(vd->vdev_spa, SCL_L2ARC, vd, RW_READER); memset(dev->l2ad_dev_hdr, 0, dev->l2ad_dev_hdr_asize); l2arc_dev_hdr_update(dev); spa_config_exit(vd->vdev_spa, SCL_L2ARC, vd); vd->vdev_trim_thread = NULL; if (vd->vdev_trim_state == VDEV_TRIM_COMPLETE) dev->l2ad_trim_all = B_FALSE; cv_broadcast(&vd->vdev_trim_cv); mutex_exit(&vd->vdev_trim_lock); thread_exit(); } /* * Punches out TRIM threads for the L2ARC devices in a spa and assigns them * to vd->vdev_trim_thread variable. This facilitates the management of * trimming the whole cache device using TRIM_TYPE_MANUAL upon addition * to a pool or pool creation or when the header of the device is invalid. */ void vdev_trim_l2arc(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); /* * Locate the spa's l2arc devices and kick off TRIM threads. */ for (int i = 0; i < spa->spa_l2cache.sav_count; i++) { vdev_t *vd = spa->spa_l2cache.sav_vdevs[i]; l2arc_dev_t *dev = l2arc_vdev_get(vd); if (dev == NULL || !dev->l2ad_trim_all) { /* * Don't attempt TRIM if the vdev is UNAVAIL or if the * cache device was not marked for whole device TRIM * (ie l2arc_trim_ahead = 0, or the L2ARC device header * is valid with trim_state = VDEV_TRIM_COMPLETE and * l2ad_log_entries > 0). */ continue; } mutex_enter(&vd->vdev_trim_lock); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); ASSERT3P(vd->vdev_trim_thread, ==, NULL); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_trim_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, 0, 0, 0); vd->vdev_trim_thread = thread_create(NULL, 0, vdev_trim_l2arc_thread, vd, 0, &p0, TS_RUN, maxclsyspri); mutex_exit(&vd->vdev_trim_lock); } } /* * A wrapper which calls vdev_trim_ranges(). It is intended to be called * on leaf vdevs. */ int vdev_trim_simple(vdev_t *vd, uint64_t start, uint64_t size) { trim_args_t ta = {0}; range_seg64_t physical_rs; int error; physical_rs.rs_start = start; physical_rs.rs_end = start + size; ASSERT(vdev_is_concrete(vd)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_top->vdev_removing); ASSERT(!vd->vdev_top->vdev_rz_expanding); ta.trim_vdev = vd; ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0); ta.trim_type = TRIM_TYPE_SIMPLE; ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max; ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE; ta.trim_flags = 0; ASSERT3U(physical_rs.rs_end, >=, physical_rs.rs_start); if (physical_rs.rs_end > physical_rs.rs_start) { range_tree_add(ta.trim_tree, physical_rs.rs_start, physical_rs.rs_end - physical_rs.rs_start); } else { ASSERT3U(physical_rs.rs_end, ==, physical_rs.rs_start); } error = vdev_trim_ranges(&ta); mutex_enter(&vd->vdev_trim_io_lock); while (vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE] > 0) { cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock); } mutex_exit(&vd->vdev_trim_io_lock); range_tree_vacate(ta.trim_tree, NULL, NULL); range_tree_destroy(ta.trim_tree); return (error); } EXPORT_SYMBOL(vdev_trim); EXPORT_SYMBOL(vdev_trim_stop); EXPORT_SYMBOL(vdev_trim_stop_all); EXPORT_SYMBOL(vdev_trim_stop_wait); EXPORT_SYMBOL(vdev_trim_restart); EXPORT_SYMBOL(vdev_autotrim); EXPORT_SYMBOL(vdev_autotrim_stop_all); EXPORT_SYMBOL(vdev_autotrim_stop_wait); EXPORT_SYMBOL(vdev_autotrim_restart); EXPORT_SYMBOL(vdev_trim_l2arc); EXPORT_SYMBOL(vdev_trim_simple); ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, extent_bytes_max, UINT, ZMOD_RW, "Max size of TRIM commands, larger will be split"); ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, extent_bytes_min, UINT, ZMOD_RW, "Min size of TRIM commands, smaller will be skipped"); ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, metaslab_skip, UINT, ZMOD_RW, "Skip metaslabs which have never been initialized"); ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, txg_batch, UINT, ZMOD_RW, "Min number of txgs to aggregate frees before issuing TRIM"); ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, queue_limit, UINT, ZMOD_RW, "Max queued TRIMs outstanding per leaf vdev"); diff --git a/sys/contrib/openzfs/module/zfs/zio.c b/sys/contrib/openzfs/module/zfs/zio.c index 414e3d4e93bd..1ba99f4d4624 100644 --- a/sys/contrib/openzfs/module/zfs/zio.c +++ b/sys/contrib/openzfs/module/zfs/zio.c @@ -1,5267 +1,5268 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2022 by Delphix. All rights reserved. * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2019, 2023, 2024, Klara Inc. * Copyright (c) 2019, Allan Jude * Copyright (c) 2021, Datto, Inc. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * ========================================================================== * I/O type descriptions * ========================================================================== */ const char *const zio_type_name[ZIO_TYPES] = { /* * Note: Linux kernel thread name length is limited * so these names will differ from upstream open zfs. */ "z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_flush", "z_trim" }; int zio_dva_throttle_enabled = B_TRUE; static int zio_deadman_log_all = B_FALSE; /* * ========================================================================== * I/O kmem caches * ========================================================================== */ static kmem_cache_t *zio_cache; static kmem_cache_t *zio_link_cache; kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; #if defined(ZFS_DEBUG) && !defined(_KERNEL) static uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; static uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; #endif /* Mark IOs as "slow" if they take longer than 30 seconds */ static uint_t zio_slow_io_ms = (30 * MILLISEC); #define BP_SPANB(indblkshift, level) \ (((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT))) #define COMPARE_META_LEVEL 0x80000000ul /* * The following actions directly effect the spa's sync-to-convergence logic. * The values below define the sync pass when we start performing the action. * Care should be taken when changing these values as they directly impact * spa_sync() performance. Tuning these values may introduce subtle performance * pathologies and should only be done in the context of performance analysis. * These tunables will eventually be removed and replaced with #defines once * enough analysis has been done to determine optimal values. * * The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that * regular blocks are not deferred. * * Starting in sync pass 8 (zfs_sync_pass_dont_compress), we disable * compression (including of metadata). In practice, we don't have this * many sync passes, so this has no effect. * * The original intent was that disabling compression would help the sync * passes to converge. However, in practice disabling compression increases * the average number of sync passes, because when we turn compression off, a * lot of block's size will change and thus we have to re-allocate (not * overwrite) them. It also increases the number of 128KB allocations (e.g. * for indirect blocks and spacemaps) because these will not be compressed. * The 128K allocations are especially detrimental to performance on highly * fragmented systems, which may have very few free segments of this size, * and may need to load new metaslabs to satisfy 128K allocations. */ /* defer frees starting in this pass */ uint_t zfs_sync_pass_deferred_free = 2; /* don't compress starting in this pass */ static uint_t zfs_sync_pass_dont_compress = 8; /* rewrite new bps starting in this pass */ static uint_t zfs_sync_pass_rewrite = 2; /* * An allocating zio is one that either currently has the DVA allocate * stage set or will have it later in its lifetime. */ #define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE) /* * Enable smaller cores by excluding metadata * allocations as well. */ int zio_exclude_metadata = 0; static int zio_requeue_io_start_cut_in_line = 1; #ifdef ZFS_DEBUG static const int zio_buf_debug_limit = 16384; #else static const int zio_buf_debug_limit = 0; #endif static inline void __zio_execute(zio_t *zio); static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t); void zio_init(void) { size_t c; zio_cache = kmem_cache_create("zio_cache", sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0); zio_link_cache = kmem_cache_create("zio_link_cache", sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0); for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { size_t size = (c + 1) << SPA_MINBLOCKSHIFT; size_t align, cflags, data_cflags; char name[32]; /* * Create cache for each half-power of 2 size, starting from * SPA_MINBLOCKSIZE. It should give us memory space efficiency * of ~7/8, sufficient for transient allocations mostly using * these caches. */ size_t p2 = size; while (!ISP2(p2)) p2 &= p2 - 1; if (!IS_P2ALIGNED(size, p2 / 2)) continue; #ifndef _KERNEL /* * If we are using watchpoints, put each buffer on its own page, * to eliminate the performance overhead of trapping to the * kernel when modifying a non-watched buffer that shares the * page with a watched buffer. */ if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE)) continue; #endif if (IS_P2ALIGNED(size, PAGESIZE)) align = PAGESIZE; else align = 1 << (highbit64(size ^ (size - 1)) - 1); cflags = (zio_exclude_metadata || size > zio_buf_debug_limit) ? KMC_NODEBUG : 0; data_cflags = KMC_NODEBUG; if (cflags == data_cflags) { /* * Resulting kmem caches would be identical. * Save memory by creating only one. */ (void) snprintf(name, sizeof (name), "zio_buf_comb_%lu", (ulong_t)size); zio_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags); zio_data_buf_cache[c] = zio_buf_cache[c]; continue; } (void) snprintf(name, sizeof (name), "zio_buf_%lu", (ulong_t)size); zio_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags); (void) snprintf(name, sizeof (name), "zio_data_buf_%lu", (ulong_t)size); zio_data_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, data_cflags); } while (--c != 0) { ASSERT(zio_buf_cache[c] != NULL); if (zio_buf_cache[c - 1] == NULL) zio_buf_cache[c - 1] = zio_buf_cache[c]; ASSERT(zio_data_buf_cache[c] != NULL); if (zio_data_buf_cache[c - 1] == NULL) zio_data_buf_cache[c - 1] = zio_data_buf_cache[c]; } zio_inject_init(); lz4_init(); } void zio_fini(void) { size_t n = SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; #if defined(ZFS_DEBUG) && !defined(_KERNEL) for (size_t i = 0; i < n; i++) { if (zio_buf_cache_allocs[i] != zio_buf_cache_frees[i]) (void) printf("zio_fini: [%d] %llu != %llu\n", (int)((i + 1) << SPA_MINBLOCKSHIFT), (long long unsigned)zio_buf_cache_allocs[i], (long long unsigned)zio_buf_cache_frees[i]); } #endif /* * The same kmem cache can show up multiple times in both zio_buf_cache * and zio_data_buf_cache. Do a wasteful but trivially correct scan to * sort it out. */ for (size_t i = 0; i < n; i++) { kmem_cache_t *cache = zio_buf_cache[i]; if (cache == NULL) continue; for (size_t j = i; j < n; j++) { if (cache == zio_buf_cache[j]) zio_buf_cache[j] = NULL; if (cache == zio_data_buf_cache[j]) zio_data_buf_cache[j] = NULL; } kmem_cache_destroy(cache); } for (size_t i = 0; i < n; i++) { kmem_cache_t *cache = zio_data_buf_cache[i]; if (cache == NULL) continue; for (size_t j = i; j < n; j++) { if (cache == zio_data_buf_cache[j]) zio_data_buf_cache[j] = NULL; } kmem_cache_destroy(cache); } for (size_t i = 0; i < n; i++) { VERIFY3P(zio_buf_cache[i], ==, NULL); VERIFY3P(zio_data_buf_cache[i], ==, NULL); } kmem_cache_destroy(zio_link_cache); kmem_cache_destroy(zio_cache); zio_inject_fini(); lz4_fini(); } /* * ========================================================================== * Allocate and free I/O buffers * ========================================================================== */ #ifdef ZFS_DEBUG static const ulong_t zio_buf_canary = (ulong_t)0xdeadc0dedead210b; #endif /* * Use empty space after the buffer to detect overflows. * * Since zio_init() creates kmem caches only for certain set of buffer sizes, * allocations of different sizes may have some unused space after the data. * Filling part of that space with a known pattern on allocation and checking * it on free should allow us to detect some buffer overflows. */ static void zio_buf_put_canary(ulong_t *p, size_t size, kmem_cache_t **cache, size_t c) { #ifdef ZFS_DEBUG size_t off = P2ROUNDUP(size, sizeof (ulong_t)); ulong_t *canary = p + off / sizeof (ulong_t); size_t asize = (c + 1) << SPA_MINBLOCKSHIFT; if (c + 1 < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT && cache[c] == cache[c + 1]) asize = (c + 2) << SPA_MINBLOCKSHIFT; for (; off < asize; canary++, off += sizeof (ulong_t)) *canary = zio_buf_canary; #endif } static void zio_buf_check_canary(ulong_t *p, size_t size, kmem_cache_t **cache, size_t c) { #ifdef ZFS_DEBUG size_t off = P2ROUNDUP(size, sizeof (ulong_t)); ulong_t *canary = p + off / sizeof (ulong_t); size_t asize = (c + 1) << SPA_MINBLOCKSHIFT; if (c + 1 < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT && cache[c] == cache[c + 1]) asize = (c + 2) << SPA_MINBLOCKSHIFT; for (; off < asize; canary++, off += sizeof (ulong_t)) { if (unlikely(*canary != zio_buf_canary)) { PANIC("ZIO buffer overflow %p (%zu) + %zu %#lx != %#lx", p, size, (canary - p) * sizeof (ulong_t), *canary, zio_buf_canary); } } #endif } /* * Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a * crashdump if the kernel panics, so use it judiciously. Obviously, it's * useful to inspect ZFS metadata, but if possible, we should avoid keeping * excess / transient data in-core during a crashdump. */ void * zio_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); #if defined(ZFS_DEBUG) && !defined(_KERNEL) atomic_add_64(&zio_buf_cache_allocs[c], 1); #endif void *p = kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE); zio_buf_put_canary(p, size, zio_buf_cache, c); return (p); } /* * Use zio_data_buf_alloc to allocate data. The data will not appear in a * crashdump if the kernel panics. This exists so that we will limit the amount * of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount * of kernel heap dumped to disk when the kernel panics) */ void * zio_data_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); void *p = kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE); zio_buf_put_canary(p, size, zio_data_buf_cache, c); return (p); } void zio_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); #if defined(ZFS_DEBUG) && !defined(_KERNEL) atomic_add_64(&zio_buf_cache_frees[c], 1); #endif zio_buf_check_canary(buf, size, zio_buf_cache, c); kmem_cache_free(zio_buf_cache[c], buf); } void zio_data_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); zio_buf_check_canary(buf, size, zio_data_buf_cache, c); kmem_cache_free(zio_data_buf_cache[c], buf); } static void zio_abd_free(void *abd, size_t size) { (void) size; abd_free((abd_t *)abd); } /* * ========================================================================== * Push and pop I/O transform buffers * ========================================================================== */ void zio_push_transform(zio_t *zio, abd_t *data, uint64_t size, uint64_t bufsize, zio_transform_func_t *transform) { zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP); zt->zt_orig_abd = zio->io_abd; zt->zt_orig_size = zio->io_size; zt->zt_bufsize = bufsize; zt->zt_transform = transform; zt->zt_next = zio->io_transform_stack; zio->io_transform_stack = zt; zio->io_abd = data; zio->io_size = size; } void zio_pop_transforms(zio_t *zio) { zio_transform_t *zt; while ((zt = zio->io_transform_stack) != NULL) { if (zt->zt_transform != NULL) zt->zt_transform(zio, zt->zt_orig_abd, zt->zt_orig_size); if (zt->zt_bufsize != 0) abd_free(zio->io_abd); zio->io_abd = zt->zt_orig_abd; zio->io_size = zt->zt_orig_size; zio->io_transform_stack = zt->zt_next; kmem_free(zt, sizeof (zio_transform_t)); } } /* * ========================================================================== * I/O transform callbacks for subblocks, decompression, and decryption * ========================================================================== */ static void zio_subblock(zio_t *zio, abd_t *data, uint64_t size) { ASSERT(zio->io_size > size); if (zio->io_type == ZIO_TYPE_READ) abd_copy(data, zio->io_abd, size); } static void zio_decompress(zio_t *zio, abd_t *data, uint64_t size) { if (zio->io_error == 0) { void *tmp = abd_borrow_buf(data, size); int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp), zio->io_abd, tmp, zio->io_size, size, &zio->io_prop.zp_complevel); abd_return_buf_copy(data, tmp, size); if (zio_injection_enabled && ret == 0) ret = zio_handle_fault_injection(zio, EINVAL); if (ret != 0) zio->io_error = SET_ERROR(EIO); } } static void zio_decrypt(zio_t *zio, abd_t *data, uint64_t size) { int ret; void *tmp; blkptr_t *bp = zio->io_bp; spa_t *spa = zio->io_spa; uint64_t dsobj = zio->io_bookmark.zb_objset; uint64_t lsize = BP_GET_LSIZE(bp); dmu_object_type_t ot = BP_GET_TYPE(bp); uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; boolean_t no_crypt = B_FALSE; ASSERT(BP_USES_CRYPT(bp)); ASSERT3U(size, !=, 0); if (zio->io_error != 0) return; /* * Verify the cksum of MACs stored in an indirect bp. It will always * be possible to verify this since it does not require an encryption * key. */ if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { zio_crypt_decode_mac_bp(bp, mac); if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) { /* * We haven't decompressed the data yet, but * zio_crypt_do_indirect_mac_checksum() requires * decompressed data to be able to parse out the MACs * from the indirect block. We decompress it now and * throw away the result after we are finished. */ tmp = zio_buf_alloc(lsize); ret = zio_decompress_data(BP_GET_COMPRESS(bp), zio->io_abd, tmp, zio->io_size, lsize, &zio->io_prop.zp_complevel); if (ret != 0) { ret = SET_ERROR(EIO); goto error; } ret = zio_crypt_do_indirect_mac_checksum(B_FALSE, tmp, lsize, BP_SHOULD_BYTESWAP(bp), mac); zio_buf_free(tmp, lsize); } else { ret = zio_crypt_do_indirect_mac_checksum_abd(B_FALSE, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp), mac); } abd_copy(data, zio->io_abd, size); if (zio_injection_enabled && ot != DMU_OT_DNODE && ret == 0) { ret = zio_handle_decrypt_injection(spa, &zio->io_bookmark, ot, ECKSUM); } if (ret != 0) goto error; return; } /* * If this is an authenticated block, just check the MAC. It would be * nice to separate this out into its own flag, but when this was done, * we had run out of bits in what is now zio_flag_t. Future cleanup * could make this a flag bit. */ if (BP_IS_AUTHENTICATED(bp)) { if (ot == DMU_OT_OBJSET) { ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp)); } else { zio_crypt_decode_mac_bp(bp, mac); ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, zio->io_abd, size, mac); if (zio_injection_enabled && ret == 0) { ret = zio_handle_decrypt_injection(spa, &zio->io_bookmark, ot, ECKSUM); } } abd_copy(data, zio->io_abd, size); if (ret != 0) goto error; return; } zio_crypt_decode_params_bp(bp, salt, iv); if (ot == DMU_OT_INTENT_LOG) { tmp = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t)); zio_crypt_decode_mac_zil(tmp, mac); abd_return_buf(zio->io_abd, tmp, sizeof (zil_chain_t)); } else { zio_crypt_decode_mac_bp(bp, mac); } ret = spa_do_crypt_abd(B_FALSE, spa, &zio->io_bookmark, BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, size, data, zio->io_abd, &no_crypt); if (no_crypt) abd_copy(data, zio->io_abd, size); if (ret != 0) goto error; return; error: /* assert that the key was found unless this was speculative */ ASSERT(ret != EACCES || (zio->io_flags & ZIO_FLAG_SPECULATIVE)); /* * If there was a decryption / authentication error return EIO as * the io_error. If this was not a speculative zio, create an ereport. */ if (ret == ECKSUM) { zio->io_error = SET_ERROR(EIO); if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(spa, &zio->io_bookmark, BP_GET_LOGICAL_BIRTH(zio->io_bp)); (void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, &zio->io_bookmark, zio, 0); } } else { zio->io_error = ret; } } /* * ========================================================================== * I/O parent/child relationships and pipeline interlocks * ========================================================================== */ zio_t * zio_walk_parents(zio_t *cio, zio_link_t **zl) { list_t *pl = &cio->io_parent_list; *zl = (*zl == NULL) ? list_head(pl) : list_next(pl, *zl); if (*zl == NULL) return (NULL); ASSERT((*zl)->zl_child == cio); return ((*zl)->zl_parent); } zio_t * zio_walk_children(zio_t *pio, zio_link_t **zl) { list_t *cl = &pio->io_child_list; ASSERT(MUTEX_HELD(&pio->io_lock)); *zl = (*zl == NULL) ? list_head(cl) : list_next(cl, *zl); if (*zl == NULL) return (NULL); ASSERT((*zl)->zl_parent == pio); return ((*zl)->zl_child); } zio_t * zio_unique_parent(zio_t *cio) { zio_link_t *zl = NULL; zio_t *pio = zio_walk_parents(cio, &zl); VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL); return (pio); } void zio_add_child(zio_t *pio, zio_t *cio) { /* * Logical I/Os can have logical, gang, or vdev children. * Gang I/Os can have gang or vdev children. * Vdev I/Os can only have vdev children. * The following ASSERT captures all of these constraints. */ ASSERT3S(cio->io_child_type, <=, pio->io_child_type); /* Parent should not have READY stage if child doesn't have it. */ IMPLY((cio->io_pipeline & ZIO_STAGE_READY) == 0 && (cio->io_child_type != ZIO_CHILD_VDEV), (pio->io_pipeline & ZIO_STAGE_READY) == 0); zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP); zl->zl_parent = pio; zl->zl_child = cio; mutex_enter(&pio->io_lock); mutex_enter(&cio->io_lock); ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0); uint64_t *countp = pio->io_children[cio->io_child_type]; for (int w = 0; w < ZIO_WAIT_TYPES; w++) countp[w] += !cio->io_state[w]; list_insert_head(&pio->io_child_list, zl); list_insert_head(&cio->io_parent_list, zl); mutex_exit(&cio->io_lock); mutex_exit(&pio->io_lock); } void zio_add_child_first(zio_t *pio, zio_t *cio) { /* * Logical I/Os can have logical, gang, or vdev children. * Gang I/Os can have gang or vdev children. * Vdev I/Os can only have vdev children. * The following ASSERT captures all of these constraints. */ ASSERT3S(cio->io_child_type, <=, pio->io_child_type); /* Parent should not have READY stage if child doesn't have it. */ IMPLY((cio->io_pipeline & ZIO_STAGE_READY) == 0 && (cio->io_child_type != ZIO_CHILD_VDEV), (pio->io_pipeline & ZIO_STAGE_READY) == 0); zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP); zl->zl_parent = pio; zl->zl_child = cio; ASSERT(list_is_empty(&cio->io_parent_list)); list_insert_head(&cio->io_parent_list, zl); mutex_enter(&pio->io_lock); ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0); uint64_t *countp = pio->io_children[cio->io_child_type]; for (int w = 0; w < ZIO_WAIT_TYPES; w++) countp[w] += !cio->io_state[w]; list_insert_head(&pio->io_child_list, zl); mutex_exit(&pio->io_lock); } static void zio_remove_child(zio_t *pio, zio_t *cio, zio_link_t *zl) { ASSERT(zl->zl_parent == pio); ASSERT(zl->zl_child == cio); mutex_enter(&pio->io_lock); mutex_enter(&cio->io_lock); list_remove(&pio->io_child_list, zl); list_remove(&cio->io_parent_list, zl); mutex_exit(&cio->io_lock); mutex_exit(&pio->io_lock); kmem_cache_free(zio_link_cache, zl); } static boolean_t zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait) { boolean_t waiting = B_FALSE; mutex_enter(&zio->io_lock); ASSERT(zio->io_stall == NULL); for (int c = 0; c < ZIO_CHILD_TYPES; c++) { if (!(ZIO_CHILD_BIT_IS_SET(childbits, c))) continue; uint64_t *countp = &zio->io_children[c][wait]; if (*countp != 0) { zio->io_stage >>= 1; ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN); zio->io_stall = countp; waiting = B_TRUE; break; } } mutex_exit(&zio->io_lock); return (waiting); } __attribute__((always_inline)) static inline void zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait, zio_t **next_to_executep) { uint64_t *countp = &pio->io_children[zio->io_child_type][wait]; int *errorp = &pio->io_child_error[zio->io_child_type]; mutex_enter(&pio->io_lock); if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE)) *errorp = zio_worst_error(*errorp, zio->io_error); pio->io_reexecute |= zio->io_reexecute; ASSERT3U(*countp, >, 0); (*countp)--; if (*countp == 0 && pio->io_stall == countp) { zio_taskq_type_t type = pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE : ZIO_TASKQ_INTERRUPT; pio->io_stall = NULL; mutex_exit(&pio->io_lock); /* * If we can tell the caller to execute this parent next, do * so. We only do this if the parent's zio type matches the * child's type. Otherwise dispatch the parent zio in its * own taskq. * * Having the caller execute the parent when possible reduces * locking on the zio taskq's, reduces context switch * overhead, and has no recursion penalty. Note that one * read from disk typically causes at least 3 zio's: a * zio_null(), the logical zio_read(), and then a physical * zio. When the physical ZIO completes, we are able to call * zio_done() on all 3 of these zio's from one invocation of * zio_execute() by returning the parent back to * zio_execute(). Since the parent isn't executed until this * thread returns back to zio_execute(), the caller should do * so promptly. * * In other cases, dispatching the parent prevents * overflowing the stack when we have deeply nested * parent-child relationships, as we do with the "mega zio" * of writes for spa_sync(), and the chain of ZIL blocks. */ if (next_to_executep != NULL && *next_to_executep == NULL && pio->io_type == zio->io_type) { *next_to_executep = pio; } else { zio_taskq_dispatch(pio, type, B_FALSE); } } else { mutex_exit(&pio->io_lock); } } static void zio_inherit_child_errors(zio_t *zio, enum zio_child c) { if (zio->io_child_error[c] != 0 && zio->io_error == 0) zio->io_error = zio->io_child_error[c]; } int zio_bookmark_compare(const void *x1, const void *x2) { const zio_t *z1 = x1; const zio_t *z2 = x2; if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset) return (-1); if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset) return (1); if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object) return (-1); if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object) return (1); if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level) return (-1); if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level) return (1); if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid) return (-1); if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid) return (1); if (z1 < z2) return (-1); if (z1 > z2) return (1); return (0); } /* * ========================================================================== * Create the various types of I/O (read, write, free, etc) * ========================================================================== */ static zio_t * zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, abd_t *data, uint64_t lsize, uint64_t psize, zio_done_func_t *done, void *private, zio_type_t type, zio_priority_t priority, zio_flag_t flags, vdev_t *vd, uint64_t offset, const zbookmark_phys_t *zb, enum zio_stage stage, enum zio_stage pipeline) { zio_t *zio; IMPLY(type != ZIO_TYPE_TRIM, psize <= SPA_MAXBLOCKSIZE); ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0); ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0); ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER)); ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER)); ASSERT(vd || stage == ZIO_STAGE_OPEN); IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW_COMPRESS) != 0); zio = kmem_cache_alloc(zio_cache, KM_SLEEP); memset(zio, 0, sizeof (zio_t)); mutex_init(&zio->io_lock, NULL, MUTEX_NOLOCKDEP, NULL); cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL); list_create(&zio->io_parent_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_parent_node)); list_create(&zio->io_child_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_child_node)); metaslab_trace_init(&zio->io_alloc_list); if (vd != NULL) zio->io_child_type = ZIO_CHILD_VDEV; else if (flags & ZIO_FLAG_GANG_CHILD) zio->io_child_type = ZIO_CHILD_GANG; else if (flags & ZIO_FLAG_DDT_CHILD) zio->io_child_type = ZIO_CHILD_DDT; else zio->io_child_type = ZIO_CHILD_LOGICAL; if (bp != NULL) { if (type != ZIO_TYPE_WRITE || zio->io_child_type == ZIO_CHILD_DDT) { zio->io_bp_copy = *bp; zio->io_bp = &zio->io_bp_copy; /* so caller can free */ } else { zio->io_bp = (blkptr_t *)bp; } zio->io_bp_orig = *bp; if (zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_logical = zio; if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp)) pipeline |= ZIO_GANG_STAGES; } zio->io_spa = spa; zio->io_txg = txg; zio->io_done = done; zio->io_private = private; zio->io_type = type; zio->io_priority = priority; zio->io_vd = vd; zio->io_offset = offset; zio->io_orig_abd = zio->io_abd = data; zio->io_orig_size = zio->io_size = psize; zio->io_lsize = lsize; zio->io_orig_flags = zio->io_flags = flags; zio->io_orig_stage = zio->io_stage = stage; zio->io_orig_pipeline = zio->io_pipeline = pipeline; zio->io_pipeline_trace = ZIO_STAGE_OPEN; zio->io_allocator = ZIO_ALLOCATOR_NONE; zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY) || (pipeline & ZIO_STAGE_READY) == 0; zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE); if (zb != NULL) zio->io_bookmark = *zb; if (pio != NULL) { zio->io_metaslab_class = pio->io_metaslab_class; if (zio->io_logical == NULL) zio->io_logical = pio->io_logical; if (zio->io_child_type == ZIO_CHILD_GANG) zio->io_gang_leader = pio->io_gang_leader; zio_add_child_first(pio, zio); } taskq_init_ent(&zio->io_tqent); return (zio); } void zio_destroy(zio_t *zio) { metaslab_trace_fini(&zio->io_alloc_list); list_destroy(&zio->io_parent_list); list_destroy(&zio->io_child_list); mutex_destroy(&zio->io_lock); cv_destroy(&zio->io_cv); kmem_cache_free(zio_cache, zio); } /* * ZIO intended to be between others. Provides synchronization at READY * and DONE pipeline stages and calls the respective callbacks. */ zio_t * zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *private, zio_flag_t flags) { zio_t *zio; zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private, ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE); return (zio); } /* * ZIO intended to be a root of a tree. Unlike null ZIO does not have a * READY pipeline stage (is ready on creation), so it should not be used * as child of any ZIO that may need waiting for grandchildren READY stage * (any other ZIO type). */ zio_t * zio_root(spa_t *spa, zio_done_func_t *done, void *private, zio_flag_t flags) { zio_t *zio; zio = zio_create(NULL, spa, 0, NULL, NULL, 0, 0, done, private, ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_ROOT_PIPELINE); return (zio); } static int zfs_blkptr_verify_log(spa_t *spa, const blkptr_t *bp, enum blk_verify_flag blk_verify, const char *fmt, ...) { va_list adx; char buf[256]; va_start(adx, fmt); (void) vsnprintf(buf, sizeof (buf), fmt, adx); va_end(adx); zfs_dbgmsg("bad blkptr at %px: " "DVA[0]=%#llx/%#llx " "DVA[1]=%#llx/%#llx " "DVA[2]=%#llx/%#llx " "prop=%#llx " "pad=%#llx,%#llx " "phys_birth=%#llx " "birth=%#llx " "fill=%#llx " "cksum=%#llx/%#llx/%#llx/%#llx", bp, (long long)bp->blk_dva[0].dva_word[0], (long long)bp->blk_dva[0].dva_word[1], (long long)bp->blk_dva[1].dva_word[0], (long long)bp->blk_dva[1].dva_word[1], (long long)bp->blk_dva[2].dva_word[0], (long long)bp->blk_dva[2].dva_word[1], (long long)bp->blk_prop, (long long)bp->blk_pad[0], (long long)bp->blk_pad[1], (long long)BP_GET_PHYSICAL_BIRTH(bp), (long long)BP_GET_LOGICAL_BIRTH(bp), (long long)bp->blk_fill, (long long)bp->blk_cksum.zc_word[0], (long long)bp->blk_cksum.zc_word[1], (long long)bp->blk_cksum.zc_word[2], (long long)bp->blk_cksum.zc_word[3]); switch (blk_verify) { case BLK_VERIFY_HALT: zfs_panic_recover("%s: %s", spa_name(spa), buf); break; case BLK_VERIFY_LOG: zfs_dbgmsg("%s: %s", spa_name(spa), buf); break; case BLK_VERIFY_ONLY: break; } return (1); } /* * Verify the block pointer fields contain reasonable values. This means * it only contains known object types, checksum/compression identifiers, * block sizes within the maximum allowed limits, valid DVAs, etc. * * If everything checks out B_TRUE is returned. The zfs_blkptr_verify * argument controls the behavior when an invalid field is detected. * * Values for blk_verify_flag: * BLK_VERIFY_ONLY: evaluate the block * BLK_VERIFY_LOG: evaluate the block and log problems * BLK_VERIFY_HALT: call zfs_panic_recover on error * * Values for blk_config_flag: * BLK_CONFIG_HELD: caller holds SCL_VDEV for writer * BLK_CONFIG_NEEDED: caller holds no config lock, SCL_VDEV will be * obtained for reader * BLK_CONFIG_SKIP: skip checks which require SCL_VDEV, for better * performance */ boolean_t zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp, enum blk_config_flag blk_config, enum blk_verify_flag blk_verify) { int errors = 0; if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px has invalid TYPE %llu", bp, (longlong_t)BP_GET_TYPE(bp)); } if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px has invalid CHECKSUM %llu", bp, (longlong_t)BP_GET_CHECKSUM(bp)); } if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px has invalid COMPRESS %llu", bp, (longlong_t)BP_GET_COMPRESS(bp)); } if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px has invalid LSIZE %llu", bp, (longlong_t)BP_GET_LSIZE(bp)); } if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px has invalid PSIZE %llu", bp, (longlong_t)BP_GET_PSIZE(bp)); } if (BP_IS_EMBEDDED(bp)) { if (BPE_GET_ETYPE(bp) >= NUM_BP_EMBEDDED_TYPES) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px has invalid ETYPE %llu", bp, (longlong_t)BPE_GET_ETYPE(bp)); } } /* * Do not verify individual DVAs if the config is not trusted. This * will be done once the zio is executed in vdev_mirror_map_alloc. */ if (!spa->spa_trust_config) return (errors == 0); switch (blk_config) { case BLK_CONFIG_HELD: ASSERT(spa_config_held(spa, SCL_VDEV, RW_WRITER)); break; case BLK_CONFIG_NEEDED: spa_config_enter(spa, SCL_VDEV, bp, RW_READER); break; case BLK_CONFIG_SKIP: return (errors == 0); default: panic("invalid blk_config %u", blk_config); } /* * Pool-specific checks. * * Note: it would be nice to verify that the logical birth * and physical birth are not too large. However, * spa_freeze() allows the birth time of log blocks (and * dmu_sync()-ed blocks that are in the log) to be arbitrarily * large. */ for (int i = 0; i < BP_GET_NDVAS(bp); i++) { const dva_t *dva = &bp->blk_dva[i]; uint64_t vdevid = DVA_GET_VDEV(dva); if (vdevid >= spa->spa_root_vdev->vdev_children) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px DVA %u has invalid VDEV %llu", bp, i, (longlong_t)vdevid); continue; } vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid]; if (vd == NULL) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px DVA %u has invalid VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_hole_ops) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px DVA %u has hole VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_missing_ops) { /* * "missing" vdevs are valid during import, but we * don't have their detailed info (e.g. asize), so * we can't perform any more checks on them. */ continue; } uint64_t offset = DVA_GET_OFFSET(dva); uint64_t asize = DVA_GET_ASIZE(dva); if (DVA_GET_GANG(dva)) asize = vdev_gang_header_asize(vd); if (offset + asize > vd->vdev_asize) { errors += zfs_blkptr_verify_log(spa, bp, blk_verify, "blkptr at %px DVA %u has invalid OFFSET %llu", bp, i, (longlong_t)offset); } } if (blk_config == BLK_CONFIG_NEEDED) spa_config_exit(spa, SCL_VDEV, bp); return (errors == 0); } boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva, const blkptr_t *bp) { (void) bp; uint64_t vdevid = DVA_GET_VDEV(dva); if (vdevid >= spa->spa_root_vdev->vdev_children) return (B_FALSE); vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid]; if (vd == NULL) return (B_FALSE); if (vd->vdev_ops == &vdev_hole_ops) return (B_FALSE); if (vd->vdev_ops == &vdev_missing_ops) { return (B_FALSE); } uint64_t offset = DVA_GET_OFFSET(dva); uint64_t asize = DVA_GET_ASIZE(dva); if (DVA_GET_GANG(dva)) asize = vdev_gang_header_asize(vd); if (offset + asize > vd->vdev_asize) return (B_FALSE); return (B_TRUE); } zio_t * zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, abd_t *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, zio_flag_t flags, const zbookmark_phys_t *zb) { zio_t *zio; zio = zio_create(pio, spa, BP_GET_BIRTH(bp), bp, data, size, size, done, private, ZIO_TYPE_READ, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE); return (zio); } zio_t * zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data, uint64_t lsize, uint64_t psize, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *children_ready, zio_done_func_t *done, void *private, zio_priority_t priority, zio_flag_t flags, const zbookmark_phys_t *zb) { zio_t *zio; ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF && zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS && zp->zp_compress >= ZIO_COMPRESS_OFF && zp->zp_compress < ZIO_COMPRESS_FUNCTIONS && DMU_OT_IS_VALID(zp->zp_type) && zp->zp_level < 32 && zp->zp_copies > 0 && zp->zp_copies <= spa_max_replication(spa)); zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private, ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE); zio->io_ready = ready; zio->io_children_ready = children_ready; zio->io_prop = *zp; /* * Data can be NULL if we are going to call zio_write_override() to * provide the already-allocated BP. But we may need the data to * verify a dedup hit (if requested). In this case, don't try to * dedup (just take the already-allocated BP verbatim). Encrypted * dedup blocks need data as well so we also disable dedup in this * case. */ if (data == NULL && (zio->io_prop.zp_dedup_verify || zio->io_prop.zp_encrypt)) { zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE; } return (zio); } zio_t * zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, zio_flag_t flags, zbookmark_phys_t *zb) { zio_t *zio; zio = zio_create(pio, spa, txg, bp, data, size, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_IO_REWRITE, NULL, 0, zb, ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE); return (zio); } void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite, boolean_t brtwrite) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio->io_stage == ZIO_STAGE_OPEN); ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa)); ASSERT(!brtwrite || !nopwrite); /* * We must reset the io_prop to match the values that existed * when the bp was first written by dmu_sync() keeping in mind * that nopwrite and dedup are mutually exclusive. */ zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup; zio->io_prop.zp_nopwrite = nopwrite; zio->io_prop.zp_brtwrite = brtwrite; zio->io_prop.zp_copies = copies; zio->io_bp_override = bp; } void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp) { (void) zfs_blkptr_verify(spa, bp, BLK_CONFIG_NEEDED, BLK_VERIFY_HALT); /* * The check for EMBEDDED is a performance optimization. We * process the free here (by ignoring it) rather than * putting it on the list and then processing it in zio_free_sync(). */ if (BP_IS_EMBEDDED(bp)) return; /* * Frees that are for the currently-syncing txg, are not going to be * deferred, and which will not need to do a read (i.e. not GANG or * DEDUP), can be processed immediately. Otherwise, put them on the * in-memory list for later processing. * * Note that we only defer frees after zfs_sync_pass_deferred_free * when the log space map feature is disabled. [see relevant comment * in spa_sync_iterate_to_convergence()] */ if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp) || txg != spa->spa_syncing_txg || (spa_sync_pass(spa) >= zfs_sync_pass_deferred_free && !spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) || brt_maybe_exists(spa, bp)) { metaslab_check_free(spa, bp); bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp); } else { VERIFY3P(zio_free_sync(NULL, spa, txg, bp, 0), ==, NULL); } } /* * To improve performance, this function may return NULL if we were able * to do the free immediately. This avoids the cost of creating a zio * (and linking it to the parent, etc). */ zio_t * zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_flag_t flags) { ASSERT(!BP_IS_HOLE(bp)); ASSERT(spa_syncing_txg(spa) == txg); if (BP_IS_EMBEDDED(bp)) return (NULL); metaslab_check_free(spa, bp); arc_freed(spa, bp); dsl_scan_freed(spa, bp); if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp) || brt_maybe_exists(spa, bp)) { /* * GANG, DEDUP and BRT blocks can induce a read (for the gang * block header, the DDT or the BRT), so issue them * asynchronously so that this thread is not tied up. */ enum zio_stage stage = ZIO_FREE_PIPELINE | ZIO_STAGE_ISSUE_ASYNC; return (zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp), BP_GET_PSIZE(bp), NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage)); } else { metaslab_free(spa, bp, txg, B_FALSE); return (NULL); } } zio_t * zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *private, zio_flag_t flags) { zio_t *zio; (void) zfs_blkptr_verify(spa, bp, (flags & ZIO_FLAG_CONFIG_WRITER) ? BLK_CONFIG_HELD : BLK_CONFIG_NEEDED, BLK_VERIFY_HALT); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); /* * A claim is an allocation of a specific block. Claims are needed * to support immediate writes in the intent log. The issue is that * immediate writes contain committed data, but in a txg that was * *not* committed. Upon opening the pool after an unclean shutdown, * the intent log claims all blocks that contain immediate write data * so that the SPA knows they're in use. * * All claims *must* be resolved in the first txg -- before the SPA * starts allocating blocks -- so that nothing is allocated twice. * If txg == 0 we just verify that the block is claimable. */ ASSERT3U(BP_GET_LOGICAL_BIRTH(&spa->spa_uberblock.ub_rootbp), <, spa_min_claim_txg(spa)); ASSERT(txg == spa_min_claim_txg(spa) || txg == 0); ASSERT(!BP_GET_DEDUP(bp) || !spa_writeable(spa)); /* zdb(8) */ zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp), BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE); ASSERT0(zio->io_queued_timestamp); return (zio); } zio_t * zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, zio_flag_t flags, enum trim_flag trim_flags) { zio_t *zio; ASSERT0(vd->vdev_children); ASSERT0(P2PHASE(offset, 1ULL << vd->vdev_ashift)); ASSERT0(P2PHASE(size, 1ULL << vd->vdev_ashift)); ASSERT3U(size, !=, 0); zio = zio_create(pio, vd->vdev_spa, 0, NULL, NULL, size, size, done, private, ZIO_TYPE_TRIM, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_TRIM_PIPELINE); zio->io_trim_flags = trim_flags; return (zio); } zio_t * zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, abd_t *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, zio_flag_t flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done, private, ZIO_TYPE_READ, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; return (zio); } zio_t * zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, abd_t *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, zio_flag_t flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { /* * zec checksums are necessarily destructive -- they modify * the end of the write buffer to hold the verifier/checksum. * Therefore, we must make a local copy in case the data is * being written to multiple places in parallel. */ abd_t *wbuf = abd_alloc_sametype(data, size); abd_copy(wbuf, data, size); zio_push_transform(zio, wbuf, size, size, NULL); } return (zio); } /* * Create a child I/O to do some work for us. */ zio_t * zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size, int type, zio_priority_t priority, zio_flag_t flags, zio_done_func_t *done, void *private) { enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE; zio_t *zio; /* * vdev child I/Os do not propagate their error to the parent. * Therefore, for correct operation the caller *must* check for * and handle the error in the child i/o's done callback. * The only exceptions are i/os that we don't care about * (OPTIONAL or REPAIR). */ ASSERT((flags & ZIO_FLAG_OPTIONAL) || (flags & ZIO_FLAG_IO_REPAIR) || done != NULL); if (type == ZIO_TYPE_READ && bp != NULL) { /* * If we have the bp, then the child should perform the * checksum and the parent need not. This pushes error * detection as close to the leaves as possible and * eliminates redundant checksums in the interior nodes. */ pipeline |= ZIO_STAGE_CHECKSUM_VERIFY; pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } if (vd->vdev_ops->vdev_op_leaf) { ASSERT0(vd->vdev_children); offset += VDEV_LABEL_START_SIZE; } flags |= ZIO_VDEV_CHILD_FLAGS(pio); /* * If we've decided to do a repair, the write is not speculative -- * even if the original read was. */ if (flags & ZIO_FLAG_IO_REPAIR) flags &= ~ZIO_FLAG_SPECULATIVE; /* * If we're creating a child I/O that is not associated with a * top-level vdev, then the child zio is not an allocating I/O. * If this is a retried I/O then we ignore it since we will * have already processed the original allocating I/O. */ if (flags & ZIO_FLAG_IO_ALLOCATING && (vd != vd->vdev_top || (flags & ZIO_FLAG_IO_RETRY))) { ASSERT(pio->io_metaslab_class != NULL); ASSERT(pio->io_metaslab_class->mc_alloc_throttle_enabled); ASSERT(type == ZIO_TYPE_WRITE); ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(flags & ZIO_FLAG_IO_REPAIR)); ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) || pio->io_child_type == ZIO_CHILD_GANG); flags &= ~ZIO_FLAG_IO_ALLOCATING; } zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size, done, private, type, priority, flags, vd, offset, &pio->io_bookmark, ZIO_STAGE_VDEV_IO_START >> 1, pipeline); ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); return (zio); } zio_t * zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size, zio_type_t type, zio_priority_t priority, zio_flag_t flags, zio_done_func_t *done, void *private) { zio_t *zio; ASSERT(vd->vdev_ops->vdev_op_leaf); zio = zio_create(NULL, vd->vdev_spa, 0, NULL, data, size, size, done, private, type, priority, flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED, vd, offset, NULL, ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE); return (zio); } /* * Send a flush command to the given vdev. Unlike most zio creation functions, * the flush zios are issued immediately. You can wait on pio to pause until * the flushes complete. */ void zio_flush(zio_t *pio, vdev_t *vd) { const zio_flag_t flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY; if (vd->vdev_nowritecache) return; if (vd->vdev_children == 0) { zio_nowait(zio_create(pio, vd->vdev_spa, 0, NULL, NULL, 0, 0, NULL, NULL, ZIO_TYPE_FLUSH, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_FLUSH_PIPELINE)); } else { for (uint64_t c = 0; c < vd->vdev_children; c++) zio_flush(pio, vd->vdev_child[c]); } } void zio_shrink(zio_t *zio, uint64_t size) { ASSERT3P(zio->io_executor, ==, NULL); ASSERT3U(zio->io_orig_size, ==, zio->io_size); ASSERT3U(size, <=, zio->io_size); /* * We don't shrink for raidz because of problems with the * reconstruction when reading back less than the block size. * Note, BP_IS_RAIDZ() assumes no compression. */ ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF); if (!BP_IS_RAIDZ(zio->io_bp)) { /* we are not doing a raw write */ ASSERT3U(zio->io_size, ==, zio->io_lsize); zio->io_orig_size = zio->io_size = zio->io_lsize = size; } } /* * Round provided allocation size up to a value that can be allocated * by at least some vdev(s) in the pool with minimum or no additional * padding and without extra space usage on others */ static uint64_t zio_roundup_alloc_size(spa_t *spa, uint64_t size) { if (size > spa->spa_min_alloc) return (roundup(size, spa->spa_gcd_alloc)); return (spa->spa_min_alloc); } /* * ========================================================================== * Prepare to read and write logical blocks * ========================================================================== */ static zio_t * zio_read_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; uint64_t psize = BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp); ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy); if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF && zio->io_child_type == ZIO_CHILD_LOGICAL && !(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) { zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize), psize, psize, zio_decompress); } if (((BP_IS_PROTECTED(bp) && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) || BP_HAS_INDIRECT_MAC_CKSUM(bp)) && zio->io_child_type == ZIO_CHILD_LOGICAL) { zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize), psize, psize, zio_decrypt); } if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) { int psize = BPE_GET_PSIZE(bp); void *data = abd_borrow_buf(zio->io_abd, psize); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; decode_embedded_bp_compressed(bp, data); abd_return_buf_copy(zio->io_abd, data, psize); } else { ASSERT(!BP_IS_EMBEDDED(bp)); } if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_pipeline = ZIO_DDT_READ_PIPELINE; return (zio); } static zio_t * zio_write_bp_init(zio_t *zio) { if (!IO_IS_ALLOCATING(zio)) return (zio); ASSERT(zio->io_child_type != ZIO_CHILD_DDT); if (zio->io_bp_override) { blkptr_t *bp = zio->io_bp; zio_prop_t *zp = &zio->io_prop; ASSERT(BP_GET_LOGICAL_BIRTH(bp) != zio->io_txg); *bp = *zio->io_bp_override; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (zp->zp_brtwrite) return (zio); ASSERT(!BP_GET_DEDUP(zio->io_bp_override)); if (BP_IS_EMBEDDED(bp)) return (zio); /* * If we've been overridden and nopwrite is set then * set the flag accordingly to indicate that a nopwrite * has already occurred. */ if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) { ASSERT(!zp->zp_dedup); ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum); zio->io_flags |= ZIO_FLAG_NOPWRITE; return (zio); } ASSERT(!zp->zp_nopwrite); if (BP_IS_HOLE(bp) || !zp->zp_dedup) return (zio); ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP) || zp->zp_dedup_verify); if (BP_GET_CHECKSUM(bp) == zp->zp_checksum && !zp->zp_encrypt) { BP_SET_DEDUP(bp, 1); zio->io_pipeline |= ZIO_STAGE_DDT_WRITE; return (zio); } /* * We were unable to handle this as an override bp, treat * it as a regular write I/O. */ zio->io_bp_override = NULL; *bp = zio->io_bp_orig; zio->io_pipeline = zio->io_orig_pipeline; } return (zio); } static zio_t * zio_write_compress(zio_t *zio) { spa_t *spa = zio->io_spa; zio_prop_t *zp = &zio->io_prop; enum zio_compress compress = zp->zp_compress; blkptr_t *bp = zio->io_bp; uint64_t lsize = zio->io_lsize; uint64_t psize = zio->io_size; uint32_t pass = 1; /* * If our children haven't all reached the ready stage, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT | ZIO_CHILD_GANG_BIT, ZIO_WAIT_READY)) { return (NULL); } if (!IO_IS_ALLOCATING(zio)) return (zio); if (zio->io_children_ready != NULL) { /* * Now that all our children are ready, run the callback * associated with this zio in case it wants to modify the * data to be written. */ ASSERT3U(zp->zp_level, >, 0); zio->io_children_ready(zio); } ASSERT(zio->io_child_type != ZIO_CHILD_DDT); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg) { /* * We're rewriting an existing block, which means we're * working on behalf of spa_sync(). For spa_sync() to * converge, it must eventually be the case that we don't * have to allocate new blocks. But compression changes * the blocksize, which forces a reallocate, and makes * convergence take longer. Therefore, after the first * few passes, stop compressing to ensure convergence. */ pass = spa_sync_pass(spa); ASSERT(zio->io_txg == spa_syncing_txg(spa)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!BP_GET_DEDUP(bp)); if (pass >= zfs_sync_pass_dont_compress) compress = ZIO_COMPRESS_OFF; /* Make sure someone doesn't change their mind on overwrites */ ASSERT(BP_IS_EMBEDDED(bp) || BP_IS_GANG(bp) || MIN(zp->zp_copies, spa_max_replication(spa)) == BP_GET_NDVAS(bp)); } /* If it's a compressed write that is not raw, compress the buffer. */ if (compress != ZIO_COMPRESS_OFF && !(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) { void *cbuf = NULL; psize = zio_compress_data(compress, zio->io_abd, &cbuf, lsize, zp->zp_complevel); if (psize == 0) { compress = ZIO_COMPRESS_OFF; } else if (psize >= lsize) { compress = ZIO_COMPRESS_OFF; if (cbuf != NULL) zio_buf_free(cbuf, lsize); } else if (!zp->zp_dedup && !zp->zp_encrypt && psize <= BPE_PAYLOAD_SIZE && zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) && spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) { encode_embedded_bp_compressed(bp, cbuf, compress, lsize, psize); BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA); BP_SET_TYPE(bp, zio->io_prop.zp_type); BP_SET_LEVEL(bp, zio->io_prop.zp_level); zio_buf_free(cbuf, lsize); BP_SET_LOGICAL_BIRTH(bp, zio->io_txg); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; ASSERT(spa_feature_is_active(spa, SPA_FEATURE_EMBEDDED_DATA)); return (zio); } else { /* * Round compressed size up to the minimum allocation * size of the smallest-ashift device, and zero the * tail. This ensures that the compressed size of the * BP (and thus compressratio property) are correct, * in that we charge for the padding used to fill out * the last sector. */ size_t rounded = (size_t)zio_roundup_alloc_size(spa, psize); if (rounded >= lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); psize = lsize; } else { abd_t *cdata = abd_get_from_buf(cbuf, lsize); abd_take_ownership_of_buf(cdata, B_TRUE); abd_zero_off(cdata, psize, rounded - psize); psize = rounded; zio_push_transform(zio, cdata, psize, lsize, NULL); } } /* * We were unable to handle this as an override bp, treat * it as a regular write I/O. */ zio->io_bp_override = NULL; *bp = zio->io_bp_orig; zio->io_pipeline = zio->io_orig_pipeline; } else if ((zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) != 0 && zp->zp_type == DMU_OT_DNODE) { /* * The DMU actually relies on the zio layer's compression * to free metadnode blocks that have had all contained * dnodes freed. As a result, even when doing a raw * receive, we must check whether the block can be compressed * to a hole. */ psize = zio_compress_data(ZIO_COMPRESS_EMPTY, zio->io_abd, NULL, lsize, zp->zp_complevel); if (psize == 0 || psize >= lsize) compress = ZIO_COMPRESS_OFF; } else if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) { /* * If we are raw receiving an encrypted dataset we should not * take this codepath because it will change the on-disk block * and decryption will fail. */ size_t rounded = MIN((size_t)zio_roundup_alloc_size(spa, psize), lsize); if (rounded != psize) { abd_t *cdata = abd_alloc_linear(rounded, B_TRUE); abd_zero_off(cdata, psize, rounded - psize); abd_copy_off(cdata, zio->io_abd, 0, 0, psize); psize = rounded; zio_push_transform(zio, cdata, psize, rounded, NULL); } } else { ASSERT3U(psize, !=, 0); } /* * The final pass of spa_sync() must be all rewrites, but the first * few passes offer a trade-off: allocating blocks defers convergence, * but newly allocated blocks are sequential, so they can be written * to disk faster. Therefore, we allow the first few passes of * spa_sync() to allocate new blocks, but force rewrites after that. * There should only be a handful of blocks after pass 1 in any case. */ if (!BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg && BP_GET_PSIZE(bp) == psize && pass >= zfs_sync_pass_rewrite) { VERIFY3U(psize, !=, 0); enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES; zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages; zio->io_flags |= ZIO_FLAG_IO_REWRITE; } else { BP_ZERO(bp); zio->io_pipeline = ZIO_WRITE_PIPELINE; } if (psize == 0) { if (BP_GET_LOGICAL_BIRTH(&zio->io_bp_orig) != 0 && spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) { BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_BIRTH(bp, zio->io_txg, 0); } zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; } else { ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER); BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_PSIZE(bp, psize); BP_SET_COMPRESS(bp, compress); BP_SET_CHECKSUM(bp, zp->zp_checksum); BP_SET_DEDUP(bp, zp->zp_dedup); BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER); if (zp->zp_dedup) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(!zp->zp_encrypt || DMU_OT_IS_ENCRYPTED(zp->zp_type)); zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE; } if (zp->zp_nopwrite) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); zio->io_pipeline |= ZIO_STAGE_NOP_WRITE; } } return (zio); } static zio_t * zio_free_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio->io_child_type == ZIO_CHILD_LOGICAL) { if (BP_GET_DEDUP(bp)) zio->io_pipeline = ZIO_DDT_FREE_PIPELINE; } ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy); return (zio); } /* * ========================================================================== * Execute the I/O pipeline * ========================================================================== */ static void zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline) { spa_t *spa = zio->io_spa; zio_type_t t = zio->io_type; int flags = (cutinline ? TQ_FRONT : 0); /* * If we're a config writer or a probe, the normal issue and * interrupt threads may all be blocked waiting for the config lock. * In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL. */ if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_PROBE)) t = ZIO_TYPE_NULL; /* * A similar issue exists for the L2ARC write thread until L2ARC 2.0. */ if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux) t = ZIO_TYPE_NULL; /* * If this is a high priority I/O, then use the high priority taskq if * available. */ if ((zio->io_priority == ZIO_PRIORITY_NOW || zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) && spa->spa_zio_taskq[t][q + 1].stqs_count != 0) q++; ASSERT3U(q, <, ZIO_TASKQ_TYPES); /* * NB: We are assuming that the zio can only be dispatched * to a single taskq at a time. It would be a grievous error * to dispatch the zio to another taskq at the same time. */ ASSERT(taskq_empty_ent(&zio->io_tqent)); spa_taskq_dispatch_ent(spa, t, q, zio_execute, zio, flags, &zio->io_tqent, zio); } static boolean_t zio_taskq_member(zio_t *zio, zio_taskq_type_t q) { spa_t *spa = zio->io_spa; taskq_t *tq = taskq_of_curthread(); for (zio_type_t t = 0; t < ZIO_TYPES; t++) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; uint_t i; for (i = 0; i < tqs->stqs_count; i++) { if (tqs->stqs_taskq[i] == tq) return (B_TRUE); } } return (B_FALSE); } static zio_t * zio_issue_async(zio_t *zio) { ASSERT((zio->io_type != ZIO_TYPE_WRITE) || ZIO_HAS_ALLOCATOR(zio)); zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (NULL); } void zio_interrupt(void *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE); } void zio_delay_interrupt(zio_t *zio) { /* * The timeout_generic() function isn't defined in userspace, so * rather than trying to implement the function, the zio delay * functionality has been disabled for userspace builds. */ #ifdef _KERNEL /* * If io_target_timestamp is zero, then no delay has been registered * for this IO, thus jump to the end of this function and "skip" the * delay; issuing it directly to the zio layer. */ if (zio->io_target_timestamp != 0) { hrtime_t now = gethrtime(); if (now >= zio->io_target_timestamp) { /* * This IO has already taken longer than the target * delay to complete, so we don't want to delay it * any longer; we "miss" the delay and issue it * directly to the zio layer. This is likely due to * the target latency being set to a value less than * the underlying hardware can satisfy (e.g. delay * set to 1ms, but the disks take 10ms to complete an * IO request). */ DTRACE_PROBE2(zio__delay__miss, zio_t *, zio, hrtime_t, now); zio_interrupt(zio); } else { taskqid_t tid; hrtime_t diff = zio->io_target_timestamp - now; clock_t expire_at_tick = ddi_get_lbolt() + NSEC_TO_TICK(diff); DTRACE_PROBE3(zio__delay__hit, zio_t *, zio, hrtime_t, now, hrtime_t, diff); if (NSEC_TO_TICK(diff) == 0) { /* Our delay is less than a jiffy - just spin */ zfs_sleep_until(zio->io_target_timestamp); zio_interrupt(zio); } else { /* * Use taskq_dispatch_delay() in the place of * OpenZFS's timeout_generic(). */ tid = taskq_dispatch_delay(system_taskq, zio_interrupt, zio, TQ_NOSLEEP, expire_at_tick); if (tid == TASKQID_INVALID) { /* * Couldn't allocate a task. Just * finish the zio without a delay. */ zio_interrupt(zio); } } } return; } #endif DTRACE_PROBE1(zio__delay__skip, zio_t *, zio); zio_interrupt(zio); } static void zio_deadman_impl(zio_t *pio, int ziodepth) { zio_t *cio, *cio_next; zio_link_t *zl = NULL; vdev_t *vd = pio->io_vd; if (zio_deadman_log_all || (vd != NULL && vd->vdev_ops->vdev_op_leaf)) { vdev_queue_t *vq = vd ? &vd->vdev_queue : NULL; zbookmark_phys_t *zb = &pio->io_bookmark; uint64_t delta = gethrtime() - pio->io_timestamp; uint64_t failmode = spa_get_deadman_failmode(pio->io_spa); zfs_dbgmsg("slow zio[%d]: zio=%px timestamp=%llu " "delta=%llu queued=%llu io=%llu " "path=%s " "last=%llu type=%d " "priority=%d flags=0x%llx stage=0x%x " "pipeline=0x%x pipeline-trace=0x%x " "objset=%llu object=%llu " "level=%llu blkid=%llu " "offset=%llu size=%llu " "error=%d", ziodepth, pio, pio->io_timestamp, (u_longlong_t)delta, pio->io_delta, pio->io_delay, vd ? vd->vdev_path : "NULL", vq ? vq->vq_io_complete_ts : 0, pio->io_type, pio->io_priority, (u_longlong_t)pio->io_flags, pio->io_stage, pio->io_pipeline, pio->io_pipeline_trace, (u_longlong_t)zb->zb_objset, (u_longlong_t)zb->zb_object, (u_longlong_t)zb->zb_level, (u_longlong_t)zb->zb_blkid, (u_longlong_t)pio->io_offset, (u_longlong_t)pio->io_size, pio->io_error); (void) zfs_ereport_post(FM_EREPORT_ZFS_DEADMAN, pio->io_spa, vd, zb, pio, 0); if (failmode == ZIO_FAILURE_MODE_CONTINUE && taskq_empty_ent(&pio->io_tqent)) { zio_interrupt(pio); } } mutex_enter(&pio->io_lock); for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); zio_deadman_impl(cio, ziodepth + 1); } mutex_exit(&pio->io_lock); } /* * Log the critical information describing this zio and all of its children * using the zfs_dbgmsg() interface then post deadman event for the ZED. */ void zio_deadman(zio_t *pio, const char *tag) { spa_t *spa = pio->io_spa; char *name = spa_name(spa); if (!zfs_deadman_enabled || spa_suspended(spa)) return; zio_deadman_impl(pio, 0); switch (spa_get_deadman_failmode(spa)) { case ZIO_FAILURE_MODE_WAIT: zfs_dbgmsg("%s waiting for hung I/O to pool '%s'", tag, name); break; case ZIO_FAILURE_MODE_CONTINUE: zfs_dbgmsg("%s restarting hung I/O for pool '%s'", tag, name); break; case ZIO_FAILURE_MODE_PANIC: fm_panic("%s determined I/O to pool '%s' is hung.", tag, name); break; } } /* * Execute the I/O pipeline until one of the following occurs: * (1) the I/O completes; (2) the pipeline stalls waiting for * dependent child I/Os; (3) the I/O issues, so we're waiting * for an I/O completion interrupt; (4) the I/O is delegated by * vdev-level caching or aggregation; (5) the I/O is deferred * due to vdev-level queueing; (6) the I/O is handed off to * another thread. In all cases, the pipeline stops whenever * there's no CPU work; it never burns a thread in cv_wait_io(). * * There's no locking on io_stage because there's no legitimate way * for multiple threads to be attempting to process the same I/O. */ static zio_pipe_stage_t *zio_pipeline[]; /* * zio_execute() is a wrapper around the static function * __zio_execute() so that we can force __zio_execute() to be * inlined. This reduces stack overhead which is important * because __zio_execute() is called recursively in several zio * code paths. zio_execute() itself cannot be inlined because * it is externally visible. */ void zio_execute(void *zio) { fstrans_cookie_t cookie; cookie = spl_fstrans_mark(); __zio_execute(zio); spl_fstrans_unmark(cookie); } /* * Used to determine if in the current context the stack is sized large * enough to allow zio_execute() to be called recursively. A minimum * stack size of 16K is required to avoid needing to re-dispatch the zio. */ static boolean_t zio_execute_stack_check(zio_t *zio) { #if !defined(HAVE_LARGE_STACKS) dsl_pool_t *dp = spa_get_dsl(zio->io_spa); /* Executing in txg_sync_thread() context. */ if (dp && curthread == dp->dp_tx.tx_sync_thread) return (B_TRUE); /* Pool initialization outside of zio_taskq context. */ if (dp && spa_is_initializing(dp->dp_spa) && !zio_taskq_member(zio, ZIO_TASKQ_ISSUE) && !zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH)) return (B_TRUE); #else (void) zio; #endif /* HAVE_LARGE_STACKS */ return (B_FALSE); } __attribute__((always_inline)) static inline void __zio_execute(zio_t *zio) { ASSERT3U(zio->io_queued_timestamp, >, 0); while (zio->io_stage < ZIO_STAGE_DONE) { enum zio_stage pipeline = zio->io_pipeline; enum zio_stage stage = zio->io_stage; zio->io_executor = curthread; ASSERT(!MUTEX_HELD(&zio->io_lock)); ASSERT(ISP2(stage)); ASSERT(zio->io_stall == NULL); do { stage <<= 1; } while ((stage & pipeline) == 0); ASSERT(stage <= ZIO_STAGE_DONE); /* * If we are in interrupt context and this pipeline stage * will grab a config lock that is held across I/O, * or may wait for an I/O that needs an interrupt thread * to complete, issue async to avoid deadlock. * * For VDEV_IO_START, we cut in line so that the io will * be sent to disk promptly. */ if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL && zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) { boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ? zio_requeue_io_start_cut_in_line : B_FALSE; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut); return; } /* * If the current context doesn't have large enough stacks * the zio must be issued asynchronously to prevent overflow. */ if (zio_execute_stack_check(zio)) { boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ? zio_requeue_io_start_cut_in_line : B_FALSE; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut); return; } zio->io_stage = stage; zio->io_pipeline_trace |= zio->io_stage; /* * The zio pipeline stage returns the next zio to execute * (typically the same as this one), or NULL if we should * stop. */ zio = zio_pipeline[highbit64(stage) - 1](zio); if (zio == NULL) return; } } /* * ========================================================================== * Initiate I/O, either sync or async * ========================================================================== */ int zio_wait(zio_t *zio) { /* * Some routines, like zio_free_sync(), may return a NULL zio * to avoid the performance overhead of creating and then destroying * an unneeded zio. For the callers' simplicity, we accept a NULL * zio and ignore it. */ if (zio == NULL) return (0); long timeout = MSEC_TO_TICK(zfs_deadman_ziotime_ms); int error; ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN); ASSERT3P(zio->io_executor, ==, NULL); zio->io_waiter = curthread; ASSERT0(zio->io_queued_timestamp); zio->io_queued_timestamp = gethrtime(); if (zio->io_type == ZIO_TYPE_WRITE) { spa_select_allocator(zio); } __zio_execute(zio); mutex_enter(&zio->io_lock); while (zio->io_executor != NULL) { error = cv_timedwait_io(&zio->io_cv, &zio->io_lock, ddi_get_lbolt() + timeout); if (zfs_deadman_enabled && error == -1 && gethrtime() - zio->io_queued_timestamp > spa_deadman_ziotime(zio->io_spa)) { mutex_exit(&zio->io_lock); timeout = MSEC_TO_TICK(zfs_deadman_checktime_ms); zio_deadman(zio, FTAG); mutex_enter(&zio->io_lock); } } mutex_exit(&zio->io_lock); error = zio->io_error; zio_destroy(zio); return (error); } void zio_nowait(zio_t *zio) { /* * See comment in zio_wait(). */ if (zio == NULL) return; ASSERT3P(zio->io_executor, ==, NULL); if (zio->io_child_type == ZIO_CHILD_LOGICAL && list_is_empty(&zio->io_parent_list)) { zio_t *pio; /* * This is a logical async I/O with no parent to wait for it. * We add it to the spa_async_root_zio "Godfather" I/O which * will ensure they complete prior to unloading the pool. */ spa_t *spa = zio->io_spa; pio = spa->spa_async_zio_root[CPU_SEQID_UNSTABLE]; zio_add_child(pio, zio); } ASSERT0(zio->io_queued_timestamp); zio->io_queued_timestamp = gethrtime(); if (zio->io_type == ZIO_TYPE_WRITE) { spa_select_allocator(zio); } __zio_execute(zio); } /* * ========================================================================== * Reexecute, cancel, or suspend/resume failed I/O * ========================================================================== */ static void zio_reexecute(void *arg) { zio_t *pio = arg; zio_t *cio, *cio_next, *gio; ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN); ASSERT(pio->io_gang_leader == NULL); ASSERT(pio->io_gang_tree == NULL); mutex_enter(&pio->io_lock); pio->io_flags = pio->io_orig_flags; pio->io_stage = pio->io_orig_stage; pio->io_pipeline = pio->io_orig_pipeline; pio->io_reexecute = 0; pio->io_flags |= ZIO_FLAG_REEXECUTED; pio->io_pipeline_trace = 0; pio->io_error = 0; pio->io_state[ZIO_WAIT_READY] = (pio->io_stage >= ZIO_STAGE_READY) || (pio->io_pipeline & ZIO_STAGE_READY) == 0; pio->io_state[ZIO_WAIT_DONE] = (pio->io_stage >= ZIO_STAGE_DONE); zio_link_t *zl = NULL; while ((gio = zio_walk_parents(pio, &zl)) != NULL) { for (int w = 0; w < ZIO_WAIT_TYPES; w++) { gio->io_children[pio->io_child_type][w] += !pio->io_state[w]; } } for (int c = 0; c < ZIO_CHILD_TYPES; c++) pio->io_child_error[c] = 0; if (IO_IS_ALLOCATING(pio)) BP_ZERO(pio->io_bp); /* * As we reexecute pio's children, new children could be created. * New children go to the head of pio's io_child_list, however, * so we will (correctly) not reexecute them. The key is that * the remainder of pio's io_child_list, from 'cio_next' onward, * cannot be affected by any side effects of reexecuting 'cio'. */ zl = NULL; for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); mutex_exit(&pio->io_lock); zio_reexecute(cio); mutex_enter(&pio->io_lock); } mutex_exit(&pio->io_lock); /* * Now that all children have been reexecuted, execute the parent. * We don't reexecute "The Godfather" I/O here as it's the * responsibility of the caller to wait on it. */ if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) { pio->io_queued_timestamp = gethrtime(); __zio_execute(pio); } } void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t reason) { if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC) fm_panic("Pool '%s' has encountered an uncorrectable I/O " "failure and the failure mode property for this pool " "is set to panic.", spa_name(spa)); cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O " "failure and has been suspended.\n", spa_name(spa)); (void) zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL, NULL, NULL, 0); mutex_enter(&spa->spa_suspend_lock); if (spa->spa_suspend_zio_root == NULL) spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); spa->spa_suspended = reason; if (zio != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); ASSERT(zio != spa->spa_suspend_zio_root); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio_unique_parent(zio) == NULL); ASSERT(zio->io_stage == ZIO_STAGE_DONE); zio_add_child(spa->spa_suspend_zio_root, zio); } mutex_exit(&spa->spa_suspend_lock); } int zio_resume(spa_t *spa) { zio_t *pio; /* * Reexecute all previously suspended i/o. */ mutex_enter(&spa->spa_suspend_lock); spa->spa_suspended = ZIO_SUSPEND_NONE; cv_broadcast(&spa->spa_suspend_cv); pio = spa->spa_suspend_zio_root; spa->spa_suspend_zio_root = NULL; mutex_exit(&spa->spa_suspend_lock); if (pio == NULL) return (0); zio_reexecute(pio); return (zio_wait(pio)); } void zio_resume_wait(spa_t *spa) { mutex_enter(&spa->spa_suspend_lock); while (spa_suspended(spa)) cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock); mutex_exit(&spa->spa_suspend_lock); } /* * ========================================================================== * Gang blocks. * * A gang block is a collection of small blocks that looks to the DMU * like one large block. When zio_dva_allocate() cannot find a block * of the requested size, due to either severe fragmentation or the pool * being nearly full, it calls zio_write_gang_block() to construct the * block from smaller fragments. * * A gang block consists of a gang header (zio_gbh_phys_t) and up to * three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like * an indirect block: it's an array of block pointers. It consumes * only one sector and hence is allocatable regardless of fragmentation. * The gang header's bps point to its gang members, which hold the data. * * Gang blocks are self-checksumming, using the bp's * as the verifier to ensure uniqueness of the SHA256 checksum. * Critically, the gang block bp's blk_cksum is the checksum of the data, * not the gang header. This ensures that data block signatures (needed for * deduplication) are independent of how the block is physically stored. * * Gang blocks can be nested: a gang member may itself be a gang block. * Thus every gang block is a tree in which root and all interior nodes are * gang headers, and the leaves are normal blocks that contain user data. * The root of the gang tree is called the gang leader. * * To perform any operation (read, rewrite, free, claim) on a gang block, * zio_gang_assemble() first assembles the gang tree (minus data leaves) * in the io_gang_tree field of the original logical i/o by recursively * reading the gang leader and all gang headers below it. This yields * an in-core tree containing the contents of every gang header and the * bps for every constituent of the gang block. * * With the gang tree now assembled, zio_gang_issue() just walks the gang tree * and invokes a callback on each bp. To free a gang block, zio_gang_issue() * calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp. * zio_claim_gang() provides a similarly trivial wrapper for zio_claim(). * zio_read_gang() is a wrapper around zio_read() that omits reading gang * headers, since we already have those in io_gang_tree. zio_rewrite_gang() * performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite() * of the gang header plus zio_checksum_compute() of the data to update the * gang header's blk_cksum as described above. * * The two-phase assemble/issue model solves the problem of partial failure -- * what if you'd freed part of a gang block but then couldn't read the * gang header for another part? Assembling the entire gang tree first * ensures that all the necessary gang header I/O has succeeded before * starting the actual work of free, claim, or write. Once the gang tree * is assembled, free and claim are in-memory operations that cannot fail. * * In the event that a gang write fails, zio_dva_unallocate() walks the * gang tree to immediately free (i.e. insert back into the space map) * everything we've allocated. This ensures that we don't get ENOSPC * errors during repeated suspend/resume cycles due to a flaky device. * * Gang rewrites only happen during sync-to-convergence. If we can't assemble * the gang tree, we won't modify the block, so we can safely defer the free * (knowing that the block is still intact). If we *can* assemble the gang * tree, then even if some of the rewrites fail, zio_dva_unallocate() will free * each constituent bp and we can allocate a new block on the next sync pass. * * In all cases, the gang tree allows complete recovery from partial failure. * ========================================================================== */ static void zio_gang_issue_func_done(zio_t *zio) { abd_free(zio->io_abd); } static zio_t * zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { if (gn != NULL) return (pio); return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset), BP_GET_PSIZE(bp), zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark)); } static zio_t * zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { zio_t *zio; if (gn != NULL) { abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * As we rewrite each gang header, the pipeline will compute * a new gang block header checksum for it; but no one will * compute a new data checksum, so we do that here. The one * exception is the gang leader: the pipeline already computed * its data checksum because that stage precedes gang assembly. * (Presently, nothing actually uses interior data checksums; * this is just good hygiene.) */ if (gn != pio->io_gang_leader->io_gang_tree) { abd_t *buf = abd_get_offset(data, offset); zio_checksum_compute(zio, BP_GET_CHECKSUM(bp), buf, BP_GET_PSIZE(bp)); abd_free(buf); } /* * If we are here to damage data for testing purposes, * leave the GBH alone so that we can detect the damage. */ if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE) zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } else { zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, abd_get_offset(data, offset), BP_GET_PSIZE(bp), zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); } return (zio); } static zio_t * zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { (void) gn, (void) data, (void) offset; zio_t *zio = zio_free_sync(pio, pio->io_spa, pio->io_txg, bp, ZIO_GANG_CHILD_FLAGS(pio)); if (zio == NULL) { zio = zio_null(pio, pio->io_spa, NULL, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)); } return (zio); } static zio_t * zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { (void) gn, (void) data, (void) offset; return (zio_claim(pio, pio->io_spa, pio->io_txg, bp, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio))); } static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = { NULL, zio_read_gang, zio_rewrite_gang, zio_free_gang, zio_claim_gang, NULL }; static void zio_gang_tree_assemble_done(zio_t *zio); static zio_gang_node_t * zio_gang_node_alloc(zio_gang_node_t **gnpp) { zio_gang_node_t *gn; ASSERT(*gnpp == NULL); gn = kmem_zalloc(sizeof (*gn), KM_SLEEP); gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE); *gnpp = gn; return (gn); } static void zio_gang_node_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) ASSERT(gn->gn_child[g] == NULL); zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE); kmem_free(gn, sizeof (*gn)); *gnpp = NULL; } static void zio_gang_tree_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; if (gn == NULL) return; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) zio_gang_tree_free(&gn->gn_child[g]); zio_gang_node_free(gnpp); } static void zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp) { zio_gang_node_t *gn = zio_gang_node_alloc(gnpp); abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); ASSERT(gio->io_gang_leader == gio); ASSERT(BP_IS_GANG(bp)); zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_tree_assemble_done, gn, gio->io_priority, ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark)); } static void zio_gang_tree_assemble_done(zio_t *zio) { zio_t *gio = zio->io_gang_leader; zio_gang_node_t *gn = zio->io_private; blkptr_t *bp = zio->io_bp; ASSERT(gio == zio_unique_parent(zio)); ASSERT(list_is_empty(&zio->io_child_list)); if (zio->io_error) return; /* this ABD was created from a linear buf in zio_gang_tree_assemble */ if (BP_SHOULD_BYTESWAP(bp)) byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size); ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh); ASSERT(zio->io_size == SPA_GANGBLOCKSIZE); ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); abd_free(zio->io_abd); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (!BP_IS_GANG(gbp)) continue; zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]); } } static void zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, abd_t *data, uint64_t offset) { zio_t *gio = pio->io_gang_leader; zio_t *zio; ASSERT(BP_IS_GANG(bp) == !!gn); ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp)); ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == gio->io_gang_tree); /* * If you're a gang header, your data is in gn->gn_gbh. * If you're a gang member, your data is in 'data' and gn == NULL. */ zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset); if (gn != NULL) { ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (BP_IS_HOLE(gbp)) continue; zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data, offset); offset += BP_GET_PSIZE(gbp); } } if (gn == gio->io_gang_tree) ASSERT3U(gio->io_size, ==, offset); if (zio != pio) zio_nowait(zio); } static zio_t * zio_gang_assemble(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree); return (zio); } static zio_t * zio_gang_issue(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT, ZIO_WAIT_DONE)) { return (NULL); } ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); if (zio->io_child_error[ZIO_CHILD_GANG] == 0) zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd, 0); else zio_gang_tree_free(&zio->io_gang_tree); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; return (zio); } static void zio_gang_inherit_allocator(zio_t *pio, zio_t *cio) { cio->io_allocator = pio->io_allocator; cio->io_wr_iss_tq = pio->io_wr_iss_tq; } static void zio_write_gang_member_ready(zio_t *zio) { zio_t *pio = zio_unique_parent(zio); dva_t *cdva = zio->io_bp->blk_dva; dva_t *pdva = pio->io_bp->blk_dva; uint64_t asize; zio_t *gio __maybe_unused = zio->io_gang_leader; if (BP_IS_HOLE(zio->io_bp)) return; ASSERT(BP_IS_HOLE(&zio->io_bp_orig)); ASSERT(zio->io_child_type == ZIO_CHILD_GANG); ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies); ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp)); ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp)); VERIFY3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp)); mutex_enter(&pio->io_lock); for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) { ASSERT(DVA_GET_GANG(&pdva[d])); asize = DVA_GET_ASIZE(&pdva[d]); asize += DVA_GET_ASIZE(&cdva[d]); DVA_SET_ASIZE(&pdva[d], asize); } mutex_exit(&pio->io_lock); } static void zio_write_gang_done(zio_t *zio) { /* * The io_abd field will be NULL for a zio with no data. The io_flags * will initially have the ZIO_FLAG_NODATA bit flag set, but we can't * check for it here as it is cleared in zio_ready. */ if (zio->io_abd != NULL) abd_free(zio->io_abd); } static zio_t * zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) { spa_t *spa = pio->io_spa; blkptr_t *bp = pio->io_bp; zio_t *gio = pio->io_gang_leader; zio_t *zio; zio_gang_node_t *gn, **gnpp; zio_gbh_phys_t *gbh; abd_t *gbh_abd; uint64_t txg = pio->io_txg; uint64_t resid = pio->io_size; uint64_t lsize; int copies = gio->io_prop.zp_copies; zio_prop_t zp; int error; boolean_t has_data = !(pio->io_flags & ZIO_FLAG_NODATA); /* * If one copy was requested, store 2 copies of the GBH, so that we * can still traverse all the data (e.g. to free or scrub) even if a * block is damaged. Note that we can't store 3 copies of the GBH in * all cases, e.g. with encryption, which uses DVA[2] for the IV+salt. */ int gbh_copies = copies; if (gbh_copies == 1) { gbh_copies = MIN(2, spa_max_replication(spa)); } ASSERT(ZIO_HAS_ALLOCATOR(pio)); int flags = METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER; if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(has_data); flags |= METASLAB_ASYNC_ALLOC; VERIFY(zfs_refcount_held(&mc->mc_allocator[pio->io_allocator]. mca_alloc_slots, pio)); /* * The logical zio has already placed a reservation for * 'copies' allocation slots but gang blocks may require * additional copies. These additional copies * (i.e. gbh_copies - copies) are guaranteed to succeed * since metaslab_class_throttle_reserve() always allows * additional reservations for gang blocks. */ VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies, pio->io_allocator, pio, flags)); } error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE, bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags, &pio->io_alloc_list, pio, pio->io_allocator); if (error) { if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(has_data); /* * If we failed to allocate the gang block header then * we remove any additional allocation reservations that * we placed here. The original reservation will * be removed when the logical I/O goes to the ready * stage. */ metaslab_class_throttle_unreserve(mc, gbh_copies - copies, pio->io_allocator, pio); } pio->io_error = error; return (pio); } if (pio == gio) { gnpp = &gio->io_gang_tree; } else { gnpp = pio->io_private; ASSERT(pio->io_ready == zio_write_gang_member_ready); } gn = zio_gang_node_alloc(gnpp); gbh = gn->gn_gbh; memset(gbh, 0, SPA_GANGBLOCKSIZE); gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE); /* * Create the gang header. */ zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_write_gang_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); zio_gang_inherit_allocator(pio, zio); /* * Create and nowait the gang children. */ for (int g = 0; resid != 0; resid -= lsize, g++) { lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g), SPA_MINBLOCKSIZE); ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid); zp.zp_checksum = gio->io_prop.zp_checksum; zp.zp_compress = ZIO_COMPRESS_OFF; zp.zp_complevel = gio->io_prop.zp_complevel; zp.zp_type = DMU_OT_NONE; zp.zp_level = 0; zp.zp_copies = gio->io_prop.zp_copies; zp.zp_dedup = B_FALSE; zp.zp_dedup_verify = B_FALSE; zp.zp_nopwrite = B_FALSE; zp.zp_encrypt = gio->io_prop.zp_encrypt; zp.zp_byteorder = gio->io_prop.zp_byteorder; memset(zp.zp_salt, 0, ZIO_DATA_SALT_LEN); memset(zp.zp_iv, 0, ZIO_DATA_IV_LEN); memset(zp.zp_mac, 0, ZIO_DATA_MAC_LEN); zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g], has_data ? abd_get_offset(pio->io_abd, pio->io_size - resid) : NULL, lsize, lsize, &zp, zio_write_gang_member_ready, NULL, zio_write_gang_done, &gn->gn_child[g], pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); zio_gang_inherit_allocator(zio, cio); if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(has_data); /* * Gang children won't throttle but we should * account for their work, so reserve an allocation * slot for them here. */ VERIFY(metaslab_class_throttle_reserve(mc, zp.zp_copies, cio->io_allocator, cio, flags)); } zio_nowait(cio); } /* * Set pio's pipeline to just wait for zio to finish. */ pio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio_nowait(zio); return (pio); } /* * The zio_nop_write stage in the pipeline determines if allocating a * new bp is necessary. The nopwrite feature can handle writes in * either syncing or open context (i.e. zil writes) and as a result is * mutually exclusive with dedup. * * By leveraging a cryptographically secure checksum, such as SHA256, we * can compare the checksums of the new data and the old to determine if * allocating a new block is required. Note that our requirements for * cryptographic strength are fairly weak: there can't be any accidental * hash collisions, but we don't need to be secure against intentional * (malicious) collisions. To trigger a nopwrite, you have to be able * to write the file to begin with, and triggering an incorrect (hash * collision) nopwrite is no worse than simply writing to the file. * That said, there are no known attacks against the checksum algorithms * used for nopwrite, assuming that the salt and the checksums * themselves remain secret. */ static zio_t * zio_nop_write(zio_t *zio) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; zio_prop_t *zp = &zio->io_prop; ASSERT(BP_IS_HOLE(bp)); ASSERT(BP_GET_LEVEL(bp) == 0); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(zp->zp_nopwrite); ASSERT(!zp->zp_dedup); ASSERT(zio->io_bp_override == NULL); ASSERT(IO_IS_ALLOCATING(zio)); /* * Check to see if the original bp and the new bp have matching * characteristics (i.e. same checksum, compression algorithms, etc). * If they don't then just continue with the pipeline which will * allocate a new bp. */ if (BP_IS_HOLE(bp_orig) || !(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags & ZCHECKSUM_FLAG_NOPWRITE) || BP_IS_ENCRYPTED(bp) || BP_IS_ENCRYPTED(bp_orig) || BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) || BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) || BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) || zp->zp_copies != BP_GET_NDVAS(bp_orig)) return (zio); /* * If the checksums match then reset the pipeline so that we * avoid allocating a new bp and issuing any I/O. */ if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) { ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_NOPWRITE); ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig)); ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig)); ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF); ASSERT3U(bp->blk_prop, ==, bp_orig->blk_prop); /* * If we're overwriting a block that is currently on an * indirect vdev, then ignore the nopwrite request and * allow a new block to be allocated on a concrete vdev. */ spa_config_enter(zio->io_spa, SCL_VDEV, FTAG, RW_READER); for (int d = 0; d < BP_GET_NDVAS(bp_orig); d++) { vdev_t *tvd = vdev_lookup_top(zio->io_spa, DVA_GET_VDEV(&bp_orig->blk_dva[d])); if (tvd->vdev_ops == &vdev_indirect_ops) { spa_config_exit(zio->io_spa, SCL_VDEV, FTAG); return (zio); } } spa_config_exit(zio->io_spa, SCL_VDEV, FTAG); *bp = *bp_orig; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio->io_flags |= ZIO_FLAG_NOPWRITE; } return (zio); } /* * ========================================================================== * Block Reference Table * ========================================================================== */ static zio_t * zio_brt_free(zio_t *zio) { blkptr_t *bp; bp = zio->io_bp; if (BP_GET_LEVEL(bp) > 0 || BP_IS_METADATA(bp) || !brt_maybe_exists(zio->io_spa, bp)) { return (zio); } if (!brt_entry_decref(zio->io_spa, bp)) { /* * This isn't the last reference, so we cannot free * the data yet. */ zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; } return (zio); } /* * ========================================================================== * Dedup * ========================================================================== */ static void zio_ddt_child_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp; zio_t *pio = zio_unique_parent(zio); mutex_enter(&pio->io_lock); ddp = ddt_phys_select(dde, bp); if (zio->io_error == 0) ddt_phys_clear(ddp); /* this ddp doesn't need repair */ if (zio->io_error == 0 && dde->dde_repair_abd == NULL) dde->dde_repair_abd = zio->io_abd; else abd_free(zio->io_abd); mutex_exit(&pio->io_lock); } static zio_t * zio_ddt_read_start(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = ddt_repair_start(ddt, bp); ddt_phys_t *ddp = dde->dde_phys; ddt_phys_t *ddp_self = ddt_phys_select(dde, bp); blkptr_t blk; ASSERT(zio->io_vsd == NULL); zio->io_vsd = dde; if (ddp_self == NULL) return (zio); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { if (ddp->ddp_phys_birth == 0 || ddp == ddp_self) continue; ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp, &blk); zio_nowait(zio_read(zio, zio->io_spa, &blk, abd_alloc_for_io(zio->io_size, B_TRUE), zio->io_size, zio_ddt_child_read_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) | ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark)); } return (zio); } zio_nowait(zio_read(zio, zio->io_spa, bp, zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark)); return (zio); } static zio_t * zio_ddt_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_DDT_BIT, ZIO_WAIT_DONE)) { return (NULL); } ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_vsd; if (ddt == NULL) { ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE); return (zio); } if (dde == NULL) { zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (NULL); } if (dde->dde_repair_abd != NULL) { abd_copy(zio->io_abd, dde->dde_repair_abd, zio->io_size); zio->io_child_error[ZIO_CHILD_DDT] = 0; } ddt_repair_done(ddt, dde); zio->io_vsd = NULL; } ASSERT(zio->io_vsd == NULL); return (zio); } static boolean_t zio_ddt_collision(zio_t *zio, ddt_t *ddt, ddt_entry_t *dde) { spa_t *spa = zio->io_spa; boolean_t do_raw = !!(zio->io_flags & ZIO_FLAG_RAW); ASSERT(!(zio->io_bp_override && do_raw)); /* * Note: we compare the original data, not the transformed data, * because when zio->io_bp is an override bp, we will not have * pushed the I/O transforms. That's an important optimization * because otherwise we'd compress/encrypt all dmu_sync() data twice. * However, we should never get a raw, override zio so in these * cases we can compare the io_abd directly. This is useful because * it allows us to do dedup verification even if we don't have access * to the original data (for instance, if the encryption keys aren't * loaded). */ for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { zio_t *lio = dde->dde_lead_zio[p]; if (lio != NULL && do_raw) { return (lio->io_size != zio->io_size || abd_cmp(zio->io_abd, lio->io_abd) != 0); } else if (lio != NULL) { return (lio->io_orig_size != zio->io_orig_size || abd_cmp(zio->io_orig_abd, lio->io_orig_abd) != 0); } } for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { ddt_phys_t *ddp = &dde->dde_phys[p]; if (ddp->ddp_phys_birth != 0 && do_raw) { blkptr_t blk = *zio->io_bp; uint64_t psize; abd_t *tmpabd; int error; ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth); psize = BP_GET_PSIZE(&blk); if (psize != zio->io_size) return (B_TRUE); ddt_exit(ddt); tmpabd = abd_alloc_for_io(psize, B_TRUE); error = zio_wait(zio_read(NULL, spa, &blk, tmpabd, psize, NULL, NULL, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_RAW, &zio->io_bookmark)); if (error == 0) { if (abd_cmp(tmpabd, zio->io_abd) != 0) error = SET_ERROR(ENOENT); } abd_free(tmpabd); ddt_enter(ddt); return (error != 0); } else if (ddp->ddp_phys_birth != 0) { arc_buf_t *abuf = NULL; arc_flags_t aflags = ARC_FLAG_WAIT; blkptr_t blk = *zio->io_bp; int error; ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth); if (BP_GET_LSIZE(&blk) != zio->io_orig_size) return (B_TRUE); ddt_exit(ddt); error = arc_read(NULL, spa, &blk, arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &aflags, &zio->io_bookmark); if (error == 0) { if (abd_cmp_buf(zio->io_orig_abd, abuf->b_data, zio->io_orig_size) != 0) error = SET_ERROR(ENOENT); arc_buf_destroy(abuf, &abuf); } ddt_enter(ddt); return (error != 0); } } return (B_FALSE); } static void zio_ddt_child_write_ready(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; zio_t *pio; if (zio->io_error) return; ddt_enter(ddt); ASSERT(dde->dde_lead_zio[p] == zio); ddt_phys_fill(ddp, zio->io_bp); zio_link_t *zl = NULL; while ((pio = zio_walk_parents(zio, &zl)) != NULL) ddt_bp_fill(ddp, pio->io_bp, zio->io_txg); ddt_exit(ddt); } static void zio_ddt_child_write_done(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { zio_link_t *zl = NULL; while (zio_walk_parents(zio, &zl) != NULL) ddt_phys_addref(ddp); } else { ddt_phys_clear(ddp); } ddt_exit(ddt); } static zio_t * zio_ddt_write(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; uint64_t txg = zio->io_txg; zio_prop_t *zp = &zio->io_prop; int p = zp->zp_copies; zio_t *cio = NULL; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum); ASSERT(BP_IS_HOLE(bp) || zio->io_bp_override); ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW))); ddt_enter(ddt); dde = ddt_lookup(ddt, bp, B_TRUE); ddp = &dde->dde_phys[p]; if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) { /* * If we're using a weak checksum, upgrade to a strong checksum * and try again. If we're already using a strong checksum, * we can't resolve it, so just convert to an ordinary write. * (And automatically e-mail a paper to Nature?) */ if (!(zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP)) { zp->zp_checksum = spa_dedup_checksum(spa); zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; BP_ZERO(bp); } else { zp->zp_dedup = B_FALSE; BP_SET_DEDUP(bp, B_FALSE); } ASSERT(!BP_GET_DEDUP(bp)); zio->io_pipeline = ZIO_WRITE_PIPELINE; ddt_exit(ddt); return (zio); } if (ddp->ddp_phys_birth != 0 || dde->dde_lead_zio[p] != NULL) { if (ddp->ddp_phys_birth != 0) ddt_bp_fill(ddp, bp, txg); if (dde->dde_lead_zio[p] != NULL) zio_add_child(zio, dde->dde_lead_zio[p]); else ddt_phys_addref(ddp); } else if (zio->io_bp_override) { ASSERT(BP_GET_LOGICAL_BIRTH(bp) == txg); ASSERT(BP_EQUAL(bp, zio->io_bp_override)); ddt_phys_fill(ddp, bp); ddt_phys_addref(ddp); } else { cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd, zio->io_orig_size, zio->io_orig_size, zp, zio_ddt_child_write_ready, NULL, zio_ddt_child_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL); dde->dde_lead_zio[p] = cio; } ddt_exit(ddt); zio_nowait(cio); return (zio); } static ddt_entry_t *freedde; /* for debugging */ static zio_t * zio_ddt_free(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ddt_enter(ddt); freedde = dde = ddt_lookup(ddt, bp, B_TRUE); if (dde) { ddp = ddt_phys_select(dde, bp); if (ddp) ddt_phys_decref(ddp); } ddt_exit(ddt); return (zio); } /* * ========================================================================== * Allocate and free blocks * ========================================================================== */ static zio_t * zio_io_to_allocate(spa_t *spa, int allocator) { zio_t *zio; ASSERT(MUTEX_HELD(&spa->spa_allocs[allocator].spaa_lock)); zio = avl_first(&spa->spa_allocs[allocator].spaa_tree); if (zio == NULL) return (NULL); ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(ZIO_HAS_ALLOCATOR(zio)); /* * Try to place a reservation for this zio. If we're unable to * reserve then we throttle. */ ASSERT3U(zio->io_allocator, ==, allocator); if (!metaslab_class_throttle_reserve(zio->io_metaslab_class, zio->io_prop.zp_copies, allocator, zio, 0)) { return (NULL); } avl_remove(&spa->spa_allocs[allocator].spaa_tree, zio); ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE); return (zio); } static zio_t * zio_dva_throttle(zio_t *zio) { spa_t *spa = zio->io_spa; zio_t *nio; metaslab_class_t *mc; /* locate an appropriate allocation class */ mc = spa_preferred_class(spa, zio->io_size, zio->io_prop.zp_type, zio->io_prop.zp_level, zio->io_prop.zp_zpl_smallblk); if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE || !mc->mc_alloc_throttle_enabled || zio->io_child_type == ZIO_CHILD_GANG || zio->io_flags & ZIO_FLAG_NODATA) { return (zio); } ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(ZIO_HAS_ALLOCATOR(zio)); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); ASSERT3U(zio->io_queued_timestamp, >, 0); ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE); int allocator = zio->io_allocator; zio->io_metaslab_class = mc; mutex_enter(&spa->spa_allocs[allocator].spaa_lock); avl_add(&spa->spa_allocs[allocator].spaa_tree, zio); nio = zio_io_to_allocate(spa, allocator); mutex_exit(&spa->spa_allocs[allocator].spaa_lock); return (nio); } static void zio_allocate_dispatch(spa_t *spa, int allocator) { zio_t *zio; mutex_enter(&spa->spa_allocs[allocator].spaa_lock); zio = zio_io_to_allocate(spa, allocator); mutex_exit(&spa->spa_allocs[allocator].spaa_lock); if (zio == NULL) return; ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE); ASSERT0(zio->io_error); zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE); } static zio_t * zio_dva_allocate(zio_t *zio) { spa_t *spa = zio->io_spa; metaslab_class_t *mc; blkptr_t *bp = zio->io_bp; int error; int flags = 0; if (zio->io_gang_leader == NULL) { ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; } ASSERT(BP_IS_HOLE(bp)); ASSERT0(BP_GET_NDVAS(bp)); ASSERT3U(zio->io_prop.zp_copies, >, 0); ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa)); ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp)); if (zio->io_flags & ZIO_FLAG_NODATA) flags |= METASLAB_DONT_THROTTLE; if (zio->io_flags & ZIO_FLAG_GANG_CHILD) flags |= METASLAB_GANG_CHILD; if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE) flags |= METASLAB_ASYNC_ALLOC; /* * if not already chosen, locate an appropriate allocation class */ mc = zio->io_metaslab_class; if (mc == NULL) { mc = spa_preferred_class(spa, zio->io_size, zio->io_prop.zp_type, zio->io_prop.zp_level, zio->io_prop.zp_zpl_smallblk); zio->io_metaslab_class = mc; } /* * Try allocating the block in the usual metaslab class. * If that's full, allocate it in the normal class. * If that's full, allocate as a gang block, * and if all are full, the allocation fails (which shouldn't happen). * * Note that we do not fall back on embedded slog (ZIL) space, to * preserve unfragmented slog space, which is critical for decent * sync write performance. If a log allocation fails, we will fall * back to spa_sync() which is abysmal for performance. */ ASSERT(ZIO_HAS_ALLOCATOR(zio)); error = metaslab_alloc(spa, mc, zio->io_size, bp, zio->io_prop.zp_copies, zio->io_txg, NULL, flags, &zio->io_alloc_list, zio, zio->io_allocator); /* * Fallback to normal class when an alloc class is full */ if (error == ENOSPC && mc != spa_normal_class(spa)) { /* * If throttling, transfer reservation over to normal class. * The io_allocator slot can remain the same even though we * are switching classes. */ if (mc->mc_alloc_throttle_enabled && (zio->io_flags & ZIO_FLAG_IO_ALLOCATING)) { metaslab_class_throttle_unreserve(mc, zio->io_prop.zp_copies, zio->io_allocator, zio); zio->io_flags &= ~ZIO_FLAG_IO_ALLOCATING; VERIFY(metaslab_class_throttle_reserve( spa_normal_class(spa), zio->io_prop.zp_copies, zio->io_allocator, zio, flags | METASLAB_MUST_RESERVE)); } zio->io_metaslab_class = mc = spa_normal_class(spa); if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) { zfs_dbgmsg("%s: metaslab allocation failure, " "trying normal class: zio %px, size %llu, error %d", spa_name(spa), zio, (u_longlong_t)zio->io_size, error); } error = metaslab_alloc(spa, mc, zio->io_size, bp, zio->io_prop.zp_copies, zio->io_txg, NULL, flags, &zio->io_alloc_list, zio, zio->io_allocator); } if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) { if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) { zfs_dbgmsg("%s: metaslab allocation failure, " "trying ganging: zio %px, size %llu, error %d", spa_name(spa), zio, (u_longlong_t)zio->io_size, error); } return (zio_write_gang_block(zio, mc)); } if (error != 0) { if (error != ENOSPC || (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC)) { zfs_dbgmsg("%s: metaslab allocation failure: zio %px, " "size %llu, error %d", spa_name(spa), zio, (u_longlong_t)zio->io_size, error); } zio->io_error = error; } return (zio); } static zio_t * zio_dva_free(zio_t *zio) { metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE); return (zio); } static zio_t * zio_dva_claim(zio_t *zio) { int error; error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg); if (error) zio->io_error = error; return (zio); } /* * Undo an allocation. This is used by zio_done() when an I/O fails * and we want to give back the block we just allocated. * This handles both normal blocks and gang blocks. */ static void zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp) { ASSERT(BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg || BP_IS_HOLE(bp)); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp)) { metaslab_free(zio->io_spa, bp, BP_GET_LOGICAL_BIRTH(bp), B_TRUE); } if (gn != NULL) { for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { zio_dva_unallocate(zio, gn->gn_child[g], &gn->gn_gbh->zg_blkptr[g]); } } } /* * Try to allocate an intent log block. Return 0 on success, errno on failure. */ int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp, uint64_t size, boolean_t *slog) { int error = 1; zio_alloc_list_t io_alloc_list; ASSERT(txg > spa_syncing_txg(spa)); metaslab_trace_init(&io_alloc_list); /* * Block pointer fields are useful to metaslabs for stats and debugging. * Fill in the obvious ones before calling into metaslab_alloc(). */ BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG); BP_SET_PSIZE(new_bp, size); BP_SET_LEVEL(new_bp, 0); /* * When allocating a zil block, we don't have information about * the final destination of the block except the objset it's part * of, so we just hash the objset ID to pick the allocator to get * some parallelism. */ int flags = METASLAB_ZIL; int allocator = (uint_t)cityhash4(0, 0, 0, os->os_dsl_dataset->ds_object) % spa->spa_alloc_count; error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1, txg, NULL, flags, &io_alloc_list, NULL, allocator); *slog = (error == 0); if (error != 0) { error = metaslab_alloc(spa, spa_embedded_log_class(spa), size, new_bp, 1, txg, NULL, flags, &io_alloc_list, NULL, allocator); } if (error != 0) { error = metaslab_alloc(spa, spa_normal_class(spa), size, new_bp, 1, txg, NULL, flags, &io_alloc_list, NULL, allocator); } metaslab_trace_fini(&io_alloc_list); if (error == 0) { BP_SET_LSIZE(new_bp, size); BP_SET_PSIZE(new_bp, size); BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF); BP_SET_CHECKSUM(new_bp, spa_version(spa) >= SPA_VERSION_SLIM_ZIL ? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG); BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG); BP_SET_LEVEL(new_bp, 0); BP_SET_DEDUP(new_bp, 0); BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER); /* * encrypted blocks will require an IV and salt. We generate * these now since we will not be rewriting the bp at * rewrite time. */ if (os->os_encrypted) { uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t salt[ZIO_DATA_SALT_LEN]; BP_SET_CRYPT(new_bp, B_TRUE); VERIFY0(spa_crypt_get_salt(spa, dmu_objset_id(os), salt)); VERIFY0(zio_crypt_generate_iv(iv)); zio_crypt_encode_params_bp(new_bp, salt, iv); } } else { zfs_dbgmsg("%s: zil block allocation failure: " "size %llu, error %d", spa_name(spa), (u_longlong_t)size, error); } return (error); } /* * ========================================================================== * Read and write to physical devices * ========================================================================== */ /* * Issue an I/O to the underlying vdev. Typically the issue pipeline * stops after this stage and will resume upon I/O completion. * However, there are instances where the vdev layer may need to * continue the pipeline when an I/O was not issued. Since the I/O * that was sent to the vdev layer might be different than the one * currently active in the pipeline (see vdev_queue_io()), we explicitly * force the underlying vdev layers to call either zio_execute() or * zio_interrupt() to ensure that the pipeline continues with the correct I/O. */ static zio_t * zio_vdev_io_start(zio_t *zio) { vdev_t *vd = zio->io_vd; uint64_t align; spa_t *spa = zio->io_spa; zio->io_delay = 0; ASSERT(zio->io_error == 0); ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0); if (vd == NULL) { if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_enter(spa, SCL_ZIO, zio, RW_READER); /* * The mirror_ops handle multiple DVAs in a single BP. */ vdev_mirror_ops.vdev_op_io_start(zio); return (NULL); } ASSERT3P(zio->io_logical, !=, zio); if (zio->io_type == ZIO_TYPE_WRITE) { ASSERT(spa->spa_trust_config); /* * Note: the code can handle other kinds of writes, * but we don't expect them. */ if (zio->io_vd->vdev_noalloc) { ASSERT(zio->io_flags & (ZIO_FLAG_PHYSICAL | ZIO_FLAG_SELF_HEAL | ZIO_FLAG_RESILVER | ZIO_FLAG_INDUCE_DAMAGE)); } } align = 1ULL << vd->vdev_top->vdev_ashift; if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) && P2PHASE(zio->io_size, align) != 0) { /* Transform logical writes to be a full physical block size. */ uint64_t asize = P2ROUNDUP(zio->io_size, align); abd_t *abuf = abd_alloc_sametype(zio->io_abd, asize); ASSERT(vd == vd->vdev_top); if (zio->io_type == ZIO_TYPE_WRITE) { abd_copy(abuf, zio->io_abd, zio->io_size); abd_zero_off(abuf, zio->io_size, asize - zio->io_size); } zio_push_transform(zio, abuf, asize, asize, zio_subblock); } /* * If this is not a physical io, make sure that it is properly aligned * before proceeding. */ if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) { ASSERT0(P2PHASE(zio->io_offset, align)); ASSERT0(P2PHASE(zio->io_size, align)); } else { /* * For physical writes, we allow 512b aligned writes and assume * the device will perform a read-modify-write as necessary. */ ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE)); ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE)); } VERIFY(zio->io_type != ZIO_TYPE_WRITE || spa_writeable(spa)); /* * If this is a repair I/O, and there's no self-healing involved -- * that is, we're just resilvering what we expect to resilver -- * then don't do the I/O unless zio's txg is actually in vd's DTL. * This prevents spurious resilvering. * * There are a few ways that we can end up creating these spurious * resilver i/os: * * 1. A resilver i/o will be issued if any DVA in the BP has a * dirty DTL. The mirror code will issue resilver writes to * each DVA, including the one(s) that are not on vdevs with dirty * DTLs. * * 2. With nested replication, which happens when we have a * "replacing" or "spare" vdev that's a child of a mirror or raidz. * For example, given mirror(replacing(A+B), C), it's likely that * only A is out of date (it's the new device). In this case, we'll * read from C, then use the data to resilver A+B -- but we don't * actually want to resilver B, just A. The top-level mirror has no * way to know this, so instead we just discard unnecessary repairs * as we work our way down the vdev tree. * * 3. ZTEST also creates mirrors of mirrors, mirrors of raidz, etc. * The same logic applies to any form of nested replication: ditto * + mirror, RAID-Z + replacing, etc. * * However, indirect vdevs point off to other vdevs which may have * DTL's, so we never bypass them. The child i/os on concrete vdevs * will be properly bypassed instead. * * Leaf DTL_PARTIAL can be empty when a legitimate write comes from * a dRAID spare vdev. For example, when a dRAID spare is first * used, its spare blocks need to be written to but the leaf vdev's * of such blocks can have empty DTL_PARTIAL. * * There seemed no clean way to allow such writes while bypassing * spurious ones. At this point, just avoid all bypassing for dRAID * for correctness. */ if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) && !(zio->io_flags & ZIO_FLAG_SELF_HEAL) && zio->io_txg != 0 && /* not a delegated i/o */ vd->vdev_ops != &vdev_indirect_ops && vd->vdev_top->vdev_ops != &vdev_draid_ops && !vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); zio_vdev_io_bypass(zio); return (zio); } /* * Select the next best leaf I/O to process. Distributed spares are * excluded since they dispatch the I/O directly to a leaf vdev after * applying the dRAID mapping. */ if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops != &vdev_draid_spare_ops && (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_TRIM)) { if (zio_handle_device_injection(vd, zio, ENOSYS) != 0) { /* * "no-op" injections return success, but do no actual * work. Just skip the remaining vdev stages. */ zio_vdev_io_bypass(zio); zio_interrupt(zio); return (NULL); } if ((zio = vdev_queue_io(zio)) == NULL) return (NULL); if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); zio_interrupt(zio); return (NULL); } zio->io_delay = gethrtime(); } vd->vdev_ops->vdev_op_io_start(zio); return (NULL); } static zio_t * zio_vdev_io_done(zio_t *zio) { vdev_t *vd = zio->io_vd; vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops; boolean_t unexpected_error = B_FALSE; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) { return (NULL); } ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_FLUSH || zio->io_type == ZIO_TYPE_TRIM); if (zio->io_delay) zio->io_delay = gethrtime() - zio->io_delay; if (vd != NULL && vd->vdev_ops->vdev_op_leaf && vd->vdev_ops != &vdev_draid_spare_ops) { if (zio->io_type != ZIO_TYPE_FLUSH) vdev_queue_io_done(zio); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_device_injections(vd, zio, EIO, EILSEQ); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_label_injection(zio, EIO); - if (zio->io_error && zio->io_type != ZIO_TYPE_TRIM) { + if (zio->io_error && zio->io_type != ZIO_TYPE_FLUSH && + zio->io_type != ZIO_TYPE_TRIM) { if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); } else { unexpected_error = B_TRUE; } } } ops->vdev_op_io_done(zio); if (unexpected_error && vd->vdev_remove_wanted == B_FALSE) VERIFY(vdev_probe(vd, zio) == NULL); return (zio); } /* * This function is used to change the priority of an existing zio that is * currently in-flight. This is used by the arc to upgrade priority in the * event that a demand read is made for a block that is currently queued * as a scrub or async read IO. Otherwise, the high priority read request * would end up having to wait for the lower priority IO. */ void zio_change_priority(zio_t *pio, zio_priority_t priority) { zio_t *cio, *cio_next; zio_link_t *zl = NULL; ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE); if (pio->io_vd != NULL && pio->io_vd->vdev_ops->vdev_op_leaf) { vdev_queue_change_io_priority(pio, priority); } else { pio->io_priority = priority; } mutex_enter(&pio->io_lock); for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); zio_change_priority(cio, priority); } mutex_exit(&pio->io_lock); } /* * For non-raidz ZIOs, we can just copy aside the bad data read from the * disk, and use that to finish the checksum ereport later. */ static void zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr, const abd_t *good_buf) { /* no processing needed */ zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE); } void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr) { void *abd = abd_alloc_sametype(zio->io_abd, zio->io_size); abd_copy(abd, zio->io_abd, zio->io_size); zcr->zcr_cbinfo = zio->io_size; zcr->zcr_cbdata = abd; zcr->zcr_finish = zio_vsd_default_cksum_finish; zcr->zcr_free = zio_abd_free; } static zio_t * zio_vdev_io_assess(zio_t *zio) { vdev_t *vd = zio->io_vd; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) { return (NULL); } if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_exit(zio->io_spa, SCL_ZIO, zio); if (zio->io_vsd != NULL) { zio->io_vsd_ops->vsd_free(zio); zio->io_vsd = NULL; } if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_fault_injection(zio, EIO); /* * If the I/O failed, determine whether we should attempt to retry it. * * On retry, we cut in line in the issue queue, since we don't want * compression/checksumming/etc. work to prevent our (cheap) IO reissue. */ if (zio->io_error && vd == NULL && !(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) { ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */ zio->io_error = 0; zio->io_flags |= ZIO_FLAG_IO_RETRY | ZIO_FLAG_DONT_AGGREGATE; zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, zio_requeue_io_start_cut_in_line); return (NULL); } /* * If we got an error on a leaf device, convert it to ENXIO * if the device is not accessible at all. */ if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf && !vdev_accessible(vd, zio)) zio->io_error = SET_ERROR(ENXIO); /* * If we can't write to an interior vdev (mirror or RAID-Z), * set vdev_cant_write so that we stop trying to allocate from it. */ if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE && vd != NULL && !vd->vdev_ops->vdev_op_leaf) { vdev_dbgmsg(vd, "zio_vdev_io_assess(zio=%px) setting " "cant_write=TRUE due to write failure with ENXIO", zio); vd->vdev_cant_write = B_TRUE; } /* * If a cache flush returns ENOTSUP or ENOTTY, we know that no future * attempts will ever succeed. In this case we set a persistent * boolean flag so that we don't bother with it in the future. */ if ((zio->io_error == ENOTSUP || zio->io_error == ENOTTY) && zio->io_type == ZIO_TYPE_FLUSH && vd != NULL) vd->vdev_nowritecache = B_TRUE; if (zio->io_error) zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; return (zio); } void zio_vdev_io_reissue(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_stage >>= 1; } void zio_vdev_io_redone(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE); zio->io_stage >>= 1; } void zio_vdev_io_bypass(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_flags |= ZIO_FLAG_IO_BYPASS; zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1; } /* * ========================================================================== * Encrypt and store encryption parameters * ========================================================================== */ /* * This function is used for ZIO_STAGE_ENCRYPT. It is responsible for * managing the storage of encryption parameters and passing them to the * lower-level encryption functions. */ static zio_t * zio_encrypt(zio_t *zio) { zio_prop_t *zp = &zio->io_prop; spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; uint64_t psize = BP_GET_PSIZE(bp); uint64_t dsobj = zio->io_bookmark.zb_objset; dmu_object_type_t ot = BP_GET_TYPE(bp); void *enc_buf = NULL; abd_t *eabd = NULL; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; boolean_t no_crypt = B_FALSE; /* the root zio already encrypted the data */ if (zio->io_child_type == ZIO_CHILD_GANG) return (zio); /* only ZIL blocks are re-encrypted on rewrite */ if (!IO_IS_ALLOCATING(zio) && ot != DMU_OT_INTENT_LOG) return (zio); if (!(zp->zp_encrypt || BP_IS_ENCRYPTED(bp))) { BP_SET_CRYPT(bp, B_FALSE); return (zio); } /* if we are doing raw encryption set the provided encryption params */ if (zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) { ASSERT0(BP_GET_LEVEL(bp)); BP_SET_CRYPT(bp, B_TRUE); BP_SET_BYTEORDER(bp, zp->zp_byteorder); if (ot != DMU_OT_OBJSET) zio_crypt_encode_mac_bp(bp, zp->zp_mac); /* dnode blocks must be written out in the provided byteorder */ if (zp->zp_byteorder != ZFS_HOST_BYTEORDER && ot == DMU_OT_DNODE) { void *bswap_buf = zio_buf_alloc(psize); abd_t *babd = abd_get_from_buf(bswap_buf, psize); ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); abd_copy_to_buf(bswap_buf, zio->io_abd, psize); dmu_ot_byteswap[DMU_OT_BYTESWAP(ot)].ob_func(bswap_buf, psize); abd_take_ownership_of_buf(babd, B_TRUE); zio_push_transform(zio, babd, psize, psize, NULL); } if (DMU_OT_IS_ENCRYPTED(ot)) zio_crypt_encode_params_bp(bp, zp->zp_salt, zp->zp_iv); return (zio); } /* indirect blocks only maintain a cksum of the lower level MACs */ if (BP_GET_LEVEL(bp) > 0) { BP_SET_CRYPT(bp, B_TRUE); VERIFY0(zio_crypt_do_indirect_mac_checksum_abd(B_TRUE, zio->io_orig_abd, BP_GET_LSIZE(bp), BP_SHOULD_BYTESWAP(bp), mac)); zio_crypt_encode_mac_bp(bp, mac); return (zio); } /* * Objset blocks are a special case since they have 2 256-bit MACs * embedded within them. */ if (ot == DMU_OT_OBJSET) { ASSERT0(DMU_OT_IS_ENCRYPTED(ot)); ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); BP_SET_CRYPT(bp, B_TRUE); VERIFY0(spa_do_crypt_objset_mac_abd(B_TRUE, spa, dsobj, zio->io_abd, psize, BP_SHOULD_BYTESWAP(bp))); return (zio); } /* unencrypted object types are only authenticated with a MAC */ if (!DMU_OT_IS_ENCRYPTED(ot)) { BP_SET_CRYPT(bp, B_TRUE); VERIFY0(spa_do_crypt_mac_abd(B_TRUE, spa, dsobj, zio->io_abd, psize, mac)); zio_crypt_encode_mac_bp(bp, mac); return (zio); } /* * Later passes of sync-to-convergence may decide to rewrite data * in place to avoid more disk reallocations. This presents a problem * for encryption because this constitutes rewriting the new data with * the same encryption key and IV. However, this only applies to blocks * in the MOS (particularly the spacemaps) and we do not encrypt the * MOS. We assert that the zio is allocating or an intent log write * to enforce this. */ ASSERT(IO_IS_ALLOCATING(zio) || ot == DMU_OT_INTENT_LOG); ASSERT(BP_GET_LEVEL(bp) == 0 || ot == DMU_OT_INTENT_LOG); ASSERT(spa_feature_is_active(spa, SPA_FEATURE_ENCRYPTION)); ASSERT3U(psize, !=, 0); enc_buf = zio_buf_alloc(psize); eabd = abd_get_from_buf(enc_buf, psize); abd_take_ownership_of_buf(eabd, B_TRUE); /* * For an explanation of what encryption parameters are stored * where, see the block comment in zio_crypt.c. */ if (ot == DMU_OT_INTENT_LOG) { zio_crypt_decode_params_bp(bp, salt, iv); } else { BP_SET_CRYPT(bp, B_TRUE); } /* Perform the encryption. This should not fail */ VERIFY0(spa_do_crypt_abd(B_TRUE, spa, &zio->io_bookmark, BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, psize, zio->io_abd, eabd, &no_crypt)); /* encode encryption metadata into the bp */ if (ot == DMU_OT_INTENT_LOG) { /* * ZIL blocks store the MAC in the embedded checksum, so the * transform must always be applied. */ zio_crypt_encode_mac_zil(enc_buf, mac); zio_push_transform(zio, eabd, psize, psize, NULL); } else { BP_SET_CRYPT(bp, B_TRUE); zio_crypt_encode_params_bp(bp, salt, iv); zio_crypt_encode_mac_bp(bp, mac); if (no_crypt) { ASSERT3U(ot, ==, DMU_OT_DNODE); abd_free(eabd); } else { zio_push_transform(zio, eabd, psize, psize, NULL); } } return (zio); } /* * ========================================================================== * Generate and verify checksums * ========================================================================== */ static zio_t * zio_checksum_generate(zio_t *zio) { blkptr_t *bp = zio->io_bp; enum zio_checksum checksum; if (bp == NULL) { /* * This is zio_write_phys(). * We're either generating a label checksum, or none at all. */ checksum = zio->io_prop.zp_checksum; if (checksum == ZIO_CHECKSUM_OFF) return (zio); ASSERT(checksum == ZIO_CHECKSUM_LABEL); } else { if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) { ASSERT(!IO_IS_ALLOCATING(zio)); checksum = ZIO_CHECKSUM_GANG_HEADER; } else { checksum = BP_GET_CHECKSUM(bp); } } zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size); return (zio); } static zio_t * zio_checksum_verify(zio_t *zio) { zio_bad_cksum_t info; blkptr_t *bp = zio->io_bp; int error; ASSERT(zio->io_vd != NULL); if (bp == NULL) { /* * This is zio_read_phys(). * We're either verifying a label checksum, or nothing at all. */ if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF) return (zio); ASSERT3U(zio->io_prop.zp_checksum, ==, ZIO_CHECKSUM_LABEL); } if ((error = zio_checksum_error(zio, &info)) != 0) { zio->io_error = error; if (error == ECKSUM && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { mutex_enter(&zio->io_vd->vdev_stat_lock); zio->io_vd->vdev_stat.vs_checksum_errors++; mutex_exit(&zio->io_vd->vdev_stat_lock); (void) zfs_ereport_start_checksum(zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, zio->io_offset, zio->io_size, &info); } } return (zio); } /* * Called by RAID-Z to ensure we don't compute the checksum twice. */ void zio_checksum_verified(zio_t *zio) { zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } /* * ========================================================================== * Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other. * An error of 0 indicates success. ENXIO indicates whole-device failure, * which may be transient (e.g. unplugged) or permanent. ECKSUM and EIO * indicate errors that are specific to one I/O, and most likely permanent. * Any other error is presumed to be worse because we weren't expecting it. * ========================================================================== */ int zio_worst_error(int e1, int e2) { static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO }; int r1, r2; for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++) if (e1 == zio_error_rank[r1]) break; for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++) if (e2 == zio_error_rank[r2]) break; return (r1 > r2 ? e1 : e2); } /* * ========================================================================== * I/O completion * ========================================================================== */ static zio_t * zio_ready(zio_t *zio) { blkptr_t *bp = zio->io_bp; zio_t *pio, *pio_next; zio_link_t *zl = NULL; if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT | ZIO_CHILD_GANG_BIT | ZIO_CHILD_DDT_BIT, ZIO_WAIT_READY)) { return (NULL); } if (zio->io_ready) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg || BP_IS_HOLE(bp) || (zio->io_flags & ZIO_FLAG_NOPWRITE)); ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0); zio->io_ready(zio); } #ifdef ZFS_DEBUG if (bp != NULL && bp != &zio->io_bp_copy) zio->io_bp_copy = *bp; #endif if (zio->io_error != 0) { zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(zio->io_metaslab_class != NULL); ASSERT(ZIO_HAS_ALLOCATOR(zio)); /* * We were unable to allocate anything, unreserve and * issue the next I/O to allocate. */ metaslab_class_throttle_unreserve( zio->io_metaslab_class, zio->io_prop.zp_copies, zio->io_allocator, zio); zio_allocate_dispatch(zio->io_spa, zio->io_allocator); } } mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_READY] = 1; pio = zio_walk_parents(zio, &zl); mutex_exit(&zio->io_lock); /* * As we notify zio's parents, new parents could be added. * New parents go to the head of zio's io_parent_list, however, * so we will (correctly) not notify them. The remainder of zio's * io_parent_list, from 'pio_next' onward, cannot change because * all parents must wait for us to be done before they can be done. */ for (; pio != NULL; pio = pio_next) { pio_next = zio_walk_parents(zio, &zl); zio_notify_parent(pio, zio, ZIO_WAIT_READY, NULL); } if (zio->io_flags & ZIO_FLAG_NODATA) { if (bp != NULL && BP_IS_GANG(bp)) { zio->io_flags &= ~ZIO_FLAG_NODATA; } else { ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE); zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } } if (zio_injection_enabled && zio->io_spa->spa_syncing_txg == zio->io_txg) zio_handle_ignored_writes(zio); return (zio); } /* * Update the allocation throttle accounting. */ static void zio_dva_throttle_done(zio_t *zio) { zio_t *lio __maybe_unused = zio->io_logical; zio_t *pio = zio_unique_parent(zio); vdev_t *vd = zio->io_vd; int flags = METASLAB_ASYNC_ALLOC; ASSERT3P(zio->io_bp, !=, NULL); ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE); ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE); ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); ASSERT(vd != NULL); ASSERT3P(vd, ==, vd->vdev_top); ASSERT(zio_injection_enabled || !(zio->io_flags & ZIO_FLAG_IO_RETRY)); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR)); ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING); ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA)); /* * Parents of gang children can have two flavors -- ones that * allocated the gang header (will have ZIO_FLAG_IO_REWRITE set) * and ones that allocated the constituent blocks. The allocation * throttle needs to know the allocating parent zio so we must find * it here. */ if (pio->io_child_type == ZIO_CHILD_GANG) { /* * If our parent is a rewrite gang child then our grandparent * would have been the one that performed the allocation. */ if (pio->io_flags & ZIO_FLAG_IO_REWRITE) pio = zio_unique_parent(pio); flags |= METASLAB_GANG_CHILD; } ASSERT(IO_IS_ALLOCATING(pio)); ASSERT(ZIO_HAS_ALLOCATOR(pio)); ASSERT3P(zio, !=, zio->io_logical); ASSERT(zio->io_logical != NULL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR)); ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE); ASSERT(zio->io_metaslab_class != NULL); mutex_enter(&pio->io_lock); metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags, pio->io_allocator, B_TRUE); mutex_exit(&pio->io_lock); metaslab_class_throttle_unreserve(zio->io_metaslab_class, 1, pio->io_allocator, pio); /* * Call into the pipeline to see if there is more work that * needs to be done. If there is work to be done it will be * dispatched to another taskq thread. */ zio_allocate_dispatch(zio->io_spa, pio->io_allocator); } static zio_t * zio_done(zio_t *zio) { /* * Always attempt to keep stack usage minimal here since * we can be called recursively up to 19 levels deep. */ const uint64_t psize = zio->io_size; zio_t *pio, *pio_next; zio_link_t *zl = NULL; /* * If our children haven't all completed, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_ALL_BITS, ZIO_WAIT_DONE)) { return (NULL); } /* * If the allocation throttle is enabled, then update the accounting. * We only track child I/Os that are part of an allocating async * write. We must do this since the allocation is performed * by the logical I/O but the actual write is done by child I/Os. */ if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING && zio->io_child_type == ZIO_CHILD_VDEV) { ASSERT(zio->io_metaslab_class != NULL); ASSERT(zio->io_metaslab_class->mc_alloc_throttle_enabled); zio_dva_throttle_done(zio); } /* * If the allocation throttle is enabled, verify that * we have decremented the refcounts for every I/O that was throttled. */ if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(zio->io_bp != NULL); ASSERT(ZIO_HAS_ALLOCATOR(zio)); metaslab_group_alloc_verify(zio->io_spa, zio->io_bp, zio, zio->io_allocator); VERIFY(zfs_refcount_not_held(&zio->io_metaslab_class-> mc_allocator[zio->io_allocator].mca_alloc_slots, zio)); } for (int c = 0; c < ZIO_CHILD_TYPES; c++) for (int w = 0; w < ZIO_WAIT_TYPES; w++) ASSERT(zio->io_children[c][w] == 0); if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) { ASSERT(zio->io_bp->blk_pad[0] == 0); ASSERT(zio->io_bp->blk_pad[1] == 0); ASSERT(memcmp(zio->io_bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 || (zio->io_bp == zio_unique_parent(zio)->io_bp)); if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) && zio->io_bp_override == NULL && !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) { ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp)); ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 || (BP_COUNT_GANG(zio->io_bp) == BP_GET_NDVAS(zio->io_bp))); } if (zio->io_flags & ZIO_FLAG_NOPWRITE) VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig)); } /* * If there were child vdev/gang/ddt errors, they apply to us now. */ zio_inherit_child_errors(zio, ZIO_CHILD_VDEV); zio_inherit_child_errors(zio, ZIO_CHILD_GANG); zio_inherit_child_errors(zio, ZIO_CHILD_DDT); /* * If the I/O on the transformed data was successful, generate any * checksum reports now while we still have the transformed data. */ if (zio->io_error == 0) { while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; uint64_t align = zcr->zcr_align; uint64_t asize = P2ROUNDUP(psize, align); abd_t *adata = zio->io_abd; if (adata != NULL && asize != psize) { adata = abd_alloc(asize, B_TRUE); abd_copy(adata, zio->io_abd, psize); abd_zero_off(adata, psize, asize - psize); } zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, adata); zfs_ereport_free_checksum(zcr); if (adata != NULL && asize != psize) abd_free(adata); } } zio_pop_transforms(zio); /* note: may set zio->io_error */ vdev_stat_update(zio, psize); /* * If this I/O is attached to a particular vdev is slow, exceeding * 30 seconds to complete, post an error described the I/O delay. * We ignore these errors if the device is currently unavailable. */ if (zio->io_delay >= MSEC2NSEC(zio_slow_io_ms)) { if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) { /* * We want to only increment our slow IO counters if * the IO is valid (i.e. not if the drive is removed). * * zfs_ereport_post() will also do these checks, but * it can also ratelimit and have other failures, so we * need to increment the slow_io counters independent * of it. */ if (zfs_ereport_is_valid(FM_EREPORT_ZFS_DELAY, zio->io_spa, zio->io_vd, zio)) { mutex_enter(&zio->io_vd->vdev_stat_lock); zio->io_vd->vdev_stat.vs_slow_ios++; mutex_exit(&zio->io_vd->vdev_stat_lock); (void) zfs_ereport_post(FM_EREPORT_ZFS_DELAY, zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0); } } } if (zio->io_error) { /* * If this I/O is attached to a particular vdev, * generate an error message describing the I/O failure * at the block level. We ignore these errors if the * device is currently unavailable. */ if (zio->io_error != ECKSUM && zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) { int ret = zfs_ereport_post(FM_EREPORT_ZFS_IO, zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0); if (ret != EALREADY) { mutex_enter(&zio->io_vd->vdev_stat_lock); if (zio->io_type == ZIO_TYPE_READ) zio->io_vd->vdev_stat.vs_read_errors++; else if (zio->io_type == ZIO_TYPE_WRITE) zio->io_vd->vdev_stat.vs_write_errors++; mutex_exit(&zio->io_vd->vdev_stat_lock); } } if ((zio->io_error == EIO || !(zio->io_flags & (ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) && zio == zio->io_logical) { /* * For logical I/O requests, tell the SPA to log the * error and generate a logical data ereport. */ spa_log_error(zio->io_spa, &zio->io_bookmark, BP_GET_LOGICAL_BIRTH(zio->io_bp)); (void) zfs_ereport_post(FM_EREPORT_ZFS_DATA, zio->io_spa, NULL, &zio->io_bookmark, zio, 0); } } if (zio->io_error && zio == zio->io_logical) { /* * Determine whether zio should be reexecuted. This will * propagate all the way to the root via zio_notify_parent(). */ ASSERT(zio->io_vd == NULL && zio->io_bp != NULL); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (IO_IS_ALLOCATING(zio) && !(zio->io_flags & ZIO_FLAG_CANFAIL)) { if (zio->io_error != ENOSPC) zio->io_reexecute |= ZIO_REEXECUTE_NOW; else zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; } if ((zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_FREE) && !(zio->io_flags & ZIO_FLAG_SCAN_THREAD) && zio->io_error == ENXIO && spa_load_state(zio->io_spa) == SPA_LOAD_NONE && spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; /* * Here is a possibly good place to attempt to do * either combinatorial reconstruction or error correction * based on checksums. It also might be a good place * to send out preliminary ereports before we suspend * processing. */ } /* * If there were logical child errors, they apply to us now. * We defer this until now to avoid conflating logical child * errors with errors that happened to the zio itself when * updating vdev stats and reporting FMA events above. */ zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL); if ((zio->io_error || zio->io_reexecute) && IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio && !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE))) zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp); zio_gang_tree_free(&zio->io_gang_tree); /* * Godfather I/Os should never suspend. */ if ((zio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND; if (zio->io_reexecute) { /* * This is a logical I/O that wants to reexecute. * * Reexecute is top-down. When an i/o fails, if it's not * the root, it simply notifies its parent and sticks around. * The parent, seeing that it still has children in zio_done(), * does the same. This percolates all the way up to the root. * The root i/o will reexecute or suspend the entire tree. * * This approach ensures that zio_reexecute() honors * all the original i/o dependency relationships, e.g. * parents not executing until children are ready. */ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); zio->io_gang_leader = NULL; mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); /* * "The Godfather" I/O monitors its children but is * not a true parent to them. It will track them through * the pipeline but severs its ties whenever they get into * trouble (e.g. suspended). This allows "The Godfather" * I/O to return status without blocking. */ zl = NULL; for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) { zio_link_t *remove_zl = zl; pio_next = zio_walk_parents(zio, &zl); if ((pio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) { zio_remove_child(pio, zio, remove_zl); /* * This is a rare code path, so we don't * bother with "next_to_execute". */ zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL); } } if ((pio = zio_unique_parent(zio)) != NULL) { /* * We're not a root i/o, so there's nothing to do * but notify our parent. Don't propagate errors * upward since we haven't permanently failed yet. */ ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE; /* * This is a rare code path, so we don't bother with * "next_to_execute". */ zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL); } else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) { /* * We'd fail again if we reexecuted now, so suspend * until conditions improve (e.g. device comes online). */ zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR); } else { /* * Reexecution is potentially a huge amount of work. * Hand it off to the otherwise-unused claim taskq. */ ASSERT(taskq_empty_ent(&zio->io_tqent)); spa_taskq_dispatch_ent(zio->io_spa, ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE, zio_reexecute, zio, 0, &zio->io_tqent, NULL); } return (NULL); } ASSERT(list_is_empty(&zio->io_child_list)); ASSERT(zio->io_reexecute == 0); ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL)); /* * Report any checksum errors, since the I/O is complete. */ while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, NULL); zfs_ereport_free_checksum(zcr); } /* * It is the responsibility of the done callback to ensure that this * particular zio is no longer discoverable for adoption, and as * such, cannot acquire any new parents. */ if (zio->io_done) zio->io_done(zio); mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); /* * We are done executing this zio. We may want to execute a parent * next. See the comment in zio_notify_parent(). */ zio_t *next_to_execute = NULL; zl = NULL; for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) { zio_link_t *remove_zl = zl; pio_next = zio_walk_parents(zio, &zl); zio_remove_child(pio, zio, remove_zl); zio_notify_parent(pio, zio, ZIO_WAIT_DONE, &next_to_execute); } if (zio->io_waiter != NULL) { mutex_enter(&zio->io_lock); zio->io_executor = NULL; cv_broadcast(&zio->io_cv); mutex_exit(&zio->io_lock); } else { zio_destroy(zio); } return (next_to_execute); } /* * ========================================================================== * I/O pipeline definition * ========================================================================== */ static zio_pipe_stage_t *zio_pipeline[] = { NULL, zio_read_bp_init, zio_write_bp_init, zio_free_bp_init, zio_issue_async, zio_write_compress, zio_encrypt, zio_checksum_generate, zio_nop_write, zio_brt_free, zio_ddt_read_start, zio_ddt_read_done, zio_ddt_write, zio_ddt_free, zio_gang_assemble, zio_gang_issue, zio_dva_throttle, zio_dva_allocate, zio_dva_free, zio_dva_claim, zio_ready, zio_vdev_io_start, zio_vdev_io_done, zio_vdev_io_assess, zio_checksum_verify, zio_done }; /* * Compare two zbookmark_phys_t's to see which we would reach first in a * pre-order traversal of the object tree. * * This is simple in every case aside from the meta-dnode object. For all other * objects, we traverse them in order (object 1 before object 2, and so on). * However, all of these objects are traversed while traversing object 0, since * the data it points to is the list of objects. Thus, we need to convert to a * canonical representation so we can compare meta-dnode bookmarks to * non-meta-dnode bookmarks. * * We do this by calculating "equivalents" for each field of the zbookmark. * zbookmarks outside of the meta-dnode use their own object and level, and * calculate the level 0 equivalent (the first L0 blkid that is contained in the * blocks this bookmark refers to) by multiplying their blkid by their span * (the number of L0 blocks contained within one block at their level). * zbookmarks inside the meta-dnode calculate their object equivalent * (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use * level + 1<<31 (any value larger than a level could ever be) for their level. * This causes them to always compare before a bookmark in their object * equivalent, compare appropriately to bookmarks in other objects, and to * compare appropriately to other bookmarks in the meta-dnode. */ int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2) { /* * These variables represent the "equivalent" values for the zbookmark, * after converting zbookmarks inside the meta dnode to their * normal-object equivalents. */ uint64_t zb1obj, zb2obj; uint64_t zb1L0, zb2L0; uint64_t zb1level, zb2level; if (zb1->zb_object == zb2->zb_object && zb1->zb_level == zb2->zb_level && zb1->zb_blkid == zb2->zb_blkid) return (0); IMPLY(zb1->zb_level > 0, ibs1 >= SPA_MINBLOCKSHIFT); IMPLY(zb2->zb_level > 0, ibs2 >= SPA_MINBLOCKSHIFT); /* * BP_SPANB calculates the span in blocks. */ zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level); zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level); if (zb1->zb_object == DMU_META_DNODE_OBJECT) { zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb1L0 = 0; zb1level = zb1->zb_level + COMPARE_META_LEVEL; } else { zb1obj = zb1->zb_object; zb1level = zb1->zb_level; } if (zb2->zb_object == DMU_META_DNODE_OBJECT) { zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb2L0 = 0; zb2level = zb2->zb_level + COMPARE_META_LEVEL; } else { zb2obj = zb2->zb_object; zb2level = zb2->zb_level; } /* Now that we have a canonical representation, do the comparison. */ if (zb1obj != zb2obj) return (zb1obj < zb2obj ? -1 : 1); else if (zb1L0 != zb2L0) return (zb1L0 < zb2L0 ? -1 : 1); else if (zb1level != zb2level) return (zb1level > zb2level ? -1 : 1); /* * This can (theoretically) happen if the bookmarks have the same object * and level, but different blkids, if the block sizes are not the same. * There is presently no way to change the indirect block sizes */ return (0); } /* * This function checks the following: given that last_block is the place that * our traversal stopped last time, does that guarantee that we've visited * every node under subtree_root? Therefore, we can't just use the raw output * of zbookmark_compare. We have to pass in a modified version of * subtree_root; by incrementing the block id, and then checking whether * last_block is before or equal to that, we can tell whether or not having * visited last_block implies that all of subtree_root's children have been * visited. */ boolean_t zbookmark_subtree_completed(const dnode_phys_t *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block) { zbookmark_phys_t mod_zb = *subtree_root; mod_zb.zb_blkid++; ASSERT0(last_block->zb_level); /* The objset_phys_t isn't before anything. */ if (dnp == NULL) return (B_FALSE); /* * We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the * data block size in sectors, because that variable is only used if * the bookmark refers to a block in the meta-dnode. Since we don't * know without examining it what object it refers to, and there's no * harm in passing in this value in other cases, we always pass it in. * * We pass in 0 for the indirect block size shift because zb2 must be * level 0. The indirect block size is only used to calculate the span * of the bookmark, but since the bookmark must be level 0, the span is * always 1, so the math works out. * * If you make changes to how the zbookmark_compare code works, be sure * to make sure that this code still works afterwards. */ return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift, 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb, last_block) <= 0); } /* * This function is similar to zbookmark_subtree_completed(), but returns true * if subtree_root is equal or ahead of last_block, i.e. still to be done. */ boolean_t zbookmark_subtree_tbd(const dnode_phys_t *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block) { ASSERT0(last_block->zb_level); if (dnp == NULL) return (B_FALSE); return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift, 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, subtree_root, last_block) >= 0); } EXPORT_SYMBOL(zio_type_name); EXPORT_SYMBOL(zio_buf_alloc); EXPORT_SYMBOL(zio_data_buf_alloc); EXPORT_SYMBOL(zio_buf_free); EXPORT_SYMBOL(zio_data_buf_free); ZFS_MODULE_PARAM(zfs_zio, zio_, slow_io_ms, INT, ZMOD_RW, "Max I/O completion time (milliseconds) before marking it as slow"); ZFS_MODULE_PARAM(zfs_zio, zio_, requeue_io_start_cut_in_line, INT, ZMOD_RW, "Prioritize requeued I/O"); ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_deferred_free, UINT, ZMOD_RW, "Defer frees starting in this pass"); ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_dont_compress, UINT, ZMOD_RW, "Don't compress starting in this pass"); ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_rewrite, UINT, ZMOD_RW, "Rewrite new bps starting in this pass"); ZFS_MODULE_PARAM(zfs_zio, zio_, dva_throttle_enabled, INT, ZMOD_RW, "Throttle block allocations in the ZIO pipeline"); ZFS_MODULE_PARAM(zfs_zio, zio_, deadman_log_all, INT, ZMOD_RW, "Log all slow ZIOs, not just those with vdevs"); diff --git a/sys/contrib/openzfs/module/zfs/zio_inject.c b/sys/contrib/openzfs/module/zfs/zio_inject.c index 1af2c26f8a43..3773e400d799 100644 --- a/sys/contrib/openzfs/module/zfs/zio_inject.c +++ b/sys/contrib/openzfs/module/zfs/zio_inject.c @@ -1,976 +1,1096 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or https://opensource.org/licenses/CDDL-1.0. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2015 by Delphix. All rights reserved. * Copyright (c) 2017, Intel Corporation. + * Copyright (c) 2024, Klara Inc. */ /* * ZFS fault injection * * To handle fault injection, we keep track of a series of zinject_record_t * structures which describe which logical block(s) should be injected with a * fault. These are kept in a global list. Each record corresponds to a given * spa_t and maintains a special hold on the spa_t so that it cannot be deleted * or exported while the injection record exists. * * Device level injection is done using the 'zi_guid' field. If this is set, it * means that the error is destined for a particular device, not a piece of * data. * * This is a rather poor data structure and algorithm, but we don't expect more * than a few faults at any one time, so it should be sufficient for our needs. */ #include #include #include #include #include #include #include uint32_t zio_injection_enabled = 0; /* * Data describing each zinject handler registered on the system, and * contains the list node linking the handler in the global zinject * handler list. */ typedef struct inject_handler { int zi_id; spa_t *zi_spa; + char *zi_spa_name; /* ZINJECT_DELAY_IMPORT only */ zinject_record_t zi_record; uint64_t *zi_lanes; int zi_next_lane; list_node_t zi_link; } inject_handler_t; /* * List of all zinject handlers registered on the system, protected by * the inject_lock defined below. */ static list_t inject_handlers; /* * This protects insertion into, and traversal of, the inject handler * list defined above; as well as the inject_delay_count. Any time a * handler is inserted or removed from the list, this lock should be * taken as a RW_WRITER; and any time traversal is done over the list * (without modification to it) this lock should be taken as a RW_READER. */ static krwlock_t inject_lock; /* * This holds the number of zinject delay handlers that have been * registered on the system. It is protected by the inject_lock defined * above. Thus modifications to this count must be a RW_WRITER of the * inject_lock, and reads of this count must be (at least) a RW_READER * of the lock. */ static int inject_delay_count = 0; /* * This lock is used only in zio_handle_io_delay(), refer to the comment * in that function for more details. */ static kmutex_t inject_delay_mtx; /* * Used to assign unique identifying numbers to each new zinject handler. */ static int inject_next_id = 1; /* * Test if the requested frequency was triggered */ static boolean_t freq_triggered(uint32_t frequency) { /* * zero implies always (100%) */ if (frequency == 0) return (B_TRUE); /* * Note: we still handle legacy (unscaled) frequency values */ uint32_t maximum = (frequency <= 100) ? 100 : ZI_PERCENTAGE_MAX; return (random_in_range(maximum) < frequency); } /* * Returns true if the given record matches the I/O in progress. */ static boolean_t zio_match_handler(const zbookmark_phys_t *zb, uint64_t type, int dva, zinject_record_t *record, int error) { /* * Check for a match against the MOS, which is based on type */ if (zb->zb_objset == DMU_META_OBJSET && record->zi_objset == DMU_META_OBJSET && record->zi_object == DMU_META_DNODE_OBJECT) { if (record->zi_type == DMU_OT_NONE || type == record->zi_type) return (freq_triggered(record->zi_freq)); else return (B_FALSE); } /* * Check for an exact match. */ if (zb->zb_objset == record->zi_objset && zb->zb_object == record->zi_object && zb->zb_level == record->zi_level && zb->zb_blkid >= record->zi_start && zb->zb_blkid <= record->zi_end && (record->zi_dvas == 0 || (dva != ZI_NO_DVA && (record->zi_dvas & (1ULL << dva)))) && error == record->zi_error) { return (freq_triggered(record->zi_freq)); } return (B_FALSE); } /* * Panic the system when a config change happens in the function * specified by tag. */ void zio_handle_panic_injection(spa_t *spa, const char *tag, uint64_t type) { inject_handler_t *handler; rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { if (spa != handler->zi_spa) continue; if (handler->zi_record.zi_type == type && strcmp(tag, handler->zi_record.zi_func) == 0) panic("Panic requested in function %s\n", tag); } rw_exit(&inject_lock); } /* * Inject a decryption failure. Decryption failures can occur in * both the ARC and the ZIO layers. */ int zio_handle_decrypt_injection(spa_t *spa, const zbookmark_phys_t *zb, uint64_t type, int error) { int ret = 0; inject_handler_t *handler; rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { if (spa != handler->zi_spa || handler->zi_record.zi_cmd != ZINJECT_DECRYPT_FAULT) continue; if (zio_match_handler(zb, type, ZI_NO_DVA, &handler->zi_record, error)) { ret = error; break; } } rw_exit(&inject_lock); return (ret); } /* * If this is a physical I/O for a vdev child determine which DVA it is * for. We iterate backwards through the DVAs matching on the offset so * that we end up with ZI_NO_DVA (-1) if we don't find a match. */ static int zio_match_dva(zio_t *zio) { int i = ZI_NO_DVA; if (zio->io_bp != NULL && zio->io_vd != NULL && zio->io_child_type == ZIO_CHILD_VDEV) { for (i = BP_GET_NDVAS(zio->io_bp) - 1; i >= 0; i--) { dva_t *dva = &zio->io_bp->blk_dva[i]; uint64_t off = DVA_GET_OFFSET(dva); vdev_t *vd = vdev_lookup_top(zio->io_spa, DVA_GET_VDEV(dva)); /* Compensate for vdev label added to leaves */ if (zio->io_vd->vdev_ops->vdev_op_leaf) off += VDEV_LABEL_START_SIZE; if (zio->io_vd == vd && zio->io_offset == off) break; } } return (i); } /* * Determine if the I/O in question should return failure. Returns the errno * to be returned to the caller. */ int zio_handle_fault_injection(zio_t *zio, int error) { int ret = 0; inject_handler_t *handler; /* * Ignore I/O not associated with any logical data. */ if (zio->io_logical == NULL) return (0); /* * Currently, we only support fault injection on reads. */ if (zio->io_type != ZIO_TYPE_READ) return (0); /* * A rebuild I/O has no checksum to verify. */ if (zio->io_priority == ZIO_PRIORITY_REBUILD && error == ECKSUM) return (0); rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { if (zio->io_spa != handler->zi_spa || handler->zi_record.zi_cmd != ZINJECT_DATA_FAULT) continue; /* If this handler matches, return the specified error */ if (zio_match_handler(&zio->io_logical->io_bookmark, zio->io_bp ? BP_GET_TYPE(zio->io_bp) : DMU_OT_NONE, zio_match_dva(zio), &handler->zi_record, error)) { ret = error; break; } } rw_exit(&inject_lock); return (ret); } /* * Determine if the zio is part of a label update and has an injection * handler associated with that portion of the label. Currently, we * allow error injection in either the nvlist or the uberblock region of * of the vdev label. */ int zio_handle_label_injection(zio_t *zio, int error) { inject_handler_t *handler; vdev_t *vd = zio->io_vd; uint64_t offset = zio->io_offset; int label; int ret = 0; if (offset >= VDEV_LABEL_START_SIZE && offset < vd->vdev_psize - VDEV_LABEL_END_SIZE) return (0); rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { uint64_t start = handler->zi_record.zi_start; uint64_t end = handler->zi_record.zi_end; if (handler->zi_record.zi_cmd != ZINJECT_LABEL_FAULT) continue; /* * The injection region is the relative offsets within a * vdev label. We must determine the label which is being * updated and adjust our region accordingly. */ label = vdev_label_number(vd->vdev_psize, offset); start = vdev_label_offset(vd->vdev_psize, label, start); end = vdev_label_offset(vd->vdev_psize, label, end); if (zio->io_vd->vdev_guid == handler->zi_record.zi_guid && (offset >= start && offset <= end)) { ret = error; break; } } rw_exit(&inject_lock); return (ret); } static int zio_inject_bitflip_cb(void *data, size_t len, void *private) { zio_t *zio = private; uint8_t *buffer = data; uint_t byte = random_in_range(len); ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ); /* flip a single random bit in an abd data buffer */ buffer[byte] ^= 1 << random_in_range(8); return (1); /* stop after first flip */ } static int zio_handle_device_injection_impl(vdev_t *vd, zio_t *zio, int err1, int err2) { inject_handler_t *handler; int ret = 0; /* * We skip over faults in the labels unless it's during device open * (i.e. zio == NULL) or a device flush (offset is meaningless) */ if (zio != NULL && zio->io_type != ZIO_TYPE_FLUSH) { uint64_t offset = zio->io_offset; if (offset < VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE) return (0); } rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { if (handler->zi_record.zi_cmd != ZINJECT_DEVICE_FAULT) continue; if (vd->vdev_guid == handler->zi_record.zi_guid) { if (handler->zi_record.zi_failfast && (zio == NULL || (zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)))) { continue; } /* Handle type specific I/O failures */ if (zio != NULL && handler->zi_record.zi_iotype != ZIO_TYPES && handler->zi_record.zi_iotype != zio->io_type) continue; if (handler->zi_record.zi_error == err1 || handler->zi_record.zi_error == err2) { /* * limit error injection if requested */ if (!freq_triggered(handler->zi_record.zi_freq)) continue; /* * For a failed open, pretend like the device * has gone away. */ if (err1 == ENXIO) vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; /* * Treat these errors as if they had been * retried so that all the appropriate stats * and FMA events are generated. */ if (!handler->zi_record.zi_failfast && zio != NULL) zio->io_flags |= ZIO_FLAG_IO_RETRY; /* * EILSEQ means flip a bit after a read */ if (handler->zi_record.zi_error == EILSEQ) { if (zio == NULL) break; /* locate buffer data and flip a bit */ (void) abd_iterate_func(zio->io_abd, 0, zio->io_size, zio_inject_bitflip_cb, zio); break; } ret = handler->zi_record.zi_error; break; } if (handler->zi_record.zi_error == ENXIO) { ret = SET_ERROR(EIO); break; } } } rw_exit(&inject_lock); return (ret); } int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error) { return (zio_handle_device_injection_impl(vd, zio, error, INT_MAX)); } int zio_handle_device_injections(vdev_t *vd, zio_t *zio, int err1, int err2) { return (zio_handle_device_injection_impl(vd, zio, err1, err2)); } /* * Simulate hardware that ignores cache flushes. For requested number * of seconds nix the actual writing to disk. */ void zio_handle_ignored_writes(zio_t *zio) { inject_handler_t *handler; rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { /* Ignore errors not destined for this pool */ if (zio->io_spa != handler->zi_spa || handler->zi_record.zi_cmd != ZINJECT_IGNORED_WRITES) continue; /* * Positive duration implies # of seconds, negative * a number of txgs */ if (handler->zi_record.zi_timer == 0) { if (handler->zi_record.zi_duration > 0) handler->zi_record.zi_timer = ddi_get_lbolt64(); else handler->zi_record.zi_timer = zio->io_txg; } /* Have a "problem" writing 60% of the time */ if (random_in_range(100) < 60) zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; break; } rw_exit(&inject_lock); } void spa_handle_ignored_writes(spa_t *spa) { inject_handler_t *handler; if (zio_injection_enabled == 0) return; rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { if (spa != handler->zi_spa || handler->zi_record.zi_cmd != ZINJECT_IGNORED_WRITES) continue; if (handler->zi_record.zi_duration > 0) { VERIFY(handler->zi_record.zi_timer == 0 || ddi_time_after64( (int64_t)handler->zi_record.zi_timer + handler->zi_record.zi_duration * hz, ddi_get_lbolt64())); } else { /* duration is negative so the subtraction here adds */ VERIFY(handler->zi_record.zi_timer == 0 || handler->zi_record.zi_timer - handler->zi_record.zi_duration >= spa_syncing_txg(spa)); } } rw_exit(&inject_lock); } hrtime_t zio_handle_io_delay(zio_t *zio) { vdev_t *vd = zio->io_vd; inject_handler_t *min_handler = NULL; hrtime_t min_target = 0; rw_enter(&inject_lock, RW_READER); /* * inject_delay_count is a subset of zio_injection_enabled that * is only incremented for delay handlers. These checks are * mainly added to remind the reader why we're not explicitly * checking zio_injection_enabled like the other functions. */ IMPLY(inject_delay_count > 0, zio_injection_enabled > 0); IMPLY(zio_injection_enabled == 0, inject_delay_count == 0); /* * If there aren't any inject delay handlers registered, then we * can short circuit and simply return 0 here. A value of zero * informs zio_delay_interrupt() that this request should not be * delayed. This short circuit keeps us from acquiring the * inject_delay_mutex unnecessarily. */ if (inject_delay_count == 0) { rw_exit(&inject_lock); return (0); } /* * Each inject handler has a number of "lanes" associated with * it. Each lane is able to handle requests independently of one * another, and at a latency defined by the inject handler * record's zi_timer field. Thus if a handler in configured with * a single lane with a 10ms latency, it will delay requests * such that only a single request is completed every 10ms. So, * if more than one request is attempted per each 10ms interval, * the average latency of the requests will be greater than * 10ms; but if only a single request is submitted each 10ms * interval the average latency will be 10ms. * * We need to acquire this mutex to prevent multiple concurrent * threads being assigned to the same lane of a given inject * handler. The mutex allows us to perform the following two * operations atomically: * * 1. determine the minimum handler and minimum target * value of all the possible handlers * 2. update that minimum handler's lane array * * Without atomicity, two (or more) threads could pick the same * lane in step (1), and then conflict with each other in step * (2). This could allow a single lane handler to process * multiple requests simultaneously, which shouldn't be possible. */ mutex_enter(&inject_delay_mtx); for (inject_handler_t *handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) { if (handler->zi_record.zi_cmd != ZINJECT_DELAY_IO) continue; if (!freq_triggered(handler->zi_record.zi_freq)) continue; if (vd->vdev_guid != handler->zi_record.zi_guid) continue; if (handler->zi_record.zi_iotype != ZIO_TYPES && handler->zi_record.zi_iotype != zio->io_type) continue; /* * Defensive; should never happen as the array allocation * occurs prior to inserting this handler on the list. */ ASSERT3P(handler->zi_lanes, !=, NULL); /* * This should never happen, the zinject command should * prevent a user from setting an IO delay with zero lanes. */ ASSERT3U(handler->zi_record.zi_nlanes, !=, 0); ASSERT3U(handler->zi_record.zi_nlanes, >, handler->zi_next_lane); /* * We want to issue this IO to the lane that will become * idle the soonest, so we compare the soonest this * specific handler can complete the IO with all other * handlers, to find the lowest value of all possible * lanes. We then use this lane to submit the request. * * Since each handler has a constant value for its * delay, we can just use the "next" lane for that * handler; as it will always be the lane with the * lowest value for that particular handler (i.e. the * lane that will become idle the soonest). This saves a * scan of each handler's lanes array. * * There's two cases to consider when determining when * this specific IO request should complete. If this * lane is idle, we want to "submit" the request now so * it will complete after zi_timer milliseconds. Thus, * we set the target to now + zi_timer. * * If the lane is busy, we want this request to complete * zi_timer milliseconds after the lane becomes idle. * Since the 'zi_lanes' array holds the time at which * each lane will become idle, we use that value to * determine when this request should complete. */ hrtime_t idle = handler->zi_record.zi_timer + gethrtime(); hrtime_t busy = handler->zi_record.zi_timer + handler->zi_lanes[handler->zi_next_lane]; hrtime_t target = MAX(idle, busy); if (min_handler == NULL) { min_handler = handler; min_target = target; continue; } ASSERT3P(min_handler, !=, NULL); ASSERT3U(min_target, !=, 0); /* * We don't yet increment the "next lane" variable since * we still might find a lower value lane in another * handler during any remaining iterations. Once we're * sure we've selected the absolute minimum, we'll claim * the lane and increment the handler's "next lane" * field below. */ if (target < min_target) { min_handler = handler; min_target = target; } } /* * 'min_handler' will be NULL if no IO delays are registered for * this vdev, otherwise it will point to the handler containing * the lane that will become idle the soonest. */ if (min_handler != NULL) { ASSERT3U(min_target, !=, 0); min_handler->zi_lanes[min_handler->zi_next_lane] = min_target; /* * If we've used all possible lanes for this handler, * loop back and start using the first lane again; * otherwise, just increment the lane index. */ min_handler->zi_next_lane = (min_handler->zi_next_lane + 1) % min_handler->zi_record.zi_nlanes; } mutex_exit(&inject_delay_mtx); rw_exit(&inject_lock); return (min_target); } +static void +zio_handle_pool_delay(spa_t *spa, hrtime_t elapsed, zinject_type_t command) +{ + inject_handler_t *handler; + hrtime_t delay = 0; + int id = 0; + + rw_enter(&inject_lock, RW_READER); + + for (handler = list_head(&inject_handlers); + handler != NULL && handler->zi_record.zi_cmd == command; + handler = list_next(&inject_handlers, handler)) { + ASSERT3P(handler->zi_spa_name, !=, NULL); + if (strcmp(spa_name(spa), handler->zi_spa_name) == 0) { + uint64_t pause = + SEC2NSEC(handler->zi_record.zi_duration); + if (pause > elapsed) { + delay = pause - elapsed; + } + id = handler->zi_id; + break; + } + } + + rw_exit(&inject_lock); + + if (delay) { + if (command == ZINJECT_DELAY_IMPORT) { + spa_import_progress_set_notes(spa, "injecting %llu " + "sec delay", (u_longlong_t)NSEC2SEC(delay)); + } + zfs_sleep_until(gethrtime() + delay); + } + if (id) { + /* all done with this one-shot handler */ + zio_clear_fault(id); + } +} + +/* + * For testing, inject a delay during an import + */ +void +zio_handle_import_delay(spa_t *spa, hrtime_t elapsed) +{ + zio_handle_pool_delay(spa, elapsed, ZINJECT_DELAY_IMPORT); +} + +/* + * For testing, inject a delay during an export + */ +void +zio_handle_export_delay(spa_t *spa, hrtime_t elapsed) +{ + zio_handle_pool_delay(spa, elapsed, ZINJECT_DELAY_EXPORT); +} + static int zio_calculate_range(const char *pool, zinject_record_t *record) { dsl_pool_t *dp; dsl_dataset_t *ds; objset_t *os = NULL; dnode_t *dn = NULL; int error; /* * Obtain the dnode for object using pool, objset, and object */ error = dsl_pool_hold(pool, FTAG, &dp); if (error) return (error); error = dsl_dataset_hold_obj(dp, record->zi_objset, FTAG, &ds); dsl_pool_rele(dp, FTAG); if (error) return (error); error = dmu_objset_from_ds(ds, &os); dsl_dataset_rele(ds, FTAG); if (error) return (error); error = dnode_hold(os, record->zi_object, FTAG, &dn); if (error) return (error); /* * Translate the range into block IDs */ if (record->zi_start != 0 || record->zi_end != -1ULL) { record->zi_start >>= dn->dn_datablkshift; record->zi_end >>= dn->dn_datablkshift; } if (record->zi_level > 0) { if (record->zi_level >= dn->dn_nlevels) { dnode_rele(dn, FTAG); return (SET_ERROR(EDOM)); } if (record->zi_start != 0 || record->zi_end != 0) { int shift = dn->dn_indblkshift - SPA_BLKPTRSHIFT; for (int level = record->zi_level; level > 0; level--) { record->zi_start >>= shift; record->zi_end >>= shift; } } } dnode_rele(dn, FTAG); return (0); } +static boolean_t +zio_pool_handler_exists(const char *name, zinject_type_t command) +{ + boolean_t exists = B_FALSE; + + rw_enter(&inject_lock, RW_READER); + for (inject_handler_t *handler = list_head(&inject_handlers); + handler != NULL; handler = list_next(&inject_handlers, handler)) { + if (command != handler->zi_record.zi_cmd) + continue; + + const char *pool = (handler->zi_spa_name != NULL) ? + handler->zi_spa_name : spa_name(handler->zi_spa); + if (strcmp(name, pool) == 0) { + exists = B_TRUE; + break; + } + } + rw_exit(&inject_lock); + + return (exists); +} /* * Create a new handler for the given record. We add it to the list, adding * a reference to the spa_t in the process. We increment zio_injection_enabled, * which is the switch to trigger all fault injection. */ int zio_inject_fault(char *name, int flags, int *id, zinject_record_t *record) { inject_handler_t *handler; int error; spa_t *spa; /* * If this is pool-wide metadata, make sure we unload the corresponding * spa_t, so that the next attempt to load it will trigger the fault. * We call spa_reset() to unload the pool appropriately. */ if (flags & ZINJECT_UNLOAD_SPA) if ((error = spa_reset(name)) != 0) return (error); if (record->zi_cmd == ZINJECT_DELAY_IO) { /* * A value of zero for the number of lanes or for the * delay time doesn't make sense. */ if (record->zi_timer == 0 || record->zi_nlanes == 0) return (SET_ERROR(EINVAL)); /* * The number of lanes is directly mapped to the size of * an array used by the handler. Thus, to ensure the * user doesn't trigger an allocation that's "too large" * we cap the number of lanes here. */ if (record->zi_nlanes >= UINT16_MAX) return (SET_ERROR(EINVAL)); } /* * If the supplied range was in bytes -- calculate the actual blkid */ if (flags & ZINJECT_CALC_RANGE) { error = zio_calculate_range(name, record); if (error != 0) return (error); } if (!(flags & ZINJECT_NULL)) { /* - * spa_inject_ref() will add an injection reference, which will - * prevent the pool from being removed from the namespace while - * still allowing it to be unloaded. + * Pool delays for import or export don't take an + * injection reference on the spa. Instead they + * rely on matching by name. */ - if ((spa = spa_inject_addref(name)) == NULL) - return (SET_ERROR(ENOENT)); + if (record->zi_cmd == ZINJECT_DELAY_IMPORT || + record->zi_cmd == ZINJECT_DELAY_EXPORT) { + if (record->zi_duration <= 0) + return (SET_ERROR(EINVAL)); + /* + * Only one import | export delay handler per pool. + */ + if (zio_pool_handler_exists(name, record->zi_cmd)) + return (SET_ERROR(EEXIST)); + + mutex_enter(&spa_namespace_lock); + boolean_t has_spa = spa_lookup(name) != NULL; + mutex_exit(&spa_namespace_lock); + + if (record->zi_cmd == ZINJECT_DELAY_IMPORT && has_spa) + return (SET_ERROR(EEXIST)); + if (record->zi_cmd == ZINJECT_DELAY_EXPORT && !has_spa) + return (SET_ERROR(ENOENT)); + spa = NULL; + } else { + /* + * spa_inject_ref() will add an injection reference, + * which will prevent the pool from being removed + * from the namespace while still allowing it to be + * unloaded. + */ + if ((spa = spa_inject_addref(name)) == NULL) + return (SET_ERROR(ENOENT)); + } handler = kmem_alloc(sizeof (inject_handler_t), KM_SLEEP); - - handler->zi_spa = spa; + handler->zi_spa = spa; /* note: can be NULL */ handler->zi_record = *record; if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) { handler->zi_lanes = kmem_zalloc( sizeof (*handler->zi_lanes) * handler->zi_record.zi_nlanes, KM_SLEEP); handler->zi_next_lane = 0; } else { handler->zi_lanes = NULL; handler->zi_next_lane = 0; } + if (handler->zi_spa == NULL) + handler->zi_spa_name = spa_strdup(name); + else + handler->zi_spa_name = NULL; + rw_enter(&inject_lock, RW_WRITER); /* * We can't move this increment into the conditional * above because we need to hold the RW_WRITER lock of * inject_lock, and we don't want to hold that while * allocating the handler's zi_lanes array. */ if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) { ASSERT3S(inject_delay_count, >=, 0); inject_delay_count++; ASSERT3S(inject_delay_count, >, 0); } *id = handler->zi_id = inject_next_id++; list_insert_tail(&inject_handlers, handler); atomic_inc_32(&zio_injection_enabled); rw_exit(&inject_lock); } /* * Flush the ARC, so that any attempts to read this data will end up * going to the ZIO layer. Note that this is a little overkill, but * we don't have the necessary ARC interfaces to do anything else, and * fault injection isn't a performance critical path. */ if (flags & ZINJECT_FLUSH_ARC) /* * We must use FALSE to ensure arc_flush returns, since * we're not preventing concurrent ARC insertions. */ arc_flush(NULL, FALSE); return (0); } /* * Returns the next record with an ID greater than that supplied to the * function. Used to iterate over all handlers in the system. */ int zio_inject_list_next(int *id, char *name, size_t buflen, zinject_record_t *record) { inject_handler_t *handler; int ret; mutex_enter(&spa_namespace_lock); rw_enter(&inject_lock, RW_READER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) if (handler->zi_id > *id) break; if (handler) { *record = handler->zi_record; *id = handler->zi_id; - (void) strlcpy(name, spa_name(handler->zi_spa), buflen); + ASSERT(handler->zi_spa || handler->zi_spa_name); + if (handler->zi_spa != NULL) + (void) strlcpy(name, spa_name(handler->zi_spa), buflen); + else + (void) strlcpy(name, handler->zi_spa_name, buflen); ret = 0; } else { ret = SET_ERROR(ENOENT); } rw_exit(&inject_lock); mutex_exit(&spa_namespace_lock); return (ret); } /* * Clear the fault handler with the given identifier, or return ENOENT if none * exists. */ int zio_clear_fault(int id) { inject_handler_t *handler; rw_enter(&inject_lock, RW_WRITER); for (handler = list_head(&inject_handlers); handler != NULL; handler = list_next(&inject_handlers, handler)) if (handler->zi_id == id) break; if (handler == NULL) { rw_exit(&inject_lock); return (SET_ERROR(ENOENT)); } if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) { ASSERT3S(inject_delay_count, >, 0); inject_delay_count--; ASSERT3S(inject_delay_count, >=, 0); } list_remove(&inject_handlers, handler); rw_exit(&inject_lock); if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) { ASSERT3P(handler->zi_lanes, !=, NULL); kmem_free(handler->zi_lanes, sizeof (*handler->zi_lanes) * handler->zi_record.zi_nlanes); } else { ASSERT3P(handler->zi_lanes, ==, NULL); } - spa_inject_delref(handler->zi_spa); + if (handler->zi_spa_name != NULL) + spa_strfree(handler->zi_spa_name); + + if (handler->zi_spa != NULL) + spa_inject_delref(handler->zi_spa); kmem_free(handler, sizeof (inject_handler_t)); atomic_dec_32(&zio_injection_enabled); return (0); } void zio_inject_init(void) { rw_init(&inject_lock, NULL, RW_DEFAULT, NULL); mutex_init(&inject_delay_mtx, NULL, MUTEX_DEFAULT, NULL); list_create(&inject_handlers, sizeof (inject_handler_t), offsetof(inject_handler_t, zi_link)); } void zio_inject_fini(void) { list_destroy(&inject_handlers); mutex_destroy(&inject_delay_mtx); rw_destroy(&inject_lock); } #if defined(_KERNEL) EXPORT_SYMBOL(zio_injection_enabled); EXPORT_SYMBOL(zio_inject_fault); EXPORT_SYMBOL(zio_inject_list_next); EXPORT_SYMBOL(zio_clear_fault); EXPORT_SYMBOL(zio_handle_fault_injection); EXPORT_SYMBOL(zio_handle_device_injection); EXPORT_SYMBOL(zio_handle_label_injection); #endif diff --git a/sys/contrib/openzfs/rpm/redhat/zfs-kmod.spec.in b/sys/contrib/openzfs/rpm/redhat/zfs-kmod.spec.in index 9c836786baea..876c198c64de 100644 --- a/sys/contrib/openzfs/rpm/redhat/zfs-kmod.spec.in +++ b/sys/contrib/openzfs/rpm/redhat/zfs-kmod.spec.in @@ -1,112 +1,112 @@ %bcond_with debug %bcond_with debuginfo Name: @PACKAGE@-kmod Version: @VERSION@ Release: @RELEASE@%{?dist} Summary: Kernel module(s) Group: System Environment/Kernel License: @ZFS_META_LICENSE@ URL: https://github.com/openzfs/zfs BuildRequires: %kernel_module_package_buildreqs Source0: @PACKAGE@-%{version}.tar.gz BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-root-%(%{__id_u} -n) # Additional dependency information for the kmod sub-package must be specified # by generating a preamble text file which kmodtool can append to the spec file. %(/bin/echo -e "\ Requires: @PACKAGE@ = %{version}\n\ -Conflicts: @PACKAGE@-dkms) +Conflicts: @PACKAGE@-dkms" > %{_sourcedir}/kmod-preamble) # LDFLAGS are not sanitized by arch/*/Makefile for these architectures. %ifarch ppc ppc64 ppc64le aarch64 %global __global_ldflags %{nil} %endif %description This package contains the ZFS kernel modules. %define kmod_name @PACKAGE@ %kernel_module_package -n %{kmod_name} -p %{_sourcedir}/kmod-preamble %define ksrc %{_usrsrc}/kernels/%{kverrel} %define kobj %{ksrc} %package -n kmod-%{kmod_name}-devel Summary: ZFS kernel module(s) devel common Group: System Environment/Kernel %description -n kmod-%{kmod_name}-devel This package provides the header files and objects to build kernel modules. %prep if ! [ -d "%{ksrc}" ]; then echo "Kernel build directory isn't set properly, cannot continue" exit 1 fi %if %{with debug} %define debug --enable-debug %else %define debug --disable-debug %endif %if %{with debuginfo} %define debuginfo --enable-debuginfo %else %define debuginfo --disable-debuginfo %endif %setup -n %{kmod_name}-%{version} %build %configure \ --with-config=kernel \ --with-linux=%{ksrc} \ --with-linux-obj=%{kobj} \ %{debug} \ %{debuginfo} \ %{?kernel_cc} \ %{?kernel_ld} \ %{?kernel_llvm} make %{?_smp_mflags} # Module signing (modsign) # # This must be run _after_ find-debuginfo.sh runs, otherwise that will strip # the signature off of the modules. # (Based on Fedora's kernel.spec workaround) %define __modsign_install_post \ sign_pem="%{ksrc}/certs/signing_key.pem"; \ sign_x509="%{ksrc}/certs/signing_key.x509"; \ if [ -f "${sign_x509}" ]\ then \ echo "Signing kernel modules ..."; \ for kmod in $(find %{buildroot}/lib/modules/%{kverrel}/extra/ -name \*.ko); do \ %{ksrc}/scripts/sign-file sha256 ${sign_pem} ${sign_x509} ${kmod}; \ done \ fi \ %{nil} # hack to ensure signing happens after find-debuginfo.sh runs %define __spec_install_post \ %{?__debug_package:%{__debug_install_post}}\ %{__arch_install_post}\ %{__os_install_post}\ %{__modsign_install_post} %install make install \ DESTDIR=${RPM_BUILD_ROOT} \ INSTALL_MOD_DIR=extra/%{kmod_name} %{__rm} -f %{buildroot}/lib/modules/%{kverrel}/modules.* # find-debuginfo.sh only considers executables %{__chmod} u+x %{buildroot}/lib/modules/%{kverrel}/extra/*/* %clean rm -rf $RPM_BUILD_ROOT %files -n kmod-%{kmod_name}-devel %{_usrsrc}/%{kmod_name}-%{version} diff --git a/sys/contrib/openzfs/scripts/zfs-tests.sh b/sys/contrib/openzfs/scripts/zfs-tests.sh index b5b3e4ab351f..c25903ea1bee 100755 --- a/sys/contrib/openzfs/scripts/zfs-tests.sh +++ b/sys/contrib/openzfs/scripts/zfs-tests.sh @@ -1,758 +1,766 @@ #!/bin/sh # shellcheck disable=SC2154 # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License, Version 1.0 only # (the "License"). You may not use this file except in compliance # with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2020 OmniOS Community Edition (OmniOSce) Association. # SCRIPT_COMMON=${SCRIPT_COMMON:-${0%/*}/common.sh} . "${SCRIPT_COMMON}" || exit PROG=zfs-tests.sh VERBOSE="no" QUIET="" +DEBUG="" CLEANUP="yes" CLEANUPALL="no" KMSG="" LOOPBACK="yes" STACK_TRACER="no" FILESIZE="4G" DEFAULT_RUNFILES="common.run,$(uname | tr '[:upper:]' '[:lower:]').run" RUNFILES=${RUNFILES:-$DEFAULT_RUNFILES} FILEDIR=${FILEDIR:-/var/tmp} DISKS=${DISKS:-""} SINGLETEST="" SINGLETESTUSER="root" TAGS="" ITERATIONS=1 ZFS_DBGMSG="$STF_SUITE/callbacks/zfs_dbgmsg.ksh" ZFS_DMESG="$STF_SUITE/callbacks/zfs_dmesg.ksh" UNAME=$(uname) RERUN="" KMEMLEAK="" # Override some defaults if on FreeBSD if [ "$UNAME" = "FreeBSD" ] ; then TESTFAIL_CALLBACKS=${TESTFAIL_CALLBACKS:-"$ZFS_DMESG"} LOSETUP=/sbin/mdconfig DMSETUP=/sbin/gpart else ZFS_MMP="$STF_SUITE/callbacks/zfs_mmp.ksh" TESTFAIL_CALLBACKS=${TESTFAIL_CALLBACKS:-"$ZFS_DBGMSG:$ZFS_DMESG:$ZFS_MMP"} LOSETUP=${LOSETUP:-/sbin/losetup} DMSETUP=${DMSETUP:-/sbin/dmsetup} fi # # Log an informational message when additional verbosity is enabled. # msg() { if [ "$VERBOSE" = "yes" ]; then echo "$@" fi } # # Log a failure message, cleanup, and return an error. # fail() { echo "$PROG: $1" >&2 cleanup exit 1 } cleanup_freebsd_loopback() { for TEST_LOOPBACK in ${LOOPBACKS}; do if [ -c "/dev/${TEST_LOOPBACK}" ]; then sudo "${LOSETUP}" -d -u "${TEST_LOOPBACK}" || echo "Failed to destroy: ${TEST_LOOPBACK}" fi done } cleanup_linux_loopback() { for TEST_LOOPBACK in ${LOOPBACKS}; do LOOP_DEV="${TEST_LOOPBACK##*/}" DM_DEV=$(sudo "${DMSETUP}" ls 2>/dev/null | \ awk -v l="${LOOP_DEV}" '$0 ~ l {print $1}') if [ -n "$DM_DEV" ]; then sudo "${DMSETUP}" remove "${DM_DEV}" || echo "Failed to remove: ${DM_DEV}" fi if [ -n "${TEST_LOOPBACK}" ]; then sudo "${LOSETUP}" -d "${TEST_LOOPBACK}" || echo "Failed to remove: ${TEST_LOOPBACK}" fi done } # # Attempt to remove loopback devices and files which where created earlier # by this script to run the test framework. The '-k' option may be passed # to the script to suppress cleanup for debugging purposes. # cleanup() { if [ "$CLEANUP" = "no" ]; then return 0 fi if [ "$LOOPBACK" = "yes" ]; then if [ "$UNAME" = "FreeBSD" ] ; then cleanup_freebsd_loopback else cleanup_linux_loopback fi fi # shellcheck disable=SC2086 rm -f ${FILES} >/dev/null 2>&1 if [ "$STF_PATH_REMOVE" = "yes" ] && [ -d "$STF_PATH" ]; then rm -Rf "$STF_PATH" fi } trap cleanup EXIT # # Attempt to remove all testpools (testpool.XXX), unopened dm devices, # loopback devices, and files. This is a useful way to cleanup a previous # test run failure which has left the system in an unknown state. This can # be dangerous and should only be used in a dedicated test environment. # cleanup_all() { TEST_POOLS=$(ASAN_OPTIONS=detect_leaks=false "$ZPOOL" list -Ho name | grep testpool) if [ "$UNAME" = "FreeBSD" ] ; then TEST_LOOPBACKS=$(sudo "${LOSETUP}" -l) else TEST_LOOPBACKS=$("${LOSETUP}" -a | awk -F: '/file-vdev/ {print $1}') fi TEST_FILES=$(ls "${FILEDIR}"/file-vdev* /var/tmp/file-vdev* 2>/dev/null) msg msg "--- Cleanup ---" # shellcheck disable=2116,2086 msg "Removing pool(s): $(echo ${TEST_POOLS})" for TEST_POOL in $TEST_POOLS; do sudo env ASAN_OPTIONS=detect_leaks=false "$ZPOOL" destroy "${TEST_POOL}" done if [ "$UNAME" != "FreeBSD" ] ; then msg "Removing all dm(s): $(sudo "${DMSETUP}" ls | grep loop | tr '\n' ' ')" sudo "${DMSETUP}" remove_all fi # shellcheck disable=2116,2086 msg "Removing loopback(s): $(echo ${TEST_LOOPBACKS})" for TEST_LOOPBACK in $TEST_LOOPBACKS; do if [ "$UNAME" = "FreeBSD" ] ; then sudo "${LOSETUP}" -d -u "${TEST_LOOPBACK}" else sudo "${LOSETUP}" -d "${TEST_LOOPBACK}" fi done # shellcheck disable=2116,2086 msg "Removing files(s): $(echo ${TEST_FILES})" # shellcheck disable=2086 sudo rm -f ${TEST_FILES} } # # Takes a name as the only arguments and looks for the following variations # on that name. If one is found it is returned. # # $RUNFILE_DIR/ # $RUNFILE_DIR/.run # # .run # find_runfile() { NAME=$1 if [ -f "$RUNFILE_DIR/$NAME" ]; then echo "$RUNFILE_DIR/$NAME" elif [ -f "$RUNFILE_DIR/$NAME.run" ]; then echo "$RUNFILE_DIR/$NAME.run" elif [ -f "$NAME" ]; then echo "$NAME" elif [ -f "$NAME.run" ]; then echo "$NAME.run" else return 1 fi } # # Symlink file if it appears under any of the given paths. # create_links() { dir_list="$1" file_list="$2" [ -n "$STF_PATH" ] || fail "STF_PATH wasn't correctly set" for i in $file_list; do for j in $dir_list; do [ ! -e "$STF_PATH/$i" ] || continue if [ ! -d "$j/$i" ] && [ -e "$j/$i" ]; then ln -sf "$j/$i" "$STF_PATH/$i" || \ fail "Couldn't link $i" break fi done [ ! -e "$STF_PATH/$i" ] && \ STF_MISSING_BIN="$STF_MISSING_BIN $i" done STF_MISSING_BIN=${STF_MISSING_BIN# } } # # Constrain the path to limit the available binaries to a known set. # When running in-tree a top level ./bin/ directory is created for # convenience, otherwise a temporary directory is used. # constrain_path() { . "$STF_SUITE/include/commands.cfg" # On FreeBSD, base system zfs utils are in /sbin and OpenZFS utils # install to /usr/local/sbin. To avoid testing the wrong utils we # need /usr/local to come before / in the path search order. SYSTEM_DIRS="/usr/local/bin /usr/local/sbin" SYSTEM_DIRS="$SYSTEM_DIRS /usr/bin /usr/sbin /bin /sbin $LIBEXEC_DIR" if [ "$INTREE" = "yes" ]; then # Constrained path set to $(top_builddir)/tests/zfs-tests/bin STF_PATH="$BIN_DIR" STF_PATH_REMOVE="no" STF_MISSING_BIN="" if [ ! -d "$STF_PATH" ]; then mkdir "$STF_PATH" chmod 755 "$STF_PATH" || fail "Couldn't chmod $STF_PATH" fi # Special case links for standard zfs utilities create_links "$CMD_DIR" "$ZFS_FILES" # Special case links for zfs test suite utilities create_links "$CMD_DIR/tests/zfs-tests/cmd" "$ZFSTEST_FILES" else # Constrained path set to /var/tmp/constrained_path.* SYSTEMDIR=${SYSTEMDIR:-/var/tmp/constrained_path.XXXXXX} STF_PATH=$(mktemp -d "$SYSTEMDIR") STF_PATH_REMOVE="yes" STF_MISSING_BIN="" chmod 755 "$STF_PATH" || fail "Couldn't chmod $STF_PATH" # Special case links for standard zfs utilities create_links "$SYSTEM_DIRS" "$ZFS_FILES" # Special case links for zfs test suite utilities create_links "$STF_SUITE/bin" "$ZFSTEST_FILES" fi # Standard system utilities SYSTEM_FILES="$SYSTEM_FILES_COMMON" if [ "$UNAME" = "FreeBSD" ] ; then SYSTEM_FILES="$SYSTEM_FILES $SYSTEM_FILES_FREEBSD" else SYSTEM_FILES="$SYSTEM_FILES $SYSTEM_FILES_LINUX" fi create_links "$SYSTEM_DIRS" "$SYSTEM_FILES" # Exceptions if [ "$UNAME" = "Linux" ] ; then ln -fs /sbin/fsck.ext4 "$STF_PATH/fsck" ln -fs /sbin/mkfs.ext4 "$STF_PATH/newfs" ln -fs "$STF_PATH/gzip" "$STF_PATH/compress" ln -fs "$STF_PATH/gunzip" "$STF_PATH/uncompress" elif [ "$UNAME" = "FreeBSD" ] ; then ln -fs /usr/local/bin/ksh93 "$STF_PATH/ksh" fi } # # Output a useful usage message. # usage() { cat << EOF USAGE: $0 [-hvqxkfS] [-s SIZE] [-r RUNFILES] [-t PATH] [-u USER] DESCRIPTION: ZFS Test Suite launch script OPTIONS: -h Show this message -v Verbose zfs-tests.sh output -q Quiet test-runner output + -D Debug; show all test output immediately (noisy) -x Remove all testpools, dm, lo, and files (unsafe) -k Disable cleanup after test failure -K Log test names to /dev/kmsg -f Use files only, disables block device tests -S Enable stack tracer (negative performance impact) -c Only create and populate constrained path -R Automatically rerun failing tests -m Enable kmemleak reporting (Linux only) -n NFSFILE Use the nfsfile to determine the NFS configuration -I NUM Number of iterations -d DIR Use world-writable DIR for files and loopback devices -s SIZE Use vdevs of SIZE (default: 4G) -r RUNFILES Run tests in RUNFILES (default: ${DEFAULT_RUNFILES}) -t PATH|NAME Run single test at PATH relative to test suite, or search for test by NAME -T TAGS Comma separated list of tags (default: 'functional') -u USER Run single test as USER (default: root) EXAMPLES: # Run the default ($(echo "${DEFAULT_RUNFILES}" | sed 's/\.run//')) suite of tests and output the configuration used. $0 -v # Run a smaller suite of tests designed to run more quickly. $0 -r linux-fast # Run a single test $0 -t tests/functional/cli_root/zfs_bookmark/zfs_bookmark_cliargs.ksh # Run a single test by name $0 -t zfs_bookmark_cliargs # Cleanup a previous run of the test suite prior to testing, run the # default ($(echo "${DEFAULT_RUNFILES}" | sed 's/\.run//')) suite of tests and perform no cleanup on exit. $0 -x EOF } -while getopts 'hvqxkKfScRmn:d:s:r:?t:T:u:I:' OPTION; do +while getopts 'hvqxkKfScRmn:d:Ds:r:?t:T:u:I:' OPTION; do case $OPTION in h) usage exit 1 ;; v) VERBOSE="yes" ;; q) QUIET="yes" ;; x) CLEANUPALL="yes" ;; k) CLEANUP="no" ;; K) KMSG="yes" ;; f) LOOPBACK="no" ;; S) STACK_TRACER="yes" ;; c) constrain_path exit ;; R) RERUN="yes" ;; m) KMEMLEAK="yes" ;; n) nfsfile=$OPTARG [ -f "$nfsfile" ] || fail "Cannot read file: $nfsfile" export NFS=1 . "$nfsfile" ;; d) FILEDIR="$OPTARG" ;; + D) + DEBUG="yes" + ;; I) ITERATIONS="$OPTARG" if [ "$ITERATIONS" -le 0 ]; then fail "Iterations must be greater than 0." fi ;; s) FILESIZE="$OPTARG" ;; r) RUNFILES="$OPTARG" ;; t) if [ -n "$SINGLETEST" ]; then fail "-t can only be provided once." fi SINGLETEST="$OPTARG" ;; T) TAGS="$OPTARG" ;; u) SINGLETESTUSER="$OPTARG" ;; ?) usage exit ;; *) ;; esac done shift $((OPTIND-1)) FILES=${FILES:-"$FILEDIR/file-vdev0 $FILEDIR/file-vdev1 $FILEDIR/file-vdev2"} LOOPBACKS=${LOOPBACKS:-""} if [ -n "$SINGLETEST" ]; then if [ -n "$TAGS" ]; then fail "-t and -T are mutually exclusive." fi RUNFILE_DIR="/var/tmp" RUNFILES="zfs-tests.$$.run" [ -n "$QUIET" ] && SINGLEQUIET="True" || SINGLEQUIET="False" cat >"${RUNFILE_DIR}/${RUNFILES}" << EOF [DEFAULT] pre = quiet = $SINGLEQUIET pre_user = root user = $SINGLETESTUSER timeout = 600 post_user = root post = outputdir = /var/tmp/test_results EOF if [ "$SINGLETEST" = "${SINGLETEST%/*}" ] ; then NEWSINGLETEST=$(find "$STF_SUITE" -name "$SINGLETEST*" -print -quit) if [ -z "$NEWSINGLETEST" ] ; then fail "couldn't find test matching '$SINGLETEST'" fi SINGLETEST=$NEWSINGLETEST fi SINGLETESTDIR="${SINGLETEST%/*}" SETUPDIR="$SINGLETESTDIR" [ "${SETUPDIR#/}" = "$SETUPDIR" ] && SETUPDIR="$STF_SUITE/$SINGLETESTDIR" [ -x "$SETUPDIR/setup.ksh" ] && SETUPSCRIPT="setup" || SETUPSCRIPT= [ -x "$SETUPDIR/cleanup.ksh" ] && CLEANUPSCRIPT="cleanup" || CLEANUPSCRIPT= SINGLETESTFILE="${SINGLETEST##*/}" cat >>"${RUNFILE_DIR}/${RUNFILES}" << EOF [$SINGLETESTDIR] tests = ['$SINGLETESTFILE'] pre = $SETUPSCRIPT post = $CLEANUPSCRIPT tags = ['functional'] EOF fi # # Use default tag if none was specified # TAGS=${TAGS:='functional'} # # Attempt to locate the runfiles describing the test workload. # R="" IFS=, for RUNFILE in $RUNFILES; do if [ -n "$RUNFILE" ]; then SAVED_RUNFILE="$RUNFILE" RUNFILE=$(find_runfile "$RUNFILE") || fail "Cannot find runfile: $SAVED_RUNFILE" R="$R,$RUNFILE" fi if [ ! -r "$RUNFILE" ]; then fail "Cannot read runfile: $RUNFILE" fi done unset IFS RUNFILES=${R#,} # # This script should not be run as root. Instead the test user, which may # be a normal user account, needs to be configured such that it can # run commands via sudo passwordlessly. # if [ "$(id -u)" = "0" ]; then fail "This script must not be run as root." fi if [ "$(sudo id -un)" != "root" ]; then fail "Passwordless sudo access required." fi # # Constrain the available binaries to a known set. # constrain_path # # Check if ksh exists # if [ "$UNAME" = "FreeBSD" ]; then sudo ln -fs /usr/local/bin/ksh93 /bin/ksh fi [ -e "$STF_PATH/ksh" ] || fail "This test suite requires ksh." [ -e "$STF_SUITE/include/default.cfg" ] || fail \ "Missing $STF_SUITE/include/default.cfg file." # # Verify the ZFS module stack is loaded. # if [ "$STACK_TRACER" = "yes" ]; then sudo "${ZFS_SH}" -S >/dev/null 2>&1 else sudo "${ZFS_SH}" >/dev/null 2>&1 fi # # Attempt to cleanup all previous state for a new test run. # if [ "$CLEANUPALL" = "yes" ]; then cleanup_all fi # # By default preserve any existing pools # if [ -z "${KEEP}" ]; then KEEP="$(ASAN_OPTIONS=detect_leaks=false "$ZPOOL" list -Ho name | tr -s '[:space:]' ' ')" if [ -z "${KEEP}" ]; then KEEP="rpool" fi else KEEP="$(echo "$KEEP" | tr -s '[:space:]' ' ')" fi # # NOTE: The following environment variables are undocumented # and should be used for testing purposes only: # # __ZFS_POOL_EXCLUDE - don't iterate over the pools it lists # __ZFS_POOL_RESTRICT - iterate only over the pools it lists # # See libzfs/libzfs_config.c for more information. # __ZFS_POOL_EXCLUDE="$KEEP" . "$STF_SUITE/include/default.cfg" # # No DISKS have been provided so a basic file or loopback based devices # must be created for the test suite to use. # if [ -z "${DISKS}" ]; then # # If this is a performance run, prevent accidental use of # loopback devices. # [ "$TAGS" = "perf" ] && fail "Running perf tests without disks." # # Create sparse files for the test suite. These may be used # directory or have loopback devices layered on them. # for TEST_FILE in ${FILES}; do [ -f "$TEST_FILE" ] && fail "Failed file exists: ${TEST_FILE}" truncate -s "${FILESIZE}" "${TEST_FILE}" || fail "Failed creating: ${TEST_FILE} ($?)" done # # If requested setup loopback devices backed by the sparse files. # if [ "$LOOPBACK" = "yes" ]; then test -x "$LOSETUP" || fail "$LOSETUP utility must be installed" for TEST_FILE in ${FILES}; do if [ "$UNAME" = "FreeBSD" ] ; then MDDEVICE=$(sudo "${LOSETUP}" -a -t vnode -f "${TEST_FILE}") if [ -z "$MDDEVICE" ] ; then fail "Failed: ${TEST_FILE} -> loopback" fi DISKS="$DISKS $MDDEVICE" LOOPBACKS="$LOOPBACKS $MDDEVICE" else TEST_LOOPBACK=$(sudo "${LOSETUP}" --show -f "${TEST_FILE}") || fail "Failed: ${TEST_FILE} -> ${TEST_LOOPBACK}" BASELOOPBACK="${TEST_LOOPBACK##*/}" DISKS="$DISKS $BASELOOPBACK" LOOPBACKS="$LOOPBACKS $TEST_LOOPBACK" fi done DISKS=${DISKS# } LOOPBACKS=${LOOPBACKS# } else DISKS="$FILES" fi fi # # It may be desirable to test with fewer disks than the default when running # the performance tests, but the functional tests require at least three. # NUM_DISKS=$(echo "${DISKS}" | awk '{print NF}') if [ "$TAGS" != "perf" ]; then [ "$NUM_DISKS" -lt 3 ] && fail "Not enough disks ($NUM_DISKS/3 minimum)" fi # # Disable SELinux until the ZFS Test Suite has been updated accordingly. # if command -v setenforce >/dev/null; then sudo setenforce permissive >/dev/null 2>&1 fi # # Enable internal ZFS debug log and clear it. # if [ -e /sys/module/zfs/parameters/zfs_dbgmsg_enable ]; then sudo sh -c "echo 1 >/sys/module/zfs/parameters/zfs_dbgmsg_enable" sudo sh -c "echo 0 >/proc/spl/kstat/zfs/dbgmsg" fi msg msg "--- Configuration ---" msg "Runfiles: $RUNFILES" msg "STF_TOOLS: $STF_TOOLS" msg "STF_SUITE: $STF_SUITE" msg "STF_PATH: $STF_PATH" msg "FILEDIR: $FILEDIR" msg "FILES: $FILES" msg "LOOPBACKS: $LOOPBACKS" msg "DISKS: $DISKS" msg "NUM_DISKS: $NUM_DISKS" msg "FILESIZE: $FILESIZE" msg "ITERATIONS: $ITERATIONS" msg "TAGS: $TAGS" msg "STACK_TRACER: $STACK_TRACER" msg "Keep pool(s): $KEEP" msg "Missing util(s): $STF_MISSING_BIN" msg "" export STF_TOOLS export STF_SUITE export STF_PATH export DISKS export FILEDIR export KEEP export __ZFS_POOL_EXCLUDE export TESTFAIL_CALLBACKS mktemp_file() { if [ "$UNAME" = "FreeBSD" ]; then mktemp -u "${FILEDIR}/$1.XXXXXX" else mktemp -ut "$1.XXXXXX" -p "$FILEDIR" fi } mkdir -p "$FILEDIR" || : RESULTS_FILE=$(mktemp_file zts-results) REPORT_FILE=$(mktemp_file zts-report) # # Run all the tests as specified. # msg "${TEST_RUNNER}" \ "${QUIET:+-q}" \ + "${DEBUG:+-D}" \ "${KMEMLEAK:+-m}" \ "${KMSG:+-K}" \ "-c \"${RUNFILES}\"" \ "-T \"${TAGS}\"" \ "-i \"${STF_SUITE}\"" \ "-I \"${ITERATIONS}\"" { PATH=$STF_PATH \ ${TEST_RUNNER} \ ${QUIET:+-q} \ + ${DEBUG:+-D} \ ${KMEMLEAK:+-m} \ ${KMSG:+-K} \ -c "${RUNFILES}" \ -T "${TAGS}" \ -i "${STF_SUITE}" \ -I "${ITERATIONS}" \ 2>&1; echo $? >"$REPORT_FILE"; } | tee "$RESULTS_FILE" read -r RUNRESULT <"$REPORT_FILE" # # Analyze the results. # ${ZTS_REPORT} ${RERUN:+--no-maybes} "$RESULTS_FILE" >"$REPORT_FILE" RESULT=$? if [ "$RESULT" -eq "2" ] && [ -n "$RERUN" ]; then MAYBES="$($ZTS_REPORT --list-maybes)" TEMP_RESULTS_FILE=$(mktemp_file zts-results-tmp) TEST_LIST=$(mktemp_file test-list) grep "^Test:.*\[FAIL\]" "$RESULTS_FILE" >"$TEMP_RESULTS_FILE" for test_name in $MAYBES; do grep "$test_name " "$TEMP_RESULTS_FILE" >>"$TEST_LIST" done { PATH=$STF_PATH \ ${TEST_RUNNER} \ ${QUIET:+-q} \ + ${DEBUG:+-D} \ ${KMEMLEAK:+-m} \ -c "${RUNFILES}" \ -T "${TAGS}" \ -i "${STF_SUITE}" \ -I "${ITERATIONS}" \ -l "${TEST_LIST}" \ 2>&1; echo $? >"$REPORT_FILE"; } | tee "$RESULTS_FILE" read -r RUNRESULT <"$REPORT_FILE" # # Analyze the results. # ${ZTS_REPORT} --no-maybes "$RESULTS_FILE" >"$REPORT_FILE" RESULT=$? fi cat "$REPORT_FILE" RESULTS_DIR=$(awk '/^Log directory/ { print $3 }' "$RESULTS_FILE") if [ -d "$RESULTS_DIR" ]; then cat "$RESULTS_FILE" "$REPORT_FILE" >"$RESULTS_DIR/results" fi rm -f "$RESULTS_FILE" "$REPORT_FILE" "$TEST_LIST" "$TEMP_RESULTS_FILE" if [ -n "$SINGLETEST" ]; then rm -f "$RUNFILES" >/dev/null 2>&1 fi [ "$RUNRESULT" -gt 3 ] && exit "$RUNRESULT" || exit "$RESULT" diff --git a/sys/contrib/openzfs/tests/runfiles/common.run b/sys/contrib/openzfs/tests/runfiles/common.run index 558cd425afd8..0586d991b802 100644 --- a/sys/contrib/openzfs/tests/runfiles/common.run +++ b/sys/contrib/openzfs/tests/runfiles/common.run @@ -1,1035 +1,1036 @@ # # This file and its contents are supplied under the terms of the # Common Development and Distribution License ("CDDL"), version 1.0. # You may only use this file in accordance with the terms of version # 1.0 of the CDDL. # # A full copy of the text of the CDDL should have accompanied this # source. A copy of the CDDL is also available via the Internet at # http://www.illumos.org/license/CDDL. # # This run file contains all of the common functional tests. When # adding a new test consider also adding it to the sanity.run file # if the new test runs to completion in only a few seconds. # # Approximate run time: 4-5 hours # [DEFAULT] pre = setup quiet = False pre_user = root user = root timeout = 600 post_user = root post = cleanup failsafe_user = root failsafe = callbacks/zfs_failsafe outputdir = /var/tmp/test_results tags = ['functional'] [tests/functional/acl/off] tests = ['dosmode', 'posixmode'] tags = ['functional', 'acl'] [tests/functional/alloc_class] tests = ['alloc_class_001_pos', 'alloc_class_002_neg', 'alloc_class_003_pos', 'alloc_class_004_pos', 'alloc_class_005_pos', 'alloc_class_006_pos', 'alloc_class_007_pos', 'alloc_class_008_pos', 'alloc_class_009_pos', 'alloc_class_010_pos', 'alloc_class_011_neg', 'alloc_class_012_pos', 'alloc_class_013_pos', 'alloc_class_014_neg', 'alloc_class_015_pos'] tags = ['functional', 'alloc_class'] [tests/functional/append] tests = ['file_append', 'threadsappend_001_pos'] tags = ['functional', 'append'] [tests/functional/arc] tests = ['dbufstats_001_pos', 'dbufstats_002_pos', 'dbufstats_003_pos', 'arcstats_runtime_tuning'] tags = ['functional', 'arc'] [tests/functional/atime] tests = ['atime_001_pos', 'atime_002_neg', 'root_atime_off', 'root_atime_on'] tags = ['functional', 'atime'] [tests/functional/bclone] tests = ['bclone_crossfs_corner_cases_limited', 'bclone_crossfs_data', 'bclone_crossfs_embedded', 'bclone_crossfs_hole', 'bclone_diffprops_all', 'bclone_diffprops_checksum', 'bclone_diffprops_compress', 'bclone_diffprops_copies', 'bclone_diffprops_recordsize', 'bclone_prop_sync', 'bclone_samefs_corner_cases_limited', 'bclone_samefs_data', 'bclone_samefs_embedded', 'bclone_samefs_hole'] tags = ['functional', 'bclone'] timeout = 7200 [tests/functional/block_cloning] tests = ['block_cloning_clone_mmap_cached', 'block_cloning_copyfilerange', 'block_cloning_copyfilerange_partial', 'block_cloning_copyfilerange_fallback', 'block_cloning_disabled_copyfilerange', 'block_cloning_copyfilerange_cross_dataset', 'block_cloning_cross_enc_dataset', 'block_cloning_copyfilerange_fallback_same_txg', 'block_cloning_replay', 'block_cloning_replay_encrypted', 'block_cloning_lwb_buffer_overflow', 'block_cloning_clone_mmap_write'] tags = ['functional', 'block_cloning'] [tests/functional/bootfs] tests = ['bootfs_001_pos', 'bootfs_002_neg', 'bootfs_003_pos', 'bootfs_004_neg', 'bootfs_005_neg', 'bootfs_006_pos', 'bootfs_007_pos', 'bootfs_008_pos'] tags = ['functional', 'bootfs'] [tests/functional/btree] tests = ['btree_positive', 'btree_negative'] tags = ['functional', 'btree'] pre = post = [tests/functional/cache] tests = ['cache_001_pos', 'cache_002_pos', 'cache_003_pos', 'cache_004_neg', 'cache_005_neg', 'cache_006_pos', 'cache_007_neg', 'cache_008_neg', 'cache_009_pos', 'cache_010_pos', 'cache_011_pos', 'cache_012_pos'] tags = ['functional', 'cache'] [tests/functional/cachefile] tests = ['cachefile_001_pos', 'cachefile_002_pos', 'cachefile_003_pos', 'cachefile_004_pos'] tags = ['functional', 'cachefile'] [tests/functional/casenorm] tests = ['case_all_values', 'norm_all_values', 'mixed_create_failure', 'sensitive_none_lookup', 'sensitive_none_delete', 'sensitive_formd_lookup', 'sensitive_formd_delete', 'insensitive_none_lookup', 'insensitive_none_delete', 'insensitive_formd_lookup', 'insensitive_formd_delete', 'mixed_none_lookup', 'mixed_none_lookup_ci', 'mixed_none_delete', 'mixed_formd_lookup', 'mixed_formd_lookup_ci', 'mixed_formd_delete'] tags = ['functional', 'casenorm'] [tests/functional/channel_program/lua_core] tests = ['tst.args_to_lua', 'tst.divide_by_zero', 'tst.exists', 'tst.integer_illegal', 'tst.integer_overflow', 'tst.language_functions_neg', 'tst.language_functions_pos', 'tst.large_prog', 'tst.libraries', 'tst.memory_limit', 'tst.nested_neg', 'tst.nested_pos', 'tst.nvlist_to_lua', 'tst.recursive_neg', 'tst.recursive_pos', 'tst.return_large', 'tst.return_nvlist_neg', 'tst.return_nvlist_pos', 'tst.return_recursive_table', 'tst.stack_gsub', 'tst.timeout'] tags = ['functional', 'channel_program', 'lua_core'] [tests/functional/channel_program/synctask_core] tests = ['tst.destroy_fs', 'tst.destroy_snap', 'tst.get_count_and_limit', 'tst.get_index_props', 'tst.get_mountpoint', 'tst.get_neg', 'tst.get_number_props', 'tst.get_string_props', 'tst.get_type', 'tst.get_userquota', 'tst.get_written', 'tst.inherit', 'tst.list_bookmarks', 'tst.list_children', 'tst.list_clones', 'tst.list_holds', 'tst.list_snapshots', 'tst.list_system_props', 'tst.list_user_props', 'tst.parse_args_neg','tst.promote_conflict', 'tst.promote_multiple', 'tst.promote_simple', 'tst.rollback_mult', 'tst.rollback_one', 'tst.set_props', 'tst.snapshot_destroy', 'tst.snapshot_neg', 'tst.snapshot_recursive', 'tst.snapshot_rename', 'tst.snapshot_simple', 'tst.bookmark.create', 'tst.bookmark.copy', 'tst.terminate_by_signal' ] tags = ['functional', 'channel_program', 'synctask_core'] [tests/functional/checksum] tests = ['run_edonr_test', 'run_sha2_test', 'run_skein_test', 'run_blake3_test', 'filetest_001_pos', 'filetest_002_pos'] tags = ['functional', 'checksum'] [tests/functional/clean_mirror] tests = [ 'clean_mirror_001_pos', 'clean_mirror_002_pos', 'clean_mirror_003_pos', 'clean_mirror_004_pos'] tags = ['functional', 'clean_mirror'] [tests/functional/cli_root/zinject] tests = ['zinject_args'] pre = post = tags = ['functional', 'cli_root', 'zinject'] [tests/functional/cli_root/zdb] tests = ['zdb_002_pos', 'zdb_003_pos', 'zdb_004_pos', 'zdb_005_pos', 'zdb_006_pos', 'zdb_args_neg', 'zdb_args_pos', 'zdb_block_size_histogram', 'zdb_checksum', 'zdb_decompress', 'zdb_display_block', 'zdb_encrypted', 'zdb_label_checksum', 'zdb_object_range_neg', 'zdb_object_range_pos', 'zdb_objset_id', 'zdb_decompress_zstd', 'zdb_recover', 'zdb_recover_2', 'zdb_backup'] pre = post = tags = ['functional', 'cli_root', 'zdb'] timeout = 1200 [tests/functional/cli_root/zfs] tests = ['zfs_001_neg', 'zfs_002_pos'] tags = ['functional', 'cli_root', 'zfs'] [tests/functional/cli_root/zfs_bookmark] tests = ['zfs_bookmark_cliargs'] tags = ['functional', 'cli_root', 'zfs_bookmark'] [tests/functional/cli_root/zfs_change-key] tests = ['zfs_change-key', 'zfs_change-key_child', 'zfs_change-key_format', 'zfs_change-key_inherit', 'zfs_change-key_load', 'zfs_change-key_location', 'zfs_change-key_pbkdf2iters', 'zfs_change-key_clones'] tags = ['functional', 'cli_root', 'zfs_change-key'] [tests/functional/cli_root/zfs_clone] tests = ['zfs_clone_001_neg', 'zfs_clone_002_pos', 'zfs_clone_003_pos', 'zfs_clone_004_pos', 'zfs_clone_005_pos', 'zfs_clone_006_pos', 'zfs_clone_007_pos', 'zfs_clone_008_neg', 'zfs_clone_009_neg', 'zfs_clone_010_pos', 'zfs_clone_encrypted', 'zfs_clone_deeply_nested', 'zfs_clone_rm_nested'] tags = ['functional', 'cli_root', 'zfs_clone'] [tests/functional/cli_root/zfs_copies] tests = ['zfs_copies_001_pos', 'zfs_copies_002_pos', 'zfs_copies_003_pos', 'zfs_copies_004_neg', 'zfs_copies_005_neg', 'zfs_copies_006_pos'] tags = ['functional', 'cli_root', 'zfs_copies'] [tests/functional/cli_root/zfs_create] tests = ['zfs_create_001_pos', 'zfs_create_002_pos', 'zfs_create_003_pos', 'zfs_create_004_pos', 'zfs_create_005_pos', 'zfs_create_006_pos', 'zfs_create_007_pos', 'zfs_create_008_neg', 'zfs_create_009_neg', 'zfs_create_010_neg', 'zfs_create_011_pos', 'zfs_create_012_pos', 'zfs_create_013_pos', 'zfs_create_014_pos', 'zfs_create_encrypted', 'zfs_create_crypt_combos', 'zfs_create_dryrun', 'zfs_create_nomount', 'zfs_create_verbose'] tags = ['functional', 'cli_root', 'zfs_create'] [tests/functional/cli_root/zfs_destroy] tests = ['zfs_clone_livelist_condense_and_disable', 'zfs_clone_livelist_condense_races', 'zfs_clone_livelist_dedup', 'zfs_destroy_001_pos', 'zfs_destroy_002_pos', 'zfs_destroy_003_pos', 'zfs_destroy_004_pos', 'zfs_destroy_005_neg', 'zfs_destroy_006_neg', 'zfs_destroy_007_neg', 'zfs_destroy_008_pos', 'zfs_destroy_009_pos', 'zfs_destroy_010_pos', 'zfs_destroy_011_pos', 'zfs_destroy_012_pos', 'zfs_destroy_013_neg', 'zfs_destroy_014_pos', 'zfs_destroy_015_pos', 'zfs_destroy_016_pos', 'zfs_destroy_clone_livelist', 'zfs_destroy_dev_removal', 'zfs_destroy_dev_removal_condense'] tags = ['functional', 'cli_root', 'zfs_destroy'] [tests/functional/cli_root/zfs_diff] tests = ['zfs_diff_changes', 'zfs_diff_cliargs', 'zfs_diff_timestamp', 'zfs_diff_types', 'zfs_diff_encrypted', 'zfs_diff_mangle'] tags = ['functional', 'cli_root', 'zfs_diff'] [tests/functional/cli_root/zfs_get] tests = ['zfs_get_001_pos', 'zfs_get_002_pos', 'zfs_get_003_pos', 'zfs_get_004_pos', 'zfs_get_005_neg', 'zfs_get_006_neg', 'zfs_get_007_neg', 'zfs_get_008_pos', 'zfs_get_009_pos', 'zfs_get_010_neg'] tags = ['functional', 'cli_root', 'zfs_get'] [tests/functional/cli_root/zfs_ids_to_path] tests = ['zfs_ids_to_path_001_pos'] tags = ['functional', 'cli_root', 'zfs_ids_to_path'] [tests/functional/cli_root/zfs_inherit] tests = ['zfs_inherit_001_neg', 'zfs_inherit_002_neg', 'zfs_inherit_003_pos', 'zfs_inherit_mountpoint'] tags = ['functional', 'cli_root', 'zfs_inherit'] [tests/functional/cli_root/zfs_load-key] tests = ['zfs_load-key', 'zfs_load-key_all', 'zfs_load-key_file', 'zfs_load-key_https', 'zfs_load-key_location', 'zfs_load-key_noop', 'zfs_load-key_recursive'] tags = ['functional', 'cli_root', 'zfs_load-key'] [tests/functional/cli_root/zfs_mount] tests = ['zfs_mount_001_pos', 'zfs_mount_002_pos', 'zfs_mount_003_pos', 'zfs_mount_004_pos', 'zfs_mount_005_pos', 'zfs_mount_007_pos', 'zfs_mount_009_neg', 'zfs_mount_010_neg', 'zfs_mount_011_neg', 'zfs_mount_012_pos', 'zfs_mount_all_001_pos', 'zfs_mount_encrypted', 'zfs_mount_remount', 'zfs_mount_all_fail', 'zfs_mount_all_mountpoints', 'zfs_mount_test_race', 'zfs_mount_recursive'] tags = ['functional', 'cli_root', 'zfs_mount'] [tests/functional/cli_root/zfs_program] tests = ['zfs_program_json'] tags = ['functional', 'cli_root', 'zfs_program'] [tests/functional/cli_root/zfs_promote] tests = ['zfs_promote_001_pos', 'zfs_promote_002_pos', 'zfs_promote_003_pos', 'zfs_promote_004_pos', 'zfs_promote_005_pos', 'zfs_promote_006_neg', 'zfs_promote_007_neg', 'zfs_promote_008_pos', 'zfs_promote_encryptionroot'] tags = ['functional', 'cli_root', 'zfs_promote'] [tests/functional/cli_root/zfs_property] tests = ['zfs_written_property_001_pos'] tags = ['functional', 'cli_root', 'zfs_property'] [tests/functional/cli_root/zfs_receive] tests = ['zfs_receive_001_pos', 'zfs_receive_002_pos', 'zfs_receive_003_pos', 'zfs_receive_004_neg', 'zfs_receive_005_neg', 'zfs_receive_006_pos', 'zfs_receive_007_neg', 'zfs_receive_008_pos', 'zfs_receive_009_neg', 'zfs_receive_010_pos', 'zfs_receive_011_pos', 'zfs_receive_012_pos', 'zfs_receive_013_pos', 'zfs_receive_014_pos', 'zfs_receive_015_pos', 'zfs_receive_016_pos', 'receive-o-x_props_override', 'receive-o-x_props_aliases', 'zfs_receive_from_encrypted', 'zfs_receive_to_encrypted', 'zfs_receive_raw', 'zfs_receive_raw_incremental', 'zfs_receive_-e', 'zfs_receive_raw_-d', 'zfs_receive_from_zstd', 'zfs_receive_new_props', 'zfs_receive_-wR-encrypted-mix', 'zfs_receive_corrective', 'zfs_receive_compressed_corrective', 'zfs_receive_large_block_corrective'] tags = ['functional', 'cli_root', 'zfs_receive'] [tests/functional/cli_root/zfs_rename] tests = ['zfs_rename_001_pos', 'zfs_rename_002_pos', 'zfs_rename_003_pos', 'zfs_rename_004_neg', 'zfs_rename_005_neg', 'zfs_rename_006_pos', 'zfs_rename_007_pos', 'zfs_rename_008_pos', 'zfs_rename_009_neg', 'zfs_rename_010_neg', 'zfs_rename_011_pos', 'zfs_rename_012_neg', 'zfs_rename_013_pos', 'zfs_rename_014_neg', 'zfs_rename_encrypted_child', 'zfs_rename_to_encrypted', 'zfs_rename_mountpoint', 'zfs_rename_nounmount'] tags = ['functional', 'cli_root', 'zfs_rename'] [tests/functional/cli_root/zfs_reservation] tests = ['zfs_reservation_001_pos', 'zfs_reservation_002_pos'] tags = ['functional', 'cli_root', 'zfs_reservation'] [tests/functional/cli_root/zfs_rollback] tests = ['zfs_rollback_001_pos', 'zfs_rollback_002_pos', 'zfs_rollback_003_neg', 'zfs_rollback_004_neg'] tags = ['functional', 'cli_root', 'zfs_rollback'] [tests/functional/cli_root/zfs_send] tests = ['zfs_send_001_pos', 'zfs_send_002_pos', 'zfs_send_003_pos', 'zfs_send_004_neg', 'zfs_send_005_pos', 'zfs_send_006_pos', 'zfs_send_007_pos', 'zfs_send_encrypted', 'zfs_send_encrypted_unloaded', 'zfs_send_raw', 'zfs_send_sparse', 'zfs_send-b', 'zfs_send_skip_missing'] tags = ['functional', 'cli_root', 'zfs_send'] [tests/functional/cli_root/zfs_set] tests = ['cache_001_pos', 'cache_002_neg', 'canmount_001_pos', 'canmount_002_pos', 'canmount_003_pos', 'canmount_004_pos', 'checksum_001_pos', 'compression_001_pos', 'mountpoint_001_pos', 'mountpoint_002_pos', 'reservation_001_neg', 'user_property_002_pos', 'share_mount_001_neg', 'snapdir_001_pos', 'onoffs_001_pos', 'user_property_001_pos', 'user_property_003_neg', 'readonly_001_pos', 'user_property_004_pos', 'version_001_neg', 'zfs_set_001_neg', 'zfs_set_002_neg', 'zfs_set_003_neg', 'property_alias_001_pos', 'mountpoint_003_pos', 'ro_props_001_pos', 'zfs_set_keylocation', 'zfs_set_feature_activation', 'zfs_set_nomount'] tags = ['functional', 'cli_root', 'zfs_set'] [tests/functional/cli_root/zfs_share] tests = ['zfs_share_001_pos', 'zfs_share_002_pos', 'zfs_share_003_pos', 'zfs_share_004_pos', 'zfs_share_006_pos', 'zfs_share_008_neg', 'zfs_share_010_neg', 'zfs_share_011_pos', 'zfs_share_concurrent_shares', 'zfs_share_after_mount'] tags = ['functional', 'cli_root', 'zfs_share'] [tests/functional/cli_root/zfs_snapshot] tests = ['zfs_snapshot_001_neg', 'zfs_snapshot_002_neg', 'zfs_snapshot_003_neg', 'zfs_snapshot_004_neg', 'zfs_snapshot_005_neg', 'zfs_snapshot_006_pos', 'zfs_snapshot_007_neg', 'zfs_snapshot_008_neg', 'zfs_snapshot_009_pos'] tags = ['functional', 'cli_root', 'zfs_snapshot'] [tests/functional/cli_root/zfs_unload-key] tests = ['zfs_unload-key', 'zfs_unload-key_all', 'zfs_unload-key_recursive'] tags = ['functional', 'cli_root', 'zfs_unload-key'] [tests/functional/cli_root/zfs_unmount] tests = ['zfs_unmount_001_pos', 'zfs_unmount_002_pos', 'zfs_unmount_003_pos', 'zfs_unmount_004_pos', 'zfs_unmount_005_pos', 'zfs_unmount_006_pos', 'zfs_unmount_007_neg', 'zfs_unmount_008_neg', 'zfs_unmount_009_pos', 'zfs_unmount_all_001_pos', 'zfs_unmount_nested', 'zfs_unmount_unload_keys'] tags = ['functional', 'cli_root', 'zfs_unmount'] [tests/functional/cli_root/zfs_unshare] tests = ['zfs_unshare_001_pos', 'zfs_unshare_002_pos', 'zfs_unshare_003_pos', 'zfs_unshare_004_neg', 'zfs_unshare_005_neg', 'zfs_unshare_006_pos', 'zfs_unshare_007_pos'] tags = ['functional', 'cli_root', 'zfs_unshare'] [tests/functional/cli_root/zfs_upgrade] tests = ['zfs_upgrade_001_pos', 'zfs_upgrade_002_pos', 'zfs_upgrade_003_pos', 'zfs_upgrade_004_pos', 'zfs_upgrade_005_pos', 'zfs_upgrade_006_neg', 'zfs_upgrade_007_neg'] tags = ['functional', 'cli_root', 'zfs_upgrade'] [tests/functional/cli_root/zfs_wait] tests = ['zfs_wait_deleteq', 'zfs_wait_getsubopt'] tags = ['functional', 'cli_root', 'zfs_wait'] [tests/functional/cli_root/zhack] tests = ['zhack_label_repair_001', 'zhack_label_repair_002', 'zhack_label_repair_003', 'zhack_label_repair_004'] pre = post = tags = ['functional', 'cli_root', 'zhack'] [tests/functional/cli_root/zpool] tests = ['zpool_001_neg', 'zpool_002_pos', 'zpool_003_pos', 'zpool_colors'] tags = ['functional', 'cli_root', 'zpool'] [tests/functional/cli_root/zpool_add] tests = ['zpool_add_001_pos', 'zpool_add_002_pos', 'zpool_add_003_pos', 'zpool_add_004_pos', 'zpool_add_006_pos', 'zpool_add_007_neg', 'zpool_add_008_neg', 'zpool_add_009_neg', 'zpool_add_010_pos', 'add-o_ashift', 'add_prop_ashift', 'zpool_add_dryrun_output', 'zpool_add--allow-ashift-mismatch'] tags = ['functional', 'cli_root', 'zpool_add'] [tests/functional/cli_root/zpool_attach] tests = ['zpool_attach_001_neg', 'attach-o_ashift'] tags = ['functional', 'cli_root', 'zpool_attach'] [tests/functional/cli_root/zpool_clear] tests = ['zpool_clear_001_pos', 'zpool_clear_002_neg', 'zpool_clear_003_neg', 'zpool_clear_readonly'] tags = ['functional', 'cli_root', 'zpool_clear'] [tests/functional/cli_root/zpool_create] tests = ['zpool_create_001_pos', 'zpool_create_002_pos', 'zpool_create_003_pos', 'zpool_create_004_pos', 'zpool_create_005_pos', 'zpool_create_006_pos', 'zpool_create_007_neg', 'zpool_create_008_pos', 'zpool_create_009_neg', 'zpool_create_010_neg', 'zpool_create_011_neg', 'zpool_create_012_neg', 'zpool_create_014_neg', 'zpool_create_015_neg', 'zpool_create_017_neg', 'zpool_create_018_pos', 'zpool_create_019_pos', 'zpool_create_020_pos', 'zpool_create_021_pos', 'zpool_create_022_pos', 'zpool_create_023_neg', 'zpool_create_024_pos', 'zpool_create_encrypted', 'zpool_create_crypt_combos', 'zpool_create_draid_001_pos', 'zpool_create_draid_002_pos', 'zpool_create_draid_003_pos', 'zpool_create_draid_004_pos', 'zpool_create_features_001_pos', 'zpool_create_features_002_pos', 'zpool_create_features_003_pos', 'zpool_create_features_004_neg', 'zpool_create_features_005_pos', 'zpool_create_features_006_pos', 'zpool_create_features_007_pos', 'zpool_create_features_008_pos', 'zpool_create_features_009_pos', 'create-o_ashift', 'zpool_create_tempname', 'zpool_create_dryrun_output'] tags = ['functional', 'cli_root', 'zpool_create'] [tests/functional/cli_root/zpool_destroy] tests = ['zpool_destroy_001_pos', 'zpool_destroy_002_pos', 'zpool_destroy_003_neg'] pre = post = tags = ['functional', 'cli_root', 'zpool_destroy'] [tests/functional/cli_root/zpool_detach] tests = ['zpool_detach_001_neg'] tags = ['functional', 'cli_root', 'zpool_detach'] [tests/functional/cli_root/zpool_events] tests = ['zpool_events_clear', 'zpool_events_cliargs', 'zpool_events_follow', 'zpool_events_poolname', 'zpool_events_errors', 'zpool_events_duplicates', 'zpool_events_clear_retained'] tags = ['functional', 'cli_root', 'zpool_events'] [tests/functional/cli_root/zpool_export] tests = ['zpool_export_001_pos', 'zpool_export_002_pos', 'zpool_export_003_neg', 'zpool_export_004_pos'] tags = ['functional', 'cli_root', 'zpool_export'] [tests/functional/cli_root/zpool_get] tests = ['zpool_get_001_pos', 'zpool_get_002_pos', 'zpool_get_003_pos', 'zpool_get_004_neg', 'zpool_get_005_pos', 'vdev_get_001_pos'] tags = ['functional', 'cli_root', 'zpool_get'] [tests/functional/cli_root/zpool_history] tests = ['zpool_history_001_neg', 'zpool_history_002_pos'] tags = ['functional', 'cli_root', 'zpool_history'] [tests/functional/cli_root/zpool_import] tests = ['zpool_import_001_pos', 'zpool_import_002_pos', 'zpool_import_003_pos', 'zpool_import_004_pos', 'zpool_import_005_pos', 'zpool_import_006_pos', 'zpool_import_007_pos', 'zpool_import_008_pos', 'zpool_import_009_neg', 'zpool_import_010_pos', 'zpool_import_011_neg', 'zpool_import_012_pos', 'zpool_import_013_neg', 'zpool_import_014_pos', 'zpool_import_015_pos', 'zpool_import_016_pos', 'zpool_import_017_pos', 'zpool_import_features_001_pos', 'zpool_import_features_002_neg', 'zpool_import_features_003_pos', 'zpool_import_missing_001_pos', 'zpool_import_missing_002_pos', 'zpool_import_missing_003_pos', 'zpool_import_rename_001_pos', 'zpool_import_all_001_pos', 'zpool_import_encrypted', 'zpool_import_encrypted_load', 'zpool_import_errata3', 'zpool_import_errata4', 'import_cachefile_device_added', 'import_cachefile_device_removed', 'import_cachefile_device_replaced', 'import_cachefile_mirror_attached', 'import_cachefile_mirror_detached', 'import_cachefile_paths_changed', 'import_cachefile_shared_device', 'import_devices_missing', 'import_log_missing', 'import_paths_changed', 'import_rewind_config_changed', 'import_rewind_device_replaced', - 'zpool_import_status'] + 'zpool_import_status', 'zpool_import_parallel_pos', + 'zpool_import_parallel_neg', 'zpool_import_parallel_admin'] tags = ['functional', 'cli_root', 'zpool_import'] timeout = 1200 [tests/functional/cli_root/zpool_labelclear] tests = ['zpool_labelclear_active', 'zpool_labelclear_exported', 'zpool_labelclear_removed', 'zpool_labelclear_valid'] pre = post = tags = ['functional', 'cli_root', 'zpool_labelclear'] [tests/functional/cli_root/zpool_initialize] tests = ['zpool_initialize_attach_detach_add_remove', 'zpool_initialize_fault_export_import_online', 'zpool_initialize_import_export', 'zpool_initialize_offline_export_import_online', 'zpool_initialize_online_offline', 'zpool_initialize_split', 'zpool_initialize_start_and_cancel_neg', 'zpool_initialize_start_and_cancel_pos', 'zpool_initialize_suspend_resume', 'zpool_initialize_uninit', 'zpool_initialize_unsupported_vdevs', 'zpool_initialize_verify_checksums', 'zpool_initialize_verify_initialized'] pre = tags = ['functional', 'cli_root', 'zpool_initialize'] [tests/functional/cli_root/zpool_offline] tests = ['zpool_offline_001_pos', 'zpool_offline_002_neg', 'zpool_offline_003_pos'] tags = ['functional', 'cli_root', 'zpool_offline'] [tests/functional/cli_root/zpool_online] tests = ['zpool_online_001_pos', 'zpool_online_002_neg'] tags = ['functional', 'cli_root', 'zpool_online'] [tests/functional/cli_root/zpool_remove] tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos', 'zpool_remove_003_pos'] tags = ['functional', 'cli_root', 'zpool_remove'] [tests/functional/cli_root/zpool_replace] tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift'] tags = ['functional', 'cli_root', 'zpool_replace'] [tests/functional/cli_root/zpool_resilver] tests = ['zpool_resilver_bad_args', 'zpool_resilver_restart', 'zpool_resilver_concurrent'] tags = ['functional', 'cli_root', 'zpool_resilver'] [tests/functional/cli_root/zpool_scrub] tests = ['zpool_scrub_001_neg', 'zpool_scrub_002_pos', 'zpool_scrub_003_pos', 'zpool_scrub_004_pos', 'zpool_scrub_005_pos', 'zpool_scrub_encrypted_unloaded', 'zpool_scrub_print_repairing', 'zpool_scrub_offline_device', 'zpool_scrub_multiple_copies', 'zpool_error_scrub_001_pos', 'zpool_error_scrub_002_pos', 'zpool_error_scrub_003_pos', 'zpool_error_scrub_004_pos'] tags = ['functional', 'cli_root', 'zpool_scrub'] [tests/functional/cli_root/zpool_set] tests = ['zpool_set_001_pos', 'zpool_set_002_neg', 'zpool_set_003_neg', 'zpool_set_ashift', 'zpool_set_features', 'vdev_set_001_pos', 'user_property_001_pos', 'user_property_002_neg'] tags = ['functional', 'cli_root', 'zpool_set'] [tests/functional/cli_root/zpool_split] tests = ['zpool_split_cliargs', 'zpool_split_devices', 'zpool_split_encryption', 'zpool_split_props', 'zpool_split_vdevs', 'zpool_split_resilver', 'zpool_split_indirect', 'zpool_split_dryrun_output'] tags = ['functional', 'cli_root', 'zpool_split'] [tests/functional/cli_root/zpool_status] tests = ['zpool_status_001_pos', 'zpool_status_002_pos', 'zpool_status_003_pos', 'zpool_status_004_pos', 'zpool_status_005_pos', 'zpool_status_006_pos', 'zpool_status_007_pos', 'zpool_status_008_pos', 'zpool_status_features_001_pos'] tags = ['functional', 'cli_root', 'zpool_status'] [tests/functional/cli_root/zpool_sync] tests = ['zpool_sync_001_pos', 'zpool_sync_002_neg'] tags = ['functional', 'cli_root', 'zpool_sync'] [tests/functional/cli_root/zpool_trim] tests = ['zpool_trim_attach_detach_add_remove', 'zpool_trim_fault_export_import_online', 'zpool_trim_import_export', 'zpool_trim_multiple', 'zpool_trim_neg', 'zpool_trim_offline_export_import_online', 'zpool_trim_online_offline', 'zpool_trim_partial', 'zpool_trim_rate', 'zpool_trim_rate_neg', 'zpool_trim_secure', 'zpool_trim_split', 'zpool_trim_start_and_cancel_neg', 'zpool_trim_start_and_cancel_pos', 'zpool_trim_suspend_resume', 'zpool_trim_unsupported_vdevs', 'zpool_trim_verify_checksums', 'zpool_trim_verify_trimmed'] tags = ['functional', 'zpool_trim'] [tests/functional/cli_root/zpool_upgrade] tests = ['zpool_upgrade_001_pos', 'zpool_upgrade_002_pos', 'zpool_upgrade_003_pos', 'zpool_upgrade_004_pos', 'zpool_upgrade_005_neg', 'zpool_upgrade_006_neg', 'zpool_upgrade_007_pos', 'zpool_upgrade_008_pos', 'zpool_upgrade_009_neg', 'zpool_upgrade_features_001_pos'] tags = ['functional', 'cli_root', 'zpool_upgrade'] [tests/functional/cli_root/zpool_wait] tests = ['zpool_wait_discard', 'zpool_wait_freeing', 'zpool_wait_initialize_basic', 'zpool_wait_initialize_cancel', 'zpool_wait_initialize_flag', 'zpool_wait_multiple', 'zpool_wait_no_activity', 'zpool_wait_remove', 'zpool_wait_remove_cancel', 'zpool_wait_trim_basic', 'zpool_wait_trim_cancel', 'zpool_wait_trim_flag', 'zpool_wait_usage'] tags = ['functional', 'cli_root', 'zpool_wait'] [tests/functional/cli_root/zpool_wait/scan] tests = ['zpool_wait_replace_cancel', 'zpool_wait_rebuild', 'zpool_wait_resilver', 'zpool_wait_scrub_cancel', 'zpool_wait_replace', 'zpool_wait_scrub_basic', 'zpool_wait_scrub_flag'] tags = ['functional', 'cli_root', 'zpool_wait'] [tests/functional/cli_user/misc] tests = ['zdb_001_neg', 'zfs_001_neg', 'zfs_allow_001_neg', 'zfs_clone_001_neg', 'zfs_create_001_neg', 'zfs_destroy_001_neg', 'zfs_get_001_neg', 'zfs_inherit_001_neg', 'zfs_mount_001_neg', 'zfs_promote_001_neg', 'zfs_receive_001_neg', 'zfs_rename_001_neg', 'zfs_rollback_001_neg', 'zfs_send_001_neg', 'zfs_set_001_neg', 'zfs_share_001_neg', 'zfs_snapshot_001_neg', 'zfs_unallow_001_neg', 'zfs_unmount_001_neg', 'zfs_unshare_001_neg', 'zfs_upgrade_001_neg', 'zpool_001_neg', 'zpool_add_001_neg', 'zpool_attach_001_neg', 'zpool_clear_001_neg', 'zpool_create_001_neg', 'zpool_destroy_001_neg', 'zpool_detach_001_neg', 'zpool_export_001_neg', 'zpool_get_001_neg', 'zpool_history_001_neg', 'zpool_import_001_neg', 'zpool_import_002_neg', 'zpool_offline_001_neg', 'zpool_online_001_neg', 'zpool_remove_001_neg', 'zpool_replace_001_neg', 'zpool_scrub_001_neg', 'zpool_set_001_neg', 'zpool_status_001_neg', 'zpool_upgrade_001_neg', 'arcstat_001_pos', 'arc_summary_001_pos', 'arc_summary_002_neg', 'zpool_wait_privilege', 'zilstat_001_pos'] user = tags = ['functional', 'cli_user', 'misc'] [tests/functional/cli_user/zfs_list] tests = ['zfs_list_001_pos', 'zfs_list_002_pos', 'zfs_list_003_pos', 'zfs_list_004_neg', 'zfs_list_005_neg', 'zfs_list_007_pos', 'zfs_list_008_neg'] user = tags = ['functional', 'cli_user', 'zfs_list'] [tests/functional/cli_user/zpool_iostat] tests = ['zpool_iostat_001_neg', 'zpool_iostat_002_pos', 'zpool_iostat_003_neg', 'zpool_iostat_004_pos', 'zpool_iostat_005_pos', 'zpool_iostat_-c_disable', 'zpool_iostat_-c_homedir', 'zpool_iostat_-c_searchpath'] user = tags = ['functional', 'cli_user', 'zpool_iostat'] [tests/functional/cli_user/zpool_list] tests = ['zpool_list_001_pos', 'zpool_list_002_neg'] user = tags = ['functional', 'cli_user', 'zpool_list'] [tests/functional/cli_user/zpool_status] tests = ['zpool_status_003_pos', 'zpool_status_-c_disable', 'zpool_status_-c_homedir', 'zpool_status_-c_searchpath'] user = tags = ['functional', 'cli_user', 'zpool_status'] [tests/functional/compression] tests = ['compress_001_pos', 'compress_002_pos', 'compress_003_pos', 'l2arc_compressed_arc', 'l2arc_compressed_arc_disabled', 'l2arc_encrypted', 'l2arc_encrypted_no_compressed_arc'] tags = ['functional', 'compression'] [tests/functional/cp_files] tests = ['cp_files_001_pos', 'cp_files_002_pos', 'cp_stress'] tags = ['functional', 'cp_files'] [tests/functional/crtime] tests = ['crtime_001_pos' ] tags = ['functional', 'crtime'] [tests/functional/ctime] tests = ['ctime_001_pos' ] tags = ['functional', 'ctime'] [tests/functional/deadman] tests = ['deadman_ratelimit', 'deadman_sync', 'deadman_zio'] pre = post = tags = ['functional', 'deadman'] [tests/functional/delegate] tests = ['zfs_allow_001_pos', 'zfs_allow_002_pos', 'zfs_allow_003_pos', 'zfs_allow_004_pos', 'zfs_allow_005_pos', 'zfs_allow_006_pos', 'zfs_allow_007_pos', 'zfs_allow_008_pos', 'zfs_allow_009_neg', 'zfs_allow_010_pos', 'zfs_allow_011_neg', 'zfs_allow_012_neg', 'zfs_unallow_001_pos', 'zfs_unallow_002_pos', 'zfs_unallow_003_pos', 'zfs_unallow_004_pos', 'zfs_unallow_005_pos', 'zfs_unallow_006_pos', 'zfs_unallow_007_neg', 'zfs_unallow_008_neg'] tags = ['functional', 'delegate'] [tests/functional/exec] tests = ['exec_001_pos', 'exec_002_neg'] tags = ['functional', 'exec'] [tests/functional/fallocate] tests = ['fallocate_punch-hole'] tags = ['functional', 'fallocate'] [tests/functional/features/async_destroy] tests = ['async_destroy_001_pos'] tags = ['functional', 'features', 'async_destroy'] [tests/functional/features/large_dnode] tests = ['large_dnode_001_pos', 'large_dnode_003_pos', 'large_dnode_004_neg', 'large_dnode_005_pos', 'large_dnode_007_neg', 'large_dnode_009_pos'] tags = ['functional', 'features', 'large_dnode'] [tests/functional/grow] pre = post = tests = ['grow_pool_001_pos', 'grow_replicas_001_pos'] tags = ['functional', 'grow'] [tests/functional/history] tests = ['history_001_pos', 'history_002_pos', 'history_003_pos', 'history_004_pos', 'history_005_neg', 'history_006_neg', 'history_007_pos', 'history_008_pos', 'history_009_pos', 'history_010_pos'] tags = ['functional', 'history'] [tests/functional/hkdf] pre = post = tests = ['hkdf_test'] tags = ['functional', 'hkdf'] [tests/functional/inheritance] tests = ['inherit_001_pos'] pre = tags = ['functional', 'inheritance'] [tests/functional/io] tests = ['sync', 'psync', 'posixaio', 'mmap'] tags = ['functional', 'io'] [tests/functional/inuse] tests = ['inuse_004_pos', 'inuse_005_pos', 'inuse_008_pos', 'inuse_009_pos'] post = tags = ['functional', 'inuse'] [tests/functional/large_files] tests = ['large_files_001_pos', 'large_files_002_pos'] tags = ['functional', 'large_files'] [tests/functional/limits] tests = ['filesystem_count', 'filesystem_limit', 'snapshot_count', 'snapshot_limit'] tags = ['functional', 'limits'] [tests/functional/link_count] tests = ['link_count_001', 'link_count_root_inode'] tags = ['functional', 'link_count'] [tests/functional/migration] tests = ['migration_001_pos', 'migration_002_pos', 'migration_003_pos', 'migration_004_pos', 'migration_005_pos', 'migration_006_pos', 'migration_007_pos', 'migration_008_pos', 'migration_009_pos', 'migration_010_pos', 'migration_011_pos', 'migration_012_pos'] tags = ['functional', 'migration'] [tests/functional/mmap] tests = ['mmap_mixed', 'mmap_read_001_pos', 'mmap_seek_001_pos', 'mmap_sync_001_pos', 'mmap_write_001_pos'] tags = ['functional', 'mmap'] [tests/functional/mount] tests = ['umount_001', 'umountall_001'] tags = ['functional', 'mount'] [tests/functional/mv_files] tests = ['mv_files_001_pos', 'mv_files_002_pos', 'random_creation'] tags = ['functional', 'mv_files'] [tests/functional/nestedfs] tests = ['nestedfs_001_pos'] tags = ['functional', 'nestedfs'] [tests/functional/no_space] tests = ['enospc_001_pos', 'enospc_002_pos', 'enospc_003_pos', 'enospc_df', 'enospc_ganging', 'enospc_rm'] tags = ['functional', 'no_space'] [tests/functional/nopwrite] tests = ['nopwrite_copies', 'nopwrite_mtime', 'nopwrite_negative', 'nopwrite_promoted_clone', 'nopwrite_recsize', 'nopwrite_sync', 'nopwrite_varying_compression', 'nopwrite_volume'] tags = ['functional', 'nopwrite'] [tests/functional/online_offline] tests = ['online_offline_001_pos', 'online_offline_002_neg', 'online_offline_003_neg'] tags = ['functional', 'online_offline'] [tests/functional/pool_checkpoint] tests = ['checkpoint_after_rewind', 'checkpoint_big_rewind', 'checkpoint_capacity', 'checkpoint_conf_change', 'checkpoint_discard', 'checkpoint_discard_busy', 'checkpoint_discard_many', 'checkpoint_indirect', 'checkpoint_invalid', 'checkpoint_lun_expsz', 'checkpoint_open', 'checkpoint_removal', 'checkpoint_rewind', 'checkpoint_ro_rewind', 'checkpoint_sm_scale', 'checkpoint_twice', 'checkpoint_vdev_add', 'checkpoint_zdb', 'checkpoint_zhack_feat'] tags = ['functional', 'pool_checkpoint'] timeout = 1800 [tests/functional/pool_names] tests = ['pool_names_001_pos', 'pool_names_002_neg'] pre = post = tags = ['functional', 'pool_names'] [tests/functional/poolversion] tests = ['poolversion_001_pos', 'poolversion_002_pos'] tags = ['functional', 'poolversion'] [tests/functional/pyzfs] tests = ['pyzfs_unittest'] pre = post = tags = ['functional', 'pyzfs'] [tests/functional/quota] tests = ['quota_001_pos', 'quota_002_pos', 'quota_003_pos', 'quota_004_pos', 'quota_005_pos', 'quota_006_neg'] tags = ['functional', 'quota'] [tests/functional/redacted_send] tests = ['redacted_compressed', 'redacted_contents', 'redacted_deleted', 'redacted_disabled_feature', 'redacted_embedded', 'redacted_holes', 'redacted_incrementals', 'redacted_largeblocks', 'redacted_many_clones', 'redacted_mixed_recsize', 'redacted_mounts', 'redacted_negative', 'redacted_origin', 'redacted_panic', 'redacted_props', 'redacted_resume', 'redacted_size', 'redacted_volume'] tags = ['functional', 'redacted_send'] [tests/functional/raidz] tests = ['raidz_001_neg', 'raidz_002_pos', 'raidz_expand_001_pos', 'raidz_expand_002_pos', 'raidz_expand_003_neg', 'raidz_expand_003_pos', 'raidz_expand_004_pos', 'raidz_expand_005_pos', 'raidz_expand_006_neg', 'raidz_expand_007_neg'] tags = ['functional', 'raidz'] timeout = 1200 [tests/functional/redundancy] tests = ['redundancy_draid', 'redundancy_draid1', 'redundancy_draid2', 'redundancy_draid3', 'redundancy_draid_damaged1', 'redundancy_draid_damaged2', 'redundancy_draid_spare1', 'redundancy_draid_spare2', 'redundancy_draid_spare3', 'redundancy_mirror', 'redundancy_raidz', 'redundancy_raidz1', 'redundancy_raidz2', 'redundancy_raidz3', 'redundancy_stripe'] tags = ['functional', 'redundancy'] timeout = 1200 [tests/functional/refquota] tests = ['refquota_001_pos', 'refquota_002_pos', 'refquota_003_pos', 'refquota_004_pos', 'refquota_005_pos', 'refquota_006_neg', 'refquota_007_neg', 'refquota_008_neg'] tags = ['functional', 'refquota'] [tests/functional/refreserv] tests = ['refreserv_001_pos', 'refreserv_002_pos', 'refreserv_003_pos', 'refreserv_004_pos', 'refreserv_005_pos', 'refreserv_multi_raidz', 'refreserv_raidz'] tags = ['functional', 'refreserv'] [tests/functional/removal] pre = tests = ['removal_all_vdev', 'removal_cancel', 'removal_check_space', 'removal_condense_export', 'removal_multiple_indirection', 'removal_nopwrite', 'removal_remap_deadlists', 'removal_resume_export', 'removal_sanity', 'removal_with_add', 'removal_with_create_fs', 'removal_with_dedup', 'removal_with_errors', 'removal_with_export', 'removal_with_indirect', 'removal_with_ganging', 'removal_with_faulted', 'removal_with_remove', 'removal_with_scrub', 'removal_with_send', 'removal_with_send_recv', 'removal_with_snapshot', 'removal_with_write', 'removal_with_zdb', 'remove_expanded', 'remove_mirror', 'remove_mirror_sanity', 'remove_raidz', 'remove_indirect', 'remove_attach_mirror', 'removal_reservation'] tags = ['functional', 'removal'] [tests/functional/rename_dirs] tests = ['rename_dirs_001_pos'] tags = ['functional', 'rename_dirs'] [tests/functional/replacement] tests = ['attach_import', 'attach_multiple', 'attach_rebuild', 'attach_resilver', 'detach', 'rebuild_disabled_feature', 'rebuild_multiple', 'rebuild_raidz', 'replace_import', 'replace_rebuild', 'replace_resilver', 'resilver_restart_001', 'resilver_restart_002', 'scrub_cancel'] tags = ['functional', 'replacement'] [tests/functional/reservation] tests = ['reservation_001_pos', 'reservation_002_pos', 'reservation_003_pos', 'reservation_004_pos', 'reservation_005_pos', 'reservation_006_pos', 'reservation_007_pos', 'reservation_008_pos', 'reservation_009_pos', 'reservation_010_pos', 'reservation_011_pos', 'reservation_012_pos', 'reservation_013_pos', 'reservation_014_pos', 'reservation_015_pos', 'reservation_016_pos', 'reservation_017_pos', 'reservation_018_pos', 'reservation_019_pos', 'reservation_020_pos', 'reservation_021_neg', 'reservation_022_pos'] tags = ['functional', 'reservation'] [tests/functional/rootpool] tests = ['rootpool_002_neg', 'rootpool_003_neg', 'rootpool_007_pos'] tags = ['functional', 'rootpool'] [tests/functional/rsend] tests = ['recv_dedup', 'recv_dedup_encrypted_zvol', 'rsend_001_pos', 'rsend_002_pos', 'rsend_003_pos', 'rsend_004_pos', 'rsend_005_pos', 'rsend_006_pos', 'rsend_007_pos', 'rsend_008_pos', 'rsend_009_pos', 'rsend_010_pos', 'rsend_011_pos', 'rsend_012_pos', 'rsend_013_pos', 'rsend_014_pos', 'rsend_016_neg', 'rsend_019_pos', 'rsend_020_pos', 'rsend_021_pos', 'rsend_022_pos', 'rsend_024_pos', 'rsend_025_pos', 'rsend_026_neg', 'rsend_027_pos', 'rsend_028_neg', 'rsend_029_neg', 'rsend_030_pos', 'rsend_031_pos', 'send-c_verify_ratio', 'send-c_verify_contents', 'send-c_props', 'send-c_incremental', 'send-c_volume', 'send-c_zstream_recompress', 'send-c_zstreamdump', 'send-c_lz4_disabled', 'send-c_recv_lz4_disabled', 'send-c_mixed_compression', 'send-c_stream_size_estimate', 'send-c_embedded_blocks', 'send-c_resume', 'send-cpL_varied_recsize', 'send-c_recv_dedup', 'send-L_toggle', 'send_encrypted_incremental', 'send_encrypted_freeobjects', 'send_encrypted_hierarchy', 'send_encrypted_props', 'send_encrypted_truncated_files', 'send_freeobjects', 'send_realloc_files', 'send_realloc_encrypted_files', 'send_spill_block', 'send_holds', 'send_hole_birth', 'send_mixed_raw', 'send-wR_encrypted_zvol', 'send_partial_dataset', 'send_invalid', 'send_doall', 'send_raw_spill_block', 'send_raw_ashift', 'send_raw_large_blocks'] tags = ['functional', 'rsend'] [tests/functional/scrub_mirror] tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos', 'scrub_mirror_003_pos', 'scrub_mirror_004_pos'] tags = ['functional', 'scrub_mirror'] [tests/functional/slog] tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos', 'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg', 'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg', 'slog_013_pos', 'slog_014_pos', 'slog_015_neg', 'slog_replay_fs_001', 'slog_replay_fs_002', 'slog_replay_volume', 'slog_016_pos'] tags = ['functional', 'slog'] [tests/functional/snapshot] tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos', 'rollback_003_pos', 'snapshot_001_pos', 'snapshot_002_pos', 'snapshot_003_pos', 'snapshot_004_pos', 'snapshot_005_pos', 'snapshot_006_pos', 'snapshot_007_pos', 'snapshot_008_pos', 'snapshot_009_pos', 'snapshot_010_pos', 'snapshot_011_pos', 'snapshot_012_pos', 'snapshot_013_pos', 'snapshot_014_pos', 'snapshot_017_pos', 'snapshot_018_pos'] tags = ['functional', 'snapshot'] [tests/functional/snapused] tests = ['snapused_001_pos', 'snapused_002_pos', 'snapused_003_pos', 'snapused_004_pos', 'snapused_005_pos'] tags = ['functional', 'snapused'] [tests/functional/sparse] tests = ['sparse_001_pos'] tags = ['functional', 'sparse'] [tests/functional/stat] tests = ['stat_001_pos'] tags = ['functional', 'stat'] [tests/functional/suid] tests = ['suid_write_to_suid', 'suid_write_to_sgid', 'suid_write_to_suid_sgid', 'suid_write_to_none', 'suid_write_zil_replay'] tags = ['functional', 'suid'] [tests/functional/trim] tests = ['autotrim_integrity', 'autotrim_config', 'autotrim_trim_integrity', 'trim_integrity', 'trim_config', 'trim_l2arc'] tags = ['functional', 'trim'] [tests/functional/truncate] tests = ['truncate_001_pos', 'truncate_002_pos', 'truncate_timestamps'] tags = ['functional', 'truncate'] [tests/functional/upgrade] tests = ['upgrade_userobj_001_pos', 'upgrade_readonly_pool'] tags = ['functional', 'upgrade'] [tests/functional/userquota] tests = [ 'userquota_001_pos', 'userquota_002_pos', 'userquota_003_pos', 'userquota_004_pos', 'userquota_005_neg', 'userquota_006_pos', 'userquota_007_pos', 'userquota_008_pos', 'userquota_009_pos', 'userquota_010_pos', 'userquota_011_pos', 'userquota_012_neg', 'userspace_001_pos', 'userspace_002_pos', 'userspace_encrypted', 'userspace_send_encrypted', 'userspace_encrypted_13709'] tags = ['functional', 'userquota'] [tests/functional/vdev_disk:Linux] pre = post = tests = ['page_alignment'] tags = ['functional', 'vdev_disk'] [tests/functional/vdev_zaps] tests = ['vdev_zaps_001_pos', 'vdev_zaps_002_pos', 'vdev_zaps_003_pos', 'vdev_zaps_004_pos', 'vdev_zaps_005_pos', 'vdev_zaps_006_pos', 'vdev_zaps_007_pos'] tags = ['functional', 'vdev_zaps'] [tests/functional/write_dirs] tests = ['write_dirs_001_pos', 'write_dirs_002_pos'] tags = ['functional', 'write_dirs'] [tests/functional/xattr] tests = ['xattr_001_pos', 'xattr_002_neg', 'xattr_003_neg', 'xattr_004_pos', 'xattr_005_pos', 'xattr_006_pos', 'xattr_007_neg', 'xattr_011_pos', 'xattr_012_pos', 'xattr_013_pos', 'xattr_compat'] tags = ['functional', 'xattr'] [tests/functional/zvol/zvol_ENOSPC] tests = ['zvol_ENOSPC_001_pos'] tags = ['functional', 'zvol', 'zvol_ENOSPC'] [tests/functional/zvol/zvol_cli] tests = ['zvol_cli_001_pos', 'zvol_cli_002_pos', 'zvol_cli_003_neg'] tags = ['functional', 'zvol', 'zvol_cli'] [tests/functional/zvol/zvol_misc] tests = ['zvol_misc_002_pos', 'zvol_misc_hierarchy', 'zvol_misc_rename_inuse', 'zvol_misc_snapdev', 'zvol_misc_trim', 'zvol_misc_volmode', 'zvol_misc_zil'] tags = ['functional', 'zvol', 'zvol_misc'] [tests/functional/zvol/zvol_stress] tests = ['zvol_stress'] tags = ['functional', 'zvol', 'zvol_stress'] [tests/functional/zvol/zvol_swap] tests = ['zvol_swap_001_pos', 'zvol_swap_002_pos', 'zvol_swap_004_pos'] tags = ['functional', 'zvol', 'zvol_swap'] [tests/functional/libzfs] tests = ['many_fds', 'libzfs_input'] tags = ['functional', 'libzfs'] [tests/functional/log_spacemap] tests = ['log_spacemap_import_logs'] pre = post = tags = ['functional', 'log_spacemap'] [tests/functional/l2arc] tests = ['l2arc_arcstats_pos', 'l2arc_mfuonly_pos', 'l2arc_l2miss_pos', 'persist_l2arc_001_pos', 'persist_l2arc_002_pos', 'persist_l2arc_003_neg', 'persist_l2arc_004_pos', 'persist_l2arc_005_pos'] tags = ['functional', 'l2arc'] [tests/functional/zpool_influxdb] tests = ['zpool_influxdb'] tags = ['functional', 'zpool_influxdb'] diff --git a/sys/contrib/openzfs/tests/test-runner/bin/test-runner.py.in b/sys/contrib/openzfs/tests/test-runner/bin/test-runner.py.in index 422ebd7bc8bf..65247f4f06fc 100755 --- a/sys/contrib/openzfs/tests/test-runner/bin/test-runner.py.in +++ b/sys/contrib/openzfs/tests/test-runner/bin/test-runner.py.in @@ -1,1156 +1,1165 @@ #!/usr/bin/env @PYTHON_SHEBANG@ # # This file and its contents are supplied under the terms of the # Common Development and Distribution License ("CDDL"), version 1.0. # You may only use this file in accordance with the terms of version # 1.0 of the CDDL. # # A full copy of the text of the CDDL should have accompanied this # source. A copy of the CDDL is also available via the Internet at # http://www.illumos.org/license/CDDL. # # # Copyright (c) 2012, 2018 by Delphix. All rights reserved. # Copyright (c) 2019 Datto Inc. # # This script must remain compatible with Python 3.6+. # import os import sys import ctypes import re import configparser from datetime import datetime from optparse import OptionParser from pwd import getpwnam from pwd import getpwuid from select import select from subprocess import PIPE from subprocess import Popen from subprocess import check_output from threading import Timer from time import time, CLOCK_MONOTONIC from os.path import exists BASEDIR = '/var/tmp/test_results' TESTDIR = '/usr/share/zfs/' KMEMLEAK_FILE = '/sys/kernel/debug/kmemleak' KILL = 'kill' TRUE = 'true' SUDO = 'sudo' LOG_FILE = 'LOG_FILE' LOG_OUT = 'LOG_OUT' LOG_ERR = 'LOG_ERR' LOG_FILE_OBJ = None try: from time import monotonic as monotonic_time except ImportError: class timespec(ctypes.Structure): _fields_ = [ ('tv_sec', ctypes.c_long), ('tv_nsec', ctypes.c_long) ] librt = ctypes.CDLL('librt.so.1', use_errno=True) clock_gettime = librt.clock_gettime clock_gettime.argtypes = [ctypes.c_int, ctypes.POINTER(timespec)] def monotonic_time(): t = timespec() if clock_gettime(CLOCK_MONOTONIC, ctypes.pointer(t)) != 0: errno_ = ctypes.get_errno() raise OSError(errno_, os.strerror(errno_)) return t.tv_sec + t.tv_nsec * 1e-9 class Result(object): total = 0 runresults = {'PASS': 0, 'FAIL': 0, 'SKIP': 0, 'KILLED': 0, 'RERAN': 0} def __init__(self): self.starttime = None self.returncode = None self.runtime = '' self.stdout = [] self.stderr = [] self.kmemleak = '' self.result = '' def done(self, proc, killed, reran): """ Finalize the results of this Cmd. """ Result.total += 1 m, s = divmod(monotonic_time() - self.starttime, 60) self.runtime = '%02d:%02d' % (m, s) self.returncode = proc.returncode if reran is True: Result.runresults['RERAN'] += 1 if killed: self.result = 'KILLED' Result.runresults['KILLED'] += 1 elif len(self.kmemleak) > 0: self.result = 'FAIL' Result.runresults['FAIL'] += 1 elif self.returncode == 0: self.result = 'PASS' Result.runresults['PASS'] += 1 elif self.returncode == 4: self.result = 'SKIP' Result.runresults['SKIP'] += 1 elif self.returncode != 0: self.result = 'FAIL' Result.runresults['FAIL'] += 1 class Output(object): """ This class is a slightly modified version of the 'Stream' class found here: http://goo.gl/aSGfv """ - def __init__(self, stream): + def __init__(self, stream, debug=False): self.stream = stream + self.debug = debug self._buf = b'' self.lines = [] def fileno(self): return self.stream.fileno() def read(self, drain=0): """ Read from the file descriptor. If 'drain' set, read until EOF. """ while self._read() is not None: if not drain: break def _read(self): """ Read up to 4k of data from this output stream. Collect the output up to the last newline, and append it to any leftover data from a previous call. The lines are stored as a (timestamp, data) tuple for easy sorting/merging later. """ fd = self.fileno() buf = os.read(fd, 4096) if not buf: return None + if self.debug: + os.write(sys.stderr.fileno(), buf) if b'\n' not in buf: self._buf += buf return [] buf = self._buf + buf tmp, rest = buf.rsplit(b'\n', 1) self._buf = rest now = datetime.now() rows = tmp.split(b'\n') self.lines += [(now, r) for r in rows] class Cmd(object): verified_users = [] def __init__(self, pathname, identifier=None, outputdir=None, timeout=None, user=None, tags=None): self.pathname = pathname self.identifier = identifier self.outputdir = outputdir or 'BASEDIR' """ The timeout for tests is measured in wall-clock time """ self.timeout = timeout self.user = user or '' self.killed = False self.reran = None self.result = Result() if self.timeout is None: self.timeout = 60 def __str__(self): return '''\ Pathname: %s Identifier: %s Outputdir: %s Timeout: %d User: %s ''' % (self.pathname, self.identifier, self.outputdir, self.timeout, self.user) def kill_cmd(self, proc, options, kmemleak, keyboard_interrupt=False): """ Kill a running command due to timeout, or ^C from the keyboard. If sudo is required, this user was verified previously. """ self.killed = True do_sudo = len(self.user) != 0 signal = '-TERM' cmd = [SUDO, KILL, signal, str(proc.pid)] if not do_sudo: del cmd[0] try: kp = Popen(cmd) kp.wait() except Exception: pass """ If this is not a user-initiated kill and the test has not been reran before we consider if the test needs to be reran: If the test has spent some time hibernating and didn't run the whole length of time before being timed out we will rerun the test. """ if keyboard_interrupt is False and self.reran is None: runtime = monotonic_time() - self.result.starttime if int(self.timeout) > runtime: self.killed = False self.reran = False self.run(options, dryrun=False, kmemleak=kmemleak) self.reran = True def update_cmd_privs(self, cmd, user): """ If a user has been specified to run this Cmd and we're not already running as that user, prepend the appropriate sudo command to run as that user. """ me = getpwuid(os.getuid()) if not user or user is me: if os.path.isfile(cmd+'.ksh') and os.access(cmd+'.ksh', os.X_OK): cmd += '.ksh' if os.path.isfile(cmd+'.sh') and os.access(cmd+'.sh', os.X_OK): cmd += '.sh' return cmd if not os.path.isfile(cmd): if os.path.isfile(cmd+'.ksh') and os.access(cmd+'.ksh', os.X_OK): cmd += '.ksh' if os.path.isfile(cmd+'.sh') and os.access(cmd+'.sh', os.X_OK): cmd += '.sh' ret = '%s -E -u %s %s' % (SUDO, user, cmd) return ret.split(' ') - def collect_output(self, proc): + def collect_output(self, proc, debug=False): """ Read from stdout/stderr as data becomes available, until the process is no longer running. Return the lines from the stdout and stderr Output objects. """ - out = Output(proc.stdout) - err = Output(proc.stderr) + out = Output(proc.stdout, debug) + err = Output(proc.stderr, debug) res = [] while proc.returncode is None: proc.poll() res = select([out, err], [], [], .1) for fd in res[0]: fd.read() for fd in res[0]: fd.read(drain=1) return out.lines, err.lines def run(self, options, dryrun=None, kmemleak=None): """ This is the main function that runs each individual test. Determine whether or not the command requires sudo, and modify it if needed. Run the command, and update the result object. """ if dryrun is None: dryrun = options.dryrun if dryrun is True: print(self) return if kmemleak is None: kmemleak = options.kmemleak privcmd = self.update_cmd_privs(self.pathname, self.user) try: old = os.umask(0) if not os.path.isdir(self.outputdir): os.makedirs(self.outputdir, mode=0o777) os.umask(old) except OSError as e: fail('%s' % e) """ Log each test we run to /dev/kmsg (on Linux), so if there's a kernel warning we'll be able to match it up to a particular test. """ if options.kmsg is True and exists("/dev/kmsg"): try: kp = Popen([SUDO, "sh", "-c", f"echo ZTS run {self.pathname} > /dev/kmsg"]) kp.wait() except Exception: pass self.result.starttime = monotonic_time() if kmemleak: cmd = f'{SUDO} sh -c "echo clear > {KMEMLEAK_FILE}"' check_output(cmd, shell=True) proc = Popen(privcmd, stdout=PIPE, stderr=PIPE) # Allow a special timeout value of 0 to mean infinity if int(self.timeout) == 0: self.timeout = sys.maxsize / (10 ** 9) t = Timer( int(self.timeout), self.kill_cmd, [proc, options, kmemleak] ) try: t.start() - self.result.stdout, self.result.stderr = self.collect_output(proc) + + out, err = self.collect_output(proc, options.debug) + self.result.stdout = out + self.result.stderr = err if kmemleak: cmd = f'{SUDO} sh -c "echo scan > {KMEMLEAK_FILE}"' check_output(cmd, shell=True) cmd = f'{SUDO} cat {KMEMLEAK_FILE}' self.result.kmemleak = check_output(cmd, shell=True) except KeyboardInterrupt: self.kill_cmd(proc, options, kmemleak, True) fail('\nRun terminated at user request.') finally: t.cancel() if self.reran is not False: self.result.done(proc, self.killed, self.reran) def skip(self): """ Initialize enough of the test result that we can log a skipped command. """ Result.total += 1 Result.runresults['SKIP'] += 1 self.result.stdout = self.result.stderr = [] self.result.starttime = monotonic_time() m, s = divmod(monotonic_time() - self.result.starttime, 60) self.result.runtime = '%02d:%02d' % (m, s) self.result.result = 'SKIP' def log(self, options, suppress_console=False): """ This function is responsible for writing all output. This includes the console output, the logfile of all results (with timestamped merged stdout and stderr), and for each test, the unmodified stdout/stderr/merged in its own file. """ logname = getpwuid(os.getuid()).pw_name rer = '' if self.reran is True: rer = ' (RERAN)' user = ' (run as %s)' % (self.user if len(self.user) else logname) if self.identifier: msga = 'Test (%s): %s%s ' % (self.identifier, self.pathname, user) else: msga = 'Test: %s%s ' % (self.pathname, user) msgb = '[%s] [%s]%s\n' % (self.result.runtime, self.result.result, rer) pad = ' ' * (80 - (len(msga) + len(msgb))) result_line = msga + pad + msgb # The result line is always written to the log file. If -q was # specified only failures are written to the console, otherwise # the result line is written to the console. The console output # may be suppressed by calling log() with suppress_console=True. write_log(bytearray(result_line, encoding='utf-8'), LOG_FILE) if not suppress_console: if not options.quiet: write_log(result_line, LOG_OUT) elif options.quiet and self.result.result != 'PASS': write_log(result_line, LOG_OUT) lines = sorted(self.result.stdout + self.result.stderr, key=lambda x: x[0]) # Write timestamped output (stdout and stderr) to the logfile for dt, line in lines: timestamp = bytearray(dt.strftime("%H:%M:%S.%f ")[:11], encoding='utf-8') write_log(b'%s %s\n' % (timestamp, line), LOG_FILE) # Write the separate stdout/stderr/merged files, if the data exists if len(self.result.stdout): with open(os.path.join(self.outputdir, 'stdout'), 'wb') as out: for _, line in self.result.stdout: os.write(out.fileno(), b'%s\n' % line) if len(self.result.stderr): with open(os.path.join(self.outputdir, 'stderr'), 'wb') as err: for _, line in self.result.stderr: os.write(err.fileno(), b'%s\n' % line) if len(self.result.stdout) and len(self.result.stderr): with open(os.path.join(self.outputdir, 'merged'), 'wb') as merged: for _, line in lines: os.write(merged.fileno(), b'%s\n' % line) if len(self.result.kmemleak): with open(os.path.join(self.outputdir, 'kmemleak'), 'wb') as kmem: kmem.write(self.result.kmemleak) class Test(Cmd): props = ['outputdir', 'timeout', 'user', 'pre', 'pre_user', 'post', 'post_user', 'failsafe', 'failsafe_user', 'tags'] def __init__(self, pathname, pre=None, pre_user=None, post=None, post_user=None, failsafe=None, failsafe_user=None, tags=None, **kwargs): super(Test, self).__init__(pathname, **kwargs) self.pre = pre or '' self.pre_user = pre_user or '' self.post = post or '' self.post_user = post_user or '' self.failsafe = failsafe or '' self.failsafe_user = failsafe_user or '' self.tags = tags or [] def __str__(self): post_user = pre_user = failsafe_user = '' if len(self.pre_user): pre_user = ' (as %s)' % (self.pre_user) if len(self.post_user): post_user = ' (as %s)' % (self.post_user) if len(self.failsafe_user): failsafe_user = ' (as %s)' % (self.failsafe_user) return '''\ Pathname: %s Identifier: %s Outputdir: %s Timeout: %d User: %s Pre: %s%s Post: %s%s Failsafe: %s%s Tags: %s ''' % (self.pathname, self.identifier, self.outputdir, self.timeout, self.user, self.pre, pre_user, self.post, post_user, self.failsafe, failsafe_user, self.tags) def verify(self): """ Check the pre/post/failsafe scripts, user and Test. Omit the Test from this run if there are any problems. """ files = [self.pre, self.pathname, self.post, self.failsafe] users = [self.pre_user, self.user, self.post_user, self.failsafe_user] for f in [f for f in files if len(f)]: if not verify_file(f): write_log("Warning: Test '%s' not added to this run because" " it failed verification.\n" % f, LOG_ERR) return False for user in [user for user in users if len(user)]: if not verify_user(user): write_log("Not adding Test '%s' to this run.\n" % self.pathname, LOG_ERR) return False return True def run(self, options, dryrun=None, kmemleak=None): """ Create Cmd instances for the pre/post/failsafe scripts. If the pre script doesn't pass, skip this Test. Run the post script regardless. If the Test is killed, also run the failsafe script. """ odir = os.path.join(self.outputdir, os.path.basename(self.pre)) pretest = Cmd(self.pre, identifier=self.identifier, outputdir=odir, timeout=self.timeout, user=self.pre_user) test = Cmd(self.pathname, identifier=self.identifier, outputdir=self.outputdir, timeout=self.timeout, user=self.user) odir = os.path.join(self.outputdir, os.path.basename(self.failsafe)) failsafe = Cmd(self.failsafe, identifier=self.identifier, outputdir=odir, timeout=self.timeout, user=self.failsafe_user) odir = os.path.join(self.outputdir, os.path.basename(self.post)) posttest = Cmd(self.post, identifier=self.identifier, outputdir=odir, timeout=self.timeout, user=self.post_user) cont = True if len(pretest.pathname): pretest.run(options, kmemleak=False) cont = pretest.result.result == 'PASS' pretest.log(options) if cont: test.run(options, kmemleak=kmemleak) if test.result.result == 'KILLED' and len(failsafe.pathname): failsafe.run(options, kmemleak=False) failsafe.log(options, suppress_console=True) else: test.skip() test.log(options) if len(posttest.pathname): posttest.run(options, kmemleak=False) posttest.log(options) class TestGroup(Test): props = Test.props + ['tests'] def __init__(self, pathname, tests=None, **kwargs): super(TestGroup, self).__init__(pathname, **kwargs) self.tests = tests or [] def __str__(self): post_user = pre_user = failsafe_user = '' if len(self.pre_user): pre_user = ' (as %s)' % (self.pre_user) if len(self.post_user): post_user = ' (as %s)' % (self.post_user) if len(self.failsafe_user): failsafe_user = ' (as %s)' % (self.failsafe_user) return '''\ Pathname: %s Identifier: %s Outputdir: %s Tests: %s Timeout: %s User: %s Pre: %s%s Post: %s%s Failsafe: %s%s Tags: %s ''' % (self.pathname, self.identifier, self.outputdir, self.tests, self.timeout, self.user, self.pre, pre_user, self.post, post_user, self.failsafe, failsafe_user, self.tags) def filter(self, keeplist): self.tests = [x for x in self.tests if x in keeplist] def verify(self): """ Check the pre/post/failsafe scripts, user and tests in this TestGroup. Omit the TestGroup entirely, or simply delete the relevant tests in the group, if that's all that's required. """ # If the pre/post/failsafe scripts are relative pathnames, convert to # absolute, so they stand a chance of passing verification. if len(self.pre) and not os.path.isabs(self.pre): self.pre = os.path.join(self.pathname, self.pre) if len(self.post) and not os.path.isabs(self.post): self.post = os.path.join(self.pathname, self.post) if len(self.failsafe) and not os.path.isabs(self.failsafe): self.post = os.path.join(self.pathname, self.post) auxfiles = [self.pre, self.post, self.failsafe] users = [self.pre_user, self.user, self.post_user, self.failsafe_user] for f in [f for f in auxfiles if len(f)]: if f != self.failsafe and self.pathname != os.path.dirname(f): write_log("Warning: TestGroup '%s' not added to this run. " "Auxiliary script '%s' exists in a different " "directory.\n" % (self.pathname, f), LOG_ERR) return False if not verify_file(f): write_log("Warning: TestGroup '%s' not added to this run. " "Auxiliary script '%s' failed verification.\n" % (self.pathname, f), LOG_ERR) return False for user in [user for user in users if len(user)]: if not verify_user(user): write_log("Not adding TestGroup '%s' to this run.\n" % self.pathname, LOG_ERR) return False # If one of the tests is invalid, delete it, log it, and drive on. for test in self.tests: if not verify_file(os.path.join(self.pathname, test)): del self.tests[self.tests.index(test)] write_log("Warning: Test '%s' removed from TestGroup '%s' " "because it failed verification.\n" % (test, self.pathname), LOG_ERR) return len(self.tests) != 0 def run(self, options, dryrun=None, kmemleak=None): """ Create Cmd instances for the pre/post/failsafe scripts. If the pre script doesn't pass, skip all the tests in this TestGroup. Run the post script regardless. Run the failsafe script when a test is killed. """ # tags assigned to this test group also include the test names if options.tags and not set(self.tags).intersection(set(options.tags)): return odir = os.path.join(self.outputdir, os.path.basename(self.pre)) pretest = Cmd(self.pre, outputdir=odir, timeout=self.timeout, user=self.pre_user, identifier=self.identifier) odir = os.path.join(self.outputdir, os.path.basename(self.post)) posttest = Cmd(self.post, outputdir=odir, timeout=self.timeout, user=self.post_user, identifier=self.identifier) cont = True if len(pretest.pathname): pretest.run(options, dryrun=dryrun, kmemleak=False) cont = pretest.result.result == 'PASS' pretest.log(options) for fname in self.tests: odir = os.path.join(self.outputdir, fname) test = Cmd(os.path.join(self.pathname, fname), outputdir=odir, timeout=self.timeout, user=self.user, identifier=self.identifier) odir = os.path.join(odir, os.path.basename(self.failsafe)) failsafe = Cmd(self.failsafe, outputdir=odir, timeout=self.timeout, user=self.failsafe_user, identifier=self.identifier) if cont: test.run(options, dryrun=dryrun, kmemleak=kmemleak) if test.result.result == 'KILLED' and len(failsafe.pathname): failsafe.run(options, dryrun=dryrun, kmemleak=False) failsafe.log(options, suppress_console=True) else: test.skip() test.log(options) if len(posttest.pathname): posttest.run(options, dryrun=dryrun, kmemleak=False) posttest.log(options) class TestRun(object): - props = ['quiet', 'outputdir'] + props = ['quiet', 'outputdir', 'debug'] def __init__(self, options): self.tests = {} self.testgroups = {} self.starttime = time() self.timestamp = datetime.now().strftime('%Y%m%dT%H%M%S') self.outputdir = os.path.join(options.outputdir, self.timestamp) self.setup_logging(options) self.defaults = [ ('outputdir', BASEDIR), ('quiet', False), ('timeout', 60), ('user', ''), ('pre', ''), ('pre_user', ''), ('post', ''), ('post_user', ''), ('failsafe', ''), ('failsafe_user', ''), - ('tags', []) + ('tags', []), + ('debug', False) ] def __str__(self): s = 'TestRun:\n outputdir: %s\n' % self.outputdir s += 'TESTS:\n' for key in sorted(self.tests.keys()): s += '%s%s' % (self.tests[key].__str__(), '\n') s += 'TESTGROUPS:\n' for key in sorted(self.testgroups.keys()): s += '%s%s' % (self.testgroups[key].__str__(), '\n') return s def addtest(self, pathname, options): """ Create a new Test, and apply any properties that were passed in from the command line. If it passes verification, add it to the TestRun. """ test = Test(pathname) for prop in Test.props: setattr(test, prop, getattr(options, prop)) if test.verify(): self.tests[pathname] = test def addtestgroup(self, dirname, filenames, options): """ Create a new TestGroup, and apply any properties that were passed in from the command line. If it passes verification, add it to the TestRun. """ if dirname not in self.testgroups: testgroup = TestGroup(dirname) for prop in Test.props: setattr(testgroup, prop, getattr(options, prop)) # Prevent pre/post/failsafe scripts from running as regular tests for f in [testgroup.pre, testgroup.post, testgroup.failsafe]: if f in filenames: del filenames[filenames.index(f)] self.testgroups[dirname] = testgroup self.testgroups[dirname].tests = sorted(filenames) testgroup.verify() def filter(self, keeplist): for group in list(self.testgroups.keys()): if group not in keeplist: del self.testgroups[group] continue g = self.testgroups[group] if g.pre and os.path.basename(g.pre) in keeplist[group]: continue g.filter(keeplist[group]) for test in list(self.tests.keys()): directory, base = os.path.split(test) if directory not in keeplist or base not in keeplist[directory]: del self.tests[test] def read(self, options): """ Read in the specified runfiles, and apply the TestRun properties listed in the 'DEFAULT' section to our TestRun. Then read each section, and apply the appropriate properties to the Test or TestGroup. Properties from individual sections override those set in the 'DEFAULT' section. If the Test or TestGroup passes verification, add it to the TestRun. """ config = configparser.RawConfigParser() parsed = config.read(options.runfiles) failed = options.runfiles - set(parsed) if len(failed): files = ' '.join(sorted(failed)) fail("Couldn't read config files: %s" % files) for opt in TestRun.props: if config.has_option('DEFAULT', opt): setattr(self, opt, config.get('DEFAULT', opt)) self.outputdir = os.path.join(self.outputdir, self.timestamp) testdir = options.testdir for section in config.sections(): if 'tests' in config.options(section): parts = section.split(':', 1) sectiondir = parts[0] identifier = parts[1] if len(parts) == 2 else None if os.path.isdir(sectiondir): pathname = sectiondir elif os.path.isdir(os.path.join(testdir, sectiondir)): pathname = os.path.join(testdir, sectiondir) else: pathname = sectiondir testgroup = TestGroup(os.path.abspath(pathname), identifier=identifier) for prop in TestGroup.props: for sect in ['DEFAULT', section]: if config.has_option(sect, prop): if prop == 'tags': setattr(testgroup, prop, eval(config.get(sect, prop))) elif prop == 'failsafe': failsafe = config.get(sect, prop) setattr(testgroup, prop, os.path.join(testdir, failsafe)) else: setattr(testgroup, prop, config.get(sect, prop)) # Repopulate tests using eval to convert the string to a list testgroup.tests = eval(config.get(section, 'tests')) if testgroup.verify(): self.testgroups[section] = testgroup else: test = Test(section) for prop in Test.props: for sect in ['DEFAULT', section]: if config.has_option(sect, prop): if prop == 'failsafe': failsafe = config.get(sect, prop) setattr(test, prop, os.path.join(testdir, failsafe)) else: setattr(test, prop, config.get(sect, prop)) if test.verify(): self.tests[section] = test def write(self, options): """ Create a configuration file for editing and later use. The 'DEFAULT' section of the config file is created from the properties that were specified on the command line. Tests are simply added as sections that inherit everything from the 'DEFAULT' section. TestGroups are the same, except they get an option including all the tests to run in that directory. """ defaults = dict([(prop, getattr(options, prop)) for prop, _ in self.defaults]) config = configparser.RawConfigParser(defaults) for test in sorted(self.tests.keys()): config.add_section(test) for prop in Test.props: if prop not in self.props: config.set(test, prop, getattr(self.tests[test], prop)) for testgroup in sorted(self.testgroups.keys()): config.add_section(testgroup) config.set(testgroup, 'tests', self.testgroups[testgroup].tests) for prop in TestGroup.props: if prop not in self.props: config.set(testgroup, prop, getattr(self.testgroups[testgroup], prop)) try: with open(options.template, 'w') as f: return config.write(f) except IOError: fail('Could not open \'%s\' for writing.' % options.template) def complete_outputdirs(self): """ Collect all the pathnames for Tests, and TestGroups. Work backwards one pathname component at a time, to create a unique directory name in which to deposit test output. Tests will be able to write output files directly in the newly modified outputdir. TestGroups will be able to create one subdirectory per test in the outputdir, and are guaranteed uniqueness because a group can only contain files in one directory. Pre and post tests will create a directory rooted at the outputdir of the Test or TestGroup in question for their output. Failsafe scripts will create a directory rooted at the outputdir of each Test for their output. """ done = False components = 0 tmp_dict = dict(list(self.tests.items()) + list(self.testgroups.items())) total = len(tmp_dict) base = self.outputdir while not done: paths = [] components -= 1 for testfile in list(tmp_dict.keys()): uniq = '/'.join(testfile.split('/')[components:]).lstrip('/') if uniq not in paths: paths.append(uniq) tmp_dict[testfile].outputdir = os.path.join(base, uniq) else: break done = total == len(paths) def setup_logging(self, options): """ This function creates the output directory and gets a file object for the logfile. This function must be called before write_log() can be used. """ if options.dryrun is True: return global LOG_FILE_OBJ if not options.template: try: old = os.umask(0) os.makedirs(self.outputdir, mode=0o777) os.umask(old) filename = os.path.join(self.outputdir, 'log') LOG_FILE_OBJ = open(filename, buffering=0, mode='wb') except OSError as e: fail('%s' % e) def run(self, options): """ Walk through all the Tests and TestGroups, calling run(). """ try: os.chdir(self.outputdir) except OSError: fail('Could not change to directory %s' % self.outputdir) # make a symlink to the output for the currently running test logsymlink = os.path.join(self.outputdir, '../current') if os.path.islink(logsymlink): os.unlink(logsymlink) if not os.path.exists(logsymlink): os.symlink(self.outputdir, logsymlink) else: write_log('Could not make a symlink to directory %s\n' % self.outputdir, LOG_ERR) if options.kmemleak: cmd = f'{SUDO} -c "echo scan=0 > {KMEMLEAK_FILE}"' check_output(cmd, shell=True) iteration = 0 while iteration < options.iterations: for test in sorted(self.tests.keys()): self.tests[test].run(options) for testgroup in sorted(self.testgroups.keys()): self.testgroups[testgroup].run(options) iteration += 1 def summary(self): if Result.total == 0: return 2 print('\nResults Summary') for key in list(Result.runresults.keys()): if Result.runresults[key] != 0: print('%s\t% 4d' % (key, Result.runresults[key])) m, s = divmod(time() - self.starttime, 60) h, m = divmod(m, 60) print('\nRunning Time:\t%02d:%02d:%02d' % (h, m, s)) print('Percent passed:\t%.1f%%' % ((float(Result.runresults['PASS']) / float(Result.total)) * 100)) print('Log directory:\t%s' % self.outputdir) if Result.runresults['FAIL'] > 0: return 1 if Result.runresults['KILLED'] > 0: return 1 if Result.runresults['RERAN'] > 0: return 3 return 0 def write_log(msg, target): """ Write the provided message to standard out, standard error or the logfile. If specifying LOG_FILE, then `msg` must be a bytes like object. This way we can still handle output from tests that may be in unexpected encodings. """ if target == LOG_OUT: os.write(sys.stdout.fileno(), bytearray(msg, encoding='utf-8')) elif target == LOG_ERR: os.write(sys.stderr.fileno(), bytearray(msg, encoding='utf-8')) elif target == LOG_FILE: os.write(LOG_FILE_OBJ.fileno(), msg) else: fail('log_msg called with unknown target "%s"' % target) def verify_file(pathname): """ Verify that the supplied pathname is an executable regular file. """ if os.path.isdir(pathname) or os.path.islink(pathname): return False for ext in '', '.ksh', '.sh': script_path = pathname + ext if os.path.isfile(script_path) and os.access(script_path, os.X_OK): return True return False def verify_user(user): """ Verify that the specified user exists on this system, and can execute sudo without being prompted for a password. """ testcmd = [SUDO, '-n', '-u', user, TRUE] if user in Cmd.verified_users: return True try: getpwnam(user) except KeyError: write_log("Warning: user '%s' does not exist.\n" % user, LOG_ERR) return False p = Popen(testcmd) p.wait() if p.returncode != 0: write_log("Warning: user '%s' cannot use passwordless sudo.\n" % user, LOG_ERR) return False else: Cmd.verified_users.append(user) return True def find_tests(testrun, options): """ For the given list of pathnames, add files as Tests. For directories, if do_groups is True, add the directory as a TestGroup. If False, recursively search for executable files. """ for p in sorted(options.pathnames): if os.path.isdir(p): for dirname, _, filenames in os.walk(p): if options.do_groups: testrun.addtestgroup(dirname, filenames, options) else: for f in sorted(filenames): testrun.addtest(os.path.join(dirname, f), options) else: testrun.addtest(p, options) def filter_tests(testrun, options): try: fh = open(options.logfile, "r") except Exception as e: fail('%s' % e) failed = {} while True: line = fh.readline() if not line: break m = re.match(r'Test: .*(tests/.*)/(\S+).*\[FAIL\]', line) if not m: continue group, test = m.group(1, 2) try: failed[group].append(test) except KeyError: failed[group] = [test] fh.close() testrun.filter(failed) def fail(retstr, ret=1): print('%s: %s' % (sys.argv[0], retstr)) exit(ret) def kmemleak_cb(option, opt_str, value, parser): if not os.path.exists(KMEMLEAK_FILE): fail(f"File '{KMEMLEAK_FILE}' doesn't exist. " + "Enable CONFIG_DEBUG_KMEMLEAK in kernel configuration.") setattr(parser.values, option.dest, True) def options_cb(option, opt_str, value, parser): path_options = ['outputdir', 'template', 'testdir', 'logfile'] if opt_str in parser.rargs: fail('%s may only be specified once.' % opt_str) if option.dest == 'runfiles': parser.values.cmd = 'rdconfig' value = set(os.path.abspath(p) for p in value.split(',')) if option.dest == 'tags': value = [x.strip() for x in value.split(',')] if option.dest in path_options: setattr(parser.values, option.dest, os.path.abspath(value)) else: setattr(parser.values, option.dest, value) def parse_args(): parser = OptionParser() parser.add_option('-c', action='callback', callback=options_cb, type='string', dest='runfiles', metavar='runfiles', help='Specify tests to run via config files.') parser.add_option('-d', action='store_true', default=False, dest='dryrun', help='Dry run. Print tests, but take no other action.') + parser.add_option('-D', action='store_true', default=False, dest='debug', + help='Write all test output to stdout as it arrives.') parser.add_option('-l', action='callback', callback=options_cb, default=None, dest='logfile', metavar='logfile', type='string', help='Read logfile and re-run tests which failed.') parser.add_option('-g', action='store_true', default=False, dest='do_groups', help='Make directories TestGroups.') parser.add_option('-o', action='callback', callback=options_cb, default=BASEDIR, dest='outputdir', type='string', metavar='outputdir', help='Specify an output directory.') parser.add_option('-i', action='callback', callback=options_cb, default=TESTDIR, dest='testdir', type='string', metavar='testdir', help='Specify a test directory.') parser.add_option('-K', action='store_true', default=False, dest='kmsg', help='Log tests names to /dev/kmsg') parser.add_option('-m', action='callback', callback=kmemleak_cb, default=False, dest='kmemleak', help='Enable kmemleak reporting (Linux only)') parser.add_option('-p', action='callback', callback=options_cb, default='', dest='pre', metavar='script', type='string', help='Specify a pre script.') parser.add_option('-P', action='callback', callback=options_cb, default='', dest='post', metavar='script', type='string', help='Specify a post script.') parser.add_option('-q', action='store_true', default=False, dest='quiet', help='Silence on the console during a test run.') parser.add_option('-s', action='callback', callback=options_cb, default='', dest='failsafe', metavar='script', type='string', help='Specify a failsafe script.') parser.add_option('-S', action='callback', callback=options_cb, default='', dest='failsafe_user', metavar='failsafe_user', type='string', help='Specify a user to execute the failsafe script.') parser.add_option('-t', action='callback', callback=options_cb, default=60, dest='timeout', metavar='seconds', type='int', help='Timeout (in seconds) for an individual test.') parser.add_option('-u', action='callback', callback=options_cb, default='', dest='user', metavar='user', type='string', help='Specify a different user name to run as.') parser.add_option('-w', action='callback', callback=options_cb, default=None, dest='template', metavar='template', type='string', help='Create a new config file.') parser.add_option('-x', action='callback', callback=options_cb, default='', dest='pre_user', metavar='pre_user', type='string', help='Specify a user to execute the pre script.') parser.add_option('-X', action='callback', callback=options_cb, default='', dest='post_user', metavar='post_user', type='string', help='Specify a user to execute the post script.') parser.add_option('-T', action='callback', callback=options_cb, default='', dest='tags', metavar='tags', type='string', help='Specify tags to execute specific test groups.') parser.add_option('-I', action='callback', callback=options_cb, default=1, dest='iterations', metavar='iterations', type='int', help='Number of times to run the test run.') (options, pathnames) = parser.parse_args() if options.runfiles and len(pathnames): fail('Extraneous arguments.') options.pathnames = [os.path.abspath(path) for path in pathnames] return options def main(): options = parse_args() testrun = TestRun(options) if options.runfiles: testrun.read(options) else: find_tests(testrun, options) if options.logfile: filter_tests(testrun, options) if options.template: testrun.write(options) exit(0) testrun.complete_outputdirs() testrun.run(options) exit(testrun.summary()) if __name__ == '__main__': main() diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am index f182a2825cd6..dc447e042225 100644 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am @@ -1,2117 +1,2120 @@ CLEANFILES = dist_noinst_DATA = include $(top_srcdir)/config/Substfiles.am datadir_zfs_tests_testsdir = $(datadir)/$(PACKAGE)/zfs-tests/tests nobase_dist_datadir_zfs_tests_tests_DATA = \ perf/nfs-sample.cfg \ perf/perf.shlib \ \ perf/fio/mkfiles.fio \ perf/fio/random_reads.fio \ perf/fio/random_readwrite.fio \ perf/fio/random_readwrite_fixed.fio \ perf/fio/random_writes.fio \ perf/fio/sequential_reads.fio \ perf/fio/sequential_readwrite.fio \ perf/fio/sequential_writes.fio nobase_dist_datadir_zfs_tests_tests_SCRIPTS = \ perf/regression/random_reads.ksh \ perf/regression/random_readwrite.ksh \ perf/regression/random_readwrite_fixed.ksh \ perf/regression/random_writes.ksh \ perf/regression/random_writes_zil.ksh \ perf/regression/sequential_reads_arc_cached_clone.ksh \ perf/regression/sequential_reads_arc_cached.ksh \ perf/regression/sequential_reads_dbuf_cached.ksh \ perf/regression/sequential_reads.ksh \ perf/regression/sequential_writes.ksh \ perf/regression/setup.ksh \ \ perf/scripts/prefetch_io.sh # These lists can be regenerated by running make regen-tests at the root, or, on a *clean* source: # find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po' ! -executable -name '*.in' | sort | sed 's/\.in$//;s/^/\t/;$!s/$/ \\/' # find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po' -executable -name '*.in' | sort | sed 's/\.in$//;s/^/\t/;$!s/$/ \\/' # find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po' ! -name '*.in' ! -name '*.c' | grep -Fe /simd -e /tmpfile | sort | sed 's/^/\t/;$!s/$/ \\/' # find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po' ! -executable ! -name '*.in' ! -name '*.c' | grep -vFe /simd -e /tmpfile | sort | sed 's/^/\t/;$!s/$/ \\/' # find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po' -executable ! -name '*.in' ! -name '*.c' | grep -vFe /simd -e /tmpfile | sort | sed 's/^/\t/;$!s/$/ \\/' # # simd and tmpfile are Linux-only and not installed elsewhere # # C programs are specced in ../Makefile.am above as part of the main Makefile find_common := find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po' regen: @$(MAKE) -C $(top_builddir) clean @$(MAKE) clean $(SED) $(ac_inplace) '/^# -- >8 --/q' Makefile.am echo >> Makefile.am echo 'nobase_nodist_datadir_zfs_tests_tests_DATA = \' >> Makefile.am $(find_common) ! -executable -name '*.in' | sort | sed 's/\.in$$//;s/^/\t/;$$!s/$$/ \\/' >> Makefile.am echo 'nobase_nodist_datadir_zfs_tests_tests_SCRIPTS = \' >> Makefile.am $(find_common) -executable -name '*.in' | sort | sed 's/\.in$$//;s/^/\t/;$$!s/$$/ \\/' >> Makefile.am echo >> Makefile.am echo 'SUBSTFILES += $$(nobase_nodist_datadir_zfs_tests_tests_DATA) $$(nobase_nodist_datadir_zfs_tests_tests_SCRIPTS)' >> Makefile.am echo >> Makefile.am echo 'if BUILD_LINUX' >> Makefile.am echo 'nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \' >> Makefile.am $(find_common) ! -name '*.in' ! -name '*.c' | grep -Fe /simd -e /tmpfile | sort | sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am echo 'endif' >> Makefile.am echo >> Makefile.am echo 'nobase_dist_datadir_zfs_tests_tests_DATA += \' >> Makefile.am $(find_common) ! -executable ! -name '*.in' ! -name '*.c' | grep -vFe /simd -e /tmpfile | sort | sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am echo >> Makefile.am echo 'nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \' >> Makefile.am $(find_common) -executable ! -name '*.in' ! -name '*.c' | grep -vFe /simd -e /tmpfile | sort | sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am # -- >8 -- nobase_nodist_datadir_zfs_tests_tests_DATA = \ functional/pam/utilities.kshlib nobase_nodist_datadir_zfs_tests_tests_SCRIPTS = \ functional/pyzfs/pyzfs_unittest.ksh SUBSTFILES += $(nobase_nodist_datadir_zfs_tests_tests_DATA) $(nobase_nodist_datadir_zfs_tests_tests_SCRIPTS) if BUILD_LINUX nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \ functional/simd/simd_supported.ksh \ functional/tmpfile/cleanup.ksh \ functional/tmpfile/setup.ksh endif nobase_dist_datadir_zfs_tests_tests_DATA += \ functional/acl/acl.cfg \ functional/acl/acl_common.kshlib \ functional/alloc_class/alloc_class.cfg \ functional/alloc_class/alloc_class.kshlib \ functional/atime/atime.cfg \ functional/atime/atime_common.kshlib \ functional/bclone/bclone.cfg \ functional/bclone/bclone_common.kshlib \ functional/bclone/bclone_corner_cases.kshlib \ functional/block_cloning/block_cloning.kshlib \ functional/cache/cache.cfg \ functional/cache/cache.kshlib \ functional/cachefile/cachefile.cfg \ functional/cachefile/cachefile.kshlib \ functional/casenorm/casenorm.cfg \ functional/casenorm/casenorm.kshlib \ functional/channel_program/channel_common.kshlib \ functional/channel_program/lua_core/tst.args_to_lua.out \ functional/channel_program/lua_core/tst.args_to_lua.zcp \ functional/channel_program/lua_core/tst.divide_by_zero.err \ functional/channel_program/lua_core/tst.divide_by_zero.zcp \ functional/channel_program/lua_core/tst.exists.zcp \ functional/channel_program/lua_core/tst.large_prog.out \ functional/channel_program/lua_core/tst.large_prog.zcp \ functional/channel_program/lua_core/tst.lib_base.lua \ functional/channel_program/lua_core/tst.lib_coroutine.lua \ functional/channel_program/lua_core/tst.lib_strings.lua \ functional/channel_program/lua_core/tst.lib_table.lua \ functional/channel_program/lua_core/tst.nested_neg.zcp \ functional/channel_program/lua_core/tst.nested_pos.zcp \ functional/channel_program/lua_core/tst.recursive.zcp \ functional/channel_program/lua_core/tst.return_large.zcp \ functional/channel_program/lua_core/tst.return_recursive_table.zcp \ functional/channel_program/lua_core/tst.stack_gsub.err \ functional/channel_program/lua_core/tst.stack_gsub.zcp \ functional/channel_program/lua_core/tst.timeout.zcp \ functional/channel_program/synctask_core/tst.bookmark.copy.zcp \ functional/channel_program/synctask_core/tst.bookmark.create.zcp \ functional/channel_program/synctask_core/tst.get_index_props.out \ functional/channel_program/synctask_core/tst.get_index_props.zcp \ functional/channel_program/synctask_core/tst.get_number_props.out \ functional/channel_program/synctask_core/tst.get_number_props.zcp \ functional/channel_program/synctask_core/tst.get_string_props.out \ functional/channel_program/synctask_core/tst.get_string_props.zcp \ functional/channel_program/synctask_core/tst.promote_conflict.zcp \ functional/channel_program/synctask_core/tst.set_props.zcp \ functional/channel_program/synctask_core/tst.snapshot_destroy.zcp \ functional/channel_program/synctask_core/tst.snapshot_neg.zcp \ functional/channel_program/synctask_core/tst.snapshot_recursive.zcp \ functional/channel_program/synctask_core/tst.snapshot_rename.zcp \ functional/channel_program/synctask_core/tst.snapshot_simple.zcp \ functional/checksum/default.cfg \ functional/clean_mirror/clean_mirror_common.kshlib \ functional/clean_mirror/default.cfg \ functional/cli_root/cli_common.kshlib \ functional/cli_root/zfs_copies/zfs_copies.cfg \ functional/cli_root/zfs_copies/zfs_copies.kshlib \ functional/cli_root/zfs_create/properties.kshlib \ functional/cli_root/zfs_create/zfs_create.cfg \ functional/cli_root/zfs_create/zfs_create_common.kshlib \ functional/cli_root/zfs_destroy/zfs_destroy.cfg \ functional/cli_root/zfs_destroy/zfs_destroy_common.kshlib \ functional/cli_root/zfs_get/zfs_get_common.kshlib \ functional/cli_root/zfs_get/zfs_get_list_d.kshlib \ functional/cli_root/zfs_jail/jail.conf \ functional/cli_root/zfs_load-key/HEXKEY \ functional/cli_root/zfs_load-key/PASSPHRASE \ functional/cli_root/zfs_load-key/RAWKEY \ functional/cli_root/zfs_load-key/zfs_load-key.cfg \ functional/cli_root/zfs_load-key/zfs_load-key_common.kshlib \ functional/cli_root/zfs_mount/zfs_mount.cfg \ functional/cli_root/zfs_mount/zfs_mount.kshlib \ functional/cli_root/zfs_promote/zfs_promote.cfg \ functional/cli_root/zfs_receive/zstd_test_data.txt \ functional/cli_root/zfs_rename/zfs_rename.cfg \ functional/cli_root/zfs_rename/zfs_rename.kshlib \ functional/cli_root/zfs_rollback/zfs_rollback.cfg \ functional/cli_root/zfs_rollback/zfs_rollback_common.kshlib \ functional/cli_root/zfs_send/zfs_send.cfg \ functional/cli_root/zfs_set/zfs_set_common.kshlib \ functional/cli_root/zfs_share/zfs_share.cfg \ functional/cli_root/zfs_snapshot/zfs_snapshot.cfg \ functional/cli_root/zfs_unmount/zfs_unmount.cfg \ functional/cli_root/zfs_unmount/zfs_unmount.kshlib \ functional/cli_root/zfs_upgrade/zfs_upgrade.kshlib \ functional/cli_root/zfs_wait/zfs_wait.kshlib \ functional/cli_root/zpool_add/zpool_add.cfg \ functional/cli_root/zpool_add/zpool_add.kshlib \ functional/cli_root/zpool_clear/zpool_clear.cfg \ functional/cli_root/zpool_create/draidcfg.gz \ functional/cli_root/zpool_create/zpool_create.cfg \ functional/cli_root/zpool_create/zpool_create.shlib \ functional/cli_root/zpool_destroy/zpool_destroy.cfg \ functional/cli_root/zpool_events/zpool_events.cfg \ functional/cli_root/zpool_events/zpool_events.kshlib \ functional/cli_root/zpool_expand/zpool_expand.cfg \ functional/cli_root/zpool_export/zpool_export.cfg \ functional/cli_root/zpool_export/zpool_export.kshlib \ functional/cli_root/zpool_get/vdev_get.cfg \ functional/cli_root/zpool_get/zpool_get.cfg \ functional/cli_root/zpool_get/zpool_get_parsable.cfg \ functional/cli_root/zpool_import/blockfiles/cryptv0.dat.bz2 \ functional/cli_root/zpool_import/blockfiles/missing_ivset.dat.bz2 \ functional/cli_root/zpool_import/blockfiles/unclean_export.dat.bz2 \ functional/cli_root/zpool_import/zpool_import.cfg \ functional/cli_root/zpool_import/zpool_import.kshlib \ functional/cli_root/zpool_initialize/zpool_initialize.kshlib \ functional/cli_root/zpool_labelclear/labelclear.cfg \ functional/cli_root/zpool_remove/zpool_remove.cfg \ functional/cli_root/zpool_reopen/zpool_reopen.cfg \ functional/cli_root/zpool_reopen/zpool_reopen.shlib \ functional/cli_root/zpool_resilver/zpool_resilver.cfg \ functional/cli_root/zpool_scrub/zpool_scrub.cfg \ functional/cli_root/zpool_split/zpool_split.cfg \ functional/cli_root/zpool_trim/zpool_trim.kshlib \ functional/cli_root/zpool_upgrade/blockfiles/zfs-broken-mirror1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-broken-mirror2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v10.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v11.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v12.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v13.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v14.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v15.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz21.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz22.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz23.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe1.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe2.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe3.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v4.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v5.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v6.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v7.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v8.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v999.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v9.dat.bz2 \ functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-vBROKEN.dat.bz2 \ functional/cli_root/zpool_upgrade/zpool_upgrade.cfg \ functional/cli_root/zpool_upgrade/zpool_upgrade.kshlib \ functional/cli_root/zpool_wait/zpool_wait.kshlib \ functional/cli_root/zhack/library.kshlib \ functional/cli_user/misc/misc.cfg \ functional/cli_user/zfs_list/zfs_list.cfg \ functional/cli_user/zfs_list/zfs_list.kshlib \ functional/compression/compress.cfg \ functional/compression/testpool_zstd.tar.gz \ functional/deadman/deadman.cfg \ functional/delegate/delegate.cfg \ functional/delegate/delegate_common.kshlib \ functional/devices/devices.cfg \ functional/devices/devices_common.kshlib \ functional/events/events.cfg \ functional/events/events_common.kshlib \ functional/fault/fault.cfg \ functional/grow/grow.cfg \ functional/history/history.cfg \ functional/history/history_common.kshlib \ functional/history/i386.migratedpool.DAT.Z \ functional/history/i386.orig_history.txt \ functional/history/sparc.migratedpool.DAT.Z \ functional/history/sparc.orig_history.txt \ functional/history/zfs-pool-v4.dat.Z \ functional/inheritance/config001.cfg \ functional/inheritance/config002.cfg \ functional/inheritance/config003.cfg \ functional/inheritance/config004.cfg \ functional/inheritance/config005.cfg \ functional/inheritance/config006.cfg \ functional/inheritance/config007.cfg \ functional/inheritance/config008.cfg \ functional/inheritance/config009.cfg \ functional/inheritance/config010.cfg \ functional/inheritance/config011.cfg \ functional/inheritance/config012.cfg \ functional/inheritance/config013.cfg \ functional/inheritance/config014.cfg \ functional/inheritance/config015.cfg \ functional/inheritance/config016.cfg \ functional/inheritance/config017.cfg \ functional/inheritance/config018.cfg \ functional/inheritance/config019.cfg \ functional/inheritance/config020.cfg \ functional/inheritance/config021.cfg \ functional/inheritance/config022.cfg \ functional/inheritance/config023.cfg \ functional/inheritance/config024.cfg \ functional/inheritance/inherit.kshlib \ functional/inheritance/README.config \ functional/inheritance/README.state \ functional/inheritance/state001.cfg \ functional/inheritance/state002.cfg \ functional/inheritance/state003.cfg \ functional/inheritance/state004.cfg \ functional/inheritance/state005.cfg \ functional/inheritance/state006.cfg \ functional/inheritance/state007.cfg \ functional/inheritance/state008.cfg \ functional/inheritance/state009.cfg \ functional/inheritance/state010.cfg \ functional/inheritance/state011.cfg \ functional/inheritance/state012.cfg \ functional/inheritance/state013.cfg \ functional/inheritance/state014.cfg \ functional/inheritance/state015.cfg \ functional/inheritance/state016.cfg \ functional/inheritance/state017.cfg \ functional/inheritance/state018.cfg \ functional/inheritance/state019.cfg \ functional/inheritance/state020.cfg \ functional/inheritance/state021.cfg \ functional/inheritance/state022.cfg \ functional/inheritance/state023.cfg \ functional/inheritance/state024.cfg \ functional/inuse/inuse.cfg \ functional/io/io.cfg \ functional/l2arc/l2arc.cfg \ functional/largest_pool/largest_pool.cfg \ functional/migration/migration.cfg \ functional/migration/migration.kshlib \ functional/mmap/mmap.cfg \ functional/mmp/mmp.cfg \ functional/mmp/mmp.kshlib \ functional/mv_files/mv_files.cfg \ functional/mv_files/mv_files_common.kshlib \ functional/nopwrite/nopwrite.shlib \ functional/no_space/enospc.cfg \ functional/online_offline/online_offline.cfg \ functional/pool_checkpoint/pool_checkpoint.kshlib \ functional/projectquota/projectquota.cfg \ functional/projectquota/projectquota_common.kshlib \ functional/quota/quota.cfg \ functional/quota/quota.kshlib \ functional/redacted_send/redacted.cfg \ functional/redacted_send/redacted.kshlib \ functional/redundancy/redundancy.cfg \ functional/redundancy/redundancy.kshlib \ functional/refreserv/refreserv.cfg \ functional/removal/removal.kshlib \ functional/replacement/replacement.cfg \ functional/reservation/reservation.cfg \ functional/reservation/reservation.shlib \ functional/rsend/dedup_encrypted_zvol.bz2 \ functional/rsend/dedup_encrypted_zvol.zsend.bz2 \ functional/rsend/dedup.zsend.bz2 \ functional/rsend/fs.tar.gz \ functional/rsend/rsend.cfg \ functional/rsend/rsend.kshlib \ functional/scrub_mirror/default.cfg \ functional/scrub_mirror/scrub_mirror_common.kshlib \ functional/slog/slog.cfg \ functional/slog/slog.kshlib \ functional/snapshot/snapshot.cfg \ functional/snapused/snapused.kshlib \ functional/sparse/sparse.cfg \ functional/trim/trim.cfg \ functional/trim/trim.kshlib \ functional/truncate/truncate.cfg \ functional/upgrade/upgrade_common.kshlib \ functional/user_namespace/user_namespace.cfg \ functional/user_namespace/user_namespace_common.kshlib \ functional/userquota/13709_reproducer.bz2 \ functional/userquota/userquota.cfg \ functional/userquota/userquota_common.kshlib \ functional/vdev_zaps/vdev_zaps.kshlib \ functional/xattr/xattr.cfg \ functional/xattr/xattr_common.kshlib \ functional/zvol/zvol.cfg \ functional/zvol/zvol_cli/zvol_cli.cfg \ functional/zvol/zvol_common.shlib \ functional/zvol/zvol_ENOSPC/zvol_ENOSPC.cfg \ functional/zvol/zvol_misc/zvol_misc_common.kshlib \ functional/zvol/zvol_swap/zvol_swap.cfg \ functional/idmap_mount/idmap_mount.cfg \ functional/idmap_mount/idmap_mount_common.kshlib nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \ functional/acl/off/cleanup.ksh \ functional/acl/off/dosmode.ksh \ functional/acl/off/posixmode.ksh \ functional/acl/off/setup.ksh \ functional/acl/posix/cleanup.ksh \ functional/acl/posix/posix_001_pos.ksh \ functional/acl/posix/posix_002_pos.ksh \ functional/acl/posix/posix_003_pos.ksh \ functional/acl/posix/posix_004_pos.ksh \ functional/acl/posix-sa/cleanup.ksh \ functional/acl/posix-sa/posix_001_pos.ksh \ functional/acl/posix-sa/posix_002_pos.ksh \ functional/acl/posix-sa/posix_003_pos.ksh \ functional/acl/posix-sa/posix_004_pos.ksh \ functional/acl/posix-sa/setup.ksh \ functional/acl/posix/setup.ksh \ functional/alloc_class/alloc_class_001_pos.ksh \ functional/alloc_class/alloc_class_002_neg.ksh \ functional/alloc_class/alloc_class_003_pos.ksh \ functional/alloc_class/alloc_class_004_pos.ksh \ functional/alloc_class/alloc_class_005_pos.ksh \ functional/alloc_class/alloc_class_006_pos.ksh \ functional/alloc_class/alloc_class_007_pos.ksh \ functional/alloc_class/alloc_class_008_pos.ksh \ functional/alloc_class/alloc_class_009_pos.ksh \ functional/alloc_class/alloc_class_010_pos.ksh \ functional/alloc_class/alloc_class_011_neg.ksh \ functional/alloc_class/alloc_class_012_pos.ksh \ functional/alloc_class/alloc_class_013_pos.ksh \ functional/alloc_class/alloc_class_014_neg.ksh \ functional/alloc_class/alloc_class_015_pos.ksh \ functional/alloc_class/cleanup.ksh \ functional/alloc_class/setup.ksh \ functional/append/file_append.ksh \ functional/append/threadsappend_001_pos.ksh \ functional/append/cleanup.ksh \ functional/append/setup.ksh \ functional/arc/arcstats_runtime_tuning.ksh \ functional/arc/cleanup.ksh \ functional/arc/dbufstats_001_pos.ksh \ functional/arc/dbufstats_002_pos.ksh \ functional/arc/dbufstats_003_pos.ksh \ functional/arc/setup.ksh \ functional/atime/atime_001_pos.ksh \ functional/atime/atime_002_neg.ksh \ functional/atime/atime_003_pos.ksh \ functional/atime/cleanup.ksh \ functional/atime/root_atime_off.ksh \ functional/atime/root_atime_on.ksh \ functional/atime/root_relatime_on.ksh \ functional/atime/setup.ksh \ functional/bclone/bclone_crossfs_corner_cases.ksh \ functional/bclone/bclone_crossfs_corner_cases_limited.ksh \ functional/bclone/bclone_crossfs_data.ksh \ functional/bclone/bclone_crossfs_embedded.ksh \ functional/bclone/bclone_crossfs_hole.ksh \ functional/bclone/bclone_diffprops_all.ksh \ functional/bclone/bclone_diffprops_checksum.ksh \ functional/bclone/bclone_diffprops_compress.ksh \ functional/bclone/bclone_diffprops_copies.ksh \ functional/bclone/bclone_diffprops_recordsize.ksh \ functional/bclone/bclone_prop_sync.ksh \ functional/bclone/bclone_samefs_corner_cases.ksh \ functional/bclone/bclone_samefs_corner_cases_limited.ksh \ functional/bclone/bclone_samefs_data.ksh \ functional/bclone/bclone_samefs_embedded.ksh \ functional/bclone/bclone_samefs_hole.ksh \ functional/bclone/cleanup.ksh \ functional/bclone/setup.ksh \ functional/block_cloning/cleanup.ksh \ functional/block_cloning/setup.ksh \ functional/block_cloning/block_cloning_clone_mmap_cached.ksh \ functional/block_cloning/block_cloning_clone_mmap_write.ksh \ functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh \ functional/block_cloning/block_cloning_copyfilerange_fallback.ksh \ functional/block_cloning/block_cloning_copyfilerange_fallback_same_txg.ksh \ functional/block_cloning/block_cloning_copyfilerange.ksh \ functional/block_cloning/block_cloning_copyfilerange_partial.ksh \ functional/block_cloning/block_cloning_disabled_copyfilerange.ksh \ functional/block_cloning/block_cloning_disabled_ficlone.ksh \ functional/block_cloning/block_cloning_disabled_ficlonerange.ksh \ functional/block_cloning/block_cloning_ficlone.ksh \ functional/block_cloning/block_cloning_ficlonerange.ksh \ functional/block_cloning/block_cloning_ficlonerange_partial.ksh \ functional/block_cloning/block_cloning_cross_enc_dataset.ksh \ functional/block_cloning/block_cloning_replay.ksh \ functional/block_cloning/block_cloning_replay_encrypted.ksh \ functional/block_cloning/block_cloning_lwb_buffer_overflow.ksh \ functional/bootfs/bootfs_001_pos.ksh \ functional/bootfs/bootfs_002_neg.ksh \ functional/bootfs/bootfs_003_pos.ksh \ functional/bootfs/bootfs_004_neg.ksh \ functional/bootfs/bootfs_005_neg.ksh \ functional/bootfs/bootfs_006_pos.ksh \ functional/bootfs/bootfs_007_pos.ksh \ functional/bootfs/bootfs_008_pos.ksh \ functional/bootfs/cleanup.ksh \ functional/bootfs/setup.ksh \ functional/btree/btree_negative.ksh \ functional/btree/btree_positive.ksh \ functional/cache/cache_001_pos.ksh \ functional/cache/cache_002_pos.ksh \ functional/cache/cache_003_pos.ksh \ functional/cache/cache_004_neg.ksh \ functional/cache/cache_005_neg.ksh \ functional/cache/cache_006_pos.ksh \ functional/cache/cache_007_neg.ksh \ functional/cache/cache_008_neg.ksh \ functional/cache/cache_009_pos.ksh \ functional/cache/cache_010_pos.ksh \ functional/cache/cache_011_pos.ksh \ functional/cache/cache_012_pos.ksh \ functional/cache/cleanup.ksh \ functional/cachefile/cachefile_001_pos.ksh \ functional/cachefile/cachefile_002_pos.ksh \ functional/cachefile/cachefile_003_pos.ksh \ functional/cachefile/cachefile_004_pos.ksh \ functional/cachefile/cleanup.ksh \ functional/cachefile/setup.ksh \ functional/cache/setup.ksh \ functional/casenorm/case_all_values.ksh \ functional/casenorm/cleanup.ksh \ functional/casenorm/insensitive_formd_delete.ksh \ functional/casenorm/insensitive_formd_lookup.ksh \ functional/casenorm/insensitive_none_delete.ksh \ functional/casenorm/insensitive_none_lookup.ksh \ functional/casenorm/mixed_create_failure.ksh \ functional/casenorm/mixed_formd_delete.ksh \ functional/casenorm/mixed_formd_lookup_ci.ksh \ functional/casenorm/mixed_formd_lookup.ksh \ functional/casenorm/mixed_none_delete.ksh \ functional/casenorm/mixed_none_lookup_ci.ksh \ functional/casenorm/mixed_none_lookup.ksh \ functional/casenorm/norm_all_values.ksh \ functional/casenorm/sensitive_formd_delete.ksh \ functional/casenorm/sensitive_formd_lookup.ksh \ functional/casenorm/sensitive_none_delete.ksh \ functional/casenorm/sensitive_none_lookup.ksh \ functional/casenorm/setup.ksh \ functional/channel_program/lua_core/cleanup.ksh \ functional/channel_program/lua_core/setup.ksh \ functional/channel_program/lua_core/tst.args_to_lua.ksh \ functional/channel_program/lua_core/tst.divide_by_zero.ksh \ functional/channel_program/lua_core/tst.exists.ksh \ functional/channel_program/lua_core/tst.integer_illegal.ksh \ functional/channel_program/lua_core/tst.integer_overflow.ksh \ functional/channel_program/lua_core/tst.language_functions_neg.ksh \ functional/channel_program/lua_core/tst.language_functions_pos.ksh \ functional/channel_program/lua_core/tst.large_prog.ksh \ functional/channel_program/lua_core/tst.libraries.ksh \ functional/channel_program/lua_core/tst.memory_limit.ksh \ functional/channel_program/lua_core/tst.nested_neg.ksh \ functional/channel_program/lua_core/tst.nested_pos.ksh \ functional/channel_program/lua_core/tst.nvlist_to_lua.ksh \ functional/channel_program/lua_core/tst.recursive_neg.ksh \ functional/channel_program/lua_core/tst.recursive_pos.ksh \ functional/channel_program/lua_core/tst.return_large.ksh \ functional/channel_program/lua_core/tst.return_nvlist_neg.ksh \ functional/channel_program/lua_core/tst.return_nvlist_pos.ksh \ functional/channel_program/lua_core/tst.return_recursive_table.ksh \ functional/channel_program/lua_core/tst.stack_gsub.ksh \ functional/channel_program/lua_core/tst.timeout.ksh \ functional/channel_program/synctask_core/cleanup.ksh \ functional/channel_program/synctask_core/setup.ksh \ functional/channel_program/synctask_core/tst.bookmark.copy.ksh \ functional/channel_program/synctask_core/tst.bookmark.create.ksh \ functional/channel_program/synctask_core/tst.destroy_fs.ksh \ functional/channel_program/synctask_core/tst.destroy_snap.ksh \ functional/channel_program/synctask_core/tst.get_count_and_limit.ksh \ functional/channel_program/synctask_core/tst.get_index_props.ksh \ functional/channel_program/synctask_core/tst.get_mountpoint.ksh \ functional/channel_program/synctask_core/tst.get_neg.ksh \ functional/channel_program/synctask_core/tst.get_number_props.ksh \ functional/channel_program/synctask_core/tst.get_string_props.ksh \ functional/channel_program/synctask_core/tst.get_type.ksh \ functional/channel_program/synctask_core/tst.get_userquota.ksh \ functional/channel_program/synctask_core/tst.get_written.ksh \ functional/channel_program/synctask_core/tst.inherit.ksh \ functional/channel_program/synctask_core/tst.list_bookmarks.ksh \ functional/channel_program/synctask_core/tst.list_children.ksh \ functional/channel_program/synctask_core/tst.list_clones.ksh \ functional/channel_program/synctask_core/tst.list_holds.ksh \ functional/channel_program/synctask_core/tst.list_snapshots.ksh \ functional/channel_program/synctask_core/tst.list_system_props.ksh \ functional/channel_program/synctask_core/tst.list_user_props.ksh \ functional/channel_program/synctask_core/tst.parse_args_neg.ksh \ functional/channel_program/synctask_core/tst.promote_conflict.ksh \ functional/channel_program/synctask_core/tst.promote_multiple.ksh \ functional/channel_program/synctask_core/tst.promote_simple.ksh \ functional/channel_program/synctask_core/tst.rollback_mult.ksh \ functional/channel_program/synctask_core/tst.rollback_one.ksh \ functional/channel_program/synctask_core/tst.set_props.ksh \ functional/channel_program/synctask_core/tst.snapshot_destroy.ksh \ functional/channel_program/synctask_core/tst.snapshot_neg.ksh \ functional/channel_program/synctask_core/tst.snapshot_recursive.ksh \ functional/channel_program/synctask_core/tst.snapshot_rename.ksh \ functional/channel_program/synctask_core/tst.snapshot_simple.ksh \ functional/channel_program/synctask_core/tst.terminate_by_signal.ksh \ functional/chattr/chattr_001_pos.ksh \ functional/chattr/chattr_002_neg.ksh \ functional/chattr/cleanup.ksh \ functional/chattr/setup.ksh \ functional/checksum/cleanup.ksh \ functional/checksum/filetest_001_pos.ksh \ functional/checksum/filetest_002_pos.ksh \ functional/checksum/run_blake3_test.ksh \ functional/checksum/run_edonr_test.ksh \ functional/checksum/run_sha2_test.ksh \ functional/checksum/run_skein_test.ksh \ functional/checksum/setup.ksh \ functional/clean_mirror/clean_mirror_001_pos.ksh \ functional/clean_mirror/clean_mirror_002_pos.ksh \ functional/clean_mirror/clean_mirror_003_pos.ksh \ functional/clean_mirror/clean_mirror_004_pos.ksh \ functional/clean_mirror/cleanup.ksh \ functional/clean_mirror/setup.ksh \ functional/cli_root/zinject/zinject_args.ksh \ functional/cli_root/zdb/zdb_002_pos.ksh \ functional/cli_root/zdb/zdb_003_pos.ksh \ functional/cli_root/zdb/zdb_004_pos.ksh \ functional/cli_root/zdb/zdb_005_pos.ksh \ functional/cli_root/zdb/zdb_006_pos.ksh \ functional/cli_root/zdb/zdb_args_neg.ksh \ functional/cli_root/zdb/zdb_args_pos.ksh \ functional/cli_root/zdb/zdb_backup.ksh \ functional/cli_root/zdb/zdb_block_size_histogram.ksh \ functional/cli_root/zdb/zdb_checksum.ksh \ functional/cli_root/zdb/zdb_decompress.ksh \ functional/cli_root/zdb/zdb_decompress_zstd.ksh \ functional/cli_root/zdb/zdb_display_block.ksh \ functional/cli_root/zdb/zdb_encrypted.ksh \ functional/cli_root/zdb/zdb_label_checksum.ksh \ functional/cli_root/zdb/zdb_object_range_neg.ksh \ functional/cli_root/zdb/zdb_object_range_pos.ksh \ functional/cli_root/zdb/zdb_objset_id.ksh \ functional/cli_root/zdb/zdb_recover_2.ksh \ functional/cli_root/zdb/zdb_recover.ksh \ functional/cli_root/zfs_bookmark/cleanup.ksh \ functional/cli_root/zfs_bookmark/setup.ksh \ functional/cli_root/zfs_bookmark/zfs_bookmark_cliargs.ksh \ functional/cli_root/zfs_change-key/cleanup.ksh \ functional/cli_root/zfs_change-key/setup.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_child.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_clones.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_format.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_inherit.ksh \ functional/cli_root/zfs_change-key/zfs_change-key.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_load.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_location.ksh \ functional/cli_root/zfs_change-key/zfs_change-key_pbkdf2iters.ksh \ functional/cli_root/zfs/cleanup.ksh \ functional/cli_root/zfs_clone/cleanup.ksh \ functional/cli_root/zfs_clone/setup.ksh \ functional/cli_root/zfs_clone/zfs_clone_001_neg.ksh \ functional/cli_root/zfs_clone/zfs_clone_002_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_003_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_004_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_005_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_006_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_007_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_008_neg.ksh \ functional/cli_root/zfs_clone/zfs_clone_009_neg.ksh \ functional/cli_root/zfs_clone/zfs_clone_010_pos.ksh \ functional/cli_root/zfs_clone/zfs_clone_deeply_nested.ksh \ functional/cli_root/zfs_clone/zfs_clone_encrypted.ksh \ functional/cli_root/zfs_clone/zfs_clone_rm_nested.ksh \ functional/cli_root/zfs_copies/cleanup.ksh \ functional/cli_root/zfs_copies/setup.ksh \ functional/cli_root/zfs_copies/zfs_copies_001_pos.ksh \ functional/cli_root/zfs_copies/zfs_copies_002_pos.ksh \ functional/cli_root/zfs_copies/zfs_copies_003_pos.ksh \ functional/cli_root/zfs_copies/zfs_copies_004_neg.ksh \ functional/cli_root/zfs_copies/zfs_copies_005_neg.ksh \ functional/cli_root/zfs_copies/zfs_copies_006_pos.ksh \ functional/cli_root/zfs_create/cleanup.ksh \ functional/cli_root/zfs_create/setup.ksh \ functional/cli_root/zfs_create/zfs_create_001_pos.ksh \ functional/cli_root/zfs_create/zfs_create_002_pos.ksh \ functional/cli_root/zfs_create/zfs_create_003_pos.ksh \ functional/cli_root/zfs_create/zfs_create_004_pos.ksh \ functional/cli_root/zfs_create/zfs_create_005_pos.ksh \ functional/cli_root/zfs_create/zfs_create_006_pos.ksh \ functional/cli_root/zfs_create/zfs_create_007_pos.ksh \ functional/cli_root/zfs_create/zfs_create_008_neg.ksh \ functional/cli_root/zfs_create/zfs_create_009_neg.ksh \ functional/cli_root/zfs_create/zfs_create_010_neg.ksh \ functional/cli_root/zfs_create/zfs_create_011_pos.ksh \ functional/cli_root/zfs_create/zfs_create_012_pos.ksh \ functional/cli_root/zfs_create/zfs_create_013_pos.ksh \ functional/cli_root/zfs_create/zfs_create_014_pos.ksh \ functional/cli_root/zfs_create/zfs_create_crypt_combos.ksh \ functional/cli_root/zfs_create/zfs_create_dryrun.ksh \ functional/cli_root/zfs_create/zfs_create_encrypted.ksh \ functional/cli_root/zfs_create/zfs_create_nomount.ksh \ functional/cli_root/zfs_create/zfs_create_verbose.ksh \ functional/cli_root/zfs_destroy/cleanup.ksh \ functional/cli_root/zfs_destroy/setup.ksh \ functional/cli_root/zfs_destroy/zfs_clone_livelist_condense_and_disable.ksh \ functional/cli_root/zfs_destroy/zfs_clone_livelist_condense_races.ksh \ functional/cli_root/zfs_destroy/zfs_clone_livelist_dedup.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_001_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_002_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_003_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_004_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_005_neg.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_006_neg.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_007_neg.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_008_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_009_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_010_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_011_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_012_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_013_neg.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_014_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_015_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_016_pos.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_clone_livelist.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_dev_removal_condense.ksh \ functional/cli_root/zfs_destroy/zfs_destroy_dev_removal.ksh \ functional/cli_root/zfs_diff/cleanup.ksh \ functional/cli_root/zfs_diff/setup.ksh \ functional/cli_root/zfs_diff/zfs_diff_changes.ksh \ functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh \ functional/cli_root/zfs_diff/zfs_diff_encrypted.ksh \ functional/cli_root/zfs_diff/zfs_diff_mangle.ksh \ functional/cli_root/zfs_diff/zfs_diff_timestamp.ksh \ functional/cli_root/zfs_diff/zfs_diff_types.ksh \ functional/cli_root/zfs_get/cleanup.ksh \ functional/cli_root/zfs_get/setup.ksh \ functional/cli_root/zfs_get/zfs_get_001_pos.ksh \ functional/cli_root/zfs_get/zfs_get_002_pos.ksh \ functional/cli_root/zfs_get/zfs_get_003_pos.ksh \ functional/cli_root/zfs_get/zfs_get_004_pos.ksh \ functional/cli_root/zfs_get/zfs_get_005_neg.ksh \ functional/cli_root/zfs_get/zfs_get_006_neg.ksh \ functional/cli_root/zfs_get/zfs_get_007_neg.ksh \ functional/cli_root/zfs_get/zfs_get_008_pos.ksh \ functional/cli_root/zfs_get/zfs_get_009_pos.ksh \ functional/cli_root/zfs_get/zfs_get_010_neg.ksh \ functional/cli_root/zfs_ids_to_path/cleanup.ksh \ functional/cli_root/zfs_ids_to_path/setup.ksh \ functional/cli_root/zfs_ids_to_path/zfs_ids_to_path_001_pos.ksh \ functional/cli_root/zfs_inherit/cleanup.ksh \ functional/cli_root/zfs_inherit/setup.ksh \ functional/cli_root/zfs_inherit/zfs_inherit_001_neg.ksh \ functional/cli_root/zfs_inherit/zfs_inherit_002_neg.ksh \ functional/cli_root/zfs_inherit/zfs_inherit_003_pos.ksh \ functional/cli_root/zfs_inherit/zfs_inherit_mountpoint.ksh \ functional/cli_root/zfs_jail/cleanup.ksh \ functional/cli_root/zfs_jail/setup.ksh \ functional/cli_root/zfs_jail/zfs_jail_001_pos.ksh \ functional/cli_root/zfs_load-key/cleanup.ksh \ functional/cli_root/zfs_load-key/setup.ksh \ functional/cli_root/zfs_load-key/zfs_load-key_all.ksh \ functional/cli_root/zfs_load-key/zfs_load-key_file.ksh \ functional/cli_root/zfs_load-key/zfs_load-key_https.ksh \ functional/cli_root/zfs_load-key/zfs_load-key.ksh \ functional/cli_root/zfs_load-key/zfs_load-key_location.ksh \ functional/cli_root/zfs_load-key/zfs_load-key_noop.ksh \ functional/cli_root/zfs_load-key/zfs_load-key_recursive.ksh \ functional/cli_root/zfs_mount/cleanup.ksh \ functional/cli_root/zfs_mount/setup.ksh \ functional/cli_root/zfs_mount/zfs_mount_001_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_002_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_003_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_004_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_005_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_006_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_007_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_008_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_009_neg.ksh \ functional/cli_root/zfs_mount/zfs_mount_010_neg.ksh \ functional/cli_root/zfs_mount/zfs_mount_011_neg.ksh \ functional/cli_root/zfs_mount/zfs_mount_012_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_013_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_014_neg.ksh \ functional/cli_root/zfs_mount/zfs_mount_all_001_pos.ksh \ functional/cli_root/zfs_mount/zfs_mount_all_fail.ksh \ functional/cli_root/zfs_mount/zfs_mount_all_mountpoints.ksh \ functional/cli_root/zfs_mount/zfs_mount_encrypted.ksh \ functional/cli_root/zfs_mount/zfs_mount_recursive.ksh \ functional/cli_root/zfs_mount/zfs_mount_remount.ksh \ functional/cli_root/zfs_mount/zfs_mount_test_race.ksh \ functional/cli_root/zfs_mount/zfs_multi_mount.ksh \ functional/cli_root/zfs_program/cleanup.ksh \ functional/cli_root/zfs_program/setup.ksh \ functional/cli_root/zfs_program/zfs_program_json.ksh \ functional/cli_root/zfs_promote/cleanup.ksh \ functional/cli_root/zfs_promote/setup.ksh \ functional/cli_root/zfs_promote/zfs_promote_001_pos.ksh \ functional/cli_root/zfs_promote/zfs_promote_002_pos.ksh \ functional/cli_root/zfs_promote/zfs_promote_003_pos.ksh \ functional/cli_root/zfs_promote/zfs_promote_004_pos.ksh \ functional/cli_root/zfs_promote/zfs_promote_005_pos.ksh \ functional/cli_root/zfs_promote/zfs_promote_006_neg.ksh \ functional/cli_root/zfs_promote/zfs_promote_007_neg.ksh \ functional/cli_root/zfs_promote/zfs_promote_008_pos.ksh \ functional/cli_root/zfs_promote/zfs_promote_encryptionroot.ksh \ functional/cli_root/zfs_property/cleanup.ksh \ functional/cli_root/zfs_property/setup.ksh \ functional/cli_root/zfs_property/zfs_written_property_001_pos.ksh \ functional/cli_root/zfs_receive/cleanup.ksh \ functional/cli_root/zfs_receive/receive-o-x_props_aliases.ksh \ functional/cli_root/zfs_receive/receive-o-x_props_override.ksh \ functional/cli_root/zfs_receive/setup.ksh \ functional/cli_root/zfs_receive/zfs_receive_001_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_002_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_003_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_004_neg.ksh \ functional/cli_root/zfs_receive/zfs_receive_005_neg.ksh \ functional/cli_root/zfs_receive/zfs_receive_006_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_007_neg.ksh \ functional/cli_root/zfs_receive/zfs_receive_008_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_009_neg.ksh \ functional/cli_root/zfs_receive/zfs_receive_010_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_011_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_012_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_013_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_014_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_015_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_016_pos.ksh \ functional/cli_root/zfs_receive/zfs_receive_-e.ksh \ functional/cli_root/zfs_receive/zfs_receive_from_encrypted.ksh \ functional/cli_root/zfs_receive/zfs_receive_from_zstd.ksh \ functional/cli_root/zfs_receive/zfs_receive_new_props.ksh \ functional/cli_root/zfs_receive/zfs_receive_raw_-d.ksh \ functional/cli_root/zfs_receive/zfs_receive_raw_incremental.ksh \ functional/cli_root/zfs_receive/zfs_receive_raw.ksh \ functional/cli_root/zfs_receive/zfs_receive_to_encrypted.ksh \ functional/cli_root/zfs_receive/zfs_receive_-wR-encrypted-mix.ksh \ functional/cli_root/zfs_receive/zfs_receive_corrective.ksh \ functional/cli_root/zfs_receive/zfs_receive_compressed_corrective.ksh \ functional/cli_root/zfs_receive/zfs_receive_large_block_corrective.ksh \ functional/cli_root/zfs_rename/cleanup.ksh \ functional/cli_root/zfs_rename/setup.ksh \ functional/cli_root/zfs_rename/zfs_rename_001_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_002_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_003_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_004_neg.ksh \ functional/cli_root/zfs_rename/zfs_rename_005_neg.ksh \ functional/cli_root/zfs_rename/zfs_rename_006_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_007_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_008_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_009_neg.ksh \ functional/cli_root/zfs_rename/zfs_rename_010_neg.ksh \ functional/cli_root/zfs_rename/zfs_rename_011_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_012_neg.ksh \ functional/cli_root/zfs_rename/zfs_rename_013_pos.ksh \ functional/cli_root/zfs_rename/zfs_rename_014_neg.ksh \ functional/cli_root/zfs_rename/zfs_rename_encrypted_child.ksh \ functional/cli_root/zfs_rename/zfs_rename_mountpoint.ksh \ functional/cli_root/zfs_rename/zfs_rename_nounmount.ksh \ functional/cli_root/zfs_rename/zfs_rename_to_encrypted.ksh \ functional/cli_root/zfs_reservation/cleanup.ksh \ functional/cli_root/zfs_reservation/setup.ksh \ functional/cli_root/zfs_reservation/zfs_reservation_001_pos.ksh \ functional/cli_root/zfs_reservation/zfs_reservation_002_pos.ksh \ functional/cli_root/zfs_rollback/cleanup.ksh \ functional/cli_root/zfs_rollback/setup.ksh \ functional/cli_root/zfs_rollback/zfs_rollback_001_pos.ksh \ functional/cli_root/zfs_rollback/zfs_rollback_002_pos.ksh \ functional/cli_root/zfs_rollback/zfs_rollback_003_neg.ksh \ functional/cli_root/zfs_rollback/zfs_rollback_004_neg.ksh \ functional/cli_root/zfs_send/cleanup.ksh \ functional/cli_root/zfs_send/setup.ksh \ functional/cli_root/zfs_send/zfs_send_001_pos.ksh \ functional/cli_root/zfs_send/zfs_send_002_pos.ksh \ functional/cli_root/zfs_send/zfs_send_003_pos.ksh \ functional/cli_root/zfs_send/zfs_send_004_neg.ksh \ functional/cli_root/zfs_send/zfs_send_005_pos.ksh \ functional/cli_root/zfs_send/zfs_send_006_pos.ksh \ functional/cli_root/zfs_send/zfs_send_007_pos.ksh \ functional/cli_root/zfs_send/zfs_send-b.ksh \ functional/cli_root/zfs_send/zfs_send_encrypted.ksh \ functional/cli_root/zfs_send/zfs_send_encrypted_unloaded.ksh \ functional/cli_root/zfs_send/zfs_send_raw.ksh \ functional/cli_root/zfs_send/zfs_send_skip_missing.ksh \ functional/cli_root/zfs_send/zfs_send_sparse.ksh \ functional/cli_root/zfs_set/cache_001_pos.ksh \ functional/cli_root/zfs_set/cache_002_neg.ksh \ functional/cli_root/zfs_set/canmount_001_pos.ksh \ functional/cli_root/zfs_set/canmount_002_pos.ksh \ functional/cli_root/zfs_set/canmount_003_pos.ksh \ functional/cli_root/zfs_set/canmount_004_pos.ksh \ functional/cli_root/zfs_set/checksum_001_pos.ksh \ functional/cli_root/zfs_set/cleanup.ksh \ functional/cli_root/zfs_set/compression_001_pos.ksh \ functional/cli_root/zfs_set/mountpoint_001_pos.ksh \ functional/cli_root/zfs_set/mountpoint_002_pos.ksh \ functional/cli_root/zfs_set/mountpoint_003_pos.ksh \ functional/cli_root/zfs_set/onoffs_001_pos.ksh \ functional/cli_root/zfs_set/property_alias_001_pos.ksh \ functional/cli_root/zfs_set/readonly_001_pos.ksh \ functional/cli_root/zfs_set/reservation_001_neg.ksh \ functional/cli_root/zfs_set/ro_props_001_pos.ksh \ functional/cli_root/zfs_set/setup.ksh \ functional/cli_root/zfs_set/share_mount_001_neg.ksh \ functional/cli_root/zfs_set/snapdir_001_pos.ksh \ functional/cli_root/zfs/setup.ksh \ functional/cli_root/zfs_set/user_property_001_pos.ksh \ functional/cli_root/zfs_set/user_property_002_pos.ksh \ functional/cli_root/zfs_set/user_property_003_neg.ksh \ functional/cli_root/zfs_set/user_property_004_pos.ksh \ functional/cli_root/zfs_set/version_001_neg.ksh \ functional/cli_root/zfs_set/zfs_set_001_neg.ksh \ functional/cli_root/zfs_set/zfs_set_002_neg.ksh \ functional/cli_root/zfs_set/zfs_set_003_neg.ksh \ functional/cli_root/zfs_set/zfs_set_feature_activation.ksh \ functional/cli_root/zfs_set/zfs_set_keylocation.ksh \ functional/cli_root/zfs_set/zfs_set_nomount.ksh \ functional/cli_root/zfs_share/cleanup.ksh \ functional/cli_root/zfs_share/setup.ksh \ functional/cli_root/zfs_share/zfs_share_001_pos.ksh \ functional/cli_root/zfs_share/zfs_share_002_pos.ksh \ functional/cli_root/zfs_share/zfs_share_003_pos.ksh \ functional/cli_root/zfs_share/zfs_share_004_pos.ksh \ functional/cli_root/zfs_share/zfs_share_005_pos.ksh \ functional/cli_root/zfs_share/zfs_share_006_pos.ksh \ functional/cli_root/zfs_share/zfs_share_007_neg.ksh \ functional/cli_root/zfs_share/zfs_share_008_neg.ksh \ functional/cli_root/zfs_share/zfs_share_009_neg.ksh \ functional/cli_root/zfs_share/zfs_share_010_neg.ksh \ functional/cli_root/zfs_share/zfs_share_011_pos.ksh \ functional/cli_root/zfs_share/zfs_share_012_pos.ksh \ functional/cli_root/zfs_share/zfs_share_013_pos.ksh \ functional/cli_root/zfs_share/zfs_share_concurrent_shares.ksh \ functional/cli_root/zfs_share/zfs_share_after_mount.ksh \ functional/cli_root/zfs_snapshot/cleanup.ksh \ functional/cli_root/zfs_snapshot/setup.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_001_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_002_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_003_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_004_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_005_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_006_pos.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_007_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_008_neg.ksh \ functional/cli_root/zfs_snapshot/zfs_snapshot_009_pos.ksh \ functional/cli_root/zfs_sysfs/cleanup.ksh \ functional/cli_root/zfs_sysfs/setup.ksh \ functional/cli_root/zfs_sysfs/zfeature_set_unsupported.ksh \ functional/cli_root/zfs_sysfs/zfs_get_unsupported.ksh \ functional/cli_root/zfs_sysfs/zfs_set_unsupported.ksh \ functional/cli_root/zfs_sysfs/zfs_sysfs_live.ksh \ functional/cli_root/zfs_sysfs/zpool_get_unsupported.ksh \ functional/cli_root/zfs_sysfs/zpool_set_unsupported.ksh \ functional/cli_root/zfs_unload-key/cleanup.ksh \ functional/cli_root/zfs_unload-key/setup.ksh \ functional/cli_root/zfs_unload-key/zfs_unload-key_all.ksh \ functional/cli_root/zfs_unload-key/zfs_unload-key.ksh \ functional/cli_root/zfs_unload-key/zfs_unload-key_recursive.ksh \ functional/cli_root/zfs_unmount/cleanup.ksh \ functional/cli_root/zfs_unmount/setup.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_001_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_002_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_003_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_004_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_005_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_006_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_007_neg.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_008_neg.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_009_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_all_001_pos.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_nested.ksh \ functional/cli_root/zfs_unmount/zfs_unmount_unload_keys.ksh \ functional/cli_root/zfs_unshare/cleanup.ksh \ functional/cli_root/zfs_unshare/setup.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_001_pos.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_002_pos.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_003_pos.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_004_neg.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_005_neg.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_006_pos.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_007_pos.ksh \ functional/cli_root/zfs_unshare/zfs_unshare_008_pos.ksh \ functional/cli_root/zfs_upgrade/cleanup.ksh \ functional/cli_root/zfs_upgrade/setup.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_001_pos.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_002_pos.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_003_pos.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_004_pos.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_005_pos.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_006_neg.ksh \ functional/cli_root/zfs_upgrade/zfs_upgrade_007_neg.ksh \ functional/cli_root/zfs_wait/cleanup.ksh \ functional/cli_root/zfs_wait/setup.ksh \ functional/cli_root/zfs_wait/zfs_wait_deleteq.ksh \ functional/cli_root/zfs_wait/zfs_wait_getsubopt.ksh \ functional/cli_root/zfs/zfs_001_neg.ksh \ functional/cli_root/zfs/zfs_002_pos.ksh \ functional/cli_root/zfs/zfs_003_neg.ksh \ functional/cli_root/zhack/zhack_label_repair_001.ksh \ functional/cli_root/zhack/zhack_label_repair_002.ksh \ functional/cli_root/zhack/zhack_label_repair_003.ksh \ functional/cli_root/zhack/zhack_label_repair_004.ksh \ functional/cli_root/zpool_add/add_nested_replacing_spare.ksh \ functional/cli_root/zpool_add/add-o_ashift.ksh \ functional/cli_root/zpool_add/add_prop_ashift.ksh \ functional/cli_root/zpool_add/cleanup.ksh \ functional/cli_root/zpool_add/setup.ksh \ functional/cli_root/zpool_add/zpool_add--allow-ashift-mismatch.ksh \ functional/cli_root/zpool_add/zpool_add_001_pos.ksh \ functional/cli_root/zpool_add/zpool_add_002_pos.ksh \ functional/cli_root/zpool_add/zpool_add_003_pos.ksh \ functional/cli_root/zpool_add/zpool_add_004_pos.ksh \ functional/cli_root/zpool_add/zpool_add_005_pos.ksh \ functional/cli_root/zpool_add/zpool_add_006_pos.ksh \ functional/cli_root/zpool_add/zpool_add_007_neg.ksh \ functional/cli_root/zpool_add/zpool_add_008_neg.ksh \ functional/cli_root/zpool_add/zpool_add_009_neg.ksh \ functional/cli_root/zpool_add/zpool_add_010_pos.ksh \ functional/cli_root/zpool_add/zpool_add_dryrun_output.ksh \ functional/cli_root/zpool_attach/attach-o_ashift.ksh \ functional/cli_root/zpool_attach/cleanup.ksh \ functional/cli_root/zpool_attach/setup.ksh \ functional/cli_root/zpool_attach/zpool_attach_001_neg.ksh \ functional/cli_root/zpool/cleanup.ksh \ functional/cli_root/zpool_clear/cleanup.ksh \ functional/cli_root/zpool_clear/setup.ksh \ functional/cli_root/zpool_clear/zpool_clear_001_pos.ksh \ functional/cli_root/zpool_clear/zpool_clear_002_neg.ksh \ functional/cli_root/zpool_clear/zpool_clear_003_neg.ksh \ functional/cli_root/zpool_clear/zpool_clear_readonly.ksh \ functional/cli_root/zpool_create/cleanup.ksh \ functional/cli_root/zpool_create/create-o_ashift.ksh \ functional/cli_root/zpool_create/setup.ksh \ functional/cli_root/zpool_create/zpool_create_001_pos.ksh \ functional/cli_root/zpool_create/zpool_create_002_pos.ksh \ functional/cli_root/zpool_create/zpool_create_003_pos.ksh \ functional/cli_root/zpool_create/zpool_create_004_pos.ksh \ functional/cli_root/zpool_create/zpool_create_005_pos.ksh \ functional/cli_root/zpool_create/zpool_create_006_pos.ksh \ functional/cli_root/zpool_create/zpool_create_007_neg.ksh \ functional/cli_root/zpool_create/zpool_create_008_pos.ksh \ functional/cli_root/zpool_create/zpool_create_009_neg.ksh \ functional/cli_root/zpool_create/zpool_create_010_neg.ksh \ functional/cli_root/zpool_create/zpool_create_011_neg.ksh \ functional/cli_root/zpool_create/zpool_create_012_neg.ksh \ functional/cli_root/zpool_create/zpool_create_014_neg.ksh \ functional/cli_root/zpool_create/zpool_create_015_neg.ksh \ functional/cli_root/zpool_create/zpool_create_016_pos.ksh \ functional/cli_root/zpool_create/zpool_create_017_neg.ksh \ functional/cli_root/zpool_create/zpool_create_018_pos.ksh \ functional/cli_root/zpool_create/zpool_create_019_pos.ksh \ functional/cli_root/zpool_create/zpool_create_020_pos.ksh \ functional/cli_root/zpool_create/zpool_create_021_pos.ksh \ functional/cli_root/zpool_create/zpool_create_022_pos.ksh \ functional/cli_root/zpool_create/zpool_create_023_neg.ksh \ functional/cli_root/zpool_create/zpool_create_024_pos.ksh \ functional/cli_root/zpool_create/zpool_create_crypt_combos.ksh \ functional/cli_root/zpool_create/zpool_create_draid_001_pos.ksh \ functional/cli_root/zpool_create/zpool_create_draid_002_pos.ksh \ functional/cli_root/zpool_create/zpool_create_draid_003_pos.ksh \ functional/cli_root/zpool_create/zpool_create_draid_004_pos.ksh \ functional/cli_root/zpool_create/zpool_create_dryrun_output.ksh \ functional/cli_root/zpool_create/zpool_create_encrypted.ksh \ functional/cli_root/zpool_create/zpool_create_features_001_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_002_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_003_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_004_neg.ksh \ functional/cli_root/zpool_create/zpool_create_features_005_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_006_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_007_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_008_pos.ksh \ functional/cli_root/zpool_create/zpool_create_features_009_pos.ksh \ functional/cli_root/zpool_create/zpool_create_tempname.ksh \ functional/cli_root/zpool_destroy/zpool_destroy_001_pos.ksh \ functional/cli_root/zpool_destroy/zpool_destroy_002_pos.ksh \ functional/cli_root/zpool_destroy/zpool_destroy_003_neg.ksh \ functional/cli_root/zpool_detach/cleanup.ksh \ functional/cli_root/zpool_detach/setup.ksh \ functional/cli_root/zpool_detach/zpool_detach_001_neg.ksh \ functional/cli_root/zpool_events/cleanup.ksh \ functional/cli_root/zpool_events/setup.ksh \ functional/cli_root/zpool_events/zpool_events_clear.ksh \ functional/cli_root/zpool_events/zpool_events_clear_retained.ksh \ functional/cli_root/zpool_events/zpool_events_cliargs.ksh \ functional/cli_root/zpool_events/zpool_events_duplicates.ksh \ functional/cli_root/zpool_events/zpool_events_errors.ksh \ functional/cli_root/zpool_events/zpool_events_follow.ksh \ functional/cli_root/zpool_events/zpool_events_poolname.ksh \ functional/cli_root/zpool_expand/cleanup.ksh \ functional/cli_root/zpool_expand/setup.ksh \ functional/cli_root/zpool_expand/zpool_expand_001_pos.ksh \ functional/cli_root/zpool_expand/zpool_expand_002_pos.ksh \ functional/cli_root/zpool_expand/zpool_expand_003_neg.ksh \ functional/cli_root/zpool_expand/zpool_expand_004_pos.ksh \ functional/cli_root/zpool_expand/zpool_expand_005_pos.ksh \ functional/cli_root/zpool_export/cleanup.ksh \ functional/cli_root/zpool_export/setup.ksh \ functional/cli_root/zpool_export/zpool_export_001_pos.ksh \ functional/cli_root/zpool_export/zpool_export_002_pos.ksh \ functional/cli_root/zpool_export/zpool_export_003_neg.ksh \ functional/cli_root/zpool_export/zpool_export_004_pos.ksh \ functional/cli_root/zpool_get/cleanup.ksh \ functional/cli_root/zpool_get/setup.ksh \ functional/cli_root/zpool_get/vdev_get_001_pos.ksh \ functional/cli_root/zpool_get/zpool_get_001_pos.ksh \ functional/cli_root/zpool_get/zpool_get_002_pos.ksh \ functional/cli_root/zpool_get/zpool_get_003_pos.ksh \ functional/cli_root/zpool_get/zpool_get_004_neg.ksh \ functional/cli_root/zpool_get/zpool_get_005_pos.ksh \ functional/cli_root/zpool_history/cleanup.ksh \ functional/cli_root/zpool_history/setup.ksh \ functional/cli_root/zpool_history/zpool_history_001_neg.ksh \ functional/cli_root/zpool_history/zpool_history_002_pos.ksh \ functional/cli_root/zpool_import/cleanup.ksh \ functional/cli_root/zpool_import/import_cachefile_device_added.ksh \ functional/cli_root/zpool_import/import_cachefile_device_removed.ksh \ functional/cli_root/zpool_import/import_cachefile_device_replaced.ksh \ functional/cli_root/zpool_import/import_cachefile_mirror_attached.ksh \ functional/cli_root/zpool_import/import_cachefile_mirror_detached.ksh \ functional/cli_root/zpool_import/import_cachefile_paths_changed.ksh \ functional/cli_root/zpool_import/import_cachefile_shared_device.ksh \ functional/cli_root/zpool_import/import_devices_missing.ksh \ functional/cli_root/zpool_import/import_log_missing.ksh \ functional/cli_root/zpool_import/import_paths_changed.ksh \ functional/cli_root/zpool_import/import_rewind_config_changed.ksh \ functional/cli_root/zpool_import/import_rewind_device_replaced.ksh \ functional/cli_root/zpool_import/setup.ksh \ functional/cli_root/zpool_import/zpool_import_001_pos.ksh \ functional/cli_root/zpool_import/zpool_import_002_pos.ksh \ functional/cli_root/zpool_import/zpool_import_003_pos.ksh \ functional/cli_root/zpool_import/zpool_import_004_pos.ksh \ functional/cli_root/zpool_import/zpool_import_005_pos.ksh \ functional/cli_root/zpool_import/zpool_import_006_pos.ksh \ functional/cli_root/zpool_import/zpool_import_007_pos.ksh \ functional/cli_root/zpool_import/zpool_import_008_pos.ksh \ functional/cli_root/zpool_import/zpool_import_009_neg.ksh \ functional/cli_root/zpool_import/zpool_import_010_pos.ksh \ functional/cli_root/zpool_import/zpool_import_011_neg.ksh \ functional/cli_root/zpool_import/zpool_import_012_pos.ksh \ functional/cli_root/zpool_import/zpool_import_013_neg.ksh \ functional/cli_root/zpool_import/zpool_import_014_pos.ksh \ functional/cli_root/zpool_import/zpool_import_015_pos.ksh \ functional/cli_root/zpool_import/zpool_import_016_pos.ksh \ functional/cli_root/zpool_import/zpool_import_017_pos.ksh \ functional/cli_root/zpool_import/zpool_import_all_001_pos.ksh \ functional/cli_root/zpool_import/zpool_import_encrypted.ksh \ functional/cli_root/zpool_import/zpool_import_encrypted_load.ksh \ functional/cli_root/zpool_import/zpool_import_errata3.ksh \ functional/cli_root/zpool_import/zpool_import_errata4.ksh \ functional/cli_root/zpool_import/zpool_import_features_001_pos.ksh \ functional/cli_root/zpool_import/zpool_import_features_002_neg.ksh \ functional/cli_root/zpool_import/zpool_import_features_003_pos.ksh \ functional/cli_root/zpool_import/zpool_import_hostid_changed.ksh \ functional/cli_root/zpool_import/zpool_import_hostid_changed_unclean_export.ksh \ functional/cli_root/zpool_import/zpool_import_hostid_changed_cachefile.ksh \ functional/cli_root/zpool_import/zpool_import_hostid_changed_cachefile_unclean_export.ksh \ functional/cli_root/zpool_import/zpool_import_missing_001_pos.ksh \ functional/cli_root/zpool_import/zpool_import_missing_002_pos.ksh \ functional/cli_root/zpool_import/zpool_import_missing_003_pos.ksh \ functional/cli_root/zpool_import/zpool_import_rename_001_pos.ksh \ functional/cli_root/zpool_import/zpool_import_status.ksh \ + functional/cli_root/zpool_import/zpool_import_parallel_admin.ksh \ + functional/cli_root/zpool_import/zpool_import_parallel_neg.ksh \ + functional/cli_root/zpool_import/zpool_import_parallel_pos.ksh \ functional/cli_root/zpool_initialize/cleanup.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_attach_detach_add_remove.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_fault_export_import_online.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_import_export.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_offline_export_import_online.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_online_offline.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_split.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_start_and_cancel_neg.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_start_and_cancel_pos.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_suspend_resume.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_uninit.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_unsupported_vdevs.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_verify_checksums.ksh \ functional/cli_root/zpool_initialize/zpool_initialize_verify_initialized.ksh \ functional/cli_root/zpool_labelclear/zpool_labelclear_active.ksh \ functional/cli_root/zpool_labelclear/zpool_labelclear_exported.ksh \ functional/cli_root/zpool_labelclear/zpool_labelclear_removed.ksh \ functional/cli_root/zpool_labelclear/zpool_labelclear_valid.ksh \ functional/cli_root/zpool_offline/cleanup.ksh \ functional/cli_root/zpool_offline/setup.ksh \ functional/cli_root/zpool_offline/zpool_offline_001_pos.ksh \ functional/cli_root/zpool_offline/zpool_offline_002_neg.ksh \ functional/cli_root/zpool_offline/zpool_offline_003_pos.ksh \ functional/cli_root/zpool_online/cleanup.ksh \ functional/cli_root/zpool_online/setup.ksh \ functional/cli_root/zpool_online/zpool_online_001_pos.ksh \ functional/cli_root/zpool_online/zpool_online_002_neg.ksh \ functional/cli_root/zpool_remove/cleanup.ksh \ functional/cli_root/zpool_remove/setup.ksh \ functional/cli_root/zpool_remove/zpool_remove_001_neg.ksh \ functional/cli_root/zpool_remove/zpool_remove_002_pos.ksh \ functional/cli_root/zpool_remove/zpool_remove_003_pos.ksh \ functional/cli_root/zpool_reopen/cleanup.ksh \ functional/cli_root/zpool_reopen/setup.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_001_pos.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_002_pos.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_003_pos.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_004_pos.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_005_pos.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_006_neg.ksh \ functional/cli_root/zpool_reopen/zpool_reopen_007_pos.ksh \ functional/cli_root/zpool_replace/cleanup.ksh \ functional/cli_root/zpool_replace/replace-o_ashift.ksh \ functional/cli_root/zpool_replace/replace_prop_ashift.ksh \ functional/cli_root/zpool_replace/setup.ksh \ functional/cli_root/zpool_replace/zpool_replace_001_neg.ksh \ functional/cli_root/zpool_resilver/cleanup.ksh \ functional/cli_root/zpool_resilver/setup.ksh \ functional/cli_root/zpool_resilver/zpool_resilver_bad_args.ksh \ functional/cli_root/zpool_resilver/zpool_resilver_restart.ksh \ functional/cli_root/zpool_resilver/zpool_resilver_concurrent.ksh \ functional/cli_root/zpool_scrub/cleanup.ksh \ functional/cli_root/zpool_scrub/setup.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_001_neg.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_002_pos.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_003_pos.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_004_pos.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_005_pos.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_encrypted_unloaded.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_multiple_copies.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_offline_device.ksh \ functional/cli_root/zpool_scrub/zpool_scrub_print_repairing.ksh \ functional/cli_root/zpool_scrub/zpool_error_scrub_001_pos.ksh \ functional/cli_root/zpool_scrub/zpool_error_scrub_002_pos.ksh \ functional/cli_root/zpool_scrub/zpool_error_scrub_003_pos.ksh \ functional/cli_root/zpool_scrub/zpool_error_scrub_004_pos.ksh \ functional/cli_root/zpool_set/cleanup.ksh \ functional/cli_root/zpool_set/setup.ksh \ functional/cli_root/zpool/setup.ksh \ functional/cli_root/zpool_set/vdev_set_001_pos.ksh \ functional/cli_root/zpool_set/zpool_set_common.kshlib \ functional/cli_root/zpool_set/zpool_set_001_pos.ksh \ functional/cli_root/zpool_set/zpool_set_002_neg.ksh \ functional/cli_root/zpool_set/zpool_set_003_neg.ksh \ functional/cli_root/zpool_set/zpool_set_ashift.ksh \ functional/cli_root/zpool_set/user_property_001_pos.ksh \ functional/cli_root/zpool_set/user_property_002_neg.ksh \ functional/cli_root/zpool_set/zpool_set_features.ksh \ functional/cli_root/zpool_split/cleanup.ksh \ functional/cli_root/zpool_split/setup.ksh \ functional/cli_root/zpool_split/zpool_split_cliargs.ksh \ functional/cli_root/zpool_split/zpool_split_devices.ksh \ functional/cli_root/zpool_split/zpool_split_dryrun_output.ksh \ functional/cli_root/zpool_split/zpool_split_encryption.ksh \ functional/cli_root/zpool_split/zpool_split_indirect.ksh \ functional/cli_root/zpool_split/zpool_split_props.ksh \ functional/cli_root/zpool_split/zpool_split_resilver.ksh \ functional/cli_root/zpool_split/zpool_split_vdevs.ksh \ functional/cli_root/zpool_split/zpool_split_wholedisk.ksh \ functional/cli_root/zpool_status/cleanup.ksh \ functional/cli_root/zpool_status/setup.ksh \ functional/cli_root/zpool_status/zpool_status_001_pos.ksh \ functional/cli_root/zpool_status/zpool_status_002_pos.ksh \ functional/cli_root/zpool_status/zpool_status_003_pos.ksh \ functional/cli_root/zpool_status/zpool_status_004_pos.ksh \ functional/cli_root/zpool_status/zpool_status_005_pos.ksh \ functional/cli_root/zpool_status/zpool_status_006_pos.ksh \ functional/cli_root/zpool_status/zpool_status_007_pos.ksh \ functional/cli_root/zpool_status/zpool_status_008_pos.ksh \ functional/cli_root/zpool_status/zpool_status_features_001_pos.ksh \ functional/cli_root/zpool_sync/cleanup.ksh \ functional/cli_root/zpool_sync/setup.ksh \ functional/cli_root/zpool_sync/zpool_sync_001_pos.ksh \ functional/cli_root/zpool_sync/zpool_sync_002_neg.ksh \ functional/cli_root/zpool_trim/cleanup.ksh \ functional/cli_root/zpool_trim/setup.ksh \ functional/cli_root/zpool_trim/zpool_trim_attach_detach_add_remove.ksh \ functional/cli_root/zpool_trim/zpool_trim_fault_export_import_online.ksh \ functional/cli_root/zpool_trim/zpool_trim_import_export.ksh \ functional/cli_root/zpool_trim/zpool_trim_multiple.ksh \ functional/cli_root/zpool_trim/zpool_trim_neg.ksh \ functional/cli_root/zpool_trim/zpool_trim_offline_export_import_online.ksh \ functional/cli_root/zpool_trim/zpool_trim_online_offline.ksh \ functional/cli_root/zpool_trim/zpool_trim_partial.ksh \ functional/cli_root/zpool_trim/zpool_trim_rate.ksh \ functional/cli_root/zpool_trim/zpool_trim_rate_neg.ksh \ functional/cli_root/zpool_trim/zpool_trim_secure.ksh \ functional/cli_root/zpool_trim/zpool_trim_split.ksh \ functional/cli_root/zpool_trim/zpool_trim_start_and_cancel_neg.ksh \ functional/cli_root/zpool_trim/zpool_trim_start_and_cancel_pos.ksh \ functional/cli_root/zpool_trim/zpool_trim_suspend_resume.ksh \ functional/cli_root/zpool_trim/zpool_trim_unsupported_vdevs.ksh \ functional/cli_root/zpool_trim/zpool_trim_verify_checksums.ksh \ functional/cli_root/zpool_trim/zpool_trim_verify_trimmed.ksh \ functional/cli_root/zpool_upgrade/cleanup.ksh \ functional/cli_root/zpool_upgrade/setup.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_001_pos.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_002_pos.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_003_pos.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_004_pos.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_005_neg.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_006_neg.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_007_pos.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_008_pos.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_009_neg.ksh \ functional/cli_root/zpool_upgrade/zpool_upgrade_features_001_pos.ksh \ functional/cli_root/zpool_wait/cleanup.ksh \ functional/cli_root/zpool_wait/scan/cleanup.ksh \ functional/cli_root/zpool_wait/scan/setup.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_rebuild.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_replace_cancel.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_replace.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_resilver.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_scrub_basic.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_scrub_cancel.ksh \ functional/cli_root/zpool_wait/scan/zpool_wait_scrub_flag.ksh \ functional/cli_root/zpool_wait/setup.ksh \ functional/cli_root/zpool_wait/zpool_wait_discard.ksh \ functional/cli_root/zpool_wait/zpool_wait_freeing.ksh \ functional/cli_root/zpool_wait/zpool_wait_initialize_basic.ksh \ functional/cli_root/zpool_wait/zpool_wait_initialize_cancel.ksh \ functional/cli_root/zpool_wait/zpool_wait_initialize_flag.ksh \ functional/cli_root/zpool_wait/zpool_wait_multiple.ksh \ functional/cli_root/zpool_wait/zpool_wait_no_activity.ksh \ functional/cli_root/zpool_wait/zpool_wait_remove_cancel.ksh \ functional/cli_root/zpool_wait/zpool_wait_remove.ksh \ functional/cli_root/zpool_wait/zpool_wait_trim_basic.ksh \ functional/cli_root/zpool_wait/zpool_wait_trim_cancel.ksh \ functional/cli_root/zpool_wait/zpool_wait_trim_flag.ksh \ functional/cli_root/zpool_wait/zpool_wait_usage.ksh \ functional/cli_root/zpool/zpool_001_neg.ksh \ functional/cli_root/zpool/zpool_002_pos.ksh \ functional/cli_root/zpool/zpool_003_pos.ksh \ functional/cli_root/zpool/zpool_colors.ksh \ functional/cli_user/misc/arcstat_001_pos.ksh \ functional/cli_user/misc/arc_summary_001_pos.ksh \ functional/cli_user/misc/arc_summary_002_neg.ksh \ functional/cli_user/misc/zilstat_001_pos.ksh \ functional/cli_user/misc/cleanup.ksh \ functional/cli_user/misc/setup.ksh \ functional/cli_user/misc/zdb_001_neg.ksh \ functional/cli_user/misc/zfs_001_neg.ksh \ functional/cli_user/misc/zfs_allow_001_neg.ksh \ functional/cli_user/misc/zfs_clone_001_neg.ksh \ functional/cli_user/misc/zfs_create_001_neg.ksh \ functional/cli_user/misc/zfs_destroy_001_neg.ksh \ functional/cli_user/misc/zfs_get_001_neg.ksh \ functional/cli_user/misc/zfs_inherit_001_neg.ksh \ functional/cli_user/misc/zfs_mount_001_neg.ksh \ functional/cli_user/misc/zfs_promote_001_neg.ksh \ functional/cli_user/misc/zfs_receive_001_neg.ksh \ functional/cli_user/misc/zfs_rename_001_neg.ksh \ functional/cli_user/misc/zfs_rollback_001_neg.ksh \ functional/cli_user/misc/zfs_send_001_neg.ksh \ functional/cli_user/misc/zfs_set_001_neg.ksh \ functional/cli_user/misc/zfs_share_001_neg.ksh \ functional/cli_user/misc/zfs_snapshot_001_neg.ksh \ functional/cli_user/misc/zfs_unallow_001_neg.ksh \ functional/cli_user/misc/zfs_unmount_001_neg.ksh \ functional/cli_user/misc/zfs_unshare_001_neg.ksh \ functional/cli_user/misc/zfs_upgrade_001_neg.ksh \ functional/cli_user/misc/zpool_001_neg.ksh \ functional/cli_user/misc/zpool_add_001_neg.ksh \ functional/cli_user/misc/zpool_attach_001_neg.ksh \ functional/cli_user/misc/zpool_clear_001_neg.ksh \ functional/cli_user/misc/zpool_create_001_neg.ksh \ functional/cli_user/misc/zpool_destroy_001_neg.ksh \ functional/cli_user/misc/zpool_detach_001_neg.ksh \ functional/cli_user/misc/zpool_export_001_neg.ksh \ functional/cli_user/misc/zpool_get_001_neg.ksh \ functional/cli_user/misc/zpool_history_001_neg.ksh \ functional/cli_user/misc/zpool_import_001_neg.ksh \ functional/cli_user/misc/zpool_import_002_neg.ksh \ functional/cli_user/misc/zpool_offline_001_neg.ksh \ functional/cli_user/misc/zpool_online_001_neg.ksh \ functional/cli_user/misc/zpool_remove_001_neg.ksh \ functional/cli_user/misc/zpool_replace_001_neg.ksh \ functional/cli_user/misc/zpool_scrub_001_neg.ksh \ functional/cli_user/misc/zpool_set_001_neg.ksh \ functional/cli_user/misc/zpool_status_001_neg.ksh \ functional/cli_user/misc/zpool_upgrade_001_neg.ksh \ functional/cli_user/misc/zpool_wait_privilege.ksh \ functional/cli_user/zfs_list/cleanup.ksh \ functional/cli_user/zfs_list/setup.ksh \ functional/cli_user/zfs_list/zfs_list_001_pos.ksh \ functional/cli_user/zfs_list/zfs_list_002_pos.ksh \ functional/cli_user/zfs_list/zfs_list_003_pos.ksh \ functional/cli_user/zfs_list/zfs_list_004_neg.ksh \ functional/cli_user/zfs_list/zfs_list_005_neg.ksh \ functional/cli_user/zfs_list/zfs_list_007_pos.ksh \ functional/cli_user/zfs_list/zfs_list_008_neg.ksh \ functional/cli_user/zpool_iostat/cleanup.ksh \ functional/cli_user/zpool_iostat/setup.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_001_neg.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_002_pos.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_003_neg.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_004_pos.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_005_pos.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_-c_disable.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_-c_homedir.ksh \ functional/cli_user/zpool_iostat/zpool_iostat_-c_searchpath.ksh \ functional/cli_user/zpool_list/cleanup.ksh \ functional/cli_user/zpool_list/setup.ksh \ functional/cli_user/zpool_list/zpool_list_001_pos.ksh \ functional/cli_user/zpool_list/zpool_list_002_neg.ksh \ functional/cli_user/zpool_status/cleanup.ksh \ functional/cli_user/zpool_status/setup.ksh \ functional/cli_user/zpool_status/zpool_status_003_pos.ksh \ functional/cli_user/zpool_status/zpool_status_-c_disable.ksh \ functional/cli_user/zpool_status/zpool_status_-c_homedir.ksh \ functional/cli_user/zpool_status/zpool_status_-c_searchpath.ksh \ functional/compression/cleanup.ksh \ functional/compression/compress_001_pos.ksh \ functional/compression/compress_002_pos.ksh \ functional/compression/compress_003_pos.ksh \ functional/compression/compress_004_pos.ksh \ functional/compression/compress_zstd_bswap.ksh \ functional/compression/l2arc_compressed_arc_disabled.ksh \ functional/compression/l2arc_compressed_arc.ksh \ functional/compression/l2arc_encrypted.ksh \ functional/compression/l2arc_encrypted_no_compressed_arc.ksh \ functional/compression/setup.ksh \ functional/cp_files/cleanup.ksh \ functional/cp_files/cp_files_001_pos.ksh \ functional/cp_files/cp_files_002_pos.ksh \ functional/cp_files/cp_stress.ksh \ functional/cp_files/setup.ksh \ functional/crtime/cleanup.ksh \ functional/crtime/crtime_001_pos.ksh \ functional/crtime/setup.ksh \ functional/ctime/cleanup.ksh \ functional/ctime/ctime_001_pos.ksh \ functional/ctime/setup.ksh \ functional/deadman/deadman_ratelimit.ksh \ functional/deadman/deadman_sync.ksh \ functional/deadman/deadman_zio.ksh \ functional/delegate/cleanup.ksh \ functional/delegate/setup.ksh \ functional/delegate/zfs_allow_001_pos.ksh \ functional/delegate/zfs_allow_002_pos.ksh \ functional/delegate/zfs_allow_003_pos.ksh \ functional/delegate/zfs_allow_004_pos.ksh \ functional/delegate/zfs_allow_005_pos.ksh \ functional/delegate/zfs_allow_006_pos.ksh \ functional/delegate/zfs_allow_007_pos.ksh \ functional/delegate/zfs_allow_008_pos.ksh \ functional/delegate/zfs_allow_009_neg.ksh \ functional/delegate/zfs_allow_010_pos.ksh \ functional/delegate/zfs_allow_011_neg.ksh \ functional/delegate/zfs_allow_012_neg.ksh \ functional/delegate/zfs_unallow_001_pos.ksh \ functional/delegate/zfs_unallow_002_pos.ksh \ functional/delegate/zfs_unallow_003_pos.ksh \ functional/delegate/zfs_unallow_004_pos.ksh \ functional/delegate/zfs_unallow_005_pos.ksh \ functional/delegate/zfs_unallow_006_pos.ksh \ functional/delegate/zfs_unallow_007_neg.ksh \ functional/delegate/zfs_unallow_008_neg.ksh \ functional/devices/cleanup.ksh \ functional/devices/devices_001_pos.ksh \ functional/devices/devices_002_neg.ksh \ functional/devices/devices_003_pos.ksh \ functional/devices/setup.ksh \ functional/dos_attributes/cleanup.ksh \ functional/dos_attributes/read_dos_attrs_001.ksh \ functional/dos_attributes/setup.ksh \ functional/dos_attributes/write_dos_attrs_001.ksh \ functional/events/cleanup.ksh \ functional/events/events_001_pos.ksh \ functional/events/events_002_pos.ksh \ functional/events/setup.ksh \ functional/events/zed_cksum_config.ksh \ functional/events/zed_cksum_reported.ksh \ functional/events/zed_fd_spill.ksh \ functional/events/zed_io_config.ksh \ functional/events/zed_rc_filter.ksh \ functional/events/zed_slow_io.ksh \ functional/events/zed_slow_io_many_vdevs.ksh \ functional/exec/cleanup.ksh \ functional/exec/exec_001_pos.ksh \ functional/exec/exec_002_neg.ksh \ functional/exec/setup.ksh \ functional/fadvise/cleanup.ksh \ functional/fadvise/fadvise_sequential.ksh \ functional/fadvise/setup.ksh \ functional/fallocate/cleanup.ksh \ functional/fallocate/fallocate_prealloc.ksh \ functional/fallocate/fallocate_punch-hole.ksh \ functional/fallocate/fallocate_zero-range.ksh \ functional/fallocate/setup.ksh \ functional/fault/auto_offline_001_pos.ksh \ functional/fault/auto_online_001_pos.ksh \ functional/fault/auto_online_002_pos.ksh \ functional/fault/auto_replace_001_pos.ksh \ functional/fault/auto_replace_002_pos.ksh \ functional/fault/auto_spare_001_pos.ksh \ functional/fault/auto_spare_002_pos.ksh \ functional/fault/auto_spare_ashift.ksh \ functional/fault/auto_spare_multiple.ksh \ functional/fault/auto_spare_shared.ksh \ functional/fault/cleanup.ksh \ functional/fault/decompress_fault.ksh \ functional/fault/decrypt_fault.ksh \ functional/fault/scrub_after_resilver.ksh \ functional/fault/setup.ksh \ functional/fault/zpool_status_-s.ksh \ functional/features/async_destroy/async_destroy_001_pos.ksh \ functional/features/async_destroy/cleanup.ksh \ functional/features/async_destroy/setup.ksh \ functional/features/large_dnode/cleanup.ksh \ functional/features/large_dnode/large_dnode_001_pos.ksh \ functional/features/large_dnode/large_dnode_002_pos.ksh \ functional/features/large_dnode/large_dnode_003_pos.ksh \ functional/features/large_dnode/large_dnode_004_neg.ksh \ functional/features/large_dnode/large_dnode_005_pos.ksh \ functional/features/large_dnode/large_dnode_006_pos.ksh \ functional/features/large_dnode/large_dnode_007_neg.ksh \ functional/features/large_dnode/large_dnode_008_pos.ksh \ functional/features/large_dnode/large_dnode_009_pos.ksh \ functional/features/large_dnode/setup.ksh \ functional/grow/grow_pool_001_pos.ksh \ functional/grow/grow_replicas_001_pos.ksh \ functional/history/cleanup.ksh \ functional/history/history_001_pos.ksh \ functional/history/history_002_pos.ksh \ functional/history/history_003_pos.ksh \ functional/history/history_004_pos.ksh \ functional/history/history_005_neg.ksh \ functional/history/history_006_neg.ksh \ functional/history/history_007_pos.ksh \ functional/history/history_008_pos.ksh \ functional/history/history_009_pos.ksh \ functional/history/history_010_pos.ksh \ functional/history/setup.ksh \ functional/inheritance/cleanup.ksh \ functional/inheritance/inherit_001_pos.ksh \ functional/inuse/inuse_001_pos.ksh \ functional/inuse/inuse_003_pos.ksh \ functional/inuse/inuse_004_pos.ksh \ functional/inuse/inuse_005_pos.ksh \ functional/inuse/inuse_006_pos.ksh \ functional/inuse/inuse_007_pos.ksh \ functional/inuse/inuse_008_pos.ksh \ functional/inuse/inuse_009_pos.ksh \ functional/inuse/setup.ksh \ functional/io/cleanup.ksh \ functional/io/io_uring.ksh \ functional/io/libaio.ksh \ functional/io/mmap.ksh \ functional/io/posixaio.ksh \ functional/io/psync.ksh \ functional/io/setup.ksh \ functional/io/sync.ksh \ functional/l2arc/cleanup.ksh \ functional/l2arc/l2arc_arcstats_pos.ksh \ functional/l2arc/l2arc_l2miss_pos.ksh \ functional/l2arc/l2arc_mfuonly_pos.ksh \ functional/l2arc/persist_l2arc_001_pos.ksh \ functional/l2arc/persist_l2arc_002_pos.ksh \ functional/l2arc/persist_l2arc_003_neg.ksh \ functional/l2arc/persist_l2arc_004_pos.ksh \ functional/l2arc/persist_l2arc_005_pos.ksh \ functional/l2arc/setup.ksh \ functional/large_files/cleanup.ksh \ functional/large_files/large_files_001_pos.ksh \ functional/large_files/large_files_002_pos.ksh \ functional/large_files/setup.ksh \ functional/largest_pool/largest_pool_001_pos.ksh \ functional/libzfs/cleanup.ksh \ functional/libzfs/libzfs_input.ksh \ functional/libzfs/setup.ksh \ functional/limits/cleanup.ksh \ functional/limits/filesystem_count.ksh \ functional/limits/filesystem_limit.ksh \ functional/limits/setup.ksh \ functional/limits/snapshot_count.ksh \ functional/limits/snapshot_limit.ksh \ functional/link_count/cleanup.ksh \ functional/link_count/link_count_001.ksh \ functional/link_count/link_count_root_inode.ksh \ functional/link_count/setup.ksh \ functional/log_spacemap/log_spacemap_import_logs.ksh \ functional/migration/cleanup.ksh \ functional/migration/migration_001_pos.ksh \ functional/migration/migration_002_pos.ksh \ functional/migration/migration_003_pos.ksh \ functional/migration/migration_004_pos.ksh \ functional/migration/migration_005_pos.ksh \ functional/migration/migration_006_pos.ksh \ functional/migration/migration_007_pos.ksh \ functional/migration/migration_008_pos.ksh \ functional/migration/migration_009_pos.ksh \ functional/migration/migration_010_pos.ksh \ functional/migration/migration_011_pos.ksh \ functional/migration/migration_012_pos.ksh \ functional/migration/setup.ksh \ functional/mmap/cleanup.ksh \ functional/mmap/mmap_libaio_001_pos.ksh \ functional/mmap/mmap_mixed.ksh \ functional/mmap/mmap_read_001_pos.ksh \ functional/mmap/mmap_seek_001_pos.ksh \ functional/mmap/mmap_sync_001_pos.ksh \ functional/mmap/mmap_write_001_pos.ksh \ functional/mmap/setup.ksh \ functional/mmp/cleanup.ksh \ functional/mmp/mmp_active_import.ksh \ functional/mmp/mmp_exported_import.ksh \ functional/mmp/mmp_hostid.ksh \ functional/mmp/mmp_inactive_import.ksh \ functional/mmp/mmp_interval.ksh \ functional/mmp/mmp_on_off.ksh \ functional/mmp/mmp_on_thread.ksh \ functional/mmp/mmp_on_uberblocks.ksh \ functional/mmp/mmp_on_zdb.ksh \ functional/mmp/mmp_reset_interval.ksh \ functional/mmp/mmp_write_distribution.ksh \ functional/mmp/mmp_write_uberblocks.ksh \ functional/mmp/multihost_history.ksh \ functional/mmp/setup.ksh \ functional/mount/cleanup.ksh \ functional/mount/setup.ksh \ functional/mount/umount_001.ksh \ functional/mount/umountall_001.ksh \ functional/mount/umount_unlinked_drain.ksh \ functional/mv_files/cleanup.ksh \ functional/mv_files/mv_files_001_pos.ksh \ functional/mv_files/mv_files_002_pos.ksh \ functional/mv_files/random_creation.ksh \ functional/mv_files/setup.ksh \ functional/nestedfs/cleanup.ksh \ functional/nestedfs/nestedfs_001_pos.ksh \ functional/nestedfs/setup.ksh \ functional/nopwrite/cleanup.ksh \ functional/nopwrite/nopwrite_copies.ksh \ functional/nopwrite/nopwrite_mtime.ksh \ functional/nopwrite/nopwrite_negative.ksh \ functional/nopwrite/nopwrite_promoted_clone.ksh \ functional/nopwrite/nopwrite_recsize.ksh \ functional/nopwrite/nopwrite_sync.ksh \ functional/nopwrite/nopwrite_varying_compression.ksh \ functional/nopwrite/nopwrite_volume.ksh \ functional/nopwrite/setup.ksh \ functional/no_space/cleanup.ksh \ functional/no_space/enospc_001_pos.ksh \ functional/no_space/enospc_002_pos.ksh \ functional/no_space/enospc_003_pos.ksh \ functional/no_space/enospc_df.ksh \ functional/no_space/enospc_ganging.ksh \ functional/no_space/enospc_rm.ksh \ functional/no_space/setup.ksh \ functional/online_offline/cleanup.ksh \ functional/online_offline/online_offline_001_pos.ksh \ functional/online_offline/online_offline_002_neg.ksh \ functional/online_offline/online_offline_003_neg.ksh \ functional/online_offline/setup.ksh \ functional/pam/cleanup.ksh \ functional/pam/pam_basic.ksh \ functional/pam/pam_change_unmounted.ksh \ functional/pam/pam_nounmount.ksh \ functional/pam/pam_recursive.ksh \ functional/pam/pam_short_password.ksh \ functional/pam/setup.ksh \ functional/pool_checkpoint/checkpoint_after_rewind.ksh \ functional/pool_checkpoint/checkpoint_big_rewind.ksh \ functional/pool_checkpoint/checkpoint_capacity.ksh \ functional/pool_checkpoint/checkpoint_conf_change.ksh \ functional/pool_checkpoint/checkpoint_discard_busy.ksh \ functional/pool_checkpoint/checkpoint_discard.ksh \ functional/pool_checkpoint/checkpoint_discard_many.ksh \ functional/pool_checkpoint/checkpoint_indirect.ksh \ functional/pool_checkpoint/checkpoint_invalid.ksh \ functional/pool_checkpoint/checkpoint_lun_expsz.ksh \ functional/pool_checkpoint/checkpoint_open.ksh \ functional/pool_checkpoint/checkpoint_removal.ksh \ functional/pool_checkpoint/checkpoint_rewind.ksh \ functional/pool_checkpoint/checkpoint_ro_rewind.ksh \ functional/pool_checkpoint/checkpoint_sm_scale.ksh \ functional/pool_checkpoint/checkpoint_twice.ksh \ functional/pool_checkpoint/checkpoint_vdev_add.ksh \ functional/pool_checkpoint/checkpoint_zdb.ksh \ functional/pool_checkpoint/checkpoint_zhack_feat.ksh \ functional/pool_checkpoint/cleanup.ksh \ functional/pool_checkpoint/setup.ksh \ functional/pool_names/pool_names_001_pos.ksh \ functional/pool_names/pool_names_002_neg.ksh \ functional/poolversion/cleanup.ksh \ functional/poolversion/poolversion_001_pos.ksh \ functional/poolversion/poolversion_002_pos.ksh \ functional/poolversion/setup.ksh \ functional/privilege/cleanup.ksh \ functional/privilege/privilege_001_pos.ksh \ functional/privilege/privilege_002_pos.ksh \ functional/privilege/setup.ksh \ functional/procfs/cleanup.ksh \ functional/procfs/pool_state.ksh \ functional/procfs/procfs_list_basic.ksh \ functional/procfs/procfs_list_concurrent_readers.ksh \ functional/procfs/procfs_list_stale_read.ksh \ functional/procfs/setup.ksh \ functional/projectquota/cleanup.ksh \ functional/projectquota/projectid_001_pos.ksh \ functional/projectquota/projectid_002_pos.ksh \ functional/projectquota/projectid_003_pos.ksh \ functional/projectquota/projectquota_001_pos.ksh \ functional/projectquota/projectquota_002_pos.ksh \ functional/projectquota/projectquota_003_pos.ksh \ functional/projectquota/projectquota_004_neg.ksh \ functional/projectquota/projectquota_005_pos.ksh \ functional/projectquota/projectquota_006_pos.ksh \ functional/projectquota/projectquota_007_pos.ksh \ functional/projectquota/projectquota_008_pos.ksh \ functional/projectquota/projectquota_009_pos.ksh \ functional/projectquota/projectspace_001_pos.ksh \ functional/projectquota/projectspace_002_pos.ksh \ functional/projectquota/projectspace_003_pos.ksh \ functional/projectquota/projectspace_004_pos.ksh \ functional/projectquota/projecttree_001_pos.ksh \ functional/projectquota/projecttree_002_pos.ksh \ functional/projectquota/projecttree_003_neg.ksh \ functional/projectquota/setup.ksh \ functional/quota/cleanup.ksh \ functional/quota/quota_001_pos.ksh \ functional/quota/quota_002_pos.ksh \ functional/quota/quota_003_pos.ksh \ functional/quota/quota_004_pos.ksh \ functional/quota/quota_005_pos.ksh \ functional/quota/quota_006_neg.ksh \ functional/quota/setup.ksh \ functional/raidz/cleanup.ksh \ functional/raidz/raidz_001_neg.ksh \ functional/raidz/raidz_002_pos.ksh \ functional/raidz/raidz_expand_001_pos.ksh \ functional/raidz/raidz_expand_002_pos.ksh \ functional/raidz/raidz_expand_003_neg.ksh \ functional/raidz/raidz_expand_003_pos.ksh \ functional/raidz/raidz_expand_004_pos.ksh \ functional/raidz/raidz_expand_005_pos.ksh \ functional/raidz/raidz_expand_006_neg.ksh \ functional/raidz/raidz_expand_007_neg.ksh \ functional/raidz/setup.ksh \ functional/redacted_send/cleanup.ksh \ functional/redacted_send/redacted_compressed.ksh \ functional/redacted_send/redacted_contents.ksh \ functional/redacted_send/redacted_deleted.ksh \ functional/redacted_send/redacted_disabled_feature.ksh \ functional/redacted_send/redacted_embedded.ksh \ functional/redacted_send/redacted_holes.ksh \ functional/redacted_send/redacted_incrementals.ksh \ functional/redacted_send/redacted_largeblocks.ksh \ functional/redacted_send/redacted_many_clones.ksh \ functional/redacted_send/redacted_mixed_recsize.ksh \ functional/redacted_send/redacted_mounts.ksh \ functional/redacted_send/redacted_negative.ksh \ functional/redacted_send/redacted_origin.ksh \ functional/redacted_send/redacted_panic.ksh \ functional/redacted_send/redacted_props.ksh \ functional/redacted_send/redacted_resume.ksh \ functional/redacted_send/redacted_size.ksh \ functional/redacted_send/redacted_volume.ksh \ functional/redacted_send/setup.ksh \ functional/redundancy/cleanup.ksh \ functional/redundancy/redundancy_draid1.ksh \ functional/redundancy/redundancy_draid2.ksh \ functional/redundancy/redundancy_draid3.ksh \ functional/redundancy/redundancy_draid_damaged1.ksh \ functional/redundancy/redundancy_draid_damaged2.ksh \ functional/redundancy/redundancy_draid.ksh \ functional/redundancy/redundancy_draid_spare1.ksh \ functional/redundancy/redundancy_draid_spare2.ksh \ functional/redundancy/redundancy_draid_spare3.ksh \ functional/redundancy/redundancy_mirror.ksh \ functional/redundancy/redundancy_raidz1.ksh \ functional/redundancy/redundancy_raidz2.ksh \ functional/redundancy/redundancy_raidz3.ksh \ functional/redundancy/redundancy_raidz.ksh \ functional/redundancy/redundancy_stripe.ksh \ functional/redundancy/setup.ksh \ functional/refquota/cleanup.ksh \ functional/refquota/refquota_001_pos.ksh \ functional/refquota/refquota_002_pos.ksh \ functional/refquota/refquota_003_pos.ksh \ functional/refquota/refquota_004_pos.ksh \ functional/refquota/refquota_005_pos.ksh \ functional/refquota/refquota_006_neg.ksh \ functional/refquota/refquota_007_neg.ksh \ functional/refquota/refquota_008_neg.ksh \ functional/refquota/setup.ksh \ functional/refreserv/cleanup.ksh \ functional/refreserv/refreserv_001_pos.ksh \ functional/refreserv/refreserv_002_pos.ksh \ functional/refreserv/refreserv_003_pos.ksh \ functional/refreserv/refreserv_004_pos.ksh \ functional/refreserv/refreserv_005_pos.ksh \ functional/refreserv/refreserv_multi_raidz.ksh \ functional/refreserv/refreserv_raidz.ksh \ functional/refreserv/setup.ksh \ functional/removal/cleanup.ksh \ functional/removal/removal_all_vdev.ksh \ functional/removal/removal_cancel.ksh \ functional/removal/removal_check_space.ksh \ functional/removal/removal_condense_export.ksh \ functional/removal/removal_multiple_indirection.ksh \ functional/removal/removal_nopwrite.ksh \ functional/removal/removal_remap_deadlists.ksh \ functional/removal/removal_reservation.ksh \ functional/removal/removal_resume_export.ksh \ functional/removal/removal_sanity.ksh \ functional/removal/removal_with_add.ksh \ functional/removal/removal_with_create_fs.ksh \ functional/removal/removal_with_dedup.ksh \ functional/removal/removal_with_errors.ksh \ functional/removal/removal_with_export.ksh \ functional/removal/removal_with_faulted.ksh \ functional/removal/removal_with_ganging.ksh \ functional/removal/removal_with_indirect.ksh \ functional/removal/removal_with_remove.ksh \ functional/removal/removal_with_scrub.ksh \ functional/removal/removal_with_send.ksh \ functional/removal/removal_with_send_recv.ksh \ functional/removal/removal_with_snapshot.ksh \ functional/removal/removal_with_write.ksh \ functional/removal/removal_with_zdb.ksh \ functional/removal/remove_attach_mirror.ksh \ functional/removal/remove_expanded.ksh \ functional/removal/remove_indirect.ksh \ functional/removal/remove_mirror.ksh \ functional/removal/remove_mirror_sanity.ksh \ functional/removal/remove_raidz.ksh \ functional/rename_dirs/cleanup.ksh \ functional/rename_dirs/rename_dirs_001_pos.ksh \ functional/rename_dirs/setup.ksh \ functional/renameat2/cleanup.ksh \ functional/renameat2/setup.ksh \ functional/renameat2/renameat2_exchange.ksh \ functional/renameat2/renameat2_noreplace.ksh \ functional/renameat2/renameat2_whiteout.ksh \ functional/replacement/attach_import.ksh \ functional/replacement/attach_multiple.ksh \ functional/replacement/attach_rebuild.ksh \ functional/replacement/attach_resilver.ksh \ functional/replacement/cleanup.ksh \ functional/replacement/detach.ksh \ functional/replacement/rebuild_disabled_feature.ksh \ functional/replacement/rebuild_multiple.ksh \ functional/replacement/rebuild_raidz.ksh \ functional/replacement/replace_import.ksh \ functional/replacement/replace_rebuild.ksh \ functional/replacement/replace_resilver.ksh \ functional/replacement/resilver_restart_001.ksh \ functional/replacement/resilver_restart_002.ksh \ functional/replacement/scrub_cancel.ksh \ functional/replacement/setup.ksh \ functional/reservation/cleanup.ksh \ functional/reservation/reservation_001_pos.ksh \ functional/reservation/reservation_002_pos.ksh \ functional/reservation/reservation_003_pos.ksh \ functional/reservation/reservation_004_pos.ksh \ functional/reservation/reservation_005_pos.ksh \ functional/reservation/reservation_006_pos.ksh \ functional/reservation/reservation_007_pos.ksh \ functional/reservation/reservation_008_pos.ksh \ functional/reservation/reservation_009_pos.ksh \ functional/reservation/reservation_010_pos.ksh \ functional/reservation/reservation_011_pos.ksh \ functional/reservation/reservation_012_pos.ksh \ functional/reservation/reservation_013_pos.ksh \ functional/reservation/reservation_014_pos.ksh \ functional/reservation/reservation_015_pos.ksh \ functional/reservation/reservation_016_pos.ksh \ functional/reservation/reservation_017_pos.ksh \ functional/reservation/reservation_018_pos.ksh \ functional/reservation/reservation_019_pos.ksh \ functional/reservation/reservation_020_pos.ksh \ functional/reservation/reservation_021_neg.ksh \ functional/reservation/reservation_022_pos.ksh \ functional/reservation/setup.ksh \ functional/rootpool/cleanup.ksh \ functional/rootpool/rootpool_002_neg.ksh \ functional/rootpool/rootpool_003_neg.ksh \ functional/rootpool/rootpool_007_pos.ksh \ functional/rootpool/setup.ksh \ functional/rsend/cleanup.ksh \ functional/rsend/recv_dedup_encrypted_zvol.ksh \ functional/rsend/recv_dedup.ksh \ functional/rsend/rsend_001_pos.ksh \ functional/rsend/rsend_002_pos.ksh \ functional/rsend/rsend_003_pos.ksh \ functional/rsend/rsend_004_pos.ksh \ functional/rsend/rsend_005_pos.ksh \ functional/rsend/rsend_006_pos.ksh \ functional/rsend/rsend_007_pos.ksh \ functional/rsend/rsend_008_pos.ksh \ functional/rsend/rsend_009_pos.ksh \ functional/rsend/rsend_010_pos.ksh \ functional/rsend/rsend_011_pos.ksh \ functional/rsend/rsend_012_pos.ksh \ functional/rsend/rsend_013_pos.ksh \ functional/rsend/rsend_014_pos.ksh \ functional/rsend/rsend_016_neg.ksh \ functional/rsend/rsend_019_pos.ksh \ functional/rsend/rsend_020_pos.ksh \ functional/rsend/rsend_021_pos.ksh \ functional/rsend/rsend_022_pos.ksh \ functional/rsend/rsend_024_pos.ksh \ functional/rsend/rsend_025_pos.ksh \ functional/rsend/rsend_026_neg.ksh \ functional/rsend/rsend_027_pos.ksh \ functional/rsend/rsend_028_neg.ksh \ functional/rsend/rsend_029_neg.ksh \ functional/rsend/rsend_030_pos.ksh \ functional/rsend/rsend_031_pos.ksh \ functional/rsend/send-c_embedded_blocks.ksh \ functional/rsend/send-c_incremental.ksh \ functional/rsend/send-c_lz4_disabled.ksh \ functional/rsend/send-c_mixed_compression.ksh \ functional/rsend/send-c_props.ksh \ functional/rsend/send-c_recv_dedup.ksh \ functional/rsend/send-c_recv_lz4_disabled.ksh \ functional/rsend/send-c_resume.ksh \ functional/rsend/send-c_stream_size_estimate.ksh \ functional/rsend/send-c_verify_contents.ksh \ functional/rsend/send-c_verify_ratio.ksh \ functional/rsend/send-c_volume.ksh \ functional/rsend/send-c_zstream_recompress.ksh \ functional/rsend/send-c_zstreamdump.ksh \ functional/rsend/send-cpL_varied_recsize.ksh \ functional/rsend/send_doall.ksh \ functional/rsend/send_encrypted_incremental.ksh \ functional/rsend/send_encrypted_files.ksh \ functional/rsend/send_encrypted_freeobjects.ksh \ functional/rsend/send_encrypted_hierarchy.ksh \ functional/rsend/send_encrypted_props.ksh \ functional/rsend/send_encrypted_truncated_files.ksh \ functional/rsend/send_freeobjects.ksh \ functional/rsend/send_holds.ksh \ functional/rsend/send_hole_birth.ksh \ functional/rsend/send_invalid.ksh \ functional/rsend/send-L_toggle.ksh \ functional/rsend/send_mixed_raw.ksh \ functional/rsend/send_partial_dataset.ksh \ functional/rsend/send_raw_ashift.ksh \ functional/rsend/send_raw_spill_block.ksh \ functional/rsend/send_raw_large_blocks.ksh \ functional/rsend/send_realloc_dnode_size.ksh \ functional/rsend/send_realloc_encrypted_files.ksh \ functional/rsend/send_realloc_files.ksh \ functional/rsend/send_spill_block.ksh \ functional/rsend/send-wR_encrypted_zvol.ksh \ functional/rsend/setup.ksh \ functional/scrub_mirror/cleanup.ksh \ functional/scrub_mirror/scrub_mirror_001_pos.ksh \ functional/scrub_mirror/scrub_mirror_002_pos.ksh \ functional/scrub_mirror/scrub_mirror_003_pos.ksh \ functional/scrub_mirror/scrub_mirror_004_pos.ksh \ functional/scrub_mirror/setup.ksh \ functional/slog/cleanup.ksh \ functional/slog/setup.ksh \ functional/slog/slog_001_pos.ksh \ functional/slog/slog_002_pos.ksh \ functional/slog/slog_003_pos.ksh \ functional/slog/slog_004_pos.ksh \ functional/slog/slog_005_pos.ksh \ functional/slog/slog_006_pos.ksh \ functional/slog/slog_007_pos.ksh \ functional/slog/slog_008_neg.ksh \ functional/slog/slog_009_neg.ksh \ functional/slog/slog_010_neg.ksh \ functional/slog/slog_011_neg.ksh \ functional/slog/slog_012_neg.ksh \ functional/slog/slog_013_pos.ksh \ functional/slog/slog_014_pos.ksh \ functional/slog/slog_015_neg.ksh \ functional/slog/slog_016_pos.ksh \ functional/slog/slog_replay_fs_001.ksh \ functional/slog/slog_replay_fs_002.ksh \ functional/slog/slog_replay_volume.ksh \ functional/snapshot/cleanup.ksh \ functional/snapshot/clone_001_pos.ksh \ functional/snapshot/rollback_001_pos.ksh \ functional/snapshot/rollback_002_pos.ksh \ functional/snapshot/rollback_003_pos.ksh \ functional/snapshot/setup.ksh \ functional/snapshot/snapshot_001_pos.ksh \ functional/snapshot/snapshot_002_pos.ksh \ functional/snapshot/snapshot_003_pos.ksh \ functional/snapshot/snapshot_004_pos.ksh \ functional/snapshot/snapshot_005_pos.ksh \ functional/snapshot/snapshot_006_pos.ksh \ functional/snapshot/snapshot_007_pos.ksh \ functional/snapshot/snapshot_008_pos.ksh \ functional/snapshot/snapshot_009_pos.ksh \ functional/snapshot/snapshot_010_pos.ksh \ functional/snapshot/snapshot_011_pos.ksh \ functional/snapshot/snapshot_012_pos.ksh \ functional/snapshot/snapshot_013_pos.ksh \ functional/snapshot/snapshot_014_pos.ksh \ functional/snapshot/snapshot_015_pos.ksh \ functional/snapshot/snapshot_016_pos.ksh \ functional/snapshot/snapshot_017_pos.ksh \ functional/snapshot/snapshot_018_pos.ksh \ functional/snapused/cleanup.ksh \ functional/snapused/setup.ksh \ functional/snapused/snapused_001_pos.ksh \ functional/snapused/snapused_002_pos.ksh \ functional/snapused/snapused_003_pos.ksh \ functional/snapused/snapused_004_pos.ksh \ functional/snapused/snapused_005_pos.ksh \ functional/sparse/cleanup.ksh \ functional/sparse/setup.ksh \ functional/sparse/sparse_001_pos.ksh \ functional/stat/cleanup.ksh \ functional/stat/setup.ksh \ functional/stat/stat_001_pos.ksh \ functional/suid/cleanup.ksh \ functional/suid/setup.ksh \ functional/suid/suid_write_to_none.ksh \ functional/suid/suid_write_to_sgid.ksh \ functional/suid/suid_write_to_suid.ksh \ functional/suid/suid_write_to_suid_sgid.ksh \ functional/suid/suid_write_zil_replay.ksh \ functional/trim/autotrim_config.ksh \ functional/trim/autotrim_integrity.ksh \ functional/trim/autotrim_trim_integrity.ksh \ functional/trim/cleanup.ksh \ functional/trim/setup.ksh \ functional/trim/trim_config.ksh \ functional/trim/trim_integrity.ksh \ functional/trim/trim_l2arc.ksh \ functional/truncate/cleanup.ksh \ functional/truncate/setup.ksh \ functional/truncate/truncate_001_pos.ksh \ functional/truncate/truncate_002_pos.ksh \ functional/truncate/truncate_timestamps.ksh \ functional/upgrade/cleanup.ksh \ functional/upgrade/setup.ksh \ functional/upgrade/upgrade_projectquota_001_pos.ksh \ functional/upgrade/upgrade_readonly_pool.ksh \ functional/upgrade/upgrade_userobj_001_pos.ksh \ functional/user_namespace/cleanup.ksh \ functional/user_namespace/setup.ksh \ functional/user_namespace/user_namespace_001.ksh \ functional/user_namespace/user_namespace_002.ksh \ functional/user_namespace/user_namespace_003.ksh \ functional/user_namespace/user_namespace_004.ksh \ functional/userquota/cleanup.ksh \ functional/userquota/groupspace_001_pos.ksh \ functional/userquota/groupspace_002_pos.ksh \ functional/userquota/groupspace_003_pos.ksh \ functional/userquota/setup.ksh \ functional/userquota/userquota_001_pos.ksh \ functional/userquota/userquota_002_pos.ksh \ functional/userquota/userquota_003_pos.ksh \ functional/userquota/userquota_004_pos.ksh \ functional/userquota/userquota_005_neg.ksh \ functional/userquota/userquota_006_pos.ksh \ functional/userquota/userquota_007_pos.ksh \ functional/userquota/userquota_008_pos.ksh \ functional/userquota/userquota_009_pos.ksh \ functional/userquota/userquota_010_pos.ksh \ functional/userquota/userquota_011_pos.ksh \ functional/userquota/userquota_012_neg.ksh \ functional/userquota/userquota_013_pos.ksh \ functional/userquota/userspace_001_pos.ksh \ functional/userquota/userspace_002_pos.ksh \ functional/userquota/userspace_003_pos.ksh \ functional/userquota/userspace_encrypted.ksh \ functional/userquota/userspace_send_encrypted.ksh \ functional/userquota/userspace_encrypted_13709.ksh \ functional/vdev_zaps/cleanup.ksh \ functional/vdev_zaps/setup.ksh \ functional/vdev_zaps/vdev_zaps_001_pos.ksh \ functional/vdev_zaps/vdev_zaps_002_pos.ksh \ functional/vdev_zaps/vdev_zaps_003_pos.ksh \ functional/vdev_zaps/vdev_zaps_004_pos.ksh \ functional/vdev_zaps/vdev_zaps_005_pos.ksh \ functional/vdev_zaps/vdev_zaps_006_pos.ksh \ functional/vdev_zaps/vdev_zaps_007_pos.ksh \ functional/write_dirs/cleanup.ksh \ functional/write_dirs/setup.ksh \ functional/write_dirs/write_dirs_001_pos.ksh \ functional/write_dirs/write_dirs_002_pos.ksh \ functional/xattr/cleanup.ksh \ functional/xattr/setup.ksh \ functional/xattr/xattr_001_pos.ksh \ functional/xattr/xattr_002_neg.ksh \ functional/xattr/xattr_003_neg.ksh \ functional/xattr/xattr_004_pos.ksh \ functional/xattr/xattr_005_pos.ksh \ functional/xattr/xattr_006_pos.ksh \ functional/xattr/xattr_007_neg.ksh \ functional/xattr/xattr_008_pos.ksh \ functional/xattr/xattr_009_neg.ksh \ functional/xattr/xattr_010_neg.ksh \ functional/xattr/xattr_011_pos.ksh \ functional/xattr/xattr_012_pos.ksh \ functional/xattr/xattr_013_pos.ksh \ functional/xattr/xattr_compat.ksh \ functional/zpool_influxdb/cleanup.ksh \ functional/zpool_influxdb/setup.ksh \ functional/zpool_influxdb/zpool_influxdb.ksh \ functional/zvol/zvol_cli/cleanup.ksh \ functional/zvol/zvol_cli/setup.ksh \ functional/zvol/zvol_cli/zvol_cli_001_pos.ksh \ functional/zvol/zvol_cli/zvol_cli_002_pos.ksh \ functional/zvol/zvol_cli/zvol_cli_003_neg.ksh \ functional/zvol/zvol_ENOSPC/cleanup.ksh \ functional/zvol/zvol_ENOSPC/setup.ksh \ functional/zvol/zvol_ENOSPC/zvol_ENOSPC_001_pos.ksh \ functional/zvol/zvol_misc/cleanup.ksh \ functional/zvol/zvol_misc/setup.ksh \ functional/zvol/zvol_misc/zvol_misc_001_neg.ksh \ functional/zvol/zvol_misc/zvol_misc_002_pos.ksh \ functional/zvol/zvol_misc/zvol_misc_003_neg.ksh \ functional/zvol/zvol_misc/zvol_misc_004_pos.ksh \ functional/zvol/zvol_misc/zvol_misc_005_neg.ksh \ functional/zvol/zvol_misc/zvol_misc_006_pos.ksh \ functional/zvol/zvol_misc/zvol_misc_fua.ksh \ functional/zvol/zvol_misc/zvol_misc_hierarchy.ksh \ functional/zvol/zvol_misc/zvol_misc_rename_inuse.ksh \ functional/zvol/zvol_misc/zvol_misc_snapdev.ksh \ functional/zvol/zvol_misc/zvol_misc_trim.ksh \ functional/zvol/zvol_misc/zvol_misc_volmode.ksh \ functional/zvol/zvol_misc/zvol_misc_zil.ksh \ functional/zvol/zvol_stress/cleanup.ksh \ functional/zvol/zvol_stress/setup.ksh \ functional/zvol/zvol_stress/zvol_stress.ksh \ functional/zvol/zvol_swap/cleanup.ksh \ functional/zvol/zvol_swap/setup.ksh \ functional/zvol/zvol_swap/zvol_swap_001_pos.ksh \ functional/zvol/zvol_swap/zvol_swap_002_pos.ksh \ functional/zvol/zvol_swap/zvol_swap_003_pos.ksh \ functional/zvol/zvol_swap/zvol_swap_004_pos.ksh \ functional/zvol/zvol_swap/zvol_swap_005_pos.ksh \ functional/zvol/zvol_swap/zvol_swap_006_pos.ksh \ functional/idmap_mount/cleanup.ksh \ functional/idmap_mount/setup.ksh \ functional/idmap_mount/idmap_mount_001.ksh \ functional/idmap_mount/idmap_mount_002.ksh \ functional/idmap_mount/idmap_mount_003.ksh \ functional/idmap_mount/idmap_mount_004.ksh \ functional/idmap_mount/idmap_mount_005.ksh diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_admin.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_admin.ksh new file mode 100755 index 000000000000..c681d1b7dd23 --- /dev/null +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_admin.ksh @@ -0,0 +1,165 @@ +#!/bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or https://opensource.org/licenses/CDDL-1.0. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright 2007 Sun Microsystems, Inc. All rights reserved. +# Use is subject to license terms. +# + +# +# Copyright (c) 2023 Klara, Inc. +# + +. $STF_SUITE/include/libtest.shlib +. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.cfg +. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.kshlib + +# +# DESCRIPTION: +# Verify that admin commands to different pool are not blocked by import +# +# STRATEGY: +# 1. Create 2 pools +# 2. Export one of the pools +# 4. Import the pool with an injected delay +# 5. Execute some admin commands against both pools +# 6. Verify that the admin commands to the non-imported pool don't stall +# + +verify_runnable "global" + +function cleanup +{ + zinject -c all + destroy_pool $TESTPOOL1 + destroy_pool $TESTPOOL2 +} + +function pool_import +{ + typeset dir=$1 + typeset pool=$2 + + SECONDS=0 + errmsg=$(zpool import -d $dir -f $pool 2>&1 > /dev/null) + if [[ $? -eq 0 ]]; then + echo ${pool}: imported in $SECONDS secs + echo $SECONDS > ${DEVICE_DIR}/${pool}-import + else + echo ${pool}: import failed $errmsg in $SECONDS secs + fi +} + +function pool_add_device +{ + typeset pool=$1 + typeset device=$2 + typeset devtype=$3 + + SECONDS=0 + errmsg=$(zpool add $pool $devtype $device 2>&1 > /dev/null) + if [[ $? -eq 0 ]]; then + echo ${pool}: added $devtype vdev in $SECONDS secs + echo $SECONDS > ${DEVICE_DIR}/${pool}-add + else + echo ${pool}: add $devtype vdev failed ${errmsg}, in $SECONDS secs + fi +} + +function pool_stats +{ + typeset stats=$1 + typeset pool=$2 + + SECONDS=0 + errmsg=$(zpool $stats $pool 2>&1 > /dev/null) + if [[ $? -eq 0 ]]; then + echo ${pool}: $stats in $SECONDS secs + echo $SECONDS > ${DEVICE_DIR}/${pool}-${stats} + else + echo ${pool}: $stats failed ${errmsg}, in $SECONDS secs + fi +} + +function pool_create +{ + typeset pool=$1 + typeset device=$2 + + SECONDS=0 + errmsg=$(zpool create $pool $device 2>&1 > /dev/null) + if [[ $? -eq 0 ]]; then + echo ${pool}: created in $SECONDS secs + echo $SECONDS > ${DEVICE_DIR}/${pool}-create + else + echo ${pool}: create failed ${errmsg}, in $SECONDS secs + fi +} + +log_assert "Simple admin commands to different pool not blocked by import" + +log_onexit cleanup + +# +# create two pools and export one +# +log_must zpool create $TESTPOOL1 $VDEV0 +log_must zpool export $TESTPOOL1 +log_must zpool create $TESTPOOL2 $VDEV1 + +# +# import pool asyncronously with an injected 10 second delay +# +log_must zinject -P import -s 10 $TESTPOOL1 +pool_import $DEVICE_DIR $TESTPOOL1 & + +sleep 2 + +# +# run some admin commands on the pools while the import is in progress +# + +pool_add_device $TESTPOOL1 $VDEV2 "log" & +pool_add_device $TESTPOOL2 $VDEV3 "cache" & +pool_stats "status" $TESTPOOL1 & +pool_stats "status" $TESTPOOL2 & +pool_stats "list" $TESTPOOL1 & +pool_stats "list" $TESTPOOL2 & +pool_create $TESTPOOL1 $VDEV4 & +wait + +log_must zpool sync $TESTPOOL1 $TESTPOOL2 + +zpool history $TESTPOOL1 +zpool history $TESTPOOL2 + +log_must test "5" -lt $(<${DEVICE_DIR}/${TESTPOOL1}-import) + +# +# verify that commands to second pool did not wait for import to finish +# +log_must test "2" -gt $(<${DEVICE_DIR}/${TESTPOOL2}-status) +log_must test "2" -gt $(<${DEVICE_DIR}/${TESTPOOL2}-list) +log_must test "2" -gt $(<${DEVICE_DIR}/${TESTPOOL2}-add) +[[ -e ${DEVICE_DIR}/${TESTPOOL1}-create ]] && log_fail "unexpected pool create" + +log_pass "Simple admin commands to different pool not blocked by import" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_neg.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_neg.ksh new file mode 100755 index 000000000000..339dc2575ede --- /dev/null +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_neg.ksh @@ -0,0 +1,130 @@ +#!/bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or https://opensource.org/licenses/CDDL-1.0. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright 2007 Sun Microsystems, Inc. All rights reserved. +# Use is subject to license terms. +# + +# +# Copyright (c) 2023 Klara, Inc. +# + +. $STF_SUITE/include/libtest.shlib +. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.cfg +. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.kshlib + +# +# DESCRIPTION: +# Verify that pool imports by same name only have one winner +# +# STRATEGY: +# 1. Create 4 single disk pools with the same name +# 2. Generate some ZIL records (for a longer import) +# 3. Export the pools +# 4. Import the pools in parallel +# 5. Repeat with using matching guids +# + +verify_runnable "global" + +POOLNAME="import_pool" +DEV_DIR_PREFIX="$DEVICE_DIR/$POOLNAME" +VDEVSIZE=$((512 * 1024 * 1024)) + +log_assert "parallel pool imports by same name only have one winner" + +# each pool has its own device directory +for i in {0..3}; do + log_must mkdir -p ${DEV_DIR_PREFIX}$i + log_must truncate -s $VDEVSIZE ${DEV_DIR_PREFIX}$i/${DEVICE_FILE}$i +done + +function cleanup +{ + zinject -c all + log_must set_tunable64 KEEP_LOG_SPACEMAPS_AT_EXPORT 0 + log_must set_tunable64 METASLAB_DEBUG_LOAD 0 + + destroy_pool $POOLNAME + + log_must rm -rf $DEV_DIR_PREFIX* +} + +log_onexit cleanup + +log_must set_tunable64 KEEP_LOG_SPACEMAPS_AT_EXPORT 1 +log_must set_tunable64 METASLAB_DEBUG_LOAD 1 + +function import_pool +{ + typeset dir=$1 + typeset pool=$2 + typeset newname=$3 + + SECONDS=0 + errmsg=$(zpool import -N -d $dir -f $pool $newname 2>&1 > /dev/null) + if [[ $? -eq 0 ]]; then + touch $dir/imported + echo "imported $pool in $SECONDS secs" + elif [[ $errmsg == *"cannot import"* ]]; then + echo "pool import failed: $errmsg, waited $SECONDS secs" + touch $dir/failed + fi +} + +# +# create four exported pools with the same name +# +for i in {0..3}; do + log_must zpool create $POOLNAME ${DEV_DIR_PREFIX}$i/${DEVICE_FILE}$i + log_must zpool export $POOLNAME +done +log_must zinject -P import -s 10 $POOLNAME + +# +# import the pools in parallel, expecting only one winner +# +for i in {0..3}; do + import_pool ${DEV_DIR_PREFIX}$i $POOLNAME & +done +wait + +# check the result of background imports +typeset num_imports=0 +typeset num_cannot=0 +for i in {0..3}; do + if [[ -f ${DEV_DIR_PREFIX}$i/imported ]]; then + ((num_imports += 1)) + fi + if [[ -f ${DEV_DIR_PREFIX}$i/failed ]]; then + ((num_cannot += 1)) + loser=$i + fi +done +[[ $num_imports -eq "1" ]] || log_fail "expecting an import" +[[ $num_cannot -eq "3" ]] || \ + log_fail "expecting 3 pool exists errors, found $num_cannot" + +log_note "$num_imports imported and $num_cannot failed (expected)" + +log_pass "parallel pool imports by same name only have one winner" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_pos.ksh new file mode 100755 index 000000000000..71b2437a37ec --- /dev/null +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_pos.ksh @@ -0,0 +1,137 @@ +#!/bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or https://opensource.org/licenses/CDDL-1.0. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright 2007 Sun Microsystems, Inc. All rights reserved. +# Use is subject to license terms. +# + +# +# Copyright (c) 2023 Klara, Inc. +# + +. $STF_SUITE/include/libtest.shlib +. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.cfg +. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.kshlib + +# test uses 8 vdevs +export MAX_NUM=8 + +# +# DESCRIPTION: +# Verify that pool imports can occur in parallel +# +# STRATEGY: +# 1. Create 8 pools +# 2. Generate some ZIL records +# 3. Export the pools +# 4. Import half of the pools synchronously to baseline sequential cost +# 5. Import the other half asynchronously to demonstrate parallel savings +# 6. Export 4 pools +# 7. Test zpool import -a +# + +verify_runnable "global" + +# +# override the minimum sized vdevs +# +VDEVSIZE=$((512 * 1024 * 1024)) +increase_device_sizes $VDEVSIZE + +POOLNAME="import_pool" + +function cleanup +{ + zinject -c all + log_must set_tunable64 KEEP_LOG_SPACEMAPS_AT_EXPORT 0 + log_must set_tunable64 METASLAB_DEBUG_LOAD 0 + + for i in {0..$(($MAX_NUM - 1))}; do + destroy_pool $POOLNAME-$i + done + # reset the devices + increase_device_sizes 0 + increase_device_sizes $FILE_SIZE +} + +log_assert "Pool imports can occur in parallel" + +log_onexit cleanup + +log_must set_tunable64 KEEP_LOG_SPACEMAPS_AT_EXPORT 1 +log_must set_tunable64 METASLAB_DEBUG_LOAD 1 + + +# +# create some exported pools with import delay injectors +# +for i in {0..$(($MAX_NUM - 1))}; do + log_must zpool create $POOLNAME-$i $DEVICE_DIR/${DEVICE_FILE}$i + log_must zpool export $POOLNAME-$i + log_must zinject -P import -s 12 $POOLNAME-$i +done +wait + +# +# import half of the pools synchronously +# +SECONDS=0 +for i in {0..3}; do + log_must zpool import -d $DEVICE_DIR -f $POOLNAME-$i +done +sequential_time=$SECONDS +log_note "sequentially imported 4 pools in $sequential_time seconds" + +# +# import half of the pools in parallel +# +SECONDS=0 +for i in {4..7}; do + log_must zpool import -d $DEVICE_DIR -f $POOLNAME-$i & +done +wait +parallel_time=$SECONDS +log_note "asyncronously imported 4 pools in $parallel_time seconds" + +log_must test $parallel_time -lt $(($sequential_time / 3)) + +# +# export pools with import delay injectors +# +for i in {4..7}; do + log_must zpool export $POOLNAME-$i + log_must zinject -P import -s 12 $POOLNAME-$i +done +wait + +# +# now test zpool import -a +# +SECONDS=0 +log_must zpool import -a -d $DEVICE_DIR -f +parallel_time=$SECONDS +log_note "asyncronously imported 4 pools in $parallel_time seconds" + +log_must test $parallel_time -lt $(($sequential_time / 3)) + +log_pass "Pool imports occur in parallel" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota.kshlib b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota.kshlib index 5083415c8956..b4cfde020f3f 100644 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota.kshlib +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota.kshlib @@ -1,97 +1,104 @@ # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/tests/functional/quota/quota.cfg # BLOCK_SIZE, QUOTA_VALUE and TOLERANCE set in quota.cfg if is_linux; then readonly EDQUOT=122 elif is_freebsd; then readonly EDQUOT=69 else readonly EDQUOT=49 fi # # Function to fill the quota of a zfs filesystem # # $1 - The File system or container to fill. # $2 - The mountpoint to use. # function fill_quota { typeset FILESYSTEM="$1" typeset MNTPT="$2" log_must zfs set quota=$QUOTA_VALUE $FILESYSTEM typeset -i write_size=0 (( write_size = 2 * QUOTA_VALUE )) typeset -i zret=0 file_write -o create -f $MNTPT/$TESTFILE1 -b $BLOCK_SIZE \ -c $write_size -d 0 zret=$? [[ $zret -ne $EDQUOT ]] && \ log_fail "Returned error code: $zret. Expected: $EDQUOT." typeset -i file_size=`ls -lsk $MNTPT/$TESTFILE1 | awk '{ print $1 }'` typeset -i limit=0 (( file_size = file_size * 1024 )) (( limit = QUOTA_VALUE + TOLERANCE )) (( file_size > limit )) && \ log_fail "File was created larger than the quota value, aborting!!!" return 0 } # # Function attempts to write another file in a ZFS filesystem # that has already filled its quota # function exceed_quota { typeset FILESYSTEM="$1" typeset MNTPT="$2" log_must fill_quota $FILESYSTEM $MNTPT typeset -i write_size=0 (( write_size = 2 * QUOTA_VALUE )) typeset -i zret=0 # # Writing a file without API to access return code # log_note "Creating a file in a FS that has already exceeded its quota" file_write -o create -f $MNTPT/$TESTFILE2 \ -b $BLOCK_SIZE -c $write_size -d 0 zret=$? [[ $zret -ne $EDQUOT ]] && \ log_fail "Returned error code: $zret. Expected: $EDQUOT." return 0 } + +function reset_quota +{ + typeset FILESYSTEM="$1" + + log_must zfs set quota=none $FILESYSTEM +} diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_001_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_001_pos.ksh index d124cb26ae98..f01008a46bb1 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_001_pos.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_001_pos.ksh @@ -1,77 +1,79 @@ #! /bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/quota/quota.kshlib # # DESCRIPTION: # # A ZFS file system quota limits the amount of pool space # available to a file system. Apply a quota and verify # that no more file creates are permitted. # # STRATEGY: # 1) Apply quota to ZFS file system # 2) Create a file which is larger than the set quota # 3) Verify that the resulting file size is less than the quota limit # verify_runnable "both" log_assert "Verify that file size is limited by the file system quota" # # cleanup to be used internally as otherwise quota assertions cannot be # run independently or out of order # function cleanup { [[ -e $TESTDIR/$TESTFILE1 ]] && \ log_must rm $TESTDIR/$TESTFILE1 # # Need to allow time for space to be released back to # pool, otherwise next test will fail trying to set a # quota which is less than the space used. # wait_freeing $TESTPOOL sync_pool $TESTPOOL + + reset_quota $TESTPOOL/$TESTFS } log_onexit cleanup # # Sets the quota value and attempts to fill it with a file # twice the size of the quota # log_must fill_quota $TESTPOOL/$TESTFS $TESTDIR log_pass "File size limited by quota" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_002_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_002_pos.ksh index 3af005e874e9..bea2a5a68691 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_002_pos.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_002_pos.ksh @@ -1,76 +1,78 @@ #! /bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/quota/quota.kshlib # # DESCRIPTION: # A zfs file system quota limits the amount of pool space # available to a given ZFS file system. Once exceeded, it is impossible # to write any more files to the file system. # # STRATEGY: # 1) Apply quota to the ZFS file system # 2) Exceed the quota # 3) Attempt to write another file # 4) Verify the attempt fails with error code EDQUOTA (linux 122, others 49) # # verify_runnable "both" log_assert "Verify that a file write cannot exceed the file system quota" # # cleanup to be used internally as otherwise quota assertions cannot be # run independently or out of order # function cleanup { [[ -e $TESTDIR/$TESTFILE1 ]] && \ log_must rm $TESTDIR/$TESTFILE1 [[ -e $TESTDIR/$TESTFILE2 ]] && \ log_must rm $TESTDIR/$TESTFILE2 wait_freeing $TESTPOOL sync_pool $TESTPOOL + + reset_quota $TESTPOOL/$TESTFS } log_onexit cleanup # # Fills the quota & attempts to write another file # log_must exceed_quota $TESTPOOL/$TESTFS $TESTDIR log_pass "Could not write file. Quota limit enforced as expected" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_003_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_003_pos.ksh index de265813d55b..33f6421131fc 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_003_pos.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_003_pos.ksh @@ -1,80 +1,82 @@ #! /bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/quota/quota.kshlib # # DESCRIPTION: # A ZFS file system quota limits the amount of pool space # available to a file system dataset. Apply a quota and verify # that no more file creates are permitted. # # NOTE: THis test applies to a dataset rather than a file system. # # STRATEGY: # 1) Apply quota to ZFS file system dataset # 2) Create a file which is larger than the set quota # 3) Verify that the resulting file size is less than the quota limit # verify_runnable "both" log_assert "Verify that file size is limited by the file system quota" \ "(dataset version)" # # cleanup to be used internally as otherwise quota assertions cannot be # run independently or out of order # function cleanup { [[ -e $TESTDIR1/$TESTFILE1 ]] && \ log_must rm $TESTDIR1/$TESTFILE1 # # Need to allow time for space to be released back to # pool, otherwise next test will fail trying to set a # quota which is less than the space used. # wait_freeing $TESTPOOL sync_pool $TESTPOOL + + reset_quota $TESTPOOL/$TESTCTR/$TESTFS1 } log_onexit cleanup # # Sets the quota value and attempts to fill it with a file # twice the size of the quota # log_must fill_quota $TESTPOOL/$TESTCTR/$TESTFS1 $TESTDIR1 log_pass "File size limited by quota" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_004_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_004_pos.ksh index 8f20b533da68..682d09f080a4 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_004_pos.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_004_pos.ksh @@ -1,77 +1,79 @@ #! /bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/quota/quota.kshlib # # DESCRIPTION: # A zfs file system quota limits the amount of pool space # available to a given ZFS file system dataset. Once exceeded, it # is impossible to write any more files to the file system. # # STRATEGY: # 1) Apply quota to the ZFS file system dataset # 2) Exceed the quota # 3) Attempt to write another file # 4) Verify the attempt fails with error code EDQUOTA (linux 122, others 49) # # verify_runnable "both" log_assert "Verify that a file write cannot exceed the file system quota" \ "(dataset version)" # # cleanup to be used internally as otherwise quota assertions cannot be # run independently or out of order # function cleanup { [[ -e $TESTDIR1/$TESTFILE1 ]] && \ log_must rm $TESTDIR1/$TESTFILE1 [[ -e $TESTDIR1/$TESTFILE2 ]] && \ log_must rm $TESTDIR1/$TESTFILE2 wait_freeing $TESTPOOL sync_pool $TESTPOOL + + reset_quota $TESTPOOL/$TESTCTR/$TESTFS1 } log_onexit cleanup # # Fills the quota & attempts to write another file # log_must exceed_quota $TESTPOOL/$TESTCTR/$TESTFS1 $TESTDIR1 log_pass "Could not write file. Quota limit enforced as expected" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_005_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_005_pos.ksh index 98ee4edae650..9c4db81ca270 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_005_pos.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_005_pos.ksh @@ -1,75 +1,74 @@ #! /bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/quota/quota.kshlib # # DESCRIPTION: # # Verify that quota doesn't inherit its value from parent. # # STRATEGY: # 1) Set quota for parents # 2) Create a filesystem tree # 3) Verify that the 'quota' for descendent doesnot inherit the value. # ############################################################################### verify_runnable "both" function cleanup { datasetexists $fs_child && destroy_dataset $fs_child - log_must zfs set quota=$quota_val $fs + reset_quota $fs } log_onexit cleanup -log_assert "Verify that quota doesnot inherit its value from parent." -log_onexit cleanup +log_assert "Verify that quota does not inherit its value from parent." fs=$TESTPOOL/$TESTFS fs_child=$TESTPOOL/$TESTFS/$TESTFS space_avail=$(get_prop available $fs) quota_val=$(get_prop quota $fs) -typeset -i quotasize=$space_avail +typeset -li quotasize=$space_avail ((quotasize = quotasize * 2 )) log_must zfs set quota=$quotasize $fs log_must zfs create $fs_child quota_space=$(get_prop quota $fs_child) [[ $quota_space == $quotasize ]] && \ log_fail "The quota of child dataset inherits its value from parent." -log_pass "quota doesnot inherit its value from parent as expected." +log_pass "quota does not inherit its value from parent as expected." diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_006_neg.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_006_neg.ksh index 12105162c5b5..111d771188e3 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_006_neg.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/quota/quota_006_neg.ksh @@ -1,71 +1,71 @@ #! /bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2007 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # # # Copyright (c) 2013, 2016 by Delphix. All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/quota/quota.kshlib # # DESCRIPTION: # # Can't set a quota to less than currently being used by the dataset. # # STRATEGY: # 1) Create a filesystem # 2) Set a quota on the filesystem that is lower than the space # currently in use. # 3) Verify that the attempt fails. # verify_runnable "both" log_assert "Verify cannot set quota lower than the space currently in use" function cleanup { - log_must zfs set quota=none $TESTPOOL/$TESTFS + reset_quota $TESTPOOL/$TESTFS } log_onexit cleanup typeset -i quota_integer_size=0 typeset invalid_size="123! @456 7#89 0\$ abc123% 123%s 12%s3 %c123 123%d %x123 12%p3 \ ^def456 789&ghi" typeset -i space_used=`get_prop used $TESTPOOL/$TESTFS` (( quota_integer_size = space_used - 1 )) quota_fp_size=${quota_integer_size}.123 for size in 0 -1 $quota_integer_size -$quota_integer_size $quota_fp_size -$quota_fp_size \ $invalid_size ; do log_mustnot zfs set quota=$size $TESTPOOL/$TESTFS done log_must zfs set quota=$space_used $TESTPOOL/$TESTFS log_pass "As expected cannot set quota lower than space currently in use" diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/user_namespace/user_namespace_004.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/user_namespace/user_namespace_004.ksh index 37ef84b72377..e6ad25f23f93 100755 --- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/user_namespace/user_namespace_004.ksh +++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/user_namespace/user_namespace_004.ksh @@ -1,67 +1,67 @@ #!/bin/ksh -p # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License (the "License"). # You may not use this file except in compliance with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or https://opensource.org/licenses/CDDL-1.0. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # . $STF_SUITE/tests/functional/user_namespace/user_namespace_common.kshlib # # DESCRIPTION: # Regression test for safeguards around the delegation of datasets to # user namespaces. # # STRATEGY: # 1. Check that 'zfs zone' correctly handles the case of the first # argument being a non-namespace file. # 2. Check that 'zfs zone' correctly handles the case of the first # argument being a non-namespace and non-existent file. # verify_runnable "both" user_ns_cleanup() { if [ -n "$temp_file" ]; then log_must rm -f "$temp_file" fi log_must zfs destroy -r "$TESTPOOL/userns" } -log_onexit user_ns_cleanup - log_assert "Check zfs zone command handling of non-namespace files" # Pass if user namespaces are not supported. unshare -Urm echo test if [ "$?" -ne "0" ]; then log_unsupported "Failed to create user namespace" fi +log_onexit user_ns_cleanup + # Create the baseline datasets. log_must zfs create -o zoned=on "$TESTPOOL/userns" # 1. Try to pass a non-namespace file to zfs zone. temp_file="$(TMPDIR=$TEST_BASE_DIR mktemp)" log_mustnot zfs zone "$temp_file" "$TESTPOOL/userns" # 2. Try to pass a non-namespace and non-existent file to zfs zone. log_mustnot zfs zone "$TEMP_BASE_DIR/nonexistent" "$TESTPOOL/userns" log_pass "Check zfs zone command handling of non-namespace files" diff --git a/sys/modules/zfs/zfs_config.h b/sys/modules/zfs/zfs_config.h index 0e18c6092b9c..6e1f5de7cf9e 100644 --- a/sys/modules/zfs/zfs_config.h +++ b/sys/modules/zfs/zfs_config.h @@ -1,1218 +1,1218 @@ /* */ /* zfs_config.h. Generated from zfs_config.h.in by configure. */ /* zfs_config.h.in. Generated from configure.ac by autoheader. */ /* Define to 1 if translation of program messages to the user's native language is requested. */ /* #undef ENABLE_NLS */ /* bio_end_io_t wants 1 arg */ /* #undef HAVE_1ARG_BIO_END_IO_T */ /* lookup_bdev() wants 1 arg */ /* #undef HAVE_1ARG_LOOKUP_BDEV */ /* submit_bio() wants 1 arg */ /* #undef HAVE_1ARG_SUBMIT_BIO */ /* bdi_setup_and_register() wants 2 args */ /* #undef HAVE_2ARGS_BDI_SETUP_AND_REGISTER */ /* vfs_getattr wants 2 args */ /* #undef HAVE_2ARGS_VFS_GETATTR */ /* zlib_deflate_workspacesize() wants 2 args */ /* #undef HAVE_2ARGS_ZLIB_DEFLATE_WORKSPACESIZE */ /* bdi_setup_and_register() wants 3 args */ /* #undef HAVE_3ARGS_BDI_SETUP_AND_REGISTER */ /* vfs_getattr wants 3 args */ /* #undef HAVE_3ARGS_VFS_GETATTR */ /* vfs_getattr wants 4 args */ /* #undef HAVE_4ARGS_VFS_GETATTR */ /* kernel has access_ok with 'type' parameter */ /* #undef HAVE_ACCESS_OK_TYPE */ /* posix_acl has refcount_t */ /* #undef HAVE_ACL_REFCOUNT */ /* add_disk() returns int */ /* #undef HAVE_ADD_DISK_RET */ /* Define if host toolchain supports AES */ #define HAVE_AES 1 /* Define if you have [rt] */ #define HAVE_AIO_H 1 #ifdef __amd64__ #ifndef RESCUE /* Define if host toolchain supports AVX */ #define HAVE_AVX 1 #endif /* Define if host toolchain supports AVX2 */ #define HAVE_AVX2 1 /* Define if host toolchain supports AVX512BW */ #define HAVE_AVX512BW 1 /* Define if host toolchain supports AVX512CD */ #define HAVE_AVX512CD 1 /* Define if host toolchain supports AVX512DQ */ #define HAVE_AVX512DQ 1 /* Define if host toolchain supports AVX512ER */ #define HAVE_AVX512ER 1 /* Define if host toolchain supports AVX512F */ #define HAVE_AVX512F 1 /* Define if host toolchain supports AVX512IFMA */ #define HAVE_AVX512IFMA 1 /* Define if host toolchain supports AVX512PF */ #define HAVE_AVX512PF 1 /* Define if host toolchain supports AVX512VBMI */ #define HAVE_AVX512VBMI 1 /* Define if host toolchain supports AVX512VL */ #define HAVE_AVX512VL 1 #endif /* bdevname() is available */ /* #undef HAVE_BDEVNAME */ /* bdev_check_media_change() exists */ /* #undef HAVE_BDEV_CHECK_MEDIA_CHANGE */ /* bdev_file_open_by_path() exists */ /* #undef HAVE_BDEV_FILE_OPEN_BY_PATH */ /* bdev_*_io_acct() available */ /* #undef HAVE_BDEV_IO_ACCT_63 */ /* bdev_*_io_acct() available */ /* #undef HAVE_BDEV_IO_ACCT_OLD */ /* bdev_kobj() exists */ /* #undef HAVE_BDEV_KOBJ */ /* bdev_max_discard_sectors() is available */ /* #undef HAVE_BDEV_MAX_DISCARD_SECTORS */ /* bdev_max_secure_erase_sectors() is available */ /* #undef HAVE_BDEV_MAX_SECURE_ERASE_SECTORS */ /* bdev_open_by_path() exists */ /* #undef HAVE_BDEV_OPEN_BY_PATH */ /* bdev_release() exists */ /* #undef HAVE_BDEV_RELEASE */ /* block_device_operations->submit_bio() returns void */ /* #undef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID */ /* bdev_whole() is available */ /* #undef HAVE_BDEV_WHOLE */ /* bio_alloc() takes 4 arguments */ /* #undef HAVE_BIO_ALLOC_4ARG */ /* bio->bi_bdev->bd_disk exists */ /* #undef HAVE_BIO_BDEV_DISK */ /* bio->bi_opf is defined */ /* #undef HAVE_BIO_BI_OPF */ /* bio->bi_status exists */ /* #undef HAVE_BIO_BI_STATUS */ /* bio has bi_iter */ /* #undef HAVE_BIO_BVEC_ITER */ /* bio_*_io_acct() available */ /* #undef HAVE_BIO_IO_ACCT */ /* bio_max_segs() is implemented */ /* #undef HAVE_BIO_MAX_SEGS */ /* bio_set_dev() is available */ /* #undef HAVE_BIO_SET_DEV */ /* bio_set_dev() GPL-only */ /* #undef HAVE_BIO_SET_DEV_GPL_ONLY */ /* bio_set_dev() is a macro */ /* #undef HAVE_BIO_SET_DEV_MACRO */ /* bio_set_op_attrs is available */ /* #undef HAVE_BIO_SET_OP_ATTRS */ /* blkdev_get_by_path() exists and takes 4 args */ /* #undef HAVE_BLKDEV_GET_BY_PATH_4ARG */ /* blkdev_get_by_path() handles ERESTARTSYS */ /* #undef HAVE_BLKDEV_GET_ERESTARTSYS */ /* __blkdev_issue_discard(flags) is available */ /* #undef HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_FLAGS */ /* __blkdev_issue_discard() is available */ /* #undef HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_NOFLAGS */ /* blkdev_issue_discard(flags) is available */ /* #undef HAVE_BLKDEV_ISSUE_DISCARD_FLAGS */ /* blkdev_issue_discard() is available */ /* #undef HAVE_BLKDEV_ISSUE_DISCARD_NOFLAGS */ /* blkdev_issue_secure_erase() is available */ /* #undef HAVE_BLKDEV_ISSUE_SECURE_ERASE */ /* blkdev_put() exists */ /* #undef HAVE_BLKDEV_PUT */ /* blkdev_put() accepts void* as arg 2 */ /* #undef HAVE_BLKDEV_PUT_HOLDER */ /* blkdev_reread_part() exists */ /* #undef HAVE_BLKDEV_REREAD_PART */ /* blkg_tryget() is available */ /* #undef HAVE_BLKG_TRYGET */ /* blkg_tryget() GPL-only */ /* #undef HAVE_BLKG_TRYGET_GPL_ONLY */ /* blk_alloc_disk() exists */ /* #undef HAVE_BLK_ALLOC_DISK */ /* blk_alloc_disk() exists and takes 2 args */ /* #undef HAVE_BLK_ALLOC_DISK_2ARG */ /* blk_alloc_queue() expects request function */ /* #undef HAVE_BLK_ALLOC_QUEUE_REQUEST_FN */ /* blk_alloc_queue_rh() expects request function */ /* #undef HAVE_BLK_ALLOC_QUEUE_REQUEST_FN_RH */ /* blk_cleanup_disk() exists */ /* #undef HAVE_BLK_CLEANUP_DISK */ /* blk_mode_t is defined */ /* #undef HAVE_BLK_MODE_T */ /* block multiqueue is available */ /* #undef HAVE_BLK_MQ */ /* block multiqueue hardware context is cached in struct request */ /* #undef HAVE_BLK_MQ_RQ_HCTX */ /* blk queue backing_dev_info is dynamic */ /* #undef HAVE_BLK_QUEUE_BDI_DYNAMIC */ /* blk_queue_discard() is available */ /* #undef HAVE_BLK_QUEUE_DISCARD */ /* blk_queue_flag_clear() exists */ /* #undef HAVE_BLK_QUEUE_FLAG_CLEAR */ /* blk_queue_flag_set() exists */ /* #undef HAVE_BLK_QUEUE_FLAG_SET */ /* blk_queue_flush() is available */ /* #undef HAVE_BLK_QUEUE_FLUSH */ /* blk_queue_flush() is GPL-only */ /* #undef HAVE_BLK_QUEUE_FLUSH_GPL_ONLY */ /* blk_queue_secdiscard() is available */ /* #undef HAVE_BLK_QUEUE_SECDISCARD */ /* blk_queue_secure_erase() is available */ /* #undef HAVE_BLK_QUEUE_SECURE_ERASE */ /* blk_queue_update_readahead() exists */ /* #undef HAVE_BLK_QUEUE_UPDATE_READAHEAD */ /* blk_queue_write_cache() exists */ /* #undef HAVE_BLK_QUEUE_WRITE_CACHE */ /* blk_queue_write_cache() is GPL-only */ /* #undef HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY */ /* BLK_STS_RESV_CONFLICT is defined */ /* #undef HAVE_BLK_STS_RESV_CONFLICT */ /* Define if release() in block_device_operations takes 1 arg */ /* #undef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG */ /* Define if revalidate_disk() in block_device_operations */ /* #undef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK */ /* Define to 1 if you have the Mac OS X function CFLocaleCopyCurrent in the CoreFoundation framework. */ /* #undef HAVE_CFLOCALECOPYCURRENT */ /* Define to 1 if you have the Mac OS X function CFLocaleCopyPreferredLanguages in the CoreFoundation framework. */ /* #undef HAVE_CFLOCALECOPYPREFERREDLANGUAGES */ /* Define to 1 if you have the Mac OS X function CFPreferencesCopyAppValue in the CoreFoundation framework. */ /* #undef HAVE_CFPREFERENCESCOPYAPPVALUE */ /* check_disk_change() exists */ /* #undef HAVE_CHECK_DISK_CHANGE */ /* clear_inode() is available */ /* #undef HAVE_CLEAR_INODE */ /* dentry uses const struct dentry_operations */ /* #undef HAVE_CONST_DENTRY_OPERATIONS */ /* copy_from_iter() is available */ /* #undef HAVE_COPY_FROM_ITER */ /* copy_splice_read exists */ /* #undef HAVE_COPY_SPLICE_READ */ /* copy_to_iter() is available */ /* #undef HAVE_COPY_TO_ITER */ /* cpu_has_feature() is GPL-only */ /* #undef HAVE_CPU_HAS_FEATURE_GPL_ONLY */ /* yes */ /* #undef HAVE_CPU_HOTPLUG */ /* current_time() exists */ /* #undef HAVE_CURRENT_TIME */ /* Define if the GNU dcgettext() function is already present or preinstalled. */ /* #undef HAVE_DCGETTEXT */ /* DECLARE_EVENT_CLASS() is available */ /* #undef HAVE_DECLARE_EVENT_CLASS */ /* dentry aliases are in d_u member */ /* #undef HAVE_DENTRY_D_U_ALIASES */ /* dequeue_signal() takes 4 arguments */ /* #undef HAVE_DEQUEUE_SIGNAL_4ARG */ /* lookup_bdev() wants dev_t arg */ /* #undef HAVE_DEVT_LOOKUP_BDEV */ /* sops->dirty_inode() wants flags */ /* #undef HAVE_DIRTY_INODE_WITH_FLAGS */ /* disk_check_media_change() exists */ /* #undef HAVE_DISK_CHECK_MEDIA_CHANGE */ /* disk_*_io_acct() available */ /* #undef HAVE_DISK_IO_ACCT */ /* disk_update_readahead() exists */ /* #undef HAVE_DISK_UPDATE_READAHEAD */ /* Define to 1 if you have the header file. */ #define HAVE_DLFCN_H 1 /* d_make_root() is available */ /* #undef HAVE_D_MAKE_ROOT */ /* d_prune_aliases() is available */ /* #undef HAVE_D_PRUNE_ALIASES */ /* dops->d_revalidate() operation takes nameidata */ /* #undef HAVE_D_REVALIDATE_NAMEIDATA */ /* eops->encode_fh() wants child and parent inodes */ /* #undef HAVE_ENCODE_FH_WITH_INODE */ /* sops->evict_inode() exists */ /* #undef HAVE_EVICT_INODE */ /* Define to 1 if you have the 'execvpe' function. */ #define HAVE_EXECVPE 1 /* FALLOC_FL_ZERO_RANGE is defined */ /* #undef HAVE_FALLOC_FL_ZERO_RANGE */ /* fault_in_iov_iter_readable() is available */ /* #undef HAVE_FAULT_IN_IOV_ITER_READABLE */ /* filemap_range_has_page() is available */ /* #undef HAVE_FILEMAP_RANGE_HAS_PAGE */ /* fops->aio_fsync() exists */ /* #undef HAVE_FILE_AIO_FSYNC */ /* file_dentry() is available */ /* #undef HAVE_FILE_DENTRY */ /* fops->fadvise() exists */ /* #undef HAVE_FILE_FADVISE */ /* file_inode() is available */ /* #undef HAVE_FILE_INODE */ /* flush_dcache_page() is GPL-only */ /* #undef HAVE_FLUSH_DCACHE_PAGE_GPL_ONLY */ /* iops->follow_link() cookie */ /* #undef HAVE_FOLLOW_LINK_COOKIE */ /* iops->follow_link() nameidata */ /* #undef HAVE_FOLLOW_LINK_NAMEIDATA */ /* Define if compiler supports -Wformat-overflow */ /* #undef HAVE_FORMAT_OVERFLOW */ /* fsync_bdev() is declared in include/blkdev.h */ /* #undef HAVE_FSYNC_BDEV */ /* fops->fsync() with range */ /* #undef HAVE_FSYNC_RANGE */ /* fops->fsync() without dentry */ /* #undef HAVE_FSYNC_WITHOUT_DENTRY */ /* yes */ /* #undef HAVE_GENERIC_FADVISE */ /* generic_fillattr requires struct mnt_idmap* */ /* #undef HAVE_GENERIC_FILLATTR_IDMAP */ /* generic_fillattr requires struct mnt_idmap* and u32 request_mask */ /* #undef HAVE_GENERIC_FILLATTR_IDMAP_REQMASK */ /* generic_fillattr requires struct user_namespace* */ /* #undef HAVE_GENERIC_FILLATTR_USERNS */ /* generic_*_io_acct() 3 arg available */ /* #undef HAVE_GENERIC_IO_ACCT_3ARG */ /* generic_*_io_acct() 4 arg available */ /* #undef HAVE_GENERIC_IO_ACCT_4ARG */ /* generic_readlink is global */ /* #undef HAVE_GENERIC_READLINK */ /* generic_setxattr() exists */ /* #undef HAVE_GENERIC_SETXATTR */ /* generic_write_checks() takes kiocb */ /* #undef HAVE_GENERIC_WRITE_CHECKS_KIOCB */ /* Define if the GNU gettext() function is already present or preinstalled. */ /* #undef HAVE_GETTEXT */ /* iops->get_acl() exists */ /* #undef HAVE_GET_ACL */ /* iops->get_acl() takes rcu */ /* #undef HAVE_GET_ACL_RCU */ /* has iops->get_inode_acl() */ /* #undef HAVE_GET_INODE_ACL */ /* iops->get_link() cookie */ /* #undef HAVE_GET_LINK_COOKIE */ /* iops->get_link() delayed */ /* #undef HAVE_GET_LINK_DELAYED */ /* group_info->gid exists */ /* #undef HAVE_GROUP_INFO_GID */ /* has_capability() is available */ /* #undef HAVE_HAS_CAPABILITY */ /* iattr->ia_vfsuid and iattr->ia_vfsgid exist */ /* #undef HAVE_IATTR_VFSID */ /* Define if you have the iconv() function and it works. */ #define HAVE_ICONV 1 /* iops->getattr() takes struct mnt_idmap* */ /* #undef HAVE_IDMAP_IOPS_GETATTR */ /* iops->setattr() takes struct mnt_idmap* */ /* #undef HAVE_IDMAP_IOPS_SETATTR */ /* APIs for idmapped mount are present */ /* #undef HAVE_IDMAP_MNT_API */ /* mnt_idmap does not have user_namespace */ /* #undef HAVE_IDMAP_NO_USERNS */ /* Define if compiler supports -Wimplicit-fallthrough */ /* #undef HAVE_IMPLICIT_FALLTHROUGH */ /* Define if compiler supports -Winfinite-recursion */ /* #undef HAVE_INFINITE_RECURSION */ /* inode_get_atime() exists in linux/fs.h */ /* #undef HAVE_INODE_GET_ATIME */ /* inode_get_ctime() exists in linux/fs.h */ /* #undef HAVE_INODE_GET_CTIME */ /* inode_get_mtime() exists in linux/fs.h */ /* #undef HAVE_INODE_GET_MTIME */ /* yes */ /* #undef HAVE_INODE_LOCK_SHARED */ /* inode_owner_or_capable() exists */ /* #undef HAVE_INODE_OWNER_OR_CAPABLE */ /* inode_owner_or_capable() takes mnt_idmap */ /* #undef HAVE_INODE_OWNER_OR_CAPABLE_IDMAP */ /* inode_owner_or_capable() takes user_ns */ /* #undef HAVE_INODE_OWNER_OR_CAPABLE_USERNS */ /* inode_set_atime_to_ts() exists in linux/fs.h */ /* #undef HAVE_INODE_SET_ATIME_TO_TS */ /* inode_set_ctime_to_ts() exists in linux/fs.h */ /* #undef HAVE_INODE_SET_CTIME_TO_TS */ /* inode_set_flags() exists */ /* #undef HAVE_INODE_SET_FLAGS */ /* inode_set_iversion() exists */ /* #undef HAVE_INODE_SET_IVERSION */ /* inode_set_mtime_to_ts() exists in linux/fs.h */ /* #undef HAVE_INODE_SET_MTIME_TO_TS */ /* inode->i_*time's are timespec64 */ /* #undef HAVE_INODE_TIMESPEC64_TIMES */ /* timestamp_truncate() exists */ /* #undef HAVE_INODE_TIMESTAMP_TRUNCATE */ /* Define to 1 if you have the header file. */ #define HAVE_INTTYPES_H 1 /* in_compat_syscall() is available */ /* #undef HAVE_IN_COMPAT_SYSCALL */ /* iops->create() takes struct mnt_idmap* */ /* #undef HAVE_IOPS_CREATE_IDMAP */ /* iops->create() takes struct user_namespace* */ /* #undef HAVE_IOPS_CREATE_USERNS */ /* iops->mkdir() takes struct mnt_idmap* */ /* #undef HAVE_IOPS_MKDIR_IDMAP */ /* iops->mkdir() takes struct user_namespace* */ /* #undef HAVE_IOPS_MKDIR_USERNS */ /* iops->mknod() takes struct mnt_idmap* */ /* #undef HAVE_IOPS_MKNOD_IDMAP */ /* iops->mknod() takes struct user_namespace* */ /* #undef HAVE_IOPS_MKNOD_USERNS */ /* iops->permission() takes struct mnt_idmap* */ /* #undef HAVE_IOPS_PERMISSION_IDMAP */ /* iops->permission() takes struct user_namespace* */ /* #undef HAVE_IOPS_PERMISSION_USERNS */ /* iops->rename() takes struct mnt_idmap* */ /* #undef HAVE_IOPS_RENAME_IDMAP */ /* iops->rename() takes struct user_namespace* */ /* #undef HAVE_IOPS_RENAME_USERNS */ /* iops->setattr() exists */ /* #undef HAVE_IOPS_SETATTR */ /* iops->symlink() takes struct mnt_idmap* */ /* #undef HAVE_IOPS_SYMLINK_IDMAP */ /* iops->symlink() takes struct user_namespace* */ /* #undef HAVE_IOPS_SYMLINK_USERNS */ /* iov_iter_advance() is available */ /* #undef HAVE_IOV_ITER_ADVANCE */ /* iov_iter_count() is available */ /* #undef HAVE_IOV_ITER_COUNT */ /* iov_iter_fault_in_readable() is available */ /* #undef HAVE_IOV_ITER_FAULT_IN_READABLE */ /* iov_iter_revert() is available */ /* #undef HAVE_IOV_ITER_REVERT */ /* iov_iter_type() is available */ /* #undef HAVE_IOV_ITER_TYPE */ /* iov_iter types are available */ /* #undef HAVE_IOV_ITER_TYPES */ /* yes */ /* #undef HAVE_IO_SCHEDULE_TIMEOUT */ /* Define to 1 if you have the 'issetugid' function. */ #define HAVE_ISSETUGID 1 /* iter_iov() is available */ /* #undef HAVE_ITER_IOV */ /* kernel has kernel_fpu_* functions */ /* #undef HAVE_KERNEL_FPU */ /* kernel has asm/fpu/api.h */ /* #undef HAVE_KERNEL_FPU_API_HEADER */ /* kernel fpu internal */ /* #undef HAVE_KERNEL_FPU_INTERNAL */ /* kernel has asm/fpu/internal.h */ /* #undef HAVE_KERNEL_FPU_INTERNAL_HEADER */ /* uncached_acl_sentinel() exists */ /* #undef HAVE_KERNEL_GET_ACL_HANDLE_CACHE */ /* Define if compiler supports -Winfinite-recursion */ /* #undef HAVE_KERNEL_INFINITE_RECURSION */ /* kernel has kernel_neon_* functions */ /* #undef HAVE_KERNEL_NEON */ /* kernel does stack verification */ /* #undef HAVE_KERNEL_OBJTOOL */ /* kernel has linux/objtool.h */ /* #undef HAVE_KERNEL_OBJTOOL_HEADER */ /* kernel_read() take loff_t pointer */ /* #undef HAVE_KERNEL_READ_PPOS */ /* strlcpy() exists */ /* #undef HAVE_KERNEL_STRLCPY */ /* strscpy() exists */ /* #undef HAVE_KERNEL_STRSCPY */ /* timer_list.function gets a timer_list */ /* #undef HAVE_KERNEL_TIMER_FUNCTION_TIMER_LIST */ /* struct timer_list has a flags member */ /* #undef HAVE_KERNEL_TIMER_LIST_FLAGS */ /* timer_setup() is available */ /* #undef HAVE_KERNEL_TIMER_SETUP */ /* kernel_write() take loff_t pointer */ /* #undef HAVE_KERNEL_WRITE_PPOS */ /* kmem_cache_create_usercopy() exists */ /* #undef HAVE_KMEM_CACHE_CREATE_USERCOPY */ /* kstrtoul() exists */ /* #undef HAVE_KSTRTOUL */ /* ktime_get_coarse_real_ts64() exists */ /* #undef HAVE_KTIME_GET_COARSE_REAL_TS64 */ /* ktime_get_raw_ts64() exists */ /* #undef HAVE_KTIME_GET_RAW_TS64 */ /* kvmalloc exists */ /* #undef HAVE_KVMALLOC */ /* Define if you have [aio] */ /* #undef HAVE_LIBAIO */ /* Define if you have [blkid] */ /* #undef HAVE_LIBBLKID */ /* Define if you have [crypto] */ #define HAVE_LIBCRYPTO 1 /* Define if you have [tirpc] */ /* #undef HAVE_LIBTIRPC */ /* Define if you have [udev] */ /* #undef HAVE_LIBUDEV */ /* Define if you have [uuid] */ /* #undef HAVE_LIBUUID */ /* linux/blk-cgroup.h exists */ /* #undef HAVE_LINUX_BLK_CGROUP_HEADER */ /* lseek_execute() is available */ /* #undef HAVE_LSEEK_EXECUTE */ /* makedev() is declared in sys/mkdev.h */ /* #undef HAVE_MAKEDEV_IN_MKDEV */ /* makedev() is declared in sys/sysmacros.h */ /* #undef HAVE_MAKEDEV_IN_SYSMACROS */ /* Noting that make_request_fn() returns blk_qc_t */ /* #undef HAVE_MAKE_REQUEST_FN_RET_QC */ /* Noting that make_request_fn() returns void */ /* #undef HAVE_MAKE_REQUEST_FN_RET_VOID */ /* iops->mkdir() takes umode_t */ /* #undef HAVE_MKDIR_UMODE_T */ /* Define to 1 if you have the 'mlockall' function. */ #define HAVE_MLOCKALL 1 /* page_size() is available */ /* #undef HAVE_MM_PAGE_SIZE */ /* lookup_bdev() wants mode arg */ /* #undef HAVE_MODE_LOOKUP_BDEV */ /* Define if host toolchain supports MOVBE */ #define HAVE_MOVBE 1 /* new_sync_read()/new_sync_write() are available */ /* #undef HAVE_NEW_SYNC_READ */ /* folio_wait_bit() exists */ /* #undef HAVE_PAGEMAP_FOLIO_WAIT_BIT */ /* part_to_dev() exists */ /* #undef HAVE_PART_TO_DEV */ /* iops->getattr() takes a path */ /* #undef HAVE_PATH_IOPS_GETATTR */ /* Define if host toolchain supports PCLMULQDQ */ #define HAVE_PCLMULQDQ 1 /* percpu_counter_add_batch() is defined */ /* #undef HAVE_PERCPU_COUNTER_ADD_BATCH */ /* percpu_counter_init() wants gfp_t */ /* #undef HAVE_PERCPU_COUNTER_INIT_WITH_GFP */ /* posix_acl_chmod() exists */ /* #undef HAVE_POSIX_ACL_CHMOD */ /* posix_acl_from_xattr() needs user_ns */ /* #undef HAVE_POSIX_ACL_FROM_XATTR_USERNS */ /* posix_acl_release() is available */ /* #undef HAVE_POSIX_ACL_RELEASE */ /* posix_acl_release() is GPL-only */ /* #undef HAVE_POSIX_ACL_RELEASE_GPL_ONLY */ /* posix_acl_valid() wants user namespace */ /* #undef HAVE_POSIX_ACL_VALID_WITH_NS */ /* proc_ops structure exists */ /* #undef HAVE_PROC_OPS_STRUCT */ /* iops->put_link() cookie */ /* #undef HAVE_PUT_LINK_COOKIE */ /* iops->put_link() delayed */ /* #undef HAVE_PUT_LINK_DELAYED */ /* iops->put_link() nameidata */ /* #undef HAVE_PUT_LINK_NAMEIDATA */ /* If available, contains the Python version number currently in use. */ #define HAVE_PYTHON "3.7" /* qat is enabled and existed */ /* #undef HAVE_QAT */ /* struct reclaim_state has reclaimed */ /* #undef HAVE_RECLAIM_STATE_RECLAIMED */ /* register_shrinker is vararg */ /* #undef HAVE_REGISTER_SHRINKER_VARARG */ /* register_sysctl_table exists */ /* #undef HAVE_REGISTER_SYSCTL_TABLE */ /* iops->rename2() exists */ /* #undef HAVE_RENAME2 */ /* struct inode_operations_wrapper takes .rename2() */ /* #undef HAVE_RENAME2_OPERATIONS_WRAPPER */ /* iops->rename() wants flags */ /* #undef HAVE_RENAME_WANTS_FLAGS */ /* REQ_DISCARD is defined */ /* #undef HAVE_REQ_DISCARD */ /* REQ_FLUSH is defined */ /* #undef HAVE_REQ_FLUSH */ /* REQ_OP_DISCARD is defined */ /* #undef HAVE_REQ_OP_DISCARD */ /* REQ_OP_FLUSH is defined */ /* #undef HAVE_REQ_OP_FLUSH */ /* REQ_OP_SECURE_ERASE is defined */ /* #undef HAVE_REQ_OP_SECURE_ERASE */ /* REQ_PREFLUSH is defined */ /* #undef HAVE_REQ_PREFLUSH */ /* revalidate_disk() is available */ /* #undef HAVE_REVALIDATE_DISK */ /* revalidate_disk_size() is available */ /* #undef HAVE_REVALIDATE_DISK_SIZE */ /* struct rw_semaphore has member activity */ /* #undef HAVE_RWSEM_ACTIVITY */ /* struct rw_semaphore has atomic_long_t member count */ /* #undef HAVE_RWSEM_ATOMIC_LONG_COUNT */ /* linux/sched/signal.h exists */ /* #undef HAVE_SCHED_SIGNAL_HEADER */ /* Define to 1 if you have the header file. */ #define HAVE_SECURITY_PAM_MODULES_H 1 /* setattr_prepare() accepts mnt_idmap */ /* #undef HAVE_SETATTR_PREPARE_IDMAP */ /* setattr_prepare() is available, doesn't accept user_namespace */ /* #undef HAVE_SETATTR_PREPARE_NO_USERNS */ /* setattr_prepare() accepts user_namespace */ /* #undef HAVE_SETATTR_PREPARE_USERNS */ /* iops->set_acl() exists, takes 3 args */ /* #undef HAVE_SET_ACL */ /* iops->set_acl() takes 4 args, arg1 is struct mnt_idmap * */ /* #undef HAVE_SET_ACL_IDMAP_DENTRY */ /* iops->set_acl() takes 4 args */ /* #undef HAVE_SET_ACL_USERNS */ /* iops->set_acl() takes 4 args, arg2 is struct dentry * */ /* #undef HAVE_SET_ACL_USERNS_DENTRY_ARG2 */ /* set_cached_acl() is usable */ /* #undef HAVE_SET_CACHED_ACL_USABLE */ /* set_special_state() exists */ /* #undef HAVE_SET_SPECIAL_STATE */ /* shrinker_register exists */ /* #undef HAVE_SHRINKER_REGISTER */ /* struct shrink_control exists */ /* #undef HAVE_SHRINK_CONTROL_STRUCT */ /* kernel_siginfo_t exists */ /* #undef HAVE_SIGINFO */ /* signal_stop() exists */ /* #undef HAVE_SIGNAL_STOP */ /* new shrinker callback wants 2 args */ /* #undef HAVE_SINGLE_SHRINKER_CALLBACK */ /* cs->count_objects exists */ /* #undef HAVE_SPLIT_SHRINKER_CALLBACK */ #if defined(__amd64__) || defined(__i386__) /* Define if host toolchain supports SSE */ #define HAVE_SSE 1 /* Define if host toolchain supports SSE2 */ #define HAVE_SSE2 1 /* Define if host toolchain supports SSE3 */ #define HAVE_SSE3 1 /* Define if host toolchain supports SSE4.1 */ #define HAVE_SSE4_1 1 /* Define if host toolchain supports SSE4.2 */ #define HAVE_SSE4_2 1 /* Define if host toolchain supports SSSE3 */ #define HAVE_SSSE3 1 #endif /* STACK_FRAME_NON_STANDARD is defined */ /* #undef HAVE_STACK_FRAME_NON_STANDARD */ /* standalone exists */ /* #undef HAVE_STANDALONE_LINUX_STDARG */ /* Define to 1 if you have the header file. */ #define HAVE_STDINT_H 1 /* Define to 1 if you have the header file. */ #define HAVE_STDIO_H 1 /* Define to 1 if you have the header file. */ #define HAVE_STDLIB_H 1 /* Define to 1 if you have the header file. */ #define HAVE_STRINGS_H 1 /* Define to 1 if you have the header file. */ #define HAVE_STRING_H 1 /* Define to 1 if you have the 'strlcat' function. */ #define HAVE_STRLCAT 1 /* Define to 1 if you have the 'strlcpy' function. */ #define HAVE_STRLCPY 1 /* submit_bio is member of struct block_device_operations */ /* #undef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */ /* have super_block s_shrink */ /* #undef HAVE_SUPER_BLOCK_S_SHRINK */ /* have super_block s_shrink pointer */ /* #undef HAVE_SUPER_BLOCK_S_SHRINK_PTR */ /* super_setup_bdi_name() exits */ /* #undef HAVE_SUPER_SETUP_BDI_NAME */ /* super_block->s_user_ns exists */ /* #undef HAVE_SUPER_USER_NS */ /* sync_blockdev() is declared in include/blkdev.h */ /* #undef HAVE_SYNC_BLOCKDEV */ /* struct kobj_type has default_groups */ /* #undef HAVE_SYSFS_DEFAULT_GROUPS */ /* Define to 1 if you have the header file. */ #define HAVE_SYS_STAT_H 1 /* Define to 1 if you have the header file. */ #define HAVE_SYS_TYPES_H 1 /* i_op->tmpfile() exists */ /* #undef HAVE_TMPFILE */ /* i_op->tmpfile() uses old dentry signature */ /* #undef HAVE_TMPFILE_DENTRY */ /* i_op->tmpfile() has mnt_idmap */ /* #undef HAVE_TMPFILE_IDMAP */ /* i_op->tmpfile() has userns */ /* #undef HAVE_TMPFILE_USERNS */ /* totalhigh_pages() exists */ /* #undef HAVE_TOTALHIGH_PAGES */ /* kernel has totalram_pages() */ /* #undef HAVE_TOTALRAM_PAGES_FUNC */ /* Define to 1 if you have the 'udev_device_get_is_initialized' function. */ /* #undef HAVE_UDEV_DEVICE_GET_IS_INITIALIZED */ /* kernel has __kernel_fpu_* functions */ /* #undef HAVE_UNDERSCORE_KERNEL_FPU */ /* Define to 1 if you have the header file. */ #define HAVE_UNISTD_H 1 /* iops->getattr() takes struct user_namespace* */ /* #undef HAVE_USERNS_IOPS_GETATTR */ /* iops->setattr() takes struct user_namespace* */ /* #undef HAVE_USERNS_IOPS_SETATTR */ /* user_namespace->ns.inum exists */ /* #undef HAVE_USER_NS_COMMON_INUM */ /* iops->getattr() takes a vfsmount */ /* #undef HAVE_VFSMOUNT_IOPS_GETATTR */ /* fops->clone_file_range() is available */ /* #undef HAVE_VFS_CLONE_FILE_RANGE */ /* fops->copy_file_range() is available */ /* #undef HAVE_VFS_COPY_FILE_RANGE */ /* fops->dedupe_file_range() is available */ /* #undef HAVE_VFS_DEDUPE_FILE_RANGE */ /* aops->direct_IO() uses iovec */ /* #undef HAVE_VFS_DIRECT_IO_IOVEC */ /* aops->direct_IO() uses iov_iter without rw */ /* #undef HAVE_VFS_DIRECT_IO_ITER */ /* aops->direct_IO() uses iov_iter with offset */ /* #undef HAVE_VFS_DIRECT_IO_ITER_OFFSET */ /* aops->direct_IO() uses iov_iter with rw and offset */ /* #undef HAVE_VFS_DIRECT_IO_ITER_RW_OFFSET */ /* filemap_dirty_folio exists */ /* #undef HAVE_VFS_FILEMAP_DIRTY_FOLIO */ /* file_operations_extend takes .copy_file_range() and .clone_file_range() */ /* #undef HAVE_VFS_FILE_OPERATIONS_EXTEND */ /* generic_copy_file_range() is available */ /* #undef HAVE_VFS_GENERIC_COPY_FILE_RANGE */ /* All required iov_iter interfaces are available */ /* #undef HAVE_VFS_IOV_ITER */ /* fops->iterate() is available */ /* #undef HAVE_VFS_ITERATE */ /* fops->iterate_shared() is available */ /* #undef HAVE_VFS_ITERATE_SHARED */ /* fops->readdir() is available */ /* #undef HAVE_VFS_READDIR */ /* address_space_operations->readpages exists */ /* #undef HAVE_VFS_READPAGES */ /* read_folio exists */ /* #undef HAVE_VFS_READ_FOLIO */ /* fops->remap_file_range() is available */ /* #undef HAVE_VFS_REMAP_FILE_RANGE */ /* fops->read/write_iter() are available */ /* #undef HAVE_VFS_RW_ITERATE */ /* __set_page_dirty_nobuffers exists */ /* #undef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS */ /* splice_copy_file_range() is available */ /* #undef HAVE_VFS_SPLICE_COPY_FILE_RANGE */ /* __vmalloc page flags exists */ /* #undef HAVE_VMALLOC_PAGE_KERNEL */ /* yes */ /* #undef HAVE_WAIT_ON_BIT_ACTION */ /* wait_queue_entry_t exists */ /* #undef HAVE_WAIT_QUEUE_ENTRY_T */ /* wq_head->head and wq_entry->entry exist */ /* #undef HAVE_WAIT_QUEUE_HEAD_ENTRY */ /* int (*writepage_t)() takes struct folio* */ /* #undef HAVE_WRITEPAGE_T_FOLIO */ /* xattr_handler->get() wants dentry */ /* #undef HAVE_XATTR_GET_DENTRY */ /* xattr_handler->get() wants both dentry and inode */ /* #undef HAVE_XATTR_GET_DENTRY_INODE */ /* xattr_handler->get() wants dentry and inode and flags */ /* #undef HAVE_XATTR_GET_DENTRY_INODE_FLAGS */ /* xattr_handler->get() wants xattr_handler */ /* #undef HAVE_XATTR_GET_HANDLER */ /* xattr_handler has name */ /* #undef HAVE_XATTR_HANDLER_NAME */ /* xattr_handler->list() wants dentry */ /* #undef HAVE_XATTR_LIST_DENTRY */ /* xattr_handler->list() wants xattr_handler */ /* #undef HAVE_XATTR_LIST_HANDLER */ /* xattr_handler->list() wants simple */ /* #undef HAVE_XATTR_LIST_SIMPLE */ /* xattr_handler->set() wants dentry */ /* #undef HAVE_XATTR_SET_DENTRY */ /* xattr_handler->set() wants both dentry and inode */ /* #undef HAVE_XATTR_SET_DENTRY_INODE */ /* xattr_handler->set() wants xattr_handler */ /* #undef HAVE_XATTR_SET_HANDLER */ /* xattr_handler->set() takes mnt_idmap */ /* #undef HAVE_XATTR_SET_IDMAP */ /* xattr_handler->set() takes user_namespace */ /* #undef HAVE_XATTR_SET_USERNS */ /* Define if host toolchain supports XSAVE */ #define HAVE_XSAVE 1 /* Define if host toolchain supports XSAVEOPT */ #define HAVE_XSAVEOPT 1 /* Define if host toolchain supports XSAVES */ #define HAVE_XSAVES 1 /* ZERO_PAGE() is GPL-only */ /* #undef HAVE_ZERO_PAGE_GPL_ONLY */ /* Define if you have [z] */ #define HAVE_ZLIB 1 /* __posix_acl_chmod() exists */ /* #undef HAVE___POSIX_ACL_CHMOD */ /* kernel exports FPU functions */ /* #undef KERNEL_EXPORTS_X86_FPU */ /* TBD: fetch(3) support */ #if 0 /* whether the chosen libfetch is to be loaded at run-time */ #define LIBFETCH_DYNAMIC 1 /* libfetch is fetch(3) */ #define LIBFETCH_IS_FETCH 1 /* libfetch is libcurl */ #define LIBFETCH_IS_LIBCURL 0 /* soname of chosen libfetch */ #define LIBFETCH_SONAME "libfetch.so.6" #endif /* Define to the sub-directory where libtool stores uninstalled libraries. */ #define LT_OBJDIR ".libs/" /* make_request_fn() return type */ /* #undef MAKE_REQUEST_FN_RET */ /* struct shrink_control has nid */ /* #undef SHRINK_CONTROL_HAS_NID */ /* using complete_and_exit() instead */ /* #undef SPL_KTHREAD_COMPLETE_AND_EXIT */ /* Defined for legacy compatibility. */ #define SPL_META_ALIAS ZFS_META_ALIAS /* Defined for legacy compatibility. */ #define SPL_META_RELEASE ZFS_META_RELEASE /* Defined for legacy compatibility. */ #define SPL_META_VERSION ZFS_META_VERSION /* pde_data() is PDE_DATA() */ /* #undef SPL_PDE_DATA */ /* Define to 1 if all of the C89 standard headers exist (not just the ones required in a freestanding environment). This macro is provided for backward compatibility; new code need not use it. */ #define SYSTEM_FREEBSD 1 /* True if ZFS is to be compiled for a Linux system */ /* #undef SYSTEM_LINUX */ /* Version number of package */ /* #undef ZFS_DEBUG */ /* /dev/zfs minor */ /* #undef ZFS_DEVICE_MINOR */ /* enum node_stat_item contains NR_FILE_PAGES */ /* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_FILE_PAGES */ /* enum node_stat_item contains NR_INACTIVE_ANON */ /* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_INACTIVE_ANON */ /* enum node_stat_item contains NR_INACTIVE_FILE */ /* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_INACTIVE_FILE */ /* enum zone_stat_item contains NR_FILE_PAGES */ /* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_FILE_PAGES */ /* enum zone_stat_item contains NR_INACTIVE_ANON */ /* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_INACTIVE_ANON */ /* enum zone_stat_item contains NR_INACTIVE_FILE */ /* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_INACTIVE_FILE */ /* GENHD_FL_EXT_DEVT flag is not available */ /* #undef ZFS_GENHD_FL_EXT_DEVT */ /* GENHD_FL_NO_PART_SCAN flag is available */ /* #undef ZFS_GENHD_FL_NO_PART */ /* global_node_page_state() exists */ /* #undef ZFS_GLOBAL_NODE_PAGE_STATE */ /* global_zone_page_state() exists */ /* #undef ZFS_GLOBAL_ZONE_PAGE_STATE */ /* Define to 1 if GPL-only symbols can be used */ /* #undef ZFS_IS_GPL_COMPATIBLE */ /* Define the project alias string. */ -#define ZFS_META_ALIAS "zfs-2.2.99-440-FreeBSD_g90ba19eb7" +#define ZFS_META_ALIAS "zfs-2.2.99-456-FreeBSD_g1f940de07" /* Define the project author. */ #define ZFS_META_AUTHOR "OpenZFS" /* Define the project release date. */ /* #undef ZFS_META_DATA */ /* Define the maximum compatible kernel version. */ -#define ZFS_META_KVER_MAX "6.7" +#define ZFS_META_KVER_MAX "6.8" /* Define the minimum compatible kernel version. */ #define ZFS_META_KVER_MIN "3.10" /* Define the project license. */ #define ZFS_META_LICENSE "CDDL" /* Define the libtool library 'age' version information. */ /* #undef ZFS_META_LT_AGE */ /* Define the libtool library 'current' version information. */ /* #undef ZFS_META_LT_CURRENT */ /* Define the libtool library 'revision' version information. */ /* #undef ZFS_META_LT_REVISION */ /* Define the project name. */ #define ZFS_META_NAME "zfs" /* Define the project release. */ -#define ZFS_META_RELEASE "440-FreeBSD_g90ba19eb7" +#define ZFS_META_RELEASE "456-FreeBSD_g1f940de07" /* Define the project version. */ #define ZFS_META_VERSION "2.2.99" /* count is located in percpu_ref.data */ /* #undef ZFS_PERCPU_REF_COUNT_IN_DATA */ diff --git a/sys/modules/zfs/zfs_gitrev.h b/sys/modules/zfs/zfs_gitrev.h index 98948b94d273..e0d3575a6230 100644 --- a/sys/modules/zfs/zfs_gitrev.h +++ b/sys/modules/zfs/zfs_gitrev.h @@ -1 +1 @@ -#define ZFS_META_GITREV "zfs-2.2.99-440-g90ba19eb7" +#define ZFS_META_GITREV "zfs-2.2.99-456-g1f940de07"