Differential D41154
vm_phys_enqueue_contig: handle zero pages Authored by dougm on Jul 22 2023, 10:31 PM. Tags None Referenced Files
Details By letting vm_phys_enqueue_contig handle the case when npages == 0, the callers can stop checking it, and the compiler can stop zero-checking with every call to ffs(). Letting vm_phys_enqueue_contig call vm_phys_enqueue_contig for part of its work also saves a few bytes. The amd64 object code shrinks by 80 bytes. Peter, can you test this please?
Diff Detail
Event TimelineComment Actions Using objdump outputs to compare start addresses of vm_phys functions before and after this change Before After 0000000000000090 <vm_phys_mem_affinity>: 0000000000000090 <vm_phys_mem_affinity>: 00000000000000c0 <vm_phys_add_seg>: 00000000000000c0 <vm_phys_add_seg>: 0000000000000130 <vm_phys_create_seg>: 0000000000000130 <vm_phys_create_seg>: 0000000000000350 <vm_phys_init>: 0000000000000350 <vm_phys_init>: 0000000000000870 <vm_phys_register_domains>: 0000000000000870 <vm_phys_register_domains>: 0000000000000930 <vm_phys_alloc_npages>: 0000000000000930 <vm_phys_alloc_npages>: 0000000000000f00 <vm_phys_alloc_pages>: | 0000000000000ec0 <vm_phys_alloc_pages>: 0000000000000f70 <vm_phys_alloc_freelist_pages>: | 0000000000000f30 <vm_phys_alloc_freelist_pages>: 00000000000011d0 <vm_phys_paddr_to_vm_page>: | 0000000000001190 <vm_phys_paddr_to_vm_page>: 0000000000001250 <vm_phys_fictitious_to_vm_page>: | 0000000000001210 <vm_phys_fictitious_to_vm_page>: 00000000000012c0 <vm_phys_fictitious_reg_range>: | 0000000000001280 <vm_phys_fictitious_reg_range>: 0000000000001790 <vm_phys_fictitious_unreg_range>: | 0000000000001750 <vm_phys_fictitious_unreg_range>: 0000000000001b50 <vm_phys_free_pages>: | 0000000000001b10 <vm_phys_free_pages>: 0000000000001cf0 <vm_phys_enqueue_contig>: | 0000000000001cb0 <vm_phys_enqueue_contig>: 0000000000001e90 <vm_phys_free_contig>: | 0000000000001e40 <vm_phys_free_contig>: 0000000000001f80 <vm_phys_find_range>: | 0000000000001f30 <vm_phys_find_range>: 0000000000002060 <vm_phys_unfree_page>: | 0000000000002010 <vm_phys_unfree_page>: 00000000000021f0 <vm_phys_alloc_contig>: | 00000000000021a0 <vm_phys_alloc_contig>: 0000000000002730 <vm_phys_avail_largest>: | 00000000000026e0 <vm_phys_avail_largest>: 0000000000002790 <vm_phys_avail_size>: | 0000000000002740 <vm_phys_avail_size>: 00000000000027b0 <vm_phys_is_dumpable>: | 0000000000002760 <vm_phys_is_dumpable>: 0000000000002890 <vm_phys_early_add_seg>: | 0000000000002840 <vm_phys_early_add_seg>: 00000000000028f0 <vm_phys_early_alloc>: | 00000000000028a0 <vm_phys_early_alloc>: 0000000000002b30 <vm_phys_avail_split>: | 0000000000002ae0 <vm_phys_avail_split>: 0000000000002ce0 <vm_phys_early_startup>: | 0000000000002c90 <vm_phys_early_startup>: 0000000000002f30 <sysctl_vm_phys_free>: | 0000000000002ee0 <sysctl_vm_phys_free>: 0000000000003120 <sysctl_vm_phys_segs>: | 00000000000030d0 <sysctl_vm_phys_segs>: 0000000000003220 <sysctl_vm_phys_locality>: | 00000000000031d0 <sysctl_vm_phys_locality>: Comment Actions With D41154.125484.patch I now see: igb0: <Intel(R) I350 (Copper)> port 0xd020-0xd03f mem 0xfb320000-0xfb33ffff,0xfb344000-0xfb347fff irq 16 at device 0.0 on pci8 igb0: EEPROM V1.63-0 eTrack 0x800009fa igb0: Using 1024 TX descriptors and 1024 RX descriptors igb0: queue equality override not set, capping rx_queues at 6 and tx_queues at 6 igb0: Usin db> show panic panic: vm_phys_enq_range: page 0xfffffe0002fd4c18 and npages 0 are misaligned db> x/s version version: FreeBSD 14.0-CURRENT #0 main-n264491-8a5c836b51ce2-dirty: Thu Aug 3 06:23:37 CEST 2023\012 pho@mercat1.netperf.freebsd.org:/usr/sr db> bt Tracing pid 37 tid 100210 td 0xfffffe010e620900 kdb_enter() at kdb_enter+0x32/frame 0xfffffe010db821f0 vpanic() at vpanic+0x17a/frame 0xfffffe010db82240 panic() at panic+0x43/frame 0xfffffe010db822a0 vm_phys_enq_range() at vm_phys_enq_range+0x11d/frame 0xfffffe010db822b0 vm_phys_alloc_npages() at vm_phys_alloc_npages+0x5a6/frame 0xfffffe010db82320 vm_page_zone_import() at vm_page_zone_import+0xa7/frame 0xfffffe010db82370 zone_alloc_item() at zone_alloc_item+0x78/frame 0xfffffe010db823c0 vm_page_alloc_domain_after() at vm_page_alloc_domain_after+0x151/frame 0xfffffe010db82450 vm_page_alloc() at vm_page_alloc+0x64/frame 0xfffffe010db824b0 vm_fault_allocate() at vm_fault_allocate+0x1a5/frame 0xfffffe010db82520 vm_fault() at vm_fault+0x448/frame 0xfffffe010db82650 vm_fault_trap() at vm_fault_trap+0x6f/frame 0xfffffe010db82690 trap_pfault() at trap_pfault+0x24a/frame 0xfffffe010db82700 calltrap() at calltrap+0x8/frame 0xfffffe010db82700 --- trap 0xc, rip = 0xffffffff8104db36, rsp = 0xfffffe010db827d0, rbp = 0xfffffe010db827d0 --- copyout_nosmap_erms() at copyout_nosmap_erms+0x156/frame 0xfffffe010db827d0 elf64_load_sections() at elf64_load_sections+0x283/frame 0xfffffe010db82860 elf64_load_file() at elf64_load_file+0x280/frame 0xfffffe010db828c0 elf64_load_interp() at elf64_load_interp+0x67/frame 0xfffffe010db82900 exec_elf64_imgact() at exec_elf64_imgact+0xe01/frame 0xfffffe010db829f0 kern_execve() at kern_execve+0x772/frame 0xfffffe010db82d80 sys_execve() at sys_execve+0x58/frame 0xfffffe010db82e00 amd64_syscall() at amd64_syscall+0x14f/frame 0xfffffe010db82f30 fast_syscall_common() at fast_syscall_common+0xf8/frame 0xfffffe010db82f30 --- syscall (59, FreeBSD ELF64, execve), rip = 0x254589a5b30a, rsp = 0x2545874dfdf8, rbp = 0x2545874dff40 --- db> db> show proc Process 37 (sh) at 0xfffffe010e96b000: state: NORMAL uid: 0 gids: 0 parent: pid 20 at 0xfffffe003ce595c0 ABI: FreeBSD ELF64 flag: 0x14000002 flag2: 0 arguments: /bin/sh -o verify /etc/rc autoboot reaper: 0xfffffe00e4c65040 reapsubtree: 1 sigparent: 20 vmspace: 0xfffffe010e95a860 (map 0xfffffe010e95a860) (map.pmap 0xfffffe010e95a920) (pmap 0xfffffe010e95a990) threads: 1 100210 Run CPU 0 sh db> `` Comment Actions I ran a buildkernel and a few other test with D41154.125493.patch. |