Add support for rtld-elf on arm64
ClosedPublic
Actions

Authored by andrew on Mar 25 2015, 3:12 PM.

Details

Reviewers

kib
emaste

Commits

rS280903: Add the arm64 code to the runtime linker. It's not able to be built as we

Summary

This adds the MD parts of rtld for arm64.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Skipped

Unit

Tests Skipped

Event Timeline

andrew updated this revision to Diff 4399.Mar 25 2015, 3:12 PM

andrew retitled this revision from to Add support for rtld-elf on arm64.

andrew updated this object.

andrew edited the test plan for this revision. (Show Details)

andrew added reviewers: emaste, kib.

Herald added subscribers: andrew, imp. · View Herald TranscriptMar 25 2015, 3:12 PM

imp added inline comments.Mar 26 2015, 6:19 PM

libexec/rtld-elf/aarch64/reloc.c
2 ↗	(On Diff #4399)	How much of this was copied from Rui's work? It looks mostly new.
202 ↗	(On Diff #4399)	same alignment question.
206–207 ↗	(On Diff #4399)	same alignment question.
210 ↗	(On Diff #4399)	do you need to print out rela->r_info to help diagnose this in the future?
249 ↗	(On Diff #4399)	Same alignment question as below.
251 ↗	(On Diff #4399)	Above we handle a number of different types of relation in what looks to be the same list. Yet here we just punt. Do we need to handle the TLSDESC here too or no? If not, a comment about why might be useful.
265 ↗	(On Diff #4399)	Why not return -1 or some other error indicator to show we need to implement this?
275 ↗	(On Diff #4399)	why not return -1 or some error indicator that we'll need to implement this?
311 ↗	(On Diff #4399)	why not return -1 here?
334 ↗	(On Diff #4399)	Is the alignment of these two relocation types guaranteed to be sufficient we don't have to copy them?
349 ↗	(On Diff #4399)	So all we do is check to see if this is not in the main program, and ignore it if it is? That seems strange. However, looking above I see that it is handled specially there, so a pointer to that / comment here would help.
352 ↗	(On Diff #4399)	same alignment question.

Missing change?

--- a/libexec/rtld-elf/rtld.c
+++ b/libexec/rtld-elf/rtld.c
@@ -4351,8 +4351,7 @@ tls_get_addr_common(Elf_Addr **dtvp, int index, size_t offset)
        return (tls_get_addr_slow(dtvp, index, offset));
 }

-#if defined(__arm__) || defined(__aarch64__) || defined(__mips__) || \
-    defined(__powerpc__)
+#if defined(__arm__) || defined(__mips__) || defined(__powerpc__)

 /*
  * Allocate Static TLS using the Variant I method.

Defer symbol lookup, & fix a few issues found while adding this.
Use reloc_non_plt to relocate rtld

It is somewhat pity that ifuncs are left out, but it is probably more important to get the initial port going.

libexec/rtld-elf/aarch64/reloc.c
148 ↗	(On Diff #4519)	Shouldn't this situation also result in return (-1) ?
294 ↗	(On Diff #4519)	The break is excessive.
311 ↗	(On Diff #4519)	Remove the empy line.