[ELF] Fix missing ELF st_other field on versioned symbols

[adalava]

Summary

On PowerPC64 ELFv2 ABI, functions may have 2 entry points: global and local. The local entry point location of a function is stored in the st_other field of the symbol, as an offset relative to the global entry point.

This change compliments D56586, copying st_other bits from original symbol to its versioned alias and addresses https://bugs.llvm.org/show_bug.cgi?id=41048.

[adalava]

Example code:

test.c371 BDownload

Results Unpatched: (note missing 0x60 before symbol "openattt@FBSD_1.1")

`$ /usr/local/powerpc64-unknown-freebsd12.0/bin/objdump -t test.o
test.o: file format elf64-powerpc-freebsd

SYMBOL TABLE:
0000000000000000 l df *ABS* 0000000000000000 test.c
0000000000000000 l d .toc 0000000000000000 .toc
0000000000000000 *UND* 0000000000000000 .TOC.
0000000000000000 w F .text 000000000000002c 0x60 __impl_openattt
0000000000000000 g O .bss 0000000000000004 a
0000000000000000 g F .text 000000000000002c 0x60 openattt
0000000000000000 w F .text 000000000000002c openattt@FBSD_1.1`

Patched:

`$ /usr/local/powerpc64-unknown-freebsd12.0/bin/objdump -t test.o

test.o: file format elf64-powerpc-freebsd

SYMBOL TABLE:
0000000000000000 l df *ABS* 0000000000000000 test.c
0000000000000000 l d .toc 0000000000000000 .toc
0000000000000000 *UND* 0000000000000000 .TOC.
0000000000000000 w F .text 000000000000002c 0x60 __impl_openattt
0000000000000000 g O .bss 0000000000000004 a
0000000000000000 g F .text 000000000000002c 0x60 openattt
0000000000000000 w F .text 000000000000002c 0x60 openattt@FBSD_1.1`

[MaskRay]

Looks good, but do you have a test (e.g. in test/MC/PowerPC)?

[adalava]

add test case

[adalava]

Thank you @MaskRay, a test case was added. Please note this patch currently depends on D56586 and D61647 that need to go in first.

[MaskRay]

Thanks. The test contains some constructs that are not used, e.g. .symver .addrsig. And I think it is usually better to use llvm-readelf/llvm-readobj. I always find the output of llvm-objdump/objdump weird.. If you don't mind, I can rewrite/minimalize the test and commit for you.

[adalava]

In D59436#1494416, @MaskRay wrote:

Thanks. The test contains some constructs that are not used, e.g. .symver .addrsig. And I think it is usually better to use llvm-readelf/llvm-readobj. I always find the output of llvm-objdump/objdump weird.. If you don't mind, I can rewrite/minimalize the test and commit for you.

Sure, please go ahead, thank you!
Just a reminder that even using llvm-readobj, test case will depends on D56586 in order to work

[MaskRay]

This works when .set is placed after openattt:, but it doesn't work if .set __impl_openattt, openattt is placed before openattt:. I think this is still worth landing after we have tools to dump st_other (D61718 or a similar change in llvm-readobj).

[MaskRay]

I'll create ppc64-localentry-symver.s instead. We can add the test that .set is before .localentry when D56586 is committed.

llvm/lib/MC/ELFObjectWriter.cpp

for (const std::pair<StringRef, const MCSymbol *> &P : Asm.Symvers) {
     Asm.registerSymbol(*Alias);
     const MCExpr *Value = MCSymbolRefExpr::create(&Symbol, Asm.getContext());
     Alias->setVariableValue(Value);
 
     // Aliases defined with .symvar copy the binding from the symbol they alias.
     // This is the first place we are able to copy this information.
     Alias->setExternal(Symbol.isExternal());
     Alias->setBinding(Symbol.getBinding());
+    Alias->setOther(Symbol.getOther());
 
     if (!Symbol.isUndefined() && !Rest.startswith("@@@"))
       continue;
 
     // FIXME: Get source locations for these errors or diagnose them earlier.
     if (Symbol.isUndefined() && Rest.startswith("@@") &&
         !Rest.startswith("@@@")) {
       Asm.getContext().reportError(SMLoc(), "versioned symbol " + AliasName +