[mips][freebsd] Explicitly select PC-relative Personality and LSDA encodings.

[dsanders]

Summary

The Mips port of FreeBSD is currently using an old linker (GNU ld 2.17.50
[FreeBSD] 2007-07-03) which seems to be unable to fully support read-only
.eh_frame sections. However, there exists a patch to fix this (see PR25176)
which adds the missing R_MIPS_PC32 relocation and also seems to fix some
.eh_frame-related crashes.

This patch makes clang use appropriate encodings to support read-only .eh_frame
sections with this patched linker.

[emaste]

lib/MC/MCObjectFileInfo.cpp

374

Perhaps we should be explicit about this applying to bfd ld 2.17.50. Someone could consider reverting this in a few years once we have a contemporary LLD as the system linker.

[dsanders]

lib/MC/MCObjectFileInfo.cpp

374

Given the choice I'd prefer to change the Linux encodings to match FreeBSD at some point in the future. MIPS Linux seems to be the only target that has to ask for absolute encodings in order to get pc-relative encodings.

[• rafael]

Given the choice I'd prefer to change the Linux encodings to match FreeBSD at some point in the future. MIPS Linux seems to be the only target that has to ask for absolute encodings in order to get pc-relative encodings.

Can we do it now? I.E, always ask for pc relative?

Cheers,
Rafael

[dsanders]

ping. Is it ok to make FreeBSD ask for the right thing and leave the Linux workaround in until our tools accept the right thing?

[dsanders]

ping

[seanbruno]

Anything further I can do here?

[dsanders]

Anything further I can do here?

I think the patch is good but I need someone else to review it and say 'LGTM' before I can commit it.

[seanbruno]

So, should FreeBSD run this as a local patch or should we commit this into LLVM as a workaround for our old and crufty LD?

[llvm-commits]

It is unclear. Without knowing what is an actual gas/ld bug and what
is a reasonable limitation. Why is gas on linux rejecting indirect
pcrel? Is it at least acknowledge as a bug or are they aware of some
reason to reject it that we are not?

Cheers,
Rafael

[dsanders]

Why is gas on linux rejecting indirect pcrel? Is it at least acknowledge as a bug or are they aware of some
reason to reject it that we are not?

I don't know the reason but it's been like it a long time.

I don't understand why the behaviour of Linux should prevent us from fixing FreeBSD. Is there something I'm missing?

[seanbruno]

The testcase needs to be refreshed against -head. Otherwise, this is good.

[emaste]

See D23113

[emaste]

@brooks @arichardson is this something we want to revisit?

[emaste]

In D14700#1597957, @emaste wrote:

@brooks @arichardson is this something we want to revisit?

nvm, was committed as rL277732

[emaste]

And moved to CodeGen in rL339397

lib/MC/MCObjectFileInfo.cpp

case Triple::mips64el:
     // generated for relocation.
     PersonalityEncoding = dwarf::DW_EH_PE_indirect;
     // FIXME: The N64 ABI probably ought to use DW_EH_PE_sdata8 but we can't
     //        identify N64 from just a triple.
     TTypeEncoding = dwarf::DW_EH_PE_indirect | dwarf::DW_EH_PE_pcrel |
                     dwarf::DW_EH_PE_sdata4;
     // We don't support PC-relative LSDA references in GAS so we use the default
     // DW_EH_PE_absptr for those.
+
+    // FreeBSD must be explicit about the data size and using pcrel since it's
+    // assembler/linker won't do the automatic conversion that the Linux tools
+    // do.
+    if (T.isOSFreeBSD()) {
+      PersonalityEncoding |= dwarf::DW_EH_PE_pcrel | dwarf::DW_EH_PE_sdata4;
+      LSDAEncoding = dwarf::DW_EH_PE_pcrel | dwarf::DW_EH_PE_sdata4;
+    }
     break;
   case Triple::ppc64:
   case Triple::ppc64le:
     PersonalityEncoding = dwarf::DW_EH_PE_indirect | dwarf::DW_EH_PE_pcrel |
       dwarf::DW_EH_PE_udata8;
     LSDAEncoding = dwarf::DW_EH_PE_pcrel | dwarf::DW_EH_PE_udata8;
     TTypeEncoding = dwarf::DW_EH_PE_indirect | dwarf::DW_EH_PE_pcrel |
       dwarf::DW_EH_PE_udata8;

test/CodeGen/Mips/ehframe-indirect.ll

-; RUN: llc -mtriple=mipsel-linux-gnu < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=O32 %s
-; RUN: llc -mtriple=mipsel-linux-android < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=O32 %s
-; RUN: llc -mtriple=mips64el-linux-gnu -target-abi=n32 < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N32 %s
-; RUN: llc -mtriple=mips64el-linux-android -target-abi=n32 < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N32 %s
-; RUN: llc -mtriple=mips64el-linux-gnu < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N64 %s
-; RUN: llc -mtriple=mips64el-linux-android < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N64 %s
+; RUN: llc -mtriple=mipsel-linux-gnu < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=O32 %s
+; RUN: llc -mtriple=mipsel-linux-android < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=O32 %s
+; RUN: llc -mtriple=mips64el-linux-gnu -target-abi=n32 < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=N32 %s
+; RUN: llc -mtriple=mips64el-linux-android -target-abi=n32 < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=N32 %s
+; RUN: llc -mtriple=mips64el-linux-gnu < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=N64 %s
+; RUN: llc -mtriple=mips64el-linux-android < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=N64 %s
+; RUN: llc -mtriple=mips64el-linux-gnu < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=LINUX -check-prefix=N64 %s
+; RUN: llc -mtriple=mips64-unknown-freebsd11.0 < %s -asm-verbose | \
+; RUN:     FileCheck -check-prefix=ALL -check-prefix=FREEBSD -check-prefix=N64 %s
 
 @_ZTISt9exception = external constant i8*
 
 define i32 @main() personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
 ; ALL: .cfi_startproc
-; ALL: .cfi_personality 128, DW.ref.__gxx_personality_v0
+
+; Linux must rely on the assembler/linker converting the encodings.
+; LINUX: .cfi_personality 128, DW.ref.__gxx_personality_v0
+; LINUX: .cfi_lsda 0, $exception0
+
+; FreeBSD can (and must) be more direct about the encodings it wants.
+; FREEBSD: .cfi_personality 155, DW.ref.__gxx_personality_v0
+; FREEBSD: .cfi_lsda 27, $exception0
 
 entry:
   invoke void @foo() to label %cont unwind label %lpad
 ; ALL: foo
 ; ALL: jalr
 
 lpad:
   %0 = landingpad { i8*, i32 }