[PATCH] D18911: ELF: Emit .eh_frame_hdr entries for FDEs with duplicate PCs.

[Rafael Espíndola via llvm-commits]

How hard would it be to discard that part of the .eh_frame? Or at least not
include it in the index (reduce fde_count).

If I understand it correctly with this patch we create an empty range in
the binary search table, which is a bit odd.
On Apr 8, 2016 18:03, "Peter Collingbourne" <peter at pcc.me.uk> wrote:

> pcc created this revision.
> pcc added reviewers: ruiu, rafael.
> pcc added a subscriber: llvm-commits.
>
> It is possible to have FDEs with duplicate PCs if ICF was able to merge
> functions with FDEs, or if the input files for some reason contained
> duplicate
> FDEs.  We previously weren't handling this correctly when producing the
> contents of the .eh_frame_hdr section; we were dropping entries and leaving
> null entries at the end of the section, which confused consumers of unwind
> data, such as the backtrace() function.
>
> Fix the bug by collecting the FDEs into a std::multimap rather than a
> std::map.
>
> http://reviews.llvm.org/D18911
>
> Files:
>   ELF/OutputSections.cpp
>   test/ELF/eh-frame-hdr-icf.s
>
> Index: test/ELF/eh-frame-hdr-icf.s
> ===================================================================
> --- /dev/null
> +++ test/ELF/eh-frame-hdr-icf.s
> @@ -0,0 +1,26 @@
> +# REQUIRES: x86
> +
> +# RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t
> +# RUN: ld.lld %t -o %t2 --icf=all --eh-frame-hdr
> +# RUN: llvm-objdump -s %t2 | FileCheck %s
> +
> +# CHECK: Contents of section .eh_frame_hdr:
> +# CHECK-NEXT: 101a0 011b033b b4ffffff 02000000 600e0000
> +# CHECK-NEXT: 101b0 d0ffffff 600e0000 e8ffffff
> +#                   ^ FDE for f1      ^ FDE for f2
> +
> +.globl _start, f1, f2
> +_start:
> +  ret
> +
> +.section .text.f1, "ax"
> +f1:
> +  .cfi_startproc
> +  ret
> +  .cfi_endproc
> +
> +.section .text.f2, "ax"
> +f2:
> +  .cfi_startproc
> +  ret
> +  .cfi_endproc
> Index: ELF/OutputSections.cpp
> ===================================================================
> --- ELF/OutputSections.cpp
> +++ ELF/OutputSections.cpp
> @@ -746,11 +746,13 @@
>    Buf += 12;
>
>    // InitialPC -> Offset in .eh_frame, sorted by InitialPC.
> -  std::map<uintX_t, size_t> PcToOffset;
> +  // We use a multimap in order to handle duplicate PCs, which can happen
> if ICF
> +  // was able to merge functions with FDEs.
> +  std::multimap<uintX_t, size_t> PcToOffsets;
>    for (const FdeData &F : FdeList)
> -    PcToOffset[getFdePc(EhVA, F)] = F.Off;
> +    PcToOffsets.insert({getFdePc(EhVA, F), F.Off});
>
> -  for (auto &I : PcToOffset) {
> +  for (auto &I : PcToOffsets) {
>      // The first four bytes are an offset to the initial PC value for the
> FDE.
>      write32<E>(Buf, I.first - VA);
>      // The last four bytes are an offset to the FDE data itself.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160408/da07cde1/attachment.html>