[ELF] Add .debug* and .comment sections to the list of sections ignored by GC.

[evgeny777]

Summary

The name says it.

[• rafael]

ELF/MarkLive.cpp

72 ↗

(On Diff #71843)

Is this tested in this patch? Can in be in a followup?

[evgeny777]

ELF/MarkLive.cpp

72 ↗

(On Diff #71843)

This is needed for .debug_lines section, because it has relocations to comdat sections. I'll update test case for this change.

[evgeny777]

Now this just adds two patterns to isReserved(). Rafael, can you please look at this?

[• rafael]

Sorry, this had dropped out of my inbox.

This patch would require updating the comment

Sections listed below are special because they are used by the loader
just by being in an ELF file. They should not be garbage-collected.

since it adds another reason for not gcing.

But I think we can make it far more general and avoid yet another section name check: Just GC SHF_ALLOC sections.

In practice most non SHF_ALLOC sections are debug info. I remember some time back Cary Coutant had a proposal for how to gc them, but it will be a long time before we get there.

[evgeny777]

Diff updated.

[• rafael]

ELF/MarkLive.cpp

169 ↗

(On Diff #72296)

you need to update this comment.

test/ELF/gc-sections.s

17 ↗

(On Diff #72296)

You are only testing the case without --gc-sections

[evgeny777]

Addressed review comments

[evgeny777]

test/ELF/gc-sections.s

17 ↗

(On Diff #72296)

oops, missed that

[• rafael]

LGTM

lld/trunk/ELF/MarkLive.cpp

scanEhFrameSection(EhInputSection<ELFT> &EH,
 
   ELFFile<ELFT> &EObj = EH.getFile()->getObj();
   if (EH.RelocSection->sh_type == SHT_RELA)
     scanEhFrameSection(EH, EObj.relas(EH.RelocSection), Enqueue);
   else
     scanEhFrameSection(EH, EObj.rels(EH.RelocSection), Enqueue);
 }
 
-// Sections listed below are special because they are used by the loader
-// just by being in an ELF file. They should not be garbage-collected.
+// We do not garbage-collect two types of sections:
+// 1) Sections used by the loader (.init, .fini, .ctors, .dtors, .jcr)
+// 2) Not allocatable sections which typically contain debugging information
 template <class ELFT> static bool isReserved(InputSectionBase<ELFT> *Sec) {
   switch (Sec->getSectionHdr()->sh_type) {
   case SHT_FINI_ARRAY:
   case SHT_INIT_ARRAY:
   case SHT_NOTE:
   case SHT_PREINIT_ARRAY:
     return true;
   default:
     StringRef S = Sec->Name;
 
     // We do not want to reclaim sections if they can be referred
     // by __start_* and __stop_* symbols.
     if (isValidCIdentifier(S))
       return true;
 
-    return S.startswith(".ctors") || S.startswith(".dtors") ||
+    bool IsAllocSec = Sec->getSectionHdr()->sh_flags & SHF_ALLOC;
+    return !IsAllocSec || S.startswith(".ctors") || S.startswith(".dtors") ||
            S.startswith(".init") || S.startswith(".fini") ||
            S.startswith(".jcr");
   }
 }
 
 // This is the main function of the garbage collector.
 // Starting from GC-root sections, this function visits all reachable
 // sections to set their "Live" bits.

lld/trunk/test/ELF/gc-sections.s

 # REQUIRES: x86
 
 # RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t
 # RUN: ld.lld %t -o %t2
 # RUN: llvm-readobj -sections -symbols %t2 | FileCheck -check-prefix=NOGC %s
 # RUN: ld.lld --gc-sections %t -o %t2
 # RUN: llvm-readobj -sections -symbols %t2 | FileCheck -check-prefix=GC1 %s
 # RUN: ld.lld --export-dynamic --gc-sections %t -o %t2
 # RUN: llvm-readobj -sections -symbols %t2 | FileCheck -check-prefix=GC2 %s
 
 # NOGC: Name: .eh_frame
 # NOGC: Name: .text
 # NOGC: Name: .ctors
 # NOGC: Name: .dtors
 # NOGC: Name: .init
 # NOGC: Name: .fini
+# NOGC: Name: .debug_pubtypes
+# NOGC: Name: .comment
 # NOGC: Name: a
 # NOGC: Name: b
 # NOGC: Name: c
 # NOGC: Name: x
 # NOGC: Name: y
 # NOGC: Name: __preinit_array_start
 # NOGC: Name: __preinit_array_end
 # NOGC: Name: d
 
 # GC1:     Name: .eh_frame
 # GC1:     Name: .text
 # GC1:     Name: .ctors
 # GC1:     Name: .dtors
 # GC1:     Name: .init
 # GC1:     Name: .fini
+# GC1:     Name: .debug_pubtypes
+# GC1:     Name: .comment
 # GC1:     Name: a
 # GC1:     Name: b
 # GC1:     Name: c
 # GC1-NOT: Name: x
 # GC1-NOT: Name: y
 # GC1:     Name: __preinit_array_start
 # GC1:     Name: __preinit_array_end
 # GC1-NOT: Name: d
 
 # GC2:     Name: .eh_frame
 # GC2:     Name: .text
 # GC2:     Name: .ctors
 # GC2:     Name: .dtors
 # GC2:     Name: .init
 # GC2:     Name: .fini
+# GC2:     Name: .debug_pubtypes
+# GC2:     Name: .comment
 # GC2:     Name: a
 # GC2:     Name: b
 # GC2:     Name: c
 # GC2-NOT: Name: x
 # GC2-NOT: Name: y
 # GC2:     Name: __preinit_array_start
 # GC2:     Name: __preinit_array_end
 # GC2:     Name: d

 .section .fini,"aw",@fini_array
   .quad 0
 
 .section .preinit_array,"aw",@preinit_array
   .quad 0
 
 .section .eh_frame,"a",@unwind
   .quad 0
+
+.section .debug_pubtypes,"",@progbits
+  .quad 0
+
+.section .comment,"MS",@progbits,8
+  .quad 0