[lld][ELF] Support for zero flag section groups

[phosek]

Summary

This change introduces support for zero flag ELF section groups to lld.
lld already supports COMDAT sections, which in ELF are a special type of
ELF section groups. These are generally useful to enable linker GC where
you want a group of sections to always travel together, that is to be
either retained or discarded as a whole, but without the COMDAT
semantics. Other ELF linkers already support zero flag ELF section
groups and this change helps us reach feature parity.

[MaskRay]

lld/test/ELF/gc-sections-group-debug.s

11 ↗

(On Diff #323482)

This tests a real world use case, .debug_types which has the GRP_COMDAT flag.

You can add another section to gc-sections-group.s

[MaskRay]

LGTM.

[jhenderson]

Looks good, with a couple of small comments.

lld/ELF/InputFiles.cpp

621

I wonder whether isNew is a bit misnamed nowadays. Perhaps it should be renamed to keepGroup or something like that. Waht do you think?

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

65

[phosek]

lld/ELF/InputFiles.cpp

621

That makes sense, I've renamed this to keepGroup.

[• galapoz1996]

This comment was removed by mehdi_amini.

lld/ELF/InputFiles.cpp

for (size_t i = 0, e = objSections.size(); i < e; ++i) {
     }
 
     switch (sec.sh_type) {
     case SHT_GROUP: {
       // De-duplicate section groups by their signatures.
       StringRef signature = getShtGroupSignature(objSections, sec);
       this->sections[i] = &InputSection::discarded;
 
-
       ArrayRef<Elf_Word> entries =
           CHECK(obj.template getSectionContentsAsArray<Elf_Word>(sec), this);
       if (entries.empty())
         fatal(toString(this) + ": empty SHT_GROUP");
 
-      // The first word of a SHT_GROUP section contains flags. Currently,
-      // the standard defines only "GRP_COMDAT" flag for the COMDAT group.
-      // An group with the empty flag doesn't define anything; such sections
-      // are just skipped.
-      if (entries[0] == 0)
-        continue;
-
-      if (entries[0] != GRP_COMDAT)
+      Elf_Word flag = entries[0];
+      if (flag && flag != GRP_COMDAT)
         fatal(toString(this) + ": unsupported SHT_GROUP format");
 
       bool isNew =
-          ignoreComdats ||
+          (flag & GRP_COMDAT) == 0 || ignoreComdats ||
           symtab->comdatGroups.try_emplace(CachedHashStringRef(signature), this)
               .second;
       if (isNew) {
         if (config->relocatable)
           this->sections[i] = createInputSection(sec);
         selectedGroups.push_back(entries);
         continue;
       }

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

 # REQUIRES: x86
 ## Check that group members are retained or discarded as a unit, and
 ## non-SHF_ALLOC sections in a group are subject to garbage collection,
 ## if at least one member has the SHF_ALLOC flag.
 
 # RUN: llvm-mc -filetype=obj -triple=x86_64 %s -o %t.o
 # RUN: ld.lld --gc-sections %t.o -o %t.dead
 # RUN: llvm-readobj -S %t.dead | FileCheck %s --check-prefix=CHECK-DEAD
 
-## .mynote.bar is retained because it is not in a group.
-# CHECK-DEAD-NOT: Name: .myanote.foo
-# CHECK-DEAD-NOT: Name: .mytext.foo
-# CHECK-DEAD-NOT: Name: .mybss.foo
-# CHECK-DEAD-NOT: Name: .mynote.foo
-# CHECK-DEAD:     Name: .mynote.bar
-
-# RUN: ld.lld --gc-sections %t.o -o %t -e anote_foo
-# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE
-# RUN: ld.lld --gc-sections %t.o -o %t -e foo
-# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE
-# RUN: ld.lld --gc-sections %t.o -o %t -e bss_foo
-# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE
+## .mynote.none is retained because it is not in a group.
+# CHECK-DEAD-NOT: Name: .myanote.cmdt
+# CHECK-DEAD-NOT: Name: .mytext.cmdt
+# CHECK-DEAD-NOT: Name: .mybss.cmdt
+# CHECK-DEAD-NOT: Name: .mynote.cmdt
+# CHECK-DEAD-NOT: Name: .myanote.grp
+# CHECK-DEAD-NOT: Name: .mytext.grp
+# CHECK-DEAD-NOT: Name: .mybss.grp
+# CHECK-DEAD-NOT: Name: .mynote.grp
+# CHECK-DEAD:     Name: .mynote.none
+
+# RUN: ld.lld --gc-sections %t.o -o %t -e anote_cmdt
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-COMDAT
+# RUN: ld.lld --gc-sections %t.o -o %t -e cmdt
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-COMDAT
+# RUN: ld.lld --gc-sections %t.o -o %t -e bss_cmdt
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-COMDAT
 
-## note_foo as the entry point does not make much sense because it is defined
+## note_cmdt as the entry point does not make much sense because it is defined
 ## in a non-SHF_ALLOC section. This is just to demonstrate the behavior.
-# RUN: ld.lld --gc-sections %t.o -o %t -e note_foo
-# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE
+# RUN: ld.lld --gc-sections %t.o -o %t -e note_cmdt
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-COMDAT
 
-# CHECK-LIVE: Name: .myanote.foo
-# CHECK-LIVE: Name: .mytext.foo
-# CHECK-LIVE: Name: .mybss.foo
-# CHECK-LIVE: Name: .mynote.foo
-# CHECK-LIVE: Name: .mynote.bar
+# CHECK-LIVE-COMDAT: Name: .myanote.cmdt
+# CHECK-LIVE-COMDAT: Name: .mytext.cmdt
+# CHECK-LIVE-COMDAT: Name: .mybss.cmdt
+# CHECK-LIVE-COMDAT: Name: .mynote.cmdt
+# CHECK-LIVE-COMDAT-NOT: Name: .myanote.grp
+# CHECK-LIVE-COMDAT-NOT: Name: .mytext.grp
+# CHECK-LIVE-COMDAT-NOT: Name: .mybss.grp
+# CHECK-LIVE-COMDAT-NOT: Name: .mynote.grp
+# CHECK-LIVE-COMDAT: Name: .mynote.none
+
+# RUN: ld.lld --gc-sections %t.o -o %t -e anote_grp
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-GROUP
+# RUN: ld.lld --gc-sections %t.o -o %t -e grp
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-GROUP
+# RUN: ld.lld --gc-sections %t.o -o %t -e bss_grp
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-GROUP
 
-.globl anote_foo, foo, bss_foo, note_foo
+## note_zero as the entry point does not make much sense because it is defined
+## in a non-SHF_ALLOC section. This is just to demonstrate the behavior.
+# RUN: ld.lld --gc-sections %t.o -o %t -e note_grp
+# RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=CHECK-LIVE-GROUP
+
+# CHECK-LIVE-GROUP-NOT: Name: .myanote.cmdt
+# CHECK-LIVE-GROUP-NOT: Name: .mytext.cmdt
+# CHECK-LIVE-GROUP-NOT: Name: .mybss.cmdt
+# CHECK-LIVE-GROUP-NOT: Name: .mynote.cmdt
+# CHECK-LIVE-GROUP: Name: .myanote.grp
+# CHECK-LIVE-GROUP: Name: .mytext.grp
+# CHECK-LIVE-GROUP: Name: .mybss.grp
+# CHECK-LIVE-GROUP: Name: .mynote.grp
+# CHECK-LIVE-GROUP: Name: .mynote.none
+
+## These seections are in a COMDAT group `cmdt`.
+.globl anote_cmdt, cmdt, bss_cmdt, note_cmdt
+
+.section .myanote.cmdt,"aG",@note,cmdt,comdat
+anote_cmdt:
+.byte 0
+
+.section .mytext.cmdt,"axG",@progbits,cmdt,comdat
+cmdt:
+.byte 0
+
+.section .mybss.cmdt,"awG",@nobits,cmdt,comdat
+bss_cmdt:
+.byte 0
+
+.section .mynote.cmdt,"G",@note,cmdt,comdat
+note_cmdt:
+.byte 0
+
+## These sections are in a zero flag group `grp`.
+.globl anote_grp, grp, bss_grp, note_grp
 
-.section .myanote.foo,"aG",@note,foo,comdat
-anote_foo:
+.section .myanote.grp,"aG",@note,grp
+anote_grp:
 .byte 0
 
-.section .mytext.foo,"axG",@progbits,foo,comdat
-foo:
+.section .mytext.grp,"axG",@progbits,grp
+grp:
 .byte 0
 
-.section .mybss.foo,"awG",@nobits,foo,comdat
-bss_foo:
+.section .mybss.grp,"awG",@nobits,grp
+bss_grp:
 .byte 0
 
-.section .mynote.foo,"G",@note,foo,comdat
-note_foo:
+.section .mynote.grp,"G",@note,grp
+note_grp:
 .byte 0
 
-.section .mynote.bar,"",@note
+## This section isn't in any group.
+.section .mynote.none,"",@note
 .byte 0