[ELF] - Do not merge relocation sections by name when using --emit-relocs.

[grimar]

Summary

Previously we would merge relocation sections by name.
That did not work in some cases, like testcase shows.

Patch implements logic to merge relocation sections if their target
sections were merged into the same output section.

[ruiu]

ELF/OutputSections.cpp

345 ↗

(On Diff #101185)

Use auto.

ELF/OutputSections.h

70–71 ↗

(On Diff #101185)

Pointer to a relocation section for this section. Usually nullptr because we consume relocations, but if --emit-relocs is specified (which is rare), it may have a non-null value.

test/ELF/emit-relocs-merge.test

1 ↗

(On Diff #101185)

Can you create this using llvm-mc?

[grimar]

test/ELF/emit-relocs-merge.test

1 ↗

(On Diff #101185)

I do not think so. Idea here was to create relocation section that does not have name starting from ".rel" or ".rela".
I do not think llvm-mc allows to do that.

[ruiu]

test/ELF/emit-relocs-merge.test

1 ↗

(On Diff #101185)

But you can test this if you use a linker script, no? You are using` .zed : { *(.foo) *(.bar) }` as an example.

[grimar]

test/ELF/emit-relocs-merge.test

1 ↗

(On Diff #101185)

I sure can, but linkerscript case (a bit outdated) is a D33780. I am about to update this patch
to support linkerscript case, add testcase from D33780 with Rafael's comments addressed and remove
this yaml test then.

[grimar]

• Addressed review comments.
• Switched to linkerscript testcase.

ELF/OutputSections.cpp

345 ↗

(On Diff #101185)

Done.

ELF/OutputSections.h

70–71 ↗

(On Diff #101185)

Done.

[ruiu]

LGTM

lld/trunk/ELF/OutputSections.h

public:
   // Pointer to the first section in PT_LOAD segment, which this section
   // also resides in. This field is used to correctly compute file offset
   // of a section. When two sections share the same load segment, difference
   // between their file offsets should be equal to difference between their
   // virtual addresses. To compute some section offset we use the following
   // formula: Off = Off_first + VA - VA_first.
   OutputSection *FirstInPtLoad = nullptr;
 
+  // Pointer to a relocation section for this section. Usually nullptr because
+  // we consume relocations, but if --emit-relocs is specified (which is rare),
+  // it may have a non-null value.
+  OutputSection *RelocationSection = nullptr;
+
   // The following fields correspond to Elf_Shdr members.
   uint64_t Size = 0;
   uint64_t Offset = 0;
   uint64_t LMAOffset = 0;
   uint64_t Addr = 0;
   uint32_t ShName = 0;
 
   void addSection(InputSection *S);

lld/trunk/ELF/OutputSections.cpp

void OutputSectionFactory::addInputSec(InputSectionBase *IS,
   // through all SHT_GROUP sections without merging them because merging them
   // creates broken section contents.
   if (IS->Type == SHT_GROUP) {
     OutputSection *Out = nullptr;
     addInputSec(IS, OutsecName, Out);
     return;
   }
 
+  // Imagine .zed : { *(.foo) *(.bar) } script. Both foo and bar may have
+  // relocation sections .rela.foo and .rela.bar for example. Most tools do
+  // not allow multiple REL[A] sections for output section. Hence we
+  // should combine these relocation sections into single output.
+  // We skip synthetic sections because it can be .rela.dyn/.rela.plt or any
+  // other REL[A] sections created by linker itself.
+  if (!isa<SyntheticSection>(IS) &&
+      (IS->Type == SHT_REL || IS->Type == SHT_RELA)) {
+    auto *Sec = cast<InputSection>(IS);
+    OutputSection *Out = Sec->getRelocatedSection()->getOutputSection();
+    addInputSec(IS, OutsecName, Out->RelocationSection);
+    return;
+  }
+
   SectionKey Key = createKey(IS, OutsecName);
   OutputSection *&Sec = Map[Key];
   return addInputSec(IS, OutsecName, Sec);
 }
 
 void OutputSectionFactory::addInputSec(InputSectionBase *IS,
                                        StringRef OutsecName,
                                        OutputSection *&Sec) {

lld/trunk/ELF/Writer.cpp

StringRef elf::getOutputSectionName(StringRef Name) {
   // ".zdebug_" is a prefix for ZLIB-compressed sections.
   // Because we decompressed input sections, we want to remove 'z'.
   if (Name.startswith(".zdebug_"))
     return Saver.save("." + Name.substr(2));
 
   if (Config->Relocatable)
     return Name;
 
-  // If -emit-relocs is given (which is rare), we need to copy
-  // relocation sections to the output. If input section .foo is
-  // output as .bar, we want to rename .rel.foo .rel.bar as well.
-  if (Config->EmitRelocs) {
-    for (StringRef V : {".rel.", ".rela."}) {
-      if (Name.startswith(V)) {
-        StringRef Inner = getOutputSectionName(Name.substr(V.size() - 1));
-        return Saver.save(V.drop_back() + Inner);
-      }
-    }
-  }
-
   for (StringRef V :
        {".text.", ".rodata.", ".data.rel.ro.", ".data.", ".bss.rel.ro.",
         ".bss.", ".init_array.", ".fini_array.", ".ctors.", ".dtors.", ".tbss.",
         ".gcc_except_table.", ".tdata.", ".ARM.exidx."}) {
     StringRef Prefix = V.drop_back();
     if (Name.startswith(V) || Name == Prefix)
       return Prefix;
   }

lld/trunk/test/ELF/emit-relocs-merge.s

 # REQUIRES: x86
 # RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
 # RUN: ld.lld --emit-relocs %t.o -o %t.so -shared
 # RUN: llvm-readobj -r %t.so | FileCheck %s
 
 # CHECK:       Relocations [
 # CHECK-NEXT:    Section ({{.*}}) .rela.dyn {
 # CHECK-NEXT:     0x1000 R_X86_64_64 zed 0x0
 # CHECK-NEXT:     0x1008 R_X86_64_64 zed 0x0
 # CHECK-NEXT:   }
-# CHECK-NEXT:   Section ({{.*}}) .rela.data {
+# CHECK-NEXT:   Section ({{.*}}) .rela.data.foo {
 # CHECK-NEXT:     0x1000 R_X86_64_64 zed 0x0
 # CHECK-NEXT:     0x1008 R_X86_64_64 zed 0x0
 # CHECK-NEXT:   }
 # CHECK-NEXT: ]
 
 .section        .data.foo,"aw",%progbits
 .quad zed
 .section        .data.bar,"aw",%progbits
 .quad zed

lld/trunk/test/ELF/linkerscript/emit-relocs-multiple.s

+# REQUIRES: x86
+# RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t.o
+# RUN: echo "SECTIONS { .zed : { *(.foo) *(.bar) } }" > %t.script
+# RUN: ld.lld --emit-relocs --script %t.script %t.o -o %t1
+# RUN: llvm-readobj -r %t1 | FileCheck %s
+
+# CHECK:      Relocations [
+# CHECK-NEXT:   Section {{.*}} .rela.foo {
+# CHECK-NEXT:     0x1 R_X86_64_32 .zed 0x0
+# CHECK-NEXT:     0x6 R_X86_64_32 .zed 0x5
+# CHECK-NEXT:   }
+# CHECK-NEXT: ]
+
+.section .foo,"ax",@progbits
+aaa:
+  movl $aaa, %edx
+
+.section .bar,"ax",@progbits
+bbb:
+  movl $bbb, %edx