[ELF] Fix copy relocation when two symbols share the same Symbol instance.

[MaskRay]

Summary

In glibc libc.so.6, the multiple versions of sys_errlist share the same Symbol instance. When sys_errlist is copy relocated, we would replace SharedSymbol with Defined in the first iteration of the following loop:

for (SharedSymbol *Sym : getSymbolsAt<ELFT>(SS))

Then in the second iteration, we think the symbol (which has been changed to Defined) is still SharedSymbol and screw up (the address ends up in the Size field).

[brooksmoses]

While you're doing that, maybe also fix "refer different places" in the comment on line 447 to be "refer to different places". :)

[MaskRay]

Typo as pointed by Brooks Moses

[bkramer]

ELF/Relocations.cpp

444

SmallSet of pointers doesn't have a stable iteration order. Does that matter here?

[MaskRay]

ELF/Relocations.cpp

444

The iteration order doesn't matter here. These SharedSymbol's are already in the symbol table. This function returns a set of pointers that will be changed to Defined's.

[ruiu]

Could you add a test?

ELF/Relocations.cpp

443–444

I don't think you need this comment.

[MaskRay]

Add test

[MaskRay]

Remove a comment as suggested by ruiu@

[MaskRay]

Add test

[MaskRay]

CHECK -> CHECK-NEXT

[ruiu]

LGTM

Thanks!

[MaskRay]

Commited but somehow removed the
Differential revision: https://reviews.llvm.org/D47975
tag

ELF/Relocations.cpp

 #include "OutputSections.h"
 #include "SymbolTable.h"
 #include "Symbols.h"
 #include "SyntheticSections.h"
 #include "Target.h"
 #include "Thunks.h"
 #include "lld/Common/Memory.h"
 #include "lld/Common/Strings.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/Support/Endian.h"
 #include "llvm/Support/raw_ostream.h"
 #include <algorithm>
 
 using namespace llvm;
 using namespace llvm::ELF;
 using namespace llvm::object;
 using namespace llvm::support::endian;
if ((Phdr.p_type == ELF::PT_LOAD || Phdr.p_type == ELF::PT_GNU_RELRO) &&
       return true;
   return false;
 }
 
 // Returns symbols at the same offset as a given symbol, including SS itself.
 //
 // If two or more symbols are at the same offset, and at least one of
 // them are copied by a copy relocation, all of them need to be copied.
-// Otherwise, they would refer different places at runtime.
+// Otherwise, they would refer to different places at runtime.
 template <class ELFT>
-static std::vector<SharedSymbol *> getSymbolsAt(SharedSymbol &SS) {
+static SmallSet<SharedSymbol *, 4> getSymbolsAt(SharedSymbol &SS) {
   typedef typename ELFT::Sym Elf_Sym;
 
   SharedFile<ELFT> &File = SS.getFile<ELFT>();
 
-  std::vector<SharedSymbol *> Ret;
+  // Use a set to deduplicate symbols.
+  SmallSet<SharedSymbol *, 4> Ret;
   for (const Elf_Sym &S : File.getGlobalELFSyms()) {
     if (S.st_shndx == SHN_UNDEF || S.st_shndx == SHN_ABS ||
         S.st_value != SS.Value)
       continue;
     StringRef Name = check(S.getName(File.getStringTable()));
     Symbol *Sym = Symtab->find(Name);
     if (auto *Alias = dyn_cast_or_null<SharedSymbol>(Sym))
-      Ret.push_back(Alias);
+      Ret.insert(Alias);
   }
   return Ret;
 }
 
 // When a symbol is copy relocated or we create a canonical plt entry, it is
 // effectively a defined symbol. In the case of copy relocation the symbol is
 // in .bss and in the case of a canonical plt entry it is in .plt. This function
 // replaces the existing symbol with a Defined pointing to the appropriate

test/ELF/copy-rel-version.s

+// REQUIRES: x86
+// RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
+// RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %p/Inputs/copy-rel-version.s -o %t1.o
+// RUN: echo "v1 {}; v2 {};" > %t.ver
+// RUN: ld.lld %t1.o -shared -soname t1.so --version-script=%t.ver -o %t1.so
+// RUN: ld.lld %t.o %t1.so -o %t
+// RUN: llvm-readobj -t %t | FileCheck %s
+
+.global _start
+_start:
+  leaq foo, %rax
+
+// CHECK: Name: foo (
+// CHECK: Value:
+// CHECK: Size: 8