ELF: Use index 0 for unversioned undefined symbols (#168189)

The GNU documentation is ambiguous about the version index for
unversioned undefined symbols. The current specification at
https://sourceware.org/gnu-gabi/program-loading-and-dynamic-linking.txt
defines VER_NDX_LOCAL (0) as "The symbol is private, and is not
available outside this object."

However, this naming is misleading for undefined symbols. As suggested
in
discussions, VER_NDX_LOCAL should conceptually be VER_NDX_NONE and apply
to unversioned undefined symbols as well.

GNU ld has used index 0 for unversioned undefined symbols both before
version 2.35 (see https://sourceware.org/PR26002) and in the upcoming
2.46 release (see https://sourceware.org/PR33577). This change aligns
with GNU ld's behavior by switching from index 1 to index 0.

While here, add a test to dso-undef-extract-lazy.s that undefined
symbols of index 0 in DSO are treated as unversioned symbols.

6 files changed, 12 insertions, 8 deletions

diff --git a/lld/ELF/InputFiles.cpp b/lld/ELF/InputFiles.cpp
index a5921feb1829..240a6d0cd4b2 100644
--- a/lld/ELF/InputFiles.cpp
+++ b/lld/ELF/InputFiles.cpp
@@ -1676,8 +1676,9 @@ template <class ELFT> void SharedFile::parse() {
 
     const uint16_t ver = versyms[i], idx = ver & ~VERSYM_HIDDEN;
     if (sym.isUndefined()) {
-      // For unversioned undefined symbols, VER_NDX_GLOBAL makes more sense but
-      // as of binutils 2.34, GNU ld produces VER_NDX_LOCAL.
+      // Index 0 (VER_NDX_LOCAL) is used for unversioned undefined symbols.
+      // GNU ld versions between 2.35 and 2.45 also generate VER_NDX_GLOBAL
+      // for this case (https://sourceware.org/PR33577).
       if (ver != VER_NDX_LOCAL && ver != VER_NDX_GLOBAL) {
         if (idx >= verneeds.size()) {
           ErrAlways(ctx) << "corrupt input file: version need index " << idx
diff --git a/lld/ELF/Symbols.h b/lld/ELF/Symbols.h
index c117e3b716c1..a7d61f48ed3d 100644
--- a/lld/ELF/Symbols.h
+++ b/lld/ELF/Symbols.h
@@ -313,6 +313,8 @@ public:
   // represents the Verdef index within the input DSO, which will be converted
   // to a Verneed index in the output. Otherwise, this represents the Verdef
   // index (VER_NDX_LOCAL, VER_NDX_GLOBAL, or a named version).
+  // VER_NDX_LOCAL indicates a defined symbol that has been localized by a
+  // version script's local: directive or --exclude-libs.
   uint16_t versionId;
   LLVM_PREFERRED_TYPE(bool)
   uint8_t versionScriptAssigned : 1;
diff --git a/lld/ELF/SyntheticSections.cpp b/lld/ELF/SyntheticSections.cpp
index 19b08152ae08..ea9b87952cd8 100644
--- a/lld/ELF/SyntheticSections.cpp
+++ b/lld/ELF/SyntheticSections.cpp
@@ -3784,9 +3784,10 @@ void VersionTableSection::writeTo(uint8_t *buf) {
   buf += 2;
   for (const SymbolTableEntry &s : getPartition(ctx).dynSymTab->getSymbols()) {
     // For an unextracted lazy symbol (undefined weak), it must have been
-    // converted to Undefined and have VER_NDX_GLOBAL version here.
+    // converted to Undefined.
     assert(!s.sym->isLazy());
-    write16(ctx, buf, s.sym->versionId);
+    // Undefined symbols should use index 0 when unversioned.
+    write16(ctx, buf, s.sym->isUndefined() ? 0 : s.sym->versionId);
     buf += 2;
   }
 }
diff --git a/lld/test/ELF/linkerscript/version-script.s b/lld/test/ELF/linkerscript/version-script.s
index 52382eeb1245..6b97fede00c3 100644
--- a/lld/test/ELF/linkerscript/version-script.s
+++ b/lld/test/ELF/linkerscript/version-script.s
@@ -17,7 +17,7 @@
 # CHECK-NEXT:     Name:
 # CHECK-NEXT:   }
 # CHECK-NEXT:   Symbol {
-# CHECK-NEXT:     Version: 1
+# CHECK-NEXT:     Version: 0
 # CHECK-NEXT:     Name: und
 # CHECK-NEXT:   }
 # CHECK-NEXT:   Symbol {
diff --git a/lld/test/ELF/version-script-extern-undefined.s b/lld/test/ELF/version-script-extern-undefined.s
index 38114229e0ce..010b4d5d6b63 100644
--- a/lld/test/ELF/version-script-extern-undefined.s
+++ b/lld/test/ELF/version-script-extern-undefined.s
@@ -11,7 +11,7 @@
 # CHECK-NEXT:     Name:
 # CHECK-NEXT:   }
 # CHECK-NEXT:   Symbol {
-# CHECK-NEXT:     Version: 1
+# CHECK-NEXT:     Version: 0
 # CHECK-NEXT:     Name: _Z3abbi
 # CHECK-NEXT:   }
 # CHECK-NEXT: ]
diff --git a/llvm/include/llvm/BinaryFormat/ELF.h b/llvm/include/llvm/BinaryFormat/ELF.h
index 39e9611c7190..bfcf5dab4772 100644
--- a/llvm/include/llvm/BinaryFormat/ELF.h
+++ b/llvm/include/llvm/BinaryFormat/ELF.h
@@ -1710,8 +1710,8 @@ enum { VER_FLG_BASE = 0x1, VER_FLG_WEAK = 0x2, VER_FLG_INFO = 0x4 };
 
 // Special constants for the version table. (SHT_GNU_versym/.gnu.version)
 enum {
-  VER_NDX_LOCAL = 0,       // Unversioned local symbol
-  VER_NDX_GLOBAL = 1,      // Unversioned global symbol
+  VER_NDX_LOCAL = 0,       // Unversioned undefined or localized defined symbol
+  VER_NDX_GLOBAL = 1,      // Unversioned non-local defined symbol
   VERSYM_VERSION = 0x7fff, // Version Index mask
   VERSYM_HIDDEN = 0x8000   // Hidden bit (non-default version)
 };