[ELF] Support --{,no-}allow-shlib-undefined

Summary:
In ld.bfd/gold, --no-allow-shlib-undefined is the default when linking
an executable. This patch implements a check to error on undefined
symbols in a shared object, if all of its DT_NEEDED entries are seen.

Our approach resembles the one used in gold, achieves a good balance to
be useful but not too smart (ld.bfd traces all DSOs and emulates the
behavior of a dynamic linker to catch more cases).

The error is issued based on the symbol table, different from undefined
reference errors issued for relocations. It is most effective when there
are DSOs that were not linked with -z defs (e.g. when static sanitizers
runtime is used).

gold has a comment that some system libraries on GNU/Linux may have
spurious undefined references and thus system libraries should be
excluded (https://sourceware.org/bugzilla/show_bug.cgi?id=6811). The
story may have changed now but we make --allow-shlib-undefined the
default for now. Its interaction with -shared can be discussed in the
future.

Reviewers: ruiu, grimar, pcc, espindola

Reviewed By: ruiu

Subscribers: joerg, emaste, arichardson, llvm-commits

Differential Revision: https://reviews.llvm.org/D57385

llvm-svn: 352826

lld/ELF/Config.h

@@ -119,6 +119,7 @@ struct Configuration {
                   uint64_t>
       CallGraphProfile;
   bool AllowMultipleDefinition;
+  bool AllowShlibUndefined;
   bool AndroidPackDynRelocs;
   bool ARMHasBlx = false;
   bool ARMHasMovtMovw = false;

lld/ELF/Driver.cpp

@@ -756,6 +756,8 @@ void LinkerDriver::readConfigs(opt::InputArgList &Args) {
       Args.hasFlag(OPT_allow_multiple_definition,
                    OPT_no_allow_multiple_definition, false) ||
       hasZOption(Args, "muldefs");
+  Config->AllowShlibUndefined = Args.hasFlag(
+      OPT_allow_shlib_undefined, OPT_no_allow_shlib_undefined, true);
   Config->AuxiliaryList = args::getStrings(Args, OPT_auxiliary);
   Config->Bsymbolic = Args.hasArg(OPT_Bsymbolic);
   Config->BsymbolicFunctions = Args.hasArg(OPT_Bsymbolic_functions);

lld/ELF/InputFiles.cpp

@@ -862,7 +862,7 @@ SharedFile<ELFT>::SharedFile(MemoryBufferRef M, StringRef DefaultSoName)
 
 // Partially parse the shared object file so that we can call
 // getSoName on this object.
-template <class ELFT> void SharedFile<ELFT>::parseSoName() {
+template <class ELFT> void SharedFile<ELFT>::parseDynamic() {
   const Elf_Shdr *DynamicSec = nullptr;
   const ELFFile<ELFT> Obj = this->getObj();
   ArrayRef<Elf_Shdr> Sections = CHECK(Obj.sections(), this);
@@ -899,12 +899,16 @@ template <class ELFT> void SharedFile<ELFT>::parseSoName() {
   ArrayRef<Elf_Dyn> Arr =
       CHECK(Obj.template getSectionContentsAsArray<Elf_Dyn>(DynamicSec), this);
   for (const Elf_Dyn &Dyn : Arr) {
-    if (Dyn.d_tag == DT_SONAME) {
+    if (Dyn.d_tag == DT_NEEDED) {
+      uint64_t Val = Dyn.getVal();
+      if (Val >= this->StringTable.size())
+        fatal(toString(this) + ": invalid DT_NEEDED entry");
+      DtNeeded.push_back(this->StringTable.data() + Val);
+    } else if (Dyn.d_tag == DT_SONAME) {
       uint64_t Val = Dyn.getVal();
       if (Val >= this->StringTable.size())
         fatal(toString(this) + ": invalid DT_SONAME entry");
       SoName = this->StringTable.data() + Val;
-      return;
     }
   }
 }
@@ -972,7 +976,7 @@ uint32_t SharedFile<ELFT>::getAlignment(ArrayRef<Elf_Shdr> Sections,
   return (Ret > UINT32_MAX) ? 0 : Ret;
 }
 
-// Fully parse the shared object file. This must be called after parseSoName().
+// Fully parse the shared object file. This must be called after parseDynamic().
 //
 // This function parses symbol versions. If a DSO has version information,
 // the file has a ".gnu.version_d" section which contains symbol version

lld/ELF/InputFiles.h

@@ -332,6 +332,7 @@ template <class ELFT> class SharedFile : public ELFFileBase<ELFT> {
 
 public:
   std::vector<const Elf_Verdef *> Verdefs;
+  std::vector<StringRef> DtNeeded;
   std::string SoName;
 
   static bool classof(const InputFile *F) {
@@ -340,7 +341,7 @@ template <class ELFT> class SharedFile : public ELFFileBase<ELFT> {
 
   SharedFile(MemoryBufferRef M, StringRef DefaultSoName);
 
-  void parseSoName();
+  void parseDynamic();
   void parseRest();
   uint32_t getAlignment(ArrayRef<Elf_Shdr> Sections, const Elf_Sym &Sym);
   std::vector<const Elf_Verdef *> parseVerdefs();
@@ -358,6 +359,9 @@ template <class ELFT> class SharedFile : public ELFFileBase<ELFT> {
   // data structures in the output file.
   std::map<const Elf_Verdef *, NeededVer> VerdefMap;
 
+  // Used for --no-allow-shlib-undefined.
+  bool AllNeededIsKnown;
+
   // Used for --as-needed
   bool IsNeeded;
 };

lld/ELF/Options.td

@@ -63,6 +63,10 @@ defm allow_multiple_definition: B<"allow-multiple-definition",
     "Allow multiple definitions",
     "Do not allow multiple definitions (default)">;
 
+defm allow_shlib_undefined: B<"allow-shlib-undefined",
+    "Allow unresolved references in shared libraries (default)",
+    "Do not allow unresolved references in shared libraries">;
+
 defm apply_dynamic_relocs: B<"apply-dynamic-relocs",
     "Apply link-time values for dynamic relocations",
     "Do not apply link-time values for dynamic relocations (default)">;
@@ -492,12 +496,10 @@ def plugin_opt_thinlto: J<"plugin-opt=thinlto">;
 def plugin_opt_slash: J<"plugin-opt=/">;
 
 // Options listed below are silently ignored for now for compatibility.
-def: F<"allow-shlib-undefined">;
 def: F<"detect-odr-violations">;
 def: Flag<["-"], "g">;
 def: F<"long-plt">;
 def: F<"no-add-needed">;
-def: F<"no-allow-shlib-undefined">;
 def: F<"no-copy-dt-needed-entries">;
 def: F<"no-ctors-in-init-array">;
 def: F<"no-keep-memory">;

lld/ELF/SymbolTable.cpp

@@ -92,7 +92,7 @@ template <class ELFT> void SymbolTable::addFile(InputFile *File) {
   // .so file
   if (auto *F = dyn_cast<SharedFile<ELFT>>(File)) {
     // DSOs are uniquified not by filename but by soname.
-    F->parseSoName();
+    F->parseDynamic();
     if (errorCount())
       return;
 

lld/ELF/SymbolTable.h

@@ -79,6 +79,9 @@ class SymbolTable {
 
   void handleDynamicList();
 
+  // Set of .so files to not link the same shared object file more than once.
+  llvm::DenseMap<StringRef, InputFile *> SoNames;
+
 private:
   std::pair<Symbol *, bool> insertName(StringRef Name);
 
@@ -106,9 +109,6 @@ class SymbolTable {
   // is used to uniquify them.
   llvm::DenseSet<llvm::CachedHashStringRef> ComdatGroups;
 
-  // Set of .so files to not link the same shared object file more than once.
-  llvm::DenseMap<StringRef, InputFile *> SoNames;
-
   // A map from demangled symbol names to their symbol objects.
   // This mapping is 1:N because two symbols with different versions
   // can have the same name. We use this map to handle "extern C++ {}"

lld/ELF/Writer.cpp

@@ -1682,6 +1682,27 @@ template <class ELFT> void Writer<ELFT>::finalizeSections() {
   if (In.Iplt && !In.Iplt->empty())
     In.Iplt->addSymbols();
 
+  if (!Config->AllowShlibUndefined) {
+    // Error on undefined symbols in a shared object, if all of its DT_NEEDED
+    // entires are seen. These cases would otherwise lead to runtime errors
+    // reported by the dynamic linker.
+    //
+    // ld.bfd traces all DT_NEEDED to emulate the logic of the dynamic linker to
+    // catch more cases. That is too much for us. Our approach resembles the one
+    // used in ld.gold, achieves a good balance to be useful but not too smart.
+    for (InputFile *File : SharedFiles) {
+      SharedFile<ELFT> *F = cast<SharedFile<ELFT>>(File);
+      F->AllNeededIsKnown = llvm::all_of(F->DtNeeded, [&](StringRef Needed) {
+        return Symtab->SoNames.count(Needed);
+      });
+    }
+    for (Symbol *Sym : Symtab->getSymbols())
+      if (Sym->isUndefined() && !Sym->isWeak())
+        if (auto *F = dyn_cast_or_null<SharedFile<ELFT>>(Sym->File))
+          if (F->AllNeededIsKnown)
+            error(toString(F) + ": undefined reference to " + toString(*Sym));
+  }
+
   // Now that we have defined all possible global symbols including linker-
   // synthesized ones. Visit all symbols to give the finishing touches.
   for (Symbol *Sym : Symtab->getSymbols()) {

lld/test/ELF/allow-shlib-undefined.s

@@ -1,26 +1,28 @@
 # REQUIRES: x86
-# --allow-shlib-undefined and --no-allow-shlib-undefined are fully
-# ignored in linker implementation.
 # --allow-shlib-undefined is set by default
+
+# RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t.o
 # RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux \
-# RUN: %p/Inputs/allow-shlib-undefined.s -o %t
-# RUN: ld.lld -shared %t -o %t.so
-# RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t1
+# RUN:   %p/Inputs/allow-shlib-undefined.s -o %t1.o
+# RUN: ld.lld -shared %t1.o -o %t.so
+
+# RUN: ld.lld %t.o %t.so -o /dev/null
+# RUN: ld.lld --allow-shlib-undefined %t.o %t.so -o /dev/null
+# RUN: not ld.lld --no-allow-shlib-undefined %t.o %t.so -o /dev/null 2>&1 | FileCheck %s
 
-# Executable: should link with DSO containing undefined symbols in any case.
-# RUN: ld.lld %t1 %t.so -o %t2
-# RUN: ld.lld --no-allow-shlib-undefined %t1 %t.so -o %t2
-# RUN: ld.lld --allow-shlib-undefined %t1 %t.so -o %t2
+# RUN: echo | llvm-mc -filetype=obj -triple=x86_64-unknown-linux -o %tempty.o
+# RUN: ld.lld -shared %tempty.o -o %tempty.so
+# RUN: ld.lld -shared %t1.o %tempty.so -o %t2.so
+# RUN: ld.lld --no-allow-shlib-undefined %t.o %t2.so -o /dev/null
 
 # DSO with undefines:
 # should link with or without any of these options.
-# RUN: ld.lld -shared %t -o %t.so
-# RUN: ld.lld -shared --allow-shlib-undefined %t -o %t.so
-# RUN: ld.lld -shared --no-allow-shlib-undefined %t -o %t.so
-
-# Executable still should not link when have undefines inside.
-# RUN: not ld.lld %t -o %t.so
+# RUN: ld.lld -shared %t1.o -o /dev/null
+# RUN: ld.lld -shared --allow-shlib-undefined %t1.o -o /dev/null
+# RUN: ld.lld -shared --no-allow-shlib-undefined %t1.o -o /dev/null
 
 .globl _start
 _start:
   callq _shared@PLT
+
+# CHECK: undefined reference to _unresolved