[WebAssembly] Handle weak undefined functions with a synthetic stub

This error case is described in Linking.md. The operand for call requires
generation of a synthetic stub.

Differential Revision: https://reviews.llvm.org/D44028

llvm-svn: 327151
diff --git a/lld/wasm/Driver.cpp b/lld/wasm/Driver.cpp
index c3f53e5..98d1f5d 100644
--- a/lld/wasm/Driver.cpp
+++ b/lld/wasm/Driver.cpp
@@ -215,6 +215,44 @@
   return Arg->getValue();
 }
 
+static const uint8_t UnreachableFn[] = {
+    0x03 /* ULEB length */, 0x00 /* ULEB num locals */,
+    0x00 /* opcode unreachable */, 0x0b /* opcode end */
+};
+
+// For weak undefined functions, there may be "call" instructions that reference
+// the symbol. In this case, we need to synthesise a dummy/stub function that
+// will abort at runtime, so that relocations can still provided an operand to
+// the call instruction that passes Wasm validation.
+static void handleWeakUndefines() {
+  for (Symbol *Sym : Symtab->getSymbols()) {
+    if (!Sym->isUndefined() || !Sym->isWeak())
+      continue;
+    auto *FuncSym = dyn_cast<FunctionSymbol>(Sym);
+    if (!FuncSym)
+      continue;
+
+    // It is possible for undefined functions not to have a signature (eg. if
+    // added via "--undefined"), but weak undefined ones do have a signature.
+    assert(FuncSym->getFunctionType());
+    const WasmSignature &Sig = *FuncSym->getFunctionType();
+
+    // Add a synthetic dummy for weak undefined functions.  These dummies will
+    // be GC'd if not used as the target of any "call" instructions.
+    StringRef StubName =
+        Saver.save("undefined function " + toString(*Sym, false));
+    SyntheticFunction *Func = make<SyntheticFunction>(Sig, StubName);
+    Func->setBody(UnreachableFn);
+    // Ensure it compares equal to the null pointer, and so that table relocs
+    // don't pull in the stub body (only call-operand relocs should do that).
+    Func->setTableIndex(0);
+    Symtab->SyntheticFunctions.emplace_back(Func);
+    // Hide our dummy to prevent export.
+    uint32_t Flags = WASM_SYMBOL_VISIBILITY_HIDDEN;
+    replaceSymbol<DefinedFunction>(Sym, Sym->getName(), Flags, nullptr, Func);
+  }
+}
+
 void LinkerDriver::link(ArrayRef<const char *> ArgsArr) {
   WasmOptTable Parser;
   opt::InputArgList Args = Parser.parse(ArgsArr.slice(1));
@@ -326,6 +364,10 @@
   for (InputFile *F : Files)
     Symtab->addFile(F);
 
+  // Add synthetic dummies for weak undefined functions.
+  if (!Config->Relocatable)
+    handleWeakUndefines();
+
   // Make sure we have resolved all symbols.
   if (!Config->Relocatable && !Config->AllowUndefined) {
     Symtab->reportRemainingUndefines();
diff --git a/lld/wasm/InputFiles.cpp b/lld/wasm/InputFiles.cpp
index 2c65c7b..12ee9ec 100644
--- a/lld/wasm/InputFiles.cpp
+++ b/lld/wasm/InputFiles.cpp
@@ -65,14 +65,8 @@
 uint32_t ObjFile::calcNewValue(const WasmRelocation &Reloc) const {
   switch (Reloc.Type) {
   case R_WEBASSEMBLY_TABLE_INDEX_I32:
-  case R_WEBASSEMBLY_TABLE_INDEX_SLEB: {
-    // The null case is possible, if you take the address of a weak function
-    // that's simply not supplied.
-    FunctionSymbol *Sym = getFunctionSymbol(Reloc.Index);
-    if (Sym->hasTableIndex())
-      return Sym->getTableIndex();
-    return 0;
-  }
+  case R_WEBASSEMBLY_TABLE_INDEX_SLEB:
+    return getFunctionSymbol(Reloc.Index)->getTableIndex();
   case R_WEBASSEMBLY_MEMORY_ADDR_SLEB:
   case R_WEBASSEMBLY_MEMORY_ADDR_I32:
   case R_WEBASSEMBLY_MEMORY_ADDR_LEB:
diff --git a/lld/wasm/MarkLive.cpp b/lld/wasm/MarkLive.cpp
index 22211c1..9b72697 100644
--- a/lld/wasm/MarkLive.cpp
+++ b/lld/wasm/MarkLive.cpp
@@ -73,8 +73,25 @@
     InputChunk *C = Q.pop_back_val();
 
     for (const WasmRelocation Reloc : C->getRelocations()) {
-      if (Reloc.Type != R_WEBASSEMBLY_TYPE_INDEX_LEB)
-        Enqueue(C->File->getSymbol(Reloc.Index));
+      if (Reloc.Type == R_WEBASSEMBLY_TYPE_INDEX_LEB)
+        continue;
+      Symbol *Sym = C->File->getSymbol(Reloc.Index);
+
+      // If the function has been assigned the special index zero in the table,
+      // the relocation doesn't pull in the function body, since the function
+      // won't actually go in the table (the runtime will trap attempts to call
+      // that index, since we don't use it).  A function with a table index of
+      // zero is only reachable via "call", not via "call_indirect".  The stub
+      // functions used for weak-undefined symbols have this behaviour (compare
+      // equal to null pointer, only reachable via direct call).
+      if (Reloc.Type == R_WEBASSEMBLY_TABLE_INDEX_SLEB ||
+          Reloc.Type == R_WEBASSEMBLY_TABLE_INDEX_I32) {
+        FunctionSymbol *FuncSym = cast<FunctionSymbol>(Sym);
+        if (FuncSym->hasTableIndex() && FuncSym->getTableIndex() == 0)
+          continue;
+      }
+
+      Enqueue(Sym);
     }
   }
 
diff --git a/lld/wasm/Symbols.cpp b/lld/wasm/Symbols.cpp
index 742c887..b1bf9b3 100644
--- a/lld/wasm/Symbols.cpp
+++ b/lld/wasm/Symbols.cpp
@@ -180,10 +180,10 @@
 
 void LazySymbol::fetch() { cast<ArchiveFile>(File)->addMember(&ArchiveSymbol); }
 
-std::string lld::toString(const wasm::Symbol &Sym) {
+std::string lld::toString(const wasm::Symbol &Sym, bool QuoteDemangled) {
   if (Config->Demangle)
     if (Optional<std::string> S = demangleItanium(Sym.getName()))
-      return "`" + *S + "'";
+      return QuoteDemangled ? ("`" + *S + "'") : *S;
   return Sym.getName();
 }
 
diff --git a/lld/wasm/Symbols.h b/lld/wasm/Symbols.h
index 76a7417..92acabb 100644
--- a/lld/wasm/Symbols.h
+++ b/lld/wasm/Symbols.h
@@ -312,7 +312,7 @@
 } // namespace wasm
 
 // Returns a symbol name for an error message.
-std::string toString(const wasm::Symbol &Sym);
+std::string toString(const wasm::Symbol &Sym, bool QuoteDemangled = true);
 std::string toString(wasm::Symbol::Kind Kind);
 std::string toString(WasmSymbolType Type);