COFF: Correctly handle relocations against early discarded sections.

[pcc]

Summary

Don't crash if we encounter a reference to an early discarded section
(such as .drectve). Instead, handle them the same way as sections
discarded by comdat merging, i.e. either print an error message or
(for debug sections) silently ignore the relocation.

[ruiu]

LGTM

lld/COFF/Chunks.cpp

271 ↗

(On Diff #123540)

If you move this piece of code immediately after auto *Sym = dyn_cast_or_null.., can't you simplify code a bit? You would have to write two fatal function calls with the same message, but I don't think that's not that bad.

lld/trunk/COFF/Chunks.cpp

for (const coff_relocation &Rel : Relocs) {
     if (Rel.VirtualAddress >= InputSize)
       fatal("relocation points beyond the end of its parent section");
 
     uint8_t *Off = Buf + OutputSectionOff + Rel.VirtualAddress;
 
     // Get the output section of the symbol for this relocation.  The output
     // section is needed to compute SECREL and SECTION relocations used in debug
     // info.
-    Defined *Sym = cast<Defined>(File->getSymbol(Rel.SymbolTableIndex));
+    auto *Sym =
+        dyn_cast_or_null<Defined>(File->getSymbol(Rel.SymbolTableIndex));
+    if (!Sym) {
+      if (isCodeView() || isDWARF())
+        continue;
+      // Symbols in early discarded sections are represented using null pointers,
+      // so we need to retrieve the name from the object file.
+      COFFSymbolRef Sym =
+          check(File->getCOFFObj()->getSymbol(Rel.SymbolTableIndex));
+      StringRef Name;
+      File->getCOFFObj()->getSymbolName(Sym, Name);
+      fatal("relocation against symbol in discarded section: " + Name);
+    }
     Chunk *C = Sym->getChunk();
     OutputSection *OS = C ? C->getOutputSection() : nullptr;
 
     // Only absolute and __ImageBase symbols lack an output section. For any
     // other symbol, this indicates that the chunk was discarded.  Normally
     // relocations against discarded sections are an error.  However, debug info
     // sections are not GC roots and can end up with these kinds of relocations.
     // Skip these relocations.

 // Collect all locations that contain absolute addresses, which need to be
 // fixed by the loader if load-time relocation is needed.
 // Only called when base relocation is enabled.
 void SectionChunk::getBaserels(std::vector<Baserel> *Res) {
   for (const coff_relocation &Rel : Relocs) {
     uint8_t Ty = getBaserelType(Rel);
     if (Ty == IMAGE_REL_BASED_ABSOLUTE)
       continue;
-    if (isa<DefinedAbsolute>(File->getSymbol(Rel.SymbolTableIndex)))
+    Symbol *Target = File->getSymbol(Rel.SymbolTableIndex);
+    if (!Target || isa<DefinedAbsolute>(Target))
       continue;
     Res->emplace_back(RVA + Rel.VirtualAddress, Ty);
   }
 }
 
 bool SectionChunk::hasData() const {
   return !(Header->Characteristics & IMAGE_SCN_CNT_UNINITIALIZED_DATA);
 }

lld/trunk/COFF/MarkLive.cpp

while (!Worklist.empty()) {
     // diagnosed during relocation processing.
     if (SC->isDiscarded())
       continue;
 
     assert(SC->isLive() && "We mark as live when pushing onto the worklist!");
 
     // Mark all symbols listed in the relocation table for this section.
     for (Symbol *B : SC->symbols())
+      if (B)
         AddSym(B);
 
     // Mark associative sections if any.
     for (SectionChunk *C : SC->children())
       Enqueue(C);
   }
 }
 
 }
 }

lld/trunk/test/COFF/reloc-discarded-early.s

+# RUN: llvm-mc -triple=x86_64-windows-msvc -filetype=obj -o %t.obj %s
+# RUN: lld-link -entry:__ImageBase -subsystem:console -debug %t.obj
+
+.section .debug_info,"dr"
+.quad .Ldrectve
+
+.section .drectve
+.Ldrectve:

lld/trunk/test/COFF/reloc-discarded-early2.s

+# RUN: llvm-mc -triple=x86_64-windows-msvc -filetype=obj -o %t.obj %s
+# RUN: not lld-link -entry:__ImageBase -subsystem:console %t.obj 2>&1 | FileCheck %s
+
+.text
+# CHECK: error: relocation against symbol in discarded section: .drectve
+.quad .Ldrectve
+
+.section .drectve
+.Ldrectve: