[DwarfDebug] Refuse to emit DW_OP_LLVM_arg values wider than 64 bits

[rickyz]

Summary

DwarfExpression::addUnsignedConstant(const APInt &Value) only supports
wider-than-64-bit values when it is used to emit a top-level DWARF
expression representing the location of a variable. Before this change,
it was possible to call addUnsignedConstant on >64 bit values within a
subexpression when substituting DW_OP_LLVM_arg values.

This can trigger an assertion failure (e.g. PR52584, PR52333) when it
happens in a fragment (DW_OP_LLVM_fragment) expression, as
addUnsignedConstant on >64 bit values splits the constant into separate
DW_OP_pieces, which modifies DwarfExpression::OffsetInBits.

This change papers over the assertion errors by bailing on overly wide
DW_OP_LLVM_arg values. A more comprehensive fix might be to be to split
wide values into pointer-sized fragments.

[0] https://github.com/llvm/llvm-project/blob/e71fa03/llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp#L799-L805

[rickyz]

Minor rewording/test cleanup.

[rickyz]

Update commit message.

[aprantl]

llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp

802

Can you add a comment explaining what the more principled fix would be to the code you added?

llvm/test/DebugInfo/X86/pr52584.ll

2

Can you FileCheck for something (ideally the missing DW_AT_location) here? Otherwise this test even succeeds if you symlink llc to /bin/true.

[Mordante]

Thanks for working on this!

I just tested this patch with D70631, the review where I discovered the issue.
After I applied your patch the build issue I had has been resolved.

[rickyz]

Respond to comments, avoid emitting a partially-filled DW_AT_location on encountering an overly wide argument.

[rickyz]

Minor comment wording change.

[aprantl]

Thanks!

[rickyz]

Thanks for the review! I don't have commit access, so can you please commit this on my behalf? Thanks!

llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp

if (auto *DVal = DV.getValueLoc()) {
     assert(Expr && "Variadic Debug Value must have an Expression.");
     DIELoc *Loc = new (DIEValueAllocator) DIELoc;
     DIEDwarfExpression DwarfExpr(*Asm, *this, *Loc);
     DwarfExpr.addFragmentOffset(Expr);
     DIExpressionCursor Cursor(Expr);
     const TargetRegisterInfo &TRI = *Asm->MF->getSubtarget().getRegisterInfo();
 
     auto AddEntry = [&](const DbgValueLocEntry &Entry,
-                                            DIExpressionCursor &Cursor) {
       if (Entry.isLocation()) {
         if (!DwarfExpr.addMachineRegExpression(TRI, Cursor,
                                                Entry.getLoc().getReg()))
           return false;
       } else if (Entry.isInt()) {
         // If there is an expression, emit raw unsigned bytes.
         DwarfExpr.addUnsignedConstant(Entry.getInt());
       } else if (Entry.isConstantFP()) {
         APInt RawBytes = Entry.getConstantFP()->getValueAPF().bitcastToAPInt();
+        if (RawBytes.getBitWidth() > 64)
+          return false;
         DwarfExpr.addUnsignedConstant(RawBytes);
       } else if (Entry.isConstantInt()) {
         APInt RawBytes = Entry.getConstantInt()->getValue();
+        if (RawBytes.getBitWidth() > 64)
+          return false;
         DwarfExpr.addUnsignedConstant(RawBytes);
       } else if (Entry.isTargetIndexLocation()) {
         TargetIndexLocation Loc = Entry.getTargetIndexLocation();
         // TODO TargetIndexLocation is a target-independent. Currently only the
         // WebAssembly-specific encoding is supported.
         assert(Asm->TM.getTargetTriple().isWasm());
         DwarfExpr.addWasmLocation(Loc.Index, static_cast<uint64_t>(Loc.Offset));
       } else {

llvm/test/DebugInfo/X86/pr52584.ll

+; RUN: llc %s -mtriple=x86_64-unknown-linux-gnu -o /dev/null
+
+define dso_local void @test() !dbg !4 {
+entry:
+  call void @llvm.dbg.value(metadata !DIArgList(i128 0), metadata !7, metadata !DIExpression(DW_OP_LLVM_arg, 0, DW_OP_stack_value, DW_OP_LLVM_fragment, 0, 127)), !dbg !12
+  ret void, !dbg !12
+}
+
+declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!2, !3}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, runtimeVersion: 0, emissionKind: FullDebug, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "test.c", directory: "")
+!2 = !{i32 7, !"Dwarf Version", i32 4}
+!3 = !{i32 2, !"Debug Info Version", i32 3}
+!4 = distinct !DISubprogram(name: "test", scope: !1, file: !1, line: 1, type: !5, unit: !0)
+!5 = !DISubroutineType(types: !6)
+!6 = !{null}
+!7 = !DILocalVariable(name: "arr", scope: !4, file: !1, line: 1, type: !8)
+!8 = !DICompositeType(tag: DW_TAG_array_type, baseType: !9, size: 128, elements: !10)
+!9 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_signed_char)
+!10 = !{!11}
+!11 = !DISubrange(count: 16)
+!12 = !DILocation(line: 1, column: 1, scope: !4)