[LLD][COFF] Ignore DEBUG_S_XFGHASH_TYPE/VIRTUAL

[thieta]

Summary

These are new debug types that ships with the latest
Windows SDK and would warn and finally fail lld-link.

The symbols seems to be related to Microsoft's XFG
which is their version of CFG. We can't handle any of
this yet, so for now we can just ignore these types
so that lld doesn't fail with a new version of Windows
SDK.

Fixes: #56285

[mstorsjo]

I think this looks mostly good to me.

lld/COFF/PDB.cpp

816

You mentioned elsewhere that there doesn't seem to be any testcases for other similarly ignored ones - I think it's fair to add this without an exact testcase here too.

llvm/lib/DebugInfo/PDB/Native/FormatUtil.cpp

70

Is there any testcase for these ones, where the new types could easily be added?

[thieta]

llvm/lib/DebugInfo/PDB/Native/FormatUtil.cpp

70

I grepped for several of them and there doesn't seem to be any tests for this.

[rnk]

We spent some effort to make clang's assembly output somewhat readable. You should be able to adapt it to manually add one of these subsections and make a test case. In this example, see the subsection labels and .long directives:
https://gcc.godbolt.org/z/9o5Tra9W3

You can add subsections to an existing .debug$S section with these directives:

  .long 0xFF
  .long .Lmyend0-.Lmybeg0 # Subsection size
.Lmybeg0:
  .fill 16 # Some data, just to make it non-empty
.Lmyend0:
  .p2align 2

  .long 0x100
  .long .Lmyend1-.Lmybeg1 # Subsection size
.Lmybeg1:
  .fill 16 # Some data, just to make it non-empty
.Lmyend1:
  .p2align 2

[thieta]

Added tests and fixed pdb-unknown-subsection test

[thieta]

In D129378#3643749, @rnk wrote:

You can add subsections to an existing .debug$S section with these directives:

  .long 0xFF
  .long .Lmyend0-.Lmybeg0 # Subsection size
.Lmybeg0:
  .fill 16 # Some data, just to make it non-empty
.Lmyend0:
  .p2align 2

  .long 0x100
  .long .Lmyend1-.Lmybeg1 # Subsection size
.Lmybeg1:
  .fill 16 # Some data, just to make it non-empty
.Lmyend1:
  .p2align 2

Thanks - this put me on the right path and I was able to basically copy the pdb-unknown-subsection test and adapt it to check for the XFG subsections.

Have a look at this if you can please - I would like to include this in the 15.x release if possible.

[rnk]

lgtm, thanks again!

lld/COFF/PDB.cpp

for (const DebugSubsectionRecord &ss : subsections) {
     case DebugSubsectionKind::MergedAssemblyInput:
       // These appear to relate to .Net assembly info.
       break;
 
     case DebugSubsectionKind::CoffSymbolRVA:
       // Unclear what this is for.
       break;
 
+    case DebugSubsectionKind::XfgHashType:
+    case DebugSubsectionKind::XfgHashVirtual:
+      break;
+
     default:
       warn("ignoring unknown debug$S subsection kind 0x" +
            utohexstr(uint32_t(ss.kind())) + " in file " + toString(&file));
       break;
     }
   }
 }
 

lld/test/COFF/pdb-unknown-subsection.s

 # Check that unknown symbol subsections are ignored with a warning, and a PDB
 # is produced anyway.
 
 # REQUIRES: x86
 # RUN: llvm-mc -triple=i386-pc-win32 -filetype=obj -o %t.obj %s
 # RUN: lld-link -safeseh:no -subsystem:console -debug -nodefaultlib -entry:foo -out:%t.exe -pdb:%t.pdb %t.obj 2>&1 | FileCheck %s --check-prefix=WARNING
 # RUN: llvm-pdbutil dump -symbols %t.pdb | FileCheck %s
 
 # WARNING-NOT: ignoring unknown
-# WARNING: ignoring unknown debug$S subsection kind 0xFF
+# WARNING: ignoring unknown debug$S subsection kind 0x1FF
 # WARNING-NOT: ignoring unknown
 
 # CHECK:                           Symbols
 # CHECK:        4 | S_COMPILE3 [size = 52]
 # CHECK:            machine = intel x86-x64, Ver = clang version SENTINEL, language = c
 
 .text
 _foo:
.Ltmp7:
   .short  9000                    # Backend version
   .short  0
   .short  0
   .short  0
   .asciz  "clang version SENTINEL" # Null-terminated compiler version string
   .p2align  2
 .Ltmp8:
 .Ltmp6:
-  .long 0xFF # Unknown subsection kind
+  .long 0x1FF # Unknown subsection kind
   .long 4           # Subsection size
   .long  0
   .long 0x800000F1 # Unknown subsection kind
   .long 4           # Subsection size
   .long  0
   .long 0x800000F2 # Unknown subsection kind
   .long 4           # Subsection size
   .long  0
   .long 0x800000F3 # Unknown subsection kind
   .long 4           # Subsection size
   .long  0
   .long 0x800000F4 # Unknown subsection kind
   .long 4           # Subsection size
   .long  0

lld/test/COFF/pdb-xfg-section.s

+// Check that we ignore XFG subsections in CodeView information.
+// This test should be removed if we ever end up adding XFG support to lld.
+
+# REQUIRES: x86
+# RUN: llvm-mc -triple=i386-pc-win32 -filetype=obj -o %t.obj %s
+# RUN: lld-link -verbose -safeseh:no -subsystem:console -debug -nodefaultlib -entry:foo -out:%t.exe -pdb:%t.pdb %t.obj 2>&1 | FileCheck %s
+
+// XFGHashType should not generate a warning
+# CHECK-NOT: ignoring unknown debug$S subsection kind 0xFF
+// XFGHashVirtual should not generate a warning
+# CHECK-NOT: ignoring unknown debug$S subsection kind 0x100
+
+.text
+_foo:
+ret
+
+.global _foo
+
+.section .debug$S,"dr"
+  .p2align  2
+  .long 4                       # Debug section magic
+  .long 0xF1 # Symbol subsection
+  .long .Ltmp6-.Ltmp5           # Subsection size
+.Ltmp5:
+  .short  .Ltmp8-.Ltmp7           # Record length
+.Ltmp7:
+  .short  4412                    # Record kind: S_COMPILE3
+  .long 0                       # Flags and language
+  .short  208                     # CPUType
+  .short  9                       # Frontend version
+  .short  0
+  .short  0
+  .short  0
+  .short  9000                    # Backend version
+  .short  0
+  .short  0
+  .short  0
+  .asciz  "clang version SENTINEL" # Null-terminated compiler version string
+  .p2align  2
+.Ltmp8:
+.Ltmp6:
+  .long 0xFF # XFGHashType subsection
+  .long 4           # Subsection size
+  .long  0
+  .long 0x100 # XFGHashVirtual subsection
+  .long 4           # Subsection size
+  .long  0

llvm/include/llvm/DebugInfo/CodeView/CodeView.h

enum class DebugSubsectionKind : uint32_t {
 
   // These appear to relate to .Net assembly info.
   ILLines = 0xf9,
   FuncMDTokenMap = 0xfa,
   TypeMDTokenMap = 0xfb,
   MergedAssemblyInput = 0xfc,
 
   CoffSymbolRVA = 0xfd,
+
+  XfgHashType = 0xff,
+  XfgHashVirtual = 0x100,
 };
 
 /// Equivalent to CV_ptrtype_e.
 enum class PointerKind : uint8_t {
   Near16 = 0x00,                // 16 bit pointer
   Far16 = 0x01,                 // 16:16 far pointer
   Huge16 = 0x02,                // 16:16 huge pointer
   BasedOnSegment = 0x03,        // based on segment

llvm/lib/DebugInfo/PDB/Native/FormatUtil.cpp

switch (Kind) {
       RETURN_CASE(DebugSubsectionKind, CrossScopeImports, "xmi");
       RETURN_CASE(DebugSubsectionKind, CrossScopeExports, "xme");
       RETURN_CASE(DebugSubsectionKind, ILLines, "il lines");
       RETURN_CASE(DebugSubsectionKind, FuncMDTokenMap, "func md token map");
       RETURN_CASE(DebugSubsectionKind, TypeMDTokenMap, "type md token map");
       RETURN_CASE(DebugSubsectionKind, MergedAssemblyInput,
                   "merged assembly input");
       RETURN_CASE(DebugSubsectionKind, CoffSymbolRVA, "coff symbol rva");
+      RETURN_CASE(DebugSubsectionKind, XfgHashType, "xfg hash type");
+      RETURN_CASE(DebugSubsectionKind, XfgHashVirtual, "xfg hash virtual");
     }
   } else {
     switch (Kind) {
       RETURN_CASE(DebugSubsectionKind, None, "none");
       RETURN_CASE(DebugSubsectionKind, Symbols, "DEBUG_S_SYMBOLS");
       RETURN_CASE(DebugSubsectionKind, Lines, "DEBUG_S_LINES");
       RETURN_CASE(DebugSubsectionKind, StringTable, "DEBUG_S_STRINGTABLE");
       RETURN_CASE(DebugSubsectionKind, FileChecksums, "DEBUG_S_FILECHKSMS");
       RETURN_CASE(DebugSubsectionKind, FrameData, "DEBUG_S_FRAMEDATA");
       RETURN_CASE(DebugSubsectionKind, InlineeLines, "DEBUG_S_INLINEELINES");
       RETURN_CASE(DebugSubsectionKind, CrossScopeImports,
                   "DEBUG_S_CROSSSCOPEIMPORTS");
       RETURN_CASE(DebugSubsectionKind, CrossScopeExports,
                   "DEBUG_S_CROSSSCOPEEXPORTS");
       RETURN_CASE(DebugSubsectionKind, ILLines, "DEBUG_S_IL_LINES");
       RETURN_CASE(DebugSubsectionKind, FuncMDTokenMap,
                   "DEBUG_S_FUNC_MDTOKEN_MAP");
       RETURN_CASE(DebugSubsectionKind, TypeMDTokenMap,
                   "DEBUG_S_TYPE_MDTOKEN_MAP");
       RETURN_CASE(DebugSubsectionKind, MergedAssemblyInput,
                   "DEBUG_S_MERGED_ASSEMBLYINPUT");
       RETURN_CASE(DebugSubsectionKind, CoffSymbolRVA,
                   "DEBUG_S_COFF_SYMBOL_RVA");
+      RETURN_CASE(DebugSubsectionKind, XfgHashType,
+                  "DEBUG_S_XFGHASH_TYPE");
+      RETURN_CASE(DebugSubsectionKind, XfgHashVirtual,
+                  "DEBUG_S_XFGHASH_VIRTUAL");
+
     }
   }
   return formatUnknownEnum(Kind);
 }
 
 std::string llvm::pdb::formatSymbolKind(SymbolKind K) {
   switch (uint32_t(K)) {
 #define SYMBOL_RECORD(EnumName, value, name)                                   \