[sanitizer] Add plumbing for -fsanitize-annotate-debug-info and partly replace '-mllvm -array-bounds-pseudofn' (#138577)

thurstond · web-flow · commit 6a28d8c24a07 · 2025-05-07T15:57:01.000-07:00
@fmayer introduced '-mllvm -array-bounds-pseudofn' (#128977) to make it easier to see why crashes occurred, and to estimate with a profiler the cycles spent on these array-bounds checks. This functionality could be usefully generalized to other checks in future work. This patch adds the plumbing for -fsanitize-annotate-debug-info, and connects it to the existing array-bounds-pseudo-fn functionality i.e., -fsanitize-annotate-debug-info=array-bounds can be used as a replacement for '-mllvm -array-bounds-pseudofn', though we do not yet delete the latter. Note: we replaced '-mllvm -array-bounds-pseudofn' in clang/test/CodeGen/bounds-checking-debuginfo.c, because adding test cases would modify the line numbers in the test assertions, and therefore obscure that the test output is the same between '-mllvm -array-bounds-pseudofn' and -fsanitize-annotate-debug-info=array-bounds.
diff --git a/clang/include/clang/Basic/CodeGenOptions.h b/clang/include/clang/Basic/CodeGenOptions.h
@@ -399,6 +399,10 @@ class CodeGenOptions : public CodeGenOptionsBase {
   /// (0.0 [default] to skip none, 1.0 to skip all).
   SanitizerMaskCutoffs SanitizeSkipHotCutoffs;
 
+  /// Set of sanitizer checks, for which the instrumentation will be annotated
+  /// with extra debug info.
+  SanitizerSet SanitizeAnnotateDebugInfo;
+
   /// List of backend command-line options for -fembed-bitcode.
   std::vector<uint8_t> CmdArgs;
 
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
@@ -2540,6 +2540,31 @@ def fno_sanitize_merge_handlers : Flag<["-"], "fno-sanitize-merge">, Group<f_cla
                         Alias<fno_sanitize_merge_handlers_EQ>, AliasArgs<["all"]>,
                         Visibility<[ClangOption, CLOption]>,
                         HelpText<"Do not allow compiler to merge handlers for any sanitizers">;
+def fsanitize_annotate_debug_info_EQ
+    : CommaJoined<["-"], "fsanitize-annotate-debug-info=">,
+      Group<f_clang_Group>,
+      HelpText<"Annotate sanitizer instrumentation with extra debug info for "
+               "the specified sanitizers, if supported">;
+def fno_sanitize_annotate_debug_info_EQ
+    : CommaJoined<["-"], "fno-sanitize-annotate-debug-info=">,
+      Group<f_clang_Group>,
+      HelpText<"Do not allow compiler to annotate sanitizer instrumentation "
+               "with extra debug info for the specified sanitizers">;
+def fsanitize_annotate_debug_info
+    : Flag<["-"], "fsanitize-annotate-debug-info">,
+      Group<f_clang_Group>,
+      Alias<fsanitize_annotate_debug_info_EQ>,
+      AliasArgs<["all"]>,
+      HelpText<"Allow compiler to annotate sanitizer instrumentation with "
+               "extra debug info for all sanitizers, where supported">;
+def fno_sanitize_annotate_debug_info
+    : Flag<["-"], "fno-sanitize-annotate-debug-info">,
+      Group<f_clang_Group>,
+      Alias<fno_sanitize_annotate_debug_info_EQ>,
+      AliasArgs<["all"]>,
+      Visibility<[ClangOption, CLOption]>,
+      HelpText<"Do not allow compiler to annotate sanitizer instrumentation "
+               "with extra debug info for any sanitizers">;
 def fsanitize_undefined_trap_on_error
     : Flag<["-"], "fsanitize-undefined-trap-on-error">, Group<f_clang_Group>,
       Alias<fsanitize_trap_EQ>, AliasArgs<["undefined"]>;
diff --git a/clang/include/clang/Driver/SanitizerArgs.h b/clang/include/clang/Driver/SanitizerArgs.h
@@ -27,6 +27,7 @@ class SanitizerArgs {
   SanitizerSet TrapSanitizers;
   SanitizerSet MergeHandlers;
   SanitizerMaskCutoffs SkipHotCutoffs;
+  SanitizerSet AnnotateDebugInfo;
 
   std::vector<std::string> UserIgnorelistFiles;
   std::vector<std::string> SystemIgnorelistFiles;
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
@@ -1228,7 +1228,11 @@ void CodeGenFunction::EmitBoundsCheckImpl(const Expr *E, llvm::Value *Bound,
   SanitizerScope SanScope(this);
 
   llvm::DILocation *CheckDI = Builder.getCurrentDebugLocation();
-  if (ClArrayBoundsPseudoFn && CheckDI) {
+  auto CheckKind = SanitizerKind::SO_ArrayBounds;
+  // TODO: deprecate ClArrayBoundsPseudoFn
+  if ((ClArrayBoundsPseudoFn ||
+       CGM.getCodeGenOpts().SanitizeAnnotateDebugInfo.has(CheckKind)) &&
+      CheckDI) {
     CheckDI = getDebugInfo()->CreateSyntheticInlineAt(
         Builder.getCurrentDebugLocation(), "__ubsan_check_array_bounds");
   }
@@ -1245,8 +1249,8 @@ void CodeGenFunction::EmitBoundsCheckImpl(const Expr *E, llvm::Value *Bound,
   };
   llvm::Value *Check = Accessed ? Builder.CreateICmpULT(IndexVal, BoundVal)
                                 : Builder.CreateICmpULE(IndexVal, BoundVal);
-  EmitCheck(std::make_pair(Check, SanitizerKind::SO_ArrayBounds),
-            SanitizerHandler::OutOfBounds, StaticData, Index);
+  EmitCheck(std::make_pair(Check, CheckKind), SanitizerHandler::OutOfBounds,
+            StaticData, Index);
 }
 
 CodeGenFunction::ComplexPairTy CodeGenFunction::
diff --git a/clang/lib/Driver/SanitizerArgs.cpp b/clang/lib/Driver/SanitizerArgs.cpp
@@ -76,6 +76,7 @@ static const SanitizerMask MergeDefault =
     SanitizerKind::Undefined | SanitizerKind::Vptr;
 static const SanitizerMask TrappingDefault =
     SanitizerKind::CFI | SanitizerKind::LocalBounds;
+static const SanitizerMask AnnotateDebugInfoDefault;
 static const SanitizerMask CFIClasses =
     SanitizerKind::CFIVCall | SanitizerKind::CFINVCall |
     SanitizerKind::CFIMFCall | SanitizerKind::CFIDerivedCast |
@@ -738,6 +739,13 @@ SanitizerArgs::SanitizerArgs(const ToolChain &TC,
   // Parse -fno-sanitize-top-hot flags
   SkipHotCutoffs = parseSanitizeSkipHotCutoffArgs(D, Args, DiagnoseErrors);
 
+  // Parse -f(no-)?sanitize-annotate-debug-info flags
+  SanitizerMask AnnotateDebugInfoKinds =
+      parseSanitizeArgs(D, Args, DiagnoseErrors, AnnotateDebugInfoDefault, {},
+                        {}, options::OPT_fsanitize_annotate_debug_info_EQ,
+                        options::OPT_fno_sanitize_annotate_debug_info_EQ);
+  AnnotateDebugInfoKinds &= Kinds;
+
   // Setup ignorelist files.
   // Add default ignorelist from resource directory for activated sanitizers,
   // and validate special case lists format.
@@ -1168,6 +1176,8 @@ SanitizerArgs::SanitizerArgs(const ToolChain &TC,
 
   MergeHandlers.Mask |= MergeKinds;
 
+  AnnotateDebugInfo.Mask |= AnnotateDebugInfoKinds;
+
   // Zero out SkipHotCutoffs for unused sanitizers
   SkipHotCutoffs.clear(~Sanitizers.Mask);
 }
@@ -1351,6 +1361,10 @@ void SanitizerArgs::addArgs(const ToolChain &TC, const llvm::opt::ArgList &Args,
     CmdArgs.push_back(
         Args.MakeArgString("-fsanitize-skip-hot-cutoff=" + SkipHotCutoffsStr));
 
+  if (!AnnotateDebugInfo.empty())
+    CmdArgs.push_back(Args.MakeArgString("-fsanitize-annotate-debug-info=" +
+                                         toString(AnnotateDebugInfo)));
+
   addSpecialCaseListOpt(Args, CmdArgs,
                         "-fsanitize-ignorelist=", UserIgnorelistFiles);
   addSpecialCaseListOpt(Args, CmdArgs,
@@ -1534,7 +1548,10 @@ SanitizerMask parseArgValues(const Driver &D, const llvm::opt::Arg *A,
        A->getOption().matches(options::OPT_fsanitize_trap_EQ) ||
        A->getOption().matches(options::OPT_fno_sanitize_trap_EQ) ||
        A->getOption().matches(options::OPT_fsanitize_merge_handlers_EQ) ||
-       A->getOption().matches(options::OPT_fno_sanitize_merge_handlers_EQ)) &&
+       A->getOption().matches(options::OPT_fno_sanitize_merge_handlers_EQ) ||
+       A->getOption().matches(options::OPT_fsanitize_annotate_debug_info_EQ) ||
+       A->getOption().matches(
+           options::OPT_fno_sanitize_annotate_debug_info_EQ)) &&
       "Invalid argument in parseArgValues!");
   SanitizerMask Kinds;
   for (int i = 0, n = A->getNumValues(); i != n; ++i) {
diff --git a/clang/lib/Frontend/CompilerInvocation.cpp b/clang/lib/Frontend/CompilerInvocation.cpp
@@ -1838,6 +1838,10 @@ void CompilerInvocationBase::GenerateCodeGenArgs(const CodeGenOptions &Opts,
   for (std::string Sanitizer : Values)
     GenerateArg(Consumer, OPT_fsanitize_skip_hot_cutoff_EQ, Sanitizer);
 
+  for (StringRef Sanitizer :
+       serializeSanitizerKinds(Opts.SanitizeAnnotateDebugInfo))
+    GenerateArg(Consumer, OPT_fsanitize_annotate_debug_info_EQ, Sanitizer);
+
   if (!Opts.EmitVersionIdentMetadata)
     GenerateArg(Consumer, OPT_Qn);
 
@@ -2332,6 +2336,11 @@ bool CompilerInvocation::ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args,
       "-fsanitize-skip-hot-cutoff=",
       Args.getAllArgValues(OPT_fsanitize_skip_hot_cutoff_EQ), Diags);
 
+  parseSanitizerKinds(
+      "-fsanitize-annotate-debug-info=",
+      Args.getAllArgValues(OPT_fsanitize_annotate_debug_info_EQ), Diags,
+      Opts.SanitizeAnnotateDebugInfo);
+
   Opts.EmitVersionIdentMetadata = Args.hasFlag(OPT_Qy, OPT_Qn, true);
 
   if (!LangOpts->CUDAIsDevice)
diff --git a/clang/test/CodeGen/bounds-checking-debuginfo.c b/clang/test/CodeGen/bounds-checking-debuginfo.c
@@ -1,7 +1,7 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
-// RUN: %clang_cc1 -mllvm -array-bounds-pseudofn -emit-llvm -fdebug-prefix-map=%S/= -fno-ident -fdebug-compilation-dir=%S -fsanitize=array-bounds -fsanitize-trap=array-bounds -triple x86_64 -debug-info-kind=limited %s -o - | FileCheck --check-prefix=CHECK-TRAP %s
-// RUN: %clang_cc1 -mllvm -array-bounds-pseudofn -emit-llvm -fdebug-prefix-map=%S/= -fno-ident -fdebug-compilation-dir=%S -fsanitize=array-bounds                              -triple x86_64 -debug-info-kind=limited %s -o - | FileCheck --check-prefix=CHECK-NOTRAP %s
 
+// RUN: %clang_cc1 -emit-llvm -fdebug-prefix-map=%S/= -fno-ident -fdebug-compilation-dir=%S -fsanitize=array-bounds -fsanitize-trap=array-bounds -fsanitize-annotate-debug-info=array-bounds -triple x86_64 -debug-info-kind=limited %s -o - | FileCheck --check-prefix=CHECK-TRAP %s
+// RUN: %clang_cc1 -emit-llvm -fdebug-prefix-map=%S/= -fno-ident -fdebug-compilation-dir=%S -fsanitize=array-bounds                              -fsanitize-annotate-debug-info=array-bounds -triple x86_64 -debug-info-kind=limited %s -o - | FileCheck --check-prefix=CHECK-NOTRAP %s
 
 int f();
 void d(double*);
diff --git a/clang/test/Driver/fsanitize.c b/clang/test/Driver/fsanitize.c
