[clang] [sanitizer] add pseudofunction to indicate array-bounds check #128977

fmayer · 2025-02-27T01:11:12Z

With this, we can:

use profilers to estimate how many cycles we spend on these checks (subject to caveats),
more easily see why we crashed.

Created using spr 1.3.4

llvmbot · 2025-03-04T22:30:23Z

@llvm/pr-subscribers-debuginfo

@llvm/pr-subscribers-clang

Author: Florian Mayer (fmayer)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/128977.diff

4 Files Affected:

(modified) clang/lib/CodeGen/CGDebugInfo.cpp (+12-7)
(modified) clang/lib/CodeGen/CGDebugInfo.h (+10-3)
(modified) clang/lib/CodeGen/CGExpr.cpp (+14)
(modified) clang/test/CodeGen/bounds-checking-debuginfo.c (+22-16)

diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp
index 46ad11e64c4d5..0f71795be592b 100644
--- a/clang/lib/CodeGen/CGDebugInfo.cpp
+++ b/clang/lib/CodeGen/CGDebugInfo.cpp
@@ -1771,12 +1771,12 @@ llvm::DIType *CGDebugInfo::createFieldType(
 }
 
 llvm::DISubprogram *
-CGDebugInfo::createInlinedTrapSubprogram(StringRef FuncName,
-                                         llvm::DIFile *FileScope) {
+CGDebugInfo::createInlinedSubprogram(StringRef FuncName,
+                                     llvm::DIFile *FileScope) {
   // We are caching the subprogram because we don't want to duplicate
   // subprograms with the same message. Note that `SPFlagDefinition` prevents
   // subprograms from being uniqued.
-  llvm::DISubprogram *&SP = InlinedTrapFuncMap[FuncName];
+  llvm::DISubprogram *&SP = InlinedSubprogramMap[FuncName];
 
   if (!SP) {
     llvm::DISubroutineType *DIFnTy = DBuilder.createSubroutineType(nullptr);
@@ -3598,6 +3598,14 @@ llvm::DIMacroFile *CGDebugInfo::CreateTempMacroFile(llvm::DIMacroFile *Parent,
   return DBuilder.createTempMacroFile(Parent, Line, FName);
 }
 
+llvm::DILocation *CGDebugInfo::CreateSyntheticInline(llvm::DebugLoc Location,
+                                                     StringRef FuncName) {
+  llvm::DISubprogram *TrapSP =
+      createInlinedSubprogram(FuncName, Location->getFile());
+  return llvm::DILocation::get(CGM.getLLVMContext(), /*Line=*/0, /*Column=*/0,
+                               /*Scope=*/TrapSP, /*InlinedAt=*/Location);
+}
+
 llvm::DILocation *CGDebugInfo::CreateTrapFailureMessageFor(
     llvm::DebugLoc TrapLocation, StringRef Category, StringRef FailureMsg) {
   // Create a debug location from `TrapLocation` that adds an artificial inline
@@ -3609,10 +3617,7 @@ llvm::DILocation *CGDebugInfo::CreateTrapFailureMessageFor(
   FuncName += "$";
   FuncName += FailureMsg;
 
-  llvm::DISubprogram *TrapSP =
-      createInlinedTrapSubprogram(FuncName, TrapLocation->getFile());
-  return llvm::DILocation::get(CGM.getLLVMContext(), /*Line=*/0, /*Column=*/0,
-                               /*Scope=*/TrapSP, /*InlinedAt=*/TrapLocation);
+  return CreateSyntheticInline(TrapLocation, FuncName);
 }
 
 static QualType UnwrapTypeForDebugInfo(QualType T, const ASTContext &C) {
diff --git a/clang/lib/CodeGen/CGDebugInfo.h b/clang/lib/CodeGen/CGDebugInfo.h
index 38f73eca561b7..cdca1452e2dcd 100644
--- a/clang/lib/CodeGen/CGDebugInfo.h
+++ b/clang/lib/CodeGen/CGDebugInfo.h
@@ -355,12 +355,12 @@ class CGDebugInfo {
       llvm::ArrayRef<llvm::Metadata *> PreviousFieldsDI, const RecordDecl *RD);
 
   /// A cache that maps names of artificial inlined functions to subprograms.
-  llvm::StringMap<llvm::DISubprogram *> InlinedTrapFuncMap;
+  llvm::StringMap<llvm::DISubprogram *> InlinedSubprogramMap;
 
   /// A function that returns the subprogram corresponding to the artificial
   /// inlined function for traps.
-  llvm::DISubprogram *createInlinedTrapSubprogram(StringRef FuncName,
-                                                  llvm::DIFile *FileScope);
+  llvm::DISubprogram *createInlinedSubprogram(StringRef FuncName,
+                                              llvm::DIFile *FileScope);
 
   /// Helpers for collecting fields of a record.
   /// @{
@@ -635,6 +635,13 @@ class CGDebugInfo {
   llvm::DILocation *CreateTrapFailureMessageFor(llvm::DebugLoc TrapLocation,
                                                 StringRef Category,
                                                 StringRef FailureMsg);
+  /// Create a debug location from `Location` that adds an artificial inline
+  /// frame where the frame name is FuncName
+  ///
+  /// This is used to indiciate instructions that come from compiler
+  /// instrumentation.
+  llvm::DILocation *CreateSyntheticInline(llvm::DebugLoc Location,
+                                          StringRef FuncName);
 
 private:
   /// Emit call to llvm.dbg.declare for a variable declaration.
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 191912ca7d800..7426c45a52e0f 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -31,6 +31,7 @@
 #include "clang/AST/StmtVisitor.h"
 #include "clang/Basic/Builtins.h"
 #include "clang/Basic/CodeGenOptions.h"
+#include "clang/Basic/Module.h"
 #include "clang/Basic/SourceManager.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/ScopeExit.h"
@@ -60,8 +61,14 @@ namespace clang {
 llvm::cl::opt<bool> ClSanitizeGuardChecks(
     "ubsan-guard-checks", llvm::cl::Optional,
     llvm::cl::desc("Guard UBSAN checks with `llvm.allow.ubsan.check()`."));
+
 } // namespace clang
 
+static llvm::cl::opt<bool> ClArrayBoundsPseudoFn(
+    "array-bounds-pseudofn", llvm::cl::Hidden, llvm::cl::Optional,
+    llvm::cl::desc("Emit debug info that places array-bounds instrumentation "
+                   "in an inline function called __ubsan_check_array_bounds."));
+
 //===--------------------------------------------------------------------===//
 //                        Defines for metadata
 //===--------------------------------------------------------------------===//
@@ -1215,6 +1222,13 @@ void CodeGenFunction::EmitBoundsCheckImpl(const Expr *E, llvm::Value *Bound,
 
   SanitizerScope SanScope(this);
 
+  llvm::DILocation *CheckDI = Builder.getCurrentDebugLocation();
+  if (ClArrayBoundsPseudoFn && CheckDI) {
+    CheckDI = getDebugInfo()->CreateSyntheticInline(
+        Builder.getCurrentDebugLocation(), "__ubsan_check_array_bounds");
+  }
+  ApplyDebugLocation ApplyTrapDI(*this, CheckDI);
+
   bool IndexSigned = IndexType->isSignedIntegerOrEnumerationType();
   llvm::Value *IndexVal = Builder.CreateIntCast(Index, SizeTy, IndexSigned);
   llvm::Value *BoundVal = Builder.CreateIntCast(Bound, SizeTy, false);
diff --git a/clang/test/CodeGen/bounds-checking-debuginfo.c b/clang/test/CodeGen/bounds-checking-debuginfo.c
index 61fa56590a2a2..61c7af6e7c5b8 100644
--- a/clang/test/CodeGen/bounds-checking-debuginfo.c
+++ b/clang/test/CodeGen/bounds-checking-debuginfo.c
@@ -1,6 +1,6 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
-// RUN: %clang_cc1 -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 -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 -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
 
 
 int f();
@@ -28,10 +28,10 @@ void d(double*);
 // CHECK-TRAP-NEXT:    call void @llvm.ubsantrap(i8 18) #[[ATTR3:[0-9]+]], !dbg [[DBG23]], !nosanitize [[META10]]
 // CHECK-TRAP-NEXT:    unreachable, !dbg [[DBG23]], !nosanitize [[META10]]
 // CHECK-TRAP:       [[CONT]]:
-// CHECK-TRAP-NEXT:    [[IDXPROM:%.*]] = sext i32 [[CALL]] to i64, !dbg [[DBG23]]
-// CHECK-TRAP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x double], ptr [[A]], i64 0, i64 [[IDXPROM]], !dbg [[DBG23]]
-// CHECK-TRAP-NEXT:    [[TMP2:%.*]] = load double, ptr [[ARRAYIDX]], align 8, !dbg [[DBG23]]
-// CHECK-TRAP-NEXT:    ret double [[TMP2]], !dbg [[DBG24:![0-9]+]]
+// CHECK-TRAP-NEXT:    [[IDXPROM:%.*]] = sext i32 [[CALL]] to i64, !dbg [[DBG26:![0-9]+]]
+// CHECK-TRAP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x double], ptr [[A]], i64 0, i64 [[IDXPROM]], !dbg [[DBG26]]
+// CHECK-TRAP-NEXT:    [[TMP2:%.*]] = load double, ptr [[ARRAYIDX]], align 8, !dbg [[DBG26]]
+// CHECK-TRAP-NEXT:    ret double [[TMP2]], !dbg [[DBG27:![0-9]+]]
 //
 // CHECK-NOTRAP-LABEL: define dso_local double @f1(
 // CHECK-NOTRAP-SAME: i32 noundef [[B:%.*]], i32 noundef [[I:%.*]]) #[[ATTR0:[0-9]+]] !dbg [[DBG4:![0-9]+]] {
@@ -49,16 +49,16 @@ void d(double*);
 // CHECK-NOTRAP-NEXT:    [[CALL:%.*]] = call i32 (...) @f(), !dbg [[DBG22:![0-9]+]]
 // CHECK-NOTRAP-NEXT:    [[TMP0:%.*]] = sext i32 [[CALL]] to i64, !dbg [[DBG23:![0-9]+]], !nosanitize [[META10:![0-9]+]]
 // CHECK-NOTRAP-NEXT:    [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 10, !dbg [[DBG23]], !nosanitize [[META10]]
-// CHECK-NOTRAP-NEXT:    br i1 [[TMP1]], label %[[CONT:.*]], label %[[HANDLER_OUT_OF_BOUNDS:.*]], !dbg [[DBG23]], !prof [[PROF24:![0-9]+]], !nosanitize [[META10]]
+// CHECK-NOTRAP-NEXT:    br i1 [[TMP1]], label %[[CONT:.*]], label %[[HANDLER_OUT_OF_BOUNDS:.*]], !dbg [[DBG23]], !prof [[PROF27:![0-9]+]], !nosanitize [[META10]]
 // CHECK-NOTRAP:       [[HANDLER_OUT_OF_BOUNDS]]:
 // CHECK-NOTRAP-NEXT:    [[TMP2:%.*]] = zext i32 [[CALL]] to i64, !dbg [[DBG23]], !nosanitize [[META10]]
 // CHECK-NOTRAP-NEXT:    call void @__ubsan_handle_out_of_bounds_abort(ptr @[[GLOB2:[0-9]+]], i64 [[TMP2]]) #[[ATTR3:[0-9]+]], !dbg [[DBG23]], !nosanitize [[META10]]
 // CHECK-NOTRAP-NEXT:    unreachable, !dbg [[DBG23]], !nosanitize [[META10]]
 // CHECK-NOTRAP:       [[CONT]]:
-// CHECK-NOTRAP-NEXT:    [[IDXPROM:%.*]] = sext i32 [[CALL]] to i64, !dbg [[DBG23]]
-// CHECK-NOTRAP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x double], ptr [[A]], i64 0, i64 [[IDXPROM]], !dbg [[DBG23]]
-// CHECK-NOTRAP-NEXT:    [[TMP3:%.*]] = load double, ptr [[ARRAYIDX]], align 8, !dbg [[DBG23]]
-// CHECK-NOTRAP-NEXT:    ret double [[TMP3]], !dbg [[DBG25:![0-9]+]]
+// CHECK-NOTRAP-NEXT:    [[IDXPROM:%.*]] = sext i32 [[CALL]] to i64, !dbg [[DBG26:![0-9]+]]
+// CHECK-NOTRAP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x double], ptr [[A]], i64 0, i64 [[IDXPROM]], !dbg [[DBG26]]
+// CHECK-NOTRAP-NEXT:    [[TMP3:%.*]] = load double, ptr [[ARRAYIDX]], align 8, !dbg [[DBG26]]
+// CHECK-NOTRAP-NEXT:    ret double [[TMP3]], !dbg [[DBG28:![0-9]+]]
 //
 double f1(int b, int i) {
   double a[10];
@@ -88,8 +88,11 @@ double f1(int b, int i) {
 // CHECK-TRAP: [[DBG20]] = !DILocation(line: 65, column: 5, scope: [[DBG4]])
 // CHECK-TRAP: [[DBG21]] = !DILocation(line: 65, column: 3, scope: [[DBG4]])
 // CHECK-TRAP: [[DBG22]] = !DILocation(line: 66, column: 12, scope: [[DBG4]])
-// CHECK-TRAP: [[DBG23]] = !DILocation(line: 66, column: 10, scope: [[DBG4]])
-// CHECK-TRAP: [[DBG24]] = !DILocation(line: 66, column: 3, scope: [[DBG4]])
+// CHECK-TRAP: [[DBG23]] = !DILocation(line: 0, scope: [[META24:![0-9]+]], inlinedAt: [[DBG26]])
+// CHECK-TRAP: [[META24]] = distinct !DISubprogram(name: "__ubsan_check_array_bounds", scope: [[META5]], file: [[META5]], type: [[META25:![0-9]+]], flags: DIFlagArtificial, spFlags: DISPFlagDefinition, unit: [[META0]])
+// CHECK-TRAP: [[META25]] = !DISubroutineType(types: null)
+// CHECK-TRAP: [[DBG26]] = !DILocation(line: 66, column: 10, scope: [[DBG4]])
+// CHECK-TRAP: [[DBG27]] = !DILocation(line: 66, column: 3, scope: [[DBG4]])
 //.
 // CHECK-NOTRAP: [[META0:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C11, file: [[META1:![0-9]+]], isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, splitDebugInlining: false, nameTableKind: None)
 // CHECK-NOTRAP: [[META1]] = !DIFile(filename: "<stdin>", directory: {{.*}})
@@ -112,7 +115,10 @@ double f1(int b, int i) {
 // CHECK-NOTRAP: [[DBG20]] = !DILocation(line: 65, column: 5, scope: [[DBG4]])
 // CHECK-NOTRAP: [[DBG21]] = !DILocation(line: 65, column: 3, scope: [[DBG4]])
 // CHECK-NOTRAP: [[DBG22]] = !DILocation(line: 66, column: 12, scope: [[DBG4]])
-// CHECK-NOTRAP: [[DBG23]] = !DILocation(line: 66, column: 10, scope: [[DBG4]])
-// CHECK-NOTRAP: [[PROF24]] = !{!"branch_weights", i32 1048575, i32 1}
-// CHECK-NOTRAP: [[DBG25]] = !DILocation(line: 66, column: 3, scope: [[DBG4]])
+// CHECK-NOTRAP: [[DBG23]] = !DILocation(line: 0, scope: [[META24:![0-9]+]], inlinedAt: [[DBG26]])
+// CHECK-NOTRAP: [[META24]] = distinct !DISubprogram(name: "__ubsan_check_array_bounds", scope: [[META5]], file: [[META5]], type: [[META25:![0-9]+]], flags: DIFlagArtificial, spFlags: DISPFlagDefinition, unit: [[META0]])
+// CHECK-NOTRAP: [[META25]] = !DISubroutineType(types: null)
+// CHECK-NOTRAP: [[DBG26]] = !DILocation(line: 66, column: 10, scope: [[DBG4]])
+// CHECK-NOTRAP: [[PROF27]] = !{!"branch_weights", i32 1048575, i32 1}
+// CHECK-NOTRAP: [[DBG28]] = !DILocation(line: 66, column: 3, scope: [[DBG4]])
 //.

Michael137 · 2025-03-05T10:40:04Z

Could you elaborate on how this will be used/the motivation for this?

delcypher · 2025-03-05T19:39:29Z

@fmayer The usual approach for indicating instrumentation in Clang is to use opt-remarks. This is the approach we use for -fbounds-safety.

In -fbounds-safety we embed "trap reasons" in debug info so that debuggers and symbolication tools can better understand the reason for trapping.

What's the reason for using debug info, instead of opt-remarks here?

fmayer · 2025-03-05T19:44:49Z

@fmayer The usual approach for indicating instrumentation in Clang is to use opt-remarks. This is the approach we use for -fbounds-safety.

In -fbounds-safety we embed "trap reasons" in debug info so that debuggers and symbolication tools can better understand the reason for trapping.

What's the reason for using debug info, instead of opt-remarks here?

The commit description is maybe not very clear. This is not for compiile time as opt remarks, but for run time. By doing this, we can

(more importantly) use profilers to estimate how many cycles we spend on these checks (subject to caveats),
more easily see why we crashed.

delcypher

Thanks for explaining the purpose.

Regarding the "more easily see why we crashed." please be aware I have a GSoC proposal to basically do this using the createTrapFailureMessageFor. So if possible please don't tackle what I describe in the proposal before a GSoC student has had a chance to do this. To be clear what you've done in this PR is different from I'm proposing so they don't conflict.

Also

Please give a chance for Clang Debug Info contributors to look over this (CC @adrian-prantl) before landing this.
When possible add as reviewers previous people who worked on the code. In this particular case I believe this was @ahatanak
Please make sure the final commit message clearly explains the intent and why opt-remarks are not a good fit for what you're trying to do.

delcypher · 2025-03-05T21:18:19Z

clang/lib/CodeGen/CGDebugInfo.cpp

@@ -3598,6 +3598,14 @@ llvm::DIMacroFile *CGDebugInfo::CreateTempMacroFile(llvm::DIMacroFile *Parent,
  return DBuilder.createTempMacroFile(Parent, Line, FName);
 }

+llvm::DILocation *CGDebugInfo::CreateSyntheticInline(llvm::DebugLoc Location,
+                                                     StringRef FuncName) {
+  llvm::DISubprogram *TrapSP =


Nit. The name TrapSP doesn't make much sense here given this function isn't specifically for traps.

delcypher · 2025-03-05T21:22:53Z

clang/lib/CodeGen/CGDebugInfo.h

+  ///
+  /// This is used to indiciate instructions that come from compiler
+  /// instrumentation.
+  llvm::DILocation *CreateSyntheticInline(llvm::DebugLoc Location,


Nit. Maybe call it CreateSyntheticInlineAt ? Those who know more about Clang debug info generation (e.g. @adrian-prantl @felipepiovezan @Michael137 ) might have ideas on a better name though.

vitalybuka · 2025-03-09T02:40:06Z

clang/lib/CodeGen/CGDebugInfo.h

@@ -355,12 +355,12 @@ class CGDebugInfo {
      llvm::ArrayRef<llvm::Metadata *> PreviousFieldsDI, const RecordDecl *RD);

  /// A cache that maps names of artificial inlined functions to subprograms.
-  llvm::StringMap<llvm::DISubprogram *> InlinedTrapFuncMap;
+  llvm::StringMap<llvm::DISubprogram *> InlinedSubprogramMap;


Can you rename in a separate NFC patch?

Created using spr 1.3.4

fmayer · 2025-03-25T22:56:53Z

given no response from other reviewers for 3 weeks, I will go ahead and merge this. happy to address any feedback people might have in a followup.

…#128977) With this, we can: * use profilers to estimate how many cycles we spend on these checks (subject to caveats), * more easily see why we crashed.

fmayer · 2025-03-28T20:22:52Z

now actually committed to main: c0952a9

…ly replace '-mllvm -array-bounds-pseudofn' Florian1 introduced '-mllvm -array-bounds-pseudofn' (llvm#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-add-pseudo-functions, and connects it to the existing array-bounds-pseudo-fn functionality i.e., -fsanitize-add-pseudo-functions=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-add-pseudo-functions=array-bounds.

@fmayer

…y replace '-mllvm -array-bounds-pseudofn' (#138577) @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.

@fmayer

…o and partly replace '-mllvm -array-bounds-pseudofn' (#138577) @fmayer introduced '-mllvm -array-bounds-pseudofn' (llvm/llvm-project#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.

@fmayer

…y replace '-mllvm -array-bounds-pseudofn' (llvm#138577) @fmayer introduced '-mllvm -array-bounds-pseudofn' (llvm#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.

This connects the -fsanitize-annotate-debug-info plumbing (llvm#138577) to CFI check codegen. Updates the tests from llvm#139149. A side effect is that __ubsan_check_array_bounds is renamed to __ubsan_check_array-bounds. This affects clang/test/CodeGen/bounds-checking-debuginfo.c from llvm#128977

This extends llvm#138577 to more UBSan checks. Note that the annotations are less detailed: they will always be __ubsan_check_singularity, rather than using the SanitizerKind (previous behavior, which is not always possible for all UBSan checks) or SanitizerHandler. This is a (minor) regression compared to llvm#128977 and llvm#139809. Updates the tests from llvm#128976, llvm#139149 and llvm#141814.

fmayer added 7 commits February 26, 2025 17:11

[𝘀𝗽𝗿] initial version

7c73db2

Created using spr 1.3.4

fmt

0fe2ba3

Created using spr 1.3.4

rename & fmt

a16b7a8

Created using spr 1.3.4

naming

ff406cc

Created using spr 1.3.4

reb

694b23b

Created using spr 1.3.4

rebase

d0c5ee9

Created using spr 1.3.4

name

cf4d310

Created using spr 1.3.4

fmayer marked this pull request as ready for review March 4, 2025 22:29

llvmbot added clang Clang issues not falling into any other category clang:codegen IR generation bugs: mangling, exceptions, etc. debuginfo labels Mar 4, 2025

fmayer requested a review from vitalybuka March 4, 2025 22:30

delcypher requested review from delcypher and ahatanak March 5, 2025 19:33

delcypher requested review from adrian-prantl and Michael137 March 5, 2025 21:15

delcypher approved these changes Mar 5, 2025

View reviewed changes

vitalybuka reviewed Mar 9, 2025

View reviewed changes

vitalybuka approved these changes Mar 9, 2025

View reviewed changes

fmayer added 2 commits March 25, 2025 14:58

nits

d4370dc

Created using spr 1.3.4

rebase

aaffd83

Created using spr 1.3.4

fmayer merged commit bf4da50 into users/fmayer/spr/main.clang-sanitizer-add-pseudofunction-to-indicate-array-bounds-check Mar 25, 2025
11 of 14 checks passed

fmayer deleted the users/fmayer/spr/clang-sanitizer-add-pseudofunction-to-indicate-array-bounds-check branch March 25, 2025 22:57

thurstond mentioned this pull request May 5, 2025

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

Merged

thurstond mentioned this pull request May 13, 2025

[cfi] Enable -fsanitize-annotate-debug-info functionality for CFI checks #139809

Merged

thurstond mentioned this pull request May 29, 2025

[ubsan] Add more -fsanitize-annotate-debug-info checks #141997

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[clang] [sanitizer] add pseudofunction to indicate array-bounds check #128977

[clang] [sanitizer] add pseudofunction to indicate array-bounds check #128977

Uh oh!

fmayer commented Feb 27, 2025 •

edited

Loading

Uh oh!

llvmbot commented Mar 4, 2025 •

edited

Loading

Uh oh!

Michael137 commented Mar 5, 2025

Uh oh!

delcypher commented Mar 5, 2025

Uh oh!

fmayer commented Mar 5, 2025

Uh oh!

delcypher left a comment •

edited

Loading

Uh oh!

delcypher Mar 5, 2025

Uh oh!

fmayer Mar 25, 2025

Uh oh!

delcypher Mar 5, 2025

Uh oh!

fmayer Mar 25, 2025

Uh oh!

vitalybuka Mar 9, 2025

Uh oh!

fmayer Mar 25, 2025

Uh oh!

fmayer commented Mar 25, 2025

Uh oh!

Uh oh!

fmayer commented Mar 28, 2025

Uh oh!

Uh oh!

[clang] [sanitizer] add pseudofunction to indicate array-bounds check #128977

[clang] [sanitizer] add pseudofunction to indicate array-bounds check #128977

Uh oh!

Conversation

fmayer commented Feb 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Mar 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Michael137 commented Mar 5, 2025

Uh oh!

delcypher commented Mar 5, 2025

Uh oh!

fmayer commented Mar 5, 2025

Uh oh!

delcypher left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

delcypher Mar 5, 2025

Choose a reason for hiding this comment

Uh oh!

fmayer Mar 25, 2025

Choose a reason for hiding this comment

Uh oh!

delcypher Mar 5, 2025

Choose a reason for hiding this comment

Uh oh!

fmayer Mar 25, 2025

Choose a reason for hiding this comment

Uh oh!

vitalybuka Mar 9, 2025

Choose a reason for hiding this comment

Uh oh!

fmayer Mar 25, 2025

Choose a reason for hiding this comment

Uh oh!

fmayer commented Mar 25, 2025

Uh oh!

Uh oh!

fmayer commented Mar 28, 2025

Uh oh!

Uh oh!

fmayer commented Feb 27, 2025 •

edited

Loading

llvmbot commented Mar 4, 2025 •

edited

Loading

delcypher left a comment •

edited

Loading