[BuildLibCalls] Add noundef to the returned pointers of allocators and argument of free

[aqjune]

Summary

This patch adds noundef to the returned pointers of allocators (malloc, calloc, ...)
and the pointer argument of free.
The returned pointer of allocators cannot be poison or (partially) undef.
Since the pointer that is given to free should precisely have zero offset,
it cannot be poison or (partially) undef too.

For the size arguments of allocators, noundef wasn't attached simply because
I wasn't sure whether attaching it is okay or not.

[aqjune]

If it is desirable to clarify the behavior of malloc and free in LangRef as well, I'm willing to do it.

The definition of malloc that is consistent with this patch is:

• If an undefined value is given to malloc, it allocates a heap with one of possible values nondeterministically chosen. For example, malloc(16 + (undef & 4)) allocates a heap whose size is one of 16 ~ 19.
• If a poison value is given to malloc, it raises UB.

The definition of free is as follows:

• If an undefined value or poison is given to free, it is UB.

I remember that malloc also had an issue about when it is returning null. This issue is orthogonal with this patch, but if needed I'm happy to discuss it as well.

[jdoerfert]

LGTM. I think this is the right way to handle this. the size can be partially undef (="non-deterministic & variable") but everything else better is not.

[aqjune]

Thanks!

llvm/lib/Transforms/Utils/BuildLibCalls.cpp

 //===- BuildLibCalls.cpp - Utility builder for libcalls -------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //

 static bool setOnlyReadsMemory(Function &F, unsigned ArgNo) {
   if (F.hasParamAttribute(ArgNo, Attribute::ReadOnly))
     return false;
   F.addParamAttr(ArgNo, Attribute::ReadOnly);
   ++NumReadOnlyArg;
   return true;
 }
 
-static bool setRetAndArgsNoUndef(Function &F) {
-  bool Changed = false;
+static bool setRetNoUndef(Function &F) {
   if (!F.getReturnType()->isVoidTy() &&
       !F.hasAttribute(AttributeList::ReturnIndex, Attribute::NoUndef)) {
     F.addAttribute(AttributeList::ReturnIndex, Attribute::NoUndef);
     ++NumNoUndef;
-    Changed = true;
-  }
+    return true;
+  }
+  return false;
+
+static bool setArgsNoUndef(Function &F) {
+  bool Changed = false;
   for (unsigned ArgNo = 0; ArgNo < F.arg_size(); ++ArgNo) {
     if (!F.hasParamAttribute(ArgNo, Attribute::NoUndef)) {
       F.addParamAttr(ArgNo, Attribute::NoUndef);
       ++NumNoUndef;
       Changed = true;
     }
   }
   return Changed;
 }
 
+static bool setRetAndArgsNoUndef(Function &F) {
+  return setRetNoUndef(F) | setArgsNoUndef(F);
+}
+
 static bool setRetNonNull(Function &F) {
   assert(F.getReturnType()->isPointerTy() &&
          "nonnull applies only to pointers");
   if (F.hasAttribute(AttributeList::ReturnIndex, Attribute::NonNull))
     return false;
   F.addAttribute(AttributeList::ReturnIndex, Attribute::NonNull);
   ++NumNonNull;
   return true;
case LibFunc_setitimer:
     return Changed;
   case LibFunc_system:
     // May throw; "system" is a valid pthread cancellation point.
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotCapture(F, 0);
     Changed |= setOnlyReadsMemory(F, 0);
     return Changed;
   case LibFunc_malloc:
+    Changed |= setRetNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setRetDoesNotAlias(F);
     return Changed;
   case LibFunc_memcmp:
     Changed |= setOnlyAccessesArgMemory(F);
     Changed |= setOnlyReadsMemory(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
case LibFunc_mkdir:
     Changed |= setOnlyReadsMemory(F, 0);
     return Changed;
   case LibFunc_mktime:
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     return Changed;
   case LibFunc_realloc:
+    Changed |= setRetNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setRetDoesNotAlias(F);
     Changed |= setDoesNotCapture(F, 0);
     return Changed;
+  case LibFunc_reallocf:
+    Changed |= setRetNoUndef(F);
+    return Changed;
   case LibFunc_read:
     // May throw; "read" is a valid pthread cancellation point.
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotCapture(F, 1);
     return Changed;
   case LibFunc_rewind:
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
case LibFunc_readlink:
     return Changed;
   case LibFunc_write:
     // May throw; "write" is a valid pthread cancellation point.
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotCapture(F, 1);
     Changed |= setOnlyReadsMemory(F, 1);
     return Changed;
   case LibFunc_aligned_alloc:
+    Changed |= setRetNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setRetDoesNotAlias(F);
     return Changed;
   case LibFunc_bcopy:
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     Changed |= setDoesNotCapture(F, 1);
     Changed |= setOnlyReadsMemory(F, 0);
     return Changed;
   case LibFunc_bcmp:
     Changed |= setDoesNotThrow(F);
     Changed |= setOnlyAccessesArgMemory(F);
     Changed |= setOnlyReadsMemory(F);
     Changed |= setDoesNotCapture(F, 0);
     Changed |= setDoesNotCapture(F, 1);
     return Changed;
   case LibFunc_bzero:
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     return Changed;
   case LibFunc_calloc:
+    Changed |= setRetNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setRetDoesNotAlias(F);
     return Changed;
   case LibFunc_chmod:
   case LibFunc_chown:
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
case LibFunc_fdopen:
     Changed |= setOnlyReadsMemory(F, 1);
     return Changed;
   case LibFunc_feof:
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     return Changed;
   case LibFunc_free:
+    Changed |= setArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     return Changed;
   case LibFunc_fseek:
   case LibFunc_ftell:
   case LibFunc_fgetc:
   case LibFunc_fgetc_unlocked:
   case LibFunc_fseeko:
bool llvm::inferLibFuncAttributes(Function &F, const TargetLibraryInfo &TLI) {
   case LibFunc_vfscanf:
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     Changed |= setDoesNotCapture(F, 1);
     Changed |= setOnlyReadsMemory(F, 1);
     return Changed;
   case LibFunc_valloc:
+    Changed |= setRetNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setRetDoesNotAlias(F);
     return Changed;
   case LibFunc_vprintf:
     Changed |= setRetAndArgsNoUndef(F);
     Changed |= setDoesNotThrow(F);
     Changed |= setDoesNotCapture(F, 0);
     Changed |= setOnlyReadsMemory(F, 0);
bool llvm::inferLibFuncAttributes(Function &F, const TargetLibraryInfo &TLI) {
   case LibFunc_Znwm: // new(unsigned long)
   case LibFunc_Znaj: // new[](unsigned int)
   case LibFunc_Znam: // new[](unsigned long)
   case LibFunc_msvc_new_int: // new(unsigned int)
   case LibFunc_msvc_new_longlong: // new(unsigned long long)
   case LibFunc_msvc_new_array_int: // new[](unsigned int)
   case LibFunc_msvc_new_array_longlong: // new[](unsigned long long)
     // Operator new always returns a nonnull noalias pointer
+    Changed |= setRetNoUndef(F);
     Changed |= setRetNonNull(F);
     Changed |= setRetDoesNotAlias(F);
     return Changed;
   // TODO: add LibFunc entries for:
   // case LibFunc_memset_pattern4:
   // case LibFunc_memset_pattern8:
   case LibFunc_memset_pattern16:
     Changed |= setOnlyAccessesArgMemory(F);

llvm/test/Transforms/InferFunctionAttrs/annotate.ll

 ; RUN: opt < %s -mtriple=x86_64-- -inferattrs -S | FileCheck -check-prefix=CHECK-UNKNOWN %s
 ; RUN: opt < %s -mtriple=x86_64-- -passes=inferattrs -S | FileCheck -check-prefix=CHECK-UNKNOWN %s
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -inferattrs -S | FileCheck -check-prefix=CHECK -check-prefix=CHECK-DARWIN %s
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -inferattrs -S | FileCheck -check-prefix=CHECK -check-prefix=CHECK-LINUX %s
 ; RUN: opt < %s -mtriple=nvptx -inferattrs -S | FileCheck -check-prefix=CHECK-NVPTX %s
 
 ; operator new routines
 declare i8* @_Znwj(i64 )
-; CHECK: declare noalias nonnull i8* @_Znwj(i64) [[G0:#[0-9]+]]
+; CHECK: declare noalias noundef nonnull i8* @_Znwj(i64) [[G0:#[0-9]+]]
 declare i8* @_Znwm(i64)
-; CHECK: declare noalias nonnull i8* @_Znwm(i64) [[G0]]
+; CHECK: declare noalias noundef nonnull i8* @_Znwm(i64) [[G0]]
 
 declare i32 @__nvvm_reflect(i8*)
 ; CHECK-NVPTX: declare noundef i32 @__nvvm_reflect(i8* noundef) [[G0:#[0-9]+]]
 ; CHECK-NVPTX: attributes [[G0]] = { nofree nounwind readnone }
 
 
 ; Check all the libc functions (thereby also exercising the prototype check).
 ; Note that it's OK to modify these as attributes might be missing. These checks

 declare i32 @bcmp(i8*, i8*, i64)
 
 ; CHECK: declare void @bcopy(i8* nocapture readonly, i8* nocapture, i64) [[G1]]
 declare void @bcopy(i8*, i8*, i64)
 
 ; CHECK: declare void @bzero(i8* nocapture, i64) [[G1]]
 declare void @bzero(i8*, i64)
 
-; CHECK: declare noalias i8* @calloc(i64, i64) [[G1]]
+; CHECK: declare noalias noundef i8* @calloc(i64, i64) [[G1]]
 declare i8* @calloc(i64, i64)
 
 ; CHECK: declare double @cbrt(double) [[G0]]
 declare double @cbrt(double)
 
 ; CHECK: declare float @cbrtf(float) [[G0]]
 declare float @cbrtf(float)
 

 declare i32 @fputc(i32, %opaque*)
 
 ; CHECK: declare noundef i32 @fputs(i8* nocapture noundef readonly, %opaque* nocapture noundef) [[G1]]
 declare i32 @fputs(i8*, %opaque*)
 
 ; CHECK: declare noundef i64 @fread(i8* nocapture noundef, i64 noundef, i64 noundef, %opaque* nocapture noundef) [[G1]]
 declare i64 @fread(i8*, i64, i64, %opaque*)
 
-; CHECK: declare void @free(i8* nocapture) [[G3:#[0-9]+]]
+; CHECK: declare void @free(i8* nocapture noundef) [[G3:#[0-9]+]]
 declare void @free(i8*)
 
 ; CHECK: declare double @frexp(double, i32* nocapture) [[G1]]
 declare double @frexp(double, i32*)
 
 ; CHECK: declare float @frexpf(float, i32* nocapture) [[G1]]
 declare float @frexpf(float, i32*)
 

 declare x86_fp80 @logl(x86_fp80)
 
 ; CHECK: declare noundef i32 @lstat(i8* nocapture noundef readonly, %opaque* nocapture noundef) [[G1]]
 declare i32 @lstat(i8*, %opaque*)
 
 ; CHECK-LINUX: declare noundef i32 @lstat64(i8* nocapture noundef readonly, %opaque* nocapture noundef) [[G1]]
 declare i32 @lstat64(i8*, %opaque*)
 
-; CHECK: declare noalias i8* @malloc(i64) [[G1]]
+; CHECK: declare noalias noundef i8* @malloc(i64) [[G1]]
 declare i8* @malloc(i64)
 
 ; CHECK-LINUX: declare noalias i8* @memalign(i64, i64) [[G0]]
 declare i8* @memalign(i64, i64)
 
 ; CHECK: declare i8* @memccpy(i8* noalias, i8* noalias nocapture readonly, i32, i64) [[ARGMEMONLY_NOFREE_NOUNWIND:#[0-9]+]]
 declare i8* @memccpy(i8*, i8*, i32, i64)
 

 declare void @qsort(i8*, i64, i64, i32 (i8*, i8*)*)
 
 ; CHECK: declare noundef i64 @read(i32 noundef, i8* nocapture noundef, i64 noundef) [[G0]]
 declare i64 @read(i32, i8*, i64)
 
 ; CHECK: declare noundef i64 @readlink(i8* nocapture noundef readonly, i8* nocapture noundef, i64 noundef) [[G1]]
 declare i64 @readlink(i8*, i8*, i64)
 
-; CHECK: declare noalias i8* @realloc(i8* nocapture, i64) [[G3]]
+; CHECK: declare noalias noundef i8* @realloc(i8* nocapture, i64) [[G3]]
 declare i8* @realloc(i8*, i64)
 
-; CHECK: declare i8* @reallocf(i8*, i64)
+; CHECK: declare noundef i8* @reallocf(i8*, i64)
 declare i8* @reallocf(i8*, i64)
 
 ; CHECK: declare noundef i8* @realpath(i8* nocapture noundef readonly, i8* noundef) [[G1]]
 declare i8* @realpath(i8*, i8*)
 
 ; CHECK: declare noundef i32 @remove(i8* nocapture noundef readonly) [[G1]]
 declare i32 @remove(i8*)
 

 declare i32 @unsetenv(i8*)
 
 ; CHECK: declare noundef i32 @utime(i8* nocapture noundef readonly, %opaque* nocapture noundef readonly) [[G1]]
 declare i32 @utime(i8*, %opaque*)
 
 ; CHECK: declare noundef i32 @utimes(i8* nocapture noundef readonly, %opaque* nocapture noundef readonly) [[G1]]
 declare i32 @utimes(i8*, %opaque*)
 
-; CHECK: declare noalias i8* @valloc(i64) [[G1]]
+; CHECK: declare noalias noundef i8* @valloc(i64) [[G1]]
 declare i8* @valloc(i64)
 
 ; CHECK: declare noundef i32 @vfprintf(%opaque* nocapture noundef, i8* nocapture noundef readonly, %opaque* noundef) [[G1]]
 declare i32 @vfprintf(%opaque*, i8*, %opaque*)
 
 ; CHECK: declare noundef i32 @vfscanf(%opaque* nocapture noundef, i8* nocapture noundef readonly, %opaque* noundef) [[G1]]
 declare i32 @vfscanf(%opaque*, i8*, %opaque*)