Skip to content

[llvm] Bail out when meeting pointer with negative offset instead of … #120424

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Dec 20, 2024
Merged

[llvm] Bail out when meeting pointer with negative offset instead of … #120424

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ab73245
[llvm] Bail out when meeting pointer with negative offset instead of …
serge-sans-paille Dec 18, 2024
b228a0b
fixup! [llvm] Bail out when meeting pointer with negative offset inst…
serge-sans-paille Dec 18, 2024
a2ef6ff
fixup! [llvm] Bail out when meeting pointer with negative offset inst…
serge-sans-paille Dec 19, 2024
0a71d26
fixup! [llvm] Bail out when meeting pointer with negative offset inst…
serge-sans-paille Dec 19, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions llvm/lib/Analysis/MemoryBuiltins.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -838,9 +838,9 @@ OffsetSpan ObjectSizeOffsetVisitor::computeImpl(Value *V) {

// We end up pointing on a location that's outside of the original object.
if (ORT.knownBefore() && ORT.Before.isNegative()) {
// This means that we *may* be accessing memory before the allocation. It's
// unsure though, so bail out instead of returning a potentially misleading
// result.
// This means that we *may* be accessing memory before the allocation.
// Conservatively return an unknown size.
//
// TODO: working with ranges instead of value would make it possible to take
// a better decision.
if (Options.EvalMode == ObjectSizeOpts::Mode::Min ||
Expand Down
6 changes: 4 additions & 2 deletions llvm/test/Transforms/LowerConstantIntrinsics/builtin-object-size-range.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -107,8 +107,8 @@ define i64 @select_gep_offsets(i1 %cond) {
ret i64 %res
}

define i64 @select_gep_oob_offsets(i1 %cond) {
; CHECK-LABEL: @select_gep_oob_offsets(
define i64 @select_gep_oob_overapproximated_offsets(i1 %cond) {
; CHECK-LABEL: @select_gep_oob_overapproximated_offsets(
; CHECK-NEXT: [[BASE1:%.*]] = alloca [288 x i8], align 16
; CHECK-NEXT: [[SELECT0:%.*]] = select i1 [[COND:%.*]], i64 -4, i64 -64
; CHECK-NEXT: [[SELECT1:%.*]] = select i1 [[COND]], i64 16, i64 64
Expand All @@ -119,6 +119,8 @@ define i64 @select_gep_oob_offsets(i1 %cond) {
%base1 = alloca [288 x i8], align 16
%select0 = select i1 %cond, i64 -4, i64 -64
%select1 = select i1 %cond, i64 16, i64 64
; This nevers actually goes oob, but because we approcimate each select
Copy link
Member

@mstorsjo mstorsjo Dec 19, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Typo: approximate

; independently, this actually ranges in [16 - 64 ; 64 - 4] instead of [64 - 64; 16 - 4]
%gep0 = getelementptr inbounds nuw i8, ptr %base1, i64 %select1
%gep1 = getelementptr inbounds i8, ptr %gep0, i64 %select0
%call = call i64 @llvm.objectsize.i64.p0(ptr %gep1, i1 false, i1 true, i1 false)
Expand Down
Loading