Add uwtable annotation to modules when required by Darksonn · Pull Request #156973 · rust-lang/rust

Darksonn · 2026-05-26T15:05:32Z

When unwind tables are enabled with -Cforce-unwind-tables=y, Rust will annotate all functions with the uwtable annotation. However, this annotation is missing on modules, which leads to incorrect unwind tables being generated by LLVM for constructors (such as asan.module_ctor).

This was discovered because it leads to a crash in Linux when KASAN and dynamic shadow call stack are both enabled. In this scenario, the kernel uses the unwind tables to locate the paciasp and autiasp instructions in each function and patches the machine code at boot to use the shadow call stack instructions instead. However, LLVM's AArch64PointerAuth pass emits DWARF info for paciasp whenever -g is passed, but only emits DWARF info for autiasp when the uwtable attribute is present. Since the uwtable annotation is missing for modules, the relevant directives are generated for only the autiasp instruction in asan.module_ctor, and not for the paciasp instruction. This causes the kernel's dynamic SCS logic to patch the prolouge of asan.module_ctor, but not the epilogue. This leads to a crash as the shadow call stack becomes unbalanced.

The fact that LLVM doesn't use the same condition for whether to emit DWARF information for both instructions may be a separate bug in LLVM.

Relevant issue: llvm/llvm-project#188234

AI assistance was used to determine the root cause of this crash from the observed symptoms, and to write the tests. Also thanks to @samitolvanen and @maurer for debugging this issue.

Similar to this previous PR of mine: #130824

rustbot · 2026-05-26T15:05:40Z

r? @nnethercote

rustbot has assigned @nnethercote.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: compiler
compiler expanded to 73 candidates
Random selection from 18 candidates

Darksonn · 2026-05-26T18:10:42Z

Verified that passing -Zllvm_module_flag=uwtable:u32:2:max to the Linux kernel build fixes the boot failure that triggered this PR.

nnethercote · 2026-05-27T07:57:20Z

    DllExport = 2, // Function to be accessible from DLL.
 }

+/// Must match the layout of `UWTableKind`.


This struct is UWTableKind. Did you mean to name a different type here?

View changes since the review

I'm referring to the C++ enum of the same name on the LLVM side.

Can you add that information to the comment? It wasn't clear to me. In nearby cases the LLVM type has a different name to the Rust type.

I checked a few other constants, and half of them just add an LLVM prefix even though it doesn't have that prefix in LLVM. The other half does have the prefix on the LLVM side. Anyways, I adjusted the comment.

nnethercote · 2026-05-27T11:09:35Z

@bors r+ rollup

rust-bors · 2026-05-27T11:09:38Z

📌 Commit 90fe8cc has been approved by nnethercote

It is now in the queue for this repository.

@samitolvanen

…nethercote Add uwtable annotation to modules when required When unwind tables are enabled with `-Cforce-unwind-tables=y`, Rust will annotate all functions with the `uwtable` annotation. However, this annotation is missing on modules, which leads to incorrect unwind tables being generated by LLVM for constructors (such as `asan.module_ctor`). This was discovered because it leads to a crash in Linux when KASAN and dynamic shadow call stack are both enabled. In this scenario, the kernel uses the unwind tables to locate the `paciasp` and `autiasp` instructions in each function and patches the machine code at boot to use the shadow call stack instructions instead. However, LLVM's AArch64PointerAuth pass emits DWARF info for `paciasp` whenever `-g` is passed, but only emits DWARF info for `autiasp` when the `uwtable` attribute is present. Since the `uwtable` annotation is missing for modules, the relevant directives are generated for only the `autiasp` instruction in `asan.module_ctor`, and not for the `paciasp` instruction. This causes the kernel's dynamic SCS logic to patch the prolouge of `asan.module_ctor`, but not the epilogue. This leads to a crash as the shadow call stack becomes unbalanced. The fact that LLVM doesn't use the same condition for whether to emit DWARF information for both instructions may be a separate bug in LLVM. Relevant issue: llvm/llvm-project#188234 AI assistance was used to determine the root cause of this crash from the observed symptoms, and to write the tests. Also thanks to @samitolvanen and @maurer for debugging this issue. Similar to this previous PR of mine: rust-lang#130824

bjorn3 · 2026-05-27T12:15:58Z

If uwtable is emitted at the module level, is it still necessary to emit it at the function level too?

Darksonn · 2026-05-27T12:18:26Z

Yes I believe the only effect of setting it on the module level is to inherit it for compiler-generated functions such as asan.module_ctor. Ordinary functions do not inherit it.

Furthermore, this matches clang.

Rollup of 8 pull requests Successful merges: - #156970 (coverage: Use original HIR info for synthetic by-move coroutine bodies) - #156390 (Constify Iterator-related methods and functions) - #156401 (rustdoc: deterministic sorting for `doc_cfg` badges) - #156845 (Clarify "infinite size" in cyclic-type diagnostic refers to the type name) - #156973 (Add uwtable annotation to modules when required) - #156985 (Limit the additional DLL to Windows) - #156988 (interpret/validity: properly treat zero-variant enums so that we do not have to check layout.is_uninhabited) - #157002 (std: Fix thread::available_parallelism on Redox targets)

Rollup of 10 pull requests Successful merges: - #156970 (coverage: Use original HIR info for synthetic by-move coroutine bodies) - #157022 (MIR inlining: allow backends to opt-in to inlining intrinsics) - #157026 (miri subtree update) - #156390 (Constify Iterator-related methods and functions) - #156845 (Clarify "infinite size" in cyclic-type diagnostic refers to the type name) - #156955 (Fix const-eval of shared generic reborrows) - #156973 (Add uwtable annotation to modules when required) - #156985 (Limit the additional DLL to Windows) - #156988 (interpret/validity: properly treat zero-variant enums so that we do not have to check layout.is_uninhabited) - #157002 (std: Fix thread::available_parallelism on Redox targets)

JonathanBrouwer · 2026-05-28T10:36:57Z

@rust-timer build c42badd

Rollup of 10 pull requests Successful merges: - rust-lang/rust#156970 (coverage: Use original HIR info for synthetic by-move coroutine bodies) - rust-lang/rust#157022 (MIR inlining: allow backends to opt-in to inlining intrinsics) - rust-lang/rust#157026 (miri subtree update) - rust-lang/rust#156390 (Constify Iterator-related methods and functions) - rust-lang/rust#156845 (Clarify "infinite size" in cyclic-type diagnostic refers to the type name) - rust-lang/rust#156955 (Fix const-eval of shared generic reborrows) - rust-lang/rust#156973 (Add uwtable annotation to modules when required) - rust-lang/rust#156985 (Limit the additional DLL to Windows) - rust-lang/rust#156988 (interpret/validity: properly treat zero-variant enums so that we do not have to check layout.is_uninhabited) - rust-lang/rust#157002 (std: Fix thread::available_parallelism on Redox targets)

rust-timer · 2026-05-28T13:05:48Z

Finished benchmarking commit (c42badd): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

This perf run didn't have relevant results for this metric.

Max RSS (memory usage)

Results (secondary 6.3%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	6.3%	[5.3%, 7.3%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Cycles

Results (secondary 3.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	3.1%	[2.4%, 3.8%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 511.088s -> 514.114s (0.59%)
Artifact size: 400.69 MiB -> 400.76 MiB (0.02%)

…hercote Add frame pointer annotations to the module While working on rust-lang#156973 I found llvm/llvm-project#188234 which points out that the frame pointer annotation is also missing.

Rollup merge of #156980 - Darksonn:frame-ptrs-module, r=nnethercote Add frame pointer annotations to the module While working on #156973 I found llvm/llvm-project#188234 which points out that the frame pointer annotation is also missing.

Due to a rustc bug [1] the -Cforce-unwind-tables=y flag only emits the uwtable annotation for functions, but not for the module. This means that compiler-generated functions such as 'asan.module_ctor' do not receive the uwtable annotation. When CONFIG_UNWIND_PATCH_PAC_INTO_SCS is enabled, this leads to boot failures because the dwarf information emitted for the kasan constructors is wrong, which causes the SCS boot patching code to patch the constructor in an illegal manner. Specifically, the paciasp instruction is patched, but the autiasp instruction is not. This mismatch leads to a crash when the constructor is called during boot. ================================================================== BUG: KASAN: global-out-of-bounds in do_basic_setup+0x4c/0x90 Read of size 8 at addr ffffffe3cc7eb488 by task swapper/0/1 Specifically the faulting instruction is the (*fn)() to invoke the constructor in do_ctors() of the init/main.c file. Once the fix lands in rustc, this flag can be made conditional on the rustc version. Note that passing the flag on a rustc with the fix present has no effect. [ The fix [1] has landed for Rust 1.98.0 (expected release on 2026-08-20). Thus add a version check as discussed. - Miguel ] Fixes: d077242 ("rust: support for shadow call stack sanitizer") Cc: stable@kernel.org Link: rust-lang/rust#156973 [1] Reported-by: Bo Ye <bo.ye@mediatek.com> Debugged-by: Isaac Manjarres <isaacmanjarres@google.com> Debugged-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Isaac Manjarres <isaacmanjarres@google.com> Signed-off-by: Alice Ryhl <aliceryhl@google.com> Link: https://patch.msgid.link/20260527-uwtable-module-flag-v1-1-caa41342be4b@google.com [ Adjusted link and comment. - Miguel ] Signed-off-by: Miguel Ojeda <ojeda@kernel.org>

Add frame pointer annotations to the module While working on rust-lang/rust#156973 I found llvm/llvm-project#188234 which points out that the frame pointer annotation is also missing.

Darksonn added 2 commits May 26, 2026 14:32

Emit uwtable annotation for modules

f46fade

Add tests for uwtable annotations on modules

581bf3c

rustbot assigned nnethercote May 26, 2026

Darksonn added the A-rust-for-linux Relevant for the Rust-for-Linux project label May 26, 2026

Darksonn mentioned this pull request May 26, 2026

Add frame pointer annotations to the module #156980

Merged

nnethercote reviewed May 27, 2026

View reviewed changes

Adjust UWTableKind comment ref to llvm type

90fe8cc

nnethercote approved these changes May 27, 2026

View reviewed changes

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels May 27, 2026

GuillaumeGomez mentioned this pull request May 27, 2026

Rollup of 8 pull requests #157020

Closed

GuillaumeGomez mentioned this pull request May 27, 2026

Rollup of 10 pull requests #157031

Merged

rust-bors Bot merged commit 0846622 into rust-lang:main May 27, 2026
12 checks passed

rustbot added this to the 1.98.0 milestone May 27, 2026

ojeda mentioned this pull request May 28, 2026

rustc wanted features & bugfixes Rust-for-Linux/linux#355

Open

75 tasks

Darksonn deleted the unwind-tables-module branch May 28, 2026 05:53

This comment has been minimized.

Sign in to view

Darksonn removed the S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. label May 29, 2026

Uh oh!

Conversation

Darksonn commented May 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented May 26, 2026

Uh oh!

Darksonn commented May 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nnethercote May 27, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Darksonn May 27, 2026

Choose a reason for hiding this comment

Uh oh!

nnethercote May 27, 2026

Choose a reason for hiding this comment

Uh oh!

Darksonn May 27, 2026

Choose a reason for hiding this comment

Uh oh!

nnethercote commented May 27, 2026

Uh oh!

rust-bors Bot commented May 27, 2026

Uh oh!

bjorn3 commented May 27, 2026

Uh oh!

Darksonn commented May 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

JonathanBrouwer commented May 28, 2026

Uh oh!

This comment has been minimized.

rust-timer commented May 28, 2026

Overall result: no relevant changes - no action needed

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Darksonn commented May 26, 2026 •

edited

Loading

Darksonn commented May 26, 2026 •

edited

Loading

nnethercote May 27, 2026 •

edited by rustbot

Loading

Darksonn commented May 27, 2026 •

edited

Loading