alexcrichton assigned fitzgen to PR #9420.
alexcrichton requested fitzgen for a review on PR #9420.
alexcrichton opened PR #9420 from alexcrichton:stop-backtrace-with-fp
to bytecodealliance:main
(assigned to fitzgen):
Prior to this commit entry trampolines into wasm would record their
stack pointer at the time of the function call to wasm and then this
stack pointer was used to halt the stack walking process. The problem
with this though is that due to thetail
ABI it's possible that the
callee will update the caller's stack pointer temporarily. This means
that the recorded stack pointer at the time the trampoline called wasm
may differ from the callee's idea of what the stack pointer is when a
backtrace happens.To handle this condition when stack walking the frame pointer instead of
the stack pointer is now recorded when wasm is invoked. This frame
pointer is a trusted value as it's managed by Cranelift itself. This
additionally enables the stop condition for frame walking to be a
precise "it must be this value" condition.Put together this commit fixes an issue where when
return_call
is used
it's possible for the initial few frames of the stack to get lost in
stack traces. After this the frame pointer chain should always be
precisely walked in its entirety, even in the face of different numbers
of arguments and parameters asreturn_call
instructions are executed.
alexcrichton requested wasmtime-fuzz-reviewers for a review on PR #9420.
alexcrichton requested wasmtime-core-reviewers for a review on PR #9420.
github-actions[bot] commented on PR #9420:
Subscribe to Label Action
cc @fitzgen
<details>
This issue or pull request has been labeled: "fuzzing", "wasmtime:api"Thus the following users have been cc'd because of the following labels:
- fitzgen: fuzzing
To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.
Learn more.
</details>
fitzgen submitted PR review:
Very nice clean up
fitzgen merged PR #9420.
Last updated: Nov 22 2024 at 16:03 UTC