Fix GCStress regression #119028
Fix GCStress regression #119028
Conversation
There was a problem hiding this comment.
Pull Request Overview
This PR fixes a GCStress regression where the original version of PInvoke stub code could not be found after changes made in PR #117901. The fix compensates for the MethodDesc adjustment done for PInvoke stubs during code allocation by ensuring GC instrumentation uses the correct code version that matches the CodeHeader.
- Adds logic to detect when a PInvoke stub's MethodDesc has been adjusted during code allocation
- Retrieves the actual MethodDesc from the CodeHeader and creates the appropriate NativeCodeVersion
- Ensures GC coverage setup uses the correct code version for proper instrumentation
github-actions
bot
added
the
needs-area-label
|
Context #118879 (comment) |
|
/azp run runtime-coreclr gcstress-extra, runtime-coreclr gcstress0x3-gcstress0xc |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr gcstress-extra, runtime-coreclr gcstress0x3-gcstress0xc |
jkotas
added
area-VM-coreclr
and removed
needs-area-label
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
Tagging subscribers to this area: @mangod9 |
GCStress is not able to find the original version of the code for PInvoke stubs after dotnet#117901. We need to compensate for the MethodDesc adjustment done for PInvoke stubs when storing the original version of the code during GC stress.
Interop marshalling of varargs needs MethodDesc calling convention to support ldftn <PInvoke method with varargs>. It is not possible to smugle the target MethodDesc* via vararg cookie in this case.
|
/azp run runtime-coreclr gcstress-extra, runtime-coreclr gcstress0x3-gcstress0xc |
|
Azure Pipelines successfully started running 2 pipeline(s). |
Co-authored-by: Aaron Robinson <[email protected]>
|
I have combined this PR with the change that makes GC stress faster in #119037 - to rule out timeouts. It looks like the combined change passes stress tests, except on (There is also one failure in |
Yes, there was a bug on arm32. It should be fixed by last commit. |
|
/azp run runtime-coreclr gcstress-extra, runtime-coreclr gcstress0x3-gcstress0xc |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr gcstress-extra, runtime-coreclr gcstress0x3-gcstress0xc |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr gcstress-extra, runtime-coreclr gcstress0x3-gcstress0xc |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
stress is passing on linux-arm 🎉 |
Co-authored-by: Aaron Robinson <[email protected]>
3 participants
GCStress is not able to find the original version of the code for PInvoke stubs after #117901. We need to compensate for the MethodDesc adjustment done for PInvoke stubs when storing the original version of the code during GC stress.