-
Notifications
You must be signed in to change notification settings - Fork 730
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.
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
JDK21 serviceability_jvmti_j9_0_FAILED serviceability/jvmti/RedefineClasses/RedefineRunningMethods.java Segmentation error vmState=0x00040000 #18260
Comments
Might be a GC issue @dmitripivkine can you please take a quick look |
The invocation of the GC is caused by #18236 but the crash is not caused by the change - it's in the GC verbose code. |
Subscribed to Verbose code has crashed an attempt to get |
|
|
@gacholio Would you please help to understand how acquireExclusiveVMAccess code did not set omrVM->exclusiveVMAccessStats.lastResponder? |
My understanding is in this case |
I bet we have safepoint exclusive, which doesn't update those stats. @dmitripivkine can you verify that in the core? I'll need to think about how to address this. Prior to #18236 this situation could not occur. |
Are you talking about these? Should I check something else? |
I have core uploaded to Bluebird |
The @dmitripivkine The GC should probably be updated to handle the NULL case. The field is NULL in this case because safepoint does not call the stat initializing code. I'll add that in right away as a workaround and continue to look into adding the proper tracking code to safepoint, but that will need to be done very carefully (as with anything dealing with exclusive). |
Set a valid default requestor and lastResponder in the exclusive stats when acquiring safepoint exclusive. Fixes: eclipse-openj9#18260 Signed-off-by: Graham Chapman <[email protected]>
@dmitripivkine What is the GC doing in this case for metronome? There weren't a mass of metronome failures when #18236 was merged, so it's presumably working. |
Looks like metronome uses the same code. So, I think it is harder to reproduce. |
This needs to be fixed in the GC (the fields can validly be NULL even before #18236) - the quick VM fix I had suggested won't work properly. |
Failure link
From an internal build(
rhel9-aarch64-1
):Rerun in Grinder - Change TARGET to run only the failed test targets.
Optional info
Failure output (captured from console output)
50x internal grinder - all failed
Also seen JDK21 aarch64_mac JDK21 ppc64_aix JDK21 ppc64le_linux
The text was updated successfully, but these errors were encountered: