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

Implement stack walking without AsyncGetCallTrace #66

Open
apangin opened this Issue Nov 2, 2017 · 2 comments

Comments

Projects
None yet
3 participants
@apangin
Member

apangin commented Nov 2, 2017

There is a number of problems with AsyncGetCallTrace API.

  • It is not accurate: sometimes it can point to neighbour methods. See #51 and JDK-8022893.

  • It incorrectly handles certain corner cases. See JDK-8178287.

  • It fails to walk through VM frames in Oracle JDK 9. See #60.

  • It requires preloading of all jmethodIDs thus resulting in startup time overhead. See #14 (comment).

  • It does not distinguish between interpreted, compiled and inlined frames.

  • In rare cases it may crash JVM. See #73.

  • -XX:MaxJavaStackTraceDepth unobviously affects collected stack traces. See #89.

  • Unloaded methods appears in the profile as jvmtiError 23.

  • = has workaround in async-profiler

The idea is to implement Java stack walking on our own without relying on AGCT. Since the agent runs in the context of JVM, it can access VM structures, especially those exported through VMStructs. It should be possible to replicate stack walking logic of the VM inside async-profiler, though it might be challenging. The main risk is that differrent versions of JVM may have different stack layout, but VMStructs along with special handling of the known versions is likely to help.

@toaler

This comment has been minimized.

toaler commented Jun 11, 2018

Hi @apangin any thoughts on when the features required to close this issue would be delivered? How potentially contributors help with the open work?

@apangin

This comment has been minimized.

Member

apangin commented Jun 12, 2018

@toaler Not sure, but unlikely in the nearest release. Are you interested in some particular problem of AGCT?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment