Data race in EventLoop

[shmuelr]

Hey there,

Similar to #2660, our internal Java TSAN tests in Google picked up a race in EventLoop.common.kt

Write of size 4 at
kotlinx.coroutines.EventLoopImplBase$DelayedTask.setHeap(Lkotlinx/coroutines/internal/ThreadSafeHeap;)V EventLoop.common.kt:419
versus read of size 4 at
kotlinx.coroutines.EventLoopImplBase$DelayedTask.dispose()V EventLoop.common.kt:480

The race in EventLoop seems to be here
https://github.com/Kotlin/kotlinx.coroutines/blob/master/kotlinx-coroutines-core/common/src/EventLoop.common.kt#L417-L420
that write is not thread safe, but the dispose function is marked as synchronized
https://github.com/Kotlin/kotlinx.coroutines/blob/master/kotlinx-coroutines-core/common/src/EventLoop.common.kt#L479-L485
leading to the race in _heap reading & writing.

Is this race safe?

Thanks!

[qwwdfsad]

Could you please provide more detailed stacktraces? E.g. on which code path exactly setHeap is racing

[qwwdfsad]

Oh, I see, both dispose and setHeap are synchronized, but they actually use different monitors.

This is definitely a bug, I already can see an interleaving that leads to an error, working on the fix.
I must admit that TSAN testing is amazingly valuable as such bugs are really hard to catch with non-targeting stress test or code-review

[shmuelr]

Nice! Thanks for working on a fix, I'm happy we were able to help here :D

[LifeIsStrange]

@shmuelr Hi :) What is this Java/kotlin TSAN you're talking about??
I have seen the openjdk TSAN JEP here:
https://bugs.openjdk.java.net/browse/JDK-8208520
https://openjdk.java.net/projects/tsan/

However the implementation still seems to be a work in progress and has not received updates since 2019? :/
How to use it, does the flag work on default openjdk or do I need a special OpenJDK build? Is there kotlin support out of the box?
Or do you have a custom solution at google?
I'm looking forward to your answer as the OpenJDK ecosystem would strongly benefit from a TSAN tool.