[jdk21u-dev] RFR: 8319137: release _object in ObjectMonitor dtor to avoid races
Roman Kennke
rkennke at openjdk.org
Tue Dec 19 17:28:54 UTC 2023
On Wed, 13 Dec 2023 11:02:46 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:
> Transplanted from https://github.com/openjdk/jdk21u/pull/401
>
> This effectively reverts [JDK-8256302](https://bugs.openjdk.org/browse/JDK-8256302), which introduced race condition in JDK 21. The patch does not apply cleanly for two reasons:
> 1. Deleted `release_object()` has a different form in JDK 21 due to slight WeakHandle API change. This is not relevant for this patch, just a backporting conflict. I removed `release_object()` by hand.
> 2. Without [JDK-8319896](https://bugs.openjdk.org/browse/JDK-8319896) we cannot do `JavaThread::cast` in the incoming changeset, because we can be called by a non-Java thread. I suspect JDK-8319896 would come with its own bugtail, so I would prefer not to backport it now. I have replaced the code back to original pre- JDK 21 form in a separate commit: [Modifications to fit JDK 21](https://github.com/openjdk/jdk21u-dev/commit/65251865632a5624fec70527b7bc1f4dd1d0d637)
>
> Additional testing:
> - [x] MacOS AArch64 server fastdebug, `runtime/Monitor` passes
> - [x] Linux AArch64 server fastdebug, `tier1 tier2 tier3 tier4`
> - [x] Linux x86_64 server fastdebug, `tier1 tier2 tier3 tier4`
Looks correct to me. Thanks!
-------------
Marked as reviewed by rkennke (Reviewer).
PR Review: https://git.openjdk.org/jdk21u-dev/pull/12#pullrequestreview-1789384411
More information about the jdk-updates-dev
mailing list