[jdk25u-dev] RFR: 8372835: WorkQueue::push is missing an acquire-fence

[Aleksey Shipilev]

On Tue, 13 Jan 2026 11:25:48 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> This one is a little weird process-wise. There is `25.0.3-oracle` fix for it already, but no mainline fix (yet). Mainline would get it via [JDK-8375130](https://bugs.openjdk.org/browse/JDK-8375130). I think it is easier to "backport" this change instead of partially pick up JDK-8375130. This also addresses parity more thoroughly. There is no significant correctness risk here, as we "only" add a new barrier. Might have performance implications, but correctness trumps it.
> 
> Additional testing:
>  - [x] Linux x86_64 server fastdebug, `java/util/concurrent`
>  - [x] Linux x86_64 server fastdebug, `tier1`

Hey @GoeLin, I am going to do it this way, see PR description. Any reason why we should do something else?

-------------

PR Comment: https://git.openjdk.org/jdk25u-dev/pull/141#issuecomment-3744094405