RFR: 8319137: release _object in ObjectMonitor dtor to avoid races [v3]

[Daniel D. Daugherty]

On Mon, 20 Nov 2023 17:37:15 GMT, Patricio Chilano Mateo <pchilanomate at openjdk.org> wrote:

>> Please review the following fix which moves the release of the _object WeakHandle back to the ObjectMonitor dtor. Releasing it earlier as we do today introduces a race which can lead to crashes. Tested running tiers123 in mach5.
>> 
>> This fix is an effective backout of [JDK-8256302](https://bugs.openjdk.org/browse/JDK-8256302).
>> 
>> Thanks,
>> Patricio
>
> Patricio Chilano Mateo has updated the pull request incrementally with one additional commit since the last revision:
> 
>   remove redundant branches

This is the equivalent of a [BACKOUT] of [JDK-8256302](https://bugs.openjdk.org/browse/JDK-8256302)
with a couple of adaptations in keeping with the current state of the code:
- the static `delete_monitors()` function is kept
- only the MonitorDeflationThread deflates ObjectMonitors any more so some logic that predates JDK-8256302 is no longer needed

I recommend redoing the Tier[1-3] round of testing. I don't
think you need to do Tier[4-6] testing with this [BACKOUT].

JDK-8256302 was integrated in jdk-21+26 (2023.06.07).
JDK-8319137 was found in jdk-22+21 stress testing (2023.10.28)
almost 5 months later so shaking out bugs in this area of code is
difficult.

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

Marked as reviewed by dcubed (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/16738#pullrequestreview-1740400310