[jdk22u] RFR: 8329528: G1 does not update TAMS correctly when dropping retained regions during Concurrent Start pause

Chris Hegarty chegar999 at gmail.com
Wed Apr 17 11:45:49 UTC 2024


Thanks for backporting this Thomas.

Given the severe nature of this bug and its impact, it renders JDK 22 
and JDK 22u1 effectively useless (or even worse dangerous) to many 
applications.

I would like to propose a 22.0.1.1 be considered to get this fix into 
releases prior to 22u2.

-Chris.

On 17/04/2024 10:59, Chris Hegarty wrote:
> On Tue, 16 Apr 2024 08:32:26 GMT, Thomas Schatzl <tschatzl at openjdk.org> wrote:
> 
>> Hi all,
>>
>>    please review this backport for https://bugs.openjdk.org/browse/JDK-8329528; it applies cleanly apart from the hunk now in `inline void HeapRegion::note_start_of_marking()` - in jdk23 this method moved from there to concurrent marking code, although the change itself is the same.
>>
>> This issue causes random crashes in G1. There is no workaround.
>>
>> The change baked in jdk23 for a week or so, with no issues. Risk is little at this point as the patch is small and well understood, and there is a regression test.
>>
>> Testing: gha, tier1-5
> 
> LGTM
> 
> -------------
> 
> Marked as reviewed by chegar (Reviewer).
> 
> PR Review: https://git.openjdk.org/jdk22u/pull/145#pullrequestreview-2005584588


More information about the jdk-updates-dev mailing list