Re: [PATCH] Fix ProcKill lock-group vs procLatch recycle race

Vlad Lesin <vladlesin@gmail.com>

From: Vlad Lesin <vladlesin@gmail.com>

To: Michael Paquier <michael@paquier.xyz>

Cc: Andrey Borodin <x4mmm@yandex-team.ru>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>

Date: 2026-05-18T10:11:56Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Fix procLatch ownership race in ProcKill()
- 84b9d6bceab6 19 (unreleased) cited

Attachments

vVL3-0001-add-prockill-lockgroup-regression-test.patch (text/x-patch) patch 0001
vVL3-0002-fix-lockgroup-double-push-and-leak.patch (text/x-patch) patch 0002
vVL3-0003-fix-prockill-lockgroup-procLatch-recycle-race.patch (text/x-patch) patch 0003

On 5/18/26 06:15, Michael Paquier wrote:
>> Additionally, I noticed that if several processes are waiting
>> on the same injection point, only one of them will be awakened by a single
>> injection_points_wakeup() call. I am not sure if this behavior is
>> intentional; please let me know.
> 
> Yep, I recall that as being intentional, hence I don't feel that 0002
> is a good thing to do, even worse doing so in the back-branches.
Ok, I fixed the test to use distinct injection points for the lock group 
leader and follower.

-- 
Best regards,
Vlad