On 5/13/26 2:28 PM, Ilya Maximets wrote:
> On 5/11/26 5:54 PM, Ilya Maximets wrote:
>> This set addresses some annoying and fairly misleading error logs and
>> attempts to make the code a little more clear.  The races are getting
>> hit in real ovn-kubernetes deployments as well as sometimes in OVN
>> tests in GitHub Actions CI.
>>
>> Ilya Maximets (3):
>>   ofproto-dpif: Avoid race between recirc id free and the leak check.
>>   ofproto-dpif: Rename recirc_free_ofproto to better match the code.
>>   ofproto-dpif: Remove unused rule->recirc_id.
>>
>>  ofproto/ofproto-dpif-rid.c |  8 +++++---
>>  ofproto/ofproto-dpif-rid.h |  3 ++-
>>  ofproto/ofproto-dpif.c     | 32 +-------------------------------
>>  ofproto/ofproto-dpif.h     |  8 --------
>>  4 files changed, 8 insertions(+), 43 deletions(-)
>>
> 
> Thanks, Eeclo, for review!
> 

Hi Eelco, Ilya,

> Since we do actually have a reproducer, there is no need to wait for
> OVN folks to try this (OVN tests are not reproducing reliably anyway).
> So, I applied the set and backported down to 3.3.  The backport of
> the last two patches wasn't strictly necessary, but I took them along
> since I was backporting the first one anyway.
> 

Sorry for the delay.  I did run this in my fork with tests that I knew
were failing quite often and it looked green in 9 runs:

https://github.com/dceara/ovn/actions/runs/25803721553
https://github.com/dceara/ovn/actions/runs/25803729053
https://github.com/dceara/ovn/actions/runs/25803734575

Thanks for fixing this!  Ilya, would you have time to post OVS submodule
bumps for OVN?  Otherwise I can look into it next week.

Thanks,
Dumitru

> Best regards, Ilya Maximets.
> 

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to