On 5/21/26 12:33 AM, Ilya Maximets wrote:
> On 5/21/26 12:23 AM, Ilya Maximets wrote:
>> On 5/21/26 12:05 AM, Dumitru Ceara wrote:
>>> On 5/20/26 8:59 PM, Mark Michelson via dev wrote:
>>>> The following OVS commit is included in this update since it helps to
>>>> eliminate intermittent CI failures in OVN:
>>>>
>>>> 566b3ea85add ("ofproto-dpif: Avoid race between recirc id free and the
>>>> leak check.")
>>>>
>>>> However, instead of just updating the submodule to that commit, we are
>>>> including up to
>>>>
>>>> 929c3e98ce31 ("ofproto-dpif: Remove unused rule->recirc_id.")
>>>>
>>>> since it is the final patch in a series intended to fix a particular
>>>> defect in OVS.
>>>>
>>>> Signed-off-by: Mark Michelson <[email protected]>
>>>> ---
>>>> ovs | 2 +-
>>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/ovs b/ovs
>>>> index da6b84d69..929c3e98c 160000
>>>> --- a/ovs
>>>> +++ b/ovs
>>>> @@ -1 +1 @@
>>>> -Subproject commit da6b84d691593e57951bcc5f2f916336c6e0d5cc
>>>> +Subproject commit 929c3e98ce3102bc0e5fe8594e26228f29fb9c20
>>>
>>> Hi Mark,
>>>
>>> Thanks for this patch (and the other bumps)!
>>>
>>> We do have some very red CI results with this 25.09 OVS bump:
>>> https://github.com/ovsrobot/ovn/actions/runs/26185033314
>>>
>>> e.g.:
>>>
>>> https://github.com/ovsrobot/ovn/actions/runs/26185033314/job/77038788364#step:12:5248
>>>
>>> in the "ovn.at:8241: testing spine-leaf: 3 HVs, 3 LSs, connected via
>>> distributed spine switch" test:
>>>
>>> ./ovn.at:8241: as hv1 ovs-appctl ofproto/trace --names \
>>> br-int in_port=vif1 $packet > ofproto-trace-1
>>> ./ovn.at:8241: grep 'Megaflow:' ofproto-trace-1
>>> ./ovn.at:8241: cat ofproto-trace-1 | tail -1 \
>>> | grep -oE 'tnl_push|userspace|clone|br-phys_n1|vif[0-9]'
>>> --- - 2026-05-20 20:14:59.595640708 +0000
>>> +++ /workspace/ovn-tmp/tests/testsuite.dir/at-groups/108/stdout
>>> 2026-05-20 20:14:59.591340802 +0000
>>> @@ -1,9 +1,12 @@
>>> userspace
>>> +clone
>>> userspace
>>> clone
>>> tnl_push
>>> br-phys_n1
>>> +clone
>>> userspace
>>> +clone
>>> tnl_push
>>> br-phys_n1
>>>
>>> CC-ing Ilya too in case he has ideas.
>>
>> Likely coming from this:
>>
>> https://patchwork.ozlabs.org/project/openvswitch/patch/[email protected]/
>
> We should fix that commit so it doesn't emit clones when cl_clear is
> not emitted in the datapath. So maybe wait a little with the bump...
>
Thanks, Ilya, for looking into this! I'll move the submodule bump
patches to "Changes Requested" while we wait for an OVS fix.
Regards,
Dumitru
>>
>> I assume there are some ct_clear actions in there.
>>
>> Best regards, Ilya Maximets.
>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
