On 5/20/26 9:31 PM, Mark Michelson wrote: > Hi Dumitru, thanks for the patch. >
Hi Mark, Thanks for the review! > Acked-by: Mark Michelson <[email protected]> > Applied to main and all stable branches down to 25.03. I didn't go further because on 24.09 and 24.03 we'd need other patches backported and that's not straightforward. We can revisit that in the future if needed I guess. I did have to make a small change to the test though as it was sometimes unstable in CI. I hope that's OK. diff --git a/tests/ovn-controller.at b/tests/ovn-controller.at index d4d666b699..fd63399f17 100644 --- a/tests/ovn-controller.at +++ b/tests/ovn-controller.at @@ -1096,14 +1096,10 @@ check ovn-nbctl --wait=hv sync dnl Re-add the localnet port. The patch port already exists, dnl so non_vif_data should not change, and pflow_output should dnl be handled incrementally. -check as hv2 ovn-appctl -t ovn-controller inc-engine/clear-stats check ovn-nbctl --wait=hv \ -- lsp-add-localnet-port ls-underlay ln-underlay phys \ -- set logical_switch_port ln-underlay tag_request=1000 -dnl Verify pflow_output was NOT recomputed. -check_controller_engine_stats hv2 pflow_output norecompute compute - dnl Verify the CR bridged redirect flow is back. OVS_WAIT_UNTIL([as hv2 ovs-ofctl dump-flows br-int table=OFTABLE_LOCAL_OUTPUT | grep -q "reg15=0x${cr_key},metadata=0x${router_dp_key}"]) > With regards to the CI failure, I posted patches for bumping the OVS > submodule: > > https://patchwork.ozlabs.org/project/ovn/patch/[email protected]/ > https://patchwork.ozlabs.org/project/ovn/patch/[email protected]/ > https://patchwork.ozlabs.org/project/ovn/patch/[email protected]/ > https://patchwork.ozlabs.org/project/ovn/patch/[email protected]/ > > Hopefully those will help with CI stability. > Unfortunately, as discussed on the 25.09 bump patch, these introduce other issues that need to be fixed in ovs: https://mail.openvswitch.org/pipermail/ovs-dev/2026-May/432516.html > On Wed, May 20, 2026 at 9:36 AM Dumitru Ceara via dev > <[email protected]> wrote: >> >> On 5/20/26 12:25 PM, Dumitru Ceara via dev wrote: >>> physical_handle_flows_for_lport() handles the forward >>> dependency (PATCH/EXTERNAL/L3GATEWAY -> localnet) but >>> not the reverse: when a localnet port is added or >>> updated, chassisredirect ports on peer router datapaths >>> whose bridged redirect flows depend on that localnet >>> port (via put_remote_port_redirect_bridged() -> >>> get_localnet_port()) are not re-evaluated. >>> >>> This causes the CR bridged redirect flow in >>> OFTABLE_LOCAL_OUTPUT to be permanently missing when the >>> localnet port binding is processed incrementally without >>> a full recompute. >>> >>> Fix this by iterating peer router datapaths when a >>> localnet port is processed and re-evaluating any CR port >>> bindings found there. >>> >>> Add a test that deterministically exposes the bug by >>> pre-creating the OVS patch port so that non_vif_data >>> does not change when the localnet port binding arrives, >>> forcing pflow_output to use the incremental path. >>> >>> Fixes: 3ae8470edc64 ("I-P: Handle runtime data changes for pflow_output >>> engine.") >>> Assisted-by: Claude Opus 4.6, Claude Code >>> Signed-off-by: Dumitru Ceara <[email protected]> >>> --- >>> NOTE: >>> - on 25.03 this patch would need some minor changes: >>> https://github.com/dceara/ovn/commit/54089ca >>> - that would allow us to also backport b408eedf6d9d >>> ("ovn-controller: Skip type-update check for new port bindings.") >>> to 25.03 >>> - more context here: >>> https://mail.openvswitch.org/pipermail/ovs-dev/2026-May/432414.html >>> --- >> >> Known CI failure, we need an OVS submodule bump to pick up the OVS fixes >> for this. >> >> --- /dev/null 2026-05-20 10:47:47.504261528 +0000 >> +++ /workspace/ovn-tmp/tests/testsuite.dir/at-groups/154/stdout >> 2026-05-20 10:57:31.808115116 +0000 >> @@ -0,0 +1,2 @@ >> +2026-05-20T10:57:31.587Z|00483|ofproto_dpif_rid|ERR|recirc_id 2 left >> allocated when ofproto (br-int) is destructed >> +2026-05-20T10:57:31.587Z|00484|ofproto_dpif_rid|ERR|recirc_id 4 left >> allocated when ofproto (br-int) is destructed >> related-ports-diff: >> 154. ovn.at:10203: 154. ARP/ND from localnet -- proxy reply on resident >> chassis only -- parallelization=yes -- ovn_monitor_all=yes >> (ovn.at:10203): FAILED (ovn.at:10203) >> >> Recheck-request: github-robot-_Build_and_Test >> >> _______________________________________________ >> dev mailing list >> [email protected] >> https://mail.openvswitch.org/mailman/listinfo/ovs-dev >> > Regards, Dumitru _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
