On 11/13/26 11:16 AM, Ales Musil via dev wrote:
> Normal traffic finishes router ingress before tunneling to a remote
> transit port and does not execute its egress pipeline locally.  Omit
> ordinary remote-inport and remote egress flows, while retaining disabled-
> inport rejection before pruning and shared tunneled-reply learning.
> 
> Keep address-global router handling, DHCP responses, peer-datapath
> neighbor resolution, routes and policies selecting remote outputs, and
> ECMP stateful egress commits.  Those conntrack commits and remote-output
> MTU checks execute in ingress before physical tunneling.  Remote network
> ID assignments are reachable but redundant: their only consumer is
> logical egress SNAT and the ID is not preserved across the tunnel.
> 
> Require a remote binding type as well as a remote chassis when
> classifying transit ports.  Northd can retain the old remote SB chassis
> on a patch binding after requested-chassis changes; that must not
> suppress restored ordinary port flows.
> 
> Reported-at: https://redhat.atlassian.net/browse/FDP-4694
> Assisted-by: GPT-6.1 Sol, OpenCode
> Signed-off-by: Ales Musil <[email protected]>
> ---

Hi Ales,

Thanks for the patch, looks good to me!

Applied to main, 26.09 and 26.03.

Regards,
Dumitru

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

Reply via email to