On 10/6/26 2:31 PM, Aaron Conole wrote: > Ilya Maximets via dev <[email protected]> writes: > >> Ports that do not participate in STP or RSTP while they are enabled >> on a bridge can easily become not floodable, because the bundle update >> checks the STP/RSTP config on the bridge and not on the port and these >> ports have their default forwarding disabled state. So, every time >> a bundle update runs, all bundles that do not participate in STP/RSTP >> are marked as not floodable, this means all the internal ports, >> including the bridge port, become non-floodable, as they do not >> participate by design. >> >> The issue was introduced quite some time ago, but became much easier >> to hit since the fix that made bundle update run on every >> reconfiguration to make sure the floodable flags are updated. >> >> Reported-at: >> https://mail.openvswitch.org/pipermail/ovs-discuss/2026-September/054022.html >> Reported-by: Svenne Krap <[email protected]> >> Fixes: 4b5f19962adc ("stp,rstp: disable learning and forwarding in STP/RSTP >> disabled state.") >> Fixes: 332ca1dad732 ("ofproto-dpif: Fix bundle floodable flag when disabling >> STP/RSTP.") >> Signed-off-by: Ilya Maximets <[email protected]> >> --- > > Surprised I didn't hit it in my own testing, but I guess I would have > been mostly focusing on ports which are participating in R/STP > > Reviewed-by: Aaron Conole <[email protected]>
Thanks, Aaron! Applied and backported down to 3.3. Best regards, Ilya Maximets. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
