When we are processing a bond member's protodown we get from
the dataplane, check to make sure we haven't already queued
up a set. If we have, it's likely this is just a notification
we get from the kernel after we set protodown and before we have
processed the result in our dplane pthread.
This change is needed now that we set protodown via the dplane
pthread.
Signed-off-by: Stephen Worley <sworley@nvidia.com>
zif->flags &= ~ZIF_FLAG_PROTODOWN;
if (zebra_evpn_is_es_bond_member(zif->ifp)) {
+ /* Check it's not already being sent to the dplane first */
+ if (protodown && (zif->flags & ZIF_FLAG_SET_PROTODOWN))
+ return;
+
+ if (!protodown && (zif->flags & ZIF_FLAG_UNSET_PROTODOWN))
+ return;
+
if (IS_ZEBRA_DEBUG_EVPN_MH_ES || IS_ZEBRA_DEBUG_KERNEL)
zlog_debug(
"bond mbr %s re-instate protdown %s in the dplane",