On 11/5/26 6:54 AM, Bekei Park wrote:
> The conntrack flush of a removed ECMP next-hop is queued in front of
> the flow bundle that takes the next-hop out of the flows. A packet that
> the switch handles in between finds no conntrack entry and goes through
> the old flows, which can still pick the removed next-hop and commit a
> new entry with its MAC in ct_label. Nothing flushes that entry later.
>
> Queue the flush after the bundle instead, next to the LB 5-tuple
> flushes. The test now checks that the flush follows the bundle commit.
>
> CC: Lorenzo Bianconi <[email protected]>
> CC: Dumitru Ceara <[email protected]>
> Fixes: b4cacd5c2142 ("ofctrl: Introduce ecmp_nexthop_monitor.")
> Assisted-by: Claude Opus 5.5, Claude Code
> Signed-off-by: Bekei Park <[email protected]>
> ---
Hi Bekei Park,
This patch looks good to me but it depends on 1/2 and I requested some
changes on that one.
Looking forward to v3!
Regards,
Dumitru
> v2: No change.
>
> controller/ofctrl.c | 15 +++++++++------
> tests/ovn.at | 6 +++++-
> 2 files changed, 14 insertions(+), 7 deletions(-)
>
> diff --git a/controller/ofctrl.c b/controller/ofctrl.c
> index 12da7f7..194a0c0 100644
> --- a/controller/ofctrl.c
> +++ b/controller/ofctrl.c
> @@ -2821,12 +2821,6 @@ ofctrl_put(struct ovn_desired_flow_table *lflow_table,
> /* OpenFlow messages to send to the switch to bring it up-to-date. */
> struct ovs_list msgs = OVS_LIST_INITIALIZER(&msgs);
>
> - /* Removed ECMP next-hops stay pending while the connection is down,
> - * because rconn drops the messages queued then. */
> - if (rconn_is_connected(swconn)) {
> - ecmp_nexthop_monitor_flush(swconn, &msgs);
> - }
> -
> /* Iterate through ct zones that need to be flushed. */
> struct shash_node *iter;
> SHASH_FOR_EACH(iter, pending_ct_zones) {
> @@ -3024,6 +3018,15 @@ ofctrl_put(struct ovn_desired_flow_table *lflow_table,
> }
> }
>
> + /* Flush the conntrack entries of removed ECMP next-hops after the flow
> + * bundle, so that a packet in between cannot commit a new entry towards
> + * a removed next-hop through the old flows. Removed next-hops stay
> + * pending while the connection is down, because rconn drops the
> + * messages queued then. */
> + if (rconn_is_connected(swconn)) {
> + ecmp_nexthop_monitor_flush(swconn, &msgs);
> + }
> +
> acl_ids_flush_expired(tracked_acl_ids, rconn_get_version(swconn), &msgs);
>
> if (!ovs_list_is_empty(&msgs)) {
> diff --git a/tests/ovn.at b/tests/ovn.at
> index 2ba8514..8f544a9 100644
> --- a/tests/ovn.at
> +++ b/tests/ovn.at
> @@ -48718,7 +48718,7 @@ check test -n "$zone"
> flush_10="NXT_CT_FLUSH .* zone=$zone labels=0xf0000000101000000000/"
> flush_20="NXT_CT_FLUSH .* zone=$zone labels=0xf0000000202000000000/"
>
> -AS_BOX([Removed next-hop is flushed])
> +AS_BOX([Removed next-hop is flushed after the flow bundle])
> n=$(wc -l < hv1/ovs-vswitchd.log)
> check ovn-nbctl --wait=hv lr-route-del lr0 1.0.0.0/24 192.168.10.20
> tail -n +$((n + 1)) hv1/ovs-vswitchd.log > vswitchd-del.log
> @@ -48726,6 +48726,10 @@ AT_CHECK([grep -c "$flush_20" vswitchd-del.log],
> [0], [1
> ])
> AT_CHECK([grep -c "$flush_10" vswitchd-del.log], [1], [0
> ])
> +commit=$(grep -n "type=COMMIT_REQUEST" vswitchd-del.log | head -1 | cut -d:
> -f1)
> +flush=$(grep -n "$flush_20" vswitchd-del.log | cut -d: -f1)
> +check test -n "$commit"
> +check test "$commit" -lt "$flush"
>
> AS_BOX([Removed next-hop is flushed when ofctrl could not send at removal])
> check ovn-nbctl --ecmp-symmetric-reply lr-route-add lr0 1.0.0.0/24
> 192.168.10.20
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev