On Tue Sep 8, 2026 at 5:19 PM CEST, Dumitru Ceara wrote:
> Commit 1aa8852b30f5 ("ofctrl: Don't defer flow update acknowledgement.")
> addressed the indefinitely long potential OVS flow update
> acknowledgement delay in ovn-controller's ofctrl implementation. We
> have the same behavior in ovn-br-controller as its implementation was
> based on that of ovn-controller, so we have the same bug.
>
> Fix it in the same way: when the xid of the _first_ instance of req_cfg
> has been acked, consider req_cfg relevant flows installed.
>
> Signed-off-by: Dumitru Ceara <[email protected]>
Hi Dumitru,
looks good to me. Thanks!
Acked-by: Felix Moebius <[email protected]>
> ---
> br-controller/br-ofctrl.c | 11 -----------
> 1 file changed, 11 deletions(-)
>
> diff --git a/br-controller/br-ofctrl.c b/br-controller/br-ofctrl.c
> index 4f79ba908d..a068736114 100644
> --- a/br-controller/br-ofctrl.c
> +++ b/br-controller/br-ofctrl.c
> @@ -699,17 +699,6 @@ br_ofctrl_put(struct br_ofctrl *br_ofctrl, uint64_t
> req_cfg,
> fup->req_cfg, req_cfg);
> ovs_list_remove(&fup->list_node);
> free(fup);
> - } else if (req_cfg == fup->req_cfg) {
> - /* This br_ofctrl_flow_update is for the same configuration
> as
> - * 'req_cfg'. Probably, some change to the physical topology
> - * means that we had to revise the OpenFlow flow table even
> - * though the logical topology did not change. Update
> fp->xid,
> - * so that we don't send a notification that we're up-to-date
> - * until we're really caught up. */
> - VLOG_DBG("advanced xid target for req_cfg=%"PRId64, req_cfg);
> - fup->xid = xid_;
> -
> - return;
> } else {
> break;
> }
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
