"dev" <dev-boun...@openvswitch.org> wrote on 08/06/2016 06:43:23 AM:
> From: nickcooper-zhangtonghao <nickcooper-zhangtong...@opencloud.tech>
> To: dev@openvswitch.org
> Cc: nickcooper-zhangtonghao <nickcooper-zhangtong...@opencloud.tech>
> Date: 08/06/2016 06:43 AM
> Subject: [ovs-dev] [PATCH] ovn-controller: Combine two Logical_Flow
> loops into one.
> Sent by: "dev" <dev-boun...@openvswitch.org>
>
> Signed-off-by: nickcooper-zhangtonghao <nickcooper-
> zhangtong...@opencloud.tech>
> ---
> ovn/controller/lflow.c | 14 +++++---------
> 1 file changed, 5 insertions(+), 9 deletions(-)
>
> diff --git a/ovn/controller/lflow.c b/ovn/controller/lflow.c
> index fda10eb..0cc92a7 100644
> --- a/ovn/controller/lflow.c
> +++ b/ovn/controller/lflow.c
> @@ -412,18 +412,14 @@ add_logical_flows(struct controller_ctx *ctx,
> const struct lport_index *lports,
> }
> full_logical_flow_processing = false;
> } else {
> - /* First, remove any flows that should be removed. */
> SBREC_LOGICAL_FLOW_FOR_EACH_TRACKED (lflow, ctx->ovnsb_idl) {
> + /* Remove any flows that should be removed. */
> if (sbrec_logical_flow_is_deleted(lflow)) {
> ofctrl_remove_flows(&lflow->header_.uuid);
> - }
> - }
> -
> - /* Now, add/modify existing flows. */
> - SBREC_LOGICAL_FLOW_FOR_EACH_TRACKED (lflow, ctx->ovnsb_idl) {
> - if (!sbrec_logical_flow_is_deleted(lflow)) {
> - /* if the logical flow is a modification, just remove
> - * the flows for this row, and then add new flows. */
> + } else {
> + /* Now, add/modify existing flows. If the logical
> + * flow is a modification, just remove the flows
> + * for this row, and then add new flows. */
> if (!sbrec_logical_flow_is_new(lflow)) {
> ofctrl_remove_flows(&lflow->header_.uuid);
> }
> --
I had to go back through the July archives to remember why we ended
up with two logical flow loops and this change was always on my
radar, but I just forgot about it while chasing other higher priority
items. So, I went ahead and ran through the ovn unit tests with
-j4 50 times to verify that the test failure rate I'm seeing is
reasonable (I didn't get one...), so
Acked-by: Ryan Moats <rmo...@us.ibm.com>
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
