On Fri, Aug 16, 2019 at 5:10 PM Darrell Ball <dlu...@gmail.com> wrote:
>
> Thanks for the patch
>
> Pls let me know if the following incremental works for you.
>
> diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c
> index 244155a..cb8b51e 100644
> --- a/ofproto/ofproto-dpif.c
> +++ b/ofproto/ofproto-dpif.c
> @@ -168,6 +168,12 @@ struct ct_timeout_policy {
> * "ct_tp_kill_list" list. */
> };
>
> +/* Periodically try to purge deleted timeout policies from the datapath.
> Retry
> + * may be necessary if the kernel datapath has a non-zero datapath flow
> + * reference count for the timeout policy. */
> +#define TIMEOUT_POLICY_CLEANUP_INTERVAL (300000) /* 5 minutes. */
> +static long long int timeout_policy_cleanup_timer;
Thanks for adding the clean up debounce and makes the comment clearer.
I will fold in your diff and the following minor change in the next version.
#define TIMEOUT_POLICY_CLEANUP_INTERVAL (20000) /* 20 seconds. */
static long long int timeout_policy_cleanup_timer = LLONG_MIN;
I changed the interval to be two times of the revlidataion cycle
because we should be able to remove the unused timeout policies in the
kernel datapath after the next flow revalidation cycle.
Thanks,
-Yi-Hung
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
