On 3 April 2014 20:54, YAMAMOTO Takashi <yamam...@valinux.co.jp> wrote:
> > On 3 April 2014 10:16, Joe Stringer <joestrin...@nicira.com> wrote:
> >
> >> I had noticed that this logic was in the codepath for when we delete a
> >> flow from the datapath, but not in this codepath. Thanks for picking
> this
> >> up!
> >>
> >
> > On second thought, I suppose that this means that we will expire the
> > netflow immediately when we discover that the flow is invalid, rather
> than
> > when we delete the flow. How does this impact netflow logging for traffic
> > on the flow between the time that the flow is revalidated up until the
> flow
> > is deleted?
>
> i'm not sure if i understand what you mean.
>
> the only caller of revalidate_ukey looks like the following.
>
> if (!revalidate_ukey(udpif, udump, ukey)) {
> dpif_flow_del(udpif->dpif, udump->key, udump->key_len, NULL);
> ukey_delete(revalidator, ukey);
> }
>
> ie. "when we discover that the flow is invalid" (revalidate_ukey returns
> false)
> and "when we delete the flow" (dpif_flow_del) are almost same.
>
> ideally we should use stats from dpif_flow_del, though.
> do you mean this?
Sorry, I was mistaken. I have been working on patches in this area and did
not look back at master to see how it currently works. I was referring to
push_dump_ops(), which is deliberately not used in this case.
I think that your patch is also applicable to branch-2.1.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
