On 8 July 2016 at 10:51, Ryan Moats <rmo...@us.ibm.com> wrote:
> "dev" <dev-boun...@openvswitch.org> wrote on 07/08/2016 02:38:06 AM:
>
> > From: Gurucharan Shetty <g...@ovn.org>
> > To: dev@openvswitch.org
> > Date: 07/08/2016 12:36 PM
> > Subject: [ovs-dev] [PATCH 2/2] ovn-controller: Change strategy for
> > gateway conntrack zone allocation.
> > Sent by: "dev" <dev-boun...@openvswitch.org>
> >
> > Commit 263064aeaa31e7 (Convert binding_run to incremental processing.)
> > changed the way patched_datapaths were handled. Previously we would
> > destroy the datastructure in every run and re-create it fresh. The new
> > way causes problems with the way conntrack zones are allocated as now
> > we can have stale port_binding entries causing segmentation faults.
> >
> > With this commit, we simply don't depend on port_binding records in
> > conntrack zone allocation and instead store the UUID as a string in
> > the patch_datapath datastructure.
> >
> > Fixes: 263064aeaa31e7 ("Convert binding_run to incremental processing.")
> > Signed-off-by: Gurucharan Shetty <g...@ovn.org>
> > ---
>
> I like what this is doing, but I'd like it more if it had a test case
> that would fail without this patch, but pass with it, so that we don't
> regress...
>
Ryan,
With the above referenced patch that this issue fixes, stale patched
datapaths were simply not removed. And none of the OVN tests caught it. We
can't catch this unless we start adding negative tests (for e.g: delete
records and look to see if ovn-controller deleted patch ports in br-int).
Or do you have something else in mind?
>
> Ryan
>
> _______________________________________________
> dev mailing list
> dev@openvswitch.org
> http://openvswitch.org/mailman/listinfo/dev
>
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
