On Tue, Jun 7, 2016 at 11:52 AM, Ryan Moats <rmo...@us.ibm.com> wrote:
> As a side effect, tunnel context is persisted.
>
> Signed-off-by: Ryan Moats <rmo...@us.ibm.com>

When I was working on something else in this area, I noticed what
appear to be some regressions as a result of this change.

The one that I actually noticed from testing is that updates to the
encap table aren't propagated in most cases. That's because we loop
through the chassis table using SBREC_CHASSIS_FOR_EACH_TRACKED but
while encaps are linked to the chassis table, they are separate. As a
result, a change in an existing encap won't have an effect.

While looking at the code, it also seemed to me that while we will
notice if a port is added to the OVS bridge we would notice and do
process_full_encaps, the same isn't true is a port is deleted. It also
seemed to me that the port_name created in check_and_update_tunnel()
isn't used and as a result leaked.

Would you mind taking a look and see if I'm missing something?
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Reply via email to