On Feb 15, 2013, at 4:04 PM, Ben Pfaff <b...@nicira.com> wrote: > On Fri, Feb 15, 2013 at 02:00:59PM -0800, Ethan Jackson wrote: >>> I don't think so. If you look, the code which is assigning iter->odp_port >>> first >>> checks if node is NULL. >> >> Actually no I think he's right. In the condition where the node is in >> tmp_simap, we need to re-look it up after we've put it into >> tnl_backers. Otherwise later node will be non-null, but deleted, in >> which case we will dereference it incorrectly. > > That's what I meant, yes. > I see that now. I fixed this with the latest patch I just sent.
>>>> I think that ofproto destruction will no longer delete tunnel ports. >>>> I don't know whether that matters. >> >> I'm not completely sure what the correct behavior is here. If you >> have multiple ofprotos sharing a backer, and only one is deleted. I >> think our current behavior is correct because we'll end up garbage >> collecting the old backers. If we're deleting the last ofproto using >> a backer we'll end up leaving junk tnl backers around. We could >> special case, this, but I'm not sure if it matters. Thoughts? > > I don't think it matters. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
