On Feb 15, 2013, at 3:51 PM, Ben Pfaff <b...@nicira.com> wrote: > On Fri, Feb 15, 2013 at 12:49:47PM -0800, Ethan Jackson wrote: >> I think the main loop of this version still has some bugs. It doesn't >> properly >> update 'iter''s odp_port, nor do a tnl_port_reconfigure() when 'iter''s >> backer >> changes (in some cases). What about something like the following? I think >> it's a little bit more straight forward. I haven't tested it at all though. > > The name tmp_simap is not good. Perhaps tmp_backers or old_backers > or…? > I will change this in the patch.
> I believe that your code is doing simap_delete() on 'node' then > dereferencing that same 'node' a few lines later. > I don't think so. If you look, the code which is assigning iter->odp_port first checks if node is NULL. > I think that ofproto destruction will no longer delete tunnel ports. > I don't know whether that matters. > I think the idea is for them to be deleted in type_run() instead. > Thanks, > One other thing I found with this code is that odp_port was not initialized to UINT32_MAX. I just fixed that and am testing it again. Thanks, Kyle > Ben. > _______________________________________________ > dev mailing list > dev@openvswitch.org > http://openvswitch.org/mailman/listinfo/dev _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
