On Thu, Dec 27, 2012 at 02:46:04PM -0500, Jesse Gross wrote: > On Thu, Dec 27, 2012 at 2:36 PM, Ben Pfaff <b...@nicira.com> wrote: > > On Thu, Dec 27, 2012 at 12:37:27PM -0500, Jesse Gross wrote: > >> On Wed, Dec 26, 2012 at 8:16 PM, Ethan Jackson <et...@nicira.com> wrote: > >> > Theoretically, it's possible for netdev_get_status() to be called > >> > on a netdev-vport which hasn't had it's configuration set yet. In > >> > this case, netdev-vport would dereference a null pointer. This > >> > problem was found by Jesse Gross <je...@nicira.com> in review. > >> > > >> > Signed-off-by: Ethan Jackson <et...@nicira.com> > >> > >> Shouldn't we just fetch the config? It's pretty easy and seems a lot > >> less likely to cause surprises in the future. > > > > As long as we're talking about theoretical situations, it's possible > > that the vport doesn't exist in the kernel yet. We let vports be > > created and configured in userspace before they get added to a > > datapath. In between, they're just userspace configuration data. > > That's true and in that case if we don't have a config then this patch > clearly does the right thing. However, we also use the netdev library > to also pull the current config from the kernel. If we used this for > something like ovs-dpctl (which we don't) then it would give the wrong > result. > > I think making it fetch the config is the right thing to do, although > since the code is going away and it's not a current problem it may not > be worth it (although I think the necessary refactoring is minimal).
I don't object to either solution. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
