On Tue, May 20, 2014 at 03:55:50PM -0700, Ben Pfaff wrote: > On Tue, May 20, 2014 at 03:53:26PM -0700, Ben Pfaff wrote: > > I suspect that unconditionally taking the write-lock in > > update_mcast_snooping_table() is going to serialize flow translation, > > in practice. That's why update_learning_table() checks whether any > > change is needed under the read-lock first, then grabs the write-lock > > if it's really needed. But in the long run RCU is better (and we > > should probably adapt mac-learning to use it, or we could switch to a > > fatlock there as an easy first step). > > Looking at the next patch, I guess it only matters for multicast > packets. Maybe not worth bothering.
The problem is that if someone configures a port to be flooded while that is running, we could have duplicated packets (normal action + flood port). Yes, it's specific for multicast packets and I have an idea to improve this for v2. Flavio _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
