On Wed, Mar 12, 2014 at 09:18:31AM -0800, Alex Wang wrote: > > > > As written, flow_hash_5tuple() will incorporate ICMP type and code > > > > into the hash (because those are stored into tp_src and tp_dst). Is > > > > that desirable? > > > > > > > > > > > > > I'm okay with that. I think counting ICMP type and code in hash will not > > > cause > > > particular handler receiving unfairly more ICMP related upcalls. (Is > > this > > > what you > > > concerned?) > > > > I am not concerned about fairness in this case. I am wondering whether > > one should ensure ordered delivery of ICMP with different type and > > code. My guess is that it does not matter, so let's leave this as-is > > unless anyone speaks up. > > This will require the datapath to distinguish such special packets > and send them to the same PID.
I guess I wasn't clear? I'm saying that you can keep it the way you had it. But it would be simple to cause them to be ordered: in the case where nw_proto == IPPROTO_ICMP, don't include tp_src or tp_dst in the hash. > Also, can you give me an example that we will have a userspace > action to send packet like ICMP to userspace? I can only think of > sampling. And I cannot think of other example where the reordering > would happen. Any flow that sends packets to the controller will send them to userspace. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
