This looks good to me. I have three minor comments and then it will be ready.
1) It doesn't seem to apply to master for me. Would you please rebase it and resend it? 2) Why are we using 2000 milliseconds as the stat collection time? In ofproto-dpif we're guaranteed to pull stands at least once every 1000ms. 3) I think we can ditch the decay_rx_count variable and increment bfd->rx_packets directly. Perhaps I'm missing something though, in which case what's there is fine. 4) The comment above bfd_try_decay is more about how the function works than what it does. For this reason I think it should be within the function body not outside. Thanks On Fri, Aug 16, 2013 at 5:57 AM, Ethan Jackson <et...@nicira.com> wrote: >> There are two solutions I can think of: >> 1. use "(2000 / bfd->cfg_min_rx + 10)" to compensate for the jitter. > > Sure I think this is fine as it gives us a little bit of wiggle room. > Seems a bit cleaner than option (2). > > Ethan > >> >> >> 2. if I can share the "push_timer" in ofproto-dpif.c with bfd.c then, I can >> tell exactly how much the jitter should be. >> _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
