On Jan 18, 2016, at 8:41 PM, FUJITA Tomonori <[email protected]> wrote: > > I see. I wait for the updated patches. Please let me know if you have > a patch that you want me to apply now.
Thanks for your patience. ;) I will continue to test my patch in our environment until 1/25, just to make sure everything behaves “as expected.” The changes Zhao required altered things slightly - but not so much that I am truly worried. I just want to let them “burn in” on our hardware. The patch Zhao received was about 33K. Presuming no more changes, that’s a little large to apply in one shot. I will break it up into a series, with an explanatory email, when I submit. There were a large number of subtle issues, and I made some changes that *might* be controversial. For example - I have introduced logic for enforcing a maximum allowed handler execution time (that is configurable). Ryu handlers are not supposed to block, or run for terribly long - otherwise the event queue might not be processed as it needs to be. This is documented, but nothing enforced it. My added logic enforces it, provides a backtrace for the handler that ran too long, and allows the event loop to continue processing. Another potentially controversial change: I have split the event queue into two, with OFPStateChange events being processed separately from (and in preference to) all other events. OFPStateChange events are special, in that they are internally generated and they need to be processed quickly (so that the state machine governing the actions of a Datapath proceeds in a timely manner). Without the two queues and the maximum execution timer, deadlocks could occur between event processing and datapath state transitions, especially when the event queue was saturated with PacketIn events. Anyhow - I’ll explain everything more fully when I submit the patch series, after I give the patch a bit more time to run. Zhao: if you run into any issues, *please* let me know. I’ve been working hard to stamp out event loop deadlocks, leaked sockets, leaked datapaths, etc. If you find anything problematic, I’d be grateful for the feedback (and, if one is available, a backtrace - a screenshot like you sent was fine). Best, Victor -- Victor J. Orlikowski <> vjo@[cs.]duke.edu ------------------------------------------------------------------------------ Site24x7 APM Insight: Get Deep Visibility into Application Performance APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month Monitor end-to-end web transactions and take corrective actions now Troubleshoot faster and improve end-user experience. Signup Now! http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140 _______________________________________________ Ryu-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ryu-devel
