>To be honest I haven't read this in detail. However it seems to have >generated a lot of discussion. Has a consensus been reached?
Just to make sure, here's my summary of the discussions so far. Anyone who disagrees can jump in. There was discussion on this being 2.6.18 material. The consensus is to let it get wider testing in svn first. The fix / feature isn't critical enough to push upstream immediately, so waiting until 2.6.19 seems reasonable. There was also discussion on whether the CM should track local QPNs on both the active and passive sides, or just the passive side. To properly track timewait and eventually handle path migration, tracking on both sides is needed, which is what the patch does. Or asked about removing the ib_cm_establish() call from the API. It was left as part of the API to avoid breaking the ABI, and still allow users to force connection establishment in case they poll a receive WC before the COMM_EST event occurs. There was also some side discussion about the COMM_EST event in general and trying to respond to request messages while the QP is still in the RTR state. A discussion on possible changes to the spec spawned off from this... - Sean _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
