On Wed, 2004-11-10 at 13:43, Sean Hefty wrote: > Hal Rosenstock wrote: > > > Currently if no matching send request is found, the received MAD is > > freed (around line 1035 of the current mad.c). > > > > In this case, timeout too short, etc., is this the correct behavior ? > > Or should the receive packet be given to a matching MAD agent with a > > receive handler (perhaps with a different status) ? The latter would > > allow for an additional send model for requests which I don't think is > > supported now at the cost of having the client throw away these receives > > based on a new status code (perhaps some sort of timeout). > > I think that this is the behavior that you'd want, but I can see your > view, and I'm open to changing it. From a client's perspective, > dropping an unmatched MAD keeps the client from having to handle receive > MADs without having a send outstanding. That is, I would think that a > client that could make use of this MAD would have to be fairly complex.
I don't know whether the SM or other managers would use this model so it's just a thought to keep in mind for the future. > I see a couple of cases where this would happen. The first is the one > you mention, where the timeout was too short. If the client retries the > request, then they would need to deal with an unmatched response coming > in before they issued the retry, while the retry is active (where the > retry is sent after the received had checked for a match), or after the > retry completed (with the need to handle multiple unmatched responses.) > > The second case where I can see this happening is if the client canceled > the send, and I'm not sure that we'd want to give the client an > unmatched response in this case. So we would also need to timeout send MAD cancellations (and not eliminate them totally immediately) so we wouldn't give a receive back in that case. -- Hal _______________________________________________ openib-general mailing list [EMAIL PROTECTED] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general