On Wed, Mar 4, 2009 at 10:24 AM, Hal Rosenstock <[email protected]> wrote: > Sasha, > > In commits 663255db91ecd2baf78b28d77b4397314e5349ca and > e59489bf06bfed7491b7c3da64ca22ed326b1563, there is the potential for > trap 144 to be generated. We've discussed some aspects of this when > the patches were first on the list. There's another aspect of this > that comes to light now that I've started playing with this. In > osm_req.c, osm_send_trap144 does not set (or for that matter currently > clear) the resp_expected flag. It seems to me that a more robust > strategy would be to set resp_expected to try to make sure that the > other side sees the trap
by handling the (trap repress) response matched by transaction ID > (and the timeout/retry strategy is used) if the trap repress is not received > There are a number of underpinnings to change to do this. Would you > accept a patch along those lines ? > > -- Hal > _______________________________________________ > general mailing list > [email protected] > http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general > > To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general > _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
