On 13:34 Sun 14 Dec , Eli Dorfman wrote: > >> > Did you mean to have it osm_report_notice()? Actually it is where OpenSM > >> > sends notices, not where OpenSM gets traps. Trap receiver processor is > >> > located in osm_trap_rcv.c. > >> > >> Yes that's what i meant. > >> When OpenSM receives traps it calls osm_report_notice(). > >> It is also call for OpenSM initiated traps (e.g. GID IN/OUT and MC > >> CREATE/DELETE). > > > > Ok. I see your point. Then why it should be limited by generic notice > > types? > > No special reason. Just at the moment we handle only generic traps. > we may want to report vendor specific event with other event id
OpenSM event plugin is generic API, I think we should report any trap. > > Also wouldn't it be better to call plugin report callback after > > notice was actually processed (eg. at end of this function)? > > there is no correlation between reporting an event and whether it was > already forwarded. This is maybe true from plugin perspective, but reporting an event could (at least potentially) slow down core functionality - trap sending. After all I think it would be better to report OSM_EVENT_ID_TRAP unconditionally and to make this at end of osm_report_notice() function. Agreed? Sasha _______________________________________________ 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
