Sean Hefty wrote:
On IB, resolve route is done using unreliable datagram with no lower level timeout or retry.
Hi Sean,
I wasn't sure to follow on your comment, since the sa_query kernel code does deliver the resolve_route timeout param down to the mad code, and with some enhancement it could also program the retries field of the ib_send_mad_buf structure, this way both the timeout (as today) and the retries would be done in the lowest layer (mad).
Or. _______________________________________________ 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
