>Always letting the HCA driver process the MADs, similar to the linux >code, makes sense. But MAD handling in the HCA driver is done >synchronously, and not all calls into the MAD layer are blockable. >Probably the best you could do is add a new 'do_you_want_this_mad()' >call to ci_interface_t that returned whether the HCA wanted the MAD or >not, but did not process the MAD.
Both of the HCA drivers contain functionality that looks like it could be extracted out and placed into a do_you_want_this_mad() call. The functionality could come from the mthca_process_mad() and mlx4_ib_process_mad() routines. I can see what it would take to do this, and remove the local routing determination from ibal. Leonid, do you have a preference on which direction you would prefer? - Sean _______________________________________________ ofw mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
