During the first post the issue of providing also the SL (and/or other
params) from the broadcast group as part of the path query was raised,
and I kind of failed to follow all the discussion that evolved... Can
you clarify if the consensus was that based on the pkey and traffic
class, the SA should return the --same-- SL (and/or other params) on
this path query as of the broadcast group?
I used TClass in case the multicast group spanned subnets. Since TCLass
is defined as the end-to-end service level, I think it is sufficient.
I don't know that the SM *must* have an N:1 TClass->SL mapping, but that
doesn't seem unreasonable.
memcpy(path->pathrec.dgid.raw, gid, sizeof (union ib_gid));
- path->pathrec.sgid = priv->local_gid;
- path->pathrec.pkey = cpu_to_be16(priv->pkey);
- path->pathrec.numb_path = 1;
+ path->pathrec.sgid = priv->local_gid;
+ path->pathrec.pkey = cpu_to_be16(priv->pkey);
+ path->pathrec.numb_path = 1;
Did you just wanted to add space/tab here? also some lines are broken at
least as my email see this patch, maybe you had some problem?
I don't see the broken lines, but, yes, I was just adjusting the spacing
here to line up with setting traffic_class (below).
+ path->pathrec.traffic_class =
priv->broadcast->mcmember.traffic_class;
For this to take effect, don't you need to set the
IB_SA_PATH_REC_TRAFFIC_CLASS bit in the component mask?
Err... yes. I've updated the patch. Thanks
- Sean
_______________________________________________
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
