On Thursday 27 November 2008 15:26, Moni Shoua wrote:
> I'm also changing the action in case ib_query_port() fails.
> Instead of ignoring the failure I now assume the worst (i.e: LID_CHANGE)
>
OK. This will not be worse than the current situation.
(actually, it may in certain cases, because you can generate the both
LID_CHANGE event and
a CLIENT_REREGISTER event, where before only one was generated. However,
if query_port fails, we will probably see other failures as well, so what
the heck).
BTW,
The condition I sent in my last post is not enough.
It should be:
if ((in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED ||
in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) &&
in_mad->mad_hdr.method == IB_MGMT_METHOD_SET &&
in_mad->mad_hdr.attr_id == IB_SMP_ATTR_PORT_INFO &&
!ib_query_port(ibdev, port_num, &pattr))
prev_lid = pattr.lid;
since the query_port response data is only relevant for the
IB_SMP_ATTR_PORT_INFO
path in smp_snoop.
- Jack
_______________________________________________
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
