[openib-general] sl2vl tables

[Dale Purdy]

We are experimenting with LASH.  It appears that the SL2VL tables
don't get initialized unless QoS is enabled on the opensm command line
(-Q).  Enabling this seems to rectify the problem.  So it would appear
that LASH needs to enable this also.

Dale

On Thu, 15 Feb 2007, Hal Rosenstock wrote:

> On Thu, 2007-02-15 at 17:08, Dale Purdy wrote:
> > We are experimenting with OFED 1.2 (alpha1) and have dumped the SL2VL
> > tables for both a switch port and an HCA port using the smpqueury
> > command:
> >
> > switch:
> > # SL2VL table: Lid 103
> > #                 SL: | 0| 1| 2| 3| 4| 5| 6| 7| 8| 9|10|11|12|13|14|15|
> > ports: in  0, out  1: | 0| 1| 2| 3| 0| 1| 2| 3| 0| 1| 2| 3| 0| 1| 2| 3|
> > ports: in  1, out  1: | 0| 1| 2| 3| 0| 1| 2| 3| 0| 1| 2| 3| 0| 1| 2| 3|
> > ...
> >
> > HCA:
> > # SL2VL table: Lid 37
> > #                 SL: | 0| 1| 2| 3| 4| 5| 6| 7| 8| 9|10|11|12|13|14|15|
> > ports: in  0, out  0: | 3| 2| 1| 0| 3| 2| 1| 0| 3| 2| 1| 0| 3| 2| 1| 0|
> >
> > I would have expected the behavior that the switch describes - a one
> > to one mapping mod the supported number of VLs.  But I can't explain
> > why the HCA VLs are in reverse order to the SL.  If this were a host
> > endian issue I would have expected both to behave the same.  Can
> > someone explain what is going on?
>
> Is this on powerup of HCA node or after some SM potentially programs the
> HCA (for this) ?
>
> I typically see (for HCAs):
> # SL2VL table: Lid 10
> #                 SL: | 0| 1| 2| 3| 4| 5| 6| 7| 8| 9|10|11|12|13|14|15|
> ports: in  0, out  0: | 0| 1| 2| 3| 4| 5| 6| 7| 0| 1| 2| 3| 4| 5| 6| 7|
>
> -- Hal
>
> > Dale
> >
> > _______________________________________________
> > openib-general mailing list
> > openib-general at openib.org
> > http://openib.org/mailman/listinfo/openib-general
> >
> > To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
> >
>
>