[openib-general] sl2vl tables

[Hal Rosenstock]

On Thu, 2007-02-15 at 17:48, Dale Purdy wrote:
> 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.

You can run with -Q or set no_qos to FALSE in the opensm.opts file.

I wonder whether we should tie LASH to this so this isn't needed.

-- Hal

> 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
> > >
> >
> >