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@openib.org > > > http://openib.org/mailman/listinfo/openib-general > > > > > > To unsubscribe, please visit > > > http://openib.org/mailman/listinfo/openib-general > > > > > > > _______________________________________________ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general