[openib-general] sl2vl tables

[Dale Purdy]

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?

Dale