[ofa-general] IBCM and ordering of OpenFabrics devices

[Jeff Squyres]

On Jul 21, 2008, at 2:30 PM, Hefty, Sean wrote:

>> 1. Technically, mlx4_0 is the first IB device. Should it therefore be
>> using ucm0? I.e., is libibcm wrong for trying to use ucm1? (note that
>> OMPI's openib BTL is currently replicating the logic from libibcm to
>> check for the Right ucm* file so that we can silently fail before
>> ib_cm_open_device() fails with a warning message -- so if libibcm's
>> logic to find the Right ucm* file changes, we'll also need to change
>> MPI's logic to mirror it. OMPI's logic becomes moot in newer libibcm
>> versions where Sean removed the warning message, though).
>
> I think what may be happening is that uverbs1 should map to ucm0, but
> the library opens ucm1 instead.  And uverbs2 should map to ucm1, but  
> an
> attempt is made to open ucm2, which causes the failure that you're
> seeing.  The libibcm needs to determine the correct file to open using
> something other than just the uverbs name.
>
> That was just a convenient way to get the correct file name, but was
> never updated once libibverbs supported non-IB devices.


Gotcha.

Do you have a guesstimate on when you could have a patch available?

-- 
Jeff Squyres
Cisco Systems