[openib-general] Howto setup rping and krping for OpenIB with AMSO1100

[Hal Rosenstock]

On Thu, 2006-05-18 at 20:46, Roland Dreier wrote:
>     ChristopherX> Hello, After spending some time installing OpenIB
>     ChristopherX> with AMSO1100, I have managed to get rping and
>     ChristopherX> krping to work.  Here is a howto document describing
>     ChristopherX> the steps I took in the process.  This does not get
>     ChristopherX> opensm to work, as I am still getting a port guid
>     ChristopherX> error on that front.
> 
> Thanks for the documentation.  Perhaps you could add it to the
> openib.org wiki?
> 
> It doesn't make sense to run OpenSM with the Ammasso device or with
> any iWARP/ethernet device.  Subnet managers are strictly an InfiniBand
> concept.
> 
> Maybe OpenSM should look at the node_type entry in sysfs and print a
> more informative message if someone tries to run it on an iWARP RNIC?

I don't think OpenSM ever tries to run on an iWARP RNIC. There are 2
modes of starting up OpenSM. One where it finds the "first" port (active
port state, then physical state link up, and finally physical state not
disabled) on an IB NIC (no -g specified) and the other where an explicit
GUID is chosen by the user (admin). In the first case, any iWARP RNICs
are bypassed. If there are no available IB NICs, then an error should be
indicated by libibumad. 

In the second case, here's what I get when OpenSM is attempted to be
explictly started on the AMSO1100:

ibstat shows:
??? 'amso0'
        ??? type: AMSO1100
        Number of ports: 1
        Firmware version: 0.0.0
        Hardware version: 0
        Node GUID: 0x000db200066d0000
        System image GUID: 0x000db200066d0000
        Port 1:
                State: Active
                Physical state: No state change
                Rate: 2
                Base lid: 0
                LMC: 0
                SM lid: 0
                Capability mask: 0x009f0000
                Port GUID: 0x73e01580ffffffff

opensm -g=0x73e01580ffffffff
Error from osm_opensm_bind (0x2A)

The error messages in osm.log show:
Jan 27 17:56:15 064682 [AB002D00] -> osm_vendor_bind: Binding to port 0x73e01580ffffffff.
Jan 27 17:56:15 065370 [AB002D00] -> osm_vendor_bind: ERR 5426: Unable to register class 129 version 1 
Jan 27 17:56:15 065396 [AB002D00] -> osm_sm_mad_ctrl_bind: ERR 3118: Vendor specific bind failed
Jan 27 17:56:15 065410 [AB002D00] -> osm_sm_bind: ERR 2E10: SM MAD Controller bind failed (IB_ERROR)
Jan 27 17:56:15 065447 [AB002D00] -> osm_sa_mad_ctrl_unbind: ERR 1A11: No previous bind

Guess I could add something explictly here for this case to make it more obvious.

-- Hal

>  - R.
> _______________________________________________
> 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