[openib-general] opensm trouble

[Roland Fehrenbacher]

>>>>> "Eitan" == Eitan Zahavi <eitan at mellanox.co.il> writes:

Hi Eitan,

    >     Choose a local port number with which to bind: 
    > 
    >             1: GUID = 0x       0, lid = 0x03C9, state = INIT 
    >             2: GUID = 0x       0, lid = 0x03CA, state = DOWN 
    > 
    >     Enter choice (1-2): 1 
    >     SM port is down.
    
    Eitan> The fact the displayed GUID is 0 means that somehow OpenSM
    Eitan> did not get the "correct" (in it's own eyes) to
    Eitan> VAPI_query_hca_gid_tbl.

    Eitan> I assume you were able to do vstat ?

yes. vstat output is:

# vstat
1 HCA found:
        hca_id=InfiniHost0
        vendor_id=0x02C9
        vendor_part_id=0x5A44
        hw_ver=0xA1
        fw_ver=0x300020000
        num_phys_ports=2
                port=1
                port_state=PORT_INITIALIZE
                sm_lid=0x0000
                port_lid=0x03c9
                port_lmc=0x00
                max_mtu=2048

                port=2
                port_state=PORT_DOWN
                sm_lid=0x0000
                port_lid=0x03ca
                port_lmc=0x00
                max_mtu=2048


    Eitan>   Seems there is a bug either in the OpenSM vendor layer:
    Eitan> osm_vendor_mlx_hca.c

    Eitan> Or there is a miss-match between the evapi.h that OpenSM is
    Eitan> linked with and the one that vstat is linked with.

    Eitan> I will be able to try and reproduce the problem in our lab
    Eitan> only Sunday.  But I guess if you have a chance to look into
    Eitan> the above file it will not be very hard to find too.

Roland