[openib-general] ibv_devinfo

[Roland Dreier]

 > Here is what I mean (partial output of "ibv_devinfo -v"):

 > phys_state:             invalid physical state (0)

 > The phys_state above does not correspond to what is in 
 > /sys/class/infiniband/ehca0/ports/1 
 > 
 > However, the "phys state" that ibstatus displays is indeed correct and as 
 > expected (see below).

 >         phys state:      0: <unknown>

umm... I'm not sure that either "unknown" or "invalid" can really be
considered an incorrect translation of physical state 0.  The IB spec
does not give 0 as one of the defined values for the physical state
field, so I think it is perfectly fine to say the state is "invalid."

I think the real problem is that the ehca driver in the kernel does
not fill in a real value for the phys_state field in its query_port
method.  This is also why the port's capability flags are (incorrectly
I assume) shown as 0x00000000 by ibv_devinfo, etc.

 - R.