[openib-general] Re: SDP hello ack header

[Michael S. Tsirkin]

Quoting r. Sean Hefty <sean.hefty at intel.com>:
> I think that a better solution is to have SDP set the version information for
> all headers.  The CMA can then check the version to see if it can support it,
> and set the other fields appropriately.  Thoughts?

Fine. I'll add that to SDP.
This will prevent the hard to catch bug like what I had with MajV unset.

BTW, does CMA MajV in incoming messages? It does not seem to.
If not this needs to be corrected:

CA4-15: The accepting peer shall reject the connection if MajV in the HH
does not match its local value.

and

CA4-22: The connecting peer shall terminate the connection attempt if
MajV does not match its local value, i.e., it sends a REJ back to the remote
peer, instead of RTU.

And I think this test clearly belongs in CMA.

But please also note that CA4-16/CA4-23 say that MinV should not be checked.

-- 
MST