[openib-general] rdma_cm.h: comment nits.

[Caitlin Bestler]

openib-general-bounces at openib.org wrote:
> Tom Tucker wrote:
>>> Its OK to call rdma_reject on active side as well, isn't it?
>> 
>> You'll get -EINVAL on iWARP if you do this....
> 
> For IB, rdma_reject can be called on the active side if the
> user is managing their own QP states, or is SDP.  How does iWarp
> support userspace QPs? 
> 

The assumption is that the connection is established if the
passive side indicated to proceed knowing what the active 
side requested. That doesn't mean that it was a take it or
leave it. The passive side's response could still have 
reduced the requested resource reservations. But if the
active side does not like it, the only real recourse is
to break the connection.

IT-API has some good abstractions and write-ups on two-step
vs. three-step private data exchanges in connection setup.
The bottom line is that two-way is portable, three-way
is InfiniBand specific.

My assumption has been that an application that truly required
three-way exchanges is probably doing something very specific
with IB resources, and hence would use IB-specific connection
setup.

I'm not following your question about iWARP support for
user mode QPs. The caller (user or kernel) supplies the QP
and what they want done. The only real difference is what
the resources behind a "connection request" are. With iWARP
there is an actual TCP connection by the time the private
data has been collected (from the MPA Request frame).