[openib-general] Re: [PATCH] CMA: allow/require bind beforeconnect

[Sean Hefty]

I'm still missing a bunch of openib mail, so I eventually went out to the openib 
web site to read Steve and Tom's full messages.  The fact that NFSoRDMA would 
break without a port number is reason enough to add it.

Caitlin Bestler wrote:
> Currently an application endpoint is identified by IP address
> and protocol/port, and a connection is a pairing of two endpoints.

I view RDMA endpoints as being identified by QPs, rather than addresses.  I 
guess for iWarp, a QP maps to a unique port-address pair (?), but the same isn't 
necessary for IB.

> Allocating a port number from one of the existing portspaces
> so that existing APIs can be complied with is a decision that
> can be made within this project.

I'm not sure it's as simple as that.  The only interfaces that I'm aware of for 
allocating a port require a socket structure.  Attempting to use those 
interfaces in an acceptable manner could require substantial changes.

> So, do you have any reason why allocating a port number for
> the active endpoint is so onerous that it is worth co-ordinating
> the evolution of an L4 endpoint with an IP address to accommodate
> that endpoint being an IB QP? Isn't assigning a port number to the
> IB QP far simpler?

The question is the domain from which the port number is allocated.

- Sean