[ofa-general] implicit address binding by rdma_resolve_addr

[Sean Hefty]

> Correct me if I am wrong, but implementing this scenario for UD will
> look like this: 
> /* accept handling not shown... */
> rdma_bind_addr(id, ...);
> rdma_listen(id)
> while(1) {
> 	read_request();
> 	do_work();
> 	if(new_client) { /* check if we already connected to client */
> 		/* following is not going to work because local address
> is already bound */
> 		rdma_resolve_addr(my_local_already_bound_addr,
> client_addr);

There could very well be a bug with the state management in the rdma cm 
that disallows this.  (Have you tried this and seen it fail?) 
rdma_resolve_addr() only binds if the id is not already bound.  So, call 
sequence such as:

rdma_bind_addr(id..)
rdma_resolve_addr(id..)

should work fine.

Likewise,

rdma_resolve_addr(id..)
rdma_resolve_route(id..)
rdma_connect(id..)

*should* work, but may not if the connection state isn't being reset 
properly for UD.

Note that you don't have to use the same id that you send on to resolve 
the address.  You can use a temporary ID that you allocate for the 
purpose of obtaining the remote address handle.

As an alternative to calling rdma_resolve_addr() from the server, you 
can use the receive completion from the client to create an address 
handle.  See the udaddy.c sample program with the librdmacm - 
specifically create_reply_ah().

- Sean