[openib-general][PATCH][RFC]: CMA header

[Roland Dreier]

This isn't horrible, but you seem to have ignored most of the
discussion from last month:

    > int ib_cma_get_device(struct sockaddr *remote_address,
    > 		      enum ib_qos qos, struct ib_device **device, u8 *port);

How are you dealing with hotplug and object lifetime issues here?

    > int ib_cma_get_src_ip(void *cma_id, ib_cma_addr_handler addr_handler,
    > 		      void *context);

There's no point in making this asynchronous, since we're putting the
source/dest information in the CM REQ private data.

    > int ib_cma_create_qp(struct ib_pd *pd, u8 port, struct ib_qp **qp,
    > 		     struct ib_qp_init_attr *init_attr);

What's the point of this function?

    > int ib_cma_listen(struct ib_device *device, struct sockaddr *address,
    > 		  __be64 service_id, void *context,
    > 		  ib_cma_listen_handler cm_listen_handler,
    > 		  void **cma_id);

A minor point, but there's no need for the service_id parameter.
We'll just use the sin_port (or sin6_port) member of the sockaddr.
Similarly struct ib_cma_conn doesn't need the service_id member either.

 - R.