[openib-general] userspace CM API for per device handling

[Sean Hefty]

Roland Dreier wrote:
>>ibv_get_async_event(int fd, struct ibv_async_event *event);
>>ibv_get_cq_event(int fd, struct ibv_cq **cq, void **cq_context);
> 
> This seems like mostly pain with little gain to me.  A consumer doing
> a poll or something with multiple file descriptors still needs some
> mapping to some per-fd context so that it knows which fds are CQ event
> fds, which ones are async event fds, and which ones are neither.  So
> it's pretty easy to go back to a verbs context.

I'm not sold on this change either.  Right now I'm just trying to find a decent 
API for the CM, and the one you mentioned works just as well.

As for mapping fd's to context, I think this depends on how the user groups 
multiple file descriptors together and their threading model.  A user could poll 
only fd's associated with CQs; although, I don't think that the current 
implementation of DAPL does this.

> If we wanted to be more symmetrical, we could have a CM API like
> 
>     struct ib_cm_context *ib_cm_get_context(struct ibv_context *dev_context);
>     int ib_cm_get_event(struct ib_cm_context *context, struct ib_cm_event **event);

This API still gives the benefits that I was looking for, so I will go with 
something like this for now.  Thanks.

- Sean