Re: [RFC] Vendor-specific QPs

[Jason Gunthorpe <jgg-uk2M96/98Pc@public.gmane.org>]

On Thu, Nov 23, 2017 at 11:14:40AM +0200, Moni Shoua wrote:

> Since all dv_create_qp() API functions call the ibv_cmd_create_qp()
> and put driver data at the end of the core data it will be simplest to
> manage this in one function with one additional driver data structure
> (that includes a comp_mask to describe the content of the block of
> data). Also, the signature of the mlx5_dv_create_dct() should take an
> extra parameter (besides ib_qp_init_attr and mlx4_dct_init_attr) that
> is common to all dv_create_qp() functions to represent the shared
> driver data. We can avoid all this if we declare only one API function
> to create a driver QP.

I don't see why it makes any difference. Shuffle the data around
inside a mlx5_dv_create_dct to get to the internal format then call
the internal create_qp helper.

Why should it accept a 'shared driver data'? What does that even mean
from the user facing API?

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html