Re: Status of RXE/Soft-RoCE driver?

[Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>]

On Mon, May 02, 2016 at 11:31:51AM -0700, Christoph Hellwig wrote:
> On Mon, May 02, 2016 at 11:57:09AM -0600, Jason Gunthorpe wrote:
> > The other take away from the Collab discussion is that there is a
> > reasonable opinion that linux-rdma is not the right forum to decide if
> > major changes to the multi-vendor common-verbs APIs are
> > OK. I don't know if we reached a consensus on this or not.
> 
> Who had that discussion, and why do you believe it's viable?  In fact
> I'd like to state that any other forum than linux-rdma simply is not
> credible, just like the relevant list is the only credible place to
> discuss the ABIs for any other Linux subsystem.

I would say mostly from people looking at it from a hardware design
perspective.

The unique issue with verbs is that nearly all the APIs match directly
to some feature in silicon, and cross-vendor silicon focused agreements
are rarely done on Linux focused mailing lists. Particularly when
these features are expected to be multi-operating system.

That is my fundamental concern every time a uAPI change comes up:
These proposed API changes *directly* require other vendors to
implement very specific things in their sillicon. This is not a
software-only discussion, as the majority of other Linux uAPI things
are.

* And I specifically separate the nitty gritty details of the API/ABI
from the overarching ideas: eg, introducing a dis-aggregated QP
concept with WQ objects, and defining it how that interacts with
everything else.

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