Thread (52 messages) 52 messages, 4 authors, 2021-04-18

Re: [PATCH v4 01/23] iidc: Introduce iidc.h

flat view

From: Jason Gunthorpe <jgg@nvidia.com>
Date: 2021-04-12 16:12:26
Also in: linux-rdma

On Mon, Apr 12, 2021 at 02:50:43PM +0000, Saleem, Shiraz wrote:
Because it is an unnecessary thing to force a user to build/load drivers for
which they don't have the HW for? 
Happens automatically in all distros, so I don't agree with this.
The problem gets compounded if we have to do it for all future HW
Intel PCI drivers, i.e. depends on ICE && ....
Then someday build a proper pluggable abstraction and put all your
ethernet drivers under it. 

Today you haven't done that and all we have a set of ops that only
work with one eth driver and a totally different set of functions that
only work with a different driver.

It is all just dead code until it gets finished and process is to not
merge dead code.
There is a near-term Intel ethernet VF driver which will use IIDC to
provide RDMA in the VF, and implement some of these .ops
callbacks. There is also intent to move i40e to IIDC.
"near-term" We are now on year three of Intel talking about this
driver!

Get the bulk of the thing merged and deal with the rest in followup
patches.
But in a unified irdma driver model connecting to multiple PCI gen
drivers, I do think it serves a reason.
It is fine as a high level idea, but the implementation has to meet
kernel standards.

Jason
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help