Re: [resend/standalone PATCH v4] Add auxiliary bus support

[Dan Williams]

On Thu, Dec 3, 2020 at 6:34 PM Jason Gunthorpe <jgg@xxxxxxxxxx> wrote:
>
> On Thu, Dec 03, 2020 at 04:06:24PM +0100, Greg KH wrote:
>
> > > ...for all the independent drivers to have a common commit baseline. It
> > > is not there yet pending Greg's Ack.
> >
> > I have been trying to carve out some time to review this.  At my initial
> > glance, I still have objections, so please, give me a few more days to
> > get this done...
>
> There are still several more days till the merge window, but I am
> going to ask Leon to get the mlx5 series, and this version of the
> auxbus patch it depends on, into linux-next with the intention to
> forward it to Linus if there are no substantive comments.
>
> Regardless of fault or reason this whole 1.5 year odyssey seems to have
> brought misery to everyone involved and it really is time to move on.
>
> Leon and his team did a good deed 6 weeks ago to quickly turn around
> and build another user example. For their efforts they have been
> rewarded with major merge conflicts and alot of delayed work due to
> the invasive nature of the mlx5 changes. To continue to push this out
> is disrespectful to him and his team's efforts.
>
> A major part of my time as RDMA maintainer has been to bring things
> away from vendor trees and into a common opensource community.  Intel
> shipping a large out of tree RDMA driver and abandoning their intree
> driver is really harmful. This auxbus is a substantial blocker to them
> normalizing their operations, thus I view it as important to
> resolve. Even after this it is going to take a long time and alot of
> effort to review their new RDMA driver.

When you have 3 independent driver teams (mlx5, i40e, sof) across 2
companies (NVIDIA and Intel), and multiple subsystem maintainers with
a positive track record of upstream engagement all agreeing on a piece
of infrastructure, I struggle to imagine a stronger case for merging.
I did get word of a fixup needed in the shutdown code, I'll get that
folded. Then, barring a concrete objection, I'll look to publish a
commit that others can pull in to start building soak time in -next
this time tomorrow.