Re: [PATCH V7 mlx5-next 15/15] vfio: Extend the device migration protocol with PRE_COPY

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Feb 17, 2022 at 10:15:54AM -0700, Alex Williamson wrote:

> I feel obligated to ask, is PRE_COPY support essentially RFC at this
> point since we have no proposed in-kernel users?

Yes, it is included here because the kernel in v1 had PRE_COPY, so it
seemed essential to show how this could continue to look to evaluate
v2.

NVIDIA has an out of tree driver that implemented PRE_COPY in the v1
protocol, and we have some future plan to use it in a in-tree driver.

> It seems like we're winding down comments on the remainder of the
> series and I feel ok with where it's headed and the options we have
> available for future extensions.  

Thanks, it was a lot of work for everyone to get here!

Yishai has all the revisions from Kevin included, he will sent it on
Sunday. Based on this Leon will make a formal PR next week so it can
go into linux-next through your tree. We have to stay co-ordinated
with our netdev driver branch..

I will ping the acc team and make it priority to review their next
vresion. Let's try to include their driver as well.

We'll start to make a more review ready qemu series.

> PS - Why is this a stand-alone ioctl rather than a DEVICE_FEATURE?

You asked for the ioctl to be on the data_fd, so there is no
DEVICE_FEATURE infrastructure and I think it doesn't make sense to put
a multiplexor there. We have lots of ioctl numbers and don't want this
to be complicated for performance.

Thanks,
Jason