[PATCH 1/3] drm/i915/gvt: Move mdev attribute groups into kvmgt module

[Jason Gunthorpe]

On Fri, May 14, 2021 at 03:08:08PM +0800, Zhenyu Wang wrote:
> On 2021.05.13 09:02:49 -0300, Jason Gunthorpe wrote:
> > On Thu, May 13, 2021 at 12:56:47PM +0800, Zhenyu Wang wrote:
> > > On 2021.05.12 09:47:39 -0300, Jason Gunthorpe wrote:
> > > > On Wed, May 12, 2021 at 10:31:41AM +0800, Zhenyu Wang wrote:
> > > > 
> > > > > > This need to go into the vfio tree in some way, either directly
> > > > > > through it, via rc or otherwise
> > > > > 
> > > > > As this is only for i915/gvt, once drm/i915 backmerge with linus master,
> > > > > it should still go through normal i915/gvt merge path.
> > > > 
> > > > Don't do this, you will create conflicts with ongoing vfio work.
> > > > 
> > > 
> > > Sure, there always could be conflict, which means you need to rebase onto
> > > this cleanup. Would that a problem? 
> > 
> > Yes.
> > 
> > > I'd want to fix current workaround in 5.13-rc.
> > 
> > This doesn't seem like a rc candiate to me, but going to -rc is also
> > fine.
> > 
> > > Merging i915/gvt only change through vfio doesn't make sense to me,
> > 
> > You need to do it to avoid major conflicts for stuff that will go into
> > the vfio tree this cycle.
> 
> Looks apply to vfio/for-linus is fine, and vfio/next missed gvt change to
> apply...but not conflict with any new stuff. Alex, pls let me know if you
> have any concern of this.

The conflicting stuff hasn't landed to a tree yet, I am helping you
plan things out so you don't have to explain to Linus why your tree
has conflicts all over the place.

> > VFIO drivers should not be outside drivers/vfio/ in the first place,
> > and this shows why.
> > 
> 
> Well, I can't agree, otherwise that'll be dependency nightmare for device
> driver writer. ;)

Not really

Jason