Re: [PATCH v8 0/5] Migrate on fault for device pages

[Matthew Brost <matthew.brost@intel.com>]

On Fri, Apr 24, 2026 at 07:50:37PM +0300, Mika Penttilä wrote:
> On 4/24/26 19:23, Andrew Morton wrote:
> 
> > On Tue, 14 Apr 2026 07:12:21 +0300 mpenttil@redhat.com wrote:
> >
> >> From: Mika Penttilä <mpenttil@redhat.com>
> >>
> >> Currently, the way device page faulting and migration works
> >> is not optimal, if you want to do both fault handling and
> >> migration at once.
> >>
> >> Being able to migrate not present pages (or pages mapped with incorrect
> >> permissions, eg. COW) to the GPU requires doing either of the
> >> following sequences:
> >>
> >> ...
> >>
> > Via which tree would it be best to upstream this?
> >
> >>  include/linux/hmm.h                    |  19 +-
> >>  include/linux/migrate.h                |  26 +-
> >>  lib/test_hmm.c                         | 101 ++-
> >>  lib/test_hmm_uapi.h                    |  19 +-
> >>  mm/Kconfig                             |   2 +
> >>  mm/hmm.c                               | 821 +++++++++++++++++++++++--
> >>  mm/migrate_device.c                    | 589 +++---------------
> >>  tools/testing/selftests/mm/hmm-tests.c |  54 ++
> >>  8 files changed, 1053 insertions(+), 578 deletions(-)
> > Looks like mm.git.
> >
> > It might be more useful to add it to a DRM tree if there's ongoing
> > development against this.  But please no upstreaming without good
> > review!
> >
> > Anyway, this is a bit dated and Sashiko AI review wasn't able to apply
> > the patches.  Nor was it able to apply v7 or v6.  So please refresh,
> > retest, resend?
> 
> It is against Linus tree v7.0, base-commit: 028ef9c96e96197026887c0f092424679298aae8,
> but applies and passes hmm selftests on mm trees as is.
> 
> I plan to do v9 refresh and maybe minor updates in week or so.
> 

It might be good to send a version which applies to drm-tip to
intel-xe@lists.freedesktop.org to trigger our CI too. I know tested an
earlier rev and everything seemed to be working but another round of
testing before merging is likely a good idea.

Matt

> --Mika
>