[Intel-gfx] [PATCH 7/7] drm/i915/gem: Acquire all vma/objects under reservation_ww_class

[Dave Airlie]

On Tue, 23 Jun 2020 at 20:03, Chris Wilson <chris at chris-wilson.co.uk> wrote:
>
> Quoting Thomas Hellström (Intel) (2020-06-23 10:33:20)
> > Hi, Chris!
> >
> > On 6/22/20 11:59 AM, Chris Wilson wrote:
> > > In order to actually handle eviction and what not, we need to process
> > > all the objects together under a common lock, reservation_ww_class. As
> > > such, do a memory reservation pass after looking up the object/vma,
> > > which then feeds into the rest of execbuf [relocation, cmdparsing,
> > > flushing and ofc execution].
> > >
> > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > ---
> > >   .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 91 ++++++++++++++-----
> > >   1 file changed, 70 insertions(+), 21 deletions(-)
> > >
> > Which tree is this against? The series doesn't apply cleanly against
> > drm-tip?
>
> It's continuing on from the scheduler patches, the bug fixes and the
> iris-deferred-fence work. I thought throwing all of those old patches
> into the pile would have been distracting.
>
> > ...
> >
> > > +static int eb_reserve_mm(struct i915_execbuffer *eb)
> > > +{
> > > +     const u64 idx = eb->context->timeline->fence_context;
> > > +     struct ww_acquire_ctx acquire;
> > > +     struct eb_vma *ev;
> > > +     int err;
> > > +
> > > +     eb->mm_fence = __dma_fence_create_proxy(0, 0);
> > > +     if (!eb->mm_fence)
> > > +             return -ENOMEM;
> >
> > Where are the proxy fence functions defined?
>
> In dma-fence-proxy.c ;)

The dma-fence-proxy that Christian NAKed before?

Dave.