Re: [RFC 3/4] drm/atomic-helper: Set fence deadline for vblank

[Rob Clark <robdclark@chromium.org>]

On Tue, Jul 27, 2021 at 3:44 AM Michel Dänzer <michel@daenzer.net> wrote:
>
> On 2021-07-27 1:38 a.m., Rob Clark wrote:
> > From: Rob Clark <robdclark@chromium.org>
> >
> > For an atomic commit updating a single CRTC (ie. a pageflip) calculate
> > the next vblank time, and inform the fence(s) of that deadline.
> >
> > Signed-off-by: Rob Clark <robdclark@chromium.org>
> > ---
> >  drivers/gpu/drm/drm_atomic_helper.c | 36 +++++++++++++++++++++++++++++
> >  1 file changed, 36 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> > index bc3487964fb5..f81b20775b15 100644
> > --- a/drivers/gpu/drm/drm_atomic_helper.c
> > +++ b/drivers/gpu/drm/drm_atomic_helper.c
> > @@ -1406,6 +1406,40 @@ void drm_atomic_helper_commit_modeset_enables(struct drm_device *dev,
> >  }
> >  EXPORT_SYMBOL(drm_atomic_helper_commit_modeset_enables);
> >
> > +/*
> > + * For atomic updates which touch just a single CRTC, calculate the time of the
> > + * next vblank, and inform all the fences of the of the deadline.
> > + */
> > +static void set_fence_deadline(struct drm_device *dev,
> > +                            struct drm_atomic_state *state)
> > +{
> > +     struct drm_crtc *crtc, *wait_crtc = NULL;
> > +     struct drm_crtc_state *new_crtc_state;
> > +     struct drm_plane *plane;
> > +     struct drm_plane_state *new_plane_state;
> > +     ktime_t vbltime;
> > +     int i;
> > +
> > +     for_each_new_crtc_in_state (state, crtc, new_crtc_state, i) {
> > +             if (!wait_crtc)
> > +                     return;
>
> Either this return or the next one below would always be taken, I doubt this was intended.

oops, the condition here is mistakenly inverted, it was meant to bail
if there is more than a single CRTC

>
> > +             wait_crtc = crtc;
> > +     }
> > +
> > +     /* If no CRTCs updated, then nothing to do: */
> > +     if (!wait_crtc)
> > +             return;
> > +
> > +     if (drm_crtc_next_vblank_time(wait_crtc, &vbltime))
> > +             return;
> > +
> > +     for_each_new_plane_in_state (state, plane, new_plane_state, i) {
> > +             if (!new_plane_state->fence)
> > +                     continue;
> > +             dma_fence_set_deadline(new_plane_state->fence, vbltime);
> > +     }
>
> vblank timestamps correspond to the end of vertical blank, the deadline should be the start of vertical blank though.
>

hmm, I suppose this depends on whether the hw actually has separate
irq's for frame-done and vblank (and whether the driver
differentiates).. and if the display controller is doing some
buffering, the point at which it wants to flip could be a bit earlier
still.  Maybe we just want a kms driver provided offset for how early
it wants the deadline relative to vblank?

BR,
-R

>
> --
> Earthling Michel Dänzer               |               https://redhat.com
> Libre software enthusiast             |             Mesa and X developer