[PATCH v2 4/8] drm/crtc: Add a generic infrastructure to fake VBLANK events

[Boris Brezillon]

On Mon, 2 Jul 2018 10:37:22 +0200
Daniel Vetter <daniel at ffwll.ch> wrote:

> On Mon, Jul 02, 2018 at 10:14:51AM +0200, Boris Brezillon wrote:
> > On Mon, 2 Jul 2018 10:02:52 +0200
> > Daniel Vetter <daniel at ffwll.ch> wrote:
> >   
> > > On Fri, Jun 29, 2018 at 01:17:17PM +0200, Boris Brezillon wrote:  
> > > > In some cases CRTCs are active but are not able to generating events, at
> > > > least not at every frame at it's expected to.
> > > > This is typically the case when the CRTC is feeding a writeback connector
> > > > that has no job queued. In this situation the CRTC is usually stopped
> > > > until a new job is queued, and this can lead to timeouts when part of
> > > > the pipeline is updated but no new jobs are queued to the active
> > > > writeback connector.
> > > > 
> > > > In order to solve that, we add a ->no_vblank flag to drm_crtc_state
> > > > and ask the CRTC drivers to set it to true when they know they're not
> > > > able to generate VBLANK events. The core drm_atomic_helper_fake_vblank()
> > > > helper can then be used to fake VBLANKs at commit time.
> > > > 
> > > > Signed-off-by: Boris Brezillon <boris.brezillon at bootlin.com>
> > > > ---
> > > >  drivers/gpu/drm/drm_atomic_helper.c | 40 +++++++++++++++++++++++++++++++++++++
> > > >  include/drm/drm_atomic_helper.h     |  1 +
> > > >  include/drm/drm_crtc.h              | 15 ++++++++++++++
> > > >  3 files changed, 56 insertions(+)
> > > > 
> > > > diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> > > > index 69063bcf2334..ca586993c2a2 100644
> > > > --- a/drivers/gpu/drm/drm_atomic_helper.c
> > > > +++ b/drivers/gpu/drm/drm_atomic_helper.c
> > > > @@ -2051,6 +2051,46 @@ void drm_atomic_helper_wait_for_dependencies(struct drm_atomic_state *old_state)
> > > >  }
> > > >  EXPORT_SYMBOL(drm_atomic_helper_wait_for_dependencies);
> > > >  
> > > > +/**
> > > > + * drm_atomic_helper_fake_vblank - fake VBLANK events if needed
> > > > + * @old_state: atomic state object with old state structures
> > > > + *
> > > > + * This function walks all CRTCs and fake VBLANK events on those with
> > > > + * &drm_crtc_state.no_vblank set to true and &drm_crtc_state.event != NULL.
> > > > + * The primary use of this function is writeback connectors working in oneshot
> > > > + * mode and faking VBLANK events. In this case they only fake the VBLANK event
> > > > + * when a job is queued, and any change to the pipeline that does not touch the
> > > > + * connector is leading to timeouts when calling
> > > > + * drm_atomic_helper_wait_for_vblanks() or
> > > > + * drm_atomic_helper_wait_for_flip_done().
> > > > + *
> > > > + * This is part of the atomic helper support for nonblocking commits, see
> > > > + * drm_atomic_helper_setup_commit() for an overview.
> > > > + */
> > > > +void drm_atomic_helper_fake_vblank(struct drm_atomic_state *old_state)
> > > > +{
> > > > +	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
> > > > +	struct drm_crtc *crtc;
> > > > +	int i;
> > > > +
> > > > +	for_each_oldnew_crtc_in_state(old_state, crtc, old_crtc_state,
> > > > +				      new_crtc_state, i) {
> > > > +		unsigned long flags;
> > > > +
> > > > +		if (!new_crtc_state->no_vblank && !old_crtc_state->no_vblank)    
> > > 
> > > Uh, this essentially makes it impossible to reset no_vblank.  
> > 
> > I don't want ->no_vblank to be reset by the core. It's up to the CRTC
> > driver to clear/set it when something changes in the pipeline.
> >   
> > > For control
> > > flow state bits we only check the new state for it (see e.g. the various
> > > *_changed or plane_bits or whatever).  
> > 
> > I tried with !new_crtc_state->no_vblank only, but then it does not
> > handle the case where the CRTC and connector are being disabled, and I
> > end up with a timeout.  
> 
> Why that? You should have a new_crtc_state even when you disable the crtc,
> and you can set the ->no_vblank on that one too.

Hm, right. I'll have to check why I didn't have ->no_blank set to true
when I disable the CRTC. Probably a problem in vc4_crtc.c.


-- 
Boris Brezillon, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com