Hi Emil,

On Thu, May 28, 2020 at 04:51:31PM +0100, Emil Velikov wrote:
> Have you considered splitting the series into several parts and
> focusing on merging one at a time?
> IIRC this the longest series _ever_ submitted to dri-devel, plus it
> seems to be growing with each revision.
> 
> Due to the sheer volume, it's likely to miss various points - large or
> small (like below).

Yeah, I know, I wasn't really happy about it either :/ I already removed
some parts of it to reduce it, but I'll try to be more aggressive.

> On Thu, 28 May 2020 at 08:47, Maxime Ripard <maxime@cerno.tech> wrote:
> 
> > +static int vc4_txp_enable_vblank(struct drm_crtc *crtc)
> > +{
> > +       return 0;
> > +}
> > +
> > +static void vc4_txp_disable_vblank(struct drm_crtc *crtc) {}
> > +
> Core should handle if these are NULL, so the stubs should not be needed.

I'm really not sure actually. In the general case, you'll want to have
vblank functions, since it doesn't really make sense to have a driver
without it.

We could argue that with writeback, it might be optional like being done
here, but you don't really know if you're going to use a writeback
connector in the CRTC at initialisation time, and all the other
writeback encoders actually seem to use a real CRTC that can output to a
real encoder, and therefore has some proper vblank handling too.

So yeah, it really looks like the check is valid, and that driver is
just the odd case.

Maxime