Some more VC4 vblank fixes and timestamping improvements.

[Daniel Vetter]

On Thu, Jul 21, 2016 at 12:58:16PM -0700, Eric Anholt wrote:
> Mario Kleiner <mario.kleiner.de at gmail.com> writes:
> 
> > Hi Eric and all,
> >
> > this batch prevents setting modes one shouldn't set,
> > adds precise vblank timestamping for interlaced video
> > modes, and one fix for vblank en/disable during crtc
> > en/disable.
> >
> > All successfully tested, also with timing measurement equipment,
> > on the RPi 2B.
> 
> This all seems pretty reasonable to me.  My question for modesetting
> gurus would be: Is mode_fixup the place to be filtering out
> unsupportable modes from the user (or their displays)?

Yup. And I wanted to forever add some glue so that drivers could share the
same validation logic between mode_fixup (for modesets) and mode_valid
(for probing). But never came up with a helper scaffolding that looked
satisfactory. The icky bit is that mode_fixup wants an adjusted mode, and
we'd need to fake that somehow to reuse the same code in mode_valid. It's
even worse for atomic, where atomic_check (the new mode_fixup) gets an
entire atomic state. But I still think for simpler hw (anything which
doesn't have funny crtc/encoder specific restrictions) we should be able
to create something that just works.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch