[PATCH] drm/imx: imx-ldb: do not try to dereference crtc->state->state in encoder mode_set

[Daniel Vetter]

On Mon, Jul 25, 2016 at 12:42:50PM +0200, Philipp Zabel wrote:
> Am Freitag, den 22.07.2016, 18:49 +0200 schrieb Daniel Vetter:
> > On Fri, Jul 22, 2016 at 12:57:15PM +0200, Philipp Zabel wrote:
> > > Am Freitag, den 22.07.2016, 11:35 +0200 schrieb Daniel Vetter:
> > > [...]
> > > > Proper fix would be to roll out atomic_ versions of all teh encoder
> > > > callbacks where we additionally pass both the crtc state and the connector
> > > > state. Then there's no need for walking connector lists like that. And in
> > > > the atomic helpers those two states are always readily available, and
> > > > passing them down to callbacks is also what we will do in i915. I'd be
> > > > happy to merge such a patch.
> > > 
> > > Thanks, that is a good idea. Which encoder callbacks besides mode_set ->
> > > atomic_mode_set are you thinking of, though? enable/disable?
> > 
> > Yes. The only other one is mode_fixup, and we already have atomic_check as
> > the fancy replacement for that. Please align the parameter ordering with
> > encoder->atomic_check for consistency.
> 
> Which crtc state should be passed to encoder atomic_disable?
> old_crtc_state?

Yes.

> > Also when you do that patch, pls don't forget the vtable kerneldoc
> > comments, and make sure there's no warnings/issues and it renders nicely
> > using
> > 
> > $ make htmldocs
> > 
> > Since 4.8 will feature a new sphinx-based toolchain, pls run that command
> > on top of linux-next (for the latest sphinx fixes from docs-next).
> > -Daniel
> 
> Ok, I'll have a look.

Otoh, if there's not yet a need then maybe it doesn't yet make sense to
add the full set of hooks. Might be worth it to skim driver's encoder
callbacks for connector loops and similar things.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch