[Intel-gfx] [RFC PATCH 2/2] drm/i915: respect the VBT minimum backlight brightness

[Jesse Barnes]

On Wed, 4 Jun 2014 08:04:53 -0700
Stéphane Marchesin <marcheu at chromium.org> wrote:

> > We could have dpms standby mean backlight off, everything else on,
> > similarly to what you want by 0 backlight meaning off. So you could
> > switch between dpms on/standby more quickly. The point in that is that
> > it's the right API for doing panel power sequencing.
> >
> > You see, even with defining backlight 0 = off, we'd have to respect the
> > panel sequencing and delays for backlight (admittedly we currently don't
> > do that, which is why it's fast - and broken).  
> 
> There are multiple delays at play here, and typically the longest
> panel delays have nothing to do with the backlight but they are the
> delays for turning the panel on/off (usually the panel on/off delays
> are much bigger, up to 500ms on some panels). These are the delays I'm
> trying to avoid.
> 
> > Doing the panel
> > sequencing properly from the backlight interface gets *much* harder
> > because it's all backwards in the stack then.  
> 
> If you implement some interface which does "set backlight to 0 on
> platforms which support it, and do panel off on other platforms",
> that's fine by me; however I don't want to regress all the other
> platforms just because BYT has issues. In the current state of things,
> the functionality set which is made available to user space happens to
> shrink with your patch.

According to the LVDS and eDP timing specs, we *ought* to be able to
disable the backlight and then the PWM source and leave everything else
alone, as long as we apply the appropriate delays between backlight off
and PWM off, and then again when we turn them back on (PWM first,
delay, then backlight on).

IIRC the issue on BYT was that we saw the panel power line go down when
we disabled the backlight and PWM which would obviously cause all
sorts of problems.  But that could have been user error on the scope,
so we should confirm that before baking a full DPMS cycle into the BYT
implementation or the interface in general.

-- 
Jesse Barnes, Intel Open Source Technology Center