[Intel-gfx] [PATCH 2/2] drm/i915: quirk disable i915 backlight on Dell XPS 13z

[Daniel Vetter]

On Fri, Apr 27, 2012 at 01:56:09PM -0700, Kamal Mostafa wrote:
> On Thu, 2012-04-26 at 22:07 +0200, Daniel Vetter wrote:
> > ...
> > To cut things short: This time around I want more justification for the
> > quirk than just "this makes this one machine work somehow".
> 
> 
> A bit more detail...
> 
> On this Dell XPS 13z "Ultrabook" (sandybridge_m 0x0116) when
> intel_backlight/brightness gets stuffed with any value except 0 or
> max_brightness, the backlight cycles between flashing (like a strobe
> light!) and then pulsating (bright to dim to bright).  That
> flashing/pulsating cycle repeats continuously, about every 2 seconds.
> The behavior is affected by the value stuffed into brightness more or
> less along the lines of:
> 
>    0: very dim, totally stable
>    1: flashes like crazy for about 1.5 sec, then pulsates for 0.5 sec
> 1000: flashes for about 0.5 sec, then pulsates for about 1.5 sec
> 2000: flashes very briefly, then pulsates for about 2 sec
> 3000: flickers, then pulsates for about 2 sec
> 4000  pulsates continuously, every 2 sec
> 4882: (max_brightness) full brightness, stable
> 
> This behavior manifests both in X and in a text VT, and occurs with or
> without the presence of other backlight interfaces besides
> intel_backlight.  It does not appear to me to be a userspace problem.
> 
> The additional wrinkle is that this machine presents an acpi_video0
> backlight interface as well, and it even works properly -- but *only*
> after you specifically stuff 0 into intel_backlight/brightness (or if
> intel_backlight is disabled by the proposed quirk).
> 
> Any non-zero intel_backlight/brightness value prevents acpi_video0 from
> working.  When intel_backlight is set to max_brightness (like at boot),
> acpi_video0/brightness seems to have no effect at all; when
> intel_backlight is set to other non-zero values, the flashing/pulsating
> behavior occurs.
> 
> I'd be (quite) happy to test a proper intel_backlight fix, but in the
> meantime disabling it by quirk seems appropriate for this machine, since
> that allows the acpi_video0 interface to work out of the box.
> 
> Thanks for considering it,

Hm. Can you please install intel-gpu-tools and attach the output of
intel_reg_dumper? I have an idea ...

Thanks, Daniel
-- 
Daniel Vetter
Mail: daniel at ffwll.ch
Mobile: +41 (0)79 365 57 48