[Intel-gfx] [PATCH] drm/i915: Fix eDP blank screen after S3 resume on HP desktops

[Daniel Vetter]

On Wed, Jun 20, 2012 at 11:21:11AM +0200, Takashi Iwai wrote:
> At Wed, 20 Jun 2012 10:05:12 +0200,
> Daniel Vetter wrote:
> > 
> > On Wed, Jun 20, 2012 at 09:17:41AM +0200, Takashi Iwai wrote:
> > > This patch fixes the problem on some HP desktop machines with eDP
> > > which give blank screens after S3 resume.
> > > 
> > > The problem looks like a timing issue.  Although BLC_PWM_CPU_CTL
> > > register is already restored at the beginning of resume, it doesn't
> > > seem to take effect.  Simply re-issuing the  register write restores
> > > the backlight gracefully.
> > > 
> > > Tested with 3.5-rc3 kernel.
> > > 
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=49233
> > > 
> > > Signed-off-by: Takashi Iwai <tiwai at suse.de>
> > 
> > This patch looks very fishy as-is, simply because I don't like adding
> > random calls to random functions at pretty much random places without any
> > hint why it works.
> 
> It's difficult to know exactly why it works.  We can only guess, but
> nothing more than that, just like between husband and wife :)

Yeah I know, but I've been burnt a bit recently with too much
cargo-culting. Hence why I'm so dense about this stuff ;-)

> > Just a few weeks ago we've had tons of fun with writes
> > that don't stick when the ring isn't yet set up. Can you please check with
> > a bunch of printks what exactly happens to the value of BLC_PWM_CPU_CTL?
> > I.e. whether the write doesn't stick or whether it gets reset somewhere
> > between the register restore functions and the new place.
> 
> The register value is already set, and not touched after that.
> I've already checked.

So immediately before the backlight_update_status call we have the
_correct_ value in BLC_PWM_CPU_CTL? That would be indeed puzzling ...

> > I suspect the opregion_init call could allow the bios to frob with these
> > registers. If that's the case, the patch is missing a comment that to that
> > effect.
> 
> Right.  That's also my suspect.
> 
> > Essentially I want to know why this place here works and why not move the
> > call a few lines up or down.
> 
> A few lines down works definitely.  Even just writing to sysfs works.
> A few lines up, well, I need to check where is the border line.

Thanks, that would be interesting. Especially if it's opregion that we
need to have set up before restoring the backlight properly.
-Daniel

> 
> 
> Takashi
> 
> > -Daniel
> > 
> > > ---
> > >  drivers/gpu/drm/i915/i915_drv.c |    3 +++
> > >  1 file changed, 3 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> > > index 9fe9ebe..2ba1350 100644
> > > --- a/drivers/gpu/drm/i915/i915_drv.c
> > > +++ b/drivers/gpu/drm/i915/i915_drv.c
> > > @@ -658,6 +658,9 @@ static int i915_drm_thaw(struct drm_device *dev)
> > >  
> > >  	intel_opregion_init(dev);
> > >  
> > > +	if (dev_priv->backlight)
> > > +		backlight_update_status(dev_priv->backlight);
> > > +
> > >  	dev_priv->modeset_on_lid = 0;
> > >  
> > >  	console_lock();
> > > -- 
> > > 1.7.10.4
> > > 
> > > _______________________________________________
> > > Intel-gfx mailing list
> > > Intel-gfx at lists.freedesktop.org
> > > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> > 
> > -- 
> > Daniel Vetter
> > Mail: daniel at ffwll.ch
> > Mobile: +41 (0)79 365 57 48
> > 

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