[Intel-gfx] [3.14.0-rc4] regression: drm FIFO underruns

[Ville Syrjälä]

On Tue, May 13, 2014 at 06:38:32PM +0200, Daniel Vetter wrote:
> On Tue, May 13, 2014 at 05:21:49PM +0200, Jörg Otte wrote:
> > 2014-05-13 15:22 GMT+02:00 Daniel Vetter <daniel at ffwll.ch>:
> > > On Tue, May 13, 2014 at 12:38:41PM +0200, Daniel Vetter wrote:
> > >> On Tue, May 13, 2014 at 12:29 PM, Jörg Otte <jrg.otte at gmail.com> wrote:
> > >> >>> Branch drm-intel-nightly as of
> > >> >>> ed60c27 drm-intel-nightly: 2014y-05m-09d-21h-51m-45s integration manifest
> > >> >>> looks badly:
> > >> >>>    - KDE splash screen on boot-up is not visible
> > >> >>>    - x-windows don't have title and menu bars
> > >> >>>    - KDE system menu is not visible
> > >> >>>    - moving windows around destroys its content
> > >> >>
> > >> >> Ugh, that's ugly. Nothing else change like e.g. the version of
> > >> >> xfree-video-intel?
> > >> >
> > >> >  (II) Loading /usr/lib/xorg/modules/drivers/intel_drv.so
> > >> >  (II) Module intel: vendor="X.Org Foundation"
> > >> >  compiled for 1.11.3, module version = 2.17.0
> > >> >  Module class: X.Org Video Driver
> > >> >  ABI class: X.Org Video Driver, version 11.0
> > >>
> > >> Chris, any ideas? It's an ivybridge apparently.
> > >>
> > >> For the fifo underruns I think we've fully confirmed that they only
> > >> happen on boot-up. I'll try to come up with some ideas what could have
> > >> gone wrong there.
> > >
> > > Please test the below patch.
> > > -Daniel
> > >
> > > diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> > > index b10fbde1d5ee..63ced2dee027 100644
> > > --- a/drivers/gpu/drm/i915/i915_irq.c
> > > +++ b/drivers/gpu/drm/i915/i915_irq.c
> > > @@ -427,9 +427,6 @@ bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev,
> > >
> > >         ret = !intel_crtc->cpu_fifo_underrun_disabled;
> > >
> > > -       if (enable == ret)
> > > -               goto done;
> > > -
> > >         intel_crtc->cpu_fifo_underrun_disabled = !enable;
> > >
> > >         if (enable && (INTEL_INFO(dev)->gen < 5 || IS_VALLEYVIEW(dev)))
> > > @@ -441,7 +438,6 @@ bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev,
> > >         else if (IS_GEN8(dev))
> > >                 broadwell_set_fifo_underrun_reporting(dev, pipe, enable);
> > >
> > > -done:
> > >         return ret;
> > >  }
> > >
> > > --
> > 
> > Doesn't work for me, I still have an underrun at boot-up.
> 
> I'm at a loss tbh with ideas. We successfully disable both pipes, then
> enable pipe A and it all works.
> 
> Then we enable pipe B and _both_ pipes underrun immediately afterwards.
> Really strange. Can you please reproduce the issue again on
> drm-intel-nightly (latest -nightly should also have the display
> corruptions fixed, so good to retest anyway) and attach a new dmesg with
> drm.debug=0xe.

I see a few underrun on my IVB as well. But it seems to be limited to
cases that involve the VGA connector, which doesn't actually exist on
this machine so I can't be sure if it's really properly set up on the
PCH. But so far with just two HDMI connectors I was unable to reproduce
it.

> 
> Meanwhile I'll try to come up with new theories and ideas.

I was thinking that we might frob with the PCH refclk during driver init
and that might cause the PCH underrun for Jörg, but it looks like the
underruns really happen at the modeset time which is much later than the
PCH refclock init.

For the 1<->n pipe transition, I don't think we handle it correctly at
the moment. I have a fix as part of my remaining watermark patches. I
rebased those and I'll repost them soon. In the meantime I pushed them
to [1]. Jörg, can you give that branch a go?

[1] git://gitorious.org/vsyrjala/linux.git watermarks_intm_31_notrace

-- 
Ville Syrjälä
Intel OTC