[Intel-gfx] [PATCH] drm/i915: Fix initial pipe underrun state tracking

[Daniel Vetter]

On Mon, Mar 24, 2014 at 10:22:59AM +0200, Jani Nikula wrote:
> On Mon, 24 Mar 2014, Daniel Vetter <daniel.vetter at ffwll.ch> wrote:
> > Since
> >
> > commit 5c673b60a9b3b23486f4eda75c72e91d31d26a2b
> > Author: Daniel Vetter <daniel.vetter at ffwll.ch>
> > Date:   Fri Mar 7 20:34:46 2014 +0100
> >
> >     drm/i915: Don't enable display error interrupts from the start
> >
> > we don't enable underrun interrupts any more at takeover time.
> > Unfortunately I've forgotten to also adjust the sw-side tracking.
> >
> > Since the code assumes that disabled pipes have underrun reporting
> > enabled set the disable flag on all pipes which are active at takeover
> > time. Without this underrun reporting wasn't enabled correctly on the
> > first modeset. Note that for fastboot this is another piece of state
> > that needs to be fixed up by enabling the underrung reporting after
> > watermarks have beend fixed up.
> >
> > On ivb/hsw an additional effect of this regression was that also all
> > cpu crc reporting stopped working since the master error interrupt it
> > shared across all pipes and sources.
> >
> > Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > Cc: Jani Nikula <jani.nikula at intel.com>
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76150
> > Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
> > ---
> >  drivers/gpu/drm/i915/intel_display.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index 7e4ea8d4e388..9b24ae4fb7bd 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -11502,6 +11502,14 @@ static void intel_sanitize_crtc(struct intel_crtc *crtc)
> >  			encoder->base.crtc = NULL;
> >  		}
> >  	}
> > +	if (crtc->active) {
> > +		/*
> > +		 * We start out with underrun reporting disabled to avoid races.
> > +		 * For correct bookkeeping mark this on active crtcs.
> > +		 */
> 
> Why only on active crtcs? The state remains in inactive crtcs. Won't the
> problem appear when such an crtc is actived?

Setting this for inactive crtcs breaks the code, which I've learned the
hard way ;-) The issue is that on ivb/hsw we can _only_ enable the err
interrupt if all crtcs allow it. But if some disabled crtc disallows it
that condition is never true until userspace has lit up all for crtcs at
least once. I've tried to capture this in the commit message, but seem to
have failed. Can you please suggest a rewording to clarify things?

> Quoth a comment in struct intel_crtc:
> 
> /* Access to these should be protected by dev_priv->irq_lock. */
> 
> Either you need to hold the lock or explain in a comment why it's not
> necessary.

At worst the interrupt handler runs concurrently (on platforms where we
don't yet hide fifo underruns from the start), also setting this to false
in the case of an underrun. What about:

"No protection against concurrent access is required - at wors a fifo
underrun happens which also sets this to false."

Thanks, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch