[Intel-gfx] [RFC 2/6] drm/i915: Remove (pipe == crtc->index) asummption

[Ville Syrjälä]

On Thu, Jan 30, 2020 at 03:35:20PM +0200, Ville Syrjälä wrote:
> On Thu, Jan 30, 2020 at 05:32:01PM +0530, Anshuman Gupta wrote:
> > On 2020-01-23 at 15:40:57 +0200, Ville Syrjälä wrote:
> > > On Thu, Jan 23, 2020 at 06:56:55PM +0530, Anshuman Gupta wrote:
> > > > we can't have (pipe == crtc->index) assumption in
> > > > driver in order to support 3 non-contiguous
> > > > display pipe system.
> > > > 
> > > > Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > > Signed-off-by: Anshuman Gupta <anshuman.gupta at intel.com>
> > > > ---
> > > >  drivers/gpu/drm/i915/display/intel_display.c | 10 ++++------
> > > >  1 file changed, 4 insertions(+), 6 deletions(-)
> > > > 
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> > > > index 878d331b9e8c..afd8d43160c6 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > > @@ -14070,11 +14070,11 @@ verify_single_dpll_state(struct drm_i915_private *dev_priv,
> > > >  	if (new_crtc_state->hw.active)
> > > >  		I915_STATE_WARN(!(pll->active_mask & crtc_mask),
> > > >  				"pll active mismatch (expected pipe %c in active mask 0x%02x)\n",
> > > > -				pipe_name(drm_crtc_index(&crtc->base)), pll->active_mask);
> > > > +				pipe_name(crtc->pipe), pll->active_mask);
> > > >  	else
> > > >  		I915_STATE_WARN(pll->active_mask & crtc_mask,
> > > >  				"pll active mismatch (didn't expect pipe %c in active mask 0x%02x)\n",
> > > > -				pipe_name(drm_crtc_index(&crtc->base)), pll->active_mask);
> > > > +				pipe_name(crtc->pipe), pll->active_mask);
> > > >  
> > > >  	I915_STATE_WARN(!(pll->state.crtc_mask & crtc_mask),
> > > >  			"pll enabled crtcs mismatch (expected 0x%x in 0x%02x)\n",
> > > > @@ -14103,10 +14103,10 @@ verify_shared_dpll_state(struct intel_crtc *crtc,
> > > >  
> > > >  		I915_STATE_WARN(pll->active_mask & crtc_mask,
> > > >  				"pll active mismatch (didn't expect pipe %c in active mask)\n",
> > > > -				pipe_name(drm_crtc_index(&crtc->base)));
> > > > +				pipe_name(crtc->pipe));
> > > >  		I915_STATE_WARN(pll->state.crtc_mask & crtc_mask,
> > > >  				"pll enabled crtcs mismatch (found %x in enabled mask)\n",
> > > > -				pipe_name(drm_crtc_index(&crtc->base)));
> > > > +				pipe_name(crtc->pipe));
> > > >  	}
> > > >  }
> > > >  
> > > > @@ -16485,8 +16485,6 @@ static int intel_crtc_init(struct drm_i915_private *dev_priv, enum pipe pipe)
> > > >  
> > > >  	intel_color_init(crtc);
> > > >  
> > > > -	WARN_ON(drm_crtc_index(&crtc->base) != crtc->pipe);
> > > > -
> > > 
> > > The first and second hunks don't really have anything to do with
> > > each other. Also the WARN_ON() should not be removed until all the
> > > assumptions are fixed.
> > True there can be other assumptions as well, there are few, i have come to know
> > drm_handle_vblank(&dev_priv->drm, pipe) in gen8_de_irq_handler()
> 
> In fact it's in all irq handlers.
> 
> > drm_wait_one_vblank(&dev_priv->drm, pipe) in intel_wait_for_vblank(),
> 
> Good catch. Totally forgot about these.
> 
> > i will fix these assumptions is next update, are there any other similar kind of
> > assumption on which u can throw some light to look for?
> > I am not sure how does above WARN_ON helps to know all such kind of 
> > assumptions, but it make sense to have it with FIXME.
> 
> It doesn't help finding them, what it does is make people realize
> that they're running a driver which is known to be broken.

Just remembered another borked thing: trans_offsets[]. Some places use
that to check if the transcoder is present, and we don't take fusing
into account when filling that. Though looks like
intel_display_capture_error_state() is the only place where can
actually do the wrong thing (assuming EDP/DSI transcoders are never
fused off).

-- 
Ville Syrjälä
Intel