[Intel-gfx] [PATCH 1/3] drm/i915: Don't check modeset state in the hw state force restore path

[Ander Conselvan De Oliveira]

On Mon, 2015-06-15 at 19:00 -0700, Matt Roper wrote:
> On Fri, Jun 12, 2015 at 10:19:40AM +0300, Ander Conselvan de Oliveira wrote:
> > Since the force restore logic will restore the CRTCs state one at a
> > time, it is possible that the state will be inconsistent until the whole
> > operation finishes. A call to intel_modeset_check_state() is done once
> > it's over, so don't check the state multiple times in between. This
> > regression was introduced in:
> > 
> > commit 7f27126ea3db6ade886f18fd39caf0ff0cd1d37f
> > Author: Jesse Barnes <jbarnes at virtuousgeek.org>
> > Date:   Wed Nov 5 14:26:06 2014 -0800
> > 
> >     drm/i915: factor out compute_config from __intel_set_mode v3
> > 
> > Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=94431
> > Cc: Jesse Barnes <jbarnes at virtuousgeek.org>
> > Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira at intel.com>
> > ---
> >  drivers/gpu/drm/i915/intel_display.c | 21 ++++++++++++---------
> >  1 file changed, 12 insertions(+), 9 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index 4e3f302..6ef57e6 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -87,7 +87,8 @@ static void ironlake_pch_clock_get(struct intel_crtc *crtc,
> >  				   struct intel_crtc_state *pipe_config);
> >  
> >  static int intel_set_mode(struct drm_crtc *crtc,
> > -			  struct drm_atomic_state *state);
> > +			  struct drm_atomic_state *state,
> > +			  bool check);
> >  static int intel_framebuffer_init(struct drm_device *dev,
> >  				  struct intel_framebuffer *ifb,
> >  				  struct drm_mode_fb_cmd2 *mode_cmd,
> > @@ -10096,7 +10097,7 @@ retry:
> >  
> >  	drm_mode_copy(&crtc_state->base.mode, mode);
> >  
> > -	if (intel_set_mode(crtc, state)) {
> > +	if (intel_set_mode(crtc, state, true)) {
> >  		DRM_DEBUG_KMS("failed to set mode on load-detect pipe\n");
> >  		if (old->release_fb)
> >  			old->release_fb->funcs->destroy(old->release_fb);
> > @@ -10170,7 +10171,7 @@ void intel_release_load_detect_pipe(struct drm_connector *connector,
> >  		if (ret)
> >  			goto fail;
> >  
> > -		ret = intel_set_mode(crtc, state);
> > +		ret = intel_set_mode(crtc, state, true);
> >  		if (ret)
> >  			goto fail;
> >  
> > @@ -12646,20 +12647,22 @@ static int __intel_set_mode(struct drm_crtc *modeset_crtc,
> >  }
> >  
> >  static int intel_set_mode_with_config(struct drm_crtc *crtc,
> > -				      struct intel_crtc_state *pipe_config)
> > +				      struct intel_crtc_state *pipe_config,
> > +				      bool check)
> 
> This parameter just controls whether you check the state or not in this
> patch, but in patch #2 it also starts having more of a behavioral impact
> (i.e., "don't updated staged output configuration").  I wonder if
> picking a different name for this parameter would help avoid any
> confusion?

Thanks for reviewing! I just sent a v2 with the parameter name changed
to force_restore and the memory leak fixed.

> Otherwise, this patch looks good:
>    Reviewed-by: Matt Roper <matthew.d.roper at intel.com>
> 
> Off-topic, but speaking of 'check' I wonder whether we should also
> rename the intel_modeset_check_state function and related functions at
> some point in the future.  Every time I see that name in the code it
> bothers me because I expect it to be related to atomic check (i.e.,
> something that runs before we touch hardware) rather than confirming
> that hardware programming was successful.  It feels more like an
> assert/confirm/verify function to me.

I'm not sure what would be a better name, but I agree the current one is
confusing.

Ander

> 
> Matt
> 
> >  {
> >  	int ret;
> >  
> >  	ret = __intel_set_mode(crtc, pipe_config);
> >  
> > -	if (ret == 0)
> > +	if (ret == 0 && check)
> >  		intel_modeset_check_state(crtc->dev);
> >  
> >  	return ret;
> >  }
> >  
> >  static int intel_set_mode(struct drm_crtc *crtc,
> > -			  struct drm_atomic_state *state)
> > +			  struct drm_atomic_state *state,
> > +			  bool check)
> >  {
> >  	struct intel_crtc_state *pipe_config;
> >  	int ret = 0;
> > @@ -12670,7 +12673,7 @@ static int intel_set_mode(struct drm_crtc *crtc,
> >  		goto out;
> >  	}
> >  
> > -	ret = intel_set_mode_with_config(crtc, pipe_config);
> > +	ret = intel_set_mode_with_config(crtc, pipe_config, check);
> >  	if (ret)
> >  		goto out;
> >  
> > @@ -12747,7 +12750,7 @@ void intel_crtc_restore_mode(struct drm_crtc *crtc)
> >  	intel_modeset_setup_plane_state(state, crtc, &crtc->mode,
> >  					crtc->primary->fb, crtc->x, crtc->y);
> >  
> > -	ret = intel_set_mode(crtc, state);
> > +	ret = intel_set_mode(crtc, state, false);
> >  	if (ret)
> >  		drm_atomic_state_free(state);
> >  }
> > @@ -12947,7 +12950,7 @@ static int intel_crtc_set_config(struct drm_mode_set *set)
> >  
> >  	primary_plane_was_visible = primary_plane_visible(set->crtc);
> >  
> > -	ret = intel_set_mode_with_config(set->crtc, pipe_config);
> > +	ret = intel_set_mode_with_config(set->crtc, pipe_config, true);
> >  
> >  	if (ret == 0 &&
> >  	    pipe_config->base.enable &&
> > -- 
> > 2.1.0
> > 
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>