Re: [PATCH v2 03/10] drm/i915: move checks of intel_crtc_cursor_set_obj() out

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Tue, Sep 23, 2014 at 12:41:56PM -0300, Gustavo Padovan wrote:
> 2014-09-23 Ville Syrjälä <ville.syrjala@linux.intel.com>:
> 
> > On Mon, Sep 22, 2014 at 07:23:10PM -0300, Gustavo Padovan wrote:
> > > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> > > 
> > > Move checks inside intel_crtc_cursor_set_obj() to
> > > intel_check_cursor_plane(), we only use they there so move them out to
> > > make the merge of intel_crtc_cursor_set_obj() into
> > > intel_check_cursor_plane() easier.
> > > 
> > > This is another step toward the atomic modesetting support and unification
> > > of plane operations such pin/unpin of fb objects on i915.
> > > 
> > > v2: take Ville's comment: move crtc_{w,h} assignment a bit down in the
> > > code
> > > 
> > > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> > > ---
> > >  drivers/gpu/drm/i915/intel_display.c | 61 ++++++++++++++++++++++++------------
> > >  1 file changed, 41 insertions(+), 20 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > > index 2ef1836..3f37e93 100644
> > > --- a/drivers/gpu/drm/i915/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/intel_display.c
> > > @@ -8373,7 +8373,7 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc,
> > >  	struct drm_i915_private *dev_priv = dev->dev_private;
> > >  	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
> > >  	enum pipe pipe = intel_crtc->pipe;
> > > -	unsigned old_width, stride;
> > > +	unsigned old_width;
> > >  	uint32_t addr;
> > >  	int ret;
> > >  
> > > @@ -8385,29 +8385,11 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc,
> > >  		goto finish;
> > >  	}
> > >  
> > > -	/* Check for which cursor types we support */
> > > -	if (!cursor_size_ok(dev, width, height)) {
> > > -		DRM_DEBUG("Cursor dimension not supported\n");
> > > -		return -EINVAL;
> > > -	}
> > > -
> > > -	stride = roundup_pow_of_two(width) * 4;
> > > -	if (obj->base.size < stride * height) {
> > > -		DRM_DEBUG_KMS("buffer is too small\n");
> > > -		return -ENOMEM;
> > > -	}
> > > -
> > >  	/* we only need to pin inside GTT if cursor is non-phy */
> > >  	mutex_lock(&dev->struct_mutex);
> > >  	if (!INTEL_INFO(dev)->cursor_needs_physical) {
> > >  		unsigned alignment;
> > >  
> > > -		if (obj->tiling_mode) {
> > > -			DRM_DEBUG_KMS("cursor cannot be tiled\n");
> > > -			ret = -EINVAL;
> > > -			goto fail_locked;
> > > -		}
> > > -
> > >  		/*
> > >  		 * Global gtt pte registers are special registers which actually
> > >  		 * forward writes to a chunk of system memory. Which means that
> > > @@ -11826,16 +11808,55 @@ intel_check_cursor_plane(struct drm_plane *plane,
> > >  			 struct intel_plane_state *state)
> > >  {
> > >  	struct drm_crtc *crtc = state->crtc;
> > > +	struct drm_device *dev = crtc->dev;
> > >  	struct drm_framebuffer *fb = state->fb;
> > >  	struct drm_rect *dest = &state->dst;
> > >  	struct drm_rect *src = &state->src;
> > >  	const struct drm_rect *clip = &state->clip;
> > > +	struct drm_i915_gem_object *obj = intel_fb_obj(fb);
> > > +	int crtc_w, crtc_h;
> > > +	unsigned stride;
> > > +	int ret;
> > >  
> > > -	return drm_plane_helper_check_update(plane, crtc, fb,
> > > +	ret = drm_plane_helper_check_update(plane, crtc, fb,
> > >  					    src, dest, clip,
> > >  					    DRM_PLANE_HELPER_NO_SCALING,
> > >  					    DRM_PLANE_HELPER_NO_SCALING,
> > >  					    true, true, &state->visible);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +
> > > +	/* if we want to turn off the cursor ignore width and height */
> > > +	if (!obj)
> > > +		return 0;
> > > +
> > > +	if (fb == crtc->cursor->fb)
> > > +		return 0;
> > 
> > Hmm. This check needs to be after the cursor/obj size checks. Otherwise
> > we wouldn't reject invalid sized cursors when the fb didn't change. I
> > suppose we could also just drop this check, but it can still save us from
> > doing the tiling check since that can't have changed if the fb hasn't
> > changed so maybe it's worth keeping. But either solution is fine by me.
> 
> Only if this was a bug before this patch, if you look at the original
> intel_commit_cursor_plane() we do two things when the fbs are the same:
>                
> 	intel_crtc_update_cursor();
> 	intel_frontbuffer_flip();
> 
> No checks are performed, that is why I put the fbs check before any other
> check to make the code as similar as possible to what it was before.

I guess we already had the bug then. Not entirely unsurprising given the
recent history of this code. Probably better to have the fix as a separate
patch then to avoid mixing restructuring and functional changes in the
same patch.

> 
> > > +
> > > +	/* Check for which cursor types we support */
> > > +	crtc_w = drm_rect_width(&state->orig_dst);
> > > +	crtc_h = drm_rect_height(&state->orig_dst);
> > > +	if (!cursor_size_ok(dev, crtc_w, crtc_h)) {
> > > +		DRM_DEBUG("Cursor dimension not supported\n");
> > > +		return -EINVAL;
> > > +	}
> > > +
> > > +	stride = roundup_pow_of_two(crtc_w) * 4;
> > > +	if (obj->base.size < stride * crtc_h) {
> > > +		DRM_DEBUG_KMS("buffer is too small\n");
> > > +		return -ENOMEM;
> > > +	}
> > > +
> > > +	/* we only need to pin inside GTT if cursor is non-phy */
> > > +	mutex_lock(&dev->struct_mutex);
> > > +	if (!INTEL_INFO(dev)->cursor_needs_physical && obj->tiling_mode) {
> > > +		DRM_DEBUG_KMS("cursor cannot be tiled\n");
> > > +		ret = -EINVAL;
> > > +	}
> > > +	mutex_unlock(&dev->struct_mutex);
> > > +
> > > +	return ret;
> > 
> > 'return 0' seems better here.
> 
> ret can be -EINVAL, the code inside the mutex 3 lines above can set ret.

Oh, somehow I missed that one. You can leave the patch as is then.

> 
> 	Gustavo

-- 
Ville Syrjälä
Intel OTC