[Intel-gfx] [PATCH 10/23] drm/i915: Add gamma correction handlers

[Daniel Vetter]

On Wed, Sep 23, 2015 at 01:52:21PM +0530, Sharma, Shashank wrote:
> Regards
> Shashank
> 
> On 9/22/2015 6:45 PM, Daniel Vetter wrote:
> >On Wed, Sep 16, 2015 at 11:07:07PM +0530, Shashank Sharma wrote:
> >>I915 driver registers gamma correction as palette correction
> >>property with DRM layer. This patch adds set_property() and get_property()
> >>handlers for pipe level gamma correction.
> >>
> >>The set function attaches the Gamma correction blob to CRTC state, these
> >>values will be committed during atomic commit.
> >>
> >>Signed-off-by: Shashank Sharma <shashank.sharma at intel.com>
> >>Signed-off-by: Kausal Malladi <kausalmalladi at gmail.com>
> >>---
> >>  drivers/gpu/drm/i915/intel_atomic.c        | 20 ++++++++++++++++++++
> >>  drivers/gpu/drm/i915/intel_color_manager.c | 21 +++++++++++++++++++++
> >>  drivers/gpu/drm/i915/intel_drv.h           |  5 +++++
> >>  3 files changed, 46 insertions(+)
> >>
> >>diff --git a/drivers/gpu/drm/i915/intel_atomic.c b/drivers/gpu/drm/i915/intel_atomic.c
> >>index 500d2998..0b61fef 100644
> >>--- a/drivers/gpu/drm/i915/intel_atomic.c
> >>+++ b/drivers/gpu/drm/i915/intel_atomic.c
> >>@@ -315,6 +315,13 @@ int intel_crtc_atomic_set_property(struct drm_crtc *crtc,
> >>  				   struct drm_property *property,
> >>  				   uint64_t val)
> >>  {
> >>+	struct drm_device *dev = crtc->dev;
> >>+	struct drm_mode_config *config = &dev->mode_config;
> >>+
> >>+	if (property == config->cm_palette_after_ctm_property)
> >>+		return intel_color_manager_set_pipe_gamma(dev, state,
> >>+				&crtc->base, val);
> >>+
> >>  	DRM_DEBUG_KMS("Unknown crtc property '%s'\n", property->name);
> >>  	return -EINVAL;
> >>  }
> >>@@ -324,6 +331,19 @@ int intel_crtc_atomic_get_property(struct drm_crtc *crtc,
> >>  				   struct drm_property *property,
> >>  				   uint64_t *val)
> >>  {
> >>+	struct drm_device *dev = crtc->dev;
> >>+	struct drm_mode_config *config = &dev->mode_config;
> >>+
> >>+	if (property == config->cm_palette_after_ctm_property) {
> >>+		*val = (state->palette_after_ctm_blob) ?
> >>+			state->palette_after_ctm_blob->base.id : 0;
> >>+		goto found;
> >>+	}
> >
> >Since color manager properties are meant as a new standardize KMS
> >extension (we put them into the core drm_crtc_state) the get/set support
> >should also be in the core. See e.g. how the rotation property is handled
> >in drm_atomic_plane_get/set_property. So all this code should be added to
> >drm_atomic_crtc_get/set_property.
> Thanks, sounds like a good one. Will move this.

When moving this please don't forget to add the blob->length sanity checks
mentioned in another thread in this discussion.

> >
> >
> >>+
> >>  	DRM_DEBUG_KMS("Unknown crtc property '%s'\n", property->name);
> >>  	return -EINVAL;
> >>+
> >>+found:
> >>+	DRM_DEBUG_KMS("Found property %s\n", property->name);
> >>+	return 0;
> >>  }
> >>diff --git a/drivers/gpu/drm/i915/intel_color_manager.c b/drivers/gpu/drm/i915/intel_color_manager.c
> >>index 77f58f2..9421bb6 100644
> >>--- a/drivers/gpu/drm/i915/intel_color_manager.c
> >>+++ b/drivers/gpu/drm/i915/intel_color_manager.c
> >>@@ -27,6 +27,27 @@
> >>
> >>  #include "intel_color_manager.h"
> >>
> >>+int intel_color_manager_set_pipe_gamma(struct drm_device *dev,
> >>+		struct drm_crtc_state *crtc_state,
> >>+		struct drm_mode_object *obj, uint32_t blob_id)
> >>+{
> >>+	struct drm_property_blob *blob;
> >>+
> >>+	blob = drm_property_lookup_blob(dev, blob_id);
> >>+	if (!blob) {
> >>+		DRM_DEBUG_KMS("Invalid Blob ID\n");
> >>+		return -EINVAL;
> >>+	}
> >>+
> >>+	if (crtc_state->palette_after_ctm_blob)
> >>+		drm_property_unreference_blob(
> >>+			crtc_state->palette_after_ctm_blob);
> >>+
> >>+	/* Attach the blob to be committed in state */
> >>+	crtc_state->palette_after_ctm_blob = blob;
> >>+	return 0;
> >>+}
> >
> >What is this used for? It looks a bit like legacy property code, and we
> >have a generic helper to make that happen
> >(drm_atomic_helper_crtc_set_property).
> >
> No Daniel, its not the legacy part. Please check atomic_begin() part in the
> next patches. This is like the top half of atomic set_property for CRTC and
> will be applicable for any property. As you already know, to set a property:
> - userspace creates a blob, and sends the blob id via set_property()
> interface
> - set_property() saves the blob_id in the corresponding pointer space in the
> crtc_state()
> - From the atomic_commit_begin() we will extract the values from blob and
> start committing to the registers. Its kind of bottom_half() of atomic
> set_property for CRTC.
> >Generally please don't add functions/structs without also adding a user,
> >it means that reviewers have to constantly jump around in your patch
> >series to figure out how something is used. Instead if you want to split
> >things up really fine add a stub function frist (but including relevant
> >callers) and then fill out the bits separately.
> I think the above explanation should make it clear.

In my defence I'd like to say that I was still jetlagged ;-) Reading
things now again it all looks good.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch