Skip to contentSkip to conversation

plane update with damage

PATCH In review Deepak Rawat 6 participants 37 messages Sep 5, 2018 – Sep 11, 2018

Subsystems (4)
lore
Share & tools
lore
#1 lore raw

PATCH 1/14 drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Deepak Rawat

From: Lukasz Spintzyk <[email protected]>

FB_DAMAGE_CLIPS is an optional plane property to mark damaged regions on the plane in framebuffer coordinates of the framebuffer attached to the plane.

The layout of blob data is simply an array of "struct drm_mode_rect" with maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips are not in 16.16 fixed point. As plane src in framebuffer cannot be negative so are damage clips. In damage clip, x1/y1 are inclusive and x2/y2 are exclusive.

This patch also exports the kernel internal drm_rect to userspace as drm_mode_rect. This is because "struct drm_clip_rect" is not sufficient to represent damage for current plane size.

Upper limit is set on the maximum number of damage clips to avoid overflow by user-space.

Driver which are interested in enabling FB_DAMAGE_CLIPS property for a plane should enable this property using drm_plane_enable_damage_clips.

---
 Documentation/gpu/drm-kms.rst       |  9 +++
 drivers/gpu/drm/Makefile            |  2 +-
 drivers/gpu/drm/drm_atomic.c        | 13 ++++
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++
 drivers/gpu/drm/drm_damage_helper.c | 92 +++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_mode_config.c   |  6 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |  1 +
 include/drm/drm_damage_helper.h     | 63 ++++++++++++++++++++
 include/drm/drm_mode_config.h       | 10 ++++
 include/drm/drm_plane.h             |  8 +++
 include/uapi/drm/drm_mode.h         | 19 ++++++
 11 files changed, 226 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Documentation/gpu/drm-kms.rst +9 -0
index 5dee6b8a4c12..78b66e628857 100644
@@ -542,6 +542,15 @@ Plane Composition Properties
 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
    :export:
 
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:
+
 Color Management Properties
 ---------------------------
 
drivers/gpu/drm/Makefile +1 -1
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \
 		drm_plane_helper.o drm_dp_mst_topology.o drm_atomic_helper.o \
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o \
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o drm_damage_helper.o
 
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fb_helper.o
drivers/gpu/drm/drm_atomic.c +13 -0
index 3eb061e11e2e..652e78ca1f81 100644
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb = drm_framebuffer_lookup(dev, NULL, val);
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane *plane,
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state, property, val);
 	} else {
drivers/gpu/drm/drm_atomic_helper.c +4 -0
index 80be74df7ba6..be83e2763c18 100644
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct drm_plane *plane)
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property->values[1];
+		plane->state->fb_damage_clips = NULL;
 	}
 }
 EXPORT_SYMBOL(drm_atomic_helper_plane_reset);
@@ -3598,6 +3599,7 @@ void __drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,
 
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void __drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)
 
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
drivers/gpu/drm/drm_damage_helper.c new file +92 -0

92 lines not on this page · raw message

drivers/gpu/drm/drm_mode_config.c +6 -0
index 21e353bd3948..506f5a52733f 100644
@@ -298,6 +298,12 @@ static int drm_mode_create_standard_properties(struct drm_device *dev)
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB, "FB_DAMAGE_CLIPS",
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev, DRM_MODE_PROP_ATOMIC,
 			"ACTIVE");
 	if (!prop)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
index 23beff5d8e3c..1edbae73d6d6 100644
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane *plane)
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
include/drm/drm_damage_helper.h new file +63 -0

63 lines not on this page · raw message

include/drm/drm_mode_config.h +10 -0
index a0b202e1d69a..8ccb5ddcd99d 100644
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark damaged
+	 * regions on the plane in framebuffer coordinates of the framebuffer
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +8 -0
index 8a152dc16ea5..bb8b0934119c 100644
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that changed
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
include/uapi/drm/drm_mode.h +19 -0
index 8d67243952f4..862ea0893e2e 100644
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area this
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property FB_DAMAGE_CLIPS.
+ */
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#11 lore raw

PATCH 2/14 drm: add helper iterator functions for plane fb_damage_clips blob

In reply to Deepak Rawat

With fb_damage_clips blob property in drm_plane_state, this patch adds helper iterator to traverse the damage clips that lie inside plane src. Iterator will return full plane src as damage in case need full plane update or damage is not specified.

---
 drivers/gpu/drm/drm_damage_helper.c | 93 +++++++++++++++++++++++++++++
 include/drm/drm_damage_helper.h     | 20 +++++++
 2 files changed, 113 insertions(+)
drivers/gpu/drm/drm_damage_helper.c +93 -0
index 3e7de5afe7f6..7d70f6001889 100644
@@ -30,6 +30,7 @@
  **************************************************************************/
 
 #include <drm/drmP.h>
+#include <drm/drm_atomic.h>
 #include <drm/drm_damage_helper.h>
 
 /**
@@ -75,6 +76,11 @@
  * While kernel does not error for overlapped damage clips, it is discouraged.
  */
 
+static int convert_fixed_to_32(int fixed)
+{
+	return ((fixed >> 15) & 1) + (fixed >> 16);
+}
+
 /**
  * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
  * @plane: plane on which to enable damage clips property
@@ -90,3 +96,90 @@ void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
 				   0);
 }
 EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
+
+/**
+ * drm_atomic_helper_damage_iter_init - initialize the damage iterator
+ * @iter: The iterator to initialize.
+ * @old_state: old plane state for validation.
+ * @new_state: plane state from which to iterate the damage clips.
+ *
+ * Initialize an iterator that clip framebuffer damage in plane fb_damage_clips
+ * blob to plane src clip. The iterator returns full plane src in case needing
+ * full update e.g. during full modeset.
+ *
+ * With this helper iterator, drivers which enabled fb_damage_clips property can
+ * iterate over the damage clips that falls inside plane src during plane
+ * update.
+ *
+ * Returns: 0 on success and negative error code on error. If an error code is
+ * returned then it means the plane state shouldn't update with attached fb.
+ */
+int
+drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state)
+{
+	if (!new_state || !new_state->crtc || !new_state->fb)
+		return -EINVAL;
+
+	if (!new_state->visible)
+		return -EINVAL;
+
+	memset(iter, 0, sizeof(*iter));
+	iter->clips = drm_plane_get_damage_clips(new_state);
+	iter->num_clips = drm_plane_get_damage_clips_count(new_state);
+
+	if (!iter->clips)
+		iter->full_update = true;
+
+	if (!drm_rect_equals(&new_state->src, &old_state->src))
+		iter->full_update = true;
+
+	iter->plane_src.x1 = convert_fixed_to_32(new_state->src.x1);
+	iter->plane_src.y1 = convert_fixed_to_32(new_state->src.y1);
+	iter->plane_src.x2 = convert_fixed_to_32(new_state->src.x2);
+	iter->plane_src.y2 = convert_fixed_to_32(new_state->src.y2);
+
+	if (iter->full_update) {
+		iter->clips = 0;
+		iter->curr_clip = 0;
+		iter->num_clips = 0;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_init);
+
+/**
+ * drm_atomic_helper_damage_iter_next - advance the damage iterator
+ * @iter: The iterator to advance.
+ * @rect: Return a rectangle in fb coordinate clipped to plane src.
+ *
+ * Returns:  true if the output is valid, false if we've reached the end of the
+ * rectangle list.
+ */
+bool
+drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
+				   struct drm_rect *rect)
+{
+	bool ret = false;
+
+	if (iter->full_update) {
+		*rect = iter->plane_src;
+		iter->full_update = false;
+		return true;
+	}
+
+	while (iter->curr_clip < iter->num_clips) {
+		*rect = iter->clips[iter->curr_clip];
+		iter->curr_clip++;
+
+		if (drm_rect_intersect(rect, &iter->plane_src)) {
+			ret = true;
+			break;
+		}
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_next);
include/drm/drm_damage_helper.h +20 -0
index 217694e9168c..f1282b459a4f 100644
@@ -32,6 +32,19 @@
 #ifndef DRM_DAMAGE_HELPER_H_
 #define DRM_DAMAGE_HELPER_H_
 
+/**
+ * struct drm_atomic_helper_damage_iter - damage clip iterator
+ *
+ * This iterator tracks state needed to walk the list of damage clips.
+ */
+struct drm_atomic_helper_damage_iter {
+	const struct drm_rect *clips;
+	struct drm_rect plane_src;
+	uint32_t num_clips;
+	uint32_t curr_clip;
+	bool full_update;
+};
+
 /**
  * drm_plane_get_damage_clips_count - returns damage clips count
  * @state: Plane state
@@ -59,5 +72,12 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 }
 
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+int
+drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state);
+bool
+drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
+				   struct drm_rect *rect);
 
 #endif
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#21 lore raw

PATCH 4/14 drm: add helper to implement legacy dirtyfb

In reply to Deepak Rawat

From: Rob Clark <[email protected]>

Add an atomic helper to implement dirtyfb support. This is needed to support DSI command-mode panels with x11 userspace (ie. when we can't rely on pageflips to trigger a flush to the panel).

v2: Modified the helper to use plane fb_damage_clips property and removed plane_state::dirty flag.

---
 drivers/gpu/drm/drm_damage_helper.c | 116 ++++++++++++++++++++++++++++
 include/drm/drm_damage_helper.h     |   4 +
 2 files changed, 120 insertions(+)
drivers/gpu/drm/drm_damage_helper.c +116 -0
index 7d70f6001889..b016ec375295 100644
@@ -26,6 +26,7 @@
  *
  * Authors:
  * Deepak Rawat <[email protected]>
+ * Rob Clark <[email protected]>
  *
  **************************************************************************/
 
@@ -81,6 +82,21 @@ static int convert_fixed_to_32(int fixed)
 	return ((fixed >> 15) & 1) + (fixed >> 16);
 }
 
+static void convert_clip_rect_to_rect(const struct drm_clip_rect *src,
+				      struct drm_rect *dest,
+				      uint32_t num_clips)
+{
+	while (num_clips > 0) {
+		dest->x1 = src->x1;
+		dest->y1 = src->y1;
+		dest->x2 = src->x2;
+		dest->y2 = src->y2;
+		src++;
+		dest++;
+		num_clips--;
+	}
+}
+
 /**
  * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
  * @plane: plane on which to enable damage clips property
@@ -183,3 +199,103 @@ drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
 	return ret;
 }
 EXPORT_SYMBOL(drm_atomic_helper_damage_iter_next);
+
+/**
+ * drm_atomic_helper_dirtyfb - helper for dirtyfb
+ * @fb: DRM framebuffer
+ * @file_priv: drm file for the ioctl call
+ * @flags: dirty fb annotate flags
+ * @color: color for annotate fill
+ * @clips: dirty region
+ * @num_clips: count of clip in clips
+ *
+ * A helper to implement drm_framebuffer_funcs::dirty. This helper will do a
+ * full update in case of annotate flags.
+ *
+ * Returns: Zero on success, negative errno on failure.
+ */
+int drm_atomic_helper_dirtyfb(struct drm_framebuffer *fb,
+			      struct drm_file *file_priv, unsigned flags,
+			      unsigned color, struct drm_clip_rect *clips,
+			      unsigned num_clips)
+{
+	struct drm_modeset_acquire_ctx ctx;
+	struct drm_property_blob *damage = NULL;
+	struct drm_atomic_state *state;
+	struct drm_plane *plane;
+	struct drm_rect *rects = NULL;
+	int ret = 0;
+
+	/*
+	 * When called from ioctl, we are interruptable, but not when called
+	 * internally (ie. defio worker)
+	 */
+	drm_modeset_acquire_init(&ctx,
+		file_priv ? DRM_MODESET_ACQUIRE_INTERRUPTIBLE : 0);
+
+	state = drm_atomic_state_alloc(fb->dev);
+	if (!state) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	state->acquire_ctx = &ctx;
+
+	WARN_ONCE(flags & DRM_MODE_FB_DIRTY_FLAGS,
+		  "Full update for dirtyfb with DRM_MODE_FB_DIRTY_FLAGS\n");
+
+	if (!(flags & DRM_MODE_FB_DIRTY_FLAGS)) {
+		rects = kcalloc(num_clips, sizeof(*rects), GFP_KERNEL);
+		if (!rects) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		convert_clip_rect_to_rect(clips, rects, num_clips);
+		damage = drm_property_create_blob(fb->dev,
+						  num_clips * sizeof(*rects),
+						  rects);
+		if (IS_ERR(damage)) {
+			ret = PTR_ERR(damage);
+			damage = NULL;
+			goto out;
+		}
+	}
+
+retry:
+	drm_for_each_plane(plane, fb->dev) {
+		struct drm_plane_state *plane_state;
+
+		if (plane->state->fb != fb)
+			continue;
+
+		plane_state = drm_atomic_get_plane_state(state, plane);
+		if (IS_ERR(plane_state)) {
+			ret = PTR_ERR(plane_state);
+			goto out;
+		}
+
+		drm_property_replace_blob(&plane_state->fb_damage_clips,
+					  damage);
+	}
+
+	ret = drm_atomic_commit(state);
+
+out:
+	if (ret == -EDEADLK) {
+		drm_atomic_state_clear(state);
+		ret = drm_modeset_backoff(&ctx);
+		if (!ret)
+			goto retry;
+	}
+
+	drm_property_blob_put(damage);
+	kfree(rects);
+	drm_atomic_state_put(state);
+
+	drm_modeset_drop_locks(&ctx);
+	drm_modeset_acquire_fini(&ctx);
+
+	return ret;
+
+}
+EXPORT_SYMBOL(drm_atomic_helper_dirtyfb);
include/drm/drm_damage_helper.h +4 -0
index 1f988f7fdd72..9d5c9444761b 100644
@@ -89,5 +89,9 @@ drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
 bool
 drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
 				   struct drm_rect *rect);
+int drm_atomic_helper_dirtyfb(struct drm_framebuffer *fb,
+			      struct drm_file *file_priv, unsigned flags,
+			      unsigned color, struct drm_clip_rect *clips,
+			      unsigned num_clips);
 
 #endif
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#23 lore raw

PATCH 5/14 drm/vmwgfx: add a new interface for plane update on a display unit

In reply to Deepak Rawat

Add a new struct vmw_du_update_plane similar to vmw_kms_dirty which represent the flow of operations needed to update a display unit from surface or bo(blit a new framebuffer).

---
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 108 +++++++++++++++++++++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.h | 143 ++++++++++++++++++++++++++++
 2 files changed, 251 insertions(+)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +108 -0
index 1edbae73d6d6..c1de8f609bf4 100644
@@ -30,6 +30,7 @@
 #include <drm/drm_atomic.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_rect.h>
+#include <drm/drm_damage_helper.h>
 
 /* Might need a hrtimer here? */
 #define VMWGFX_PRESENT_RATE ((HZ / 60 > 0) ? HZ / 60 : 1)
@@ -3077,3 +3078,110 @@ void vmw_kms_lost_device(struct drm_device *dev)
 {
 	drm_atomic_helper_shutdown(dev);
 }
+
+/**
+ * vmw_du_helper_plane_update - helper to do plane update on display unit
+ *
+ * @update: The closure structure.
+ *
+ * RETURNS:
+ *
+ * 0 on success or a negative error code on failure.
+ */
+int vmw_du_helper_plane_update(struct vmw_du_update_plane *update)
+{
+	struct drm_plane_state *state = update->plane->state;
+	struct drm_plane_state *old_state = update->old_state;
+	struct drm_atomic_helper_damage_iter iter;
+	struct drm_rect clip;
+	struct drm_rect bb;
+	uint32_t reserved_size = 0;
+	uint32_t submit_size = 0;
+	uint32_t curr_size = 0;
+	uint32_t num_hits = 0;
+	void *cmd_start;
+	char *cmd_next;
+	int ret;
+
+	/*
+	 * Iterate in advance to check if really need plane update and find the
+	 * number of clips that actually are in plane src for fifo allocation.
+	 * Iterator init returns negative error code if no need to do plane
+	 * update.
+	 */
+	ret = drm_atomic_helper_damage_iter_init(&iter, old_state, state);
+	if (ret)
+		return 0;
+
+	while (drm_atomic_helper_damage_iter_next(&iter, &clip))
+		num_hits++;
+
+	/*
+	 * If all clips are outside plane src no need to do plane update on
+	 * display unit
+	 */
+	if (num_hits == 0)
+		return 0;
+
+	ret = update->prepare(update);
+	if (ret) {
+		DRM_ERROR("Failed to validate FB for command submission\n");
+		return ret;
+	}
+
+	reserved_size = update->calc_fifo_size(update, num_hits);
+	cmd_start = vmw_fifo_reserve(update->dev_priv, reserved_size);
+	if (!cmd_start) {
+		DRM_ERROR("Failed to allocate FIFO space for plane update\n");
+		update->revert(update);
+		return -ENOMEM;
+	}
+	cmd_next = cmd_start;
+
+	if (update->post_prepare) {
+		curr_size = update->post_prepare(update, cmd_next);
+		cmd_next += curr_size;
+		submit_size += curr_size;
+	}
+
+	if (update->pre_clip) {
+		curr_size = update->pre_clip(update, cmd_next, num_hits);
+		cmd_next += curr_size;
+		submit_size += curr_size;
+	}
+
+	bb.x1 = INT_MAX;
+	bb.y1 = INT_MAX;
+	bb.x2 = INT_MIN;
+	bb.y2 = INT_MIN;
+
+	drm_atomic_helper_damage_iter_init(&iter, old_state, state);
+	while (drm_atomic_helper_damage_iter_next(&iter, &clip)) {
+		uint32_t fb_x = clip.x1;
+		uint32_t fb_y = clip.y1;
+
+		vmw_du_translate_to_crtc(state, &clip);
+		if (update->clip) {
+			curr_size = update->clip(update, cmd_next, &clip, fb_x,
+						 fb_y);
+			cmd_next += curr_size;
+			submit_size += curr_size;
+		}
+		bb.x1 = min_t(int, bb.x1, clip.x1);
+		bb.y1 = min_t(int, bb.y1, clip.y1);
+		bb.x2 = max_t(int, bb.x2, clip.x2);
+		bb.y2 = max_t(int, bb.y2, clip.y2);
+	}
+
+	curr_size = update->post_clip(update, cmd_next, &bb);
+	submit_size += curr_size;
+
+	if (reserved_size < submit_size)
+		submit_size = 0;
+
+	vmw_fifo_commit(update->dev_priv, submit_size);
+
+	update->finish(update);
+
+	return ret;
+}
drivers/gpu/drm/vmwgfx/vmwgfx_kms.h +143 -0

154 lines not on this page · raw message

Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#26 lore raw

PATCH 6/14 drm/vmwgfx: implement STDU plane update for surface backed fb

In reply to Deepak Rawat

Using the new interface implement STDU plane update for surface backed fb.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.h  |  11 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 209 ++++++++++++++++++++++++++-
 2 files changed, 219 insertions(+), 1 deletion(-)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.h +11 -0
index 6bda8d3fb52f..0ecc5231b617 100644
@@ -160,6 +160,17 @@ struct vmw_du_update_plane {
 	struct vmw_fence_obj **out_fence;
 };
 
+/**
+ * struct vmw_du_update_plane_surface - closure structure for surface
+ *
+ * @base: base closure structure.
+ * @ctx: validation context.
+ */
+struct vmw_du_update_plane_surface {
+	struct vmw_du_update_plane base;
+	struct vmw_validation_ctx *ctx;
+};
+
 /**
  * struct vmw_kms_dirty - closure structure for the vmw_kms_helper_dirty
  * function.
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +208 -1

228 lines not on this page · raw message

Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#28 lore raw

PATCH 8/14 drm/vmwgfx: use the new interface for STDU plane update

In reply to Deepak Rawat

With new interface to do plane update on STDU available, use that instead of old kms_dirty. Update the commet to sync with code.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 38 +++++++---------------------
 1 file changed, 9 insertions(+), 29 deletions(-)
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +9 -29
index 8c1ba874512a..167190b75e2f 100644
@@ -1706,14 +1706,10 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
 	struct vmw_private *dev_priv;
 	int ret;
 
-	/*
-	 * We cannot really fail this function, so if we do, then output an
-	 * error and maintain consistent atomic state.
-	 */
+	/* If somehow gets a device error, maintain consistent atomic state */
 	if (crtc && plane->state->fb) {
 		struct vmw_framebuffer *vfb =
 			vmw_framebuffer_to_vfb(plane->state->fb);
-		struct drm_vmw_rect vclips;
 		stdu = vmw_crtc_to_stdu(crtc);
 		dev_priv = vmw_priv(crtc->dev);
 
@@ -1721,23 +1717,17 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
 		stdu->content_fb_type = vps->content_fb_type;
 		stdu->cpp = vps->cpp;
 
-		vclips.x = crtc->x;
-		vclips.y = crtc->y;
-		vclips.w = crtc->mode.hdisplay;
-		vclips.h = crtc->mode.vdisplay;
-
 		ret = vmw_stdu_bind_st(dev_priv, stdu, &stdu->display_srf->res);
 		if (ret)
 			DRM_ERROR("Failed to bind surface to STDU.\n");
 
 		if (vfb->bo)
-			ret = vmw_kms_stdu_dma(dev_priv, NULL, vfb, NULL, NULL,
-					       &vclips, 1, 1, true, false,
-					       crtc);
+			ret = vmw_stdu_plane_update_bo(dev_priv, plane,
+						       old_state, vfb);
 		else
-			ret = vmw_kms_stdu_surface_dirty(dev_priv, vfb, NULL,
-							 &vclips, NULL, 0, 0,
-							 1, 1, NULL, crtc);
+			ret = vmw_stdu_plane_update_surface(dev_priv, plane,
+							    old_state, vfb,
+							    NULL);
 		if (ret)
 			DRM_ERROR("Failed to update STDU.\n");
 	} else {
@@ -1745,12 +1735,7 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
 		stdu = vmw_crtc_to_stdu(crtc);
 		dev_priv = vmw_priv(crtc->dev);
 
-		/*
-		 * When disabling a plane, CRTC and FB should always be NULL
-		 * together, otherwise it's an error.
-		 * Here primary plane is being disable so blank the screen
-		 * target display unit, if not already done.
-		 */
+		/* Blank STDU when fb and crtc are NULL */
 		if (!stdu->defined)
 			return;
 
@@ -1765,20 +1750,15 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
 		return;
 	}
 
+	/* In case of error vblank event is sent in vmw_du_crtc_atomic_flush */
 	event = crtc->state->event;
-	/*
-	 * In case of failure and other cases, vblank event will be sent in
-	 * vmw_du_crtc_atomic_flush.
-	 */
 	if (event && (ret == 0)) {
 		struct vmw_fence_obj *fence = NULL;
 		struct drm_file *file_priv = event->base.file_priv;
 
 		vmw_execbuf_fence_commands(NULL, dev_priv, &fence, NULL);
 
-		/*
-		 * If fence is NULL, then already sync.
-		 */
+		/* If fence is NULL, then already sync. */
 		if (fence) {
 			ret = vmw_event_fence_action_queue(
 				file_priv, fence, &event->base,
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#30 lore raw

PATCH 9/14 drm/vmwgfx: enable FB_DAMAGE_CLIPS property for STDU primary plane

In reply to Deepak Rawat

STDU primary plane now support damage clips, enable it for user-space.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 1 +
 1 file changed, 1 insertion(+)
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +1 -0
index 167190b75e2f..11a76aa96302 100644
@@ -1883,6 +1883,7 @@ static int vmw_stdu_init(struct vmw_private *dev_priv, unsigned unit)
 	}
 
 	drm_plane_helper_add(primary, &vmw_stdu_primary_plane_helper_funcs);
+	drm_plane_enable_fb_damage_clips(primary);
 
 	/* Initialize cursor plane */
 	vmw_du_plane_reset(cursor);
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#32 lore raw

PATCH 10/14 drm/vmwgfx: implement SOU plane update for surface backed fb

In reply to Deepak Rawat

Using the new interface implement SOU plane update for surface backed fb.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.h  |   3 +
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c | 187 +++++++++++++++++++++++++++
 2 files changed, 190 insertions(+)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.h +3 -0
index 1c3dd0e14414..45b12448e138 100644
@@ -165,10 +165,13 @@ struct vmw_du_update_plane {
  *
  * @base: base closure structure.
  * @ctx: validation context.
+ * @cmd_start: (optional) FIFO command start address (used by SOU only).
  */
 struct vmw_du_update_plane_surface {
 	struct vmw_du_update_plane base;
 	struct vmw_validation_ctx *ctx;
+	/* This member is to handle special case SOU surface update */
+	void *cmd_start;
 };
 
 /**
drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c +187 -0

193 lines not on this page · raw message

Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#36 lore raw

PATCH 14/14 drm/vmwgfx: use atomic helper function for dirty fb IOCTL

In reply to Deepak Rawat

With new atomic helper for dirty fb IOCTL is available which uses damage interface, use that for dirty fb IOCTL.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 116 +---------------------------
 1 file changed, 2 insertions(+), 114 deletions(-)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +2 -114
index c1de8f609bf4..adf3227fb0dc 100644
@@ -850,58 +850,6 @@ static void vmw_framebuffer_surface_destroy(struct drm_framebuffer *framebuffer)
 	kfree(vfbs);
 }
 
-static int vmw_framebuffer_surface_dirty(struct drm_framebuffer *framebuffer,
-				  struct drm_file *file_priv,
-				  unsigned flags, unsigned color,
-				  struct drm_clip_rect *clips,
-				  unsigned num_clips)
-{
-	struct vmw_private *dev_priv = vmw_priv(framebuffer->dev);
-	struct vmw_framebuffer_surface *vfbs =
-		vmw_framebuffer_to_vfbs(framebuffer);
-	struct drm_clip_rect norect;
-	int ret, inc = 1;
-
-	/* Legacy Display Unit does not support 3D */
-	if (dev_priv->active_display_unit == vmw_du_legacy)
-		return -EINVAL;
-
-	drm_modeset_lock_all(dev_priv->dev);
-
-	ret = ttm_read_lock(&dev_priv->reservation_sem, true);
-	if (unlikely(ret != 0)) {
-		drm_modeset_unlock_all(dev_priv->dev);
-		return ret;
-	}
-
-	if (!num_clips) {
-		num_clips = 1;
-		clips = &norect;
-		norect.x1 = norect.y1 = 0;
-		norect.x2 = framebuffer->width;
-		norect.y2 = framebuffer->height;
-	} else if (flags & DRM_MODE_FB_DIRTY_ANNOTATE_COPY) {
-		num_clips /= 2;
-		inc = 2; /* skip source rects */
-	}
-
-	if (dev_priv->active_display_unit == vmw_du_screen_object)
-		ret = vmw_kms_sou_do_surface_dirty(dev_priv, &vfbs->base,
-						   clips, NULL, NULL, 0, 0,
-						   num_clips, inc, NULL, NULL);
-	else
-		ret = vmw_kms_stdu_surface_dirty(dev_priv, &vfbs->base,
-						 clips, NULL, NULL, 0, 0,
-						 num_clips, inc, NULL, NULL);
-
-	vmw_fifo_flush(dev_priv, false);
-	ttm_read_unlock(&dev_priv->reservation_sem);
-
-	drm_modeset_unlock_all(dev_priv->dev);
-
-	return 0;
-}
-
 /**
  * vmw_kms_readback - Perform a readback from the screen system to
  * a buffer-object backed framebuffer.
@@ -945,7 +893,7 @@ int vmw_kms_readback(struct vmw_private *dev_priv,
 
 static const struct drm_framebuffer_funcs vmw_framebuffer_surface_funcs = {
 	.destroy = vmw_framebuffer_surface_destroy,
-	.dirty = vmw_framebuffer_surface_dirty,
+	.dirty = drm_atomic_helper_dirtyfb,
 };
 
 static int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
@@ -1056,69 +1004,9 @@ static void vmw_framebuffer_bo_destroy(struct drm_framebuffer *framebuffer)
 	kfree(vfbd);
 }
 
-static int vmw_framebuffer_bo_dirty(struct drm_framebuffer *framebuffer,
-				    struct drm_file *file_priv,
-				    unsigned int flags, unsigned int color,
-				    struct drm_clip_rect *clips,
-				    unsigned int num_clips)
-{
-	struct vmw_private *dev_priv = vmw_priv(framebuffer->dev);
-	struct vmw_framebuffer_bo *vfbd =
-		vmw_framebuffer_to_vfbd(framebuffer);
-	struct drm_clip_rect norect;
-	int ret, increment = 1;
-
-	drm_modeset_lock_all(dev_priv->dev);
-
-	ret = ttm_read_lock(&dev_priv->reservation_sem, true);
-	if (unlikely(ret != 0)) {
-		drm_modeset_unlock_all(dev_priv->dev);
-		return ret;
-	}
-
-	if (!num_clips) {
-		num_clips = 1;
-		clips = &norect;
-		norect.x1 = norect.y1 = 0;
-		norect.x2 = framebuffer->width;
-		norect.y2 = framebuffer->height;
-	} else if (flags & DRM_MODE_FB_DIRTY_ANNOTATE_COPY) {
-		num_clips /= 2;
-		increment = 2;
-	}
-
-	switch (dev_priv->active_display_unit) {
-	case vmw_du_screen_target:
-		ret = vmw_kms_stdu_dma(dev_priv, NULL, &vfbd->base, NULL,
-				       clips, NULL, num_clips, increment,
-				       true, true, NULL);
-		break;
-	case vmw_du_screen_object:
-		ret = vmw_kms_sou_do_bo_dirty(dev_priv, &vfbd->base,
-					      clips, NULL, num_clips,
-					      increment, true, NULL, NULL);
-		break;
-	case vmw_du_legacy:
-		ret = vmw_kms_ldu_do_bo_dirty(dev_priv, &vfbd->base, 0, 0,
-					      clips, num_clips, increment);
-		break;
-	default:
-		ret = -EINVAL;
-		WARN_ONCE(true, "Dirty called with invalid display system.\n");
-		break;
-	}
-
-	vmw_fifo_flush(dev_priv, false);
-	ttm_read_unlock(&dev_priv->reservation_sem);
-
-	drm_modeset_unlock_all(dev_priv->dev);
-
-	return ret;
-}
-
 static const struct drm_framebuffer_funcs vmw_framebuffer_bo_funcs = {
 	.destroy = vmw_framebuffer_bo_destroy,
-	.dirty = vmw_framebuffer_bo_dirty,
+	.dirty = drm_atomic_helper_dirtyfb,
 };
 
 /**
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#0 lore raw

cover plane update with damage

Hi,

Set of patches for a new plane property FB_DAMAGE_CLIPS to send damage during plane update. I have incorporated review comments since last RFC. Also with this patch series vmwgfx implementation of plane damage.

Based on Rob Clark earlier patch, implemented legacy dirtyfb with new damage clips property.

I will be sending a separate patch series for unit test case for damage interface using igt-gpu-tools. Also a separate patch series for a rudimentary weston implementation, which for now only send damage for default scanout plane.

Testing: vmware xorg driver mainly use dirty_fb ioctl to update the screen so it is a good overall test of damage interface. Weston implementation also works as intended.

Let me know what you think. Thanks,
Deepak

Deepak Rawat (12):
drm: add helper iterator functions for plane fb_damage_clips blob
drm: clear plane damage during full modeset
drm/vmwgfx: add a new interface for plane update on a display unit drm/vmwgfx: implement STDU plane update for surface backed fb drm/vmwgfx: implement STDU plane update for BO backed fb drm/vmwgfx: use the new interface for STDU plane update drm/vmwgfx: enable FB_DAMAGE_CLIPS property for STDU primary plane drm/vmwgfx: implement SOU plane update for surface backed fb drm/vmwgfx: implement SOU plane update for BO backed fb drm/vmwgfx: use the new interface for SOU plane update drm/vmwgfx: enable FB_DAMAGE_CLIPS property for SOU primary plane drm/vmwgfx: use atomic helper function for dirty fb IOCTL

Lukasz Spintzyk (1):
drm: add new plane property FB_DAMAGE_CLIPS to send damage during
plane update

Rob Clark (1):
drm: add helper to implement legacy dirtyfb

 Documentation/gpu/drm-kms.rst        |   9 +
 drivers/gpu/drm/Makefile             |   2 +-
 drivers/gpu/drm/drm_atomic.c         |  13 +
 drivers/gpu/drm/drm_atomic_helper.c  |   8 +
 drivers/gpu/drm/drm_damage_helper.c  | 301 +++++++++++++++++
 drivers/gpu/drm/drm_mode_config.c    |   6 +
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c  | 225 ++++++-------
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.h  | 171 ++++++++++
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c | 361 ++++++++++++++++++--
 drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 483 +++++++++++++++++++++++++--
 include/drm/drm_damage_helper.h      |  97 ++++++
 include/drm/drm_mode_config.h        |  10 +
 include/drm/drm_plane.h              |   8 +
 include/uapi/drm/drm_mode.h          |  19 ++
 14 files changed, 1539 insertions(+), 174 deletions(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#33 lore raw

PATCH 11/14 drm/vmwgfx: implement SOU plane update for BO backed fb

Using the new interface implement SOU plane update for BO backed fb.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c | 134 +++++++++++++++++++++++++++
 1 file changed, 134 insertions(+)
drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c +134 -0
index 8427a51cde3f..a68a2c8dab47 100644
@@ -76,6 +76,11 @@ struct vmw_kms_sou_dirty_cmd {
 	SVGA3dCmdBlitSurfaceToScreen body;
 };
 
+struct vmw_kms_sou_define_gmrfb {
+	uint32_t header;
+	SVGAFifoCmdDefineGMRFB body;
+};
+
 /**
  * Display unit using screen objects.
  */
@@ -499,6 +504,135 @@ vmw_sou_primary_plane_prepare_fb(struct drm_plane *plane,
 	return vmw_bo_pin_in_vram(dev_priv, vps->bo, true);
 }
 
+static int vmw_sou_bo_prepare(struct vmw_du_update_plane *update)
+{
+	struct vmw_du_update_plane_buffer *bo_update;
+	struct vmw_framebuffer_bo *vfbbo;
+
+	bo_update = container_of(update, typeof(*bo_update), base);
+	vfbbo = container_of(update->vfb, typeof(*vfbbo), base);
+
+	return vmw_kms_helper_buffer_prepare(update->dev_priv, vfbbo->buffer,
+					     true, false, bo_update->cpu_blit);
+}
+
+static uint32_t vmw_sou_bo_fifo_size(struct vmw_du_update_plane *update,
+				     uint32_t num_hits)
+{
+	return sizeof(struct vmw_kms_sou_define_gmrfb) +
+		sizeof(struct vmw_kms_sou_bo_blit) * num_hits;
+}
+
+static uint32_t vmw_sou_bo_define_gmrfb(struct vmw_du_update_plane *update,
+					void *cmd)
+{
+	struct vmw_framebuffer_bo *vfbbo =
+		container_of(update->vfb, typeof(*vfbbo), base);
+	struct vmw_kms_sou_define_gmrfb *gmr = cmd;
+	int depth = update->vfb->base.format->depth;
+
+	/* Emulate RGBA support, contrary to svga_reg.h this is not
+	 * supported by hosts. This is only a problem if we are reading
+	 * this value later and expecting what we uploaded back.
+	 */
+	if (depth == 32)
+		depth = 24;
+
+	gmr->header = SVGA_CMD_DEFINE_GMRFB;
+
+	gmr->body.format.bitsPerPixel = update->vfb->base.format->cpp[0] * 8;
+	gmr->body.format.colorDepth = depth;
+	gmr->body.format.reserved = 0;
+	gmr->body.bytesPerLine = update->vfb->base.pitches[0];
+	vmw_bo_get_guest_ptr(&vfbbo->buffer->base, &gmr->body.ptr);
+
+	return sizeof(*gmr);
+}
+
+static uint32_t vmw_sou_bo_populate_clip(struct vmw_du_update_plane  *update,
+					 void *cmd, struct drm_rect *clip,
+					 uint32_t fb_x, uint32_t fb_y)
+{
+	struct vmw_kms_sou_bo_blit *blit = cmd;
+
+	blit->header = SVGA_CMD_BLIT_GMRFB_TO_SCREEN;
+	blit->body.destScreenId = update->du->unit;
+	blit->body.srcOrigin.x = fb_x;
+	blit->body.srcOrigin.y = fb_y;
+	blit->body.destRect.left = clip->x1;
+	blit->body.destRect.top = clip->y1;
+	blit->body.destRect.right = clip->x2;
+	blit->body.destRect.bottom = clip->y2;
+
+	return sizeof(*blit);
+}
+
+static uint32_t vmw_stud_bo_post_clip(struct vmw_du_update_plane  *update,
+				      void *cmd, struct drm_rect *bb)
+{
+	return 0;
+}
+
+static void vmw_sou_bo_finish(struct vmw_du_update_plane *update)
+{
+	struct vmw_framebuffer_bo *vfbbo =
+		container_of(update->vfb, typeof(*vfbbo), base);
+
+	vmw_kms_helper_buffer_finish(update->dev_priv, NULL, vfbbo->buffer,
+				     update->out_fence, NULL);
+}
+
+static void vmw_sou_bo_revert(struct vmw_du_update_plane *update)
+{
+	struct vmw_framebuffer_bo *vfbbo =
+		container_of(update->vfb, typeof(*vfbbo), base);
+
+	vmw_kms_helper_buffer_revert(vfbbo->buffer);
+}
+
+/**
+ * vmw_sou_plane_update_bo - update display unit for bo backed fb
+ * @dev_priv: device private
+ * @plane: plane state
+ * @old_state: old plane state
+ * @vfb: framebuffer which is blitted to display unit
+ * @out_fence: (optional) If non-NULL, will return a ref-counted pointer to a
+ * struct vmw_fence_obj. The returned fence pointer may be NULL in which case
+ * the device has already synchronized.
+ *
+ * RETURNS:
+ *
+ * 0 on success or a negative error code on failure.
+ */
+static int vmw_sou_plane_update_bo(struct vmw_private *dev_priv,
+				   struct drm_plane *plane,
+				   struct drm_plane_state *old_state,
+				   struct vmw_framebuffer *vfb,
+				   struct vmw_fence_obj **out_fence)
+{
+	struct vmw_du_update_plane_buffer bo_update;
+
+	memset(&bo_update, 0, sizeof(struct vmw_du_update_plane_buffer));
+	bo_update.base.plane = plane;
+	bo_update.base.old_state = old_state;
+	bo_update.base.dev_priv = dev_priv;
+	bo_update.base.du = vmw_crtc_to_du(plane->state->crtc);
+	bo_update.base.vfb = vfb;
+	bo_update.base.out_fence = out_fence;
+
+	bo_update.cpu_blit = false;
+
+	bo_update.base.prepare = vmw_sou_bo_prepare;
+	bo_update.base.calc_fifo_size = vmw_sou_bo_fifo_size;
+	bo_update.base.post_prepare = vmw_sou_bo_define_gmrfb;
+	bo_update.base.clip = vmw_sou_bo_populate_clip;
+	bo_update.base.post_clip = vmw_stud_bo_post_clip;
+	bo_update.base.finish = vmw_sou_bo_finish;
+	bo_update.base.revert = vmw_sou_bo_revert;
+
+	return vmw_du_helper_plane_update(&bo_update.base);
+}
+
 static int vmw_sou_surface_prepare(struct vmw_du_update_plane *update)
 {
 	struct vmw_du_update_plane_surface *srf_update;
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#34 lore raw

PATCH 12/14 drm/vmwgfx: use the new interface for SOU plane update

In reply to Deepak Rawat

With new interface to do plane update on SOU available, use that instead of old kms_dirty. Updated the comments to sync with code.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c | 38 +++++++---------------------
 1 file changed, 9 insertions(+), 29 deletions(-)
drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c +9 -29
index a68a2c8dab47..290d550d1b77 100644
@@ -830,52 +830,32 @@ vmw_sou_primary_plane_atomic_update(struct drm_plane *plane,
 	struct vmw_fence_obj *fence = NULL;
 	int ret;
 
+	 /* If somehow gets a device error maintain consistent atomic state */
 	if (crtc && plane->state->fb) {
 		struct vmw_private *dev_priv = vmw_priv(crtc->dev);
 		struct vmw_framebuffer *vfb =
 			vmw_framebuffer_to_vfb(plane->state->fb);
-		struct drm_vmw_rect vclips;
-
-		vclips.x = crtc->x;
-		vclips.y = crtc->y;
-		vclips.w = crtc->mode.hdisplay;
-		vclips.h = crtc->mode.vdisplay;
 
 		if (vfb->bo)
-			ret = vmw_kms_sou_do_bo_dirty(dev_priv, vfb, NULL,
-						      &vclips, 1, 1, true,
-						      &fence, crtc);
+			ret = vmw_sou_plane_update_bo(dev_priv, plane,
+						      old_state, vfb, &fence);
 		else
-			ret = vmw_kms_sou_do_surface_dirty(dev_priv, vfb, NULL,
-							   &vclips, NULL, 0, 0,
-							   1, 1, &fence, crtc);
-
-		/*
-		 * We cannot really fail this function, so if we do, then output
-		 * an error and maintain consistent atomic state.
-		 */
+			ret = vmw_sou_plane_update_surface(dev_priv, plane,
+							   old_state, vfb,
+							   &fence);
 		if (ret != 0)
 			DRM_ERROR("Failed to update screen.\n");
 	} else {
-		/*
-		 * When disabling a plane, CRTC and FB should always be NULL
-		 * together, otherwise it's an error.
-		 * Here primary plane is being disable so should really blank
-		 * the screen object display unit, if not already done.
-		 */
+		/* Do nothing when fb and crtc is NULL (blank crtc) */
 		return;
 	}
 
+	/* For error case vblank event is sent from vmw_du_crtc_atomic_flush */
 	event = crtc->state->event;
-	/*
-	 * In case of failure and other cases, vblank event will be sent in
-	 * vmw_du_crtc_atomic_flush.
-	 */
 	if (event && fence) {
 		struct drm_file *file_priv = event->base.file_priv;
 
-		ret = vmw_event_fence_action_queue(file_priv,
-						   fence,
+		ret = vmw_event_fence_action_queue(file_priv, fence,
 						   &event->base,
 						   &event->event.vbl.tv_sec,
 						   &event->event.vbl.tv_usec,
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#35 lore raw

PATCH 13/14 drm/vmwgfx: enable FB_DAMAGE_CLIPS property for SOU primary plane

In reply to Deepak Rawat

SOU primary plane now support damage clips, enable it for user-space.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c +2 -0
index 290d550d1b77..c4966652cdb7 100644
@@ -29,6 +29,7 @@
 #include <drm/drm_plane_helper.h>
 #include <drm/drm_atomic.h>
 #include <drm/drm_atomic_helper.h>
+#include <drm/drm_damage_helper.h>
 
 
 #define vmw_crtc_to_sou(x) \
@@ -967,6 +968,7 @@ static int vmw_sou_init(struct vmw_private *dev_priv, unsigned unit)
 	}
 
 	drm_plane_helper_add(primary, &vmw_sou_primary_plane_helper_funcs);
+	drm_plane_enable_fb_damage_clips(primary);
 
 	/* Initialize cursor plane */
 	vmw_du_plane_reset(cursor);
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#15 lore raw

PATCH 3/14 drm: clear plane damage during full modeset

In reply to Deepak Rawat

Plane damage is irrelevant when full modeset happens so clear the damage blob property(If set by user-space). With damage blob cleared damage helper iterator will return full plane src as damage clip.

---
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++++
 include/drm/drm_damage_helper.h     | 10 ++++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +4 -0
index be83e2763c18..e06d2d5d582f 100644
@@ -31,6 +31,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_writeback.h>
+#include <drm/drm_damage_helper.h>
 #include <linux/dma-fence.h>
 
 #include "drm_crtc_helper_internal.h"
@@ -88,6 +89,9 @@ drm_atomic_helper_plane_changed(struct drm_atomic_state *state,
 			return;
 
 		crtc_state->planes_changed = true;
+
+		if (drm_atomic_crtc_needs_modeset(crtc_state))
+			drm_plane_clear_damage(plane_state);
 	}
 }
 
include/drm/drm_damage_helper.h +10 -0
index f1282b459a4f..1f988f7fdd72 100644
@@ -71,6 +71,16 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 				   state->fb_damage_clips->data : NULL);
 }
 
+/**
+ * drm_plane_clear_damage - clears damage blob in a plane state
+ * @state: Plane state
+ */
+static inline void drm_plane_clear_damage(struct drm_plane_state *state)
+{
+	drm_property_blob_put(state->fb_damage_clips);
+	state->fb_damage_clips = NULL;
+}
+
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
 int
 drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#27 lore raw

PATCH 7/14 drm/vmwgfx: implement STDU plane update for BO backed fb

In reply to Deepak Rawat

Using the new interface implement STDU plane update for BO backed fb.

---
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.h  |  14 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 235 +++++++++++++++++++++++++++
 2 files changed, 249 insertions(+)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.h +14 -0
index 0ecc5231b617..1c3dd0e14414 100644
@@ -171,6 +171,20 @@ struct vmw_du_update_plane_surface {
 	struct vmw_validation_ctx *ctx;
 };
 
+/**
+ * struct vmw_du_update_plane_buffer - closure structure for buffer object
+ *
+ * @base: base closure structure.
+ * @fb_left: x1 for fb damage bounding box.
+ * @fb_top: y1 for fb damage bounding box.
+ * @cpu_blit: true if need cpu blit.
+ */
+struct vmw_du_update_plane_buffer {
+	struct vmw_du_update_plane base;
+	int fb_left, fb_top;
+	bool cpu_blit;
+};
+
 /**
  * struct vmw_kms_dirty - closure structure for the vmw_kms_helper_dirty
  * function.
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +235 -0

241 lines not on this page · raw message

Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#2 lore raw

Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Deepak Rawat
Quoted 46 lines · Deepak Rawat

From: Lukasz Spintzyk <[email protected]>

FB_DAMAGE_CLIPS is an optional plane property to mark damaged regions on the plane in framebuffer coordinates of the framebuffer attached to the plane.

The layout of blob data is simply an array of "struct drm_mode_rect" with maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips are not in 16.16 fixed point. As plane src in framebuffer cannot be negative so are damage clips. In damage clip, x1/y1 are inclusive and x2/y2 are exclusive.

This patch also exports the kernel internal drm_rect to userspace as drm_mode_rect. This is because "struct drm_clip_rect" is not sufficient to represent damage for current plane size.

Upper limit is set on the maximum number of damage clips to avoid overflow by user-space.

Driver which are interested in enabling FB_DAMAGE_CLIPS property for a plane should enable this property using drm_plane_enable_damage_clips.

---
 Documentation/gpu/drm-kms.rst       |  9 +++
 drivers/gpu/drm/Makefile            |  2 +-
 drivers/gpu/drm/drm_atomic.c        | 13 ++++
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++
 drivers/gpu/drm/drm_damage_helper.c | 92 +++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_mode_config.c   |  6 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |  1 +
 include/drm/drm_damage_helper.h     | 63 ++++++++++++++++++++
 include/drm/drm_mode_config.h       | 10 ++++
 include/drm/drm_plane.h             |  8 +++
 include/uapi/drm/drm_mode.h         | 19 ++++++
 11 files changed, 226 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Documentation/gpu/drm-kms.rst +0 -0
index 5dee6b8a4c12..78b66e628857 100644
@@ -542,6 +542,15 @@ Plane Composition Properties
 .. kernel-doc:: drivers/gpu/drm/drm_blend.c

Documentation/gpu/drm-kms.rst Plane Composition Properties · line 552

Quoted 10 lines
Documentation/gpu/drm-kms.rst +8 -0
    :export:
 
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:

You forgot to include your nice kerneldoc from the header. Please run

$ make htmldocs

and make sure the generated stuff looks all nice and is complete.

Quoted 60 lines
Documentation/gpu/drm-kms.rst +1 -0
+
 Color Management Properties
 ---------------------------
 
drivers/gpu/drm/Makefile +1 -1
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \
 		drm_plane_helper.o drm_dp_mst_topology.o drm_atomic_helper.o \
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o \
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o drm_damage_helper.o
 
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fb_helper.o
drivers/gpu/drm/drm_atomic.c +13 -0
index 3eb061e11e2e..652e78ca1f81 100644
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb = drm_framebuffer_lookup(dev, NULL, val);
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane *plane,
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state, property, val);
 	} else {
drivers/gpu/drm/drm_atomic_helper.c +0 -0
index 80be74df7ba6..be83e2763c18 100644

drivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct drm_plane *plane) · line 3579

Quoted 4 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct drm_plane *plane)
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property->values[1];
+		plane->state->fb_damage_clips = NULL;

No need to set to 0, we require kzalloc.

Quoted 88 lines
drivers/gpu/drm/drm_atomic_helper.c +3 -0
 	}
 }
 EXPORT_SYMBOL(drm_atomic_helper_plane_reset);
@@ -3598,6 +3599,7 @@ void __drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,
 
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void __drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)
 
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
drivers/gpu/drm/drm_damage_helper.c new file +62 -0
new file mode 100644
index 000000000000..3e7de5afe7f6
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/**************************************************************************
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+ **************************************************************************/
+
+#include <drm/drmP.h>
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a means to
+ * specify a list of damage rectangles on a plane in framebuffer coordinates of
+ * the framebuffer attached to the plane. In current context damage is the area
+ * of plane framebuffer (excluding the framebuffer area which is outside of
+ * plane src) that has changed since last plane update (also called page-flip).
+ * Only the area inside damage rectangles will be considered different whether
+ * currently attached framebuffer is same as framebuffer attached during last
+ * plane update or not.
+ *
+ * FB_DAMAGE_CLIPS is a hint to kernel which could be helpful for some drivers
+ * to optimize internally especially for virtual devices where each framebuffer
+ * change needs to be transmitted over network, usb, etc.
+ *
+ * Since FB_DAMAGE_CLIPS is a hint so it is an optional property. User-space can
+ * ignore the damage clips property and in that case kernel will do a full plane
+ * update. In case damage clips are provided then it is guaranteed that the area
+ * inside damage clips will be updated to plane. For efficiency kernel can do
+ * full update or more area than specified in damage clips.
+ *
+ * FB_DAMAGE_CLIPS is a blob property with the layout of blob data is simply an
+ * array of drm_mode_rect(internally drm_rect) with maximum array size limited
+ * by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips
+ * are not in 16.16 fixed point. Similar to plane src in framebuffer, damage
+ * clips cannot be negative. In damage clip, x1/y1 are inclusive and x2/y2 are
+ * exclusive.

drivers/gpu/drm/drm_damage_helper.c line 72

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c new file +10 -0
+ *
+ * All the damage clips are in framebuffer coordinates and should be inside
+ * plane src and if any clip falls outside plane src it will be ignored and
+ * user-space won't be notified of the same. As mentioned above, sometimes
+ * kernel will do full plane update due to change in properties which can affect
+ * full plane e.g. color management properties. Also during full modeset damage
+ * is irrelevant so if provided by user-space it simply will be ignored.
+ * Whenever damage clips are ignored by kernel, user-space will not be informed.
+ * If a user-space provides damage clips which doesn't encompass the actual
+ * damage to framebuffer (since last plane update) will result in incorrect

s/will/can/ - it's an optimization

drivers/gpu/drm/drm_damage_helper.c line 75

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+ * rendering during plane update.
+ *
+ * While kernel does not error for overlapped damage clips, it is discouraged.

Some comments on the doc:

- You kinda explain it, but in convoluted way: I'd add a clear sentence that userspace _must_ always render the entire visible fb, since the driver is free to read more. Otherwise there can be corruptions.

- Why no input validation on the damage clips against the framebuffer size? Ime not validating just leads to funny driver bugs down the road, so what's the use-case you have in mind here?

- A short paragraph on how drivers are supposed to implement this would be good, with references to functions (will be automatically converted to hyperlinks)

Quoted 175 lines
drivers/gpu/drm/drm_damage_helper.c new file +17 -0
+ */
+
+/**
+ * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
+ * @plane: plane on which to enable damage clips property
+ *
+ * This function lets driver to enable the damage clips property on a plane.
+ */
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
+{
+	struct drm_device *dev = plane->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	drm_object_attach_property(&plane->base, config->prop_fb_damage_clips,
+				   0);
+}
+EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
drivers/gpu/drm/drm_mode_config.c +6 -0
index 21e353bd3948..506f5a52733f 100644
@@ -298,6 +298,12 @@ static int drm_mode_create_standard_properties(struct drm_device *dev)
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB, "FB_DAMAGE_CLIPS",
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev, DRM_MODE_PROP_ATOMIC,
 			"ACTIVE");
 	if (!prop)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
index 23beff5d8e3c..1edbae73d6d6 100644
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane *plane)
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
include/drm/drm_damage_helper.h new file +63 -0
new file mode 100644
index 000000000000..217694e9168c
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */
+/**************************************************************************
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+ **************************************************************************/
+
+#ifndef DRM_DAMAGE_HELPER_H_
+#define DRM_DAMAGE_HELPER_H_
+
+/**
+ * drm_plane_get_damage_clips_count - returns damage clips count
+ * @state: Plane state
+ *
+ * Returns: Number of clips in plane fb_damage_clips blob property.
+ */
+static inline uint32_t
+drm_plane_get_damage_clips_count(const struct drm_plane_state *state)
+{
+	return (state && state->fb_damage_clips) ?
+		state->fb_damage_clips->length/sizeof(struct drm_rect) : 0;
+}
+
+/**
+ * drm_plane_get_damage_clips - returns damage clips
+ * @state: Plane state
+ *
+ * Returns: Clips in plane fb_damage_clips blob property.
+ */
+static inline struct drm_rect *
+drm_plane_get_damage_clips(const struct drm_plane_state *state)
+{
+	return (struct drm_rect *)((state && state->fb_damage_clips) ?
+				   state->fb_damage_clips->data : NULL);
+}
+
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+
+#endif
include/drm/drm_mode_config.h +10 -0
index a0b202e1d69a..8ccb5ddcd99d 100644
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark damaged
+	 * regions on the plane in framebuffer coordinates of the framebuffer
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +8 -0
index 8a152dc16ea5..bb8b0934119c 100644
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that changed
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
include/uapi/drm/drm_mode.h +12 -0
index 8d67243952f4..862ea0893e2e 100644
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area this
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property FB_DAMAGE_CLIPS.
+ */

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease { · line 911

Quoted 10 lines
include/uapi/drm/drm_mode.h +7 -0
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif

Besides the small nits this looks good to me.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#12 lore raw

Re: [PATCH 02/14] drm: add helper iterator functions for plane fb_damage_clips blob

In reply to Deepak Rawat
Quoted 57 lines · Deepak Rawat

With fb_damage_clips blob property in drm_plane_state, this patch adds helper iterator to traverse the damage clips that lie inside plane src. Iterator will return full plane src as damage in case need full plane update or damage is not specified.

---
 drivers/gpu/drm/drm_damage_helper.c | 93 +++++++++++++++++++++++++++++
 include/drm/drm_damage_helper.h     | 20 +++++++
 2 files changed, 113 insertions(+)
drivers/gpu/drm/drm_damage_helper.c +24 -0
index 3e7de5afe7f6..7d70f6001889 100644
@@ -30,6 +30,7 @@
  **************************************************************************/
 
 #include <drm/drmP.h>
+#include <drm/drm_atomic.h>
 #include <drm/drm_damage_helper.h>
 
 /**
@@ -75,6 +76,11 @@
  * While kernel does not error for overlapped damage clips, it is discouraged.
  */
 
+static int convert_fixed_to_32(int fixed)
+{
+	return ((fixed >> 15) & 1) + (fixed >> 16);
+}
+
 /**
  * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
  * @plane: plane on which to enable damage clips property
@@ -90,3 +96,90 @@ void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
 				   0);
 }
 EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
+
+/**
+ * drm_atomic_helper_damage_iter_init - initialize the damage iterator
+ * @iter: The iterator to initialize.
+ * @old_state: old plane state for validation.
+ * @new_state: plane state from which to iterate the damage clips.
+ *
+ * Initialize an iterator that clip framebuffer damage in plane fb_damage_clips
+ * blob to plane src clip. The iterator returns full plane src in case needing
+ * full update e.g. during full modeset.
+ *
+ * With this helper iterator, drivers which enabled fb_damage_clips property can
+ * iterate over the damage clips that falls inside plane src during plane
+ * update.
+ *
+ * Returns: 0 on success and negative error code on error. If an error code is
+ * returned then it means the plane state shouldn't update with attached fb.
+ */

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane) · line 126

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+int
+drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state)
+{
+	if (!new_state || !new_state->crtc || !new_state->fb)
+		return -EINVAL;
+
+	if (!new_state->visible)
+		return -EINVAL;

Can't we handle this by iterating 0 damage rects instead? Would make the code a bit cleaner I think.

Quoted 5 lines
drivers/gpu/drm/drm_damage_helper.c +5 -0
+
+	memset(iter, 0, sizeof(*iter));
+	iter->clips = drm_plane_get_damage_clips(new_state);
+	iter->num_clips = drm_plane_get_damage_clips_count(new_state);
+

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane) · line 141

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+	if (!iter->clips)
+		iter->full_update = true;
+
+	if (!drm_rect_equals(&new_state->src, &old_state->src))
+		iter->full_update = true;
+
+	iter->plane_src.x1 = convert_fixed_to_32(new_state->src.x1);
+	iter->plane_src.y1 = convert_fixed_to_32(new_state->src.y1);
+	iter->plane_src.x2 = convert_fixed_to_32(new_state->src.x2);
+	iter->plane_src.y2 = convert_fixed_to_32(new_state->src.y2);

I think you want to clip with the clipped rectangles here, not with the ones userspace provides.

Also I think you're rounding is wrong here - I think you need to round down for x/y1, and round up for x/y2 to make sure you catch all the pixels?

Unit tests for this, in the form of a drm selftest (like we have for drm_mm.c already, but for kms helpers) would be perfect. Much easier to review a testcase than do bitmath in my head :-)

Quoted 69 lines
drivers/gpu/drm/drm_damage_helper.c +44 -0
+
+	if (iter->full_update) {
+		iter->clips = 0;
+		iter->curr_clip = 0;
+		iter->num_clips = 0;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_init);
+
+/**
+ * drm_atomic_helper_damage_iter_next - advance the damage iterator
+ * @iter: The iterator to advance.
+ * @rect: Return a rectangle in fb coordinate clipped to plane src.
+ *
+ * Returns:  true if the output is valid, false if we've reached the end of the
+ * rectangle list.
+ */
+bool
+drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
+				   struct drm_rect *rect)
+{
+	bool ret = false;
+
+	if (iter->full_update) {
+		*rect = iter->plane_src;
+		iter->full_update = false;
+		return true;
+	}
+
+	while (iter->curr_clip < iter->num_clips) {
+		*rect = iter->clips[iter->curr_clip];
+		iter->curr_clip++;
+
+		if (drm_rect_intersect(rect, &iter->plane_src)) {
+			ret = true;
+			break;
+		}
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_next);
include/drm/drm_damage_helper.h +13 -0
index 217694e9168c..f1282b459a4f 100644
@@ -32,6 +32,19 @@
 #ifndef DRM_DAMAGE_HELPER_H_
 #define DRM_DAMAGE_HELPER_H_
 
+/**
+ * struct drm_atomic_helper_damage_iter - damage clip iterator
+ *
+ * This iterator tracks state needed to walk the list of damage clips.
+ */
+struct drm_atomic_helper_damage_iter {
+	const struct drm_rect *clips;
+	struct drm_rect plane_src;
+	uint32_t num_clips;
+	uint32_t curr_clip;
+	bool full_update;
+};
+
 /**
  * drm_plane_get_damage_clips_count - returns damage clips count
  * @state: Plane state

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 78

Quoted 7 lines
include/drm/drm_damage_helper.h +4 -0
@@ -59,5 +72,12 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 }
 
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+int
+drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state);

I think a for_each_damage macro would be sweet on top of these here. But easy to add later on.
-Daniel

Quoted 8 lines
include/drm/drm_damage_helper.h +3 -0
+bool
+drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
+				   struct drm_rect *rect);
 
 #endif
Signature · 1 line
2.17.1
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Reviewed-by on v1 3/14
#16 lore raw

Re: [PATCH 03/14] drm: clear plane damage during full modeset

In reply to Deepak Rawat
Quoted 23 lines · Deepak Rawat

Plane damage is irrelevant when full modeset happens so clear the damage blob property(If set by user-space). With damage blob cleared damage helper iterator will return full plane src as damage clip.

---
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++++
 include/drm/drm_damage_helper.h     | 10 ++++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +1 -0
index be83e2763c18..e06d2d5d582f 100644
@@ -31,6 +31,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_writeback.h>
+#include <drm/drm_damage_helper.h>
 #include <linux/dma-fence.h>
 
 #include "drm_crtc_helper_internal.h"

drivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct drm_atomic_state *state, · line 94

Quoted 6 lines
drivers/gpu/drm/drm_atomic_helper.c +3 -0
@@ -88,6 +89,9 @@ drm_atomic_helper_plane_changed(struct drm_atomic_state *state,
 			return;
 
 		crtc_state->planes_changed = true;
+
+		if (drm_atomic_crtc_needs_modeset(crtc_state))
+			drm_plane_clear_damage(plane_state);

I'm not 100% sure this is the best place to put this. I'm also wondering whether we should clear damage when moving planes between crtc on top of this. But I guess vmwgfx doesn't allow that, we can figure this out when the first driver with moveable planes adds damage support.

Quoted 8 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 	}
 }
 
include/drm/drm_damage_helper.h +0 -0
index f1282b459a4f..1f988f7fdd72 100644

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 76

Quoted 6 lines
include/drm/drm_damage_helper.h +3 -0
@@ -71,6 +71,16 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 				   state->fb_damage_clips->data : NULL);
 }
 
+/**
+ * drm_plane_clear_damage - clears damage blob in a plane state
+ * @state: Plane state

A bit more kerneldoc would be good. Maybe explain how that impacts the damage iterator - you get full damage after calling this, which is a bit confusing for a function called clear_damage. So definitely worth explaining.

With the kerneldoc beefed up:

Quoted 13 lines
include/drm/drm_damage_helper.h +7 -0
+ */
+static inline void drm_plane_clear_damage(struct drm_plane_state *state)
+{
+	drm_property_blob_put(state->fb_damage_clips);
+	state->fb_damage_clips = NULL;
+}
+
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
 int
 drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
Signature · 1 line
2.17.1
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#6 lore raw

Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Deepak Rawat
Quoted 318 lines · Deepak Rawat

From: Lukasz Spintzyk <[email protected]>

FB_DAMAGE_CLIPS is an optional plane property to mark damaged regions on the plane in framebuffer coordinates of the framebuffer attached to the plane.

The layout of blob data is simply an array of "struct drm_mode_rect" with maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips are not in 16.16 fixed point. As plane src in framebuffer cannot be negative so are damage clips. In damage clip, x1/y1 are inclusive and x2/y2 are exclusive.

This patch also exports the kernel internal drm_rect to userspace as drm_mode_rect. This is because "struct drm_clip_rect" is not sufficient to represent damage for current plane size.

Upper limit is set on the maximum number of damage clips to avoid overflow by user-space.

Driver which are interested in enabling FB_DAMAGE_CLIPS property for a plane should enable this property using drm_plane_enable_damage_clips.

---
 Documentation/gpu/drm-kms.rst       |  9 +++
 drivers/gpu/drm/Makefile            |  2 +-
 drivers/gpu/drm/drm_atomic.c        | 13 ++++
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++
 drivers/gpu/drm/drm_damage_helper.c | 92 +++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_mode_config.c   |  6 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |  1 +
 include/drm/drm_damage_helper.h     | 63 ++++++++++++++++++++
 include/drm/drm_mode_config.h       | 10 ++++
 include/drm/drm_plane.h             |  8 +++
 include/uapi/drm/drm_mode.h         | 19 ++++++
 11 files changed, 226 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Documentation/gpu/drm-kms.rst +9 -0
index 5dee6b8a4c12..78b66e628857 100644
@@ -542,6 +542,15 @@ Plane Composition Properties
 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
    :export:
 
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:
+
 Color Management Properties
 ---------------------------
 
drivers/gpu/drm/Makefile +1 -1
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \
 		drm_plane_helper.o drm_dp_mst_topology.o drm_atomic_helper.o \
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o \
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o drm_damage_helper.o
 
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fb_helper.o
drivers/gpu/drm/drm_atomic.c +13 -0
index 3eb061e11e2e..652e78ca1f81 100644
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb = drm_framebuffer_lookup(dev, NULL, val);
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane *plane,
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state, property, val);
 	} else {
drivers/gpu/drm/drm_atomic_helper.c +4 -0
index 80be74df7ba6..be83e2763c18 100644
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct drm_plane *plane)
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property->values[1];
+		plane->state->fb_damage_clips = NULL;
 	}
 }
 EXPORT_SYMBOL(drm_atomic_helper_plane_reset);
@@ -3598,6 +3599,7 @@ void __drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,
 
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void __drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)
 
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
drivers/gpu/drm/drm_damage_helper.c new file +92 -0
new file mode 100644
index 000000000000..3e7de5afe7f6
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/**************************************************************************
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+ **************************************************************************/
+
+#include <drm/drmP.h>
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a means to
+ * specify a list of damage rectangles on a plane in framebuffer coordinates of
+ * the framebuffer attached to the plane. In current context damage is the area
+ * of plane framebuffer (excluding the framebuffer area which is outside of
+ * plane src) that has changed since last plane update (also called page-flip).
+ * Only the area inside damage rectangles will be considered different whether
+ * currently attached framebuffer is same as framebuffer attached during last
+ * plane update or not.
+ *
+ * FB_DAMAGE_CLIPS is a hint to kernel which could be helpful for some drivers
+ * to optimize internally especially for virtual devices where each framebuffer
+ * change needs to be transmitted over network, usb, etc.
+ *
+ * Since FB_DAMAGE_CLIPS is a hint so it is an optional property. User-space can
+ * ignore the damage clips property and in that case kernel will do a full plane
+ * update. In case damage clips are provided then it is guaranteed that the area
+ * inside damage clips will be updated to plane. For efficiency kernel can do
+ * full update or more area than specified in damage clips.
+ *
+ * FB_DAMAGE_CLIPS is a blob property with the layout of blob data is simply an
+ * array of drm_mode_rect(internally drm_rect) with maximum array size limited
+ * by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips
+ * are not in 16.16 fixed point. Similar to plane src in framebuffer, damage
+ * clips cannot be negative. In damage clip, x1/y1 are inclusive and x2/y2 are
+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should be inside
+ * plane src and if any clip falls outside plane src it will be ignored and
+ * user-space won't be notified of the same. As mentioned above, sometimes
+ * kernel will do full plane update due to change in properties which can affect
+ * full plane e.g. color management properties. Also during full modeset damage
+ * is irrelevant so if provided by user-space it simply will be ignored.
+ * Whenever damage clips are ignored by kernel, user-space will not be informed.
+ * If a user-space provides damage clips which doesn't encompass the actual
+ * damage to framebuffer (since last plane update) will result in incorrect
+ * rendering during plane update.
+ *
+ * While kernel does not error for overlapped damage clips, it is discouraged.
+ */
+
+/**
+ * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
+ * @plane: plane on which to enable damage clips property
+ *
+ * This function lets driver to enable the damage clips property on a plane.
+ */
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
+{
+	struct drm_device *dev = plane->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	drm_object_attach_property(&plane->base, config->prop_fb_damage_clips,
+				   0);
+}
+EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
drivers/gpu/drm/drm_mode_config.c +6 -0
index 21e353bd3948..506f5a52733f 100644
@@ -298,6 +298,12 @@ static int drm_mode_create_standard_properties(struct drm_device *dev)
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB, "FB_DAMAGE_CLIPS",
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev, DRM_MODE_PROP_ATOMIC,
 			"ACTIVE");
 	if (!prop)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
index 23beff5d8e3c..1edbae73d6d6 100644
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane *plane)
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
include/drm/drm_damage_helper.h new file +45 -0
new file mode 100644
index 000000000000..217694e9168c
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */
+/**************************************************************************
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+ **************************************************************************/
+
+#ifndef DRM_DAMAGE_HELPER_H_
+#define DRM_DAMAGE_HELPER_H_
+
+/**
+ * drm_plane_get_damage_clips_count - returns damage clips count
+ * @state: Plane state
+ *
+ * Returns: Number of clips in plane fb_damage_clips blob property.
+ */
+static inline uint32_t
+drm_plane_get_damage_clips_count(const struct drm_plane_state *state)
+{
+	return (state && state->fb_damage_clips) ?
+		state->fb_damage_clips->length/sizeof(struct drm_rect) : 0;

include/drm/drm_damage_helper.h line 55

Quoted 10 lines
include/drm/drm_damage_helper.h new file +10 -0
+}
+
+/**
+ * drm_plane_get_damage_clips - returns damage clips
+ * @state: Plane state
+ *
+ * Returns: Clips in plane fb_damage_clips blob property.
+ */
+static inline struct drm_rect *
+drm_plane_get_damage_clips(const struct drm_plane_state *state)

Wrong type, this needs to be the uapi drm_mode_rect, not the internal rect. The damge_iter needs to then convert this. Noticed while reviewing patch 4, which also uses the wrong type.
-Daniel

Quoted 81 lines
include/drm/drm_damage_helper.h new file +8 -0
+{
+	return (struct drm_rect *)((state && state->fb_damage_clips) ?
+				   state->fb_damage_clips->data : NULL);
+}
+
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+
+#endif
include/drm/drm_mode_config.h +10 -0
index a0b202e1d69a..8ccb5ddcd99d 100644
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark damaged
+	 * regions on the plane in framebuffer coordinates of the framebuffer
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +8 -0
index 8a152dc16ea5..bb8b0934119c 100644
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that changed
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
include/uapi/drm/drm_mode.h +19 -0
index 8d67243952f4..862ea0893e2e 100644
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area this
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property FB_DAMAGE_CLIPS.
+ */
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif
Signature · 1 line
2.17.1
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#22 lore raw

Re: [PATCH 04/14] drm: add helper to implement legacy dirtyfb

In reply to Deepak Rawat
Quoted 29 lines · Deepak Rawat

From: Rob Clark <[email protected]>

Add an atomic helper to implement dirtyfb support. This is needed to support DSI command-mode panels with x11 userspace (ie. when we can't rely on pageflips to trigger a flush to the panel).

v2: Modified the helper to use plane fb_damage_clips property and removed plane_state::dirty flag.

---
 drivers/gpu/drm/drm_damage_helper.c | 116 ++++++++++++++++++++++++++++
 include/drm/drm_damage_helper.h     |   4 +
 2 files changed, 120 insertions(+)
drivers/gpu/drm/drm_damage_helper.c +1 -0
index 7d70f6001889..b016ec375295 100644
@@ -26,6 +26,7 @@
  *
  * Authors:
  * Deepak Rawat <[email protected]>
+ * Rob Clark <[email protected]>
  *
  **************************************************************************/
 

drivers/gpu/drm/drm_damage_helper.c static int convert_fixed_to_32(int fixed) · line 87

Quoted 6 lines
drivers/gpu/drm/drm_damage_helper.c +3 -0
@@ -81,6 +82,21 @@ static int convert_fixed_to_32(int fixed)
 	return ((fixed >> 15) & 1) + (fixed >> 16);
 }
 
+static void convert_clip_rect_to_rect(const struct drm_clip_rect *src,
+				      struct drm_rect *dest,
+				      uint32_t num_clips)

I think we need to use drm_mode_rect here, because it's essentially the uapi one.

Quoted 52 lines
drivers/gpu/drm/drm_damage_helper.c +45 -0
+{
+	while (num_clips > 0) {
+		dest->x1 = src->x1;
+		dest->y1 = src->y1;
+		dest->x2 = src->x2;
+		dest->y2 = src->y2;
+		src++;
+		dest++;
+		num_clips--;
+	}
+}
+
 /**
  * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
  * @plane: plane on which to enable damage clips property
@@ -183,3 +199,103 @@ drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
 	return ret;
 }
 EXPORT_SYMBOL(drm_atomic_helper_damage_iter_next);
+
+/**
+ * drm_atomic_helper_dirtyfb - helper for dirtyfb
+ * @fb: DRM framebuffer
+ * @file_priv: drm file for the ioctl call
+ * @flags: dirty fb annotate flags
+ * @color: color for annotate fill
+ * @clips: dirty region
+ * @num_clips: count of clip in clips
+ *
+ * A helper to implement drm_framebuffer_funcs::dirty. This helper will do a
+ * full update in case of annotate flags.
+ *
+ * Returns: Zero on success, negative errno on failure.
+ */
+int drm_atomic_helper_dirtyfb(struct drm_framebuffer *fb,
+			      struct drm_file *file_priv, unsigned flags,
+			      unsigned color, struct drm_clip_rect *clips,
+			      unsigned num_clips)
+{
+	struct drm_modeset_acquire_ctx ctx;
+	struct drm_property_blob *damage = NULL;
+	struct drm_atomic_state *state;
+	struct drm_plane *plane;
+	struct drm_rect *rects = NULL;
+	int ret = 0;
+
+	/*
+	 * When called from ioctl, we are interruptable, but not when called
+	 * internally (ie. defio worker)
+	 */
+	drm_modeset_acquire_init(&ctx,
+		file_priv ? DRM_MODESET_ACQUIRE_INTERRUPTIBLE : 0);

drivers/gpu/drm/drm_damage_helper.c drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter, · line 244

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+
+	state = drm_atomic_state_alloc(fb->dev);
+	if (!state) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	state->acquire_ctx = &ctx;
+
+	WARN_ONCE(flags & DRM_MODE_FB_DIRTY_FLAGS,
+		  "Full update for dirtyfb with DRM_MODE_FB_DIRTY_FLAGS\n");

User-triggerable backtrace isn't good. I think you just need to handle the FILL/COPY meaning.

For COPY we need to ignore all the odd rects (iirc those are the sources). FILL we can just handle. See e.g. what tinydrm_merge_clips does.

Otherwise I think this looks good.
-Daniel

Quoted 74 lines
drivers/gpu/drm/drm_damage_helper.c +57 -0
+
+	if (!(flags & DRM_MODE_FB_DIRTY_FLAGS)) {
+		rects = kcalloc(num_clips, sizeof(*rects), GFP_KERNEL);
+		if (!rects) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		convert_clip_rect_to_rect(clips, rects, num_clips);
+		damage = drm_property_create_blob(fb->dev,
+						  num_clips * sizeof(*rects),
+						  rects);
+		if (IS_ERR(damage)) {
+			ret = PTR_ERR(damage);
+			damage = NULL;
+			goto out;
+		}
+	}
+
+retry:
+	drm_for_each_plane(plane, fb->dev) {
+		struct drm_plane_state *plane_state;
+
+		if (plane->state->fb != fb)
+			continue;
+
+		plane_state = drm_atomic_get_plane_state(state, plane);
+		if (IS_ERR(plane_state)) {
+			ret = PTR_ERR(plane_state);
+			goto out;
+		}
+
+		drm_property_replace_blob(&plane_state->fb_damage_clips,
+					  damage);
+	}
+
+	ret = drm_atomic_commit(state);
+
+out:
+	if (ret == -EDEADLK) {
+		drm_atomic_state_clear(state);
+		ret = drm_modeset_backoff(&ctx);
+		if (!ret)
+			goto retry;
+	}
+
+	drm_property_blob_put(damage);
+	kfree(rects);
+	drm_atomic_state_put(state);
+
+	drm_modeset_drop_locks(&ctx);
+	drm_modeset_acquire_fini(&ctx);
+
+	return ret;
+
+}
+EXPORT_SYMBOL(drm_atomic_helper_dirtyfb);
include/drm/drm_damage_helper.h +4 -0
index 1f988f7fdd72..9d5c9444761b 100644
@@ -89,5 +89,9 @@ drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
 bool
 drm_atomic_helper_damage_iter_next(struct drm_atomic_helper_damage_iter *iter,
 				   struct drm_rect *rect);
+int drm_atomic_helper_dirtyfb(struct drm_framebuffer *fb,
+			      struct drm_file *file_priv, unsigned flags,
+			      unsigned color, struct drm_clip_rect *clips,
+			      unsigned num_clips);
 
 #endif
Signature · 1 line
2.17.1
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#7 lore raw

Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Deepak Rawat
Quoted 178 lines · Deepak Rawat

From: Lukasz Spintzyk <[email protected]>

FB_DAMAGE_CLIPS is an optional plane property to mark damaged regions on the plane in framebuffer coordinates of the framebuffer attached to the plane.

The layout of blob data is simply an array of "struct drm_mode_rect" with maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips are not in 16.16 fixed point. As plane src in framebuffer cannot be negative so are damage clips. In damage clip, x1/y1 are inclusive and x2/y2 are exclusive.

This patch also exports the kernel internal drm_rect to userspace as drm_mode_rect. This is because "struct drm_clip_rect" is not sufficient to represent damage for current plane size.

Upper limit is set on the maximum number of damage clips to avoid overflow by user-space.

Driver which are interested in enabling FB_DAMAGE_CLIPS property for a plane should enable this property using drm_plane_enable_damage_clips.

---
 Documentation/gpu/drm-kms.rst       |  9 +++
 drivers/gpu/drm/Makefile            |  2 +-
 drivers/gpu/drm/drm_atomic.c        | 13 ++++
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++
 drivers/gpu/drm/drm_damage_helper.c | 92 +++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_mode_config.c   |  6 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |  1 +
 include/drm/drm_damage_helper.h     | 63 ++++++++++++++++++++
 include/drm/drm_mode_config.h       | 10 ++++
 include/drm/drm_plane.h             |  8 +++
 include/uapi/drm/drm_mode.h         | 19 ++++++
 11 files changed, 226 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Documentation/gpu/drm-kms.rst +9 -0
index 5dee6b8a4c12..78b66e628857 100644
@@ -542,6 +542,15 @@ Plane Composition Properties
 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
    :export:
 
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:
+
 Color Management Properties
 ---------------------------
 
drivers/gpu/drm/Makefile +1 -1
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \
 		drm_plane_helper.o drm_dp_mst_topology.o drm_atomic_helper.o \
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o \
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o drm_damage_helper.o
 
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fb_helper.o
drivers/gpu/drm/drm_atomic.c +13 -0
index 3eb061e11e2e..652e78ca1f81 100644
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb = drm_framebuffer_lookup(dev, NULL, val);
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane,
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane *plane,
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state, property, val);
 	} else {
drivers/gpu/drm/drm_atomic_helper.c +4 -0
index 80be74df7ba6..be83e2763c18 100644
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct drm_plane *plane)
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property->values[1];
+		plane->state->fb_damage_clips = NULL;
 	}
 }
 EXPORT_SYMBOL(drm_atomic_helper_plane_reset);
@@ -3598,6 +3599,7 @@ void __drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,
 
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void __drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)
 
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
drivers/gpu/drm/drm_damage_helper.c new file +32 -0
new file mode 100644
index 000000000000..3e7de5afe7f6
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/**************************************************************************
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+ **************************************************************************/
+
+#include <drm/drmP.h>

drivers/gpu/drm/drm_damage_helper.c line 42

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c new file +10 -0
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a means to
+ * specify a list of damage rectangles on a plane in framebuffer coordinates of
+ * the framebuffer attached to the plane. In current context damage is the area
+ * of plane framebuffer (excluding the framebuffer area which is outside of
+ * plane src)

Not sure why the plane src coordinates need to be mentioned here. The damage is just the part of the fb that has changed. Whether or not it extends past the src coordinates is totally irrelevant as far as I can see.

Quoted 175 lines

that has changed since last plane update (also called page-flip).
+ * Only the area inside damage rectangles will be considered different whether
+ * currently attached framebuffer is same as framebuffer attached during last
+ * plane update or not.
+ *
+ * FB_DAMAGE_CLIPS is a hint to kernel which could be helpful for some drivers
+ * to optimize internally especially for virtual devices where each framebuffer
+ * change needs to be transmitted over network, usb, etc.
+ *
+ * Since FB_DAMAGE_CLIPS is a hint so it is an optional property. User-space can
+ * ignore the damage clips property and in that case kernel will do a full plane
+ * update. In case damage clips are provided then it is guaranteed that the area
+ * inside damage clips will be updated to plane. For efficiency kernel can do
+ * full update or more area than specified in damage clips.
+ *
+ * FB_DAMAGE_CLIPS is a blob property with the layout of blob data is simply an
+ * array of drm_mode_rect(internally drm_rect) with maximum array size limited
+ * by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates, damage clips
+ * are not in 16.16 fixed point. Similar to plane src in framebuffer, damage
+ * clips cannot be negative. In damage clip, x1/y1 are inclusive and x2/y2 are
+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should be inside
+ * plane src and if any clip falls outside plane src it will be ignored and
+ * user-space won't be notified of the same. As mentioned above, sometimes
+ * kernel will do full plane update due to change in properties which can affect
+ * full plane e.g. color management properties. Also during full modeset damage
+ * is irrelevant so if provided by user-space it simply will be ignored.
+ * Whenever damage clips are ignored by kernel, user-space will not be informed.
+ * If a user-space provides damage clips which doesn't encompass the actual
+ * damage to framebuffer (since last plane update) will result in incorrect
+ * rendering during plane update.
+ *
+ * While kernel does not error for overlapped damage clips, it is discouraged.
+ */

drivers/gpu/drm/drm_damage_helper.c new file +16 -0
+
+/**
+ * drm_plane_enable_fb_damage_clips - enables plane fb damage clips property
+ * @plane: plane on which to enable damage clips property
+ *
+ * This function lets driver to enable the damage clips property on a plane.
+ */
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
+{
+	struct drm_device *dev = plane->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	drm_object_attach_property(&plane->base, config->prop_fb_damage_clips,
+				   0);
+}
+EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
drivers/gpu/drm/drm_mode_config.c +6 -0
index 21e353bd3948..506f5a52733f 100644
@@ -298,6 +298,12 @@ static int drm_mode_create_standard_properties(struct drm_device *dev)
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB, "FB_DAMAGE_CLIPS",
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev, DRM_MODE_PROP_ATOMIC,
 			"ACTIVE");
 	if (!prop)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
index 23beff5d8e3c..1edbae73d6d6 100644
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane *plane)
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
include/drm/drm_damage_helper.h new file +63 -0
new file mode 100644
index 000000000000..217694e9168c
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */
+/**************************************************************************
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+ **************************************************************************/
+
+#ifndef DRM_DAMAGE_HELPER_H_
+#define DRM_DAMAGE_HELPER_H_
+
+/**
+ * drm_plane_get_damage_clips_count - returns damage clips count
+ * @state: Plane state
+ *
+ * Returns: Number of clips in plane fb_damage_clips blob property.
+ */
+static inline uint32_t
+drm_plane_get_damage_clips_count(const struct drm_plane_state *state)
+{
+	return (state && state->fb_damage_clips) ?
+		state->fb_damage_clips->length/sizeof(struct drm_rect) : 0;
+}
+
+/**
+ * drm_plane_get_damage_clips - returns damage clips
+ * @state: Plane state
+ *
+ * Returns: Clips in plane fb_damage_clips blob property.
+ */
+static inline struct drm_rect *
+drm_plane_get_damage_clips(const struct drm_plane_state *state)
+{
+	return (struct drm_rect *)((state && state->fb_damage_clips) ?
+				   state->fb_damage_clips->data : NULL);
+}
+
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+
+#endif
include/drm/drm_mode_config.h +10 -0
index a0b202e1d69a..8ccb5ddcd99d 100644
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark damaged
+	 * regions on the plane in framebuffer coordinates of the framebuffer
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +0 -0
index 8a152dc16ea5..bb8b0934119c 100644

include/drm/drm_plane.h struct drm_plane_state { · line 191

Quoted 10 lines
include/drm/drm_plane.h +7 -0
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that changed
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;

Don't we have a better place for this? Next to some other props etc. in the plane state?

What does "in framebuffer coordinates" mean? Integers, some fixed point representation?

Quoted 41 lines
include/drm/drm_plane.h +1 -0
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
include/uapi/drm/drm_mode.h +19 -0
index 8d67243952f4..862ea0893e2e 100644
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area this
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property FB_DAMAGE_CLIPS.
+ */
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#18 lore raw

Re: [PATCH 03/14] drm: clear plane damage during full modeset

In reply to Deepak Rawat
Quoted 23 lines · Deepak Rawat

Plane damage is irrelevant when full modeset happens so clear the damage blob property(If set by user-space). With damage blob cleared damage helper iterator will return full plane src as damage clip.

---
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++++
 include/drm/drm_damage_helper.h     | 10 ++++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +1 -0
index be83e2763c18..e06d2d5d582f 100644
@@ -31,6 +31,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_writeback.h>
+#include <drm/drm_damage_helper.h>
 #include <linux/dma-fence.h>
 
 #include "drm_crtc_helper_internal.h"

drivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct drm_atomic_state *state, · line 94

Quoted 6 lines
drivers/gpu/drm/drm_atomic_helper.c +3 -0
@@ -88,6 +89,9 @@ drm_atomic_helper_plane_changed(struct drm_atomic_state *state,
 			return;
 
 		crtc_state->planes_changed = true;
+
+		if (drm_atomic_crtc_needs_modeset(crtc_state))
+			drm_plane_clear_damage(plane_state);

So if we sometimes magically clear it, where do we reinitialize it with the user provided damage for the next update?

Quoted 9 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 	}
 }
 
include/drm/drm_damage_helper.h +0 -0
index f1282b459a4f..1f988f7fdd72 100644
@@ -71,6 +71,16 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 				   state->fb_damage_clips->data : NULL);

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 81

Quoted 10 lines
include/drm/drm_damage_helper.h +8 -0
 }
 
+/**
+ * drm_plane_clear_damage - clears damage blob in a plane state
+ * @state: Plane state
+ */
+static inline void drm_plane_clear_damage(struct drm_plane_state *state)
+{
+	drm_property_blob_put(state->fb_damage_clips);
+	state->fb_damage_clips = NULL;

Ah. So you're trying to clear out the user provided damage here. That doesn't really seem like sane sematics to me. Either the user provided damage should stick always instead of just sometimes, or it should always be cleared once the update is done (ie. could clear in duplicate_state() in that case).

Quoted 12 lines
include/drm/drm_damage_helper.h +2 -0
+}
+
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
 int
 drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#19 lore raw

Re: [PATCH 03/14] drm: clear plane damage during full modeset

Quoted 45 lines · Ville Syrjälä
Quoted 23 lines · Deepak Rawat

Plane damage is irrelevant when full modeset happens so clear the damage blob property(If set by user-space). With damage blob cleared damage helper iterator will return full plane src as damage clip.

---
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++++
 include/drm/drm_damage_helper.h     | 10 ++++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +1 -0
index be83e2763c18..e06d2d5d582f 100644
@@ -31,6 +31,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_writeback.h>
+#include <drm/drm_damage_helper.h>
 #include <linux/dma-fence.h>
 
 #include "drm_crtc_helper_internal.h"

drivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct drm_atomic_state *state, · line 94

Quoted 6 lines
drivers/gpu/drm/drm_atomic_helper.c +3 -0
@@ -88,6 +89,9 @@ drm_atomic_helper_plane_changed(struct drm_atomic_state *state,
 			return;
 
 		crtc_state->planes_changed = true;
+
+		if (drm_atomic_crtc_needs_modeset(crtc_state))
+			drm_plane_clear_damage(plane_state);

So if we sometimes magically clear it, where do we reinitialize it with the user provided damage for the next update?

Quoted 11 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 	}
 }
 
include/drm/drm_damage_helper.h +0 -0
index f1282b459a4f..1f988f7fdd72 100644
@@ -71,6 +71,16 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 				   state->fb_damage_clips->data : NULL);
 }
 
Quoted 14 lines

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 81

Quoted 8 lines
include/drm/drm_damage_helper.h +8 -0
+/**
+ * drm_plane_clear_damage - clears damage blob in a plane state
+ * @state: Plane state
+ */
+static inline void drm_plane_clear_damage(struct drm_plane_state *state)
+{
+	drm_property_blob_put(state->fb_damage_clips);
+	state->fb_damage_clips = NULL;

Ah. So you're trying to clear out the user provided damage here. That doesn't really seem like sane sematics to me. Either the user provided damage should stick always instead of just sometimes, or it should always be cleared once the update is done (ie. could clear in duplicate_state() in that case).

Read patch 2 more careful, we always reset damage when duplicating state. So it's one-shot (like fences).

But you might also want to ignore damage (and do a full upload) in other cases, like a modeset, or when you move your plane between crtc. This function here is for that.

And yes this should be explained in the kerneldoc. And _clear_damage is not a great name for what it's meant to do, that's just what it does. I think in the past I've suggested _full_damage() or something like that. -Daniel

Quoted 21 lines
Quoted 13 lines
include/drm/drm_damage_helper.h +2 -0
+}
+
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
 int
 drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#20 lore raw

Re: [PATCH 03/14] drm: clear plane damage during full modeset

Quoted 61 lines · Daniel Vetter
Quoted 45 lines · Ville Syrjälä
Quoted 23 lines · Deepak Rawat

Plane damage is irrelevant when full modeset happens so clear the damage blob property(If set by user-space). With damage blob cleared damage helper iterator will return full plane src as damage clip.

---
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++++
 include/drm/drm_damage_helper.h     | 10 ++++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +1 -0
index be83e2763c18..e06d2d5d582f 100644
@@ -31,6 +31,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_writeback.h>
+#include <drm/drm_damage_helper.h>
 #include <linux/dma-fence.h>
 
 #include "drm_crtc_helper_internal.h"

drivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct drm_atomic_state *state, · line 94

Quoted 6 lines
drivers/gpu/drm/drm_atomic_helper.c +3 -0
@@ -88,6 +89,9 @@ drm_atomic_helper_plane_changed(struct drm_atomic_state *state,
 			return;
 
 		crtc_state->planes_changed = true;
+
+		if (drm_atomic_crtc_needs_modeset(crtc_state))
+			drm_plane_clear_damage(plane_state);

So if we sometimes magically clear it, where do we reinitialize it with the user provided damage for the next update?

Quoted 11 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 	}
 }
 
include/drm/drm_damage_helper.h +0 -0
index f1282b459a4f..1f988f7fdd72 100644
@@ -71,6 +71,16 @@ drm_plane_get_damage_clips(const struct drm_plane_state *state)
 				   state->fb_damage_clips->data : NULL);
 }
 
Quoted 14 lines

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 81

Quoted 8 lines
include/drm/drm_damage_helper.h +8 -0
+/**
+ * drm_plane_clear_damage - clears damage blob in a plane state
+ * @state: Plane state
+ */
+static inline void drm_plane_clear_damage(struct drm_plane_state *state)
+{
+	drm_property_blob_put(state->fb_damage_clips);
+	state->fb_damage_clips = NULL;

Ah. So you're trying to clear out the user provided damage here. That doesn't really seem like sane sematics to me. Either the user provided damage should stick always instead of just sometimes, or it should always be cleared once the update is done (ie. could clear in duplicate_state() in that case).

Quoted 2 lines

Read patch 2 more careful, we always reset damage when duplicating state. So it's one-shot (like fences).

Ah, thanks for pointing it out. Was actually in patch 1 :)

Quoted 36 lines

But you might also want to ignore damage (and do a full upload) in other cases, like a modeset, or when you move your plane between crtc. This function here is for that.

And yes this should be explained in the kerneldoc. And _clear_damage is not a great name for what it's meant to do, that's just what it does. I think in the past I've suggested _full_damage() or something like that. -Daniel

Quoted 21 lines
Quoted 13 lines
include/drm/drm_damage_helper.h +2 -0
+}
+
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
 int
 drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter,
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Signature · 6 lines
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#3 lore raw

RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Daniel Vetter
Quoted 13 lines

Documentation/gpu/drm-kms.rst Plane Composition Properties · line 552

Quoted 10 lines
Documentation/gpu/drm-kms.rst +8 -0
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:

You forgot to include your nice kerneldoc from the header. Please run

Quoted 3 lines
$ make htmldocs

and make sure the generated stuff looks all nice and is complete.

Thanks for the review Daniel, oops I added those docs later in header and forgot to update here.

Quoted 222 lines

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o

Quoted 10 lines
Documentation/gpu/drm-kms.rst +1 -0
+
 Color Management Properties
 ---------------------------
drivers/gpu/drm/Makefile +0 -0
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o

drm_dp_helper.o drm_probe_helper.o \

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 35

Quoted 1 line
drivers/gpu/drm/Makefile +0 -0
 		drm_plane_helper.o drm_dp_mst_topology.o

drm_atomic_helper.o \

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 36

Quoted 1 line
drivers/gpu/drm/Makefile +0 -0
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o

\

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \

Quoted 3 lines
drivers/gpu/drm/Makefile +1 -1
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o

drm_damage_helper.o

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 40

Quoted 3 lines
drivers/gpu/drm/Makefile +0 -0
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) +=

drm_fb_helper.o

Review

Quoted 1 line
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic.c

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct

Quoted 4 lines

index 3eb061e11e2e..652e78ca1f81 100644

drivers/gpu/drm/drm_atomic.c +0 -0
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct

drm_plane *plane,

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct drm_plane *plane,

Quoted 8 lines
drivers/gpu/drm/drm_atomic.c +2 -0
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb =

drm_framebuffer_lookup(dev, NULL, val);

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct

Quoted 1 line
drivers/gpu/drm/drm_atomic.c +0 -0
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct

drm_plane *plane,

drivers/gpu/drm/drm_atomic.c drm_atomic_plane_get_property(struct drm_plane

Quoted 15 lines
drivers/gpu/drm/drm_atomic.c +8 -0
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane

*plane,

drivers/gpu/drm/drm_atomic.c drm_atomic_plane_get_property(struct drm_plane *plane,

Quoted 8 lines
drivers/gpu/drm/drm_atomic.c +3 -0
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state,

property, val);

Review

Quoted 2 lines
drivers/gpu/drm/drm_atomic.c +0 -0
 	} else {
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic_helper.c

drivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct

Quoted 4 lines

index 80be74df7ba6..be83e2763c18 100644

drivers/gpu/drm/drm_atomic_helper.c +0 -0
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct

drm_plane *plane)

drivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct drm_plane *plane) · line 3579

Quoted 5 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property-
 values[1];
+		plane->state->fb_damage_clips = NULL;

No need to set to 0, we require kzalloc.

drivers/gpu/drm/drm_atomic_helper.c void

Quoted 4 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 	}
 }

EXPORT_SYMBOL(drm_atomic_helper_plane_reset);

drivers/gpu/drm/drm_atomic_helper.c +0 -0
@@ -3598,6 +3599,7 @@ void

__drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,

drivers/gpu/drm/drm_atomic_helper.c void

Quoted 8 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void

__drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)

Review

Quoted 9 lines
drivers/gpu/drm/drm_atomic_helper.c +2 -0
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_damage_helper.c

drivers/gpu/drm/drm_damage_helper.c line 1

Quoted 7 lines

new file mode 100644
index 000000000000..3e7de5afe7f6

drivers/gpu/drm/drm_damage_helper.c new file +1 -0
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT

+/*********************************************************
*****************

Quoted 7 lines
drivers/gpu/drm/drm_damage_helper.c new file +7 -0
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including

drivers/gpu/drm/drm_damage_helper.c

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c new file +10 -0
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR

drivers/gpu/drm/drm_damage_helper.c line 19

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY,

drivers/gpu/drm/drm_damage_helper.c line 20

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN

NO EVENT SHALL

drivers/gpu/drm/drm_damage_helper.c line 21

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE

FOR ANY CLAIM,

drivers/gpu/drm/drm_damage_helper.c line 22

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF

CONTRACT, TORT OR

drivers/gpu/drm/drm_damage_helper.c line 23

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

SOFTWARE OR THE

drivers/gpu/drm/drm_damage_helper.c line 31

Quoted 6 lines
drivers/gpu/drm/drm_damage_helper.c new file +6 -0
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+

**********************************************************
****************/

drivers/gpu/drm/drm_damage_helper.c

Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c new file +8 -0
+
+#include <drm/drmP.h>
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a

means to

drivers/gpu/drm/drm_damage_helper.c line 38

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * specify a list of damage rectangles on a plane in framebuffer

coordinates of

drivers/gpu/drm/drm_damage_helper.c line 39

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * the framebuffer attached to the plane. In current context damage is

the area

drivers/gpu/drm/drm_damage_helper.c line 40

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * of plane framebuffer (excluding the framebuffer area which is outside

of

drivers/gpu/drm/drm_damage_helper.c line 41

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * plane src) that has changed since last plane update (also called page-

flip).

drivers/gpu/drm/drm_damage_helper.c line 42

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * Only the area inside damage rectangles will be considered different

whether

drivers/gpu/drm/drm_damage_helper.c line 43

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * currently attached framebuffer is same as framebuffer attached during

last

drivers/gpu/drm/drm_damage_helper.c

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+ * plane update or not.
+ *
+ * FB_DAMAGE_CLIPS is a hint to kernel which could be helpful for some

drivers

drivers/gpu/drm/drm_damage_helper.c line 47

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * to optimize internally especially for virtual devices where each

framebuffer

drivers/gpu/drm/drm_damage_helper.c

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+ * change needs to be transmitted over network, usb, etc.
+ *
+ * Since FB_DAMAGE_CLIPS is a hint so it is an optional property. User-

space can

drivers/gpu/drm/drm_damage_helper.c line 51

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * ignore the damage clips property and in that case kernel will do a full

plane

drivers/gpu/drm/drm_damage_helper.c line 52

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * update. In case damage clips are provided then it is guaranteed that the

area

drivers/gpu/drm/drm_damage_helper.c line 53

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * inside damage clips will be updated to plane. For efficiency kernel can

do

drivers/gpu/drm/drm_damage_helper.c

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+ * full update or more area than specified in damage clips.
+ *
+ * FB_DAMAGE_CLIPS is a blob property with the layout of blob data is

simply an

drivers/gpu/drm/drm_damage_helper.c line 57

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * array of drm_mode_rect(internally drm_rect) with maximum array size

limited

drivers/gpu/drm/drm_damage_helper.c line 58

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates,

damage clips

drivers/gpu/drm/drm_damage_helper.c line 60

Quoted 2 lines
drivers/gpu/drm/drm_damage_helper.c new file +2 -0
+ * are not in 16.16 fixed point. Similar to plane src in framebuffer, damage
+ * clips cannot be negative. In damage clip, x1/y1 are inclusive and x2/y2

are

drivers/gpu/drm/drm_damage_helper.c

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should be

inside

drivers/gpu/drm/drm_damage_helper.c line 65

Quoted 2 lines
drivers/gpu/drm/drm_damage_helper.c new file +2 -0
+ * plane src and if any clip falls outside plane src it will be ignored and
+ * user-space won't be notified of the same. As mentioned above,

sometimes

drivers/gpu/drm/drm_damage_helper.c line 66

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * kernel will do full plane update due to change in properties which can

affect

drivers/gpu/drm/drm_damage_helper.c line 67

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * full plane e.g. color management properties. Also during full modeset

damage

drivers/gpu/drm/drm_damage_helper.c line 69

Quoted 2 lines
drivers/gpu/drm/drm_damage_helper.c new file +2 -0
+ * is irrelevant so if provided by user-space it simply will be ignored.
+ * Whenever damage clips are ignored by kernel, user-space will not be

informed.

drivers/gpu/drm/drm_damage_helper.c line 70

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * If a user-space provides damage clips which doesn't encompass the

actual

drivers/gpu/drm/drm_damage_helper.c line 72

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * damage to framebuffer (since last plane update) will result in incorrect

s/will/can/ - it's an optimization

drivers/gpu/drm/drm_damage_helper.c

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+ * rendering during plane update.
+ *
+ * While kernel does not error for overlapped damage clips, it is
Quoted 7 lines

discouraged.

Some comments on the doc:

- You kinda explain it, but in convoluted way: I'd add a clear sentence that userspace _must_ always render the entire visible fb, since the driver is free to read more. Otherwise there can be corruptions.

Agreed , will make it more clear for V2

Quoted 4 lines

- Why no input validation on the damage clips against the framebuffer size? Ime not validating just leads to funny driver bugs down the road, so what's the use-case you have in mind here?

My motivation behind not informing user-space if damage is outside plane src, sent during modeset(where it should be provided) or damage is outside framebuffer coordinate (simply put damage clips are invalid) is that it's a hint. The contract between kernel and user-space is simple, if damage clips are valid, kernel might use them (but not always) otherwise they will be ignored. Damage can be ignored deep in the stack (like network), so all the input validation for nothing.

Also, what all to consider in input validation ?
* clips are outside plane src
* damage clips sent during modeset should error ?
* any other properties.

This will lead to a complex input validation during modeset check like if some drivers are OK with damage while scaling whereas others cannot support, should this error?

However I am alright doing input validation if this becomes clear what kind of validation to actually do. My understanding of drm core is limited.

Quoted 4 lines

- A short paragraph on how drivers are supposed to implement this would be good, with references to functions (will be automatically converted to hyperlinks)

Agreed

Quoted 217 lines

drivers/gpu/drm/drm_damage_helper.c

Quoted 5 lines
drivers/gpu/drm/drm_damage_helper.c new file +4 -0
+ */
+
+/**
+ * drm_plane_enable_fb_damage_clips - enables plane fb damage clips

property

Review

Quoted 15 lines
drivers/gpu/drm/drm_damage_helper.c new file +10 -0
+ * @plane: plane on which to enable damage clips property
+ *
+ * This function lets driver to enable the damage clips property on a plane.
+ */
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
+{
+	struct drm_device *dev = plane->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	drm_object_attach_property(&plane->base, config-

prop_fb_damage_clips,

drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+				   0);
+}
+EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_mode_config.c

drivers/gpu/drm/drm_mode_config.c static int

Quoted 4 lines

index 21e353bd3948..506f5a52733f 100644

drivers/gpu/drm/drm_mode_config.c +0 -0
@@ -298,6 +298,12 @@ static int

drm_mode_create_standard_properties(struct drm_device *dev)

drivers/gpu/drm/drm_mode_config.c static int drm_mode_create_standard_properties(struct drm_device *dev)

Quoted 4 lines
drivers/gpu/drm/drm_mode_config.c +1 -0
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB,

"FB_DAMAGE_CLIPS",

drivers/gpu/drm/drm_mode_config.c static int drm_mode_create_standard_properties(struct drm_device *dev)

Quoted 6 lines
drivers/gpu/drm/drm_mode_config.c +5 -0
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev,

DRM_MODE_PROP_ATOMIC,

Review

Quoted 3 lines
drivers/gpu/drm/drm_mode_config.c +0 -0
 			"ACTIVE");
 	if (!prop)
(quoted diff) +0 -0

b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_du_plane_reset(struct drm_plane

Quoted 4 lines

index 23beff5d8e3c..1edbae73d6d6 100644

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +0 -0
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane

*plane)

Review

Quoted 8 lines
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
(quoted diff) +0 -0

b/include/drm/drm_damage_helper.h

include/drm/drm_damage_helper.h line 1

Quoted 7 lines

new file mode 100644
index 000000000000..217694e9168c

include/drm/drm_damage_helper.h new file +1 -0
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */

+/*********************************************************
*****************

Quoted 7 lines
include/drm/drm_damage_helper.h new file +7 -0
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including

include/drm/drm_damage_helper.h line 18

Quoted 10 lines
include/drm/drm_damage_helper.h new file +10 -0
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR

include/drm/drm_damage_helper.h line 19

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY,

include/drm/drm_damage_helper.h line 20

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN

NO EVENT SHALL

include/drm/drm_damage_helper.h line 21

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE

FOR ANY CLAIM,

include/drm/drm_damage_helper.h line 22

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF

CONTRACT, TORT OR

include/drm/drm_damage_helper.h line 23

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

SOFTWARE OR THE

include/drm/drm_damage_helper.h line 29

Quoted 6 lines
include/drm/drm_damage_helper.h new file +6 -0
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+

**********************************************************
****************/

include/drm/drm_damage_helper.h

Quoted 12 lines
include/drm/drm_damage_helper.h new file +12 -0
+
+#ifndef DRM_DAMAGE_HELPER_H_
+#define DRM_DAMAGE_HELPER_H_
+
+/**
+ * drm_plane_get_damage_clips_count - returns damage clips count
+ * @state: Plane state
+ *
+ * Returns: Number of clips in plane fb_damage_clips blob property.
+ */
+static inline uint32_t
+drm_plane_get_damage_clips_count(const struct drm_plane_state

*state)

Review

Quoted 22 lines
include/drm/drm_damage_helper.h new file +21 -0
+{
+	return (state && state->fb_damage_clips) ?
+		state->fb_damage_clips->length/sizeof(struct drm_rect) : 0;
+}
+
+/**
+ * drm_plane_get_damage_clips - returns damage clips
+ * @state: Plane state
+ *
+ * Returns: Clips in plane fb_damage_clips blob property.
+ */
+static inline struct drm_rect *
+drm_plane_get_damage_clips(const struct drm_plane_state *state)
+{
+	return (struct drm_rect *)((state && state->fb_damage_clips) ?
+				   state->fb_damage_clips->data : NULL);
+}
+
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+
+#endif
(quoted diff) +0 -0

b/include/drm/drm_mode_config.h

include/drm/drm_mode_config.h struct drm_mode_config { · line 631

Quoted 9 lines

index a0b202e1d69a..8ccb5ddcd99d 100644

include/drm/drm_mode_config.h +2 -0
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark

damaged

include/drm/drm_mode_config.h struct drm_mode_config { · line 632

Quoted 1 line
include/drm/drm_mode_config.h +1 -0
+	 * regions on the plane in framebuffer coordinates of the

framebuffer

include/drm/drm_mode_config.h struct drm_mode_config {

Quoted 4 lines
include/drm/drm_mode_config.h +4 -0
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by

DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike

Quoted 11 lines
include/drm/drm_mode_config.h +3 -0
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +0 -0
index 8a152dc16ea5..bb8b0934119c 100644

include/drm/drm_plane.h struct drm_plane_state { · line 188

Quoted 7 lines
include/drm/drm_plane.h +4 -0
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that

changed

Review

Quoted 8 lines
include/drm/drm_plane.h +4 -0
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
(quoted diff) +0 -0

b/include/uapi/drm/drm_mode.h

Quoted 5 lines

index 8d67243952f4..862ea0893e2e 100644

include/uapi/drm/drm_mode.h +0 -0
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease { · line 897

Quoted 10 lines
include/uapi/drm/drm_mode.h +8 -0
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area

this

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease {

Quoted 3 lines
include/uapi/drm/drm_mode.h +3 -0
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property

FB_DAMAGE_CLIPS.

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease { · line 911

Quoted 11 lines
include/uapi/drm/drm_mode.h +8 -0
+ */
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif

Besides the small nits this looks good to me.
-Daniel

Quoted 8 lines

--
Daniel Vetter
Software Engineer, Intel Corporation
https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fblog.ff
wll.ch&amp;data=02%7C01%7Cdrawat%40vmware.com%7C0f5c199805444f2 6d01508d613cc5abe%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C 636718165752004800&amp;sdata=zKJnKe90%2BMeBsSpb4t4HfkOlLYkgXkyxsc GwVOHCyro%3D&amp;reserved=0

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#13 lore raw

RE: [PATCH 02/14] drm: add helper iterator functions for plane fb_damage_clips blob

In reply to Daniel Vetter
Quoted 48 lines

Review line 76

Quoted 6 lines

#include <drm/drm_damage_helper.h>

/**

(quoted diff) +0 -0
@@ -75,6 +76,11 @@
  * While kernel does not error for overlapped damage clips, it is

discouraged.

drivers/gpu/drm/drm_damage_helper.c

Quoted 9 lines
drivers/gpu/drm/drm_damage_helper.c +5 -0
  */
 
+static int convert_fixed_to_32(int fixed)
+{
+	return ((fixed >> 15) & 1) + (fixed >> 16);
+}
+
 /**
  * drm_plane_enable_fb_damage_clips - enables plane fb damage clips

property

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct

Quoted 2 lines
drivers/gpu/drm/drm_damage_helper.c +0 -0
  * @plane: plane on which to enable damage clips property
@@ -90,3 +96,90 @@ void drm_plane_enable_fb_damage_clips(struct

drm_plane *plane)

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 11 lines
drivers/gpu/drm/drm_damage_helper.c +8 -0
 				   0);
 }
 EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
+
+/**
+ * drm_atomic_helper_damage_iter_init - initialize the damage iterator
+ * @iter: The iterator to initialize.
+ * @old_state: old plane state for validation.
+ * @new_state: plane state from which to iterate the damage clips.
+ *
+ * Initialize an iterator that clip framebuffer damage in plane

fb_damage_clips

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 4 lines
drivers/gpu/drm/drm_damage_helper.c +4 -0
+ * blob to plane src clip. The iterator returns full plane src in case needing
+ * full update e.g. during full modeset.
+ *
+ * With this helper iterator, drivers which enabled fb_damage_clips

property can

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 4 lines
drivers/gpu/drm/drm_damage_helper.c +4 -0
+ * iterate over the damage clips that falls inside plane src during plane
+ * update.
+ *
+ * Returns: 0 on success and negative error code on error. If an error code

is

Review void drm_plane_enable_fb_damage_clips(struct · line 115

Quoted 1 line
(quoted diff) +1 -0
+ * returned then it means the plane state shouldn't update with attached

fb.

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c +3 -0
+ */
+int
+drm_atomic_helper_damage_iter_init(struct

drm_atomic_helper_damage_iter *iter,

Quoted 11 lines

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane) · line 126

Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c +8 -0
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state)
+{
+	if (!new_state || !new_state->crtc || !new_state->fb)
+		return -EINVAL;
+
+	if (!new_state->visible)
+		return -EINVAL;

Can't we handle this by iterating 0 damage rects instead? Would make the code a bit cleaner I think.

Agreed.

Quoted 11 lines
Quoted 11 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+
+	memset(iter, 0, sizeof(*iter));
+	iter->clips = drm_plane_get_damage_clips(new_state);
+	iter->num_clips = drm_plane_get_damage_clips_count(new_state);
+
+	if (!iter->clips)
+		iter->full_update = true;
+
+	if (!drm_rect_equals(&new_state->src, &old_state->src))
+		iter->full_update = true;
Quoted 8 lines

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane) · line 141

Quoted 5 lines
drivers/gpu/drm/drm_damage_helper.c +5 -0
+
+	iter->plane_src.x1 = convert_fixed_to_32(new_state->src.x1);
+	iter->plane_src.y1 = convert_fixed_to_32(new_state->src.y1);
+	iter->plane_src.x2 = convert_fixed_to_32(new_state->src.x2);
+	iter->plane_src.y2 = convert_fixed_to_32(new_state->src.y2);

I think you want to clip with the clipped rectangles here, not with the ones userspace provides.

new_state->src.x1 is the clipped one, clipped in the function drm_atomic_helper_check_plane_state. Or am I missing something ?

Quoted 8 lines

Also I think you're rounding is wrong here - I think you need to round down for x/y1, and round up for x/y2 to make sure you catch all the pixels?

Unit tests for this, in the form of a drm selftest (like we have for drm_mm.c already, but for kms helpers) would be perfect. Much easier to review a testcase than do bitmath in my head :-)

Sure I will work on a unit test case for this scenario and work on round up of plane_src.

Quoted 92 lines
Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c +7 -0
+
+	if (iter->full_update) {
+		iter->clips = 0;
+		iter->curr_clip = 0;
+		iter->num_clips = 0;
+	}
+

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+	return 0;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_init);
+
+/**
+ * drm_atomic_helper_damage_iter_next - advance the damage iterator
+ * @iter: The iterator to advance.
+ * @rect: Return a rectangle in fb coordinate clipped to plane src.
+ *
+ * Returns:  true if the output is valid, false if we've reached the end of

the

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 4 lines
drivers/gpu/drm/drm_damage_helper.c +4 -0
+ * rectangle list.
+ */
+bool
+drm_atomic_helper_damage_iter_next(struct

drm_atomic_helper_damage_iter *iter,

Review

Quoted 24 lines
drivers/gpu/drm/drm_damage_helper.c +23 -0
+				   struct drm_rect *rect)
+{
+	bool ret = false;
+
+	if (iter->full_update) {
+		*rect = iter->plane_src;
+		iter->full_update = false;
+		return true;
+	}
+
+	while (iter->curr_clip < iter->num_clips) {
+		*rect = iter->clips[iter->curr_clip];
+		iter->curr_clip++;
+
+		if (drm_rect_intersect(rect, &iter->plane_src)) {
+			ret = true;
+			break;
+		}
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_next);
(quoted diff) +0 -0

b/include/drm/drm_damage_helper.h

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct

Quoted 24 lines

index 217694e9168c..f1282b459a4f 100644

include/drm/drm_damage_helper.h +13 -0
@@ -32,6 +32,19 @@
 #ifndef DRM_DAMAGE_HELPER_H_
 #define DRM_DAMAGE_HELPER_H_
 
+/**
+ * struct drm_atomic_helper_damage_iter - damage clip iterator
+ *
+ * This iterator tracks state needed to walk the list of damage clips.
+ */
+struct drm_atomic_helper_damage_iter {
+	const struct drm_rect *clips;
+	struct drm_rect plane_src;
+	uint32_t num_clips;
+	uint32_t curr_clip;
+	bool full_update;
+};
+
 /**
  * drm_plane_get_damage_clips_count - returns damage clips count
  * @state: Plane state
@@ -59,5 +72,12 @@ drm_plane_get_damage_clips(const struct

drm_plane_state *state)

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state)

Quoted 5 lines
include/drm/drm_damage_helper.h +2 -0
 }
 
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+int
+drm_atomic_helper_damage_iter_init(struct

drm_atomic_helper_damage_iter *iter,

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 78

Quoted 2 lines
include/drm/drm_damage_helper.h +2 -0
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state);

I think a for_each_damage macro would be sweet on top of these here. But easy to add later on.
-Daniel

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct · line 80

Quoted 2 lines
include/drm/drm_damage_helper.h +2 -0
+bool
+drm_atomic_helper_damage_iter_next(struct

drm_atomic_helper_damage_iter *iter,

Quoted 2 lines
include/drm/drm_damage_helper.h +1 -0
+				   struct drm_rect *rect);
 
Quoted 13 lines
Quoted 4 lines
include/drm/drm_damage_helper.h +0 -1
 #endif
--

2.17.1

--
Daniel Vetter
Software Engineer, Intel Corporation
https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fblog.ff
wll.ch&amp;data=02%7C01%7Cdrawat%40vmware.com%7C117739774a7341f d8c0208d613cd83d5%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C 636718170756943488&amp;sdata=tNK2zTqJrq8U5U4W96wp49FZ8LAB9fl2WV UixbOSWsU%3D&amp;reserved=0

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#17 lore raw

RE: [PATCH 03/14] drm: clear plane damage during full modeset

In reply to Daniel Vetter
Quoted 21 lines

Review

Quoted 5 lines
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++++
 include/drm/drm_damage_helper.h     | 10 ++++++++++
 2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic_helper.c

drivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct

Quoted 12 lines

index be83e2763c18..e06d2d5d582f 100644

drivers/gpu/drm/drm_atomic_helper.c +1 -0
@@ -31,6 +31,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_writeback.h>
+#include <drm/drm_damage_helper.h>
 #include <linux/dma-fence.h>
 
 #include "drm_crtc_helper_internal.h"
@@ -88,6 +89,9 @@ drm_atomic_helper_plane_changed(struct

drm_atomic_state *state,

Quoted 2 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 			return;
 
Quoted 10 lines

drivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct drm_atomic_state *state, · line 94

Quoted 4 lines
drivers/gpu/drm/drm_atomic_helper.c +3 -0
 		crtc_state->planes_changed = true;
+
+		if (drm_atomic_crtc_needs_modeset(crtc_state))
+			drm_plane_clear_damage(plane_state);

I'm not 100% sure this is the best place to put this. I'm also wondering whether we should clear damage when moving planes between crtc on top of
this. But I guess vmwgfx doesn't allow that, we can figure this out when the first driver with moveable planes adds damage support.

Yes I agree this is not the best place but it was the one with minimal code change. IMO should have a separate function for clearing damage.

Quoted 17 lines

Review

Quoted 4 lines
drivers/gpu/drm/drm_atomic_helper.c +0 -0
 	}
 }
 
(quoted diff) +0 -0

b/include/drm/drm_damage_helper.h

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct

Quoted 4 lines

index f1282b459a4f..1f988f7fdd72 100644

include/drm/drm_damage_helper.h +0 -0
@@ -71,6 +71,16 @@ drm_plane_get_damage_clips(const struct

drm_plane_state *state)

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 76

Quoted 6 lines
include/drm/drm_damage_helper.h +3 -0
 				   state->fb_damage_clips->data : NULL);
 }
 
+/**
+ * drm_plane_clear_damage - clears damage blob in a plane state
+ * @state: Plane state
Quoted 6 lines

A bit more kerneldoc would be good. Maybe explain how that impacts the damage iterator - you get full damage after calling this, which is a bit confusing for a function called clear_damage. So definitely worth explaining.

With the kerneldoc beefed up:

Agreed.

Quoted 6 lines

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct · line 78

Quoted 2 lines
include/drm/drm_damage_helper.h +2 -0
+ */
+static inline void drm_plane_clear_damage(struct drm_plane_state

*state)

Quoted 11 lines

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state)

Quoted 8 lines
include/drm/drm_damage_helper.h +5 -0
+{
+	drm_property_blob_put(state->fb_damage_clips);
+	state->fb_damage_clips = NULL;
+}
+
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
 int
 drm_atomic_helper_damage_iter_init(struct

drm_atomic_helper_damage_iter *iter,

Quoted 2 lines

--
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#8 lore raw

RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Ville Syrjälä
Quoted 221 lines · Deepak Rawat
Quoted 8 lines

From: Lukasz Spintzyk <[email protected]>

FB_DAMAGE_CLIPS is an optional plane property to mark damaged regions on the plane in framebuffer coordinates of the framebuffer attached to the plane.

The layout of blob data is simply an array of "struct drm_mode_rect" with maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS.

Unlike

Quoted 14 lines

plane src coordinates, damage clips are not in 16.16 fixed point. As plane src in framebuffer cannot be negative so are damage clips. In damage clip, x1/y1 are inclusive and x2/y2 are exclusive.

This patch also exports the kernel internal drm_rect to userspace as drm_mode_rect. This is because "struct drm_clip_rect" is not sufficient to represent damage for current plane size.

Upper limit is set on the maximum number of damage clips to avoid overflow by user-space.

Driver which are interested in enabling FB_DAMAGE_CLIPS property for a plane should enable this property using drm_plane_enable_damage_clips.

Quoted 8 lines
---
 Documentation/gpu/drm-kms.rst       |  9 +++
 drivers/gpu/drm/Makefile            |  2 +-
 drivers/gpu/drm/drm_atomic.c        | 13 ++++
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++
 drivers/gpu/drm/drm_damage_helper.c | 92

+++++++++++++++++++++++++++++

Documentation/gpu/drm-

Quoted 11 lines
 drivers/gpu/drm/drm_mode_config.c   |  6 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |  1 +
 include/drm/drm_damage_helper.h     | 63 ++++++++++++++++++++
 include/drm/drm_mode_config.h       | 10 ++++
 include/drm/drm_plane.h             |  8 +++
 include/uapi/drm/drm_mode.h         | 19 ++++++
 11 files changed, 226 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Documentation/gpu/drm- renamed from Documentation/gpu/drm-kms.rst +0 -0

kms.rst

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o

Quoted 24 lines

index 5dee6b8a4c12..78b66e628857 100644

Documentation/gpu/drm-kms.rst +9 -0
@@ -542,6 +542,15 @@ Plane Composition Properties
 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
    :export:
 
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:
+
 Color Management Properties
 ---------------------------
 
drivers/gpu/drm/Makefile +0 -0
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o

drm_dp_helper.o drm_probe_helper.o \

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 35

Quoted 1 line
drivers/gpu/drm/Makefile +0 -0
 		drm_plane_helper.o drm_dp_mst_topology.o

drm_atomic_helper.o \

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 36

Quoted 1 line
drivers/gpu/drm/Makefile +0 -0
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o

\

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \

Quoted 3 lines
drivers/gpu/drm/Makefile +1 -1
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o

drm_damage_helper.o

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 40

Quoted 3 lines
drivers/gpu/drm/Makefile +0 -0
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) +=

drm_fb_helper.o

Review

Quoted 1 line
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic.c

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct

Quoted 4 lines

index 3eb061e11e2e..652e78ca1f81 100644

drivers/gpu/drm/drm_atomic.c +0 -0
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct

drm_plane *plane,

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct drm_plane *plane,

Quoted 8 lines
drivers/gpu/drm/drm_atomic.c +2 -0
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb =

drm_framebuffer_lookup(dev, NULL, val);

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct

Quoted 1 line
drivers/gpu/drm/drm_atomic.c +0 -0
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct

drm_plane *plane,

drivers/gpu/drm/drm_atomic.c drm_atomic_plane_get_property(struct drm_plane

Quoted 15 lines
drivers/gpu/drm/drm_atomic.c +8 -0
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane

*plane,

drivers/gpu/drm/drm_atomic.c drm_atomic_plane_get_property(struct drm_plane *plane,

Quoted 8 lines
drivers/gpu/drm/drm_atomic.c +3 -0
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state,

property, val);

Review

Quoted 2 lines
drivers/gpu/drm/drm_atomic.c +0 -0
 	} else {
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic_helper.c

drivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct

Quoted 4 lines

index 80be74df7ba6..be83e2763c18 100644

drivers/gpu/drm/drm_atomic_helper.c +0 -0
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct

drm_plane *plane)

drivers/gpu/drm/drm_atomic_helper.c void

Quoted 9 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property-
 values[1];
+		plane->state->fb_damage_clips = NULL;
 	}
 }

EXPORT_SYMBOL(drm_atomic_helper_plane_reset);

drivers/gpu/drm/drm_atomic_helper.c +0 -0
@@ -3598,6 +3599,7 @@ void

__drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,

drivers/gpu/drm/drm_atomic_helper.c void

Quoted 8 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void

__drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)

Review

Quoted 9 lines
drivers/gpu/drm/drm_atomic_helper.c +2 -0
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_damage_helper.c

drivers/gpu/drm/drm_damage_helper.c line 1

Quoted 7 lines

new file mode 100644
index 000000000000..3e7de5afe7f6

drivers/gpu/drm/drm_damage_helper.c new file +1 -0
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT

+/*********************************************************
*****************

Quoted 7 lines
drivers/gpu/drm/drm_damage_helper.c new file +7 -0
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including

drivers/gpu/drm/drm_damage_helper.c

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c new file +10 -0
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR

drivers/gpu/drm/drm_damage_helper.c line 19

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY,

drivers/gpu/drm/drm_damage_helper.c line 20

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN

NO EVENT SHALL

drivers/gpu/drm/drm_damage_helper.c line 21

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE

FOR ANY CLAIM,

drivers/gpu/drm/drm_damage_helper.c line 22

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF

CONTRACT, TORT OR

drivers/gpu/drm/drm_damage_helper.c line 23

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

SOFTWARE OR THE

drivers/gpu/drm/drm_damage_helper.c line 31

Quoted 6 lines
drivers/gpu/drm/drm_damage_helper.c new file +6 -0
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+

**********************************************************
****************/

drivers/gpu/drm/drm_damage_helper.c

Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c new file +8 -0
+
+#include <drm/drmP.h>
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a

means to

drivers/gpu/drm/drm_damage_helper.c line 38

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * specify a list of damage rectangles on a plane in framebuffer

coordinates of

drivers/gpu/drm/drm_damage_helper.c line 39

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * the framebuffer attached to the plane. In current context damage is

the area

Quoted 8 lines

drivers/gpu/drm/drm_damage_helper.c line 40

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * of plane framebuffer (excluding the framebuffer area which is outside

of

drivers/gpu/drm/drm_damage_helper.c line 41

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * plane src)

Not sure why the plane src coordinates need to be mentioned here. The damage is just the part of the fb that has changed. Whether or not it extends past the src coordinates is totally irrelevant as far as I can see.

Thanks Ville for the review,

Well this is plane damage property and only those clips which falls inside will be used for plane update, so outside plane src is not required.

Quoted 227 lines
Quoted 3 lines

that has changed since last plane update (also called page-flip).
+ * Only the area inside damage rectangles will be considered different

whether

Quoted 1 line

+ * currently attached framebuffer is same as framebuffer attached during

last

Quoted 3 lines

+ * plane update or not.
+ *
+ * FB_DAMAGE_CLIPS is a hint to kernel which could be helpful for some

drivers

Quoted 1 line

+ * to optimize internally especially for virtual devices where each

framebuffer

Quoted 3 lines

+ * change needs to be transmitted over network, usb, etc.
+ *
+ * Since FB_DAMAGE_CLIPS is a hint so it is an optional property. User-

space can

Quoted 1 line

+ * ignore the damage clips property and in that case kernel will do a full

plane

Quoted 1 line

+ * update. In case damage clips are provided then it is guaranteed that the

area

Quoted 1 line

+ * inside damage clips will be updated to plane. For efficiency kernel can

do

Quoted 3 lines

+ * full update or more area than specified in damage clips.
+ *
+ * FB_DAMAGE_CLIPS is a blob property with the layout of blob data is

simply an

Quoted 1 line

+ * array of drm_mode_rect(internally drm_rect) with maximum array size

limited

Quoted 1 line

+ * by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates,

damage clips

Quoted 2 lines

+ * are not in 16.16 fixed point. Similar to plane src in framebuffer, damage
+ * clips cannot be negative. In damage clip, x1/y1 are inclusive and x2/y2

are

Quoted 3 lines

+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should be

inside

Quoted 2 lines

+ * plane src and if any clip falls outside plane src it will be ignored and
+ * user-space won't be notified of the same. As mentioned above,

sometimes

Quoted 1 line

+ * kernel will do full plane update due to change in properties which can

affect

Quoted 1 line

+ * full plane e.g. color management properties. Also during full modeset

damage

Quoted 2 lines

+ * is irrelevant so if provided by user-space it simply will be ignored.
+ * Whenever damage clips are ignored by kernel, user-space will not be

informed.

Quoted 1 line

+ * If a user-space provides damage clips which doesn't encompass the

actual

Quoted 4 lines

+ * damage to framebuffer (since last plane update) will result in incorrect
+ * rendering during plane update.
+ *
+ * While kernel does not error for overlapped damage clips, it is

discouraged.

drivers/gpu/drm/drm_damage_helper.c

Quoted 4 lines

+ */

drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+
+/**
+ * drm_plane_enable_fb_damage_clips - enables plane fb damage clips

property

Review

Quoted 15 lines

+ * @plane: plane on which to enable damage clips property
+ *
+ * This function lets driver to enable the damage clips property on a plane.
+ */

drivers/gpu/drm/drm_damage_helper.c new file +6 -0
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
+{
+	struct drm_device *dev = plane->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	drm_object_attach_property(&plane->base, config-

prop_fb_damage_clips,

drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+				   0);
+}
+EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_mode_config.c

drivers/gpu/drm/drm_mode_config.c static int

Quoted 4 lines

index 21e353bd3948..506f5a52733f 100644

drivers/gpu/drm/drm_mode_config.c +0 -0
@@ -298,6 +298,12 @@ static int

drm_mode_create_standard_properties(struct drm_device *dev)

drivers/gpu/drm/drm_mode_config.c static int drm_mode_create_standard_properties(struct drm_device *dev)

Quoted 4 lines
drivers/gpu/drm/drm_mode_config.c +1 -0
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB,

"FB_DAMAGE_CLIPS",

drivers/gpu/drm/drm_mode_config.c static int drm_mode_create_standard_properties(struct drm_device *dev)

Quoted 6 lines
drivers/gpu/drm/drm_mode_config.c +5 -0
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev,

DRM_MODE_PROP_ATOMIC,

Review

Quoted 3 lines
drivers/gpu/drm/drm_mode_config.c +0 -0
 			"ACTIVE");
 	if (!prop)
(quoted diff) +0 -0

b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_du_plane_reset(struct drm_plane

Quoted 4 lines

index 23beff5d8e3c..1edbae73d6d6 100644

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +0 -0
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane

*plane)

Review

Quoted 8 lines
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
(quoted diff) +0 -0

b/include/drm/drm_damage_helper.h

include/drm/drm_damage_helper.h line 1

Quoted 7 lines

new file mode 100644
index 000000000000..217694e9168c

include/drm/drm_damage_helper.h new file +1 -0
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */

+/*********************************************************
*****************

Quoted 7 lines
include/drm/drm_damage_helper.h new file +7 -0
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including

include/drm/drm_damage_helper.h line 18

Quoted 10 lines
include/drm/drm_damage_helper.h new file +10 -0
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR

include/drm/drm_damage_helper.h line 19

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY,

include/drm/drm_damage_helper.h line 20

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN

NO EVENT SHALL

include/drm/drm_damage_helper.h line 21

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE

FOR ANY CLAIM,

include/drm/drm_damage_helper.h line 22

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF

CONTRACT, TORT OR

include/drm/drm_damage_helper.h line 23

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

SOFTWARE OR THE

include/drm/drm_damage_helper.h line 29

Quoted 6 lines
include/drm/drm_damage_helper.h new file +6 -0
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+

**********************************************************
****************/

include/drm/drm_damage_helper.h

Quoted 12 lines
include/drm/drm_damage_helper.h new file +12 -0
+
+#ifndef DRM_DAMAGE_HELPER_H_
+#define DRM_DAMAGE_HELPER_H_
+
+/**
+ * drm_plane_get_damage_clips_count - returns damage clips count
+ * @state: Plane state
+ *
+ * Returns: Number of clips in plane fb_damage_clips blob property.
+ */
+static inline uint32_t
+drm_plane_get_damage_clips_count(const struct drm_plane_state

*state)

Review

Quoted 22 lines
include/drm/drm_damage_helper.h new file +21 -0
+{
+	return (state && state->fb_damage_clips) ?
+		state->fb_damage_clips->length/sizeof(struct drm_rect) : 0;
+}
+
+/**
+ * drm_plane_get_damage_clips - returns damage clips
+ * @state: Plane state
+ *
+ * Returns: Clips in plane fb_damage_clips blob property.
+ */
+static inline struct drm_rect *
+drm_plane_get_damage_clips(const struct drm_plane_state *state)
+{
+	return (struct drm_rect *)((state && state->fb_damage_clips) ?
+				   state->fb_damage_clips->data : NULL);
+}
+
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+
+#endif
(quoted diff) +0 -0

b/include/drm/drm_mode_config.h

include/drm/drm_mode_config.h struct drm_mode_config { · line 631

Quoted 9 lines

index a0b202e1d69a..8ccb5ddcd99d 100644

include/drm/drm_mode_config.h +2 -0
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark

damaged

include/drm/drm_mode_config.h struct drm_mode_config { · line 632

Quoted 1 line
include/drm/drm_mode_config.h +1 -0
+	 * regions on the plane in framebuffer coordinates of the

framebuffer

include/drm/drm_mode_config.h struct drm_mode_config {

Quoted 4 lines
include/drm/drm_mode_config.h +4 -0
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by

DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike

Quoted 11 lines
include/drm/drm_mode_config.h +3 -0
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +0 -0
index 8a152dc16ea5..bb8b0934119c 100644

include/drm/drm_plane.h struct drm_plane_state { · line 188

Quoted 7 lines
include/drm/drm_plane.h +4 -0
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that

changed

Quoted 6 lines

include/drm/drm_plane.h struct drm_plane_state { · line 191

Quoted 3 lines
include/drm/drm_plane.h +3 -0
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;

Don't we have a better place for this? Next to some other props etc. in the plane state?

This is in plane state, sorry am I missing something ?

Quoted 3 lines

What does "in framebuffer coordinates" mean? Integers, some fixed point representation?

I guess better wording would be "in framebuffer coordinates of the frambebuffer attached to plane". I would make it more clear for v2,

Quoted 52 lines

Review

Quoted 6 lines
include/drm/drm_plane.h +1 -0
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
(quoted diff) +0 -0

b/include/uapi/drm/drm_mode.h

Quoted 5 lines

index 8d67243952f4..862ea0893e2e 100644

include/uapi/drm/drm_mode.h +0 -0
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease { · line 897

Quoted 10 lines
include/uapi/drm/drm_mode.h +8 -0
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area

this

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease {

Quoted 3 lines
include/uapi/drm/drm_mode.h +3 -0
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property

FB_DAMAGE_CLIPS.

Quoted 14 lines
include/uapi/drm/drm_mode.h +8 -1
+ */
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif
--

2.17.1

Quoted 4 lines

_______________________________________________
dri-devel mailing list
[email protected]

https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.fr
eedesktop.org%2Fmailman%2Flistinfo%2Fdri- devel&amp;data=02%7C01%7Cdrawat%40vmware.com%7C5c06b27d7534412 8cb0f08d613f0534a%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C6 36718320248896001&amp;sdata=2a2pS5CILMUyXDJgjIms59UUOso3LJRrpFW Rypik40Y%3D&amp;reserved=0

Quoted 3 lines

--
Ville Syrjälä
Intel

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#4 lore raw

Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Deepak Singh Rawat
Quoted 18 lines · On Thu, 6 Sep 2018 21:36:51 +0000 Deepak Singh Rawat <[email protected]> wrote:
Quoted 4 lines

- Why no input validation on the damage clips against the framebuffer size? Ime not validating just leads to funny driver bugs down the road, so what's the use-case you have in mind here?

My motivation behind not informing user-space if damage is outside plane src, sent during modeset(where it should be provided) or damage is outside framebuffer coordinate (simply put damage clips are invalid) is that it's a hint. The contract between kernel and user-space is simple, if damage clips are valid, kernel might use them (but not always) otherwise they will be ignored. Damage can be ignored deep in the stack (like network), so all the input validation for nothing.

Also, what all to consider in input validation ?
* clips are outside plane src
* damage clips sent during modeset should error ?
* any other properties.

Quoted 6 lines

This will lead to a complex input validation during modeset check like if some drivers are OK with damage while scaling whereas others cannot support, should this error?

However I am alright doing input validation if this becomes clear what kind of validation to actually do. My understanding of drm core is limited.

Hi,

I think validating that the area of any clip rectangle does not extend beyond the framebuffer size would be good. A userspace violating that is arguably broken, so it would help catch userspace bugs.

I also would not assume that damage is irrelevant on modeset. Userspace does not always know if an update will be a modeset or not: drivers vary there, and userspace that just wants the update through will allow modesets always while still providing damage. Userspace could also change video mode timings or pixel format while keeping the resolution the same, and in that case damage could be meaningful to a driver.

Thanks,
pq

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#9 lore raw

Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Deepak Singh Rawat
Quoted 230 lines · Deepak Singh Rawat
Quoted 221 lines · Deepak Rawat
Quoted 8 lines

From: Lukasz Spintzyk <[email protected]>

FB_DAMAGE_CLIPS is an optional plane property to mark damaged regions on the plane in framebuffer coordinates of the framebuffer attached to the plane.

The layout of blob data is simply an array of "struct drm_mode_rect" with maximum array size limited by DRM_MODE_FB_DIRTY_MAX_CLIPS.

Unlike

Quoted 14 lines

plane src coordinates, damage clips are not in 16.16 fixed point. As plane src in framebuffer cannot be negative so are damage clips. In damage clip, x1/y1 are inclusive and x2/y2 are exclusive.

This patch also exports the kernel internal drm_rect to userspace as drm_mode_rect. This is because "struct drm_clip_rect" is not sufficient to represent damage for current plane size.

Upper limit is set on the maximum number of damage clips to avoid overflow by user-space.

Driver which are interested in enabling FB_DAMAGE_CLIPS property for a plane should enable this property using drm_plane_enable_damage_clips.

Quoted 8 lines
---
 Documentation/gpu/drm-kms.rst       |  9 +++
 drivers/gpu/drm/Makefile            |  2 +-
 drivers/gpu/drm/drm_atomic.c        | 13 ++++
 drivers/gpu/drm/drm_atomic_helper.c |  4 ++
 drivers/gpu/drm/drm_damage_helper.c | 92

+++++++++++++++++++++++++++++

Documentation/gpu/drm-

Quoted 11 lines
 drivers/gpu/drm/drm_mode_config.c   |  6 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |  1 +
 include/drm/drm_damage_helper.h     | 63 ++++++++++++++++++++
 include/drm/drm_mode_config.h       | 10 ++++
 include/drm/drm_plane.h             |  8 +++
 include/uapi/drm/drm_mode.h         | 19 ++++++
 11 files changed, 226 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/drm_damage_helper.c
 create mode 100644 include/drm/drm_damage_helper.h
Documentation/gpu/drm- renamed from Documentation/gpu/drm-kms.rst +0 -0

kms.rst

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o

Quoted 24 lines

index 5dee6b8a4c12..78b66e628857 100644

Documentation/gpu/drm-kms.rst +9 -0
@@ -542,6 +542,15 @@ Plane Composition Properties
 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
    :export:
 
+FB_DAMAGE_CLIPS
+~~~~~~~~~~~~~~~
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :doc: overview
+
+.. kernel-doc:: drivers/gpu/drm/drm_damage_helper.c
+   :export:
+
 Color Management Properties
 ---------------------------
 
drivers/gpu/drm/Makefile +0 -0
index a6771cef85e2..ca5be0decb3f 100644
@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.o

drm_dp_helper.o drm_probe_helper.o \

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 35

Quoted 1 line
drivers/gpu/drm/Makefile +0 -0
 		drm_plane_helper.o drm_dp_mst_topology.o

drm_atomic_helper.o \

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 36

Quoted 1 line
drivers/gpu/drm/Makefile +0 -0
 		drm_kms_helper_common.o drm_dp_dual_mode_helper.o

\

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \

Quoted 3 lines
drivers/gpu/drm/Makefile +1 -1
 		drm_simple_kms_helper.o drm_modeset_helper.o \
-		drm_scdc_helper.o drm_gem_framebuffer_helper.o
+		drm_scdc_helper.o drm_gem_framebuffer_helper.o

drm_damage_helper.o

drivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o · line 40

Quoted 3 lines
drivers/gpu/drm/Makefile +0 -0
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) +=

drm_fb_helper.o

Review

Quoted 1 line
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic.c

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct

Quoted 4 lines

index 3eb061e11e2e..652e78ca1f81 100644

drivers/gpu/drm/drm_atomic.c +0 -0
@@ -857,6 +857,8 @@ static int drm_atomic_plane_set_property(struct

drm_plane *plane,

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct drm_plane *plane,

Quoted 8 lines
drivers/gpu/drm/drm_atomic.c +2 -0
 {
 	struct drm_device *dev = plane->dev;
 	struct drm_mode_config *config = &dev->mode_config;
+	bool replaced = false;
+	int ret;
 
 	if (property == config->prop_fb_id) {
 		struct drm_framebuffer *fb =

drm_framebuffer_lookup(dev, NULL, val);

drivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct

Quoted 1 line
drivers/gpu/drm/drm_atomic.c +0 -0
@@ -908,6 +910,14 @@ static int drm_atomic_plane_set_property(struct

drm_plane *plane,

drivers/gpu/drm/drm_atomic.c drm_atomic_plane_get_property(struct drm_plane

Quoted 15 lines
drivers/gpu/drm/drm_atomic.c +8 -0
 		state->color_encoding = val;
 	} else if (property == plane->color_range_property) {
 		state->color_range = val;
+	} else if (property == config->prop_fb_damage_clips) {
+		ret = drm_atomic_replace_property_blob_from_id(dev,
+					&state->fb_damage_clips,
+					val,
+					-1,
+					sizeof(struct drm_rect),
+					&replaced);
+		return ret;
 	} else if (plane->funcs->atomic_set_property) {
 		return plane->funcs->atomic_set_property(plane, state,
 				property, val);
@@ -976,6 +986,9 @@ drm_atomic_plane_get_property(struct drm_plane

*plane,

drivers/gpu/drm/drm_atomic.c drm_atomic_plane_get_property(struct drm_plane *plane,

Quoted 8 lines
drivers/gpu/drm/drm_atomic.c +3 -0
 		*val = state->color_encoding;
 	} else if (property == plane->color_range_property) {
 		*val = state->color_range;
+	} else if (property == config->prop_fb_damage_clips) {
+		*val = (state->fb_damage_clips) ?
+			state->fb_damage_clips->base.id : 0;
 	} else if (plane->funcs->atomic_get_property) {
 		return plane->funcs->atomic_get_property(plane, state,

property, val);

Review

Quoted 2 lines
drivers/gpu/drm/drm_atomic.c +0 -0
 	} else {
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_atomic_helper.c

drivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct

Quoted 4 lines

index 80be74df7ba6..be83e2763c18 100644

drivers/gpu/drm/drm_atomic_helper.c +0 -0
@@ -3576,6 +3576,7 @@ void drm_atomic_helper_plane_reset(struct

drm_plane *plane)

drivers/gpu/drm/drm_atomic_helper.c void

Quoted 9 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
 		/* Reset the alpha value to fully opaque if it matters */
 		if (plane->alpha_property)
 			plane->state->alpha = plane->alpha_property-
 values[1];
+		plane->state->fb_damage_clips = NULL;
 	}
 }

EXPORT_SYMBOL(drm_atomic_helper_plane_reset);

drivers/gpu/drm/drm_atomic_helper.c +0 -0
@@ -3598,6 +3599,7 @@ void

__drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane,

drivers/gpu/drm/drm_atomic_helper.c void

Quoted 8 lines
drivers/gpu/drm/drm_atomic_helper.c +1 -0
 	state->fence = NULL;
 	state->commit = NULL;
+	state->fb_damage_clips = NULL;
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_duplicate_state);
 
@@ -3642,6 +3644,8 @@ void

__drm_atomic_helper_plane_destroy_state(struct drm_plane_state *state)

Review

Quoted 9 lines
drivers/gpu/drm/drm_atomic_helper.c +2 -0
 	if (state->commit)
 		drm_crtc_commit_put(state->commit);
+
+	drm_property_blob_put(state->fb_damage_clips);
 }
 EXPORT_SYMBOL(__drm_atomic_helper_plane_destroy_state);
 
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_damage_helper.c

drivers/gpu/drm/drm_damage_helper.c line 1

Quoted 7 lines

new file mode 100644
index 000000000000..3e7de5afe7f6

drivers/gpu/drm/drm_damage_helper.c new file +1 -0
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT

+/*********************************************************
*****************

Quoted 7 lines
drivers/gpu/drm/drm_damage_helper.c new file +7 -0
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including

drivers/gpu/drm/drm_damage_helper.c

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c new file +10 -0
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR

drivers/gpu/drm/drm_damage_helper.c line 19

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY,

drivers/gpu/drm/drm_damage_helper.c line 20

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN

NO EVENT SHALL

drivers/gpu/drm/drm_damage_helper.c line 21

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE

FOR ANY CLAIM,

drivers/gpu/drm/drm_damage_helper.c line 22

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF

CONTRACT, TORT OR

drivers/gpu/drm/drm_damage_helper.c line 23

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

SOFTWARE OR THE

drivers/gpu/drm/drm_damage_helper.c line 31

Quoted 6 lines
drivers/gpu/drm/drm_damage_helper.c new file +6 -0
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+

**********************************************************
****************/

drivers/gpu/drm/drm_damage_helper.c

Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c new file +8 -0
+
+#include <drm/drmP.h>
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a

means to

drivers/gpu/drm/drm_damage_helper.c line 38

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * specify a list of damage rectangles on a plane in framebuffer

coordinates of

drivers/gpu/drm/drm_damage_helper.c line 39

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * the framebuffer attached to the plane. In current context damage is

the area

Quoted 8 lines

drivers/gpu/drm/drm_damage_helper.c line 40

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * of plane framebuffer (excluding the framebuffer area which is outside

of

drivers/gpu/drm/drm_damage_helper.c line 41

Quoted 1 line
drivers/gpu/drm/drm_damage_helper.c new file +1 -0
+ * plane src)

Not sure why the plane src coordinates need to be mentioned here. The damage is just the part of the fb that has changed. Whether or not it extends past the src coordinates is totally irrelevant as far as I can see.

Quoted 4 lines

Thanks Ville for the review,

Well this is plane damage property and only those clips which falls inside will be used for plane update, so outside plane src is not required.

Actually we've never specified that properly. My usual interpretation is that the plane is allowed to sample past the edge of the src rectangle (to get nicer looking edges). But not sure whether that would be too surprising to people though. So we might want follow the GL linear filter rules by default, and if someone wants better looking edges we could add a property that allows the filter to reach further past the edge.

Quoted 228 lines
Quoted 228 lines
Quoted 4 lines

that has changed since last plane update (also called page-flip).
+ * Only the area inside damage rectangles will be considered different

whether

Quoted 1 line

+ * currently attached framebuffer is same as framebuffer attached during

last

Quoted 3 lines

+ * plane update or not.
+ *
+ * FB_DAMAGE_CLIPS is a hint to kernel which could be helpful for some

drivers

Quoted 1 line

+ * to optimize internally especially for virtual devices where each

framebuffer

Quoted 3 lines

+ * change needs to be transmitted over network, usb, etc.
+ *
+ * Since FB_DAMAGE_CLIPS is a hint so it is an optional property. User-

space can

Quoted 1 line

+ * ignore the damage clips property and in that case kernel will do a full

plane

Quoted 1 line

+ * update. In case damage clips are provided then it is guaranteed that the

area

Quoted 1 line

+ * inside damage clips will be updated to plane. For efficiency kernel can

do

Quoted 3 lines

+ * full update or more area than specified in damage clips.
+ *
+ * FB_DAMAGE_CLIPS is a blob property with the layout of blob data is

simply an

Quoted 1 line

+ * array of drm_mode_rect(internally drm_rect) with maximum array size

limited

Quoted 1 line

+ * by DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike plane src coordinates,

damage clips

Quoted 2 lines

+ * are not in 16.16 fixed point. Similar to plane src in framebuffer, damage
+ * clips cannot be negative. In damage clip, x1/y1 are inclusive and x2/y2

are

Quoted 3 lines

+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should be

inside

Quoted 2 lines

+ * plane src and if any clip falls outside plane src it will be ignored and
+ * user-space won't be notified of the same. As mentioned above,

sometimes

Quoted 1 line

+ * kernel will do full plane update due to change in properties which can

affect

Quoted 1 line

+ * full plane e.g. color management properties. Also during full modeset

damage

Quoted 2 lines

+ * is irrelevant so if provided by user-space it simply will be ignored.
+ * Whenever damage clips are ignored by kernel, user-space will not be

informed.

Quoted 1 line

+ * If a user-space provides damage clips which doesn't encompass the

actual

Quoted 4 lines

+ * damage to framebuffer (since last plane update) will result in incorrect
+ * rendering during plane update.
+ *
+ * While kernel does not error for overlapped damage clips, it is

discouraged.

drivers/gpu/drm/drm_damage_helper.c

Quoted 4 lines

+ */

drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+
+/**
+ * drm_plane_enable_fb_damage_clips - enables plane fb damage clips

property

Review

Quoted 15 lines

+ * @plane: plane on which to enable damage clips property
+ *
+ * This function lets driver to enable the damage clips property on a plane.
+ */

drivers/gpu/drm/drm_damage_helper.c new file +6 -0
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)
+{
+	struct drm_device *dev = plane->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	drm_object_attach_property(&plane->base, config-

prop_fb_damage_clips,

drivers/gpu/drm/drm_damage_helper.c new file +3 -0
+				   0);
+}
+EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
(quoted diff) +0 -0

b/drivers/gpu/drm/drm_mode_config.c

drivers/gpu/drm/drm_mode_config.c static int

Quoted 4 lines

index 21e353bd3948..506f5a52733f 100644

drivers/gpu/drm/drm_mode_config.c +0 -0
@@ -298,6 +298,12 @@ static int

drm_mode_create_standard_properties(struct drm_device *dev)

drivers/gpu/drm/drm_mode_config.c static int drm_mode_create_standard_properties(struct drm_device *dev)

Quoted 4 lines
drivers/gpu/drm/drm_mode_config.c +1 -0
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create(dev, DRM_MODE_PROP_BLOB,

"FB_DAMAGE_CLIPS",

drivers/gpu/drm/drm_mode_config.c static int drm_mode_create_standard_properties(struct drm_device *dev)

Quoted 6 lines
drivers/gpu/drm/drm_mode_config.c +5 -0
+				   0);
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_fb_damage_clips = prop;
+
 	prop = drm_property_create_bool(dev,

DRM_MODE_PROP_ATOMIC,

Review

Quoted 3 lines
drivers/gpu/drm/drm_mode_config.c +0 -0
 			"ACTIVE");
 	if (!prop)
(quoted diff) +0 -0

b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_du_plane_reset(struct drm_plane

Quoted 4 lines

index 23beff5d8e3c..1edbae73d6d6 100644

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +0 -0
@@ -723,6 +723,7 @@ void vmw_du_plane_reset(struct drm_plane

*plane)

Review

Quoted 8 lines
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
 	plane->state = &vps->base;
 	plane->state->plane = plane;
 	plane->state->rotation = DRM_MODE_ROTATE_0;
+	plane->state->fb_damage_clips = NULL;
 }
 
 
(quoted diff) +0 -0

b/include/drm/drm_damage_helper.h

include/drm/drm_damage_helper.h line 1

Quoted 7 lines

new file mode 100644
index 000000000000..217694e9168c

include/drm/drm_damage_helper.h new file +1 -0
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */

+/*********************************************************
*****************

Quoted 7 lines
include/drm/drm_damage_helper.h new file +7 -0
+ *
+ * Copyright (c) 2018 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including

include/drm/drm_damage_helper.h line 18

Quoted 10 lines
include/drm/drm_damage_helper.h new file +10 -0
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR

include/drm/drm_damage_helper.h line 19

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY,

include/drm/drm_damage_helper.h line 20

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN

NO EVENT SHALL

include/drm/drm_damage_helper.h line 21

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE

FOR ANY CLAIM,

include/drm/drm_damage_helper.h line 22

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF

CONTRACT, TORT OR

include/drm/drm_damage_helper.h line 23

Quoted 1 line
include/drm/drm_damage_helper.h new file +1 -0
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

SOFTWARE OR THE

include/drm/drm_damage_helper.h line 29

Quoted 6 lines
include/drm/drm_damage_helper.h new file +6 -0
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ * Deepak Rawat <[email protected]>
+ *
+

**********************************************************
****************/

include/drm/drm_damage_helper.h

Quoted 12 lines
include/drm/drm_damage_helper.h new file +12 -0
+
+#ifndef DRM_DAMAGE_HELPER_H_
+#define DRM_DAMAGE_HELPER_H_
+
+/**
+ * drm_plane_get_damage_clips_count - returns damage clips count
+ * @state: Plane state
+ *
+ * Returns: Number of clips in plane fb_damage_clips blob property.
+ */
+static inline uint32_t
+drm_plane_get_damage_clips_count(const struct drm_plane_state

*state)

Review

Quoted 22 lines
include/drm/drm_damage_helper.h new file +21 -0
+{
+	return (state && state->fb_damage_clips) ?
+		state->fb_damage_clips->length/sizeof(struct drm_rect) : 0;
+}
+
+/**
+ * drm_plane_get_damage_clips - returns damage clips
+ * @state: Plane state
+ *
+ * Returns: Clips in plane fb_damage_clips blob property.
+ */
+static inline struct drm_rect *
+drm_plane_get_damage_clips(const struct drm_plane_state *state)
+{
+	return (struct drm_rect *)((state && state->fb_damage_clips) ?
+				   state->fb_damage_clips->data : NULL);
+}
+
+void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+
+#endif
(quoted diff) +0 -0

b/include/drm/drm_mode_config.h

include/drm/drm_mode_config.h struct drm_mode_config { · line 631

Quoted 9 lines

index a0b202e1d69a..8ccb5ddcd99d 100644

include/drm/drm_mode_config.h +2 -0
@@ -627,6 +627,16 @@ struct drm_mode_config {
 	 * &drm_crtc.
 	 */
 	struct drm_property *prop_crtc_id;
+	/**
+	 * @prop_fb_damage_clips: Optional plane property to mark

damaged

include/drm/drm_mode_config.h struct drm_mode_config { · line 632

Quoted 1 line
include/drm/drm_mode_config.h +1 -0
+	 * regions on the plane in framebuffer coordinates of the

framebuffer

include/drm/drm_mode_config.h struct drm_mode_config {

Quoted 4 lines
include/drm/drm_mode_config.h +4 -0
+	 * attached to the plane.
+	 *
+	 * The layout of blob data is simply an array of drm_mode_rect with
+	 * maximum array size limited by

DRM_MODE_FB_DIRTY_MAX_CLIPS. Unlike

Quoted 11 lines
include/drm/drm_mode_config.h +3 -0
+	 * plane src coordinates, damage clips are not in 16.16 fixed point.
+	 */
+	struct drm_property *prop_fb_damage_clips;
 	/**
 	 * @prop_active: Default atomic CRTC property to control the active
 	 * state, which is the simplified implementation for DPMS in atomic
include/drm/drm_plane.h +0 -0
index 8a152dc16ea5..bb8b0934119c 100644

include/drm/drm_plane.h struct drm_plane_state { · line 188

Quoted 7 lines
include/drm/drm_plane.h +4 -0
@@ -182,6 +182,14 @@ struct drm_plane_state {
 	 */
 	struct drm_crtc_commit *commit;
 
+	/**
+	 * @fb_damage_clips:
+	 *
+	 * Blob representing damage (area in plane framebuffer that

changed

Quoted 8 lines
Quoted 6 lines

include/drm/drm_plane.h struct drm_plane_state { · line 191

Quoted 3 lines
include/drm/drm_plane.h +3 -0
+	 * since last page flip) as array of drm_rect in framebuffer coodinates.
+	 */
+	struct drm_property_blob *fb_damage_clips;

Don't we have a better place for this? Next to some other props etc. in the plane state?

This is in plane state, sorry am I missing something ?

Just the placement inside the plane state looks rather random. More logical to group it next to some other property stuff I think.

Quoted 7 lines
Quoted 4 lines

What does "in framebuffer coordinates" mean? Integers, some fixed point representation?

I guess better wording would be "in framebuffer coordinates of the frambebuffer attached to plane". I would make it more clear for v2,

I'm basically asking whether one is supposed to do '.x1 = x' or '.x1 = x<<16' or something else, where x is an interger pixel coordinate.

Quoted 56 lines
Quoted 56 lines

Review

Quoted 7 lines
include/drm/drm_plane.h +1 -0
+
 	/** @state: backpointer to global drm_atomic_state */
 	struct drm_atomic_state *state;
 };
(quoted diff) +0 -0

b/include/uapi/drm/drm_mode.h

Quoted 5 lines

index 8d67243952f4..862ea0893e2e 100644

include/uapi/drm/drm_mode.h +0 -0
@@ -887,6 +887,25 @@ struct drm_mode_revoke_lease {
 	__u32 lessee_id;

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease { · line 897

Quoted 10 lines
include/uapi/drm/drm_mode.h +8 -0
 };
 
+/**
+ * struct drm_mode_rect - two dimensional rectangle
+ * @x1: horizontal starting coordinate (inclusive)
+ * @y1: vertical starting coordinate (inclusive)
+ * @x2: horizontal ending coordinate (exclusive)
+ * @y2: vertical ending coordinate (exclusive)
+ *
+ * With drm subsystem using struct drm_rect to manage rectangular area

this

include/uapi/drm/drm_mode.h struct drm_mode_revoke_lease {

Quoted 3 lines
include/uapi/drm/drm_mode.h +3 -0
+ * export it to user-space.
+ *
+ * Currently used by drm_mode_atomic blob property

FB_DAMAGE_CLIPS.

Quoted 14 lines
include/uapi/drm/drm_mode.h +8 -1
+ */
+struct drm_mode_rect {
+	__s32 x1;
+	__s32 y1;
+	__s32 x2;
+	__s32 y2;
+};
+
 #if defined(__cplusplus)
 }
 #endif
--

2.17.1

Quoted 4 lines

_______________________________________________
dri-devel mailing list
[email protected]

https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.fr
eedesktop.org%2Fmailman%2Flistinfo%2Fdri- devel&amp;data=02%7C01%7Cdrawat%40vmware.com%7C5c06b27d7534412 8cb0f08d613f0534a%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C6 36718320248896001&amp;sdata=2a2pS5CILMUyXDJgjIms59UUOso3LJRrpFW Rypik40Y%3D&amp;reserved=0

--
Ville Syrjälä
Intel

Signature · 6 lines
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#14 lore raw

Re: [PATCH 02/14] drm: add helper iterator functions for plane fb_damage_clips blob

In reply to Deepak Singh Rawat
Quoted 82 lines · Deepak Singh Rawat
Quoted 48 lines

Review line 76

Quoted 6 lines

#include <drm/drm_damage_helper.h>

/**

(quoted diff) +0 -0
@@ -75,6 +76,11 @@
  * While kernel does not error for overlapped damage clips, it is

discouraged.

drivers/gpu/drm/drm_damage_helper.c

Quoted 9 lines
drivers/gpu/drm/drm_damage_helper.c +5 -0
  */
 
+static int convert_fixed_to_32(int fixed)
+{
+	return ((fixed >> 15) & 1) + (fixed >> 16);
+}
+
 /**
  * drm_plane_enable_fb_damage_clips - enables plane fb damage clips

property

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct

Quoted 2 lines
drivers/gpu/drm/drm_damage_helper.c +0 -0
  * @plane: plane on which to enable damage clips property
@@ -90,3 +96,90 @@ void drm_plane_enable_fb_damage_clips(struct

drm_plane *plane)

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 11 lines
drivers/gpu/drm/drm_damage_helper.c +8 -0
 				   0);
 }
 EXPORT_SYMBOL(drm_plane_enable_fb_damage_clips);
+
+/**
+ * drm_atomic_helper_damage_iter_init - initialize the damage iterator
+ * @iter: The iterator to initialize.
+ * @old_state: old plane state for validation.
+ * @new_state: plane state from which to iterate the damage clips.
+ *
+ * Initialize an iterator that clip framebuffer damage in plane

fb_damage_clips

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 4 lines
drivers/gpu/drm/drm_damage_helper.c +4 -0
+ * blob to plane src clip. The iterator returns full plane src in case needing
+ * full update e.g. during full modeset.
+ *
+ * With this helper iterator, drivers which enabled fb_damage_clips

property can

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 4 lines
drivers/gpu/drm/drm_damage_helper.c +4 -0
+ * iterate over the damage clips that falls inside plane src during plane
+ * update.
+ *
+ * Returns: 0 on success and negative error code on error. If an error code

is

Review void drm_plane_enable_fb_damage_clips(struct · line 115

Quoted 1 line
(quoted diff) +1 -0
+ * returned then it means the plane state shouldn't update with attached

fb.

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 3 lines
drivers/gpu/drm/drm_damage_helper.c +3 -0
+ */
+int
+drm_atomic_helper_damage_iter_init(struct

drm_atomic_helper_damage_iter *iter,

Quoted 11 lines

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane) · line 126

Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c +8 -0
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state)
+{
+	if (!new_state || !new_state->crtc || !new_state->fb)
+		return -EINVAL;
+
+	if (!new_state->visible)
+		return -EINVAL;

Can't we handle this by iterating 0 damage rects instead? Would make the code a bit cleaner I think.

Agreed.

Quoted 11 lines
Quoted 11 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+
+	memset(iter, 0, sizeof(*iter));
+	iter->clips = drm_plane_get_damage_clips(new_state);
+	iter->num_clips = drm_plane_get_damage_clips_count(new_state);
+
+	if (!iter->clips)
+		iter->full_update = true;
+
+	if (!drm_rect_equals(&new_state->src, &old_state->src))
+		iter->full_update = true;
Quoted 8 lines

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane) · line 141

Quoted 5 lines
drivers/gpu/drm/drm_damage_helper.c +5 -0
+
+	iter->plane_src.x1 = convert_fixed_to_32(new_state->src.x1);
+	iter->plane_src.y1 = convert_fixed_to_32(new_state->src.y1);
+	iter->plane_src.x2 = convert_fixed_to_32(new_state->src.x2);
+	iter->plane_src.y2 = convert_fixed_to_32(new_state->src.y2);

I think you want to clip with the clipped rectangles here, not with the ones userspace provides.

Quoted 2 lines

new_state->src.x1 is the clipped one, clipped in the function drm_atomic_helper_check_plane_state. Or am I missing something ?

You're right, I misremembered. This looks correct. Would be good to explain this in the kerneldoc, and that you must call drm_atomic_helper_check_plane_state() first before calling this.

Cheers, Daniel

Quoted 116 lines
Quoted 7 lines

Also I think you're rounding is wrong here - I think you need to round down for x/y1, and round up for x/y2 to make sure you catch all the pixels?

Unit tests for this, in the form of a drm selftest (like we have for drm_mm.c already, but for kms helpers) would be perfect. Much easier to review a testcase than do bitmath in my head :-)

Sure I will work on a unit test case for this scenario and work on round up of plane_src.

Quoted 105 lines
Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c +7 -0
+
+	if (iter->full_update) {
+		iter->clips = 0;
+		iter->curr_clip = 0;
+		iter->num_clips = 0;
+	}
+

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 10 lines
drivers/gpu/drm/drm_damage_helper.c +10 -0
+	return 0;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_init);
+
+/**
+ * drm_atomic_helper_damage_iter_next - advance the damage iterator
+ * @iter: The iterator to advance.
+ * @rect: Return a rectangle in fb coordinate clipped to plane src.
+ *
+ * Returns:  true if the output is valid, false if we've reached the end of

the

drivers/gpu/drm/drm_damage_helper.c void drm_plane_enable_fb_damage_clips(struct drm_plane *plane)

Quoted 4 lines
drivers/gpu/drm/drm_damage_helper.c +4 -0
+ * rectangle list.
+ */
+bool
+drm_atomic_helper_damage_iter_next(struct

drm_atomic_helper_damage_iter *iter,

Review

Quoted 24 lines
drivers/gpu/drm/drm_damage_helper.c +23 -0
+				   struct drm_rect *rect)
+{
+	bool ret = false;
+
+	if (iter->full_update) {
+		*rect = iter->plane_src;
+		iter->full_update = false;
+		return true;
+	}
+
+	while (iter->curr_clip < iter->num_clips) {
+		*rect = iter->clips[iter->curr_clip];
+		iter->curr_clip++;
+
+		if (drm_rect_intersect(rect, &iter->plane_src)) {
+			ret = true;
+			break;
+		}
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(drm_atomic_helper_damage_iter_next);
(quoted diff) +0 -0

b/include/drm/drm_damage_helper.h

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct

Quoted 24 lines

index 217694e9168c..f1282b459a4f 100644

include/drm/drm_damage_helper.h +13 -0
@@ -32,6 +32,19 @@
 #ifndef DRM_DAMAGE_HELPER_H_
 #define DRM_DAMAGE_HELPER_H_
 
+/**
+ * struct drm_atomic_helper_damage_iter - damage clip iterator
+ *
+ * This iterator tracks state needed to walk the list of damage clips.
+ */
+struct drm_atomic_helper_damage_iter {
+	const struct drm_rect *clips;
+	struct drm_rect plane_src;
+	uint32_t num_clips;
+	uint32_t curr_clip;
+	bool full_update;
+};
+
 /**
  * drm_plane_get_damage_clips_count - returns damage clips count
  * @state: Plane state
@@ -59,5 +72,12 @@ drm_plane_get_damage_clips(const struct

drm_plane_state *state)

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state)

Quoted 5 lines
include/drm/drm_damage_helper.h +2 -0
 }
 
 void drm_plane_enable_fb_damage_clips(struct drm_plane *plane);
+int
+drm_atomic_helper_damage_iter_init(struct

drm_atomic_helper_damage_iter *iter,

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct drm_plane_state *state) · line 78

Quoted 2 lines
include/drm/drm_damage_helper.h +2 -0
+				   const struct drm_plane_state *old_state,
+				   const struct drm_plane_state *new_state);

I think a for_each_damage macro would be sweet on top of these here. But easy to add later on.
-Daniel

include/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct · line 80

Quoted 2 lines
include/drm/drm_damage_helper.h +2 -0
+bool
+drm_atomic_helper_damage_iter_next(struct

drm_atomic_helper_damage_iter *iter,

Quoted 6 lines
include/drm/drm_damage_helper.h +1 -1
+				   struct drm_rect *rect);
 
 #endif
--

2.17.1

--
Daniel Vetter
Software Engineer, Intel Corporation
https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fblog.ff
wll.ch&amp;data=02%7C01%7Cdrawat%40vmware.com%7C117739774a7341f d8c0208d613cd83d5%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C 636718170756943488&amp;sdata=tNK2zTqJrq8U5U4W96wp49FZ8LAB9fl2WV UixbOSWsU%3D&amp;reserved=0

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#24 lore raw

Re: [PATCH 05/14] drm/vmwgfx: add a new interface for plane update on a display unit

In reply to Deepak Rawat
Quoted 26 lines · Deepak Rawat

Add a new struct vmw_du_update_plane similar to vmw_kms_dirty which represent the flow of operations needed to update a display unit from surface or bo(blit a new framebuffer).

---
  drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 108 +++++++++++++++++++++
  drivers/gpu/drm/vmwgfx/vmwgfx_kms.h | 143 ++++++++++++++++++++++++++++
  2 files changed, 251 insertions(+)
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +1 -0
index 1edbae73d6d6..c1de8f609bf4 100644
@@ -30,6 +30,7 @@
  #include <drm/drm_atomic.h>
  #include <drm/drm_atomic_helper.h>
  #include <drm/drm_rect.h>
+#include <drm/drm_damage_helper.h>
 
  /* Might need a hrtimer here? */
  #define VMWGFX_PRESENT_RATE ((HZ / 60 > 0) ? HZ / 60 : 1)
@@ -3077,3 +3078,110 @@ void vmw_kms_lost_device(struct drm_device *dev)
  {
  	drm_atomic_helper_shutdown(dev);
  }

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_kms_lost_device(struct drm_device *dev) · line 3090

Quoted 10 lines
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +10 -0
+
+/**
+ * vmw_du_helper_plane_update - helper to do plane update on display unit
+ *
+ * @update: The closure structure.
+ *
+ * RETURNS:
+ *
+ * 0 on success or a negative error code on failure.
+ */

We should try to comply better with the kerneldoc description
https://www.kernel.org/doc/Documentation/kernel-doc-nano-HOWTO.txt

Otherwise This patch LGTM.

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#29 lore raw

Re: [PATCH 08/14] drm/vmwgfx: use the new interface for STDU plane update

In reply to Deepak Rawat
Quoted 2 lines · Deepak Rawat

With new interface to do plane update on STDU available, use that instead of old kms_dirty. Update the commet to sync with code.

s/commet/comment/

Quoted 79 lines
---
  drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 38 +++++++---------------------
  1 file changed, 9 insertions(+), 29 deletions(-)
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +8 -26
index 8c1ba874512a..167190b75e2f 100644
@@ -1706,14 +1706,10 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
  	struct vmw_private *dev_priv;
  	int ret;
 
-	/*
-	 * We cannot really fail this function, so if we do, then output an
-	 * error and maintain consistent atomic state.
-	 */
+	/* If somehow gets a device error, maintain consistent atomic state */
  	if (crtc && plane->state->fb) {
  		struct vmw_framebuffer *vfb =
  			vmw_framebuffer_to_vfb(plane->state->fb);
-		struct drm_vmw_rect vclips;
  		stdu = vmw_crtc_to_stdu(crtc);
  		dev_priv = vmw_priv(crtc->dev);
 
@@ -1721,23 +1717,17 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
  		stdu->content_fb_type = vps->content_fb_type;
  		stdu->cpp = vps->cpp;
 
-		vclips.x = crtc->x;
-		vclips.y = crtc->y;
-		vclips.w = crtc->mode.hdisplay;
-		vclips.h = crtc->mode.vdisplay;
-
  		ret = vmw_stdu_bind_st(dev_priv, stdu, &stdu->display_srf->res);
  		if (ret)
  			DRM_ERROR("Failed to bind surface to STDU.\n");
 
  		if (vfb->bo)
-			ret = vmw_kms_stdu_dma(dev_priv, NULL, vfb, NULL, NULL,
-					       &vclips, 1, 1, true, false,
-					       crtc);
+			ret = vmw_stdu_plane_update_bo(dev_priv, plane,
+						       old_state, vfb);
  		else
-			ret = vmw_kms_stdu_surface_dirty(dev_priv, vfb, NULL,
-							 &vclips, NULL, 0, 0,
-							 1, 1, NULL, crtc);
+			ret = vmw_stdu_plane_update_surface(dev_priv, plane,
+							    old_state, vfb,
+							    NULL);
  		if (ret)
  			DRM_ERROR("Failed to update STDU.\n");
  	} else {
@@ -1745,12 +1735,7 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
  		stdu = vmw_crtc_to_stdu(crtc);
  		dev_priv = vmw_priv(crtc->dev);
 
-		/*
-		 * When disabling a plane, CRTC and FB should always be NULL
-		 * together, otherwise it's an error.
-		 * Here primary plane is being disable so blank the screen
-		 * target display unit, if not already done.
-		 */
+		/* Blank STDU when fb and crtc are NULL */
  		if (!stdu->defined)
  			return;
 
@@ -1765,20 +1750,15 @@ vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane,
  		return;
  	}
 
+	/* In case of error vblank event is sent in vmw_du_crtc_atomic_flush */
  	event = crtc->state->event;
-	/*
-	 * In case of failure and other cases, vblank event will be sent in
-	 * vmw_du_crtc_atomic_flush.
-	 */

drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane, · line 1761

Quoted 10 lines
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +1 -3
  	if (event && (ret == 0)) {
  		struct vmw_fence_obj *fence = NULL;
  		struct drm_file *file_priv = event->base.file_priv;
 
  		vmw_execbuf_fence_commands(NULL, dev_priv, &fence, NULL);
 
-		/*
-		 * If fence is NULL, then already sync.
-		 */
+		/* If fence is NULL, then already sync. */

If comment changes are not directly related to the functional changes, please use a separate patch for them.

drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c vmw_stdu_primary_plane_atomic_update(struct drm_plane *plane, · line 1764

Quoted 3 lines
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +0 -0
  		if (fence) {
  			ret = vmw_event_fence_action_queue(
  				file_priv, fence, &event->base,

Thanks,
Thomas

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#25 lore raw

Re: [PATCH 05/14] drm/vmwgfx: add a new interface for plane update on a display unit

In reply to Thomas Hellstrom
Quoted 10 lines · Thomas Hellstrom
Quoted 9 lines · Deepak Rawat

+/**
+ * vmw_du_helper_plane_update - helper to do plane update on display unit
+ *
+ * @update: The closure structure.
+ *
+ * RETURNS:
+ *
+ * 0 on success or a negative error code on failure.
+ */

Quoted 2 lines

We should try to comply better with the kerneldoc description
https://www.kernel.org/doc/Documentation/kernel-doc-nano-HOWTO.txt

Up-to-date version at [1].

BR,
Jani.

[1] https://www.kernel.org/doc/html/latest/doc-guide/kernel-doc.html

Signature · 5 lines
Jani Nikula, Intel Open Source Graphics Center
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#31 lore raw

Re: [PATCH 09/14] drm/vmwgfx: enable FB_DAMAGE_CLIPS property for STDU primary plane

In reply to Deepak Rawat
Quoted 12 lines · Deepak Rawat

STDU primary plane now support damage clips, enable it for user-space.

---
  drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 1 +
  1 file changed, 1 insertion(+)
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +0 -0
index 167190b75e2f..11a76aa96302 100644

drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c static int vmw_stdu_init(struct vmw_private *dev_priv, unsigned unit) · line 1886

Quoted 4 lines
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +1 -0
@@ -1883,6 +1883,7 @@ static int vmw_stdu_init(struct vmw_private *dev_priv, unsigned unit)
  	}
 
  	drm_plane_helper_add(primary, &vmw_stdu_primary_plane_helper_funcs);
+	drm_plane_enable_fb_damage_clips(primary);

LGTM. We should, as a second pass, have a look at cursors as well, to avoid having to send the cursor image on each cursor position update..

drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c static int vmw_stdu_init(struct vmw_private *dev_priv, unsigned unit) · line 1889

Quoted 3 lines
drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +0 -0
  	/* Initialize cursor plane */
  	vmw_du_plane_reset(cursor);

/Thomas

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#5 lore raw

RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Pekka Paalanen
Quoted 36 lines
Quoted 18 lines · On Thu, 6 Sep 2018 21:36:51 +0000 Deepak Singh Rawat <[email protected]> wrote:
Quoted 4 lines

- Why no input validation on the damage clips against the framebuffer size? Ime not validating just leads to funny driver bugs down the road, so what's the use-case you have in mind here?

My motivation behind not informing user-space if damage is outside plane src, sent during modeset(where it should be provided) or damage is outside framebuffer coordinate (simply put damage clips are invalid) is that it's a hint. The contract between kernel and user-space is simple, if damage clips are valid, kernel might use them (but not always) otherwise they will be ignored. Damage can be ignored deep in the stack (like network), so all the input validation for nothing.

Also, what all to consider in input validation ?
* clips are outside plane src
* damage clips sent during modeset should error ?
* any other properties.

Quoted 6 lines

This will lead to a complex input validation during modeset check like if some drivers are OK with damage while scaling whereas others cannot support, should this error?

However I am alright doing input validation if this becomes clear what kind of validation to actually do. My understanding of drm core is limited.

Hi,

I think validating that the area of any clip rectangle does not extend beyond the framebuffer size would be good. A userspace violating that is arguably broken, so it would help catch userspace bugs.

Quoted 7 lines

I also would not assume that damage is irrelevant on modeset. Userspace does not always know if an update will be a modeset or not: drivers vary there, and userspace that just wants the update through will allow modesets always while still providing damage. Userspace could also change video mode timings or pixel format while keeping the resolution the same, and in that case damage could be meaningful to a driver.

Thanks Pekka, for explanation. I agree clips outside framebuffer should be errored and also it's not a lot for drm core. Also, will look into the cases where drm core should clear the damage clips, like resolution change, etc.

May be the better approach would be drm core just provide the helper and driver takes care of driver specific scenarios.
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#10 lore raw

RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update

In reply to Ville Syrjälä
Quoted 27 lines
Quoted 22 lines
Quoted 13 lines

drivers/gpu/drm/drm_damage_helper.c

Quoted 8 lines
drivers/gpu/drm/drm_damage_helper.c new file +7 -0
+#include <drm/drmP.h>
+#include <drm/drm_damage_helper.h>
+
+/**
+ * DOC: overview
+ *
+ * FB_DAMAGE_CLIPS is an optional plane property which provides a

means to

Quoted 1 line

+ * specify a list of damage rectangles on a plane in framebuffer

coordinates of

Quoted 1 line

+ * the framebuffer attached to the plane. In current context damage is

the area

Quoted 8 lines
Quoted 1 line

+ * of plane framebuffer (excluding the framebuffer area which is outside

of

Quoted 1 line

+ * plane src)

Not sure why the plane src coordinates need to be mentioned here. The damage is just the part of the fb that has changed. Whether or not it extends past the src coordinates is totally irrelevant as far as I can see.

Quoted 4 lines

Thanks Ville for the review,

Well this is plane damage property and only those clips which falls inside will be used for plane update, so outside plane src is not required.

Quoted 6 lines

Actually we've never specified that properly. My usual interpretation is that the plane is allowed to sample past the edge of the src rectangle (to get nicer looking edges). But not sure whether that would be too surprising to people though. So we might want follow the GL linear filter rules by default, and if someone wants better looking edges we could add a property that allows the filter to reach further past the edge.

Ok, it makes sense not to specify plane src part, anyway with current implementation kernel do not error if damage clips are outside plane src. The helper iterator clip to plane_src but depending on the above use case can have other implementations as well.

Thanks,
Deepak

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel