plane update with damage
Subsystems (4)
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
PATCH 1/14 drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
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
@@ -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
@@ -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
@@ -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
@@ -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
@@ -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
@@ -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
@@ -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
@@ -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
@@ -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
Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
Quoted 46 lines · Deepak Rawat
From: Lukasz Spintzyk <[email protected]>
FB_DAMAGE_CLIPSis 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 byDRM_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_rectto userspace asdrm_mode_rect. This is because "structdrm_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_CLIPSproperty for a plane should enable this property usingdrm_plane_enable_damage_clips.Signed-off-by Lukasz Spintzyk <[email protected]>Signed-off-by Deepak Rawat <[email protected]>--- 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.hDocumentation/gpu/drm-kms.rst +0
-0@@ -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@@ -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.odrivers/gpu/drm/drm_atomic.c +13
-0@@ -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
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@@ -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@@ -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@@ -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@@ -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@@ -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@@ -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@@ -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
RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
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 htmldocsand 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@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.odrm_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
-0drm_plane_helper.o drm_dp_mst_topology.odrm_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
-0drm_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
-1drm_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.odrm_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
-0drm_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.cdrivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct
Quoted 4 lines
index
3eb061e11e2e..652e78ca1f81100644drivers/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
-0state->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.cdrivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct
Quoted 4 lines
index
80be74df7ba6..be83e2763c18100644drivers/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
-0state->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
-0if (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.cdrivers/gpu/drm/drm_damage_helper.c line 1
Quoted 7 lines
new file mode 100644
index 000000000000..3e7de5afe7f6drivers/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.cdrivers/gpu/drm/drm_mode_config.c static int
Quoted 4 lines
index
21e353bd3948..506f5a52733f100644drivers/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
-0return -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.cdrivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_du_plane_reset(struct drm_plane
Quoted 4 lines
index
23beff5d8e3c..1edbae73d6d6100644drivers/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
-0plane->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.hinclude/drm/drm_damage_helper.h line 1
Quoted 7 lines
new file mode 100644
index 000000000000..217694e9168cinclude/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.hinclude/drm/drm_mode_config.h struct drm_mode_config { · line 631
Quoted 9 lines
index
a0b202e1d69a..8ccb5ddcd99d100644include/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. UnlikeQuoted 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
-0include/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.hQuoted 5 lines
index
8d67243952f4..862ea0893e2e100644include/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&data=02%7C01%7Cdrawat%40vmware.com%7C0f5c199805444f2 6d01508d613cc5abe%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C 636718165752004800&sdata=zKJnKe90%2BMeBsSpb4t4HfkOlLYkgXkyxsc GwVOHCyro%3D&reserved=0
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
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
RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
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
Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
In reply to Deepak RawatQuoted 318 lines · Deepak Rawat
From: Lukasz Spintzyk <[email protected]>
FB_DAMAGE_CLIPSis 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 byDRM_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_rectto userspace asdrm_mode_rect. This is because "structdrm_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_CLIPSproperty for a plane should enable this property usingdrm_plane_enable_damage_clips.Signed-off-by Lukasz Spintzyk <[email protected]>Signed-off-by Deepak Rawat <[email protected]>--- 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.hDocumentation/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 +1
-1@@ -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.odrivers/gpu/drm/drm_atomic.c +13
-0@@ -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@@ -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@@ -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@@ -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@@ -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@@ -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@@ -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@@ -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@@ -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
Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
In reply to Deepak RawatQuoted 178 lines · Deepak Rawat
From: Lukasz Spintzyk <[email protected]>
FB_DAMAGE_CLIPSis 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 byDRM_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_rectto userspace asdrm_mode_rect. This is because "structdrm_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_CLIPSproperty for a plane should enable this property usingdrm_plane_enable_damage_clips.Signed-off-by Lukasz Spintzyk <[email protected]>Signed-off-by Deepak Rawat <[email protected]>--- 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.hDocumentation/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 +1
-1@@ -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.odrivers/gpu/drm/drm_atomic.c +13
-0@@ -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@@ -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@@ -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_CLIPSis 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.
+ *
+ * SinceFB_DAMAGE_CLIPSis 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_CLIPSis a blob property with the layout of blob data is simply an
+ * array ofdrm_mode_rect(internally drm_rect)with maximum array size limited
+ * byDRM_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@@ -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@@ -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@@ -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@@ -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
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@@ -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
RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
Quoted 221 lines · Deepak Rawat
Quoted 8 lines
From: Lukasz Spintzyk <[email protected]>
FB_DAMAGE_CLIPSis 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 byDRM_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_rectto userspace asdrm_mode_rect. This is because "structdrm_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_CLIPSproperty for a plane should enable this property usingdrm_plane_enable_damage_clips.Quoted 8 lines
Signed-off-by Lukasz Spintzyk <[email protected]>Signed-off-by Deepak Rawat <[email protected]>--- 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.hDocumentation/gpu/drm- renamed from Documentation/gpu/drm-kms.rst +0
-0
kms.rstdrivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o
Quoted 24 lines
index
5dee6b8a4c12..78b66e628857100644Documentation/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@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.odrm_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
-0drm_plane_helper.o drm_dp_mst_topology.odrm_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
-0drm_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
-1drm_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.odrm_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
-0drm_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.cdrivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct
Quoted 4 lines
index
3eb061e11e2e..652e78ca1f81100644drivers/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
-0state->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.cdrivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct
Quoted 4 lines
index
80be74df7ba6..be83e2763c18100644drivers/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
-0state->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
-0if (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.cdrivers/gpu/drm/drm_damage_helper.c line 1
Quoted 7 lines
new file mode 100644
index 000000000000..3e7de5afe7f6drivers/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 differentwhether
Quoted 1 line
+ * currently attached framebuffer is same as framebuffer attached during
last
Quoted 3 lines
+ * plane update or not.
+ *
+ *FB_DAMAGE_CLIPSis a hint to kernel which could be helpful for somedrivers
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.
+ *
+ * SinceFB_DAMAGE_CLIPSis 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_CLIPSis a blob property with the layout of blob data issimply an
Quoted 1 line
+ * array of
drm_mode_rect(internally drm_rect)with maximum array sizelimited
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/y2are
Quoted 3 lines
+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should beinside
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 beinformed.
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 isdiscouraged.
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.cdrivers/gpu/drm/drm_mode_config.c static int
Quoted 4 lines
index
21e353bd3948..506f5a52733f100644drivers/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
-0return -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.cdrivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_du_plane_reset(struct drm_plane
Quoted 4 lines
index
23beff5d8e3c..1edbae73d6d6100644drivers/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
-0plane->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.hinclude/drm/drm_damage_helper.h line 1
Quoted 7 lines
new file mode 100644
index 000000000000..217694e9168cinclude/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.hinclude/drm/drm_mode_config.h struct drm_mode_config { · line 631
Quoted 9 lines
index
a0b202e1d69a..8ccb5ddcd99d100644include/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. UnlikeQuoted 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
-0include/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.hQuoted 5 lines
index
8d67243952f4..862ea0893e2e100644include/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&data=02%7C01%7Cdrawat%40vmware.com%7C5c06b27d7534412 8cb0f08d613f0534a%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C6 36718320248896001&sdata=2a2pS5CILMUyXDJgjIms59UUOso3LJRrpFW Rypik40Y%3D&reserved=0
Quoted 3 lines
--
Ville Syrjälä
Intel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
Quoted 230 lines · Deepak Singh Rawat
Quoted 221 lines · Deepak Rawat
Quoted 8 lines
From: Lukasz Spintzyk <[email protected]>
FB_DAMAGE_CLIPSis 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 byDRM_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_rectto userspace asdrm_mode_rect. This is because "structdrm_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_CLIPSproperty for a plane should enable this property usingdrm_plane_enable_damage_clips.Quoted 8 lines
Signed-off-by Lukasz Spintzyk <[email protected]>Signed-off-by Deepak Rawat <[email protected]>--- 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.hDocumentation/gpu/drm- renamed from Documentation/gpu/drm-kms.rst +0
-0
kms.rstdrivers/gpu/drm/Makefile drm_kms_helper-y := drm_crtc_helper.o
Quoted 24 lines
index
5dee6b8a4c12..78b66e628857100644Documentation/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@@ -35,7 +35,7 @@ drm_kms_helper-y := drm_crtc_helper.odrm_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
-0drm_plane_helper.o drm_dp_mst_topology.odrm_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
-0drm_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
-1drm_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.odrm_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
-0drm_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.cdrivers/gpu/drm/drm_atomic.c static int drm_atomic_plane_set_property(struct
Quoted 4 lines
index
3eb061e11e2e..652e78ca1f81100644drivers/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
-0state->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.cdrivers/gpu/drm/drm_atomic_helper.c void drm_atomic_helper_plane_reset(struct
Quoted 4 lines
index
80be74df7ba6..be83e2763c18100644drivers/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
-0state->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
-0if (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.cdrivers/gpu/drm/drm_damage_helper.c line 1
Quoted 7 lines
new file mode 100644
index 000000000000..3e7de5afe7f6drivers/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 differentwhether
Quoted 1 line
+ * currently attached framebuffer is same as framebuffer attached during
last
Quoted 3 lines
+ * plane update or not.
+ *
+ *FB_DAMAGE_CLIPSis a hint to kernel which could be helpful for somedrivers
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.
+ *
+ * SinceFB_DAMAGE_CLIPSis 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_CLIPSis a blob property with the layout of blob data issimply an
Quoted 1 line
+ * array of
drm_mode_rect(internally drm_rect)with maximum array sizelimited
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/y2are
Quoted 3 lines
+ * exclusive.
+ *
+ * All the damage clips are in framebuffer coordinates and should beinside
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 beinformed.
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 isdiscouraged.
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.cdrivers/gpu/drm/drm_mode_config.c static int
Quoted 4 lines
index
21e353bd3948..506f5a52733f100644drivers/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
-0return -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.cdrivers/gpu/drm/vmwgfx/vmwgfx_kms.c void vmw_du_plane_reset(struct drm_plane
Quoted 4 lines
index
23beff5d8e3c..1edbae73d6d6100644drivers/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
-0plane->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.hinclude/drm/drm_damage_helper.h line 1
Quoted 7 lines
new file mode 100644
index 000000000000..217694e9168cinclude/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.hinclude/drm/drm_mode_config.h struct drm_mode_config { · line 631
Quoted 9 lines
index
a0b202e1d69a..8ccb5ddcd99d100644include/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. UnlikeQuoted 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
-0include/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.hQuoted 5 lines
index
8d67243952f4..862ea0893e2e100644include/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&data=02%7C01%7Cdrawat%40vmware.com%7C5c06b27d7534412 8cb0f08d613f0534a%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C6 36718320248896001&sdata=2a2pS5CILMUyXDJgjIms59UUOso3LJRrpFW Rypik40Y%3D&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
RE: [PATCH 01/14] drm: add new plane property FB_DAMAGE_CLIPS to send damage during plane update
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
PATCH 2/14 drm: add helper iterator functions for plane fb_damage_clips blob
In reply to Deepak RawatWith 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
@@ -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
@@ -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
Re: [PATCH 02/14] drm: add helper iterator functions for plane fb_damage_clips blob
Quoted 57 lines · Deepak Rawat
With
fb_damage_clipsblob property indrm_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.Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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@@ -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
RE: [PATCH 02/14] drm: add helper iterator functions for plane fb_damage_clips blob
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 isdiscouraged.
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
-00); }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_clipsdrivers/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 attachedfb.
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.calready, 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.hinclude/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct
Quoted 24 lines
index
217694e9168c..f1282b459a4f100644include/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_damagemacro would be sweet on top of these here. But easy to add later on.
-Danielinclude/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&data=02%7C01%7Cdrawat%40vmware.com%7C117739774a7341f d8c0208d613cd83d5%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C 636718170756943488&sdata=tNK2zTqJrq8U5U4W96wp49FZ8LAB9fl2WV UixbOSWsU%3D&reserved=0
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 02/14] drm: add helper iterator functions for plane fb_damage_clips blob
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 isdiscouraged.
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
-00); }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_clipsdrivers/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 attachedfb.
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.x1is the clipped one, clipped in the functiondrm_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.calready, 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.hinclude/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct
Quoted 24 lines
index
217694e9168c..f1282b459a4f100644include/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_damagemacro would be sweet on top of these here. But easy to add later on.
-Danielinclude/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&data=02%7C01%7Cdrawat%40vmware.com%7C117739774a7341f d8c0208d613cd83d5%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C1%7C0%7C 636718170756943488&sdata=tNK2zTqJrq8U5U4W96wp49FZ8LAB9fl2WV UixbOSWsU%3D&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
PATCH 3/14 drm: clear plane damage during full modeset
In reply to Deepak RawatPlane 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
@@ -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
@@ -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
Re: [PATCH 03/14] drm: clear plane damage during full modeset
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.
Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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
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
RE: [PATCH 03/14] drm: clear plane damage during full modeset
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.cdrivers/gpu/drm/drm_atomic_helper.c drm_atomic_helper_plane_changed(struct
Quoted 12 lines
index
be83e2763c18..e06d2d5d582f100644drivers/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
-0return ;
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
-0crtc_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.hinclude/drm/drm_damage_helper.h drm_plane_get_damage_clips(const struct
Quoted 4 lines
index
f1282b459a4f..1f988f7fdd72100644include/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
-0state->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
Reviewed-by Daniel Vetter <[email protected]>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
Re: [PATCH 03/14] drm: clear plane damage during full modeset
In reply to Deepak RawatQuoted 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.
Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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@@ -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
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.
Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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@@ -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-develSignature · 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
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.
Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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@@ -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_damageis 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. -DanielQuoted 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-develSignature · 6 lines
Ville Syrjälä Intel _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-develSignature · 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
PATCH 4/14 drm: add helper to implement legacy dirtyfb
In reply to Deepak RawatFrom: 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
@@ -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
@@ -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
Re: [PATCH 04/14] drm: add helper to implement legacy dirtyfb
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_clipsproperty and removedplane_state::dirtyflag.Signed-off-by Rob Clark <[email protected]>Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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@@ -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
PATCH 5/14 drm/vmwgfx: add a new interface for plane update on a display unit
In reply to Deepak RawatAdd 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
@@ -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
Re: [PATCH 05/14] drm/vmwgfx: add a new interface for plane update on a display unit
Quoted 26 lines · Deepak Rawat
Add a new struct
vmw_du_update_planesimilar tovmw_kms_dirtywhich represent the flow of operations needed to update a display unit from surface or bo(blit a new framebuffer).Signed-off-by Deepak Rawat <[email protected]>--- 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@@ -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
Re: [PATCH 05/14] drm/vmwgfx: add a new interface for plane update on a display unit
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
PATCH 6/14 drm/vmwgfx: implement STDU plane update for surface backed fb
In reply to Deepak RawatUsing 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
@@ -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
PATCH 7/14 drm/vmwgfx: implement STDU plane update for BO backed fb
In reply to Deepak RawatUsing 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
@@ -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
PATCH 8/14 drm/vmwgfx: use the new interface for STDU plane update
In reply to Deepak RawatWith 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
@@ -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
Re: [PATCH 08/14] drm/vmwgfx: use the new interface for STDU plane update
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
Signed-off-by Deepak Rawat <[email protected]>--- drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 38 +++++++--------------------- 1 file changed, 9 insertions(+), 29 deletions(-)drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +8
-26@@ -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
-3if (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
-0if (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
PATCH 9/14 drm/vmwgfx: enable FB_DAMAGE_CLIPS property for STDU primary plane
In reply to Deepak RawatSTDU 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
@@ -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
Re: [PATCH 09/14] drm/vmwgfx: enable FB_DAMAGE_CLIPS property for STDU primary plane
Quoted 12 lines · Deepak Rawat
STDU primary plane now support damage clips, enable it for user-space.
Signed-off-by Deepak Rawat <[email protected]>--- drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c | 1 + 1 file changed, 1 insertion(+)drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c +0
-0
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
PATCH 10/14 drm/vmwgfx: implement SOU plane update for surface backed fb
In reply to Deepak RawatUsing 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
@@ -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
PATCH 11/14 drm/vmwgfx: implement SOU plane update for BO backed fb
In reply to Deepak RawatUsing 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
@@ -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
PATCH 12/14 drm/vmwgfx: use the new interface for SOU plane update
In reply to Deepak RawatWith 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
@@ -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
PATCH 13/14 drm/vmwgfx: enable FB_DAMAGE_CLIPS property for SOU primary plane
In reply to Deepak RawatSOU 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
@@ -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
PATCH 14/14 drm/vmwgfx: use atomic helper function for dirty fb IOCTL
In reply to Deepak RawatWith 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
@@ -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