[RFC,v6,00/10] Support for Solid Fill Planes

Message

Jessica Zhang Aug. 29, 2023, 12:05 a.m. UTC

Some drivers support hardware that have optimizations for solid fill
planes. This series aims to expose these capabilities to userspace as
some compositors have a solid fill flag (ex. SOLID_COLOR in the Android
hardware composer HAL) that can be set by apps like the Android Gears
app.

In order to expose this capability to userspace, this series will:

- Introduce solid_fill and pixel_source properties to allow userspace to
  toggle between FB and solid fill sources
- Loosen NULL FB checks within the DRM atomic commit callstack to allow
  for NULL FB when solid fill is enabled.
- Add NULL FB checks in methods where FB was previously assumed to be
  non-NULL
- Have MSM DPU driver use drm_plane_state.solid_fill instead of
  dpu_plane_state.color_fill

Note: The solid fill planes feature depends on both the solid_fill *and*
pixel_source properties.

To use this feature, userspace can set the solid_fill property to a blob
containing the appropriate version number and solid fill color (in
RGB323232 format) and and setting the pixel_source property to
DRM_PLANE_PIXEL_SOURCE_COLOR. This will disable memory fetch and the
resulting plane will display the color specified by the solid_fill blob.

Currently, there's only one version of the solid_fill blob property.
However if other drivers want to support a similar feature, but require
more than just the solid fill color, they can extend this feature by
creating additional versions of the drm_solid_fill struct.

This 2 property approach was chosen because passing in a special 1x1 FB
with the necessary color information would have unecessary overhead that
does not reflect the behavior of the solid fill feature. In addition,
assigning the solid fill blob to FB_ID would require loosening some core
drm_property checks that might cause unwanted side effects elsewhere.

---
Changes in v6:
- Have _dpu_plane_color_fill() take in a single ABGR8888 color instead
  of having separate alpha and BGR color parameters (Dmitry)
- Drop plane->state->pixel_source != DRM_PLANE_PIXEL_SOURCE_FB check
  in SetPlane ioctl (Dmitry)
- Add DRM_PLANE_PIXEL_SOURCE_NONE as a default pixel source (Sebastian)
- Dropped versioning from solid fill property blob (Dmitry)
- Use DRM_ENUM_NAME_FN (Dmitry)
- Use drm_atomic_replace_property_blob_from_id() (Dmitry)
- drm_atomic_check_fb -> drm_atomic_plane_check_fb (Dmitry)
- Group redundant NULL FB checks (Dmitry)
- Squashed drm_plane_needs_disable() implementation with 
  DRM_PLANE_PIXEL_SOURCE_NONE declaration (Sebastian)
- Add comment to support RGBA solid fill color in the future (Dmitry)
- Link to v5: https://lore.kernel.org/r/[e-mail]

Changes in v5:
- Added support for PIXEL_SOURCE_NONE (Sebastian)
- Added WARN_ON() in drm_plane_has_visible_data() if pixel_source isn't
  set (Dmitry)
- Added debugfs support for both properties (Dmitry)
- Corrected u32 to u8 conversion (Pekka)
- Moved drm_solid_fill_info struct and related documentation to
  include/uapi (Pekka)
- Changed drm_solid_fill_info.version to __u32 for data alignment (Pekka)
- Added more detailed UAPI and kernel documentation (Pekka)
- Reordered patch series so that the pixel_source property is introduced
  before solid_fill (Dmitry)
- Fixed inconsistent ABGR8888/RGBA8888 format declaration (Pekka)
- Reset pixel_source to FB in drm_mode_setplane() (Dmitry)
- Rename supported_sources to extra_sources (Dmitry)
- Only destroy old solid_fill blob state if new state is valid (Pekka)
- Link to v4: https://lore.kernel.org/r/[e-mail]

Changes in v4:
- Rebased onto latest kernel
- Reworded cover letter for clarity (Dmitry)
- Reworded commit messages for clarity
- Split existing changes into smaller commits
- Added pixel_source enum property (Dmitry, Pekka, Ville)
- Updated drm-kms comment docs with pixel_source and solid_fill
  properties (Dmitry)
- Inlined drm_atomic_convert_solid_fill_info() (Dmitry)
- Passed in plane state alpha value to _dpu_plane_color_fill_pipe()
- Link to v3: https://lore.kernel.org/r/[e-mail]

Changes in v3:
- Fixed some logic errors in atomic checks (Dmitry)
- Introduced drm_plane_has_visible_data() and drm_atomic_check_fb() helper
  methods (Dmitry)
- Fixed typo in drm_solid_fill struct documentation
- Created drm_plane_has_visible_data() helper and corrected CRTC and FB
  NULL-check logic (Dmitry)
- Merged `if (fb)` blocks in drm_atomic_plane_check() and abstracted
  them into helper method (Dmitry)
- Inverted `if (solid_fill_enabled) else if (fb)` check order (Dmitry)
- Fixed indentation (Dmitry)

Changes in v2:
- Dropped SOLID_FILL_FORMAT property (Simon)
- Switched to implementing solid_fill property as a blob (Simon, Dmitry)
- Added drm_solid_fill and drm_solid_fill_info structs (Simon)
- Changed to checks for if solid_fill_blob is set (Dmitry)
- Abstracted (plane_state && !solid_fill_blob) checks to helper method
  (Dmitry)
- Removed DPU_PLANE_COLOR_FILL_FLAG
- Fixed whitespace and indentation issues (Dmitry)
- Changed to checks for if solid_fill_blob is set (Dmitry)
- Abstracted (plane_state && !solid_fill_blob) checks to helper method
  (Dmitry)
- Fixed dropped 'const' warning
- Added helper to convert color fill to BGR888 (Rob)
- Fixed indentation issue (Dmitry)
- Added support for solid fill on planes of varying sizes

---
Jessica Zhang (10):
      drm: Introduce pixel_source DRM plane property
      drm: Introduce solid fill DRM plane property
      drm: Add solid fill pixel source
      drm/atomic: Add pixel source to plane state dump
      drm/atomic: Add solid fill data to plane state dump
      drm/atomic: Move framebuffer checks to helper
      drm/atomic: Loosen FB atomic checks
      drm/msm/dpu: Allow NULL FBs in atomic commit
      drm/msm/dpu: Use DRM solid_fill property
      drm/msm/dpu: Add solid fill and pixel source properties

 drivers/gpu/drm/drm_atomic.c              | 147 +++++++++++++++++-------------
 drivers/gpu/drm/drm_atomic_helper.c       |  36 ++++----
 drivers/gpu/drm/drm_atomic_state_helper.c |  10 ++
 drivers/gpu/drm/drm_atomic_uapi.c         |  30 ++++++
 drivers/gpu/drm/drm_blend.c               | 129 ++++++++++++++++++++++++++
 drivers/gpu/drm/drm_crtc_internal.h       |   1 +
 drivers/gpu/drm/drm_plane.c               |  27 +++++-
 drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c  |   9 +-
 drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c |  80 ++++++++++------
 include/drm/drm_atomic_helper.h           |   4 +-
 include/drm/drm_blend.h                   |   3 +
 include/drm/drm_plane.h                   |  90 ++++++++++++++++++
 include/uapi/drm/drm_mode.h               |  24 +++++
 13 files changed, 478 insertions(+), 112 deletions(-)
---
base-commit: 00ee72279c963989ab435b0bc90b5dc05a9aab79
change-id: 20230404-solid-fill-05016175db36

Best regards,

Comments

Sebastian Wick Aug. 29, 2023, 3:42 p.m. UTC | #1

On Mon, Aug 28, 2023 at 05:05:06PM -0700, Jessica Zhang wrote:
> Some drivers support hardware that have optimizations for solid fill
> planes. This series aims to expose these capabilities to userspace as
> some compositors have a solid fill flag (ex. SOLID_COLOR in the Android
> hardware composer HAL) that can be set by apps like the Android Gears
> app.
> 
> In order to expose this capability to userspace, this series will:
> 
> - Introduce solid_fill and pixel_source properties to allow userspace to
>   toggle between FB and solid fill sources
> - Loosen NULL FB checks within the DRM atomic commit callstack to allow
>   for NULL FB when solid fill is enabled.
> - Add NULL FB checks in methods where FB was previously assumed to be
>   non-NULL
> - Have MSM DPU driver use drm_plane_state.solid_fill instead of
>   dpu_plane_state.color_fill
> 
> Note: The solid fill planes feature depends on both the solid_fill *and*
> pixel_source properties.
> 
> To use this feature, userspace can set the solid_fill property to a blob
> containing the appropriate version number and solid fill color (in
> RGB323232 format) and and setting the pixel_source property to
> DRM_PLANE_PIXEL_SOURCE_COLOR. This will disable memory fetch and the
> resulting plane will display the color specified by the solid_fill blob.
> 
> Currently, there's only one version of the solid_fill blob property.
> However if other drivers want to support a similar feature, but require
> more than just the solid fill color, they can extend this feature by
> creating additional versions of the drm_solid_fill struct.
> 
> This 2 property approach was chosen because passing in a special 1x1 FB
> with the necessary color information would have unecessary overhead that
> does not reflect the behavior of the solid fill feature. In addition,
> assigning the solid fill blob to FB_ID would require loosening some core
> drm_property checks that might cause unwanted side effects elsewhere.

The cover letter is a bit outdated by now. Anyway, with Pekkas issues
addressed the core drm parts are

Acked-by: Sebastian Wick <[e-mail]>
 
> ---
> Changes in v6:
> - Have _dpu_plane_color_fill() take in a single ABGR8888 color instead
>   of having separate alpha and BGR color parameters (Dmitry)
> - Drop plane->state->pixel_source != DRM_PLANE_PIXEL_SOURCE_FB check
>   in SetPlane ioctl (Dmitry)
> - Add DRM_PLANE_PIXEL_SOURCE_NONE as a default pixel source (Sebastian)
> - Dropped versioning from solid fill property blob (Dmitry)
> - Use DRM_ENUM_NAME_FN (Dmitry)
> - Use drm_atomic_replace_property_blob_from_id() (Dmitry)
> - drm_atomic_check_fb -> drm_atomic_plane_check_fb (Dmitry)
> - Group redundant NULL FB checks (Dmitry)
> - Squashed drm_plane_needs_disable() implementation with 
>   DRM_PLANE_PIXEL_SOURCE_NONE declaration (Sebastian)
> - Add comment to support RGBA solid fill color in the future (Dmitry)
> - Link to v5: https://lore.kernel.org/r/[e-mail]
> 
> Changes in v5:
> - Added support for PIXEL_SOURCE_NONE (Sebastian)
> - Added WARN_ON() in drm_plane_has_visible_data() if pixel_source isn't
>   set (Dmitry)
> - Added debugfs support for both properties (Dmitry)
> - Corrected u32 to u8 conversion (Pekka)
> - Moved drm_solid_fill_info struct and related documentation to
>   include/uapi (Pekka)
> - Changed drm_solid_fill_info.version to __u32 for data alignment (Pekka)
> - Added more detailed UAPI and kernel documentation (Pekka)
> - Reordered patch series so that the pixel_source property is introduced
>   before solid_fill (Dmitry)
> - Fixed inconsistent ABGR8888/RGBA8888 format declaration (Pekka)
> - Reset pixel_source to FB in drm_mode_setplane() (Dmitry)
> - Rename supported_sources to extra_sources (Dmitry)
> - Only destroy old solid_fill blob state if new state is valid (Pekka)
> - Link to v4: https://lore.kernel.org/r/[e-mail]
> 
> Changes in v4:
> - Rebased onto latest kernel
> - Reworded cover letter for clarity (Dmitry)
> - Reworded commit messages for clarity
> - Split existing changes into smaller commits
> - Added pixel_source enum property (Dmitry, Pekka, Ville)
> - Updated drm-kms comment docs with pixel_source and solid_fill
>   properties (Dmitry)
> - Inlined drm_atomic_convert_solid_fill_info() (Dmitry)
> - Passed in plane state alpha value to _dpu_plane_color_fill_pipe()
> - Link to v3: https://lore.kernel.org/r/[e-mail]
> 
> Changes in v3:
> - Fixed some logic errors in atomic checks (Dmitry)
> - Introduced drm_plane_has_visible_data() and drm_atomic_check_fb() helper
>   methods (Dmitry)
> - Fixed typo in drm_solid_fill struct documentation
> - Created drm_plane_has_visible_data() helper and corrected CRTC and FB
>   NULL-check logic (Dmitry)
> - Merged `if (fb)` blocks in drm_atomic_plane_check() and abstracted
>   them into helper method (Dmitry)
> - Inverted `if (solid_fill_enabled) else if (fb)` check order (Dmitry)
> - Fixed indentation (Dmitry)
> 
> Changes in v2:
> - Dropped SOLID_FILL_FORMAT property (Simon)
> - Switched to implementing solid_fill property as a blob (Simon, Dmitry)
> - Added drm_solid_fill and drm_solid_fill_info structs (Simon)
> - Changed to checks for if solid_fill_blob is set (Dmitry)
> - Abstracted (plane_state && !solid_fill_blob) checks to helper method
>   (Dmitry)
> - Removed DPU_PLANE_COLOR_FILL_FLAG
> - Fixed whitespace and indentation issues (Dmitry)
> - Changed to checks for if solid_fill_blob is set (Dmitry)
> - Abstracted (plane_state && !solid_fill_blob) checks to helper method
>   (Dmitry)
> - Fixed dropped 'const' warning
> - Added helper to convert color fill to BGR888 (Rob)
> - Fixed indentation issue (Dmitry)
> - Added support for solid fill on planes of varying sizes
> 
> ---
> Jessica Zhang (10):
>       drm: Introduce pixel_source DRM plane property
>       drm: Introduce solid fill DRM plane property
>       drm: Add solid fill pixel source
>       drm/atomic: Add pixel source to plane state dump
>       drm/atomic: Add solid fill data to plane state dump
>       drm/atomic: Move framebuffer checks to helper
>       drm/atomic: Loosen FB atomic checks
>       drm/msm/dpu: Allow NULL FBs in atomic commit
>       drm/msm/dpu: Use DRM solid_fill property
>       drm/msm/dpu: Add solid fill and pixel source properties
> 
>  drivers/gpu/drm/drm_atomic.c              | 147 +++++++++++++++++-------------
>  drivers/gpu/drm/drm_atomic_helper.c       |  36 ++++----
>  drivers/gpu/drm/drm_atomic_state_helper.c |  10 ++
>  drivers/gpu/drm/drm_atomic_uapi.c         |  30 ++++++
>  drivers/gpu/drm/drm_blend.c               | 129 ++++++++++++++++++++++++++
>  drivers/gpu/drm/drm_crtc_internal.h       |   1 +
>  drivers/gpu/drm/drm_plane.c               |  27 +++++-
>  drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c  |   9 +-
>  drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c |  80 ++++++++++------
>  include/drm/drm_atomic_helper.h           |   4 +-
>  include/drm/drm_blend.h                   |   3 +
>  include/drm/drm_plane.h                   |  90 ++++++++++++++++++
>  include/uapi/drm/drm_mode.h               |  24 +++++
>  13 files changed, 478 insertions(+), 112 deletions(-)
> ---
> base-commit: 00ee72279c963989ab435b0bc90b5dc05a9aab79
> change-id: 20230404-solid-fill-05016175db36
> 
> Best regards,
> -- 
> Jessica Zhang <[e-mail]>
>

Harry Wentland Sept. 26, 2023, 3:25 p.m. UTC | #2

On 2023-08-28 20:05, Jessica Zhang wrote:
> Some drivers support hardware that have optimizations for solid fill
> planes. This series aims to expose these capabilities to userspace as
> some compositors have a solid fill flag (ex. SOLID_COLOR in the Android
> hardware composer HAL) that can be set by apps like the Android Gears
> app.
> 
> In order to expose this capability to userspace, this series will:
> 
> - Introduce solid_fill and pixel_source properties to allow userspace to
>   toggle between FB and solid fill sources
> - Loosen NULL FB checks within the DRM atomic commit callstack to allow
>   for NULL FB when solid fill is enabled.
> - Add NULL FB checks in methods where FB was previously assumed to be
>   non-NULL
> - Have MSM DPU driver use drm_plane_state.solid_fill instead of
>   dpu_plane_state.color_fill
> 
> Note: The solid fill planes feature depends on both the solid_fill *and*
> pixel_source properties.
> 
> To use this feature, userspace can set the solid_fill property to a blob
> containing the appropriate version number and solid fill color (in
> RGB323232 format) and and setting the pixel_source property to
> DRM_PLANE_PIXEL_SOURCE_COLOR. This will disable memory fetch and the
> resulting plane will display the color specified by the solid_fill blob.
> 
> Currently, there's only one version of the solid_fill blob property.
> However if other drivers want to support a similar feature, but require
> more than just the solid fill color, they can extend this feature by
> creating additional versions of the drm_solid_fill struct.
> 
> This 2 property approach was chosen because passing in a special 1x1 FB
> with the necessary color information would have unecessary overhead that
> does not reflect the behavior of the solid fill feature. In addition,
> assigning the solid fill blob to FB_ID would require loosening some core
> drm_property checks that might cause unwanted side effects elsewhere.
> 

I didn't have a detailed review of this patchset but at a high-level this
change makes sense to me.

Feel free to add my
Acked-by: Harry Wentland <[e-mail]>
to patches 1-5.

Harry

> ---
> Changes in v6:
> - Have _dpu_plane_color_fill() take in a single ABGR8888 color instead
>   of having separate alpha and BGR color parameters (Dmitry)
> - Drop plane->state->pixel_source != DRM_PLANE_PIXEL_SOURCE_FB check
>   in SetPlane ioctl (Dmitry)
> - Add DRM_PLANE_PIXEL_SOURCE_NONE as a default pixel source (Sebastian)
> - Dropped versioning from solid fill property blob (Dmitry)
> - Use DRM_ENUM_NAME_FN (Dmitry)
> - Use drm_atomic_replace_property_blob_from_id() (Dmitry)
> - drm_atomic_check_fb -> drm_atomic_plane_check_fb (Dmitry)
> - Group redundant NULL FB checks (Dmitry)
> - Squashed drm_plane_needs_disable() implementation with 
>   DRM_PLANE_PIXEL_SOURCE_NONE declaration (Sebastian)
> - Add comment to support RGBA solid fill color in the future (Dmitry)
> - Link to v5: https://lore.kernel.org/r/[e-mail]
> 
> Changes in v5:
> - Added support for PIXEL_SOURCE_NONE (Sebastian)
> - Added WARN_ON() in drm_plane_has_visible_data() if pixel_source isn't
>   set (Dmitry)
> - Added debugfs support for both properties (Dmitry)
> - Corrected u32 to u8 conversion (Pekka)
> - Moved drm_solid_fill_info struct and related documentation to
>   include/uapi (Pekka)
> - Changed drm_solid_fill_info.version to __u32 for data alignment (Pekka)
> - Added more detailed UAPI and kernel documentation (Pekka)
> - Reordered patch series so that the pixel_source property is introduced
>   before solid_fill (Dmitry)
> - Fixed inconsistent ABGR8888/RGBA8888 format declaration (Pekka)
> - Reset pixel_source to FB in drm_mode_setplane() (Dmitry)
> - Rename supported_sources to extra_sources (Dmitry)
> - Only destroy old solid_fill blob state if new state is valid (Pekka)
> - Link to v4: https://lore.kernel.org/r/[e-mail]
> 
> Changes in v4:
> - Rebased onto latest kernel
> - Reworded cover letter for clarity (Dmitry)
> - Reworded commit messages for clarity
> - Split existing changes into smaller commits
> - Added pixel_source enum property (Dmitry, Pekka, Ville)
> - Updated drm-kms comment docs with pixel_source and solid_fill
>   properties (Dmitry)
> - Inlined drm_atomic_convert_solid_fill_info() (Dmitry)
> - Passed in plane state alpha value to _dpu_plane_color_fill_pipe()
> - Link to v3: https://lore.kernel.org/r/[e-mail]
> 
> Changes in v3:
> - Fixed some logic errors in atomic checks (Dmitry)
> - Introduced drm_plane_has_visible_data() and drm_atomic_check_fb() helper
>   methods (Dmitry)
> - Fixed typo in drm_solid_fill struct documentation
> - Created drm_plane_has_visible_data() helper and corrected CRTC and FB
>   NULL-check logic (Dmitry)
> - Merged `if (fb)` blocks in drm_atomic_plane_check() and abstracted
>   them into helper method (Dmitry)
> - Inverted `if (solid_fill_enabled) else if (fb)` check order (Dmitry)
> - Fixed indentation (Dmitry)
> 
> Changes in v2:
> - Dropped SOLID_FILL_FORMAT property (Simon)
> - Switched to implementing solid_fill property as a blob (Simon, Dmitry)
> - Added drm_solid_fill and drm_solid_fill_info structs (Simon)
> - Changed to checks for if solid_fill_blob is set (Dmitry)
> - Abstracted (plane_state && !solid_fill_blob) checks to helper method
>   (Dmitry)
> - Removed DPU_PLANE_COLOR_FILL_FLAG
> - Fixed whitespace and indentation issues (Dmitry)
> - Changed to checks for if solid_fill_blob is set (Dmitry)
> - Abstracted (plane_state && !solid_fill_blob) checks to helper method
>   (Dmitry)
> - Fixed dropped 'const' warning
> - Added helper to convert color fill to BGR888 (Rob)
> - Fixed indentation issue (Dmitry)
> - Added support for solid fill on planes of varying sizes
> 
> ---
> Jessica Zhang (10):
>       drm: Introduce pixel_source DRM plane property
>       drm: Introduce solid fill DRM plane property
>       drm: Add solid fill pixel source
>       drm/atomic: Add pixel source to plane state dump
>       drm/atomic: Add solid fill data to plane state dump
>       drm/atomic: Move framebuffer checks to helper
>       drm/atomic: Loosen FB atomic checks
>       drm/msm/dpu: Allow NULL FBs in atomic commit
>       drm/msm/dpu: Use DRM solid_fill property
>       drm/msm/dpu: Add solid fill and pixel source properties
> 
>  drivers/gpu/drm/drm_atomic.c              | 147 +++++++++++++++++-------------
>  drivers/gpu/drm/drm_atomic_helper.c       |  36 ++++----
>  drivers/gpu/drm/drm_atomic_state_helper.c |  10 ++
>  drivers/gpu/drm/drm_atomic_uapi.c         |  30 ++++++
>  drivers/gpu/drm/drm_blend.c               | 129 ++++++++++++++++++++++++++
>  drivers/gpu/drm/drm_crtc_internal.h       |   1 +
>  drivers/gpu/drm/drm_plane.c               |  27 +++++-
>  drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c  |   9 +-
>  drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c |  80 ++++++++++------
>  include/drm/drm_atomic_helper.h           |   4 +-
>  include/drm/drm_blend.h                   |   3 +
>  include/drm/drm_plane.h                   |  90 ++++++++++++++++++
>  include/uapi/drm/drm_mode.h               |  24 +++++
>  13 files changed, 478 insertions(+), 112 deletions(-)
> ---
> base-commit: 00ee72279c963989ab435b0bc90b5dc05a9aab79
> change-id: 20230404-solid-fill-05016175db36
> 
> Best regards,