Hi, Userspace currently has no atomic way to bring a display pipeline back to a pristine state. A compositor that wants to start from a known baseline must explicitly set every property on every KMS object to its default value, which requires tracking which properties exist and what their defaults are. This is fragile and must be updated every time a new property is added to the kernel.
This series introduces a new DRM_MODE_ATOMIC_RESET flag for the atomic ioctl. When set, the kernel fills the commit with default states for all KMS objects before applying the properties supplied in the request. Properties not explicitly included remain at their defaults. This allows userspace to describe the desired end state declaratively, without caring about the current state or the full set of properties. A driver-specific property is added to vkms planes for testing. Standard properties are impractical for this: compositors like Mutter enumerate all known properties, making it hard to isolate the reset behavior in a test. Mutter has a working implementation using the new flag, with a vkms-based test exercising the reset behavior: https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/5309 And we have IGT tests as well: https://lore.kernel.org/r/[email protected]/ Let me know what you think, Maxime Signed-off-by: Maxime Ripard <[email protected]> --- Changes in v11: - Reorder object state allocations to be more robust against lease revocation - Bring back lease revocation check - Link to v10: https://lore.kernel.org/r/[email protected] Changes in v10: - Stop trying to make partial lease work, they aren't meant to exist, so we can simply error out when they happen - More tests for zpos and color management - Fix DPU1 issues reported by sashiko - Link to v9: https://lore.kernel.org/r/[email protected] Changes in v9: - Add a flag to know if leases have been revoked or are empty - Handle the situation where a leased plane or connector being reset would disable an unleased CRTC - Add more kunit tests to cover the new issues reported by sashiko - Fix reset-related bugs in amdgpu, vc4 and msm DPU - Add a vc4 kunit test for its reset related bug - Link to v8: https://lore.kernel.org/r/[email protected] Changes in v8: - Fix more leases issues - Drop applied patches - Link to v7: https://lore.kernel.org/r/[email protected] Changes in v7: - Fix a whole bunch of sashiko issues to deal with color_mgmt_changed, leases, partially leased devices (like a plane leased but it's crtc isn't) and the interaction with PAGE_FLIP_EVENT better - Fix crtc create_state implementation for ssd130x - Actually make it possible for a compositor to recreate the same whole state without passing ALLOW_MODESET - Fix EDEADLK bugs with kunit tests - Don't report DRM_CAP_ATOMIC_RESET on !atomic drivers - Fix error handling in vkms for the new property - Add more kunit tests for the new corner cases - Link to v6: https://lore.kernel.org/r/[email protected] Changes in v6: - If using a lease, only considered the leased objects - Fix amdgpu plane atomic_create_state implementation - Link to v5: https://lore.kernel.org/r/[email protected] Changes in v5: - Fix bisection issue - Improve the documentation - Fix connector hotplugging issue - Added kunit tests for drm_atomic_commit_fill_with_defaults() - Link to v4: https://lore.kernel.org/r/[email protected] Changes in v4: - Improve the drm_atomic_get_private_obj_state and drm_atomic_get_connector_state array reallocations - Return void on functions to insert an object state into a commit when it cannot fail - Rename drm_atomic_can_create_state() into drm_atomic_implements_create_state() - Move capability check from drm_atomic_commit_fill_with_defaults to its caller - Added IGT test - Link to v3: https://lore.kernel.org/r/[email protected] Changes in v3: - Rebased on top of the latest atomic_create_state series - Sebastian Wick created and tested this work with mutter and vkms - Fix colorop locking - Link to v2: https://lore.kernel.org/r/[email protected] Changes in v2: - Fix bisection - Add capability to let userspace know if it can reset the state - Link to v1: https://lore.kernel.org/r/[email protected] --- Maxime Ripard (8): drm/lease: Add a helper to tell whether a lease was revoked drm/atomic: Add drm_atomic_implements_create_state() helper drm/atomic: Allow filling a commit with pristine object states drm/tests: atomic: Add tests for drm_atomic_commit_fill_with_defaults() drm/amd/display: Take over the current stream and dc_state drm/vc4: Carry the HVS channel over to created CRTC states drm/msm/dpu1: Carry the mixers over to created CRTC states drm/vkms: Add driver-specific plane property for testing Sebastian Wick (1): drm/atomic-uapi: Add DRM_MODE_ATOMIC_RESET flag drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 50 +++ drivers/gpu/drm/drm_atomic.c | 373 +++++++++++++++++++ drivers/gpu/drm/drm_atomic_uapi.c | 44 ++- drivers/gpu/drm/drm_ioctl.c | 5 + drivers/gpu/drm/drm_lease.c | 26 ++ drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c | 23 ++ drivers/gpu/drm/tests/drm_atomic_test.c | 435 +++++++++++++++++++++- drivers/gpu/drm/vc4/tests/vc4_test_pv_muxing.c | 105 ++++++ drivers/gpu/drm/vc4/vc4_kms.c | 16 +- drivers/gpu/drm/vkms/vkms_drv.h | 3 + drivers/gpu/drm/vkms/vkms_plane.c | 63 +++- include/drm/drm_atomic.h | 4 + include/drm/drm_lease.h | 2 + include/uapi/drm/drm.h | 10 + include/uapi/drm/drm_mode.h | 22 +- 15 files changed, 1171 insertions(+), 10 deletions(-) --- base-commit: 02d46912debff86b107519f9557557b3f9c13df4 change-id: 20260708-drm-reset-state-flag-2fb2b5711f97 Best regards, -- Maxime Ripard <[email protected]>
