On 8/31/26 16:05, Mario Limonciello wrote:
At Display Next Hackfest 2026 we reviewed progress moving brightness control into the DRM connector properties. There is a range LUMINANCE property that will default to 0->0. Once a driver attaches a backlight it will be updated to 1->max. If the panel supports the minimum backlight turning off the display the range can later be updated to 0->max instead of 1->max. The legacy sysfs interface is synchronized with the DRM connector. When a compositor using this feature is loaded, sysfs writes are disabled to prevent legacy tools from going out of sync with the compositor. This has an implementation initially for amdgpu, i915, and Xe with eDP connectors. It can be extended to other connectors like DP for displays that can be controlled via DDC as well later. The following compositors have implemented matching support: * Kwin: https://invent.kde.org/plasma/kwin/-/merge_requests/9298 * Mutter: https://gitlab.gnome.org/swick/mutter/-/commits/wip/kms-luminance-prop * Wlroots: https://gitlab.freedesktop.org/wlroots/wlroots/-/merge_requests/5384 v6 -> v7: Core rework (feedback from Maxime Ripard): * Make the DRM backlight core backend-agnostic. Add a struct drm_backlight_funcs indirection so the backlight subsystem is just one backend; other backends (DDC/CI, MIPI-DCS, ...) can be added later without touching the core. * Stop dynamically allocating the DRM backlight. struct drm_backlight is now embedded in struct drm_connector and initialized by the core, so drivers no longer allocate it or handle allocation failure. * Drop the "drm/amd/display: Allow backlight registration to fail" patch; it is no longer needed now that nothing in the alloc path can fail. * Update the LUMINANCE property only through the atomic path. Remove the synchronous backlight writes from the property-set and legacy paths; the hardware is now touched solely from the atomic commit/enable path (via a workqueue, so slow backends never stall a commit). * Move the inlined luminance/DPMS logic into helpers and fix the comment style to match kernel conventions. Per-connector property (fixes a multi-panel bug): * Replace the single device-wide LUMINANCE property with a per-connector property created from the linked backend's range. Previously all connectors shared one property object, so linking a second panel corrupted the range reported for the others. * Never mutate the property min/max after creation (avoids racing GETPROPERTY). Add a kernel-internal drm_property.is_luminance flag to accept value 0 (DPMS off) without a device-wide pointer comparison. Kconfig (feedback from Thomas Zimmermann): * Drop "select BACKLIGHT_CLASS_DEVICE" from DRM. Add config DRM_BACKLIGHT which "depends on" BACKLIGHT_CLASS_DEVICE and provides no-op stubs when disabled, so DRM no longer forces the backlight subsystem into the kernel. Builds verified with the option both enabled and disabled. Documentation (feedback from Hans de Goede): * Document the LUMINANCE range table (1-N vs 0-N), that 1 is the minimum visible brightness, and that 0 means the display is turned off (with its vblank/pageflip implications). * Document BACKLIGHT_UPDATE_DRM. Misc: * backlight: fix a copy/paste kerneldoc on backlight_unregister_notifier and add a struct notifier_block forward declaration in backlight.h. * Split the old "drm: link connectors to backlight devices" patch into "drm/property: add a per-connector luminance flag" and "drm: add connector backlight (LUMINANCE) infrastructure", and rework the capability patch into "drm: add DRM_CLIENT_CAP_LUMINANCE". * Drop all Tested-by tags; the series has changed substantially and needs to be re-tested. Mario Limonciello (12): Revert "backlight: Remove notifier" backlight: add kernel-internal backlight API drm/property: add a per-connector luminance flag drm: add connector backlight (LUMINANCE) infrastructure drm: add DRM_CLIENT_CAP_LUMINANCE drm/amd/display: Pass up errors reading actual brightness drm/amd: Indicate driver supports luminance drm/amd/display: use drm backlight drm/bridge: auto-link panel backlight in bridge connector drm/xe: Indicate support for luminance on the connector drm/i915: Indicate support for luminance on the connector drm/i915/display: use drm backlight drivers/gpu/drm/Kconfig | 18 + drivers/gpu/drm/Makefile | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 1 + .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 5 +- .../display/amdgpu_dm/amdgpu_dm_backlight.c | 15 +- .../display/amdgpu_dm/amdgpu_dm_backlight.h | 2 +- .../display/amdgpu_dm/amdgpu_dm_connector.c | 2 + .../tests/amdgpu_dm_backlight_test.c | 4 +- drivers/gpu/drm/bridge/panel.c | 15 + .../gpu/drm/display/drm_bridge_connector.c | 11 +- drivers/gpu/drm/drm_atomic_helper.c | 39 ++ drivers/gpu/drm/drm_atomic_uapi.c | 49 +- drivers/gpu/drm/drm_backlight.c | 515 ++++++++++++++++++ drivers/gpu/drm/drm_connector.c | 56 ++ drivers/gpu/drm/drm_drv.c | 8 + drivers/gpu/drm/drm_file.c | 5 + drivers/gpu/drm/drm_ioctl.c | 17 + drivers/gpu/drm/drm_mode_config.c | 1 + drivers/gpu/drm/drm_property.c | 6 + drivers/gpu/drm/drm_sysfs.c | 26 +- .../gpu/drm/i915/display/intel_backlight.c | 4 + drivers/gpu/drm/i915/display/intel_display.c | 7 +- drivers/gpu/drm/i915/display/intel_dp.c | 1 + drivers/gpu/drm/i915/i915_driver.c | 1 + drivers/gpu/drm/xe/xe_device.c | 3 +- drivers/video/backlight/backlight.c | 99 ++++ include/drm/drm_atomic_helper.h | 2 + include/drm/drm_backlight.h | 158 ++++++ include/drm/drm_bridge.h | 1 + include/drm/drm_connector.h | 20 + include/drm/drm_drv.h | 14 + include/drm/drm_file.h | 8 + include/drm/drm_property.h | 10 + include/linux/backlight.h | 63 +++ include/uapi/drm/drm.h | 22 + 35 files changed, 1195 insertions(+), 15 deletions(-) create mode 100644 drivers/gpu/drm/drm_backlight.c create mode 100644 include/drm/drm_backlight.h
I don't think everyone else received it, but there was a good share of Sashiko generated feedback from this and a lot of it looks valid.
Don't worry about reviewing v7 of the series. I'll take into account the feedback, re-test and post a v8 when I'm done.
