Re: [RFC] drm/kms: control display brightness through drm_connector properties

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 4/14/22 15:10, Jani Nikula wrote:
> On Thu, 07 Apr 2022, Hans de Goede <hdegoede@redhat.com> wrote:
>> As discussed already several times in the past:
>>  https://www.x.org/wiki/Events/XDC2014/XDC2014GoedeBacklight/
>>  https://lore.kernel.org/all/4b17ba08-39f3-57dd-5aad-d37d844b02c6@linux.intel.com/
>>
>> The current userspace API for brightness control offered by
>> /sys/class/backlight devices has various issues, the biggest 2 being:
>>
>> 1. There is no way to map the backlight device to a specific
>>    display-output / panel (1)
>> 2. Controlling the brightness requires root-rights requiring
>>    desktop-environments to use suid-root helpers for this.
>>
>> As already discussed on various conference's hallway tracks
>> and as has been proposed on the dri-devel list once before (2),
>> it seems that there is consensus that the best way to to solve these
>> 2 issues is to add support for controlling a video-output's brightness
>> through properties on the drm_connector.
>>
>> This RFC outlines my plan to try and actually implement this,
>> which has 3 phases:
>>
>>
>> Phase 1: Stop registering multiple /sys/class/backlight devs for a single display
>> =================================================================================
>>
>> On x86 there can be multiple firmware + direct-hw-access methods
>> for controlling the backlight and in some cases the kernel registers
>> multiple backlight-devices for a single internal laptop LCD panel:
>>
>> a) i915 and nouveau unconditionally register their "native" backlight dev
>>    even on devices where /sys/class/backlight/acpi_video0 must be used
>>    to control the backlight, relying on userspace to prefer the "firmware"
>>    acpi_video0 device over "native" devices.
>> b) amdgpu and nouveau rely on the acpi_video driver initializing before
>>    them, which currently causes /sys/class/backlight/acpi_video0 to usually
>>    show up and then they register their own native backlight driver after
>>    which the drivers/acpi/video_detect.c code unregisters the acpi_video0
>>    device. This means that userspace briefly sees 2 devices and the
>>    disappearing of acpi_video0 after a brief time confuses the systemd
>>    backlight level save/restore code, see e.g.:
>>    https://bbs.archlinux.org/viewtopic.php?id=269920
>>
>> I already have a pretty detailed plan to tackle this, which I will
>> post in a separate RFC email. I plan to start working on this right
>> away, as it will be good to have this fixed regardless.
>>
>>
>> Phase 2: Add drm_connector properties mirroring the matching backlight device
>> =============================================================================
>>
>> The plan is to add a drm_connector helper function, which optionally takes
>> a pointer to the backlight device for the GPU's native backlight device,
>> which will then mirror the backlight settings from the backlight device
>> in a set of read/write brightness* properties on the connector.
>>
>> This function can then be called by GPU drivers for the drm_connector for
>> the internal panel and it will then take care of everything. When there
>> is no native GPU backlight device, or when it should not be used then
>> (on x86) the helper will use the acpi_video_get_backlight_type() to
>> determine which backlight-device should be used instead and it will find
>> + mirror that one.
>>
>>
>> Phase 3: Deprecate /sys/class/backlight uAPI
>> ============================================
>>
>> Once most userspace has moved over to using the new drm_connector
>> brightness props, a Kconfig option can be added to stop exporting
>> the backlight-devices under /sys/class/backlight. The plan is to
>> just disable the sysfs interface and keep the existing backlight-device
>> internal kernel abstraction as is, since some abstraction for (non GPU
>> native) backlight devices will be necessary regardless.
>>
>> An alternative to disabling the sysfs class entirely, would be
>> to allow setting it to read-only through Kconfig.
>>
>>
>> What scale to use for the drm_connector bl_brightness property?
>> ===============================================================
>>
>> The tricky part of this plan is phase 2 and then esp. defining what the
>> new brightness properties will look like and how they will work.
>>
>> The biggest challenge here is to decide on a fixed scale for the main
>> brightness property, say 0-65535, using scaling where the actual hw scale
>> is different, or if this should simply be a 1:1 mirror of the current
>> backlight interface, with the actual hw scale / brightness_max value
>> exposed as a drm_connector property.
>>
>> 1:1 advantages / 0-65535 disadvantages
>> - Userspace will likely move over to the connector-props quite slowly and
>>   we can expect various userspace bits, esp. also custom user scripts, to
>>   keep using the old uAPI for a long time. Using the 2 APIs are intermixed
>>   is fine when using a 1:1 brightness scale mapping. But if we end up doing
>>   a scaling round-trip all the time then eventually the brightness is going
>>   do drift. This can even happen if the user never changes the brightness
>>   when userspace saves it over suspend/resume or reboots.
>> - Almost all laptops have brightness up/down hotkeys. E.g GNOME decides
>>   on a step size for the hotkeys by doing min(brightness_max/20, 1).
>>   Some of the vendor specific backlight fw APIs (e.g. dell-laptop) have
>>   only 8 steps. When giving userspace the actual max_brightness value, then
>>   this will all work just fine. When hardcode brightness_max to 65535 OTOH
>>   then in this case GNOME will still give the user 20 steps where only 1
>>   in every 2-3 steps actually changes the brightness which IMHO is
>>   an unacceptably bad user experience.
>>
>> 0-65535 advantages / 1:1 disadvantages
>> - Without a fixed scale for the brightness property the brightness_max
>>   value may change after an userspace application's initial enumeration
>>   of the drm_connector. This can happen when neither the native GPU nor
>>   the acpi_video backlight devices are present/usable in this case
>>   acpi_video_get_backlight_type() will _assume_ a vendor specific fw API
>>   will be used for backlight control and the driver proving the "vendor"
>>   backlight device will show up much later and may even never show-up,
>>   so waiting for it is not an option. With a fixed 0-65535 scale userspace
>>   can just always assume this and the drm_connector backlight props helper
>>   code can even cache writes and send it to the actual backlight device
>>   when it shows up. With a 1:1 mapping userspace needs to listen for
>>   a uevent and then update the brightness range on such an event.
>>
>> I believe that the 1:1 mapping advantages out way the disadvantages
>> here. Also note that current userspace already blindly assumes that
>> all relevant drivers are loaded before the graphical-environment
>> starts and all the desktop environments as such already only do
>> a single scan of /sys/class/backlight when they start. So when
>> userspace forgets to add code to listen for the uevent when switching
>> to the new API nothing changes; and with the uevent userspace actually
>> gets a good mechanism to detect backlight drivers loading after
>> the graphical-environment has already started.
>>
>> So based on this here is my proposal for a set of new brightness
>> properties on the drm_connector object. Note these are all prefixed with
>> bl which stands for backlight, which is technically not correct for OLED.
>> But we need a prefix to avoid a name collision with the "brightness"
>> attribute which is part of the existing TV specific properties and IMHO
>> it is good to have a common prefix to make it clear that these all
>> belong together.
>>
>>
>> The drm_connector brightness properties
>> =======================================
>>
>> bl_brightness: rw 0-int32_max property controlling the brightness setting
>> of the connected display. The actual maximum of this will be less then
>> int32_max and is given in bl_brightness_max.
>>
>> bl_brightness_max: ro 0-int32_max property giving the actual maximum
>> of the display's brightness setting. This will report 0 when brightness
>> control is not available (yet).
>>
>> bl_brightness_0_is_min_brightness: ro, boolean
>> When this is set to true then it is safe to set brightness to 0
>> without worrying that this completely turns the backlight off causing
>> the screen to become unreadable. When this is false setting brightness
>> to 0 may turn the backlight off, but this is _not_ guaranteed.
>> This will e.g. be true when directly driving a PWM and the video-BIOS
>> has provided a minimum (non 0) duty-cycle below which the driver will
>> never go.
>>
>> bl_brightness_control_method: ro, enum, possible values:
>> none:     The GPU driver expects brightness control to be provided by another
>>           driver and that driver has not loaded yet.
>> unknown:  The underlying control mechanism is unknown.
>> pwm:      The brightness property directly controls the duty-cycle of a PWM
>>           output.
>> firmware: The brightness is controlled through firmware calls.
>> DDC/CI:   The brightness is controlled through the DDC/CI protocol.
>> gmux:     The brightness is controlled by the GMUX.
>> Note this enum may be extended in the future, so other values may
>> be read, these should be treated as "unknown".
> 
> Some eDP panels support brightness control via DPCD, in complex
> ways. Some of them support mixed modes via both DPCD and PWM
> simultaneously. Some of them support luminance based control.
> 
> DSI command mode panels support brightness control via DCS commands.

Right the whole bl_brightness_control_method is a bad idea and
I'll drop for the actual implementation of this.

> 
>> When brightness control becomes available after being reported
>> as not available before (bl_brightness_control_method=="none")
>> a uevent with CONNECTOR=<connector-id> and
>> PROPERTY=<bl_brightness_control_method-id> will be generated
>> at this point all the properties must be re-read.
>>
>> When/once brightness control is available then all the read-only
>> properties are fixed and will never change.
>>
>> Besides the "none" value for no driver having loaded yet,
>> the different bl_brightness_control_method values are intended for
>> (userspace) heuristics for such things as the brightness setting
>> linearly controlling electrical power or setting perceived brightness.
> 
> I'm not sure if it's a good idea to expose this with the goal that it's
> to be used for heuristics. We usually don't even know if we're
> controlling actual backlight brightness or just the OLED
> brightness. Basically any of the methods could be used to control OLED,
> or some HDR display with luminance based controls, and the heuristics
> would be off.

Ack, as said I plan to drop both the bl_brightness_0_is_min_brightness
(discussed elsewhere in the thread) and bl_brightness_control_method
properties, leaving just bl_brightness + bl_brightness_max.

> There are some cases where we can actually get a rough PWM/luminance
> curve from i915 opregion. I think maybe 16 data points. We've never
> exposed that. My idea was that you'd have a property where you could add
> data points for the curve, it could get pre-populated by the kernel if
> the kernel knows how to do it, defaulting to linear, but it could also
> be set or adjusted by userspace. The point would be that the userspace
> adjusts brightness linearly, and the kernel would use the curve data
> points to adjust it non-linearly. The userspace could have completely
> separated brightness adjustment and curve adjustment, and the brightness
> adjustment would be dead simple.

Interesting, I guess this could be a future feature addition on top
of my work.

> Finally, I'd drop "backlight" as a term throughout. It's brightness
> we're setting, and backlight is just a panel implementation detail.

Right, I'm fine with dropping backlight, but we do need a prefix
for the brightness property because there already is a plain
"brightness" property which is part of the existing TV specific
properties.

So how about: display_brightness or panel_brightness ?

I'm not sure which one I like better myself...

Regards,

Hans