Re: API for setting colors of RGB backlit keyboard zones (was [PATCH V3] platform/x86: hp-wmi: Support omen backlight control wmi-acpi methods)

[Hans de Goede]

Hi,

On 2/18/23 12:48, Pavel Machek wrote:
> Hi!
> 
> 
>>> I do agree with you that we need to avoid kbd_backlight in the name to avoid causing existing upower code to have weird interactions with this (it supports / assumes there is only 1 kbd_backlight LED class device).
>>>
>>> So lets go with just these 4:
>>>
>>> /sys/class/leds/hp_omen::kbd_zoned_backlight-1/
>>> /sys/class/leds/hp_omen::kbd_zoned_backlight-2/
>>> /sys/class/leds/hp_omen::kbd_zoned_backlight-3/
>>> /sys/class/leds/hp_omen::kbd_zoned_backlight-4/
>>>
>>> Using the _zoned_ between kbd and baclight to avoid confusing the existing upower code. Then once this has landed we can look into extending upower support for this.
>>>
>>> Note the requested documentation patch should probably also explain that the _zoned_ was done deliberately to make current upower code ignore the devices.
>>>
> 
>>
>> This makes sense, I agree that the global LED file will cause more confusion
>> and hacks in the code. I'll start working on the  _zoned_ naming scheme with
>> 4 files + documentation changes and make a patch for this soon!
>>
> 
> /sys/class/leds/:rgb:kbd_zoned_backlight-4/ is better than what was
> suggested above.

Ah yes using rgb for the color part of the name makes sense.

> But we already use _1 suffix to deduplicate the, so
> I'm not sure this is best naming.



I guess we could try to actually name the zones, something like
(no idea if this are indeed the 4 zones):

:rgb:kbd_zoned_backlight-main
:rgb:kbd_zoned_backlight-wasd
:rgb:kbd_zoned_backlight-cursor
:rgb:kbd_zoned_backlight-numpad

Rishit any comments on this or improvements to it.

> There are keyboards with per-key backlight. How do you suggest to
> solve those?

I really think those fall into a separate category, currently AFAIK
all support for those use /dev/hidraw directly from userspace.

And any kernel API would need to likely be ioctl based, allowing
setting all the LEDs in a single syscall otherwise setting the
LEDs becomes to expensive / introduces to much latency when
doing software driven animations. So I think the best thing to
do there is to declare these out-of-scope for the classic
sysfs based LED class API.

Regards,

Hans