Various LED complexities was Re: [PATCH v5 05/26] leds: core: Add support for composing LED class device names

From: Pavel Machek
Date: Wed Jul 03 2019 - 18:05:32 EST


On Fri 2019-06-28 13:30:30, David Laight wrote:
> From: Linus Walleij
> > Sent: 28 June 2019 09:46
> ...
> > A problem with LEDs is that it invites bikeshedding because it is too
> > relateable.
>
> Bikeshedding leds :-)
>
> It also isn't at all clear how to handle bi-colour and tri-colour leds.
> ISTR the usual interface lets you set the brightness, but more often
> leds are single brightness but multi-colour.
> Eg the ethernet 'speed' led which is (usually) off/orange/green.
>
> Changing the brightness either means changing the current or using PWM.
> Both really require more hardware support than changing colours.
>
> I've done some led driving (for a front panel) from a PLD (small FPGA).
> As well as the obvious things I did:
> - dim: 1/8th on at 80Hz.
> - flash: 1/8th on at 4Hz.
> - orange: 50-50 red-green at 80Hz on an RGB led.
>
> There was also the 'ethernet activity' led which could either be driven
> by the hardware, or forced on/off/flash by the driver.
> If driven by the hardware, the software could read the current state.
>
> None of this really fitted the Linux leds interface.

Well, we are working on some of those :-). But lets discuss that in
separate threads.

In particular we are working on triggers and RGB LEDs.

bi-color LEDs seem to handled as two separate LEDs. Not much expected
to change there.

Pavel

--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Attachment: signature.asc
Description: Digital signature