Re: [PATCH 3/3] backlight: lp8864: Convert from LED to backlight class driver

[Daniel Thompson]

On Tue, Jun 16, 2026 at 07:17:23AM +0000, Sverdlin, Alexander wrote:
> Hi Andrew!
> 
> On Mon, 2026-06-15 at 14:51 -0500, Andrew Davis wrote:
> > > Move the TI LP8864/LP8866 driver from drivers/leds/ to
> > > drivers/video/backlight/
> > 
> > Why move it? You can register a backlight device from any directory.
> 
> I'm personally fine with the driver residing in drivers/leds, it's
> just that currently there are no combined drivers there, the combined
> drivers providing both interfaces only live in video/backlight.
> 
> But if it's OK from the maintainers' perspective, it will be even
> more consistent regarding Kconfig symbol.
> 
> > > and convert it to register a backlight class
> > > device as its primary interface.
> > > 
> > 
> > What do you mean by "primary"? You should be able to register with
> > both frameworks and have the driver interop between as needed.
> 
> Well, I only meant the user's (or my own) perspective, sorry for confusion.
> 
> > > The motivation is a use case on a hot-pluggable segment of an I2C bus.
> > > The generic led-backlight driver (drivers/video/backlight/led_bl.c) is a
> > > platform driver and as such inherently non-hotpluggable.
> > 
> > That isn't strictly true, there is platform_device_{del,unregister}(), so
> > whatever your mechanism for removing the I2C device would be, the same
> > could be done to the led_bl device before then removing the I2C device.
> 
> led_bl is not really designed to act on dynamically instantiated devices,
> it's very much device-tree affine (of_count_phandle_with_args(), etc...)
> 
> > We don't want to have to move every LED driver that could possibly
> > be used as a backlight to the backlight framework, the led_bl.c
> > handles adapting LED->backlight as needed. So what you really need
> > here is to de-couple led_bl.c from DT so it can better handle dynamic
> > add/remove. Then this LED driver simply could register a "led-backlight"
> > platform driver to handle the backlight interface, and remove the
> > backlight device when it itself (the LED device) is removed.
> 
> The mechanism we have regarding hot plugging currently is just I2C bridge,
> which de-registers and registers the bridged bus. So no additional drivers
> are required, as long as I2C devices are self-contained and not glued with
> platform devices.
> 
> So bottom line is, I'd prefer to just add the backlight interface to the
> existing driver, no matter where it would live in the future.

Is there any reasion that LP8864/LP8866 is unique in appearing on a 
hotplugged I2C bus? In other words if support for dynamism is added
specifically to leds-lp8864.c rather than in led_bl.c then what will stop
the same dynamic tricks from being adde to other LED drivers?


Daniel.