Re: [PATCH v3 2/8] mfd / platform: cros_ec: move lightbar attributes to its own driver.

From: Lee Jones
Date: Wed Dec 05 2018 - 03:09:41 EST


On Tue, 04 Dec 2018, Guenter Roeck wrote:
> On Tue, Dec 4, 2018 at 3:52 AM Enric Balletbo i Serra
> <enric.balletbo@xxxxxxxxxxxxx> wrote:
> > On 4/12/18 10:21, Lee Jones wrote:
> > > On Mon, 03 Dec 2018, Enric Balletbo i Serra wrote:
> > >> On 3/12/18 11:36, Lee Jones wrote:
> > >>> On Tue, 27 Nov 2018, Enric Balletbo i Serra wrote:
> > >>>
> > >>>> The entire way how cros sysfs attibutes are created is broken.
> > >>>> cros_ec_lightbar should be its own driver and its attributes should be
> > >>>> associated with a lightbar driver not the mfd driver. In order to retain
> > >>>> the path, the lightbar attributes are attached to the cros_class.
> > >>>
> > >>> I'm not exactly clear on what a lightbar is, but shouldn't it live in
> > >>> the appropriate subsystem. Like LED for example?
> > >>>
> > >>
> > >> The lightbar is a four-color indicator available on some Chromebook, but the
> > >> fact that can you can program this lightbar with different sequences, including
> > >> user defined sequences makes the device a bit special and very chrome platform
> > >> specific. The same happens with the VBC driver.
> > >
> > > Being Chrome specific doesn't necessarily mean that these drivers
> > > shouldn't reside in a proper subsystem. A lot of drivers in the
> > > kernel are only relevant to very specific hardware/platforms.
> >
> > Agree, and we try to put these drivers in their subsystem when we think it is
> > appropriate (we have in rtc, power, iio, keyboard, etc.)
> >
> > > IMHO code in drivers/platform should pertain only to the core platform
> > > itself. Any drivers for leaf hardware/functionality should be split
> > > out into the subsystems, however niche you think they are.
> > >
> >
> > To be honest, I don't have a hard opinion yet on if the drivers/platform should
> > pertain only to the core platform itself.
> >
> > The cros_ec_lightbar.c file already exists in drivers/platform, the patch just
> > converts it to a kernel module (only adds few lines). The main purpose of the se
> > patches was have cros-ec mfd code and their subdrivers separated instead of
> > having crossed calls.
> >
> > I don't mind to move to another subsystem (I need to discuss with the chromium
> > guys and I am still not sure if LED fits very well in this case, I can look in
> > more detail) but shouldn't be this a follow up patch?
>
> Separate patch, please, if anything.

Agreed.

> I would not know which subsystem to move this to, though, and moving
> it to misc just for the sake of it would seem odd, since this most
> definitely _is_ platform related code. What is platform for if not for
> platform specific code ?

"platform" has very broad and varying meanings in Computer Science
nomenclature. Unfortunately, it can be used to describe quite a lot
of different aspects of the (I wanted to say platform then) hardware.

All drivers are 'platform' device drivers, that's why we have the
platform_device_*/platform_driver_* API. Saying that "platform
specific code" should live in */platform reminds me of the
arch/*/<platform> days when similar arguments were put forward with
respect to jamming all "platform specific code" into arch/*. A great
deal of effort was put in to relocate it all into the proper
subsystems.

What I see here is the backslide into the same mind-set.

My guess is that defining what 'platform' means in the context of
drivers/platform is not going to be straight forward, however; I have
always seen it as a place to put *core* platform code, perhaps an API
akin to the one used for the EC in the Chromium context is a good
example. Drivers required to control peripherals that, for instance,
operate a set of LEDs, I would not consider core platform code.

In my mind, *any* and *all* leaf and peripheral drivers should be
relocated into a proper subsystem. Even if that is drivers/misc.

> > I am also worried on how this could affect the current user interface. Well,
> > something to look, right.
>
> No ABI changes, please.

Without seeing the way in which a user controls this stuff, it's
difficult to comment, but a change in subsystem usually means a shift
in control. Particularly if that control is operated via SYSFS.

--
Lee Jones [æçæ]
Linaro Services Technical Lead
Linaro.org â Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog