Re: [PATCH] Input: adp5588-keys: Remove unused driver

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Fri, May 06, 2022 at 11:57:57AM +0000, Hennerich, Michael wrote:
> 
> 
> > -----Original Message-----
> > From: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
> > Sent: Donnerstag, 5. Mai 2022 09:50
> > To: Hennerich, Michael <Michael.Hennerich@analog.com>
> > Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>; Bogdan, Dragos
> > <Dragos.Bogdan@analog.com>; Sa, Nuno <Nuno.Sa@analog.com>; Arnd
> > Bergmann <arnd@arndb.de>; kernel@pengutronix.de; linux-
> > input@vger.kernel.org
> > Subject: Re: [PATCH] Input: adp5588-keys: Remove unused driver
> > 
> > 
> > Hello Michael,
> > 
> > On Thu, May 05, 2022 at 06:20:22AM +0000, Hennerich, Michael wrote:
> > > > -----Original Message-----
> > > > From: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
> > > > Sent: Mittwoch, 4. Mai 2022 10:46
> > > > To: Dmitry Torokhov <dmitry.torokhov@gmail.com>; Hennerich, Michael
> > > > <Michael.Hennerich@analog.com>
> > > > Cc: linux-input@vger.kernel.org; kernel@pengutronix.de; Arnd
> > > > Bergmann <arnd@arndb.de>
> > > > Subject: [PATCH] Input: adp5588-keys: Remove unused driver
> > > >
> > > > The last user is gone since 2018 (commit 4ba66a976072 ("arch: remove
> > > > blackfin port")). This is an i2c driver, so it could be used on a
> > > > non-blackfin machine, but this driver requires platform data, so it
> > > > cannot be bound using device tree.
> > >
> > > Hi Uwe,
> > >
> > > If we start removing drivers which obviously don't have a mainline
> > > in-tree user, we would upset up many users of these drivers.
> > > I agree on updating this driver to make platform data optional.
> > > We could provide a patch in a few days.
> > 
> > Just to add some background why I stumbled over this driver: On of my current
> > quests is to make i2c remove callbacks return void. As a preparation for that I
> > work on updating all i2c drivers to return 0 in
> > .remove() to make the change to void have no side effects.
> > 
> > One of the offenders is drivers/gpio/gpio-adp5588.c, which in the presence of a
> > pdata->teardown callback might return a non-zero value from .remove(). While
> > looking at the pdata of possible devices I only found
> > drivers/input/keyboard/adp5588-keys.c.
> > 
> > So the options for my quest are in increasing impact order:
> > 
> >  a) just warn if struct adp5588_gpio_platform_data::teardown fails and
> >     still return 0 from .remove()
> >  b) make struct adp5588_gpio_platform_data::teardown return void
> >  c) drop teardown support from adp5588_gpio_platform_data
> >  d) drop platform support from gpio-adp5588
> >  e) drop gpio-adp5588
> > 
> > Currently I'd go for at least d).
> > 
> > Having said that I think e) has a net benefit. If there is no user left it reduces
> > maintainance burden. If there is a user left, they hopefully will tell us, we can
> > restore the driver from git history and then at least know a tester for future
> > cleanups and changes.
> 
> Hi Uwe,
> 
> Thanks for the explanation.
> 
> I know that there are users of this driver. But I admit, we should have earlier
> made platform_data support optional and also add proper dt bindings.
> We're in progress doing so. And in the meanwhile, I would prefer a less
> disruptive intermediate change. For example c) with the promise we're working on d).

I am looking at the 2 drivers (adp5588-keys and gpio-adp5588) and I
think we need to add the missing functionality to adp5588-keys (and make
keyboard part optional) and drop the gpio one.

Thanks.

-- 
Dmitry