Hi Arnd, Andy,

On 9-Aug-25 9:44 PM, Arnd Bergmann wrote:
> On Sat, Aug 9, 2025, at 12:00, Andy Shevchenko wrote:
>> On Fri, Aug 08, 2025 at 05:17:49PM +0200, Arnd Bergmann wrote:
>>> From: Arnd Bergmann <arnd@arndb.de>
>>>
>>> A few old machines have not been converted away from the old-style
>>> gpiolib interfaces. Make these select the new CONFIG_GPIOLIB_LEGACY
>>> symbol so the code still works where it is needed but can be left
>>> out otherwise.
>>
>>> --- a/drivers/platform/x86/x86-android-tablets/Kconfig
>>> +++ b/drivers/platform/x86/x86-android-tablets/Kconfig
>>> @@ -8,6 +8,7 @@ config X86_ANDROID_TABLETS
>>>  	depends on I2C && SPI && SERIAL_DEV_BUS
>>>  	depends on GPIOLIB && PMIC_OPREGION
>>>  	depends on ACPI && EFI && PCI
>>> +	select GPIOLIB_LEGACY
>>>  	select NEW_LEDS
>>>  	select LEDS_CLASS
>>>  	select POWER_SUPPLY
>>
>> Hmm... This is a surprising change. But I leave it to Hans.

Yes I was surprised by this myself since I explicitly removed
all legacy GPIO use from the x86-android-tablets code a while
ago (or so I thought).

> I think the only function that still needs it is
> x86_android_tablet_probe() doing
> 
> static struct gpio_keys_button *buttons;
> 
>                 for (i = 0; i < dev_info->gpio_button_count; i++) {
>                         ret = x86_android_tablet_get_gpiod(dev_info->gpio_button[i].chip,
>                                                            dev_info->gpio_button[i].pin,
>                                                            dev_info->gpio_button[i].button.desc,
>                                                            false, GPIOD_IN, &gpiod);
> 
>                         buttons[i] = dev_info->gpio_button[i].button;
>                         buttons[i].gpio = desc_to_gpio(gpiod);
>                         /* Release GPIO descriptor so that gpio-keys can request it */
>                         devm_gpiod_put(&x86_android_tablet_device->dev, gpiod);
>                 }
> 
> So the driver itself uses gpio descriptors, but it passes
> some of them into another driver by number. There is probably
> an easy workaround that I did not see.

Ah I see, so this is basically in the same boat as
drivers/input/misc/soc_button_array.c which also first
gets a gpio_desc and then calls desc_to_gpio() to store
the GPIO number in struct gpio_keys_button which is passed
as platform_data to drivers/input/keyboard/gpio_keys.c

The gpio_keys driver then converts things back
into a gpio_desc in gpio_keys_setup_key()
using devm_gpio_request_one() + gpio_to_desc()

So it looks like we need to add a gpiod member to
struct gpio_keys_button (include/linux/gpio_keys.h)
and modify gpio_keys.c to prefer that over using
button->gpio, something like the attached patch
basically.

I won't have time to work on this until September,
so if someone wants to take the attached patch and run
with it go for it.

Note the x86-android-tablets / soc_button_array code
will become responsible for requesting / releasing
the gpiod when using the new gpio_keys_button.gpiod
member.

For the x86-android-tablets code this is easy, just drop
these 2 lines:

                        /* Release GPIO descriptor so that gpio-keys can request it */
                        devm_gpiod_put(&x86_android_tablet_device->dev, gpiod);

And for soc_button_array.c it is _probably_ just a matter
of switching to devm_gpiod_get_index() and drop the
gpiod_put().

I have hardware to test both the x86-android-tablets
code as well as the soc_button_array code. I might be
able to do a quick test on August 22nd or 29th.

Regards,

Hans


p.s.

It looks like this will also help with:

drivers/platform/x86/meraki-mx100.c
drivers/platform/x86/pcengines-apuv2.c
drivers/platform/x86/barco-p50-gpio.c

I do not have hardware to test these, but if the changes
work for x86-android-tablets + soc_button_array then
hopefully they will be fine here too.