Re: [PATCH] mmc: core: Add DT bindings for card detect debounce time

[Linus Walleij <linus.walleij@linaro.org>]

On Sat, Jun 21, 2014 at 9:22 AM, Alexandre Courbot <gnurou@gmail.com> wrote:

>> I have added Linus Walleij and Alexandre Courbot, the maintainers of
>> gpio. Let's see if they can point us in a direction.
>
> I agree it would be nice if the debounce value could be handled by the
> GPIO framework.

I agree too.

> I just wonder what would be the correct way of doing
> it? Contrary to ACTIVE_LOW and other flags which can be specified with
> the GPIO phandle, debounce is a numeric value. Maybe using a different
> property, e.g.:
>
> cd-gpios = <&gpio 1 GPIO_ACTIVE_HIGH>;
> cd-gpios-debounce = <10>;
>
> When looking up a GPIO through gpiod_get(), the GPIO framework could
> then check for -debounce property and set the debounce time
> accordingly. At first sight I'd say that would work and could be used
> for MMC and all other subsystems that need something similar.

Yes, but we also need to write generic debounce handling into
the gpiolib, so it doesn't matter if the GPIO driver can or cannot
handle this itself. Some hardware has the capability to debounce
lines in *hardware*.

Right now a call to gpiod_set_debounce() will fail unless the
hardware has a debounce option.

What we should really do is make gpiod_set_debounce() always
work, put the debounce value into the gpio_desc, and move some
logic similar to what exists in drivers/input/keyboard/gpio_keys.c
into the gpiolib, then get rid of any local implementations like
in gpio_keys.

Then we can rely on the gpiolib handling this at all times, and also
to get the debounce config from DT.

Dmitry, opinions on this?

Anyone up for implementing the timer in gpiolib and moving the
keyboard driver over to using the gpiolib debounce set?

Yours,
Linus Walleij