Re: [PATCH-next 20/20] gpio: gpio-omap: clean up register access in omap2_set_gpio_debounce()

[Linus Walleij]

On Mon, Jun 10, 2019 at 7:13 PM Grygorii Strashko
<grygorii.strashko@xxxxxx> wrote:

> From: Russell King <rmk+kernel@xxxxxxxxxxxxxxx>
>
> Signed-off-by: Russell King <rmk+kernel@xxxxxxxxxxxxxxx>
> Signed-off-by: Grygorii Strashko <grygorii.strashko@xxxxxx>

Patch applied.

At your convenience please look at the debounce function a bit
closer because I think there is a bug/unpredictable behavior:

IIUC the whole bank/block of GPIOs share the same debounce
timer setting, and it is currently handled in a "last caller wins"
manner, so if the different GPIOs in the bank has different
debounce settings, the call order decides what debounce time
is used across all of them.

In drivers/gpio/gpio-ftgpio.c function ftgpio_gpio_set_config()
I simply reject a dounce time setting
different from the currently configured if any GPIOs are
currently using the deounce feature.

(It's the semantic I came up with but maybe there are other
ideas here.)

Yours,
Linus Walleij