Re: [PATCH] pinctrl: cherryview: Preserve CHV_PADCTRL1_INVRXTX_TXDATA flag on GPIOs

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 9/4/20 8:09 PM, Andy Shevchenko wrote:
> On Fri, Sep 04, 2020 at 07:21:41PM +0200, Hans de Goede wrote:
>> One some devices the GPIO should output the inverted value from what
> 
>> Fixes: a7d4b171660c ("Input: goodix - add support for getting IRQ + reset GPIOs on Cherry Trail devices")
> 
> Hans, a side note. Can you clean up that driver using ACPI GPIO quirks and the
> idea that IRQ line will be retrieved in client->irq?

Nope, the chip uses the IRQ line in a bi-directional manner,
when the chip is reset (by us) the IRQ line is a bootstrap pin
and we need to drive it either high or low to make sure the chip
re-appears on the same i2c-address as it had before. So we need
the get gpiod and then call gpiod_to_irq abstraction so that
we also have the gpiod to call allow calling
gpiod_direction_input() and gpiod_direction_output() on the
irq pin.

And to make things more interesting on some generations x86 hw
(apollo lake) the IRQ is actually listed as an Interrupt resource
in the ACPI device resources (_CRS) and there are GDIX specific
ACPI methods on the ACPI device which directly poke the GPIO
chip to allow setting it high/low (for during reset) or config
it as input. So the goodix.c has its own internal abstraction
around gpiod_direction_input(() and gpiod_direction_output() for
controlling the IRQ pin which does different things on different
x86 hw generations (while also maintaining device-tree compat).

So all in all I'm happy that we have this all working atm
(and it also is reasonably clean already IMHO) and I would
rather not muck with this.

Regards,

Hans