Re: [PATCH 1/6] dt-bindings: mfd: Add ST Multi-Function eXpander driver

[Linus Walleij]

On Mon, Feb 19, 2018 at 4:59 PM, Amelie DELAUNAY <amelie.delaunay@xxxxxx> wrote:
> On 02/19/2018 12:19 AM, Rob Herring wrote:
>> On Thu, Feb 08, 2018 at 03:27:32PM +0100, Amelie Delaunay wrote:

>>> +- interrupts = must be <0>
>>> +- gpio-controller: marks the device node as a GPIO controller
>>> +- #gpio-cells: should be <2>, the first cell is the GPIO offset on this GPIO
>>> +  controller, the second cell is the gpio flags in accordance with
>>> +  <dt-bindings/gpio/st-mfx-gpio.h>.
>>
>> Custom flags? Use standard flags.
>>
>> DT binding headers should be part of this patch.
>>
> Custom flags because MFX GPIOs have several types:
> - Output open drain with internal pull-up resistor
> - Output open drain without internal pull-up resistor
> - Output push pull without internal pull resistor
> - Input with internal pull-up resistor
> - Input with internal pull-down resistor
> - Input floating
> - Input analog.
> Standard flags don't have pull up or down information. That's why I am
> using custom flags, they overloads standard flags.

This is because pull up/down and tristate/high z (floating)
also known as PIN_CONFIG_BIAS_HIGH_IMPEDANCE,
is controlled by the pin control subsystem, not GPIO.

If your hardware does even more pin control (like multiplexing)
then I suggest that you create a pin control driver back-end for
this and put the resulting driver in drivers/pinctrl.

Some recent additions of expanders in drivers/pinctrl makes
for great inspiration for this work. See for example:
drivers/pinctrl/pinctrl-sx150x.c
drivers/pinctrl/pinctrl-mcp23s08.c

These are both combined pin control and GPIO drivers that
we moved from drivers/gpio because we concluded that they
do more than just GPIO. The GPIO lines are matches to
pins using the GPIO ranges from the call to
gpiochip_add_pin_range() and using gpiochip_generic_config()
as the gpiochip .set_config() callback.

It has been discussed to expose subsets of pin config to
GPIO. Indeed, we already handle open drain/source and
debounce by simply calling into the pin control back-end
or handling it directly in the GPIO driver using the standard
pin config properties.

I am reluctant about this, I think the split of responsibilities
is pretty nice.

I'd recommend looking at the sx150x pin control driver and
check if you can maybe take this direction with the patch?

Yours,
Linus Walleij