Re: [PATCH 12/18] gpio: add support for the sl28cpld GPIO controller

[Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>]

pt., 27 mar 2020 o 16:28 Michael Walle <michael@xxxxxxxx> napisał(a):
>
> Am 2020-03-27 11:20, schrieb Linus Walleij:
> > On Thu, Mar 26, 2020 at 9:06 PM Michael Walle <michael@xxxxxxxx> wrote:
> >> Am 2020-03-25 12:50, schrieb Bartosz Golaszewski:
> >
> >> > In that case maybe you should use the disable_locking option in
> >> > regmap_config and provide your own callbacks that you can use in the
> >> > irqchip code too?
> >>
> >> But how would that solve problem (1). And keep in mind, that the
> >> reqmap_irqchip is actually used for the interrupt controller, which
> >> is not this gpio controller.
> >>
> >> Ie. the interrupt controller of the sl28cpld uses the regmap_irqchip
> >> and all interrupt phandles pointing to the interrupt controller will
> >> reference the toplevel node. Any phandles pointing to the gpio
> >> controller will reference the GPIO subnode.
> >
> > Ideally we would create something generic that has been on my
> > mind for some time, like a generic GPIO regmap irqchip now that
> > there are a few controllers like that.
> >
> > I don't know how feasible it is or how much work it would be. But
> > as with GPIO_GENERIC (for MMIO) it would be helpful since we
> > can then implement things like .set_multiple() and .get_multiple()
> > for everyone.
>
> For starters, would that be a drivers/gpio/gpio-regmap.c or a
> drivers/base/regmap/regmap-gpio.c? I would assume the first,
> because the stuff in drivers/base/regmap operates on a given
> regmap and we'd just be using one, correct? On the other hand
> there is also the reqmap-irq.c. But as pointed out before, it
> will add an interrupt controller to the regmap, not a device
> so to speak.
>
> -michael

This has been on my TODO list for so long, but I've never been able to
find the time... I'd really appreciate any effort in that direction as
I believe it would allow us to slowly port a big part of the GPIO
expander drivers over to it and make large portions of our codebase
generic.

Best regards,
Bartosz Golaszewski