Re: [PATCH] gpio: Allow user to customise maximum number of GPIOs

[Linus Walleij <linus.walleij@xxxxxxxxxx>]

On Thu, Aug 18, 2022 at 2:46 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
> On Thu, Aug 18, 2022 at 2:25 PM Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:

> > git grep 'base = -1' yields these suspects:
> >
> > arch/arm/common/sa1111.c:       sachip->gc.base = -1;
> > arch/arm/common/scoop.c:        devptr->gpio.base = -1;
> > arch/powerpc/platforms/52xx/mpc52xx_gpt.c:      gpt->gc.base = -1;
> > arch/powerpc/platforms/83xx/mcu_mpc8349emitx.c: gc->base = -1;
> >
> > That's all! We could just calculate these to 512-ngpios and
> > hardcode that instead.
>
> How do the consumers find the numbers for these four?

For SA1111 the chip gets named "sa1111" and some consumers actually
use proper machine descriptions, maybe all?

arch/arm/mach-sa1100/jornada720.c:              GPIO_LOOKUP("sa1111",
0, "s0-power", GPIO_ACTIVE_HIGH),
arch/arm/mach-sa1100/jornada720.c:              GPIO_LOOKUP("sa1111",
1, "s1-power", GPIO_ACTIVE_HIGH),
(...)

For Scoop it is conditionally overridden in the code. I guess always
overridden.

For powerpc these seem to be using (old but working) device tree
lookups, so should not be an issue.

Sadly I'm not 100% sure that there are no random hard-coded
GPIO numbers referring to whatever the framework gave them
at the time the code was written :(

Another reason the base is assigned from above (usually
from 512 and downward) is that the primary SoC GPIO usually
want to be at base 0 and there is no guarantee that it will
get probed first. So hard-coded GPIO bases go from 0 -> n
and dynamically allocateed GPIO bases from n <- 512.

Then we hope they don't meet and overlap in the middle...

> > and in that case it is better to delete the use of this function
> > altogether since it can not fail.
>
> S32_MAX might be a better upper bound. That allows to
> just have no number assigned to a gpio chip. Any driver
> code calling desc_to_gpio() could then get back -1
> or a negative error code.
>
> Making the ones that are invalid today valid sounds like
> a step backwards to me if the goal is to stop using
> gpio numbers and most consumers no longer need them.

OK I get it...

Now: who wants to write this patch? :)

Christophe? Will you take a stab at it?

Yours,
Linus Walleij