[PATCH 3/4] gpiolib: Add support for GPIO "hog" nodes

[Sascha Hauer]

On Tue, May 23, 2017 at 04:25:03PM -0700, Andrey Smirnov wrote:
> On Mon, May 22, 2017 at 11:52 PM, Nikita Yushchenko
> <nikita.yoush at cogentembedded.com> wrote:
> >> +     ret = of_property_read_u32(chip_np, "#gpio-cells", &gpio_cells);
> >> +     if (ret)
> >> +             return ret;
> >> +
> >> +     if (WARN_ON(gpio_cells != 2))
> >> +             return -ENOTSUPP;
> >> +
> >> +     ret = of_property_read_u32_index(np, "gpios", idx * gpio_cells,
> >> +                                      &gpio_num);
> >> +     if (ret)
> >> +             return ret;
> >> +
> >> +     ret = of_property_read_u32_index(np, "gpios", idx * gpio_cells + 1,
> >> +                                      &gpio_flags);
> >> +     if (ret)
> >> +             return ret;
> >
> > Doesn't this hardcode interpretation of device tree words in gpio
> > specification - while this is intended to be gpio-provider specific and
> > that's why #gpio-cells exist?
> >
> 
> It does and yes that's my understanding of the purpose of #gpio-cells
> as well. The reason I did in such a primitive way was because
> Barebox's GPIO subsystem doesn't have any translation plumbing to be
> able to handle anything more than a simple one dimensional offset.
> Given the fact that of_get_named_gpio_flags() make similar assumption
> I thought that there are no real consumers of that functionality and
> left proper implementation as a future improvement that can be made
> once the need arises.
> 
> >
> >> +static int of_gpiochip_scan_gpios(struct gpio_chip *chip)
> >
> > Not best choice of name for routine that scans hogs?
> >
> > (although I understand that it comes from linux counterpart)
> >
> 
> Eh, I don't have any strong opinion on this one, I am more than happy
> to rename it if you think there are better alternatives.
> 
> >> -     return 0;
> >> +     return of_gpiochip_scan_gpios(chip);
> >
> > Should we fail gpiochip registration on failure to claim hogs?
> > I don't know.
> 
> I couldn't think of a use-case where it wasn't basically all or
> nothing: either I get everything working or I need to go back and fix
> my DT. Sascha, do you have an opinion on this one?

I can't think of any good reason why the hog initialization should fail
and we still want to keep the gpio chip.
Let's keep it like this until someone delivers us a good reason.

Sasch

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |