Hi Linus,

On 01/30/2017 04:19 PM, Linus Walleij wrote:

On Fri, Jan 27, 2017 at 5:15 PM, Alexandre TORGUE
[off-list ref] wrote:

quoted

Use device tree entries to declare gpio range. It will allow to use
no contiguous gpio bank and holes inside a bank.

Signed-off-by: Alexandre TORGUE <redacted>

(...)

quoted

+       of_property_read_string(np, "st,bank-name", &bank->gpio_chip.label);
+
+       if (!of_parse_phandle_with_fixed_args(np, "gpio-ranges", 3, 0, &args))
+               bank_nr = args.args[1] / STM32_GPIO_PINS_PER_BANK;
+       else {
+               range->name = bank->gpio_chip.label;
+               range->id = bank_nr;
+               range->pin_base = range->id * STM32_GPIO_PINS_PER_BANK;
+               range->base = range->id * STM32_GPIO_PINS_PER_BANK;
+               range->npins = npins;
+               range->gc = &bank->gpio_chip;
+               pinctrl_add_gpio_range(pctl->pctl_dev,
+                                      &pctl->banks[bank_nr].range);
+       }

I Keep the old way to get range in order to keep compatibility with 
older device tree (if no gpio ranges are defined).

Why are you doing this?

There is already code in drivers/gpio/gpiolib-of.c to pick ranges
from the device tree and add when you're adding the GPIO chips.

Yes, If I declare gpio-ranges inside DT node, the range is well handled 
by the gpiolib. But in this condition I use gpio-ranges entry to get the 
pin base offset for the bank we are registering. For example, if there 
an hole between 2 banks (GPIOA, GPIOD, ...) I need to know at which pin 
offset the bank D starts. It is the only way I found.

The "else" statement is here to get compatibility with older device tree 
(which don't use gpio-ranges entry).

Regards
Alex

Please use that or figure out what is needed to make it work for
you instead of reimplementing it.

Yours,
Linus Walleij