Re: [PATCH 2/4] gpiolib: use const parameters when possible

["Janusz Użycki" <j.uzycki@elproma.com.pl>]

Hello,

W dniu 2014-11-25 o 08:37, Alexandre Courbot pisze:
> On Wed, Nov 19, 2014 at 7:09 PM, Uwe Kleine-König
> <u.kleine-koenig@pengutronix.de> wrote:
>> Hello Alexandre,
>>
>> On Wed, Nov 19, 2014 at 06:07:50PM +0900, Alexandre Courbot wrote:
>>>   On Wed, Nov 19, 2014 at 6:02 PM, Uwe Kleine-König
>>> <u.kleine-koenig@pengutronix.de> wrote:
>>>> I'd make gpiod_get_direction static and only use it to fill in
>>>> /sys/kernel/debug/gpio.
>>> That's very tempting. I see only atmel_serial.c using this function,
>>> and there is no gpio_get_direction() declared anywhere so no user of
>>> this either. I'm not sure what I was thinking when I decided to export
>>> it?
>> In next there is also drivers/tty/serial/mxs-auart.c.
> Ok. I think we can get rid of gpiod_get_direction() if
> serial_mctrl_gpio implements its own way to query the direction (maybe
> using a bitmap in the mctrl_gpios and a mctrl_gpio_get_direction()
> function that queries that map). Using gpiod_get_direction() is just
> too unreliable.

good idea

>
> Janusz, since your change for mxs-auart is in -next, would you mind
> amending it to do this? Then we could do the same for atmel_serial and
> remove gpiod_get_direction() from the public GPIO interface. This
> function would do more harm than good anyway.

After a discussion I prepared RFC patch set to avoid the function.
You will find it here [1]:
("[RFC PATCH 1/2] serial: mctrl-gpio: Add irqs helpers for input lines")
("[RFC PATCH 2/2] serial: mxs-auart: use helpers for gpio irqs")
and the function was moved for debug only:
("[PATCH v3] gpio: mxs: implement get_direction callback")

Affected drivers by the patch set in the next are:
- atmel_serial
- mxs-auart
- clps711x (very basic usage)
The RFC patch set removes the inconvenient function from mxs-auart as 
example.
I didn't prepared a patch for atmel_serial because I wait for comments about
mctrl_gpio_is_gpio() function to differentiate if line is GPIO or native in
enable/disable_ms() callbacks.
I also can't test it for other devices than mxs (mx28).

[1] http://www.spinics.net/lists/arm-kernel/msg378444.html

best regards
Janusz