Re: [PATCH v1 3/7] gpiolib: Change type of lflags in gpiod_hog() and gpiod_configure_flags()

[Linus Walleij <linus.walleij@xxxxxxxxxx>]

On Thu, Apr 4, 2019 at 4:22 PM Andy Shevchenko
<andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:>
> On Thu, Apr 04, 2019 at 11:52:20AM +0300, Mika Westerberg wrote:
> > On Thu, Apr 04, 2019 at 11:37:35AM +0300, Andy Shevchenko wrote:
> > > On Wed, Apr 03, 2019 at 09:36:24PM +0300, Mika Westerberg wrote:
> > > > On Tue, Apr 02, 2019 at 01:57:32PM +0300, Andy Shevchenko wrote:
> > > > > Most of the code inside GPIO library is using enum gpio_lookup_flags.
> > > > > Some of the function still operate with unsigned long.
> > > > >
> > > > > In order to be more consistent and better type checking, convert
> > > > > gpiod_hog() and gpiod_configure_flags() to use enum gpio_lookup_flags
> > > > > instead of unsigned long.
> > > > >
> > > > > Signed-off-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> > > > > ---
> > > > >  drivers/gpio/gpiolib.c | 4 ++--
> > > > >  drivers/gpio/gpiolib.h | 4 ++--
> > > > >  2 files changed, 4 insertions(+), 4 deletions(-)
> > > > >
> > > > > diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> > > > > index a3fe41a2c6c0..ea0d38164b06 100644
> > > > > --- a/drivers/gpio/gpiolib.c
> > > > > +++ b/drivers/gpio/gpiolib.c
> > > > > @@ -4078,7 +4078,7 @@ EXPORT_SYMBOL_GPL(gpiod_get_optional);
> > > > >   * occurred while trying to acquire the GPIO.
> > > > >   */
> > > > >  int gpiod_configure_flags(struct gpio_desc *desc, const char *con_id,
> > > > > -               unsigned long lflags, enum gpiod_flags dflags)
> > > > > +               enum gpio_lookup_flags lflags, enum gpiod_flags dflags)
> > > >
> > > > I actually think unsigned long here is more correct because enum is
> > > > supposed to present a single enumerated value and here we pass several of
> > > > them as bitmask.
> > >
> > > There are more functions that are using enum to keep bit mask, for example,
> > > gpiochip_request_own_desc(). Are you suggesting to change them all to unsigned
> > > long?
> >
> > No just wanted to mention it. For example enum gpiod_flags used by
> > gpiochip_request_own_desc() encodes bitmask as part of the enumeration
> > value which is fine but enum gpio_lookup_flags on the other hand does
> > not. So for the latter I would prefer to use unsigned int/long instead.
>
> Ah, okay, so, basically using enum gpio_lookup_flags as a parameter to
> functions doesn't feel right.

The enum gpio_lookup_flags is essentially something I came up with
for <linux/gpio/machine.h> to use used by struct gpiod_lookup.

It looks like it does (not just 0, 1, 2, 3... ) because it mirrors the defines
for e.g. <linux/gpio.h> so if we can get rid of that legacy file we can
change the values for enum gpio_lookup_flags.

Maybe it was stupid of me. We could change the enum gpio_lookup_flags
to just
#define GPIO_ACTIVE_HIGH 0
#define GPIO_ACTIVE_LOW BIT(0)
etc.

And then just have some unsigned long in struct gpiod_lookup.

What do you folks think makes most sense?

Yours,
Linus Walleij