Re: [PATCH v2 01/13] gpiolib: add gpiod_multi_set_value_cansleep()

[Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>]

Hi David,

On Fri, 7 Feb 2025 at 17:29, David Lechner <dlechner@xxxxxxxxxxxx> wrote:
> On 2/7/25 3:10 AM, Geert Uytterhoeven wrote:
> > On Thu, 6 Feb 2025 at 23:48, David Lechner <dlechner@xxxxxxxxxxxx> wrote:
> >> Add a new gpiod_multi_set_value_cansleep() helper function with fewer
> >> parameters than gpiod_set_array_value_cansleep().
> >>
> >> Calling gpiod_set_array_value_cansleep() can get quite verbose. In many
> >> cases, the first arguments all come from the same struct gpio_descs, so
> >> having a separate function where we can just pass that cuts down on the
> >> boilerplate.
> >>
> >> Reviewed-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
> >> Signed-off-by: David Lechner <dlechner@xxxxxxxxxxxx>
> >
> > Thanks for your patch!
> >
> >> --- a/include/linux/gpio/consumer.h
> >> +++ b/include/linux/gpio/consumer.h
> >> @@ -655,4 +655,11 @@ static inline void gpiod_unexport(struct gpio_desc *desc)
> >>
> >>  #endif /* CONFIG_GPIOLIB && CONFIG_GPIO_SYSFS */
> >>
> >> +static inline int gpiod_multi_set_value_cansleep(struct gpio_descs *descs,
> >> +                                                unsigned long *value_bitmap)
> >> +{
> >> +       return gpiod_set_array_value_cansleep(descs->ndescs, descs->desc,
> >> +                                             descs->info, value_bitmap);
> >
> > I am wondering whether this needs a check for !IS_ERR_OR_NULL(descs),
> > to handle the !CONFIG_GPIOLIB and gpiod_get_array_optional() cases?
>
> I don't think it is strictly needed, but could be convenient for future use
> cases. If we add it, should it be:
>
>         if (IS_ERR_OR_NULL(descs))
>                 return PTR_ERR(descs);

return PTR_ERR_OR_ZERO(descs);

(the compiler should optimize away checks common to IS_ERR_OR_NULL()
 and PTR_ERR_OR_ZERO()).

> or:
>
>         if (!descs)
>                 return -EINVAL;
>
>         if (IS_ERR(descs))
>                 return PTR_ERR(descs);
>
> ?

The former.

> For comparison, gpiod_set_array_value_cansleep() will return -EINVAL if the
> first argument is NULL.

That function cannot take an argument returned by a *_optional() call,
if I'm not mistaken.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds