On Thu, Aug 29, 2019 at 01:08:35PM +0100, Andrew Murray wrote:
> On Thu, Aug 29, 2019 at 01:46:03PM +0200, Thierry Reding wrote:

> > If regulator_get_optional() returned NULL for absent optional supplies,
> > this could be unified across all drivers. And it would allow treating
> > NULL regulators special, if that's something you'd be willing to do.

> > In either case, the number of abuses shows that people clearly don't
> > understand how to use this. So there are two options: a) fix abuse every
> > time we come across it or b) try to change the API to make it more
> > difficult to abuse.

> Sure. I think we end up with something like:

> diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
> index e0c0cf462004..67e2a6d7abf6 100644
> --- a/drivers/regulator/core.c
> +++ b/drivers/regulator/core.c
> @@ -1868,6 +1868,9 @@ struct regulator *_regulator_get(struct device *dev, const char *id,
>                 }
>  
>                 switch (get_type) {
> +               case OPTIONAL_GET:
> +                       return NULL;
> +

Implementing returning NULL is not hard.  How returning NULL discourages
people from using regulator_get_optional() when they shouldn't be using
it in the first place is not clear to me.