Re: [PATCH] mfd: syscon: Add syscon_regmap_lookup_by_phandle_optional() function.

[Arnd Bergmann]

I noticed that my earlier reply was rejected by some mail servers,
resending from @kernel.org to make sure everyone has it.

On Tue, Oct 27, 2020 at 10:57 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
>
> On Tue, Oct 27, 2020 at 10:11 PM Enric Balletbo i Serra
> <enric.balletbo@xxxxxxxxxxxxx> wrote:
> >
> > This adds syscon_regmap_lookup_by_phandle_optional() function to get an
> > optional regmap.
> >
> > It behaves the same as syscon_regmap_lookup_by_phandle() except where
> > there is no regmap phandle. In this case, instead of returning -ENODEV,
> > the function returns NULL. This makes error checking simpler when the
> > regmap phandle is optional.
> >
> > Suggested-by: Nicolas Boichat <drinkcat@xxxxxxxxxxxx>
> > Signed-off-by: Enric Balletbo i Serra <enric.balletbo@xxxxxxxxxxxxx>
>
> Looks good in principle.
>
> > @@ -255,6 +255,19 @@ struct regmap *syscon_regmap_lookup_by_phandle_args(struct device_node *np,
> >  }
> >  EXPORT_SYMBOL_GPL(syscon_regmap_lookup_by_phandle_args);
> >
> > +struct regmap *syscon_regmap_lookup_by_phandle_optional(struct device_node *np,
> > +                                       const char *property)
> > +{
> > +       struct regmap *regmap;
> > +
> > +       regmap = syscon_regmap_lookup_by_phandle(np, property);
> > +       if (IS_ERR(regmap) && PTR_ERR(regmap) == -ENODEV)
> > +               return NULL;
> > +
> > +       return regmap;
> > +}
>
> I think the explanation from the patch description is needed here as well,
> as an interface that might either return an error code or NULL is generally
> a really bad idea. I understand what this is for, but it's easy to misread.
>
> >  static inline struct regmap *device_node_to_regmap(struct device_node *np)
> >  {
> > @@ -59,6 +62,14 @@ static inline struct regmap *syscon_regmap_lookup_by_phandle_args(
> >  {
> >         return ERR_PTR(-ENOTSUPP);
> >  }
> > +
> > +static inline struct regmap *syscon_regmap_lookup_by_phandle_optional(
> > +                                       struct device_node *np,
> > +                                       const char *property)
> > +{
> > +       return ERR_PTR(-ENOTSUPP);
> > +}
>
> Shouldn't this also return NULL then?
>
>           Arnd