On Fri, 2022-01-28 at 13:04 +0000, Robin Murphy wrote:
> On 2022-01-28 08:11, Yong Wu wrote:
> [...]
> > diff --git a/include/linux/component.h b/include/linux/component.h
> > index 16de18f473d7..5a7468ea827c 100644
> > --- a/include/linux/component.h
> > +++ b/include/linux/component.h
> > @@ -2,6 +2,8 @@
> > #ifndef COMPONENT_H
> > #define COMPONENT_H
> >
> > +#include <linux/device.h>
> > +#include <linux/of.h>
> > #include <linux/stddef.h>
> >
> >
> > @@ -82,6 +84,22 @@ struct component_master_ops {
> > void (*unbind)(struct device *master);
> > };
> >
> > +/* A set common helpers for compare/release functions */
> > +static inline int compare_of(struct device *dev, void *data)
> > +{
> > + return dev->of_node == data;
> > +}
>
> Note that this is effectively just device_match_of_node(), although
> I
> guess there is an argument that having a nice consistent set of
> component_match API helpers might be worth more than a tiny code
> saving
> by borrowing one from a different API.
Thanks for this tip. I could try use this, Let the maintainer decide.
>
> Either way, however, I don't think there's any good argument for
> instantiating separate copies of these functions in every driver
> that
> uses them. If they're used as callbacks then they can't actually be
> inlined anyway, so they may as well be exported from component.c as
oh. The MTK iommu keeps this inline for long time... OK, I will try
to move into the component.c and export them.
(Sorry for the late reply due to New Year holidays.)
> normal so that the code really is shared (plus then there's nice
> symmetry with the aforementioned device_match API helpers too).
>
> Thanks,
> Robin.
