On Wed, Jul 26, 2023 at 10:56:41AM -0700, Dmitry Torokhov wrote:
> On Wed, Jul 26, 2023 at 07:44:17PM +0300, Andy Shevchenko wrote:
> > On Wed, Jul 26, 2023 at 02:08:04PM +0100, Biju Das wrote:
...
> > > +static const void *i2c_device_get_match_data(const struct device *dev)
> > > +{
> > > + const struct i2c_client *client = (dev->type == &i2c_client_type) ?
> > > + to_i2c_client(dev) : NULL;
> >
> > There is an API i2c_verify_client() or something like this, I don't remember
> > by heart.
>
> It's been discussed in a separate thread. i2c_verify_client() needs a
> non-const pointer. It would be nice to clean up i2c_verify_client() to
> accept both variants, but that can be done later.
Then this code needs a TODO comment:
/* TODO: use i2c_verify_client() when it accepts const pointer */
> > > + if (!dev->driver)
> > > + return NULL;
> > > +
> > > + return i2c_get_match_data_helper(to_i2c_driver(dev->driver), client);
> > > +}
...
> > Side question, what is the idea for i2c_of_match_device()? Shouldn't you also
> > take it into consideration?
>
> Good call. I think we need to add something like
>
> if (!data && driver->driver.of_match_table) {
> match =
> i2c_of_match_device_sysfs(driver->driver.of_match_table, client);
> if (match)
> data = match->data;
> }
>
> to i2c_device_get_match_data().
Haven't checked myself, by I trust your suggestion. Let's see it in v3 then.
--
With Best Regards,
Andy Shevchenko
