On Mon, Aug 29, 2022 at 06:00:22PM +0100, Jonathan Cameron wrote: > On Mon, 29 Aug 2022 14:24:04 +0300 > Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote: ... > > - yas5xx->chip_info = &yas5xx_chip_info_tbl[id->driver_data]; > > - ci = yas5xx->chip_info; > > + ci = device_get_match_data(dev); > > + yas5xx->chip_info = ci; > > Am I missing a path by which device_get_match_data() can use the i2c_device_id values? > I'd expect to see a fallback to that if ci == NULL to cover the non firmware causes > of probe(). I've seen a few patches without that fallback path recently so wondering > if some magic has gotten hooked up and I've missed it (something to push that via > a swnode perhaps?) No, there is no such magic behind. The idea is that we use DT / ACPI overlays when device needs to be established. But I have got your point. -- With Best Regards, Andy Shevchenko