On September 25, 2016 3:23 AM, Mark Brown wrote:
> On Wed, Sep 21, 2016 at 01:44:39PM +0900, Eric Jeong wrote:
>
> > +#ifdef CONFIG_OF
> > +static const struct of_device_id pv88080_dt_ids[] = {
> > + { .compatible = "pvs,pv88080-aa", .data = (void *)TYPE_PV88080_AA },
> > + { .compatible = "pvs,pv88080-ba", .data = (void *)TYPE_PV88080_BA },
> > + {},
> > +};
> > +MODULE_DEVICE_TABLE(of, pv88080_dt_ids); #endif
>
> This doesn't list the old compatible string and therefore ends up breaking
> since the code defaults to assuming something it hasn't heard of is BB when
> it seems more likely that old DTs using the old compatible string will be for
> boards that also use the old silicon. It would be much better practice to
> explicitly list the old compatible string and how we handle it, that won't
> break
> in future and avoids this kind of error.
>
> Is it really not possible to read the chip revision from the device using ID
> registers? That'd be even better.
The address of ID register is different between each chip revision.
So, It is difficult to read the chip revision from the device using ID register.
And, I will send a patch with old compatible.
