Re: [PATCH v3 2/7] mfd: retu: Add OF device ID table

[Aaro Koskinen]

Hi,

On Mon, Apr 03, 2017 at 06:24:39PM -0400, Javier Martinez Canillas wrote:
> On 04/03/2017 06:20 PM, Rob Herring wrote:
> > On Mon, Apr 03, 2017 at 11:45:14AM -0400, Javier Martinez Canillas wrote:
> >> Hello Lee,
> >>
> >> On 04/03/2017 07:15 AM, Lee Jones wrote:
> >>
> >> [snip]
> >>
> >>>>  
> >>>> +static const struct of_device_id retu_of_match[] = {
> >>>> +	{ .compatible = "nokia,retu-mfd" },
> >>>> +	{ .compatible = "nokia,tahvo-mfd" },
> >>>
> >>> Please drop the "-mfd".
> >>>
> >>
> >> Yes, I also didn't like it but I didn't want to change it since that would
> >> mean that backward compatiblity and bisect-ability will be broken by this
> >> change.
> >>
> >> In other words, just adding a vendor prefix won't cause an issue if patches
> >> are merged independently since if DTS patches are merged before, the driver
> >> will still lookup using the I2C device ID table. And if the drivers patches
> >> are picked before, the DTS will match using the OF device ID table.
> >>
> >> But changing to "nokia,retu" and "nokia,tahvo" means that you will need to
> >> pick all patches and also that the DTS and drivers changes will have to be
> >> done in the same patch. If you are OK with that, then I can change in the
> >> next version.
> > 
> > tahvo is not documented nor used in any dts (in the kernel at least). 

True, there are no known DT users of Tahvo.

> > retu is used by 1 board and happened to work, but was never documented. 
> > So I think it is okay to change unless the N800 folks object.
> 
> I'm fine with changing it (in fact I just want to fix the I2C of modalias
> reporting). Does this mean that backward compatibility and bisect-ability
> should be preserved? Or it's OK to split the changes in different patches?

There are 2 boards actually, N800 and N810. Retu is critical, because
if retu-mfd/watchdog fails to probe the device will power off soon after
boot. So for bisect-ability you should make changes in a single patch.

A.