Re: [PATCH v10 4/5] rtc: mt6397: Add support for the MediaTek MT6358 RTC

[Alexandre Belloni]

On 13/03/2020 07:22:30+0000, Lee Jones wrote:
> > > >  struct mt6397_rtc {
> > > >  	struct device           *dev;
> > > >  	struct rtc_device       *rtc_dev;
> > > > @@ -74,6 +80,7 @@ struct mt6397_rtc {
> > > >  	struct regmap           *regmap;
> > > >  	int                     irq;
> > > >  	u32                     addr_base;
> > > > +	const struct mtk_rtc_data *data;
> > > 
> > > 'data' is a terrible variable name.
> > > 
> > > Why do you need to store this?
> > > 
> > > It's one variable which is used once AFAICT.
> > 
> > I would rename 'data' to 'config'.
> > 
> > This struct will be extended in future patches to achieve more PMIC chip
> > compatibility.
> 
> On closer inspection, it looks like wrtgr (also not a great name for a
> variable by the way) is a register address.  Is that correct?
> Initially I thought it was a model number, which would have been a
> suitable candidate for entry into OF .data.
> 
> However, describing register addresses in OF .data does not sound like
> good practice.  It is usually used to identify a platform in the cases
> where platforms cannot be otherwise dynamically interrogated for model
> number via a register read.
> 
> Describing register maps via 'config' data is a slippery slope.

I'm not sure I get what you mean, there are dozens if not hundreds of
drivers doing it exactly that way. What is the difference between having
.data pointing to a register map and having .data containing a model
number and then use that model number to get the register map?

of_device_id::data definitively isn't config data as the DT describes
the hardware, not the configuration.

-- 
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com