Re: [PATCH v2 02/10] dt-bindings: iio: adc: ad7606: Make corrections on spi conditions

[Jonathan Cameron <jic23@kernel.org>]

On Wed, 25 Sep 2024 17:28:30 +0200
Guillaume Stols <gstols@baylibre.com> wrote:

> On 9/24/24 16:59, Conor Dooley wrote:
> > On Tue, Sep 24, 2024 at 04:41:50PM +0200, Guillaume Stols wrote:  
> >> On 9/21/24 23:55, Conor Dooley wrote:  
> >>> On Fri, Sep 20, 2024 at 05:33:22PM +0000, Guillaume Stols wrote:  
> >>>> The SPI conditions are not always required, because there is also a
> >>>> parallel interface. The way used to detect that the SPI interface is
> >>>> used is to check if the reg value is between 0 and 256.  
> >>> And, yaknow, not that the bus you're on is a spi bus? I don't think this
> >>> comment is relevant to the binding, especially given you have a property
> >>> for it.  
> >> Apologies, I missed to change the commit message, it will be fixed in the
> >> next series.
> >>
> >> Since Jonathan did not like very much inferring the interface with the reg's
> >> value that I used i the previous verison, I introduced this flag.
> >>
> >> However this is only intended to be use in bindings, to determine whether or
> >> not spi properties should be added.  
> > To be honest, if it is not needed by software to understand what bus the
> > device is on, it shouldn't be in the bindings at all. What was Jonathan
> > opposed to? Doing an if reg < 1000: do y, otherwise do x?
> > I'd not bother with any of that, and just make cpha (or w/e it was)
> > optional with a description explaining the circumstances in which is it
> > needed.  
> OK, it will be removed from the series and sent as a side patch because 
> it anyways does not really belong to this series.
I can't remember the original thread (or immediately find it).
So I may have this totally wrong. 
- I don't want checks on reg value to change how the binding works as that
  is a wieird corner and in theory this device could be at a small address anyway.

- Fine to do as Conor suggests and just add a comment for this
  corner case rather than making it required.

Jonathan
> >> In the driver side of things, the bus interface is inferred by the parent's
> >> node (SPI driver is an module_spi_driver while parallel driver is
> >> module_platform_driver).