Re: [PATCH 1/2] dt-bindings: iio: ti,tmp117: add binding for the TMP119

[Conor Dooley <conor@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 2426 bytes --]

On Fri, Jun 05, 2026 at 03:20:05PM +0100, Jonathan Cameron wrote:
> On Fri, 05 Jun 2026 14:24:45 +0200
> Romain Gantois <romain.gantois@bootlin.com> wrote:
> 
> > Hi Jonathan,
> > 
> > On Friday, 5 June 2026 13:53:32 CEST Jonathan Cameron wrote:
> > > On Fri, 05 Jun 2026 11:01:35 +0200
> > > 
> > > Romain Gantois <romain.gantois@bootlin.com> wrote:  
> > > > The TMP119 has the same register layout as the TMP117, and a better local
> > > > sensor accuracy.  
> > > 
> > > Sounds like (and the driver changes pretty much confirm) that this could
> > > use a fallback compatible to the tmp117.
> > >   
> > 
> > What do you mean by "fallback compatible" in this case? Do you mean we should 
> > do something like this in the bindings:
> > 
> > ```
> > enum:
> >        - ti,tmp116
> >        - ti,tmp119
> > const:
> >        - ti,tmp117
> > ```
> Something like (I can never remember the syntax so this is modified from
> another binding doc and untested)
> 
>     oneOf:
>       - enum:
>           - ti,tmp116
>           - ti,tmp117
>       - items:
>           - enum:  #enum here because once we have one we tend to get more ;)
>               - ti,tmp119
>           - const: ti,tmp117

Other than the cheeky comment, this looks okay.

> 
> > 
> > But wouldn't this invalidate existing device trees which only have:
> > 
> > `compatible = "ti,tmp116";`
> > 
> > > The second patch then just resolves the print that would happen on older
> > > kernels using that fallback due to ID mismatch.  That's fine but nice
> > > if older kernels can work with this part.  
> > 
> > Sorry I don't quite understand: since this is a feature patch, I don't expect 
> > it would get backported onto older kernels. Therefore why would the second 
> > patch resolve the print happening on older kernels?
> > 
> The point of a fallback compatible is that we don't have to backport anything.
> New device tree comes along that lists
> 
> compatible = "ti,tmp119", "ti,tmp117"
> 
> will just work. Older kernel will think it has a tmp117 which is fine as they
> are register interface compatible.  It will print a message to say that it
> found an unknown ID but then assume it is what the firmware told it, here the
> tmp117 fallback.  With your second patch in place, that harmless message will
> disappear for new kernels.
> 
> Jonathan
> 
> 
> > Thanks,
> > 
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]