Re: [PATCH v4 1/3] dt-bindings: iio: adc: sophgo,cv18xx-saradc.yaml: Add Sophgo CV18XX SARADC binding

[Conor Dooley <conor@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 4335 bytes --]

On Wed, Aug 21, 2024 at 09:41:50AM +0200, Miquel Raynal wrote:
> > > > > +      Represents the channels of the ADC.
> > > > > +
> > > > > +    properties:
> > > > > +      reg:
> > > > > +        description: |
> > > > > +          The channel number. It can have up to 3 channels numbered from 0 to 2.
> > > > > +        items:
> > > > > +          - minimum: 0
> > > > > +            maximum: 2  
> > > > 
> > > > Is this sufficient to limit the number of channels to 3? Aren't you relying
> > > > on the unique unit addresses warning in dtc to limit it, rather than
> > > > actually limiting with min/maxItems?
> > > >   
> > > It seems like I can't use min/maxItems on this property. I think that it is
> > > using size-cells + address-cells to deduce that the number of items should
> > > be equal to 1.  
> 
> Looking at dt-schema, I couldn't personally understand from where did
> the error messages reported by Thomas came from. There are clear

I think the complaints are on a more meta level than that. He provided
an items list
     properties:
       reg:
         maxItems: 1
         items:
           - minimum: 0
             maximum: 2
but this list only has one entry as there's one -. The first complaint
from dt_binding_check is that having maxItems is not needed with an
items list, because the items list contains the maximum number of
elements.

The second one comes from cell.yaml:
https://github.com/devicetree-org/dt-schema/blob/main/dtschema/meta-schemas/cell.yaml

It either allows a single item, with maxItems: 1 or multiple items, in
which case maxitems must be greater than 1. That's where the "anyOf
conditonal failed" and "1 is less than the minimum of 2" stuff comes
from.

I hope that helps?

> constraints over minItems/maxItems regarding the use of {#address-cells,
> #sizez-cells} being {1, 1}, {2, 2} and {2, 1} (in reg.yaml), but nothing
> explicit regarding the other situations, namely {1, 0} in this case
> which enforces maxItems to 1 is not clearly stated in any of the core
> yaml files. Any idea where to look at? Although, I'm convinced there is
> something defined because renaming the property from 'reg' to 'foo'
> silences these warnings.
> 
> > I think I was mistaken in talking about mix/max items here. I had the
> > right idea, but mentioned an incorrect solution - sorry about that. I
> > wasn't talking about the number of elements in the reg property, what I
> > meant was limiting the number of channel nodes in the first place -
> > something which min/maxItems cannot do. As examples of the problem I was
> > thinking of, see the below two examples:
> > 
> >     adc@30f0000 {
> >         compatible = "sophgo,cv1800b-saradc";
> >         reg = <0x030f0000 0x1000>;
> >         clocks = <&clk CLK_SARADC>;
> >         interrupts = <100 IRQ_TYPE_LEVEL_HIGH>;
> >         #address-cells = <1>;
> >         #size-cells = <0>;
> > 
> >         channel@0 {
> >             reg = <0>;
> >         };
> >         channel@2 {
> >             reg = <2>;
> >         };
> >         channel@22 {
> >             reg = <2>;
> >         };
> >     };
> > 
> >     adc@30f0000 {
> >         compatible = "sophgo,cv1800b-saradc";
> >         reg = <0x030f0000 0x1000>;
> >         clocks = <&clk CLK_SARADC>;
> >         interrupts = <100 IRQ_TYPE_LEVEL_HIGH>;
> >         #address-cells = <1>;
> >         #size-cells = <0>;
> > 
> >         channel@0 {
> >             reg = <0>;
> >         };
> >         channel@2 {
> >             reg = <2>;
> >         };
> >         channel@22 {
> >             reg = <2>;
> >         };

I noticed that I pasted two of the same example. I must have just
yoinked the latter to a vim buffer rather than to my clipboard. At least
it didn't matter in the end.

Cheers,
Conor.

> >     };
> > 
> > The solution is simple, remove the + from the regex. Sorry for sending
> > you on the wrong track Thomas.
> 
> Ah! Thanks Conor for the details, now it makes full sense :-) BTW Thomas
> the regex is
> 
> 	^channel@[0-3]+$
> 
> and I guess it should instead be
> 
> 	^channel@[0-2]$
>                     ^
> 
> in order to fully match the real indexing constraints you're enforcing
> with minimum/maximum.
> 
> Thanks,
> Miquèl

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