On Fri, Sep 24, 2021 at 10:18 AM Oskar Senft <osk@xxxxxxxxxx> wrote:
>
> Hi Rob
>
> > > +maintainers:
> > > + - Guenter Roeck <linux@xxxxxxxxxxxx>
> >
> > Should be someone that cares about this h/w, not who applies patches.
>
> Hmm, ok. After talking with Guenter, I thought that would be him. But
> I can add myself, too, since we're obviously using that HW. Is that
> what you mean?
Okay, seems it is Guenter in this case.
>
> > > + properties:
> > > + ltd:
> > > + type: object
> > > + description: Internal Temperature Sensor ("LTD")
> >
> > No child properties?
>
> Yes. We really just want the ability to enable / disable that sensor.
> What's the correct way in the YAML to describe that? Same for RTD3.
Okay, you need a 'additionalProperties: false' in the schema. (status
will automagically be allowed)
>
> > > + "type":
> > > + description: Sensor type (3=thermal diode, 4=thermistor).
> >
> > 2nd time I've seen this property this week[1]. Needs to be more specific
> > than just 'type'.
>
> Ha yes, the example in [1] came from this patch. I went with this name
> to stay in-line with the sysfs name, being "tempX_type". In the
> hardware this would be called "mode".
>
> My original proposal [2] was to have this property a string list named
> "nuvoton,rtd-modes" with a set of accepted values, i.e. basically an
> enum. Splitting this string list into individual sensors makes sense.
>
> The other question that remains open (at least in my view), is whether
> naming the sensors "ltd, rtd1, rtd2, rtd3" is the right approach or if
> we should really go to naming them "sensor@X" with a reg property set
> to X. Note that ltd and rtd3 do not accept any additional
> configuration beyond "is enabled" (i.e. "status").
If X is not made up numbering (i.e. corresponds to something in the
datasheet), then using addresses and generic node names are preferred.
Alignment with other similar h/w is also preferred.
> > > + temperature-sensors {
> > > + ltd {
> > > + status = "disabled";
> >
> > Don't show status in examples.
> Hmm, ok. I found it useful to make clear that a sensor can be
> disabled, but maybe that's just always the case?
Yeah, this case is a bit special. The node not being present also disables it.
The problem is generally we don't want disabled examples as that turns
off some validation. I have a check for this that I plan to add, but I
don't have a way to have exceptions.
What you could do is just comment out the node. Then you show it, but
don't compile it.
Rob
