Re: [PATCH v5 1/2] dt-bindings: iio: vadc: Update example to include unit address for node 'usb-id-nopull'

[Jonathan Cameron <jonathan.cameron@huawei.com>]

On Tue, 30 Oct 2018 11:00:16 -0500
Rob Herring <robh@kernel.org> wrote:

> On Sun, Oct 21, 2018 at 9:17 AM Jonathan Cameron <jic23@kernel.org> wrote:
> >
> > On Thu, 18 Oct 2018 12:40:10 -0700
> > Matthias Kaehlcke <mka@chromium.org> wrote:
> >  
> > > On Fri, Oct 12, 2018 at 10:15:23AM -0700, Matthias Kaehlcke wrote:  
> > > > On Fri, Oct 05, 2018 at 03:47:43PM -0500, Rob Herring wrote:  
> > > > > On Wed, Oct 03, 2018 at 05:14:31PM -0700, Matthias Kaehlcke wrote:  
> > > > > > The node has a reg property, therefore its name should include a unit
> > > > > > address.
> > > > > >
> > > > > > Also change the name from 'usb_id_nopull' to 'usb-id-nopull' to follow
> > > > > > DT conventions.  
> > > > >
> > > > > This is ADC channels? If so, then DT convention would really be
> > > > > "adc@...".  
> > > >
> > > > Is it really? A grep for 'adc@' in arch/${ARCH}/boot/dts yields
> > > > mostly ADC controller not channel nodes.
> > > >
> > > > I'm totally fine with changing the name to 'adc@...' if that's the
> > > > preference/convention, just want to reconfirm since the actual use is
> > > > a bit ambiguous.  
> > >
> > > Could we please reach a conclusion on this?
> > >
> > > Summarizing the options on the table so far are:
> > >
> > > 1. usb-id-nopull@VADC_LR_MUX10_USB_ID
> > > 2. usb-id-nopull@57
> > > 3. adc@VADC_LR_MUX10_USB_ID
> > > 4. adc@57
> > >
> > > My personal preference goes to something <node name>@<define>
> > > since the unit address doesn't just resolve to an ADC channel number
> > > but also includes configuation information. A literal like '57'
> > > conveys less information than the define, it's easier to introduce
> > > errors and these errors are harder to spot.  
> >
> > I agree that to my mind this is the most sensible option.  
> 
> If you really want the defines, then fine. Of course, that only works
> if the function is fixed. It won't work if the function is defined per
> board.
> 
> Eventually, examples using defines will have to also include the
> headers. I plan to make the examples build-able.
> 
> > > If 'adc@...' really was the convention (or should be) I'd be clearly
> > > in favor of following it. As mentioned above, in practice the use of
> > > the 'adc@...' node name seems to be more prevalent for ADC controllers
> > > than channels, so I'm more inclined towards 'usb-id-nopull@...' or
> > > similar.
> > >
> > > All that said, these are just my preferences for the reasons outlined
> > > above, if DT maintainers really want it to be 'adc@57' or some
> > > variation of that, I'm fine with that too. Please let me know and we
> > > can move forward with this trivial series.  
> >
> > Rob, what's your view on this?  
> 
> I want node names that reflect the class of the node (not a specific
> model) and consistency across bindings. What that looks like for
> multi-channel ADCs is really up to you. There was another binding
> recently which mapped sub-nodes to inputs rather than channels. Maybe
> that's needed too if you have more inputs than simultaneous channels.
>From a DT point of view the simultaneous channels vs inputs isn't meant
to be currently visible. It's a userspace problem and from IIO point
of view a 'channel' is an input (you just might not be able to enable
all channels at once!)

channel@ would work for me, though that might then match other types
of device where this has a different meaning.  Perhaps
adc_chan@ ?

> 
> Also, if your goal is to just quiet dtc warnings, then I'd prefer you
> not. They often point to bigger issues even if they can be fixed with
> trivial changes. Of course, if not fixed someone else will come along
> and try the trivial fix again.

Agreed.  Would be good to have these consistent however as it is a fairly
common thing to represent.

Jonathan

> 
> Rob