Re: [PATCH v2] dt-bindings: i2c: Move i2c-omap.txt to YAML format

[Rob Herring]

On Fri, May 07, 2021 at 08:24:59PM +0300, Grygorii Strashko wrote:
> 
> 
> On 07/05/2021 17:36, Andreas Kemnade wrote:
> > On Fri, 7 May 2021 19:45:45 +0530
> > Vignesh Raghavendra <vigneshr@xxxxxx> wrote:
> > 
> > > On 5/7/21 12:24 PM, Grygorii Strashko wrote:
> > > > 
> > > > 
> > > > On 06/05/2021 17:00, Vignesh Raghavendra wrote:
> > > > > Convert i2c-omap.txt to YAML schema for better checks and documentation.
> > > > > 
> > > > > Following properties were used in DT but were not documented in txt
> > > > > bindings and has been included in YAML schema:
> > > > > 1. Include ti,am4372-i2c compatible
> > > > > 2. Include dmas property used in few OMAP dts files
> > > > 
> > > > The DMA is not supported by i2c-omap driver, so wouldn't be better to
> > > > just drop dmas from DTBs to avoid confusions?
> > > > It can be added later.
> > > 
> > > Will do.. I will also send patches dropping dmas from dts that currently
> > > have them populated.
> > > 
> > hmm, we have
> > - DO attempt to make bindings complete even if a driver doesn't support some
> >    features. For example, if a device has an interrupt, then include the
> >    'interrupts' property even if the driver is only polled mode.
> > 
> > in Documentation/devicetree/bindings/writing-bindings.rst
> > Shouln't the dma stay there if the hardware supports it? Devicetree
> > should describe the hardware not the driver if I understood things
> > right.
> 
> True.  But my above statement is also valid - it introduces confusion from user point of view.

In my OS, 'robOS', the driver supports DMA.

> More over, 'dmas' is not part of original binding and were randomly added to some SoCs.
> And it's much more easy to extend binding (in the future) then remove something after.

In this case, probably given that how it would be extended is already 
known, but it depends how you extend a binding. My above statement was 
born out of incomplete MFD and system controller bindings for the most 
part.

> I leave it to Vignesh, Tony to decide.

Fine with me.

Actually, for DMA with I2C I'd like to see someone show a usecase 
and data where it's actually beneficial. 

Rob