Re: [Linux-stm32] [PATCH v2 1/4] ASoC: dt-bindings: stm32: i2s: add audio-graph-card port

[Lee Jones <lee.jones@linaro.org>]

On Wed, 08 Dec 2021, Alexandre TORGUE wrote:

> Hi Ahmad
> 
> On 12/7/21 2:59 PM, Ahmad Fatoum wrote:
> > Hello Alex,
> > 
> > On 07.12.21 14:52, Alexandre TORGUE wrote:
> > > Hi Rob
> > > 
> > > On 12/1/21 11:34 PM, Rob Herring wrote:
> > > > On Fri, Nov 26, 2021 at 11:25:27AM +0100, Olivier MOYSAN wrote:
> > > > > Hi Rob,
> > > > > 
> > > > > On 11/25/21 10:26 PM, Rob Herring wrote:
> > > > > > On Thu, 25 Nov 2021 15:40:50 +0100, Olivier Moysan wrote:
> > > > > > > The STM2 I2S DAI can be connected via the audio-graph-card.
> > > > > > > Add port entry into the bindings.
> > > > > > > 
> > > > > > > Signed-off-by: Olivier Moysan <olivier.moysan@foss.st.com>
> > > > > > > ---
> > > > > > >     Documentation/devicetree/bindings/sound/st,stm32-i2s.yaml | 5 +++++
> > > > > > >     1 file changed, 5 insertions(+)
> > > > > > > 
> > > > > > 
> > > > > > Running 'make dtbs_check' with the schema in this patch gives the
> > > > > > following warnings. Consider if they are expected or the schema is
> > > > > > incorrect. These may not be new warnings.
> > > > > > 
> > > > > > Note that it is not yet a requirement to have 0 warnings for dtbs_check.
> > > > > > This will change in the future.
> > > > > > 
> > > > > > Full log is available here: https://patchwork.ozlabs.org/patch/1559750
> > > > > > 
> > > > > > 
> > > > > > audio-controller@4000b000: 'port' does not match any of the regexes: '^port@[0-9]', 'pinctrl-[0-9]+'
> > > > > >      arch/arm/boot/dts/stm32mp157a-dk1.dt.yaml
> > > > > >      arch/arm/boot/dts/stm32mp157c-dk2.dt.yaml
> > > > > > 
> > > > > 
> > > > > This warning is not a new one.
> > > > > 
> > > > > The i2s2 node in stm32mp15xx-dkx.dtsi would require the following binding:
> > > > > port:
> > > > >      $ref: audio-graph-port.yaml#
> > > > >      unevaluatedProperties: false
> > > > > 
> > > > > However the spi binding requires to introduce a unit address:
> > > > > patternProperties:
> > > > >     '^port@[0-9]':
> > > > >       $ref: audio-graph-port.yaml#
> > > > >       unevaluatedProperties: false
> > > > > 
> > > > > The warning can be removed by re-ordering the bindings patches in the serie,
> > > > > as "additionalProperties: true" makes the check more tolerant on extra
> > > > > properties.
> > > > 
> > > > That's never right.
> > > > 
> > > > > The patch "ASoC: dt-bindings: stm32: i2s: add audio-graph-card port" can
> > > > > even be merely dropped.
> > > > > So, I suggest to resend the serie without audio-graph-card patch.
> > > > 
> > > > Only if you aren't using audio-graph-card.
> > > > 
> > > > > 
> > > > > Does it sound too permissive to you ?
> > > > 
> > > > I think perhaps you need to combine the schemas into 1. Or you need to
> > > > restructure your dtsi files such that you only add spi specific
> > > > properties when spi mode is enabled and only add i2s specific properties
> > > > when i2s mode is enabled. Or use the /delete-property/ directive.
> > > 
> > > Initially the aim of this series was to fix a "make W=1" warnings seen on spi and i2s nodes (duplicate unit-address). Moving both nodes in a common node + using a different compatible depending on SPI or I2S usage sounded good) but it is not enough. In this series the common node is named as following: "spi2s2: spi@4000b000". It is fine for a spi usage but if we want to use this "common node" with I2S compatible and specific bindings, the node name remains spi@... and then specific spi checks are done. For this with this series applied we got this issue reported by spi-controller.yaml:
> > > 
> > > spi@4000b000: port@0: 'compatible' is a required property
> > > 
> > > So, if we use two separates nodes we got W=1 warning and if we use a common node we got yaml check issue. One possibility would be to use a common node with a new node name (for example i2spi@...) but I think it is not acceptable.
> > > 
> > > How to progress ?
> > 
> > Atmel Flexcom can be configured to be either UART, SPI or i2c. Functions
> > are child nodes of the flexcom node and the MFD driver matching against it,
> > just configure the operating mode and then calls of_platform_populate.
> > 
> > Would something along these lines fit here as well?
> 
> Yes it could but in my mind it was not a MFD as both feature cannot be used
> at the same time: it is either SPI or I2S and choice is done "statically" in
> device tree depending board usage.
> 
> Lee, what it is your feeling about that ? Will you accept to add a MFD
> driver for this SPI/I2S peripheral whose prurpose is only to populate child
> node (either SPI or I2S) ?

From your description, this doesn't sound like a good fit for MFD.

-- 
Lee Jones [李琼斯]
Senior Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs
Follow Linaro: Facebook | Twitter | Blog