Re: [PATCH 2/3] dt-bindings: iio: proximity: Add Nicera D3-323-AA PIR sensor

[Krzysztof Kozlowski <krzk@kernel.org>]

On 16/05/2025 19:15, Waqar Hameed wrote:
> On Fri, May 09, 2025 at 17:06 +0200 Krzysztof Kozlowski <krzk@kernel.org> wrote:
> 
>> On 09/05/2025 17:03, Waqar Hameed wrote:
>>> Nicera D3-323-AA is a PIR sensor for human detection. It has support for
>>> raw data measurements and detection notification. The communication
>>> protocol is custom made and therefore needs to be GPIO bit banged.
>>>
>>> Add devicetree bindings requiring the compatible string and the various
>>> GPIOs needed for operation. Some of the GPIOs have multiple use-cases
>>> depending on device state. Describe these thoroughly.
>>
>>
>> Drop redundant parts of description. Describe the hardware. 
> 
> I'll reformulate and incorporate some information of the pins and how it
> is used in the hardware.
> 
>> Entire last paragraph is pretty pointless.
> 
> I'll remove it then. (Some sub-system maintainers want a description of
> what the patch does, in imperative form. But I can see why it might not
> add any value when it comes to dt-bindings.)
> 
>>
>>> +
>>> +properties:
>>> +  compatible:
>>> +    const: nicera,d3323aa
>>> +
>>> +  vdd-gpio:
>>> +    maxItems: 1
>>> +    description:
>>> +      GPIO for supply voltage (1.8 to 5.5 V).
>>
>> This is not how pins are represented in the kernel. Either you have here
>> regulator (supply) or reset gpios. 
> 
> I'll change it to `vdd-supply`.
> 
>> Plus 'gpio' suffix is not valid, btw.
> 
> I actually `grep`ed before writing this to see if there were other
> dt-bindings with this suffix. Because the GPIO framework supports both
> `gpio` and `gpios` as suffixes (see `gpio_suffixes[]` in `gpiolib.c`).
> However, since `-gpios` are clearly in majority, we should go for that.


One is deprecated. It is always, always gpios.

> 
> [...]
> 
>>> +      This GPIO will be driven low by the driver in order to cut the supply and
>>> +      reset the device (driving it then back to high to power it on).
>>> +
>>> +  clk-vout-gpio:
>>
>> No, for the similar reasons. Which pin is this?
> 
> This pin is a little weird actually. As described below, right after
> power on, it is used as an interrupt to signal "ready for
> configuration". Then, it used as a bit banged clock signal for
> configuration. Finally, it is back as interrupt pin for threshold PIR
> sensor detections.
> 
> So, I'm not really sure what to call this (just opted for what it's
> called in the data sheet: "Vout (CLK)"). Just `clk-gpios` wouldn't be
> correct either, right? Should we prefix it with the vendor `nicera,`? Or
> any other suggestions?

Call it by the name of pin, so vout-clk-gpios. I guess from SoC/CPU side
this cannot be anything else than GPIO.

> 
>>
>>> +    maxItems: 1
>>> +    description:
>>> +      GPIO for clock and detection.
>>> +      After reset, the device signals with two falling edges on this pin that it
>>> +      is ready for configuration (within 1.2 s), which the driver listens for as
>>> +      interrupts.
>>> +      During configuration, it is used as clock for data reading and writing (on
>>> +      data-gpio). The driver drives this pin with the frequency of 1 kHz (bit
>>> +      banging).
>>> +      After all this, the device is now in operational mode and signals on this
>>> +      pin for any detections. The driver listens for this as interrupts.
>>> +
>>> +  data-gpio:
>>
>> There is no such pin.
> 
> You mean to change it to `data-gpios`? (There are some using that, e.g.
> `sensirion,sht15.yaml`).

No, I meant I opened datasheet and found no such pin. Unless you meant
DO, but then mention in description the actual name of the pin, if you
are using some more descriptive property name for it.

> 
>>
>>> +    maxItems: 1
>>> +    description:
>>> +      GPIO for data reading and writing.
>>> +      During configuration, configuration data will be written and read back
>>> +      with bit banging (together with clk-vout-gpio as clock).
>>> +      After this, during operational mode, the device will output serial data on
>>> +      this GPIO. However, the driver currently doesn't read this.

BTW, drop all references to driver here and in other places. If driver
change, you will change binding? If my driver behaves differently, why
binding would be claiming something else?


Best regards,
Krzysztof