Re: [PATCH 1/2] dt-bindings: iio: adc: ad4080: add AD4880 support

[David Lechner <dlechner@baylibre.com>]

On 2/3/26 9:42 AM, Miclaus, Antoniu wrote:
> ...

...

>>> @@ -98,4 +127,22 @@ examples:
>>>            io-backends = <&iio_backend>;
>>>          };
>>>      };
>>> +  - |
>>> +    spi {
>>> +        #address-cells = <1>;
>>> +        #size-cells = <0>;
>>> +
>>> +        adc@0 {
>>> +          compatible = "adi,ad4880";
>>> +          reg = <0>;
>>> +          adi,aux-spi-cs = <1>;
>>
>> We can already have multiple reg for a multiple CS device.
>>
>>              reg = <0>, <1>;
>>
>> So we shouldn't need a new adi,aux-spi-cs property for that.
> 
> Sure, that was the initial plan but for some reason going for multiple regs gets me to this:
> 
> [   11.396833]  spi_new_ancillary_device from ad4080_probe+0x4b8/0x5a4
> [   11.396861]  ad4080_probe from really_probe+0xc8/0x2c8
> [   11.397281] ad4080 spi1.0: chipselect 1 already in use
> [   11.397297] ad4080 spi1.0: failed to register ancillary device
> [   11.397317] ad4080 spi1.0: probe with driver ad4080 failed with error -16
> 
> While the aux-spi-cs works fine. Any tips?

The one existing user of spi_new_ancillary_device() used the multiple
regs like I suggested. So I wonder if that is broken too. The changes
for multiple CS for SPI parallel memories in the SPI core was made 2
years after that driver was added. So I could see how it could have
been broken without anyone noticing.

I didn't look at it too deeply yet. But my guess is that we need to
modify __spi_add_device() to change how it verifies the chip select.

If it is being called from spi_add_device(), don't change the behavior.
If it is being called from spi_new_ancillary_device(), then the
chip select should be verified as one assigned to the main device
rather than verifying that it is unused. I.e. don't call spi_dev_check_cs().