Re: [PATCH v5 3/3] iio: magnetometer: add ti tmag5273 driver

[Gerald Loacker <gerald.loacker@wolfvision.net>]

Am 30.11.2022 um 16:31 schrieb Andy Shevchenko:
> On Wed, Nov 30, 2022 at 03:53:56PM +0100, Gerald Loacker wrote:
>> Add support for TI TMAG5273 Low-Power Linear 3D Hall-Effect Sensor.
>> Additionally to temperature and magnetic X, Y and Z-axes the angle and
>> magnitude are reported.
>> The sensor is operating in continuous measurement mode and changes to sleep
>> mode if not used for 5 seconds.
> 
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> 
> But a couple of comments to address.
> 
>> +static int tmag5273_check_device_id(struct tmag5273_data *data)
>> +{
>> +	__le16 devid;
>> +	int val, ret;
>> +
>> +	ret = regmap_read(data->map, TMAG5273_DEVICE_ID, &val);
>> +	if (ret)
>> +		return dev_err_probe(data->dev, ret, "failed to power on device\n");
>> +	data->version = FIELD_PREP(TMAG5273_VERSION_MASK, val);
>> +
>> +	ret = regmap_bulk_read(data->map, TMAG5273_MANUFACTURER_ID_LSB, &devid,
>> +			       sizeof(devid));
>> +	if (ret)
>> +		return dev_err_probe(data->dev, ret, "failed to read device ID\n");
>> +	data->devid = le16_to_cpu(devid);
>> +
>> +	switch (data->devid) {
>> +	case TMAG5273_MANUFACTURER_ID:
>> +		/*
>> +		 * The device name matches the orderable part number. 'x' stands
>> +		 * for A, B, C or D devices, which have different I2C addresses.
>> +		 * Versions 1 or 2 (0 and 3 is reserved) stands for different
>> +		 * magnetic strengths.
>> +		 */
>> +		snprintf(data->name, sizeof(data->name), "tmag5273x%1u", data->version);
>> +		if (data->version < 1 || data->version > 2)
>> +			dev_warn(data->dev, "Unsupported device %s\n", data->name);
>> +		return 0;
>> +	default:
> 
>> +		dev_warn(data->dev, "Unknown device ID 0x%x\n", data->devid);
>> +		return 0;
> 
> And we still continue?! Wouldn't be a problem if that ID drastically changed in
> terms of programming model and may actually be broken by a wrong sequence?
> 

It was suggested by Jonathan to just print a warning instead of
returning with -ENODEV. Reason was "Often manufacturers spin new
versions of chips that are compatible enough that we give them a
fallback compatible in device tree.". For me this makes sense.

>> +	}
>> +}
>> +

Regards,
Gerald