[1/3] iio: adc: ad7923: fix channel readings for some variants

Some of the supported devices have 4 or 2 LSB trailing bits that should
not be taken into account. Hence we need to shift these bits out which
fits perfectly on the scan type shift property. This change fixes both
raw and buffered reads.

Fixes: f2f7a449707e ("iio:adc:ad7923: Add support for the ad7904/ad7914/ad7924")
Fixes: 851644a60d20 ("iio: adc: ad7923: Add support for the ad7908/ad7918/ad7928")
Signed-off-by: Nuno Sá <nuno.sa@analog.com>
---
 drivers/iio/adc/ad7923.c | 46 +++++++++++++++++++++-------------------
 1 file changed, 24 insertions(+), 22 deletions(-)

On Fri, 9 Sep 2022 17:14:11 +0200
Nuno Sá <nuno.sa@analog.com> wrote:

> Some of the supported devices have 4 or 2 LSB trailing bits that should
> not be taken into account. Hence we need to shift these bits out which
> fits perfectly on the scan type shift property. This change fixes both
> raw and buffered reads.

Hi Nuno,

Seems that all the values of shift are 12 - realbits.
If that's the case, can we reduce the noise this patch creates by just
updating AD7923_V_CHAN() to set .shift = 12 - (bits) ?

I guess that's not as flexible if anyone adds support for a device
with different shifts, but I suspect that may never happen.

Given we want a fix to be minimal (and hence as likely as possible to backport
cleanly) I think that approach would be a cleaner choice.

Thanks,

Jonathan
> 
> Fixes: f2f7a449707e ("iio:adc:ad7923: Add support for the ad7904/ad7914/ad7924")
> Fixes: 851644a60d20 ("iio: adc: ad7923: Add support for the ad7908/ad7918/ad7928")
> Signed-off-by: Nuno Sá <nuno.sa@analog.com>
> ---
>  drivers/iio/adc/ad7923.c | 46 +++++++++++++++++++++-------------------
>  1 file changed, 24 insertions(+), 22 deletions(-)
> 
> diff --git a/drivers/iio/adc/ad7923.c b/drivers/iio/adc/ad7923.c
> index edad1f30121d..910cf05e75cd 100644
> --- a/drivers/iio/adc/ad7923.c
> +++ b/drivers/iio/adc/ad7923.c
> @@ -80,7 +80,7 @@ enum ad7923_id {
>  	AD7928
>  };
>  
> -#define AD7923_V_CHAN(index, bits)					\
> +#define AD7923_V_CHAN(index, bits, _shift)				\
>  	{								\
>  		.type = IIO_VOLTAGE,					\
>  		.indexed = 1,						\
> @@ -93,38 +93,39 @@ enum ad7923_id {
>  			.sign = 'u',					\
>  			.realbits = (bits),				\
>  			.storagebits = 16,				\
> +			.shift = (_shift),				\
>  			.endianness = IIO_BE,				\
>  		},							\
>  	}
>  
> -#define DECLARE_AD7923_CHANNELS(name, bits) \
> +#define DECLARE_AD7923_CHANNELS(name, bits, shift) \
>  const struct iio_chan_spec name ## _channels[] = { \
> -	AD7923_V_CHAN(0, bits), \
> -	AD7923_V_CHAN(1, bits), \
> -	AD7923_V_CHAN(2, bits), \
> -	AD7923_V_CHAN(3, bits), \
> +	AD7923_V_CHAN(0, bits, shift), \
> +	AD7923_V_CHAN(1, bits, shift), \
> +	AD7923_V_CHAN(2, bits, shift), \
> +	AD7923_V_CHAN(3, bits, shift), \
>  	IIO_CHAN_SOFT_TIMESTAMP(4), \
>  }
>  
> -#define DECLARE_AD7908_CHANNELS(name, bits) \
> +#define DECLARE_AD7908_CHANNELS(name, bits, shift) \
>  const struct iio_chan_spec name ## _channels[] = { \
> -	AD7923_V_CHAN(0, bits), \
> -	AD7923_V_CHAN(1, bits), \
> -	AD7923_V_CHAN(2, bits), \
> -	AD7923_V_CHAN(3, bits), \
> -	AD7923_V_CHAN(4, bits), \
> -	AD7923_V_CHAN(5, bits), \
> -	AD7923_V_CHAN(6, bits), \
> -	AD7923_V_CHAN(7, bits), \
> +	AD7923_V_CHAN(0, bits, shift), \
> +	AD7923_V_CHAN(1, bits, shift), \
> +	AD7923_V_CHAN(2, bits, shift), \
> +	AD7923_V_CHAN(3, bits, shift), \
> +	AD7923_V_CHAN(4, bits, shift), \
> +	AD7923_V_CHAN(5, bits, shift), \
> +	AD7923_V_CHAN(6, bits, shift), \
> +	AD7923_V_CHAN(7, bits, shift), \
>  	IIO_CHAN_SOFT_TIMESTAMP(8), \
>  }
>  
> -static DECLARE_AD7923_CHANNELS(ad7904, 8);
> -static DECLARE_AD7923_CHANNELS(ad7914, 10);
> -static DECLARE_AD7923_CHANNELS(ad7924, 12);
> -static DECLARE_AD7908_CHANNELS(ad7908, 8);
> -static DECLARE_AD7908_CHANNELS(ad7918, 10);
> -static DECLARE_AD7908_CHANNELS(ad7928, 12);
> +static DECLARE_AD7923_CHANNELS(ad7904, 8, 4);
> +static DECLARE_AD7923_CHANNELS(ad7914, 10, 2);
> +static DECLARE_AD7923_CHANNELS(ad7924, 12, 0);
> +static DECLARE_AD7908_CHANNELS(ad7908, 8, 4);
> +static DECLARE_AD7908_CHANNELS(ad7918, 10, 2);
> +static DECLARE_AD7908_CHANNELS(ad7928, 12, 0);
>  
>  static const struct ad7923_chip_info ad7923_chip_info[] = {
>  	[AD7904] = {
> @@ -268,7 +269,8 @@ static int ad7923_read_raw(struct iio_dev *indio_dev,
>  			return ret;
>  
>  		if (chan->address == EXTRACT(ret, 12, 4))
> -			*val = EXTRACT(ret, 0, 12);
> +			*val = EXTRACT(ret, chan->scan_type.shift,
> +				       chan->scan_type.realbits);
>  		else
>  			return -EIO;
>

On Sun, 2022-09-11 at 12:26 +0100, Jonathan Cameron wrote:
> On Fri, 9 Sep 2022 17:14:11 +0200
> Nuno Sá <nuno.sa@analog.com> wrote:
> 
> > Some of the supported devices have 4 or 2 LSB trailing bits that
> > should
> > not be taken into account. Hence we need to shift these bits out
> > which
> > fits perfectly on the scan type shift property. This change fixes
> > both
> > raw and buffered reads.
> 
> Hi Nuno,

Hi Jonathan,

> 
> Seems that all the values of shift are 12 - realbits.
> If that's the case, can we reduce the noise this patch creates by
> just
> updating AD7923_V_CHAN() to set .shift = 12 - (bits) ?
> 

Yes, it should be pretty much the same... As I don't have any strong
feelings I can do as you suggest.

> I guess that's not as flexible if anyone adds support for a device
> with different shifts, but I suspect that may never happen.
> 

Or a device with realbits > 12. But yeah, I'm also fairly positive we
won't see that happening...

- Nuno Sá