[PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Giorgi Tchankvetadze]

The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
path treats it as MICRO. Since no write_raw_get_fmt is provided, the
IIO core defaults to MICRO when parsing userspace input.

This means reading calibscale and writing it back results in a ~1000x
gain error.

Change the read path and available range to use MICRO to match the
write path.

Fixes: 0cb8b324852f ("iio: adc: ad4030: add driver for ad4030-24")
Signed-off-by: Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com>
---
 drivers/iio/adc/ad4030.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/iio/adc/ad4030.c b/drivers/iio/adc/ad4030.c
index def3e1d01ceb..3d823648371d 100644
--- a/drivers/iio/adc/ad4030.c
+++ b/drivers/iio/adc/ad4030.c
@@ -423,10 +423,10 @@ static int ad4030_get_chan_calibscale(struct iio_dev *indio_dev,
 
 	/* From datasheet: multiplied output = input × gain word/0x8000 */
 	*val = gain / AD4030_GAIN_MIDLE_POINT;
-	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, NANO,
+	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, MICRO,
 				AD4030_GAIN_MIDLE_POINT);
 
-	return IIO_VAL_INT_PLUS_NANO;
+	return IIO_VAL_INT_PLUS_MICRO;
 }
 
 /* Returns the offset where 1 LSB = (VREF/2^precision_bits - 1)/gain */
@@ -720,8 +720,8 @@ static irqreturn_t ad4030_trigger_handler(int irq, void *p)
 
 static const int ad4030_gain_avail[3][2] = {
 	{ 0, 0 },
-	{ 0, 30518 },
-	{ 1, 999969482 },
+	{ 0, 30 },
+	{ 1, 999969 },
 };
 
 static int ad4030_read_avail(struct iio_dev *indio_dev,
@@ -739,7 +739,7 @@ static int ad4030_read_avail(struct iio_dev *indio_dev,
 
 	case IIO_CHAN_INFO_CALIBSCALE:
 		*vals = (void *)ad4030_gain_avail;
-		*type = IIO_VAL_INT_PLUS_NANO;
+		*type = IIO_VAL_INT_PLUS_MICRO;
 		return IIO_AVAIL_RANGE;
 
 	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
-- 
2.52.0

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Marcelo Schmitt]

On 02/25, Giorgi Tchankvetadze wrote:
> The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
> path treats it as MICRO. Since no write_raw_get_fmt is provided, the
> IIO core defaults to MICRO when parsing userspace input.
> 
> This means reading calibscale and writing it back results in a ~1000x
> gain error.
> 
> Change the read path and available range to use MICRO to match the
> write path.

The updates to ad4030 driver will add write_raw_get_fmt() [1].
[1]: https://lore.kernel.org/linux-iio/516cccc47e917bd26be29b016907f50a244a68b9.1771865684.git.marcelo.schmitt@analog.com/

At first glance, I think this could instead add a case to write_raw_get_fmt(),
keeping calibscale nano precision (unless nano precision doesn't make sense
for ad4030 calibscale (don't recall from top of my mind)).

> 
> Fixes: 0cb8b324852f ("iio: adc: ad4030: add driver for ad4030-24")
> Signed-off-by: Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com>
> ---
>  drivers/iio/adc/ad4030.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/iio/adc/ad4030.c b/drivers/iio/adc/ad4030.c
> index def3e1d01ceb..3d823648371d 100644
> --- a/drivers/iio/adc/ad4030.c
> +++ b/drivers/iio/adc/ad4030.c
> @@ -423,10 +423,10 @@ static int ad4030_get_chan_calibscale(struct iio_dev *indio_dev,
>  
>  	/* From datasheet: multiplied output = input × gain word/0x8000 */
>  	*val = gain / AD4030_GAIN_MIDLE_POINT;
> -	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, NANO,
> +	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, MICRO,
>  				AD4030_GAIN_MIDLE_POINT);
>  
> -	return IIO_VAL_INT_PLUS_NANO;
> +	return IIO_VAL_INT_PLUS_MICRO;
>  }
>  
>  /* Returns the offset where 1 LSB = (VREF/2^precision_bits - 1)/gain */
> @@ -720,8 +720,8 @@ static irqreturn_t ad4030_trigger_handler(int irq, void *p)
>  
>  static const int ad4030_gain_avail[3][2] = {
>  	{ 0, 0 },
> -	{ 0, 30518 },
> -	{ 1, 999969482 },
> +	{ 0, 30 },
> +	{ 1, 999969 },
>  };
>  
>  static int ad4030_read_avail(struct iio_dev *indio_dev,
> @@ -739,7 +739,7 @@ static int ad4030_read_avail(struct iio_dev *indio_dev,
>  
>  	case IIO_CHAN_INFO_CALIBSCALE:
>  		*vals = (void *)ad4030_gain_avail;
> -		*type = IIO_VAL_INT_PLUS_NANO;
> +		*type = IIO_VAL_INT_PLUS_MICRO;
>  		return IIO_AVAIL_RANGE;
>  
>  	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
> -- 
> 2.52.0
> 
> 

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Giorgi Tchankvetadze]

Hi Marcelo, Thanks for the pointer to your pending series. As for
keeping NANO precision, the hardware register is 16-bit with a step of
1/0x8000 ≈ 30 µ, so NANO doesn't provide any additional precision that
the hardware can actually represent. MICRO is sufficient.

Btw, I checked and found that no driver in the entire IIO tree uses
NANO for calibscale.
Drivers use either IIO_VAL_INT (raw register value) or IIO_VAL_INT_PLUS_MICRO.

That said, if your series is already adding write_raw_get_fmt(), I'm
happy to drop this patch and let yours handle it - as long as the
mismatch gets fixed either way.

Regards, Giorgi

On Wed, Feb 25, 2026 at 3:53 PM Marcelo Schmitt
<marcelo.schmitt1@gmail.com> wrote:
>
> On 02/25, Giorgi Tchankvetadze wrote:
> > The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
> > path treats it as MICRO. Since no write_raw_get_fmt is provided, the
> > IIO core defaults to MICRO when parsing userspace input.
> >
> > This means reading calibscale and writing it back results in a ~1000x
> > gain error.
> >
> > Change the read path and available range to use MICRO to match the
> > write path.
>
> The updates to ad4030 driver will add write_raw_get_fmt() [1].
> [1]: https://lore.kernel.org/linux-iio/516cccc47e917bd26be29b016907f50a244a68b9.1771865684.git.marcelo.schmitt@analog.com/
>
> At first glance, I think this could instead add a case to write_raw_get_fmt(),
> keeping calibscale nano precision (unless nano precision doesn't make sense
> for ad4030 calibscale (don't recall from top of my mind)).
>
> >
> > Fixes: 0cb8b324852f ("iio: adc: ad4030: add driver for ad4030-24")
> > Signed-off-by: Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com>
> > ---
> >  drivers/iio/adc/ad4030.c | 10 +++++-----
> >  1 file changed, 5 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/iio/adc/ad4030.c b/drivers/iio/adc/ad4030.c
> > index def3e1d01ceb..3d823648371d 100644
> > --- a/drivers/iio/adc/ad4030.c
> > +++ b/drivers/iio/adc/ad4030.c
> > @@ -423,10 +423,10 @@ static int ad4030_get_chan_calibscale(struct iio_dev *indio_dev,
> >
> >       /* From datasheet: multiplied output = input в gain word/0x8000 */
> >       *val = gain / AD4030_GAIN_MIDLE_POINT;
> > -     *val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, NANO,
> > +     *val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, MICRO,
> >                               AD4030_GAIN_MIDLE_POINT);
> >
> > -     return IIO_VAL_INT_PLUS_NANO;
> > +     return IIO_VAL_INT_PLUS_MICRO;
> >  }
> >
> >  /* Returns the offset where 1 LSB = (VREF/2^precision_bits - 1)/gain */
> > @@ -720,8 +720,8 @@ static irqreturn_t ad4030_trigger_handler(int irq, void *p)
> >
> >  static const int ad4030_gain_avail[3][2] = {
> >       { 0, 0 },
> > -     { 0, 30518 },
> > -     { 1, 999969482 },
> > +     { 0, 30 },
> > +     { 1, 999969 },
> >  };
> >
> >  static int ad4030_read_avail(struct iio_dev *indio_dev,
> > @@ -739,7 +739,7 @@ static int ad4030_read_avail(struct iio_dev *indio_dev,
> >
> >       case IIO_CHAN_INFO_CALIBSCALE:
> >               *vals = (void *)ad4030_gain_avail;
> > -             *type = IIO_VAL_INT_PLUS_NANO;
> > +             *type = IIO_VAL_INT_PLUS_MICRO;
> >               return IIO_AVAIL_RANGE;
> >
> >       case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
> > --
> > 2.52.0
> >
> >

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Andy Shevchenko]

On Wed, Feb 25, 2026 at 04:21:53PM +0400, Giorgi Tchankvetadze wrote:
> Hi Marcelo, Thanks for the pointer to your pending series. As for
> keeping NANO precision, the hardware register is 16-bit with a step of
> 1/0x8000 ≈ 30 µ, so NANO doesn't provide any additional precision that
> the hardware can actually represent. MICRO is sufficient.
> 
> Btw, I checked and found that no driver in the entire IIO tree uses
> NANO for calibscale.
> Drivers use either IIO_VAL_INT (raw register value) or IIO_VAL_INT_PLUS_MICRO.
> 
> That said, if your series is already adding write_raw_get_fmt(), I'm
> happy to drop this patch and let yours handle it - as long as the
> mismatch gets fixed either way.

- Please, do not top-post!
- Remove context which is unrelated to the replies.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[David Lechner]

On 2/25/26 5:53 AM, Marcelo Schmitt wrote:
> On 02/25, Giorgi Tchankvetadze wrote:
>> The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
>> path treats it as MICRO. Since no write_raw_get_fmt is provided, the
>> IIO core defaults to MICRO when parsing userspace input.
>>
>> This means reading calibscale and writing it back results in a ~1000x
>> gain error.
>>
>> Change the read path and available range to use MICRO to match the
>> write path.
> 
> The updates to ad4030 driver will add write_raw_get_fmt() [1].
> [1]: https://lore.kernel.org/linux-iio/516cccc47e917bd26be29b016907f50a244a68b9.1771865684.git.marcelo.schmitt@analog.com/
> 
> At first glance, I think this could instead add a case to write_raw_get_fmt(),
> keeping calibscale nano precision (unless nano precision doesn't make sense
> for ad4030 calibscale (don't recall from top of my mind)).

Since this is a fix and needs to be backported, it probably make sense
to not depend on a patch that is adding new features.

I'm sure Jonathan will have an opinion about how he would like to handle
a conflict between the fixes and testing branches though.

> 
>>
>> Fixes: 0cb8b324852f ("iio: adc: ad4030: add driver for ad4030-24")
>> Signed-off-by: Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com>
>> ---
>>  drivers/iio/adc/ad4030.c | 10 +++++-----
>>  1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/iio/adc/ad4030.c b/drivers/iio/adc/ad4030.c
>> index def3e1d01ceb..3d823648371d 100644
>> --- a/drivers/iio/adc/ad4030.c
>> +++ b/drivers/iio/adc/ad4030.c
>> @@ -423,10 +423,10 @@ static int ad4030_get_chan_calibscale(struct iio_dev *indio_dev,
>>  
>>  	/* From datasheet: multiplied output = input × gain word/0x8000 */
>>  	*val = gain / AD4030_GAIN_MIDLE_POINT;
>> -	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, NANO,
>> +	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, MICRO,
>>  				AD4030_GAIN_MIDLE_POINT);
>>  
>> -	return IIO_VAL_INT_PLUS_NANO;
>> +	return IIO_VAL_INT_PLUS_MICRO;
>>  }
>>  
>>  /* Returns the offset where 1 LSB = (VREF/2^precision_bits - 1)/gain */
>> @@ -720,8 +720,8 @@ static irqreturn_t ad4030_trigger_handler(int irq, void *p)
>>  
>>  static const int ad4030_gain_avail[3][2] = {
>>  	{ 0, 0 },
>> -	{ 0, 30518 },
>> -	{ 1, 999969482 },
>> +	{ 0, 30 },
>> +	{ 1, 999969 },

This loses precision. Instead I would change the write to IIO_VAL_INT_PLUS_NANO
to match the read.

>>  };
>>  
>>  static int ad4030_read_avail(struct iio_dev *indio_dev,
>> @@ -739,7 +739,7 @@ static int ad4030_read_avail(struct iio_dev *indio_dev,
>>  
>>  	case IIO_CHAN_INFO_CALIBSCALE:
>>  		*vals = (void *)ad4030_gain_avail;
>> -		*type = IIO_VAL_INT_PLUS_NANO;
>> +		*type = IIO_VAL_INT_PLUS_MICRO;
>>  		return IIO_AVAIL_RANGE;
>>  
>>  	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
>> -- 
>> 2.52.0
>>
>>

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Nuno Sá]

On Wed, 2026-02-25 at 09:20 -0600, David Lechner wrote:
> On 2/25/26 5:53 AM, Marcelo Schmitt wrote:
> > On 02/25, Giorgi Tchankvetadze wrote:
> > > The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
> > > path treats it as MICRO. Since no write_raw_get_fmt is provided, the
> > > IIO core defaults to MICRO when parsing userspace input.
> > > 
> > > This means reading calibscale and writing it back results in a ~1000x
> > > gain error.
> > > 
> > > Change the read path and available range to use MICRO to match the
> > > write path.
> > 
> > The updates to ad4030 driver will add write_raw_get_fmt() [1].
> > [1]:
> > https://lore.kernel.org/linux-iio/516cccc47e917bd26be29b016907f50a244a68b9.1771865684.git.marcelo.schmitt@analog.com/
> > 
> > At first glance, I think this could instead add a case to write_raw_get_fmt(),
> > keeping calibscale nano precision (unless nano precision doesn't make sense
> > for ad4030 calibscale (don't recall from top of my mind)).
> 
> Since this is a fix and needs to be backported, it probably make sense
> to not depend on a patch that is adding new features.
> 

+1

- Nuno Sá

> I'm sure Jonathan will have an opinion about how he would like to handle
> a conflict between the fixes and testing branches though.
> 
> > 
> > > 
> > > Fixes: 0cb8b324852f ("iio: adc: ad4030: add driver for ad4030-24")
> > > Signed-off-by: Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com>
> > > ---
> > >  drivers/iio/adc/ad4030.c | 10 +++++-----
> > >  1 file changed, 5 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/drivers/iio/adc/ad4030.c b/drivers/iio/adc/ad4030.c
> > > index def3e1d01ceb..3d823648371d 100644
> > > --- a/drivers/iio/adc/ad4030.c
> > > +++ b/drivers/iio/adc/ad4030.c
> > > @@ -423,10 +423,10 @@ static int ad4030_get_chan_calibscale(struct iio_dev *indio_dev,
> > >  
> > >  	/* From datasheet: multiplied output = input × gain word/0x8000 */
> > >  	*val = gain / AD4030_GAIN_MIDLE_POINT;
> > > -	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, NANO,
> > > +	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, MICRO,
> > >  				AD4030_GAIN_MIDLE_POINT);
> > >  
> > > -	return IIO_VAL_INT_PLUS_NANO;
> > > +	return IIO_VAL_INT_PLUS_MICRO;
> > >  }
> > >  
> > >  /* Returns the offset where 1 LSB = (VREF/2^precision_bits - 1)/gain */
> > > @@ -720,8 +720,8 @@ static irqreturn_t ad4030_trigger_handler(int irq, void *p)
> > >  
> > >  static const int ad4030_gain_avail[3][2] = {
> > >  	{ 0, 0 },
> > > -	{ 0, 30518 },
> > > -	{ 1, 999969482 },
> > > +	{ 0, 30 },
> > > +	{ 1, 999969 },
> 
> This loses precision. Instead I would change the write to IIO_VAL_INT_PLUS_NANO
> to match the read.
> 
> > >  };
> > >  
> > >  static int ad4030_read_avail(struct iio_dev *indio_dev,
> > > @@ -739,7 +739,7 @@ static int ad4030_read_avail(struct iio_dev *indio_dev,
> > >  
> > >  	case IIO_CHAN_INFO_CALIBSCALE:
> > >  		*vals = (void *)ad4030_gain_avail;
> > > -		*type = IIO_VAL_INT_PLUS_NANO;
> > > +		*type = IIO_VAL_INT_PLUS_MICRO;
> > >  		return IIO_AVAIL_RANGE;
> > >  
> > >  	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
> > > -- 
> > > 2.52.0
> > > 
> > > 

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Jonathan Cameron]

On Wed, 25 Feb 2026 15:30:39 +0400
Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com> wrote:

> The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
> path treats it as MICRO. Since no write_raw_get_fmt is provided, the
> IIO core defaults to MICRO when parsing userspace input.
> 
> This means reading calibscale and writing it back results in a ~1000x
> gain error.
I'm not following this part.  What I'd expect to see something like
reading avail gets you
[0.0 0.0000030518 1.99996982]

Reading the channel gets you maybe 1.1333333333 and then when you write
back because of the lack of write_fmt it gets truncated and you
are effectively reading 1.133333 

So I'd not expect to see a gain error, just a loss of precision.

What am I missing? Please provide a numeric example to motivate the
patch.

Thanks,

Jonathan 
> 
> Change the read path and available range to use MICRO to match the
> write path.
> 
> Fixes: 0cb8b324852f ("iio: adc: ad4030: add driver for ad4030-24")
> Signed-off-by: Giorgi Tchankvetadze <giorgitchankvetadze1997@gmail.com>
> ---
>  drivers/iio/adc/ad4030.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/iio/adc/ad4030.c b/drivers/iio/adc/ad4030.c
> index def3e1d01ceb..3d823648371d 100644
> --- a/drivers/iio/adc/ad4030.c
> +++ b/drivers/iio/adc/ad4030.c
> @@ -423,10 +423,10 @@ static int ad4030_get_chan_calibscale(struct iio_dev *indio_dev,
>  
>  	/* From datasheet: multiplied output = input × gain word/0x8000 */
>  	*val = gain / AD4030_GAIN_MIDLE_POINT;
> -	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, NANO,
> +	*val2 = mul_u64_u32_div(gain % AD4030_GAIN_MIDLE_POINT, MICRO,
>  				AD4030_GAIN_MIDLE_POINT);
>  
> -	return IIO_VAL_INT_PLUS_NANO;
> +	return IIO_VAL_INT_PLUS_MICRO;
>  }
>  
>  /* Returns the offset where 1 LSB = (VREF/2^precision_bits - 1)/gain */
> @@ -720,8 +720,8 @@ static irqreturn_t ad4030_trigger_handler(int irq, void *p)
>  
>  static const int ad4030_gain_avail[3][2] = {
>  	{ 0, 0 },
> -	{ 0, 30518 },
> -	{ 1, 999969482 },
> +	{ 0, 30 },
> +	{ 1, 999969 },
>  };
>  
>  static int ad4030_read_avail(struct iio_dev *indio_dev,
> @@ -739,7 +739,7 @@ static int ad4030_read_avail(struct iio_dev *indio_dev,
>  
>  	case IIO_CHAN_INFO_CALIBSCALE:
>  		*vals = (void *)ad4030_gain_avail;
> -		*type = IIO_VAL_INT_PLUS_NANO;
> +		*type = IIO_VAL_INT_PLUS_MICRO;
>  		return IIO_AVAIL_RANGE;
>  
>  	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:

Re: [PATCH RFC] iio: adc: ad4030: fix calibscale read/write unit mismatch

[Jonathan Cameron]

On Thu, 26 Feb 2026 09:32:45 +0000
Nuno Sá <noname.nuno@gmail.com> wrote:

> On Wed, 2026-02-25 at 09:20 -0600, David Lechner wrote:
> > On 2/25/26 5:53 AM, Marcelo Schmitt wrote:  
> > > On 02/25, Giorgi Tchankvetadze wrote:  
> > > > The read path returns calibscale in IIO_VAL_INT_PLUS_NANO but the write
> > > > path treats it as MICRO. Since no write_raw_get_fmt is provided, the
> > > > IIO core defaults to MICRO when parsing userspace input.
> > > > 
> > > > This means reading calibscale and writing it back results in a ~1000x
> > > > gain error.
> > > > 
> > > > Change the read path and available range to use MICRO to match the
> > > > write path.  
> > > 
> > > The updates to ad4030 driver will add write_raw_get_fmt() [1].
> > > [1]:
> > > https://lore.kernel.org/linux-iio/516cccc47e917bd26be29b016907f50a244a68b9.1771865684.git.marcelo.schmitt@analog.com/
> > > 
> > > At first glance, I think this could instead add a case to write_raw_get_fmt(),
> > > keeping calibscale nano precision (unless nano precision doesn't make sense
> > > for ad4030 calibscale (don't recall from top of my mind)).  
> > 
> > Since this is a fix and needs to be backported, it probably make sense
> > to not depend on a patch that is adding new features.
> >   
> 
> +1
> 
> - Nuno Sá
> 
> > I'm sure Jonathan will have an opinion about how he would like to handle
> > a conflict between the fixes and testing branches though.

Fixes in theory always go first.  That'll mean I end up holding the
other series back if this one is ready in the near future.

I'm not sure I fully understand the bug though!

Jonathan