Re: [PATCH 4/5] staging: iio: ad9832: convert to iio channels and ext_info attrs

[Marcelo Schmitt <marcelo.schmitt1@gmail.com>]

Hi Tomas,

The updates look mostly good to me. A few comments inline.

On 12/15, Tomas Borquez wrote:
> Convert ad9832 from custom sysfs attributes to standard channel interface
> using a single IIO_ALTCURRENT channel with ext_info attributes, as this
> device is a current source DAC with one output, as well as removing the
> dds.h header.
> 
> Changes:
> - Add single iio_chan_spec with ext_info for frequency0/1 and phase0-3
> - Phase attributes accept radians directly, driver converts internally
> - Frequency attributes accept Hz (unchanged)
> - Cache frequency and phase values in driver state for readback
> - Remove dependency on dds.h macros
I'm not sure, was the dds stuff being used before this patch? Maybe dds.h
removal would be better as another separate clean up patch.

> - Rename symbol attributes to frequency_symbol and phase_symbol
It's nice to have a change log between patch versions. Though, it's usually
provided as part of extra patch info, not commit message.

> 
> The pincontrol_en attribute is kept temporarily with a TODO noting it
> should become a DT property during staging graduation.
> 
> NOTE: This changes the ABI from out_altvoltage0_* to out_altcurrent0_*
> with different attribute organization.
> 
> Signed-off-by: Tomas Borquez <tomasborquez13@gmail.com>
> ---

The change log is usually provided here, below the '---'.

>  drivers/staging/iio/frequency/ad9832.c | 279 +++++++++++++++++++------
>  1 file changed, 215 insertions(+), 64 deletions(-)

This patch fails to apply? I've tried getting it applied on top of current
IIO testing branch with b4 shazam, git am <individual patches>, and
git apply <patch4 on top of applied patch3>, but patch 4 fails to apply either way.
Couldn't figure out how to fix that.

> 
> diff --git a/drivers/staging/iio/frequency/ad9832.c b/drivers/staging/iio/frequency/ad9832.c
> index 8d04f1b44f..454a317732 100644
> --- a/drivers/staging/iio/frequency/ad9832.c
> +++ b/drivers/staging/iio/frequency/ad9832.c
...
> +static const u32 ad9832_phase_addr[] = {
> +	AD9832_PHASE0H, AD9832_PHASE1H, AD9832_PHASE2H, AD9832_PHASE3H
> +};
> +
> +static ssize_t ad9832_write_phase(struct iio_dev *indio_dev,
> +				  uintptr_t private,
> +				  struct iio_chan_spec const *chan,
> +				  const char *buf, size_t len)
>  {
> +	struct ad9832_state *st = iio_priv(indio_dev);
>  	u8 phase_bytes[2];
>  	u16 phase_cmd;
> +	u32 phase_urad, phase;
> +	int val, val2, ret;
>  
> -	if (phase >= BIT(AD9832_PHASE_BITS))
> +	if (private >= ARRAY_SIZE(ad9832_phase_addr))
>  		return -EINVAL;
>  
> +	ret = iio_str_to_fixpoint(buf, 100000, &val, &val2);
Maybe I'm missing something but, why 100000 here? Should it be MICRO instead?

> +	if (ret)
> +		return ret;
> +
> +	if (val < 0 || val2 < 0)
> +		return -EINVAL;
> +
> +	phase_urad = val * 1000000 + val2;
We can use macros to make it easier to read values that are related to metric units.
	phase_urad = val * MICRO + val2;
then
#include <linux/units.h>

...
> +static ssize_t ad9832_read_phase(struct iio_dev *indio_dev,
> +				 uintptr_t private,
> +				 struct iio_chan_spec const *chan,
> +				 char *buf)
> +{
> +	struct ad9832_state *st = iio_priv(indio_dev);
> +	u32 phase_urad;
> +
> +	if (private >= ARRAY_SIZE(ad9832_phase_addr))
> +		return -EINVAL;
> +
> +	guard(mutex)(&st->lock);
> +	phase_urad = ((u64)st->phase[private] * AD9832_2PI_URAD) >> AD9832_PHASE_BITS;
> +
> +	return sysfs_emit(buf, "%u.%06u\n", phase_urad / 1000000, phase_urad % 1000000);
	return sysfs_emit(buf, "%u.%06u\n", phase_urad / MICRO, phase_urad % MICRO);
?

>  }
>  
...
> @@ -234,49 +322,110 @@ static ssize_t ad9832_write(struct device *dev, struct device_attribute *attr,
>  		st->data = cpu_to_be16(FIELD_PREP(AD9832_CMD_MSK, AD9832_CMD_SLEEPRESCLR) |
>  						  st->ctrl_src);
>  		ret = spi_sync(st->spi, &st->msg);
> +		if (ret)
> +			return ret;
> +
> +		st->output_en = val;
> +		break;
> +	case AD9832_PINCTRL_EN:
> +		if (val != 1 && val != 0)
> +			return -EINVAL;
> +
> +		st->ctrl_ss &= ~AD9832_SELSRC;
> +		st->ctrl_ss |= FIELD_PREP(AD9832_SELSRC, val ? 0 : 1);
this would then be only
		st->ctrl_ss |= FIELD_PREP(AD9832_SELSRC, val);
so we keep consistency with previous FIELD_PREP() in ad9832_store().

> +
> +		st->data = cpu_to_be16(FIELD_PREP(AD9832_CMD_MSK, AD9832_CMD_SYNCSELSRC) |
> +						 st->ctrl_ss);
> +		ret = spi_sync(st->spi, &st->msg);
> +		if (ret)
> +			return ret;
> +
> +		st->pinctrl_en = val;
>  		break;
>  	default:
>  		return -ENODEV;
>  	}
>  
> -	return ret ? ret : len;
> +	return len;
>  }
>  
...
> +#define AD9832_CHAN_FREQ(_name, _channel) {			\
> +	.name = _name,						\
> +	.write = ad9832_write_frequency,			\
> +	.read = ad9832_read_frequency,				\
> +	.private = _channel,					\
Since the DAC has only one output channel, doesn't it look somewhat misleading
to call private data _channel?
Based on data sheet naming, I'd call it _select. Maybe _fselect _psel?

> +	.shared = IIO_SEPARATE,					\
> +}
> +
> +#define AD9832_CHAN_PHASE(_name, _channel) {			\
> +	.name = _name,						\
> +	.write = ad9832_write_phase,				\
> +	.read = ad9832_read_phase,				\
> +	.private = _channel,					\
same here

> +	.shared = IIO_SEPARATE,					\
> +}
> +

Best regards,
Marcelo