mbox series

[0/2] iio: potentiometer: Add driver support for AD5110

Message ID 20210807050900.10075-1-dmugil2000@gmail.com
Headers show
Series iio: potentiometer: Add driver support for AD5110 | expand

Message

Mugilraj Dhavachelvan Aug. 7, 2021, 5:08 a.m. UTC 
Add dt-bindings and driver support for AD5110, a Nonvolatile 
Digital Potentiometer.

Mugilraj Dhavachelvan (2):
  dt-bindings: iio: potentiometer: Add AD5110 in trivial-devices
  iio: potentiometer: Add driver support for AD5110

 .../devicetree/bindings/trivial-devices.yaml  |   2 +
 MAINTAINERS                                   |   6 +
 drivers/iio/potentiometer/Kconfig             |  10 +
 drivers/iio/potentiometer/Makefile            |   1 +
 drivers/iio/potentiometer/ad5110.c            | 333 ++++++++++++++++++
 5 files changed, 352 insertions(+)
 create mode 100644 drivers/iio/potentiometer/ad5110.c

Comments

Lars-Peter Clausen Aug. 7, 2021, 12:11 p.m. UTC | #1 
On 8/7/21 7:08 AM, Mugilraj Dhavachelvan wrote:
> The AD5110/AD5112/AD5114 provide a nonvolatile solution
> for 128-/64-/32-position adjustment applications, offering
> guaranteed low resistor tolerance errors of ±8% and up to
> ±6 mA current density.
>
> Datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/AD5110_5112_5114.pdf
> Signed-off-by: Mugilraj Dhavachelvan <dmugil2000@gmail.com>

Thanks for the patch. This looks really good!


> [...]
>
> +static int ad5110_write(struct ad5110_data *data, u8 cmd, u8 val)
> +{
> +	int ret;
> +
> +	mutex_lock(&data->lock);
> +	data->buf[0] = cmd;
> +	data->buf[1] = val;
> +
> +	ret = i2c_master_send_dmasafe(data->client, data->buf, sizeof(data->buf));
> +	mutex_unlock(&data->lock);

This returns the number of bytes written, which can be less then the 
number of bytes requested if there was an error. This should check if 
the right amount of bytes was written and return -EIO otherwise. Same 
for the other places that read/write from I2C.

> +
> +	return ret < 0 ? ret : 0;
> +}
> +
> +static int ad5110_resistor_tol(struct ad5110_data *data, u8 cmd, int val)
> +{
> +	int ret;
> +
> +	ret = ad5110_read(data, cmd, &val);
> +	if (ret)
> +		return ret;
> +
> +	data->tol = FIELD_GET(GENMASK(6, 3), val);
> +	data->tol = ((val & GENMASK(2, 0)) * 1000 / 8) + data->tol * 1000;
> +	data->tol = data->cfg->kohms * data->tol / 100;

This is not wrong, but it can be implemented a bit simpler. The 
tolerance is encoded as a fixed point number, you don't have to treat 
them as two independent fields, but can treat it as one fixed point number.

     data->tol = data->cfg->kohms * (val & GENMASK(6, 0)) * 1000 / 100 / 8;


> +	if (!(val & BIT(7)))
> +		data->tol *= -1;
> +
> +	return 0;
> +}
> +
> +static ssize_t ad5110_eeprom_read(struct device *dev,
> +					  struct device_attribute *attr,
> +					  char *buf)
> +{
> +	struct iio_dev *indio_dev = dev_to_iio_dev(dev);
> +	struct ad5110_data *data = iio_priv(indio_dev);
> +	int val = AD5110_WIPER_POS;
> +	int ret;
> +
> +	ret = ad5110_read(data, AD5110_EEPROM_RD, &val);
> +	if (ret)
> +		return ret;
Maybe apply shift to get consistent behavior with `raw`.
> +
> +	return iio_format_value(buf, IIO_VAL_INT, 1, &val);
> +}
> +
> +static ssize_t ad5110_eeprom_write(struct device *dev,
> +					   struct device_attribute *attr,
> +					   const char *buf, size_t len)
> +{
> +	struct iio_dev *indio_dev = dev_to_iio_dev(dev);
> +	struct ad5110_data *data = iio_priv(indio_dev);
> +	int ret;
> +
> +	ret = ad5110_write(data, AD5110_EEPROM_WR, 0);
> +	if (ret) {
> +		dev_err(&data->client->dev, "RDAC to EEPROM write failed\n");
> +		return ret;
> +	}
> +	msleep(20);
> +
> +	return len;
> +}
> +
> +static IIO_DEVICE_ATTR(wiper_pos_eeprom, 0644,
> +		       ad5110_eeprom_read,
> +		       ad5110_eeprom_write, 0);
This is new custom ABI and needs to be documented.
> +static int ad5110_write_raw(struct iio_dev *indio_dev,
> +			    struct iio_chan_spec const *chan,
> +			    int val, int val2, long mask)
> +{
> +	struct ad5110_data *data = iio_priv(indio_dev);
> +	int ret;
> +
> +	switch (mask) {
> +	case IIO_CHAN_INFO_RAW:
> +		if (val >= data->cfg->max_pos || val < 0)
val == data->cfg->max_pos is a valid setting. Writing max_pos puts it in 
top-scale mode which gives maximum resistance.
> +			return -EINVAL;
> +
> +		return ad5110_write(data, AD5110_RDAC_WR, val << data->cfg->shift);
> +	case IIO_CHAN_INFO_ENABLE:
> +		if (val < 0 || val > 1)
> +			return -EINVAL;
> +		if (data->enable == val)
> +			return 0;
> +		ret = ad5110_write(data, AD5110_SHUTDOWN, val);
Doesn't val have to be inverted to get the right behavior?
> +		if (ret)
> +			return ret;
> +		data->enable = val;
> +		return 0;
> +	default:
> +		return -EINVAL;
> +	}
> +}
[...]
Lars-Peter Clausen Aug. 7, 2021, 12:17 p.m. UTC | #2 
On 8/7/21 7:08 AM, Mugilraj Dhavachelvan wrote:
> +static const struct iio_chan_spec ad5110_channels[] = {
> +	{
> +		.type = IIO_RESISTANCE,
> +		.output = 1,
> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | BIT(IIO_CHAN_INFO_OFFSET),
> +		.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE) |
> +						BIT(IIO_CHAN_INFO_ENABLE),
There is only on channel in this device. Why are some `separate` and 
others `shared_by_type`?
Mugilraj Dhavachelvan Aug. 7, 2021, 5:26 p.m. UTC | #3 
On 07/08/21 5:41 pm, Lars-Peter Clausen wrote:
> On 8/7/21 7:08 AM, Mugilraj Dhavachelvan wrote:
>> The AD5110/AD5112/AD5114 provide a nonvolatile solution
>> for 128-/64-/32-position adjustment applications, offering
>> guaranteed low resistor tolerance errors of ±8% and up to
>> ±6 mA current density.
>>
>> Datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/AD5110_5112_5114.pdf
>> Signed-off-by: Mugilraj Dhavachelvan <dmugil2000@gmail.com>
> 
> Thanks for the patch. This looks really good> 
> 
>> [...]
>>
>> +static int ad5110_write(struct ad5110_data *data, u8 cmd, u8 val)
>> +{
>> +    int ret;
>> +
>> +    mutex_lock(&data->lock);
>> +    data->buf[0] = cmd;
>> +    data->buf[1] = val;
>> +
>> +    ret = i2c_master_send_dmasafe(data->client, data->buf, sizeof(data->buf));
>> +    mutex_unlock(&data->lock);
> 
> This returns the number of bytes written, which can be less then the number of bytes requested if there was an error. This should check if the right amount of bytes was written and return -EIO otherwise. Same for the other places that read/write from I2C.

Fixed in v2.
> 
>> +
>> +    return ret < 0 ? ret : 0;
>> +}
>> +
>> +static int ad5110_resistor_tol(struct ad5110_data *data, u8 cmd, int val)
>> +{
>> +    int ret;
>> +
>> +    ret = ad5110_read(data, cmd, &val);
>> +    if (ret)
>> +        return ret;
>> +
>> +    data->tol = FIELD_GET(GENMASK(6, 3), val);
>> +    data->tol = ((val & GENMASK(2, 0)) * 1000 / 8) + data->tol * 1000;
>> +    data->tol = data->cfg->kohms * data->tol / 100;
> 
> This is not wrong, but it can be implemented a bit simpler. The tolerance is encoded as a fixed point number, you don't have to treat them as two independent fields, but can treat it as one fixed point number.
> 
>     data->tol = data->cfg->kohms * (val & GENMASK(6, 0)) * 1000 / 100 / 8;
> 
Great, Fixed in v2.
> 
>> +    if (!(val & BIT(7)))
>> +        data->tol *= -1;
>> +
>> +    return 0;
>> +}
>> +
>> +static ssize_t ad5110_eeprom_read(struct device *dev,
>> +                      struct device_attribute *attr,
>> +                      char *buf)
>> +{
>> +    struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>> +    struct ad5110_data *data = iio_priv(indio_dev);
>> +    int val = AD5110_WIPER_POS;
>> +    int ret;
>> +
>> +    ret = ad5110_read(data, AD5110_EEPROM_RD, &val);
>> +    if (ret)
>> +        return ret;
> Maybe apply shift to get consistent behavior with `raw`.
Fixed in v2.
>> +
>> +    return iio_format_value(buf, IIO_VAL_INT, 1, &val);
>> +}
>> +
>> +static ssize_t ad5110_eeprom_write(struct device *dev,
>> +                       struct device_attribute *attr,
>> +                       const char *buf, size_t len)
>> +{
>> +    struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>> +    struct ad5110_data *data = iio_priv(indio_dev);
>> +    int ret;
>> +
>> +    ret = ad5110_write(data, AD5110_EEPROM_WR, 0);
>> +    if (ret) {
>> +        dev_err(&data->client->dev, "RDAC to EEPROM write failed\n");
>> +        return ret;
>> +    }
>> +    msleep(20);
>> +
>> +    return len;
>> +}
>> +
>> +static IIO_DEVICE_ATTR(wiper_pos_eeprom, 0644,
>> +               ad5110_eeprom_read,
>> +               ad5110_eeprom_write, 0);
> This is new custom ABI and needs to be documentedI'm not aware of this, fixed in v2.
>> +static int ad5110_write_raw(struct iio_dev *indio_dev,
>> +                struct iio_chan_spec const *chan,
>> +                int val, int val2, long mask)
>> +{
>> +    struct ad5110_data *data = iio_priv(indio_dev);
>> +    int ret;
>> +
>> +    switch (mask) {
>> +    case IIO_CHAN_INFO_RAW:
>> +        if (val >= data->cfg->max_pos || val < 0)
> val == data->cfg->max_pos is a valid setting. Writing max_pos puts it in top-scale mode which gives maximum resistance.
Fixed in v2.
>> +            return -EINVAL;
>> +
>> +        return ad5110_write(data, AD5110_RDAC_WR, val << data->cfg->shift);
>> +    case IIO_CHAN_INFO_ENABLE:
>> +        if (val < 0 || val > 1)
>> +            return -EINVAL;
>> +        if (data->enable == val)
>> +            return 0;
>> +        ret = ad5110_write(data, AD5110_SHUTDOWN, val);
> Doesn't val have to be inverted to get the right behaviorI just replicated the datasheet operation. 
You mean,
 1 - shutdown off
 0 - shutdown on
If yes, then the user won't get confused with the datasheet and the behavior of the driver?
Or Is it work like this? But yeah even I like this change it's more convenient.
>> +        if (ret)
>> +            return ret;
>> +        data->enable = val;
>> +        return 0;
>> +    default:
>> +        return -EINVAL;
>> +    }
>> +}
> [...]
Thanks for feedback!!
Mugilraj Dhavachelvan Aug. 7, 2021, 5:28 p.m. UTC | #4 
On 07/08/21 5:47 pm, Lars-Peter Clausen wrote:
> On 8/7/21 7:08 AM, Mugilraj Dhavachelvan wrote:
>> +static const struct iio_chan_spec ad5110_channels[] = {
>> +    {
>> +        .type = IIO_RESISTANCE,
>> +        .output = 1,
>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | BIT(IIO_CHAN_INFO_OFFSET),
>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE) |
>> +                        BIT(IIO_CHAN_INFO_ENABLE),
> There is only on channel in this device. Why are some `separate` and others `shared_by_type`?
> 
No special reason, I'll put everything into 'separate'. Fixed in v2.
>
Mugilraj Dhavachelvan Aug. 7, 2021, 5:34 p.m. UTC | #5 
Sorry some formatting issues happened in my previous mail.

On 07/08/21 10:56 pm, Mugilraj Dhavachelvan wrote:
> 
> 
> On 07/08/21 5:41 pm, Lars-Peter Clausen wrote:
>> On 8/7/21 7:08 AM, Mugilraj Dhavachelvan wrote:
>>> The AD5110/AD5112/AD5114 provide a nonvolatile solution
>>> for 128-/64-/32-position adjustment applications, offering
>>> guaranteed low resistor tolerance errors of ±8% and up to
>>> ±6 mA current density.
>>>
>>> Datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/AD5110_5112_5114.pdf
>>> Signed-off-by: Mugilraj Dhavachelvan <dmugil2000@gmail.com>
>>
>> Thanks for the patch. This looks really good> 
>>
>>> [...]
>>>
>>> +static int ad5110_write(struct ad5110_data *data, u8 cmd, u8 val)
>>> +{
>>> +    int ret;
>>> +
>>> +    mutex_lock(&data->lock);
>>> +    data->buf[0] = cmd;
>>> +    data->buf[1] = val;
>>> +
>>> +    ret = i2c_master_send_dmasafe(data->client, data->buf, sizeof(data->buf));
>>> +    mutex_unlock(&data->lock);
>>
>> This returns the number of bytes written, which can be less then the number of bytes requested if there was an error. This should check if the right amount of bytes was written and return -EIO otherwise. Same for the other places that read/write from I2C.
> 
> Fixed in v2.
>>
>>> +
>>> +    return ret < 0 ? ret : 0;
>>> +}
>>> +
>>> +static int ad5110_resistor_tol(struct ad5110_data *data, u8 cmd, int val)
>>> +{
>>> +    int ret;
>>> +
>>> +    ret = ad5110_read(data, cmd, &val);
>>> +    if (ret)
>>> +        return ret;
>>> +
>>> +    data->tol = FIELD_GET(GENMASK(6, 3), val);
>>> +    data->tol = ((val & GENMASK(2, 0)) * 1000 / 8) + data->tol * 1000;
>>> +    data->tol = data->cfg->kohms * data->tol / 100;
>>
>> This is not wrong, but it can be implemented a bit simpler. The tolerance is encoded as a fixed point number, you don't have to treat them as two independent fields, but can treat it as one fixed point number.
>>
>>     data->tol = data->cfg->kohms * (val & GENMASK(6, 0)) * 1000 / 100 / 8;
>>
> Great, Fixed in v2.
>>
>>> +    if (!(val & BIT(7)))
>>> +        data->tol *= -1;
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +static ssize_t ad5110_eeprom_read(struct device *dev,
>>> +                      struct device_attribute *attr,
>>> +                      char *buf)
>>> +{
>>> +    struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>>> +    struct ad5110_data *data = iio_priv(indio_dev);
>>> +    int val = AD5110_WIPER_POS;
>>> +    int ret;
>>> +
>>> +    ret = ad5110_read(data, AD5110_EEPROM_RD, &val);
>>> +    if (ret)
>>> +        return ret;
>> Maybe apply shift to get consistent behavior with `raw`.
> Fixed in v2.
>>> +
>>> +    return iio_format_value(buf, IIO_VAL_INT, 1, &val);
>>> +}
>>> +
>>> +static ssize_t ad5110_eeprom_write(struct device *dev,
>>> +                       struct device_attribute *attr,
>>> +                       const char *buf, size_t len)
>>> +{
>>> +    struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>>> +    struct ad5110_data *data = iio_priv(indio_dev);
>>> +    int ret;
>>> +
>>> +    ret = ad5110_write(data, AD5110_EEPROM_WR, 0);
>>> +    if (ret) {
>>> +        dev_err(&data->client->dev, "RDAC to EEPROM write failed\n");
>>> +        return ret;
>>> +    }
>>> +    msleep(20);
>>> +
>>> +    return len;
>>> +}
>>> +
>>> +static IIO_DEVICE_ATTR(wiper_pos_eeprom, 0644,
>>> +               ad5110_eeprom_read,
>>> +               ad5110_eeprom_write, 0);
>> This is new custom ABI and needs to be documented

> I'm not aware of this, fixed in v2.
>>> +static int ad5110_write_raw(struct iio_dev *indio_dev,
>>> +                struct iio_chan_spec const *chan,
>>> +                int val, int val2, long mask)
>>> +{
>>> +    struct ad5110_data *data = iio_priv(indio_dev);
>>> +    int ret;
>>> +
>>> +    switch (mask) {
>>> +    case IIO_CHAN_INFO_RAW:
>>> +        if (val >= data->cfg->max_pos || val < 0)
>> val == data->cfg->max_pos is a valid setting. Writing max_pos puts it in top-scale mode which gives maximum resistance.
> Fixed in v2.
>>> +            return -EINVAL;
>>> +
>>> +        return ad5110_write(data, AD5110_RDAC_WR, val << data->cfg->shift);
>>> +    case IIO_CHAN_INFO_ENABLE:
>>> +        if (val < 0 || val > 1)
>>> +            return -EINVAL;
>>> +        if (data->enable == val)
>>> +            return 0;
>>> +        ret = ad5110_write(data, AD5110_SHUTDOWN, val);
>> Doesn't val have to be inverted to get the right behavior

> I just replicated the datasheet operation. 
> You mean,
>  1 - shutdown off
>  0 - shutdown on
> If yes, then the user won't get confused with the datasheet and the behavior of the driver?
> Or Is it work like this? But yeah even I like this change it's more convenient.
>>> +        if (ret)
>>> +            return ret;
>>> +        data->enable = val;
>>> +        return 0;
>>> +    default:
>>> +        return -EINVAL;
>>> +    }
>>> +}
>> [...]
> Thanks for feedback!!
>
Jonathan Cameron Aug. 8, 2021, 4:22 p.m. UTC | #6 
On Sat, 7 Aug 2021 23:04:18 +0530
Mugilraj Dhavachelvan <dmugil2000@gmail.com> wrote:

> Sorry some formatting issues happened in my previous mail.
> 
> On 07/08/21 10:56 pm, Mugilraj Dhavachelvan wrote:
> > 
> > 
> > On 07/08/21 5:41 pm, Lars-Peter Clausen wrote:  
> >> On 8/7/21 7:08 AM, Mugilraj Dhavachelvan wrote:  
> >>> The AD5110/AD5112/AD5114 provide a nonvolatile solution
> >>> for 128-/64-/32-position adjustment applications, offering
> >>> guaranteed low resistor tolerance errors of ±8% and up to
> >>> ±6 mA current density.
> >>>
> >>> Datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/AD5110_5112_5114.pdf
> >>> Signed-off-by: Mugilraj Dhavachelvan <dmugil2000@gmail.com>  
> >>
> >> Thanks for the patch. This looks really good> 
> >>  
>

...

> >>> +
> >>> +static IIO_DEVICE_ATTR(wiper_pos_eeprom, 0644,
> >>> +               ad5110_eeprom_read,
> >>> +               ad5110_eeprom_write, 0);  
> >> This is new custom ABI and needs to be documented  

We have existing similar ABI in dac/mcp4725 which is simply
called store_eeprom

It's in the main docs
Documentation/ABI/testing/sysfs-bus-iio as storing
device configuration.  I'm guessing this device doesn't have
other configuration so that description will work?


> 
> > I'm not aware of this, fixed in v2.  
> >>> +static int ad5110_write_raw(struct iio_dev *indio_dev,
> >>> +                struct iio_chan_spec const *chan,
> >>> +                int val, int val2, long mask)
> >>> +{
> >>> +    struct ad5110_data *data = iio_priv(indio_dev);
> >>> +    int ret;
> >>> +
> >>> +    switch (mask) {
> >>> +    case IIO_CHAN_INFO_RAW:
> >>> +        if (val >= data->cfg->max_pos || val < 0)  
> >> val == data->cfg->max_pos is a valid setting. Writing max_pos puts it in top-scale mode which gives maximum resistance.  
> > Fixed in v2.  
> >>> +            return -EINVAL;
> >>> +
> >>> +        return ad5110_write(data, AD5110_RDAC_WR, val << data->cfg->shift);
> >>> +    case IIO_CHAN_INFO_ENABLE:
> >>> +        if (val < 0 || val > 1)
> >>> +            return -EINVAL;
> >>> +        if (data->enable == val)
> >>> +            return 0;
> >>> +        ret = ad5110_write(data, AD5110_SHUTDOWN, val);  
> >> Doesn't val have to be inverted to get the right behavior  
> 
> > I just replicated the datasheet operation. 
> > You mean,
> >  1 - shutdown off
> >  0 - shutdown on
> > If yes, then the user won't get confused with the datasheet and the behavior of the driver?
> > Or Is it work like this? But yeah even I like this change it's more convenient.  

ABI has to be consistent and writing an enable attribute with 1 has to mean enabling it whatever
approach the datasheet takes to describe things.  Most users don't read
datasheets so interface needs to be intuitive.

> >>> +        if (ret)
> >>> +            return ret;
> >>> +        data->enable = val;
> >>> +        return 0;
> >>> +    default:
> >>> +        return -EINVAL;
> >>> +    }
> >>> +}  
> >> [...]  
> > Thanks for feedback!!
> >