[PATCH] iio: adc: ti-ads8688: use the auto sequence feature

[PATCH] iio: adc: ti-ads8688: use the auto sequence feature

[Sean Nyekjaer]

The TI ADS8688 have a auto sequence feature that allows
to continuesly reading new values from the chosen channels.
Lets enable it and we will be able to sample at the double speed
vs the old implementation

Signed-off-by: Sean Nyekjaer <sean.nyekjaer@prevas.dk>
---
 drivers/iio/adc/ti-ads8688.c | 60 +++++++++++++++++++++++++++++++++++-
 1 file changed, 59 insertions(+), 1 deletion(-)

diff --git a/drivers/iio/adc/ti-ads8688.c b/drivers/iio/adc/ti-ads8688.c
index 184d686ebd99..8e18c0caa1d7 100644
--- a/drivers/iio/adc/ti-ads8688.c
+++ b/drivers/iio/adc/ti-ads8688.c
@@ -25,10 +25,12 @@
 #define ADS8688_CMD_REG(x)		(x << 8)
 #define ADS8688_CMD_REG_NOOP		0x00
 #define ADS8688_CMD_REG_RST		0x85
+#define ADS8688_CMD_REG_AUTO_RST	0xA0
 #define ADS8688_CMD_REG_MAN_CH(chan)	(0xC0 | (4 * chan))
 #define ADS8688_CMD_DONT_CARE_BITS	16
 
 #define ADS8688_PROG_REG(x)		(x << 9)
+#define ADS8688_AUTO_SEQ_EN		0x01
 #define ADS8688_PROG_REG_RANGE_CH(chan)	(0x05 + chan)
 #define ADS8688_PROG_WR_BIT		BIT(8)
 #define ADS8688_PROG_DONT_CARE_BITS	8
@@ -210,6 +212,39 @@ static int ads8688_reset(struct iio_dev *indio_dev)
 	return spi_write(st->spi, &st->data[0].d8[0], 4);
 }
 
+static int ads8688_start_autoseq(struct iio_dev *indio_dev)
+{
+	struct ads8688_state *st = iio_priv(indio_dev);
+	u32 tmp;
+
+	tmp = ADS8688_CMD_REG(ADS8688_CMD_REG_AUTO_RST);
+	tmp <<= ADS8688_CMD_DONT_CARE_BITS;
+	st->data[0].d32 = cpu_to_be32(tmp);
+
+	return spi_write(st->spi, &st->data[0].d8[0], 4);
+}
+
+static int ads8688_read_autoseq(struct iio_dev *indio_dev)
+{
+	struct ads8688_state *st = iio_priv(indio_dev);
+	int ret;
+	u32 tmp;
+	struct spi_transfer t = {
+		.tx_buf = &st->data[0].d8[0],
+		.len = 4,
+	};
+
+	tmp = ADS8688_CMD_REG(ADS8688_CMD_REG_NOOP);
+	tmp <<= ADS8688_CMD_DONT_CARE_BITS;
+	st->data[0].d32 = cpu_to_be32(tmp);
+
+	ret = spi_sync_transfer(st->spi, &t, 1);
+	if (ret < 0)
+		return ret;
+
+	return be32_to_cpu(st->data[0].d32) & 0xffff;
+}
+
 static int ads8688_read(struct iio_dev *indio_dev, unsigned int chan)
 {
 	struct ads8688_state *st = iio_priv(indio_dev);
@@ -251,6 +286,10 @@ static int ads8688_read_raw(struct iio_dev *indio_dev,
 
 	struct ads8688_state *st = iio_priv(indio_dev);
 
+	/* Block raw reading when in triggered mode */
+	if (iio_buffer_enabled(indio_dev))
+		return -EBUSY;
+
 	mutex_lock(&st->lock);
 	switch (m) {
 	case IIO_CHAN_INFO_RAW:
@@ -299,6 +338,10 @@ static int ads8688_write_raw(struct iio_dev *indio_dev,
 	unsigned int scale = 0;
 	int ret = -EINVAL, i, offset = 0;
 
+	/* Block setup when in triggered mode */
+	if (iio_buffer_enabled(indio_dev))
+		return -EBUSY;
+
 	mutex_lock(&st->lock);
 	switch (mask) {
 	case IIO_CHAN_INFO_SCALE:
@@ -374,10 +417,25 @@ static int ads8688_write_raw_get_fmt(struct iio_dev *indio_dev,
 	return -EINVAL;
 }
 
+static int ads8688_update_scan_mode(struct iio_dev *indio_dev,
+	const unsigned long *active_scan_mask)
+{
+	int scan_count;
+
+	scan_count = bitmap_weight(active_scan_mask, indio_dev->masklength);
+
+	ads8688_prog_write(indio_dev, ADS8688_AUTO_SEQ_EN, (int)active_scan_mask);
+
+	ads8688_start_autoseq(indio_dev);
+
+	return 0;
+}
+
 static const struct iio_info ads8688_info = {
 	.read_raw = &ads8688_read_raw,
 	.write_raw = &ads8688_write_raw,
 	.write_raw_get_fmt = &ads8688_write_raw_get_fmt,
+	.update_scan_mode = ads8688_update_scan_mode,
 	.attrs = &ads8688_attribute_group,
 };
 
@@ -391,7 +449,7 @@ static irqreturn_t ads8688_trigger_handler(int irq, void *p)
 	for (i = 0; i < indio_dev->masklength; i++) {
 		if (!test_bit(i, indio_dev->active_scan_mask))
 			continue;
-		buffer[j] = ads8688_read(indio_dev, i);
+		buffer[j] = ads8688_read_autoseq(indio_dev);
 		j++;
 	}
 
-- 
2.17.0

Re: [PATCH] iio: adc: ti-ads8688: use the auto sequence feature

[Jonathan Cameron]

On Wed, 30 May 2018 23:18:21 +0200
Sean Nyekjaer <sean.nyekjaer@prevas.dk> wrote:

> The TI ADS8688 have a auto sequence feature that allows
> to continuesly reading new values from the chosen channels.
> Lets enable it and we will be able to sample at the double speed
> vs the old implementation
> 
> Signed-off-by: Sean Nyekjaer <sean.nyekjaer@prevas.dk>

hmm. Technically it's an ABI change as someone may be relying on the
ability to do sysfs reads whilst we are also using the chardev interface.
However, I doubt anyone is so lest see if anyone notices ;)

One comment inline. Otherwise looks fine to me.

With care, we could gotten the speed up without using autosequencing
but that doesn't really matter as that isn't what we were doing previously.

Thanks,

Jonathan

> ---
>  drivers/iio/adc/ti-ads8688.c | 60 +++++++++++++++++++++++++++++++++++-
>  1 file changed, 59 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/iio/adc/ti-ads8688.c b/drivers/iio/adc/ti-ads8688.c
> index 184d686ebd99..8e18c0caa1d7 100644
> --- a/drivers/iio/adc/ti-ads8688.c
> +++ b/drivers/iio/adc/ti-ads8688.c
> @@ -25,10 +25,12 @@
>  #define ADS8688_CMD_REG(x)		(x << 8)
>  #define ADS8688_CMD_REG_NOOP		0x00
>  #define ADS8688_CMD_REG_RST		0x85
> +#define ADS8688_CMD_REG_AUTO_RST	0xA0
>  #define ADS8688_CMD_REG_MAN_CH(chan)	(0xC0 | (4 * chan))
>  #define ADS8688_CMD_DONT_CARE_BITS	16
>  
>  #define ADS8688_PROG_REG(x)		(x << 9)
> +#define ADS8688_AUTO_SEQ_EN		0x01
>  #define ADS8688_PROG_REG_RANGE_CH(chan)	(0x05 + chan)
>  #define ADS8688_PROG_WR_BIT		BIT(8)
>  #define ADS8688_PROG_DONT_CARE_BITS	8
> @@ -210,6 +212,39 @@ static int ads8688_reset(struct iio_dev *indio_dev)
>  	return spi_write(st->spi, &st->data[0].d8[0], 4);
>  }
>  
> +static int ads8688_start_autoseq(struct iio_dev *indio_dev)
> +{
> +	struct ads8688_state *st = iio_priv(indio_dev);
> +	u32 tmp;
> +
> +	tmp = ADS8688_CMD_REG(ADS8688_CMD_REG_AUTO_RST);
> +	tmp <<= ADS8688_CMD_DONT_CARE_BITS;
> +	st->data[0].d32 = cpu_to_be32(tmp);
> +
> +	return spi_write(st->spi, &st->data[0].d8[0], 4);
> +}
> +
> +static int ads8688_read_autoseq(struct iio_dev *indio_dev)
> +{
> +	struct ads8688_state *st = iio_priv(indio_dev);
> +	int ret;
> +	u32 tmp;
> +	struct spi_transfer t = {
> +		.tx_buf = &st->data[0].d8[0],
> +		.len = 4,
> +	};
> +
> +	tmp = ADS8688_CMD_REG(ADS8688_CMD_REG_NOOP);
> +	tmp <<= ADS8688_CMD_DONT_CARE_BITS;
> +	st->data[0].d32 = cpu_to_be32(tmp);
> +
> +	ret = spi_sync_transfer(st->spi, &t, 1);
> +	if (ret < 0)
> +		return ret;
> +
> +	return be32_to_cpu(st->data[0].d32) & 0xffff;
> +}
> +
>  static int ads8688_read(struct iio_dev *indio_dev, unsigned int chan)
>  {
>  	struct ads8688_state *st = iio_priv(indio_dev);
> @@ -251,6 +286,10 @@ static int ads8688_read_raw(struct iio_dev *indio_dev,
>  
>  	struct ads8688_state *st = iio_priv(indio_dev);
>  
> +	/* Block raw reading when in triggered mode */
> +	if (iio_buffer_enabled(indio_dev))
> +		return -EBUSY;
> +
>  	mutex_lock(&st->lock);
>  	switch (m) {
>  	case IIO_CHAN_INFO_RAW:
> @@ -299,6 +338,10 @@ static int ads8688_write_raw(struct iio_dev *indio_dev,
>  	unsigned int scale = 0;
>  	int ret = -EINVAL, i, offset = 0;
>  
> +	/* Block setup when in triggered mode */
> +	if (iio_buffer_enabled(indio_dev))
> +		return -EBUSY;
> +
>  	mutex_lock(&st->lock);
>  	switch (mask) {
>  	case IIO_CHAN_INFO_SCALE:
> @@ -374,10 +417,25 @@ static int ads8688_write_raw_get_fmt(struct iio_dev *indio_dev,
>  	return -EINVAL;
>  }
>  
> +static int ads8688_update_scan_mode(struct iio_dev *indio_dev,
> +	const unsigned long *active_scan_mask)
> +{
> +	int scan_count;
> +
> +	scan_count = bitmap_weight(active_scan_mask, indio_dev->masklength);

What is scan_count for?

> +
> +	ads8688_prog_write(indio_dev, ADS8688_AUTO_SEQ_EN, (int)active_scan_mask);
> +
> +	ads8688_start_autoseq(indio_dev);
> +
> +	return 0;
> +}
> +
>  static const struct iio_info ads8688_info = {
>  	.read_raw = &ads8688_read_raw,
>  	.write_raw = &ads8688_write_raw,
>  	.write_raw_get_fmt = &ads8688_write_raw_get_fmt,
> +	.update_scan_mode = ads8688_update_scan_mode,
>  	.attrs = &ads8688_attribute_group,
>  };
>  
> @@ -391,7 +449,7 @@ static irqreturn_t ads8688_trigger_handler(int irq, void *p)
>  	for (i = 0; i < indio_dev->masklength; i++) {
>  		if (!test_bit(i, indio_dev->active_scan_mask))
>  			continue;
> -		buffer[j] = ads8688_read(indio_dev, i);
> +		buffer[j] = ads8688_read_autoseq(indio_dev);

>  		j++;
>  	}
>  

Re: [PATCH] iio: adc: ti-ads8688: use the auto sequence feature

[Sean Nyekjær]

On 2018-06-03 17:28, Jonathan Cameron wrote:
> On Wed, 30 May 2018 23:18:21 +0200
> Sean Nyekjaer <sean.nyekjaer@prevas.dk> wrote:
> 
>> The TI ADS8688 have a auto sequence feature that allows
>> to continuesly reading new values from the chosen channels.
>> Lets enable it and we will be able to sample at the double speed
>> vs the old implementation
>>
>> Signed-off-by: Sean Nyekjaer <sean.nyekjaer@prevas.dk>
> 
> hmm. Technically it's an ABI change as someone may be relying on the
> ability to do sysfs reads whilst we are also using the chardev interface.
> However, I doubt anyone is so lest see if anyone notices ;)
> 
> One comment inline. Otherwise looks fine to me.
> 
> With care, we could gotten the speed up without using autosequencing
> but that doesn't really matter as that isn't what we were doing previously.
> 
> Thanks,
> 
> Jonathan
> 
>> ---
Oh i din't know that we must always be able to read from sysfs, with 
this adc the auto sequencing will stop if a manual read is done :-(
Maybe I could change it so the auto trigger mode will still be calling 
the channels manually? That way we will have the same amount of 
channels, but greater freedom to be able to do sysfs reads...

>> +static int ads8688_update_scan_mode(struct iio_dev *indio_dev,
>> +	const unsigned long *active_scan_mask)
>> +{
>> +	int scan_count;
>> +
>> +	scan_count = bitmap_weight(active_scan_mask, indio_dev->masklength);
> 
> What is scan_count for?
Woops, just some cp for debugging :-) Will remove it

/Sean

Re: [PATCH] iio: adc: ti-ads8688: use the auto sequence feature

[Jonathan Cameron]

On Mon, 4 Jun 2018 07:21:13 +0200
Sean Nyekj=E6r <sean.nyekjaer@prevas.dk> wrote:

> On 2018-06-03 17:28, Jonathan Cameron wrote:
> > On Wed, 30 May 2018 23:18:21 +0200
> > Sean Nyekjaer <sean.nyekjaer@prevas.dk> wrote:
> >  =20
> >> The TI ADS8688 have a auto sequence feature that allows
> >> to continuesly reading new values from the chosen channels.
> >> Lets enable it and we will be able to sample at the double speed
> >> vs the old implementation
> >>
> >> Signed-off-by: Sean Nyekjaer <sean.nyekjaer@prevas.dk> =20
> >=20
> > hmm. Technically it's an ABI change as someone may be relying on the
> > ability to do sysfs reads whilst we are also using the chardev interfac=
e.
> > However, I doubt anyone is so lest see if anyone notices ;)
> >=20
> > One comment inline. Otherwise looks fine to me.
> >=20
> > With care, we could gotten the speed up without using autosequencing
> > but that doesn't really matter as that isn't what we were doing previou=
sly.
> >=20
> > Thanks,
> >=20
> > Jonathan
> >  =20
> >> --- =20
> Oh i din't know that we must always be able to read from sysfs, with=20
> this adc the auto sequencing will stop if a manual read is done :-(
We don't have to be able to always read from sysfs in general.  The potenti=
al
issue here is that we may have userspace code in the wild using this device
that will expect that to work (because it does before this patch).

Hence we are looking at potential usespace breakage.  However, rule 1 of
changing userspace visible kernel features - if no one notices it isn't
a regression ;)  So here we'll just go for it.  If anyone complains
we may need to do something 'clever' to fake sampling on demand.

> Maybe I could change it so the auto trigger mode will still be calling=20
> the channels manually? That way we will have the same amount of=20
> channels, but greater freedom to be able to do sysfs reads...

The fun would be dealing with channels that aren't enabled.  Either we slow
the device down by always sampling all of the channels or we have it fail
on any that aren't enabled.   This has always been an issue with trying
to support the two modes and is why lots of drivers don't do so (but typica=
lly
they never did so there is no chance of a regression!)


>=20
> >> +static int ads8688_update_scan_mode(struct iio_dev *indio_dev,
> >> +	const unsigned long *active_scan_mask)
> >> +{
> >> +	int scan_count;
> >> +
> >> +	scan_count =3D bitmap_weight(active_scan_mask, indio_dev->masklength=
); =20
> >=20
> > What is scan_count for? =20
> Woops, just some cp for debugging :-) Will remove it

Cool

Jonathan
>=20
> /Sean
> --
> To unsubscribe from this list: send the line "unsubscribe linux-iio" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html