Add support for ADAQ776x-1 ADC Family

This adds support for the ADAQ7767-1, ADAQ7768-1 and ADAQ7769-1 devices.

The ADAQ7768-1 and ADAQ7769-1 integrate a programmable gain amplifier (PGA) with 7 and 8 gain options, respectively. The ADAQ7767-1 and ADAQ7769-1 also feature a 3-pin selectable Anti-aliasing filter (AAF) gain.

---
Changes in v5:
* Added PERCENT macro along with BASIS_POINTS in units.h.
* Added new patch to allow writing attributes without direct mode claim.
* Write scale attribute without direct mode claim to avoid deadlock
when using GPIOs from the device's own controller.
* Link to v4: https://lore.kernel.org/linux-iio/[email protected]/T/#t

Changes in v4:
* Removed u64_fract type patch, as it is not used. u32_fract is sufficient
for the PGA gain calculations.
* Added new patch adding BASIS_POINTS macro to units.h.
* Included pga-gpios property for parts that support PGA gain.
Before we were using the internal gpio controller to manage the PGA pins,
but still exposing the controller for external use (causing possible conflicts).
* Addressed other review comments, see individual patches.
* Link to v3: https://lore.kernel.org/linux-iio/[email protected]/T/#t

Changes in v3:
* Renamed adi,gain-milli to adi,aaf-gain-bp. Now it represents basis points (one hundredth of a percent).
* ad7768_channel_masks removed along with available_masks element in ad7768_chip_info struct. It does not add anything for single channels, so not needed, at least for now.
* New patch adding 64-bit fractional number types to math.h.
* Moved aaf gain parsing to its own function, and now returning after warning to avoid setting a variable when it shouldn't (avoid confusion).
* ad7768_set_pga_gain(): removed the pgia enable check, relying on the regmap cache.
* Addressed other review comments, see individual patches.

Changes in v2:
* adi,aaf-gain property renamed to adi,gain-milli. Default value added.
* fixed some commit messages.
* Added 'select RATIONAL' to Kconfig.
* Added lock to protect PGA value access.
* rewrote AAF gain check and replaced error returns with warnings.
* Addressed other review comments, see individual patches.
* Link to v1: https://lore.kernel.org/linux-iio/[email protected]/T/#t
---

Jonathan Santos (5):
dt-bindings: iio: adc: ad7768-1: add new supported parts
iio: adc: ad7768-1: introduce chip info for future multidevice support
units: add PERCENT and BASIS_POINTS macros
iio: adc: ad7768-1: squash __ad7768_write_raw() back to

    ad7768_write_raw()

iio: adc: ad7768-1: add support for ADAQ776x-1 ADC Family

 .../bindings/iio/adc/adi,ad7768-1.yaml        |  64 ++-
 drivers/iio/adc/Kconfig                       |   1 +
 drivers/iio/adc/ad7768-1.c                    | 428 ++++++++++++++++--
 include/linux/units.h                         |  14 +
 4 files changed, 453 insertions(+), 54 deletions(-)

base-commit: daea3a394a8b425a2dd206ab09eb37f0d1087d35

2.34.1

Add compatibles for supported parts in the ad7768-1 family:

	ADAQ7767-1, ADAQ7768-1 and ADAQ7769-1

Add property and checks for AFF gain, supported by ADAQ7767-1 and ADAQ7769-1, and for PGA gain, supported by ADAQ7768-1 and ADAQ7769-1:
adi,aaf-gain-bp
pga-gpios

---
v5 Changes:
* None.

v4 Changes:
* Inspired by [1], Included pga-gpios property for parts that support PGA gain. Before we were using the internal gpio controller to manage the PGA pins, but still exposing the controller for external use (causing possible conflicts). Using pga-gpios we can let the consumer define the pins to be used for PGA, even from the internal gpio controller (but not limited by that). The problem with that approach is the deadlock described in the last patch from this set.

v3 Changes:
* Renamed adi,gain-milli to adi,aaf-gain-bp. Now it represents basis points (one hundredth of a percent) as suggested by Krzysztof. Description was adjusted.
Note: permille (1/1000) was also suggested as unit for this property.

v2 Changes:
* adi,aaf-gain property renamed to adi,gain-milli. Description was simplified.
* default value add to adi,gain-milli.

[1]: https://lore.kernel.org/linux-iio/318c31e023ebe30cc99b8743e87e869bf5e1f12b.1760984107.git.marcelo.schmitt@analog.com/

---
 .../bindings/iio/adc/adi,ad7768-1.yaml        | 64 +++++++++++++++++--
 1 file changed, 60 insertions(+), 4 deletions(-)

index c06d0fc791d3..dfa2d7fa9fb3 100644
@@ -4,18 +4,26 @@
 $id: http://devicetree.org/schemas/iio/adc/adi,ad7768-1.yaml#
 $schema: http://devicetree.org/meta-schemas/core.yaml#
 
-title: Analog Devices AD7768-1 ADC device driver
+title: Analog Devices AD7768-1 ADC family
 
 maintainers:
   - Michael Hennerich <[email protected]>
 
 description: |
-  Datasheet at:
-    https://www.analog.com/media/en/technical-documentation/data-sheets/ad7768-1.pdf
+  Analog Devices AD7768-1 24-Bit Single Channel Low Power sigma-delta ADC family
+
+  https://www.analog.com/media/en/technical-documentation/data-sheets/ad7768-1.pdf
+  https://www.analog.com/media/en/technical-documentation/data-sheets/adaq7767-1.pdf
+  https://www.analog.com/media/en/technical-documentation/data-sheets/adaq7768-1.pdf
+  https://www.analog.com/media/en/technical-documentation/data-sheets/adaq7769-1.pdf
 
 properties:
   compatible:
-    const: adi,ad7768-1
+    enum:
+      - adi,ad7768-1
+      - adi,adaq7767-1
+      - adi,adaq7768-1
+      - adi,adaq7769-1
 
   reg:
     maxItems: 1

@@ -58,6 +66,25 @@ properties:
     description:
       ADC reference voltage supply
 
+  adi,aaf-gain-bp:
+    description: |
+       Specifies the gain applied by the Analog Anti-Aliasing Filter (AAF)
+       to the ADC input in basis points (one hundredth of a percent).
+       The hardware gain is determined by which input pin(s) the signal goes
+       through into the AAF. The possible connections are:
+       * For the ADAQ7767-1: Input connected to IN1±, IN2± or IN3±.
+       * For the ADAQ7769-1: OUT_PGA pin connected to IN1_AAF+, IN2_AAF+,
+         or IN3_AAF+.
+    enum: [1430, 3640, 10000]
+    default: 10000
+
+  pga-gpios:
+    description:
+      GAIN 0, GAIN1 and GAIN2 pins for gain selection. For devices that have
+      PGA configuration input pins, pga-gpios must be defined.
+    minItems: 3
+    maxItems: 3
+
   adi,sync-in-gpios:
     maxItems: 1
     description:

@@ -147,6 +174,35 @@ patternProperties:
 allOf:
   - $ref: /schemas/spi/spi-peripheral-props.yaml#
 
+  # AAF Gain property only applies to ADAQ7767-1 and ADAQ7769-1 devices
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - adi,adaq7767-1
+              - adi,adaq7769-1
+    then:
+      required:
+        - adi,aaf-gain-bp
+    else:
+      properties:
+        adi,aaf-gain-bp: false
+
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - adi,adaq7768-1
+              - adi,adaq7769-1
+    then:
+      required:
+        - pga-gpios
+    else:
+      properties:
+        pga-gpios: false
+
 unevaluatedProperties: false
 
 examples:

2.34.1

Add Chip info struct in SPI device to store channel information for each supported part.

---
v5 Changes:
* None.

v4 Changes:
* Dropped chip info check and moved the st->chip assignment early in the probe function.
* Addressed indentation inconsistencies in AD7768_CHAN macro.

v3 Changes:
* ad7768_channel_masks removed along with available_masks element in ad7768_chip_info struct. It does not add anything for single channels, so not needed, at least for now.
* fixed inconsistency in spaces before \ in AD7768_CHAN macro.

v2 Changes:
* removed AD7768_CHAN_INFO_NONE macro.
* reordered fields in ad7768_chip_info struct.
* removed trailing comma.

---
 drivers/iio/adc/ad7768-1.c | 64 ++++++++++++++++++++++++--------------
 1 file changed, 41 insertions(+), 23 deletions(-)

index d96802b7847a..89b0ca8f584c 100644
@@ -213,6 +213,12 @@ static const struct iio_scan_type ad7768_scan_type[] = {
 	},
 };
 
+struct ad7768_chip_info {
+	const char *name;
+	const struct iio_chan_spec *channel_spec;
+	int num_channels;
+};
+
 struct ad7768_state {
 	struct spi_device *spi;
 	struct regmap *regmap;

@@ -234,6 +240,7 @@ struct ad7768_state {
 	struct gpio_desc *gpio_reset;
 	const char *labels[AD7768_MAX_CHANNELS];
 	struct gpio_chip gpiochip;
+	const struct ad7768_chip_info *chip;
 	bool en_spi_sync;
 	/*
 	 * DMA (thus cache coherency maintenance) may require the

@@ -748,24 +755,28 @@ static const struct iio_chan_spec_ext_info ad7768_ext_info[] = {
 	{ }
 };
 
+#define AD7768_CHAN(_idx, _msk_avail) \
+{ \
+	.type = IIO_VOLTAGE, \
+	.info_mask_separate_available = _msk_avail, \
+	.info_mask_separate = BIT(IIO_CHAN_INFO_RAW), \
+	.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE) | \
+				    BIT(IIO_CHAN_INFO_LOW_PASS_FILTER_3DB_FREQUENCY) | \
+				    BIT(IIO_CHAN_INFO_OVERSAMPLING_RATIO), \
+	.info_mask_shared_by_type_available = BIT(IIO_CHAN_INFO_OVERSAMPLING_RATIO), \
+	.info_mask_shared_by_all = BIT(IIO_CHAN_INFO_SAMP_FREQ), \
+	.info_mask_shared_by_all_available = BIT(IIO_CHAN_INFO_SAMP_FREQ), \
+	.ext_info = ad7768_ext_info, \
+	.indexed = 1, \
+	.channel = _idx, \
+	.scan_index = _idx, \
+	.has_ext_scan_type = 1, \
+	.ext_scan_type = ad7768_scan_type, \
+	.num_ext_scan_type = ARRAY_SIZE(ad7768_scan_type), \
+}
+
 static const struct iio_chan_spec ad7768_channels[] = {
-	{
-		.type = IIO_VOLTAGE,
-		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
-		.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE) |
-					    BIT(IIO_CHAN_INFO_LOW_PASS_FILTER_3DB_FREQUENCY) |
-					    BIT(IIO_CHAN_INFO_OVERSAMPLING_RATIO),
-		.info_mask_shared_by_type_available = BIT(IIO_CHAN_INFO_OVERSAMPLING_RATIO),
-		.info_mask_shared_by_all = BIT(IIO_CHAN_INFO_SAMP_FREQ),
-		.info_mask_shared_by_all_available = BIT(IIO_CHAN_INFO_SAMP_FREQ),
-		.ext_info = ad7768_ext_info,
-		.indexed = 1,
-		.channel = 0,
-		.scan_index = 0,
-		.has_ext_scan_type = 1,
-		.ext_scan_type = ad7768_scan_type,
-		.num_ext_scan_type = ARRAY_SIZE(ad7768_scan_type),
-	},
+	AD7768_CHAN(0, 0),
 };
 
 static int ad7768_read_raw(struct iio_dev *indio_dev,

@@ -1321,6 +1332,12 @@ static int ad7768_register_regulators(struct device *dev, struct ad7768_state *s
 	return 0;
 }
 
+static const struct ad7768_chip_info ad7768_chip_info = {
+	.name = "ad7768-1",
+	.channel_spec = ad7768_channels,
+	.num_channels = ARRAY_SIZE(ad7768_channels),
+};
+
 static int ad7768_probe(struct spi_device *spi)
 {
 	struct ad7768_state *st;

@@ -1347,6 +1364,7 @@ static int ad7768_probe(struct spi_device *spi)
 			return ret;
 	}
 
+	st->chip = spi_get_device_match_data(spi);
 	st->spi = spi;
 
 	st->regmap = devm_regmap_init_spi(spi, &ad7768_regmap_config);

@@ -1371,9 +1389,9 @@ static int ad7768_probe(struct spi_device *spi)
 
 	st->mclk_freq = clk_get_rate(st->mclk);
 
-	indio_dev->channels = ad7768_channels;
-	indio_dev->num_channels = ARRAY_SIZE(ad7768_channels);
-	indio_dev->name = spi_get_device_id(spi)->name;
+	indio_dev->channels = st->chip->channel_spec;
+	indio_dev->num_channels = st->chip->num_channels;
+	indio_dev->name = st->chip->name;
 	indio_dev->info = &ad7768_info;
 	indio_dev->modes = INDIO_DIRECT_MODE;
 

@@ -1390,7 +1408,7 @@ static int ad7768_probe(struct spi_device *spi)
 
 	init_completion(&st->completion);
 
-	ret = ad7768_set_channel_label(indio_dev, ARRAY_SIZE(ad7768_channels));
+	ret = ad7768_set_channel_label(indio_dev, st->chip->num_channels);
 	if (ret)
 		return ret;
 

@@ -1409,13 +1427,13 @@ static int ad7768_probe(struct spi_device *spi)
 }
 
 static const struct spi_device_id ad7768_id_table[] = {
-	{ "ad7768-1", 0 },
+	{ "ad7768-1", (kernel_ulong_t)&ad7768_chip_info },
 	{ }
 };
 MODULE_DEVICE_TABLE(spi, ad7768_id_table);
 
 static const struct of_device_id ad7768_of_match[] = {
-	{ .compatible = "adi,ad7768-1" },
+	{ .compatible = "adi,ad7768-1", .data = &ad7768_chip_info },
 	{ }
 };
 MODULE_DEVICE_TABLE(of, ad7768_of_match);

2.34.1

Add macros for percentage related units, with basis points defined as 1/100th of a percent. Basis points are commonly used in finance and engineering to express small percentage changes with precision.

---
v5 Changes:
* Included PERCENT macro along with BASIS_POINTS.
* Adjusted commit description and comment in the code to add more context and examples.

v4 Changes:
* New patch.

---
 include/linux/units.h | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

index 00e15de33eca..9c2fbcf04c81 100644
@@ -21,6 +21,20 @@
 #define PICO	1000000000000ULL
 #define FEMTO	1000000000000000ULL
 
+/*
+ * Percentage and basis point units
+ *
+ * Basis points are 1/100th of a percent (1/100), commonly used in finance,
+ * engineering or other applications that require precise percentage
+ * calculations.
+ *
+ * Examples:
+ *   100% = 10000 basis points = BASIS_POINTS
+ *   1%   = 100 basis points   = PERCENT
+ */
+#define PERCENT		100UL
+#define BASIS_POINTS	10000UL
+
 #define NANOHZ_PER_HZ		1000000000UL
 #define MICROHZ_PER_HZ		1000000UL
 #define MILLIHZ_PER_HZ		1000UL

2.34.1

...

Be careful with UL. What would be 10% from -555 (minus five hundreds fifty five)?

With Best Regards,
Andy Shevchenko

...

Also there are pp (percentage points) and others like per mille

https://en.wikipedia.org/wiki/Percentage_point (see "Related units" section).

With Best Regards,
Andy Shevchenko

...

And given most of them are in the PER* namespace I would use those names instead of variants.

With Best Regards,
Andy Shevchenko

Hi, Andy. Thanks for the suggestions,but I am still questioning how the percentage points would be implemented, since it is defined as the difference between two percentages and we cannot use floating points.

Also, should we keep this approach of unit conversion? Example:
value / PERCENT -> converts a value from percent to ratio value * PERCENT -> converts a value from ratio to percent

Just to make sure it is consistent with the rest.

...

My comment was more about the prefix (PER) and the naming (not BASIS_POINTS, but something from PER*), and for the comprehension I would rather see all three defined (pp is _not_ part of these as you mentioned its semantics is different).

Sounds good at the first glance.

With Best Regards,
Andy Shevchenko

Squash __ad7768_write_raw() back to ad7768_write_raw() to allow the addition of new attributes without requiring a direct mode claim.

---
v5 Changes:
* new patch suggested by Jonathan Cameron to allow attributes without the direct mode claim, necessary for fixing the deadlock issue reported in v4.

---
 drivers/iio/adc/ad7768-1.c | 50 ++++++++++++++++++++------------------
 1 file changed, 26 insertions(+), 24 deletions(-)

index 89b0ca8f584c..bd4b2e090c5b 100644
@@ -742,6 +742,19 @@ static int ad7768_get_filter_type_attr(struct iio_dev *dev,
 	return ad7768_filter_regval_to_type[FIELD_GET(mask, mode)];
 }
 
+static int ad7768_update_dec_rate(struct iio_dev *dev, unsigned int dec_rate)
+{
+	struct ad7768_state *st = iio_priv(dev);
+	int ret;
+
+	ret = ad7768_configure_dig_fil(dev, st->filter_type, dec_rate);
+	if (ret)
+		return ret;
+
+	/* Update sampling frequency */
+	return ad7768_set_freq(st, st->samp_freq);
+}
+
 static const struct iio_enum ad7768_filter_type_iio_enum = {
 	.items = ad7768_filter_enum,
 	.num_items = ARRAY_SIZE(ad7768_filter_enum),

@@ -867,44 +880,33 @@ static int ad7768_read_avail(struct iio_dev *indio_dev,
 	}
 }
 
-static int __ad7768_write_raw(struct iio_dev *indio_dev,
-			      struct iio_chan_spec const *chan,
-			      int val, int val2, long info)
+static int ad7768_write_raw(struct iio_dev *indio_dev,
+			    struct iio_chan_spec const *chan,
+			    int val, int val2, long info)
 {
 	struct ad7768_state *st = iio_priv(indio_dev);
 	int ret;
 
 	switch (info) {
 	case IIO_CHAN_INFO_SAMP_FREQ:
-		return ad7768_set_freq(st, val);
+		if (!iio_device_claim_direct(indio_dev))
+			return -EBUSY;
 
+		ret = ad7768_set_freq(st, val);
+		iio_device_release_direct(indio_dev);
+		return ret;
 	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
-		ret = ad7768_configure_dig_fil(indio_dev, st->filter_type, val);
-		if (ret)
-			return ret;
+		if (!iio_device_claim_direct(indio_dev))
+			return -EBUSY;
 
-		/* Update sampling frequency */
-		return ad7768_set_freq(st, st->samp_freq);
+		ret = ad7768_update_dec_rate(indio_dev, val);
+		iio_device_release_direct(indio_dev);
+		return ret;
 	default:
 		return -EINVAL;
 	}
 }
 
-static int ad7768_write_raw(struct iio_dev *indio_dev,
-			    struct iio_chan_spec const *chan,
-			    int val, int val2, long info)
-{
-	int ret;
-
-	if (!iio_device_claim_direct(indio_dev))
-		return -EBUSY;
-
-	ret = __ad7768_write_raw(indio_dev, chan, val, val2, info);
-	iio_device_release_direct(indio_dev);
-
-	return ret;
-}
-
 static int ad7768_read_label(struct iio_dev *indio_dev,
 	const struct iio_chan_spec *chan, char *label)
 {

2.34.1

Add support for ADAQ7767/68/69-1 series, which includes PGIA and Anti-aliasing filter (AAF) gains. Unlike the AD7768-1, they do not provide a VCM regulator interface.

The PGA gain is configured in run-time through the scale attribute, if supported by the device. PGA is controlled by GPIOs provided in the device tree.

The AAF gain is defined by hardware connections and should be specified in the device tree.

---
v5 Changes:
* Write scale attribute without direct mode claim to avoid deadlock when using GPIOs from the device's own controller.

v4 Changes:
* replaced shift_right() with '>>' operator in the ad7768_fill_scale_tbl() function.
* Refactored ad7768_parse_aaf_gain () as requested.
* renamed ad7768_register_regulators() to ad7768_register_vcm_regulator() to better reflect its purpose (not sure if this is ok to do).
* Replaced u64_fract with u32_fract -> after reviewing the numbers again, I realized that u32_fract is sufficient for these calculations.
* addressed minor suggestions.
* Adopted a new approach to manage the PGA GPIOs, using pga-gpios property. This avoids possible conflicts when the internal gpio controller is used externally (and also allows hardwiring, as soon as the gpio interface supports it). However, using GPIOs from the device's own controller causes a deadlock when claiming direct mode in the ad7768_gpio_get() function. This happens because the direct mode remains locked by the ad7768_write_raw() function. I have kept this approach for now to discuss a way around this problem. It would be good to have the flexibility provided by pga-gpios property.

v3 Changes:
* Fixed trailing comma issues.
* Addressed other minor issues related to dead code, variable declaration, etc.
* removed unnecessary comments and relocating some local variables.
* replaced mutex_init() with devm_mutex_init().
* adopted different variables for the input and output of rational_best_approximation(). Also used a u64_fract for the inputs, but kept the unsigned long for the outputs, because could not create a unsigned long fraction number type.
* ad7768_set_pga_gain(): removed the pgia enable check, relying on the regmap cache.
* Moved aaf gain parsing to its own function, and now returning after warning to avoid setting a variable when it shouldn't (avoid confusion).
* AAF gain is now in basis point units, so related multipliers and dividers are adjusted accordingly.

v2 Changes:
* Added more details to the commit message.
* Some devices does not provide VCM regulator, so a new field in the chip info struct was added to indicate this.
* Added 'select RATIONAL' to Kconfig. Kernel test robot pointed out compilation error due to undefined reference to rational_best_approximation().
* Added lock to protect PGA value access.
* precision in the PGA calculation is now dependent of the channel sign (signed or unsigned).
* went back to the original scale computation: (st->vref_uv * 2) / 2^n instead of st->vref_uv / 2^(n-1).
* rewrote AAF gain check and replaced error returns with warnings. I the AAF gain is not provided, a default value is used.
* Addressed other minor suggestions.

---
 drivers/iio/adc/Kconfig    |   1 +
 drivers/iio/adc/ad7768-1.c | 316 ++++++++++++++++++++++++++++++++++++-
 2 files changed, 313 insertions(+), 4 deletions(-)

index 9c4c1e23090a..e5794cb26da9 100644
@@ -387,6 +387,7 @@ config AD7768_1
 	depends on SPI
 	select REGULATOR
 	select REGMAP_SPI
+	select RATIONAL
 	select IIO_BUFFER
 	select IIO_TRIGGER
 	select IIO_TRIGGERED_BUFFER

2.34.1

Hi Jonathan

I noted one minor area where I think the code could be slightly cleaner. If we didn't have the outstanding question about the comment in the BP definitions patch I'd just have merged it as it stands. I don't mind that much if you prefer the current form.

Jonathan

	if (ret == -EINVAL) {
		/* If controllable, use default */
		if (st->chip->has_variable_aaf)
			st->aaf_gain = AD7768_AAF_IN1;
		return 0;
	}
	if (ret)
		return dev_err_probe(dev, ret, "Failed to get AAF gain value\n");

	if (!st->chip->has_variable_aaf)
		return dev_err_probe(dev,, -EOPNOTSUPP,
		     "AAF gain provided, but variable AFF gain not supported for %s\n", ...)

	if (ret == -EINVAL) {
		if (!st->chip->has_variable_aaf)
			return 0;

		val = 10000; /* Matches the default from DT */
	} else if (ret) {
		return dev_err_probe(dev, ret, "Failed to get AAF gain value\n");
	} else if (!st->chip->has_variable_aaf) {
		return dev_err_probe(dev,, -EOPNOTSUPP,
		     "AAF gain provided, but variable AFF gain not supported for %s\n", ...)
	}