[GSoC 2024] Add support for AD7294 #2578

ArchUsr64 · 2024-08-08T08:58:40Z

PR Description

Adds an IIO driver, devicetree binding and overlay for the AD7294 addac.
Datasheet

TODO

Support for differential channels

PR Type

Bug fix (a change that fixes an issue)
New feature (a change that adds new functionality)
Breaking change (a change that affects other repos or cause CIs to fail)

PR Checklist

I have conducted a self-review of my own code changes
I have tested the changes on the relevant hardware
I have updated the documentation outside this repo accordingly (if there is the case)

machschmitt · 2024-08-14T20:13:35Z

arch/arm/boot/dts/overlays/rpi-ad7294-overlay.dts

+	fragment@2 {
+		target = <&ad7294>;
+		__dormant__ {
+			compatible = "adi,ad7294-2";
+		};


This doesn't make much sense. You just introduced ad7294 node above. Why not have compatible = "adi,ad7294-2"; in ad7294 node? I suggest removing this fragment entirely.

machschmitt · 2024-08-14T20:16:13Z

arch/arm/boot/dts/overlays/rpi-ad7294-overlay.dts

+			#size-cells = <0>;
+			status = "okay";
+			ad7294: ad7294@62 {
+				compatible = "adi,ad7294";


If you have AD7294-2 then just use the compatible for that directly.
compatible = "adi,ad7294-2";

machschmitt · 2024-08-14T20:20:11Z

drivers/iio/addac/Kconfig

@@ -19,6 +19,16 @@ config AD74115
 	  To compile this driver as a module, choose M here: the
 	  module will be called ad74115.

+config AD7294
+	tristate "Analog Devices AD7294/AD7294-2 driver"


Nitpick: I would pick only one device to name the driver. For example, "AD7294 driver" or "AD7294-2 driver". Someone will see the driver supports both devices reading the help text or looking at the driver code.

machschmitt · 2024-08-15T18:10:36Z

drivers/iio/addac/ad7294.c

+#define AD7294_VOLTAGE_CHENNEL_COUNT 4
+
+struct ad7294_state {
+	struct mutex lock;


it is good practice to comment on what concurrency contention mechanisms protect. e.g.

struct mutex lock; /* Protect device write operations */

machschmitt · 2024-08-15T18:34:00Z

drivers/iio/addac/ad7294.c

+	int ret;
+	struct i2c_client *client = context;
+	unsigned char buffer[3] = { reg };
+	int reg_size = ad7294_reg_size(reg);


Nitpick: Linux maintainers often ask to make local variable declarations in reverse xmas tree style.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/maintainer-tip.rst?h=v6.11-rc3#n650
So this would look better in the following order

struct i2c_client *client = context; int reg_size = ad7294_reg_size(reg); unsigned char buffer[3] = { reg }; int ret;

machschmitt · 2024-08-16T00:40:10Z

drivers/iio/addac/ad7294.c

+#define AD7294_VOLTAGE_CHAN(_type, _chan_id) {                \
+	.type = _type,                                        \


Can a voltage channel be of a different type than voltage? If not, I'd suggest removing the _type parameter and just having .type = IIO_VOLTAGE within the macro.
Same for the current channel macro.

machschmitt · 2024-08-16T00:43:34Z

drivers/iio/addac/ad7294.c

+	struct iio_dev *indio_dev = private;
+	struct ad7294_state *st = iio_priv(indio_dev);
+	unsigned int voltage_status, temp_status, current_status;
+	s64 timestamp = iio_get_time_ns(indio_dev);


reverse fir/xmas tree order

machschmitt · 2024-08-16T03:35:03Z

drivers/iio/addac/ad7294.c

+				       IIO_UNMOD_EVENT_CODE(IIO_VOLTAGE, i,
+							    IIO_EV_TYPE_THRESH,
+							    IIO_EV_DIR_FALLING),
+				       timestamp);


Nitpick: align timestamp with the other arguments.

iio_push_event(indio_dev, IIO_UNMOD_EVENT_CODE(IIO_VOLTAGE, i, IIO_EV_TYPE_THRESH, IIO_EV_DIR_FALLING), timestamp);

timestamp is actually the argument for iio_push_event not the IIO_UNMOD_EVENT_CODE macro, should it still be indented as like the arguments for IIO_UNMOD_EVENT_CODE?

iio_push_event(indio_dev, IIO_UNMOD_EVENT_CODE(IIO_VOLTAGE, i, IIO_EV_TYPE_THRESH, IIO_EV_DIR_FALLING), timestamp);

oh, that's correct. Uncommon code pattern this function call has. Sure, forget what I said and keep timestamp aligned with iio_push_event.

machschmitt · 2024-08-16T04:08:07Z

drivers/iio/addac/ad7294.c

+			goto adc_read;
+		case IIO_VOLTAGE:
+			if (chan->output) {
+				*val = st->dac_value[chan->channel];
+				return IIO_VAL_INT;
+			}
+adc_read:
+			ret = regmap_write(st->regmap, AD7294_REG_CMD,
+					   BIT(chan->channel + base));
+			if (ret)
+				return ret;
+			ret = regmap_read(st->regmap, AD7294_REG_RESULT,
+					  &regval);
+			if (ret)
+				return ret;
+			*val = regval & AD7294_ADC_VALUE_MASK;
+			return IIO_VAL_INT;


Jumping from one case to another with goto is uncommon in IIO.
I'd suggest put the common part of both cases into a separate function and call that when needed.

machschmitt · 2024-08-16T04:17:48Z

drivers/iio/addac/ad7294.c

+			if (chan->output) {
+				if (st->dac_vref_reg) {
+					ret = regulator_get_voltage(
+						st->dac_vref_reg);
+					if (ret < 0)
+						return ret;
+					*val = ret / MILLI;
+				} else {
+					*val = AD7294_DAC_INTERNAL_VREF_MV;
+				}
+			} else {
+				if (st->adc_vref_reg) {
+					ret = regulator_get_voltage(
+						st->adc_vref_reg);
+					if (ret < 0)
+						return ret;
+					*val = ret / MILLI;
+				} else {
+					*val = AD7294_ADC_INTERNAL_VREF_MV;
+				}
+			}


It's so many levels of indentation here that you can't even get gain much in readability by breaking regulator_get_voltage() call. Try do the scale read logic in a separate function to reduce overall indentation.

machschmitt

Driver looks pretty good overall. There are some parts that can be improved (mainly code style stuff) and some other changes we discussed this morning but this is already a fairly comprehensive driver.
Keep up the good work @ArchUsr64 and this driver will soon be complete.

machschmitt · 2024-08-16T17:23:30Z

drivers/iio/addac/ad7294.c

+				       IIO_UNMOD_EVENT_CODE(IIO_VOLTAGE, i,
+							    IIO_EV_TYPE_THRESH,
+							    IIO_EV_DIR_FALLING),
+				       timestamp);


oh, that's correct. Uncommon code pattern this function call has. Sure, forget what I said and keep timestamp aligned with iio_push_event.

machschmitt · 2024-08-16T18:29:01Z

drivers/iio/addac/ad7294.c

+#define AD7294_REG_DATA_HIGH(x)	     ((x) * 3 + 0x0C)
+#define AD7294_REG_HYSTERESIS(x)     ((x) * 3 + 0x0D)
+
+#define AD7294_TEMP_ALERT_MASK	     GENMASK(6, 0)


If not handling overtemp alert (bit D6) this can be GENMASK(5, 0)

machschmitt · 2024-08-16T20:37:07Z

drivers/iio/addac/ad7294.c

+#define AD7294_ADC_INTERNAL_VREF_MV  2500
+#define AD7294_DAC_INTERNAL_VREF_MV  2500
+#define AD7294_RESOLUTION	     12
+#define AD7294_VOLTAGE_CHENNEL_COUNT 4


Nitpick: chennel -> channel
#define AD7294_VOLTAGE_CHANNEL_COUNT 4

machschmitt · 2024-08-16T20:54:21Z

drivers/iio/addac/ad7294.c

+	for_each_set_bit(i, (long *)&temp_status, AD7294_TEMP_ALERT_MASK) {
+		if (i & 1)
+			iio_push_event(indio_dev,
+				       IIO_UNMOD_EVENT_CODE(IIO_TEMP, i >> 1,
+							    IIO_EV_TYPE_THRESH,
+							    IIO_EV_DIR_RISING),
+				       timestamp);
+		else
+			iio_push_event(indio_dev,
+				       IIO_UNMOD_EVENT_CODE(IIO_TEMP, i >> 1,
+							    IIO_EV_TYPE_THRESH,
+							    IIO_EV_DIR_FALLING),
+				       timestamp);
+	}


Not ideal to use the mask as for_each_set_bit() size argument. Declare a define for max temperature channels (like you have for the voltage channels) and use that here. Also, to me, it looks like this part of the code doesn't really work. The bit/i local variable will have "the bit number for the next set bit" (0 (if D0), 1(if D1), 3 (if not D2 but D3 set), ...) so (i & 1) will only be true if D0 is 1?

I'm wondering if it would be better to use a normal for statement to traverse the bits from Alert Status Register B and Alert Status Register C. Declare macros similar to AD7294_ALERT_LOW() to check the alert bit. The will be no need to shift i >> 1 and this gets more readable (in my opinion).

machschmitt · 2024-08-16T21:59:20Z

drivers/iio/addac/ad7294.c

+	switch (chan->type) {
+	case IIO_VOLTAGE:
+		offset = chan->channel;
+		break;
+	case IIO_CURRENT:
+		offset = chan->channel + 4;
+		break;
+	case IIO_TEMP:
+		offset = chan->channel + 6;
+		break;
+	default:
+		return 0;
+	}
+
+	switch (info) {
+	case IIO_EV_INFO_VALUE:
+		if (dir == IIO_EV_DIR_FALLING)
+			return AD7294_REG_DATA_LOW(offset);
+		else
+			return AD7294_REG_DATA_HIGH(offset);
+	case IIO_EV_INFO_HYSTERESIS:
+		return AD7294_REG_HYSTERESIS(offset);


I see the effort to reuse code and how it allows concise read/write_event_value() functions but these register offsets can look a bit mysterious at first glance. I would probably declare separate macros for those

#define AD7294_REG_VOLTAGE_DATA_LOW(x) ((x) * 3 + 0x0B) #define AD7294_REG_VOLTAGE_DATA_HIGH(x) ((x) * 3 + 0x0C) #define AD7294_REG_VOLTAGE_HYSTERESIS(x) ((x) * 3 + 0x0D) #define AD7294_REG_CURRENT_DATA_LOW(x) ((x) * 3 + 0x17) #define AD7294_REG_CURRENT_DATA_HIGH(x) ((x) * 3 + 0x18) #define AD7294_REG_CURRENT_HYSTERESIS(x) ((x) * 3 + 0x19) #define AD7294_REG_TEMP_DATA_LOW(x) ((x) * 3 + 0x1D) #define AD7294_REG_TEMP_DATA_HIGH(x) ((x) * 3 + 0x1E) #define AD7294_REG_TEMP_HYSTERESIS(x) ((x) * 3 + 0x1F)

so there would be no extra offset besides the ones used within the macros

switch (chan->type) { case IIO_VOLTAGE: switch (info) { case IIO_EV_INFO_VALUE: if (dir == IIO_EV_DIR_FALLING) return AD7294_REG_VOLTAGE_DATA_LOW(chan->channel); else return AD7294_REG_VOLTAGE_DATA_HIGH(chan->channel); case IIO_EV_INFO_HYSTERESIS: return AD7294_REG_VOLTAGE_HYSTERESIS(chan->channel); break; case IIO_CURRENT: switch (info) { case IIO_EV_INFO_VALUE: if (dir == IIO_EV_DIR_FALLING) return AD7294_REG_CURRENT_DATA_LOW(chan->channel); else return AD7294_REG_CURRENT_DATA_HIGH(chan->channel); case IIO_EV_INFO_HYSTERESIS: return AD7294_REG_CURRENT_HYSTERESIS(chan->channel); case IIO_TEMP: switch (info) { case IIO_EV_INFO_VALUE: if (dir == IIO_EV_DIR_FALLING) return AD7294_REG_TEMP_DATA_LOW(chan->channel); else return AD7294_REG_TEMP_DATA_HIGH(chan->channel); case IIO_EV_INFO_HYSTERESIS: return AD7294_REG_TEMP_HYSTERESIS(chan->channel); break; default: return 0; }

it becomes a bit longer, but it's clearer what register is being returned in each case.

machschmitt · 2024-08-16T22:25:36Z

drivers/iio/addac/ad7294.c

+	if (IS_ERR(st->regmap))
+		return PTR_ERR(st->regmap);
+
+	ret = regmap_read(st->regmap, AD7294_REG_PWDN, &pwdn_config);


We know the default value of power-down register on power up (0x70).
This read would not be needed if using the default value of power-down reg.

#define AD7294_REG_POWER_DOWN_DEFAULT 0x70 ... int pwdn_config = AD7294_REG_POWER_DOWN_DEFAULT;

The default values varies at Bit 6 for AD7294 and AD7294-2, with default values of 0x60 and 0x70 respectively. The bit is reserved so functionality shouldn't be affected regardless. Should I go ahead with 0x70 as default values?

ArchUsr64 added 15 commits August 8, 2024 14:20

iio: addac: ad7294: add basic DAC support

8a90c0c

iio: addac: ad7294: add basic ADC support

c6f1f25

arch: dts: overlays: add ad7294 overlay

fb46b7f

iio: addac: ad7294: add regulator support

acd6a57

iio: addac: ad7294: code refactor

595d123

iio: addac: ad7294: add temperature sensing

1ddcfa3

iio: addac: ad7294: code refactor

6757d47

dt-bindings: iio: addac: add support for AD7294

e3ae35e

iio: addac: ad7294: use regmap for 8-bit registers

a8bd7c3

iio: addac: ad7294: use reg_[read/write] in regmap

d75120e

iio: addac: ad7294: temperature sensing refactor

72f8894

iio: addac: ad7294: add events for handling alerts

cd38a5b

arch: dts: overlays: add interrupts for ad7294

a62a9d8

iio: addac: ad7294: add current scale

6904726

dt-bindings: iio: addac: add shunt-resistor-ohms

d771574

ArchUsr64 requested review from nunojsa, mhennerich, dbogdan, btogorean, chegbeli, buha, amiclaus, rbolboac, danmois, CiprianRegus and machschmitt as code owners August 8, 2024 08:58

iio: addac: ad7294: code refactor

f23d91e

ArchUsr64 force-pushed the rpi-6.1.y branch from e85a7df to f23d91e Compare August 8, 2024 09:32

machschmitt reviewed Aug 16, 2024

View reviewed changes

iio: addac: ad7294: differential channel support

1196a46

machschmitt requested changes Aug 16, 2024

View reviewed changes

ArchUsr64 added 3 commits August 29, 2024 16:52

iio: addac: ad7294: addressed reviewer comments

8905a7f

iio: addac: Kconfig: remove mention of AD7294-2

e1316fc

arch: dts: overlay: update overlay for AD7294-2

aa7ad79

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[GSoC 2024] Add support for AD7294 #2578

[GSoC 2024] Add support for AD7294 #2578

ArchUsr64 commented Aug 8, 2024

machschmitt Aug 14, 2024

machschmitt Aug 14, 2024

machschmitt Aug 14, 2024

machschmitt Aug 15, 2024

machschmitt Aug 15, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

ArchUsr64 Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt left a comment

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

machschmitt Aug 16, 2024

ArchUsr64 Aug 29, 2024

		#define AD7294_VOLTAGE_CHAN(_type, _chan_id) { \
		.type = _type, \

[GSoC 2024] Add support for AD7294 #2578

Are you sure you want to change the base?

[GSoC 2024] Add support for AD7294 #2578

Conversation

ArchUsr64 commented Aug 8, 2024

PR Description

TODO

PR Type

PR Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

machschmitt left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment