Skip to contentSkip to conversation

iio: adc: ad7191: Don't check for specific errors when parsing properties

PATCH In review Andy Shevchenko 4 participants 9 messages Feb 19 – Apr 14

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: adc: ad7191: Don't check for specific errors when parsing properties

Instead of checking for the specific error codes (that can be considered a layering violation to some extent) check for the property existence first and then either parse it, or apply a default value.

---
 drivers/iio/adc/ad7191.c | 63 +++++++++++++++++++++-------------------
 1 file changed, 33 insertions(+), 30 deletions(-)
drivers/iio/adc/ad7191.c +33 -30
index d9cd903ffdd2..51ec199fb06f 100644
@@ -154,27 +154,18 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
 	const u32 gain[4] = { 1, 8, 64, 128 };
 	static u32 scale_buffer[4][2];
 	int odr_value, odr_index = 0, pga_value, pga_index = 0, i, ret;
+	const char *propname;
 	u64 scale_uv;
 
 	st->samp_freq_index = 0;
 	st->scale_index = 0;
 
-	ret = device_property_read_u32(dev, "adi,odr-value", &odr_value);
-	if (ret && ret != -EINVAL)
-		return dev_err_probe(dev, ret, "Failed to get odr value.\n");
+	propname = "adi,odr-value";
+	if (device_property_present(dev, propname)) {
+		ret = device_property_read_u32(dev, propname, &odr_value);
+		if (ret)
+			return dev_err_probe(dev, ret, "Failed to get %s.\n", propname);
 
-	if (ret == -EINVAL) {
-		st->odr_gpios = devm_gpiod_get_array(dev, "odr", GPIOD_OUT_LOW);
-		if (IS_ERR(st->odr_gpios))
-			return dev_err_probe(dev, PTR_ERR(st->odr_gpios),
-					     "Failed to get odr gpios.\n");
-
-		if (st->odr_gpios->ndescs != 2)
-			return dev_err_probe(dev, -EINVAL, "Expected 2 odr gpio pins.\n");
-
-		st->samp_freq_avail = samp_freq;
-		st->samp_freq_avail_size = ARRAY_SIZE(samp_freq);
-	} else {
 		for (i = 0; i < ARRAY_SIZE(samp_freq); i++) {
 			if (odr_value != samp_freq[i])
 				continue;
@@ -186,6 +177,17 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
 		st->samp_freq_avail_size = 1;
 
 		st->odr_gpios = NULL;
+	} else {
+		st->odr_gpios = devm_gpiod_get_array(dev, "odr", GPIOD_OUT_LOW);
+		if (IS_ERR(st->odr_gpios))
+			return dev_err_probe(dev, PTR_ERR(st->odr_gpios),
+					     "Failed to get odr gpios.\n");
+
+		if (st->odr_gpios->ndescs != 2)
+			return dev_err_probe(dev, -EINVAL, "Expected 2 odr gpio pins.\n");
+
+		st->samp_freq_avail = samp_freq;
+		st->samp_freq_avail_size = ARRAY_SIZE(samp_freq);
 	}
 
 	mutex_lock(&st->lock);
@@ -200,22 +202,12 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
 
 	mutex_unlock(&st->lock);
 
-	ret = device_property_read_u32(dev, "adi,pga-value", &pga_value);
-	if (ret && ret != -EINVAL)
-		return dev_err_probe(dev, ret, "Failed to get pga value.\n");
+	propname = "adi,pga-value";
+	if (device_property_present(dev, propname)) {
+		ret = device_property_read_u32(dev, propname, &pga_value);
+		if (ret)
+			return dev_err_probe(dev, ret, "Failed to get %s.\n", propname);
 
-	if (ret == -EINVAL) {
-		st->pga_gpios = devm_gpiod_get_array(dev, "pga", GPIOD_OUT_LOW);
-		if (IS_ERR(st->pga_gpios))
-			return dev_err_probe(dev, PTR_ERR(st->pga_gpios),
-					     "Failed to get pga gpios.\n");
-
-		if (st->pga_gpios->ndescs != 2)
-			return dev_err_probe(dev, -EINVAL, "Expected 2 pga gpio pins.\n");
-
-		st->scale_avail = scale_buffer;
-		st->scale_avail_size = ARRAY_SIZE(scale_buffer);
-	} else {
 		for (i = 0; i < ARRAY_SIZE(gain); i++) {
 			if (pga_value != gain[i])
 				continue;
@@ -227,6 +219,17 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
 		st->scale_avail_size = 1;
 
 		st->pga_gpios = NULL;
+	} else {
+		st->pga_gpios = devm_gpiod_get_array(dev, "pga", GPIOD_OUT_LOW);
+		if (IS_ERR(st->pga_gpios))
+			return dev_err_probe(dev, PTR_ERR(st->pga_gpios),
+					     "Failed to get pga gpios.\n");
+
+		if (st->pga_gpios->ndescs != 2)
+			return dev_err_probe(dev, -EINVAL, "Expected 2 pga gpio pins.\n");
+
+		st->scale_avail = scale_buffer;
+		st->scale_avail_size = ARRAY_SIZE(scale_buffer);
 	}
 
 	st->temp_gpio = devm_gpiod_get(dev, "temp", GPIOD_OUT_LOW);
Signature · 1 line
2.50.1
#1 lore raw
Quoted 6 lines · Andy Shevchenko

Instead of checking for the specific error codes (that can be considered a layering violation to some extent) check for the property existence first and then either parse it, or apply a default value.

---

Not really sure how I feel about this one. Checking for specific errors is a very common pattern and this change just makes it we check for the property presence twice. That said, this makes it more "future proof" (though I find it very unlikely for ret value o change).

Anyways, even if we choose to go down this route, I don't see much benefit in starting converting the drivers with the pattern below (which should be a considerable number).

- Nuno Sá

Quoted 105 lines

 drivers/iio/adc/ad7191.c | 63 +++++++++++++++++++++-------------------

 1 file changed, 33 insertions(+), 30 deletions(-)
drivers/iio/adc/ad7191.c +1 -0
index d9cd903ffdd2..51ec199fb06f 100644
@@ -154,27 +154,18 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
  	const u32 gain[4] = { 1, 8, 64, 128 };
  	static u32 scale_buffer[4][2];
  	int odr_value, odr_index = 0, pga_value, pga_index = 0, i, ret;
+	const char *propname;
  	u64 scale_uv;


  st->samp_freq_index = 0;
  st->scale_index = 0;

drivers/iio/adc/ad7191.c +5 -3
-	ret = device_property_read_u32(dev, "adi,odr-value", &odr_value);
-	if (ret && ret != -EINVAL)
-		return dev_err_probe(dev, ret, "Failed to get odr value.\n");
+	propname = "adi,odr-value";
+	if (device_property_present(dev, propname)) {
+		ret = device_property_read_u32(dev, propname, &odr_value);
+		if (ret)
+			return dev_err_probe(dev, ret, "Failed to get %s.\n", propname);

drivers/iio/adc/ad7191.c +0 -12
-	if (ret == -EINVAL) {
-		st->odr_gpios = devm_gpiod_get_array(dev, "odr", GPIOD_OUT_LOW);
-		if (IS_ERR(st->odr_gpios))
-			return dev_err_probe(dev, PTR_ERR(st->odr_gpios),
-					     "Failed to get odr gpios.\n");
-
-		if (st->odr_gpios->ndescs != 2)
-			return dev_err_probe(dev, -EINVAL, "Expected 2 odr gpio pins.\n");
-
-		st->samp_freq_avail = samp_freq;
-		st->samp_freq_avail_size = ARRAY_SIZE(samp_freq);
-	} else {

  for (i = 0; i < ARRAY_SIZE(samp_freq); i++) { if (odr_value != samp_freq[i]) continue;

drivers/iio/adc/ad7191.c +0 -0
@@ -186,6 +177,17 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
  		st->samp_freq_avail_size = 1;


  st->odr_gpios = NULL;

drivers/iio/adc/ad7191.c +11 -0
+	} else {
+		st->odr_gpios = devm_gpiod_get_array(dev, "odr", GPIOD_OUT_LOW);
+		if (IS_ERR(st->odr_gpios))
+			return dev_err_probe(dev, PTR_ERR(st->odr_gpios),
+					     "Failed to get odr gpios.\n");
+
+		if (st->odr_gpios->ndescs != 2)
+			return dev_err_probe(dev, -EINVAL, "Expected 2 odr gpio pins.\n");
+
+		st->samp_freq_avail = samp_freq;
+		st->samp_freq_avail_size = ARRAY_SIZE(samp_freq);

  }

  mutex_lock(&st->lock);

drivers/iio/adc/ad7191.c +0 -0
@@ -200,22 +202,12 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)


  mutex_unlock(&st->lock);

drivers/iio/adc/ad7191.c +5 -3
-	ret = device_property_read_u32(dev, "adi,pga-value", &pga_value);
-	if (ret && ret != -EINVAL)
-		return dev_err_probe(dev, ret, "Failed to get pga value.\n");
+	propname = "adi,pga-value";
+	if (device_property_present(dev, propname)) {
+		ret = device_property_read_u32(dev, propname, &pga_value);
+		if (ret)
+			return dev_err_probe(dev, ret, "Failed to get %s.\n", propname);

drivers/iio/adc/ad7191.c +0 -12
-	if (ret == -EINVAL) {
-		st->pga_gpios = devm_gpiod_get_array(dev, "pga", GPIOD_OUT_LOW);
-		if (IS_ERR(st->pga_gpios))
-			return dev_err_probe(dev, PTR_ERR(st->pga_gpios),
-					     "Failed to get pga gpios.\n");
-
-		if (st->pga_gpios->ndescs != 2)
-			return dev_err_probe(dev, -EINVAL, "Expected 2 pga gpio pins.\n");
-
-		st->scale_avail = scale_buffer;
-		st->scale_avail_size = ARRAY_SIZE(scale_buffer);
-	} else {

  for (i = 0; i < ARRAY_SIZE(gain); i++) { if (pga_value != gain[i]) continue;

drivers/iio/adc/ad7191.c +0 -0
@@ -227,6 +219,17 @@ static int ad7191_config_setup(struct iio_dev *indio_dev)
  		st->scale_avail_size = 1;


  st->pga_gpios = NULL;

drivers/iio/adc/ad7191.c +11 -0
+	} else {
+		st->pga_gpios = devm_gpiod_get_array(dev, "pga", GPIOD_OUT_LOW);
+		if (IS_ERR(st->pga_gpios))
+			return dev_err_probe(dev, PTR_ERR(st->pga_gpios),
+					     "Failed to get pga gpios.\n");
+
+		if (st->pga_gpios->ndescs != 2)
+			return dev_err_probe(dev, -EINVAL, "Expected 2 pga gpio pins.\n");
+
+		st->scale_avail = scale_buffer;
+		st->scale_avail_size = ARRAY_SIZE(scale_buffer);

  }

  st->temp_gpio = devm_gpiod_get(dev, "temp", GPIOD_OUT_LOW);

#2 lore raw
Quoted 8 lines · Nuno Sá
Quoted 3 lines · Andy Shevchenko

Instead of checking for the specific error codes (that can be considered a layering violation to some extent) check for the property existence first and then either parse it, or apply a default value.

Not really sure how I feel about this one. Checking for specific errors is a very common pattern and this change just makes it we check for the property presence twice. That said, this makes it more "future proof" (though I find it very unlikely for ret value o change).

I already have an answer to this:
https://lore.kernel.org/r/[email protected]

Quoted 2 lines

Anyways, even if we choose to go down this route, I don't see much benefit in starting converting the drivers with the pattern below (which should be a considerable number).

There is not a big number of them, so I prefer to have common patterns without exact error code checks.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 6 lines · Andy Shevchenko
Quoted 5 lines · Nuno Sá
Quoted 3 lines · Andy Shevchenko

Instead of checking for the specific error codes (that can be considered a layering violation to some extent) check for the property existence first and then either parse it, or apply a default value.

Quoted 6 lines
Quoted 3 lines

Not really sure how I feel about this one. Checking for specific errors is a very common pattern and this change just makes it we check for the property presence twice. That said, this makes it more "future proof" (though I find it very unlikely for ret value o change).

I already have an answer to this:
https://lore.kernel.org/r/[email protected]

Does it help?

Quoted 5 lines
Quoted 2 lines

Anyways, even if we choose to go down this route, I don't see much benefit in starting converting the drivers with the pattern below (which should be a considerable number).

There is not a big number of them, so I prefer to have common patterns without exact error code checks.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
maintainer
#4 lore raw
Quoted 7 lines · On Tue, 17 Mar 2026 12:42:07 +0200 Andy Shevchenko…
Quoted 6 lines · Andy Shevchenko
Quoted 5 lines · Nuno Sá
Quoted 3 lines · Andy Shevchenko

Instead of checking for the specific error codes (that can be considered a layering violation to some extent) check for the property existence first and then either parse it, or apply a default value.

Quoted 15 lines
Quoted 6 lines
Quoted 3 lines

Not really sure how I feel about this one. Checking for specific errors is a very common pattern and this change just makes it we check for the property presence twice. That said, this makes it more "future proof" (though I find it very unlikely for ret value o change).

I already have an answer to this:
https://lore.kernel.org/r/[email protected]

Does it help?

Quoted 5 lines
Quoted 2 lines

Anyways, even if we choose to go down this route, I don't see much benefit in starting converting the drivers with the pattern below (which should be a considerable number).

There is not a big number of them, so I prefer to have common patterns without exact error code checks.

I left this one for a while to see if the discussion would continue but seems not. I'm not sure it is always the case, but in this particular example I think the resulting code is a little nicer to read so applied.

So for me, case by case basis for this sort of change.

Thanks,

Jonathan

#5 lore raw
Quoted 27 lines · Jonathan Cameron
Quoted 7 lines · On Tue, 17 Mar 2026 12:42:07 +0200 Andy Shevchenko…
Quoted 6 lines · Andy Shevchenko
Quoted 5 lines · Nuno Sá
Quoted 3 lines · Andy Shevchenko

Instead of checking for the specific error codes (that can be considered a layering violation to some extent) check for the property existence first and then either parse it, or apply a default value.

Quoted 15 lines
Quoted 6 lines
Quoted 3 lines

Not really sure how I feel about this one. Checking for specific errors is a very common pattern and this change just makes it we check for the property presence twice. That said, this makes it more "future proof" (though I find it very unlikely for ret value o change).

I already have an answer to this:
https://lore.kernel.org/r/[email protected]

Does it help?

Quoted 5 lines
Quoted 2 lines

Anyways, even if we choose to go down this route, I don't see much benefit in starting converting the drivers with the pattern below (which should be a considerable number).

There is not a big number of them, so I prefer to have common patterns without exact error code checks.

Quoted 5 lines

I left this one for a while to see if the discussion would continue but seems not. I'm not sure it is always the case, but in this particular example I think the resulting code is a little nicer to read so applied.

So for me, case by case basis for this sort of change.

Yeah, I missed this one! Anyways seems the way will be to explicitly use device_property_present() to check property presence. Still don't love the dual call thing so I might just stop checking for -EINVAL (was not doing it religiously anyways).

Maybe we could have a set of optional variants of the API like device_property_read_*_optional() kind of thing where we just return 0 if the property is not present. But might also be too noisy...

- Nuno Sá

Quoted 4 lines

Thanks,

Jonathan

#6 lore raw
Quoted 3 lines · Nuno Sá
Quoted 2 lines · Jonathan Cameron

On Tue, 17 Mar 2026 12:42:07 +0200
Andy Shevchenko <[email protected]> wrote:

...

Quoted 6 lines
Quoted 5 lines

I left this one for a while to see if the discussion would continue but seems not. I'm not sure it is always the case, but in this particular example I think the resulting code is a little nicer to read so applied.

So for me, case by case basis for this sort of change.

Quoted 8 lines

Yeah, I missed this one! Anyways seems the way will be to explicitly use device_property_present() to check property presence. Still don't love the dual call thing so I might just stop checking for -EINVAL (was not doing it religiously anyways).

Maybe we could have a set of optional variants of the API like device_property_read_*_optional() kind of thing where we just return 0 if the property is not present. But might also be too noisy...

I think that's what usually Sakari likes (tons of wrappers for each case :-). I prefer be on the compromise side.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#7 lore raw
Quoted 4 lines · Andy Shevchenko
Quoted 3 lines · Nuno Sá
Quoted 2 lines · Jonathan Cameron

On Tue, 17 Mar 2026 12:42:07 +0200
Andy Shevchenko <[email protected]> wrote:

...

Quoted 11 lines
Quoted 11 lines
Quoted 5 lines

I left this one for a while to see if the discussion would continue but seems not. I'm not sure it is always the case, but in this particular example I think the resulting code is a little nicer to read so applied.

So for me, case by case basis for this sort of change.

Yeah, I missed this one! Anyways seems the way will be to explicitly use device_property_present() to check property presence. Still don't love the dual call thing so I might just stop checking for -EINVAL (was not doing it religiously anyways).

Quoted 6 lines
Quoted 3 lines

Maybe we could have a set of optional variants of the API like device_property_read_*_optional() kind of thing where we just return 0 if the property is not present. But might also be too noisy...

I think that's what usually Sakari likes (tons of wrappers for each case :-). I prefer be on the compromise side.

Ah, and note, now we have kernel-doc updated to point out that _present is recommended way of unambiguously checking for the property presence.

70fa0c308aa2 ("device property: Document how to check for the property presence")

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#8 lore raw
Quoted 23 lines · Andy Shevchenko
Quoted 4 lines · Andy Shevchenko
Quoted 3 lines · Nuno Sá
Quoted 2 lines · Jonathan Cameron

On Tue, 17 Mar 2026 12:42:07 +0200
Andy Shevchenko <[email protected]> wrote:

...

Quoted 15 lines
Quoted 14 lines
Quoted 5 lines

I left this one for a while to see if the discussion would continue but seems not. I'm not sure it is always the case, but in this particular example I think the resulting code is a little nicer to read so applied.

So for me, case by case basis for this sort of change.

Yeah, I missed this one! Anyways seems the way will be to explicitly use device_property_present() to check property presence. Still don't love the dual call thing so I might just stop checking for -EINVAL (was not doing it religiously anyways).

Maybe we could have a set of optional variants of the API like device_property_read_*_optional() kind of thing where we just return 0 if the property is not present. But might also be too noisy...

Quoted 7 lines
Quoted 2 lines

I think that's what usually Sakari likes (tons of wrappers for each case :-). I prefer be on the compromise side.

Ah, and note, now we have kernel-doc updated to point out that _present is recommended way of unambiguously checking for the property presence.

70fa0c308aa2 ("device property: Document how to check for the property presence")

Yes, I did looked at it! That's why I said

"Anyways seems the way will be to explicitly use device_property_present()..."

- Nuno Sá

Quoted 7 lines
Signature · 3 lines
-- 
With Best Regards,
Andy Shevchenko