Skip to contentSkip to conversation

iio: adc: ad4170-4: Use ERR_PTR() with %pe to improve error logging

PATCH In review Salah Triki 5 participants 9 messages Aug 7, 2025 – Aug 8, 2025

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

PATCH v2 1/1 iio: adc: ad4170-4: Use ERR_PTR() with %pe to improve error logging

Use ERR_PTR(ret) with %pe in ad4170_read_sample() to properly display symbolic error codes (e.g. -ENOMEM) instead of raw integers (e.g. -12), improving readability and debug clarity.

---
Changes in v2:
- Improve commit title

 drivers/iio/adc/ad4170-4.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/iio/adc/ad4170-4.c +2 -2
index 6cd84d6fb08b..6296b5dc930b 100644
@@ -1253,11 +1253,11 @@ static int ad4170_read_sample(struct iio_dev *indio_dev,
 
 	ret = __ad4170_read_sample(indio_dev, chan, val);
 	if (ret) {
-		dev_err(dev, "failed to read sample: %d\n", ret);
+		dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
 		ret2 = ad4170_set_channel_enable(st, chan->address, false);
 		if (ret2)
-			dev_err(dev, "failed to disable channel: %d\n", ret2);
+			dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
 
 		return ret;
 	}
Signature · 1 line
2.43.0
reviewer
#1 lore raw
Quoted 20 lines · Salah Triki

Use ERR_PTR(ret) with %pe in ad4170_read_sample() to properly display symbolic error codes (e.g. -ENOMEM) instead of raw integers (e.g. -12), improving readability and debug clarity.

---
Changes in v2:
- Improve commit title

 drivers/iio/adc/ad4170-4.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/iio/adc/ad4170-4.c +0 -0
index 6cd84d6fb08b..6296b5dc930b 100644
@@ -1253,11 +1253,11 @@ static int ad4170_read_sample(struct iio_dev *indio_dev,
 
 	ret = __ad4170_read_sample(indio_dev, chan, val);
 	if (ret) {

drivers/iio/adc/ad4170-4.c static int ad4170_read_sample(struct iio_dev *indio_dev, · line 1263

Quoted 10 lines
drivers/iio/adc/ad4170-4.c +2 -2
-		dev_err(dev, "failed to read sample: %d\n", ret);
+		dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
 		ret2 = ad4170_set_channel_enable(st, chan->address, false);
 		if (ret2)
-			dev_err(dev, "failed to disable channel: %d\n", ret2);
+			dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
 
 		return ret;
 	}

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

And if we are doing this in more places, it would make sense to have a new format specifier for integer error values instead of casting them to pointers.

#2 lore raw
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 10 lines
Quoted 10 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
drivers/iio/adc/ad4170-4.c +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
Quoted 6 lines
Quoted 2 lines
              return ret;
      }

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

Quoted 3 lines

And if we are doing this in more places, it would make sense to have a new format specifier for integer error values instead of casting them to pointers.

Will _very unlikely_ to happen. This has to be a C standard for that, otherwise you are suggesting to always have a kernel warning for each of these cases. The only way we can customize specifiers w/o introducing a compiler warnings is to continue (and still carefully) using %p extensions.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 23 lines
Quoted 10 lines
Quoted 10 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
drivers/iio/adc/ad4170-4.c +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
Quoted 6 lines
Quoted 2 lines
              return ret;
      }

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

Quoted 3 lines

And if we are doing this in more places, it would make sense to have a new format specifier for integer error values instead of casting them to pointers.

Quoted 5 lines

Will _very unlikely_ to happen. This has to be a C standard for that, otherwise you are suggesting to always have a kernel warning for each of these cases. The only way we can customize specifiers w/o introducing a compiler warnings is to continue (and still carefully) using %p extensions.

And to be clear: I am not in favour of this change exactly due to a bit weird (for the reader) castings just for the sake of use of %pe.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
reviewer
#4 lore raw
Quoted 17 lines · Andy Shevchenko
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
drivers/iio/adc/ad4170-4.c +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
              return ret;
      }

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

I meant that it would be strange to make this change just in one driver and not do the same everywhere else.

Quoted 10 lines
Quoted 10 lines
Quoted 4 lines

And if we are doing this in more places, it would make sense to have a new format specifier for integer error values instead of casting them to pointers.

Will _very unlikely_ to happen. This has to be a C standard for that, otherwise you are suggesting to always have a kernel warning for each of these cases. The only way we can customize specifiers w/o introducing a compiler warnings is to continue (and still carefully) using %p extensions.

OK, makes sense.

Quoted 5 lines

And to be clear: I am not in favour of this change exactly due to a bit weird (for the reader) castings just for the sake of use of %pe.

#5 lore raw
Quoted 5 lines · David Lechner
Quoted 4 lines · Andy Shevchenko
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
Quoted 12 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
drivers/iio/adc/ad4170-4.c +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
 
              return ret;
      }
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

I meant that it would be strange to make this change just in one driver and not do the same everywhere else.

But it's how we usually do things in the kernel, subsystem-by-subsystem or even driver-by-driver depending on the cases. And currently it's used in many already. But if we are talking about IIO, we need to agree which cases are okay to move to and which are not (besides the fact that this should be applicable only for error pointers without additional castings).

Quoted 2 lines
Quoted 2 lines

And to be clear: I am not in favour of this change exactly due to a bit weird (for the reader) castings just for the sake of use of %pe.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
Reviewed-by
#6 lore raw
In reply to David Lechner

Hi,

Quoted 21 lines · David Lechner
Quoted 20 lines · Andy Shevchenko
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 13 lines
Quoted 13 lines
Quoted 12 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
drivers/iio/adc/ad4170-4.c +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
 
              return ret;
      }
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

I meant that it would be strange to make this change just in one driver and not do the same everywhere else.

Casting error values to pointers is already being done by many IIO drivers if we consider the use of dev_err_probe(). __dev_probe_failed() does the casting from within dev_err_probe()
https://elixir.bootlin.com/linux/v6.15.9/source/drivers/base/core.c#L5026

Thus, I think this patch makes the error messaging from ad4170 more consistent and, because of that, I also see this as a good change.

Though, I'm also totally fine if maintainers prefer not to take this change for whatever reason.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

And if we are doing this in more places, it would make sense to have a new format specifier for integer error values instead of casting them to pointers.

Quoted 15 lines
Quoted 5 lines
Quoted 5 lines

Will _very unlikely_ to happen. This has to be a C standard for that, otherwise you are suggesting to always have a kernel warning for each of these cases. The only way we can customize specifiers w/o introducing a compiler warnings is to continue (and still carefully) using %p extensions.

OK, makes sense.

Quoted 5 lines

And to be clear: I am not in favour of this change exactly due to a bit weird (for the reader) castings just for the sake of use of %pe.

Best regards,
Marcelo

#7 lore raw
Quoted 6 lines · Marcelo Schmitt
Quoted 5 lines · David Lechner
Quoted 4 lines · Andy Shevchenko
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 19 lines
Quoted 19 lines
Quoted 18 lines
Quoted 18 lines
Quoted 10 lines
Quoted 10 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
drivers/iio/adc/ad4170-4.c +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
Quoted 6 lines
Quoted 2 lines
              return ret;
      }

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

Quoted 7 lines
Quoted 2 lines

I meant that it would be strange to make this change just in one driver and not do the same everywhere else.

Casting error values to pointers is already being done by many IIO drivers if we consider the use of dev_err_probe(). __dev_probe_failed() does the casting from within dev_err_probe()
https://elixir.bootlin.com/linux/v6.15.9/source/drivers/base/core.c#L5026

This is a manipulation. The dev_err_probe() and __dev_probe_failed() are parts of the core where we specifically bend the rules for all in one place just for a reason to avoid this spreading (and avoid creating specific APIs).

Quoted 4 lines

Thus, I think this patch makes the error messaging from ad4170 more consistent and, because of that, I also see this as a good change.

Thanks for the review...

Quoted 2 lines

Though, I'm also totally fine if maintainers prefer not to take this change for whatever reason.

...but the below still stays...

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

And to be clear: I am not in favour of this change exactly due to a bit weird (for the reader) castings just for the sake of use of %pe.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#8 lore raw
Quoted 46 lines · On Fri, 8 Aug 2025 14:30:53 +0200 Andy Shevchenko <[email protected]>…
Quoted 6 lines · Marcelo Schmitt
Quoted 5 lines · David Lechner
Quoted 4 lines · Andy Shevchenko
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · David Lechner

On 8/7/25 3:05 AM, Salah Triki wrote:

...

Quoted 19 lines
Quoted 19 lines
Quoted 18 lines
Quoted 18 lines
Quoted 10 lines
Quoted 10 lines
      ret = __ad4170_read_sample(indio_dev, chan, val);
      if (ret) {
(quoted diff) +2 -2
-             dev_err(dev, "failed to read sample: %d\n", ret);
+             dev_err(dev, "failed to read sample: %pe\n", ERR_PTR(ret));
 
              ret2 = ad4170_set_channel_enable(st, chan->address, false);
              if (ret2)
-                     dev_err(dev, "failed to disable channel: %d\n", ret2);
+                     dev_err(dev, "failed to disable channel: %pe\n", ERR_PTR(ret2));
Quoted 6 lines
Quoted 2 lines
              return ret;
      }  

Interesting, I didn't know we had this format specifier. But I think this is something we would want to do kernel-wide or not at all to stay consistent.

I'm sorry but I didn't follow. This is a kernel-wide format specifier.

Quoted 7 lines
Quoted 2 lines

I meant that it would be strange to make this change just in one driver and not do the same everywhere else.

Casting error values to pointers is already being done by many IIO drivers if we consider the use of dev_err_probe(). __dev_probe_failed() does the casting from within dev_err_probe()
https://elixir.bootlin.com/linux/v6.15.9/source/drivers/base/core.c#L5026

This is a manipulation. The dev_err_probe() and __dev_probe_failed() are parts of the core where we specifically bend the rules for all in one place just for a reason to avoid this spreading (and avoid creating specific APIs).

Quoted 3 lines

Thus, I think this patch makes the error messaging from ad4170 more consistent and, because of that, I also see this as a good change.

Quoted 12 lines
Quoted 1 line

Thanks for the review...

Quoted 2 lines

Though, I'm also totally fine if maintainers prefer not to take this change for whatever reason.

...but the below still stays...

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

And to be clear: I am not in favour of this change exactly due to a bit weird (for the reader) castings just for the sake of use of %pe.

Agreed. To me having this cast broadly used in drivers is just too weird.

I'd prefer we dig down a little further and use what %pe is using directly.

#include <linux/errname.h>

printk(... %s\n", errname(ret));

It's not much used so far though but there is precedence via a wrapper in bcachefs. We'd probably want to select CONFIG_SYMBOLIC_ERRNAME for IIO though (rather than every driver).

Jonathan