Skip to contentSkip to conversation

iio: magnetometer: ak8974: Silence deferred-probe error

PATCH Superseded Dmitry Osipenko 3 participants 7 messages Apr 14, 2020 – Apr 18, 2020

Subsystems (2)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: magnetometer: ak8974: Silence deferred-probe error

It's not uncommon that voltage regulator becomes available later during kernel's boot process, in this case there is no need to print a noisy error message. This patch moves the message about unavailable regulator to the debug level in a case of the deferred-probe error and also amends the message with error code.

---
 drivers/iio/magnetometer/ak8974.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/iio/magnetometer/ak8974.c +6 -1
index d32996702110..cc3861f97d42 100644
@@ -718,6 +718,7 @@ static const struct regmap_config ak8974_regmap_config = {
 static int ak8974_probe(struct i2c_client *i2c,
 			const struct i2c_device_id *id)
 {
+	const char *level = KERN_ERR;
 	struct iio_dev *indio_dev;
 	struct ak8974 *ak8974;
 	unsigned long irq_trig;
@@ -746,7 +747,11 @@ static int ak8974_probe(struct i2c_client *i2c,
 				      ARRAY_SIZE(ak8974->regs),
 				      ak8974->regs);
 	if (ret < 0) {
-		dev_err(&i2c->dev, "cannot get regulators\n");
+		if (ret == -EPROBE_DEFER)
+			level = KERN_DEBUG;
+
+		dev_printk(level, &i2c->dev, "cannot get regulators: %d\n",
+			   ret);
 		return ret;
 	}
 
Signature · 1 line
2.26.0
#1 lore raw

Hi Dmitry,

thanks for your patch!

Quoted 24 lines · Dmitry Osipenko

It's not uncommon that voltage regulator becomes available later during kernel's boot process, in this case there is no need to print a noisy error message. This patch moves the message about unavailable regulator to the debug level in a case of the deferred-probe error and also amends the message with error code.

---
 drivers/iio/magnetometer/ak8974.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/iio/magnetometer/ak8974.c +1 -0
index d32996702110..cc3861f97d42 100644
@@ -718,6 +718,7 @@ static const struct regmap_config ak8974_regmap_config = {
 static int ak8974_probe(struct i2c_client *i2c,
                        const struct i2c_device_id *id)
 {
+       const char *level = KERN_ERR;
        struct iio_dev *indio_dev;
        struct ak8974 *ak8974;
        unsigned long irq_trig;

drivers/iio/magnetometer/ak8974.c static int ak8974_probe(struct i2c_client *i2c, · line 753

Quoted 8 lines
drivers/iio/magnetometer/ak8974.c +4 -1
@@ -746,7 +747,11 @@ static int ak8974_probe(struct i2c_client *i2c,
                                      ARRAY_SIZE(ak8974->regs),
                                      ak8974->regs);
        if (ret < 0) {
-               dev_err(&i2c->dev, "cannot get regulators\n");
+               if (ret == -EPROBE_DEFER)
+                       level = KERN_DEBUG;
+
+               dev_printk(level, &i2c->dev, "cannot get regulators: %d\n",

This misses some important aspects of dev_dbg(), notably this:

#if defined(CONFIG_DYNAMIC_DEBUG)

#define dev_dbg(dev, fmt, ...)                                          \
        dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
#elif defined(DEBUG)
#define dev_dbg(dev, fmt, ...)                                          \
        dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__)
#else
#define dev_dbg(dev, fmt, ...)                                          \
({                                                                      \
        if (0)                                                          \
                dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__); \
})

#endif

If DEBUG is not defined the entire dev_dbg() message is enclodes in if (0) and compiled out of the kernel, saving space. The above does not fulfil that.

Yours,
Linus Walleij

#2 lore raw

16.04.2020 14:33, Linus Walleij пишет:

Quoted 53 lines

Hi Dmitry,

thanks for your patch!

Quoted 24 lines · Dmitry Osipenko

It's not uncommon that voltage regulator becomes available later during kernel's boot process, in this case there is no need to print a noisy error message. This patch moves the message about unavailable regulator to the debug level in a case of the deferred-probe error and also amends the message with error code.

---
 drivers/iio/magnetometer/ak8974.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/iio/magnetometer/ak8974.c +1 -0
index d32996702110..cc3861f97d42 100644
@@ -718,6 +718,7 @@ static const struct regmap_config ak8974_regmap_config = {
 static int ak8974_probe(struct i2c_client *i2c,
                        const struct i2c_device_id *id)
 {
+       const char *level = KERN_ERR;
        struct iio_dev *indio_dev;
        struct ak8974 *ak8974;
        unsigned long irq_trig;

drivers/iio/magnetometer/ak8974.c static int ak8974_probe(struct i2c_client *i2c, · line 753

Quoted 8 lines
drivers/iio/magnetometer/ak8974.c +4 -1
@@ -746,7 +747,11 @@ static int ak8974_probe(struct i2c_client *i2c,
                                      ARRAY_SIZE(ak8974->regs),
                                      ak8974->regs);
        if (ret < 0) {
-               dev_err(&i2c->dev, "cannot get regulators\n");
+               if (ret == -EPROBE_DEFER)
+                       level = KERN_DEBUG;
+
+               dev_printk(level, &i2c->dev, "cannot get regulators: %d\n",

This misses some important aspects of dev_dbg(), notably this:

#if defined(CONFIG_DYNAMIC_DEBUG)

#define dev_dbg(dev, fmt, ...)                                          \
        dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
#elif defined(DEBUG)
#define dev_dbg(dev, fmt, ...)                                          \
        dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__)
#else
#define dev_dbg(dev, fmt, ...)                                          \
({                                                                      \
        if (0)                                                          \
                dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__); \
})
Quoted 5 lines

#endif

If DEBUG is not defined the entire dev_dbg() message is enclodes in if (0) and compiled out of the kernel, saving space. The above does not fulfil that.

Hello Linus,

After some recent discussions in regards to the EPROBE_DEFER handling, Thierry Reding suggested the form which is used in my patch and we started to use it recently in the Tegra DRM driver [1]. The reason is that we don't want to miss any deferred-probe messages under any circumstances, for example like in a case of a disabled DYNAMIC_DEBUG. The debug messages are usually disabled in a release-build and when not a very experienced person hands you KMSG for diagnosing a problem, the KMSG is pretty much useless if error is hidden silently.

By moving the message to a debug level, we reduce the noise in the KMSG because usually people look for a bold-red error messages. Secondly, we don't introduce an additional overhead to the kernel size since the same text is reused for all error conditions.

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e32c8c2a5fbe1514e4ae9063a49e76ecf486ffb8

I can change this patch to something like this:

if (ret == -EPROBE_DEFER)

	dev_dbg("deferring probe\n");
else
	dev_err("cannot...");

but in general this should be a less useful variant, don't you agree?

#3 lore raw
Quoted 22 lines · Dmitry Osipenko

16.04.2020 14:33, Linus Walleij пишет:

Quoted 14 lines

This misses some important aspects of dev_dbg(), notably this:

#if defined(CONFIG_DYNAMIC_DEBUG)

#define dev_dbg(dev, fmt, ...)                                          \
        dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
#elif defined(DEBUG)
#define dev_dbg(dev, fmt, ...)                                          \
        dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__)
#else
#define dev_dbg(dev, fmt, ...)                                          \
({                                                                      \
        if (0)                                                          \
                dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__); \
})
Quoted 5 lines

#endif

If DEBUG is not defined the entire dev_dbg() message is enclodes in if (0) and compiled out of the kernel, saving space. The above does not fulfil that.

Quoted 7 lines

Hello Linus,

After some recent discussions in regards to the EPROBE_DEFER handling, Thierry Reding suggested the form which is used in my patch and we started to use it recently in the Tegra DRM driver [1]. The reason is that we don't want to miss any deferred-probe messages under any circumstances, for example like in a case of a disabled DYNAMIC_DEBUG.

I have a hard time to accept this reasoning.

Who doesn't feel that way about their subsystem? If you don't want to miss the message under any circumstances then use dev_info(). Don't override the default behaviour of dev_dbg().

Quoted 3 lines

The debug messages are usually disabled in a release-build and when not a very experienced person hands you KMSG for diagnosing a problem, the KMSG is pretty much useless if error is hidden silently.

So use dev_info().

Quoted 4 lines

By moving the message to a debug level, we reduce the noise in the KMSG because usually people look for a bold-red error messages. Secondly, we don't introduce an additional overhead to the kernel size since the same text is reused for all error conditions.

dev_info() is not supposed to be an error message, it is supposed to be information, so use that.

Yours,
Linus Walleij

#4 lore raw

16.04.2020 19:51, Linus Walleij пишет:

Quoted 43 lines · Dmitry Osipenko
Quoted 22 lines

16.04.2020 14:33, Linus Walleij пишет:

Quoted 14 lines

This misses some important aspects of dev_dbg(), notably this:

#if defined(CONFIG_DYNAMIC_DEBUG)

#define dev_dbg(dev, fmt, ...)                                          \
        dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
#elif defined(DEBUG)
#define dev_dbg(dev, fmt, ...)                                          \
        dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__)
#else
#define dev_dbg(dev, fmt, ...)                                          \
({                                                                      \
        if (0)                                                          \
                dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__); \
})
Quoted 5 lines

#endif

If DEBUG is not defined the entire dev_dbg() message is enclodes in if (0) and compiled out of the kernel, saving space. The above does not fulfil that.

Quoted 7 lines

Hello Linus,

After some recent discussions in regards to the EPROBE_DEFER handling, Thierry Reding suggested the form which is used in my patch and we started to use it recently in the Tegra DRM driver [1]. The reason is that we don't want to miss any deferred-probe messages under any circumstances, for example like in a case of a disabled DYNAMIC_DEBUG.

I have a hard time to accept this reasoning.

Who doesn't feel that way about their subsystem? If you don't want to miss the message under any circumstances then use dev_info(). Don't override the default behaviour of dev_dbg().

Quoted 3 lines

The debug messages are usually disabled in a release-build and when not a very experienced person hands you KMSG for diagnosing a problem, the KMSG is pretty much useless if error is hidden silently.

So use dev_info().

Quoted 7 lines
Quoted 4 lines

By moving the message to a debug level, we reduce the noise in the KMSG because usually people look for a bold-red error messages. Secondly, we don't introduce an additional overhead to the kernel size since the same text is reused for all error conditions.

dev_info() is not supposed to be an error message, it is supposed to be information, so use that.

Okay, I'll make a v2. Thank you for the review.

maintainer
#5 lore raw
Quoted 49 lines · On Thu, 16 Apr 2020 20:35:56 +0300 Dmitry Osipenko <[email protected]> wrote:

16.04.2020 19:51, Linus Walleij пишет:

Quoted 48 lines · Dmitry Osipenko
Quoted 22 lines

16.04.2020 14:33, Linus Walleij пишет:

Quoted 14 lines

This misses some important aspects of dev_dbg(), notably this:

#if defined(CONFIG_DYNAMIC_DEBUG)

#define dev_dbg(dev, fmt, ...)                                          \
        dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
#elif defined(DEBUG)
#define dev_dbg(dev, fmt, ...)                                          \
        dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__)
#else
#define dev_dbg(dev, fmt, ...)                                          \
({                                                                      \
        if (0)                                                          \
                dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__); \
})
Quoted 5 lines

#endif

If DEBUG is not defined the entire dev_dbg() message is enclodes in if (0) and compiled out of the kernel, saving space. The above does not fulfil that.

Quoted 7 lines

Hello Linus,

After some recent discussions in regards to the EPROBE_DEFER handling, Thierry Reding suggested the form which is used in my patch and we started to use it recently in the Tegra DRM driver [1]. The reason is that we don't want to miss any deferred-probe messages under any circumstances, for example like in a case of a disabled DYNAMIC_DEBUG.

I have a hard time to accept this reasoning.

Who doesn't feel that way about their subsystem? If you don't want to miss the message under any circumstances then use dev_info(). Don't override the default behaviour of dev_dbg().

Quoted 3 lines

The debug messages are usually disabled in a release-build and when not a very experienced person hands you KMSG for diagnosing a problem, the KMSG is pretty much useless if error is hidden silently.

So use dev_info().

Quoted 4 lines

By moving the message to a debug level, we reduce the noise in the KMSG because usually people look for a bold-red error messages. Secondly, we don't introduce an additional overhead to the kernel size since the same text is reused for all error conditions.

Quoted 4 lines
Quoted 2 lines

dev_info() is not supposed to be an error message, it is supposed to be information, so use that.

Okay, I'll make a v2. Thank you for the review.

Ah I commented on this in v2 - now I see why you did it :) Nope to dev_info. That will often spam normal logs and as Andy pointed out for v2 that can be dozens of entries on a sophisticated board. Much better to stick to dev_dbg but I'd like to see it done explicitly in the form you mention with the if / else

Thanks,

Jonathan

#6 lore raw

18.04.2020 17:37, Jonathan Cameron пишет:

Quoted 55 lines
Quoted 52 lines · On Thu, 16 Apr 2020 20:35:56 +0300 Dmitry Osipenko <[email protected]> wrote:

16.04.2020 19:51, Linus Walleij пишет:

Quoted 50 lines · Dmitry Osipenko
Quoted 22 lines

16.04.2020 14:33, Linus Walleij пишет:

Quoted 14 lines

This misses some important aspects of dev_dbg(), notably this:

#if defined(CONFIG_DYNAMIC_DEBUG)

#define dev_dbg(dev, fmt, ...)                                          \
        dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
#elif defined(DEBUG)
#define dev_dbg(dev, fmt, ...)                                          \
        dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__)
#else
#define dev_dbg(dev, fmt, ...)                                          \
({                                                                      \
        if (0)                                                          \
                dev_printk(KERN_DEBUG, dev, dev_fmt(fmt), ##__VA_ARGS__); \
})
Quoted 5 lines

#endif

If DEBUG is not defined the entire dev_dbg() message is enclodes in if (0) and compiled out of the kernel, saving space. The above does not fulfil that.

Quoted 7 lines

Hello Linus,

After some recent discussions in regards to the EPROBE_DEFER handling, Thierry Reding suggested the form which is used in my patch and we started to use it recently in the Tegra DRM driver [1]. The reason is that we don't want to miss any deferred-probe messages under any circumstances, for example like in a case of a disabled DYNAMIC_DEBUG.

I have a hard time to accept this reasoning.

Who doesn't feel that way about their subsystem? If you don't want to miss the message under any circumstances then use dev_info(). Don't override the default behaviour of dev_dbg().

Quoted 3 lines

The debug messages are usually disabled in a release-build and when not a very experienced person hands you KMSG for diagnosing a problem, the KMSG is pretty much useless if error is hidden silently.

So use dev_info().

Quoted 4 lines

By moving the message to a debug level, we reduce the noise in the KMSG because usually people look for a bold-red error messages. Secondly, we don't introduce an additional overhead to the kernel size since the same text is reused for all error conditions.

dev_info() is not supposed to be an error message, it is supposed to be information, so use that.

Quoted 7 lines
Quoted 1 line

Okay, I'll make a v2. Thank you for the review.

Ah I commented on this in v2 - now I see why you did it :) Nope to dev_info. That will often spam normal logs and as Andy pointed out for v2 that can be dozens of entries on a sophisticated board. Much better to stick to dev_dbg but I'd like to see it done explicitly in the form you mention with the if / else

Alright, I'll make a v3.