Skip to contentSkip to conversation

panic: only warn about deprecated panic_print on write access

PATCH Applied Gal Pressman 3 participants 9 messages Jan 6 – Jan 12

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 panic: only warn about deprecated panic_print on write access

The panic_print_deprecated() warning is being triggered on both read and write operations to the panic_print parameter.

This causes spurious warnings when users run 'sysctl -a' to list all sysctl values, since that command reads /proc/sys/kernel/panic_print and triggers the deprecation notice.

Modify the handlers to only emit the deprecation warning when the parameter is actually being set:

- sysctl_panic_print_handler(): check 'write' flag before warning.
- panic_print_get(): remove the deprecation call entirely.

This way, users are only warned when they actively try to use the deprecated parameter, not when passively querying system state.

Fixes ee13240cd78b ("panic: add note that panic_print sysctl interface is deprecated")
Fixes 2683df6539cb ("panic: add note that 'panic_print' parameter is deprecated")
Cc Feng Tang <[email protected]>
---
 kernel/panic.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/panic.c +2 -2
index 0d52210a9e2b..0c20fcaae98a 100644
@@ -131,7 +131,8 @@ static int proc_taint(const struct ctl_table *table, int write,
 static int sysctl_panic_print_handler(const struct ctl_table *table, int write,
 			   void *buffer, size_t *lenp, loff_t *ppos)
 {
-	panic_print_deprecated();
+	if (write)
+		panic_print_deprecated();
 	return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
 }
 
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();
 	return  param_get_ulong(val, kp);
 }
 
Signature · 1 line
2.40.1
#1 lore raw

Hi Gal,

Thanks for the fix!

Quoted 31 lines · Gal Pressman

The panic_print_deprecated() warning is being triggered on both read and write operations to the panic_print parameter.

This causes spurious warnings when users run 'sysctl -a' to list all sysctl values, since that command reads /proc/sys/kernel/panic_print and triggers the deprecation notice.

Modify the handlers to only emit the deprecation warning when the parameter is actually being set:

- sysctl_panic_print_handler(): check 'write' flag before warning.
- panic_print_get(): remove the deprecation call entirely.

This way, users are only warned when they actively try to use the deprecated parameter, not when passively querying system state.

Fixes ee13240cd78b ("panic: add note that panic_print sysctl interface is deprecated")
Fixes 2683df6539cb ("panic: add note that 'panic_print' parameter is deprecated")
Cc Feng Tang <[email protected]>
---
 kernel/panic.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/panic.c +0 -0
index 0d52210a9e2b..0c20fcaae98a 100644

kernel/panic.c static int proc_taint(const struct ctl_table *table, int write, · line 135

Quoted 6 lines
kernel/panic.c +2 -1
@@ -131,7 +131,8 @@ static int proc_taint(const struct ctl_table *table, int write,
 static int sysctl_panic_print_handler(const struct ctl_table *table, int write,
 			   void *buffer, size_t *lenp, loff_t *ppos)
 {
-	panic_print_deprecated();
+	if (write)
+		panic_print_deprecated();

This makes sense to me, as 'sysctl -a' is a very common usage.

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 8 lines
kernel/panic.c +0 -1
 	return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
 }
 
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

Thanks,
Feng

Quoted 5 lines
kernel/panic.c +0 -0
 	return  param_get_ulong(val, kp);
 }
 
Signature · 1 line
2.40.1
#2 lore raw
Quoted 9 lines · Feng Tang

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
kernel/panic.c +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

#3 lore raw
Quoted 10 lines · Gal Pressman
Quoted 9 lines · Feng Tang

Review static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
(quoted diff) +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 5 lines
Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

When people run 'sysctl -a', in 99.9% cases, the users don't care 'panic_print' or even don't know what 'panic_print' is, that's why I think removing it makes sense.

But for a user running 'cat /sys/module/kernel/parameters/panic_rint', giving a warning is meaningful.

If you insist so, the cleaner way would be

kernel/panic.c +1 -7
index 0d52210a9e2b..86b927a695a0 100644
@@ -1012,15 +1012,9 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 	return  param_set_ulong(val, kp);
 }
 
-static int panic_print_get(char *val, const struct kernel_param *kp)
-{
-	panic_print_deprecated();
-	return  param_get_ulong(val, kp);
-}
-
 static const struct kernel_param_ops panic_print_ops = {
 	.set	= panic_print_set,
-	.get	= panic_print_get,
+	.get	= param_get_ulong,
 };
 __core_param_cb(panic_print, &panic_print_ops, &panic_print, 0644);
 
#4 lore raw
Quoted 11 lines · Feng Tang
Quoted 10 lines · Gal Pressman
Quoted 9 lines · Feng Tang

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
kernel/panic.c +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

I see.

Quoted 6 lines

When people run 'sysctl -a', in 99.9% cases, the users don't care 'panic_print' or even don't know what 'panic_print' is, that's why I think removing it makes sense.

But for a user running 'cat /sys/module/kernel/parameters/panic_rint', giving a warning is meaningful.

Makes perfect sense.

We need to make people aware that "panic_print" will eventually go away. 'sysctl -a' is different because it prints all values and there is big chance that the caller is not interested in "panic_print" at all.

It makes sense to remove the warning from sysctl read. But I would keep it in sysfs read.

Best Regards,
Petr

#5 lore raw
Quoted 29 lines · Petr Mladek
Quoted 11 lines · Feng Tang
Quoted 10 lines · Gal Pressman
Quoted 9 lines · Feng Tang

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
kernel/panic.c +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

I see.

Quoted 6 lines

When people run 'sysctl -a', in 99.9% cases, the users don't care 'panic_print' or even don't know what 'panic_print' is, that's why I think removing it makes sense.

But for a user running 'cat /sys/module/kernel/parameters/panic_rint', giving a warning is meaningful.

Makes perfect sense.

Quoted 7 lines

We need to make people aware that "panic_print" will eventually go away. 'sysctl -a' is different because it prints all values and there is big chance that the caller is not interested in "panic_print" at all.

It makes sense to remove the warning from sysctl read. But I would keep it in sysfs read.

The sysfs entry exhibits the same issue, just different command:

# systool -m kernel -v

Or:

# grep . /sys/module/kernel/parameters/*
#6 lore raw
Quoted 38 lines · Gal Pressman
Quoted 29 lines · Petr Mladek
Quoted 11 lines · Feng Tang
Quoted 10 lines · Gal Pressman
Quoted 9 lines · Feng Tang

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
kernel/panic.c +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

I see.

Quoted 6 lines

When people run 'sysctl -a', in 99.9% cases, the users don't care 'panic_print' or even don't know what 'panic_print' is, that's why I think removing it makes sense.

But for a user running 'cat /sys/module/kernel/parameters/panic_rint', giving a warning is meaningful.

Makes perfect sense.

Quoted 7 lines

We need to make people aware that "panic_print" will eventually go away. 'sysctl -a' is different because it prints all values and there is big chance that the caller is not interested in "panic_print" at all.

It makes sense to remove the warning from sysctl read. But I would keep it in sysfs read.

Quoted 7 lines

The sysfs entry exhibits the same issue, just different command:

# systool -m kernel -v

Or:

# grep . /sys/module/kernel/parameters/*

Yes, when user checks the kernel parameters, I think it's better to give them a notice (it's a pr_info_once(), which only prints once for the whole power cycle).

Thanks,
Feng

#7 lore raw
Quoted 43 lines · Feng Tang
Quoted 42 lines · Gal Pressman
Quoted 29 lines · Petr Mladek
Quoted 11 lines · Feng Tang
Quoted 10 lines · Gal Pressman
Quoted 9 lines · Feng Tang

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
kernel/panic.c +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

I see.

Quoted 6 lines

When people run 'sysctl -a', in 99.9% cases, the users don't care 'panic_print' or even don't know what 'panic_print' is, that's why I think removing it makes sense.

But for a user running 'cat /sys/module/kernel/parameters/panic_rint', giving a warning is meaningful.

Makes perfect sense.

Quoted 7 lines

We need to make people aware that "panic_print" will eventually go away. 'sysctl -a' is different because it prints all values and there is big chance that the caller is not interested in "panic_print" at all.

It makes sense to remove the warning from sysctl read. But I would keep it in sysfs read.

The sysfs entry exhibits the same issue, just different command:

# systool -m kernel -v
Quoted 7 lines
Quoted 3 lines

Or:

# grep . /sys/module/kernel/parameters/*

Yes, when user checks the kernel parameters, I think it's better to give them a notice (it's a pr_info_once(), which only prints once for the whole power cycle).

This log message causes a lot of noise for people who have no idea what panic_print is.
We started seeing this print internally, and it took a long time and significant resources to figure out what changed and if we were doing anything wrong in our environments.

Seeing a deprecation notice is alarming, especially if you don't know what is being deprecated or how it affects you. In my opinion, we shouldn't print this unless someone actually uses panic_print.

#8 lore raw
Quoted 40 lines · Gal Pressman
Quoted 39 lines · Feng Tang
Quoted 38 lines · Gal Pressman
Quoted 36 lines · Petr Mladek
Quoted 11 lines · Feng Tang
Quoted 10 lines · Gal Pressman
Quoted 9 lines · Feng Tang

kernel/panic.c static int panic_print_set(const char *val, const struct kernel_param *kp) · line 1017

Quoted 5 lines
kernel/panic.c +0 -1
@@ -1014,7 +1015,6 @@ static int panic_print_set(const char *val, const struct kernel_param *kp)
 
 static int panic_print_get(char *val, const struct kernel_param *kp)
 {
-	panic_print_deprecated();

Actually this was intentional, in one of the patch version, this panic_print_get() was not there but reusing the param_get_ulong().

Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

It was added later as sometimes developer do want to runtime check the current 'panic_print' setting through /sys/module/kernel/parameters/ interface, and I thought it may be better to give the warning.

I figured it would make sense to keep the behaviors consistent.

I see.

Quoted 6 lines

When people run 'sysctl -a', in 99.9% cases, the users don't care 'panic_print' or even don't know what 'panic_print' is, that's why I think removing it makes sense.

But for a user running 'cat /sys/module/kernel/parameters/panic_rint', giving a warning is meaningful.

Makes perfect sense.

We need to make people aware that "panic_print" will eventually go away. 'sysctl -a' is different because it prints all values and there is big chance that the caller is not interested in "panic_print" at all.

It makes sense to remove the warning from sysctl read. But I would keep it in sysfs read.

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

The sysfs entry exhibits the same issue, just different command:

# systool -m kernel -v

Or:

# grep . /sys/module/kernel/parameters/*

I see.

Quoted 9 lines
Quoted 3 lines

Yes, when user checks the kernel parameters, I think it's better to give them a notice (it's a pr_info_once(), which only prints once for the whole power cycle).

This log message causes a lot of noise for people who have no idea what panic_print is.
We started seeing this print internally, and it took a long time and significant resources to figure out what changed and if we were doing anything wrong in our environments.

Quoted 3 lines

Seeing a deprecation notice is alarming, especially if you don't know what is being deprecated or how it affects you. In my opinion, we shouldn't print this unless someone actually uses panic_print.

I am fine with printing the warning only when the value is modified (set/write).

Best Regards,
Petr

PS: Sigh, it is hard to change or remove an interface. It is always
better to think twice before introducing one.