Skip to contentSkip to conversation

Fix generic WARN_ON message

PATCH In review Jeremy Fitzhardinge 5 participants 7 messages Oct 18, 2006 – Oct 25, 2006 Generic include/asm header files

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix generic WARN_ON message

A warning is a warning, not a BUG.

include/asm-generic/bug.h +0 -0
diff -r 6c118d4e4240 include/asm-generic/bug.h
@@ -35,7 +35,7 @@ struct bug_entry {
#define WARN_ON(condition) ({						\
	typeof(condition) __ret_warn_on = (condition);			\
	if (unlikely(__ret_warn_on)) {					\
-		printk("BUG: warning at %s:%d/%s()\n", __FILE__,	\
+		printk("WARNING at %s:%d %s()\n", __FILE__,	\
			__LINE__, __FUNCTION__);			\
		dump_stack();						\
	}								\
#1 lore raw

Review

Quoted 4 lines · Jeremy Fitzhardinge

A warning is a warning, not a BUG.

(quoted diff) +1 -1
-		printk("BUG: warning at %s:%d/%s()\n", __FILE__,	\
+		printk("WARNING at %s:%d %s()\n", __FILE__,	\

i'm not really happy about this change.

Firstly, most WARN_ON()s are /bugs/, not warnings ... If it's a real warning, a KERN_INFO printk should be done.

Secondly, the reason i changed it to the 'BUG: ...' format is that i tried to make it easier for automated tools (and for users) to figure out that a kernel bug happened.

Ingo

#2 lore raw
Quoted 2 lines · Ingo Molnar

Firstly, most WARN_ON()s are /bugs/, not warnings ... If it's a real warning, a KERN_INFO printk should be done.

It seems to me that either the warnings are really bugs, in which case they should be using BUG/BUG_ON, or they're not really bugs, in which case they should be presented differently.

Quoted 2 lines

Secondly, the reason i changed it to the 'BUG: ...' format is that i tried to make it easier for automated tools (and for users) to figure out that a kernel bug happened.

Well, are they bugs or not? I think people are more confused by the "BUG" prefix and stacktrace than helped by it (even an experienced eye will glance-parse a BUG+stack trace as a serious oops-level problem rather than a warning).

J

J

#3 lore raw
Quoted 7 lines · Jeremy Fitzhardinge
Quoted 2 lines · Ingo Molnar

Firstly, most WARN_ON()s are /bugs/, not warnings ... If it's a real warning, a KERN_INFO printk should be done.

It seems to me that either the warnings are really bugs, in which case they should be using BUG/BUG_ON, or they're not really bugs, in which case they should be presented differently.

No. A BUG() will terminate the current process which, aside from the loss of userspace data, can tangle up the kernel badly and deadlock or panic it.

If a bug can be fixed up or otherwise will not result in unstable behaviour with continued operation, then it should be a WARN.

Quoted 6 lines
Quoted 3 lines

Secondly, the reason i changed it to the 'BUG: ...' format is that i tried to make it easier for automated tools (and for users) to figure out that a kernel bug happened.

Well, are they bugs or not? I think people are more confused by the "BUG" prefix and stacktrace than helped by it (even an experienced eye will glance-parse a BUG+stack trace as a serious oops-level problem rather than a warning).

Definitely a bug. If the condition is not a bug then the code calling WARN is, so it is a bug no matter how you look at it ;)

Signature · 2 lines
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
#4 lore raw
In reply to Ingo Molnar

Hi!

Quoted 9 lines · Jeremy Fitzhardinge

Review

Quoted 5 lines

A warning is a warning, not a BUG.

(quoted diff) +1 -1
-		printk("BUG: warning at %s:%d/%s()\n", __FILE__,	\
+		printk("WARNING at %s:%d %s()\n", __FILE__,	\

i'm not really happy about this change.

Quoted 6 lines

Firstly, most WARN_ON()s are /bugs/, not warnings ... If it's a real warning, a KERN_INFO printk should be done.

Secondly, the reason i changed it to the 'BUG: ...' format is that i tried to make it easier for automated tools (and for users) to figure out that a kernel bug happened.

Well... but the message is really bad. It leads to users telling us "I hit BUG in kernel"...
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#5 lore raw
Quoted 14 lines · Pavel Machek

Hi!

Quoted 12 lines · Jeremy Fitzhardinge

Review

Quoted 5 lines

A warning is a warning, not a BUG.

(quoted diff) +1 -1
-		printk("BUG: warning at %s:%d/%s()\n", __FILE__,	\
+		printk("WARNING at %s:%d %s()\n", __FILE__,	\

i'm not really happy about this change.

Firstly, most WARN_ON()s are /bugs/, not warnings ... If it's a real warning, a KERN_INFO printk should be done.

Quoted 6 lines
Quoted 3 lines

Secondly, the reason i changed it to the 'BUG: ...' format is that i tried to make it easier for automated tools (and for users) to figure out that a kernel bug happened.

Well... but the message is really bad. It leads to users telling us "I hit BUG in kernel"...

But they *did* hit a BUG. It just so happens that the BUG was fixable. We want this reported because a WARN_ON should *never* be hit unless there's a bug. If people start getting "WARNING" messages, they will more likely not be reporting them.

As Ingo already said, if it is just a "warning" then a normal printk should be used.

-- Steve

#6 lore raw
Quoted 22 lines · Steven Rostedt
Quoted 14 lines · Pavel Machek

Hi!

Quoted 12 lines · Jeremy Fitzhardinge

Review

Quoted 5 lines

A warning is a warning, not a BUG.

(quoted diff) +1 -1
-		printk("BUG: warning at %s:%d/%s()\n", __FILE__,	\
+		printk("WARNING at %s:%d %s()\n", __FILE__,	\

i'm not really happy about this change.

Firstly, most WARN_ON()s are /bugs/, not warnings ... If it's a real warning, a KERN_INFO printk should be done.

Quoted 6 lines
Quoted 3 lines

Secondly, the reason i changed it to the 'BUG: ...' format is that i tried to make it easier for automated tools (and for users) to figure out that a kernel bug happened.

Well... but the message is really bad. It leads to users telling us "I hit BUG in kernel"...

Quoted 7 lines

But they *did* hit a BUG. It just so happens that the BUG was fixable. We want this reported because a WARN_ON should *never* be hit unless there's a bug. If people start getting "WARNING" messages, they will more likely not be reporting them.

As Ingo already said, if it is just a "warning" then a normal printk should be used.

Fine, then why is the macro called WARN_ON()? That's certainly highly confusing.

NONFATAL_BUG_ON()?

I hate people reporting BUG (or BUG()) when they hit WARN_ON(), and current wording certainly makes it easy.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html