[PATCH] kmod: prevent kmod_loop_msg overflow in __request_module()

[PATCH] kmod: prevent kmod_loop_msg overflow in __request_module()

[Jiri Kosina]

Due to post-decrement in condition of kmod_loop_msg in __request_module(),
the system log can be spammed by much more than 5 instances of the 'runaway
loop' message if the number of events triggering it makes the kmod_loop_msg
to overflow.

Fix that by making sure we never increment it past the threshold.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
---
 kernel/kmod.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/kernel/kmod.c b/kernel/kmod.c
index ddc7644..a4bea97 100644
--- a/kernel/kmod.c
+++ b/kernel/kmod.c
@@ -114,10 +114,12 @@ int __request_module(bool wait, const char *fmt, ...)
 	atomic_inc(&kmod_concurrent);
 	if (atomic_read(&kmod_concurrent) > max_modprobes) {
 		/* We may be blaming an innocent here, but unlikely */
-		if (kmod_loop_msg++ < 5)
+		if (kmod_loop_msg < 5) {
 			printk(KERN_ERR
 			       "request_module: runaway loop modprobe %s\n",
 			       module_name);
+			kmod_loop_msg++;
+		}
 		atomic_dec(&kmod_concurrent);
 		return -ENOMEM;
 	}
-- 
1.7.3.1

Re: [PATCH] kmod: prevent kmod_loop_msg overflow in __request_module()

[Andrew Morton]

On Thu, 29 Sep 2011 17:50:45 +0200 (CEST)
Jiri Kosina <jkosina@suse.cz> wrote:

> Due to post-decrement

(post-increment)

> in condition of kmod_loop_msg in __request_module(),
> the system log can be spammed by much more than 5 instances of the 'runaway
> loop' message if the number of events triggering it makes the kmod_loop_msg
> to overflow.
> 
> Fix that by making sure we never increment it past the threshold.
> 
> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
> ---
>  kernel/kmod.c |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/kernel/kmod.c b/kernel/kmod.c
> index ddc7644..a4bea97 100644
> --- a/kernel/kmod.c
> +++ b/kernel/kmod.c
> @@ -114,10 +114,12 @@ int __request_module(bool wait, const char *fmt, ...)
>  	atomic_inc(&kmod_concurrent);
>  	if (atomic_read(&kmod_concurrent) > max_modprobes) {
>  		/* We may be blaming an innocent here, but unlikely */
> -		if (kmod_loop_msg++ < 5)
> +		if (kmod_loop_msg < 5) {
>  			printk(KERN_ERR
>  			       "request_module: runaway loop modprobe %s\n",
>  			       module_name);
> +			kmod_loop_msg++;
> +		}
>  		atomic_dec(&kmod_concurrent);
>  		return -ENOMEM;
>  	}

Well, it will take two billion passes through here to cause the
overflow.  I really hope you found this problem by inspection!

Re: [PATCH] kmod: prevent kmod_loop_msg overflow in __request_module()

[Jiri Kosina]

On Thu, 29 Sep 2011, Andrew Morton wrote:

> > Due to post-decrement
> 
> (post-increment)

Indeed, thanks for noticing.

> > in condition of kmod_loop_msg in __request_module(),
> > the system log can be spammed by much more than 5 instances of the 'runaway
> > loop' message if the number of events triggering it makes the kmod_loop_msg
> > to overflow.
> > 
> > Fix that by making sure we never increment it past the threshold.
> > 
> > Signed-off-by: Jiri Kosina <jkosina@suse.cz>
> > ---
> >  kernel/kmod.c |    4 +++-
> >  1 files changed, 3 insertions(+), 1 deletions(-)
> > 
> > diff --git a/kernel/kmod.c b/kernel/kmod.c
> > index ddc7644..a4bea97 100644
> > --- a/kernel/kmod.c
> > +++ b/kernel/kmod.c
> > @@ -114,10 +114,12 @@ int __request_module(bool wait, const char *fmt, ...)
> >  	atomic_inc(&kmod_concurrent);
> >  	if (atomic_read(&kmod_concurrent) > max_modprobes) {
> >  		/* We may be blaming an innocent here, but unlikely */
> > -		if (kmod_loop_msg++ < 5)
> > +		if (kmod_loop_msg < 5) {
> >  			printk(KERN_ERR
> >  			       "request_module: runaway loop modprobe %s\n",
> >  			       module_name);
> > +			kmod_loop_msg++;
> > +		}
> >  		atomic_dec(&kmod_concurrent);
> >  		return -ENOMEM;
> >  	}
> 
> Well, it will take two billion passes through here to cause the
> overflow.  I really hope you found this problem by inspection!

Unfortunately not, it was more painful experience than code reading :)

Thanks,

-- 
Jiri Kosina
SUSE Labs

Re: [PATCH] kmod: prevent kmod_loop_msg overflow in __request_module()

[Rusty Russell]

On Fri, 30 Sep 2011 00:18:17 +0200 (CEST), Jiri Kosina <jkosina@suse.cz> wrote:
> On Thu, 29 Sep 2011, Andrew Morton wrote:
> > Well, it will take two billion passes through here to cause the
> > overflow.  I really hope you found this problem by inspection!
> 
> Unfortunately not, it was more painful experience than code reading :)

Cool!

Changing it to unsigned would have helped too.  But I've applied
this....

Thanks,
Rusty.

Re: [PATCH] kmod: prevent kmod_loop_msg overflow in __request_module()

[Jiri Kosina]

On Tue, 4 Oct 2011, Rusty Russell wrote:

> > > Well, it will take two billion passes through here to cause the
> > > overflow.  I really hope you found this problem by inspection!
> > 
> > Unfortunately not, it was more painful experience than code reading :)
> 
> Cool!
> 
> Changing it to unsigned would have helped too.  

Still woudln't have helped much in cases of __request_module() being 
called incredibly quickly in an infinite loop. Changing it to unsigned 
will cause the messages being printed in bunches of five, but the effect 
will be the same -- the log being infinitely filled by the printk()s. 
Computers can do infinite loops really quickly today :)

-- 
Jiri Kosina
SUSE Labs