[PATCH] Don't oops when unregistering unknown kprobes

[PATCH] Don't oops when unregistering unknown kprobes

[Frederik Deweerdt]

[-- Attachment #1: Type: text/plain, Size: 455 bytes --]

Hi Prasanna,

Here's a patch that handles gracefully attempts to unregister
unregistered kprobes (ie. a warning message instead of an oops).
Patch is against 2.6.12-rc3

Signed-off-by: Frederik Deweerdt <frederik.deweerdt@laposte.net>

Regards,
Frederik

-- 
o----------------------------------------------o
| http://open-news.net : l'info alternative    |
| Tech - Sciences - Politique - International  |
o----------------------------------------------o

[-- Attachment #2: dont.oops.on.unregister.unknown.kprobe.patch --]
[-- Type: text/plain, Size: 508 bytes --]

--- linux-2.6.12-rc3/kernel/kprobes.c	2005-04-26 16:35:22.000000000 +0200
+++ linux-2.6.12-rc3-devel/kernel/kprobes.c	2005-04-26 16:44:57.000000000 +0200
@@ -107,6 +107,13 @@ rm_kprobe:
 void unregister_kprobe(struct kprobe *p)
 {
 	unsigned long flags;
+
+	if (!get_kprobe(p)) {
+		printk(KERN_WARNING "Warning: Attempt to unregister "
+					"unknown kprobe (addr:0x%lx)\n",
+					(unsigned long) p);
+		return;
+	}
 	arch_remove_kprobe(p);
 	spin_lock_irqsave(&kprobe_lock, flags);
 	*p->addr = p->opcode;

Re: [PATCH] Don't oops when unregistering unknown kprobes

[Prasanna S Panchamukhi]

> @@ -107,6 +107,13 @@ rm_kprobe:
>  void unregister_kprobe(struct kprobe *p)
>  {
>  	unsigned long flags;
> +
> +	if (!get_kprobe(p)) {
> +		printk(KERN_WARNING "Warning: Attempt to unregister "
> +					"unknown kprobe (addr:0x%lx)\n",
> +					(unsigned long) p);
> +		return;
> +	}

This is wrong. You should call get_kprobe() with spin_lock().
 

-Prasanna

-- 

Prasanna S Panchamukhi
Linux Technology Center
India Software Labs, IBM Bangalore
Ph: 91-80-25044636
<prasanna@in.ibm.com>

Re: [PATCH] Don't oops when unregistering unknown kprobes

[Chris Wedgwood]

On Tue, Apr 26, 2005 at 06:22:03PM +0200, Frederik Deweerdt wrote:

> Here's a patch that handles gracefully attempts to unregister
> unregistered kprobes (ie. a warning message instead of an oops).
> Patch is against 2.6.12-rc3

Why not -EINVAL instead of the printk?

Re: [PATCH] Don't oops when unregistering unknown kprobes

[Frederik Deweerdt]

[-- Attachment #1: Type: text/plain, Size: 487 bytes --]

Le 26/04/05 21:57 +0530, Prasanna S Panchamukhi écrivit:
> This is wrong. You should call get_kprobe() with spin_lock().
>  
Right, corrected patch attached. It also sets flags to zero.

Signed-off-by: Frederik Deweerdt <frederik.deweerdt@laposte.net>

Regards,
Frederik

-- 
o----------------------------------------------o
| http://open-news.net : l'info alternative    |
| Tech - Sciences - Politique - International  |
o----------------------------------------------o

[-- Attachment #2: dont.oops.on.unregister.unknown.kprobe.patch --]
[-- Type: text/plain, Size: 769 bytes --]

--- linux-2.6.12-rc3/kernel/kprobes.c	2005-04-26 16:35:22.000000000 +0200
+++ linux-2.6.12-rc3-devel/kernel/kprobes.c	2005-04-26 23:18:47.000000000 +0200
@@ -106,13 +106,22 @@ rm_kprobe:
 
 void unregister_kprobe(struct kprobe *p)
 {
-	unsigned long flags;
+	unsigned long flags = 0;
+
+	spin_lock_irqsave(&kprobe_lock, flags);
+	if (!get_kprobe(p)) {
+		printk(KERN_WARNING "Warning: Attempt to unregister "
+					"unknown kprobe (addr:0x%lx)\n",
+					(unsigned long) p);
+		goto out;
+	}
 	arch_remove_kprobe(p);
 	spin_lock_irqsave(&kprobe_lock, flags);
 	*p->addr = p->opcode;
 	hlist_del(&p->hlist);
 	flush_icache_range((unsigned long) p->addr,
 			   (unsigned long) p->addr + sizeof(kprobe_opcode_t));
+out:
 	spin_unlock_irqrestore(&kprobe_lock, flags);
 }
 

Re: [PATCH] Don't oops when unregistering unknown kprobes

[Jesper Juhl]

On Tue, 26 Apr 2005, Frederik Deweerdt wrote:

> Le 26/04/05 21:57 +0530, Prasanna S Panchamukhi écrivit:
> > This is wrong. You should call get_kprobe() with spin_lock().
> >  
> Right, corrected patch attached. It also sets flags to zero.
> Signed-off-by: Frederik Deweerdt <frederik.deweerdt@laposte.net>

> --- linux-2.6.12-rc3/kernel/kprobes.c   2005-04-26 16:35:22.000000000 +0200
> +++ linux-2.6.12-rc3-devel/kernel/kprobes.c     2005-04-26 23:18:47.000000000 +0200
> @@ -106,13 +106,22 @@ rm_kprobe:
> 
>  void unregister_kprobe(struct kprobe *p)
>  {
> -       unsigned long flags;
> +       unsigned long flags = 0;
> +
> +       spin_lock_irqsave(&kprobe_lock, flags);
          ^^^
           +one...

> +       if (!get_kprobe(p)) {
> +               printk(KERN_WARNING "Warning: Attempt to unregister "
> +                                       "unknown kprobe (addr:0x%lx)\n",
> +                                       (unsigned long) p);
> +               goto out;
> +       }
>         arch_remove_kprobe(p);
>         spin_lock_irqsave(&kprobe_lock, flags);
          ^^^
           +two...

>         *p->addr = p->opcode;
>         hlist_del(&p->hlist);
>         flush_icache_range((unsigned long) p->addr,
>                            (unsigned long) p->addr + sizeof(kprobe_opcode_t));
> +out:
>         spin_unlock_irqrestore(&kprobe_lock, flags);
          ^^^
           -one...

>  }

Seems to me this will end up calling spin_lock_irqsave() twice, but only 
spin_unlock_irqrestore once in the non-failing case... hmm..

Also, as Chris Wedgwood asked, why not simply  return -EINVAL;  instead of 
the printk()?  Does the user really care that we tried to unregister a 
nonexisting kprobe? and if you really think someone would like to know 
then I'd personally say that KERN_DEBUG should be sufficient.

I'd suggest something like this :

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>

--- linux-2.6.12-rc2-mm3-orig/kernel/kprobes.c	2005-04-11 21:20:56.000000000 +0200
+++ linux-2.6.12-rc2-mm3/kernel/kprobes.c	2005-04-27 00:04:23.000000000 +0200
@@ -108,16 +108,24 @@ rm_kprobe:
 	return ret;
 }
 
-void unregister_kprobe(struct kprobe *p)
+int unregister_kprobe(struct kprobe *p)
 {
-	unsigned long flags;
-	arch_remove_kprobe(p);
+	unsigned long flags = 0;
+	int ret = 0;
+
 	spin_lock_irqsave(&kprobe_lock, flags);
+	if (!get_kprobe(p)) {
+		ret = -EINVAL;
+		goto out;
+	}
+	arch_remove_kprobe(p);
 	*p->addr = p->opcode;
 	hlist_del(&p->hlist);
 	flush_icache_range((unsigned long) p->addr,
 			   (unsigned long) p->addr + sizeof(kprobe_opcode_t));
+out:
 	spin_unlock_irqrestore(&kprobe_lock, flags);
+	return ret;
 }
 
 


And if you really want that printk in there, then this should make you 
happy - but personally I don't see the big need : 
+       if (!get_kprobe(p)) {
+               printk(KERN_DEBUG "Warning: Attempt to unregister "
+                                 "unknown kprobe (addr:0x%lx)\n",
+                                 (unsigned long) p);
+               ret = -EINVAL;
+               goto out;
+       }


Ohh and btw, would you mind posting patches inline? Having to save the 
patch, add quotes to it and then read it back into the reply mail to 
comment on it is a pain...

And don't trim the CC: list when replying to mails on LKML, please.


-- 
Jesper Juhl

Re: [PATCH] Don't oops when unregistering unknown kprobes

[Frederik Deweerdt]

Le 27/04/05 00:09 +0200, Jesper Juhl écrivit:
> On Tue, 26 Apr 2005, Frederik Deweerdt wrote:
> Seems to me this will end up calling spin_lock_irqsave() twice, but only 
> spin_unlock_irqrestore once in the non-failing case... hmm..

Indeed, I made an obvious mistake there, sorry.

> 
> Also, as Chris Wedgwood asked, why not simply  return -EINVAL;  instead of 
> the printk()?  Does the user really care that we tried to unregister a 
> nonexisting kprobe? and if you really think someone would like to know 
> then I'd personally say that KERN_DEBUG should be sufficient.

I wanted to make the patch minimal, but it does make sense. 

> I'd suggest something like this :
> [ ... ]

You should also change the prototype in include/kernel/kprobes.h:

--- linux-2.6.12-rc3/include/linux/kprobes.h    2005-03-02 08:37:50.000000000 +0100
+++ linux-2.6.12-rc3-devel/include/linux/kprobes.h      2005-04-27 00:23:09.000000000 +0200
@@ -103,7 +103,7 @@ extern void show_registers(struct pt_reg
 struct kprobe *get_kprobe(void *addr);
 
 int register_kprobe(struct kprobe *p);
-void unregister_kprobe(struct kprobe *p);
+int unregister_kprobe(struct kprobe *p);
 int setjmp_pre_handler(struct kprobe *, struct pt_regs *);
 int longjmp_break_handler(struct kprobe *, struct pt_regs *);
 int register_jprobe(struct jprobe *p);

Regards,
Frederik
PS: I've fixed my mailing habits, sorry for inconvenience :)

-- 
o----------------------------------------------o
| http://open-news.net : l'info alternative    |
| Tech - Sciences - Politique - International  |
o----------------------------------------------o

Re: [PATCH] Don't oops when unregistering unknown kprobes

[Prasanna S Panchamukhi]

Chris,
On Tue, Apr 26, 2005 at 12:48:41PM -0700, Chris Wedgwood wrote:
> On Tue, Apr 26, 2005 at 06:22:03PM +0200, Frederik Deweerdt wrote:
> 
> > Here's a patch that handles gracefully attempts to unregister
> > unregistered kprobes (ie. a warning message instead of an oops).
> > Patch is against 2.6.12-rc3
> 
> Why not -EINVAL instead of the printk?

This problem has been already fixed, pls see the URL below.
http://lkml.org/lkml/2005/4/11/112

Currently this patch is in -mm tree. But this patch does not have the
prink reporting failure.

Thanks
Prasanna

-- 

Prasanna S Panchamukhi
Linux Technology Center
India Software Labs, IBM Bangalore
Ph: 91-80-25044636
<prasanna@in.ibm.com>