[PATCH v2] Return proper error code from register_kprobe()

[PATCH v2] Return proper error code from register_kprobe()

[Prashanth Nageshappa]

Missed lkml in my previous posting, hence resending....

register_kprobe() aborts if the address of the new request falls in a
prohibited area (such as ftrace pouch, __kprobes annotated functions,
non-kernel text addresses, jump label text). We however don't return the
right error on this abort, resulting in a silent failure - incorrect 
adding/reporting of kprobes ('perf probe do_fork+18' or
'perf probe mcount' for instance).

In V2 we are incorporating Masami Hiramatsu's  feedback.

This patch fixes it by returning -EINVAL upon failure.

While we are here, rename the label used for exit to be more appropriate.

Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Signed-off-by: Prashanth K Nageshappa <prashanth@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
---

 kernel/kprobes.c |   12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

Index: linux-14feb/kernel/kprobes.c
===================================================================
--- linux-14feb.orig/kernel/kprobes.c
+++ linux-14feb/kernel/kprobes.c
@@ -1334,8 +1334,10 @@ int __kprobes register_kprobe(struct kpr
 	if (!kernel_text_address((unsigned long) p->addr) ||
 	    in_kprobes_functions((unsigned long) p->addr) ||
 	    ftrace_text_reserved(p->addr, p->addr) ||
-	    jump_label_text_reserved(p->addr, p->addr))
-		goto fail_with_jump_label;
+	    jump_label_text_reserved(p->addr, p->addr)) {
+		ret = -EINVAL;
+		goto cannot_probe;
+	}

 	/* User can pass only KPROBE_FLAG_DISABLED to register_kprobe */
 	p->flags &= KPROBE_FLAG_DISABLED;
@@ -1352,7 +1354,7 @@ int __kprobes register_kprobe(struct kpr
 		 * its code to prohibit unexpected unloading.
 		 */
 		if (unlikely(!try_module_get(probed_mod)))
-			goto fail_with_jump_label;
+			goto cannot_probe;

 		/*
 		 * If the module freed .init.text, we couldn't insert
@@ -1361,7 +1363,7 @@ int __kprobes register_kprobe(struct kpr
 		if (within_module_init((unsigned long)p->addr, probed_mod) &&
 		    probed_mod->state != MODULE_STATE_COMING) {
 			module_put(probed_mod);
-			goto fail_with_jump_label;
+			goto cannot_probe;
 		}
 		/* ret will be updated by following code */
 	}
@@ -1409,7 +1411,7 @@ int __kprobes register_kprobe(struct kpr

 	return ret;

-fail_with_jump_label:
+cannot_probe:
 	preempt_enable();
 	jump_label_unlock();
 	return ret;

Re: [PATCH v2] Return proper error code from register_kprobe()

[Andrew Morton]

On Wed, 15 Feb 2012 12:46:03 +0530
Prashanth Nageshappa <prashanth@linux.vnet.ibm.com> wrote:

> register_kprobe() aborts if the address of the new request falls in a
> prohibited area (such as ftrace pouch, __kprobes annotated functions,

I'm sitting here wondering what you meant to type when you typed
"ftrace pouch".  I'm stumped!  But you're not allowed to tell us - that
would take all the fun out of it.

> non-kernel text addresses, jump label text). We however don't return the
> right error on this abort, resulting in a silent failure - incorrect 
> adding/reporting of kprobes ('perf probe do_fork+18' or
> 'perf probe mcount' for instance).
> 
> In V2 we are incorporating Masami Hiramatsu's  feedback.
> 
> This patch fixes it by returning -EINVAL upon failure.
> 
> While we are here, rename the label used for exit to be more appropriate.
> 
> Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
> Signed-off-by: Prashanth K Nageshappa <prashanth@linux.vnet.ibm.com>
> Cc: stable@vger.kernel.org

So we want this for 3.3 with a -stable backport.