Skip to contentSkip to conversation

wait_for_helper: SIGCHLD from user-space can lead to use-after-free

PATCH In review Oleg Nesterov 3 participants 5 messages Mar 10, 2010

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 wait_for_helper: SIGCHLD from user-space can lead to use-after-free

(this patch is orthogonal to other umh changes in -mm)

1. wait_for_helper() does allow_signal(SIGCHLD) to ensure the child can't autoreap itself.

However, this means that a spurious SIGCHILD from user-space can set TIF_SIGPENDING and:

- kernel_thread() or sys_wait4() can fail due to signal_pending()

- worse, wait4() can fail before ____call_usermodehelper() execs or exits. In this case the caller may kfree(subprocess_info) while the child still uses this memory.

Change the code to use SIG_DFL instead of magic "(void __user *)2" set by allow_signal(). This means that SIGCHLD won't be delivered, yet the child won't autoreap itsefl.

The problem is minor, only root can send a signal to this kthread.

2. If sys_wait4(&ret) fails it doesn't populate "ret", in this case wait_for_helper() reports a random value from uninitialized var.

With this patch sys_wait4() should never fail, but still it makes sense to initialize ret = -ECHILD so that the caller can notice the problem.

---

 kernel/kmod.c |    9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
kernel/kmod.c renamed from kernel/kmod.c~4_SIGCHLD_RACE +5 -4
@@ -194,15 +194,16 @@ static int wait_for_helper(void *data)
 	struct subprocess_info *sub_info = data;
 	pid_t pid;
 
-	/* Install a handler: if SIGCLD isn't handled sys_wait4 won't
-	 * populate the status, but will return -ECHILD. */
-	allow_signal(SIGCHLD);
+	/* If SIGCLD is ignored sys_wait4 won't populate the status. */
+	spin_lock_irq(&current->sighand->siglock);
+	current->sighand->action[SIGCHLD-1].sa.sa_handler = SIG_DFL;
+	spin_unlock_irq(&current->sighand->siglock);
 
 	pid = kernel_thread(____call_usermodehelper, sub_info, SIGCHLD);
 	if (pid < 0) {
 		sub_info->retval = pid;
 	} else {
-		int ret;
+		int ret = -ECHILD;
 
 		/*
 		 * Normally it is bogus to call wait4() from in-kernel because
Acked-by on v1 1/1
#1 lore raw
Quoted 20 lines · Oleg Nesterov

(this patch is orthogonal to other umh changes in -mm)

1. wait_for_helper() does allow_signal(SIGCHLD) to ensure the child can't autoreap itself.

However, this means that a spurious SIGCHILD from user-space can set TIF_SIGPENDING and:

- kernel_thread() or sys_wait4() can fail due to signal_pending()

- worse, wait4() can fail before ____call_usermodehelper() execs or exits. In this case the caller may kfree(subprocess_info) while the child still uses this memory.

Change the code to use SIG_DFL instead of magic "(void __user *)2" set by allow_signal(). This means that SIGCHLD won't be delivered, yet the child won't autoreap itsefl.

The problem is minor, only root can send a signal to this kthread.

Quoted 8 lines

2. If sys_wait4(&ret) fails it doesn't populate "ret", in this case wait_for_helper() reports a random value from uninitialized var.

With this patch sys_wait4() should never fail, but still it makes sense to initialize ret = -ECHILD so that the caller can notice the problem.

Quoted 1 line
#2 lore raw
In reply to Oleg Nesterov

SIGCHLD being blocked doesn't affect reaping, so SIG_IGN or sa_flags & SA_NOCLDWAIT is the only thing that would do this. How does that come about here in this kthread? Is it inherited from the instigating user process? If so, then SA_NOCLDWAIT is as much a problem as SIG_IGN. Or I guess maybe it's from ignore_signals() in kthreadd()? In that case SIG_IGN is indeed all that matters. (I don't really know all the kthread/kmod/userhelper code organization.)

Perhaps it would be cleaner to do:

	flush_signal_handlers(current, 1);

in wait_for_helper. That should make it redundant in ____call_usermodehelper, so it could be removed from there.

Thanks,
Roland

#3 lore raw
Quoted 8 lines · Roland McGrath

SIGCHLD being blocked doesn't affect reaping, so SIG_IGN or sa_flags & SA_NOCLDWAIT is the only thing that would do this. How does that come about here in this kthread? Is it inherited from the instigating user process? If so, then SA_NOCLDWAIT is as much a problem as SIG_IGN. Or I guess maybe it's from ignore_signals() in kthreadd()? In that case SIG_IGN is indeed all that matters. (I don't really know all the kthread/kmod/userhelper code organization.)

Yes. kthreads run with all signal ignored, this is inherited from kthreadd() which does ignore_signal().

Quoted 5 lines

Perhaps it would be cleaner to do:

	flush_signal_handlers(current, 1);

in wait_for_helper

I don't think this can work. SIG_DFL for SIGCHLD is OK because it is sig_kernel_ignore(). But, say, SIGHUP and other signals still should be ignored, otherwise we have the same problems with the unwanted signal_pending() this patch tries to avoid.

But even if we could do this,

Quoted 2 lines

That should make it redundant in ____call_usermodehelper, so it could be removed from there.

Please note that __call_usermodehelper() forks ____call_usermodehelper() too.

Oleg.

#4 lore raw
Quoted 2 lines

Yes. kthreads run with all signal ignored, this is inherited from kthreadd() which does ignore_signal().

Ok.

Quoted 4 lines

I don't think this can work. SIG_DFL for SIGCHLD is OK because it is sig_kernel_ignore(). But, say, SIGHUP and other signals still should be ignored, otherwise we have the same problems with the unwanted signal_pending() this patch tries to avoid.

Yes, I see.

Quoted 6 lines

But even if we could do this,

Quoted 2 lines

That should make it redundant in ____call_usermodehelper, so it could be removed from there.

Please note that __call_usermodehelper() forks ____call_usermodehelper() too.

Ok.

Thanks,
Roland