[PATCH] pidns: remove recursion from free_pid_ns() v5

[PATCH] pidns: remove recursion from free_pid_ns() v5

[Cyrill Gorcunov]

The free_pid_ns function done in recursion fashion:

free_pid_ns(parent)
  put_pid_ns(parent)
    kref_put(&ns->kref, free_pid_ns);
      free_pid_ns

thus if there was a huge nesting of namespaces the userspace
may trigger avalanche calling of free_pid_ns leading to
kernel stack exhausting and a panic eventually.

This patch turns the recursion into iterative loop.

v5 (from oleg@):
 - Drop @ret variable
 - Make put_pid_ns non-inline since it grows in size,
   in turn make free_pid_ns static

Based-on-patch-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Greg KH <greg@kroah.com>
---
 include/linux/pid_namespace.h |    8 +-------
 kernel/pid_namespace.c        |   19 +++++++++++++------
 2 files changed, 14 insertions(+), 13 deletions(-)

Index: linux-2.6.git/include/linux/pid_namespace.h
===================================================================
--- linux-2.6.git.orig/include/linux/pid_namespace.h
+++ linux-2.6.git/include/linux/pid_namespace.h
@@ -47,15 +47,9 @@ static inline struct pid_namespace *get_
 }
 
 extern struct pid_namespace *copy_pid_ns(unsigned long flags, struct pid_namespace *ns);
-extern void free_pid_ns(struct kref *kref);
 extern void zap_pid_ns_processes(struct pid_namespace *pid_ns);
 extern int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd);
-
-static inline void put_pid_ns(struct pid_namespace *ns)
-{
-	if (ns != &init_pid_ns)
-		kref_put(&ns->kref, free_pid_ns);
-}
+extern void put_pid_ns(struct pid_namespace *ns);
 
 #else /* !CONFIG_PID_NS */
 #include <linux/err.h>
Index: linux-2.6.git/kernel/pid_namespace.c
===================================================================
--- linux-2.6.git.orig/kernel/pid_namespace.c
+++ linux-2.6.git/kernel/pid_namespace.c
@@ -132,17 +132,24 @@ struct pid_namespace *copy_pid_ns(unsign
 	return create_pid_namespace(old_ns);
 }
 
-void free_pid_ns(struct kref *kref)
+static void free_pid_ns(struct kref *kref)
 {
-	struct pid_namespace *ns, *parent;
+	struct pid_namespace *ns;
 
 	ns = container_of(kref, struct pid_namespace, kref);
-
-	parent = ns->parent;
 	destroy_pid_namespace(ns);
+}
+
+void put_pid_ns(struct pid_namespace *ns)
+{
+	struct pid_namespace *parent;
 
-	if (parent != NULL)
-		put_pid_ns(parent);
+	while (ns != &init_pid_ns) {
+		parent = ns->parent;
+		if (!kref_put(&ns->kref, free_pid_ns))
+			break;
+		ns = parent;
+	}
 }
 
 void zap_pid_ns_processes(struct pid_namespace *pid_ns)

Re: [PATCH] pidns: remove recursion from free_pid_ns() v5

[Andrew Morton]

On Thu, 11 Oct 2012 00:42:56 +0400
Cyrill Gorcunov <gorcunov@openvz.org> wrote:

> The free_pid_ns function done in recursion fashion:
> 
> free_pid_ns(parent)
>   put_pid_ns(parent)
>     kref_put(&ns->kref, free_pid_ns);
>       free_pid_ns
> 
> thus if there was a huge nesting of namespaces the userspace
> may trigger avalanche calling of free_pid_ns leading to
> kernel stack exhausting and a panic eventually.
> 
> This patch turns the recursion into iterative loop.
> 
> v5 (from oleg@):
>  - Drop @ret variable
>  - Make put_pid_ns non-inline since it grows in size,
>    in turn make free_pid_ns static

OK, let's try that.  I'll sit on this until -rc2 to give it a bit of
time to cook.

A -stable backport might be needed.  What capabilities does userspace
need to be able to trigger the kernel stack overflow?

Re: [PATCH] pidns: remove recursion from free_pid_ns() v5

[Eric W. Biederman]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Thu, 11 Oct 2012 00:42:56 +0400
> Cyrill Gorcunov <gorcunov@openvz.org> wrote:
>
>> The free_pid_ns function done in recursion fashion:
>> 
>> free_pid_ns(parent)
>>   put_pid_ns(parent)
>>     kref_put(&ns->kref, free_pid_ns);
>>       free_pid_ns
>> 
>> thus if there was a huge nesting of namespaces the userspace
>> may trigger avalanche calling of free_pid_ns leading to
>> kernel stack exhausting and a panic eventually.
>> 
>> This patch turns the recursion into iterative loop.
>> 
>> v5 (from oleg@):
>>  - Drop @ret variable
>>  - Make put_pid_ns non-inline since it grows in size,
>>    in turn make free_pid_ns static
>
> OK, let's try that.  I'll sit on this until -rc2 to give it a bit of
> time to cook.
>
> A -stable backport might be needed.  What capabilities does userspace
> need to be able to trigger the kernel stack overflow?

CAP_SYS_ADMIN is required to create a new pid namespace today.

With a little luck the user namespace bits that allow unprivelged
creation of pid namespaces will be ready for 3.8.

Eric

Re: [PATCH] pidns: remove recursion from free_pid_ns() v5

[Cyrill Gorcunov]

On Wed, Oct 10, 2012 at 01:54:08PM -0700, Andrew Morton wrote:
> On Thu, 11 Oct 2012 00:42:56 +0400
> Cyrill Gorcunov <gorcunov@openvz.org> wrote:
> 
> > The free_pid_ns function done in recursion fashion:
> > 
> > free_pid_ns(parent)
> >   put_pid_ns(parent)
> >     kref_put(&ns->kref, free_pid_ns);
> >       free_pid_ns
> > 
> > thus if there was a huge nesting of namespaces the userspace
> > may trigger avalanche calling of free_pid_ns leading to
> > kernel stack exhausting and a panic eventually.
> > 
> > This patch turns the recursion into iterative loop.
> > 
> > v5 (from oleg@):
> >  - Drop @ret variable
> >  - Make put_pid_ns non-inline since it grows in size,
> >    in turn make free_pid_ns static
> 
> OK, let's try that.  I'll sit on this until -rc2 to give it a bit of
> time to cook.
> 
> A -stable backport might be needed.  What capabilities does userspace
> need to be able to trigger the kernel stack overflow?

I believe it'll apply on stable even in current form. As Eric mentioned
CAP_SYS_ADMIN is required (so it's not that urgent i think).