pidns: Outlaw thread creation after unshare(CLONE_NEWPID)
authorEric W. Biederman <ebiederm@xmission.com>
Fri, 21 Dec 2012 03:26:06 +0000 (19:26 -0800)
committerEric W. Biederman <ebiederm@xmission.com>
Tue, 25 Dec 2012 06:53:14 +0000 (22:53 -0800)
The sequence:
unshare(CLONE_NEWPID)
clone(CLONE_THREAD|CLONE_SIGHAND|CLONE_VM)

Creates a new process in the new pid namespace without setting
pid_ns->child_reaper.  After forking this results in a NULL
pointer dereference.

Avoid this and other nonsense scenarios that can show up after
creating a new pid namespace with unshare by adding a new
check in copy_prodcess.

Pointed-out-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
kernel/fork.c

index a31b823b3c2d6d4e6254128f00e36c6c1b048c37..65ca6d27f24e1065013a428f12f33935b1b18490 100644 (file)
@@ -1166,6 +1166,14 @@ static struct task_struct *copy_process(unsigned long clone_flags,
                                current->signal->flags & SIGNAL_UNKILLABLE)
                return ERR_PTR(-EINVAL);
 
+       /*
+        * If the new process will be in a different pid namespace
+        * don't allow the creation of threads.
+        */
+       if ((clone_flags & (CLONE_VM|CLONE_NEWPID)) &&
+           (task_active_pid_ns(current) != current->nsproxy->pid_ns))
+               return ERR_PTR(-EINVAL);
+
        retval = security_task_create(clone_flags);
        if (retval)
                goto fork_out;