proc-ns: use d_set_d_op() API to set dentry ops in proc_ns_instantiat…

…e().

The namespace cleanup path leaks a dentry which holds a reference count
on a network namespace.  Keeping that network namespace from being freed
when the last user goes away.  Leaving things like vlan devices in the
leaked network namespace.

If you use ip netns add for much real work this problem becomes apparent
pretty quickly.  It light testing the problem hides because frequently
you simply don't notice the leak.

Use d_set_d_op() so that DCACHE_OP_* flags are set correctly.

This issue exists back to 3.0.

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Reported-by: Justin Pettit <jpettit@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Cc: David Miller <davem@davemloft.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/proc/namespaces.c

@@ -53,7 +53,7 @@ static struct dentry *proc_ns_instantiate(struct inode *dir,
 	ei->ns_ops    = ns_ops;
 	ei->ns	      = ns;
 
-	dentry->d_op = &pid_dentry_operations;
+	d_set_d_op(dentry, &pid_dentry_operations);
 	d_add(dentry, inode);
 	/* Close the race of the process dying before we return the dentry */
 	if (pid_revalidate(dentry, NULL))