proc: Use nd_jump_link in proc_ns_follow_link

commit db04dc67 upstream.

Update proc_ns_follow_link to use nd_jump_link instead of just
manually updating nd.path.dentry.

This fixes the BUG_ON(nd->inode != parent->d_inode) reported by Dave
Jones and reproduced trivially with mkdir /proc/self/ns/uts/a.

Sigh it looks like the VFS change to require use of nd_jump_link
happend while proc_ns_follow_link was baking and since the common case
of proc_ns_follow_link continued to work without problems the need for
making this change was overlooked.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs/proc/namespaces.c

@@ -118,7 +118,7 @@ static void *proc_ns_follow_link(struct dentry *dentry, struct nameidata *nd)
 	struct super_block *sb = inode->i_sb;
 	struct proc_inode *ei = PROC_I(inode);
 	struct task_struct *task;
-	struct dentry *ns_dentry;
+	struct path ns_path;
 	void *error = ERR_PTR(-EACCES);
 
 	task = get_proc_task(inode);
@@ -128,14 +128,14 @@ static void *proc_ns_follow_link(struct dentry *dentry, struct nameidata *nd)
 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
 		goto out_put_task;
 
-	ns_dentry = proc_ns_get_dentry(sb, task, ei->ns_ops);
-	if (IS_ERR(ns_dentry)) {
-		error = ERR_CAST(ns_dentry);
+	ns_path.dentry = proc_ns_get_dentry(sb, task, ei->ns_ops);
+	if (IS_ERR(ns_path.dentry)) {
+		error = ERR_CAST(ns_path.dentry);
 		goto out_put_task;
 	}
 
-	dput(nd->path.dentry);
-	nd->path.dentry = ns_dentry;
+	ns_path.mnt = mntget(nd->path.mnt);
+	nd_jump_link(nd, &ns_path);
 	error = NULL;
 
 out_put_task: