net: rds: acquire refcount on TCP sockets

syzbot is reporting use-after-free read in tcp_retransmit_timer() [1],
for TCP socket used by RDS is accessing sock_net() without acquiring a
refcount on net namespace. Since TCP's retransmission can happen after
a process which created net namespace terminated, we need to explicitly
acquire a refcount.

Link: https://syzkaller.appspot.com/bug?extid=694120e1002c117747ed [1]
Reported-by: syzbot <syzbot+694120e1002c117747ed@syzkaller.appspotmail.com>
Fixes: 26abe14 ("net: Modify sk_alloc to not reference count the netns of kernel sockets.")
Fixes: 8a68173 ("net: sk_clone_lock() should only do get_net() if the parent is not a kernel socket")
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Tested-by: syzbot <syzbot+694120e1002c117747ed@syzkaller.appspotmail.com>
Link: https://lore.kernel.org/r/a5fb1fc4-2284-3359-f6a0-e4e390239d7b@I-love.SAKURA.ne.jp
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net/rds/tcp.c

@@ -495,6 +495,14 @@ void rds_tcp_tune(struct socket *sock)
 
 	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
+	/* TCP timer functions might access net namespace even after
+	 * a process which created this net namespace terminated.
+	 */
+	if (!sk->sk_net_refcnt) {
+		sk->sk_net_refcnt = 1;
+		get_net_track(net, &sk->ns_tracker, GFP_KERNEL);
+		sock_inuse_add(net, 1);
+	}
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
 		sk->sk_userlocks |= SOCK_SNDBUF_LOCK;