tcp: fix possible deadlock in TCP stack vs BPF filter

Filtering the ACK packet was not put at the right place.

At this place, we already allocated a child and put it
into accept queue.

We absolutely need to call tcp_child_process() to release
its spinlock, or we will deadlock at accept() or close() time.

Found by syzkaller team (Thanks a lot !)

Fixes: 8fac365f ("tcp: Add a tcp_filter hook before handle ack packet")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Chenbo Feng <fengc@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/ipv4/tcp_ipv4.c

@@ -1722,6 +1722,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		 */
 		sock_hold(sk);
 		refcounted = true;
+		if (tcp_filter(sk, skb))
+			goto discard_and_relse;
 		nsk = tcp_check_req(sk, skb, req, false);
 		if (!nsk) {
 			reqsk_put(req);
@@ -1729,8 +1731,6 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		}
 		if (nsk == sk) {
 			reqsk_put(req);
-		} else if (tcp_filter(sk, skb)) {
-			goto discard_and_relse;
 		} else if (tcp_child_process(sk, nsk, skb)) {
 			tcp_v4_send_reset(nsk, skb);
 			goto discard_and_relse;

net/ipv6/tcp_ipv6.c

@@ -1456,6 +1456,8 @@ static int tcp_v6_rcv(struct sk_buff *skb)
 		}
 		sock_hold(sk);
 		refcounted = true;
+		if (tcp_filter(sk, skb))
+			goto discard_and_relse;
 		nsk = tcp_check_req(sk, skb, req, false);
 		if (!nsk) {
 			reqsk_put(req);
@@ -1464,8 +1466,6 @@ static int tcp_v6_rcv(struct sk_buff *skb)
 		if (nsk == sk) {
 			reqsk_put(req);
 			tcp_v6_restore_cb(skb);
-		} else if (tcp_filter(sk, skb)) {
-			goto discard_and_relse;
 		} else if (tcp_child_process(sk, nsk, skb)) {
 			tcp_v6_send_reset(nsk, skb);
 			goto discard_and_relse;