tcp: prepare fastopen code for upcoming listener changes

While auditing TCP stack for upcoming 'lockless' listener changes,
I found I had to change fastopen_init_queue() to properly init the object
before publishing it.

Otherwise an other cpu could try to lock the spinlock before it gets
properly initialized.

Instead of adding appropriate barriers, just remove dynamic memory
allocations :
- Structure is 28 bytes on 64bit arches. Using additional 8 bytes
  for holding a pointer seems overkill.
- Two listeners can share same cache line and performance would suffer.

If we really want to save few bytes, we would instead dynamically allocate
whole struct request_sock_queue in the future.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 19 insertions, 34 deletions

diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 8a55664..3af85ee 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -219,17 +219,13 @@ int inet_listen(struct socket *sock, int backlog)
 		 * shutdown() (rather than close()).
 		 */
 		if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
-		    !inet_csk(sk)->icsk_accept_queue.fastopenq) {
+		    !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
 			if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
-				err = fastopen_init_queue(sk, backlog);
+				fastopen_queue_tune(sk, backlog);
 			else if ((sysctl_tcp_fastopen &
 				  TFO_SERVER_WO_SOCKOPT2) != 0)
-				err = fastopen_init_queue(sk,
+				fastopen_queue_tune(sk,
 				    ((uint)sysctl_tcp_fastopen) >> 16);
-			else
-				err = 0;
-			if (err)
-				goto out;
 
 			tcp_fastopen_init_key_once(true);
 		}
diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
index 694a5e8..e152788 100644
--- a/net/ipv4/inet_connection_sock.c
+++ b/net/ipv4/inet_connection_sock.c
@@ -335,9 +335,8 @@ struct sock *inet_csk_accept(struct sock *sk, int flags, int *err)
 
 	sk_acceptq_removed(sk);
 	if (sk->sk_protocol == IPPROTO_TCP &&
-	    tcp_rsk(req)->tfo_listener &&
-	    queue->fastopenq) {
-		spin_lock_bh(&queue->fastopenq->lock);
+	    tcp_rsk(req)->tfo_listener) {
+		spin_lock_bh(&queue->fastopenq.lock);
 		if (tcp_rsk(req)->tfo_listener) {
 			/* We are still waiting for the final ACK from 3WHS
 			 * so can't free req now. Instead, we set req->sk to
@@ -348,7 +347,7 @@ struct sock *inet_csk_accept(struct sock *sk, int flags, int *err)
 			req->sk = NULL;
 			req = NULL;
 		}
-		spin_unlock_bh(&queue->fastopenq->lock);
+		spin_unlock_bh(&queue->fastopenq.lock);
 	}
 out:
 	release_sock(sk);
@@ -886,12 +885,12 @@ void inet_csk_listen_stop(struct sock *sk)
 		sk_acceptq_removed(sk);
 		reqsk_put(req);
 	}
-	if (queue->fastopenq) {
+	if (queue->fastopenq.rskq_rst_head) {
 		/* Free all the reqs queued in rskq_rst_head. */
-		spin_lock_bh(&queue->fastopenq->lock);
-		acc_req = queue->fastopenq->rskq_rst_head;
-		queue->fastopenq->rskq_rst_head = NULL;
-		spin_unlock_bh(&queue->fastopenq->lock);
+		spin_lock_bh(&queue->fastopenq.lock);
+		acc_req = queue->fastopenq.rskq_rst_head;
+		queue->fastopenq.rskq_rst_head = NULL;
+		spin_unlock_bh(&queue->fastopenq.lock);
 		while ((req = acc_req) != NULL) {
 			acc_req = req->dl_next;
 			reqsk_put(req);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index b8b8fa1..3c96fa8 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2253,13 +2253,6 @@ int tcp_disconnect(struct sock *sk, int flags)
 }
 EXPORT_SYMBOL(tcp_disconnect);
 
-void tcp_sock_destruct(struct sock *sk)
-{
-	inet_sock_destruct(sk);
-
-	kfree(inet_csk(sk)->icsk_accept_queue.fastopenq);
-}
-
 static inline bool tcp_can_repair_sock(const struct sock *sk)
 {
 	return ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN) &&
@@ -2581,7 +2574,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		    TCPF_LISTEN))) {
 			tcp_fastopen_init_key_once(true);
 
-			err = fastopen_init_queue(sk, val);
+			fastopen_queue_tune(sk, val);
 		} else {
 			err = -EINVAL;
 		}
@@ -2849,10 +2842,7 @@ static int do_tcp_getsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_FASTOPEN:
-		if (icsk->icsk_accept_queue.fastopenq)
-			val = icsk->icsk_accept_queue.fastopenq->max_qlen;
-		else
-			val = 0;
+		val = icsk->icsk_accept_queue.fastopenq.max_qlen;
 		break;
 
 	case TCP_TIMESTAMP:
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index db43c62..f69f436 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -142,9 +142,9 @@ static struct sock *tcp_fastopen_create_child(struct sock *sk,
 	if (!child)
 		return NULL;
 
-	spin_lock(&queue->fastopenq->lock);
-	queue->fastopenq->qlen++;
-	spin_unlock(&queue->fastopenq->lock);
+	spin_lock(&queue->fastopenq.lock);
+	queue->fastopenq.qlen++;
+	spin_unlock(&queue->fastopenq.lock);
 
 	/* Initialize the child socket. Have to fix some values to take
 	 * into account the child is a Fast Open socket and is created
@@ -237,8 +237,8 @@ static bool tcp_fastopen_queue_check(struct sock *sk)
 	 * between qlen overflow causing Fast Open to be disabled
 	 * temporarily vs a server not supporting Fast Open at all.
 	 */
-	fastopenq = inet_csk(sk)->icsk_accept_queue.fastopenq;
-	if (!fastopenq || fastopenq->max_qlen == 0)
+	fastopenq = &inet_csk(sk)->icsk_accept_queue.fastopenq;
+	if (fastopenq->max_qlen == 0)
 		return false;
 
 	if (fastopenq->qlen >= fastopenq->max_qlen) {
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index f551e9e..64ece71 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -2186,7 +2186,7 @@ static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
 	const struct tcp_sock *tp = tcp_sk(sk);
 	const struct inet_connection_sock *icsk = inet_csk(sk);
 	const struct inet_sock *inet = inet_sk(sk);
-	struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq;
+	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
 	__be32 dest = inet->inet_daddr;
 	__be32 src = inet->inet_rcv_saddr;
 	__u16 destp = ntohs(inet->inet_dport);