14 files changed, 125 insertions, 46 deletions

diff --git a/net/rds/ib_cm.c b/net/rds/ib_cm.c
index 310cabc..7c2a65a 100644
--- a/net/rds/ib_cm.c
+++ b/net/rds/ib_cm.c
@@ -111,7 +111,7 @@ void rds_ib_cm_connect_complete(struct rds_connection *conn, struct rdma_cm_even
 		}
 	}
 
-	if (conn->c_version < RDS_PROTOCOL(3,1)) {
+	if (conn->c_version < RDS_PROTOCOL(3, 1)) {
 		printk(KERN_NOTICE "RDS/IB: Connection to %pI4 version %u.%u failed,"
 		       " no longer supported\n",
 		       &conn->c_faddr,
diff --git a/net/rds/loop.c b/net/rds/loop.c
index 6b12b68..814173b 100644
--- a/net/rds/loop.c
+++ b/net/rds/loop.c
@@ -95,8 +95,9 @@ out:
  */
 static void rds_loop_inc_free(struct rds_incoming *inc)
 {
-        struct rds_message *rm = container_of(inc, struct rds_message, m_inc);
-        rds_message_put(rm);
+	struct rds_message *rm = container_of(inc, struct rds_message, m_inc);
+
+	rds_message_put(rm);
 }
 
 /* we need to at least give the thread something to succeed */
diff --git a/net/rds/rds.h b/net/rds/rds.h
index 80256b0..387df5f 100644
--- a/net/rds/rds.h
+++ b/net/rds/rds.h
@@ -74,6 +74,7 @@ enum {
 	RDS_CONN_CONNECTING,
 	RDS_CONN_DISCONNECTING,
 	RDS_CONN_UP,
+	RDS_CONN_RESETTING,
 	RDS_CONN_ERROR,
 };
 
@@ -813,6 +814,7 @@ void rds_connect_worker(struct work_struct *);
 void rds_shutdown_worker(struct work_struct *);
 void rds_send_worker(struct work_struct *);
 void rds_recv_worker(struct work_struct *);
+void rds_connect_path_complete(struct rds_connection *conn, int curr);
 void rds_connect_complete(struct rds_connection *conn);
 
 /* transport.c */
diff --git a/net/rds/recv.c b/net/rds/recv.c
index c0be1ec..8413f6c 100644
--- a/net/rds/recv.c
+++ b/net/rds/recv.c
@@ -561,5 +561,7 @@ void rds_inc_info_copy(struct rds_incoming *inc,
 		minfo.fport = inc->i_hdr.h_dport;
 	}
 
+	minfo.flags = 0;
+
 	rds_info_copy(iter, &minfo, sizeof(minfo));
 }
diff --git a/net/rds/send.c b/net/rds/send.c
index c9cdb35..b1962f8 100644
--- a/net/rds/send.c
+++ b/net/rds/send.c
@@ -99,6 +99,7 @@ void rds_send_reset(struct rds_connection *conn)
 	list_splice_init(&conn->c_retrans, &conn->c_send_queue);
 	spin_unlock_irqrestore(&conn->c_lock, flags);
 }
+EXPORT_SYMBOL_GPL(rds_send_reset);
 
 static int acquire_in_xmit(struct rds_connection *conn)
 {
diff --git a/net/rds/sysctl.c b/net/rds/sysctl.c
index c173f69..e381bbc 100644
--- a/net/rds/sysctl.c
+++ b/net/rds/sysctl.c
@@ -102,7 +102,8 @@ int rds_sysctl_init(void)
 	rds_sysctl_reconnect_min = msecs_to_jiffies(1);
 	rds_sysctl_reconnect_min_jiffies = rds_sysctl_reconnect_min;
 
-	rds_sysctl_reg_table = register_net_sysctl(&init_net,"net/rds", rds_sysctl_rds_table);
+	rds_sysctl_reg_table =
+		register_net_sysctl(&init_net, "net/rds", rds_sysctl_rds_table);
 	if (!rds_sysctl_reg_table)
 		return -ENOMEM;
 	return 0;
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 86187da..74ee126 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -126,9 +126,81 @@ void rds_tcp_restore_callbacks(struct socket *sock,
 }
 
 /*
- * This is the only path that sets tc->t_sock.  Send and receive trust that
- * it is set.  The RDS_CONN_UP bit protects those paths from being
- * called while it isn't set.
+ * rds_tcp_reset_callbacks() switches the to the new sock and
+ * returns the existing tc->t_sock.
+ *
+ * The only functions that set tc->t_sock are rds_tcp_set_callbacks
+ * and rds_tcp_reset_callbacks.  Send and receive trust that
+ * it is set.  The absence of RDS_CONN_UP bit protects those paths
+ * from being called while it isn't set.
+ */
+void rds_tcp_reset_callbacks(struct socket *sock,
+			     struct rds_connection *conn)
+{
+	struct rds_tcp_connection *tc = conn->c_transport_data;
+	struct socket *osock = tc->t_sock;
+
+	if (!osock)
+		goto newsock;
+
+	/* Need to resolve a duelling SYN between peers.
+	 * We have an outstanding SYN to this peer, which may
+	 * potentially have transitioned to the RDS_CONN_UP state,
+	 * so we must quiesce any send threads before resetting
+	 * c_transport_data. We quiesce these threads by setting
+	 * c_state to something other than RDS_CONN_UP, and then
+	 * waiting for any existing threads in rds_send_xmit to
+	 * complete release_in_xmit(). (Subsequent threads entering
+	 * rds_send_xmit() will bail on !rds_conn_up().
+	 *
+	 * However an incoming syn-ack at this point would end up
+	 * marking the conn as RDS_CONN_UP, and would again permit
+	 * rds_send_xmi() threads through, so ideally we would
+	 * synchronize on RDS_CONN_UP after lock_sock(), but cannot
+	 * do that: waiting on !RDS_IN_XMIT after lock_sock() may
+	 * end up deadlocking with tcp_sendmsg(), and the RDS_IN_XMIT
+	 * would not get set. As a result, we set c_state to
+	 * RDS_CONN_RESETTTING, to ensure that rds_tcp_state_change
+	 * cannot mark rds_conn_path_up() in the window before lock_sock()
+	 */
+	atomic_set(&conn->c_state, RDS_CONN_RESETTING);
+	wait_event(conn->c_waitq, !test_bit(RDS_IN_XMIT, &conn->c_flags));
+	lock_sock(osock->sk);
+	/* reset receive side state for rds_tcp_data_recv() for osock  */
+	if (tc->t_tinc) {
+		rds_inc_put(&tc->t_tinc->ti_inc);
+		tc->t_tinc = NULL;
+	}
+	tc->t_tinc_hdr_rem = sizeof(struct rds_header);
+	tc->t_tinc_data_rem = 0;
+	tc->t_sock = NULL;
+
+	write_lock_bh(&osock->sk->sk_callback_lock);
+
+	osock->sk->sk_user_data = NULL;
+	osock->sk->sk_data_ready = tc->t_orig_data_ready;
+	osock->sk->sk_write_space = tc->t_orig_write_space;
+	osock->sk->sk_state_change = tc->t_orig_state_change;
+	write_unlock_bh(&osock->sk->sk_callback_lock);
+	release_sock(osock->sk);
+	sock_release(osock);
+newsock:
+	rds_send_reset(conn);
+	lock_sock(sock->sk);
+	write_lock_bh(&sock->sk->sk_callback_lock);
+	tc->t_sock = sock;
+	sock->sk->sk_user_data = conn;
+	sock->sk->sk_data_ready = rds_tcp_data_ready;
+	sock->sk->sk_write_space = rds_tcp_write_space;
+	sock->sk->sk_state_change = rds_tcp_state_change;
+
+	write_unlock_bh(&sock->sk->sk_callback_lock);
+	release_sock(sock->sk);
+}
+
+/* Add tc to rds_tcp_tc_list and set tc->t_sock. See comments
+ * above rds_tcp_reset_callbacks for notes about synchronization
+ * with data path
  */
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_connection *conn)
 {
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 41c2283..7940bab 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,6 +50,7 @@ struct rds_tcp_statistics {
 void rds_tcp_tune(struct socket *sock);
 void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_connection *conn);
+void rds_tcp_reset_callbacks(struct socket *sock, struct rds_connection *conn);
 void rds_tcp_restore_callbacks(struct socket *sock,
 			       struct rds_tcp_connection *tc);
 u32 rds_tcp_snd_nxt(struct rds_tcp_connection *tc);
@@ -82,7 +83,7 @@ int rds_tcp_inc_copy_to_user(struct rds_incoming *inc, struct iov_iter *to);
 void rds_tcp_xmit_prepare(struct rds_connection *conn);
 void rds_tcp_xmit_complete(struct rds_connection *conn);
 int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
-	         unsigned int hdr_off, unsigned int sg, unsigned int off);
+		 unsigned int hdr_off, unsigned int sg, unsigned int off);
 void rds_tcp_write_space(struct sock *sk);
 
 /* tcp_stats.c */
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index fb82e0a..f6e95d6 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -54,19 +54,19 @@ void rds_tcp_state_change(struct sock *sk)
 
 	rdsdebug("sock %p state_change to %d\n", tc->t_sock, sk->sk_state);
 
-	switch(sk->sk_state) {
-		/* ignore connecting sockets as they make progress */
-		case TCP_SYN_SENT:
-		case TCP_SYN_RECV:
-			break;
-		case TCP_ESTABLISHED:
-			rds_connect_complete(conn);
-			break;
-		case TCP_CLOSE_WAIT:
-		case TCP_CLOSE:
-			rds_conn_drop(conn);
-		default:
-			break;
+	switch (sk->sk_state) {
+	/* ignore connecting sockets as they make progress */
+	case TCP_SYN_SENT:
+	case TCP_SYN_RECV:
+		break;
+	case TCP_ESTABLISHED:
+		rds_connect_path_complete(conn, RDS_CONN_CONNECTING);
+		break;
+	case TCP_CLOSE_WAIT:
+	case TCP_CLOSE:
+		rds_conn_drop(conn);
+	default:
+		break;
 	}
 out:
 	read_unlock_bh(&sk->sk_callback_lock);
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 4bf4bef..245542c 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -78,7 +78,6 @@ int rds_tcp_accept_one(struct socket *sock)
 	struct inet_sock *inet;
 	struct rds_tcp_connection *rs_tcp = NULL;
 	int conn_state;
-	struct sock *nsk;
 
 	if (!sock) /* module unload or netns delete in progress */
 		return -ENETUNREACH;
@@ -136,26 +135,21 @@ int rds_tcp_accept_one(struct socket *sock)
 		    !conn->c_outgoing) {
 			goto rst_nsk;
 		} else {
-			atomic_set(&conn->c_state, RDS_CONN_CONNECTING);
-			wait_event(conn->c_waitq,
-				   !test_bit(RDS_IN_XMIT, &conn->c_flags));
-			rds_tcp_restore_callbacks(rs_tcp->t_sock, rs_tcp);
+			rds_tcp_reset_callbacks(new_sock, conn);
 			conn->c_outgoing = 0;
+			/* rds_connect_path_complete() marks RDS_CONN_UP */
+			rds_connect_path_complete(conn, RDS_CONN_RESETTING);
 		}
+	} else {
+		rds_tcp_set_callbacks(new_sock, conn);
+		rds_connect_path_complete(conn, RDS_CONN_CONNECTING);
 	}
-	rds_tcp_set_callbacks(new_sock, conn);
-	rds_connect_complete(conn); /* marks RDS_CONN_UP */
 	new_sock = NULL;
 	ret = 0;
 	goto out;
 rst_nsk:
 	/* reset the newly returned accept sock and bail */
-	nsk = new_sock->sk;
-	rds_tcp_stats_inc(s_tcp_listen_closed_stale);
-	nsk->sk_user_data = NULL;
-	nsk->sk_prot->disconnect(nsk, 0);
-	tcp_done(nsk);
-	new_sock = NULL;
+	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
 	if (rs_tcp)
diff --git a/net/rds/tcp_recv.c b/net/rds/tcp_recv.c
index c3196f9..6e6a711 100644
--- a/net/rds/tcp_recv.c
+++ b/net/rds/tcp_recv.c
@@ -171,7 +171,7 @@ static int rds_tcp_data_recv(read_descriptor_t *desc, struct sk_buff *skb,
 	while (left) {
 		if (!tinc) {
 			tinc = kmem_cache_alloc(rds_tcp_incoming_slab,
-					        arg->gfp);
+						arg->gfp);
 			if (!tinc) {
 				desc->error = -ENOMEM;
 				goto out;
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 22d0f20..618be69 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -66,19 +66,19 @@ void rds_tcp_xmit_complete(struct rds_connection *conn)
 static int rds_tcp_sendmsg(struct socket *sock, void *data, unsigned int len)
 {
 	struct kvec vec = {
-                .iov_base = data,
-                .iov_len = len,
+		.iov_base = data,
+		.iov_len = len,
+	};
+	struct msghdr msg = {
+		.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
 	};
-        struct msghdr msg = {
-                .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
-        };
 
 	return kernel_sendmsg(sock, &msg, &vec, 1, vec.iov_len);
 }
 
 /* the core send_sem serializes this with other xmit and shutdown */
 int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
-	         unsigned int hdr_off, unsigned int sg, unsigned int off)
+		 unsigned int hdr_off, unsigned int sg, unsigned int off)
 {
 	struct rds_tcp_connection *tc = conn->c_transport_data;
 	int done = 0;
@@ -196,7 +196,7 @@ void rds_tcp_write_space(struct sock *sk)
 	tc->t_last_seen_una = rds_tcp_snd_una(tc);
 	rds_send_drop_acked(conn, rds_tcp_snd_una(tc), rds_tcp_is_acked);
 
-        if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf)
+	if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf)
 		queue_delayed_work(rds_wq, &conn->c_send_w, 0);
 
 out:
diff --git a/net/rds/threads.c b/net/rds/threads.c
index 454aa6d..4a32304 100644
--- a/net/rds/threads.c
+++ b/net/rds/threads.c
@@ -71,9 +71,9 @@
 struct workqueue_struct *rds_wq;
 EXPORT_SYMBOL_GPL(rds_wq);
 
-void rds_connect_complete(struct rds_connection *conn)
+void rds_connect_path_complete(struct rds_connection *conn, int curr)
 {
-	if (!rds_conn_transition(conn, RDS_CONN_CONNECTING, RDS_CONN_UP)) {
+	if (!rds_conn_transition(conn, curr, RDS_CONN_UP)) {
 		printk(KERN_WARNING "%s: Cannot transition to state UP, "
 				"current state is %d\n",
 				__func__,
@@ -90,6 +90,12 @@ void rds_connect_complete(struct rds_connection *conn)
 	queue_delayed_work(rds_wq, &conn->c_send_w, 0);
 	queue_delayed_work(rds_wq, &conn->c_recv_w, 0);
 }
+EXPORT_SYMBOL_GPL(rds_connect_path_complete);
+
+void rds_connect_complete(struct rds_connection *conn)
+{
+	rds_connect_path_complete(conn, RDS_CONN_CONNECTING);
+}
 EXPORT_SYMBOL_GPL(rds_connect_complete);
 
 /*
diff --git a/net/rds/transport.c b/net/rds/transport.c
index f3afd1d..2ffd3e30 100644
--- a/net/rds/transport.c
+++ b/net/rds/transport.c
@@ -140,8 +140,7 @@ unsigned int rds_trans_stats_info_copy(struct rds_info_iterator *iter,
 	rds_info_iter_unmap(iter);
 	down_read(&rds_trans_sem);
 
-	for (i = 0; i < RDS_TRANS_COUNT; i++)
-	{
+	for (i = 0; i < RDS_TRANS_COUNT; i++) {
 		trans = transports[i];
 		if (!trans || !trans->stats_info_copy)
 			continue;