[MPTCP] [PATCH RFC mptcp 3/4] mptcp: add and use mptcp_subflow_hold

[MPTCP] [PATCH RFC mptcp 3/4] mptcp: add and use mptcp_subflow_hold

[Florian Westphal]

[-- Attachment #1: Type: text/plain, Size: 4185 bytes --]

subflow sockets already have lifetime managed by RCU, so we can
switch to atimic_inc_not_zero and skip/pretend we did not find
such socket in the mptcp subflow list.

This is required to get rid of synchronize_rcu() from mptcp_close().

Signed-off-by: Florian Westphal <fw(a)strlen.de>
---
 net/mptcp/protocol.c | 63 ++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 26 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 12857eeded8b..91db8a0c222d 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -31,14 +31,35 @@ static inline bool before64(__u64 seq1, __u64 seq2)
 }
 #define after64(seq2, seq1)	before64(seq1, seq2)
 
+static bool mptcp_subflow_hold(struct subflow_sock *sk)
+{
+	struct sock *_sk = (struct sock *)sk;
+
+	return refcount_inc_not_zero(&_sk->sk_refcnt);
+}
+
+static struct sock *mptcp_subflow_get_ref(const struct mptcp_sock *msk)
+{
+	struct subflow_sock *subflow;
+
+	rcu_read_lock();
+	mptcp_for_each_subflow(msk, subflow) {
+		if (mptcp_subflow_hold(subflow)) {
+			rcu_read_unlock();
+			return (struct sock *)subflow;
+		}
+	}
+
+	rcu_read_unlock();
+	return NULL;
+}
+
 static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
 	int mss_now, size_goal, poffset, ret;
 	struct mptcp_skb_cb *mcb = NULL;
-	struct subflow_sock *subflow;
 	struct page *page = NULL;
-	struct hlist_node *node;
 	struct sk_buff *skb;
 	struct sock *ssk;
 	size_t psize;
@@ -49,20 +70,17 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 		return sock_sendmsg(msk->subflow, msg);
 	}
 
-	rcu_read_lock();
-	node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
-	subflow = hlist_entry(node, struct subflow_sock, node);
-	ssk = sock_sk(subflow);
-	sock_hold(ssk);
-	rcu_read_unlock();
+	ssk = mptcp_subflow_get_ref(msk);
+	if (!ssk)
+		return -ENOTCONN;
 
 	if (!msg_data_left(msg)) {
 		pr_debug("empty send");
-		ret = sock_sendmsg(sock_sk(subflow)->sk_socket, msg);
+		ret = sock_sendmsg(ssk->sk_socket, msg);
 		goto put_out;
 	}
 
-	pr_debug("conn_list->subflow=%p", subflow);
+	pr_debug("conn_list->subflow=%p", ssk);
 
 	if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL)) {
 		ret = -ENOTSUPP;
@@ -279,7 +297,6 @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 	struct mptcp_sock *msk = mptcp_sk(sk);
 	struct subflow_sock *subflow;
 	struct mptcp_read_arg arg;
-	struct hlist_node *node;
 	read_descriptor_t desc;
 	struct tcp_sock *tp;
 	struct sock *ssk;
@@ -291,13 +308,11 @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		return sock_recvmsg(msk->subflow, msg, flags);
 	}
 
-	rcu_read_lock();
-	node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
-	subflow = hlist_entry(node, struct subflow_sock, node);
-	ssk = sock_sk(subflow);
-	sock_hold(ssk);
-	rcu_read_unlock();
+	ssk = mptcp_subflow_get_ref(msk);
+	if (!ssk)
+		return -ENOTCONN;
 
+	subflow = subflow_sk(ssk);
 	tp = tcp_sk(ssk);
 
 	desc.arg.data = &arg;
@@ -715,8 +730,6 @@ int mptcp_stream_connect(struct socket *sock, struct sockaddr *uaddr,
 int mptcp_stream_getname(struct socket *sock, struct sockaddr *uaddr, int peer)
 {
 	struct mptcp_sock *msk = mptcp_sk(sock->sk);
-	struct subflow_sock *subflow;
-	struct hlist_node *node;
 	struct sock *ssk;
 	int ret;
 
@@ -731,14 +744,12 @@ int mptcp_stream_getname(struct socket *sock, struct sockaddr *uaddr, int peer)
 	 * is connected and there are multiple subflows is not defined.
 	 * For now just use the first subflow on the list.
 	 */
-	rcu_read_lock();
-	node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
-	subflow = hlist_entry(node, struct subflow_sock, node);
-	ssk = sock_sk(subflow);
-	sock_hold(ssk);
-	rcu_read_unlock();
 
-	ret = inet_getname(sock_sk(subflow)->sk_socket, uaddr, peer);
+	ssk = mptcp_subflow_get_ref(msk);
+	if (!ssk)
+		return -ENOTCONN;
+
+	ret = inet_getname(ssk->sk_socket, uaddr, peer);
 	sock_put(ssk);
 	return ret;
 }
-- 
2.19.2