mptcp: deliver ssk errors to msk

Currently all errors received on msk subflows are ignored. We need to catch at least the errors on connect() and on fallback sockets. Use a custom sk_error_report callback at subflow level, and do the real action under the msk socket lock - via the usual sock_owned_by_user()/release_callback() schema. Reviewed-by: Mat Martineau <[email protected]> Signed-off-by: Paolo Abeni <[email protected]>
multipath-tcp · Feb 12, 2021 · 68b3a93 · 68b3a93
1 parent 359a2ca
commit 68b3a93
Show file tree

Hide file tree

Showing 3 changed files with 54 additions and 0 deletions.
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
@@ -2954,6 +2954,8 @@ static void mptcp_release_cb(struct sock *sk)
 		mptcp_push_pending(sk, 0);
 		spin_lock_bh(&sk->sk_lock.slock);
 	}
+	if (test_and_clear_bit(MPTCP_ERROR_REPORT, &mptcp_sk(sk)->flags))
+		__mptcp_error_report(sk);
 
 	/* clear any wmem reservation and errors */
 	__mptcp_update_wmem(sk);
@@ -3366,6 +3368,11 @@ static __poll_t mptcp_poll(struct file *file, struct socket *sock,
 	if (sk->sk_shutdown & RCV_SHUTDOWN)
 		mask |= EPOLLIN | EPOLLRDNORM | EPOLLRDHUP;
 
+	/* This barrier is coupled with smp_wmb() in tcp_reset() */
+	smp_rmb();
+	if (sk->sk_err)
+		mask |= EPOLLERR;
+
 	return mask;
 }
 

diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
@@ -102,6 +102,7 @@
 #define MPTCP_WORK_CLOSE_SUBFLOW 5
 #define MPTCP_PUSH_PENDING	6
 #define MPTCP_CLEAN_UNA		7
+#define MPTCP_ERROR_REPORT	8
 
 static inline bool before64(__u64 seq1, __u64 seq2)
 {
@@ -436,6 +437,7 @@ struct mptcp_subflow_context {
 	void	(*tcp_data_ready)(struct sock *sk);
 	void	(*tcp_state_change)(struct sock *sk);
 	void	(*tcp_write_space)(struct sock *sk);
+	void	(*tcp_error_report)(struct sock *sk);
 
 	struct	rcu_head rcu;
 };
@@ -560,6 +562,7 @@ static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
 	sk->sk_data_ready = ctx->tcp_data_ready;
 	sk->sk_state_change = ctx->tcp_state_change;
 	sk->sk_write_space = ctx->tcp_write_space;
+	sk->sk_error_report = ctx->tcp_error_report;
 
 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
 }
@@ -587,6 +590,7 @@ bool mptcp_finish_join(struct sock *sk);
 bool mptcp_schedule_work(struct sock *sk);
 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
 void __mptcp_data_acked(struct sock *sk);
+void __mptcp_error_report(struct sock *sk);
 void mptcp_subflow_eof(struct sock *sk);
 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
 void __mptcp_flush_join_list(struct mptcp_sock *msk);

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
@@ -1124,6 +1124,46 @@ static void subflow_write_space(struct sock *ssk)
 	mptcp_write_space(sk);
 }
 
+void __mptcp_error_report(struct sock *sk)
+{
+	struct mptcp_subflow_context *subflow;
+	struct mptcp_sock *msk = mptcp_sk(sk);
+
+	mptcp_for_each_subflow(msk, subflow) {
+		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
+		int err = sock_error(ssk);
+
+		if (!err)
+			continue;
+
+		/* only propagate errors on fallen-back sockets or
+		 * on MPC connect
+		 */
+		if (sk->sk_state != TCP_SYN_SENT && !__mptcp_check_fallback(msk))
+			continue;
+
+		inet_sk_state_store(sk, inet_sk_state_load(ssk));
+		sk->sk_err = -err;
+
+		/* This barrier is coupled with smp_rmb() in mptcp_poll() */
+		smp_wmb();
+		sk->sk_error_report(sk);
+		break;
+	}
+}
+
+static void subflow_error_report(struct sock *ssk)
+{
+	struct sock *sk = mptcp_subflow_ctx(ssk)->conn;
+
+	mptcp_data_lock(sk);
+	if (!sock_owned_by_user(sk))
+		__mptcp_error_report(sk);
+	else
+		set_bit(MPTCP_ERROR_REPORT,  &mptcp_sk(sk)->flags);
+	mptcp_data_unlock(sk);
+}
+
 static struct inet_connection_sock_af_ops *
 subflow_default_af_ops(struct sock *sk)
 {
@@ -1470,9 +1510,11 @@ static int subflow_ulp_init(struct sock *sk)
 	ctx->tcp_data_ready = sk->sk_data_ready;
 	ctx->tcp_state_change = sk->sk_state_change;
 	ctx->tcp_write_space = sk->sk_write_space;
+	ctx->tcp_error_report = sk->sk_error_report;
 	sk->sk_data_ready = subflow_data_ready;
 	sk->sk_write_space = subflow_write_space;
 	sk->sk_state_change = subflow_state_change;
+	sk->sk_error_report = subflow_error_report;
 out:
 	return err;
 }
@@ -1526,6 +1568,7 @@ static void subflow_ulp_clone(const struct request_sock *req,
 	new_ctx->tcp_data_ready = old_ctx->tcp_data_ready;
 	new_ctx->tcp_state_change = old_ctx->tcp_state_change;
 	new_ctx->tcp_write_space = old_ctx->tcp_write_space;
+	new_ctx->tcp_error_report = old_ctx->tcp_error_report;
 	new_ctx->rel_write_seq = 1;
 	new_ctx->tcp_sock = newsk;