Merge branch 'mptcp-fix-MP_JOIN-failure-handling'

Paolo Abeni says:

====================
mptcp: fix MP_JOIN failure handling

Currently if we hit an MP_JOIN failure on the third ack, the child socket is
closed with reset, but the request socket is not deleted, causing weird
behaviors.

The main problem is that MPTCP's MP_JOIN code needs to plug it's own
'valid 3rd ack' checks and the current TCP callbacks do not allow that.

This series tries to address the above shortcoming introducing a new MPTCP
specific bit in a 'struct tcp_request_sock' hole, and leveraging that to allow
tcp_check_req releasing the request socket when needed.

The above allows cleaning-up a bit current MPTCP hooking in tcp_check_req().

An alternative solution, possibly cleaner but more invasive, would be
changing the 'bool *own_req' syn_recv_sock() argument into 'int *req_status'
and let MPTCP set it to 'REQ_DROP'.

v1 -> v2:
- be more conservative about drop_req initialization

RFC -> v1:
- move the drop_req bit inside tcp_request_sock (Eric)
====================

Signed-off-by: David S. Miller <[e-mail]>

include/linux/tcp.h

···
 	u64				snt_synack; /* first SYNACK sent time */
 	bool				tfo_listener;
 	bool				is_mptcp;
+#if IS_ENABLED(CONFIG_MPTCP)
+	bool				drop_req;
+#endif
 	u32				txhash;
 	u32				rcv_isn;
 	u32				snt_isn;

include/net/inet_connection_sock.h

···
 void inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req);
 void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req);
 
+static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk)
+{
+	/* The below has to be done to allow calling inet_csk_destroy_sock */
+	sock_set_flag(sk, SOCK_DEAD);
+	percpu_counter_inc(sk->sk_prot->orphan_count);
+	inet_sk(sk)->inet_num = 0;
+}
+
 void inet_csk_destroy_sock(struct sock *sk);
 void inet_csk_prepare_forced_close(struct sock *sk);
 

include/net/mptcp.h

···
 	return tcp_rsk(req)->is_mptcp;
 }
 
+static inline bool rsk_drop_req(const struct request_sock *req)
+{
+	return tcp_rsk(req)->is_mptcp && tcp_rsk(req)->drop_req;
+}
+
 void mptcp_space(const struct sock *ssk, int *space, int *full_space);
 bool mptcp_syn_options(struct sock *sk, const struct sk_buff *skb,
 		       unsigned int *size, struct mptcp_out_options *opts);
···
 				 skb_ext_find(from, SKB_EXT_MPTCP));
 }
 
-bool mptcp_sk_is_subflow(const struct sock *sk);
-
 void mptcp_seq_show(struct seq_file *seq);
 #else
 
···
 }
 
 static inline bool rsk_is_mptcp(const struct request_sock *req)
+{
+	return false;
+}
+
+static inline bool rsk_drop_req(const struct request_sock *req)
 {
 	return false;
 }
···
 					  const struct sk_buff *from)
 {
 	return true;
-}
-
-static inline bool mptcp_sk_is_subflow(const struct sock *sk)
-{
-	return false;
 }
 
 static inline void mptcp_space(const struct sock *ssk, int *s, int *fs) { }

net/ipv4/inet_connection_sock.c

···
 	/* sk_clone_lock locked the socket and set refcnt to 2 */
 	bh_unlock_sock(sk);
 	sock_put(sk);
-
-	/* The below has to be done to allow calling inet_csk_destroy_sock */
-	sock_set_flag(sk, SOCK_DEAD);
-	percpu_counter_inc(sk->sk_prot->orphan_count);
-	inet_sk(sk)->inet_num = 0;
+	inet_csk_prepare_for_destroy_sock(sk);
 }
 EXPORT_SYMBOL(inet_csk_prepare_forced_close);
 

net/ipv4/tcp_minisocks.c

···
 	if (!child)
 		goto listen_overflow;
 
-	if (own_req && sk_is_mptcp(child) && mptcp_sk_is_subflow(child)) {
+	if (own_req && rsk_drop_req(req)) {
 		reqsk_queue_removed(&inet_csk(sk)->icsk_accept_queue, req);
 		inet_csk_reqsk_queue_drop_and_put(sk, req);
 		return child;

net/mptcp/protocol.c

···
 	return ret;
 }
 
-bool mptcp_sk_is_subflow(const struct sock *sk)
-{
-	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
-
-	return subflow->mp_join == 1;
-}
-
 static bool mptcp_memory_free(const struct sock *sk, int wake)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);

net/mptcp/subflow.c

···
 	if (child && *own_req) {
 		struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(child);
 
+		tcp_rsk(req)->drop_req = false;
+
 		/* we need to fallback on ctx allocation failure and on pre-reqs
 		 * checking above. In the latter scenario we additionally need
 		 * to reset the context to non MPTCP status.
 		 */
 		if (!ctx || fallback) {
 			if (fallback_is_fatal)
-				goto close_child;
+				goto dispose_child;
 
 			if (ctx) {
 				subflow_ulp_fallback(child, ctx);
···
 
 			owner = mptcp_token_get_sock(ctx->token);
 			if (!owner)
-				goto close_child;
+				goto dispose_child;
 
 			ctx->conn = (struct sock *)owner;
 			if (!mptcp_finish_join(child))
-				goto close_child;
+				goto dispose_child;
 
 			SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINACKRX);
+			tcp_rsk(req)->drop_req = true;
 		}
 	}
 
···
 		      !mptcp_subflow_ctx(child)->conn));
 	return child;
 
-close_child:
+dispose_child:
+	tcp_rsk(req)->drop_req = true;
 	tcp_send_active_reset(child, GFP_ATOMIC);
-	inet_csk_prepare_forced_close(child);
+	inet_csk_prepare_for_destroy_sock(child);
 	tcp_done(child);
-	return NULL;
+
+	/* The last child reference will be released by the caller */
+	return child;
 }
 
 static struct inet_connection_sock_af_ops subflow_specific;