currently, inbound TCP connections increment MPTcpExtMPCapableFallbackACK
when the server accepts them. As reported by Cristoph, this is inaccurate
because the counter becomes greater than MPTcpExtMPCapableSYNRX when many
non-MPC TCP connections are accepted.
Change the semantic of MPTcpExtMPCapableFallbackACK to increment when the
subflow context of a inbound MPC connection attempt is dropped.

Reported-by: Cristoph Paasch <cpaasch@apple.com>
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/449
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
---
 net/mptcp/protocol.c | 2 --
 net/mptcp/subflow.c  | 2 ++
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index cdf9ec67795e..556b3b95c537 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -3937,8 +3937,6 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
 				mptcp_set_state(newsk, TCP_CLOSE);
 		}
 	} else {
-		MPTCP_INC_STATS(sock_net(ssk),
-				MPTCP_MIB_MPCAPABLEPASSIVEFALLBACK);
 tcpfallback:
 		newsk->sk_kern_sock = kern;
 		lock_sock(newsk);
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 1626dd20c68f..6e3fe38f057d 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -905,6 +905,8 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 	return child;
 
 fallback:
+	if (child)
+		SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MPCAPABLEPASSIVEFALLBACK);
 	mptcp_subflow_drop_ctx(child);
 	return child;
 }
-- 
2.43.0