Currently MPTCP drops partial packets for no good reason at all.
Instead we should just skip the already acked bytes.

Also add a missing check for zero window, which in turn requires
properly initializing the rcv_wnd_sent at connection creation time.

Fixes: ab174ad8ef76 ("mptcp: move ooo skbs into msk out of order queue.")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
Notes:
- We should also add a MIB for zerowin drop, but that should be a follow-up
patch, I think.
- based on top of 'mptcp: autotune related improvement', but targeting net,
  will have a conflict in __mptcp_move_skb().
---
 net/mptcp/protocol.c | 27 ++++++++++++++++++++-------
 net/mptcp/subflow.c  |  3 ++-
 2 files changed, 22 insertions(+), 8 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 7b966f105f89..e3330141a6da 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -378,6 +378,10 @@ static bool __mptcp_move_skb(struct sock *sk, struct sk_buff *skb)
 
 	mptcp_borrow_fwdmem(sk, skb);
 
+	/* Check for zero window.*/
+	if (atomic64_read(&msk->rcv_wnd_sent) == msk->ack_seq)
+		goto drop;
+
 	if (MPTCP_SKB_CB(skb)->map_seq == msk->ack_seq) {
 		/* in sequence */
 		msk->bytes_received += copy_len;
@@ -386,18 +390,27 @@ static bool __mptcp_move_skb(struct sock *sk, struct sk_buff *skb)
 		if (tail && mptcp_try_coalesce(sk, tail, skb))
 			return true;
 
-		skb_set_owner_r(skb, sk);
-		__skb_queue_tail(&sk->sk_receive_queue, skb);
-		return true;
+		goto enqueue;
 	} else if (after64(MPTCP_SKB_CB(skb)->map_seq, msk->ack_seq)) {
 		mptcp_data_queue_ofo(msk, skb);
 		return false;
 	}
 
-	/* old data, keep it simple and drop the whole pkt, sender
-	 * will retransmit as needed, if needed.
-	 */
-	MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_DUPDATA);
+	/* Check for old data. */
+	if (!after64(MPTCP_SKB_CB(skb)->end_seq, msk->ack_seq)) {
+		MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_DUPDATA);
+		goto drop;
+	}
+
+	/* Partial packet, seq < rcv_next < end_seq. */
+	MPTCP_SKB_CB(skb)->offset += msk->ack_seq - MPTCP_SKB_CB(skb)->map_seq;
+
+enqueue:
+	skb_set_owner_r(skb, sk);
+	__skb_queue_tail(&sk->sk_receive_queue, skb);
+	return true;
+
+drop:
 	mptcp_drop(sk, skb);
 	return false;
 }
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 39bf69e73975..0db832b5ddca 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -496,7 +496,8 @@ static void subflow_set_remote_key(struct mptcp_sock *msk,
 	WRITE_ONCE(msk->remote_key, subflow->remote_key);
 	WRITE_ONCE(msk->ack_seq, subflow->iasn);
 	WRITE_ONCE(msk->can_ack, true);
-	atomic64_set(&msk->rcv_wnd_sent, subflow->iasn);
+	atomic64_set(&msk->rcv_wnd_sent, subflow->iasn +
+		     tcp_receive_window(tcp_sk(subflow->tcp_sock)));
 }
 
 static void mptcp_propagate_state(struct sock *sk, struct sock *ssk,
-- 
2.51.1