From: Marc Dionne <marc.c.dionne@gmail.com>

Peers may only send immediate acks for every 2 UDP packets received.
When sending a jumbogram, it is important to check that there is
sufficient window space to send another same sized jumbogram following
the current one, and request an ack if there isn't.  Failure to do so may
cause the call to stall waiting for an ack until the resend timer fires.

Where jumbograms are in use this causes a very significant drop in
performance.

Fixes: fe24a5494390 ("rxrpc: Send jumbo DATA packets")
Signed-off-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeffrey Altman <jaltman@auristor.com>
cc: Eric Dumazet <edumazet@google.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Simon Horman <horms@kernel.org>
cc: linux-afs@lists.infradead.org
cc: netdev@vger.kernel.org
cc: stable@kernel.org
---
 include/trace/events/rxrpc.h | 1 +
 net/rxrpc/ar-internal.h      | 2 +-
 net/rxrpc/output.c           | 2 ++
 net/rxrpc/proc.c             | 5 +++--
 4 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/include/trace/events/rxrpc.h b/include/trace/events/rxrpc.h
index f7f559204b87..578b8038b211 100644
--- a/include/trace/events/rxrpc.h
+++ b/include/trace/events/rxrpc.h
@@ -521,6 +521,7 @@
 #define rxrpc_req_ack_traces \
 	EM(rxrpc_reqack_ack_lost,		"ACK-LOST  ")	\
 	EM(rxrpc_reqack_app_stall,		"APP-STALL ")	\
+	EM(rxrpc_reqack_jumbo_win,		"JUMBO-WIN ")	\
 	EM(rxrpc_reqack_more_rtt,		"MORE-RTT  ")	\
 	EM(rxrpc_reqack_no_srv_last,		"NO-SRVLAST")	\
 	EM(rxrpc_reqack_old_rtt,		"OLD-RTT   ")	\
diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index 36d6ca0d1089..96ecb83c9071 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -117,7 +117,7 @@ struct rxrpc_net {
 	atomic_t		stat_tx_jumbo[10];
 	atomic_t		stat_rx_jumbo[10];
 
-	atomic_t		stat_why_req_ack[8];
+	atomic_t		stat_why_req_ack[9];
 
 	atomic_t		stat_io_loop;
 };
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index d70db367e358..870e59bf06af 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -479,6 +479,8 @@ static size_t rxrpc_prepare_data_subpacket(struct rxrpc_call *call,
 		why = rxrpc_reqack_old_rtt;
 	else if (!last && !after(READ_ONCE(call->send_top), txb->seq))
 		why = rxrpc_reqack_app_stall;
+	else if (call->tx_winsize <= (2 * req->n) || call->cong_cwnd <= (2 * req->n))
+		why = rxrpc_reqack_jumbo_win;
 	else
 		goto dont_set_request_ack;
 
diff --git a/net/rxrpc/proc.c b/net/rxrpc/proc.c
index 59292f7f9205..7755fca5beb8 100644
--- a/net/rxrpc/proc.c
+++ b/net/rxrpc/proc.c
@@ -518,11 +518,12 @@ int rxrpc_stats_show(struct seq_file *seq, void *v)
 		   atomic_read(&rxnet->stat_rx_acks[RXRPC_ACK_IDLE]),
 		   atomic_read(&rxnet->stat_rx_acks[0]));
 	seq_printf(seq,
-		   "Why-Req-A: acklost=%u mrtt=%u ortt=%u stall=%u\n",
+		   "Why-Req-A: acklost=%u mrtt=%u ortt=%u stall=%u jwin=%u\n",
 		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_ack_lost]),
 		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_more_rtt]),
 		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_old_rtt]),
-		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_app_stall]));
+		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_app_stall]),
+		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_jumbo_win]));
 	seq_printf(seq,
 		   "Why-Req-A: nolast=%u retx=%u slows=%u smtxw=%u\n",
 		   atomic_read(&rxnet->stat_why_req_ack[rxrpc_reqack_no_srv_last]),