From 9f512f6540c53251fa65a7c4be128f1b25f7dfbf Mon Sep 17 00:00:00 2001
From: Rusty Russell <rusty@rustcorp.com.au>
Date: Thu, 18 Aug 2016 14:25:14 +0930
Subject: [PATCH] peer: restart from the other side.

Testing this revealed that we can't just reconnect when we have something to
send, as we might be NATed; we should try to reconnect anyway.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
---
 daemon/peer.c       | 14 +++++---------
 daemon/test/test.sh | 31 +++++++++++++++++++++++++++++--
 2 files changed, 34 insertions(+), 11 deletions(-)

diff --git a/daemon/peer.c b/daemon/peer.c
index 143cb6b0f..fa37911b8 100644
--- a/daemon/peer.c
+++ b/daemon/peer.c
@@ -2321,7 +2321,7 @@ static struct io_plan *reconnect_pkt_in(struct io_conn *conn, struct peer *peer)
 		return pkt_out(conn, peer);
 	}
 
-	/* Queue prompted us to reconnect, but we need to eliminate it now. */
+	/* We need to eliminate queue now. */
 	clear_output_queue(peer);
 
 	/* They might have missed the error, tell them before hanging up */
@@ -3832,7 +3832,8 @@ static struct io_plan *peer_reconnect(struct io_conn *conn, struct peer *peer)
 				 peer->id, crypto_on_reconnect_out, peer);
 }
 
-/* Only retry when we want to send something. */ 
+/* We can't only retry when we want to send: they may want to send us
+ * something but not be able to connect (NAT).  So keep retrying.. */ 
 static void reconnect_failed(struct io_conn *conn, struct peer *peer)
 {
 	/* Already otherwise connected (ie. they connected in)? */
@@ -3841,13 +3842,8 @@ static void reconnect_failed(struct io_conn *conn, struct peer *peer)
 		return;
 	}
 
-	if (!tal_count(peer->outpkt)) {
-		log_debug(peer->log, "reconnect_failed: nothing to send");
-		return;
-	}
-
-	log_debug(peer->log, "Have packets to send, setting timer");
-	new_reltimer(peer->dstate, peer, time_from_sec(60), try_reconnect, peer);
+	log_debug(peer->log, "Setting timer to re-connect");
+	new_reltimer(peer->dstate, peer, time_from_sec(15), try_reconnect, peer);
 }
 
 static struct io_plan *init_conn(struct io_conn *conn, struct peer *peer)
diff --git a/daemon/test/test.sh b/daemon/test/test.sh
index 7279d81d1..2c96407c0 100755
--- a/daemon/test/test.sh
+++ b/daemon/test/test.sh
@@ -354,6 +354,14 @@ EOF
 
 cp $DIR2/config $DIR3/config
 
+if [ x"$RECONNECT" = xrestart ]; then
+    # Make sure node2 restarts on same port, by setting in config.
+    # Find a free TCP port.
+    port=4000
+    while netstat -ntl | grep -q ":$port "; do port=$(($port + 1)); done
+    echo port=$port >> $DIR2/config
+fi
+
 if [ -n "$DIFFERENT_FEES" ]; then
     # Simply override default fee (estimatefee fails on regtest anyway)
     CLOSE_FEE_RATE2=50000
@@ -373,11 +381,13 @@ else
     $LIGHTNINGD1 > $REDIR1 2> $REDIRERR1 &
 fi
 
+LIGHTNINGD2="$(readlink -f `pwd`/../lightningd) --lightning-dir=$DIR2"
 if [ -n "$GDB2" ]; then
-    echo Press return once you run: gdb --args daemon/lightningd --lightning-dir=$DIR2 >&2
+    echo Press return once you run: gdb --args $LIGHTNINGD2 >&2
     read REPLY
 else
-    $PREFIX ../lightningd --lightning-dir=$DIR2 > $REDIR2 2> $REDIRERR2 &
+    LIGHTNINGD2="$PREFIX $LIGHTNINGD2"
+    $LIGHTNINGD2 > $REDIR2 2> $REDIRERR2 &
 fi
 $PREFIX ../lightningd --lightning-dir=$DIR3 > $DIR3/output 2> $DIR3/errors &
 
@@ -848,6 +858,23 @@ lcli1 getpeers | tr -s '\012\011" ' ' ' | $FGREP "our_htlcs : [ { msatoshis : $H
 
 lcli2 getpeers | tr -s '\012\011" ' ' ' | $FGREP "our_htlcs : [ ], their_htlcs : [ { msatoshis : $HTLC_AMOUNT, expiry : { block : $EXPIRY }, rhash : $RHASH , state : RCVD_ADD_ACK_REVOCATION }, { msatoshis : $HTLC_AMOUNT, expiry : { block : $EXPIRY }, rhash : $RHASH2 , state : RCVD_ADD_ACK_REVOCATION } ]" || lcli2 getpeers | tr -s '\012\011" ' ' ' | $FGREP "our_htlcs : [ ], their_htlcs : [ { msatoshis : $HTLC_AMOUNT, expiry : { block : $EXPIRY }, rhash : $RHASH2 , state : RCVD_ADD_ACK_REVOCATION }, { msatoshis : $HTLC_AMOUNT, expiry : { block : $EXPIRY }, rhash : $RHASH , state : RCVD_ADD_ACK_REVOCATION } ]"
 
+# Just for once, reconnect/restart node 2.
+case "$RECONNECT" in
+    reconnect)
+	echo RECONNECTING NODE2
+	$LCLI2 dev-reconnect $ID1 >/dev/null
+	sleep 1
+	;;
+    restart)
+	echo RESTARTING NODE2
+	$LCLI2 -- dev-restart $LIGHTNINGD2 >/dev/null 2>&1 || true
+	if ! check "$LCLI2 getlog 2>/dev/null | fgrep -q Hello"; then
+	    echo "Node2 dev-restart failed!">&2
+	    exit 1
+	fi
+	;;
+esac
+
 # Node2 collects the HTLCs.
 lcli2 fulfillhtlc $ID1 $HTLCID $SECRET
 lcli2 fulfillhtlc $ID1 $HTLCID2 $SECRET2