From be2622a4ff69b93d775d2dc0b129497c93265e34 Mon Sep 17 00:00:00 2001
From: Rusty Russell <rusty@rustcorp.com.au>
Date: Sat, 9 Oct 2021 16:23:27 +1030
Subject: [PATCH] channeld: perform regular keepalive pings.

Send a ping every 15-45 seconds.  If we try to send another one and we
haven't got a reply, hang up.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Changed: Protocol: Send regular pings to detect dead connections (particularly for Tor).
---
 channeld/channeld.c      | 41 +++++++++++++++++++++++++++++++++-------
 tests/test_connection.py | 14 ++++++++++++++
 2 files changed, 48 insertions(+), 7 deletions(-)

diff --git a/channeld/channeld.c b/channeld/channeld.c
index 368c84776..93c0c775b 100644
--- a/channeld/channeld.c
+++ b/channeld/channeld.c
@@ -100,6 +100,9 @@ struct peer {
 	u64 commit_timer_attempts;
 	u32 commit_msec;
 
+	/* Random ping timer, to detect dead connections. */
+	struct oneshot *ping_timer;
+
 	/* Are we expecting a pong? */
 	bool expecting_pong;
 
@@ -1084,6 +1087,29 @@ static struct bitcoin_signature *calc_commitsigs(const tal_t *ctx,
 	return htlc_sigs;
 }
 
+/* Mutual recursion */
+static void send_ping(struct peer *peer);
+
+static void set_ping_timer(struct peer *peer)
+{
+	peer->ping_timer = new_reltimer(&peer->timers, peer,
+					time_from_sec(15 + pseudorand(30)),
+					send_ping, peer);
+}
+
+static void send_ping(struct peer *peer)
+{
+	/* Already have a ping in flight? */
+	if (peer->expecting_pong) {
+		status_debug("Last ping unreturned: hanging up");
+		exit(0);
+	}
+
+	sync_crypto_write_no_delay(peer->pps, take(make_ping(NULL, 1, 0)));
+	peer->expecting_pong = true;
+	set_ping_timer(peer);
+}
+
 /* Peer protocol doesn't want sighash flags. */
 static secp256k1_ecdsa_signature *raw_sigs(const tal_t *ctx,
 					   const struct bitcoin_signature *sigs)
@@ -2121,12 +2147,6 @@ static void peer_in(struct peer *peer, const u8 *msg)
 	 */
 	bool soft_error = peer->funding_locked[REMOTE] || peer->funding_locked[LOCAL];
 
-	/* Catch our own ping replies. */
-	if (type == WIRE_PONG && peer->expecting_pong) {
-		peer->expecting_pong = false;
-		return;
-	}
-
 	if (channeld_handle_custommsg(msg))
 		return;
 
@@ -2210,6 +2230,13 @@ static void peer_in(struct peer *peer, const u8 *msg)
 	case WIRE_INIT_RBF:
 	case WIRE_ACK_RBF:
 		break;
+	case WIRE_PONG:
+		if (peer->expecting_pong) {
+			peer->expecting_pong = false;
+			return;
+		}
+		status_debug("Unexpected pong?");
+		return;
 
 	case WIRE_CHANNEL_REESTABLISH:
 		handle_unexpected_reestablish(peer, msg);
@@ -2225,7 +2252,6 @@ static void peer_in(struct peer *peer, const u8 *msg)
 	case WIRE_GOSSIP_TIMESTAMP_FILTER:
 	case WIRE_REPLY_SHORT_CHANNEL_IDS_END:
 	case WIRE_PING:
-	case WIRE_PONG:
 	case WIRE_WARNING:
 	case WIRE_ERROR:
 	case WIRE_ONION_MESSAGE:
@@ -3856,6 +3882,7 @@ int main(int argc, char *argv[])
 	peer->expecting_pong = false;
 	timers_init(&peer->timers, time_mono());
 	peer->commit_timer = NULL;
+	set_ping_timer(peer);
 	peer->have_sigs[LOCAL] = peer->have_sigs[REMOTE] = false;
 	peer->announce_depth_reached = false;
 	peer->channel_local_active = false;
diff --git a/tests/test_connection.py b/tests/test_connection.py
index 0bc0f95c2..891a66123 100644
--- a/tests/test_connection.py
+++ b/tests/test_connection.py
@@ -3734,3 +3734,17 @@ def test_old_feerate(node_factory):
 
     # This will timeout if l2 didn't accept fee.
     l1.pay(l2, 1000)
+
+
+@pytest.mark.developer("dev-disconnect required")
+def test_ping_timeout(node_factory):
+    # Disconnects after this, but doesn't know it.
+    l1_disconnects = ['xWIRE_PING']
+
+    l1, l2 = node_factory.line_graph(2, opts=[{'dev-no-reconnect': None,
+                                               'disconnect': l1_disconnects},
+                                              {}])
+    # Takes 15-45 seconds, then another to try second ping
+    l1.daemon.wait_for_log('Last ping unreturned: hanging up',
+                           timeout=45 + 45 + 5)
+    wait_for(lambda: l1.rpc.getpeer(l2.info['id'])['connected'] is False)