lightningd: new state AWAITING_UNILATERAL.

When in this state, we send a canned error "Awaiting unilateral close".
We enter this both when we drop to chain, and when we're trying to get
them to drop to chain due to option_data_loss_protect.

As this state (unlike channel errors) is saved to the database, it means
we will *never* talk to a peer again in this state, so they can't
confuse us.

Since we set this state in channel_fail_permanent() (which is the only
place we call drop_to_chain for a unilateral close), we don't need to
save to the db: channel_set_state() does that for us.

This state change has a subtle effect: we return WIRE_UNKNOWN_NEXT_PEER
instead of WIRE_TEMPORARY_CHANNEL_FAILURE as soon as we get a failure
with a peer.  To provoke a temporary failure in test_pay_disconnect we
take the node offline.

Reported-by: Christian Decker @cdecker
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

channeld/channel.c

@@ -1894,8 +1894,7 @@ static void check_future_dataloss_fields(struct peer *peer,
 				       remote_current_per_commitment_point)));
 
 	/* We have to send them an error to trigger dropping to chain. */
-	peer_failed(&peer->cs, &peer->channel_id,
-		    "Catastrophic failure: please close channel");
+	peer_failed(&peer->cs, &peer->channel_id, "Awaiting unilateral close");
 }
 
 /* BOLT #2:

lightningd/channel.c

@@ -356,6 +356,10 @@ void channel_fail_permanent(struct channel *channel, const char *fmt, ...)
 	channel_set_owner(channel, NULL, false);
 	/* Drop non-cooperatively (unilateral) to chain. */
 	drop_to_chain(ld, channel, false);
+
+	if (channel_active(channel))
+		channel_set_state(channel, channel->state, AWAITING_UNILATERAL);
+
 	tal_free(why);
 }
 

lightningd/channel_control.c

@@ -105,7 +105,6 @@ static void peer_got_shutdown(struct channel *channel, const u8 *msg)
 static void channel_fail_fallen_behind(struct channel *channel, const u8 *msg)
 {
 	struct pubkey per_commitment_point;
-	struct channel_id cid;
 
 	if (!fromwire_channel_fail_fallen_behind(msg, &per_commitment_point)) {
 		channel_internal_error(channel,
@@ -117,17 +116,8 @@ static void channel_fail_fallen_behind(struct channel *channel, const u8 *msg)
 	channel->future_per_commitment_point
 		= tal_dup(channel, struct pubkey, &per_commitment_point);
 
-	/* TODO(cdecker) Selectively save updated fields to DB */
-	wallet_channel_save(channel->peer->ld->wallet, channel);
-
-	/* We don't fail yet, since we want daemon to send them an error
-	 * to trigger rebroadcasting.  But make sure we set error now in
-	 * case something else goes wrong! */
-	derive_channel_id(&cid,
-			  &channel->funding_txid,
-			  channel->funding_outnum);
-	channel->error = towire_errorfmt(channel, &cid,
-					 "Catastrophic failure: please close channel");
+	/* Peer sees this, so send a generic msg about unilateral close. */
+	channel_fail_permanent(channel,	"Awaiting unilateral close");
 }
 
 static void peer_start_closingd_after_shutdown(struct channel *channel,

lightningd/channel_state.h

@@ -19,6 +19,9 @@ enum channel_state {
 	/* Waiting for onchain event. */
 	CLOSINGD_COMPLETE,
 
+	/* Waiting for unilateral close to hit blockchain. */
+	AWAITING_UNILATERAL,
+
 	/* We've seen the funding spent, we're waiting for onchaind. */
 	FUNDING_SPEND_SEEN,
 

lightningd/peer_control.c

@@ -503,6 +503,19 @@ void peer_connected(struct lightningd *ld, const u8 *msg,
 			/* Channel is supposed to be active! */
 			abort();
 
+		/* We consider this "active" but we only send an error */
+		case AWAITING_UNILATERAL: {
+			struct channel_id cid;
+			derive_channel_id(&cid,
+					  &channel->funding_txid,
+					  channel->funding_outnum);
+			/* channel->error is not saved in db, so this can
+			 * happen if we restart. */
+			error = towire_errorfmt(tmpctx, &cid,
+						"Awaiting unilateral close");
+			goto send_error;
+		}
+
 		case CHANNELD_AWAITING_LOCKIN:
 		case CHANNELD_NORMAL:
 		case CHANNELD_SHUTTING_DOWN:

tests/test_connection.py

@@ -1205,7 +1205,7 @@ def test_dataloss_protection(node_factory, bitcoind):
     l2.start()
 
     # l2 should freak out!
-    l2.daemon.wait_for_log("Catastrophic failure: please close channel")
+    l2.daemon.wait_for_log("Peer permanent failure in CHANNELD_NORMAL: Awaiting unilateral close")
 
     # l1 should drop to chain.
     l1.daemon.wait_for_log('sendrawtx exit 0')

tests/test_pay.py

@@ -85,7 +85,8 @@ def test_pay0(node_factory):
 @unittest.skipIf(not DEVELOPER, "needs DEVELOPER=1")
 def test_pay_disconnect(node_factory, bitcoind):
     """If the remote node has disconnected, we fail payment, but can try again when it reconnects"""
-    l1, l2 = node_factory.line_graph(2, opts={'dev-max-fee-multiplier': 5})
+    l1, l2 = node_factory.line_graph(2, opts={'dev-max-fee-multiplier': 5,
+                                              'may_reconnect': True})
 
     inv = l2.rpc.invoice(123000, 'test_pay_disconnect', 'description')
     rhash = inv['payment_hash']
@@ -93,21 +94,27 @@ def test_pay_disconnect(node_factory, bitcoind):
     # Can't use `pay` since that'd notice that we can't route, due to disabling channel_update
     route = l1.rpc.getroute(l2.info['id'], 123000, 1)["route"]
 
-    # Make l2 upset by asking for crazy fee.
-    l1.rpc.dev_setfees('150000')
-    # Wait for l1 notice
-    l1.daemon.wait_for_log(r'Peer permanent failure in CHANNELD_NORMAL: lightning_channeld: received ERROR channel .*: update_fee 150000 outside range 1875-75000')
+    l2.stop()
+    l1.daemon.wait_for_log('Disabling channel .*, active 1 -> 0')
 
     # Can't pay while its offline.
     with pytest.raises(RpcError):
         l1.rpc.sendpay(route, rhash)
     l1.daemon.wait_for_log('failed: WIRE_TEMPORARY_CHANNEL_FAILURE \\(First peer not ready\\)')
 
-    # Should fail due to temporary channel fail
+    l2.start()
+    l1.daemon.wait_for_log('peer_out WIRE_CHANNEL_REESTABLISH')
+
+    # Make l2 upset by asking for crazy fee.
+    l1.rpc.dev_setfees('150000')
+    # Wait for l1 notice
+    l1.daemon.wait_for_log(r'Peer permanent failure in CHANNELD_NORMAL: lightning_channeld: received ERROR channel .*: update_fee 150000 outside range 1875-75000')
+
+    # Should fail due to permenant channel fail
     with pytest.raises(RpcError):
         l1.rpc.sendpay(route, rhash)
 
-    l1.daemon.wait_for_log('failed: WIRE_TEMPORARY_CHANNEL_FAILURE \\(First peer not ready\\)')
+    l1.daemon.wait_for_log('failed: WIRE_UNKNOWN_NEXT_PEER \\(First peer not ready\\)')
     assert not l1.daemon.is_in_log('Payment is still in progress')
 
     # After it sees block, someone should close channel.