From 140d2d02478d2f79ea57851c8b2309645f6725e5 Mon Sep 17 00:00:00 2001 From: SomberNight Date: Mon, 15 Jan 2024 17:31:39 +0000 Subject: [PATCH] lnpeer: fix timing issues in reestablish_channel, for dataloss case Assume Alice and Bob have a channel, and Alice is on an old state, but neither of them knows the latter yet. Timing scenarios: 1. Alice sends reest first, and Bob receives it before sending reest himself - old code: Bob realises Alice is behind, Bob will force-close, Bob won't send reest to Alice, so Alice does not learn she is behind - new code: Bob realises Alice is behind, Bob will force-close, Bob will still send reest to Alice, so Alice learns she is behind. 2. Bob sends reest first, and Alice receives it before sending reest herself - old code: Alice learns she is behind. Alice won't send reest to Bob, so Bob does not learn he is ahead, so Bob won't force-close. - new code: Alice learns she is behind. Alice will still send reest to Bob though with ctn=0 instead of actual. Bob learns he is ahead, so Bob will force close. 3. Alice and Bob both send reest, and then they both receive what the other sent - no change: Alice and Bob both learn Alice is behind. Bob will force-close. --- electrum/lnpeer.py | 70 +++++++++++++++++++++++------------ electrum/tests/test_lnpeer.py | 68 ++++++++++++++++++++++------------ 2 files changed, 90 insertions(+), 48 deletions(-) diff --git a/electrum/lnpeer.py b/electrum/lnpeer.py index 37b67579e..e3db405f7 100644 --- a/electrum/lnpeer.py +++ b/electrum/lnpeer.py @@ -1146,7 +1146,7 @@ class Peer(Logger): self.logger.info(f"tried to force-close channel {chan.get_id_for_log()} " f"but close option is not allowed. {chan.get_state()=!r}") - def on_channel_reestablish(self, chan, msg): + def on_channel_reestablish(self, chan: Channel, msg): their_next_local_ctn = msg["next_commitment_number"] their_oldest_unrevoked_remote_ctn = msg["next_revocation_number"] their_local_pcp = msg.get("my_current_per_commitment_point") @@ -1230,40 +1230,23 @@ class Peer(Logger): self.lnworker.save_channel(chan) chan.peer_state = PeerState.BAD # raise after we send channel_reestablish, so the remote can realize they are ahead - fut.set_exception(RemoteMisbehaving("remote ahead of us")) + # FIXME what if we have multiple chans with peer? timing... + fut.set_exception(GracefulDisconnect("remote ahead of us")) elif we_are_ahead: self.logger.warning(f"channel_reestablish ({chan.get_id_for_log()}): we are ahead of remote! trying to force-close.") self.schedule_force_closing(chan.channel_id) - fut.set_exception(RemoteMisbehaving("we are ahead of remote")) + # FIXME what if we have multiple chans with peer? timing... + fut.set_exception(GracefulDisconnect("we are ahead of remote")) else: # all good fut.set_result((we_must_resend_revoke_and_ack, their_next_local_ctn)) - async def reestablish_channel(self, chan: Channel): - await self.initialized + def _send_channel_reestablish(self, chan: Channel): + assert self.is_initialized() chan_id = chan.channel_id - if chan.should_request_force_close: - chan.set_state(ChannelState.REQUESTED_FCLOSE) - await self.request_force_close(chan_id) - chan.should_request_force_close = False - return - assert ChannelState.PREOPENING < chan.get_state() < ChannelState.FORCE_CLOSING - if chan.peer_state != PeerState.DISCONNECTED: - self.logger.info( - f'reestablish_channel was called but channel {chan.get_id_for_log()} ' - f'already in peer_state {chan.peer_state!r}') - return - chan.peer_state = PeerState.REESTABLISHING - util.trigger_callback('channel', self.lnworker.wallet, chan) # ctns - oldest_unrevoked_local_ctn = chan.get_oldest_unrevoked_ctn(LOCAL) - latest_local_ctn = chan.get_latest_ctn(LOCAL) next_local_ctn = chan.get_next_ctn(LOCAL) oldest_unrevoked_remote_ctn = chan.get_oldest_unrevoked_ctn(REMOTE) - latest_remote_ctn = chan.get_latest_ctn(REMOTE) - next_remote_ctn = chan.get_next_ctn(REMOTE) - # BOLT-02: "A node [...] upon disconnection [...] MUST reverse any uncommitted updates sent by the other side" - chan.hm.discard_unsigned_remote_updates() # send message assert chan.is_static_remotekey_enabled() latest_secret, latest_point = chan.get_secret_and_point(LOCAL, 0) @@ -1284,6 +1267,45 @@ class Peer(Logger): f'(next_local_ctn={next_local_ctn}, ' f'oldest_unrevoked_remote_ctn={oldest_unrevoked_remote_ctn})') + async def reestablish_channel(self, chan: Channel): + await self.initialized + chan_id = chan.channel_id + if chan.get_state() == ChannelState.WE_ARE_TOXIC: + # Depending on timing, the remote might not know we are behind. + # We should let them know, so that they force-close. + # We do "request force-close" with ctn=0, instead of leaking our actual ctns, + # to decrease the remote's confidence of actual data loss on our part. + await self.request_force_close(chan_id) + return + if chan.get_state() == ChannelState.FORCE_CLOSING: + # We likely got here because we found out that we are ahead (i.e. remote lost state). + # Depending on timing, the remote might not know they are behind. + # We should let them know: + self._send_channel_reestablish(chan) + return + if chan.should_request_force_close: + chan.set_state(ChannelState.REQUESTED_FCLOSE) + await self.request_force_close(chan_id) + chan.should_request_force_close = False + return + # if we get here, we will try to do a proper reestablish + if not (ChannelState.PREOPENING < chan.get_state() < ChannelState.FORCE_CLOSING): + raise Exception(f"unexpected {chan.get_state()=} for reestablish") + if chan.peer_state != PeerState.DISCONNECTED: + self.logger.info( + f'reestablish_channel was called but channel {chan.get_id_for_log()} ' + f'already in peer_state {chan.peer_state!r}') + return + chan.peer_state = PeerState.REESTABLISHING + util.trigger_callback('channel', self.lnworker.wallet, chan) + # ctns + oldest_unrevoked_local_ctn = chan.get_oldest_unrevoked_ctn(LOCAL) + next_local_ctn = chan.get_next_ctn(LOCAL) + oldest_unrevoked_remote_ctn = chan.get_oldest_unrevoked_ctn(REMOTE) + # BOLT-02: "A node [...] upon disconnection [...] MUST reverse any uncommitted updates sent by the other side" + chan.hm.discard_unsigned_remote_updates() + # send message + self._send_channel_reestablish(chan) # wait until we receive their channel_reestablish fut = self.channel_reestablish_msg[chan_id] await fut diff --git a/electrum/tests/test_lnpeer.py b/electrum/tests/test_lnpeer.py index fc25a8bb0..8d04474a1 100644 --- a/electrum/tests/test_lnpeer.py +++ b/electrum/tests/test_lnpeer.py @@ -566,30 +566,50 @@ class TestPeerDirect(TestPeer): await gath async def test_reestablish_with_old_state(self): - random_seed = os.urandom(32) - alice_channel, bob_channel = create_test_channels(random_seed=random_seed) - alice_channel_0, bob_channel_0 = create_test_channels(random_seed=random_seed) # these are identical - p1, p2, w1, w2, _q1, _q2 = self.prepare_peers(alice_channel, bob_channel) - lnaddr, pay_req = self.prepare_invoice(w2) - async def pay(): - result, log = await w1.pay_invoice(pay_req) - self.assertEqual(result, True) - gath.cancel() - gath = asyncio.gather(pay(), p1._message_loop(), p2._message_loop(), p1.htlc_switch(), p2.htlc_switch()) - with self.assertRaises(asyncio.CancelledError): - await gath - p1, p2, w1, w2, _q1, _q2 = self.prepare_peers(alice_channel_0, bob_channel) - for chan in (alice_channel_0, bob_channel): - chan.peer_state = PeerState.DISCONNECTED - async def reestablish(): - await asyncio.gather( - p1.reestablish_channel(alice_channel_0), - p2.reestablish_channel(bob_channel)) - gath = asyncio.gather(reestablish(), p1._message_loop(), p2._message_loop(), p1.htlc_switch(), p2.htlc_switch()) - with self.assertRaises(lnutil.RemoteMisbehaving): - await gath - self.assertEqual(alice_channel_0.peer_state, PeerState.BAD) - self.assertEqual(bob_channel._state, ChannelState.FORCE_CLOSING) + async def f(alice_slow: bool, bob_slow: bool): + random_seed = os.urandom(32) + alice_channel, bob_channel = create_test_channels(random_seed=random_seed) + alice_channel_0, bob_channel_0 = create_test_channels(random_seed=random_seed) # these are identical + p1, p2, w1, w2, _q1, _q2 = self.prepare_peers(alice_channel, bob_channel) + lnaddr, pay_req = self.prepare_invoice(w2) + async def pay(): + result, log = await w1.pay_invoice(pay_req) + self.assertEqual(result, True) + gath.cancel() + gath = asyncio.gather(pay(), p1._message_loop(), p2._message_loop(), p1.htlc_switch(), p2.htlc_switch()) + with self.assertRaises(asyncio.CancelledError): + await gath + p1, p2, w1, w2, _q1, _q2 = self.prepare_peers(alice_channel_0, bob_channel) + for chan in (alice_channel_0, bob_channel): + chan.peer_state = PeerState.DISCONNECTED + + async def alice_sends_reest(): + if alice_slow: await asyncio.sleep(0.05) + await p1.reestablish_channel(alice_channel_0) + async def bob_sends_reest(): + if bob_slow: await asyncio.sleep(0.05) + await p2.reestablish_channel(bob_channel) + + with self.assertRaises(GracefulDisconnect): + async with OldTaskGroup() as group: + await group.spawn(p1._message_loop()) + await group.spawn(p1.htlc_switch()) + await group.spawn(p2._message_loop()) + await group.spawn(p2.htlc_switch()) + await group.spawn(alice_sends_reest) + await group.spawn(bob_sends_reest) + self.assertEqual(alice_channel_0.peer_state, PeerState.BAD) + self.assertEqual(alice_channel_0._state, ChannelState.WE_ARE_TOXIC) + self.assertEqual(bob_channel._state, ChannelState.FORCE_CLOSING) + + with self.subTest(msg="both fast"): + # FIXME: we want to test the case where both Alice and Bob sends channel-reestablish before + # receiving what the other sent. This is not a reliable way to do that... + await f(alice_slow=False, bob_slow=False) + with self.subTest(msg="alice is slow"): + await f(alice_slow=True, bob_slow=False) + with self.subTest(msg="bob is slow"): + await f(alice_slow=False, bob_slow=True) @staticmethod def _send_fake_htlc(peer: Peer, chan: Channel) -> UpdateAddHtlc: