rust-lightning/fuzz/src/chanmon_consistency.rs

1209 lines
54 KiB
Rust
Raw Normal View History

// This file is Copyright its original authors, visible in version control
// history.
//
// This file is licensed under the Apache License, Version 2.0 <LICENSE-APACHE
// or http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your option.
// You may not use this file except in accordance with one or both of these
// licenses.
//! Test that monitor update failures don't get our channel state out of sync.
//! One of the biggest concern with the monitor update failure handling code is that messages
//! resent after monitor updating is restored are delivered out-of-order, resulting in
//! commitment_signed messages having "invalid signatures".
//! To test this we stand up a network of three nodes and read bytes from the fuzz input to denote
//! actions such as sending payments, handling events, or changing monitor update return values on
//! a per-node basis. This should allow it to find any cases where the ordering of actions results
//! in us getting out of sync with ourselves, and, assuming at least one of our recieve- or
//! send-side handling is correct, other peers. We consider it a failure if any action results in a
//! channel being force-closed.
use bitcoin::blockdata::block::BlockHeader;
use bitcoin::blockdata::constants::genesis_block;
use bitcoin::blockdata::transaction::{Transaction, TxOut};
use bitcoin::blockdata::script::{Builder, Script};
use bitcoin::blockdata::opcodes;
use bitcoin::network::constants::Network;
use bitcoin::hashes::Hash as TraitImport;
use bitcoin::hashes::sha256::Hash as Sha256;
2020-04-27 18:13:27 +02:00
use bitcoin::hash_types::{BlockHash, WPubkeyHash};
use lightning::chain;
use lightning::chain::{BestBlock, ChannelMonitorUpdateErr, chainmonitor, channelmonitor, Confirm, Watch};
use lightning::chain::channelmonitor::{ChannelMonitor, MonitorEvent};
use lightning::chain::transaction::OutPoint;
use lightning::chain::chaininterface::{BroadcasterInterface, ConfirmationTarget, FeeEstimator};
use lightning::chain::keysinterface::{KeyMaterial, KeysInterface, InMemorySigner, Recipient};
use lightning::ln::{PaymentHash, PaymentPreimage, PaymentSecret};
use lightning::ln::channelmanager::{ChainParameters, ChannelManager, PaymentSendFailure, ChannelManagerReadArgs};
use lightning::ln::channel::FEE_SPIKE_BUFFER_FEE_INCREASE_MULTIPLE;
use lightning::ln::features::{ChannelFeatures, InitFeatures, NodeFeatures};
use lightning::ln::msgs::{CommitmentUpdate, ChannelMessageHandler, DecodeError, UpdateAddHTLC, Init};
use lightning::ln::script::ShutdownScript;
use lightning::util::enforcing_trait_impls::{EnforcingSigner, EnforcementState};
use lightning::util::errors::APIError;
use lightning::util::events;
use lightning::util::logger::Logger;
use lightning::util::config::UserConfig;
use lightning::util::events::MessageSendEventsProvider;
use lightning::util::ser::{Readable, ReadableArgs, Writeable, Writer};
2020-05-02 09:37:38 -04:00
use lightning::routing::router::{Route, RouteHop};
use utils::test_logger::{self, Output};
use utils::test_persister::TestPersister;
2022-05-05 17:59:38 +02:00
use bitcoin::secp256k1::{PublicKey,SecretKey};
use bitcoin::secp256k1::ecdsa::RecoverableSignature;
2020-04-27 16:51:59 +02:00
use bitcoin::secp256k1::Secp256k1;
use std::mem;
use std::cmp::{self, Ordering};
use std::collections::{HashSet, hash_map, HashMap};
use std::sync::{Arc,Mutex};
use std::sync::atomic;
use std::io::Cursor;
use bitcoin::bech32::u5;
const MAX_FEE: u32 = 10_000;
struct FuzzEstimator {
ret_val: atomic::AtomicU32,
}
impl FeeEstimator for FuzzEstimator {
fn get_est_sat_per_1000_weight(&self, conf_target: ConfirmationTarget) -> u32 {
// We force-close channels if our counterparty sends us a feerate which is a small multiple
// of our HighPriority fee estimate or smaller than our Background fee estimate. Thus, we
// always return a HighPriority feerate here which is >= the maximum Normal feerate and a
// Background feerate which is <= the minimum Normal feerate.
match conf_target {
ConfirmationTarget::HighPriority => MAX_FEE,
ConfirmationTarget::Background => 253,
ConfirmationTarget::Normal => cmp::min(self.ret_val.load(atomic::Ordering::Acquire), MAX_FEE),
}
}
}
pub struct TestBroadcaster {}
impl BroadcasterInterface for TestBroadcaster {
fn broadcast_transaction(&self, _tx: &Transaction) { }
}
pub struct VecWriter(pub Vec<u8>);
impl Writer for VecWriter {
fn write_all(&mut self, buf: &[u8]) -> Result<(), ::std::io::Error> {
self.0.extend_from_slice(buf);
Ok(())
}
}
struct TestChainMonitor {
pub logger: Arc<dyn Logger>,
pub keys: Arc<KeyProvider>,
pub persister: Arc<TestPersister>,
pub chain_monitor: Arc<chainmonitor::ChainMonitor<EnforcingSigner, Arc<dyn chain::Filter>, Arc<TestBroadcaster>, Arc<FuzzEstimator>, Arc<dyn Logger>, Arc<TestPersister>>>,
// If we reload a node with an old copy of ChannelMonitors, the ChannelManager deserialization
// logic will automatically force-close our channels for us (as we don't have an up-to-date
// monitor implying we are not able to punish misbehaving counterparties). Because this test
// "fails" if we ever force-close a channel, we avoid doing so, always saving the latest
// fully-serialized monitor state here, as well as the corresponding update_id.
pub latest_monitors: Mutex<HashMap<OutPoint, (u64, Vec<u8>)>>,
pub should_update_manager: atomic::AtomicBool,
}
impl TestChainMonitor {
pub fn new(broadcaster: Arc<TestBroadcaster>, logger: Arc<dyn Logger>, feeest: Arc<FuzzEstimator>, persister: Arc<TestPersister>, keys: Arc<KeyProvider>) -> Self {
Self {
chain_monitor: Arc::new(chainmonitor::ChainMonitor::new(None, broadcaster, logger.clone(), feeest, Arc::clone(&persister))),
logger,
keys,
persister,
latest_monitors: Mutex::new(HashMap::new()),
should_update_manager: atomic::AtomicBool::new(false),
}
}
}
impl chain::Watch<EnforcingSigner> for TestChainMonitor {
fn watch_channel(&self, funding_txo: OutPoint, monitor: channelmonitor::ChannelMonitor<EnforcingSigner>) -> Result<(), chain::ChannelMonitorUpdateErr> {
let mut ser = VecWriter(Vec::new());
monitor.write(&mut ser).unwrap();
if let Some(_) = self.latest_monitors.lock().unwrap().insert(funding_txo, (monitor.get_latest_update_id(), ser.0)) {
panic!("Already had monitor pre-watch_channel");
}
self.should_update_manager.store(true, atomic::Ordering::Relaxed);
self.chain_monitor.watch_channel(funding_txo, monitor)
}
fn update_channel(&self, funding_txo: OutPoint, update: channelmonitor::ChannelMonitorUpdate) -> Result<(), chain::ChannelMonitorUpdateErr> {
let mut map_lock = self.latest_monitors.lock().unwrap();
let mut map_entry = match map_lock.entry(funding_txo) {
hash_map::Entry::Occupied(entry) => entry,
hash_map::Entry::Vacant(_) => panic!("Didn't have monitor on update call"),
};
let deserialized_monitor = <(BlockHash, channelmonitor::ChannelMonitor<EnforcingSigner>)>::
read(&mut Cursor::new(&map_entry.get().1), &*self.keys).unwrap().1;
deserialized_monitor.update_monitor(&update, &&TestBroadcaster{}, &&FuzzEstimator { ret_val: atomic::AtomicU32::new(253) }, &self.logger).unwrap();
let mut ser = VecWriter(Vec::new());
deserialized_monitor.write(&mut ser).unwrap();
map_entry.insert((update.update_id, ser.0));
self.should_update_manager.store(true, atomic::Ordering::Relaxed);
self.chain_monitor.update_channel(funding_txo, update)
}
fn release_pending_monitor_events(&self) -> Vec<(OutPoint, Vec<MonitorEvent>)> {
return self.chain_monitor.release_pending_monitor_events();
}
}
struct KeyProvider {
node_id: u8,
rand_bytes_id: atomic::AtomicU32,
enforcement_states: Mutex<HashMap<[u8;32], Arc<Mutex<EnforcementState>>>>,
}
impl KeysInterface for KeyProvider {
type Signer = EnforcingSigner;
fn get_node_secret(&self, _recipient: Recipient) -> Result<SecretKey, ()> {
Ok(SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, self.node_id]).unwrap())
}
fn get_inbound_payment_key_material(&self) -> KeyMaterial {
KeyMaterial([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, self.node_id])
}
fn get_destination_script(&self) -> Script {
let secp_ctx = Secp256k1::signing_only();
let channel_monitor_claim_key = SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, self.node_id]).unwrap();
2020-04-27 18:13:27 +02:00
let our_channel_monitor_claim_key_hash = WPubkeyHash::hash(&PublicKey::from_secret_key(&secp_ctx, &channel_monitor_claim_key).serialize());
Builder::new().push_opcode(opcodes::all::OP_PUSHBYTES_0).push_slice(&our_channel_monitor_claim_key_hash[..]).into_script()
}
fn get_shutdown_scriptpubkey(&self) -> ShutdownScript {
let secp_ctx = Secp256k1::signing_only();
let secret_key = SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, self.node_id]).unwrap();
let pubkey_hash = WPubkeyHash::hash(&PublicKey::from_secret_key(&secp_ctx, &secret_key).serialize());
ShutdownScript::new_p2wpkh(&pubkey_hash)
}
fn get_channel_signer(&self, _inbound: bool, channel_value_satoshis: u64) -> EnforcingSigner {
2020-02-04 09:15:59 -08:00
let secp_ctx = Secp256k1::signing_only();
2020-12-05 18:56:27 +01:00
let id = self.rand_bytes_id.fetch_add(1, atomic::Ordering::Relaxed);
let keys = InMemorySigner::new(
2020-02-04 09:15:59 -08:00
&secp_ctx,
self.get_node_secret(Recipient::Node).unwrap(),
2020-02-04 09:15:59 -08:00
SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, self.node_id]).unwrap(),
SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5, self.node_id]).unwrap(),
SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6, self.node_id]).unwrap(),
SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7, self.node_id]).unwrap(),
SecretKey::from_slice(&[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, self.node_id]).unwrap(),
[id as u8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 9, self.node_id],
channel_value_satoshis,
[0; 32],
2020-12-05 18:56:27 +01:00
);
let revoked_commitment = self.make_enforcement_state_cell(keys.commitment_seed);
EnforcingSigner::new_with_revoked(keys, revoked_commitment, false)
}
fn get_secure_random_bytes(&self) -> [u8; 32] {
let id = self.rand_bytes_id.fetch_add(1, atomic::Ordering::Relaxed);
let mut res = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, self.node_id];
res[30-4..30].copy_from_slice(&id.to_le_bytes());
res
}
fn read_chan_signer(&self, buffer: &[u8]) -> Result<Self::Signer, DecodeError> {
2020-12-05 18:56:27 +01:00
let mut reader = std::io::Cursor::new(buffer);
let inner: InMemorySigner = ReadableArgs::read(&mut reader, self.get_node_secret(Recipient::Node).unwrap())?;
let state = self.make_enforcement_state_cell(inner.commitment_seed);
2020-12-05 18:56:27 +01:00
Ok(EnforcingSigner {
2020-12-05 18:56:27 +01:00
inner,
state,
disable_revocation_policy_check: false,
2020-12-05 18:56:27 +01:00
})
}
fn sign_invoice(&self, _hrp_bytes: &[u8], _invoice_data: &[u5], _recipient: Recipient) -> Result<RecoverableSignature, ()> {
unreachable!()
}
2020-12-05 18:56:27 +01:00
}
impl KeyProvider {
fn make_enforcement_state_cell(&self, commitment_seed: [u8; 32]) -> Arc<Mutex<EnforcementState>> {
let mut revoked_commitments = self.enforcement_states.lock().unwrap();
2020-12-05 18:56:27 +01:00
if !revoked_commitments.contains_key(&commitment_seed) {
revoked_commitments.insert(commitment_seed, Arc::new(Mutex::new(EnforcementState::new())));
2020-12-05 18:56:27 +01:00
}
let cell = revoked_commitments.get(&commitment_seed).unwrap();
Arc::clone(cell)
}
}
#[inline]
fn check_api_err(api_err: APIError) {
match api_err {
APIError::APIMisuseError { .. } => panic!("We can't misuse the API"),
APIError::FeeRateTooHigh { .. } => panic!("We can't send too much fee?"),
APIError::RouteError { .. } => panic!("Our routes should work"),
APIError::ChannelUnavailable { err } => {
// Test the error against a list of errors we can hit, and reject
// all others. If you hit this panic, the list of acceptable errors
// is probably just stale and you should add new messages here.
match err.as_str() {
"Peer for first hop currently disconnected/pending monitor update!" => {},
_ if err.starts_with("Cannot push more than their max accepted HTLCs ") => {},
_ if err.starts_with("Cannot send value that would put us over the max HTLC value in flight our peer will accept ") => {},
_ if err.starts_with("Cannot send value that would put our balance under counterparty-announced channel reserve value") => {},
_ if err.starts_with("Cannot send value that would put counterparty balance under holder-announced channel reserve value") => {},
_ if err.starts_with("Cannot send value that would overdraw remaining funds.") => {},
_ if err.starts_with("Cannot send value that would not leave enough to pay for fees.") => {},
_ if err.starts_with("Cannot send value that would put our exposure to dust HTLCs at") => {},
_ => panic!("{}", err),
}
},
APIError::MonitorUpdateFailed => {
// We can (obviously) temp-fail a monitor update
},
APIError::IncompatibleShutdownScript { .. } => panic!("Cannot send an incompatible shutdown script"),
}
}
#[inline]
fn check_payment_err(send_err: PaymentSendFailure) {
match send_err {
PaymentSendFailure::ParameterError(api_err) => check_api_err(api_err),
PaymentSendFailure::PathParameterError(per_path_results) => {
for res in per_path_results { if let Err(api_err) = res { check_api_err(api_err); } }
},
PaymentSendFailure::AllFailedRetrySafe(per_path_results) => {
for api_err in per_path_results { check_api_err(api_err); }
},
PaymentSendFailure::PartialFailure { results, .. } => {
for res in results { if let Err(api_err) = res { check_api_err(api_err); } }
},
}
}
type ChanMan = ChannelManager<EnforcingSigner, Arc<TestChainMonitor>, Arc<TestBroadcaster>, Arc<KeyProvider>, Arc<FuzzEstimator>, Arc<dyn Logger>>;
#[inline]
fn get_payment_secret_hash(dest: &ChanMan, payment_id: &mut u8) -> Option<(PaymentSecret, PaymentHash)> {
let mut payment_hash;
for _ in 0..256 {
payment_hash = PaymentHash(Sha256::hash(&[*payment_id; 1]).into_inner());
if let Ok(payment_secret) = dest.create_inbound_payment_for_hash(payment_hash, None, 3600) {
return Some((payment_secret, payment_hash));
}
*payment_id = payment_id.wrapping_add(1);
}
None
}
#[inline]
fn send_payment(source: &ChanMan, dest: &ChanMan, dest_chan_id: u64, amt: u64, payment_id: &mut u8) -> bool {
let (payment_secret, payment_hash) =
if let Some((secret, hash)) = get_payment_secret_hash(dest, payment_id) { (secret, hash) } else { return true; };
if let Err(err) = source.send_payment(&Route {
paths: vec![vec![RouteHop {
pubkey: dest.get_our_node_id(),
node_features: NodeFeatures::known(),
short_channel_id: dest_chan_id,
channel_features: ChannelFeatures::known(),
fee_msat: amt,
cltv_expiry_delta: 200,
}]],
2022-01-21 10:39:01 +01:00
payment_params: None,
}, payment_hash, &Some(payment_secret)) {
check_payment_err(err);
false
} else { true }
}
#[inline]
fn send_hop_payment(source: &ChanMan, middle: &ChanMan, middle_chan_id: u64, dest: &ChanMan, dest_chan_id: u64, amt: u64, payment_id: &mut u8) -> bool {
let (payment_secret, payment_hash) =
if let Some((secret, hash)) = get_payment_secret_hash(dest, payment_id) { (secret, hash) } else { return true; };
if let Err(err) = source.send_payment(&Route {
paths: vec![vec![RouteHop {
pubkey: middle.get_our_node_id(),
node_features: NodeFeatures::known(),
short_channel_id: middle_chan_id,
channel_features: ChannelFeatures::known(),
fee_msat: 50000,
cltv_expiry_delta: 100,
},RouteHop {
pubkey: dest.get_our_node_id(),
node_features: NodeFeatures::known(),
short_channel_id: dest_chan_id,
channel_features: ChannelFeatures::known(),
fee_msat: amt,
cltv_expiry_delta: 200,
}]],
2022-01-21 10:39:01 +01:00
payment_params: None,
}, payment_hash, &Some(payment_secret)) {
check_payment_err(err);
false
} else { true }
}
#[inline]
pub fn do_test<Out: Output>(data: &[u8], underlying_out: Out) {
let out = SearchingOutput::new(underlying_out);
let broadcast = Arc::new(TestBroadcaster{});
macro_rules! make_node {
($node_id: expr, $fee_estimator: expr) => { {
let logger: Arc<dyn Logger> = Arc::new(test_logger::TestLogger::new($node_id.to_string(), out.clone()));
let keys_manager = Arc::new(KeyProvider { node_id: $node_id, rand_bytes_id: atomic::AtomicU32::new(0), enforcement_states: Mutex::new(HashMap::new()) });
let monitor = Arc::new(TestChainMonitor::new(broadcast.clone(), logger.clone(), $fee_estimator.clone(),
Arc::new(TestPersister { update_ret: Mutex::new(Ok(())) }), Arc::clone(&keys_manager)));
let mut config = UserConfig::default();
config.channel_options.forwarding_fee_proportional_millionths = 0;
config.channel_options.announced_channel = true;
let network = Network::Bitcoin;
let params = ChainParameters {
network,
best_block: BestBlock::from_genesis(network),
};
(ChannelManager::new($fee_estimator.clone(), monitor.clone(), broadcast.clone(), Arc::clone(&logger), keys_manager.clone(), config, params),
2020-12-05 18:56:27 +01:00
monitor, keys_manager)
} }
}
macro_rules! reload_node {
($ser: expr, $node_id: expr, $old_monitors: expr, $keys_manager: expr, $fee_estimator: expr) => { {
2020-12-05 18:56:27 +01:00
let keys_manager = Arc::clone(& $keys_manager);
let logger: Arc<dyn Logger> = Arc::new(test_logger::TestLogger::new($node_id.to_string(), out.clone()));
let chain_monitor = Arc::new(TestChainMonitor::new(broadcast.clone(), logger.clone(), $fee_estimator.clone(),
Arc::new(TestPersister { update_ret: Mutex::new(Ok(())) }), Arc::clone(& $keys_manager)));
let mut config = UserConfig::default();
config.channel_options.forwarding_fee_proportional_millionths = 0;
config.channel_options.announced_channel = true;
let mut monitors = HashMap::new();
let mut old_monitors = $old_monitors.latest_monitors.lock().unwrap();
for (outpoint, (update_id, monitor_ser)) in old_monitors.drain() {
monitors.insert(outpoint, <(BlockHash, ChannelMonitor<EnforcingSigner>)>::read(&mut Cursor::new(&monitor_ser), &*$keys_manager).expect("Failed to read monitor").1);
chain_monitor.latest_monitors.lock().unwrap().insert(outpoint, (update_id, monitor_ser));
}
let mut monitor_refs = HashMap::new();
for (outpoint, monitor) in monitors.iter_mut() {
monitor_refs.insert(*outpoint, monitor);
}
let read_args = ChannelManagerReadArgs {
keys_manager,
fee_estimator: $fee_estimator.clone(),
chain_monitor: chain_monitor.clone(),
tx_broadcaster: broadcast.clone(),
logger,
default_config: config,
channel_monitors: monitor_refs,
};
let res = (<(BlockHash, ChanMan)>::read(&mut Cursor::new(&$ser.0), read_args).expect("Failed to read manager").1, chain_monitor.clone());
for (funding_txo, mon) in monitors.drain() {
assert!(chain_monitor.chain_monitor.watch_channel(funding_txo, mon).is_ok());
}
res
} }
}
let mut channel_txn = Vec::new();
macro_rules! make_channel {
($source: expr, $dest: expr, $chan_id: expr) => { {
$source.peer_connected(&$dest.get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
$dest.peer_connected(&$source.get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
$source.create_channel($dest.get_our_node_id(), 100_000, 42, 0, None).unwrap();
let open_channel = {
let events = $source.get_and_clear_pending_msg_events();
assert_eq!(events.len(), 1);
if let events::MessageSendEvent::SendOpenChannel { ref msg, .. } = events[0] {
msg.clone()
} else { panic!("Wrong event type"); }
};
$dest.handle_open_channel(&$source.get_our_node_id(), InitFeatures::known(), &open_channel);
let accept_channel = {
let events = $dest.get_and_clear_pending_msg_events();
assert_eq!(events.len(), 1);
if let events::MessageSendEvent::SendAcceptChannel { ref msg, .. } = events[0] {
msg.clone()
} else { panic!("Wrong event type"); }
};
$source.handle_accept_channel(&$dest.get_our_node_id(), InitFeatures::known(), &accept_channel);
let funding_output;
{
let events = $source.get_and_clear_pending_events();
assert_eq!(events.len(), 1);
if let events::Event::FundingGenerationReady { ref temporary_channel_id, ref channel_value_satoshis, ref output_script, .. } = events[0] {
let tx = Transaction { version: $chan_id, lock_time: 0, input: Vec::new(), output: vec![TxOut {
value: *channel_value_satoshis, script_pubkey: output_script.clone(),
}]};
funding_output = OutPoint { txid: tx.txid(), index: 0 };
$source.funding_transaction_generated(&temporary_channel_id, &$dest.get_our_node_id(), tx.clone()).unwrap();
channel_txn.push(tx);
} else { panic!("Wrong event type"); }
}
let funding_created = {
let events = $source.get_and_clear_pending_msg_events();
assert_eq!(events.len(), 1);
if let events::MessageSendEvent::SendFundingCreated { ref msg, .. } = events[0] {
msg.clone()
} else { panic!("Wrong event type"); }
};
$dest.handle_funding_created(&$source.get_our_node_id(), &funding_created);
let funding_signed = {
let events = $dest.get_and_clear_pending_msg_events();
assert_eq!(events.len(), 1);
if let events::MessageSendEvent::SendFundingSigned { ref msg, .. } = events[0] {
msg.clone()
} else { panic!("Wrong event type"); }
};
$source.handle_funding_signed(&$dest.get_our_node_id(), &funding_signed);
funding_output
} }
}
macro_rules! confirm_txn {
($node: expr) => { {
let chain_hash = genesis_block(Network::Bitcoin).block_hash();
let mut header = BlockHeader { version: 0x20000000, prev_blockhash: chain_hash, merkle_root: Default::default(), time: 42, bits: 42, nonce: 42 };
let txdata: Vec<_> = channel_txn.iter().enumerate().map(|(i, tx)| (i + 1, tx)).collect();
$node.transactions_confirmed(&header, &txdata, 1);
for _ in 2..100 {
2020-08-25 17:12:00 -04:00
header = BlockHeader { version: 0x20000000, prev_blockhash: header.block_hash(), merkle_root: Default::default(), time: 42, bits: 42, nonce: 42 };
}
$node.best_block_updated(&header, 99);
} }
}
macro_rules! lock_fundings {
($nodes: expr) => { {
let mut node_events = Vec::new();
for node in $nodes.iter() {
node_events.push(node.get_and_clear_pending_msg_events());
}
for (idx, node_event) in node_events.iter().enumerate() {
for event in node_event {
if let events::MessageSendEvent::SendFundingLocked { ref node_id, ref msg } = event {
for node in $nodes.iter() {
if node.get_our_node_id() == *node_id {
node.handle_funding_locked(&$nodes[idx].get_our_node_id(), msg);
}
}
} else { panic!("Wrong event type"); }
}
}
for node in $nodes.iter() {
let events = node.get_and_clear_pending_msg_events();
for event in events {
if let events::MessageSendEvent::SendAnnouncementSignatures { .. } = event {
} else { panic!("Wrong event type"); }
}
}
} }
}
let fee_est_a = Arc::new(FuzzEstimator { ret_val: atomic::AtomicU32::new(253) });
let mut last_htlc_clear_fee_a = 253;
let fee_est_b = Arc::new(FuzzEstimator { ret_val: atomic::AtomicU32::new(253) });
let mut last_htlc_clear_fee_b = 253;
let fee_est_c = Arc::new(FuzzEstimator { ret_val: atomic::AtomicU32::new(253) });
let mut last_htlc_clear_fee_c = 253;
// 3 nodes is enough to hit all the possible cases, notably unknown-source-unknown-dest
// forwarding.
let (node_a, mut monitor_a, keys_manager_a) = make_node!(0, fee_est_a);
let (node_b, mut monitor_b, keys_manager_b) = make_node!(1, fee_est_b);
let (node_c, mut monitor_c, keys_manager_c) = make_node!(2, fee_est_c);
let mut nodes = [node_a, node_b, node_c];
let chan_1_funding = make_channel!(nodes[0], nodes[1], 0);
let chan_2_funding = make_channel!(nodes[1], nodes[2], 1);
for node in nodes.iter() {
confirm_txn!(node);
}
lock_fundings!(nodes);
let chan_a = nodes[0].list_usable_channels()[0].short_channel_id.unwrap();
let chan_b = nodes[2].list_usable_channels()[0].short_channel_id.unwrap();
let mut payment_id: u8 = 0;
let mut chan_a_disconnected = false;
let mut chan_b_disconnected = false;
let mut ab_events = Vec::new();
let mut ba_events = Vec::new();
let mut bc_events = Vec::new();
let mut cb_events = Vec::new();
let mut node_a_ser = VecWriter(Vec::new());
nodes[0].write(&mut node_a_ser).unwrap();
let mut node_b_ser = VecWriter(Vec::new());
nodes[1].write(&mut node_b_ser).unwrap();
let mut node_c_ser = VecWriter(Vec::new());
nodes[2].write(&mut node_c_ser).unwrap();
macro_rules! test_return {
() => { {
assert_eq!(nodes[0].list_channels().len(), 1);
assert_eq!(nodes[1].list_channels().len(), 2);
assert_eq!(nodes[2].list_channels().len(), 1);
return;
} }
}
let mut read_pos = 0;
macro_rules! get_slice {
($len: expr) => {
{
let slice_len = $len as usize;
if data.len() < read_pos + slice_len {
test_return!();
}
read_pos += slice_len;
&data[read_pos - slice_len..read_pos]
}
}
}
loop {
// Push any events from Node B onto ba_events and bc_events
macro_rules! push_excess_b_events {
($excess_events: expr, $expect_drop_node: expr) => { {
let a_id = nodes[0].get_our_node_id();
let expect_drop_node: Option<usize> = $expect_drop_node;
let expect_drop_id = if let Some(id) = expect_drop_node { Some(nodes[id].get_our_node_id()) } else { None };
for event in $excess_events {
let push_a = match event {
events::MessageSendEvent::UpdateHTLCs { ref node_id, .. } => {
if Some(*node_id) == expect_drop_id { panic!("peer_disconnected should drop msgs bound for the disconnected peer"); }
*node_id == a_id
},
events::MessageSendEvent::SendRevokeAndACK { ref node_id, .. } => {
if Some(*node_id) == expect_drop_id { panic!("peer_disconnected should drop msgs bound for the disconnected peer"); }
*node_id == a_id
},
events::MessageSendEvent::SendChannelReestablish { ref node_id, .. } => {
if Some(*node_id) == expect_drop_id { panic!("peer_disconnected should drop msgs bound for the disconnected peer"); }
*node_id == a_id
},
events::MessageSendEvent::SendFundingLocked { .. } => continue,
events::MessageSendEvent::SendAnnouncementSignatures { .. } => continue,
events::MessageSendEvent::SendChannelUpdate { ref node_id, ref msg } => {
assert_eq!(msg.contents.flags & 2, 0); // The disable bit must never be set!
if Some(*node_id) == expect_drop_id { panic!("peer_disconnected should drop msgs bound for the disconnected peer"); }
*node_id == a_id
},
_ => panic!("Unhandled message event {:?}", event),
};
if push_a { ba_events.push(event); } else { bc_events.push(event); }
}
} }
}
// While delivering messages, we select across three possible message selection processes
// to ensure we get as much coverage as possible. See the individual enum variants for more
// details.
#[derive(PartialEq)]
enum ProcessMessages {
/// Deliver all available messages, including fetching any new messages from
/// `get_and_clear_pending_msg_events()` (which may have side effects).
AllMessages,
/// Call `get_and_clear_pending_msg_events()` first, and then deliver up to one
/// message (which may already be queued).
OneMessage,
/// Deliver up to one already-queued message. This avoids any potential side-effects
/// of `get_and_clear_pending_msg_events()` (eg freeing the HTLC holding cell), which
/// provides potentially more coverage.
OnePendingMessage,
}
macro_rules! process_msg_events {
($node: expr, $corrupt_forward: expr, $limit_events: expr) => { {
let mut events = if $node == 1 {
let mut new_events = Vec::new();
mem::swap(&mut new_events, &mut ba_events);
new_events.extend_from_slice(&bc_events[..]);
bc_events.clear();
new_events
} else if $node == 0 {
let mut new_events = Vec::new();
mem::swap(&mut new_events, &mut ab_events);
new_events
} else {
let mut new_events = Vec::new();
mem::swap(&mut new_events, &mut cb_events);
new_events
};
let mut new_events = Vec::new();
if $limit_events != ProcessMessages::OnePendingMessage {
new_events = nodes[$node].get_and_clear_pending_msg_events();
}
let mut had_events = false;
let mut events_iter = events.drain(..).chain(new_events.drain(..));
let mut extra_ev = None;
for event in &mut events_iter {
had_events = true;
match event {
events::MessageSendEvent::UpdateHTLCs { node_id, updates: CommitmentUpdate { update_add_htlcs, update_fail_htlcs, update_fulfill_htlcs, update_fail_malformed_htlcs, update_fee, commitment_signed } } => {
for (idx, dest) in nodes.iter().enumerate() {
if dest.get_our_node_id() == node_id {
for update_add in update_add_htlcs.iter() {
out.locked_write(format!("Delivering update_add_htlc to node {}.\n", idx).as_bytes());
if !$corrupt_forward {
dest.handle_update_add_htlc(&nodes[$node].get_our_node_id(), update_add);
} else {
// Corrupt the update_add_htlc message so that its HMAC
// check will fail and we generate a
// update_fail_malformed_htlc instead of an
// update_fail_htlc as we do when we reject a payment.
let mut msg_ser = update_add.encode();
msg_ser[1000] ^= 0xff;
let new_msg = UpdateAddHTLC::read(&mut Cursor::new(&msg_ser)).unwrap();
dest.handle_update_add_htlc(&nodes[$node].get_our_node_id(), &new_msg);
}
}
for update_fulfill in update_fulfill_htlcs.iter() {
out.locked_write(format!("Delivering update_fulfill_htlc to node {}.\n", idx).as_bytes());
dest.handle_update_fulfill_htlc(&nodes[$node].get_our_node_id(), update_fulfill);
}
for update_fail in update_fail_htlcs.iter() {
out.locked_write(format!("Delivering update_fail_htlc to node {}.\n", idx).as_bytes());
dest.handle_update_fail_htlc(&nodes[$node].get_our_node_id(), update_fail);
}
for update_fail_malformed in update_fail_malformed_htlcs.iter() {
out.locked_write(format!("Delivering update_fail_malformed_htlc to node {}.\n", idx).as_bytes());
dest.handle_update_fail_malformed_htlc(&nodes[$node].get_our_node_id(), update_fail_malformed);
}
if let Some(msg) = update_fee {
out.locked_write(format!("Delivering update_fee to node {}.\n", idx).as_bytes());
dest.handle_update_fee(&nodes[$node].get_our_node_id(), &msg);
}
let processed_change = !update_add_htlcs.is_empty() || !update_fulfill_htlcs.is_empty() ||
!update_fail_htlcs.is_empty() || !update_fail_malformed_htlcs.is_empty();
if $limit_events != ProcessMessages::AllMessages && processed_change {
// If we only want to process some messages, don't deliver the CS until later.
extra_ev = Some(events::MessageSendEvent::UpdateHTLCs { node_id, updates: CommitmentUpdate {
update_add_htlcs: Vec::new(),
update_fail_htlcs: Vec::new(),
update_fulfill_htlcs: Vec::new(),
update_fail_malformed_htlcs: Vec::new(),
update_fee: None,
commitment_signed
} });
break;
}
out.locked_write(format!("Delivering commitment_signed to node {}.\n", idx).as_bytes());
dest.handle_commitment_signed(&nodes[$node].get_our_node_id(), &commitment_signed);
break;
}
}
},
events::MessageSendEvent::SendRevokeAndACK { ref node_id, ref msg } => {
for (idx, dest) in nodes.iter().enumerate() {
if dest.get_our_node_id() == *node_id {
out.locked_write(format!("Delivering revoke_and_ack to node {}.\n", idx).as_bytes());
dest.handle_revoke_and_ack(&nodes[$node].get_our_node_id(), msg);
}
}
},
events::MessageSendEvent::SendChannelReestablish { ref node_id, ref msg } => {
for (idx, dest) in nodes.iter().enumerate() {
if dest.get_our_node_id() == *node_id {
out.locked_write(format!("Delivering channel_reestablish to node {}.\n", idx).as_bytes());
dest.handle_channel_reestablish(&nodes[$node].get_our_node_id(), msg);
}
}
},
events::MessageSendEvent::SendFundingLocked { .. } => {
// Can be generated as a reestablish response
},
events::MessageSendEvent::SendAnnouncementSignatures { .. } => {
// Can be generated as a reestablish response
},
events::MessageSendEvent::SendChannelUpdate { ref msg, .. } => {
// When we reconnect we will resend a channel_update to make sure our
// counterparty has the latest parameters for receiving payments
// through us. We do, however, check that the message does not include
// the "disabled" bit, as we should never ever have a channel which is
// disabled when we send such an update (or it may indicate channel
// force-close which we should detect as an error).
assert_eq!(msg.contents.flags & 2, 0);
},
_ => if out.may_fail.load(atomic::Ordering::Acquire) {
return;
} else {
panic!("Unhandled message event {:?}", event)
},
}
if $limit_events != ProcessMessages::AllMessages {
break;
}
}
if $node == 1 {
push_excess_b_events!(extra_ev.into_iter().chain(events_iter), None);
} else if $node == 0 {
if let Some(ev) = extra_ev { ab_events.push(ev); }
for event in events_iter { ab_events.push(event); }
} else {
if let Some(ev) = extra_ev { cb_events.push(ev); }
for event in events_iter { cb_events.push(event); }
}
had_events
} }
}
macro_rules! drain_msg_events_on_disconnect {
($counterparty_id: expr) => { {
if $counterparty_id == 0 {
for event in nodes[0].get_and_clear_pending_msg_events() {
match event {
events::MessageSendEvent::UpdateHTLCs { .. } => {},
events::MessageSendEvent::SendRevokeAndACK { .. } => {},
events::MessageSendEvent::SendChannelReestablish { .. } => {},
events::MessageSendEvent::SendFundingLocked { .. } => {},
events::MessageSendEvent::SendAnnouncementSignatures { .. } => {},
events::MessageSendEvent::SendChannelUpdate { ref msg, .. } => {
assert_eq!(msg.contents.flags & 2, 0); // The disable bit must never be set!
},
_ => if out.may_fail.load(atomic::Ordering::Acquire) {
return;
} else {
panic!("Unhandled message event")
},
}
}
push_excess_b_events!(nodes[1].get_and_clear_pending_msg_events().drain(..), Some(0));
ab_events.clear();
ba_events.clear();
} else {
for event in nodes[2].get_and_clear_pending_msg_events() {
match event {
events::MessageSendEvent::UpdateHTLCs { .. } => {},
events::MessageSendEvent::SendRevokeAndACK { .. } => {},
events::MessageSendEvent::SendChannelReestablish { .. } => {},
events::MessageSendEvent::SendFundingLocked { .. } => {},
events::MessageSendEvent::SendAnnouncementSignatures { .. } => {},
events::MessageSendEvent::SendChannelUpdate { ref msg, .. } => {
assert_eq!(msg.contents.flags & 2, 0); // The disable bit must never be set!
},
_ => if out.may_fail.load(atomic::Ordering::Acquire) {
return;
} else {
panic!("Unhandled message event")
},
}
}
push_excess_b_events!(nodes[1].get_and_clear_pending_msg_events().drain(..), Some(2));
bc_events.clear();
cb_events.clear();
}
} }
}
macro_rules! process_events {
($node: expr, $fail: expr) => { {
// In case we get 256 payments we may have a hash collision, resulting in the
// second claim/fail call not finding the duplicate-hash HTLC, so we have to
// deduplicate the calls here.
let mut claim_set = HashSet::new();
let mut events = nodes[$node].get_and_clear_pending_events();
// Sort events so that PendingHTLCsForwardable get processed last. This avoids a
// case where we first process a PendingHTLCsForwardable, then claim/fail on a
// PaymentReceived, claiming/failing two HTLCs, but leaving a just-generated
// PaymentReceived event for the second HTLC in our pending_events (and breaking
// our claim_set deduplication).
events.sort_by(|a, b| {
if let events::Event::PaymentReceived { .. } = a {
if let events::Event::PendingHTLCsForwardable { .. } = b {
Ordering::Less
} else { Ordering::Equal }
} else if let events::Event::PendingHTLCsForwardable { .. } = a {
if let events::Event::PaymentReceived { .. } = b {
Ordering::Greater
} else { Ordering::Equal }
} else { Ordering::Equal }
});
let had_events = !events.is_empty();
for event in events.drain(..) {
match event {
events::Event::PaymentReceived { payment_hash, .. } => {
if claim_set.insert(payment_hash.0) {
if $fail {
assert!(nodes[$node].fail_htlc_backwards(&payment_hash));
} else {
nodes[$node].claim_funds(PaymentPreimage(payment_hash.0));
}
}
},
events::Event::PaymentSent { .. } => {},
events::Event::PaymentClaimed { .. } => {},
events::Event::PaymentPathSuccessful { .. } => {},
events::Event::PaymentPathFailed { .. } => {},
events::Event::PaymentForwarded { .. } if $node == 1 => {},
events::Event::PendingHTLCsForwardable { .. } => {
nodes[$node].process_pending_htlc_forwards();
},
_ => if out.may_fail.load(atomic::Ordering::Acquire) {
return;
} else {
panic!("Unhandled event")
},
}
}
had_events
} }
}
let v = get_slice!(1)[0];
out.locked_write(format!("READ A BYTE! HANDLING INPUT {:x}...........\n", v).as_bytes());
match v {
// In general, we keep related message groups close together in binary form, allowing
// bit-twiddling mutations to have similar effects. This is probably overkill, but no
// harm in doing so.
0x00 => *monitor_a.persister.update_ret.lock().unwrap() = Err(ChannelMonitorUpdateErr::TemporaryFailure),
0x01 => *monitor_b.persister.update_ret.lock().unwrap() = Err(ChannelMonitorUpdateErr::TemporaryFailure),
0x02 => *monitor_c.persister.update_ret.lock().unwrap() = Err(ChannelMonitorUpdateErr::TemporaryFailure),
0x04 => *monitor_a.persister.update_ret.lock().unwrap() = Ok(()),
0x05 => *monitor_b.persister.update_ret.lock().unwrap() = Ok(()),
0x06 => *monitor_c.persister.update_ret.lock().unwrap() = Ok(()),
0x08 => {
if let Some((id, _)) = monitor_a.latest_monitors.lock().unwrap().get(&chan_1_funding) {
monitor_a.chain_monitor.force_channel_monitor_updated(chan_1_funding, *id);
nodes[0].process_monitor_events();
}
},
0x09 => {
if let Some((id, _)) = monitor_b.latest_monitors.lock().unwrap().get(&chan_1_funding) {
monitor_b.chain_monitor.force_channel_monitor_updated(chan_1_funding, *id);
nodes[1].process_monitor_events();
}
},
0x0a => {
if let Some((id, _)) = monitor_b.latest_monitors.lock().unwrap().get(&chan_2_funding) {
monitor_b.chain_monitor.force_channel_monitor_updated(chan_2_funding, *id);
nodes[1].process_monitor_events();
}
},
0x0b => {
if let Some((id, _)) = monitor_c.latest_monitors.lock().unwrap().get(&chan_2_funding) {
monitor_c.chain_monitor.force_channel_monitor_updated(chan_2_funding, *id);
nodes[2].process_monitor_events();
}
},
0x0c => {
if !chan_a_disconnected {
nodes[0].peer_disconnected(&nodes[1].get_our_node_id(), false);
nodes[1].peer_disconnected(&nodes[0].get_our_node_id(), false);
chan_a_disconnected = true;
drain_msg_events_on_disconnect!(0);
}
},
0x0d => {
if !chan_b_disconnected {
nodes[1].peer_disconnected(&nodes[2].get_our_node_id(), false);
nodes[2].peer_disconnected(&nodes[1].get_our_node_id(), false);
chan_b_disconnected = true;
drain_msg_events_on_disconnect!(2);
}
},
0x0e => {
if chan_a_disconnected {
nodes[0].peer_connected(&nodes[1].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
nodes[1].peer_connected(&nodes[0].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
chan_a_disconnected = false;
}
},
0x0f => {
if chan_b_disconnected {
nodes[1].peer_connected(&nodes[2].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
nodes[2].peer_connected(&nodes[1].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
chan_b_disconnected = false;
}
},
0x10 => { process_msg_events!(0, true, ProcessMessages::AllMessages); },
0x11 => { process_msg_events!(0, false, ProcessMessages::AllMessages); },
0x12 => { process_msg_events!(0, true, ProcessMessages::OneMessage); },
0x13 => { process_msg_events!(0, false, ProcessMessages::OneMessage); },
0x14 => { process_msg_events!(0, true, ProcessMessages::OnePendingMessage); },
0x15 => { process_msg_events!(0, false, ProcessMessages::OnePendingMessage); },
0x16 => { process_events!(0, true); },
0x17 => { process_events!(0, false); },
0x18 => { process_msg_events!(1, true, ProcessMessages::AllMessages); },
0x19 => { process_msg_events!(1, false, ProcessMessages::AllMessages); },
0x1a => { process_msg_events!(1, true, ProcessMessages::OneMessage); },
0x1b => { process_msg_events!(1, false, ProcessMessages::OneMessage); },
0x1c => { process_msg_events!(1, true, ProcessMessages::OnePendingMessage); },
0x1d => { process_msg_events!(1, false, ProcessMessages::OnePendingMessage); },
0x1e => { process_events!(1, true); },
0x1f => { process_events!(1, false); },
0x20 => { process_msg_events!(2, true, ProcessMessages::AllMessages); },
0x21 => { process_msg_events!(2, false, ProcessMessages::AllMessages); },
0x22 => { process_msg_events!(2, true, ProcessMessages::OneMessage); },
0x23 => { process_msg_events!(2, false, ProcessMessages::OneMessage); },
0x24 => { process_msg_events!(2, true, ProcessMessages::OnePendingMessage); },
0x25 => { process_msg_events!(2, false, ProcessMessages::OnePendingMessage); },
0x26 => { process_events!(2, true); },
0x27 => { process_events!(2, false); },
0x2c => {
if !chan_a_disconnected {
nodes[1].peer_disconnected(&nodes[0].get_our_node_id(), false);
chan_a_disconnected = true;
drain_msg_events_on_disconnect!(0);
}
if monitor_a.should_update_manager.load(atomic::Ordering::Relaxed) {
node_a_ser.0.clear();
nodes[0].write(&mut node_a_ser).unwrap();
}
let (new_node_a, new_monitor_a) = reload_node!(node_a_ser, 0, monitor_a, keys_manager_a, fee_est_a);
nodes[0] = new_node_a;
monitor_a = new_monitor_a;
},
0x2d => {
if !chan_a_disconnected {
nodes[0].peer_disconnected(&nodes[1].get_our_node_id(), false);
chan_a_disconnected = true;
nodes[0].get_and_clear_pending_msg_events();
ab_events.clear();
ba_events.clear();
}
if !chan_b_disconnected {
nodes[2].peer_disconnected(&nodes[1].get_our_node_id(), false);
chan_b_disconnected = true;
nodes[2].get_and_clear_pending_msg_events();
bc_events.clear();
cb_events.clear();
}
let (new_node_b, new_monitor_b) = reload_node!(node_b_ser, 1, monitor_b, keys_manager_b, fee_est_b);
nodes[1] = new_node_b;
monitor_b = new_monitor_b;
},
0x2e => {
if !chan_b_disconnected {
nodes[1].peer_disconnected(&nodes[2].get_our_node_id(), false);
chan_b_disconnected = true;
drain_msg_events_on_disconnect!(2);
}
if monitor_c.should_update_manager.load(atomic::Ordering::Relaxed) {
node_c_ser.0.clear();
nodes[2].write(&mut node_c_ser).unwrap();
}
let (new_node_c, new_monitor_c) = reload_node!(node_c_ser, 2, monitor_c, keys_manager_c, fee_est_c);
nodes[2] = new_node_c;
monitor_c = new_monitor_c;
},
// 1/10th the channel size:
0x30 => { send_payment(&nodes[0], &nodes[1], chan_a, 10_000_000, &mut payment_id); },
0x31 => { send_payment(&nodes[1], &nodes[0], chan_a, 10_000_000, &mut payment_id); },
0x32 => { send_payment(&nodes[1], &nodes[2], chan_b, 10_000_000, &mut payment_id); },
0x33 => { send_payment(&nodes[2], &nodes[1], chan_b, 10_000_000, &mut payment_id); },
0x34 => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 10_000_000, &mut payment_id); },
0x35 => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 10_000_000, &mut payment_id); },
0x38 => { send_payment(&nodes[0], &nodes[1], chan_a, 1_000_000, &mut payment_id); },
0x39 => { send_payment(&nodes[1], &nodes[0], chan_a, 1_000_000, &mut payment_id); },
0x3a => { send_payment(&nodes[1], &nodes[2], chan_b, 1_000_000, &mut payment_id); },
0x3b => { send_payment(&nodes[2], &nodes[1], chan_b, 1_000_000, &mut payment_id); },
0x3c => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 1_000_000, &mut payment_id); },
0x3d => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 1_000_000, &mut payment_id); },
0x40 => { send_payment(&nodes[0], &nodes[1], chan_a, 100_000, &mut payment_id); },
0x41 => { send_payment(&nodes[1], &nodes[0], chan_a, 100_000, &mut payment_id); },
0x42 => { send_payment(&nodes[1], &nodes[2], chan_b, 100_000, &mut payment_id); },
0x43 => { send_payment(&nodes[2], &nodes[1], chan_b, 100_000, &mut payment_id); },
0x44 => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 100_000, &mut payment_id); },
0x45 => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 100_000, &mut payment_id); },
0x48 => { send_payment(&nodes[0], &nodes[1], chan_a, 10_000, &mut payment_id); },
0x49 => { send_payment(&nodes[1], &nodes[0], chan_a, 10_000, &mut payment_id); },
0x4a => { send_payment(&nodes[1], &nodes[2], chan_b, 10_000, &mut payment_id); },
0x4b => { send_payment(&nodes[2], &nodes[1], chan_b, 10_000, &mut payment_id); },
0x4c => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 10_000, &mut payment_id); },
0x4d => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 10_000, &mut payment_id); },
0x50 => { send_payment(&nodes[0], &nodes[1], chan_a, 1_000, &mut payment_id); },
0x51 => { send_payment(&nodes[1], &nodes[0], chan_a, 1_000, &mut payment_id); },
0x52 => { send_payment(&nodes[1], &nodes[2], chan_b, 1_000, &mut payment_id); },
0x53 => { send_payment(&nodes[2], &nodes[1], chan_b, 1_000, &mut payment_id); },
0x54 => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 1_000, &mut payment_id); },
0x55 => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 1_000, &mut payment_id); },
0x58 => { send_payment(&nodes[0], &nodes[1], chan_a, 100, &mut payment_id); },
0x59 => { send_payment(&nodes[1], &nodes[0], chan_a, 100, &mut payment_id); },
0x5a => { send_payment(&nodes[1], &nodes[2], chan_b, 100, &mut payment_id); },
0x5b => { send_payment(&nodes[2], &nodes[1], chan_b, 100, &mut payment_id); },
0x5c => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 100, &mut payment_id); },
0x5d => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 100, &mut payment_id); },
0x60 => { send_payment(&nodes[0], &nodes[1], chan_a, 10, &mut payment_id); },
0x61 => { send_payment(&nodes[1], &nodes[0], chan_a, 10, &mut payment_id); },
0x62 => { send_payment(&nodes[1], &nodes[2], chan_b, 10, &mut payment_id); },
0x63 => { send_payment(&nodes[2], &nodes[1], chan_b, 10, &mut payment_id); },
0x64 => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 10, &mut payment_id); },
0x65 => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 10, &mut payment_id); },
0x68 => { send_payment(&nodes[0], &nodes[1], chan_a, 1, &mut payment_id); },
0x69 => { send_payment(&nodes[1], &nodes[0], chan_a, 1, &mut payment_id); },
0x6a => { send_payment(&nodes[1], &nodes[2], chan_b, 1, &mut payment_id); },
0x6b => { send_payment(&nodes[2], &nodes[1], chan_b, 1, &mut payment_id); },
0x6c => { send_hop_payment(&nodes[0], &nodes[1], chan_a, &nodes[2], chan_b, 1, &mut payment_id); },
0x6d => { send_hop_payment(&nodes[2], &nodes[1], chan_b, &nodes[0], chan_a, 1, &mut payment_id); },
0x80 => {
let max_feerate = last_htlc_clear_fee_a * FEE_SPIKE_BUFFER_FEE_INCREASE_MULTIPLE as u32;
if fee_est_a.ret_val.fetch_add(250, atomic::Ordering::AcqRel) + 250 > max_feerate {
fee_est_a.ret_val.store(max_feerate, atomic::Ordering::Release);
}
nodes[0].maybe_update_chan_fees();
},
0x81 => { fee_est_a.ret_val.store(253, atomic::Ordering::Release); nodes[0].maybe_update_chan_fees(); },
0x84 => {
let max_feerate = last_htlc_clear_fee_b * FEE_SPIKE_BUFFER_FEE_INCREASE_MULTIPLE as u32;
if fee_est_b.ret_val.fetch_add(250, atomic::Ordering::AcqRel) + 250 > max_feerate {
fee_est_b.ret_val.store(max_feerate, atomic::Ordering::Release);
}
nodes[1].maybe_update_chan_fees();
},
0x85 => { fee_est_b.ret_val.store(253, atomic::Ordering::Release); nodes[1].maybe_update_chan_fees(); },
0x88 => {
let max_feerate = last_htlc_clear_fee_c * FEE_SPIKE_BUFFER_FEE_INCREASE_MULTIPLE as u32;
if fee_est_c.ret_val.fetch_add(250, atomic::Ordering::AcqRel) + 250 > max_feerate {
fee_est_c.ret_val.store(max_feerate, atomic::Ordering::Release);
}
nodes[2].maybe_update_chan_fees();
},
0x89 => { fee_est_c.ret_val.store(253, atomic::Ordering::Release); nodes[2].maybe_update_chan_fees(); },
0xff => {
// Test that no channel is in a stuck state where neither party can send funds even
// after we resolve all pending events.
// First make sure there are no pending monitor updates, resetting the error state
// and calling force_channel_monitor_updated for each monitor.
*monitor_a.persister.update_ret.lock().unwrap() = Ok(());
*monitor_b.persister.update_ret.lock().unwrap() = Ok(());
*monitor_c.persister.update_ret.lock().unwrap() = Ok(());
if let Some((id, _)) = monitor_a.latest_monitors.lock().unwrap().get(&chan_1_funding) {
monitor_a.chain_monitor.force_channel_monitor_updated(chan_1_funding, *id);
nodes[0].process_monitor_events();
}
if let Some((id, _)) = monitor_b.latest_monitors.lock().unwrap().get(&chan_1_funding) {
monitor_b.chain_monitor.force_channel_monitor_updated(chan_1_funding, *id);
nodes[1].process_monitor_events();
}
if let Some((id, _)) = monitor_b.latest_monitors.lock().unwrap().get(&chan_2_funding) {
monitor_b.chain_monitor.force_channel_monitor_updated(chan_2_funding, *id);
nodes[1].process_monitor_events();
}
if let Some((id, _)) = monitor_c.latest_monitors.lock().unwrap().get(&chan_2_funding) {
monitor_c.chain_monitor.force_channel_monitor_updated(chan_2_funding, *id);
nodes[2].process_monitor_events();
}
// Next, make sure peers are all connected to each other
if chan_a_disconnected {
nodes[0].peer_connected(&nodes[1].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
nodes[1].peer_connected(&nodes[0].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
chan_a_disconnected = false;
}
if chan_b_disconnected {
nodes[1].peer_connected(&nodes[2].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
nodes[2].peer_connected(&nodes[1].get_our_node_id(), &Init { features: InitFeatures::known(), remote_network_address: None });
chan_b_disconnected = false;
}
for i in 0..std::usize::MAX {
if i == 100 { panic!("It may take may iterations to settle the state, but it should not take forever"); }
// Then, make sure any current forwards make their way to their destination
if process_msg_events!(0, false, ProcessMessages::AllMessages) { continue; }
if process_msg_events!(1, false, ProcessMessages::AllMessages) { continue; }
if process_msg_events!(2, false, ProcessMessages::AllMessages) { continue; }
// ...making sure any pending PendingHTLCsForwardable events are handled and
// payments claimed.
if process_events!(0, false) { continue; }
if process_events!(1, false) { continue; }
if process_events!(2, false) { continue; }
break;
}
// Finally, make sure that at least one end of each channel can make a substantial payment
assert!(
send_payment(&nodes[0], &nodes[1], chan_a, 10_000_000, &mut payment_id) ||
send_payment(&nodes[1], &nodes[0], chan_a, 10_000_000, &mut payment_id));
assert!(
send_payment(&nodes[1], &nodes[2], chan_b, 10_000_000, &mut payment_id) ||
send_payment(&nodes[2], &nodes[1], chan_b, 10_000_000, &mut payment_id));
last_htlc_clear_fee_a = fee_est_a.ret_val.load(atomic::Ordering::Acquire);
last_htlc_clear_fee_b = fee_est_b.ret_val.load(atomic::Ordering::Acquire);
last_htlc_clear_fee_c = fee_est_c.ret_val.load(atomic::Ordering::Acquire);
},
_ => test_return!(),
}
node_a_ser.0.clear();
nodes[0].write(&mut node_a_ser).unwrap();
monitor_a.should_update_manager.store(false, atomic::Ordering::Relaxed);
node_b_ser.0.clear();
nodes[1].write(&mut node_b_ser).unwrap();
monitor_b.should_update_manager.store(false, atomic::Ordering::Relaxed);
node_c_ser.0.clear();
nodes[2].write(&mut node_c_ser).unwrap();
monitor_c.should_update_manager.store(false, atomic::Ordering::Relaxed);
}
}
/// We actually have different behavior based on if a certain log string has been seen, so we have
/// to do a bit more tracking.
#[derive(Clone)]
struct SearchingOutput<O: Output> {
output: O,
may_fail: Arc<atomic::AtomicBool>,
}
impl<O: Output> Output for SearchingOutput<O> {
fn locked_write(&self, data: &[u8]) {
// We hit a design limitation of LN state machine (see CONCURRENT_INBOUND_HTLC_FEE_BUFFER)
if std::str::from_utf8(data).unwrap().contains("Outbound update_fee HTLC buffer overflow - counterparty should force-close this channel") {
self.may_fail.store(true, atomic::Ordering::Release);
}
self.output.locked_write(data)
}
}
impl<O: Output> SearchingOutput<O> {
pub fn new(output: O) -> Self {
Self { output, may_fail: Arc::new(atomic::AtomicBool::new(false)) }
}
}
pub fn chanmon_consistency_test<Out: Output>(data: &[u8], out: Out) {
do_test(data, out);
}
#[no_mangle]
pub extern "C" fn chanmon_consistency_run(data: *const u8, datalen: usize) {
do_test(unsafe { std::slice::from_raw_parts(data, datalen) }, test_logger::DevNull{});
}