// This file is Copyright its original authors, visible in version control // history. // // This file is licensed under the Apache License, Version 2.0 or the MIT license // , at your option. // You may not use this file except in accordance with one or both of these // licenses. //! Further functional tests which test blockchain reorganizations. use chain::channelmonitor::{ANTI_REORG_DELAY, ChannelMonitor}; use chain::{Confirm, Watch}; use ln::channelmanager::{ChannelManager, ChannelManagerReadArgs}; use ln::features::InitFeatures; use ln::msgs::{ChannelMessageHandler, ErrorAction, HTLCFailChannelUpdate}; use util::enforcing_trait_impls::EnforcingSigner; use util::events::{Event, MessageSendEvent, MessageSendEventsProvider}; use util::test_utils; use util::ser::{ReadableArgs, Writeable}; use bitcoin::blockdata::block::{Block, BlockHeader}; use bitcoin::hash_types::BlockHash; use prelude::*; use core::mem; use ln::functional_test_utils::*; fn do_test_onchain_htlc_reorg(local_commitment: bool, claim: bool) { // Our on-chain HTLC-claim learning has a few properties worth testing: // * If an upstream HTLC is claimed with a preimage (both against our own commitment // transaction our counterparty's), we claim it backwards immediately. // * If an upstream HTLC is claimed with a timeout, we delay ANTI_REORG_DELAY before failing // it backwards to ensure our counterparty can't claim with a preimage in a reorg. // // Here we test both properties in any combination based on the two bools passed in as // arguments. // // If local_commitment is set, we first broadcast a local commitment containing an offered HTLC // and an HTLC-Timeout tx, otherwise we broadcast a remote commitment containing a received // HTLC and a local HTLC-Timeout tx spending it. // // We then either allow these transactions to confirm (if !claim) or we wait until one block // before they otherwise would and reorg them out, confirming an HTLC-Success tx instead. let chanmon_cfgs = create_chanmon_cfgs(3); let node_cfgs = create_node_cfgs(3, &chanmon_cfgs); let node_chanmgrs = create_node_chanmgrs(3, &node_cfgs, &[None, None, None]); let nodes = create_network(3, &node_cfgs, &node_chanmgrs); create_announced_chan_between_nodes(&nodes, 0, 1, InitFeatures::known(), InitFeatures::known()); let chan_2 = create_announced_chan_between_nodes(&nodes, 1, 2, InitFeatures::known(), InitFeatures::known()); // Make sure all nodes are at the same starting height connect_blocks(&nodes[0], 2*CHAN_CONFIRM_DEPTH + 1 - nodes[0].best_block_info().1); connect_blocks(&nodes[1], 2*CHAN_CONFIRM_DEPTH + 1 - nodes[1].best_block_info().1); connect_blocks(&nodes[2], 2*CHAN_CONFIRM_DEPTH + 1 - nodes[2].best_block_info().1); let (our_payment_preimage, our_payment_hash, _) = route_payment(&nodes[0], &[&nodes[1], &nodes[2]], 1000000); // Provide preimage to node 2 by claiming payment nodes[2].node.claim_funds(our_payment_preimage); check_added_monitors!(nodes[2], 1); get_htlc_update_msgs!(nodes[2], nodes[1].node.get_our_node_id()); let mut header = BlockHeader { version: 0x2000_0000, prev_blockhash: nodes[2].best_block_hash(), merkle_root: Default::default(), time: 42, bits: 42, nonce: 42 }; let claim_txn = if local_commitment { // Broadcast node 1 commitment txn to broadcast the HTLC-Timeout let node_1_commitment_txn = get_local_commitment_txn!(nodes[1], chan_2.2); assert_eq!(node_1_commitment_txn.len(), 2); // 1 local commitment tx, 1 Outbound HTLC-Timeout assert_eq!(node_1_commitment_txn[0].output.len(), 2); // to-self and Offered HTLC (to-remote/to-node-3 is dust) check_spends!(node_1_commitment_txn[0], chan_2.3); check_spends!(node_1_commitment_txn[1], node_1_commitment_txn[0]); // Give node 2 node 1's transactions and get its response (claiming the HTLC instead). connect_block(&nodes[2], &Block { header, txdata: node_1_commitment_txn.clone() }); check_added_monitors!(nodes[2], 1); check_closed_broadcast!(nodes[2], true); // We should get a BroadcastChannelUpdate (and *only* a BroadcstChannelUpdate) let node_2_commitment_txn = nodes[2].tx_broadcaster.txn_broadcasted.lock().unwrap(); assert_eq!(node_2_commitment_txn.len(), 3); // ChannelMonitor: 1 offered HTLC-Claim, ChannelManger: 1 local commitment tx, 1 Received HTLC-Claim assert_eq!(node_2_commitment_txn[1].output.len(), 2); // to-remote and Received HTLC (to-self is dust) check_spends!(node_2_commitment_txn[1], chan_2.3); check_spends!(node_2_commitment_txn[2], node_2_commitment_txn[1]); check_spends!(node_2_commitment_txn[0], node_1_commitment_txn[0]); // Make sure node 1's height is the same as the !local_commitment case connect_blocks(&nodes[1], 1); // Confirm node 1's commitment txn (and HTLC-Timeout) on node 1 header.prev_blockhash = nodes[1].best_block_hash(); connect_block(&nodes[1], &Block { header, txdata: node_1_commitment_txn.clone() }); // ...but return node 1's commitment tx in case claim is set and we're preparing to reorg vec![node_1_commitment_txn[0].clone(), node_2_commitment_txn[0].clone()] } else { // Broadcast node 2 commitment txn let node_2_commitment_txn = get_local_commitment_txn!(nodes[2], chan_2.2); assert_eq!(node_2_commitment_txn.len(), 2); // 1 local commitment tx, 1 Received HTLC-Claim assert_eq!(node_2_commitment_txn[0].output.len(), 2); // to-remote and Received HTLC (to-self is dust) check_spends!(node_2_commitment_txn[0], chan_2.3); check_spends!(node_2_commitment_txn[1], node_2_commitment_txn[0]); // Give node 1 node 2's commitment transaction and get its response (timing the HTLC out) mine_transaction(&nodes[1], &node_2_commitment_txn[0]); connect_blocks(&nodes[1], TEST_FINAL_CLTV - 1); // Confirm blocks until the HTLC expires let node_1_commitment_txn = nodes[1].tx_broadcaster.txn_broadcasted.lock().unwrap().clone(); assert_eq!(node_1_commitment_txn.len(), 2); // ChannelMonitor: 1 offered HTLC-Timeout, ChannelManger: 1 local commitment tx assert_eq!(node_1_commitment_txn[0].output.len(), 2); // to-local and Offered HTLC (to-remote is dust) check_spends!(node_1_commitment_txn[0], chan_2.3); check_spends!(node_1_commitment_txn[1], node_2_commitment_txn[0]); // Confirm node 2's commitment txn (and node 1's HTLC-Timeout) on node 1 header.prev_blockhash = nodes[1].best_block_hash(); let block = Block { header, txdata: vec![node_2_commitment_txn[0].clone(), node_1_commitment_txn[1].clone()] }; connect_block(&nodes[1], &block); // ...but return node 2's commitment tx (and claim) in case claim is set and we're preparing to reorg node_2_commitment_txn }; check_added_monitors!(nodes[1], 1); check_closed_broadcast!(nodes[1], true); // We should get a BroadcastChannelUpdate (and *only* a BroadcstChannelUpdate) // Connect ANTI_REORG_DELAY - 2 blocks, giving us a confirmation count of ANTI_REORG_DELAY - 1. connect_blocks(&nodes[1], ANTI_REORG_DELAY - 2); check_added_monitors!(nodes[1], 0); assert_eq!(nodes[1].node.get_and_clear_pending_events().len(), 0); if claim { disconnect_blocks(&nodes[1], ANTI_REORG_DELAY - 2); let block = Block { header: BlockHeader { version: 0x20000000, prev_blockhash: nodes[1].best_block_hash(), merkle_root: Default::default(), time: 42, bits: 42, nonce: 42 }, txdata: claim_txn, }; connect_block(&nodes[1], &block); // ChannelManager only polls chain::Watch::release_pending_monitor_events when we // probe it for events, so we probe non-message events here (which should still end up empty): assert_eq!(nodes[1].node.get_and_clear_pending_events().len(), 0); } else { // Confirm the timeout tx and check that we fail the HTLC backwards let block = Block { header: BlockHeader { version: 0x20000000, prev_blockhash: nodes[1].best_block_hash(), merkle_root: Default::default(), time: 42, bits: 42, nonce: 42 }, txdata: vec![], }; connect_block(&nodes[1], &block); expect_pending_htlcs_forwardable!(nodes[1]); } check_added_monitors!(nodes[1], 1); // Which should result in an immediate claim/fail of the HTLC: let htlc_updates = get_htlc_update_msgs!(nodes[1], nodes[0].node.get_our_node_id()); if claim { assert_eq!(htlc_updates.update_fulfill_htlcs.len(), 1); nodes[0].node.handle_update_fulfill_htlc(&nodes[1].node.get_our_node_id(), &htlc_updates.update_fulfill_htlcs[0]); } else { assert_eq!(htlc_updates.update_fail_htlcs.len(), 1); nodes[0].node.handle_update_fail_htlc(&nodes[1].node.get_our_node_id(), &htlc_updates.update_fail_htlcs[0]); } commitment_signed_dance!(nodes[0], nodes[1], htlc_updates.commitment_signed, false, true); if claim { expect_payment_sent!(nodes[0], our_payment_preimage); } else { let events = nodes[0].node.get_and_clear_pending_msg_events(); assert_eq!(events.len(), 1); if let MessageSendEvent::PaymentFailureNetworkUpdate { update: HTLCFailChannelUpdate::ChannelClosed { ref is_permanent, .. } } = events[0] { assert!(is_permanent); } else { panic!("Unexpected event!"); } expect_payment_failed!(nodes[0], our_payment_hash, false); } } #[test] fn test_onchain_htlc_claim_reorg_local_commitment() { do_test_onchain_htlc_reorg(true, true); } #[test] fn test_onchain_htlc_timeout_delay_local_commitment() { do_test_onchain_htlc_reorg(true, false); } #[test] fn test_onchain_htlc_claim_reorg_remote_commitment() { do_test_onchain_htlc_reorg(false, true); } #[test] fn test_onchain_htlc_timeout_delay_remote_commitment() { do_test_onchain_htlc_reorg(false, false); } fn do_test_unconf_chan(reload_node: bool, reorg_after_reload: bool, use_funding_unconfirmed: bool, connect_style: ConnectStyle) { // After creating a chan between nodes, we disconnect all blocks previously seen to force a // channel close on nodes[0] side. We also use this to provide very basic testing of logic // around freeing background events which store monitor updates during block_[dis]connected. let chanmon_cfgs = create_chanmon_cfgs(2); let node_cfgs = create_node_cfgs(2, &chanmon_cfgs); let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]); let persister: test_utils::TestPersister; let new_chain_monitor: test_utils::TestChainMonitor; let nodes_0_deserialized: ChannelManager; let mut nodes = create_network(2, &node_cfgs, &node_chanmgrs); *nodes[0].connect_style.borrow_mut() = connect_style; let chan = create_announced_chan_between_nodes(&nodes, 0, 1, InitFeatures::known(), InitFeatures::known()); let channel_state = nodes[0].node.channel_state.lock().unwrap(); assert_eq!(channel_state.by_id.len(), 1); assert_eq!(channel_state.short_to_id.len(), 1); mem::drop(channel_state); if !reorg_after_reload { if use_funding_unconfirmed { let relevant_txids = nodes[0].node.get_relevant_txids(); assert_eq!(&relevant_txids[..], &[chan.3.txid()]); nodes[0].node.transaction_unconfirmed(&relevant_txids[0]); } else { disconnect_all_blocks(&nodes[0]); } if connect_style == ConnectStyle::FullBlockViaListen && !use_funding_unconfirmed { handle_announce_close_broadcast_events(&nodes, 0, 1, true, "Funding transaction was un-confirmed. Locked at 6 confs, now have 2 confs."); } else { handle_announce_close_broadcast_events(&nodes, 0, 1, true, "Funding transaction was un-confirmed. Locked at 6 confs, now have 0 confs."); } check_added_monitors!(nodes[1], 1); { let channel_state = nodes[0].node.channel_state.lock().unwrap(); assert_eq!(channel_state.by_id.len(), 0); assert_eq!(channel_state.short_to_id.len(), 0); } } if reload_node { // Since we currently have a background event pending, it's good to test that we survive a // serialization roundtrip. Further, this tests the somewhat awkward edge-case of dropping // the Channel object from the ChannelManager, but still having a monitor event pending for // it when we go to deserialize, and then use the ChannelManager. let nodes_0_serialized = nodes[0].node.encode(); let mut chan_0_monitor_serialized = test_utils::TestVecWriter(Vec::new()); nodes[0].chain_monitor.chain_monitor.monitors.read().unwrap().iter().next().unwrap().1.write(&mut chan_0_monitor_serialized).unwrap(); persister = test_utils::TestPersister::new(); let keys_manager = &chanmon_cfgs[0].keys_manager; new_chain_monitor = test_utils::TestChainMonitor::new(Some(nodes[0].chain_source), nodes[0].tx_broadcaster.clone(), nodes[0].logger, node_cfgs[0].fee_estimator, &persister, keys_manager); nodes[0].chain_monitor = &new_chain_monitor; let mut chan_0_monitor_read = &chan_0_monitor_serialized.0[..]; let (_, mut chan_0_monitor) = <(BlockHash, ChannelMonitor)>::read( &mut chan_0_monitor_read, keys_manager).unwrap(); assert!(chan_0_monitor_read.is_empty()); let mut nodes_0_read = &nodes_0_serialized[..]; nodes_0_deserialized = { let mut channel_monitors = HashMap::new(); channel_monitors.insert(chan_0_monitor.get_funding_txo().0, &mut chan_0_monitor); <(BlockHash, ChannelManager)>::read( &mut nodes_0_read, ChannelManagerReadArgs { default_config: *nodes[0].node.get_current_default_configuration(), keys_manager, fee_estimator: node_cfgs[0].fee_estimator, chain_monitor: nodes[0].chain_monitor, tx_broadcaster: nodes[0].tx_broadcaster.clone(), logger: nodes[0].logger, channel_monitors, }).unwrap().1 }; nodes[0].node = &nodes_0_deserialized; assert!(nodes_0_read.is_empty()); if !reorg_after_reload { // If the channel is already closed when we reload the node, we'll broadcast a closing // transaction via the ChannelMonitor which is missing a corresponding channel. assert_eq!(nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().len(), 1); nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().clear(); } nodes[0].chain_monitor.watch_channel(chan_0_monitor.get_funding_txo().0.clone(), chan_0_monitor).unwrap(); check_added_monitors!(nodes[0], 1); } if reorg_after_reload { if use_funding_unconfirmed { let relevant_txids = nodes[0].node.get_relevant_txids(); assert_eq!(&relevant_txids[..], &[chan.3.txid()]); nodes[0].node.transaction_unconfirmed(&relevant_txids[0]); } else { disconnect_all_blocks(&nodes[0]); } if connect_style == ConnectStyle::FullBlockViaListen && !use_funding_unconfirmed { handle_announce_close_broadcast_events(&nodes, 0, 1, true, "Funding transaction was un-confirmed. Locked at 6 confs, now have 2 confs."); } else { handle_announce_close_broadcast_events(&nodes, 0, 1, true, "Funding transaction was un-confirmed. Locked at 6 confs, now have 0 confs."); } check_added_monitors!(nodes[1], 1); { let channel_state = nodes[0].node.channel_state.lock().unwrap(); assert_eq!(channel_state.by_id.len(), 0); assert_eq!(channel_state.short_to_id.len(), 0); } } // With expect_channel_force_closed set the TestChainMonitor will enforce that the next update // is a ChannelForcClosed on the right channel with should_broadcast set. *nodes[0].chain_monitor.expect_channel_force_closed.lock().unwrap() = Some((chan.2, true)); nodes[0].node.test_process_background_events(); // Required to free the pending background monitor update check_added_monitors!(nodes[0], 1); assert_eq!(nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().len(), 1); nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().clear(); // Now check that we can create a new channel create_announced_chan_between_nodes(&nodes, 0, 1, InitFeatures::known(), InitFeatures::known()); send_payment(&nodes[0], &[&nodes[1]], 8000000); } #[test] fn test_unconf_chan() { do_test_unconf_chan(true, true, false, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(false, true, false, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(true, false, false, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(false, false, false, ConnectStyle::BestBlockFirstSkippingBlocks); } #[test] fn test_unconf_chan_via_listen() { do_test_unconf_chan(true, true, false, ConnectStyle::FullBlockViaListen); do_test_unconf_chan(false, true, false, ConnectStyle::FullBlockViaListen); do_test_unconf_chan(true, false, false, ConnectStyle::FullBlockViaListen); do_test_unconf_chan(false, false, false, ConnectStyle::FullBlockViaListen); } #[test] fn test_unconf_chan_via_funding_unconfirmed() { do_test_unconf_chan(true, true, true, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(false, true, true, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(true, false, true, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(false, false, true, ConnectStyle::BestBlockFirstSkippingBlocks); do_test_unconf_chan(true, true, true, ConnectStyle::FullBlockViaListen); do_test_unconf_chan(false, true, true, ConnectStyle::FullBlockViaListen); do_test_unconf_chan(true, false, true, ConnectStyle::FullBlockViaListen); do_test_unconf_chan(false, false, true, ConnectStyle::FullBlockViaListen); } #[test] fn test_set_outpoints_partial_claiming() { // - remote party claim tx, new bump tx // - disconnect remote claiming tx, new bump // - disconnect tx, see no tx anymore let chanmon_cfgs = create_chanmon_cfgs(2); let node_cfgs = create_node_cfgs(2, &chanmon_cfgs); let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]); let nodes = create_network(2, &node_cfgs, &node_chanmgrs); let chan = create_announced_chan_between_nodes_with_value(&nodes, 0, 1, 1000000, 59000000, InitFeatures::known(), InitFeatures::known()); let payment_preimage_1 = route_payment(&nodes[1], &vec!(&nodes[0])[..], 3_000_000).0; let payment_preimage_2 = route_payment(&nodes[1], &vec!(&nodes[0])[..], 3_000_000).0; // Remote commitment txn with 4 outputs: to_local, to_remote, 2 outgoing HTLC let remote_txn = get_local_commitment_txn!(nodes[1], chan.2); assert_eq!(remote_txn.len(), 3); assert_eq!(remote_txn[0].output.len(), 4); assert_eq!(remote_txn[0].input.len(), 1); assert_eq!(remote_txn[0].input[0].previous_output.txid, chan.3.txid()); check_spends!(remote_txn[1], remote_txn[0]); check_spends!(remote_txn[2], remote_txn[0]); // Connect blocks on node A to advance height towards TEST_FINAL_CLTV // Provide node A with both preimage nodes[0].node.claim_funds(payment_preimage_1); nodes[0].node.claim_funds(payment_preimage_2); check_added_monitors!(nodes[0], 2); nodes[0].node.get_and_clear_pending_events(); nodes[0].node.get_and_clear_pending_msg_events(); // Connect blocks on node A commitment transaction mine_transaction(&nodes[0], &remote_txn[0]); check_closed_broadcast!(nodes[0], true); check_added_monitors!(nodes[0], 1); // Verify node A broadcast tx claiming both HTLCs { let mut node_txn = nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap(); // ChannelMonitor: claim tx, ChannelManager: local commitment tx + HTLC-Success*2 assert_eq!(node_txn.len(), 4); check_spends!(node_txn[0], remote_txn[0]); check_spends!(node_txn[1], chan.3); check_spends!(node_txn[2], node_txn[1]); check_spends!(node_txn[3], node_txn[1]); assert_eq!(node_txn[0].input.len(), 2); node_txn.clear(); } // Connect blocks on node B connect_blocks(&nodes[1], 135); check_closed_broadcast!(nodes[1], true); check_added_monitors!(nodes[1], 1); // Verify node B broadcast 2 HTLC-timeout txn let partial_claim_tx = { let node_txn = nodes[1].tx_broadcaster.txn_broadcasted.lock().unwrap(); assert_eq!(node_txn.len(), 3); check_spends!(node_txn[1], node_txn[0]); check_spends!(node_txn[2], node_txn[0]); assert_eq!(node_txn[1].input.len(), 1); assert_eq!(node_txn[2].input.len(), 1); node_txn[1].clone() }; // Broadcast partial claim on node A, should regenerate a claiming tx with HTLC dropped mine_transaction(&nodes[0], &partial_claim_tx); { let mut node_txn = nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap(); assert_eq!(node_txn.len(), 1); check_spends!(node_txn[0], remote_txn[0]); assert_eq!(node_txn[0].input.len(), 1); //dropped HTLC node_txn.clear(); } nodes[0].node.get_and_clear_pending_msg_events(); // Disconnect last block on node A, should regenerate a claiming tx with HTLC dropped disconnect_blocks(&nodes[0], 1); { let mut node_txn = nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap(); assert_eq!(node_txn.len(), 1); check_spends!(node_txn[0], remote_txn[0]); assert_eq!(node_txn[0].input.len(), 2); //resurrected HTLC node_txn.clear(); } //// Disconnect one more block and then reconnect multiple no transaction should be generated disconnect_blocks(&nodes[0], 1); connect_blocks(&nodes[0], 15); { let mut node_txn = nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap(); assert_eq!(node_txn.len(), 0); node_txn.clear(); } }