//! The logic to build claims and bump in-flight transactions until confirmations. //! //! OnchainTxHandler objetcs are fully-part of ChannelMonitor and encapsulates all //! building, tracking, bumping and notifications functions. use bitcoin::blockdata::transaction::{Transaction, TxIn, TxOut, SigHashType}; use bitcoin::blockdata::transaction::OutPoint as BitcoinOutPoint; use bitcoin::blockdata::script::Script; use bitcoin::hash_types::Txid; use bitcoin::secp256k1::{Secp256k1, Signature}; use bitcoin::secp256k1; use bitcoin::secp256k1::key::PublicKey; use ln::msgs::DecodeError; use ln::channelmonitor::{ANTI_REORG_DELAY, CLTV_SHARED_CLAIM_BUFFER, InputMaterial, ClaimRequest}; use ln::channelmanager::PaymentPreimage; use ln::chan_utils; use ln::chan_utils::{TxCreationKeys, LocalCommitmentTransaction, HTLCOutputInCommitment}; use chain::chaininterface::{FeeEstimator, BroadcasterInterface, ConfirmationTarget, MIN_RELAY_FEE_SAT_PER_1000_WEIGHT}; use chain::keysinterface::ChannelKeys; use util::logger::Logger; use util::ser::{Readable, Writer, Writeable}; use util::byte_utils; use std::collections::{HashMap, hash_map}; use std::cmp; use std::ops::Deref; const MAX_ALLOC_SIZE: usize = 64*1024; /// Upon discovering of some classes of onchain tx by ChannelMonitor, we may have to take actions on it /// once they mature to enough confirmations (ANTI_REORG_DELAY) #[derive(Clone, PartialEq)] enum OnchainEvent { /// Outpoint under claim process by our own tx, once this one get enough confirmations, we remove it from /// bump-txn candidate buffer. Claim { claim_request: Txid, }, /// Claim tx aggregate multiple claimable outpoints. One of the outpoint may be claimed by a remote party tx. /// In this case, we need to drop the outpoint and regenerate a new claim tx. By safety, we keep tracking /// the outpoint to be sure to resurect it back to the claim tx if reorgs happen. ContentiousOutpoint { outpoint: BitcoinOutPoint, input_material: InputMaterial, } } /// Cache remote basepoint to compute any transaction on /// remote outputs, either justice or preimage/timeout transactions. struct RemoteTxCache { remote_delayed_payment_base_key: PublicKey, remote_htlc_base_key: PublicKey, per_htlc: HashMap> } /// Higher-level cache structure needed to re-generate bumped claim txn if needed #[derive(Clone, PartialEq)] pub struct ClaimTxBumpMaterial { // At every block tick, used to check if pending claiming tx is taking too // much time for confirmation and we need to bump it. height_timer: Option, // Tracked in case of reorg to wipe out now-superflous bump material feerate_previous: u64, // Soonest timelocks among set of outpoints claimed, used to compute // a priority of not feerate soonest_timelock: u32, // Cache of script, pubkey, sig or key to solve claimable outputs scriptpubkey. per_input_material: HashMap, } impl Writeable for ClaimTxBumpMaterial { fn write(&self, writer: &mut W) -> Result<(), ::std::io::Error> { self.height_timer.write(writer)?; writer.write_all(&byte_utils::be64_to_array(self.feerate_previous))?; writer.write_all(&byte_utils::be32_to_array(self.soonest_timelock))?; writer.write_all(&byte_utils::be64_to_array(self.per_input_material.len() as u64))?; for (outp, tx_material) in self.per_input_material.iter() { outp.write(writer)?; tx_material.write(writer)?; } Ok(()) } } impl Readable for ClaimTxBumpMaterial { fn read(reader: &mut R) -> Result { let height_timer = Readable::read(reader)?; let feerate_previous = Readable::read(reader)?; let soonest_timelock = Readable::read(reader)?; let per_input_material_len: u64 = Readable::read(reader)?; let mut per_input_material = HashMap::with_capacity(cmp::min(per_input_material_len as usize, MAX_ALLOC_SIZE / 128)); for _ in 0 ..per_input_material_len { let outpoint = Readable::read(reader)?; let input_material = Readable::read(reader)?; per_input_material.insert(outpoint, input_material); } Ok(Self { height_timer, feerate_previous, soonest_timelock, per_input_material }) } } #[derive(PartialEq, Clone, Copy)] pub(crate) enum InputDescriptors { RevokedOfferedHTLC, RevokedReceivedHTLC, OfferedHTLC, ReceivedHTLC, RevokedOutput, // either a revoked to_local output on commitment tx, a revoked HTLC-Timeout output or a revoked HTLC-Success output } impl Writeable for InputDescriptors { fn write(&self, writer: &mut W) -> Result<(), ::std::io::Error> { match self { &InputDescriptors::RevokedOfferedHTLC => { writer.write_all(&[0; 1])?; }, &InputDescriptors::RevokedReceivedHTLC => { writer.write_all(&[1; 1])?; }, &InputDescriptors::OfferedHTLC => { writer.write_all(&[2; 1])?; }, &InputDescriptors::ReceivedHTLC => { writer.write_all(&[3; 1])?; } &InputDescriptors::RevokedOutput => { writer.write_all(&[4; 1])?; } } Ok(()) } } impl Readable for InputDescriptors { fn read(reader: &mut R) -> Result { let input_descriptor = match ::read(reader)? { 0 => { InputDescriptors::RevokedOfferedHTLC }, 1 => { InputDescriptors::RevokedReceivedHTLC }, 2 => { InputDescriptors::OfferedHTLC }, 3 => { InputDescriptors::ReceivedHTLC }, 4 => { InputDescriptors::RevokedOutput } _ => return Err(DecodeError::InvalidValue), }; Ok(input_descriptor) } } macro_rules! subtract_high_prio_fee { ($logger: ident, $fee_estimator: expr, $value: expr, $predicted_weight: expr, $used_feerate: expr) => { { $used_feerate = $fee_estimator.get_est_sat_per_1000_weight(ConfirmationTarget::HighPriority); let mut fee = $used_feerate * ($predicted_weight as u64) / 1000; if $value <= fee { $used_feerate = $fee_estimator.get_est_sat_per_1000_weight(ConfirmationTarget::Normal); fee = $used_feerate * ($predicted_weight as u64) / 1000; if $value <= fee { $used_feerate = $fee_estimator.get_est_sat_per_1000_weight(ConfirmationTarget::Background); fee = $used_feerate * ($predicted_weight as u64) / 1000; if $value <= fee { log_error!($logger, "Failed to generate an on-chain punishment tx as even low priority fee ({} sat) was more than the entire claim balance ({} sat)", fee, $value); false } else { log_warn!($logger, "Used low priority fee for on-chain punishment tx as high priority fee was more than the entire claim balance ({} sat)", $value); $value -= fee; true } } else { log_warn!($logger, "Used medium priority fee for on-chain punishment tx as high priority fee was more than the entire claim balance ({} sat)", $value); $value -= fee; true } } else { $value -= fee; true } } } } impl Readable for Option>> { fn read(reader: &mut R) -> Result { match Readable::read(reader)? { 0u8 => Ok(None), 1u8 => { let vlen: u64 = Readable::read(reader)?; let mut ret = Vec::with_capacity(cmp::min(vlen as usize, MAX_ALLOC_SIZE / ::std::mem::size_of::>())); for _ in 0..vlen { ret.push(match Readable::read(reader)? { 0u8 => None, 1u8 => Some((::read(reader)? as usize, Readable::read(reader)?)), _ => return Err(DecodeError::InvalidValue) }); } Ok(Some(ret)) }, _ => Err(DecodeError::InvalidValue), } } } impl Writeable for Option>> { fn write(&self, writer: &mut W) -> Result<(), ::std::io::Error> { match self { &Some(ref vec) => { 1u8.write(writer)?; (vec.len() as u64).write(writer)?; for opt in vec.iter() { match opt { &Some((ref idx, ref sig)) => { 1u8.write(writer)?; (*idx as u64).write(writer)?; sig.write(writer)?; }, &None => 0u8.write(writer)?, } } }, &None => 0u8.write(writer)?, } Ok(()) } } /// OnchainTxHandler receives claiming requests, aggregates them if it's sound, broadcast and /// do RBF bumping if possible. pub struct OnchainTxHandler { destination_script: Script, local_commitment: Option, // local_htlc_sigs and prev_local_htlc_sigs are in the order as they appear in the commitment // transaction outputs (hence the Option<>s inside the Vec). The first usize is the index in // the set of HTLCs in the LocalCommitmentTransaction (including those which do not appear in // the commitment transaction). local_htlc_sigs: Option>>, prev_local_commitment: Option, prev_local_htlc_sigs: Option>>, local_csv: u16, remote_tx_cache: RemoteTxCache, remote_csv: u16, key_storage: ChanSigner, // Used to track claiming requests. If claim tx doesn't confirm before height timer expiration we need to bump // it (RBF or CPFP). If an input has been part of an aggregate tx at first claim try, we need to keep it within // another bumped aggregate tx to comply with RBF rules. We may have multiple claiming txn in the flight for the // same set of outpoints. One of the outpoints may be spent by a transaction not issued by us. That's why at // block connection we scan all inputs and if any of them is among a set of a claiming request we test for set // equality between spending transaction and claim request. If true, it means transaction was one our claiming one // after a security delay of 6 blocks we remove pending claim request. If false, it means transaction wasn't and // we need to regenerate new claim request with reduced set of still-claimable outpoints. // Key is identifier of the pending claim request, i.e the txid of the initial claiming transaction generated by // us and is immutable until all outpoint of the claimable set are post-anti-reorg-delay solved. // Entry is cache of elements need to generate a bumped claiming transaction (see ClaimTxBumpMaterial) #[cfg(test)] // Used in functional_test to verify sanitization pub pending_claim_requests: HashMap, #[cfg(not(test))] pending_claim_requests: HashMap, // Used to link outpoints claimed in a connected block to a pending claim request. // Key is outpoint than monitor parsing has detected we have keys/scripts to claim // Value is (pending claim request identifier, confirmation_block), identifier // is txid of the initial claiming transaction and is immutable until outpoint is // post-anti-reorg-delay solved, confirmaiton_block is used to erase entry if // block with output gets disconnected. #[cfg(test)] // Used in functional_test to verify sanitization pub claimable_outpoints: HashMap, #[cfg(not(test))] claimable_outpoints: HashMap, onchain_events_waiting_threshold_conf: HashMap>, secp_ctx: Secp256k1, } impl OnchainTxHandler { pub(crate) fn write(&self, writer: &mut W) -> Result<(), ::std::io::Error> { self.destination_script.write(writer)?; self.local_commitment.write(writer)?; self.local_htlc_sigs.write(writer)?; self.prev_local_commitment.write(writer)?; self.prev_local_htlc_sigs.write(writer)?; self.local_csv.write(writer)?; self.remote_tx_cache.remote_delayed_payment_base_key.write(writer)?; self.remote_tx_cache.remote_htlc_base_key.write(writer)?; writer.write_all(&byte_utils::be64_to_array(self.remote_tx_cache.per_htlc.len() as u64))?; for (ref txid, ref htlcs) in self.remote_tx_cache.per_htlc.iter() { writer.write_all(&txid[..])?; writer.write_all(&byte_utils::be64_to_array(htlcs.len() as u64))?; for &ref htlc in htlcs.iter() { htlc.write(writer)?; } } self.remote_csv.write(writer)?; self.key_storage.write(writer)?; writer.write_all(&byte_utils::be64_to_array(self.pending_claim_requests.len() as u64))?; for (ref ancestor_claim_txid, claim_tx_data) in self.pending_claim_requests.iter() { ancestor_claim_txid.write(writer)?; claim_tx_data.write(writer)?; } writer.write_all(&byte_utils::be64_to_array(self.claimable_outpoints.len() as u64))?; for (ref outp, ref claim_and_height) in self.claimable_outpoints.iter() { outp.write(writer)?; claim_and_height.0.write(writer)?; claim_and_height.1.write(writer)?; } writer.write_all(&byte_utils::be64_to_array(self.onchain_events_waiting_threshold_conf.len() as u64))?; for (ref target, ref events) in self.onchain_events_waiting_threshold_conf.iter() { writer.write_all(&byte_utils::be32_to_array(**target))?; writer.write_all(&byte_utils::be64_to_array(events.len() as u64))?; for ev in events.iter() { match *ev { OnchainEvent::Claim { ref claim_request } => { writer.write_all(&[0; 1])?; claim_request.write(writer)?; }, OnchainEvent::ContentiousOutpoint { ref outpoint, ref input_material } => { writer.write_all(&[1; 1])?; outpoint.write(writer)?; input_material.write(writer)?; } } } } Ok(()) } } impl Readable for OnchainTxHandler { fn read(reader: &mut R) -> Result { let destination_script = Readable::read(reader)?; let local_commitment = Readable::read(reader)?; let local_htlc_sigs = Readable::read(reader)?; let prev_local_commitment = Readable::read(reader)?; let prev_local_htlc_sigs = Readable::read(reader)?; let local_csv = Readable::read(reader)?; let remote_tx_cache = { let remote_delayed_payment_base_key = Readable::read(reader)?; let remote_htlc_base_key = Readable::read(reader)?; let per_htlc_len: u64 = Readable::read(reader)?; let mut per_htlc = HashMap::with_capacity(cmp::min(per_htlc_len as usize, MAX_ALLOC_SIZE / 64)); for _ in 0..per_htlc_len { let txid: Txid = Readable::read(reader)?; let htlcs_count: u64 = Readable::read(reader)?; let mut htlcs = Vec::with_capacity(cmp::min(htlcs_count as usize, MAX_ALLOC_SIZE / 32)); for _ in 0..htlcs_count { let htlc = Readable::read(reader)?; htlcs.push(htlc); } if let Some(_) = per_htlc.insert(txid, htlcs) { return Err(DecodeError::InvalidValue); } } RemoteTxCache { remote_delayed_payment_base_key, remote_htlc_base_key, per_htlc, } }; let remote_csv = Readable::read(reader)?; let key_storage = Readable::read(reader)?; let pending_claim_requests_len: u64 = Readable::read(reader)?; let mut pending_claim_requests = HashMap::with_capacity(cmp::min(pending_claim_requests_len as usize, MAX_ALLOC_SIZE / 128)); for _ in 0..pending_claim_requests_len { pending_claim_requests.insert(Readable::read(reader)?, Readable::read(reader)?); } let claimable_outpoints_len: u64 = Readable::read(reader)?; let mut claimable_outpoints = HashMap::with_capacity(cmp::min(pending_claim_requests_len as usize, MAX_ALLOC_SIZE / 128)); for _ in 0..claimable_outpoints_len { let outpoint = Readable::read(reader)?; let ancestor_claim_txid = Readable::read(reader)?; let height = Readable::read(reader)?; claimable_outpoints.insert(outpoint, (ancestor_claim_txid, height)); } let waiting_threshold_conf_len: u64 = Readable::read(reader)?; let mut onchain_events_waiting_threshold_conf = HashMap::with_capacity(cmp::min(waiting_threshold_conf_len as usize, MAX_ALLOC_SIZE / 128)); for _ in 0..waiting_threshold_conf_len { let height_target = Readable::read(reader)?; let events_len: u64 = Readable::read(reader)?; let mut events = Vec::with_capacity(cmp::min(events_len as usize, MAX_ALLOC_SIZE / 128)); for _ in 0..events_len { let ev = match ::read(reader)? { 0 => { let claim_request = Readable::read(reader)?; OnchainEvent::Claim { claim_request } }, 1 => { let outpoint = Readable::read(reader)?; let input_material = Readable::read(reader)?; OnchainEvent::ContentiousOutpoint { outpoint, input_material } } _ => return Err(DecodeError::InvalidValue), }; events.push(ev); } onchain_events_waiting_threshold_conf.insert(height_target, events); } Ok(OnchainTxHandler { destination_script, local_commitment, local_htlc_sigs, prev_local_commitment, prev_local_htlc_sigs, local_csv, remote_tx_cache, remote_csv, key_storage, claimable_outpoints, pending_claim_requests, onchain_events_waiting_threshold_conf, secp_ctx: Secp256k1::new(), }) } } impl OnchainTxHandler { pub(super) fn new(destination_script: Script, keys: ChanSigner, local_csv: u16, remote_delayed_payment_base_key: PublicKey, remote_htlc_base_key: PublicKey, remote_csv: u16) -> Self { let key_storage = keys; let remote_tx_cache = RemoteTxCache { remote_delayed_payment_base_key, remote_htlc_base_key, per_htlc: HashMap::new(), }; OnchainTxHandler { destination_script, local_commitment: None, local_htlc_sigs: None, prev_local_commitment: None, prev_local_htlc_sigs: None, local_csv, remote_tx_cache, remote_csv, key_storage, pending_claim_requests: HashMap::new(), claimable_outpoints: HashMap::new(), onchain_events_waiting_threshold_conf: HashMap::new(), secp_ctx: Secp256k1::new(), } } pub(super) fn get_witnesses_weight(inputs: &[InputDescriptors]) -> usize { let mut tx_weight = 2; // count segwit flags for inp in inputs { // We use expected weight (and not actual) as signatures and time lock delays may vary tx_weight += match inp { // number_of_witness_elements + sig_length + revocation_sig + pubkey_length + revocationpubkey + witness_script_length + witness_script &InputDescriptors::RevokedOfferedHTLC => { 1 + 1 + 73 + 1 + 33 + 1 + 133 }, // number_of_witness_elements + sig_length + revocation_sig + pubkey_length + revocationpubkey + witness_script_length + witness_script &InputDescriptors::RevokedReceivedHTLC => { 1 + 1 + 73 + 1 + 33 + 1 + 139 }, // number_of_witness_elements + sig_length + remotehtlc_sig + preimage_length + preimage + witness_script_length + witness_script &InputDescriptors::OfferedHTLC => { 1 + 1 + 73 + 1 + 32 + 1 + 133 }, // number_of_witness_elements + sig_length + revocation_sig + pubkey_length + revocationpubkey + witness_script_length + witness_script &InputDescriptors::ReceivedHTLC => { 1 + 1 + 73 + 1 + 1 + 1 + 139 }, // number_of_witness_elements + sig_length + revocation_sig + true_length + op_true + witness_script_length + witness_script &InputDescriptors::RevokedOutput => { 1 + 1 + 73 + 1 + 1 + 1 + 77 }, }; } tx_weight } /// In LN, output claimed are time-sensitive, which means we have to spend them before reaching some timelock expiration. At in-channel /// output detection, we generate a first version of a claim tx and associate to it a height timer. A height timer is an absolute block /// height than once reached we should generate a new bumped "version" of the claim tx to be sure than we safely claim outputs before /// than our counterparty can do it too. If timelock expires soon, height timer is going to be scale down in consequence to increase /// frequency of the bump and so increase our bets of success. fn get_height_timer(current_height: u32, timelock_expiration: u32) -> u32 { if timelock_expiration <= current_height + 3 { return current_height + 1 } else if timelock_expiration - current_height <= 15 { return current_height + 3 } current_height + 15 } /// Lightning security model (i.e being able to redeem/timeout HTLC or penalize coutnerparty onchain) lays on the assumption of claim transactions getting confirmed before timelock expiration /// (CSV or CLTV following cases). In case of high-fee spikes, claim tx may stuck in the mempool, so you need to bump its feerate quickly using Replace-By-Fee or Child-Pay-For-Parent. fn generate_claim_tx(&mut self, height: u32, cached_claim_datas: &ClaimTxBumpMaterial, fee_estimator: F, logger: L) -> Option<(Option, u64, Transaction)> where F::Target: FeeEstimator, L::Target: Logger, { if cached_claim_datas.per_input_material.len() == 0 { return None } // But don't prune pending claiming request yet, we may have to resurrect HTLCs let mut inputs = Vec::new(); for outp in cached_claim_datas.per_input_material.keys() { log_trace!(logger, "Outpoint {}:{}", outp.txid, outp.vout); inputs.push(TxIn { previous_output: *outp, script_sig: Script::new(), sequence: 0xfffffffd, witness: Vec::new(), }); } let mut bumped_tx = Transaction { version: 2, lock_time: 0, input: inputs, output: vec![TxOut { script_pubkey: self.destination_script.clone(), value: 0 }], }; macro_rules! RBF_bump { ($amount: expr, $old_feerate: expr, $fee_estimator: expr, $predicted_weight: expr) => { { let mut used_feerate; // If old feerate inferior to actual one given back by Fee Estimator, use it to compute new fee... let new_fee = if $old_feerate < $fee_estimator.get_est_sat_per_1000_weight(ConfirmationTarget::HighPriority) { let mut value = $amount; if subtract_high_prio_fee!(logger, $fee_estimator, value, $predicted_weight, used_feerate) { // Overflow check is done in subtract_high_prio_fee $amount - value } else { log_trace!(logger, "Can't new-estimation bump new claiming tx, amount {} is too small", $amount); return None; } // ...else just increase the previous feerate by 25% (because that's a nice number) } else { let fee = $old_feerate * $predicted_weight / 750; if $amount <= fee { log_trace!(logger, "Can't 25% bump new claiming tx, amount {} is too small", $amount); return None; } fee }; let previous_fee = $old_feerate * $predicted_weight / 1000; let min_relay_fee = MIN_RELAY_FEE_SAT_PER_1000_WEIGHT * $predicted_weight / 1000; // BIP 125 Opt-in Full Replace-by-Fee Signaling // * 3. The replacement transaction pays an absolute fee of at least the sum paid by the original transactions. // * 4. The replacement transaction must also pay for its own bandwidth at or above the rate set by the node's minimum relay fee setting. let new_fee = if new_fee < previous_fee + min_relay_fee { new_fee + previous_fee + min_relay_fee - new_fee } else { new_fee }; Some((new_fee, new_fee * 1000 / $predicted_weight)) } } } // Compute new height timer to decide when we need to regenerate a new bumped version of the claim tx (if we // didn't receive confirmation of it before, or not enough reorg-safe depth on top of it). let new_timer = Some(Self::get_height_timer(height, cached_claim_datas.soonest_timelock)); let mut inputs_witnesses_weight = 0; let mut amt = 0; let mut dynamic_fee = true; for per_outp_material in cached_claim_datas.per_input_material.values() { match per_outp_material { &InputMaterial::Revoked { ref input_descriptor, ref amount, .. } => { inputs_witnesses_weight += Self::get_witnesses_weight(&[*input_descriptor]); amt += *amount; }, &InputMaterial::RemoteHTLC { ref preimage, ref amount, .. } => { inputs_witnesses_weight += Self::get_witnesses_weight(if preimage.is_some() { &[InputDescriptors::OfferedHTLC] } else { &[InputDescriptors::ReceivedHTLC] }); amt += *amount; }, &InputMaterial::LocalHTLC { .. } => { dynamic_fee = false; }, &InputMaterial::Funding { .. } => { dynamic_fee = false; } } } if dynamic_fee { let predicted_weight = bumped_tx.get_weight() + inputs_witnesses_weight; let mut new_feerate; // If old feerate is 0, first iteration of this claim, use normal fee calculation if cached_claim_datas.feerate_previous != 0 { if let Some((new_fee, feerate)) = RBF_bump!(amt, cached_claim_datas.feerate_previous, fee_estimator, predicted_weight as u64) { // If new computed fee is superior at the whole claimable amount burn all in fees if new_fee > amt { bumped_tx.output[0].value = 0; } else { bumped_tx.output[0].value = amt - new_fee; } new_feerate = feerate; } else { return None; } } else { if subtract_high_prio_fee!(logger, fee_estimator, amt, predicted_weight, new_feerate) { bumped_tx.output[0].value = amt; } else { return None; } } assert!(new_feerate != 0); for (i, (outp, per_outp_material)) in cached_claim_datas.per_input_material.iter().enumerate() { match per_outp_material { &InputMaterial::Revoked { ref per_commitment_point, ref per_commitment_key, ref input_descriptor, ref amount } => { if let Ok(chan_keys) = TxCreationKeys::new(&self.secp_ctx, &per_commitment_point, &self.remote_tx_cache.remote_delayed_payment_base_key, &self.remote_tx_cache.remote_htlc_base_key, &self.key_storage.pubkeys().revocation_basepoint, &self.key_storage.pubkeys().htlc_basepoint) { let mut this_htlc = None; if *input_descriptor != InputDescriptors::RevokedOutput { if let Some(htlcs) = self.remote_tx_cache.per_htlc.get(&outp.txid) { for htlc in htlcs { if htlc.transaction_output_index.unwrap() == outp.vout { this_htlc = Some(htlc); } } } } let witness_script = if *input_descriptor != InputDescriptors::RevokedOutput && this_htlc.is_some() { chan_utils::get_htlc_redeemscript_with_explicit_keys(&this_htlc.unwrap(), &chan_keys.a_htlc_key, &chan_keys.b_htlc_key, &chan_keys.revocation_key) } else if *input_descriptor != InputDescriptors::RevokedOutput { return None; } else { chan_utils::get_revokeable_redeemscript(&chan_keys.revocation_key, self.remote_csv, &chan_keys.a_delayed_payment_key) }; let is_htlc = *input_descriptor != InputDescriptors::RevokedOutput; if let Ok(sig) = self.key_storage.sign_justice_transaction(&bumped_tx, i, &witness_script, *amount, &per_commitment_key, &chan_keys.revocation_key, is_htlc, &self.secp_ctx) { bumped_tx.input[i].witness.push(sig.serialize_der().to_vec()); bumped_tx.input[i].witness[0].push(SigHashType::All as u8); if is_htlc { bumped_tx.input[i].witness.push(chan_keys.revocation_key.clone().serialize().to_vec()); } else { bumped_tx.input[i].witness.push(vec!(1)); } bumped_tx.input[i].witness.push(witness_script.clone().into_bytes()); } else { return None; } //TODO: panic ? log_trace!(logger, "Going to broadcast Penalty Transaction {} claiming revoked {} output {} from {} with new feerate {}...", bumped_tx.txid(), if *input_descriptor == InputDescriptors::RevokedOutput { "to_local" } else if *input_descriptor == InputDescriptors::RevokedOfferedHTLC { "offered" } else if *input_descriptor == InputDescriptors::RevokedReceivedHTLC { "received" } else { "" }, outp.vout, outp.txid, new_feerate); } }, &InputMaterial::RemoteHTLC { ref per_commitment_point, ref preimage, ref amount, ref locktime } => { if let Ok(chan_keys) = TxCreationKeys::new(&self.secp_ctx, &per_commitment_point, &self.remote_tx_cache.remote_delayed_payment_base_key, &self.remote_tx_cache.remote_htlc_base_key, &self.key_storage.pubkeys().revocation_basepoint, &self.key_storage.pubkeys().htlc_basepoint) { let mut this_htlc = None; if let Some(htlcs) = self.remote_tx_cache.per_htlc.get(&outp.txid) { for htlc in htlcs { if htlc.transaction_output_index.unwrap() == outp.vout { this_htlc = Some(htlc); } } } if this_htlc.is_none() { return None; } let witness_script = chan_utils::get_htlc_redeemscript_with_explicit_keys(&this_htlc.unwrap(), &chan_keys.a_htlc_key, &chan_keys.b_htlc_key, &chan_keys.revocation_key); if !preimage.is_some() { bumped_tx.lock_time = *locktime }; // Right now we don't aggregate time-locked transaction, if we do we should set lock_time before to avoid breaking hash computation if let Ok(sig) = self.key_storage.sign_remote_htlc_transaction(&bumped_tx, i, &witness_script, *amount, &per_commitment_point, preimage, &self.secp_ctx) { bumped_tx.input[i].witness.push(sig.serialize_der().to_vec()); bumped_tx.input[i].witness[0].push(SigHashType::All as u8); if let &Some(preimage) = preimage { bumped_tx.input[i].witness.push(preimage.0.to_vec()); } else { // Due to BIP146 (MINIMALIF) this must be a zero-length element to relay. bumped_tx.input[i].witness.push(vec![]); } bumped_tx.input[i].witness.push(witness_script.clone().into_bytes()); } log_trace!(logger, "Going to broadcast Claim Transaction {} claiming remote {} htlc output {} from {} with new feerate {}...", bumped_tx.txid(), if preimage.is_some() { "offered" } else { "received" }, outp.vout, outp.txid, new_feerate); } }, _ => unreachable!() } } log_trace!(logger, "...with timer {}", new_timer.unwrap()); assert!(predicted_weight >= bumped_tx.get_weight()); return Some((new_timer, new_feerate, bumped_tx)) } else { for (_, (outp, per_outp_material)) in cached_claim_datas.per_input_material.iter().enumerate() { match per_outp_material { &InputMaterial::LocalHTLC { ref preimage, ref amount } => { let htlc_tx = self.get_fully_signed_htlc_tx(outp, preimage); if let Some(htlc_tx) = htlc_tx { let feerate = (amount - htlc_tx.output[0].value) * 1000 / htlc_tx.get_weight() as u64; // Timer set to $NEVER given we can't bump tx without anchor outputs log_trace!(logger, "Going to broadcast Local HTLC-{} claiming HTLC output {} from {}...", if preimage.is_some() { "Success" } else { "Timeout" }, outp.vout, outp.txid); return Some((None, feerate, htlc_tx)); } return None; }, &InputMaterial::Funding { ref funding_redeemscript } => { let signed_tx = self.get_fully_signed_local_tx(funding_redeemscript).unwrap(); // Timer set to $NEVER given we can't bump tx without anchor outputs log_trace!(logger, "Going to broadcast Local Transaction {} claiming funding output {} from {}...", signed_tx.txid(), outp.vout, outp.txid); return Some((None, self.local_commitment.as_ref().unwrap().feerate_per_kw, signed_tx)); } _ => unreachable!() } } } None } pub(super) fn block_connected(&mut self, txn_matched: &[&Transaction], claimable_outpoints: Vec, height: u32, broadcaster: B, fee_estimator: F, logger: L) where B::Target: BroadcasterInterface, F::Target: FeeEstimator, L::Target: Logger, { log_trace!(logger, "Block at height {} connected with {} claim requests", height, claimable_outpoints.len()); let mut new_claims = Vec::new(); let mut aggregated_claim = HashMap::new(); let mut aggregated_soonest = ::std::u32::MAX; // Try to aggregate outputs if their timelock expiration isn't imminent (absolute_timelock // <= CLTV_SHARED_CLAIM_BUFFER) and they don't require an immediate nLockTime (aggregable). for req in claimable_outpoints { // Don't claim a outpoint twice that would be bad for privacy and may uselessly lock a CPFP input for a while if let Some(_) = self.claimable_outpoints.get(&req.outpoint) { log_trace!(logger, "Bouncing off outpoint {}:{}, already registered its claiming request", req.outpoint.txid, req.outpoint.vout); } else { log_trace!(logger, "Test if outpoint can be aggregated with expiration {} against {}", req.absolute_timelock, height + CLTV_SHARED_CLAIM_BUFFER); if req.absolute_timelock <= height + CLTV_SHARED_CLAIM_BUFFER || !req.aggregable { // Don't aggregate if outpoint absolute timelock is soon or marked as non-aggregable let mut single_input = HashMap::new(); single_input.insert(req.outpoint, req.witness_data); new_claims.push((req.absolute_timelock, single_input)); } else { aggregated_claim.insert(req.outpoint, req.witness_data); if req.absolute_timelock < aggregated_soonest { aggregated_soonest = req.absolute_timelock; } } } } new_claims.push((aggregated_soonest, aggregated_claim)); // Generate claim transactions and track them to bump if necessary at // height timer expiration (i.e in how many blocks we're going to take action). for (soonest_timelock, claim) in new_claims.drain(..) { let mut claim_material = ClaimTxBumpMaterial { height_timer: None, feerate_previous: 0, soonest_timelock, per_input_material: claim }; if let Some((new_timer, new_feerate, tx)) = self.generate_claim_tx(height, &claim_material, &*fee_estimator, &*logger) { claim_material.height_timer = new_timer; claim_material.feerate_previous = new_feerate; let txid = tx.txid(); for k in claim_material.per_input_material.keys() { log_trace!(logger, "Registering claiming request for {}:{}", k.txid, k.vout); self.claimable_outpoints.insert(k.clone(), (txid, height)); } self.pending_claim_requests.insert(txid, claim_material); log_trace!(logger, "Broadcast onchain {}", log_tx!(tx)); broadcaster.broadcast_transaction(&tx); } } let mut bump_candidates = HashMap::new(); for tx in txn_matched { // Scan all input to verify is one of the outpoint spent is of interest for us let mut claimed_outputs_material = Vec::new(); for inp in &tx.input { if let Some(first_claim_txid_height) = self.claimable_outpoints.get(&inp.previous_output) { // If outpoint has claim request pending on it... if let Some(claim_material) = self.pending_claim_requests.get_mut(&first_claim_txid_height.0) { //... we need to verify equality between transaction outpoints and claim request // outpoints to know if transaction is the original claim or a bumped one issued // by us. let mut set_equality = true; if claim_material.per_input_material.len() != tx.input.len() { set_equality = false; } else { for (claim_inp, tx_inp) in claim_material.per_input_material.keys().zip(tx.input.iter()) { if *claim_inp != tx_inp.previous_output { set_equality = false; } } } macro_rules! clean_claim_request_after_safety_delay { () => { let new_event = OnchainEvent::Claim { claim_request: first_claim_txid_height.0.clone() }; match self.onchain_events_waiting_threshold_conf.entry(height + ANTI_REORG_DELAY - 1) { hash_map::Entry::Occupied(mut entry) => { if !entry.get().contains(&new_event) { entry.get_mut().push(new_event); } }, hash_map::Entry::Vacant(entry) => { entry.insert(vec![new_event]); } } } } // If this is our transaction (or our counterparty spent all the outputs // before we could anyway with same inputs order than us), wait for // ANTI_REORG_DELAY and clean the RBF tracking map. if set_equality { clean_claim_request_after_safety_delay!(); } else { // If false, generate new claim request with update outpoint set let mut at_least_one_drop = false; for input in tx.input.iter() { if let Some(input_material) = claim_material.per_input_material.remove(&input.previous_output) { claimed_outputs_material.push((input.previous_output, input_material)); at_least_one_drop = true; } // If there are no outpoints left to claim in this request, drop it entirely after ANTI_REORG_DELAY. if claim_material.per_input_material.is_empty() { clean_claim_request_after_safety_delay!(); } } //TODO: recompute soonest_timelock to avoid wasting a bit on fees if at_least_one_drop { bump_candidates.insert(first_claim_txid_height.0.clone(), claim_material.clone()); } } break; //No need to iterate further, either tx is our or their } else { panic!("Inconsistencies between pending_claim_requests map and claimable_outpoints map"); } } } for (outpoint, input_material) in claimed_outputs_material.drain(..) { let new_event = OnchainEvent::ContentiousOutpoint { outpoint, input_material }; match self.onchain_events_waiting_threshold_conf.entry(height + ANTI_REORG_DELAY - 1) { hash_map::Entry::Occupied(mut entry) => { if !entry.get().contains(&new_event) { entry.get_mut().push(new_event); } }, hash_map::Entry::Vacant(entry) => { entry.insert(vec![new_event]); } } } } // After security delay, either our claim tx got enough confs or outpoint is definetely out of reach if let Some(events) = self.onchain_events_waiting_threshold_conf.remove(&height) { for ev in events { match ev { OnchainEvent::Claim { claim_request } => { // We may remove a whole set of claim outpoints here, as these one may have // been aggregated in a single tx and claimed so atomically if let Some(bump_material) = self.pending_claim_requests.remove(&claim_request) { for outpoint in bump_material.per_input_material.keys() { self.claimable_outpoints.remove(&outpoint); } } }, OnchainEvent::ContentiousOutpoint { outpoint, .. } => { self.claimable_outpoints.remove(&outpoint); } } } } // Check if any pending claim request must be rescheduled for (first_claim_txid, ref claim_data) in self.pending_claim_requests.iter() { if let Some(h) = claim_data.height_timer { if h == height { bump_candidates.insert(*first_claim_txid, (*claim_data).clone()); } } } // Build, bump and rebroadcast tx accordingly log_trace!(logger, "Bumping {} candidates", bump_candidates.len()); for (first_claim_txid, claim_material) in bump_candidates.iter() { if let Some((new_timer, new_feerate, bump_tx)) = self.generate_claim_tx(height, &claim_material, &*fee_estimator, &*logger) { log_trace!(logger, "Broadcast onchain {}", log_tx!(bump_tx)); broadcaster.broadcast_transaction(&bump_tx); if let Some(claim_material) = self.pending_claim_requests.get_mut(first_claim_txid) { claim_material.height_timer = new_timer; claim_material.feerate_previous = new_feerate; } } } } pub(super) fn block_disconnected(&mut self, height: u32, broadcaster: B, fee_estimator: F, logger: L) where B::Target: BroadcasterInterface, F::Target: FeeEstimator, L::Target: Logger, { let mut bump_candidates = HashMap::new(); if let Some(events) = self.onchain_events_waiting_threshold_conf.remove(&(height + ANTI_REORG_DELAY - 1)) { //- our claim tx on a commitment tx output //- resurect outpoint back in its claimable set and regenerate tx for ev in events { match ev { OnchainEvent::ContentiousOutpoint { outpoint, input_material } => { if let Some(ancestor_claimable_txid) = self.claimable_outpoints.get(&outpoint) { if let Some(claim_material) = self.pending_claim_requests.get_mut(&ancestor_claimable_txid.0) { claim_material.per_input_material.insert(outpoint, input_material); // Using a HashMap guarantee us than if we have multiple outpoints getting // resurrected only one bump claim tx is going to be broadcast bump_candidates.insert(ancestor_claimable_txid.clone(), claim_material.clone()); } } }, _ => {}, } } } for (_, claim_material) in bump_candidates.iter_mut() { if let Some((new_timer, new_feerate, bump_tx)) = self.generate_claim_tx(height, &claim_material, &*fee_estimator, &*logger) { claim_material.height_timer = new_timer; claim_material.feerate_previous = new_feerate; broadcaster.broadcast_transaction(&bump_tx); } } for (ancestor_claim_txid, claim_material) in bump_candidates.drain() { self.pending_claim_requests.insert(ancestor_claim_txid.0, claim_material); } //TODO: if we implement cross-block aggregated claim transaction we need to refresh set of outpoints and regenerate tx but // right now if one of the outpoint get disconnected, just erase whole pending claim request. let mut remove_request = Vec::new(); self.claimable_outpoints.retain(|_, ref v| if v.1 == height { remove_request.push(v.0.clone()); false } else { true }); for req in remove_request { self.pending_claim_requests.remove(&req); } } pub(super) fn provide_latest_local_tx(&mut self, tx: LocalCommitmentTransaction) -> Result<(), ()> { // To prevent any unsafe state discrepancy between offchain and onchain, once local // commitment transaction has been signed due to an event (either block height for // HTLC-timeout or channel force-closure), don't allow any further update of local // commitment transaction view to avoid delivery of revocation secret to counterparty // for the aformentionned signed transaction. if self.local_htlc_sigs.is_some() || self.prev_local_htlc_sigs.is_some() { return Err(()); } self.prev_local_commitment = self.local_commitment.take(); self.local_commitment = Some(tx); Ok(()) } fn sign_latest_local_htlcs(&mut self) { if let Some(ref local_commitment) = self.local_commitment { if let Ok(sigs) = self.key_storage.sign_local_commitment_htlc_transactions(local_commitment, self.local_csv, &self.secp_ctx) { self.local_htlc_sigs = Some(Vec::new()); let ret = self.local_htlc_sigs.as_mut().unwrap(); for (htlc_idx, (local_sig, &(ref htlc, _))) in sigs.iter().zip(local_commitment.per_htlc.iter()).enumerate() { if let Some(tx_idx) = htlc.transaction_output_index { if ret.len() <= tx_idx as usize { ret.resize(tx_idx as usize + 1, None); } ret[tx_idx as usize] = Some((htlc_idx, local_sig.expect("Did not receive a signature for a non-dust HTLC"))); } else { assert!(local_sig.is_none(), "Received a signature for a dust HTLC"); } } } } } fn sign_prev_local_htlcs(&mut self) { if let Some(ref local_commitment) = self.prev_local_commitment { if let Ok(sigs) = self.key_storage.sign_local_commitment_htlc_transactions(local_commitment, self.local_csv, &self.secp_ctx) { self.prev_local_htlc_sigs = Some(Vec::new()); let ret = self.prev_local_htlc_sigs.as_mut().unwrap(); for (htlc_idx, (local_sig, &(ref htlc, _))) in sigs.iter().zip(local_commitment.per_htlc.iter()).enumerate() { if let Some(tx_idx) = htlc.transaction_output_index { if ret.len() <= tx_idx as usize { ret.resize(tx_idx as usize + 1, None); } ret[tx_idx as usize] = Some((htlc_idx, local_sig.expect("Did not receive a signature for a non-dust HTLC"))); } else { assert!(local_sig.is_none(), "Received a signature for a dust HTLC"); } } } } } //TODO: getting lastest local transactions should be infaillible and result in us "force-closing the channel", but we may // have empty local commitment transaction if a ChannelMonitor is asked to force-close just after Channel::get_outbound_funding_created, // before providing a initial commitment transaction. For outbound channel, init ChannelMonitor at Channel::funding_signed, there is nothing // to monitor before. pub(super) fn get_fully_signed_local_tx(&mut self, funding_redeemscript: &Script) -> Option { if let Some(ref mut local_commitment) = self.local_commitment { match self.key_storage.sign_local_commitment(local_commitment, &self.secp_ctx) { Ok(sig) => Some(local_commitment.add_local_sig(funding_redeemscript, sig)), Err(_) => return None, } } else { None } } pub(super) fn provide_latest_remote_tx(&mut self, commitment_txid: Txid, htlcs: Vec) { self.remote_tx_cache.per_htlc.insert(commitment_txid, htlcs); } #[cfg(test)] pub(super) fn get_fully_signed_copy_local_tx(&mut self, funding_redeemscript: &Script) -> Option { if let Some(ref mut local_commitment) = self.local_commitment { let local_commitment = local_commitment.clone(); match self.key_storage.sign_local_commitment(&local_commitment, &self.secp_ctx) { Ok(sig) => Some(local_commitment.add_local_sig(funding_redeemscript, sig)), Err(_) => return None, } } else { None } } pub(super) fn get_fully_signed_htlc_tx(&mut self, outp: &::bitcoin::OutPoint, preimage: &Option) -> Option { let mut htlc_tx = None; if self.local_commitment.is_some() { let commitment_txid = self.local_commitment.as_ref().unwrap().txid(); if commitment_txid == outp.txid { self.sign_latest_local_htlcs(); if let &Some(ref htlc_sigs) = &self.local_htlc_sigs { let &(ref htlc_idx, ref htlc_sig) = htlc_sigs[outp.vout as usize].as_ref().unwrap(); htlc_tx = Some(self.local_commitment.as_ref().unwrap() .get_signed_htlc_tx(*htlc_idx, htlc_sig, preimage, self.local_csv)); } } } if self.prev_local_commitment.is_some() { let commitment_txid = self.prev_local_commitment.as_ref().unwrap().txid(); if commitment_txid == outp.txid { self.sign_prev_local_htlcs(); if let &Some(ref htlc_sigs) = &self.prev_local_htlc_sigs { let &(ref htlc_idx, ref htlc_sig) = htlc_sigs[outp.vout as usize].as_ref().unwrap(); htlc_tx = Some(self.prev_local_commitment.as_ref().unwrap() .get_signed_htlc_tx(*htlc_idx, htlc_sig, preimage, self.local_csv)); } } } htlc_tx } #[cfg(test)] pub(super) fn unsafe_get_fully_signed_htlc_tx(&mut self, outp: &::bitcoin::OutPoint, preimage: &Option) -> Option { let latest_had_sigs = self.local_htlc_sigs.is_some(); let prev_had_sigs = self.prev_local_htlc_sigs.is_some(); let ret = self.get_fully_signed_htlc_tx(outp, preimage); if !latest_had_sigs { self.local_htlc_sigs = None; } if !prev_had_sigs { self.prev_local_htlc_sigs = None; } ret } }