From: Matt Corallo Date: Tue, 27 Nov 2018 02:54:14 +0000 (-0500) Subject: Handle monitor update failures in msg-recv functions X-Git-Tag: v0.0.12~258^2~2 X-Git-Url: http://git.bitcoin.ninja/?a=commitdiff_plain;h=787644d795193eaae3041078030efebca1a5687c;p=rust-lightning Handle monitor update failures in msg-recv functions This adds a few TODOs around further message rebroadcasting which needs to be implemented as well as some loss of tracking of HTLCs on permanent channel failure which needs to get transferred over to the appropriate in-memory ChannelMonitor. --- diff --git a/src/ln/channel.rs b/src/ln/channel.rs index 296d97c16..c2ec53fb3 100644 --- a/src/ln/channel.rs +++ b/src/ln/channel.rs @@ -2154,7 +2154,7 @@ impl Channel { /// commitment update or a revoke_and_ack generation). The messages which were generated from /// that original call must *not* have been sent to the remote end, and must instead have been /// dropped. They will be regenerated when monitor_updating_restored is called. - pub fn monitor_update_failed(&mut self, order: RAACommitmentOrder) { + pub fn monitor_update_failed(&mut self, order: RAACommitmentOrder, mut pending_forwards: Vec<(PendingForwardHTLCInfo, u64)>, mut pending_fails: Vec<(HTLCSource, [u8; 32], HTLCFailReason)>, raa_first_dropped_cs: bool) { assert_eq!(self.channel_state & ChannelState::MonitorUpdateFailed as u32, 0); match order { RAACommitmentOrder::CommitmentFirst => { @@ -2163,9 +2163,13 @@ impl Channel { }, RAACommitmentOrder::RevokeAndACKFirst => { self.monitor_pending_revoke_and_ack = true; - self.monitor_pending_commitment_signed = false; + self.monitor_pending_commitment_signed = raa_first_dropped_cs; }, } + assert!(self.monitor_pending_forwards.is_empty()); + mem::swap(&mut pending_forwards, &mut self.monitor_pending_forwards); + assert!(self.monitor_pending_failures.is_empty()); + mem::swap(&mut pending_fails, &mut self.monitor_pending_failures); self.monitor_pending_order = Some(order); self.channel_state |= ChannelState::MonitorUpdateFailed as u32; } diff --git a/src/ln/channelmanager.rs b/src/ln/channelmanager.rs index f6ee5f8cd..87326f2be 100644 --- a/src/ln/channelmanager.rs +++ b/src/ln/channelmanager.rs @@ -442,6 +442,43 @@ macro_rules! try_chan_entry { } } +macro_rules! return_monitor_err { + ($self: expr, $err: expr, $channel_state: expr, $entry: expr, $action_type: path) => { + return_monitor_err!($self, $err, $channel_state, $entry, $action_type, Vec::new(), Vec::new()) + }; + ($self: expr, $err: expr, $channel_state: expr, $entry: expr, $action_type: path, $raa_first_dropped_cs: expr) => { + if $action_type != RAACommitmentOrder::RevokeAndACKFirst { panic!("Bad return_monitor_err call!"); } + return_monitor_err!($self, $err, $channel_state, $entry, $action_type, Vec::new(), Vec::new(), $raa_first_dropped_cs) + }; + ($self: expr, $err: expr, $channel_state: expr, $entry: expr, $action_type: path, $failed_forwards: expr, $failed_fails: expr) => { + return_monitor_err!($self, $err, $channel_state, $entry, $action_type, $failed_forwards, $failed_fails, false) + }; + ($self: expr, $err: expr, $channel_state: expr, $entry: expr, $action_type: path, $failed_forwards: expr, $failed_fails: expr, $raa_first_dropped_cs: expr) => { + match $err { + ChannelMonitorUpdateErr::PermanentFailure => { + let (channel_id, mut chan) = $entry.remove_entry(); + if let Some(short_id) = chan.get_short_channel_id() { + $channel_state.short_to_id.remove(&short_id); + } + // TODO: $failed_fails is dropped here, which will cause other channels to hit the + // chain in a confused state! We need to move them into the ChannelMonitor which + // will be responsible for failing backwards once things confirm on-chain. + // It's ok that we drop $failed_forwards here - at this point we'd rather they + // broadcast HTLC-Timeout and pay the associated fees to get their funds back than + // us bother trying to claim it just to forward on to another peer. If we're + // splitting hairs we'd prefer to claim payments that were to us, but we haven't + // given up the preimage yet, so might as well just wait until the payment is + // retried, avoiding the on-chain fees. + return Err(MsgHandleErrInternal::from_finish_shutdown("ChannelMonitor storage failure", channel_id, chan.force_shutdown(), $self.get_channel_update(&chan).ok())) + }, + ChannelMonitorUpdateErr::TemporaryFailure => { + $entry.get_mut().monitor_update_failed($action_type, $failed_forwards, $failed_fails, $raa_first_dropped_cs); + return Err(MsgHandleErrInternal::from_chan_no_close(ChannelError::Ignore("Failed to update ChannelMonitor"), *$entry.key())); + }, + } + } +} + // Does not break in case of TemporaryFailure! macro_rules! maybe_break_monitor_err { ($self: expr, $err: expr, $channel_state: expr, $entry: expr, $action_type: path) => { @@ -454,7 +491,7 @@ macro_rules! maybe_break_monitor_err { break Err(MsgHandleErrInternal::from_finish_shutdown("ChannelMonitor storage failure", channel_id, chan.force_shutdown(), $self.get_channel_update(&chan).ok())) }, ChannelMonitorUpdateErr::TemporaryFailure => { - $entry.get_mut().monitor_update_failed($action_type); + $entry.get_mut().monitor_update_failed($action_type, Vec::new(), Vec::new(), false); }, } } @@ -1681,6 +1718,13 @@ impl ChannelManager { if let Err(e) = self.monitor.add_update_monitor(chan_monitor.get_funding_txo().unwrap(), chan_monitor) { match e { ChannelMonitorUpdateErr::PermanentFailure => { + // TODO: There may be some pending HTLCs that we intended to fail + // backwards when a monitor update failed. We should make sure + // knowledge of those gets moved into the appropriate in-memory + // ChannelMonitor and they get failed backwards once we get + // on-chain confirmations. + // Note I think #198 addresses this, so once its merged a test + // should be written. if let Some(short_id) = channel.get_short_channel_id() { short_to_id.remove(&short_id); } @@ -2283,8 +2327,9 @@ impl ChannelManager { } let (revoke_and_ack, commitment_signed, closing_signed, chan_monitor) = try_chan_entry!(self, chan.get_mut().commitment_signed(&msg, &*self.fee_estimator), channel_state, chan); - if let Err(_e) = self.monitor.add_update_monitor(chan_monitor.get_funding_txo().unwrap(), chan_monitor) { - unimplemented!(); + if let Err(e) = self.monitor.add_update_monitor(chan_monitor.get_funding_txo().unwrap(), chan_monitor) { + return_monitor_err!(self, e, channel_state, chan, RAACommitmentOrder::RevokeAndACKFirst, commitment_signed.is_some()); + //TODO: Rebroadcast closing_signed if present on monitor update restoration } channel_state.pending_msg_events.push(events::MessageSendEvent::SendRevokeAndACK { node_id: their_node_id.clone(), @@ -2360,8 +2405,8 @@ impl ChannelManager { } let (commitment_update, pending_forwards, pending_failures, closing_signed, chan_monitor) = try_chan_entry!(self, chan.get_mut().revoke_and_ack(&msg, &*self.fee_estimator), channel_state, chan); - if let Err(_e) = self.monitor.add_update_monitor(chan_monitor.get_funding_txo().unwrap(), chan_monitor) { - unimplemented!(); + if let Err(e) = self.monitor.add_update_monitor(chan_monitor.get_funding_txo().unwrap(), chan_monitor) { + return_monitor_err!(self, e, channel_state, chan, RAACommitmentOrder::CommitmentFirst, pending_forwards, pending_failures); } if let Some(updates) = commitment_update { channel_state.pending_msg_events.push(events::MessageSendEvent::UpdateHTLCs { @@ -2455,11 +2500,21 @@ impl ChannelManager { if chan.get().get_their_node_id() != *their_node_id { return Err(MsgHandleErrInternal::send_err_msg_no_close("Got a message for a channel from the wrong node!", msg.channel_id)); } - let (funding_locked, revoke_and_ack, commitment_update, channel_monitor, order, shutdown) = + let (funding_locked, revoke_and_ack, commitment_update, channel_monitor, mut order, shutdown) = try_chan_entry!(self, chan.get_mut().channel_reestablish(msg), channel_state, chan); if let Some(monitor) = channel_monitor { - if let Err(_e) = self.monitor.add_update_monitor(monitor.get_funding_txo().unwrap(), monitor) { - unimplemented!(); + if let Err(e) = self.monitor.add_update_monitor(monitor.get_funding_txo().unwrap(), monitor) { + // channel_reestablish doesn't guarantee the order it returns is sensical + // for the messages it returns, but if we're setting what messages to + // re-transmit on monitor update success, we need to make sure it is sane. + if revoke_and_ack.is_none() { + order = RAACommitmentOrder::CommitmentFirst; + } + if commitment_update.is_none() { + order = RAACommitmentOrder::RevokeAndACKFirst; + } + return_monitor_err!(self, e, channel_state, chan, order); + //TODO: Resend the funding_locked if needed once we get the monitor running again } } if let Some(msg) = funding_locked {