Don't apply gossip backpressure to non-channel-announcing peers
[rust-lightning] / lightning / src / routing / utxo.rs
index 1008e6a369eff5141adbcf8eda2a37ebf72e487d..da9c65c3b12a605c385fd566fe4cf0a505df4b35 100644 (file)
 //! order to announce a channel. This module handles that checking.
 
 use bitcoin::{BlockHash, TxOut};
+use bitcoin::hashes::hex::ToHex;
+
+use crate::ln::chan_utils::make_funding_redeemscript_from_slices;
+use crate::ln::msgs::{self, LightningError, ErrorAction};
+use crate::routing::gossip::{NetworkGraph, NodeId, P2PGossipSync};
+use crate::util::events::MessageSendEvent;
+use crate::util::logger::{Level, Logger};
+use crate::util::ser::Writeable;
+
+use crate::prelude::*;
+
+use alloc::sync::{Arc, Weak};
+use crate::sync::Mutex;
+use core::ops::Deref;
 
 /// An error when accessing the chain via [`UtxoLookup`].
 #[derive(Clone, Debug)]
@@ -25,6 +39,23 @@ pub enum UtxoLookupError {
        UnknownTx,
 }
 
+/// The result of a [`UtxoLookup::get_utxo`] call. A call may resolve either synchronously,
+/// returning the `Sync` variant, or asynchronously, returning an [`UtxoFuture`] in the `Async`
+/// variant.
+pub enum UtxoResult {
+       /// A result which was resolved synchronously. It either includes a [`TxOut`] for the output
+       /// requested or a [`UtxoLookupError`].
+       Sync(Result<TxOut, UtxoLookupError>),
+       /// A result which will be resolved asynchronously. It includes a [`UtxoFuture`], a `clone` of
+       /// which you must keep locally and call [`UtxoFuture::resolve`] on once the lookup completes.
+       ///
+       /// Note that in order to avoid runaway memory usage, the number of parallel checks is limited,
+       /// but only fairly loosely. Because a pending checks block all message processing, leaving
+       /// checks pending for an extended time may cause DoS of other functions. It is recommended you
+       /// keep a tight timeout on lookups, on the order of a few seconds.
+       Async(UtxoFuture),
+}
+
 /// The `UtxoLookup` trait defines behavior for accessing on-chain UTXOs.
 pub trait UtxoLookup {
        /// Returns the transaction output of a funding transaction encoded by [`short_channel_id`].
@@ -32,5 +63,497 @@ pub trait UtxoLookup {
        /// is unknown.
        ///
        /// [`short_channel_id`]: https://github.com/lightning/bolts/blob/master/07-routing-gossip.md#definition-of-short_channel_id
-       fn get_utxo(&self, genesis_hash: &BlockHash, short_channel_id: u64) -> Result<TxOut, UtxoLookupError>;
+       fn get_utxo(&self, genesis_hash: &BlockHash, short_channel_id: u64) -> UtxoResult;
+}
+
+enum ChannelAnnouncement {
+       Full(msgs::ChannelAnnouncement),
+       Unsigned(msgs::UnsignedChannelAnnouncement),
+}
+impl ChannelAnnouncement {
+       fn node_id_1(&self) -> &NodeId {
+               match self {
+                       ChannelAnnouncement::Full(msg) => &msg.contents.node_id_1,
+                       ChannelAnnouncement::Unsigned(msg) => &msg.node_id_1,
+               }
+       }
+}
+
+enum NodeAnnouncement {
+       Full(msgs::NodeAnnouncement),
+       Unsigned(msgs::UnsignedNodeAnnouncement),
+}
+impl NodeAnnouncement {
+       fn timestamp(&self) -> u32 {
+               match self {
+                       NodeAnnouncement::Full(msg) => msg.contents.timestamp,
+                       NodeAnnouncement::Unsigned(msg) => msg.timestamp,
+               }
+       }
+}
+
+enum ChannelUpdate {
+       Full(msgs::ChannelUpdate),
+       Unsigned(msgs::UnsignedChannelUpdate),
+}
+impl ChannelUpdate {
+       fn timestamp(&self) -> u32 {
+               match self {
+                       ChannelUpdate::Full(msg) => msg.contents.timestamp,
+                       ChannelUpdate::Unsigned(msg) => msg.timestamp,
+               }
+       }
+}
+
+struct UtxoMessages {
+       complete: Option<Result<TxOut, UtxoLookupError>>,
+       channel_announce: Option<ChannelAnnouncement>,
+       latest_node_announce_a: Option<NodeAnnouncement>,
+       latest_node_announce_b: Option<NodeAnnouncement>,
+       latest_channel_update_a: Option<ChannelUpdate>,
+       latest_channel_update_b: Option<ChannelUpdate>,
+}
+
+/// Represents a future resolution of a [`UtxoLookup::get_utxo`] query resolving async.
+///
+/// See [`UtxoResult::Async`] and [`UtxoFuture::resolve`] for more info.
+#[derive(Clone)]
+pub struct UtxoFuture {
+       state: Arc<Mutex<UtxoMessages>>,
+}
+
+/// A trivial implementation of [`UtxoLookup`] which is used to call back into the network graph
+/// once we have a concrete resolution of a request.
+struct UtxoResolver(Result<TxOut, UtxoLookupError>);
+impl UtxoLookup for UtxoResolver {
+       fn get_utxo(&self, _genesis_hash: &BlockHash, _short_channel_id: u64) -> UtxoResult {
+               UtxoResult::Sync(self.0.clone())
+       }
+}
+
+impl UtxoFuture {
+       /// Builds a new future for later resolution.
+       pub fn new() -> Self {
+               Self { state: Arc::new(Mutex::new(UtxoMessages {
+                       complete: None,
+                       channel_announce: None,
+                       latest_node_announce_a: None,
+                       latest_node_announce_b: None,
+                       latest_channel_update_a: None,
+                       latest_channel_update_b: None,
+               }))}
+       }
+
+       /// Resolves this future against the given `graph` and with the given `result`.
+       ///
+       /// This is identical to calling [`UtxoFuture::resolve`] with a dummy `gossip`, disabling
+       /// forwarding the validated gossip message onwards to peers.
+       ///
+       /// Because this may cause the [`NetworkGraph`]'s [`processing_queue_high`] to flip, in order
+       /// to allow us to interact with peers again, you should call [`PeerManager::process_events`]
+       /// after this.
+       ///
+       /// [`processing_queue_high`]: crate::ln::msgs::RoutingMessageHandler::processing_queue_high
+       /// [`PeerManager::process_events`]: crate::ln::peer_handler::PeerManager::process_events
+       pub fn resolve_without_forwarding<L: Deref>(&self,
+               graph: &NetworkGraph<L>, result: Result<TxOut, UtxoLookupError>)
+       where L::Target: Logger {
+               self.do_resolve(graph, result);
+       }
+
+       /// Resolves this future against the given `graph` and with the given `result`.
+       ///
+       /// The given `gossip` is used to broadcast any validated messages onwards to all peers which
+       /// have available buffer space.
+       ///
+       /// Because this may cause the [`NetworkGraph`]'s [`processing_queue_high`] to flip, in order
+       /// to allow us to interact with peers again, you should call [`PeerManager::process_events`]
+       /// after this.
+       ///
+       /// [`processing_queue_high`]: crate::ln::msgs::RoutingMessageHandler::processing_queue_high
+       /// [`PeerManager::process_events`]: crate::ln::peer_handler::PeerManager::process_events
+       pub fn resolve<L: Deref, G: Deref<Target=NetworkGraph<L>>, U: Deref, GS: Deref<Target = P2PGossipSync<G, U, L>>>(&self,
+               graph: &NetworkGraph<L>, gossip: GS, result: Result<TxOut, UtxoLookupError>
+       ) where L::Target: Logger, U::Target: UtxoLookup {
+               let mut res = self.do_resolve(graph, result);
+               for msg_opt in res.iter_mut() {
+                       if let Some(msg) = msg_opt.take() {
+                               gossip.forward_gossip_msg(msg);
+                       }
+               }
+       }
+
+       fn do_resolve<L: Deref>(&self, graph: &NetworkGraph<L>, result: Result<TxOut, UtxoLookupError>)
+       -> [Option<MessageSendEvent>; 5] where L::Target: Logger {
+               let (announcement, node_a, node_b, update_a, update_b) = {
+                       let mut pending_checks = graph.pending_checks.internal.lock().unwrap();
+                       let mut async_messages = self.state.lock().unwrap();
+
+                       if async_messages.channel_announce.is_none() {
+                               // We raced returning to `check_channel_announcement` which hasn't updated
+                               // `channel_announce` yet. That's okay, we can set the `complete` field which it will
+                               // check once it gets control again.
+                               async_messages.complete = Some(result);
+                               return [None, None, None, None, None];
+                       }
+
+                       let announcement_msg = match async_messages.channel_announce.as_ref().unwrap() {
+                               ChannelAnnouncement::Full(signed_msg) => &signed_msg.contents,
+                               ChannelAnnouncement::Unsigned(msg) => &msg,
+                       };
+
+                       pending_checks.lookup_completed(announcement_msg, &Arc::downgrade(&self.state));
+
+                       (async_messages.channel_announce.take().unwrap(),
+                               async_messages.latest_node_announce_a.take(),
+                               async_messages.latest_node_announce_b.take(),
+                               async_messages.latest_channel_update_a.take(),
+                               async_messages.latest_channel_update_b.take())
+               };
+
+               let mut res = [None, None, None, None, None];
+               let mut res_idx = 0;
+
+               // Now that we've updated our internal state, pass the pending messages back through the
+               // network graph with a different `UtxoLookup` which will resolve immediately.
+               // Note that we ignore errors as we don't disconnect peers anyway, so there's nothing to do
+               // with them.
+               let resolver = UtxoResolver(result);
+               match announcement {
+                       ChannelAnnouncement::Full(signed_msg) => {
+                               if graph.update_channel_from_announcement(&signed_msg, &Some(&resolver)).is_ok() {
+                                       res[res_idx] = Some(MessageSendEvent::BroadcastChannelAnnouncement {
+                                               msg: signed_msg, update_msg: None,
+                                       });
+                                       res_idx += 1;
+                               }
+                       },
+                       ChannelAnnouncement::Unsigned(msg) => {
+                               let _ = graph.update_channel_from_unsigned_announcement(&msg, &Some(&resolver));
+                       },
+               }
+
+               for announce in core::iter::once(node_a).chain(core::iter::once(node_b)) {
+                       match announce {
+                               Some(NodeAnnouncement::Full(signed_msg)) => {
+                                       if graph.update_node_from_announcement(&signed_msg).is_ok() {
+                                               res[res_idx] = Some(MessageSendEvent::BroadcastNodeAnnouncement {
+                                                       msg: signed_msg,
+                                               });
+                                               res_idx += 1;
+                                       }
+                               },
+                               Some(NodeAnnouncement::Unsigned(msg)) => {
+                                       let _ = graph.update_node_from_unsigned_announcement(&msg);
+                               },
+                               None => {},
+                       }
+               }
+
+               for update in core::iter::once(update_a).chain(core::iter::once(update_b)) {
+                       match update {
+                               Some(ChannelUpdate::Full(signed_msg)) => {
+                                       if graph.update_channel(&signed_msg).is_ok() {
+                                               res[res_idx] = Some(MessageSendEvent::BroadcastChannelUpdate {
+                                                       msg: signed_msg,
+                                               });
+                                               res_idx += 1;
+                                       }
+                               },
+                               Some(ChannelUpdate::Unsigned(msg)) => {
+                                       let _ = graph.update_channel_unsigned(&msg);
+                               },
+                               None => {},
+                       }
+               }
+
+               res
+       }
+}
+
+struct PendingChecksContext {
+       channels: HashMap<u64, Weak<Mutex<UtxoMessages>>>,
+       nodes: HashMap<NodeId, Vec<Weak<Mutex<UtxoMessages>>>>,
+}
+
+impl PendingChecksContext {
+       fn lookup_completed(&mut self,
+               msg: &msgs::UnsignedChannelAnnouncement, completed_state: &Weak<Mutex<UtxoMessages>>
+       ) {
+               if let hash_map::Entry::Occupied(e) = self.channels.entry(msg.short_channel_id) {
+                       if Weak::ptr_eq(e.get(), &completed_state) {
+                               e.remove();
+                       }
+               }
+
+               if let hash_map::Entry::Occupied(mut e) = self.nodes.entry(msg.node_id_1) {
+                       e.get_mut().retain(|elem| !Weak::ptr_eq(&elem, &completed_state));
+                       if e.get().is_empty() { e.remove(); }
+               }
+               if let hash_map::Entry::Occupied(mut e) = self.nodes.entry(msg.node_id_2) {
+                       e.get_mut().retain(|elem| !Weak::ptr_eq(&elem, &completed_state));
+                       if e.get().is_empty() { e.remove(); }
+               }
+       }
+}
+
+/// A set of messages which are pending UTXO lookups for processing.
+pub(super) struct PendingChecks {
+       internal: Mutex<PendingChecksContext>,
+}
+
+impl PendingChecks {
+       pub(super) fn new() -> Self {
+               PendingChecks { internal: Mutex::new(PendingChecksContext {
+                       channels: HashMap::new(), nodes: HashMap::new(),
+               }) }
+       }
+
+       /// Checks if there is a pending `channel_update` UTXO validation for the given channel,
+       /// and, if so, stores the channel message for handling later and returns an `Err`.
+       pub(super) fn check_hold_pending_channel_update(
+               &self, msg: &msgs::UnsignedChannelUpdate, full_msg: Option<&msgs::ChannelUpdate>
+       ) -> Result<(), LightningError> {
+               let mut pending_checks = self.internal.lock().unwrap();
+               if let hash_map::Entry::Occupied(e) = pending_checks.channels.entry(msg.short_channel_id) {
+                       let is_from_a = (msg.flags & 1) == 1;
+                       match Weak::upgrade(e.get()) {
+                               Some(msgs_ref) => {
+                                       let mut messages = msgs_ref.lock().unwrap();
+                                       let latest_update = if is_from_a {
+                                                       &mut messages.latest_channel_update_a
+                                               } else {
+                                                       &mut messages.latest_channel_update_b
+                                               };
+                                       if latest_update.is_none() || latest_update.as_ref().unwrap().timestamp() < msg.timestamp {
+                                               // If the messages we got has a higher timestamp, just blindly assume the
+                                               // signatures on the new message are correct and drop the old message. This
+                                               // may cause us to end up dropping valid `channel_update`s if a peer is
+                                               // malicious, but we should get the correct ones when the node updates them.
+                                               *latest_update = Some(
+                                                       if let Some(msg) = full_msg { ChannelUpdate::Full(msg.clone()) }
+                                                       else { ChannelUpdate::Unsigned(msg.clone()) });
+                                       }
+                                       return Err(LightningError {
+                                               err: "Awaiting channel_announcement validation to accept channel_update".to_owned(),
+                                               action: ErrorAction::IgnoreAndLog(Level::Gossip),
+                                       });
+                               },
+                               None => { e.remove(); },
+                       }
+               }
+               Ok(())
+       }
+
+       /// Checks if there is a pending `node_announcement` UTXO validation for a channel with the
+       /// given node and, if so, stores the channel message for handling later and returns an `Err`.
+       pub(super) fn check_hold_pending_node_announcement(
+               &self, msg: &msgs::UnsignedNodeAnnouncement, full_msg: Option<&msgs::NodeAnnouncement>
+       ) -> Result<(), LightningError> {
+               let mut pending_checks = self.internal.lock().unwrap();
+               if let hash_map::Entry::Occupied(mut e) = pending_checks.nodes.entry(msg.node_id) {
+                       let mut found_at_least_one_chan = false;
+                       e.get_mut().retain(|node_msgs| {
+                               match Weak::upgrade(&node_msgs) {
+                                       Some(chan_mtx) => {
+                                               let mut chan_msgs = chan_mtx.lock().unwrap();
+                                               if let Some(chan_announce) = &chan_msgs.channel_announce {
+                                                       let latest_announce =
+                                                               if *chan_announce.node_id_1() == msg.node_id {
+                                                                       &mut chan_msgs.latest_node_announce_a
+                                                               } else {
+                                                                       &mut chan_msgs.latest_node_announce_b
+                                                               };
+                                                       if latest_announce.is_none() ||
+                                                               latest_announce.as_ref().unwrap().timestamp() < msg.timestamp
+                                                       {
+                                                               // If the messages we got has a higher timestamp, just blindly
+                                                               // assume the signatures on the new message are correct and drop
+                                                               // the old message. This may cause us to end up dropping valid
+                                                               // `node_announcement`s if a peer is malicious, but we should get
+                                                               // the correct ones when the node updates them.
+                                                               *latest_announce = Some(
+                                                                       if let Some(msg) = full_msg { NodeAnnouncement::Full(msg.clone()) }
+                                                                       else { NodeAnnouncement::Unsigned(msg.clone()) });
+                                                       }
+                                                       found_at_least_one_chan = true;
+                                                       true
+                                               } else {
+                                                       debug_assert!(false, "channel_announce is set before struct is added to node map");
+                                                       false
+                                               }
+                                       },
+                                       None => false,
+                               }
+                       });
+                       if e.get().is_empty() { e.remove(); }
+                       if found_at_least_one_chan {
+                               return Err(LightningError {
+                                       err: "Awaiting channel_announcement validation to accept node_announcement".to_owned(),
+                                       action: ErrorAction::IgnoreAndLog(Level::Gossip),
+                               });
+                       }
+               }
+               Ok(())
+       }
+
+       fn check_replace_previous_entry(msg: &msgs::UnsignedChannelAnnouncement,
+               full_msg: Option<&msgs::ChannelAnnouncement>, replacement: Option<Weak<Mutex<UtxoMessages>>>,
+               pending_channels: &mut HashMap<u64, Weak<Mutex<UtxoMessages>>>
+       ) -> Result<(), msgs::LightningError> {
+               match pending_channels.entry(msg.short_channel_id) {
+                       hash_map::Entry::Occupied(mut e) => {
+                               // There's already a pending lookup for the given SCID. Check if the messages
+                               // are the same and, if so, return immediately (don't bother spawning another
+                               // lookup if we haven't gotten that far yet).
+                               match Weak::upgrade(&e.get()) {
+                                       Some(pending_msgs) => {
+                                               let pending_matches = match &pending_msgs.lock().unwrap().channel_announce {
+                                                       Some(ChannelAnnouncement::Full(pending_msg)) => Some(pending_msg) == full_msg,
+                                                       Some(ChannelAnnouncement::Unsigned(pending_msg)) => pending_msg == msg,
+                                                       None => {
+                                                               // This shouldn't actually be reachable. We set the
+                                                               // `channel_announce` field under the same lock as setting the
+                                                               // channel map entry. Still, we can just treat it as
+                                                               // non-matching and let the new request fly.
+                                                               debug_assert!(false);
+                                                               false
+                                                       },
+                                               };
+                                               if pending_matches {
+                                                       return Err(LightningError {
+                                                               err: "Channel announcement is already being checked".to_owned(),
+                                                               action: ErrorAction::IgnoreDuplicateGossip,
+                                                       });
+                                               } else {
+                                                       // The earlier lookup is a different message. If we have another
+                                                       // request in-flight now replace the original.
+                                                       // Note that in the replace case whether to replace is somewhat
+                                                       // arbitrary - both results will be handled, we're just updating the
+                                                       // value that will be compared to future lookups with the same SCID.
+                                                       if let Some(item) = replacement {
+                                                               *e.get_mut() = item;
+                                                       }
+                                               }
+                                       },
+                                       None => {
+                                               // The earlier lookup already resolved. We can't be sure its the same
+                                               // so just remove/replace it and move on.
+                                               if let Some(item) = replacement {
+                                                       *e.get_mut() = item;
+                                               } else { e.remove(); }
+                                       },
+                               }
+                       },
+                       hash_map::Entry::Vacant(v) => {
+                               if let Some(item) = replacement { v.insert(item); }
+                       },
+               }
+               Ok(())
+       }
+
+       pub(super) fn check_channel_announcement<U: Deref>(&self,
+               utxo_lookup: &Option<U>, msg: &msgs::UnsignedChannelAnnouncement,
+               full_msg: Option<&msgs::ChannelAnnouncement>
+       ) -> Result<Option<u64>, msgs::LightningError> where U::Target: UtxoLookup {
+               let handle_result = |res| {
+                       match res {
+                               Ok(TxOut { value, script_pubkey }) => {
+                                       let expected_script =
+                                               make_funding_redeemscript_from_slices(msg.bitcoin_key_1.as_slice(), msg.bitcoin_key_2.as_slice()).to_v0_p2wsh();
+                                       if script_pubkey != expected_script {
+                                               return Err(LightningError{
+                                                       err: format!("Channel announcement key ({}) didn't match on-chain script ({})",
+                                                               expected_script.to_hex(), script_pubkey.to_hex()),
+                                                       action: ErrorAction::IgnoreError
+                                               });
+                                       }
+                                       Ok(Some(value))
+                               },
+                               Err(UtxoLookupError::UnknownChain) => {
+                                       Err(LightningError {
+                                               err: format!("Channel announced on an unknown chain ({})",
+                                                       msg.chain_hash.encode().to_hex()),
+                                               action: ErrorAction::IgnoreError
+                                       })
+                               },
+                               Err(UtxoLookupError::UnknownTx) => {
+                                       Err(LightningError {
+                                               err: "Channel announced without corresponding UTXO entry".to_owned(),
+                                               action: ErrorAction::IgnoreError
+                                       })
+                               },
+                       }
+               };
+
+               Self::check_replace_previous_entry(msg, full_msg, None,
+                       &mut self.internal.lock().unwrap().channels)?;
+
+               match utxo_lookup {
+                       &None => {
+                               // Tentatively accept, potentially exposing us to DoS attacks
+                               Ok(None)
+                       },
+                       &Some(ref utxo_lookup) => {
+                               match utxo_lookup.get_utxo(&msg.chain_hash, msg.short_channel_id) {
+                                       UtxoResult::Sync(res) => handle_result(res),
+                                       UtxoResult::Async(future) => {
+                                               let mut pending_checks = self.internal.lock().unwrap();
+                                               let mut async_messages = future.state.lock().unwrap();
+                                               if let Some(res) = async_messages.complete.take() {
+                                                       // In the unlikely event the future resolved before we managed to get it,
+                                                       // handle the result in-line.
+                                                       handle_result(res)
+                                               } else {
+                                                       Self::check_replace_previous_entry(msg, full_msg,
+                                                               Some(Arc::downgrade(&future.state)), &mut pending_checks.channels)?;
+                                                       async_messages.channel_announce = Some(
+                                                               if let Some(msg) = full_msg { ChannelAnnouncement::Full(msg.clone()) }
+                                                               else { ChannelAnnouncement::Unsigned(msg.clone()) });
+                                                       pending_checks.nodes.entry(msg.node_id_1)
+                                                               .or_insert(Vec::new()).push(Arc::downgrade(&future.state));
+                                                       pending_checks.nodes.entry(msg.node_id_2)
+                                                               .or_insert(Vec::new()).push(Arc::downgrade(&future.state));
+                                                       Err(LightningError {
+                                                               err: "Channel being checked async".to_owned(),
+                                                               action: ErrorAction::IgnoreAndLog(Level::Gossip),
+                                                       })
+                                               }
+                                       },
+                               }
+                       }
+               }
+       }
+
+       /// The maximum number of pending gossip checks before [`Self::too_many_checks_pending`]
+       /// returns `true`. Note that this isn't a strict upper-bound on the number of checks pending -
+       /// each peer may, at a minimum, read one more socket buffer worth of `channel_announcement`s
+       /// which we'll have to process. With a socket buffer of 4KB and a minimum
+       /// `channel_announcement` size of, roughly, 429 bytes, this may leave us with `10*our peer
+       /// count` messages to process beyond this limit. Because we'll probably have a few peers,
+       /// there's no reason for this constant to be materially less than 30 or so, and 32 in-flight
+       /// checks should be more than enough for decent parallelism.
+       const MAX_PENDING_LOOKUPS: usize = 32;
+
+       /// Returns true if there are a large number of async checks pending and future
+       /// `channel_announcement` messages should be delayed. Note that this is only a hint and
+       /// messages already in-flight may still have to be handled for various reasons.
+       pub(super) fn too_many_checks_pending(&self) -> bool {
+               let mut pending_checks = self.internal.lock().unwrap();
+               if pending_checks.channels.len() > Self::MAX_PENDING_LOOKUPS {
+                       // If we have many channel checks pending, ensure we don't have any dangling checks
+                       // (i.e. checks where the user told us they'd call back but drop'd the `AccessFuture`
+                       // instead) before we commit to applying backpressure.
+                       pending_checks.channels.retain(|_, chan| {
+                               Weak::upgrade(&chan).is_some()
+                       });
+                       pending_checks.nodes.retain(|_, channels| {
+                               channels.retain(|chan| Weak::upgrade(&chan).is_some());
+                               !channels.is_empty()
+                       });
+                       pending_checks.channels.len() > Self::MAX_PENDING_LOOKUPS
+               } else {
+                       false
+               }
+       }
 }