Relay DummyOrdered messages (#318)

* DummyOrdered relay: initial commit

* 1 ML file -> 3 files

* extracted generic message race loop

* uncommented race error procesing

* lane loop tests

* cargo fmt

* moved HeaderId to utils.rs

* restart lane loop on stall

* message delivery strategy tests

* removed obsolete code

* clippy

* Update relays/ethereum/src/message_lane_loop.rs

Co-authored-by: Tomasz Drwięga <tomusdrw@users.noreply.github.com>

* Update relays/ethereum/src/message_lane_loop.rs

Co-authored-by: Tomasz Drwięga <tomusdrw@users.noreply.github.com>

* added more races docs

* Update relays/ethereum/src/message_race_delivery.rs

Co-authored-by: Tomasz Drwięga <tomusdrw@users.noreply.github.com>

* update docs

* Update relays/ethereum/src/message_race_loop.rs

Co-authored-by: Tomasz Drwięga <tomusdrw@users.noreply.github.com>

* deal with TODOs

* removed docs

* docs

* cargo fmt --all

* Update relays/ethereum/src/message_race_loop.rs

Co-authored-by: Hernando Castano <HCastano@users.noreply.github.com>

Co-authored-by: Tomasz Drwięga <tomusdrw@users.noreply.github.com>
Co-authored-by: Hernando Castano <HCastano@users.noreply.github.com>

bridges/relays/ethereum/src/ethereum_client.rs

@@ -21,8 +21,8 @@ use crate::ethereum_types::{
 use crate::rpc::{Ethereum, EthereumRpc};
 use crate::rpc_errors::{EthereumNodeError, RpcError};
 use crate::substrate_types::{GrandpaJustification, Hash as SubstrateHash, QueuedSubstrateHeader, SubstrateHeaderId};
-use crate::sync_types::{HeaderId, SubmittedHeaders};
+use crate::sync_types::SubmittedHeaders;
-use crate::utils::MaybeConnectionError;
+use crate::utils::{HeaderId, MaybeConnectionError};
 
 use async_trait::async_trait;
 use codec::{Decode, Encode};

bridges/relays/ethereum/src/ethereum_deploy_contract.rs

@@ -21,7 +21,7 @@ use crate::instances::BridgeInstance;
 use crate::rpc::SubstrateRpc;
 use crate::substrate_client::{SubstrateConnectionParams, SubstrateRpcClient};
 use crate::substrate_types::{Hash as SubstrateHash, Header as SubstrateHeader, SubstrateHeaderId};
-use crate::sync_types::HeaderId;
+use crate::utils::HeaderId;
 
 use codec::{Decode, Encode};
 use num_traits::Zero;

bridges/relays/ethereum/src/ethereum_exchange.rs

@@ -34,7 +34,7 @@ use crate::substrate_client::{
 	SubmitEthereumExchangeTransactionProof, SubstrateConnectionParams, SubstrateRpcClient, SubstrateSigningParams,
 };
 use crate::substrate_types::into_substrate_ethereum_receipt;
-use crate::sync_types::HeaderId;
+use crate::utils::HeaderId;
 
 use async_trait::async_trait;
 use bp_currency_exchange::MaybeLockFundsTransaction;

bridges/relays/ethereum/src/ethereum_types.rs

@@ -15,7 +15,8 @@
 // along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
 
 use crate::substrate_types::{into_substrate_ethereum_header, into_substrate_ethereum_receipts};
-use crate::sync_types::{HeaderId, HeadersSyncPipeline, QueuedHeader, SourceHeader};
+use crate::sync_types::{HeadersSyncPipeline, QueuedHeader, SourceHeader};
+use crate::utils::HeaderId;
 use codec::Encode;
 
 pub use web3::types::{Address, Bytes, CallRequest, H256, U128, U256, U64};

bridges/relays/ethereum/src/exchange.rs

@@ -54,7 +54,7 @@ pub trait SourceBlock {
 	type Transaction: SourceTransaction;
 
 	/// Return hash of the block.
-	fn id(&self) -> crate::sync_types::HeaderId<Self::Hash, Self::Number>;
+	fn id(&self) -> crate::utils::HeaderId<Self::Hash, Self::Number>;
 	/// Return block transactions iterator.
 	fn transactions(&self) -> Vec<Self::Transaction>;
 }
@@ -81,7 +81,7 @@ pub type TransactionOf<P> = <<P as TransactionProofPipeline>::Block as SourceBlo
 pub type TransactionHashOf<P> = <TransactionOf<P> as SourceTransaction>::Hash;
 
 /// Header id.
-pub type HeaderId<P> = crate::sync_types::HeaderId<BlockHashOf<P>, BlockNumberOf<P>>;
+pub type HeaderId<P> = crate::utils::HeaderId<BlockHashOf<P>, BlockNumberOf<P>>;
 
 /// Source client API.
 #[async_trait]
@@ -443,7 +443,7 @@ async fn wait_header_finalized<P: TransactionProofPipeline>(
 #[cfg(test)]
 pub(crate) mod tests {
 	use super::*;
-	use crate::sync_types::HeaderId;
+	use crate::utils::HeaderId;
 
 	use parking_lot::Mutex;
 	use std::{

bridges/relays/ethereum/src/headers.rs

@@ -14,7 +14,9 @@
 // You should have received a copy of the GNU General Public License
 // along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
 
-use crate::sync_types::{HeaderId, HeaderIdOf, HeaderStatus, HeadersSyncPipeline, QueuedHeader, SourceHeader};
+use crate::sync_types::{HeaderIdOf, HeaderStatus, HeadersSyncPipeline, QueuedHeader, SourceHeader};
+use crate::utils::HeaderId;
+
 use linked_hash_map::LinkedHashMap;
 use num_traits::{One, Zero};
 use std::{
@@ -777,7 +779,7 @@ fn queued_incomplete_header<Id: Clone + Eq + std::hash::Hash, T>(
 pub(crate) mod tests {
 	use super::*;
 	use crate::ethereum_types::{EthereumHeaderId, EthereumHeadersSyncPipeline, Header, H256};
-	use crate::sync_types::{HeaderId, QueuedHeader};
+	use crate::sync_types::QueuedHeader;
 
 	pub(crate) fn header(number: u64) -> QueuedHeader<EthereumHeadersSyncPipeline> {
 		QueuedHeader::new(Header {

bridges/relays/ethereum/src/main.rs

@@ -27,6 +27,10 @@ mod exchange_loop;
 mod exchange_loop_metrics;
 mod headers;
 mod instances;
+mod message_lane;
+mod message_lane_loop;
+mod message_race_delivery;
+mod message_race_loop;
 mod metrics;
 mod rpc;
 mod rpc_errors;

bridges/relays/ethereum/src/message_lane.rs

@@ -0,0 +1,55 @@
+// Copyright 2019-2020 Parity Technologies (UK) Ltd.
+// This file is part of Parity Bridges Common.
+
+// Parity Bridges Common is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+
+// Parity Bridges Common is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
+
+//! One-way message lane types. Within single one-way lane we have three 'races' where we try to:
+//!
+//! 1) relay new messages from source to target node;
+//! 2) relay proof-of-receiving from target to source node;
+//! 3) relay proof-of-processing from target no source node.
+
+use crate::utils::HeaderId;
+
+use std::fmt::Debug;
+
+/// One-way message lane.
+pub trait MessageLane {
+	/// Name of the messages source.
+	const SOURCE_NAME: &'static str;
+	/// Name of the messages target.
+	const TARGET_NAME: &'static str;
+
+	/// Message nonce type.
+	type MessageNonce: Clone + Copy + Debug + Default + From<u32> + Ord + std::ops::Add<Output = Self::MessageNonce>;
+
+	/// Messages proof.
+	type MessagesProof: Clone;
+
+	/// Number of the source header.
+	type SourceHeaderNumber: Clone + Debug + Default + Ord + PartialEq;
+	/// Hash of the source header.
+	type SourceHeaderHash: Clone + Debug + Default + PartialEq;
+
+	/// Number of the target header.
+	type TargetHeaderNumber: Clone + Debug + Default + Ord + PartialEq;
+	/// Hash of the target header.
+	type TargetHeaderHash: Clone + Debug + Default + PartialEq;
+}
+
+/// Source header id within given one-way message lane.
+pub type SourceHeaderIdOf<P> = HeaderId<<P as MessageLane>::SourceHeaderHash, <P as MessageLane>::SourceHeaderNumber>;
+
+/// Target header id within given one-way message lane.
+pub type TargetHeaderIdOf<P> = HeaderId<<P as MessageLane>::TargetHeaderHash, <P as MessageLane>::TargetHeaderNumber>;

bridges/relays/ethereum/src/message_lane_loop.rs

@@ -0,0 +1,591 @@
+// Copyright 2019-2020 Parity Technologies (UK) Ltd.
+// This file is part of Parity Bridges Common.
+
+// Parity Bridges Common is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+
+// Parity Bridges Common is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
+
+//! Message delivery loop. Designed to work with message-lane pallet.
+//!
+//! Single relay instance delivers messages of single lane in single direction.
+//! To serve two-way lane, you would need two instances of relay.
+//! To serve N two-way lanes, you would need N*2 instances of relay.
+//!
+//! Please keep in mind that the best header in this file is actually best
+//! finalized header. I.e. when talking about headers in lane context, we
+//! only care about finalized headers.
+
+// Until there'll be actual message-lane in the runtime.
+#![allow(dead_code)]
+
+use crate::message_lane::{MessageLane, SourceHeaderIdOf, TargetHeaderIdOf};
+use crate::message_race_delivery::run as run_message_delivery_race;
+use crate::utils::{interval, process_future_result, retry_backoff, FailedClient, MaybeConnectionError};
+
+use async_trait::async_trait;
+use futures::{channel::mpsc::unbounded, future::FutureExt, stream::StreamExt};
+use std::{fmt::Debug, future::Future, ops::RangeInclusive, time::Duration};
+
+/// Source client trait.
+#[async_trait(?Send)]
+pub trait SourceClient<P: MessageLane>: Clone {
+	/// Type of error this clients returns.
+	type Error: std::fmt::Debug + MaybeConnectionError;
+
+	/// Try to reconnect to source node.
+	fn reconnect(self) -> Self;
+
+	/// Returns state of the client.
+	async fn state(&self) -> Result<SourceClientState<P>, Self::Error>;
+
+	/// Get nonce of instance of latest generated message.
+	async fn latest_generated_nonce(
+		&self,
+		id: SourceHeaderIdOf<P>,
+	) -> Result<(SourceHeaderIdOf<P>, P::MessageNonce), Self::Error>;
+
+	/// Prove messages in inclusive range [begin; end].
+	async fn prove_messages(
+		&self,
+		id: SourceHeaderIdOf<P>,
+		nonces: RangeInclusive<P::MessageNonce>,
+	) -> Result<(SourceHeaderIdOf<P>, RangeInclusive<P::MessageNonce>, P::MessagesProof), Self::Error>;
+}
+
+/// Target client trait.
+#[async_trait(?Send)]
+pub trait TargetClient<P: MessageLane>: Clone {
+	/// Type of error this clients returns.
+	type Error: std::fmt::Debug + MaybeConnectionError;
+
+	/// Try to reconnect to source node.
+	fn reconnect(self) -> Self;
+
+	/// Returns state of the client.
+	async fn state(&self) -> Result<TargetClientState<P>, Self::Error>;
+
+	/// Get nonce of latest message, which receival has been confirmed.
+	async fn latest_received_nonce(
+		&self,
+		id: TargetHeaderIdOf<P>,
+	) -> Result<(TargetHeaderIdOf<P>, P::MessageNonce), Self::Error>;
+
+	/// Submit messages proof.
+	async fn submit_messages_proof(
+		&self,
+		generated_at_header: SourceHeaderIdOf<P>,
+		nonces: RangeInclusive<P::MessageNonce>,
+		proof: P::MessagesProof,
+	) -> Result<RangeInclusive<P::MessageNonce>, Self::Error>;
+}
+
+/// State of the client.
+#[derive(Clone, Debug, Default, PartialEq)]
+pub struct ClientState<SelfHeaderId, PeerHeaderId> {
+	/// Best header id of this chain.
+	pub best_self: SelfHeaderId,
+	/// Best header id of the peer chain.
+	pub best_peer: PeerHeaderId,
+}
+
+/// State of source client in one-way message lane.
+pub type SourceClientState<P> = ClientState<SourceHeaderIdOf<P>, TargetHeaderIdOf<P>>;
+
+/// State of target client in one-way message lane.
+pub type TargetClientState<P> = ClientState<TargetHeaderIdOf<P>, SourceHeaderIdOf<P>>;
+
+/// Both clients state.
+#[derive(Debug, Default)]
+pub struct ClientsState<P: MessageLane> {
+	/// Source client state.
+	pub source: Option<SourceClientState<P>>,
+	/// Target client state.
+	pub target: Option<TargetClientState<P>>,
+}
+
+/// Run message lane service loop.
+pub fn run<P: MessageLane>(
+	mut source_client: impl SourceClient<P>,
+	source_tick: Duration,
+	mut target_client: impl TargetClient<P>,
+	target_tick: Duration,
+	reconnect_delay: Duration,
+	stall_timeout: Duration,
+	exit_signal: impl Future<Output = ()>,
+) {
+	let mut local_pool = futures::executor::LocalPool::new();
+	let exit_signal = exit_signal.shared();
+
+	local_pool.run_until(async move {
+		loop {
+			let result = run_until_connection_lost(
+				source_client.clone(),
+				source_tick,
+				target_client.clone(),
+				target_tick,
+				stall_timeout,
+				exit_signal.clone(),
+			)
+			.await;
+
+			match result {
+				Ok(()) => break,
+				Err(failed_client) => {
+					async_std::task::sleep(reconnect_delay).await;
+					if failed_client == FailedClient::Both || failed_client == FailedClient::Source {
+						source_client = source_client.reconnect();
+					}
+					if failed_client == FailedClient::Both || failed_client == FailedClient::Target {
+						target_client = target_client.reconnect();
+					}
+				}
+			}
+
+			log::debug!(
+				target: "bridge",
+				"Restarting lane {} -> {}",
+				P::SOURCE_NAME,
+				P::TARGET_NAME,
+			);
+		}
+	});
+}
+
+/// Run one-way message delivery loop until connection with target or source node is lost, or exit signal is received.
+async fn run_until_connection_lost<P: MessageLane, SC: SourceClient<P>, TC: TargetClient<P>>(
+	source_client: SC,
+	source_tick: Duration,
+	target_client: TC,
+	target_tick: Duration,
+	stall_timeout: Duration,
+	exit_signal: impl Future<Output = ()>,
+) -> Result<(), FailedClient> {
+	let mut source_retry_backoff = retry_backoff();
+	let mut source_client_is_online = false;
+	let mut source_state_required = true;
+	let source_state = source_client.state().fuse();
+	let source_go_offline_future = futures::future::Fuse::terminated();
+	let source_tick_stream = interval(source_tick).fuse();
+
+	let mut target_retry_backoff = retry_backoff();
+	let mut target_client_is_online = false;
+	let mut target_state_required = true;
+	let target_state = target_client.state().fuse();
+	let target_go_offline_future = futures::future::Fuse::terminated();
+	let target_tick_stream = interval(target_tick).fuse();
+
+	let (
+		(delivery_source_state_sender, delivery_source_state_receiver),
+		(delivery_target_state_sender, delivery_target_state_receiver),
+	) = (unbounded(), unbounded());
+	let delivery_race_loop = run_message_delivery_race(
+		source_client.clone(),
+		delivery_source_state_receiver,
+		target_client.clone(),
+		delivery_target_state_receiver,
+		stall_timeout,
+	)
+	.fuse();
+
+	let exit_signal = exit_signal.fuse();
+
+	futures::pin_mut!(
+		source_state,
+		source_go_offline_future,
+		source_tick_stream,
+		target_state,
+		target_go_offline_future,
+		target_tick_stream,
+		delivery_race_loop,
+		exit_signal
+	);
+
+	loop {
+		futures::select! {
+			new_source_state = source_state => {
+				source_state_required = false;
+
+				source_client_is_online = process_future_result(
+					new_source_state,
+					&mut source_retry_backoff,
+					|new_source_state| {
+						log::debug!(
+							target: "bridge",
+							"Received state from {} node: {:?}",
+							P::SOURCE_NAME,
+							new_source_state,
+						);
+						let _ = delivery_source_state_sender.unbounded_send(new_source_state);
+					},
+					&mut source_go_offline_future,
+					|delay| async_std::task::sleep(delay),
+					|| format!("Error retrieving state from {} node", P::SOURCE_NAME),
+				).fail_if_connection_error(FailedClient::Source)?;
+			},
+			_ = source_go_offline_future => {
+				source_client_is_online = true;
+			},
+			_ = source_tick_stream.next() => {
+				source_state_required = true;
+			},
+			new_target_state = target_state => {
+				target_state_required = false;
+
+				target_client_is_online = process_future_result(
+					new_target_state,
+					&mut target_retry_backoff,
+					|new_target_state| {
+						log::debug!(
+							target: "bridge",
+							"Received state from {} node: {:?}",
+							P::TARGET_NAME,
+							new_target_state,
+						);
+						let _ = delivery_target_state_sender.unbounded_send(new_target_state);
+					},
+					&mut target_go_offline_future,
+					|delay| async_std::task::sleep(delay),
+					|| format!("Error retrieving state from {} node", P::TARGET_NAME),
+				).fail_if_connection_error(FailedClient::Target)?;
+			},
+			_ = target_go_offline_future => {
+				target_client_is_online = true;
+			},
+			_ = target_tick_stream.next() => {
+				target_state_required = true;
+			},
+
+			delivery_error = delivery_race_loop => {
+				match delivery_error {
+					Ok(_) => unreachable!("only ends with error; qed"),
+					Err(err) => return Err(err),
+				}
+			},
+
+			() = exit_signal => {
+				return Ok(());
+			}
+		}
+
+		if source_client_is_online && source_state_required {
+			log::debug!(target: "bridge", "Asking {} node about its state", P::SOURCE_NAME);
+			source_state.set(source_client.state().fuse());
+			source_client_is_online = false;
+		}
+
+		if target_client_is_online && target_state_required {
+			log::debug!(target: "bridge", "Asking {} node about its state", P::TARGET_NAME);
+			target_state.set(target_client.state().fuse());
+			target_client_is_online = false;
+		}
+	}
+}
+
+#[cfg(test)]
+pub(crate) mod tests {
+	use super::*;
+	use crate::utils::HeaderId;
+	use futures::stream::StreamExt;
+	use parking_lot::Mutex;
+	use std::sync::Arc;
+
+	pub fn header_id(number: TestSourceHeaderNumber) -> HeaderId<TestSourceHeaderNumber, TestSourceHeaderHash> {
+		HeaderId(number, number)
+	}
+
+	pub type TestMessageNonce = u64;
+	pub type TestMessagesProof = RangeInclusive<TestMessageNonce>;
+
+	pub type TestSourceHeaderNumber = u64;
+	pub type TestSourceHeaderHash = u64;
+
+	pub type TestTargetHeaderNumber = u64;
+	pub type TestTargetHeaderHash = u64;
+
+	#[derive(Debug)]
+	pub enum TestError {
+		Logic,
+		Connection,
+	}
+
+	impl MaybeConnectionError for TestError {
+		fn is_connection_error(&self) -> bool {
+			match *self {
+				TestError::Logic => false,
+				TestError::Connection => true,
+			}
+		}
+	}
+
+	pub struct TestMessageLane;
+
+	impl MessageLane for TestMessageLane {
+		const SOURCE_NAME: &'static str = "TestSource";
+		const TARGET_NAME: &'static str = "TestTarget";
+
+		type MessageNonce = TestMessageNonce;
+		type MessagesProof = TestMessagesProof;
+
+		type SourceHeaderNumber = TestSourceHeaderNumber;
+		type SourceHeaderHash = TestSourceHeaderHash;
+
+		type TargetHeaderNumber = TestTargetHeaderNumber;
+		type TargetHeaderHash = TestTargetHeaderHash;
+	}
+
+	#[derive(Debug, Default, Clone)]
+	pub struct TestClientData {
+		is_source_fails: bool,
+		is_source_reconnected: bool,
+		source_state: SourceClientState<TestMessageLane>,
+		source_latest_generated_nonce: TestMessageNonce,
+		is_target_fails: bool,
+		is_target_reconnected: bool,
+		target_state: SourceClientState<TestMessageLane>,
+		target_latest_received_nonce: TestMessageNonce,
+		submitted_messages_proofs: Vec<TestMessagesProof>,
+	}
+
+	#[derive(Clone)]
+	pub struct TestSourceClient {
+		data: Arc<Mutex<TestClientData>>,
+		tick: Arc<dyn Fn(&mut TestClientData)>,
+	}
+
+	#[async_trait(?Send)]
+	impl SourceClient<TestMessageLane> for TestSourceClient {
+		type Error = TestError;
+
+		fn reconnect(self) -> Self {
+			{
+				let mut data = self.data.lock();
+				(self.tick)(&mut *data);
+				data.is_source_reconnected = true;
+			}
+			self
+		}
+
+		async fn state(&self) -> Result<SourceClientState<TestMessageLane>, Self::Error> {
+			let mut data = self.data.lock();
+			(self.tick)(&mut *data);
+			if data.is_source_fails {
+				return Err(TestError::Connection);
+			}
+			Ok(data.source_state.clone())
+		}
+
+		/// Get nonce of instance of latest generated message.
+		async fn latest_generated_nonce(
+			&self,
+			id: SourceHeaderIdOf<TestMessageLane>,
+		) -> Result<(SourceHeaderIdOf<TestMessageLane>, TestMessageNonce), Self::Error> {
+			let mut data = self.data.lock();
+			(self.tick)(&mut *data);
+			if data.is_source_fails {
+				return Err(TestError::Connection);
+			}
+			Ok((id, data.source_latest_generated_nonce))
+		}
+
+		async fn prove_messages(
+			&self,
+			id: SourceHeaderIdOf<TestMessageLane>,
+			nonces: RangeInclusive<TestMessageNonce>,
+		) -> Result<
+			(
+				SourceHeaderIdOf<TestMessageLane>,
+				RangeInclusive<TestMessageNonce>,
+				TestMessagesProof,
+			),
+			Self::Error,
+		> {
+			Ok((id, nonces.clone(), nonces))
+		}
+	}
+
+	#[derive(Clone)]
+	pub struct TestTargetClient {
+		data: Arc<Mutex<TestClientData>>,
+		tick: Arc<dyn Fn(&mut TestClientData)>,
+	}
+
+	#[async_trait(?Send)]
+	impl TargetClient<TestMessageLane> for TestTargetClient {
+		type Error = TestError;
+
+		fn reconnect(self) -> Self {
+			{
+				let mut data = self.data.lock();
+				(self.tick)(&mut *data);
+				data.is_target_reconnected = true;
+			}
+			self
+		}
+
+		async fn state(&self) -> Result<TargetClientState<TestMessageLane>, Self::Error> {
+			let mut data = self.data.lock();
+			(self.tick)(&mut *data);
+			if data.is_target_fails {
+				return Err(TestError::Connection);
+			}
+			Ok(data.target_state.clone())
+		}
+
+		async fn latest_received_nonce(
+			&self,
+			id: TargetHeaderIdOf<TestMessageLane>,
+		) -> Result<(TargetHeaderIdOf<TestMessageLane>, TestMessageNonce), Self::Error> {
+			let mut data = self.data.lock();
+			(self.tick)(&mut *data);
+			if data.is_target_fails {
+				return Err(TestError::Connection);
+			}
+			Ok((id, data.target_latest_received_nonce))
+		}
+
+		/// Submit messages proof.
+		async fn submit_messages_proof(
+			&self,
+			_generated_at_header: SourceHeaderIdOf<TestMessageLane>,
+			nonces: RangeInclusive<TestMessageNonce>,
+			proof: TestMessagesProof,
+		) -> Result<RangeInclusive<TestMessageNonce>, Self::Error> {
+			let mut data = self.data.lock();
+			(self.tick)(&mut *data);
+			if data.is_target_fails {
+				return Err(TestError::Connection);
+			}
+			data.target_state.best_self =
+				HeaderId(data.target_state.best_self.0 + 1, data.target_state.best_self.1 + 1);
+			data.target_latest_received_nonce = *proof.end();
+			data.submitted_messages_proofs.push(proof);
+			Ok(nonces)
+		}
+	}
+
+	fn run_loop_test(
+		data: TestClientData,
+		source_tick: Arc<dyn Fn(&mut TestClientData)>,
+		target_tick: Arc<dyn Fn(&mut TestClientData)>,
+		exit_signal: impl Future<Output = ()>,
+	) -> TestClientData {
+		async_std::task::block_on(async {
+			let data = Arc::new(Mutex::new(data));
+
+			let source_client = TestSourceClient {
+				data: data.clone(),
+				tick: source_tick,
+			};
+			let target_client = TestTargetClient {
+				data: data.clone(),
+				tick: target_tick,
+			};
+			run(
+				source_client,
+				Duration::from_millis(100),
+				target_client,
+				Duration::from_millis(100),
+				Duration::from_millis(0),
+				Duration::from_secs(60),
+				exit_signal,
+			);
+
+			let result = data.lock().clone();
+			result
+		})
+	}
+
+	#[test]
+	fn message_lane_loop_is_able_to_recover_from_connection_errors() {
+		// with this configuration, source client will return Err, making source client
+		// reconnect. Then the target client will fail with Err + reconnect. Then we finally
+		// able to deliver messages.
+		let (exit_sender, exit_receiver) = unbounded();
+		let result = run_loop_test(
+			TestClientData {
+				is_source_fails: true,
+				source_state: ClientState {
+					best_self: HeaderId(0, 0),
+					best_peer: HeaderId(0, 0),
+				},
+				source_latest_generated_nonce: 1,
+				target_state: ClientState {
+					best_self: HeaderId(0, 0),
+					best_peer: HeaderId(0, 0),
+				},
+				target_latest_received_nonce: 0,
+				..Default::default()
+			},
+			Arc::new(|data: &mut TestClientData| {
+				if data.is_source_reconnected {
+					data.is_source_fails = false;
+					data.is_target_fails = true;
+				}
+			}),
+			Arc::new(move |data: &mut TestClientData| {
+				if data.is_target_reconnected {
+					data.is_target_fails = false;
+				}
+				if data.target_state.best_peer.0 < 10 {
+					data.target_state.best_peer =
+						HeaderId(data.target_state.best_peer.0 + 1, data.target_state.best_peer.0 + 1);
+				}
+				if !data.submitted_messages_proofs.is_empty() {
+					exit_sender.unbounded_send(()).unwrap();
+				}
+			}),
+			exit_receiver.into_future().map(|(_, _)| ()),
+		);
+
+		assert_eq!(result.submitted_messages_proofs, vec![1..=1],);
+	}
+
+	#[test]
+	fn message_lane_loop_works() {
+		// with this configuration, target client must first sync headers [1; 10] and
+		// then submit proof-of-messages [0; 10] at once
+		let (exit_sender, exit_receiver) = unbounded();
+		let result = run_loop_test(
+			TestClientData {
+				source_state: ClientState {
+					best_self: HeaderId(10, 10),
+					best_peer: HeaderId(0, 0),
+				},
+				source_latest_generated_nonce: 10,
+				target_state: ClientState {
+					best_self: HeaderId(0, 0),
+					best_peer: HeaderId(0, 0),
+				},
+				target_latest_received_nonce: 0,
+				..Default::default()
+			},
+			Arc::new(|_: &mut TestClientData| {}),
+			Arc::new(move |data: &mut TestClientData| {
+				if data.target_state.best_peer.0 < 10 {
+					data.target_state.best_peer =
+						HeaderId(data.target_state.best_peer.0 + 1, data.target_state.best_peer.0 + 1);
+				}
+				if data
+					.submitted_messages_proofs
+					.last()
+					.map(|last| *last.end() == 10)
+					.unwrap_or(false)
+				{
+					exit_sender.unbounded_send(()).unwrap();
+				}
+			}),
+			exit_receiver.into_future().map(|(_, _)| ()),
+		);
+
+		assert_eq!(result.submitted_messages_proofs, vec![1..=4, 5..=8, 9..=10],);
+	}
+}

bridges/relays/ethereum/src/message_race_delivery.rs

@@ -0,0 +1,385 @@
+// Copyright 2019-2020 Parity Technologies (UK) Ltd.
+// This file is part of Parity Bridges Common.
+
+// Parity Bridges Common is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+
+// Parity Bridges Common is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+//! Message delivery race delivers proof-of-messages from lane.source to lane.target.
+
+use crate::message_lane::{MessageLane, SourceHeaderIdOf, TargetHeaderIdOf};
+use crate::message_lane_loop::{
+	SourceClient as MessageLaneSourceClient, SourceClientState, TargetClient as MessageLaneTargetClient,
+	TargetClientState,
+};
+use crate::message_race_loop::{MessageRace, RaceState, RaceStrategy, SourceClient, TargetClient};
+use crate::utils::FailedClient;
+
+use async_trait::async_trait;
+use futures::stream::FusedStream;
+use std::{collections::VecDeque, marker::PhantomData, ops::RangeInclusive, time::Duration};
+
+/// Maximal number of messages to relay in single transaction.
+const MAX_MESSAGES_TO_RELAY_IN_SINGLE_TX: u32 = 4;
+
+/// Run message delivery race.
+pub async fn run<P: MessageLane>(
+	source_client: impl MessageLaneSourceClient<P>,
+	source_state_updates: impl FusedStream<Item = SourceClientState<P>>,
+	target_client: impl MessageLaneTargetClient<P>,
+	target_state_updates: impl FusedStream<Item = TargetClientState<P>>,
+	stall_timeout: Duration,
+) -> Result<(), FailedClient> {
+	crate::message_race_loop::run(
+		MessageDeliveryRaceSource {
+			client: source_client,
+			_phantom: Default::default(),
+		},
+		source_state_updates,
+		MessageDeliveryRaceTarget {
+			client: target_client,
+			_phantom: Default::default(),
+		},
+		target_state_updates,
+		stall_timeout,
+		MessageDeliveryStrategy::<P>::default(),
+	)
+	.await
+}
+
+/// Message delivery race.
+struct MessageDeliveryRace<P>(std::marker::PhantomData<P>);
+
+impl<P: MessageLane> MessageRace for MessageDeliveryRace<P> {
+	type SourceHeaderId = SourceHeaderIdOf<P>;
+	type TargetHeaderId = TargetHeaderIdOf<P>;
+
+	type MessageNonce = P::MessageNonce;
+	type Proof = P::MessagesProof;
+
+	fn source_name() -> String {
+		format!("{}::MessagesDelivery", P::SOURCE_NAME)
+	}
+
+	fn target_name() -> String {
+		format!("{}::MessagesDelivery", P::TARGET_NAME)
+	}
+}
+
+/// Message delivery race source, which is a source of the lane.
+struct MessageDeliveryRaceSource<P: MessageLane, C> {
+	client: C,
+	_phantom: PhantomData<P>,
+}
+
+#[async_trait(?Send)]
+impl<P, C> SourceClient<MessageDeliveryRace<P>> for MessageDeliveryRaceSource<P, C>
+where
+	P: MessageLane,
+	C: MessageLaneSourceClient<P>,
+{
+	type Error = C::Error;
+
+	async fn latest_nonce(
+		&self,
+		at_block: SourceHeaderIdOf<P>,
+	) -> Result<(SourceHeaderIdOf<P>, P::MessageNonce), Self::Error> {
+		self.client.latest_generated_nonce(at_block).await
+	}
+
+	async fn generate_proof(
+		&self,
+		at_block: SourceHeaderIdOf<P>,
+		nonces: RangeInclusive<P::MessageNonce>,
+	) -> Result<(SourceHeaderIdOf<P>, RangeInclusive<P::MessageNonce>, P::MessagesProof), Self::Error> {
+		self.client.prove_messages(at_block, nonces).await
+	}
+}
+
+/// Message delivery race target, which is a target of the lane.
+struct MessageDeliveryRaceTarget<P: MessageLane, C> {
+	client: C,
+	_phantom: PhantomData<P>,
+}
+
+#[async_trait(?Send)]
+impl<P, C> TargetClient<MessageDeliveryRace<P>> for MessageDeliveryRaceTarget<P, C>
+where
+	P: MessageLane,
+	C: MessageLaneTargetClient<P>,
+{
+	type Error = C::Error;
+
+	async fn latest_nonce(
+		&self,
+		at_block: TargetHeaderIdOf<P>,
+	) -> Result<(TargetHeaderIdOf<P>, P::MessageNonce), Self::Error> {
+		self.client.latest_received_nonce(at_block).await
+	}
+
+	async fn submit_proof(
+		&self,
+		generated_at_block: SourceHeaderIdOf<P>,
+		nonces: RangeInclusive<P::MessageNonce>,
+		proof: P::MessagesProof,
+	) -> Result<RangeInclusive<P::MessageNonce>, Self::Error> {
+		self.client
+			.submit_messages_proof(generated_at_block, nonces, proof)
+			.await
+	}
+}
+
+/// Message delivery strategy.
+struct MessageDeliveryStrategy<P: MessageLane> {
+	/// All queued nonces.
+	source_queue: VecDeque<(SourceHeaderIdOf<P>, P::MessageNonce)>,
+	/// Best nonce known to target node.
+	target_nonce: P::MessageNonce,
+	/// Unused generic types dump.
+	_phantom: PhantomData<P>,
+}
+
+impl<P: MessageLane> Default for MessageDeliveryStrategy<P> {
+	fn default() -> Self {
+		MessageDeliveryStrategy {
+			source_queue: VecDeque::new(),
+			target_nonce: Default::default(),
+			_phantom: Default::default(),
+		}
+	}
+}
+
+impl<P: MessageLane> RaceStrategy<SourceHeaderIdOf<P>, TargetHeaderIdOf<P>, P::MessageNonce, P::MessagesProof>
+	for MessageDeliveryStrategy<P>
+{
+	fn is_empty(&self) -> bool {
+		self.source_queue.is_empty()
+	}
+
+	fn source_nonce_updated(&mut self, at_block: SourceHeaderIdOf<P>, nonce: P::MessageNonce) {
+		if nonce <= self.target_nonce {
+			return;
+		}
+
+		match self.source_queue.back() {
+			Some((_, prev_nonce)) if *prev_nonce < nonce => (),
+			Some(_) => return,
+			None => (),
+		}
+
+		self.source_queue.push_back((at_block, nonce))
+	}
+
+	fn target_nonce_updated(
+		&mut self,
+		nonce: P::MessageNonce,
+		race_state: &mut RaceState<SourceHeaderIdOf<P>, TargetHeaderIdOf<P>, P::MessageNonce, P::MessagesProof>,
+	) {
+		if nonce < self.target_nonce {
+			return;
+		}
+
+		while let Some(true) = self
+			.source_queue
+			.front()
+			.map(|(_, source_nonce)| *source_nonce <= nonce)
+		{
+			self.source_queue.pop_front();
+		}
+
+		let need_to_select_new_nonces = race_state
+			.nonces_to_submit
+			.as_ref()
+			.map(|(_, nonces, _)| *nonces.end() <= nonce)
+			.unwrap_or(false);
+		if need_to_select_new_nonces {
+			race_state.nonces_to_submit = None;
+		}
+
+		let need_new_nonces_to_submit = race_state
+			.nonces_submitted
+			.as_ref()
+			.map(|nonces| *nonces.end() <= nonce)
+			.unwrap_or(false);
+		if need_new_nonces_to_submit {
+			race_state.nonces_submitted = None;
+		}
+
+		self.target_nonce = nonce;
+	}
+
+	fn select_nonces_to_deliver(
+		&mut self,
+		race_state: &RaceState<SourceHeaderIdOf<P>, TargetHeaderIdOf<P>, P::MessageNonce, P::MessagesProof>,
+	) -> Option<RangeInclusive<P::MessageNonce>> {
+		// if we have already selected nonces that we want to submit, do nothing
+		if race_state.nonces_to_submit.is_some() {
+			return None;
+		}
+
+		// if we already submitted some nonces, do nothing
+		if race_state.nonces_submitted.is_some() {
+			return None;
+		}
+
+		// 1) we want to deliver all nonces, starting from `target_nonce + 1`
+		// 2) we want to deliver at most `MAX_MESSAGES_TO_RELAY_IN_SINGLE_TX` nonces in this batch
+		// 3) we can't deliver new nonce until header, that has emitted this nonce, is finalized
+		// by target client
+		let nonces_begin = self.target_nonce + 1.into();
+		let best_header_at_target = &race_state.target_state.as_ref()?.best_peer;
+		let mut nonces_end = None;
+		for i in 0..MAX_MESSAGES_TO_RELAY_IN_SINGLE_TX {
+			let nonce = nonces_begin + i.into();
+
+			// if queue is empty, we don't need to prove anything
+			let (first_queued_at, first_queued_nonce) = match self.source_queue.front() {
+				Some((first_queued_at, first_queued_nonce)) => (first_queued_at.clone(), *first_queued_nonce),
+				None => break,
+			};
+
+			// if header that has queued the message is not yet finalized at bridged chain,
+			// we can't prove anything
+			if first_queued_at.0 > best_header_at_target.0 {
+				break;
+			}
+
+			// ok, we may deliver this nonce
+			nonces_end = Some(nonce);
+
+			// probably remove it from the queue?
+			if nonce == first_queued_nonce {
+				self.source_queue.pop_front();
+			}
+		}
+
+		nonces_end.map(|nonces_end| RangeInclusive::new(nonces_begin, nonces_end))
+	}
+}
+
+#[cfg(test)]
+mod tests {
+	use super::*;
+	use crate::message_lane_loop::{
+		tests::{header_id, TestMessageLane, TestMessageNonce, TestMessagesProof},
+		ClientState,
+	};
+
+	#[test]
+	fn strategy_is_empty_works() {
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		assert_eq!(strategy.is_empty(), true);
+		strategy.source_nonce_updated(header_id(1), 1);
+		assert_eq!(strategy.is_empty(), false);
+	}
+
+	#[test]
+	fn source_nonce_is_never_lower_than_known_target_nonce() {
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		strategy.target_nonce_updated(10, &mut Default::default());
+		strategy.source_nonce_updated(header_id(1), 5);
+		assert_eq!(strategy.source_queue, vec![]);
+	}
+
+	#[test]
+	fn source_nonce_is_never_lower_than_latest_known_source_nonce() {
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		strategy.source_nonce_updated(header_id(1), 5);
+		strategy.source_nonce_updated(header_id(2), 3);
+		strategy.source_nonce_updated(header_id(2), 5);
+		assert_eq!(strategy.source_queue, vec![(header_id(1), 5)]);
+	}
+
+	#[test]
+	fn target_nonce_is_never_lower_than_latest_known_target_nonce() {
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		strategy.target_nonce_updated(10, &mut Default::default());
+		strategy.target_nonce_updated(5, &mut Default::default());
+		assert_eq!(strategy.target_nonce, 10);
+	}
+
+	#[test]
+	fn updated_target_nonce_removes_queued_entries() {
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		strategy.source_nonce_updated(header_id(1), 5);
+		strategy.source_nonce_updated(header_id(2), 10);
+		strategy.source_nonce_updated(header_id(3), 15);
+		strategy.source_nonce_updated(header_id(4), 20);
+		strategy.target_nonce_updated(15, &mut Default::default());
+		assert_eq!(strategy.source_queue, vec![(header_id(4), 20)]);
+	}
+
+	#[test]
+	fn selected_nonces_are_dropped_on_target_nonce_update() {
+		let mut state = RaceState::default();
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		state.nonces_to_submit = Some((header_id(1), 5..=10, 5..=10));
+		strategy.target_nonce_updated(7, &mut state);
+		assert!(state.nonces_to_submit.is_some());
+		strategy.target_nonce_updated(10, &mut state);
+		assert!(state.nonces_to_submit.is_none());
+	}
+
+	#[test]
+	fn submitted_nonces_are_dropped_on_target_nonce_update() {
+		let mut state = RaceState::default();
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		state.nonces_submitted = Some(5..=10);
+		strategy.target_nonce_updated(7, &mut state);
+		assert!(state.nonces_submitted.is_some());
+		strategy.target_nonce_updated(10, &mut state);
+		assert!(state.nonces_submitted.is_none());
+	}
+
+	#[test]
+	fn nothing_is_selected_if_something_is_already_selected() {
+		let mut state = RaceState::default();
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		state.nonces_to_submit = Some((header_id(1), 1..=10, 1..=10));
+		strategy.source_nonce_updated(header_id(1), 10);
+		assert_eq!(strategy.select_nonces_to_deliver(&state), None);
+	}
+
+	#[test]
+	fn nothing_is_selected_if_something_is_already_submitted() {
+		let mut state = RaceState::default();
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		state.nonces_submitted = Some(1..=10);
+		strategy.source_nonce_updated(header_id(1), 10);
+		assert_eq!(strategy.select_nonces_to_deliver(&state), None);
+	}
+
+	#[test]
+	fn select_nonces_to_deliver_works() {
+		let mut state = RaceState::<_, _, TestMessageNonce, TestMessagesProof>::default();
+		let mut strategy = MessageDeliveryStrategy::<TestMessageLane>::default();
+		strategy.source_nonce_updated(header_id(1), 1);
+		strategy.source_nonce_updated(header_id(2), 2);
+		strategy.source_nonce_updated(header_id(3), 6);
+		strategy.source_nonce_updated(header_id(5), 8);
+
+		state.target_state = Some(ClientState {
+			best_self: header_id(0),
+			best_peer: header_id(4),
+		});
+		assert_eq!(strategy.select_nonces_to_deliver(&state), Some(1..=4));
+		strategy.target_nonce_updated(4, &mut state);
+		assert_eq!(strategy.select_nonces_to_deliver(&state), Some(5..=6));
+		strategy.target_nonce_updated(6, &mut state);
+		assert_eq!(strategy.select_nonces_to_deliver(&state), None);
+
+		state.target_state = Some(ClientState {
+			best_self: header_id(0),
+			best_peer: header_id(5),
+		});
+		assert_eq!(strategy.select_nonces_to_deliver(&state), Some(7..=8));
+		strategy.target_nonce_updated(8, &mut state);
+		assert_eq!(strategy.select_nonces_to_deliver(&state), None);
+	}
+}

bridges/relays/ethereum/src/message_race_loop.rs

@@ -0,0 +1,359 @@
+// Copyright 2019-2020 Parity Technologies (UK) Ltd.
+// This file is part of Parity Bridges Common.
+
+// Parity Bridges Common is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+
+// Parity Bridges Common is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+//! Loop that is serving single race within message lane. This could be
+//! message delivery race, receiving confirmations race or processing
+//! confirmations race.
+//!
+//! The idea of the race is simple - we have `nonce`-s on source and target
+//! nodes. We're trying to prove that the source node has this nonce (and
+//! associated data - like messages, lane state, etc) to the target node by
+//! generating and submitting proof.
+
+// Until there'll be actual message-lane in the runtime.
+#![allow(dead_code)]
+
+use crate::message_lane_loop::ClientState;
+use crate::utils::{process_future_result, retry_backoff, FailedClient, MaybeConnectionError};
+
+use async_trait::async_trait;
+use futures::{
+	future::FutureExt,
+	stream::{FusedStream, StreamExt},
+};
+use std::{
+	fmt::Debug,
+	ops::RangeInclusive,
+	time::{Duration, Instant},
+};
+
+/// One of races within lane.
+pub trait MessageRace {
+	/// Header id of the race source.
+	type SourceHeaderId: Debug + Clone + PartialEq;
+	/// Header id of the race source.
+	type TargetHeaderId: Debug + Clone + PartialEq;
+
+	/// Message nonce used in the race.
+	type MessageNonce: Debug + Clone;
+	/// Proof that is generated and delivered in this race.
+	type Proof: Clone;
+
+	/// Name of the race source.
+	fn source_name() -> String;
+	/// Name of the race target.
+	fn target_name() -> String;
+}
+
+/// State of race source client.
+type SourceClientState<P> = ClientState<<P as MessageRace>::SourceHeaderId, <P as MessageRace>::TargetHeaderId>;
+
+/// State of race target client.
+type TargetClientState<P> = ClientState<<P as MessageRace>::TargetHeaderId, <P as MessageRace>::SourceHeaderId>;
+
+/// One of message lane clients, which is source client for the race.
+#[async_trait(?Send)]
+pub trait SourceClient<P: MessageRace> {
+	/// Type of error this clients returns.
+	type Error: std::fmt::Debug + MaybeConnectionError;
+
+	/// Return latest nonce that is known to the source client.
+	async fn latest_nonce(
+		&self,
+		at_block: P::SourceHeaderId,
+	) -> Result<(P::SourceHeaderId, P::MessageNonce), Self::Error>;
+	/// Generate proof for delivering to the target client.
+	async fn generate_proof(
+		&self,
+		at_block: P::SourceHeaderId,
+		nonces: RangeInclusive<P::MessageNonce>,
+	) -> Result<(P::SourceHeaderId, RangeInclusive<P::MessageNonce>, P::Proof), Self::Error>;
+}
+
+/// One of message lane clients, which is target client for the race.
+#[async_trait(?Send)]
+pub trait TargetClient<P: MessageRace> {
+	/// Type of error this clients returns.
+	type Error: std::fmt::Debug + MaybeConnectionError;
+
+	/// Return latest nonce that is known to the target client.
+	async fn latest_nonce(
+		&self,
+		at_block: P::TargetHeaderId,
+	) -> Result<(P::TargetHeaderId, P::MessageNonce), Self::Error>;
+	/// Submit proof to the target client.
+	async fn submit_proof(
+		&self,
+		generated_at_block: P::SourceHeaderId,
+		nonces: RangeInclusive<P::MessageNonce>,
+		proof: P::Proof,
+	) -> Result<RangeInclusive<P::MessageNonce>, Self::Error>;
+}
+
+/// Race strategy.
+pub trait RaceStrategy<SourceHeaderId, TargetHeaderId, MessageNonce, Proof> {
+	/// Should return true if nothing has to be synced.
+	fn is_empty(&self) -> bool;
+	/// Called when latest nonce is updated at source node of the race.
+	fn source_nonce_updated(&mut self, at_block: SourceHeaderId, nonce: MessageNonce);
+	/// Called when latest nonce is updated at target node of the race.
+	fn target_nonce_updated(
+		&mut self,
+		nonce: MessageNonce,
+		race_state: &mut RaceState<SourceHeaderId, TargetHeaderId, MessageNonce, Proof>,
+	);
+	/// Should return `Some(nonces)` if we need to deliver proof of `nonces` (and associated
+	/// data) from source to target node.
+	fn select_nonces_to_deliver(
+		&mut self,
+		race_state: &RaceState<SourceHeaderId, TargetHeaderId, MessageNonce, Proof>,
+	) -> Option<RangeInclusive<MessageNonce>>;
+}
+
+/// State of the race.
+pub struct RaceState<SourceHeaderId, TargetHeaderId, MessageNonce, Proof> {
+	/// Source state, if known.
+	pub source_state: Option<ClientState<SourceHeaderId, TargetHeaderId>>,
+	/// Target state, if known.
+	pub target_state: Option<ClientState<TargetHeaderId, SourceHeaderId>>,
+	/// Range of nonces that we have selected to submit.
+	pub nonces_to_submit: Option<(SourceHeaderId, RangeInclusive<MessageNonce>, Proof)>,
+	/// Range of nonces that is currently submitted.
+	pub nonces_submitted: Option<RangeInclusive<MessageNonce>>,
+}
+
+/// Run race loop until connection with target or source node is lost.
+pub async fn run<P: MessageRace>(
+	race_source: impl SourceClient<P>,
+	race_source_updated: impl FusedStream<Item = SourceClientState<P>>,
+	race_target: impl TargetClient<P>,
+	race_target_updated: impl FusedStream<Item = TargetClientState<P>>,
+	stall_timeout: Duration,
+	mut strategy: impl RaceStrategy<P::SourceHeaderId, P::TargetHeaderId, P::MessageNonce, P::Proof>,
+) -> Result<(), FailedClient> {
+	let mut race_state = RaceState::default();
+	let mut stall_countdown = Instant::now();
+
+	let mut source_retry_backoff = retry_backoff();
+	let mut source_client_is_online = true;
+	let mut source_latest_nonce_required = false;
+	let source_latest_nonce = futures::future::Fuse::terminated();
+	let source_generate_proof = futures::future::Fuse::terminated();
+	let source_go_offline_future = futures::future::Fuse::terminated();
+
+	let mut target_retry_backoff = retry_backoff();
+	let mut target_client_is_online = true;
+	let mut target_latest_nonce_required = false;
+	let target_latest_nonce = futures::future::Fuse::terminated();
+	let target_submit_proof = futures::future::Fuse::terminated();
+	let target_go_offline_future = futures::future::Fuse::terminated();
+
+	futures::pin_mut!(
+		race_source_updated,
+		source_latest_nonce,
+		source_generate_proof,
+		source_go_offline_future,
+		race_target_updated,
+		target_latest_nonce,
+		target_submit_proof,
+		target_go_offline_future,
+	);
+
+	loop {
+		futures::select! {
+			// when headers ids are updated
+			source_state = race_source_updated.next() => {
+				if let Some(source_state) = source_state {
+					if race_state.source_state.as_ref() != Some(&source_state) {
+						source_latest_nonce_required = true;
+						race_state.source_state = Some(source_state);
+					}
+				}
+			},
+			target_state = race_target_updated.next() => {
+				if let Some(target_state) = target_state {
+					if race_state.target_state.as_ref() != Some(&target_state) {
+						target_latest_nonce_required = true;
+						race_state.target_state = Some(target_state);
+					}
+				}
+			},
+
+			// when nonces are updated
+			latest_nonce = source_latest_nonce => {
+				source_latest_nonce_required = false;
+
+				source_client_is_online = process_future_result(
+					latest_nonce,
+					&mut source_retry_backoff,
+					|(at_block, latest_nonce)| {
+						log::debug!(
+							target: "bridge",
+							"Received latest nonce from {}: {:?}",
+							P::source_name(),
+							latest_nonce,
+						);
+
+						strategy.source_nonce_updated(at_block, latest_nonce);
+					},
+					&mut source_go_offline_future,
+					|delay| async_std::task::sleep(delay),
+					|| format!("Error retrieving latest nonce from {}", P::source_name()),
+				).fail_if_connection_error(FailedClient::Source)?;
+			},
+			latest_nonce = target_latest_nonce => {
+				target_latest_nonce_required = false;
+
+				target_client_is_online = process_future_result(
+					latest_nonce,
+					&mut target_retry_backoff,
+					|(_, latest_nonce)| {
+						log::debug!(
+							target: "bridge",
+							"Received latest nonce from {}: {:?}",
+							P::target_name(),
+							latest_nonce,
+						);
+
+						strategy.target_nonce_updated(latest_nonce, &mut race_state);
+					},
+					&mut target_go_offline_future,
+					|delay| async_std::task::sleep(delay),
+					|| format!("Error retrieving latest nonce from {}", P::target_name()),
+				).fail_if_connection_error(FailedClient::Target)?;
+			},
+
+			// proof generation and submission
+			proof = source_generate_proof => {
+				source_client_is_online = process_future_result(
+					proof,
+					&mut source_retry_backoff,
+					|(at_block, nonces_range, proof)| {
+						log::debug!(
+							target: "bridge",
+							"Received proof for nonces in range {:?} from {}",
+							nonces_range,
+							P::source_name(),
+						);
+
+						race_state.nonces_to_submit = Some((at_block, nonces_range, proof));
+					},
+					&mut source_go_offline_future,
+					|delay| async_std::task::sleep(delay),
+					|| format!("Error generating proof at {}", P::source_name()),
+				).fail_if_connection_error(FailedClient::Source)?;
+			},
+			proof_submit_result = target_submit_proof => {
+				target_client_is_online = process_future_result(
+					proof_submit_result,
+					&mut target_retry_backoff,
+					|nonces_range| {
+						log::debug!(
+							target: "bridge",
+							"Successfully submitted proof of nonces {:?} to {}",
+							nonces_range,
+							P::target_name(),
+						);
+
+						race_state.nonces_to_submit = None;
+						race_state.nonces_submitted = Some(nonces_range);
+					},
+					&mut target_go_offline_future,
+					|delay| async_std::task::sleep(delay),
+					|| format!("Error submitting proof {}", P::target_name()),
+				).fail_if_connection_error(FailedClient::Target)?;
+			}
+		}
+
+		if stall_countdown.elapsed() > stall_timeout {
+			return Err(FailedClient::Both);
+		} else if race_state.nonces_to_submit.is_none() && race_state.nonces_submitted.is_none() && strategy.is_empty()
+		{
+			stall_countdown = Instant::now();
+		}
+
+		if source_client_is_online {
+			source_client_is_online = false;
+
+			let nonces_to_deliver = race_state.source_state.as_ref().and_then(|source_state| {
+				strategy
+					.select_nonces_to_deliver(&race_state)
+					.map(|nonces_range| (source_state.best_self.clone(), nonces_range))
+			});
+
+			if let Some((at_block, nonces_range)) = nonces_to_deliver {
+				log::debug!(
+					target: "bridge",
+					"Asking {} to prove nonces in range {:?}",
+					P::source_name(),
+					nonces_range,
+				);
+				source_generate_proof.set(race_source.generate_proof(at_block, nonces_range).fuse());
+			} else if source_latest_nonce_required {
+				log::debug!(target: "bridge", "Asking {} about latest generated message nonce", P::source_name());
+				let at_block = race_state
+					.source_state
+					.as_ref()
+					.expect("source_latest_nonce_required is only true when source_state is Some; qed")
+					.best_self
+					.clone();
+				source_latest_nonce.set(race_source.latest_nonce(at_block).fuse());
+			} else {
+				source_client_is_online = true;
+			}
+		}
+
+		if target_client_is_online {
+			target_client_is_online = false;
+
+			if let Some((at_block, nonces_range, proof)) = race_state.nonces_to_submit.as_ref() {
+				log::debug!(
+					target: "bridge",
+					"Going to submit proof of messages in range {:?} to {} node",
+					nonces_range,
+					P::target_name(),
+				);
+				target_submit_proof.set(
+					race_target
+						.submit_proof(at_block.clone(), nonces_range.clone(), proof.clone())
+						.fuse(),
+				);
+			}
+			if target_latest_nonce_required {
+				log::debug!(target: "bridge", "Asking {} about latest nonce", P::target_name());
+				let at_block = race_state
+					.target_state
+					.as_ref()
+					.expect("target_latest_nonce_required is only true when target_state is Some; qed")
+					.best_self
+					.clone();
+				target_latest_nonce.set(race_target.latest_nonce(at_block).fuse());
+			} else {
+				target_client_is_online = true;
+			}
+		}
+	}
+}
+
+impl<SourceHeaderId, TargetHeaderId, MessageNonce, Proof> Default
+	for RaceState<SourceHeaderId, TargetHeaderId, MessageNonce, Proof>
+{
+	fn default() -> Self {
+		RaceState {
+			source_state: None,
+			target_state: None,
+			nonces_to_submit: None,
+			nonces_submitted: None,
+		}
+	}
+}

bridges/relays/ethereum/src/substrate_client.rs

@@ -19,7 +19,8 @@ use crate::instances::BridgeInstance;
 use crate::rpc::{Substrate, SubstrateRpc};
 use crate::rpc_errors::RpcError;
 use crate::substrate_types::{Hash, Header as SubstrateHeader, Number, SignedBlock as SignedSubstrateBlock};
-use crate::sync_types::{HeaderId, SubmittedHeaders};
+use crate::sync_types::SubmittedHeaders;
+use crate::utils::HeaderId;
 
 use async_trait::async_trait;
 use bp_eth_poa::Header as SubstrateEthereumHeader;

bridges/relays/ethereum/src/substrate_types.rs

@@ -17,7 +17,9 @@
 use crate::ethereum_types::{
 	Header as EthereumHeader, Receipt as EthereumReceipt, HEADER_ID_PROOF as ETHEREUM_HEADER_ID_PROOF,
 };
-use crate::sync_types::{HeaderId, HeadersSyncPipeline, QueuedHeader, SourceHeader};
+use crate::sync_types::{HeadersSyncPipeline, QueuedHeader, SourceHeader};
+use crate::utils::HeaderId;
+
 use codec::Encode;
 
 pub use bp_eth_poa::{

bridges/relays/ethereum/src/sync.rs

@@ -310,7 +310,8 @@ pub mod tests {
 	use super::*;
 	use crate::ethereum_types::{EthereumHeadersSyncPipeline, H256};
 	use crate::headers::tests::{header, id};
-	use crate::sync_types::{HeaderId, HeaderStatus};
+	use crate::sync_types::HeaderStatus;
+	use crate::utils::HeaderId;
 
 	fn side_hash(number: u64) -> H256 {
 		H256::from_low_u64_le(1000 + number)

bridges/relays/ethereum/src/sync_loop.rs

@@ -18,10 +18,11 @@ use crate::metrics::{start as metrics_start, GlobalMetrics, MetricsParams};
 use crate::sync::HeadersSyncParams;
 use crate::sync_loop_metrics::SyncLoopMetrics;
 use crate::sync_types::{HeaderIdOf, HeaderStatus, HeadersSyncPipeline, QueuedHeader, SubmittedHeaders};
-use crate::utils::{format_ids, retry_backoff, MaybeConnectionError, StringifiedMaybeConnectionError};
+use crate::utils::{
+	format_ids, interval, process_future_result, retry_backoff, MaybeConnectionError, StringifiedMaybeConnectionError,
+};
 
 use async_trait::async_trait;
-use backoff::{backoff::Backoff, ExponentialBackoff};
 use futures::{future::FutureExt, stream::StreamExt};
 use num_traits::{Saturating, Zero};
 use std::{
@@ -44,9 +45,6 @@ const STALL_SYNC_TIMEOUT: Duration = Duration::from_secs(5 * 60);
 /// Delay after we have seen update of best source header at target node,
 /// for us to treat sync stalled. ONLY when relay operates in backup mode.
 const BACKUP_STALL_SYNC_TIMEOUT: Duration = Duration::from_secs(10 * 60);
-/// Delay after connection-related error happened before we'll try
-/// reconnection again.
-const CONNECTION_ERROR_DELAY: Duration = Duration::from_secs(10);
 
 /// Source client trait.
 #[async_trait]
@@ -186,7 +184,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut source_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving best header number from {}", P::SOURCE_NAME),
-					);
+					).is_ok();
 				},
 				source_new_header = source_new_header_future => {
 					source_client_is_online = process_future_result(
@@ -196,7 +194,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut source_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving header from {} node", P::SOURCE_NAME),
-					);
+					).is_ok();
 				},
 				source_orphan_header = source_orphan_header_future => {
 					source_client_is_online = process_future_result(
@@ -206,7 +204,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut source_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving orphan header from {} node", P::SOURCE_NAME),
-					);
+					).is_ok();
 				},
 				source_extra = source_extra_future => {
 					source_client_is_online = process_future_result(
@@ -216,7 +214,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut source_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving extra data from {} node", P::SOURCE_NAME),
-					);
+					).is_ok();
 				},
 				source_completion = source_completion_future => {
 					source_client_is_online = process_future_result(
@@ -226,7 +224,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut source_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving completion data from {} node", P::SOURCE_NAME),
-					);
+					).is_ok();
 				},
 				source_client = source_go_offline_future => {
 					source_client_is_online = true;
@@ -277,7 +275,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut target_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving best known header from {} node", P::TARGET_NAME),
-					);
+					).is_ok();
 				},
 				incomplete_headers_ids = target_incomplete_headers_future => {
 					target_incomplete_headers_required = false;
@@ -289,7 +287,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut target_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving incomplete headers from {} node", P::TARGET_NAME),
-					);
+					).is_ok();
 				},
 				target_existence_status = target_existence_status_future => {
 					target_client_is_online = process_future_result(
@@ -301,7 +299,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut target_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving existence status from {} node", P::TARGET_NAME),
-					);
+					).is_ok();
 				},
 				submitted_headers = target_submit_header_future => {
 					// following line helps Rust understand the type of `submitted_headers` :/
@@ -329,7 +327,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut target_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error submitting headers to {} node", P::TARGET_NAME),
-					);
+					).is_ok();
 
 					log::debug!(target: "bridge", "Header submit result: {}", submitted_headers_str);
 
@@ -350,7 +348,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut target_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error completing headers at {}", P::TARGET_NAME),
-					);
+					).is_ok();
 				},
 				target_extra_check_result = target_extra_check_future => {
 					target_client_is_online = process_future_result(
@@ -362,7 +360,7 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 						&mut target_go_offline_future,
 						|delay| async_std::task::sleep(delay),
 						|| format!("Error retrieving receipts requirement from {} node", P::TARGET_NAME),
-					);
+					).is_ok();
 				},
 				target_client = target_go_offline_future => {
 					target_client_is_online = true;
@@ -557,62 +555,6 @@ pub fn run<P: HeadersSyncPipeline, TC: TargetClient<P>>(
 	});
 }
 
-/// Stream that emits item every `timeout_ms` milliseconds.
-fn interval(timeout: Duration) -> impl futures::Stream<Item = ()> {
-	futures::stream::unfold((), move |_| async move {
-		async_std::task::sleep(timeout).await;
-		Some(((), ()))
-	})
-}
-
-/// Process result of the future from a client.
-///
-/// Returns whether or not the client we're interacting with is online. In this context
-/// what online means is that the client is currently not handling any other requests
-/// that we've previously sent.
-pub(crate) fn process_future_result<TResult, TError, TGoOfflineFuture>(
-	result: Result<TResult, TError>,
-	retry_backoff: &mut ExponentialBackoff,
-	on_success: impl FnOnce(TResult),
-	go_offline_future: &mut std::pin::Pin<&mut futures::future::Fuse<TGoOfflineFuture>>,
-	go_offline: impl FnOnce(Duration) -> TGoOfflineFuture,
-	error_pattern: impl FnOnce() -> String,
-) -> bool
-where
-	TError: std::fmt::Debug + MaybeConnectionError,
-	TGoOfflineFuture: FutureExt,
-{
-	let mut client_is_online = false;
-
-	match result {
-		Ok(result) => {
-			on_success(result);
-			retry_backoff.reset();
-			client_is_online = true
-		}
-		Err(error) => {
-			let is_connection_error = error.is_connection_error();
-			let retry_delay = if is_connection_error {
-				retry_backoff.reset();
-				CONNECTION_ERROR_DELAY
-			} else {
-				retry_backoff.next_backoff().unwrap_or(CONNECTION_ERROR_DELAY)
-			};
-			go_offline_future.set(go_offline(retry_delay).fuse());
-
-			log::error!(
-				target: "bridge",
-				"{}: {:?}. Retrying in {}s",
-				error_pattern(),
-				error,
-				retry_delay.as_secs_f64(),
-			);
-		}
-	}
-
-	client_is_online
-}
-
 /// Print synchronization progress.
 fn print_sync_progress<P: HeadersSyncPipeline>(
 	progress_context: (Instant, Option<P::Number>, Option<P::Number>),

bridges/relays/ethereum/src/sync_loop_tests.rs

@@ -16,9 +16,9 @@
 
 #![cfg(test)]
 
-use crate::sync_loop::{process_future_result, run, SourceClient, TargetClient};
+use crate::sync_loop::{run, SourceClient, TargetClient};
-use crate::sync_types::{HeaderId, HeadersSyncPipeline, QueuedHeader, SourceHeader, SubmittedHeaders};
+use crate::sync_types::{HeadersSyncPipeline, QueuedHeader, SourceHeader, SubmittedHeaders};
-use crate::utils::{retry_backoff, MaybeConnectionError};
+use crate::utils::{process_future_result, retry_backoff, HeaderId, MaybeConnectionError};
 
 use async_trait::async_trait;
 use backoff::backoff::Backoff;

bridges/relays/ethereum/src/sync_types.rs

@@ -14,14 +14,10 @@
 // You should have received a copy of the GNU General Public License
 // along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
 
-use crate::utils::format_ids;
+use crate::utils::{format_ids, HeaderId};
 
 use std::{ops::Deref, sync::Arc};
 
-/// Ethereum header Id.
-#[derive(Debug, Clone, Copy, Eq, Hash, PartialEq)]
-pub struct HeaderId<Hash, Number>(pub Number, pub Hash);
-
 /// Ethereum header synchronization status.
 #[derive(Debug, Clone, Copy, PartialEq)]
 pub enum HeaderStatus {

bridges/relays/ethereum/src/utils.rs

@@ -14,12 +14,16 @@
 // You should have received a copy of the GNU General Public License
 // along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
 
-use backoff::ExponentialBackoff;
+use backoff::{backoff::Backoff, ExponentialBackoff};
+use futures::future::FutureExt;
 use std::time::Duration;
 
 /// Max delay after connection-unrelated error happened before we'll try the
 /// same request again.
 const MAX_BACKOFF_INTERVAL: Duration = Duration::from_secs(60);
+/// Delay after connection-related error happened before we'll try
+/// reconnection again.
+const CONNECTION_ERROR_DELAY: Duration = Duration::from_secs(10);
 
 /// Macro that returns (client, Err(error)) tuple from function if result is Err(error).
 #[macro_export]
@@ -43,6 +47,10 @@ macro_rules! bail_on_arg_error {
 	};
 }
 
+/// Ethereum header Id.
+#[derive(Debug, Default, Clone, Copy, Eq, Hash, PartialEq)]
+pub struct HeaderId<Hash, Number>(pub Number, pub Hash);
+
 /// Error type that can signal connection errors.
 pub trait MaybeConnectionError {
 	/// Returns true if error (maybe) represents connection error.
@@ -114,3 +122,102 @@ pub fn format_ids<Id: std::fmt::Debug>(mut ids: impl ExactSizeIterator<Item = Id
 		}
 	}
 }
+
+/// Stream that emits item every `timeout_ms` milliseconds.
+pub fn interval(timeout: Duration) -> impl futures::Stream<Item = ()> {
+	futures::stream::unfold((), move |_| async move {
+		async_std::task::sleep(timeout).await;
+		Some(((), ()))
+	})
+}
+
+/// Which client has caused error.
+#[derive(Debug, Clone, Copy, PartialEq)]
+pub enum FailedClient {
+	/// It is the source client who has caused error.
+	Source,
+	/// It is the target client who has caused error.
+	Target,
+	/// Both clients are failing, or we just encountered some other error that
+	/// should be treated like that.
+	Both,
+}
+
+/// Future process result.
+#[derive(Debug, Clone, Copy)]
+pub enum ProcessFutureResult {
+	/// Future has been processed successfully.
+	Success,
+	/// Future has failed with non-connection error.
+	Failed,
+	/// Future has failed with connection error.
+	ConnectionFailed,
+}
+
+impl ProcessFutureResult {
+	/// Returns true if result is Success.
+	pub fn is_ok(self) -> bool {
+		match self {
+			ProcessFutureResult::Success => true,
+			ProcessFutureResult::Failed | ProcessFutureResult::ConnectionFailed => false,
+		}
+	}
+
+	/// Returns Ok(true) if future has succeeded.
+	/// Returns Ok(false) if future has failed with non-connection error.
+	/// Returns Err if future is `ConnectionFailed`.
+	pub fn fail_if_connection_error(self, failed_client: FailedClient) -> Result<bool, FailedClient> {
+		match self {
+			ProcessFutureResult::Success => Ok(true),
+			ProcessFutureResult::Failed => Ok(false),
+			ProcessFutureResult::ConnectionFailed => Err(failed_client),
+		}
+	}
+}
+
+/// Process result of the future from a client.
+pub(crate) fn process_future_result<TResult, TError, TGoOfflineFuture>(
+	result: Result<TResult, TError>,
+	retry_backoff: &mut ExponentialBackoff,
+	on_success: impl FnOnce(TResult),
+	go_offline_future: &mut std::pin::Pin<&mut futures::future::Fuse<TGoOfflineFuture>>,
+	go_offline: impl FnOnce(Duration) -> TGoOfflineFuture,
+	error_pattern: impl FnOnce() -> String,
+) -> ProcessFutureResult
+where
+	TError: std::fmt::Debug + MaybeConnectionError,
+	TGoOfflineFuture: FutureExt,
+{
+	match result {
+		Ok(result) => {
+			on_success(result);
+			retry_backoff.reset();
+			ProcessFutureResult::Success
+		}
+		Err(error) if error.is_connection_error() => {
+			log::error!(
+				target: "bridge",
+				"{}: {:?}. Going to restart",
+				error_pattern(),
+				error,
+			);
+
+			retry_backoff.reset();
+			go_offline_future.set(go_offline(CONNECTION_ERROR_DELAY).fuse());
+			ProcessFutureResult::ConnectionFailed
+		}
+		Err(error) => {
+			let retry_delay = retry_backoff.next_backoff().unwrap_or(CONNECTION_ERROR_DELAY);
+			log::error!(
+				target: "bridge",
+				"{}: {:?}. Retrying in {}",
+				error_pattern(),
+				error,
+				retry_delay.as_secs_f64(),
+			);
+
+			go_offline_future.set(go_offline(retry_delay).fuse());
+			ProcessFutureResult::Failed
+		}
+	}
+}