Rework the event system of sc-network (#1370)

This commit introduces a new concept called `NotificationService` which allows Polkadot protocols to communicate with the underlying notification protocol implementation directly, without routing events through `NetworkWorker`. This implies that each protocol has its own service which it uses to communicate with remote peers and that each `NotificationService` is unique with respect to the underlying notification protocol, meaning `NotificationService` for the transaction protocol can only be used to send and receive transaction-related notifications. The `NotificationService` concept introduces two additional benefits: * allow protocols to start using custom handshakes * allow protocols to accept/reject inbound peers Previously the validation of inbound connections was solely the responsibility of `ProtocolController`. This caused issues with light peers and `SyncingEngine` as `ProtocolController` would accept more peers than `SyncingEngine` could accept which caused peers to have differing views of their own states. `SyncingEngine` would reject excess peers but these rejections were not properly communicated to those peers causing them to assume that they were accepted. With `NotificationService`, the local handshake is not sent to remote peer if peer is rejected which allows it to detect that it was rejected. This commit also deprecates the use of `NetworkEventStream` for all notification-related events and going forward only DHT events are provided through `NetworkEventStream`. If protocols wish to follow each other's events, they must introduce additional abtractions, as is done for GRANDPA and transactions protocols by following the syncing protocol through `SyncEventStream`. Fixes https://github.com/paritytech/polkadot-sdk/issues/512 Fixes https://github.com/paritytech/polkadot-sdk/issues/514 Fixes https://github.com/paritytech/polkadot-sdk/issues/515 Fixes https://github.com/paritytech/polkadot-sdk/issues/554 Fixes https://github.com/paritytech/polkadot-sdk/issues/556 --- These changes are transferred from https://github.com/paritytech/substrate/pull/14197 but there are no functional changes compared to that PR --------- Co-authored-by: Dmitry Markin <dmitry@markin.tech> Co-authored-by: Alexandru Vasile <60601340+lexnv@users.noreply.github.com>
2026-07-14 03:55:53 +00:00 · 2023-11-28 20:18:52 +02:00
parent ec3a61ed86
commit e71c484d5b
102 changed files with 5694 additions and 2603 deletions
@@ -15,7 +15,7 @@
 // along with Polkadot.  If not, see <http://www.gnu.org/licenses/>.

 use super::*;
-use futures::{channel::oneshot, executor, stream::BoxStream};
+use futures::{channel::oneshot, executor};
 use overseer::jaeger;
 use polkadot_node_network_protocol::{self as net_protocol, OurView};
 use polkadot_node_subsystem::messages::NetworkBridgeEvent;
@@ -26,10 +26,13 @@ use parking_lot::Mutex;
 use std::{
 	collections::HashSet,
 	sync::atomic::{AtomicBool, Ordering},
-	task::Poll,
 };

-use sc_network::{Event as NetworkEvent, IfDisconnected, ProtocolName, ReputationChange};
+use sc_network::{
+	service::traits::{Direction, MessageSink, NotificationService},
+	IfDisconnected, Multiaddr, ObservedRole as SubstrateObservedRole, ProtocolName,
+	ReputationChange, Roles,
+};

 use polkadot_node_network_protocol::{
 	peer_set::PeerSetProtocolNames,
@@ -47,9 +50,8 @@ use polkadot_node_subsystem_test_helpers::{
 	mock::new_leaf, SingleItemSink, SingleItemStream, TestSubsystemContextHandle,
 };
 use polkadot_node_subsystem_util::metered;
-use polkadot_primitives::{AuthorityDiscoveryId, CandidateHash, Hash};
+use polkadot_primitives::{AuthorityDiscoveryId, Hash};

-use sc_network::Multiaddr;
 use sp_keyring::Sr25519Keyring;

 use crate::{network::Network, validator_discovery::AuthorityDiscovery};
@@ -64,10 +66,9 @@ pub enum NetworkAction {
 	WriteNotification(PeerId, PeerSet, Vec<u8>),
 }

-// The subsystem's view of the network - only supports a single call to `event_stream`.
+// The subsystem's view of the network.
 #[derive(Clone)]
 struct TestNetwork {
-	net_events: Arc<Mutex<Option<SingleItemStream<NetworkEvent>>>>,
 	action_tx: Arc<Mutex<metered::UnboundedMeteredSender<NetworkAction>>>,
 	protocol_names: Arc<PeerSetProtocolNames>,
 }
@@ -79,37 +80,42 @@ struct TestAuthorityDiscovery;
 // of `NetworkAction`s.
 struct TestNetworkHandle {
 	action_rx: metered::UnboundedMeteredReceiver<NetworkAction>,
-	net_tx: SingleItemSink<NetworkEvent>,
-	protocol_names: PeerSetProtocolNames,
+	validation_tx: SingleItemSink<NotificationEvent>,
+	collation_tx: SingleItemSink<NotificationEvent>,
 }

 fn new_test_network(
 	protocol_names: PeerSetProtocolNames,
-) -> (TestNetwork, TestNetworkHandle, TestAuthorityDiscovery) {
-	let (net_tx, net_rx) = polkadot_node_subsystem_test_helpers::single_item_sink();
+) -> (
+	TestNetwork,
+	TestNetworkHandle,
+	TestAuthorityDiscovery,
+	Box<dyn NotificationService>,
+	Box<dyn NotificationService>,
+) {
 	let (action_tx, action_rx) = metered::unbounded();
+	let (validation_tx, validation_rx) = polkadot_node_subsystem_test_helpers::single_item_sink();
+	let (collation_tx, collation_rx) = polkadot_node_subsystem_test_helpers::single_item_sink();
+	let action_tx = Arc::new(Mutex::new(action_tx));

 	(
 		TestNetwork {
-			net_events: Arc::new(Mutex::new(Some(net_rx))),
-			action_tx: Arc::new(Mutex::new(action_tx)),
+			action_tx: action_tx.clone(),
 			protocol_names: Arc::new(protocol_names.clone()),
 		},
-		TestNetworkHandle { action_rx, net_tx, protocol_names },
+		TestNetworkHandle { action_rx, validation_tx, collation_tx },
 		TestAuthorityDiscovery,
+		Box::new(TestNotificationService::new(
+			PeerSet::Validation,
+			action_tx.clone(),
+			validation_rx,
+		)),
+		Box::new(TestNotificationService::new(PeerSet::Collation, action_tx, collation_rx)),
 	)
 }

 #[async_trait]
 impl Network for TestNetwork {
-	fn event_stream(&mut self) -> BoxStream<'static, NetworkEvent> {
-		self.net_events
-			.lock()
-			.take()
-			.expect("Subsystem made more than one call to `event_stream`")
-			.boxed()
-	}
-
 	async fn set_reserved_peers(
 		&mut self,
 		_protocol: ProtocolName,
@@ -143,7 +149,8 @@ impl Network for TestNetwork {
 	}

 	fn disconnect_peer(&self, who: PeerId, protocol: ProtocolName) {
-		let (peer_set, _) = self.protocol_names.try_get_protocol(&protocol).unwrap();
+		let (peer_set, version) = self.protocol_names.try_get_protocol(&protocol).unwrap();
+		assert_eq!(version, peer_set.get_main_version());

 		self.action_tx
 			.lock()
@@ -151,13 +158,10 @@ impl Network for TestNetwork {
 			.unwrap();
 	}

-	fn write_notification(&self, who: PeerId, protocol: ProtocolName, message: Vec<u8>) {
-		let (peer_set, _) = self.protocol_names.try_get_protocol(&protocol).unwrap();
-
-		self.action_tx
-			.lock()
-			.unbounded_send(NetworkAction::WriteNotification(who, peer_set, message))
-			.unwrap();
+	fn peer_role(&self, _peer_id: PeerId, handshake: Vec<u8>) -> Option<SubstrateObservedRole> {
+		Roles::decode_all(&mut &handshake[..])
+			.ok()
+			.and_then(|role| Some(SubstrateObservedRole::from(role)))
 	}
 }

@@ -201,35 +205,85 @@ impl TestNetworkHandle {
 		peer_set: PeerSet,
 		role: ObservedRole,
 	) {
-		let protocol_version = ProtocolVersion::from(protocol_version);
-		self.send_network_event(NetworkEvent::NotificationStreamOpened {
-			remote: peer,
-			protocol: self.protocol_names.get_name(peer_set, protocol_version),
-			negotiated_fallback: None,
-			role: role.into(),
-			received_handshake: vec![],
-		})
-		.await;
+		fn observed_role_to_handshake(role: &ObservedRole) -> Vec<u8> {
+			match role {
+				&ObservedRole::Light => Roles::LIGHT.encode(),
+				&ObservedRole::Authority => Roles::AUTHORITY.encode(),
+				&ObservedRole::Full => Roles::FULL.encode(),
+			}
+		}
+
+		// because of how protocol negotiation works, if two peers support at least one common
+		// protocol, the protocol is negotiated over the main protocol (`ValidationVersion::V2`) but
+		// if either one of the peers used a fallback protocol for the negotiation (meaning they
+		// don't support the main protocol but some older version of it ), `negotiated_fallback` is
+		// set to that protocol.
+		let negotiated_fallback = match protocol_version {
+			ValidationVersion::V2 => None,
+			ValidationVersion::V1 => match peer_set {
+				PeerSet::Validation => Some(ProtocolName::from("/polkadot/validation/1")),
+				PeerSet::Collation => Some(ProtocolName::from("/polkadot/collation/1")),
+			},
+			ValidationVersion::VStaging => match peer_set {
+				PeerSet::Validation => Some(ProtocolName::from("/polkadot/validation/3")),
+				PeerSet::Collation => unreachable!(),
+			},
+		};
+
+		match peer_set {
+			PeerSet::Validation => {
+				self.validation_tx
+					.send(NotificationEvent::NotificationStreamOpened {
+						peer,
+						direction: Direction::Inbound,
+						handshake: observed_role_to_handshake(&role),
+						negotiated_fallback,
+					})
+					.await
+					.expect("subsystem concluded early");
+			},
+			PeerSet::Collation => {
+				self.collation_tx
+					.send(NotificationEvent::NotificationStreamOpened {
+						peer,
+						direction: Direction::Inbound,
+						handshake: observed_role_to_handshake(&role),
+						negotiated_fallback,
+					})
+					.await
+					.expect("subsystem concluded early");
+			},
+		}
 	}

 	async fn disconnect_peer(&mut self, peer: PeerId, peer_set: PeerSet) {
-		self.send_network_event(NetworkEvent::NotificationStreamClosed {
-			remote: peer,
-			protocol: self.protocol_names.get_main_name(peer_set),
-		})
-		.await;
+		match peer_set {
+			PeerSet::Validation => self
+				.validation_tx
+				.send(NotificationEvent::NotificationStreamClosed { peer })
+				.await
+				.expect("subsystem concluded early"),
+			PeerSet::Collation => self
+				.collation_tx
+				.send(NotificationEvent::NotificationStreamClosed { peer })
+				.await
+				.expect("subsystem concluded early"),
+		}
 	}

 	async fn peer_message(&mut self, peer: PeerId, peer_set: PeerSet, message: Vec<u8>) {
-		self.send_network_event(NetworkEvent::NotificationsReceived {
-			remote: peer,
-			messages: vec![(self.protocol_names.get_main_name(peer_set), message.into())],
-		})
-		.await;
-	}
-
-	async fn send_network_event(&mut self, event: NetworkEvent) {
-		self.net_tx.send(event).await.expect("subsystem concluded early");
+		match peer_set {
+			PeerSet::Validation => self
+				.validation_tx
+				.send(NotificationEvent::NotificationReceived { peer, notification: message })
+				.await
+				.expect("subsystem concluded early"),
+			PeerSet::Collation => self
+				.collation_tx
+				.send(NotificationEvent::NotificationReceived { peer, notification: message })
+				.await
+				.expect("subsystem concluded early"),
+		}
 	}
 }

@@ -240,6 +294,121 @@ fn assert_network_actions_contains(actions: &[NetworkAction], action: &NetworkAc
 	}
 }

+struct TestNotificationService {
+	peer_set: PeerSet,
+	action_tx: Arc<Mutex<metered::UnboundedMeteredSender<NetworkAction>>>,
+	rx: SingleItemStream<NotificationEvent>,
+}
+
+impl std::fmt::Debug for TestNotificationService {
+	fn fmt(&self, _: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+		Ok(())
+	}
+}
+
+impl TestNotificationService {
+	pub fn new(
+		peer_set: PeerSet,
+		action_tx: Arc<Mutex<metered::UnboundedMeteredSender<NetworkAction>>>,
+		rx: SingleItemStream<NotificationEvent>,
+	) -> Self {
+		Self { peer_set, action_tx, rx }
+	}
+}
+
+struct TestMessageSink {
+	peer: PeerId,
+	peer_set: PeerSet,
+	action_tx: Arc<Mutex<metered::UnboundedMeteredSender<NetworkAction>>>,
+}
+
+impl TestMessageSink {
+	fn new(
+		peer: PeerId,
+		peer_set: PeerSet,
+		action_tx: Arc<Mutex<metered::UnboundedMeteredSender<NetworkAction>>>,
+	) -> TestMessageSink {
+		Self { peer, peer_set, action_tx }
+	}
+}
+
+#[async_trait::async_trait]
+impl MessageSink for TestMessageSink {
+	fn send_sync_notification(&self, notification: Vec<u8>) {
+		self.action_tx
+			.lock()
+			.unbounded_send(NetworkAction::WriteNotification(
+				self.peer,
+				self.peer_set,
+				notification,
+			))
+			.unwrap();
+	}
+
+	async fn send_async_notification(
+		&self,
+		_notification: Vec<u8>,
+	) -> Result<(), sc_network::error::Error> {
+		unimplemented!();
+	}
+}
+
+#[async_trait::async_trait]
+impl NotificationService for TestNotificationService {
+	/// Instruct `Notifications` to open a new substream for `peer`.
+	async fn open_substream(&mut self, _peer: PeerId) -> Result<(), ()> {
+		unimplemented!();
+	}
+
+	/// Instruct `Notifications` to close substream for `peer`.
+	async fn close_substream(&mut self, _peer: PeerId) -> Result<(), ()> {
+		unimplemented!();
+	}
+
+	/// Send synchronous `notification` to `peer`.
+	fn send_sync_notification(&self, _peer: &PeerId, _notification: Vec<u8>) {
+		unimplemented!();
+	}
+
+	/// Send asynchronous `notification` to `peer`, allowing sender to exercise backpressure.
+	async fn send_async_notification(
+		&self,
+		_peer: &PeerId,
+		_notification: Vec<u8>,
+	) -> Result<(), sc_network::error::Error> {
+		unimplemented!();
+	}
+
+	/// Set handshake for the notification protocol replacing the old handshake.
+	async fn set_handshake(&mut self, _handshake: Vec<u8>) -> Result<(), ()> {
+		unimplemented!();
+	}
+
+	fn try_set_handshake(&mut self, _handshake: Vec<u8>) -> Result<(), ()> {
+		unimplemented!();
+	}
+
+	/// Get next event from the `Notifications` event stream.
+	async fn next_event(&mut self) -> Option<NotificationEvent> {
+		self.rx.next().await
+	}
+
+	// Clone [`NotificationService`]
+	fn clone(&mut self) -> Result<Box<dyn NotificationService>, ()> {
+		unimplemented!();
+	}
+
+	/// Get protocol name.
+	fn protocol(&self) -> &ProtocolName {
+		unimplemented!();
+	}
+
+	/// Get notification sink of the peer.
+	fn message_sink(&self, peer: &PeerId) -> Option<Box<dyn MessageSink>> {
+		Some(Box::new(TestMessageSink::new(*peer, self.peer_set, self.action_tx.clone())))
+	}
+}
+
 #[derive(Clone)]
 struct TestSyncOracle {
 	is_major_syncing: Arc<AtomicBool>,
@@ -335,10 +504,11 @@ fn test_harness<T: Future<Output = VirtualOverseer>>(
 	let peerset_protocol_names = PeerSetProtocolNames::new(genesis_hash, fork_id);

 	let pool = sp_core::testing::TaskExecutor::new();
-	let (mut network, network_handle, discovery) = new_test_network(peerset_protocol_names.clone());
+	let (network, network_handle, discovery, validation_service, collation_service) =
+		new_test_network(peerset_protocol_names.clone());
 	let (context, virtual_overseer) =
 		polkadot_node_subsystem_test_helpers::make_subsystem_context(pool);
-	let network_stream = network.event_stream();
+	let notification_sinks = Arc::new(Mutex::new(HashMap::new()));
 	let shared = Shared::default();

 	let bridge = NetworkBridgeRx {
@@ -348,9 +518,12 @@ fn test_harness<T: Future<Output = VirtualOverseer>>(
 		sync_oracle,
 		shared: shared.clone(),
 		peerset_protocol_names,
+		validation_service,
+		collation_service,
+		notification_sinks,
 	};

-	let network_bridge = run_network_in(bridge, context, network_stream)
+	let network_bridge = run_network_in(bridge, context)
 		.map_err(|_| panic!("subsystem execution failed"))
 		.map(|_| ());

@@ -942,8 +1115,6 @@ fn relays_collation_protocol_messages() {
 			.await;
 		}

-		// peer A gets reported for sending a collation message.
-
 		let collator_protocol_message = protocol_v1::CollatorProtocolMessage::Declare(
 			Sr25519Keyring::Alice.public().into(),
 			Default::default(),
@@ -953,19 +1124,23 @@ fn relays_collation_protocol_messages() {
 		let message_v1 =
 			protocol_v1::CollationProtocol::CollatorProtocol(collator_protocol_message.clone());

-		network_handle
-			.peer_message(
-				peer_a,
-				PeerSet::Collation,
-				WireMessage::ProtocolMessage(message_v1.clone()).encode(),
-			)
-			.await;
+		// peer A gets reported for sending a collation message.
+		// NOTE: this is not possible since peer A cannot send
+		// a collation message if it has not opened a collation protocol

-		let actions = network_handle.next_network_actions(3).await;
-		assert_network_actions_contains(
-			&actions,
-			&NetworkAction::ReputationChange(peer_a, UNCONNECTED_PEERSET_COST.into()),
-		);
+		// network_handle
+		// 	.peer_message(
+		// 		peer_a,
+		// 		PeerSet::Collation,
+		// 		WireMessage::ProtocolMessage(message_v1.clone()).encode(),
+		// 	)
+		// 	.await;
+
+		// let actions = network_handle.next_network_actions(3).await;
+		// assert_network_actions_contains(
+		// 	&actions,
+		// 	&NetworkAction::ReputationChange(peer_a, UNCONNECTED_PEERSET_COST.into()),
+		// );

 		// peer B has the message relayed.

@@ -1212,7 +1387,7 @@ fn our_view_updates_decreasing_order_and_limited_to_max() {
 fn network_protocol_versioning_view_update() {
 	let (oracle, handle) = make_sync_oracle(false);
 	test_harness(Box::new(oracle), |test_harness| async move {
-		let TestHarness { mut network_handle, mut virtual_overseer, .. } = test_harness;
+		let TestHarness { mut network_handle, mut virtual_overseer, shared } = test_harness;

 		let peer_ids: Vec<_> = (0..4).map(|_| PeerId::random()).collect();
 		let peers = [
@@ -1231,12 +1406,22 @@ fn network_protocol_versioning_view_update() {

 		handle.await_mode_switch().await;

+		let mut total_validation_peers = 0;
+		let mut total_collation_peers = 0;
+
 		for &(peer_id, peer_set, version) in &peers {
 			network_handle
 				.connect_peer(peer_id, version, peer_set, ObservedRole::Full)
 				.await;
+
+			match peer_set {
+				PeerSet::Validation => total_validation_peers += 1,
+				PeerSet::Collation => total_collation_peers += 1,
+			}
 		}

+		await_peer_connections(&shared, total_validation_peers, total_collation_peers).await;
+
 		let view = view![head];
 		let actions = network_handle.next_network_actions(4).await;

@@ -1264,15 +1449,19 @@ fn network_protocol_versioning_view_update() {

 #[test]
 fn network_protocol_versioning_subsystem_msg() {
+	use polkadot_primitives::CandidateHash;
+	use std::task::Poll;
+
 	let (oracle, _handle) = make_sync_oracle(false);
 	test_harness(Box::new(oracle), |test_harness| async move {
-		let TestHarness { mut network_handle, mut virtual_overseer, .. } = test_harness;
+		let TestHarness { mut network_handle, mut virtual_overseer, shared } = test_harness;

 		let peer = PeerId::random();

 		network_handle
 			.connect_peer(peer, ValidationVersion::V2, PeerSet::Validation, ObservedRole::Full)
 			.await;
+		await_peer_connections(&shared, 1, 0).await;

 		// bridge will inform about all connected peers.
 		{