Sync block justifications (#1410)

* core: sync protocol for justifications

* core: basic test for justification sync

* core: pass block number with justification

* grandpa: request justifications when importing change blocks

* core: pass finality notifications to chain sync

* core: require justifications for pending change blocks on start

* core: avoid requesting justifications from previous failed peers

* core: timeout block justification requests

* core: add some docs

* core: fix unused variables warning

* core: tick pending justifications fetch periodically

* grandpa: add test for syncing justifications

* core: early exit dispatch of pending justifications

* core: style fix

* core: grandpa: change logging level

* core: sync: add missing docs

* core: network: report peer on bad justification

* core: replace mem::replace with Option::take

* core: revert authority set changes on failed block finalization

* core: grandpa: add docs to import_justification

* core: warn on re-finalization of last finalized block

* core: only notify sync with last finality notification

* core: style fix

* core: add docs for PendingJustifications

* core: network: use BlockRequest messages for justification requests

* core: reference issues in todo comments

* core: grandpa: revert authority set changes on db

* core: grandpa: remove inconsistent state warning

substrate/core/client/src/client.rs

@@ -728,7 +728,11 @@ impl<B, E, Block, RA> Client<B, E, Block, RA> where
 		// find tree route from last finalized to given block.
 		let last_finalized = self.backend.blockchain().last_finalized()?;
 
-		if block == last_finalized { return Ok(()) }
+		if block == last_finalized {
+			warn!("Possible safety violation: attempted to re-finalize last finalized block {:?} ", last_finalized);
+			return Ok(());
+		}
+
 		let route_from_finalized = crate::blockchain::tree_route(
 			self.backend.blockchain(),
 			BlockId::Hash(last_finalized),
@@ -1093,7 +1097,7 @@ impl<B, E, Block, RA> consensus::BlockImport<Block> for Client<B, E, Block, RA>
 		match self.backend.blockchain().status(BlockId::Hash(parent_hash)) {
 			Ok(blockchain::BlockStatus::InChain) => {},
 			Ok(blockchain::BlockStatus::Unknown) => return Ok(ImportResult::UnknownParent),
-            Err(e) => return Err(ConsensusErrorKind::ClientImport(e.to_string()).into())
+			Err(e) => return Err(ConsensusErrorKind::ClientImport(e.to_string()).into()),
 		}
 
 		let import_headers = if post_digests.is_empty() {
@@ -1131,6 +1135,19 @@ impl<B, E, Block, RA> consensus::BlockImport<Block> for Client<B, E, Block, RA>
 		);
 		result.map_err(|e| ConsensusErrorKind::ClientImport(e.to_string()).into())
 	}
+
+	/// Import a block justification and finalize the block. The justification
+	/// isn't interpreted by the client and is assumed to have been validated
+	/// previously. The block is finalized unconditionally.
+	fn import_justification(
+		&self,
+		hash: Block::Hash,
+		_number: NumberFor<Block>,
+		justification: Justification,
+	) -> Result<(), Self::Error> {
+		self.finalize_block(BlockId::Hash(hash), Some(justification), true)
+			.map_err(|_| ConsensusErrorKind::InvalidJustification.into())
+	}
 }
 
 impl<B, E, Block, RA> consensus::Authorities<Block> for Client<B, E, Block, RA> where

substrate/core/consensus/common/src/block_import.rs

@@ -16,7 +16,7 @@
 
 //! Block import helpers.
 
-use runtime_primitives::traits::{AuthorityIdFor, Block as BlockT, Header as HeaderT, DigestItemFor};
+use runtime_primitives::traits::{AuthorityIdFor, Block as BlockT, DigestItemFor, Header as HeaderT, NumberFor};
 use runtime_primitives::Justification;
 use std::borrow::Cow;
 
@@ -33,6 +33,9 @@ pub enum ImportResult {
 	KnownBad,
 	/// Block parent is not in the chain.
 	UnknownParent,
+	/// Added to the import queue but must be justified
+	/// (usually required to safely enact consensus changes).
+	NeedsJustification,
 }
 
 /// Block data origin.
@@ -140,9 +143,22 @@ impl<Block: BlockT> ImportBlock<Block> {
 /// Block import trait.
 pub trait BlockImport<B: BlockT> {
 	type Error: ::std::error::Error + Send + 'static;
-	/// Import a Block alongside the new authorities valid form this block forward
-	fn import_block(&self,
+
+	/// Called by the import queue when it is started.
+	fn on_start(&self, _link: &::import_queue::Link<B>) { }
+
+	/// Import a Block alongside the new authorities valid from this block forward
+	fn import_block(
+		&self,
 		block: ImportBlock<B>,
-		new_authorities: Option<Vec<AuthorityIdFor<B>>>
+		new_authorities: Option<Vec<AuthorityIdFor<B>>>,
 	) -> Result<ImportResult, Self::Error>;
+
+	/// Import a Block justification and finalize the given block.
+	fn import_justification(
+		&self,
+		hash: B::Hash,
+		number: NumberFor<B>,
+		justification: Justification,
+	) -> Result<(), Self::Error>;
 }

substrate/core/consensus/common/src/import_queue.rs

@@ -92,6 +92,8 @@ pub trait ImportQueue<B: BlockT>: Send + Sync {
 	fn is_importing(&self, hash: &B::Hash) -> bool;
 	/// Import bunch of blocks.
 	fn import_blocks(&self, origin: BlockOrigin, blocks: Vec<IncomingBlock<B>>);
+	/// Import a block justification.
+	fn import_justification(&self, hash: B::Hash, number: NumberFor<B>, justification: Justification) -> bool;
 }
 
 /// Import queue status. It isn't completely accurate.
@@ -161,6 +163,7 @@ impl<B: BlockT, V: 'static + Verifier<B>> ImportQueue<B> for BasicQueue<B, V> {
 		let verifier = self.verifier.clone();
 		let block_import = self.block_import.clone();
 		*self.handle.lock() = Some(::std::thread::Builder::new().name("ImportQueue".into()).spawn(move || {
+			block_import.on_start(&link);
 			import_thread(block_import, link, qdata, verifier)
 		})?);
 		Ok(())
@@ -218,6 +221,10 @@ impl<B: BlockT, V: 'static + Verifier<B>> ImportQueue<B> for BasicQueue<B, V> {
 		queue.push_back((origin, blocks));
 		self.data.signal.notify_one();
 	}
+
+	fn import_justification(&self, hash: B::Hash, number: NumberFor<B>, justification: Justification) -> bool {
+		self.block_import.import_justification(hash, number, justification).is_ok()
+	}
 }
 
 impl<B: BlockT, V: 'static + Verifier<B>> Drop for BasicQueue<B, V> {
@@ -279,6 +286,8 @@ fn import_thread<B: BlockT, L: Link<B>, V: Verifier<B>>(
 pub trait Link<B: BlockT>: Send {
 	/// Block imported.
 	fn block_imported(&self, _hash: &B::Hash, _number: NumberFor<B>) { }
+	/// Request a justification for the given block.
+	fn request_justification(&self, _hash: &B::Hash, _number: NumberFor<B>) { }
 	/// Maintain sync.
 	fn maintain_sync(&self) { }
 	/// Disconnect from peer.
@@ -296,6 +305,8 @@ pub enum BlockImportResult<H: ::std::fmt::Debug + PartialEq, N: ::std::fmt::Debu
 	ImportedKnown(H, N),
 	/// Imported unknown block.
 	ImportedUnknown(H, N),
+	/// Imported unjustified block that requires one.
+	ImportedUnjustified(H, N),
 }
 
 /// Block import error.
@@ -409,6 +420,10 @@ pub fn import_single_block<B: BlockT, V: Verifier<B>>(
 			trace!(target: "sync", "Block queued {}: {:?}", number, hash);
 			Ok(BlockImportResult::ImportedUnknown(hash, number))
 		},
+		Ok(ImportResult::NeedsJustification) => {
+			trace!(target: "sync", "Block queued but requires justification {}: {:?}", number, hash);
+			Ok(BlockImportResult::ImportedUnjustified(hash, number))
+		},
 		Ok(ImportResult::UnknownParent) => {
 			debug!(target: "sync", "Block with unknown parent {}: {:?}, parent: {:?}", number, hash, parent);
 			Err(BlockImportError::UnknownParent)
@@ -416,7 +431,7 @@ pub fn import_single_block<B: BlockT, V: Verifier<B>>(
 		Ok(ImportResult::KnownBad) => {
 			debug!(target: "sync", "Peer gave us a bad block {}: {:?}", number, hash);
 			Err(BlockImportError::BadBlock(peer)) //TODO: use persistent ID
-		}
+		},
 		Err(e) => {
 			debug!(target: "sync", "Error importing block {}: {:?}: {:?}", number, hash, e);
 			Err(BlockImportError::Error)
@@ -439,6 +454,11 @@ pub fn process_import_result<B: BlockT>(
 			link.block_imported(&hash, number);
 			1
 		},
+		Ok(BlockImportResult::ImportedUnjustified(hash, number)) => {
+			link.block_imported(&hash, number);
+			link.request_justification(&hash, number);
+			1
+		},
 		Err(BlockImportError::IncompleteHeader(who)) => {
 			if let Some(peer) = who {
 				link.useless_peer(peer, "Sent block with incomplete header to import");

substrate/core/finality-grandpa/src/authorities.rs

@@ -146,7 +146,6 @@ where
 	}
 
 	/// Inspect pending changes.
-	#[cfg(test)]
 	pub(crate) fn pending_changes(&self) -> &[PendingChange<H, N>] {
 		&self.pending_changes
 	}
@@ -261,7 +260,7 @@ pub(crate) struct PendingChange<H, N> {
 
 impl<H, N: Add<Output=N> + Clone> PendingChange<H, N> {
 	/// Returns the effective number this change will be applied at.
-	fn effective_number(&self) -> N {
+	pub fn effective_number(&self) -> N {
 		self.canon_height.clone() + self.finalization_depth.clone()
 	}
 }

substrate/core/finality-grandpa/src/lib.rs

@@ -92,6 +92,7 @@ use client::{
 use client::blockchain::HeaderBackend;
 use codec::{Encode, Decode};
 use consensus_common::{BlockImport, Error as ConsensusError, ErrorKind as ConsensusErrorKind, ImportBlock, ImportResult, Authorities};
+use runtime_primitives::Justification;
 use runtime_primitives::traits::{
 	NumberFor, Block as BlockT, Header as HeaderT, DigestFor, ProvideRuntimeApi, Hash as HashT,
 	DigestItemFor, DigestItem, As, Zero,
@@ -315,7 +316,7 @@ impl<B, E, Block: BlockT<Hash=H256>, RA> BlockStatus<Block> for Arc<Client<B, E,
 }
 
 /// Consensus-related data changes tracker.
-#[derive(Debug, Encode, Decode)]
+#[derive(Clone, Debug, Encode, Decode)]
 struct ConsensusChanges<H, N> {
 	pending_changes: Vec<(N, H)>,
 }
@@ -796,6 +797,13 @@ fn finalize_block<B, Block: BlockT<Hash=H256>, E, RA>(
 {
 	// lock must be held through writing to DB to avoid race
 	let mut authority_set = authority_set.inner().write();
+
+	// TODO [andre]: clone only when changed (#1483)
+	let old_authority_set = authority_set.clone();
+	// needed in case there is an authority set change, used for reverting in
+	// case of error
+	let mut old_last_completed = None;
+
 	let mut consensus_changes = consensus_changes.lock();
 	let status = authority_set.apply_changes(number, |canon_number| {
 		canonical_at_height(client, (hash, number), canon_number)
@@ -813,6 +821,8 @@ fn finalize_block<B, Block: BlockT<Hash=H256>, E, RA>(
 			let last_completed: LastCompleted<_, _> = (0, round_state);
 			let encoded = last_completed.encode();
 
+			old_last_completed = Backend::get_aux(&**client.backend(), LAST_COMPLETED_KEY)?;
+
 			Backend::insert_aux(
 				&**client.backend(),
 				&[
@@ -836,7 +846,13 @@ fn finalize_block<B, Block: BlockT<Hash=H256>, E, RA>(
 	// check if this is this is the first finalization of some consensus changes
 	let (alters_consensus_changes, finalizes_consensus_changes) = consensus_changes
 		.finalize((number, hash), |at_height| canonical_at_height(client, (hash, number), at_height))?;
+
+	// holds the old consensus changes in case it is changed below, needed for
+	// reverting in case of failure
+	let mut old_consensus_changes = None;
 	if alters_consensus_changes {
+		old_consensus_changes = Some(consensus_changes.clone());
+
 		let encoded = consensus_changes.encode();
 		let write_result = Backend::insert_aux(&**client.backend(), &[(CONSENSUS_CHANGES_KEY, &encoded[..])], &[]);
 		if let Err(e) = write_result {
@@ -847,75 +863,115 @@ fn finalize_block<B, Block: BlockT<Hash=H256>, E, RA>(
 		}
 	}
 
-	// NOTE: this code assumes that honest voters will never vote past a
-	// transition block, thus we don't have to worry about the case where
-	// we have a transition with `effective_block = N`, but we finalize
-	// `N+1`. this assumption is required to make sure we store
-	// justifications for transition blocks which will be requested by
-	// syncing clients.
-	let justification = match justification_or_commit {
-		JustificationOrCommit::Justification(justification) => Some(justification.encode()),
-		JustificationOrCommit::Commit((round_number, commit)) => {
-			let mut justification_required =
-				// justification is always required when block that enacts new authorities
-				// set is finalized
-				status.new_set_block.is_some() ||
-				// justification is required when consensus changes are finalized
-				finalizes_consensus_changes;
+	let aux = |authority_set: &authorities::AuthoritySet<Block::Hash, NumberFor<Block>>| {
+		// NOTE: this code assumes that honest voters will never vote past a
+		// transition block, thus we don't have to worry about the case where
+		// we have a transition with `effective_block = N`, but we finalize
+		// `N+1`. this assumption is required to make sure we store
+		// justifications for transition blocks which will be requested by
+		// syncing clients.
+		let justification = match justification_or_commit {
+			JustificationOrCommit::Justification(justification) => Some(justification.encode()),
+			JustificationOrCommit::Commit((round_number, commit)) => {
+				let mut justification_required =
+					// justification is always required when block that enacts new authorities
+					// set is finalized
+					status.new_set_block.is_some() ||
+					// justification is required when consensus changes are finalized
+					finalizes_consensus_changes;
 
-			// justification is required every N blocks to be able to prove blocks
-			// finalization to remote nodes
-			if !justification_required {
-				if let Some(justification_period) = justification_period {
-					let last_finalized_number = client.info()?.chain.finalized_number;
-					justification_required = (!last_finalized_number.is_zero() ||
-						number - last_finalized_number == justification_period) &&
-						(last_finalized_number / justification_period != number / justification_period);
+				// justification is required every N blocks to be able to prove blocks
+				// finalization to remote nodes
+				if !justification_required {
+					if let Some(justification_period) = justification_period {
+						let last_finalized_number = client.info()?.chain.finalized_number;
+						justification_required =
+							(!last_finalized_number.is_zero() || number - last_finalized_number == justification_period) &&
+							(last_finalized_number / justification_period != number / justification_period);
+					}
 				}
-			}
 
-			if justification_required {
-				let justification = GrandpaJustification::from_commit(
-					client,
-					round_number,
-					commit,
-				)?;
+				if justification_required {
+					let justification = GrandpaJustification::from_commit(
+						client,
+						round_number,
+						commit,
+					)?;
 
-				Some(justification.encode())
+					Some(justification.encode())
+				} else {
+					None
+				}
+			},
+		};
+
+		debug!(target: "afg", "Finalizing blocks up to ({:?}, {})", number, hash);
+
+		// ideally some handle to a synchronization oracle would be used
+		// to avoid unconditionally notifying.
+		client.finalize_block(BlockId::Hash(hash), justification, true).map_err(|e| {
+			warn!(target: "finality", "Error applying finality to block {:?}: {:?}", (hash, number), e);
+			e
+		})?;
+
+		if let Some((canon_hash, canon_number)) = status.new_set_block {
+			// the authority set has changed.
+			let (new_id, set_ref) = authority_set.current();
+
+			if set_ref.len() > 16 {
+				info!("Applying GRANDPA set change to new set with {} authorities", set_ref.len());
 			} else {
-				None
+				info!("Applying GRANDPA set change to new set {:?}", set_ref);
 			}
-		},
+
+			Err(ExitOrError::AuthoritiesChanged(NewAuthoritySet {
+				canon_hash,
+				canon_number,
+				set_id: new_id,
+				authorities: set_ref.to_vec(),
+			}))
+		} else {
+			Ok(())
+		}
 	};
 
-	debug!(target: "afg", "Finalizing blocks up to ({:?}, {})", number, hash);
+	match aux(&authority_set) {
+		Err(ExitOrError::Error(err)) => {
+			debug!(target: "afg", "Reverting authority set and/or consensus changes after block finalization error: {:?}", err);
 
-	// ideally some handle to a synchronization oracle would be used
-	// to avoid unconditionally notifying.
-	client.finalize_block(BlockId::Hash(hash), justification, true).map_err(|e| {
-		warn!(target: "finality", "Error applying finality to block {:?}: {:?}", (hash, number), e);
-		warn!(target: "finality", "Node is in a potentially inconsistent state.");
-		e
-	})?;
+			let mut revert_aux = Vec::new();
 
-	if let Some((canon_hash, canon_number)) = status.new_set_block {
-		// the authority set has changed.
-		let (new_id, set_ref) = authority_set.current();
+			if status.changed {
+				revert_aux.push((AUTHORITY_SET_KEY, old_authority_set.encode()));
+				if let Some(old_last_completed) = old_last_completed {
+					revert_aux.push((LAST_COMPLETED_KEY, old_last_completed));
+				}
 
-		if set_ref.len() > 16 {
-			info!("Applying GRANDPA set change to new set with {} authorities", set_ref.len());
-		} else {
-			info!("Applying GRANDPA set change to new set {:?}", set_ref);
-		}
+				*authority_set = old_authority_set.clone();
+			}
 
-		Err(ExitOrError::AuthoritiesChanged(NewAuthoritySet {
-			canon_hash,
-			canon_number,
-			set_id: new_id,
-			authorities: set_ref.to_vec(),
-		}))
-	} else {
-		Ok(())
+			if let Some(old_consensus_changes) = old_consensus_changes {
+				revert_aux.push((CONSENSUS_CHANGES_KEY, old_consensus_changes.encode()));
+
+				*consensus_changes = old_consensus_changes;
+			}
+
+			let write_result = Backend::insert_aux(
+				&**client.backend(),
+				revert_aux.iter().map(|(k, v)| (*k, &**v)).collect::<Vec<_>>().iter(),
+				&[],
+			);
+
+			if let Err(e) = write_result {
+				warn!(target: "finality", "Failed to revert consensus changes to disk. Bailing.");
+				warn!(target: "finality", "Node is in a potentially inconsistent state.");
+
+				return Err(e.into());
+			}
+
+			Err(ExitOrError::Error(err))
+		},
+		res => res,
 	}
 }
 
@@ -949,6 +1005,33 @@ impl<B, E, Block: BlockT<Hash=H256>, RA, PRA> BlockImport<Block>
 {
 	type Error = ConsensusError;
 
+	fn on_start(&self, link: &::consensus_common::import_queue::Link<Block>) {
+		let chain_info = match self.inner.info() {
+			Ok(info) => info.chain,
+			_ => return,
+		};
+
+		// request justifications for all pending changes for which change blocks have already been imported
+		for pending_change in self.authority_set.inner().read().pending_changes() {
+			if pending_change.effective_number() > chain_info.finalized_number &&
+				pending_change.effective_number() <= chain_info.best_number
+			{
+				let effective_block_hash = self.inner.best_containing(
+					pending_change.canon_hash,
+					Some(pending_change.effective_number()),
+				);
+
+				if let Ok(Some(hash)) = effective_block_hash {
+					if let Ok(Some(header)) = self.inner.header(&BlockId::Hash(hash)) {
+						if *header.number() == pending_change.effective_number() {
+							link.request_justification(&header.hash(), *header.number());
+						}
+					}
+				}
+			}
+		}
+	}
+
 	fn import_block(&self, mut block: ImportBlock<Block>, new_authorities: Option<Vec<Ed25519AuthorityId>>)
 		-> Result<ImportResult, Self::Error>
 	{
@@ -1045,52 +1128,7 @@ impl<B, E, Block: BlockT<Hash=H256>, RA, PRA> BlockImport<Block>
 
 		match justification {
 			Some(justification) => {
-				let justification = GrandpaJustification::decode_and_verify(
-					justification,
-					self.authority_set.set_id(),
-					&self.authority_set.current_authorities(),
-				);
-
-				let justification = match justification {
-					Err(e) => return Err(ConsensusErrorKind::ClientImport(e.to_string()).into()),
-					Ok(justification) => justification,
-				};
-
-				let result = finalize_block(
-					&*self.inner,
-					&self.authority_set,
-					&self.consensus_changes,
-					None,
-					hash,
-					number,
-					justification.into(),
-				);
-
-				match result {
-					Ok(_) => {
-						assert!(!enacts_change, "returns Ok when no authority set change should be enacted; qed;");
-					},
-					Err(ExitOrError::AuthoritiesChanged(new)) => {
-						assert!(
-							enacts_change,
-							"returns AuthoritiesChanged when authority set change should be enacted; qed;"
-						);
-
-						debug!(target: "finality", "Imported justified block #{} that enacts authority set change, signalling voter.", number);
-						if let Err(e) = self.authority_set_change.unbounded_send(new) {
-							return Err(ConsensusErrorKind::ClientImport(e.to_string()).into());
-						}
-					},
-					Err(ExitOrError::Error(e)) => {
-						match e {
-							Error::Grandpa(error) => return Err(ConsensusErrorKind::ClientImport(error.to_string()).into()),
-							Error::Network(error) => return Err(ConsensusErrorKind::ClientImport(error).into()),
-							Error::Blockchain(error) => return Err(ConsensusErrorKind::ClientImport(error).into()),
-							Error::Client(error) => return Err(ConsensusErrorKind::ClientImport(error.to_string()).into()),
-							Error::Timer(error) => return Err(ConsensusErrorKind::ClientImport(error.to_string()).into()),
-						}
-					},
-				}
+				self.import_justification(hash, number, justification, enacts_change)?;
 			},
 			None => {
 				if enacts_change {
@@ -1106,11 +1144,87 @@ impl<B, E, Block: BlockT<Hash=H256>, RA, PRA> BlockImport<Block>
 				if enacts_consensus_change {
 					self.consensus_changes.lock().note_change((number, hash));
 				}
-			},
+
+				return Ok(ImportResult::NeedsJustification);
+			}
 		}
 
 		Ok(import_result)
 	}
+
+	fn import_justification(
+		&self,
+		hash: Block::Hash,
+		number: NumberFor<Block>,
+		justification: Justification,
+	) -> Result<(), Self::Error> {
+		self.import_justification(hash, number, justification, false)
+	}
+}
+
+impl<B, E, Block: BlockT<Hash=H256>, RA, PRA>
+	GrandpaBlockImport<B, E, Block, RA, PRA> where
+		NumberFor<Block>: grandpa::BlockNumberOps,
+		B: Backend<Block, Blake2Hasher> + 'static,
+		E: CallExecutor<Block, Blake2Hasher> + 'static + Clone + Send + Sync,
+		RA: Send + Sync,
+{
+
+	/// Import a block justification and finalize the block.
+	///
+	/// If `enacts_change` is set to true, then finalizing this block *must*
+	/// enact an authority set change, the function will panic otherwise.
+	fn import_justification(
+		&self,
+		hash: Block::Hash,
+		number: NumberFor<Block>,
+		justification: Justification,
+		enacts_change: bool,
+	) -> Result<(), ConsensusError> {
+		let justification = GrandpaJustification::decode_and_verify(
+			justification,
+			self.authority_set.set_id(),
+			&self.authority_set.current_authorities(),
+		);
+
+		let justification = match justification {
+			Err(e) => return Err(ConsensusErrorKind::ClientImport(e.to_string()).into()),
+			Ok(justification) => justification,
+		};
+
+		let result = finalize_block(
+			&*self.inner,
+			&self.authority_set,
+			&self.consensus_changes,
+			None,
+			hash,
+			number,
+			justification.into(),
+		);
+
+		match result {
+			Err(ExitOrError::AuthoritiesChanged(new)) => {
+				info!(target: "finality", "Imported justification for block #{} that enacts authority set change, signalling voter.", number);
+				if let Err(e) = self.authority_set_change.unbounded_send(new) {
+					return Err(ConsensusErrorKind::ClientImport(e.to_string()).into());
+				}
+			},
+			Err(ExitOrError::Error(e)) => {
+				return Err(match e {
+					Error::Grandpa(error) => ConsensusErrorKind::ClientImport(error.to_string()),
+					Error::Network(error) => ConsensusErrorKind::ClientImport(error),
+					Error::Blockchain(error) => ConsensusErrorKind::ClientImport(error),
+					Error::Client(error) => ConsensusErrorKind::ClientImport(error.to_string()),
+					Error::Timer(error) => ConsensusErrorKind::ClientImport(error.to_string()),
+				}.into());
+			},
+			Ok(_) => {
+				assert!(!enacts_change, "returns Ok when no authority set change should be enacted; qed;");
+			},
+		}
+
+		Ok(())
+	}
 }
 
 /// Using the given base get the block at the given height on this chain. The

substrate/core/finality-grandpa/src/tests.rs

@@ -494,7 +494,7 @@ fn transition_3_voters_twice_1_observer() {
 	let transitions = api.scheduled_changes.clone();
 	let net = Arc::new(Mutex::new(GrandpaTestNet::new(api, 8)));
 
-	let mut runtime = tokio::runtime::Runtime::new().unwrap();
+	let mut runtime = current_thread::Runtime::new().unwrap();
 
 	net.lock().peer(0).push_blocks(1, false);
 	net.lock().sync();
@@ -619,6 +619,7 @@ fn transition_3_voters_twice_1_observer() {
 	let drive_to_completion = ::tokio::timer::Interval::new_interval(TEST_ROUTING_INTERVAL)
 		.for_each(move |_| {
 			net.lock().send_import_notifications();
+			net.lock().send_finality_notifications();
 			net.lock().sync();
 			Ok(())
 		})
@@ -682,4 +683,55 @@ fn consensus_changes_works() {
 	changes.note_change((1, 1.into()));
 	changes.note_change((1, 101.into()));
 	assert_eq!(changes.finalize((10, 10.into()), |_| Ok(Some(1.into()))).unwrap(), (true, true));
-}
+}
+
+#[test]
+fn sync_justifications_on_change_blocks() {
+	::env_logger::init();
+
+	let peers_a = &[Keyring::Alice, Keyring::Bob, Keyring::Charlie];
+	let peers_b = &[Keyring::Alice, Keyring::Bob];
+	let voters = make_ids(peers_b);
+
+	// 4 peers, 3 of them are authorities and participate in grandpa
+	let api = TestApi::new(voters);
+	let transitions = api.scheduled_changes.clone();
+	let mut net = GrandpaTestNet::new(api, 4);
+
+	// add 20 blocks
+	net.peer(0).push_blocks(20, false);
+
+	// at block 21 we do add a transition which is instant
+	net.peer(0).generate_blocks(1, BlockOrigin::File, |builder| {
+		let block = builder.bake().unwrap();
+		transitions.lock().insert(*block.header.parent_hash(), ScheduledChange {
+			next_authorities: make_ids(peers_b),
+			delay: 0,
+		});
+		block
+	});
+
+	// add more blocks on top of it (until we have 25)
+	net.peer(0).push_blocks(4, false);
+	net.sync();
+
+	for i in 0..4 {
+		assert_eq!(net.peer(i).client().info().unwrap().chain.best_number, 25,
+			"Peer #{} failed to sync", i);
+	}
+
+	let net = Arc::new(Mutex::new(net));
+	run_to_completion(25, net.clone(), peers_a);
+
+	// the first 3 peers are grandpa voters and therefore have already finalized
+	// block 21 and stored a justification
+	for i in 0..3 {
+		assert!(net.lock().peer(i).client().justification(&BlockId::Number(21)).unwrap().is_some());
+	}
+
+	// the last peer should get the justification by syncing from other peers
+	assert!(net.lock().peer(3).client().justification(&BlockId::Number(21)).unwrap().is_none());
+	while net.lock().peer(3).client().justification(&BlockId::Number(21)).unwrap().is_none() {
+		net.lock().sync_steps(100);
+	}
+}

substrate/core/network/src/message.rs

@@ -48,7 +48,6 @@ pub type BlockRequest<B> = generic::BlockRequest<
 	<<B as BlockT>::Header as HeaderT>::Number,
 >;
 
-
 /// Type alias for using the BlockData type using block type parameters.
 pub type BlockData<B> = generic::BlockData<
 	<B as BlockT>::Header,

substrate/core/network/src/protocol.rs

@@ -15,7 +15,7 @@
 // along with Substrate.  If not, see <http://www.gnu.org/licenses/>.
 
 use std::collections::{HashMap, HashSet, BTreeMap};
-use std::{mem, cmp};
+use std::cmp;
 use std::sync::Arc;
 use std::time;
 use parking_lot::RwLock;
@@ -273,7 +273,7 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 					let mut peers = self.context_data.peers.write();
 					if let Some(ref mut peer) = peers.get_mut(&who) {
 						peer.request_timestamp = None;
-						match mem::replace(&mut peer.block_request, None) {
+						match peer.block_request.take() {
 							Some(r) => r,
 							None => {
 								io.report_peer(who, Severity::Bad("Unexpected response packet received from peer"));
@@ -285,10 +285,12 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 						return;
 					}
 				};
+
 				if request.id != r.id {
 					trace!(target: "sync", "Ignoring mismatched response packet from {} (expected {} got {})", who, request.id, r.id);
 					return;
 				}
+
 				self.on_block_response(io, who, request, r);
 			},
 			GenericMessage::BlockAnnounce(announce) => self.on_block_announce(io, who, announce),
@@ -330,7 +332,6 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 	pub fn on_peer_disconnected(&self, io: &mut SyncIo, peer: NodeIndex) {
 		trace!(target: "sync", "Disconnecting {}: {}", peer, io.peer_debug_info(peer));
 
-
 		// lock all the the peer lists so that add/remove peer events are in order
 		let mut sync = self.sync.write();
 		let mut spec = self.specialization.write();
@@ -351,7 +352,15 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 	}
 
 	fn on_block_request(&self, io: &mut SyncIo, peer: NodeIndex, request: message::BlockRequest<B>) {
-		trace!(target: "sync", "BlockRequest {} from {}: from {:?} to {:?} max {:?}", request.id, peer, request.from, request.to, request.max);
+		trace!(target: "sync", "BlockRequest {} from {} with fields {:?}: from {:?} to {:?} max {:?}",
+			request.id,
+			peer,
+			request.fields,
+			request.from,
+			request.to,
+			request.max,
+		);
+
 		let mut blocks = Vec::new();
 		let mut id = match request.from {
 			message::FromBlock::Hash(h) => BlockId::Hash(h),
@@ -409,25 +418,36 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 		trace!(target: "sync", "BlockResponse {} from {} with {} blocks{}",
 			response.id, peer, response.blocks.len(), blocks_range);
 
-		// import_queue.import_blocks also acquires sync.write();
-		// Break the cycle by doing these separately from the outside;
-		let new_blocks = {
+		// TODO [andre]: move this logic to the import queue so that
+		// justifications are imported asynchronously (#1482)
+		if request.fields == message::BlockAttributes::JUSTIFICATION {
 			let mut sync = self.sync.write();
-			sync.on_block_data(&mut ProtocolContext::new(&self.context_data, io), peer, request, response)
-		};
+			sync.on_block_justification_data(
+				&mut ProtocolContext::new(&self.context_data, io),
+				peer,
+				request,
+				response,
+			);
+		} else {
+			// import_queue.import_blocks also acquires sync.write();
+			// Break the cycle by doing these separately from the outside;
+			let new_blocks = {
+				let mut sync = self.sync.write();
+				sync.on_block_data(&mut ProtocolContext::new(&self.context_data, io), peer, request, response)
+			};
 
-		if let Some((origin, new_blocks)) = new_blocks {
-			let import_queue = self.sync.read().import_queue();
-			import_queue.import_blocks(origin, new_blocks);
+			if let Some((origin, new_blocks)) = new_blocks {
+				let import_queue = self.sync.read().import_queue();
+				import_queue.import_blocks(origin, new_blocks);
+			}
 		}
-
-
 	}
 
 	/// Perform time based maintenance.
 	pub fn tick(&self, io: &mut SyncIo) {
 		self.consensus_gossip.write().collect_garbage(|_| true);
 		self.maintain_peers(io);
+		self.sync.write().tick(&mut ProtocolContext::new(&self.context_data, io));
 		self.on_demand.as_ref().map(|s| s.maintain_peers(io));
 	}
 
@@ -439,7 +459,8 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 			let handshaking_peers = self.handshaking_peers.read();
 			for (who, timestamp) in peers.iter()
 				.filter_map(|(id, peer)| peer.request_timestamp.as_ref().map(|r| (id, r)))
-				.chain(handshaking_peers.iter()) {
+				.chain(handshaking_peers.iter())
+			{
 				if (tick - *timestamp).as_secs() > REQUEST_TIMEOUT_SEC {
 					trace!(target: "sync", "Timeout {}", who);
 					aborting.push(*who);
@@ -648,6 +669,10 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 		}
 	}
 
+	pub fn on_block_finalized(&self, _io: &mut SyncIo, hash: B::Hash, header: &B::Header) {
+		self.sync.write().block_finalized(&hash, *header.number());
+	}
+
 	fn on_remote_call_request(&self, io: &mut SyncIo, who: NodeIndex, request: message::RemoteCallRequest<B::Hash>) {
 		trace!(target: "sync", "Remote call request {} from {} ({} at {})", request.id, who, request.method, request.block);
 		let proof = match self.context_data.chain.execution_proof(&request.block, &request.method, &request.data) {
@@ -752,8 +777,8 @@ impl<B: BlockT, S: NetworkSpecialization<B>, H: ExHashT> Protocol<B, S, H> {
 }
 
 fn send_message<B: BlockT, H: ExHashT>(peers: &RwLock<HashMap<NodeIndex, Peer<B, H>>>, io: &mut SyncIo, who: NodeIndex, mut message: Message<B>) {
-	match &mut message {
-		&mut GenericMessage::BlockRequest(ref mut r) => {
+	match message {
+		GenericMessage::BlockRequest(ref mut r) => {
 			let mut peers = peers.write();
 			if let Some(ref mut peer) = peers.get_mut(&who) {
 				r.id = peer.next_request_id;

substrate/core/network/src/service.rs

@@ -94,6 +94,10 @@ impl<B: BlockT, E: ExecuteInContext<B>> Link<B> for NetworkLink<B, E> {
 		self.with_sync(|sync, _| sync.block_imported(&hash, number))
 	}
 
+	fn request_justification(&self, hash: &B::Hash, number: NumberFor<B>) {
+		self.with_sync(|sync, protocol| sync.request_justification(hash, number, protocol))
+	}
+
 	fn maintain_sync(&self) {
 		self.with_sync(|sync, protocol| sync.maintain_sync(protocol))
 	}
@@ -174,6 +178,11 @@ impl<B: BlockT + 'static, S: NetworkSpecialization<B>, H: ExHashT> Service<B, S,
 		self.handler.on_block_imported(&mut NetSyncIo::new(&self.network, self.protocol_id), hash, header)
 	}
 
+	/// Called when a new block is finalized by the client.
+	pub fn on_block_finalized(&self, hash: B::Hash, header: &B::Header) {
+		self.handler.on_block_finalized(&mut NetSyncIo::new(&self.network, self.protocol_id), hash, header)
+	}
+
 	/// Called when new transactons are imported by the client.
 	pub fn trigger_repropagate(&self) {
 		self.handler.propagate_extrinsics(&mut NetSyncIo::new(&self.network, self.protocol_id));

substrate/core/network/src/sync.rs

@@ -14,8 +14,9 @@
 // You should have received a copy of the GNU General Public License
 // along with Substrate.  If not, see <http://www.gnu.org/licenses/>.
 
-use std::collections::HashMap;
+use std::collections::{HashMap, HashSet, VecDeque};
 use std::sync::Arc;
+use std::time::{Duration, Instant};
 use protocol::Context;
 use network_libp2p::{Severity, NodeIndex};
 use client::{BlockStatus, ClientInfo};
@@ -23,6 +24,7 @@ use consensus::BlockOrigin;
 use consensus::import_queue::{ImportQueue, IncomingBlock};
 use client::error::Error as ClientError;
 use blocks::BlockCollection;
+use runtime_primitives::Justification;
 use runtime_primitives::traits::{Block as BlockT, Header as HeaderT, As, NumberFor, Zero};
 use runtime_primitives::generic::BlockId;
 use message::{self, generic::Message as GenericMessage};
@@ -34,6 +36,8 @@ const MAX_BLOCKS_TO_REQUEST: usize = 128;
 const MAX_IMPORTING_BLOCKS: usize = 2048;
 // Number of blocks in the queue that prevents ancestry search.
 const MAJOR_SYNC_BLOCKS: usize = 5;
+// Time to wait before trying to get a justification from the same peer.
+const JUSTIFICATION_RETRY_WAIT: Duration = Duration::from_secs(10);
 
 struct PeerSync<B: BlockT> {
 	pub common_number: NumberFor<B>,
@@ -48,6 +52,180 @@ enum PeerSyncState<B: BlockT> {
 	Available,
 	DownloadingNew(NumberFor<B>),
 	DownloadingStale(B::Hash),
+	DownloadingJustification(B::Hash),
+}
+
+/// Pending justification request for the given block (hash and number).
+type PendingJustification<B> = (<B as BlockT>::Hash, NumberFor<B>);
+
+/// Manages pending block justification requests.
+struct PendingJustifications<B: BlockT> {
+	justifications: HashSet<PendingJustification<B>>,
+	pending_requests: VecDeque<PendingJustification<B>>,
+	peer_requests: HashMap<NodeIndex, PendingJustification<B>>,
+	previous_requests: HashMap<PendingJustification<B>, Vec<(NodeIndex, Instant)>>,
+}
+
+impl<B: BlockT> PendingJustifications<B> {
+	fn new() -> PendingJustifications<B> {
+		PendingJustifications {
+			justifications: HashSet::new(),
+			pending_requests: VecDeque::new(),
+			peer_requests: HashMap::new(),
+			previous_requests: HashMap::new(),
+		}
+	}
+
+	/// Dispatches all possible pending requests to the given peers. Peers are
+	/// filtered according to the current known best block (i.e. we won't send a
+	/// justification request for block #10 to a peer at block #2), and we also
+	/// throttle requests to the same peer if a previous justification request
+	/// yielded no results.
+	fn dispatch(&mut self, peers: &mut HashMap<NodeIndex, PeerSync<B>>, protocol: &mut Context<B>) {
+		if self.pending_requests.is_empty() {
+			return;
+		}
+
+		// clean up previous failed requests so we can retry again
+		for (_, requests) in self.previous_requests.iter_mut() {
+			requests.retain(|(_, instant)| instant.elapsed() < JUSTIFICATION_RETRY_WAIT);
+		}
+
+		let mut available_peers = peers.iter().filter_map(|(peer, sync)| {
+			// don't request to any peers that already have pending requests
+			if let PeerSyncState::Available = sync.state {
+				Some((*peer, sync.best_number))
+			} else {
+				None
+			}
+		}).collect::<VecDeque<_>>();
+
+		let mut last_peer = available_peers.back().map(|p| p.0);
+		let mut unhandled_requests = VecDeque::new();
+
+		loop {
+			let (peer, peer_best_number) = match available_peers.pop_front() {
+				Some(p) => p,
+				_ => break,
+			};
+
+			// only ask peers that have synced past the block number that we're
+			// asking the justification for and to whom we haven't already made
+			// the same request recently
+			let peer_eligible = {
+				let request = match self.pending_requests.front() {
+					Some(r) => r.clone(),
+					_ => break,
+				};
+
+				peer_best_number >= request.1 &&
+					!self.previous_requests
+						 .get(&request)
+						 .map(|requests| requests.iter().any(|i| i.0 == peer))
+						 .unwrap_or(false)
+			};
+
+			if !peer_eligible {
+				available_peers.push_back((peer, peer_best_number));
+
+				// we tried all peers and none can answer this request
+				if Some(peer) == last_peer {
+					last_peer = available_peers.back().map(|p| p.0);
+
+					let request = self.pending_requests.pop_front()
+						.expect("verified to be Some in the beginning of the loop; qed");
+
+					unhandled_requests.push_back(request);
+				}
+
+				continue;
+			}
+
+			last_peer = available_peers.back().map(|p| p.0);
+
+			let request = self.pending_requests.pop_front()
+				.expect("verified to be Some in the beginning of the loop; qed");
+
+			self.peer_requests.insert(peer, request);
+
+			peers.get_mut(&peer)
+				.expect("peer was is taken from available_peers; available_peers is a subset of peers; qed")
+				.state = PeerSyncState::DownloadingJustification(request.0);
+
+			let request = message::generic::BlockRequest {
+				id: 0,
+				fields: message::BlockAttributes::JUSTIFICATION,
+				from: message::FromBlock::Hash(request.0),
+				to: None,
+				direction: message::Direction::Ascending,
+				max: Some(1),
+			};
+
+			protocol.send_message(peer, GenericMessage::BlockRequest(request));
+		}
+
+		self.pending_requests.append(&mut unhandled_requests);
+	}
+
+	/// Queue a justification request (without dispatching it).
+	fn queue_request(&mut self, justification: &PendingJustification<B>) {
+		if !self.justifications.insert(*justification) {
+			return;
+		}
+		self.pending_requests.push_back(*justification);
+	}
+
+	/// Retry any pending request if a peer disconnected.
+	fn peer_disconnected(&mut self, who: NodeIndex) {
+		if let Some(request) = self.peer_requests.remove(&who) {
+			self.pending_requests.push_front(request);
+		}
+	}
+
+	/// Processes the response for the request previously sent to the given
+	/// peer. Queues a retry in case the import fails or the given justification
+	/// was `None`.
+	fn on_response(
+		&mut self,
+		who: NodeIndex,
+		justification: Option<Justification>,
+		protocol: &mut Context<B>,
+		import_queue: &ImportQueue<B>,
+	) {
+		// we assume that the request maps to the given response, this is
+		// currently enforced by the outer network protocol before passing on
+		// messages to chain sync.
+		if let Some(request) = self.peer_requests.remove(&who) {
+			if let Some(justification) = justification {
+				if import_queue.import_justification(request.0, request.1, justification) {
+					self.justifications.remove(&request);
+					self.previous_requests.remove(&request);
+					return;
+				} else {
+					protocol.report_peer(
+						who,
+						Severity::Bad(&format!("Invalid justification provided for #{}", request.0)),
+					);
+				}
+			} else {
+				self.previous_requests
+					.entry(request)
+					.or_insert(Vec::new())
+					.push((who, Instant::now()));
+			}
+
+			self.pending_requests.push_front(request);
+		}
+	}
+
+	/// Removes any pending justification requests for blocks lower than the
+	/// given best finalized.
+	fn collect_garbage(&mut self, best_finalized: NumberFor<B>) {
+		self.justifications.retain(|(_, n)| *n > best_finalized);
+		self.pending_requests.retain(|(_, n)| *n > best_finalized);
+		self.peer_requests.retain(|_, (_, n)| *n > best_finalized);
+		self.previous_requests.retain(|(_, n), _| *n > best_finalized);
+	}
 }
 
 /// Relay chain sync strategy.
@@ -59,6 +237,7 @@ pub struct ChainSync<B: BlockT> {
 	best_queued_hash: B::Hash,
 	required_block_attributes: message::BlockAttributes,
 	import_queue: Arc<ImportQueue<B>>,
+	justifications: PendingJustifications<B>,
 }
 
 /// Reported sync state.
@@ -104,6 +283,7 @@ impl<B: BlockT> ChainSync<B> {
 			blocks: BlockCollection::new(),
 			best_queued_hash: info.best_queued_hash.unwrap_or(info.chain.best_hash),
 			best_queued_number: info.best_queued_number.unwrap_or(info.chain.best_number),
+			justifications: PendingJustifications::new(),
 			required_block_attributes,
 			import_queue,
 		}
@@ -192,6 +372,7 @@ impl<B: BlockT> ChainSync<B> {
 		}
 	}
 
+	/// Handle new block data.
 	pub(crate) fn on_block_data(
 		&mut self,
 		protocol: &mut Context<B>,
@@ -269,10 +450,10 @@ impl<B: BlockT> ChainSync<B> {
 						}
 					}
 				},
-				PeerSyncState::Available => Vec::new(),
+				PeerSyncState::Available | PeerSyncState::DownloadingJustification(..) => Vec::new(),
 			}
 		} else {
-			vec![]
+			Vec::new()
 		};
 
 		let best_seen = self.best_seen_block();
@@ -288,17 +469,87 @@ impl<B: BlockT> ChainSync<B> {
 		Some((origin, new_blocks))
 	}
 
+	/// Handle new justification data.
+	pub(crate) fn on_block_justification_data(
+		&mut self,
+		protocol: &mut Context<B>,
+		who: NodeIndex,
+		_request: message::BlockRequest<B>,
+		response: message::BlockResponse<B>,
+	) {
+		if let Some(ref mut peer) = self.peers.get_mut(&who) {
+			if let PeerSyncState::DownloadingJustification(hash) = peer.state {
+				peer.state = PeerSyncState::Available;
+
+				// we only request one justification at a time
+				match response.blocks.into_iter().next() {
+					Some(response) => {
+						if hash != response.hash {
+							let msg = format!(
+								"Invalid block justification provided: requested: {:?} got: {:?}",
+								hash,
+								response.hash,
+							);
+
+							protocol.report_peer(who, Severity::Bad(&msg));
+							return;
+						}
+
+						self.justifications.on_response(
+							who,
+							response.justification,
+							protocol,
+							&*self.import_queue,
+						);
+					},
+					None => {
+						let msg = format!(
+							"Provided empty response for justification request {:?}",
+							hash,
+						);
+
+						protocol.report_peer(who, Severity::Useless(&msg));
+						return;
+					},
+				}
+			}
+		}
+
+		self.maintain_sync(protocol);
+	}
+
+	/// Maintain the sync process (download new blocks, fetch justifications).
 	pub fn maintain_sync(&mut self, protocol: &mut Context<B>) {
 		let peers: Vec<NodeIndex> = self.peers.keys().map(|p| *p).collect();
 		for peer in peers {
 			self.download_new(protocol, peer);
 		}
+		self.justifications.dispatch(&mut self.peers, protocol);
 	}
 
+	/// Called periodically to perform any time-based actions.
+	pub fn tick(&mut self, protocol: &mut Context<B>) {
+		self.justifications.dispatch(&mut self.peers, protocol);
+	}
+
+	/// Request a justification for the given block.
+	///
+	/// Queues a new justification request and tries to dispatch all pending requests.
+	pub fn request_justification(&mut self, hash: &B::Hash, number: NumberFor<B>, protocol: &mut Context<B>) {
+		self.justifications.queue_request(&(*hash, number));
+		self.justifications.dispatch(&mut self.peers, protocol);
+	}
+
+	/// Notify about successful import of the given block.
 	pub fn block_imported(&mut self, hash: &B::Hash, number: NumberFor<B>) {
 		trace!(target: "sync", "Block imported successfully {} ({})", number, hash);
 	}
 
+	/// Notify about finalization of the given block.
+	pub fn block_finalized(&mut self, _hash: &B::Hash, number: NumberFor<B>) {
+		self.justifications.collect_garbage(number);
+	}
+
 	fn block_queued(&mut self, hash: &B::Hash, number: NumberFor<B>) {
 		if number > self.best_queued_number {
 			self.best_queued_number = number;
@@ -324,6 +575,7 @@ impl<B: BlockT> ChainSync<B> {
 		self.block_queued(&hash, best_header.number().clone())
 	}
 
+	/// Handle new block announcement.
 	pub(crate) fn on_block_announce(&mut self, protocol: &mut Context<B>, who: NodeIndex, hash: B::Hash, header: &B::Header) {
 		let number = *header.number();
 		if number <= As::sa(0) {
@@ -376,12 +628,15 @@ impl<B: BlockT> ChainSync<B> {
 		block_status(&*protocol.client(), &*self.import_queue, *hash).ok().map_or(false, |s| s != BlockStatus::Unknown)
 	}
 
+	/// Handle disconnected peer.
 	pub(crate) fn peer_disconnected(&mut self, protocol: &mut Context<B>, who: NodeIndex) {
 		self.blocks.clear_peer_download(who);
 		self.peers.remove(&who);
+		self.justifications.peer_disconnected(who);
 		self.maintain_sync(protocol);
 	}
 
+	/// Restart the sync process.
 	pub(crate) fn restart(&mut self, protocol: &mut Context<B>) {
 		self.import_queue.clear();
 		self.blocks.clear();
@@ -403,6 +658,7 @@ impl<B: BlockT> ChainSync<B> {
 		}
 	}
 
+	/// Clear all sync data.
 	pub(crate) fn clear(&mut self) {
 		self.blocks.clear();
 		self.peers.clear();

substrate/core/network/src/test/mod.rs

@@ -30,7 +30,7 @@ use client::block_builder::BlockBuilder;
 use primitives::Ed25519AuthorityId;
 use runtime_primitives::Justification;
 use runtime_primitives::generic::BlockId;
-use runtime_primitives::traits::{Block as BlockT, Zero, Header, Digest, DigestItem, AuthorityIdFor};
+use runtime_primitives::traits::{AuthorityIdFor, Block as BlockT, Digest, DigestItem, Header, NumberFor, Zero};
 use io::SyncIo;
 use protocol::{Context, Protocol, ProtocolContext};
 use config::ProtocolConfig;
@@ -190,6 +190,15 @@ impl<B: 'static + BlockT, V: 'static + Verifier<B>> ImportQueue<B> for SyncImpor
 	fn import_blocks(&self, origin: BlockOrigin, blocks: Vec<IncomingBlock<B>>) {
 		self.link.call(origin, blocks);
 	}
+
+	fn import_justification(
+		&self,
+		hash: B::Hash,
+		number: NumberFor<B>,
+		justification: Justification,
+	) -> bool {
+		self.block_import.import_justification(hash, number, justification).is_ok()
+	}
 }
 
 struct DummyContextExecutor(Arc<Protocol<Block, DummySpecialization, Hash>>, Arc<RwLock<VecDeque<TestPacket>>>);
@@ -366,6 +375,13 @@ impl<V: 'static + Verifier<Block>, D> Peer<V, D> {
 		self.sync.on_block_imported(&mut TestIo::new(&self.queue, None), info.chain.best_hash, &header);
 	}
 
+	/// Send block finalization notifications.
+	pub fn send_finality_notifications(&self) {
+		let info = self.client.info().expect("In-mem client does not fail");
+		let header = self.client.header(&BlockId::Hash(info.chain.finalized_hash)).unwrap().unwrap();
+		self.sync.on_block_finalized(&mut TestIo::new(&self.queue, None), info.chain.finalized_hash, &header);
+	}
+
 	/// Restart sync for a peer.
 	fn restart_sync(&self) {
 		self.sync.abort();
@@ -380,6 +396,14 @@ impl<V: 'static + Verifier<Block>, D> Peer<V, D> {
 		self.sync.gossip_consensus_message(&mut TestIo::new(&self.queue, None), topic, data, broadcast);
 	}
 
+	/// Request a justification for the given block.
+	#[cfg(test)]
+	fn request_justification(&self, hash: &::primitives::H256, number: NumberFor<Block>) {
+		self.executor.execute_in_context(|context| {
+			self.sync.sync().write().request_justification(hash, number, context);
+		})
+	}
+
 	/// Add blocks to the peer -- edit the block before adding
 	pub fn generate_blocks<F>(&self, count: usize, origin: BlockOrigin, mut edit_block: F)
 		where F: FnMut(BlockBuilder<Block, (), PeersClient>) -> Block
@@ -601,6 +625,15 @@ pub trait TestNetFactory: Sized {
 		})
 	}
 
+	/// Send block finalization notifications for all peers.
+	fn send_finality_notifications(&mut self) {
+		self.mut_peers(|peers| {
+			for peer in peers {
+				peer.send_finality_notifications();
+			}
+		})
+	}
+
 	/// Restart sync for a peer.
 	fn restart_peer(&mut self, i: usize) {
 		self.peers()[i].restart_sync();

substrate/core/network/src/test/sync.rs

@@ -65,6 +65,27 @@ fn sync_no_common_longer_chain_fails() {
 	assert!(!net.peer(0).client.backend().blockchain().canon_equals_to(net.peer(1).client.backend().blockchain()));
 }
 
+#[test]
+fn sync_justifications() {
+	::env_logger::init().ok();
+	let mut net = TestNet::new(3);
+	net.peer(0).push_blocks(20, false);
+	net.sync();
+
+	// there's currently no justification for block #10
+	assert_eq!(net.peer(0).client().justification(&BlockId::Number(10)).unwrap(), None);
+	// we finalize block #10 for peer 0 with a justification
+	net.peer(0).client().finalize_block(BlockId::Number(10), Some(Vec::new()), true).unwrap();
+
+	let header = net.peer(1).client().header(&BlockId::Number(10)).unwrap().unwrap();
+	net.peer(1).request_justification(&header.hash().into(), 10);
+
+	net.sync();
+
+	assert_eq!(net.peer(0).client().justification(&BlockId::Number(10)).unwrap(), Some(Vec::new()));
+	assert_eq!(net.peer(1).client().justification(&BlockId::Number(10)).unwrap(), Some(Vec::new()));
+}
+
 #[test]
 fn sync_after_fork_works() {
 	::env_logger::init().ok();

substrate/core/service/src/lib.rs

@@ -78,7 +78,7 @@ pub use self::error::{ErrorKind, Error};
 pub use config::{Configuration, Roles, PruningMode};
 pub use chain_spec::{ChainSpec, Properties};
 pub use transaction_pool::txpool::{self, Pool as TransactionPool, Options as TransactionPoolOptions, ChainApi, IntoPoolError};
-pub use client::ExecutionStrategy;
+pub use client::{ExecutionStrategy, FinalityNotifications};
 
 pub use components::{ServiceFactory, FullBackend, FullExecutor, LightBackend,
 	LightExecutor, Components, PoolApi, ComponentClient,
@@ -222,6 +222,52 @@ impl<Components: components::Components> Service<Components> {
 			task_executor.spawn(events);
 		}
 
+		{
+			// finality notifications
+			let network = Arc::downgrade(&network);
+
+			// A utility stream that drops all ready items and only returns the last one.
+			// This is used to only keep the last finality notification and avoid
+			// overloading the sync module with notifications.
+			struct MostRecentNotification<B: network::BlockT>(futures::stream::Fuse<FinalityNotifications<B>>);
+
+			impl<B: network::BlockT> Stream for MostRecentNotification<B> {
+				type Item = <FinalityNotifications<B> as Stream>::Item;
+				type Error = <FinalityNotifications<B> as Stream>::Error;
+
+				fn poll(&mut self) -> Poll<Option<Self::Item>, Self::Error> {
+					let mut last = None;
+					let last = loop {
+						match self.0.poll()? {
+							Async::Ready(Some(item)) => { last = Some(item) }
+							Async::Ready(None) => match last {
+								None => return Ok(Async::Ready(None)),
+								Some(last) => break last,
+							},
+							Async::NotReady => match last {
+								None => return Ok(Async::NotReady),
+								Some(last) => break last,
+							},
+						}
+					};
+
+					Ok(Async::Ready(Some(last)))
+				}
+			}
+
+			let events = MostRecentNotification(client.finality_notification_stream().fuse())
+				.for_each(move |notification| {
+					if let Some(network) = network.upgrade() {
+						network.on_block_finalized(notification.hash, &notification.header);
+					}
+					Ok(())
+				})
+				.select(exit.clone())
+				.then(|_| Ok(()));
+
+			task_executor.spawn(events);
+		}
+
 		{
 			// extrinsic notifications
 			let network = Arc::downgrade(&network);
@@ -554,7 +600,7 @@ macro_rules! construct_service_factory {
 
 			fn new_full(
 				config: $crate::FactoryFullConfiguration<Self>,
-				executor: $crate::TaskExecutor
+				executor: $crate::TaskExecutor,
 			) -> Result<Self::FullService, $crate::Error>
 			{
 				( $( $full_service_init )* ) (config, executor.clone()).and_then(|service| {