Files
pezkuwi-subxt/substrate/client/network/sync/src/block_request_handler.rs
T
Aaro Altonen 80616f6d03 Integrate litep2p into Polkadot SDK (#2944)
[litep2p](https://github.com/altonen/litep2p) is a libp2p-compatible P2P
networking library. It supports all of the features of `rust-libp2p`
that are currently being utilized by Polkadot SDK.

Compared to `rust-libp2p`, `litep2p` has a quite different architecture
which is why the new `litep2p` network backend is only able to use a
little of the existing code in `sc-network`. The design has been mainly
influenced by how we'd wish to structure our networking-related code in
Polkadot SDK: independent higher-levels protocols directly communicating
with the network over links that support bidirectional backpressure. A
good example would be `NotificationHandle`/`RequestResponseHandle`
abstractions which allow, e.g., `SyncingEngine` to directly communicate
with peers to announce/request blocks.

I've tried running `polkadot --network-backend litep2p` with a few
different peer configurations and there is a noticeable reduction in
networking CPU usage. For high load (`--out-peers 200`), networking CPU
usage goes down from ~110% to ~30% (80 pp) and for normal load
(`--out-peers 40`), the usage goes down from ~55% to ~18% (37 pp).

These should not be taken as final numbers because:

a) there are still some low-hanging optimization fruits, such as
enabling [receive window
auto-tuning](https://github.com/libp2p/rust-yamux/pull/176), integrating
`Peerset` more closely with `litep2p` or improving memory usage of the
WebSocket transport
b) fixing bugs/instabilities that incorrectly cause `litep2p` to do less
work will increase the networking CPU usage
c) verification in a more diverse set of tests/conditions is needed

Nevertheless, these numbers should give an early estimate for CPU usage
of the new networking backend.

This PR consists of three separate changes:
* introduce a generic `PeerId` (wrapper around `Multihash`) so that we
don't have use `NetworkService::PeerId` in every part of the code that
uses a `PeerId`
* introduce `NetworkBackend` trait, implement it for the libp2p network
stack and make Polkadot SDK generic over `NetworkBackend`
  * implement `NetworkBackend` for litep2p

The new library should be considered experimental which is why
`rust-libp2p` will remain as the default option for the time being. This
PR currently depends on the master branch of `litep2p` but I'll cut a
new release for the library once all review comments have been
addresses.

---------

Signed-off-by: Alexandru Vasile <alexandru.vasile@parity.io>
Co-authored-by: Dmitry Markin <dmitry@markin.tech>
Co-authored-by: Alexandru Vasile <60601340+lexnv@users.noreply.github.com>
Co-authored-by: Alexandru Vasile <alexandru.vasile@parity.io>
2024-04-08 16:44:13 +00:00

622 lines
18 KiB
Rust

// Copyright (C) Parity Technologies (UK) Ltd.
// This file is part of Substrate.
// Substrate 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.
// Substrate 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 Substrate. If not, see <http://www.gnu.org/licenses/>.
//! Helper for handling (i.e. answering) block requests from a remote peer via the
//! `crate::request_responses::RequestResponsesBehaviour`.
use crate::{
block_relay_protocol::{BlockDownloader, BlockRelayParams, BlockResponseError, BlockServer},
schema::v1::{
block_request::FromBlock as FromBlockSchema, BlockRequest as BlockRequestSchema,
BlockResponse as BlockResponseSchema, BlockResponse, Direction,
},
service::network::NetworkServiceHandle,
LOG_TARGET,
};
use codec::{Decode, DecodeAll, Encode};
use futures::{channel::oneshot, stream::StreamExt};
use log::debug;
use prost::Message;
use schnellru::{ByLength, LruMap};
use sc_client_api::BlockBackend;
use sc_network::{
config::ProtocolId,
request_responses::{IfDisconnected, IncomingRequest, OutgoingResponse, RequestFailure},
service::traits::RequestResponseConfig,
types::ProtocolName,
NetworkBackend,
};
use sc_network_common::sync::message::{BlockAttributes, BlockData, BlockRequest, FromBlock};
use sc_network_types::PeerId;
use sp_blockchain::HeaderBackend;
use sp_runtime::{
generic::BlockId,
traits::{Block as BlockT, Header, One, Zero},
};
use std::{
cmp::min,
hash::{Hash, Hasher},
sync::Arc,
time::Duration,
};
/// Maximum blocks per response.
pub(crate) const MAX_BLOCKS_IN_RESPONSE: usize = 128;
const MAX_BODY_BYTES: usize = 8 * 1024 * 1024;
const MAX_NUMBER_OF_SAME_REQUESTS_PER_PEER: usize = 2;
mod rep {
use sc_network::ReputationChange as Rep;
/// Reputation change when a peer sent us the same request multiple times.
pub const SAME_REQUEST: Rep = Rep::new_fatal("Same block request multiple times");
/// Reputation change when a peer sent us the same "small" request multiple times.
pub const SAME_SMALL_REQUEST: Rep =
Rep::new(-(1 << 10), "same small block request multiple times");
}
/// Generates a `RequestResponseProtocolConfig` for the block request protocol,
/// refusing incoming requests.
pub fn generate_protocol_config<
Hash: AsRef<[u8]>,
B: BlockT,
N: NetworkBackend<B, <B as BlockT>::Hash>,
>(
protocol_id: &ProtocolId,
genesis_hash: Hash,
fork_id: Option<&str>,
inbound_queue: async_channel::Sender<IncomingRequest>,
) -> N::RequestResponseProtocolConfig {
N::request_response_config(
generate_protocol_name(genesis_hash, fork_id).into(),
std::iter::once(generate_legacy_protocol_name(protocol_id).into()).collect(),
1024 * 1024,
16 * 1024 * 1024,
Duration::from_secs(20),
Some(inbound_queue),
)
}
/// Generate the block protocol name from the genesis hash and fork id.
fn generate_protocol_name<Hash: AsRef<[u8]>>(genesis_hash: Hash, fork_id: Option<&str>) -> String {
let genesis_hash = genesis_hash.as_ref();
if let Some(fork_id) = fork_id {
format!("/{}/{}/sync/2", array_bytes::bytes2hex("", genesis_hash), fork_id)
} else {
format!("/{}/sync/2", array_bytes::bytes2hex("", genesis_hash))
}
}
/// Generate the legacy block protocol name from chain specific protocol identifier.
fn generate_legacy_protocol_name(protocol_id: &ProtocolId) -> String {
format!("/{}/sync/2", protocol_id.as_ref())
}
/// The key of [`BlockRequestHandler::seen_requests`].
#[derive(Eq, PartialEq, Clone)]
struct SeenRequestsKey<B: BlockT> {
peer: PeerId,
from: BlockId<B>,
max_blocks: usize,
direction: Direction,
attributes: BlockAttributes,
support_multiple_justifications: bool,
}
#[allow(clippy::derived_hash_with_manual_eq)]
impl<B: BlockT> Hash for SeenRequestsKey<B> {
fn hash<H: Hasher>(&self, state: &mut H) {
self.peer.hash(state);
self.max_blocks.hash(state);
self.direction.hash(state);
self.attributes.hash(state);
self.support_multiple_justifications.hash(state);
match self.from {
BlockId::Hash(h) => h.hash(state),
BlockId::Number(n) => n.hash(state),
}
}
}
/// The value of [`BlockRequestHandler::seen_requests`].
enum SeenRequestsValue {
/// First time we have seen the request.
First,
/// We have fulfilled the request `n` times.
Fulfilled(usize),
}
/// The full block server implementation of [`BlockServer`]. It handles
/// the incoming block requests from a remote peer.
pub struct BlockRequestHandler<B: BlockT, Client> {
client: Arc<Client>,
request_receiver: async_channel::Receiver<IncomingRequest>,
/// Maps from request to number of times we have seen this request.
///
/// This is used to check if a peer is spamming us with the same request.
seen_requests: LruMap<SeenRequestsKey<B>, SeenRequestsValue>,
}
impl<B, Client> BlockRequestHandler<B, Client>
where
B: BlockT,
Client: HeaderBackend<B> + BlockBackend<B> + Send + Sync + 'static,
{
/// Create a new [`BlockRequestHandler`].
pub fn new<N: NetworkBackend<B, <B as BlockT>::Hash>>(
network: NetworkServiceHandle,
protocol_id: &ProtocolId,
fork_id: Option<&str>,
client: Arc<Client>,
num_peer_hint: usize,
) -> BlockRelayParams<B, N> {
// Reserve enough request slots for one request per peer when we are at the maximum
// number of peers.
let capacity = std::cmp::max(num_peer_hint, 1);
let (tx, request_receiver) = async_channel::bounded(capacity);
let protocol_config = generate_protocol_config::<_, B, N>(
protocol_id,
client
.block_hash(0u32.into())
.ok()
.flatten()
.expect("Genesis block exists; qed"),
fork_id,
tx,
);
let capacity = ByLength::new(num_peer_hint.max(1) as u32 * 2);
let seen_requests = LruMap::new(capacity);
BlockRelayParams {
server: Box::new(Self { client, request_receiver, seen_requests }),
downloader: Arc::new(FullBlockDownloader::new(
protocol_config.protocol_name().clone(),
network,
)),
request_response_config: protocol_config,
}
}
/// Run [`BlockRequestHandler`].
async fn process_requests(&mut self) {
while let Some(request) = self.request_receiver.next().await {
let IncomingRequest { peer, payload, pending_response } = request;
match self.handle_request(payload, pending_response, &peer) {
Ok(()) => debug!(target: LOG_TARGET, "Handled block request from {}.", peer),
Err(e) => debug!(
target: LOG_TARGET,
"Failed to handle block request from {}: {}", peer, e,
),
}
}
}
fn handle_request(
&mut self,
payload: Vec<u8>,
pending_response: oneshot::Sender<OutgoingResponse>,
peer: &PeerId,
) -> Result<(), HandleRequestError> {
let request = crate::schema::v1::BlockRequest::decode(&payload[..])?;
let from_block_id = match request.from_block.ok_or(HandleRequestError::MissingFromField)? {
FromBlockSchema::Hash(ref h) => {
let h = Decode::decode(&mut h.as_ref())?;
BlockId::<B>::Hash(h)
},
FromBlockSchema::Number(ref n) => {
let n = Decode::decode(&mut n.as_ref())?;
BlockId::<B>::Number(n)
},
};
let max_blocks = if request.max_blocks == 0 {
MAX_BLOCKS_IN_RESPONSE
} else {
min(request.max_blocks as usize, MAX_BLOCKS_IN_RESPONSE)
};
let direction =
i32::try_into(request.direction).map_err(|_| HandleRequestError::ParseDirection)?;
let attributes = BlockAttributes::from_be_u32(request.fields)?;
let support_multiple_justifications = request.support_multiple_justifications;
let key = SeenRequestsKey {
peer: *peer,
max_blocks,
direction,
from: from_block_id,
attributes,
support_multiple_justifications,
};
let mut reputation_change = None;
let small_request = attributes
.difference(BlockAttributes::HEADER | BlockAttributes::JUSTIFICATION)
.is_empty();
match self.seen_requests.get(&key) {
Some(SeenRequestsValue::First) => {},
Some(SeenRequestsValue::Fulfilled(ref mut requests)) => {
*requests = requests.saturating_add(1);
if *requests > MAX_NUMBER_OF_SAME_REQUESTS_PER_PEER {
reputation_change = Some(if small_request {
rep::SAME_SMALL_REQUEST
} else {
rep::SAME_REQUEST
});
}
},
None => {
self.seen_requests.insert(key.clone(), SeenRequestsValue::First);
},
}
debug!(
target: LOG_TARGET,
"Handling block request from {peer}: Starting at `{from_block_id:?}` with \
maximum blocks of `{max_blocks}`, reputation_change: `{reputation_change:?}`, \
small_request `{small_request:?}`, direction `{direction:?}` and \
attributes `{attributes:?}`.",
);
let maybe_block_response = if reputation_change.is_none() || small_request {
let block_response = self.get_block_response(
attributes,
from_block_id,
direction,
max_blocks,
support_multiple_justifications,
)?;
// If any of the blocks contains any data, we can consider it as successful request.
if block_response
.blocks
.iter()
.any(|b| !b.header.is_empty() || !b.body.is_empty() || b.is_empty_justification)
{
if let Some(value) = self.seen_requests.get(&key) {
// If this is the first time we have processed this request, we need to change
// it to `Fulfilled`.
if let SeenRequestsValue::First = value {
*value = SeenRequestsValue::Fulfilled(1);
}
}
}
Some(block_response)
} else {
None
};
debug!(
target: LOG_TARGET,
"Sending result of block request from {peer} starting at `{from_block_id:?}`: \
blocks: {:?}, data: {:?}",
maybe_block_response.as_ref().map(|res| res.blocks.len()),
maybe_block_response.as_ref().map(|res| res.encoded_len()),
);
let result = if let Some(block_response) = maybe_block_response {
let mut data = Vec::with_capacity(block_response.encoded_len());
block_response.encode(&mut data)?;
Ok(data)
} else {
Err(())
};
pending_response
.send(OutgoingResponse {
result,
reputation_changes: reputation_change.into_iter().collect(),
sent_feedback: None,
})
.map_err(|_| HandleRequestError::SendResponse)
}
fn get_block_response(
&self,
attributes: BlockAttributes,
mut block_id: BlockId<B>,
direction: Direction,
max_blocks: usize,
support_multiple_justifications: bool,
) -> Result<BlockResponse, HandleRequestError> {
let get_header = attributes.contains(BlockAttributes::HEADER);
let get_body = attributes.contains(BlockAttributes::BODY);
let get_indexed_body = attributes.contains(BlockAttributes::INDEXED_BODY);
let get_justification = attributes.contains(BlockAttributes::JUSTIFICATION);
let mut blocks = Vec::new();
let mut total_size: usize = 0;
let client_header_from_block_id =
|block_id: BlockId<B>| -> Result<Option<B::Header>, HandleRequestError> {
if let Some(hash) = self.client.block_hash_from_id(&block_id)? {
return self.client.header(hash).map_err(Into::into)
}
Ok(None)
};
while let Some(header) = client_header_from_block_id(block_id).unwrap_or_default() {
let number = *header.number();
let hash = header.hash();
let parent_hash = *header.parent_hash();
let justifications =
if get_justification { self.client.justifications(hash)? } else { None };
let (justifications, justification, is_empty_justification) =
if support_multiple_justifications {
let justifications = match justifications {
Some(v) => v.encode(),
None => Vec::new(),
};
(justifications, Vec::new(), false)
} else {
// For now we keep compatibility by selecting precisely the GRANDPA one, and not
// just the first one. When sending we could have just taken the first one,
// since we don't expect there to be any other kind currently, but when
// receiving we need to add the engine ID tag.
// The ID tag is hardcoded here to avoid depending on the GRANDPA crate, and
// will be removed once we remove the backwards compatibility.
// See: https://github.com/paritytech/substrate/issues/8172
let justification =
justifications.and_then(|just| just.into_justification(*b"FRNK"));
let is_empty_justification =
justification.as_ref().map(|j| j.is_empty()).unwrap_or(false);
let justification = justification.unwrap_or_default();
(Vec::new(), justification, is_empty_justification)
};
let body = if get_body {
match self.client.block_body(hash)? {
Some(mut extrinsics) =>
extrinsics.iter_mut().map(|extrinsic| extrinsic.encode()).collect(),
None => {
log::trace!(target: LOG_TARGET, "Missing data for block request.");
break
},
}
} else {
Vec::new()
};
let indexed_body = if get_indexed_body {
match self.client.block_indexed_body(hash)? {
Some(transactions) => transactions,
None => {
log::trace!(
target: LOG_TARGET,
"Missing indexed block data for block request."
);
// If the indexed body is missing we still continue returning headers.
// Ideally `None` should distinguish a missing body from the empty body,
// but the current protobuf based protocol does not allow it.
Vec::new()
},
}
} else {
Vec::new()
};
let block_data = crate::schema::v1::BlockData {
hash: hash.encode(),
header: if get_header { header.encode() } else { Vec::new() },
body,
receipt: Vec::new(),
message_queue: Vec::new(),
justification,
is_empty_justification,
justifications,
indexed_body,
};
let new_total_size = total_size +
block_data.body.iter().map(|ex| ex.len()).sum::<usize>() +
block_data.indexed_body.iter().map(|ex| ex.len()).sum::<usize>();
// Send at least one block, but make sure to not exceed the limit.
if !blocks.is_empty() && new_total_size > MAX_BODY_BYTES {
break
}
total_size = new_total_size;
blocks.push(block_data);
if blocks.len() >= max_blocks as usize {
break
}
match direction {
Direction::Ascending => block_id = BlockId::Number(number + One::one()),
Direction::Descending => {
if number.is_zero() {
break
}
block_id = BlockId::Hash(parent_hash)
},
}
}
Ok(BlockResponse { blocks })
}
}
#[async_trait::async_trait]
impl<B, Client> BlockServer<B> for BlockRequestHandler<B, Client>
where
B: BlockT,
Client: HeaderBackend<B> + BlockBackend<B> + Send + Sync + 'static,
{
async fn run(&mut self) {
self.process_requests().await;
}
}
#[derive(Debug, thiserror::Error)]
enum HandleRequestError {
#[error("Failed to decode request: {0}.")]
DecodeProto(#[from] prost::DecodeError),
#[error("Failed to encode response: {0}.")]
EncodeProto(#[from] prost::EncodeError),
#[error("Failed to decode block hash: {0}.")]
DecodeScale(#[from] codec::Error),
#[error("Missing `BlockRequest::from_block` field.")]
MissingFromField,
#[error("Failed to parse BlockRequest::direction.")]
ParseDirection,
#[error(transparent)]
Client(#[from] sp_blockchain::Error),
#[error("Failed to send response.")]
SendResponse,
}
/// The full block downloader implementation of [`BlockDownloader].
pub struct FullBlockDownloader {
protocol_name: ProtocolName,
network: NetworkServiceHandle,
}
impl FullBlockDownloader {
fn new(protocol_name: ProtocolName, network: NetworkServiceHandle) -> Self {
Self { protocol_name, network }
}
/// Extracts the blocks from the response schema.
fn blocks_from_schema<B: BlockT>(
&self,
request: &BlockRequest<B>,
response: BlockResponseSchema,
) -> Result<Vec<BlockData<B>>, String> {
response
.blocks
.into_iter()
.map(|block_data| {
Ok(BlockData::<B> {
hash: Decode::decode(&mut block_data.hash.as_ref())?,
header: if !block_data.header.is_empty() {
Some(Decode::decode(&mut block_data.header.as_ref())?)
} else {
None
},
body: if request.fields.contains(BlockAttributes::BODY) {
Some(
block_data
.body
.iter()
.map(|body| Decode::decode(&mut body.as_ref()))
.collect::<Result<Vec<_>, _>>()?,
)
} else {
None
},
indexed_body: if request.fields.contains(BlockAttributes::INDEXED_BODY) {
Some(block_data.indexed_body)
} else {
None
},
receipt: if !block_data.receipt.is_empty() {
Some(block_data.receipt)
} else {
None
},
message_queue: if !block_data.message_queue.is_empty() {
Some(block_data.message_queue)
} else {
None
},
justification: if !block_data.justification.is_empty() {
Some(block_data.justification)
} else if block_data.is_empty_justification {
Some(Vec::new())
} else {
None
},
justifications: if !block_data.justifications.is_empty() {
Some(DecodeAll::decode_all(&mut block_data.justifications.as_ref())?)
} else {
None
},
})
})
.collect::<Result<_, _>>()
.map_err(|error: codec::Error| error.to_string())
}
}
#[async_trait::async_trait]
impl<B: BlockT> BlockDownloader<B> for FullBlockDownloader {
async fn download_blocks(
&self,
who: PeerId,
request: BlockRequest<B>,
) -> Result<Result<(Vec<u8>, ProtocolName), RequestFailure>, oneshot::Canceled> {
// Build the request protobuf.
let bytes = BlockRequestSchema {
fields: request.fields.to_be_u32(),
from_block: match request.from {
FromBlock::Hash(h) => Some(FromBlockSchema::Hash(h.encode())),
FromBlock::Number(n) => Some(FromBlockSchema::Number(n.encode())),
},
direction: request.direction as i32,
max_blocks: request.max.unwrap_or(0),
support_multiple_justifications: true,
}
.encode_to_vec();
let (tx, rx) = oneshot::channel();
self.network.start_request(
who,
self.protocol_name.clone(),
bytes,
tx,
IfDisconnected::ImmediateError,
);
rx.await
}
fn block_response_into_blocks(
&self,
request: &BlockRequest<B>,
response: Vec<u8>,
) -> Result<Vec<BlockData<B>>, BlockResponseError> {
// Decode the response protobuf
let response_schema = BlockResponseSchema::decode(response.as_slice())
.map_err(|error| BlockResponseError::DecodeFailed(error.to_string()))?;
// Extract the block data from the protobuf
self.blocks_from_schema::<B>(request, response_schema)
.map_err(|error| BlockResponseError::ExtractionFailed(error.to_string()))
}
}