Files
lighthouse/beacon_node/execution_layer/src/lib.rs
Pawan Dhananjay 7bfdb33729 Return error if getBlobs not supported (#6911)
N/A


  Previously, we were returning an empty vec of Nones if get_blobs was not supported in the EL. This results in confusing logging where we try to process the empty list of blobs and log a bunch of Unexpected errors. See
```
Feb 03 17:32:12.383 DEBG Fetching blobs from the EL, num_expected_blobs: 6, block_root: 0x7326fe2dc1cb9036c9de7a07a662c86a339085597849016eadf061b70b7815ba, service: fetch_engine_blobs, service: beacon, module: beac
on_chain::fetch_blobs:84
Feb 03 17:32:12.384 DEBG Processing engine blobs, num_fetched_blobs: 0, block_root: 0x7326fe2dc1cb9036c9de7a07a662c86a339085597849016eadf061b70b7815ba, service: fetch_engine_blobs, service: beacon, module: beacon_c
hain::fetch_blobs:197
Feb 03 17:32:12.384 ERRO Error fetching or processing blobs from EL, block_root: 0x7326fe2dc1cb9036c9de7a07a662c86a339085597849016eadf061b70b7815ba, error: BlobProcessingError(AvailabilityCheck(Unexpected)), module
: network::network_beacon_processor:1011
```

The error we should be getting is that getBlobs is not supported, this PR adds a new error variant and returns that.
2025-02-05 01:14:41 +00:00

2340 lines
87 KiB
Rust

//! This crate provides an abstraction over one or more *execution engines*. An execution engine
//! was formerly known as an "eth1 node", like Geth, Nethermind, Erigon, etc.
//!
//! This crate only provides useful functionality for "The Merge", it does not provide any of the
//! deposit-contract functionality that the `beacon_node/eth1` crate already provides.
use crate::json_structures::BlobAndProofV1;
use crate::payload_cache::PayloadCache;
use arc_swap::ArcSwapOption;
use auth::{strip_prefix, Auth, JwtKey};
pub use block_hash::calculate_execution_block_hash;
use builder_client::BuilderHttpClient;
pub use engine_api::EngineCapabilities;
use engine_api::Error as ApiError;
pub use engine_api::*;
pub use engine_api::{http, http::deposit_methods, http::HttpJsonRpc};
use engines::{Engine, EngineError};
pub use engines::{EngineState, ForkchoiceState};
use eth2::types::FullPayloadContents;
use eth2::types::{builder_bid::SignedBuilderBid, BlobsBundle, ForkVersionedResponse};
use ethers_core::types::Transaction as EthersTransaction;
use fixed_bytes::UintExtended;
use fork_choice::ForkchoiceUpdateParameters;
use lru::LruCache;
use payload_status::process_payload_status;
pub use payload_status::PayloadStatus;
use sensitive_url::SensitiveUrl;
use serde::{Deserialize, Serialize};
use slog::{crit, debug, error, info, warn, Logger};
use slot_clock::SlotClock;
use std::collections::{hash_map::Entry, HashMap};
use std::fmt;
use std::future::Future;
use std::io::Write;
use std::num::NonZeroUsize;
use std::path::PathBuf;
use std::sync::Arc;
use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
use strum::AsRefStr;
use task_executor::TaskExecutor;
use tokio::{
sync::{Mutex, MutexGuard, RwLock},
time::sleep,
};
use tokio_stream::wrappers::WatchStream;
use tree_hash::TreeHash;
use types::beacon_block_body::KzgCommitments;
use types::builder_bid::BuilderBid;
use types::non_zero_usize::new_non_zero_usize;
use types::payload::BlockProductionVersion;
use types::{
AbstractExecPayload, BlobsList, ExecutionPayloadDeneb, ExecutionRequests, KzgProofs,
SignedBlindedBeaconBlock,
};
use types::{
BeaconStateError, BlindedPayload, ChainSpec, Epoch, ExecPayload, ExecutionPayloadBellatrix,
ExecutionPayloadCapella, ExecutionPayloadElectra, ExecutionPayloadFulu, FullPayload,
ProposerPreparationData, PublicKeyBytes, Signature, Slot,
};
mod block_hash;
mod engine_api;
pub mod engines;
mod keccak;
mod metrics;
pub mod payload_cache;
mod payload_status;
pub mod test_utils;
pub mod versioned_hashes;
/// Indicates the default jwt authenticated execution endpoint.
pub const DEFAULT_EXECUTION_ENDPOINT: &str = "http://localhost:8551/";
/// Name for the default file used for the jwt secret.
pub const DEFAULT_JWT_FILE: &str = "jwt.hex";
/// Each time the `ExecutionLayer` retrieves a block from an execution node, it stores that block
/// in an LRU cache to avoid redundant lookups. This is the size of that cache.
const EXECUTION_BLOCKS_LRU_CACHE_SIZE: NonZeroUsize = new_non_zero_usize(128);
/// A fee recipient address for use during block production. Only used as a very last resort if
/// there is no address provided by the user.
///
/// ## Note
///
/// This is *not* the zero-address, since Geth has been known to return errors for a coinbase of
/// 0x00..00.
const DEFAULT_SUGGESTED_FEE_RECIPIENT: [u8; 20] =
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1];
/// A payload alongside some information about where it came from.
pub enum ProvenancedPayload<P> {
/// A good old fashioned farm-to-table payload from your local EE.
Local(P),
/// A payload from a builder (e.g. mev-boost).
Builder(P),
}
impl<E: EthSpec> TryFrom<BuilderBid<E>> for ProvenancedPayload<BlockProposalContentsType<E>> {
type Error = Error;
fn try_from(value: BuilderBid<E>) -> Result<Self, Error> {
let block_proposal_contents = match value {
BuilderBid::Bellatrix(builder_bid) => BlockProposalContents::Payload {
payload: ExecutionPayloadHeader::Bellatrix(builder_bid.header).into(),
block_value: builder_bid.value,
},
BuilderBid::Capella(builder_bid) => BlockProposalContents::Payload {
payload: ExecutionPayloadHeader::Capella(builder_bid.header).into(),
block_value: builder_bid.value,
},
BuilderBid::Deneb(builder_bid) => BlockProposalContents::PayloadAndBlobs {
payload: ExecutionPayloadHeader::Deneb(builder_bid.header).into(),
block_value: builder_bid.value,
kzg_commitments: builder_bid.blob_kzg_commitments,
blobs_and_proofs: None,
requests: None,
},
BuilderBid::Electra(builder_bid) => BlockProposalContents::PayloadAndBlobs {
payload: ExecutionPayloadHeader::Electra(builder_bid.header).into(),
block_value: builder_bid.value,
kzg_commitments: builder_bid.blob_kzg_commitments,
blobs_and_proofs: None,
requests: Some(builder_bid.execution_requests),
},
BuilderBid::Fulu(builder_bid) => BlockProposalContents::PayloadAndBlobs {
payload: ExecutionPayloadHeader::Fulu(builder_bid.header).into(),
block_value: builder_bid.value,
kzg_commitments: builder_bid.blob_kzg_commitments,
blobs_and_proofs: None,
// TODO(fulu): update this with builder api returning the requests
requests: None,
},
};
Ok(ProvenancedPayload::Builder(
BlockProposalContentsType::Blinded(block_proposal_contents),
))
}
}
#[derive(Debug)]
pub enum Error {
NoEngine,
NoPayloadBuilder,
ApiError(ApiError),
Builder(builder_client::Error),
NoHeaderFromBuilder,
CannotProduceHeader,
EngineError(Box<EngineError>),
NotSynced,
ShuttingDown,
FeeRecipientUnspecified,
MissingLatestValidHash,
BlockHashMismatch {
computed: ExecutionBlockHash,
payload: ExecutionBlockHash,
transactions_root: Hash256,
},
ZeroLengthTransaction,
PayloadBodiesByRangeNotSupported,
GetBlobsNotSupported,
InvalidJWTSecret(String),
InvalidForkForPayload,
InvalidPayloadBody(String),
InvalidPayloadConversion,
InvalidBlobConversion(String),
BeaconStateError(BeaconStateError),
PayloadTypeMismatch,
VerifyingVersionedHashes(versioned_hashes::Error),
}
impl From<BeaconStateError> for Error {
fn from(e: BeaconStateError) -> Self {
Error::BeaconStateError(e)
}
}
impl From<ApiError> for Error {
fn from(e: ApiError) -> Self {
Error::ApiError(e)
}
}
impl From<EngineError> for Error {
fn from(e: EngineError) -> Self {
match e {
// This removes an unnecessary layer of indirection.
// TODO (mark): consider refactoring these error enums
EngineError::Api { error } => Error::ApiError(error),
_ => Error::EngineError(Box::new(e)),
}
}
}
pub enum BlockProposalContentsType<E: EthSpec> {
Full(BlockProposalContents<E, FullPayload<E>>),
Blinded(BlockProposalContents<E, BlindedPayload<E>>),
}
pub enum BlockProposalContents<E: EthSpec, Payload: AbstractExecPayload<E>> {
Payload {
payload: Payload,
block_value: Uint256,
},
PayloadAndBlobs {
payload: Payload,
block_value: Uint256,
kzg_commitments: KzgCommitments<E>,
/// `None` for blinded `PayloadAndBlobs`.
blobs_and_proofs: Option<(BlobsList<E>, KzgProofs<E>)>,
// TODO(electra): this should probably be a separate variant/superstruct
requests: Option<ExecutionRequests<E>>,
},
}
impl<E: EthSpec> From<BlockProposalContents<E, FullPayload<E>>>
for BlockProposalContents<E, BlindedPayload<E>>
{
fn from(item: BlockProposalContents<E, FullPayload<E>>) -> Self {
match item {
BlockProposalContents::Payload {
payload,
block_value,
} => BlockProposalContents::Payload {
payload: payload.execution_payload().into(),
block_value,
},
BlockProposalContents::PayloadAndBlobs {
payload,
block_value,
kzg_commitments,
blobs_and_proofs: _,
requests,
} => BlockProposalContents::PayloadAndBlobs {
payload: payload.execution_payload().into(),
block_value,
kzg_commitments,
blobs_and_proofs: None,
requests,
},
}
}
}
impl<E: EthSpec, Payload: AbstractExecPayload<E>> TryFrom<GetPayloadResponse<E>>
for BlockProposalContents<E, Payload>
{
type Error = Error;
fn try_from(response: GetPayloadResponse<E>) -> Result<Self, Error> {
let (execution_payload, block_value, maybe_bundle, maybe_requests) = response.into();
match maybe_bundle {
Some(bundle) => Ok(Self::PayloadAndBlobs {
payload: execution_payload.into(),
block_value,
kzg_commitments: bundle.commitments,
blobs_and_proofs: Some((bundle.blobs, bundle.proofs)),
requests: maybe_requests,
}),
None => Ok(Self::Payload {
payload: execution_payload.into(),
block_value,
}),
}
}
}
impl<E: EthSpec> TryFrom<GetPayloadResponseType<E>> for BlockProposalContentsType<E> {
type Error = Error;
fn try_from(response_type: GetPayloadResponseType<E>) -> Result<Self, Error> {
match response_type {
GetPayloadResponseType::Full(response) => Ok(Self::Full(response.try_into()?)),
GetPayloadResponseType::Blinded(response) => Ok(Self::Blinded(response.try_into()?)),
}
}
}
#[allow(clippy::type_complexity)]
impl<E: EthSpec, Payload: AbstractExecPayload<E>> BlockProposalContents<E, Payload> {
pub fn deconstruct(
self,
) -> (
Payload,
Option<KzgCommitments<E>>,
Option<(BlobsList<E>, KzgProofs<E>)>,
Option<ExecutionRequests<E>>,
Uint256,
) {
match self {
Self::Payload {
payload,
block_value,
} => (payload, None, None, None, block_value),
Self::PayloadAndBlobs {
payload,
block_value,
kzg_commitments,
blobs_and_proofs,
requests,
} => (
payload,
Some(kzg_commitments),
blobs_and_proofs,
requests,
block_value,
),
}
}
pub fn payload(&self) -> &Payload {
match self {
Self::Payload { payload, .. } => payload,
Self::PayloadAndBlobs { payload, .. } => payload,
}
}
pub fn to_payload(self) -> Payload {
match self {
Self::Payload { payload, .. } => payload,
Self::PayloadAndBlobs { payload, .. } => payload,
}
}
pub fn block_value(&self) -> &Uint256 {
match self {
Self::Payload { block_value, .. } => block_value,
Self::PayloadAndBlobs { block_value, .. } => block_value,
}
}
}
// This just groups together a bunch of parameters that commonly
// get passed around together in calls to get_payload.
#[derive(Clone, Copy, Debug)]
pub struct PayloadParameters<'a> {
pub parent_hash: ExecutionBlockHash,
pub parent_gas_limit: u64,
pub proposer_gas_limit: Option<u64>,
pub payload_attributes: &'a PayloadAttributes,
pub forkchoice_update_params: &'a ForkchoiceUpdateParameters,
pub current_fork: ForkName,
}
#[derive(Clone, PartialEq)]
pub struct ProposerPreparationDataEntry {
update_epoch: Epoch,
preparation_data: ProposerPreparationData,
gas_limit: Option<u64>,
}
impl ProposerPreparationDataEntry {
pub fn update(&mut self, updated: Self) -> bool {
let mut changed = false;
// Update `gas_limit` if `updated.gas_limit` is `Some` and:
// - `self.gas_limit` is `None`, or
// - both are `Some` but the values differ.
if let Some(updated_gas_limit) = updated.gas_limit {
if self.gas_limit != Some(updated_gas_limit) {
self.gas_limit = Some(updated_gas_limit);
changed = true;
}
}
// Update `update_epoch` if it differs
if self.update_epoch != updated.update_epoch {
self.update_epoch = updated.update_epoch;
changed = true;
}
// Update `preparation_data` if it differs
if self.preparation_data != updated.preparation_data {
self.preparation_data = updated.preparation_data;
changed = true;
}
changed
}
}
#[derive(Hash, PartialEq, Eq)]
pub struct ProposerKey {
slot: Slot,
head_block_root: Hash256,
}
#[derive(PartialEq, Clone)]
pub struct Proposer {
validator_index: u64,
payload_attributes: PayloadAttributes,
}
/// Information from the beacon chain that is necessary for querying the builder API.
pub struct BuilderParams {
pub pubkey: PublicKeyBytes,
pub slot: Slot,
pub chain_health: ChainHealth,
}
#[derive(PartialEq)]
pub enum ChainHealth {
Healthy,
Unhealthy(FailedCondition),
Optimistic,
PreMerge,
}
#[derive(Debug, PartialEq)]
pub enum FailedCondition {
Skips,
SkipsPerEpoch,
EpochsSinceFinalization,
}
type PayloadContentsRefTuple<'a, E> = (ExecutionPayloadRef<'a, E>, Option<&'a BlobsBundle<E>>);
struct Inner<E: EthSpec> {
engine: Arc<Engine>,
builder: ArcSwapOption<BuilderHttpClient>,
execution_engine_forkchoice_lock: Mutex<()>,
suggested_fee_recipient: Option<Address>,
proposer_preparation_data: Mutex<HashMap<u64, ProposerPreparationDataEntry>>,
execution_blocks: Mutex<LruCache<ExecutionBlockHash, ExecutionBlock>>,
proposers: RwLock<HashMap<ProposerKey, Proposer>>,
executor: TaskExecutor,
payload_cache: PayloadCache<E>,
log: Logger,
/// Track whether the last `newPayload` call errored.
///
/// This is used *only* in the informational sync status endpoint, so that a VC using this
/// node can prefer another node with a healthier EL.
last_new_payload_errored: RwLock<bool>,
}
#[derive(Debug, Default, Clone, Serialize, Deserialize)]
pub struct Config {
/// Endpoint url for EL nodes that are running the engine api.
pub execution_endpoint: Option<SensitiveUrl>,
/// Endpoint urls for services providing the builder api.
pub builder_url: Option<SensitiveUrl>,
/// The timeout value used when making a request to fetch a block header
/// from the builder api.
pub builder_header_timeout: Option<Duration>,
/// User agent to send with requests to the builder API.
pub builder_user_agent: Option<String>,
/// JWT secret for the above endpoint running the engine api.
pub secret_file: Option<PathBuf>,
/// The default fee recipient to use on the beacon node if none if provided from
/// the validator client during block preparation.
pub suggested_fee_recipient: Option<Address>,
/// An optional id for the beacon node that will be passed to the EL in the JWT token claim.
pub jwt_id: Option<String>,
/// An optional client version for the beacon node that will be passed to the EL in the JWT token claim.
pub jwt_version: Option<String>,
/// Default directory for the jwt secret if not provided through cli.
pub default_datadir: PathBuf,
pub execution_timeout_multiplier: Option<u32>,
}
/// Provides access to one execution engine and provides a neat interface for consumption by the
/// `BeaconChain`.
#[derive(Clone)]
pub struct ExecutionLayer<E: EthSpec> {
inner: Arc<Inner<E>>,
}
impl<E: EthSpec> ExecutionLayer<E> {
/// Instantiate `Self` with an Execution engine specified in `Config`, using JSON-RPC via HTTP.
pub fn from_config(config: Config, executor: TaskExecutor, log: Logger) -> Result<Self, Error> {
let Config {
execution_endpoint: url,
builder_url,
builder_user_agent,
builder_header_timeout,
secret_file,
suggested_fee_recipient,
jwt_id,
jwt_version,
default_datadir,
execution_timeout_multiplier,
} = config;
let execution_url = url.ok_or(Error::NoEngine)?;
// Use the default jwt secret path if not provided via cli.
let secret_file = secret_file.unwrap_or_else(|| default_datadir.join(DEFAULT_JWT_FILE));
let jwt_key = if secret_file.exists() {
// Read secret from file if it already exists
std::fs::read_to_string(&secret_file)
.map_err(|e| format!("Failed to read JWT secret file. Error: {:?}", e))
.and_then(|ref s| {
let secret = JwtKey::from_slice(
&hex::decode(strip_prefix(s.trim_end()))
.map_err(|e| format!("Invalid hex string: {:?}", e))?,
)?;
Ok(secret)
})
.map_err(Error::InvalidJWTSecret)
} else {
// Create a new file and write a randomly generated secret to it if file does not exist
warn!(log, "No JWT found on disk. Generating"; "path" => %secret_file.display());
std::fs::File::options()
.write(true)
.create_new(true)
.open(&secret_file)
.map_err(|e| format!("Failed to open JWT secret file. Error: {:?}", e))
.and_then(|mut f| {
let secret = auth::JwtKey::random();
f.write_all(secret.hex_string().as_bytes())
.map_err(|e| format!("Failed to write to JWT secret file: {:?}", e))?;
Ok(secret)
})
.map_err(Error::InvalidJWTSecret)
}?;
let engine: Engine = {
let auth = Auth::new(jwt_key, jwt_id, jwt_version);
debug!(log, "Loaded execution endpoint"; "endpoint" => %execution_url, "jwt_path" => ?secret_file.as_path());
let api = HttpJsonRpc::new_with_auth(execution_url, auth, execution_timeout_multiplier)
.map_err(Error::ApiError)?;
Engine::new(api, executor.clone(), &log)
};
let inner = Inner {
engine: Arc::new(engine),
builder: ArcSwapOption::empty(),
execution_engine_forkchoice_lock: <_>::default(),
suggested_fee_recipient,
proposer_preparation_data: Mutex::new(HashMap::new()),
proposers: RwLock::new(HashMap::new()),
execution_blocks: Mutex::new(LruCache::new(EXECUTION_BLOCKS_LRU_CACHE_SIZE)),
executor,
payload_cache: PayloadCache::default(),
log,
last_new_payload_errored: RwLock::new(false),
};
let el = Self {
inner: Arc::new(inner),
};
if let Some(builder_url) = builder_url {
el.set_builder_url(builder_url, builder_user_agent, builder_header_timeout)?;
}
Ok(el)
}
fn engine(&self) -> &Arc<Engine> {
&self.inner.engine
}
pub fn builder(&self) -> Option<Arc<BuilderHttpClient>> {
self.inner.builder.load_full()
}
/// Set the builder URL after initialization.
///
/// This is useful for breaking circular dependencies between mock ELs and mock builders in
/// tests.
pub fn set_builder_url(
&self,
builder_url: SensitiveUrl,
builder_user_agent: Option<String>,
builder_header_timeout: Option<Duration>,
) -> Result<(), Error> {
let builder_client = BuilderHttpClient::new(
builder_url.clone(),
builder_user_agent,
builder_header_timeout,
)
.map_err(Error::Builder)?;
info!(
self.log(),
"Using external block builder";
"builder_url" => ?builder_url,
"local_user_agent" => builder_client.get_user_agent(),
);
self.inner.builder.swap(Some(Arc::new(builder_client)));
Ok(())
}
/// Cache a full payload, keyed on the `tree_hash_root` of the payload
fn cache_payload(
&self,
payload_and_blobs: PayloadContentsRefTuple<E>,
) -> Option<FullPayloadContents<E>> {
let (payload_ref, maybe_json_blobs_bundle) = payload_and_blobs;
let payload = payload_ref.clone_from_ref();
let maybe_blobs_bundle = maybe_json_blobs_bundle
.cloned()
.map(|blobs_bundle| BlobsBundle {
commitments: blobs_bundle.commitments,
proofs: blobs_bundle.proofs,
blobs: blobs_bundle.blobs,
});
self.inner
.payload_cache
.put(FullPayloadContents::new(payload, maybe_blobs_bundle))
}
/// Attempt to retrieve a full payload from the payload cache by the payload root
pub fn get_payload_by_root(&self, root: &Hash256) -> Option<FullPayloadContents<E>> {
self.inner.payload_cache.get(root)
}
pub fn executor(&self) -> &TaskExecutor {
&self.inner.executor
}
/// Get the current difficulty of the PoW chain.
pub async fn get_current_difficulty(&self) -> Result<Uint256, ApiError> {
let block = self
.engine()
.api
.get_block_by_number(BlockByNumberQuery::Tag(LATEST_TAG))
.await?
.ok_or(ApiError::ExecutionHeadBlockNotFound)?;
Ok(block.total_difficulty)
}
/// Note: this function returns a mutex guard, be careful to avoid deadlocks.
async fn execution_blocks(
&self,
) -> MutexGuard<'_, LruCache<ExecutionBlockHash, ExecutionBlock>> {
self.inner.execution_blocks.lock().await
}
/// Gives access to a channel containing if the last engine state is online or not.
///
/// This can be called several times.
pub async fn get_responsiveness_watch(&self) -> WatchStream<EngineState> {
self.engine().watch_state().await
}
/// Note: this function returns a mutex guard, be careful to avoid deadlocks.
async fn proposer_preparation_data(
&self,
) -> MutexGuard<'_, HashMap<u64, ProposerPreparationDataEntry>> {
self.inner.proposer_preparation_data.lock().await
}
fn proposers(&self) -> &RwLock<HashMap<ProposerKey, Proposer>> {
&self.inner.proposers
}
fn log(&self) -> &Logger {
&self.inner.log
}
pub async fn execution_engine_forkchoice_lock(&self) -> MutexGuard<'_, ()> {
self.inner.execution_engine_forkchoice_lock.lock().await
}
/// Convenience function to allow spawning a task without waiting for the result.
pub fn spawn<F, U>(&self, generate_future: F, name: &'static str)
where
F: FnOnce(Self) -> U,
U: Future<Output = ()> + Send + 'static,
{
self.executor().spawn(generate_future(self.clone()), name);
}
/// Spawns a routine which attempts to keep the execution engine online.
pub fn spawn_watchdog_routine<S: SlotClock + 'static>(&self, slot_clock: S) {
let watchdog = |el: ExecutionLayer<E>| async move {
// Run one task immediately.
el.watchdog_task().await;
// Start the loop to periodically update.
loop {
el.spawn(
|el| async move { el.watchdog_task().await },
"exec_watchdog_task",
);
sleep(slot_clock.slot_duration()).await;
}
};
self.spawn(watchdog, "exec_watchdog");
}
/// Performs a single execution of the watchdog routine.
pub async fn watchdog_task(&self) {
self.engine().upcheck().await;
}
/// Spawns a routine which cleans the cached proposer data periodically.
pub fn spawn_clean_proposer_caches_routine<S: SlotClock + 'static>(&self, slot_clock: S) {
let preparation_cleaner = |el: ExecutionLayer<E>| async move {
// Start the loop to periodically clean proposer preparation cache.
loop {
if let Some(duration_to_next_epoch) =
slot_clock.duration_to_next_epoch(E::slots_per_epoch())
{
// Wait for next epoch
sleep(duration_to_next_epoch).await;
match slot_clock
.now()
.map(|slot| slot.epoch(E::slots_per_epoch()))
{
Some(current_epoch) => el
.clean_proposer_caches(current_epoch)
.await
.map_err(|e| {
error!(
el.log(),
"Failed to clean proposer preparation cache";
"error" => format!("{:?}", e)
)
})
.unwrap_or(()),
None => error!(el.log(), "Failed to get current epoch from slot clock"),
}
} else {
error!(el.log(), "Failed to read slot clock");
// If we can't read the slot clock, just wait another slot and retry.
sleep(slot_clock.slot_duration()).await;
}
}
};
self.spawn(preparation_cleaner, "exec_preparation_cleanup");
}
/// Returns `true` if the execution engine is synced and reachable.
pub async fn is_synced(&self) -> bool {
self.engine().is_synced().await
}
/// Execution nodes return a "SYNCED" response when they do not have any peers.
///
/// This function is a wrapper over `Self::is_synced` that makes an additional
/// check for the execution layer sync status. Checks if the latest block has
/// a `block_number != 0` *if* the `current_slot` is also `> 0`.
/// Returns the `Self::is_synced` response if unable to get latest block.
pub async fn is_synced_for_notifier(&self, current_slot: Slot) -> bool {
let synced = self.is_synced().await;
if synced {
if let Ok(Some(block)) = self
.engine()
.api
.get_block_by_number(BlockByNumberQuery::Tag(LATEST_TAG))
.await
{
if block.block_number == 0 && current_slot > 0 {
return false;
}
}
}
synced
}
/// Return `true` if the execution layer is offline or returning errors on `newPayload`.
///
/// This function should never be used to prevent any operation in the beacon node, but can
/// be used to give an indication on the HTTP API that the node's execution layer is struggling,
/// which can in turn be used by the VC.
pub async fn is_offline_or_erroring(&self) -> bool {
self.engine().is_offline().await || *self.inner.last_new_payload_errored.read().await
}
/// Updates the proposer preparation data provided by validators
pub async fn update_proposer_preparation<'a, I>(&self, update_epoch: Epoch, proposer_data: I)
where
I: IntoIterator<Item = (&'a ProposerPreparationData, &'a Option<u64>)>,
{
let mut proposer_preparation_data = self.proposer_preparation_data().await;
for (preparation_entry, gas_limit) in proposer_data {
let new = ProposerPreparationDataEntry {
update_epoch,
preparation_data: preparation_entry.clone(),
gas_limit: *gas_limit,
};
match proposer_preparation_data.entry(preparation_entry.validator_index) {
Entry::Occupied(mut entry) => {
if entry.get_mut().update(new) {
metrics::inc_counter(&metrics::EXECUTION_LAYER_PROPOSER_DATA_UPDATED);
}
}
Entry::Vacant(entry) => {
entry.insert(new);
metrics::inc_counter(&metrics::EXECUTION_LAYER_PROPOSER_DATA_UPDATED);
}
}
}
}
/// Delete proposer preparation data for `proposer_index`. This is only useful in tests.
pub async fn clear_proposer_preparation(&self, proposer_index: u64) {
self.proposer_preparation_data()
.await
.remove(&proposer_index);
}
/// Removes expired entries from proposer_preparation_data and proposers caches
async fn clean_proposer_caches(&self, current_epoch: Epoch) -> Result<(), Error> {
let mut proposer_preparation_data = self.proposer_preparation_data().await;
// Keep all entries that have been updated in the last 2 epochs
let retain_epoch = current_epoch.saturating_sub(Epoch::new(2));
proposer_preparation_data.retain(|_validator_index, preparation_entry| {
preparation_entry.update_epoch >= retain_epoch
});
drop(proposer_preparation_data);
let retain_slot = retain_epoch.start_slot(E::slots_per_epoch());
self.proposers()
.write()
.await
.retain(|proposer_key, _proposer| proposer_key.slot >= retain_slot);
Ok(())
}
/// Returns `true` if there have been any validators registered via
/// `Self::update_proposer_preparation`.
pub async fn has_any_proposer_preparation_data(&self) -> bool {
!self.proposer_preparation_data().await.is_empty()
}
/// Returns `true` if the `proposer_index` has registered as a local validator via
/// `Self::update_proposer_preparation`.
pub async fn has_proposer_preparation_data(&self, proposer_index: u64) -> bool {
self.proposer_preparation_data()
.await
.contains_key(&proposer_index)
}
/// Check if a proposer is registered as a local validator, *from a synchronous context*.
///
/// This method MUST NOT be called from an async task.
pub fn has_proposer_preparation_data_blocking(&self, proposer_index: u64) -> bool {
self.inner
.proposer_preparation_data
.blocking_lock()
.contains_key(&proposer_index)
}
/// Returns the fee-recipient address that should be used to build a block
pub async fn get_suggested_fee_recipient(&self, proposer_index: u64) -> Address {
if let Some(preparation_data_entry) =
self.proposer_preparation_data().await.get(&proposer_index)
{
// The values provided via the API have first priority.
preparation_data_entry.preparation_data.fee_recipient
} else if let Some(address) = self.inner.suggested_fee_recipient {
// If there has been no fee recipient provided via the API, but the BN has been provided
// with a global default address, use that.
address
} else {
// If there is no user-provided fee recipient, use a junk value and complain loudly.
crit!(
self.log(),
"Fee recipient unknown";
"msg" => "the suggested_fee_recipient was unknown during block production. \
a junk address was used, rewards were lost! \
check the --suggested-fee-recipient flag and VC configuration.",
"proposer_index" => ?proposer_index
);
Address::from_slice(&DEFAULT_SUGGESTED_FEE_RECIPIENT)
}
}
pub async fn get_proposer_gas_limit(&self, proposer_index: u64) -> Option<u64> {
self.proposer_preparation_data()
.await
.get(&proposer_index)
.and_then(|entry| entry.gas_limit)
}
/// Maps to the `engine_getPayload` JSON-RPC call.
///
/// However, it will attempt to call `self.prepare_payload` if it cannot find an existing
/// payload id for the given parameters.
///
/// ## Fallback Behavior
///
/// The result will be returned from the first node that returns successfully. No more nodes
/// will be contacted.
pub async fn get_payload(
&self,
payload_parameters: PayloadParameters<'_>,
builder_params: BuilderParams,
spec: &ChainSpec,
builder_boost_factor: Option<u64>,
block_production_version: BlockProductionVersion,
) -> Result<BlockProposalContentsType<E>, Error> {
let payload_result_type = match block_production_version {
BlockProductionVersion::V3 => match self
.determine_and_fetch_payload(
payload_parameters,
builder_params,
builder_boost_factor,
spec,
)
.await
{
Ok(payload) => payload,
Err(e) => {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_OUTCOME,
&[metrics::FAILURE],
);
return Err(e);
}
},
BlockProductionVersion::BlindedV2 => {
let _timer = metrics::start_timer_vec(
&metrics::EXECUTION_LAYER_REQUEST_TIMES,
&[metrics::GET_BLINDED_PAYLOAD],
);
self.determine_and_fetch_payload(payload_parameters, builder_params, None, spec)
.await?
}
BlockProductionVersion::FullV2 => self
.get_full_payload_with(payload_parameters, noop)
.await
.and_then(GetPayloadResponseType::try_into)
.map(ProvenancedPayload::Local)?,
};
let block_proposal_content_type = match payload_result_type {
ProvenancedPayload::Local(local_payload) => local_payload,
ProvenancedPayload::Builder(builder_payload) => builder_payload,
};
match block_proposal_content_type {
BlockProposalContentsType::Full(block_proposal_contents) => {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_OUTCOME,
&[metrics::SUCCESS],
);
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_SOURCE,
&[metrics::LOCAL],
);
if matches!(block_production_version, BlockProductionVersion::BlindedV2) {
Ok(BlockProposalContentsType::Blinded(
block_proposal_contents.into(),
))
} else {
Ok(BlockProposalContentsType::Full(block_proposal_contents))
}
}
BlockProposalContentsType::Blinded(block_proposal_contents) => {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_OUTCOME,
&[metrics::SUCCESS],
);
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_SOURCE,
&[metrics::BUILDER],
);
Ok(BlockProposalContentsType::Blinded(block_proposal_contents))
}
}
}
/// Fetches local and builder paylaods concurrently, Logs and returns results.
async fn fetch_builder_and_local_payloads(
&self,
builder: &BuilderHttpClient,
builder_params: &BuilderParams,
payload_parameters: PayloadParameters<'_>,
) -> (
Result<Option<ForkVersionedResponse<SignedBuilderBid<E>>>, builder_client::Error>,
Result<GetPayloadResponse<E>, Error>,
) {
let slot = builder_params.slot;
let pubkey = &builder_params.pubkey;
let parent_hash = payload_parameters.parent_hash;
info!(
self.log(),
"Requesting blinded header from connected builder";
"slot" => ?slot,
"pubkey" => ?pubkey,
"parent_hash" => ?parent_hash,
);
// Wait for the builder *and* local EL to produce a payload (or return an error).
let ((relay_result, relay_duration), (local_result, local_duration)) = tokio::join!(
timed_future(metrics::GET_BLINDED_PAYLOAD_BUILDER, async {
builder
.get_builder_header::<E>(slot, parent_hash, pubkey)
.await
}),
timed_future(metrics::GET_BLINDED_PAYLOAD_LOCAL, async {
self.get_full_payload_caching(payload_parameters)
.await
.and_then(|local_result_type| match local_result_type {
GetPayloadResponseType::Full(payload) => Ok(payload),
GetPayloadResponseType::Blinded(_) => Err(Error::PayloadTypeMismatch),
})
})
);
info!(
self.log(),
"Requested blinded execution payload";
"relay_fee_recipient" => match &relay_result {
Ok(Some(r)) => format!("{:?}", r.data.message.header().fee_recipient()),
Ok(None) => "empty response".to_string(),
Err(_) => "request failed".to_string(),
},
"relay_response_ms" => relay_duration.as_millis(),
"local_fee_recipient" => match &local_result {
Ok(get_payload_response) => format!("{:?}", get_payload_response.fee_recipient()),
Err(_) => "request failed".to_string()
},
"local_response_ms" => local_duration.as_millis(),
"parent_hash" => ?parent_hash,
);
(relay_result, local_result)
}
async fn determine_and_fetch_payload(
&self,
payload_parameters: PayloadParameters<'_>,
builder_params: BuilderParams,
builder_boost_factor: Option<u64>,
spec: &ChainSpec,
) -> Result<ProvenancedPayload<BlockProposalContentsType<E>>, Error> {
let Some(builder) = self.builder() else {
// no builder.. return local payload
return self
.get_full_payload_caching(payload_parameters)
.await
.and_then(GetPayloadResponseType::try_into)
.map(ProvenancedPayload::Local);
};
// check chain health
if builder_params.chain_health != ChainHealth::Healthy {
// chain is unhealthy, gotta use local payload
match builder_params.chain_health {
ChainHealth::Unhealthy(condition) => info!(
self.log(),
"Chain is unhealthy, using local payload";
"info" => "this helps protect the network. the --builder-fallback flags \
can adjust the expected health conditions.",
"failed_condition" => ?condition
),
// Intentional no-op, so we never attempt builder API proposals pre-merge.
ChainHealth::PreMerge => (),
ChainHealth::Optimistic => info!(
self.log(),
"Chain is optimistic; can't build payload";
"info" => "the local execution engine is syncing and the builder network \
cannot safely be used - unable to propose block"
),
ChainHealth::Healthy => crit!(
self.log(),
"got healthy but also not healthy.. this shouldn't happen!"
),
}
return self
.get_full_payload_caching(payload_parameters)
.await
.and_then(GetPayloadResponseType::try_into)
.map(ProvenancedPayload::Local);
}
let parent_hash = payload_parameters.parent_hash;
let (relay_result, local_result) = self
.fetch_builder_and_local_payloads(builder.as_ref(), &builder_params, payload_parameters)
.await;
match (relay_result, local_result) {
(Err(e), Ok(local)) => {
warn!(
self.log(),
"Builder error when requesting payload";
"info" => "falling back to local execution client",
"relay_error" => ?e,
"local_block_hash" => ?local.block_hash(),
"parent_hash" => ?parent_hash,
);
Ok(ProvenancedPayload::Local(BlockProposalContentsType::Full(
local.try_into()?,
)))
}
(Ok(None), Ok(local)) => {
info!(
self.log(),
"Builder did not return a payload";
"info" => "falling back to local execution client",
"local_block_hash" => ?local.block_hash(),
"parent_hash" => ?parent_hash,
);
Ok(ProvenancedPayload::Local(BlockProposalContentsType::Full(
local.try_into()?,
)))
}
(Err(relay_error), Err(local_error)) => {
crit!(
self.log(),
"Unable to produce execution payload";
"info" => "the local EL and builder both failed - unable to propose block",
"relay_error" => ?relay_error,
"local_error" => ?local_error,
"parent_hash" => ?parent_hash,
);
Err(Error::CannotProduceHeader)
}
(Ok(None), Err(local_error)) => {
crit!(
self.log(),
"Unable to produce execution payload";
"info" => "the local EL failed and the builder returned nothing - \
the block proposal will be missed",
"local_error" => ?local_error,
"parent_hash" => ?parent_hash,
);
Err(Error::CannotProduceHeader)
}
(Ok(Some(relay)), Ok(local)) => {
let header = &relay.data.message.header();
info!(
self.log(),
"Received local and builder payloads";
"relay_block_hash" => ?header.block_hash(),
"local_block_hash" => ?local.block_hash(),
"parent_hash" => ?parent_hash,
);
// check relay payload validity
if let Err(reason) =
verify_builder_bid(&relay, payload_parameters, Some(local.block_number()), spec)
{
// relay payload invalid -> return local
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_BUILDER_REJECTIONS,
&[reason.as_ref().as_ref()],
);
warn!(
self.log(),
"Builder returned invalid payload";
"info" => "using local payload",
"reason" => %reason,
"relay_block_hash" => ?header.block_hash(),
"parent_hash" => ?parent_hash,
);
return Ok(ProvenancedPayload::Local(BlockProposalContentsType::Full(
local.try_into()?,
)));
}
let relay_value = *relay.data.message.value();
let boosted_relay_value = match builder_boost_factor {
Some(builder_boost_factor) => (relay_value / Uint256::from(100))
.saturating_mul(Uint256::from(builder_boost_factor)),
None => relay_value,
};
let local_value = *local.block_value();
if local_value >= boosted_relay_value {
info!(
self.log(),
"Local block is more profitable than relay block";
"local_block_value" => %local_value,
"relay_value" => %relay_value,
"boosted_relay_value" => %boosted_relay_value,
"builder_boost_factor" => ?builder_boost_factor,
);
return Ok(ProvenancedPayload::Local(BlockProposalContentsType::Full(
local.try_into()?,
)));
}
if local.should_override_builder().unwrap_or(false) {
info!(
self.log(),
"Using local payload because execution engine suggested we ignore builder payload";
"local_block_value" => %local_value,
"relay_value" => %relay_value
);
return Ok(ProvenancedPayload::Local(BlockProposalContentsType::Full(
local.try_into()?,
)));
}
info!(
self.log(),
"Relay block is more profitable than local block";
"local_block_value" => %local_value,
"relay_value" => %relay_value,
"boosted_relay_value" => %boosted_relay_value,
"builder_boost_factor" => ?builder_boost_factor
);
Ok(ProvenancedPayload::try_from(relay.data.message)?)
}
(Ok(Some(relay)), Err(local_error)) => {
let header = &relay.data.message.header();
info!(
self.log(),
"Received builder payload with local error";
"relay_block_hash" => ?header.block_hash(),
"local_error" => ?local_error,
"parent_hash" => ?parent_hash,
);
match verify_builder_bid(&relay, payload_parameters, None, spec) {
Ok(()) => Ok(ProvenancedPayload::try_from(relay.data.message)?),
Err(reason) => {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_GET_PAYLOAD_BUILDER_REJECTIONS,
&[reason.as_ref().as_ref()],
);
crit!(
self.log(),
"Builder returned invalid payload";
"info" => "no local payload either - unable to propose block",
"reason" => %reason,
"relay_block_hash" => ?header.block_hash(),
"parent_hash" => ?parent_hash,
);
Err(Error::CannotProduceHeader)
}
}
}
}
}
/// Get a full payload and cache its result in the execution layer's payload cache.
async fn get_full_payload_caching(
&self,
payload_parameters: PayloadParameters<'_>,
) -> Result<GetPayloadResponseType<E>, Error> {
self.get_full_payload_with(payload_parameters, Self::cache_payload)
.await
}
async fn get_full_payload_with(
&self,
payload_parameters: PayloadParameters<'_>,
cache_fn: fn(
&ExecutionLayer<E>,
PayloadContentsRefTuple<E>,
) -> Option<FullPayloadContents<E>>,
) -> Result<GetPayloadResponseType<E>, Error> {
let PayloadParameters {
parent_hash,
payload_attributes,
forkchoice_update_params,
current_fork,
..
} = payload_parameters;
self.engine()
.request(move |engine| async move {
let payload_id = if let Some(id) = engine
.get_payload_id(&parent_hash, payload_attributes)
.await
{
// The payload id has been cached for this engine.
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_PRE_PREPARED_PAYLOAD_ID,
&[metrics::HIT],
);
id
} else {
// The payload id has *not* been cached. Trigger an artificial
// fork choice update to retrieve a payload ID.
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_PRE_PREPARED_PAYLOAD_ID,
&[metrics::MISS],
);
let fork_choice_state = ForkchoiceState {
head_block_hash: parent_hash,
safe_block_hash: forkchoice_update_params
.justified_hash
.unwrap_or_else(ExecutionBlockHash::zero),
finalized_block_hash: forkchoice_update_params
.finalized_hash
.unwrap_or_else(ExecutionBlockHash::zero),
};
let response = engine
.notify_forkchoice_updated(
fork_choice_state,
Some(payload_attributes.clone()),
self.log(),
)
.await?;
match response.payload_id {
Some(payload_id) => payload_id,
None => {
error!(
self.log(),
"Exec engine unable to produce payload";
"msg" => "No payload ID, the engine is likely syncing. \
This has the potential to cause a missed block proposal.",
"status" => ?response.payload_status
);
return Err(ApiError::PayloadIdUnavailable);
}
}
};
let payload_response = async {
debug!(
self.log(),
"Issuing engine_getPayload";
"suggested_fee_recipient" => ?payload_attributes.suggested_fee_recipient(),
"prev_randao" => ?payload_attributes.prev_randao(),
"timestamp" => payload_attributes.timestamp(),
"parent_hash" => ?parent_hash,
);
let _timer = metrics::start_timer_vec(
&metrics::EXECUTION_LAYER_REQUEST_TIMES,
&[metrics::GET_PAYLOAD],
);
engine.api.get_payload::<E>(current_fork, payload_id).await
}.await?;
if payload_response.execution_payload_ref().fee_recipient() != payload_attributes.suggested_fee_recipient() {
error!(
self.log(),
"Inconsistent fee recipient";
"msg" => "The fee recipient returned from the Execution Engine differs \
from the suggested_fee_recipient set on the beacon node. This could \
indicate that fees are being diverted to another address. Please \
ensure that the value of suggested_fee_recipient is set correctly and \
that the Execution Engine is trusted.",
"fee_recipient" => ?payload_response.execution_payload_ref().fee_recipient(),
"suggested_fee_recipient" => ?payload_attributes.suggested_fee_recipient(),
);
}
if cache_fn(self, (payload_response.execution_payload_ref(), payload_response.blobs_bundle().ok())).is_some() {
warn!(
self.log(),
"Duplicate payload cached, this might indicate redundant proposal \
attempts."
);
}
Ok(GetPayloadResponseType::Full(payload_response))
})
.await
.map_err(Box::new)
.map_err(Error::EngineError)
}
/// Maps to the `engine_newPayload` JSON-RPC call.
pub async fn notify_new_payload(
&self,
new_payload_request: NewPayloadRequest<'_, E>,
) -> Result<PayloadStatus, Error> {
let _timer = metrics::start_timer_vec(
&metrics::EXECUTION_LAYER_REQUEST_TIMES,
&[metrics::NEW_PAYLOAD],
);
let timer = std::time::Instant::now();
let block_number = new_payload_request.block_number();
let block_hash = new_payload_request.block_hash();
let parent_hash = new_payload_request.parent_hash();
let result = self
.engine()
.request(|engine| engine.api.new_payload(new_payload_request))
.await;
if let Ok(status) = &result {
let status_str = <&'static str>::from(status.status);
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_PAYLOAD_STATUS,
&["new_payload", status_str],
);
debug!(
self.log(),
"Processed engine_newPayload";
"status" => status_str,
"parent_hash" => ?parent_hash,
"block_hash" => ?block_hash,
"block_number" => block_number,
"response_time_ms" => timer.elapsed().as_millis()
);
}
*self.inner.last_new_payload_errored.write().await = result.is_err();
process_payload_status(block_hash, result, self.log())
.map_err(Box::new)
.map_err(Error::EngineError)
}
/// Update engine sync status.
pub async fn upcheck(&self) {
self.engine().upcheck().await;
}
/// Register that the given `validator_index` is going to produce a block at `slot`.
///
/// The block will be built atop `head_block_root` and the EL will need to prepare an
/// `ExecutionPayload` as defined by the given `payload_attributes`.
pub async fn insert_proposer(
&self,
slot: Slot,
head_block_root: Hash256,
validator_index: u64,
payload_attributes: PayloadAttributes,
) -> bool {
let proposers_key = ProposerKey {
slot,
head_block_root,
};
let existing = self.proposers().write().await.insert(
proposers_key,
Proposer {
validator_index,
payload_attributes,
},
);
if existing.is_none() {
metrics::inc_counter(&metrics::EXECUTION_LAYER_PROPOSER_INSERTED);
}
existing.is_some()
}
/// If there has been a proposer registered via `Self::insert_proposer` with a matching `slot`
/// `head_block_root`, then return the appropriate `PayloadAttributes` for inclusion in
/// `forkchoiceUpdated` calls.
pub async fn payload_attributes(
&self,
current_slot: Slot,
head_block_root: Hash256,
) -> Option<PayloadAttributes> {
let proposers_key = ProposerKey {
slot: current_slot,
head_block_root,
};
let proposer = self.proposers().read().await.get(&proposers_key).cloned()?;
debug!(
self.log(),
"Beacon proposer found";
"payload_attributes" => ?proposer.payload_attributes,
"head_block_root" => ?head_block_root,
"slot" => current_slot,
"validator_index" => proposer.validator_index,
);
Some(proposer.payload_attributes)
}
/// Maps to the `engine_consensusValidated` JSON-RPC call.
pub async fn notify_forkchoice_updated(
&self,
head_block_hash: ExecutionBlockHash,
justified_block_hash: ExecutionBlockHash,
finalized_block_hash: ExecutionBlockHash,
current_slot: Slot,
head_block_root: Hash256,
) -> Result<PayloadStatus, Error> {
let _timer = metrics::start_timer_vec(
&metrics::EXECUTION_LAYER_REQUEST_TIMES,
&[metrics::FORKCHOICE_UPDATED],
);
debug!(
self.log(),
"Issuing engine_forkchoiceUpdated";
"finalized_block_hash" => ?finalized_block_hash,
"justified_block_hash" => ?justified_block_hash,
"head_block_hash" => ?head_block_hash,
"head_block_root" => ?head_block_root,
"current_slot" => current_slot,
);
let next_slot = current_slot + 1;
let payload_attributes = self.payload_attributes(next_slot, head_block_root).await;
// Compute the "lookahead", the time between when the payload will be produced and now.
if let Some(ref payload_attributes) = payload_attributes {
if let Ok(now) = SystemTime::now().duration_since(UNIX_EPOCH) {
let timestamp = Duration::from_secs(payload_attributes.timestamp());
if let Some(lookahead) = timestamp.checked_sub(now) {
metrics::observe_duration(
&metrics::EXECUTION_LAYER_PAYLOAD_ATTRIBUTES_LOOKAHEAD,
lookahead,
);
} else {
debug!(
self.log(),
"Late payload attributes";
"timestamp" => ?timestamp,
"now" => ?now,
)
}
}
}
let forkchoice_state = ForkchoiceState {
head_block_hash,
safe_block_hash: justified_block_hash,
finalized_block_hash,
};
self.engine()
.set_latest_forkchoice_state(forkchoice_state)
.await;
let result = self
.engine()
.request(|engine| async move {
engine
.notify_forkchoice_updated(forkchoice_state, payload_attributes, self.log())
.await
})
.await;
if let Ok(status) = &result {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_PAYLOAD_STATUS,
&["forkchoice_updated", status.payload_status.status.into()],
);
}
process_payload_status(
head_block_hash,
result.map(|response| response.payload_status),
self.log(),
)
.map_err(Box::new)
.map_err(Error::EngineError)
}
/// Returns the execution engine capabilities resulting from a call to
/// engine_exchangeCapabilities. If the capabilities cache is not populated,
/// or if it is populated with a cached result of age >= `age_limit`, this
/// method will fetch the result from the execution engine and populate the
/// cache before returning it. Otherwise it will return a cached result from
/// a previous call.
///
/// Set `age_limit` to `None` to always return the cached result
/// Set `age_limit` to `Some(Duration::ZERO)` to force fetching from EE
pub async fn get_engine_capabilities(
&self,
age_limit: Option<Duration>,
) -> Result<EngineCapabilities, Error> {
self.engine()
.request(|engine| engine.get_engine_capabilities(age_limit))
.await
.map_err(Into::into)
}
/// Returns the execution engine version resulting from a call to
/// engine_clientVersionV1. If the version cache is not populated, or if it
/// is populated with a cached result of age >= `age_limit`, this method will
/// fetch the result from the execution engine and populate the cache before
/// returning it. Otherwise it will return the cached result from an earlier
/// call.
///
/// Set `age_limit` to `None` to always return the cached result
/// Set `age_limit` to `Some(Duration::ZERO)` to force fetching from EE
pub async fn get_engine_version(
&self,
age_limit: Option<Duration>,
) -> Result<Vec<ClientVersionV1>, Error> {
self.engine()
.request(|engine| engine.get_engine_version(age_limit))
.await
.map_err(Into::into)
}
/// Used during block production to determine if the merge has been triggered.
///
/// ## Specification
///
/// `get_terminal_pow_block_hash`
///
/// https://github.com/ethereum/consensus-specs/blob/v1.1.5/specs/merge/validator.md
pub async fn get_terminal_pow_block_hash(
&self,
spec: &ChainSpec,
timestamp: u64,
) -> Result<Option<ExecutionBlockHash>, Error> {
let _timer = metrics::start_timer_vec(
&metrics::EXECUTION_LAYER_REQUEST_TIMES,
&[metrics::GET_TERMINAL_POW_BLOCK_HASH],
);
let hash_opt = self
.engine()
.request(|engine| async move {
let terminal_block_hash = spec.terminal_block_hash;
if terminal_block_hash != ExecutionBlockHash::zero() {
if self
.get_pow_block(engine, terminal_block_hash)
.await?
.is_some()
{
return Ok(Some(terminal_block_hash));
} else {
return Ok(None);
}
}
let block = self.get_pow_block_at_total_difficulty(engine, spec).await?;
if let Some(pow_block) = block {
// If `terminal_block.timestamp == transition_block.timestamp`,
// we violate the invariant that a block's timestamp must be
// strictly greater than its parent's timestamp.
// The execution layer will reject a fcu call with such payload
// attributes leading to a missed block.
// Hence, we return `None` in such a case.
if pow_block.timestamp >= timestamp {
return Ok(None);
}
}
Ok(block.map(|b| b.block_hash))
})
.await
.map_err(Box::new)
.map_err(Error::EngineError)?;
if let Some(hash) = &hash_opt {
info!(
self.log(),
"Found terminal block hash";
"terminal_block_hash_override" => ?spec.terminal_block_hash,
"terminal_total_difficulty" => ?spec.terminal_total_difficulty,
"block_hash" => ?hash,
);
}
Ok(hash_opt)
}
/// This function should remain internal. External users should use
/// `self.get_terminal_pow_block` instead, since it checks against the terminal block hash
/// override.
///
/// ## Specification
///
/// `get_pow_block_at_terminal_total_difficulty`
///
/// https://github.com/ethereum/consensus-specs/blob/v1.1.5/specs/merge/validator.md
async fn get_pow_block_at_total_difficulty(
&self,
engine: &Engine,
spec: &ChainSpec,
) -> Result<Option<ExecutionBlock>, ApiError> {
let mut block = engine
.api
.get_block_by_number(BlockByNumberQuery::Tag(LATEST_TAG))
.await?
.ok_or(ApiError::ExecutionHeadBlockNotFound)?;
self.execution_blocks().await.put(block.block_hash, block);
loop {
let block_reached_ttd = block.total_difficulty >= spec.terminal_total_difficulty;
if block_reached_ttd {
if block.parent_hash == ExecutionBlockHash::zero() {
return Ok(Some(block));
}
let parent = self
.get_pow_block(engine, block.parent_hash)
.await?
.ok_or(ApiError::ExecutionBlockNotFound(block.parent_hash))?;
let parent_reached_ttd = parent.total_difficulty >= spec.terminal_total_difficulty;
if block_reached_ttd && !parent_reached_ttd {
return Ok(Some(block));
} else {
block = parent;
}
} else {
return Ok(None);
}
}
}
/// Used during block verification to check that a block correctly triggers the merge.
///
/// ## Returns
///
/// - `Some(true)` if the given `block_hash` is the terminal proof-of-work block.
/// - `Some(false)` if the given `block_hash` is certainly *not* the terminal proof-of-work
/// block.
/// - `None` if the `block_hash` or its parent were not present on the execution engine.
/// - `Err(_)` if there was an error connecting to the execution engine.
///
/// ## Fallback Behaviour
///
/// The request will be broadcast to all nodes, simultaneously. It will await a response (or
/// failure) from all nodes and then return based on the first of these conditions which
/// returns true:
///
/// - Terminal, if any node indicates it is terminal.
/// - Not terminal, if any node indicates it is non-terminal.
/// - Block not found, if any node cannot find the block.
/// - An error, if all nodes return an error.
///
/// ## Specification
///
/// `is_valid_terminal_pow_block`
///
/// https://github.com/ethereum/consensus-specs/blob/v1.1.0/specs/merge/fork-choice.md
pub async fn is_valid_terminal_pow_block_hash(
&self,
block_hash: ExecutionBlockHash,
spec: &ChainSpec,
) -> Result<Option<bool>, Error> {
let _timer = metrics::start_timer_vec(
&metrics::EXECUTION_LAYER_REQUEST_TIMES,
&[metrics::IS_VALID_TERMINAL_POW_BLOCK_HASH],
);
self.engine()
.request(|engine| async move {
if let Some(pow_block) = self.get_pow_block(engine, block_hash).await? {
if let Some(pow_parent) =
self.get_pow_block(engine, pow_block.parent_hash).await?
{
return Ok(Some(
self.is_valid_terminal_pow_block(pow_block, pow_parent, spec),
));
}
}
Ok(None)
})
.await
.map_err(Box::new)
.map_err(Error::EngineError)
}
/// This function should remain internal.
///
/// External users should use `self.is_valid_terminal_pow_block_hash`.
fn is_valid_terminal_pow_block(
&self,
block: ExecutionBlock,
parent: ExecutionBlock,
spec: &ChainSpec,
) -> bool {
let is_total_difficulty_reached = block.total_difficulty >= spec.terminal_total_difficulty;
let is_parent_total_difficulty_valid =
parent.total_difficulty < spec.terminal_total_difficulty;
is_total_difficulty_reached && is_parent_total_difficulty_valid
}
/// Maps to the `eth_getBlockByHash` JSON-RPC call.
async fn get_pow_block(
&self,
engine: &Engine,
hash: ExecutionBlockHash,
) -> Result<Option<ExecutionBlock>, ApiError> {
if let Some(cached) = self.execution_blocks().await.get(&hash).copied() {
// The block was in the cache, no need to request it from the execution
// engine.
return Ok(Some(cached));
}
// The block was *not* in the cache, request it from the execution
// engine and cache it for future reference.
if let Some(block) = engine.api.get_block_by_hash(hash).await? {
self.execution_blocks().await.put(hash, block);
Ok(Some(block))
} else {
Ok(None)
}
}
pub async fn get_payload_bodies_by_hash(
&self,
hashes: Vec<ExecutionBlockHash>,
) -> Result<Vec<Option<ExecutionPayloadBodyV1<E>>>, Error> {
self.engine()
.request(|engine: &Engine| async move {
engine.api.get_payload_bodies_by_hash_v1(hashes).await
})
.await
.map_err(Box::new)
.map_err(Error::EngineError)
}
pub async fn get_payload_bodies_by_range(
&self,
start: u64,
count: u64,
) -> Result<Vec<Option<ExecutionPayloadBodyV1<E>>>, Error> {
let _timer = metrics::start_timer(&metrics::EXECUTION_LAYER_GET_PAYLOAD_BODIES_BY_RANGE);
self.engine()
.request(|engine: &Engine| async move {
engine
.api
.get_payload_bodies_by_range_v1(start, count)
.await
})
.await
.map_err(Box::new)
.map_err(Error::EngineError)
}
/// Fetch a full payload from the execution node.
///
/// This will fail if the payload is not from the finalized portion of the chain.
pub async fn get_payload_for_header(
&self,
header: &ExecutionPayloadHeader<E>,
fork: ForkName,
) -> Result<Option<ExecutionPayload<E>>, Error> {
let block_number = header.block_number();
// Handle default payload body.
if header.block_hash() == ExecutionBlockHash::zero() {
let payload = match fork {
ForkName::Bellatrix => ExecutionPayloadBellatrix::default().into(),
ForkName::Capella => ExecutionPayloadCapella::default().into(),
ForkName::Deneb => ExecutionPayloadDeneb::default().into(),
ForkName::Electra => ExecutionPayloadElectra::default().into(),
ForkName::Fulu => ExecutionPayloadFulu::default().into(),
ForkName::Base | ForkName::Altair => {
return Err(Error::InvalidForkForPayload);
}
};
return Ok(Some(payload));
}
// Use efficient payload bodies by range method if supported.
let capabilities = self.get_engine_capabilities(None).await?;
if capabilities.get_payload_bodies_by_range_v1 {
let mut payload_bodies = self.get_payload_bodies_by_range(block_number, 1).await?;
if payload_bodies.len() != 1 {
return Ok(None);
}
let opt_payload_body = payload_bodies.pop().flatten();
opt_payload_body
.map(|body| {
body.to_payload(header.clone())
.map_err(Error::InvalidPayloadBody)
})
.transpose()
} else {
Err(Error::PayloadBodiesByRangeNotSupported)
}
}
pub async fn get_blobs(
&self,
query: Vec<Hash256>,
) -> Result<Vec<Option<BlobAndProofV1<E>>>, Error> {
let capabilities = self.get_engine_capabilities(None).await?;
if capabilities.get_blobs_v1 {
self.engine()
.request(|engine| async move { engine.api.get_blobs(query).await })
.await
.map_err(Box::new)
.map_err(Error::EngineError)
} else {
Err(Error::GetBlobsNotSupported)
}
}
pub async fn get_block_by_number(
&self,
query: BlockByNumberQuery<'_>,
) -> Result<Option<ExecutionBlock>, Error> {
self.engine()
.request(|engine| async move { engine.api.get_block_by_number(query).await })
.await
.map_err(Box::new)
.map_err(Error::EngineError)
}
pub async fn propose_blinded_beacon_block(
&self,
block_root: Hash256,
block: &SignedBlindedBeaconBlock<E>,
) -> Result<FullPayloadContents<E>, Error> {
debug!(
self.log(),
"Sending block to builder";
"root" => ?block_root,
);
if let Some(builder) = self.builder() {
let (payload_result, duration) =
timed_future(metrics::POST_BLINDED_PAYLOAD_BUILDER, async {
if builder.is_ssz_enabled() {
builder
.post_builder_blinded_blocks_ssz(block)
.await
.map_err(Error::Builder)
} else {
builder
.post_builder_blinded_blocks(block)
.await
.map_err(Error::Builder)
.map(|d| d.data)
}
})
.await;
match &payload_result {
Ok(unblinded_response) => {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_BUILDER_REVEAL_PAYLOAD_OUTCOME,
&[metrics::SUCCESS],
);
let payload = unblinded_response.payload_ref();
info!(
self.log(),
"Builder successfully revealed payload";
"relay_response_ms" => duration.as_millis(),
"block_root" => ?block_root,
"fee_recipient" => ?payload.fee_recipient(),
"block_hash" => ?payload.block_hash(),
"parent_hash" => ?payload.parent_hash()
)
}
Err(e) => {
metrics::inc_counter_vec(
&metrics::EXECUTION_LAYER_BUILDER_REVEAL_PAYLOAD_OUTCOME,
&[metrics::FAILURE],
);
warn!(
self.log(),
"Builder failed to reveal payload";
"info" => "this is common behaviour for some builders and may not indicate an issue",
"error" => ?e,
"relay_response_ms" => duration.as_millis(),
"block_root" => ?block_root,
"parent_hash" => ?block
.message()
.execution_payload()
.map(|payload| format!("{}", payload.parent_hash()))
.unwrap_or_else(|_| "unknown".to_string())
)
}
}
payload_result
} else {
Err(Error::NoPayloadBuilder)
}
}
}
#[derive(AsRefStr)]
#[strum(serialize_all = "snake_case")]
enum InvalidBuilderPayload {
ParentHash {
payload: ExecutionBlockHash,
expected: ExecutionBlockHash,
},
PrevRandao {
payload: Hash256,
expected: Hash256,
},
Timestamp {
payload: u64,
expected: u64,
},
BlockNumber {
payload: u64,
expected: Option<u64>,
},
Fork {
payload: Option<ForkName>,
expected: ForkName,
},
Signature {
signature: Signature,
pubkey: PublicKeyBytes,
},
WithdrawalsRoot {
payload: Option<Hash256>,
expected: Option<Hash256>,
},
GasLimitMismatch {
payload: u64,
expected: u64,
},
}
impl fmt::Display for InvalidBuilderPayload {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
InvalidBuilderPayload::ParentHash { payload, expected } => {
write!(f, "payload block hash was {} not {}", payload, expected)
}
InvalidBuilderPayload::PrevRandao { payload, expected } => {
write!(f, "payload prev randao was {} not {}", payload, expected)
}
InvalidBuilderPayload::Timestamp { payload, expected } => {
write!(f, "payload timestamp was {} not {}", payload, expected)
}
InvalidBuilderPayload::BlockNumber { payload, expected } => {
write!(f, "payload block number was {} not {:?}", payload, expected)
}
InvalidBuilderPayload::Fork { payload, expected } => {
write!(f, "payload fork was {:?} not {}", payload, expected)
}
InvalidBuilderPayload::Signature { signature, pubkey } => write!(
f,
"invalid payload signature {} for pubkey {}",
signature, pubkey
),
InvalidBuilderPayload::WithdrawalsRoot { payload, expected } => {
let opt_string = |opt_hash: &Option<Hash256>| {
opt_hash
.map(|hash| hash.to_string())
.unwrap_or_else(|| "None".to_string())
};
write!(
f,
"payload withdrawals root was {} not {}",
opt_string(payload),
opt_string(expected)
)
}
InvalidBuilderPayload::GasLimitMismatch { payload, expected } => {
write!(f, "payload gas limit was {} not {}", payload, expected)
}
}
}
}
/// Calculate the expected gas limit for a block.
pub fn expected_gas_limit(
parent_gas_limit: u64,
target_gas_limit: u64,
spec: &ChainSpec,
) -> Option<u64> {
// Calculate the maximum gas limit difference allowed safely
let max_gas_limit_difference = parent_gas_limit
.checked_div(spec.gas_limit_adjustment_factor)
.and_then(|result| result.checked_sub(1))
.unwrap_or(0);
// Adjust the gas limit safely
if target_gas_limit > parent_gas_limit {
let gas_diff = target_gas_limit.saturating_sub(parent_gas_limit);
parent_gas_limit.checked_add(std::cmp::min(gas_diff, max_gas_limit_difference))
} else {
let gas_diff = parent_gas_limit.saturating_sub(target_gas_limit);
parent_gas_limit.checked_sub(std::cmp::min(gas_diff, max_gas_limit_difference))
}
}
/// Perform some cursory, non-exhaustive validation of the bid returned from the builder.
fn verify_builder_bid<E: EthSpec>(
bid: &ForkVersionedResponse<SignedBuilderBid<E>>,
payload_parameters: PayloadParameters<'_>,
block_number: Option<u64>,
spec: &ChainSpec,
) -> Result<(), Box<InvalidBuilderPayload>> {
let PayloadParameters {
parent_hash,
payload_attributes,
current_fork,
parent_gas_limit,
proposer_gas_limit,
..
} = payload_parameters;
let is_signature_valid = bid.data.verify_signature(spec);
let header = &bid.data.message.header();
metrics::set_gauge_vec(
&metrics::EXECUTION_LAYER_PAYLOAD_BIDS,
&[metrics::BUILDER],
bid.data.message.value().to_i64(),
);
let expected_withdrawals_root = payload_attributes
.withdrawals()
.ok()
.cloned()
.map(|withdrawals| Withdrawals::<E>::from(withdrawals).tree_hash_root());
let payload_withdrawals_root = header.withdrawals_root().ok();
let expected_gas_limit = proposer_gas_limit
.and_then(|target_gas_limit| expected_gas_limit(parent_gas_limit, target_gas_limit, spec));
if header.parent_hash() != parent_hash {
Err(Box::new(InvalidBuilderPayload::ParentHash {
payload: header.parent_hash(),
expected: parent_hash,
}))
} else if header.prev_randao() != payload_attributes.prev_randao() {
Err(Box::new(InvalidBuilderPayload::PrevRandao {
payload: header.prev_randao(),
expected: payload_attributes.prev_randao(),
}))
} else if header.timestamp() != payload_attributes.timestamp() {
Err(Box::new(InvalidBuilderPayload::Timestamp {
payload: header.timestamp(),
expected: payload_attributes.timestamp(),
}))
} else if block_number.is_some_and(|n| n != header.block_number()) {
Err(Box::new(InvalidBuilderPayload::BlockNumber {
payload: header.block_number(),
expected: block_number,
}))
} else if bid.version != Some(current_fork) {
Err(Box::new(InvalidBuilderPayload::Fork {
payload: bid.version,
expected: current_fork,
}))
} else if !is_signature_valid {
Err(Box::new(InvalidBuilderPayload::Signature {
signature: bid.data.signature.clone(),
pubkey: *bid.data.message.pubkey(),
}))
} else if payload_withdrawals_root != expected_withdrawals_root {
Err(Box::new(InvalidBuilderPayload::WithdrawalsRoot {
payload: payload_withdrawals_root,
expected: expected_withdrawals_root,
}))
} else if expected_gas_limit
.map(|gas_limit| header.gas_limit() != gas_limit)
.unwrap_or(false)
{
Err(Box::new(InvalidBuilderPayload::GasLimitMismatch {
payload: header.gas_limit(),
expected: expected_gas_limit.unwrap_or(0),
}))
} else {
Ok(())
}
}
/// A helper function to record the time it takes to execute a future.
async fn timed_future<F: Future<Output = T>, T>(metric: &str, future: F) -> (T, Duration) {
let start = Instant::now();
let result = future.await;
let duration = start.elapsed();
metrics::observe_timer_vec(&metrics::EXECUTION_LAYER_REQUEST_TIMES, &[metric], duration);
(result, duration)
}
#[cfg(test)]
/// Returns the duration since the unix epoch.
fn timestamp_now() -> u64 {
SystemTime::now()
.duration_since(UNIX_EPOCH)
.unwrap_or_else(|_| Duration::from_secs(0))
.as_secs()
}
fn noop<E: EthSpec>(
_: &ExecutionLayer<E>,
_: PayloadContentsRefTuple<E>,
) -> Option<FullPayloadContents<E>> {
None
}
#[cfg(test)]
mod test {
use super::*;
use crate::test_utils::MockExecutionLayer as GenericMockExecutionLayer;
use task_executor::test_utils::TestRuntime;
use types::MainnetEthSpec;
type MockExecutionLayer = GenericMockExecutionLayer<MainnetEthSpec>;
#[tokio::test]
async fn produce_three_valid_pos_execution_blocks() {
let runtime = TestRuntime::default();
MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_terminal_block()
.produce_valid_execution_payload_on_head()
.await
.produce_valid_execution_payload_on_head()
.await
.produce_valid_execution_payload_on_head()
.await;
}
#[tokio::test]
async fn test_expected_gas_limit() {
let spec = ChainSpec::mainnet();
assert_eq!(
expected_gas_limit(30_000_000, 30_000_000, &spec),
Some(30_000_000)
);
assert_eq!(
expected_gas_limit(30_000_000, 40_000_000, &spec),
Some(30_029_295)
);
assert_eq!(
expected_gas_limit(30_029_295, 40_000_000, &spec),
Some(30_058_619)
);
assert_eq!(
expected_gas_limit(30_058_619, 30_000_000, &spec),
Some(30_029_266)
);
}
#[tokio::test]
async fn test_forked_terminal_block() {
let runtime = TestRuntime::default();
let (mock, block_hash) = MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_terminal_block()
.produce_forked_pow_block();
assert!(mock
.el
.is_valid_terminal_pow_block_hash(block_hash, &mock.spec)
.await
.unwrap()
.unwrap());
}
#[tokio::test]
async fn finds_valid_terminal_block_hash() {
let runtime = TestRuntime::default();
MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_block_prior_to_terminal_block()
.with_terminal_block(|spec, el, _| async move {
el.engine().upcheck().await;
assert_eq!(
el.get_terminal_pow_block_hash(&spec, timestamp_now())
.await
.unwrap(),
None
)
})
.await
.move_to_terminal_block()
.with_terminal_block(|spec, el, terminal_block| async move {
assert_eq!(
el.get_terminal_pow_block_hash(&spec, timestamp_now())
.await
.unwrap(),
Some(terminal_block.unwrap().block_hash)
)
})
.await;
}
#[tokio::test]
async fn rejects_terminal_block_with_equal_timestamp() {
let runtime = TestRuntime::default();
MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_block_prior_to_terminal_block()
.with_terminal_block(|spec, el, _| async move {
el.engine().upcheck().await;
assert_eq!(
el.get_terminal_pow_block_hash(&spec, timestamp_now())
.await
.unwrap(),
None
)
})
.await
.move_to_terminal_block()
.with_terminal_block(|spec, el, terminal_block| async move {
let timestamp = terminal_block.as_ref().map(|b| b.timestamp).unwrap();
assert_eq!(
el.get_terminal_pow_block_hash(&spec, timestamp)
.await
.unwrap(),
None
)
})
.await;
}
#[tokio::test]
async fn verifies_valid_terminal_block_hash() {
let runtime = TestRuntime::default();
MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_terminal_block()
.with_terminal_block(|spec, el, terminal_block| async move {
el.engine().upcheck().await;
assert_eq!(
el.is_valid_terminal_pow_block_hash(terminal_block.unwrap().block_hash, &spec)
.await
.unwrap(),
Some(true)
)
})
.await;
}
#[tokio::test]
async fn rejects_invalid_terminal_block_hash() {
let runtime = TestRuntime::default();
MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_terminal_block()
.with_terminal_block(|spec, el, terminal_block| async move {
el.engine().upcheck().await;
let invalid_terminal_block = terminal_block.unwrap().parent_hash;
assert_eq!(
el.is_valid_terminal_pow_block_hash(invalid_terminal_block, &spec)
.await
.unwrap(),
Some(false)
)
})
.await;
}
#[tokio::test]
async fn rejects_unknown_terminal_block_hash() {
let runtime = TestRuntime::default();
MockExecutionLayer::default_params(runtime.task_executor.clone())
.move_to_terminal_block()
.with_terminal_block(|spec, el, _| async move {
el.engine().upcheck().await;
let missing_terminal_block = ExecutionBlockHash::repeat_byte(42);
assert_eq!(
el.is_valid_terminal_pow_block_hash(missing_terminal_block, &spec)
.await
.unwrap(),
None
)
})
.await;
}
}