lighthouse/beacon_node/beacon_chain/tests/tests.rs
Jimmy Chen 8a6f171b2a
Merge branch 'unstable' into merge-unstable-to-deneb-20230822
# Conflicts:
#	beacon_node/beacon_chain/src/builder.rs
#	beacon_node/beacon_chain/tests/store_tests.rs
#	beacon_node/client/src/builder.rs
#	beacon_node/src/config.rs
#	beacon_node/store/src/hot_cold_store.rs
#	lighthouse/tests/beacon_node.rs
2023-08-22 21:20:47 +10:00

876 lines
25 KiB
Rust

#![cfg(not(debug_assertions))]
use beacon_chain::{
attestation_verification::Error as AttnError,
test_utils::{
AttestationStrategy, BeaconChainHarness, BlockStrategy, EphemeralHarnessType,
OP_POOL_DB_KEY,
},
BeaconChain, ChainConfig, NotifyExecutionLayer, StateSkipConfig, WhenSlotSkipped,
};
use lazy_static::lazy_static;
use operation_pool::PersistedOperationPool;
use state_processing::{
per_slot_processing, per_slot_processing::Error as SlotProcessingError, EpochProcessingError,
};
use types::{
BeaconState, BeaconStateError, EthSpec, Hash256, Keypair, MinimalEthSpec, RelativeEpoch, Slot,
};
// Should ideally be divisible by 3.
pub const VALIDATOR_COUNT: usize = 48;
lazy_static! {
/// A cached set of keys.
static ref KEYPAIRS: Vec<Keypair> = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
}
fn get_harness(validator_count: usize) -> BeaconChainHarness<EphemeralHarnessType<MinimalEthSpec>> {
let harness = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.chain_config(ChainConfig {
reconstruct_historic_states: true,
..ChainConfig::default()
})
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
harness.advance_slot();
harness
}
#[test]
fn massive_skips() {
let harness = get_harness(8);
let spec = &harness.chain.spec;
let mut state = harness.chain.head_beacon_state_cloned();
// Run per_slot_processing until it returns an error.
let error = loop {
match per_slot_processing(&mut state, None, spec) {
Ok(_) => continue,
Err(e) => break e,
}
};
assert!(state.slot() > 1, "the state should skip at least one slot");
assert_eq!(
error,
SlotProcessingError::EpochProcessingError(EpochProcessingError::BeaconStateError(
BeaconStateError::InsufficientValidators
)),
"should return error indicating that validators have been slashed out"
)
}
#[tokio::test]
async fn iterators() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 2 - 1;
let harness = get_harness(VALIDATOR_COUNT);
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
// No need to produce attestations for this test.
AttestationStrategy::SomeValidators(vec![]),
)
.await;
let block_roots: Vec<(Hash256, Slot)> = harness
.chain
.forwards_iter_block_roots(Slot::new(0))
.expect("should get iter")
.map(Result::unwrap)
.collect();
let state_roots: Vec<(Hash256, Slot)> = harness
.chain
.forwards_iter_state_roots(Slot::new(0))
.expect("should get iter")
.map(Result::unwrap)
.collect();
assert_eq!(
block_roots.len(),
state_roots.len(),
"should be an equal amount of block and state roots"
);
assert!(
block_roots.iter().any(|(_root, slot)| *slot == 0),
"should contain genesis block root"
);
assert!(
state_roots.iter().any(|(_root, slot)| *slot == 0),
"should contain genesis state root"
);
assert_eq!(
block_roots.len(),
num_blocks_produced as usize + 1,
"should contain all produced blocks, plus the genesis block"
);
block_roots.windows(2).for_each(|x| {
assert_eq!(
x[1].1,
x[0].1 + 1,
"block root slots should be increasing by one"
)
});
state_roots.windows(2).for_each(|x| {
assert_eq!(
x[1].1,
x[0].1 + 1,
"state root slots should be increasing by one"
)
});
let head = harness.chain.head_snapshot();
assert_eq!(
*block_roots.last().expect("should have some block roots"),
(head.beacon_block_root, head.beacon_block.slot()),
"last block root and slot should be for the head block"
);
assert_eq!(
*state_roots.last().expect("should have some state roots"),
(head.beacon_state_root(), head.beacon_state.slot()),
"last state root and slot should be for the head state"
);
}
fn find_reorg_slot(
chain: &BeaconChain<EphemeralHarnessType<MinimalEthSpec>>,
new_state: &BeaconState<MinimalEthSpec>,
new_block_root: Hash256,
) -> Slot {
let (old_state, old_block_root) = {
let head = chain.canonical_head.cached_head();
let old_state = head.snapshot.beacon_state.clone();
let old_block_root = head.head_block_root();
(old_state, old_block_root)
};
beacon_chain::canonical_head::find_reorg_slot(
&old_state,
old_block_root,
new_state,
new_block_root,
&chain.spec,
)
.unwrap()
}
#[tokio::test]
async fn find_reorgs() {
let num_blocks_produced = MinimalEthSpec::slots_per_historical_root() + 1;
let harness = get_harness(VALIDATOR_COUNT);
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
// No need to produce attestations for this test.
AttestationStrategy::SomeValidators(vec![]),
)
.await;
let head = harness.chain.head_snapshot();
let head_state = &head.beacon_state;
let head_slot = head_state.slot();
let genesis_state = harness
.chain
.state_at_slot(Slot::new(0), StateSkipConfig::WithStateRoots)
.unwrap();
// because genesis is more than `SLOTS_PER_HISTORICAL_ROOT` away, this should return with the
// finalized slot.
assert_eq!(
find_reorg_slot(
&harness.chain,
&genesis_state,
harness.chain.genesis_block_root
),
head_state
.finalized_checkpoint()
.epoch
.start_slot(MinimalEthSpec::slots_per_epoch())
);
// test head
assert_eq!(
find_reorg_slot(
&harness.chain,
&head_state,
harness.chain.head_beacon_block().canonical_root()
),
head_slot
);
// Re-org back to the slot prior to the head.
let prev_slot = head_slot - Slot::new(1);
let prev_state = harness
.chain
.state_at_slot(prev_slot, StateSkipConfig::WithStateRoots)
.unwrap();
let prev_block_root = harness
.chain
.block_root_at_slot(prev_slot, WhenSlotSkipped::None)
.unwrap()
.unwrap();
assert_eq!(
find_reorg_slot(&harness.chain, &prev_state, prev_block_root),
prev_slot
);
}
#[tokio::test]
async fn chooses_fork() {
let harness = get_harness(VALIDATOR_COUNT);
let two_thirds = (VALIDATOR_COUNT / 3) * 2;
let delay = MinimalEthSpec::default_spec().min_attestation_inclusion_delay as usize;
let honest_validators: Vec<usize> = (0..two_thirds).collect();
let faulty_validators: Vec<usize> = (two_thirds..VALIDATOR_COUNT).collect();
let initial_blocks = delay + 1;
let honest_fork_blocks = delay + 1;
let faulty_fork_blocks = delay + 2;
// Build an initial chain where all validators agree.
harness
.extend_chain(
initial_blocks,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
let (honest_head, faulty_head) = harness
.generate_two_forks_by_skipping_a_block(
&honest_validators,
&faulty_validators,
honest_fork_blocks,
faulty_fork_blocks,
)
.await;
assert_ne!(honest_head, faulty_head, "forks should be distinct");
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
assert_eq!(
state.slot(),
Slot::from(initial_blocks + honest_fork_blocks),
"head should be at the current slot"
);
assert_eq!(
harness.chain.head_snapshot().beacon_block_root,
honest_head,
"the honest chain should be the canonical chain"
);
}
#[tokio::test]
async fn finalizes_with_full_participation() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 5;
let harness = get_harness(VALIDATOR_COUNT);
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
assert_eq!(
state.slot(),
num_blocks_produced,
"head should be at the current slot"
);
assert_eq!(
state.current_epoch(),
num_blocks_produced / MinimalEthSpec::slots_per_epoch(),
"head should be at the expected epoch"
);
assert_eq!(
state.current_justified_checkpoint().epoch,
state.current_epoch() - 1,
"the head should be justified one behind the current epoch"
);
assert_eq!(
state.finalized_checkpoint().epoch,
state.current_epoch() - 2,
"the head should be finalized two behind the current epoch"
);
}
#[tokio::test]
async fn finalizes_with_two_thirds_participation() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 5;
let harness = get_harness(VALIDATOR_COUNT);
let two_thirds = (VALIDATOR_COUNT / 3) * 2;
let attesters = (0..two_thirds).collect();
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::SomeValidators(attesters),
)
.await;
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
assert_eq!(
state.slot(),
num_blocks_produced,
"head should be at the current slot"
);
assert_eq!(
state.current_epoch(),
num_blocks_produced / MinimalEthSpec::slots_per_epoch(),
"head should be at the expected epoch"
);
// Note: the 2/3rds tests are not justifying the immediately prior epochs because the
// `MIN_ATTESTATION_INCLUSION_DELAY` is preventing an adequate number of attestations being
// included in blocks during that epoch.
assert_eq!(
state.current_justified_checkpoint().epoch,
state.current_epoch() - 2,
"the head should be justified two behind the current epoch"
);
assert_eq!(
state.finalized_checkpoint().epoch,
state.current_epoch() - 4,
"the head should be finalized three behind the current epoch"
);
}
#[tokio::test]
async fn does_not_finalize_with_less_than_two_thirds_participation() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 5;
let harness = get_harness(VALIDATOR_COUNT);
let two_thirds = (VALIDATOR_COUNT / 3) * 2;
let less_than_two_thirds = two_thirds - 1;
let attesters = (0..less_than_two_thirds).collect();
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::SomeValidators(attesters),
)
.await;
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
assert_eq!(
state.slot(),
num_blocks_produced,
"head should be at the current slot"
);
assert_eq!(
state.current_epoch(),
num_blocks_produced / MinimalEthSpec::slots_per_epoch(),
"head should be at the expected epoch"
);
assert_eq!(
state.current_justified_checkpoint().epoch,
0,
"no epoch should have been justified"
);
assert_eq!(
state.finalized_checkpoint().epoch,
0,
"no epoch should have been finalized"
);
}
#[tokio::test]
async fn does_not_finalize_without_attestation() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 5;
let harness = get_harness(VALIDATOR_COUNT);
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::SomeValidators(vec![]),
)
.await;
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
assert_eq!(
state.slot(),
num_blocks_produced,
"head should be at the current slot"
);
assert_eq!(
state.current_epoch(),
num_blocks_produced / MinimalEthSpec::slots_per_epoch(),
"head should be at the expected epoch"
);
assert_eq!(
state.current_justified_checkpoint().epoch,
0,
"no epoch should have been justified"
);
assert_eq!(
state.finalized_checkpoint().epoch,
0,
"no epoch should have been finalized"
);
}
#[tokio::test]
async fn roundtrip_operation_pool() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 5;
let harness = get_harness(VALIDATOR_COUNT);
// Add some attestations
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
assert!(harness.chain.op_pool.num_attestations() > 0);
// TODO: could add some other operations
harness
.chain
.persist_op_pool()
.expect("should persist op pool");
let restored_op_pool = harness
.chain
.store
.get_item::<PersistedOperationPool<MinimalEthSpec>>(&OP_POOL_DB_KEY)
.expect("should read db")
.expect("should find op pool")
.into_operation_pool()
.unwrap();
assert_eq!(harness.chain.op_pool, restored_op_pool);
}
#[tokio::test]
async fn unaggregated_attestations_added_to_fork_choice_some_none() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() / 2;
let harness = get_harness(VALIDATOR_COUNT);
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
let mut fork_choice = harness.chain.canonical_head.fork_choice_write_lock();
// Move forward a slot so all queued attestations can be processed.
harness.advance_slot();
fork_choice
.update_time(harness.chain.slot().unwrap())
.unwrap();
let validator_slots: Vec<(usize, Slot)> = (0..VALIDATOR_COUNT)
.into_iter()
.map(|validator_index| {
let slot = state
.get_attestation_duties(validator_index, RelativeEpoch::Current)
.expect("should get attester duties")
.unwrap()
.slot;
(validator_index, slot)
})
.collect();
for (validator, slot) in validator_slots.clone() {
let latest_message = fork_choice.latest_message(validator);
if slot <= num_blocks_produced && slot != 0 {
assert_eq!(
latest_message.unwrap().1,
slot.epoch(MinimalEthSpec::slots_per_epoch()),
"Latest message epoch for {} should be equal to epoch {}.",
validator,
slot
)
} else {
assert!(
latest_message.is_none(),
"Latest message slot should be None."
)
}
}
}
#[tokio::test]
async fn attestations_with_increasing_slots() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 5;
let harness = get_harness(VALIDATOR_COUNT);
let mut attestations = vec![];
for _ in 0..num_blocks_produced {
harness
.extend_chain(
2,
BlockStrategy::OnCanonicalHead,
// Don't produce & include any attestations (we'll collect them later).
AttestationStrategy::SomeValidators(vec![]),
)
.await;
let head = harness.chain.head_snapshot();
let head_state_root = head.beacon_state_root();
attestations.extend(harness.get_unaggregated_attestations(
&AttestationStrategy::AllValidators,
&head.beacon_state,
head_state_root,
head.beacon_block_root,
head.beacon_block.slot(),
));
harness.advance_slot();
}
for (attestation, subnet_id) in attestations.into_iter().flatten() {
let res = harness
.chain
.verify_unaggregated_attestation_for_gossip(&attestation, Some(subnet_id));
let current_slot = harness.chain.slot().expect("should get slot");
let expected_attestation_slot = attestation.data.slot;
let expected_earliest_permissible_slot =
current_slot - MinimalEthSpec::slots_per_epoch() - 1;
if expected_attestation_slot < expected_earliest_permissible_slot {
assert!(matches!(
res.err().unwrap(),
AttnError::PastSlot {
attestation_slot,
earliest_permissible_slot,
}
if attestation_slot == expected_attestation_slot && earliest_permissible_slot == expected_earliest_permissible_slot
))
} else {
res.expect("should process attestation");
}
}
}
#[tokio::test]
async fn unaggregated_attestations_added_to_fork_choice_all_updated() {
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 2 - 1;
let harness = get_harness(VALIDATOR_COUNT);
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
let head = harness.chain.head_snapshot();
let state = &head.beacon_state;
let mut fork_choice = harness.chain.canonical_head.fork_choice_write_lock();
// Move forward a slot so all queued attestations can be processed.
harness.advance_slot();
fork_choice
.update_time(harness.chain.slot().unwrap())
.unwrap();
let validators: Vec<usize> = (0..VALIDATOR_COUNT).collect();
let slots: Vec<Slot> = validators
.iter()
.map(|&v| {
state
.get_attestation_duties(v, RelativeEpoch::Current)
.expect("should get attester duties")
.unwrap()
.slot
})
.collect();
let validator_slots: Vec<(&usize, Slot)> = validators.iter().zip(slots).collect();
for (validator, slot) in validator_slots {
let latest_message = fork_choice.latest_message(*validator);
assert_eq!(
latest_message.unwrap().1,
slot.epoch(MinimalEthSpec::slots_per_epoch()),
"Latest message slot should be equal to attester duty."
);
if slot != num_blocks_produced {
let block_root = state
.get_block_root(slot)
.expect("Should get block root at slot");
assert_eq!(
latest_message.unwrap().0,
*block_root,
"Latest message block root should be equal to block at slot."
);
}
}
}
async fn run_skip_slot_test(skip_slots: u64) {
let num_validators = 8;
let harness_a = get_harness(num_validators);
let harness_b = get_harness(num_validators);
for _ in 0..skip_slots {
harness_a.advance_slot();
harness_b.advance_slot();
}
harness_a
.extend_chain(
1,
BlockStrategy::OnCanonicalHead,
// No attestation required for test.
AttestationStrategy::SomeValidators(vec![]),
)
.await;
assert_eq!(
harness_a.chain.head_snapshot().beacon_block.slot(),
Slot::new(skip_slots + 1)
);
assert_eq!(
harness_b.chain.head_snapshot().beacon_block.slot(),
Slot::new(0)
);
let status = harness_b
.chain
.process_block(
harness_a.chain.head_snapshot().beacon_block_root,
harness_a.get_head_block(),
NotifyExecutionLayer::Yes,
|| Ok(()),
)
.await
.unwrap();
let root: Hash256 = status.try_into().unwrap();
assert_eq!(root, harness_a.chain.head_snapshot().beacon_block_root);
harness_b.chain.recompute_head_at_current_slot().await;
assert_eq!(
harness_b.chain.head_snapshot().beacon_block.slot(),
Slot::new(skip_slots + 1)
);
}
#[tokio::test]
async fn produces_and_processes_with_genesis_skip_slots() {
for i in 0..MinimalEthSpec::slots_per_epoch() * 4 {
run_skip_slot_test(i).await
}
}
#[tokio::test]
async fn block_roots_skip_slot_behaviour() {
let harness = get_harness(VALIDATOR_COUNT);
// Test should be longer than the block roots to ensure a DB lookup is triggered.
let chain_length = harness
.chain
.head_snapshot()
.beacon_state
.block_roots()
.len() as u64
* 3;
let skipped_slots = [1, 6, 7, 10, chain_length];
// Build a chain with some skip slots.
for i in 1..=chain_length {
if i > 1 {
harness.advance_slot();
}
let slot = harness.chain.slot().unwrap().as_u64();
if !skipped_slots.contains(&slot) {
harness
.extend_chain(
1,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
}
}
let mut prev_unskipped_root = None;
for target_slot in 0..=chain_length {
if skipped_slots.contains(&target_slot) {
/*
* A skip slot
*/
assert!(
harness
.chain
.block_root_at_slot(target_slot.into(), WhenSlotSkipped::None)
.unwrap()
.is_none(),
"WhenSlotSkipped::None should return None on a skip slot"
);
let skipped_root = harness
.chain
.block_root_at_slot(target_slot.into(), WhenSlotSkipped::Prev)
.unwrap()
.expect("WhenSlotSkipped::Prev should always return Some");
assert_eq!(
skipped_root,
prev_unskipped_root.expect("test is badly formed"),
"WhenSlotSkipped::Prev should accurately return the prior skipped block"
);
let expected_block = harness
.chain
.get_blinded_block(&skipped_root)
.unwrap()
.unwrap();
assert_eq!(
harness
.chain
.block_at_slot(target_slot.into(), WhenSlotSkipped::Prev)
.unwrap()
.unwrap(),
expected_block,
);
assert!(
harness
.chain
.block_at_slot(target_slot.into(), WhenSlotSkipped::None)
.unwrap()
.is_none(),
"WhenSlotSkipped::None should return None on a skip slot"
);
} else {
/*
* Not a skip slot
*/
let skips_none = harness
.chain
.block_root_at_slot(target_slot.into(), WhenSlotSkipped::None)
.unwrap()
.expect("WhenSlotSkipped::None should return Some for non-skipped block");
let skips_prev = harness
.chain
.block_root_at_slot(target_slot.into(), WhenSlotSkipped::Prev)
.unwrap()
.expect("WhenSlotSkipped::Prev should always return Some");
assert_eq!(
skips_none, skips_prev,
"WhenSlotSkipped::None and WhenSlotSkipped::Prev should be equal on non-skipped slot"
);
let expected_block = harness
.chain
.get_blinded_block(&skips_prev)
.unwrap()
.unwrap();
assert_eq!(
harness
.chain
.block_at_slot(target_slot.into(), WhenSlotSkipped::Prev)
.unwrap()
.unwrap(),
expected_block
);
assert_eq!(
harness
.chain
.block_at_slot(target_slot.into(), WhenSlotSkipped::None)
.unwrap()
.unwrap(),
expected_block
);
prev_unskipped_root = Some(skips_prev);
}
}
/*
* A future, non-existent slot.
*/
let future_slot = harness.chain.slot().unwrap() + 1;
assert_eq!(
harness.chain.head_snapshot().beacon_block.slot(),
future_slot - 2,
"test precondition"
);
assert!(
harness
.chain
.block_root_at_slot(future_slot, WhenSlotSkipped::None)
.unwrap()
.is_none(),
"WhenSlotSkipped::None should return None on a future slot"
);
assert!(
harness
.chain
.block_root_at_slot(future_slot, WhenSlotSkipped::Prev)
.unwrap()
.is_none(),
"WhenSlotSkipped::Prev should return None on a future slot"
);
}