lighthouse/beacon_node/operation_pool/src/lib.rs

1951 lines
75 KiB
Rust
Raw Normal View History

mod attestation;
mod attestation_id;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
mod attestation_storage;
mod attester_slashing;
mod max_cover;
mod metrics;
2019-06-18 07:55:18 +00:00
mod persistence;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
mod reward_cache;
mod sync_aggregate_id;
2019-06-18 07:55:18 +00:00
pub use attestation::AttMaxCover;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
pub use attestation_storage::{AttestationRef, SplitAttestation};
pub use max_cover::MaxCover;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
pub use persistence::{
PersistedOperationPool, PersistedOperationPoolV12, PersistedOperationPoolV5,
};
pub use reward_cache::RewardCache;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
use crate::attestation_storage::{AttestationMap, CheckpointKey};
use crate::sync_aggregate_id::SyncAggregateId;
use attester_slashing::AttesterSlashingMaxCover;
use max_cover::maximum_cover;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
use parking_lot::{RwLock, RwLockWriteGuard};
use state_processing::per_block_processing::errors::AttestationValidationError;
2019-03-06 03:46:12 +00:00
use state_processing::per_block_processing::{
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
get_slashable_indices_modular, verify_exit, VerifySignatures,
2019-03-06 03:46:12 +00:00
};
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
use state_processing::{SigVerifiedOp, VerifyOperation};
use std::collections::{hash_map::Entry, HashMap, HashSet};
use std::marker::PhantomData;
use std::ptr;
2019-03-06 03:46:12 +00:00
use types::{
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
sync_aggregate::Error as SyncAggregateError, typenum::Unsigned, Attestation, AttestationData,
AttesterSlashing, BeaconState, BeaconStateError, ChainSpec, Epoch, EthSpec, ProposerSlashing,
SignedVoluntaryExit, Slot, SyncAggregate, SyncCommitteeContribution, Validator,
2019-03-06 03:46:12 +00:00
};
type SyncContributions<T> = RwLock<HashMap<SyncAggregateId, Vec<SyncCommitteeContribution<T>>>>;
2019-06-26 03:06:08 +00:00
#[derive(Default, Debug)]
2019-05-10 04:47:09 +00:00
pub struct OperationPool<T: EthSpec + Default> {
/// Map from attestation ID (see below) to vectors of attestations.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
attestations: RwLock<AttestationMap<T>>,
/// Map from sync aggregate ID to the best `SyncCommitteeContribution`s seen for that ID.
sync_contributions: SyncContributions<T>,
/// Set of attester slashings, and the fork version they were verified against.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
attester_slashings: RwLock<HashSet<SigVerifiedOp<AttesterSlashing<T>, T>>>,
2019-03-06 03:46:12 +00:00
/// Map from proposer index to slashing.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
proposer_slashings: RwLock<HashMap<u64, SigVerifiedOp<ProposerSlashing, T>>>,
2019-03-06 03:46:12 +00:00
/// Map from exiting validator to their exit data.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
voluntary_exits: RwLock<HashMap<u64, SigVerifiedOp<SignedVoluntaryExit, T>>>,
/// Reward cache for accelerating attestation packing.
reward_cache: RwLock<RewardCache>,
_phantom: PhantomData<T>,
2019-03-06 03:46:12 +00:00
}
#[derive(Debug, PartialEq)]
pub enum OpPoolError {
GetAttestationsTotalBalanceError(BeaconStateError),
GetBlockRootError(BeaconStateError),
SyncAggregateError(SyncAggregateError),
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
RewardCacheUpdatePrevEpoch(BeaconStateError),
RewardCacheUpdateCurrEpoch(BeaconStateError),
RewardCacheGetBlockRoot(BeaconStateError),
RewardCacheWrongEpoch,
RewardCacheValidatorUnknown(BeaconStateError),
RewardCacheOutOfBounds,
IncorrectOpPoolVariant,
}
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
#[derive(Default)]
pub struct AttestationStats {
/// Total number of attestations for all committeees/indices/votes.
pub num_attestations: usize,
/// Number of unique `AttestationData` attested to.
pub num_attestation_data: usize,
/// Maximum number of aggregates for a single `AttestationData`.
pub max_aggregates_per_data: usize,
}
impl From<SyncAggregateError> for OpPoolError {
fn from(e: SyncAggregateError) -> Self {
OpPoolError::SyncAggregateError(e)
}
}
2019-05-10 04:47:09 +00:00
impl<T: EthSpec> OperationPool<T> {
2019-03-06 03:46:12 +00:00
/// Create a new operation pool.
pub fn new() -> Self {
Self::default()
}
/// Insert a sync contribution into the pool. We don't aggregate these contributions until they
/// are retrieved from the pool.
///
/// ## Note
///
/// This function assumes the given `contribution` is valid.
pub fn insert_sync_contribution(
&self,
contribution: SyncCommitteeContribution<T>,
) -> Result<(), OpPoolError> {
let aggregate_id = SyncAggregateId::new(contribution.slot, contribution.beacon_block_root);
let mut contributions = self.sync_contributions.write();
match contributions.entry(aggregate_id) {
Entry::Vacant(entry) => {
// If no contributions exist for the key, insert the given contribution.
entry.insert(vec![contribution]);
}
Entry::Occupied(mut entry) => {
// If contributions exists for this key, check whether there exists a contribution
// with a matching `subcommittee_index`. If one exists, check whether the new or
// old contribution has more aggregation bits set. If the new one does, add it to the
// pool in place of the old one.
let existing_contributions = entry.get_mut();
match existing_contributions
.iter_mut()
.find(|existing_contribution| {
existing_contribution.subcommittee_index == contribution.subcommittee_index
}) {
Some(existing_contribution) => {
// Only need to replace the contribution if the new contribution has more
// bits set.
if existing_contribution.aggregation_bits.num_set_bits()
< contribution.aggregation_bits.num_set_bits()
{
*existing_contribution = contribution;
}
}
None => {
// If there has been no previous sync contribution for this subcommittee index,
// add it to the pool.
existing_contributions.push(contribution);
}
}
}
};
Ok(())
}
/// Calculate the `SyncAggregate` from the sync contributions that exist in the pool for the
/// slot previous to the slot associated with `state`. Return the calculated `SyncAggregate` if
/// contributions exist at this slot, or else `None`.
pub fn get_sync_aggregate(
&self,
state: &BeaconState<T>,
) -> Result<Option<SyncAggregate<T>>, OpPoolError> {
// Sync aggregates are formed from the contributions from the previous slot.
let slot = state.slot().saturating_sub(1u64);
let block_root = *state
.get_block_root(slot)
.map_err(OpPoolError::GetBlockRootError)?;
let id = SyncAggregateId::new(slot, block_root);
self.sync_contributions
.read()
.get(&id)
.map(|contributions| SyncAggregate::from_contributions(contributions))
.transpose()
.map_err(|e| e.into())
}
/// Total number of sync contributions in the pool.
pub fn num_sync_contributions(&self) -> usize {
self.sync_contributions
.read()
.values()
.map(|contributions| contributions.len())
.sum()
}
/// Remove sync contributions which are too old to be included in a block.
pub fn prune_sync_contributions(&self, current_slot: Slot) {
// Prune sync contributions that are from before the previous slot.
self.sync_contributions.write().retain(|_, contributions| {
// All the contributions in this bucket have the same data, so we only need to
// check the first one.
contributions.first().map_or(false, |contribution| {
current_slot <= contribution.slot.saturating_add(Slot::new(1))
})
});
}
/// Insert an attestation into the pool, aggregating it with existing attestations if possible.
Attestation processing (#497) * Renamed fork_choice::process_attestation_from_block * Processing attestation in fork choice * Retrieving state from store and checking signature * Looser check on beacon state validity. * Cleaned up get_attestation_state * Expanded fork choice api to provide latest validator message. * Checking if the an attestation contains a latest message * Correct process_attestation error handling. * Copy paste error in comment fixed. * Tidy ancestor iterators * Getting attestation slot via helper method * Refactored attestation creation in test utils * Revert "Refactored attestation creation in test utils" This reverts commit 4d277fe4239a7194758b18fb5c00dfe0b8231306. * Integration tests for free attestation processing * Implicit conflicts resolved. * formatting * Do first pass on Grants code * Add another attestation processing test * Tidy attestation processing * Remove old code fragment * Add non-compiling half finished changes * Simplify, fix bugs, add tests for chain iters * Remove attestation processing from op pool * Fix bug with fork choice, tidy * Fix overly restrictive check in fork choice. * Ensure committee cache is build during attn proc * Ignore unknown blocks at fork choice * Various minor fixes * Make fork choice write lock in to read lock * Remove unused method * Tidy comments * Fix attestation prod. target roots change * Fix compile error in store iters * Reject any attestation prior to finalization * Fix minor PR comments * Remove duplicated attestation finalization check * Remove awkward `let` statement
2019-08-14 00:55:24 +00:00
///
/// ## Note
///
/// This function assumes the given `attestation` is valid.
pub fn insert_attestation(
&self,
Update to frozen spec ❄️ (v0.8.1) (#444) * types: first updates for v0.8 * state_processing: epoch processing v0.8.0 * state_processing: block processing v0.8.0 * tree_hash_derive: support generics in SignedRoot * types v0.8: update to use ssz_types * state_processing v0.8: use ssz_types * ssz_types: add bitwise methods and from_elem * types: fix v0.8 FIXMEs * ssz_types: add bitfield shift_up * ssz_types: iterators and DerefMut for VariableList * types,state_processing: use VariableList * ssz_types: fix BitVector Decode impl Fixed a typo in the implementation of ssz::Decode for BitVector, which caused it to be considered variable length! * types: fix test modules for v0.8 update * types: remove slow type-level arithmetic * state_processing: fix tests for v0.8 * op_pool: update for v0.8 * ssz_types: Bitfield difference length-independent Allow computing the difference of two bitfields of different lengths. * Implement compact committee support * epoch_processing: committee & active index roots * state_processing: genesis state builder v0.8 * state_processing: implement v0.8.1 * Further improve tree_hash * Strip examples, tests from cached_tree_hash * Update TreeHash, un-impl CachedTreeHash * Update bitfield TreeHash, un-impl CachedTreeHash * Update FixedLenVec TreeHash, unimpl CachedTreeHash * Update update tree_hash_derive for new TreeHash * Fix TreeHash, un-impl CachedTreeHash for ssz_types * Remove fixed_len_vec, ssz benches SSZ benches relied upon fixed_len_vec -- it is easier to just delete them and rebuild them later (when necessary) * Remove boolean_bitfield crate * Fix fake_crypto BLS compile errors * Update ef_tests for new v.8 type params * Update ef_tests submodule to v0.8.1 tag * Make fixes to support parsing ssz ef_tests * `compact_committee...` to `compact_committees...` * Derive more traits for `CompactCommittee` * Flip bitfield byte-endianness * Fix tree_hash for bitfields * Modify CLI output for ef_tests * Bump ssz crate version * Update ssz_types doc comment * Del cached tree hash tests from ssz_static tests * Tidy SSZ dependencies * Rename ssz_types crate to eth2_ssz_types * validator_client: update for v0.8 * ssz_types: update union/difference for bit order swap * beacon_node: update for v0.8, EthSpec * types: disable cached tree hash, update min spec * state_processing: fix slot bug in committee update * tests: temporarily disable fork choice harness test See #447 * committee cache: prevent out-of-bounds access In the case where we tried to access the committee of a shard that didn't have a committee in the current epoch, we were accessing elements beyond the end of the shuffling vector and panicking! This commit adds a check to make the failure safe and explicit. * fix bug in get_indexed_attestation and simplify There was a bug in our implementation of get_indexed_attestation whereby incorrect "committee indices" were used to index into the custody bitfield. The bug was only observable in the case where some bits of the custody bitfield were set to 1. The implementation has been simplified to remove the bug, and a test added. * state_proc: workaround for compact committees bug https://github.com/ethereum/eth2.0-specs/issues/1315 * v0.8: updates to make the EF tests pass * Remove redundant max operation checks. * Always supply both messages when checking attestation signatures -- allowing verification of an attestation with no signatures. * Swap the order of the fork and domain constant in `get_domain`, to match the spec. * rustfmt * ef_tests: add new epoch processing tests * Integrate v0.8 into master (compiles) * Remove unused crates, fix clippy lints * Replace v0.6.3 tags w/ v0.8.1 * Remove old comment * Ensure lmd ghost tests only run in release * Update readme
2019-07-30 02:44:51 +00:00
attestation: Attestation<T>,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
attesting_indices: Vec<u64>,
) -> Result<(), AttestationValidationError> {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.attestations
.write()
.insert(attestation, attesting_indices);
Ok(())
}
2019-03-26 07:29:02 +00:00
/// Total number of attestations in the pool, including attestations for the same data.
pub fn num_attestations(&self) -> usize {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.attestation_stats().num_attestations
2019-03-26 07:29:02 +00:00
}
pub fn attestation_stats(&self) -> AttestationStats {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.attestations.read().stats()
}
/// Return all valid attestations for the given epoch, for use in max cover.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
#[allow(clippy::too_many_arguments)]
fn get_valid_attestations_for_epoch<'a>(
&'a self,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
checkpoint_key: &'a CheckpointKey,
all_attestations: &'a AttestationMap<T>,
state: &'a BeaconState<T>,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
reward_cache: &'a RewardCache,
total_active_balance: u64,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
validity_filter: impl FnMut(&AttestationRef<'a, T>) -> bool + Send,
spec: &'a ChainSpec,
) -> impl Iterator<Item = AttMaxCover<'a, T>> + Send {
all_attestations
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.get_attestations(checkpoint_key)
.filter(|att| {
att.data.slot + spec.min_attestation_inclusion_delay <= state.slot()
&& state.slot() <= att.data.slot + T::slots_per_epoch()
})
.filter(validity_filter)
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.filter_map(move |att| {
AttMaxCover::new(att, state, reward_cache, total_active_balance, spec)
})
}
/// Get a list of attestations for inclusion in a block.
///
/// The `validity_filter` is a closure that provides extra filtering of the attestations
/// before an approximately optimal bundle is constructed. We use it to provide access
/// to the fork choice data from the `BeaconChain` struct that doesn't logically belong
/// in the operation pool.
Update to frozen spec ❄️ (v0.8.1) (#444) * types: first updates for v0.8 * state_processing: epoch processing v0.8.0 * state_processing: block processing v0.8.0 * tree_hash_derive: support generics in SignedRoot * types v0.8: update to use ssz_types * state_processing v0.8: use ssz_types * ssz_types: add bitwise methods and from_elem * types: fix v0.8 FIXMEs * ssz_types: add bitfield shift_up * ssz_types: iterators and DerefMut for VariableList * types,state_processing: use VariableList * ssz_types: fix BitVector Decode impl Fixed a typo in the implementation of ssz::Decode for BitVector, which caused it to be considered variable length! * types: fix test modules for v0.8 update * types: remove slow type-level arithmetic * state_processing: fix tests for v0.8 * op_pool: update for v0.8 * ssz_types: Bitfield difference length-independent Allow computing the difference of two bitfields of different lengths. * Implement compact committee support * epoch_processing: committee & active index roots * state_processing: genesis state builder v0.8 * state_processing: implement v0.8.1 * Further improve tree_hash * Strip examples, tests from cached_tree_hash * Update TreeHash, un-impl CachedTreeHash * Update bitfield TreeHash, un-impl CachedTreeHash * Update FixedLenVec TreeHash, unimpl CachedTreeHash * Update update tree_hash_derive for new TreeHash * Fix TreeHash, un-impl CachedTreeHash for ssz_types * Remove fixed_len_vec, ssz benches SSZ benches relied upon fixed_len_vec -- it is easier to just delete them and rebuild them later (when necessary) * Remove boolean_bitfield crate * Fix fake_crypto BLS compile errors * Update ef_tests for new v.8 type params * Update ef_tests submodule to v0.8.1 tag * Make fixes to support parsing ssz ef_tests * `compact_committee...` to `compact_committees...` * Derive more traits for `CompactCommittee` * Flip bitfield byte-endianness * Fix tree_hash for bitfields * Modify CLI output for ef_tests * Bump ssz crate version * Update ssz_types doc comment * Del cached tree hash tests from ssz_static tests * Tidy SSZ dependencies * Rename ssz_types crate to eth2_ssz_types * validator_client: update for v0.8 * ssz_types: update union/difference for bit order swap * beacon_node: update for v0.8, EthSpec * types: disable cached tree hash, update min spec * state_processing: fix slot bug in committee update * tests: temporarily disable fork choice harness test See #447 * committee cache: prevent out-of-bounds access In the case where we tried to access the committee of a shard that didn't have a committee in the current epoch, we were accessing elements beyond the end of the shuffling vector and panicking! This commit adds a check to make the failure safe and explicit. * fix bug in get_indexed_attestation and simplify There was a bug in our implementation of get_indexed_attestation whereby incorrect "committee indices" were used to index into the custody bitfield. The bug was only observable in the case where some bits of the custody bitfield were set to 1. The implementation has been simplified to remove the bug, and a test added. * state_proc: workaround for compact committees bug https://github.com/ethereum/eth2.0-specs/issues/1315 * v0.8: updates to make the EF tests pass * Remove redundant max operation checks. * Always supply both messages when checking attestation signatures -- allowing verification of an attestation with no signatures. * Swap the order of the fork and domain constant in `get_domain`, to match the spec. * rustfmt * ef_tests: add new epoch processing tests * Integrate v0.8 into master (compiles) * Remove unused crates, fix clippy lints * Replace v0.6.3 tags w/ v0.8.1 * Remove old comment * Ensure lmd ghost tests only run in release * Update readme
2019-07-30 02:44:51 +00:00
pub fn get_attestations(
&self,
state: &BeaconState<T>,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
prev_epoch_validity_filter: impl for<'a> FnMut(&AttestationRef<'a, T>) -> bool + Send,
curr_epoch_validity_filter: impl for<'a> FnMut(&AttestationRef<'a, T>) -> bool + Send,
Update to frozen spec ❄️ (v0.8.1) (#444) * types: first updates for v0.8 * state_processing: epoch processing v0.8.0 * state_processing: block processing v0.8.0 * tree_hash_derive: support generics in SignedRoot * types v0.8: update to use ssz_types * state_processing v0.8: use ssz_types * ssz_types: add bitwise methods and from_elem * types: fix v0.8 FIXMEs * ssz_types: add bitfield shift_up * ssz_types: iterators and DerefMut for VariableList * types,state_processing: use VariableList * ssz_types: fix BitVector Decode impl Fixed a typo in the implementation of ssz::Decode for BitVector, which caused it to be considered variable length! * types: fix test modules for v0.8 update * types: remove slow type-level arithmetic * state_processing: fix tests for v0.8 * op_pool: update for v0.8 * ssz_types: Bitfield difference length-independent Allow computing the difference of two bitfields of different lengths. * Implement compact committee support * epoch_processing: committee & active index roots * state_processing: genesis state builder v0.8 * state_processing: implement v0.8.1 * Further improve tree_hash * Strip examples, tests from cached_tree_hash * Update TreeHash, un-impl CachedTreeHash * Update bitfield TreeHash, un-impl CachedTreeHash * Update FixedLenVec TreeHash, unimpl CachedTreeHash * Update update tree_hash_derive for new TreeHash * Fix TreeHash, un-impl CachedTreeHash for ssz_types * Remove fixed_len_vec, ssz benches SSZ benches relied upon fixed_len_vec -- it is easier to just delete them and rebuild them later (when necessary) * Remove boolean_bitfield crate * Fix fake_crypto BLS compile errors * Update ef_tests for new v.8 type params * Update ef_tests submodule to v0.8.1 tag * Make fixes to support parsing ssz ef_tests * `compact_committee...` to `compact_committees...` * Derive more traits for `CompactCommittee` * Flip bitfield byte-endianness * Fix tree_hash for bitfields * Modify CLI output for ef_tests * Bump ssz crate version * Update ssz_types doc comment * Del cached tree hash tests from ssz_static tests * Tidy SSZ dependencies * Rename ssz_types crate to eth2_ssz_types * validator_client: update for v0.8 * ssz_types: update union/difference for bit order swap * beacon_node: update for v0.8, EthSpec * types: disable cached tree hash, update min spec * state_processing: fix slot bug in committee update * tests: temporarily disable fork choice harness test See #447 * committee cache: prevent out-of-bounds access In the case where we tried to access the committee of a shard that didn't have a committee in the current epoch, we were accessing elements beyond the end of the shuffling vector and panicking! This commit adds a check to make the failure safe and explicit. * fix bug in get_indexed_attestation and simplify There was a bug in our implementation of get_indexed_attestation whereby incorrect "committee indices" were used to index into the custody bitfield. The bug was only observable in the case where some bits of the custody bitfield were set to 1. The implementation has been simplified to remove the bug, and a test added. * state_proc: workaround for compact committees bug https://github.com/ethereum/eth2.0-specs/issues/1315 * v0.8: updates to make the EF tests pass * Remove redundant max operation checks. * Always supply both messages when checking attestation signatures -- allowing verification of an attestation with no signatures. * Swap the order of the fork and domain constant in `get_domain`, to match the spec. * rustfmt * ef_tests: add new epoch processing tests * Integrate v0.8 into master (compiles) * Remove unused crates, fix clippy lints * Replace v0.6.3 tags w/ v0.8.1 * Remove old comment * Ensure lmd ghost tests only run in release * Update readme
2019-07-30 02:44:51 +00:00
spec: &ChainSpec,
) -> Result<Vec<Attestation<T>>, OpPoolError> {
// Attestations for the current fork, which may be from the current or previous epoch.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let (prev_epoch_key, curr_epoch_key) = CheckpointKey::keys_for_state(state);
let all_attestations = self.attestations.read();
let total_active_balance = state
.get_total_active_balance()
.map_err(OpPoolError::GetAttestationsTotalBalanceError)?;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
// Update the reward cache.
let reward_timer = metrics::start_timer(&metrics::BUILD_REWARD_CACHE_TIME);
let mut reward_cache = self.reward_cache.write();
reward_cache.update(state)?;
let reward_cache = RwLockWriteGuard::downgrade(reward_cache);
drop(reward_timer);
// Split attestations for the previous & current epochs, so that we
// can optimise them individually in parallel.
let mut num_prev_valid = 0_i64;
let mut num_curr_valid = 0_i64;
let prev_epoch_att = self
.get_valid_attestations_for_epoch(
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
&prev_epoch_key,
&*all_attestations,
state,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
&*reward_cache,
total_active_balance,
prev_epoch_validity_filter,
spec,
)
.inspect(|_| num_prev_valid += 1);
let curr_epoch_att = self
.get_valid_attestations_for_epoch(
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
&curr_epoch_key,
&*all_attestations,
state,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
&*reward_cache,
total_active_balance,
curr_epoch_validity_filter,
spec,
)
.inspect(|_| num_curr_valid += 1);
let prev_epoch_limit = if let BeaconState::Base(base_state) = state {
std::cmp::min(
T::MaxPendingAttestations::to_usize()
.saturating_sub(base_state.previous_epoch_attestations.len()),
T::MaxAttestations::to_usize(),
)
} else {
T::MaxAttestations::to_usize()
};
let (prev_cover, curr_cover) = rayon::join(
move || {
let _timer = metrics::start_timer(&metrics::ATTESTATION_PREV_EPOCH_PACKING_TIME);
// If we're in the genesis epoch, just use the current epoch attestations.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
if prev_epoch_key == curr_epoch_key {
vec![]
} else {
maximum_cover(prev_epoch_att, prev_epoch_limit, "prev_epoch_attestations")
}
},
move || {
let _timer = metrics::start_timer(&metrics::ATTESTATION_CURR_EPOCH_PACKING_TIME);
maximum_cover(
curr_epoch_att,
T::MaxAttestations::to_usize(),
"curr_epoch_attestations",
)
},
);
metrics::set_gauge(&metrics::NUM_PREV_EPOCH_ATTESTATIONS, num_prev_valid);
metrics::set_gauge(&metrics::NUM_CURR_EPOCH_ATTESTATIONS, num_curr_valid);
Ok(max_cover::merge_solutions(
curr_cover,
prev_cover,
T::MaxAttestations::to_usize(),
))
}
2019-03-19 23:47:19 +00:00
/// Remove attestations which are too old to be included in a block.
pub fn prune_attestations(&self, current_epoch: Epoch) {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.attestations.write().prune(current_epoch);
}
2019-03-06 03:46:12 +00:00
/// Insert a proposer slashing into the pool.
pub fn insert_proposer_slashing(
&self,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
verified_proposer_slashing: SigVerifiedOp<ProposerSlashing, T>,
) {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.proposer_slashings.write().insert(
verified_proposer_slashing.as_inner().proposer_index(),
verified_proposer_slashing,
);
2019-03-25 05:58:20 +00:00
}
/// Insert an attester slashing into the pool.
pub fn insert_attester_slashing(
&self,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
verified_slashing: SigVerifiedOp<AttesterSlashing<T>, T>,
) {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.attester_slashings.write().insert(verified_slashing);
2019-03-25 05:58:20 +00:00
}
/// Get proposer and attester slashings for inclusion in a block.
///
/// This function computes both types of slashings together, because
/// attester slashings may be invalidated by proposer slashings included
/// earlier in the block.
pub fn get_slashings_and_exits(
2019-03-06 03:46:12 +00:00
&self,
state: &BeaconState<T>,
spec: &ChainSpec,
) -> (
Vec<ProposerSlashing>,
Vec<AttesterSlashing<T>>,
Vec<SignedVoluntaryExit>,
) {
2019-03-25 05:58:20 +00:00
let proposer_slashings = filter_limit_operations(
self.proposer_slashings.read().values(),
2019-03-06 03:46:12 +00:00
|slashing| {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
slashing.signature_is_still_valid(&state.fork())
&& state
.validators()
.get(slashing.as_inner().signed_header_1.message.proposer_index as usize)
.map_or(false, |validator| !validator.slashed)
2019-03-06 03:46:12 +00:00
},
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
|slashing| slashing.as_inner().clone(),
Update to frozen spec ❄️ (v0.8.1) (#444) * types: first updates for v0.8 * state_processing: epoch processing v0.8.0 * state_processing: block processing v0.8.0 * tree_hash_derive: support generics in SignedRoot * types v0.8: update to use ssz_types * state_processing v0.8: use ssz_types * ssz_types: add bitwise methods and from_elem * types: fix v0.8 FIXMEs * ssz_types: add bitfield shift_up * ssz_types: iterators and DerefMut for VariableList * types,state_processing: use VariableList * ssz_types: fix BitVector Decode impl Fixed a typo in the implementation of ssz::Decode for BitVector, which caused it to be considered variable length! * types: fix test modules for v0.8 update * types: remove slow type-level arithmetic * state_processing: fix tests for v0.8 * op_pool: update for v0.8 * ssz_types: Bitfield difference length-independent Allow computing the difference of two bitfields of different lengths. * Implement compact committee support * epoch_processing: committee & active index roots * state_processing: genesis state builder v0.8 * state_processing: implement v0.8.1 * Further improve tree_hash * Strip examples, tests from cached_tree_hash * Update TreeHash, un-impl CachedTreeHash * Update bitfield TreeHash, un-impl CachedTreeHash * Update FixedLenVec TreeHash, unimpl CachedTreeHash * Update update tree_hash_derive for new TreeHash * Fix TreeHash, un-impl CachedTreeHash for ssz_types * Remove fixed_len_vec, ssz benches SSZ benches relied upon fixed_len_vec -- it is easier to just delete them and rebuild them later (when necessary) * Remove boolean_bitfield crate * Fix fake_crypto BLS compile errors * Update ef_tests for new v.8 type params * Update ef_tests submodule to v0.8.1 tag * Make fixes to support parsing ssz ef_tests * `compact_committee...` to `compact_committees...` * Derive more traits for `CompactCommittee` * Flip bitfield byte-endianness * Fix tree_hash for bitfields * Modify CLI output for ef_tests * Bump ssz crate version * Update ssz_types doc comment * Del cached tree hash tests from ssz_static tests * Tidy SSZ dependencies * Rename ssz_types crate to eth2_ssz_types * validator_client: update for v0.8 * ssz_types: update union/difference for bit order swap * beacon_node: update for v0.8, EthSpec * types: disable cached tree hash, update min spec * state_processing: fix slot bug in committee update * tests: temporarily disable fork choice harness test See #447 * committee cache: prevent out-of-bounds access In the case where we tried to access the committee of a shard that didn't have a committee in the current epoch, we were accessing elements beyond the end of the shuffling vector and panicking! This commit adds a check to make the failure safe and explicit. * fix bug in get_indexed_attestation and simplify There was a bug in our implementation of get_indexed_attestation whereby incorrect "committee indices" were used to index into the custody bitfield. The bug was only observable in the case where some bits of the custody bitfield were set to 1. The implementation has been simplified to remove the bug, and a test added. * state_proc: workaround for compact committees bug https://github.com/ethereum/eth2.0-specs/issues/1315 * v0.8: updates to make the EF tests pass * Remove redundant max operation checks. * Always supply both messages when checking attestation signatures -- allowing verification of an attestation with no signatures. * Swap the order of the fork and domain constant in `get_domain`, to match the spec. * rustfmt * ef_tests: add new epoch processing tests * Integrate v0.8 into master (compiles) * Remove unused crates, fix clippy lints * Replace v0.6.3 tags w/ v0.8.1 * Remove old comment * Ensure lmd ghost tests only run in release * Update readme
2019-07-30 02:44:51 +00:00
T::MaxProposerSlashings::to_usize(),
2019-03-25 05:58:20 +00:00
);
// Set of validators to be slashed, so we don't attempt to construct invalid attester
// slashings.
let mut to_be_slashed = proposer_slashings
2019-03-25 05:58:20 +00:00
.iter()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.map(|s| s.proposer_index())
.collect();
let attester_slashings = self.get_attester_slashings(state, &mut to_be_slashed);
2019-03-25 05:58:20 +00:00
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let voluntary_exits = self.get_voluntary_exits(
state,
|exit| !to_be_slashed.contains(&exit.message.validator_index),
spec,
);
(proposer_slashings, attester_slashings, voluntary_exits)
}
/// Get attester slashings taking into account already slashed validators.
///
/// This function *must* remain private.
fn get_attester_slashings(
&self,
state: &BeaconState<T>,
to_be_slashed: &mut HashSet<u64>,
) -> Vec<AttesterSlashing<T>> {
let reader = self.attester_slashings.read();
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let relevant_attester_slashings = reader.iter().flat_map(|slashing| {
if slashing.signature_is_still_valid(&state.fork()) {
AttesterSlashingMaxCover::new(slashing.as_inner(), to_be_slashed, state)
} else {
None
}
});
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
maximum_cover(
relevant_attester_slashings,
T::MaxAttesterSlashings::to_usize(),
"attester_slashings",
)
.into_iter()
.map(|cover| {
to_be_slashed.extend(cover.covering_set().keys());
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
cover.intermediate().clone()
})
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.collect()
2019-03-06 03:46:12 +00:00
}
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
/// Prune proposer slashings for validators which are exited in the finalized epoch.
pub fn prune_proposer_slashings(&self, head_state: &BeaconState<T>) {
prune_validator_hash_map(
&mut self.proposer_slashings.write(),
|validator| validator.exit_epoch <= head_state.finalized_checkpoint().epoch,
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
head_state,
);
2019-03-06 03:46:12 +00:00
}
2019-03-25 05:58:20 +00:00
/// Prune attester slashings for all slashed or withdrawn validators, or attestations on another
/// fork.
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
pub fn prune_attester_slashings(&self, head_state: &BeaconState<T>) {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
self.attester_slashings.write().retain(|slashing| {
// Check that the attestation's signature is still valid wrt the fork version.
let signature_ok = slashing.signature_is_still_valid(&head_state.fork());
// Slashings that don't slash any validators can also be dropped.
let slashing_ok =
get_slashable_indices_modular(head_state, slashing.as_inner(), |_, validator| {
// Declare that a validator is still slashable if they have not exited prior
// to the finalized epoch.
//
// We cannot check the `slashed` field since the `head` is not finalized and
// a fork could un-slash someone.
validator.exit_epoch > head_state.finalized_checkpoint().epoch
})
.map_or(false, |indices| !indices.is_empty());
signature_ok && slashing_ok
});
2019-03-25 05:58:20 +00:00
}
2019-03-06 03:46:12 +00:00
/// Total number of attester slashings in the pool.
pub fn num_attester_slashings(&self) -> usize {
self.attester_slashings.read().len()
}
/// Total number of proposer slashings in the pool.
pub fn num_proposer_slashings(&self) -> usize {
self.proposer_slashings.read().len()
}
/// Insert a voluntary exit that has previously been checked elsewhere.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
pub fn insert_voluntary_exit(&self, exit: SigVerifiedOp<SignedVoluntaryExit, T>) {
self.voluntary_exits
.write()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.insert(exit.as_inner().message.validator_index, exit);
2019-03-06 03:46:12 +00:00
}
/// Get a list of voluntary exits for inclusion in a block.
fn get_voluntary_exits<F>(
&self,
state: &BeaconState<T>,
filter: F,
spec: &ChainSpec,
) -> Vec<SignedVoluntaryExit>
where
F: Fn(&SignedVoluntaryExit) -> bool,
{
2019-03-06 03:46:12 +00:00
filter_limit_operations(
self.voluntary_exits.read().values(),
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
|exit| {
filter(exit.as_inner())
&& exit.signature_is_still_valid(&state.fork())
&& verify_exit(state, exit.as_inner(), VerifySignatures::False, spec).is_ok()
},
|exit| exit.as_inner().clone(),
Update to frozen spec ❄️ (v0.8.1) (#444) * types: first updates for v0.8 * state_processing: epoch processing v0.8.0 * state_processing: block processing v0.8.0 * tree_hash_derive: support generics in SignedRoot * types v0.8: update to use ssz_types * state_processing v0.8: use ssz_types * ssz_types: add bitwise methods and from_elem * types: fix v0.8 FIXMEs * ssz_types: add bitfield shift_up * ssz_types: iterators and DerefMut for VariableList * types,state_processing: use VariableList * ssz_types: fix BitVector Decode impl Fixed a typo in the implementation of ssz::Decode for BitVector, which caused it to be considered variable length! * types: fix test modules for v0.8 update * types: remove slow type-level arithmetic * state_processing: fix tests for v0.8 * op_pool: update for v0.8 * ssz_types: Bitfield difference length-independent Allow computing the difference of two bitfields of different lengths. * Implement compact committee support * epoch_processing: committee & active index roots * state_processing: genesis state builder v0.8 * state_processing: implement v0.8.1 * Further improve tree_hash * Strip examples, tests from cached_tree_hash * Update TreeHash, un-impl CachedTreeHash * Update bitfield TreeHash, un-impl CachedTreeHash * Update FixedLenVec TreeHash, unimpl CachedTreeHash * Update update tree_hash_derive for new TreeHash * Fix TreeHash, un-impl CachedTreeHash for ssz_types * Remove fixed_len_vec, ssz benches SSZ benches relied upon fixed_len_vec -- it is easier to just delete them and rebuild them later (when necessary) * Remove boolean_bitfield crate * Fix fake_crypto BLS compile errors * Update ef_tests for new v.8 type params * Update ef_tests submodule to v0.8.1 tag * Make fixes to support parsing ssz ef_tests * `compact_committee...` to `compact_committees...` * Derive more traits for `CompactCommittee` * Flip bitfield byte-endianness * Fix tree_hash for bitfields * Modify CLI output for ef_tests * Bump ssz crate version * Update ssz_types doc comment * Del cached tree hash tests from ssz_static tests * Tidy SSZ dependencies * Rename ssz_types crate to eth2_ssz_types * validator_client: update for v0.8 * ssz_types: update union/difference for bit order swap * beacon_node: update for v0.8, EthSpec * types: disable cached tree hash, update min spec * state_processing: fix slot bug in committee update * tests: temporarily disable fork choice harness test See #447 * committee cache: prevent out-of-bounds access In the case where we tried to access the committee of a shard that didn't have a committee in the current epoch, we were accessing elements beyond the end of the shuffling vector and panicking! This commit adds a check to make the failure safe and explicit. * fix bug in get_indexed_attestation and simplify There was a bug in our implementation of get_indexed_attestation whereby incorrect "committee indices" were used to index into the custody bitfield. The bug was only observable in the case where some bits of the custody bitfield were set to 1. The implementation has been simplified to remove the bug, and a test added. * state_proc: workaround for compact committees bug https://github.com/ethereum/eth2.0-specs/issues/1315 * v0.8: updates to make the EF tests pass * Remove redundant max operation checks. * Always supply both messages when checking attestation signatures -- allowing verification of an attestation with no signatures. * Swap the order of the fork and domain constant in `get_domain`, to match the spec. * rustfmt * ef_tests: add new epoch processing tests * Integrate v0.8 into master (compiles) * Remove unused crates, fix clippy lints * Replace v0.6.3 tags w/ v0.8.1 * Remove old comment * Ensure lmd ghost tests only run in release * Update readme
2019-07-30 02:44:51 +00:00
T::MaxVoluntaryExits::to_usize(),
2019-03-06 03:46:12 +00:00
)
}
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
/// Prune if validator has already exited at or before the finalized checkpoint of the head.
pub fn prune_voluntary_exits(&self, head_state: &BeaconState<T>) {
prune_validator_hash_map(
&mut self.voluntary_exits.write(),
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
// This condition is slightly too loose, since there will be some finalized exits that
// are missed here.
//
// We choose simplicity over the gain of pruning more exits since they are small and
// should not be seen frequently.
|validator| validator.exit_epoch <= head_state.finalized_checkpoint().epoch,
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
head_state,
);
2019-03-06 03:46:12 +00:00
}
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
/// Prune all types of transactions given the latest head state and head fork.
pub fn prune_all(&self, head_state: &BeaconState<T>, current_epoch: Epoch) {
self.prune_attestations(current_epoch);
self.prune_sync_contributions(head_state.slot());
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
self.prune_proposer_slashings(head_state);
self.prune_attester_slashings(head_state);
self.prune_voluntary_exits(head_state);
2019-03-19 23:14:31 +00:00
}
/// Total number of voluntary exits in the pool.
pub fn num_voluntary_exits(&self) -> usize {
self.voluntary_exits.read().len()
}
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
/// Returns all known `Attestation` objects.
///
/// This method may return objects that are invalid for block inclusion.
pub fn get_all_attestations(&self) -> Vec<Attestation<T>> {
self.attestations
.read()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.iter()
.map(|att| att.clone_as_attestation())
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
.collect()
}
/// Returns all known `Attestation` objects that pass the provided filter.
///
/// This method may return objects that are invalid for block inclusion.
pub fn get_filtered_attestations<F>(&self, filter: F) -> Vec<Attestation<T>>
where
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
F: Fn(&AttestationData) -> bool,
{
self.attestations
.read()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.iter()
.filter(|att| filter(&att.attestation_data()))
.map(|att| att.clone_as_attestation())
.collect()
}
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
/// Returns all known `AttesterSlashing` objects.
///
/// This method may return objects that are invalid for block inclusion.
pub fn get_all_attester_slashings(&self) -> Vec<AttesterSlashing<T>> {
self.attester_slashings
.read()
.iter()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.map(|slashing| slashing.as_inner().clone())
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
.collect()
}
/// Returns all known `ProposerSlashing` objects.
///
/// This method may return objects that are invalid for block inclusion.
pub fn get_all_proposer_slashings(&self) -> Vec<ProposerSlashing> {
self.proposer_slashings
.read()
.iter()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.map(|(_, slashing)| slashing.as_inner().clone())
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
.collect()
}
/// Returns all known `SignedVoluntaryExit` objects.
///
/// This method may return objects that are invalid for block inclusion.
pub fn get_all_voluntary_exits(&self) -> Vec<SignedVoluntaryExit> {
self.voluntary_exits
.read()
.iter()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.map(|(_, exit)| exit.as_inner().clone())
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
.collect()
}
2019-03-06 03:46:12 +00:00
}
2019-03-25 05:58:20 +00:00
/// Filter up to a maximum number of operations out of an iterator.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
fn filter_limit_operations<'a, T: 'a, V: 'a, I, F, G>(
operations: I,
filter: F,
mapping: G,
limit: usize,
) -> Vec<V>
2019-03-06 03:46:12 +00:00
where
I: IntoIterator<Item = &'a T>,
F: Fn(&T) -> bool,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
G: Fn(&T) -> V,
2019-03-06 03:46:12 +00:00
T: Clone,
{
operations
.into_iter()
.filter(|x| filter(*x))
Update to frozen spec ❄️ (v0.8.1) (#444) * types: first updates for v0.8 * state_processing: epoch processing v0.8.0 * state_processing: block processing v0.8.0 * tree_hash_derive: support generics in SignedRoot * types v0.8: update to use ssz_types * state_processing v0.8: use ssz_types * ssz_types: add bitwise methods and from_elem * types: fix v0.8 FIXMEs * ssz_types: add bitfield shift_up * ssz_types: iterators and DerefMut for VariableList * types,state_processing: use VariableList * ssz_types: fix BitVector Decode impl Fixed a typo in the implementation of ssz::Decode for BitVector, which caused it to be considered variable length! * types: fix test modules for v0.8 update * types: remove slow type-level arithmetic * state_processing: fix tests for v0.8 * op_pool: update for v0.8 * ssz_types: Bitfield difference length-independent Allow computing the difference of two bitfields of different lengths. * Implement compact committee support * epoch_processing: committee & active index roots * state_processing: genesis state builder v0.8 * state_processing: implement v0.8.1 * Further improve tree_hash * Strip examples, tests from cached_tree_hash * Update TreeHash, un-impl CachedTreeHash * Update bitfield TreeHash, un-impl CachedTreeHash * Update FixedLenVec TreeHash, unimpl CachedTreeHash * Update update tree_hash_derive for new TreeHash * Fix TreeHash, un-impl CachedTreeHash for ssz_types * Remove fixed_len_vec, ssz benches SSZ benches relied upon fixed_len_vec -- it is easier to just delete them and rebuild them later (when necessary) * Remove boolean_bitfield crate * Fix fake_crypto BLS compile errors * Update ef_tests for new v.8 type params * Update ef_tests submodule to v0.8.1 tag * Make fixes to support parsing ssz ef_tests * `compact_committee...` to `compact_committees...` * Derive more traits for `CompactCommittee` * Flip bitfield byte-endianness * Fix tree_hash for bitfields * Modify CLI output for ef_tests * Bump ssz crate version * Update ssz_types doc comment * Del cached tree hash tests from ssz_static tests * Tidy SSZ dependencies * Rename ssz_types crate to eth2_ssz_types * validator_client: update for v0.8 * ssz_types: update union/difference for bit order swap * beacon_node: update for v0.8, EthSpec * types: disable cached tree hash, update min spec * state_processing: fix slot bug in committee update * tests: temporarily disable fork choice harness test See #447 * committee cache: prevent out-of-bounds access In the case where we tried to access the committee of a shard that didn't have a committee in the current epoch, we were accessing elements beyond the end of the shuffling vector and panicking! This commit adds a check to make the failure safe and explicit. * fix bug in get_indexed_attestation and simplify There was a bug in our implementation of get_indexed_attestation whereby incorrect "committee indices" were used to index into the custody bitfield. The bug was only observable in the case where some bits of the custody bitfield were set to 1. The implementation has been simplified to remove the bug, and a test added. * state_proc: workaround for compact committees bug https://github.com/ethereum/eth2.0-specs/issues/1315 * v0.8: updates to make the EF tests pass * Remove redundant max operation checks. * Always supply both messages when checking attestation signatures -- allowing verification of an attestation with no signatures. * Swap the order of the fork and domain constant in `get_domain`, to match the spec. * rustfmt * ef_tests: add new epoch processing tests * Integrate v0.8 into master (compiles) * Remove unused crates, fix clippy lints * Replace v0.6.3 tags w/ v0.8.1 * Remove old comment * Ensure lmd ghost tests only run in release * Update readme
2019-07-30 02:44:51 +00:00
.take(limit)
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.map(mapping)
2019-03-06 03:46:12 +00:00
.collect()
}
/// Remove all entries from the given hash map for which `prune_if` returns true.
///
/// The keys in the map should be validator indices, which will be looked up
/// in the state's validator registry and then passed to `prune_if`.
/// Entries for unknown validators will be kept.
fn prune_validator_hash_map<T, F, E: EthSpec>(
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
map: &mut HashMap<u64, SigVerifiedOp<T, E>>,
prune_if: F,
Address queue congestion in migrator (#1923) ## Issue Addressed *Should* address #1917 ## Proposed Changes Stops the `BackgroupMigrator` rx channel from backing up with big `BeaconState` messages. Looking at some logs from my Medalla node, we can see a discrepancy between the head finalized epoch and the migrator finalized epoch: ``` Nov 17 16:50:21.606 DEBG Head beacon block slot: 129214, root: 0xbc7a…0b99, finalized_epoch: 4033, finalized_root: 0xf930…6562, justified_epoch: 4035, justified_root: 0x206b…9321, service: beacon Nov 17 16:50:21.626 DEBG Batch processed service: sync, processed_blocks: 43, last_block_slot: 129214, chain: 8274002112260436595, first_block_slot: 129153, batch_epoch: 4036 Nov 17 16:50:21.626 DEBG Chain advanced processing_target: 4036, new_start: 4036, previous_start: 4034, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Completed batch received awaiting_batches: 5, blocks: 47, epoch: 4048, chain: 8274002112260436595, service: sync Nov 17 16:50:22.162 DEBG Requesting batch start_slot: 129601, end_slot: 129664, downloaded: 0, processed: 0, state: Downloading(16Uiu2HAmG3C3t1McaseReECjAF694tjVVjkDoneZEbxNhWm1nZaT, 0 blocks, 1273), epoch: 4050, chain: 8274002112260436595, service: sync Nov 17 16:50:22.654 DEBG Database compaction complete service: beacon Nov 17 16:50:22.655 INFO Starting database pruning new_finalized_epoch: 2193, old_finalized_epoch: 2192, service: beacon ``` I believe this indicates that the migrator rx has a backed-up queue of `MigrationNotification` items which each contain a `BeaconState`. ## TODO - [x] Remove finalized state requirement for op-pool
2020-11-17 23:11:26 +00:00
head_state: &BeaconState<E>,
) where
F: Fn(&Validator) -> bool,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
T: VerifyOperation<E>,
{
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
map.retain(|&validator_index, op| {
op.signature_is_still_valid(&head_state.fork())
&& head_state
.validators()
.get(validator_index as usize)
.map_or(true, |validator| !prune_if(validator))
});
}
2019-06-26 03:06:08 +00:00
/// Compare two operation pools.
impl<T: EthSpec + Default> PartialEq for OperationPool<T> {
fn eq(&self, other: &Self) -> bool {
if ptr::eq(self, other) {
return true;
}
2019-06-26 03:06:08 +00:00
*self.attestations.read() == *other.attestations.read()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
&& *self.sync_contributions.read() == *other.sync_contributions.read()
2019-06-26 03:06:08 +00:00
&& *self.attester_slashings.read() == *other.attester_slashings.read()
&& *self.proposer_slashings.read() == *other.proposer_slashings.read()
&& *self.voluntary_exits.read() == *other.voluntary_exits.read()
}
}
#[cfg(all(test, not(debug_assertions)))]
mod release_tests {
use super::attestation::earliest_attestation_validators;
2019-03-06 03:46:12 +00:00
use super::*;
use beacon_chain::test_utils::{
test_spec, BeaconChainHarness, EphemeralHarnessType, RelativeSyncCommittee,
};
use lazy_static::lazy_static;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
use maplit::hashset;
use state_processing::{common::get_attesting_indices_from_state, VerifyOperation};
use std::collections::BTreeSet;
use types::consts::altair::SYNC_COMMITTEE_SUBNET_COUNT;
use types::*;
2019-03-06 03:46:12 +00:00
pub const MAX_VALIDATOR_COUNT: usize = 4 * 32 * 128;
lazy_static! {
/// A cached set of keys.
static ref KEYPAIRS: Vec<Keypair> = types::test_utils::generate_deterministic_keypairs(MAX_VALIDATOR_COUNT);
}
fn get_harness<E: EthSpec>(
validator_count: usize,
spec: Option<ChainSpec>,
) -> BeaconChainHarness<EphemeralHarnessType<E>> {
let harness = BeaconChainHarness::builder(E::default())
.spec_or_default(spec)
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.fresh_ephemeral_store()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.mock_execution_layer()
.build();
harness.advance_slot();
harness
2019-03-06 03:46:12 +00:00
}
/// Test state for attestation-related tests.
fn attestation_test_state<E: EthSpec>(
num_committees: usize,
) -> (BeaconChainHarness<EphemeralHarnessType<E>>, ChainSpec) {
let spec = test_spec::<E>();
let num_validators =
num_committees * E::slots_per_epoch() as usize * spec.target_committee_size;
let harness = get_harness::<E>(num_validators, Some(spec.clone()));
(harness, spec)
}
/// Test state for sync contribution-related tests.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
async fn sync_contribution_test_state<E: EthSpec>(
num_committees: usize,
) -> (BeaconChainHarness<EphemeralHarnessType<E>>, ChainSpec) {
let mut spec = E::default_spec();
spec.altair_fork_epoch = Some(Epoch::new(0));
let num_validators =
num_committees * E::slots_per_epoch() as usize * spec.target_committee_size;
let harness = get_harness::<E>(num_validators, Some(spec.clone()));
let state = harness.get_current_state();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.add_attested_blocks_at_slots(
state,
Hash256::zero(),
&[Slot::new(1)],
(0..num_validators).collect::<Vec<_>>().as_slice(),
)
.await;
(harness, spec)
2019-03-06 03:46:12 +00:00
}
2019-03-20 01:44:37 +00:00
#[test]
fn test_earliest_attestation() {
let (harness, ref spec) = attestation_test_state::<MainnetEthSpec>(1);
// Only run this test on the phase0 hard-fork.
if spec.altair_fork_epoch != None {
return;
}
let mut state = harness.get_current_state();
let slot = state.slot();
let committees = state
.get_beacon_committees_at_slot(slot)
.unwrap()
.into_iter()
.map(BeaconCommittee::into_owned)
.collect::<Vec<_>>();
let num_validators =
MainnetEthSpec::slots_per_epoch() as usize * spec.target_committee_size;
let attestations = harness.make_attestations(
(0..num_validators).collect::<Vec<_>>().as_slice(),
&state,
Hash256::zero(),
SignedBeaconBlockHash::from(Hash256::zero()),
slot,
);
for (atts, aggregate) in &attestations {
let att2 = aggregate.as_ref().unwrap().message.aggregate.clone();
let att1 = atts
.into_iter()
.map(|(att, _)| att)
.take(2)
.fold::<Option<Attestation<MainnetEthSpec>>, _>(None, |att, new_att| {
if let Some(mut a) = att {
a.aggregate(&new_att);
Some(a)
} else {
Some(new_att.clone())
}
})
.unwrap();
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let att1_indices = get_attesting_indices_from_state(&state, &att1).unwrap();
let att2_indices = get_attesting_indices_from_state(&state, &att2).unwrap();
let att1_split = SplitAttestation::new(att1.clone(), att1_indices);
let att2_split = SplitAttestation::new(att2.clone(), att2_indices);
assert_eq!(
att1.aggregation_bits.num_set_bits(),
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
earliest_attestation_validators(
&att1_split.as_ref(),
&state,
state.as_base().unwrap()
)
.num_set_bits()
);
state
.as_base_mut()
.unwrap()
.current_epoch_attestations
.push(PendingAttestation {
aggregation_bits: att1.aggregation_bits.clone(),
data: att1.data.clone(),
inclusion_delay: 0,
proposer_index: 0,
})
.unwrap();
assert_eq!(
committees.get(0).unwrap().committee.len() - 2,
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
earliest_attestation_validators(
&att2_split.as_ref(),
&state,
state.as_base().unwrap()
)
.num_set_bits()
2019-03-26 07:29:02 +00:00
);
2019-05-09 08:56:41 +00:00
}
}
2019-03-26 07:29:02 +00:00
/// End-to-end test of basic attestation handling.
#[test]
fn attestation_aggregation_insert_get_prune() {
let (harness, ref spec) = attestation_test_state::<MainnetEthSpec>(1);
let op_pool = OperationPool::<MainnetEthSpec>::new();
let mut state = harness.get_current_state();
let slot = state.slot();
let committees = state
.get_beacon_committees_at_slot(slot)
.unwrap()
.into_iter()
.map(BeaconCommittee::into_owned)
.collect::<Vec<_>>();
assert_eq!(
committees.len(),
1,
"we expect just one committee with this many validators"
);
let num_validators =
MainnetEthSpec::slots_per_epoch() as usize * spec.target_committee_size;
let attestations = harness.make_attestations(
(0..num_validators).collect::<Vec<_>>().as_slice(),
&state,
Hash256::zero(),
SignedBeaconBlockHash::from(Hash256::zero()),
slot,
);
for (atts, _) in attestations {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
for (att, _) in atts {
let attesting_indices = get_attesting_indices_from_state(&state, &att).unwrap();
op_pool.insert_attestation(att, attesting_indices).unwrap();
2019-03-26 07:29:02 +00:00
}
}
assert_eq!(op_pool.num_attestations(), committees.len());
2019-03-26 07:29:02 +00:00
// Before the min attestation inclusion delay, get_attestations shouldn't return anything.
assert_eq!(
op_pool
.get_attestations(&state, |_| true, |_| true, spec)
.expect("should have attestations")
.len(),
0
);
2019-03-26 07:29:02 +00:00
// Then once the delay has elapsed, we should get a single aggregated attestation.
*state.slot_mut() += spec.min_attestation_inclusion_delay;
2019-03-26 07:29:02 +00:00
let block_attestations = op_pool
.get_attestations(&state, |_| true, |_| true, spec)
.expect("Should have block attestations");
assert_eq!(block_attestations.len(), committees.len());
2019-03-26 07:29:02 +00:00
let agg_att = &block_attestations[0];
assert_eq!(
agg_att.aggregation_bits.num_set_bits(),
spec.target_committee_size as usize
);
2019-03-26 07:29:02 +00:00
// Prune attestations shouldn't do anything at this point.
op_pool.prune_attestations(state.current_epoch());
assert_eq!(op_pool.num_attestations(), committees.len());
2019-03-26 07:29:02 +00:00
// But once we advance to more than an epoch after the attestation, it should prune it
// out of existence.
*state.slot_mut() += 2 * MainnetEthSpec::slots_per_epoch();
op_pool.prune_attestations(state.current_epoch());
assert_eq!(op_pool.num_attestations(), 0);
}
2019-05-09 08:56:41 +00:00
/// Adding an attestation already in the pool should not increase the size of the pool.
#[test]
fn attestation_duplicate() {
let (harness, ref spec) = attestation_test_state::<MainnetEthSpec>(1);
2019-05-09 08:56:41 +00:00
let state = harness.get_current_state();
2019-03-26 07:29:02 +00:00
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slot = state.slot();
let committees = state
.get_beacon_committees_at_slot(slot)
.unwrap()
.into_iter()
.map(BeaconCommittee::into_owned)
.collect::<Vec<_>>();
2019-05-09 08:56:41 +00:00
let num_validators =
MainnetEthSpec::slots_per_epoch() as usize * spec.target_committee_size;
let attestations = harness.make_attestations(
(0..num_validators).collect::<Vec<_>>().as_slice(),
&state,
Hash256::zero(),
SignedBeaconBlockHash::from(Hash256::zero()),
slot,
);
for (_, aggregate) in attestations {
let att = aggregate.unwrap().message.aggregate;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let attesting_indices = get_attesting_indices_from_state(&state, &att).unwrap();
op_pool
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.insert_attestation(att.clone(), attesting_indices.clone())
.unwrap();
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attestation(att, attesting_indices).unwrap();
2019-03-26 07:29:02 +00:00
}
assert_eq!(op_pool.num_attestations(), committees.len());
}
/// Adding lots of attestations that only intersect pairwise should lead to two aggregate
/// attestations.
#[test]
fn attestation_pairwise_overlapping() {
let (harness, ref spec) = attestation_test_state::<MainnetEthSpec>(1);
2019-03-26 07:29:02 +00:00
let state = harness.get_current_state();
2019-03-26 07:29:02 +00:00
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slot = state.slot();
let committees = state
.get_beacon_committees_at_slot(slot)
.unwrap()
.into_iter()
.map(BeaconCommittee::into_owned)
.collect::<Vec<_>>();
2019-06-03 05:25:06 +00:00
let num_validators =
MainnetEthSpec::slots_per_epoch() as usize * spec.target_committee_size;
let attestations = harness.make_attestations(
(0..num_validators).collect::<Vec<_>>().as_slice(),
&state,
Hash256::zero(),
SignedBeaconBlockHash::from(Hash256::zero()),
slot,
);
let step_size = 2;
// Create attestations that overlap on `step_size` validators, like:
// {0,1,2,3}, {2,3,4,5}, {4,5,6,7}, ...
for (atts1, _) in attestations {
let atts2 = atts1.clone();
let aggs1 = atts1
.chunks_exact(step_size * 2)
.map(|chunk| {
let agg = chunk.into_iter().map(|(att, _)| att).fold::<Option<
Attestation<MainnetEthSpec>,
>, _>(
None,
|att, new_att| {
if let Some(mut a) = att {
a.aggregate(new_att);
Some(a)
} else {
Some(new_att.clone())
}
},
);
agg.unwrap()
})
.collect::<Vec<_>>();
let aggs2 = atts2
.into_iter()
.skip(step_size)
.collect::<Vec<_>>()
.as_slice()
.chunks_exact(step_size * 2)
.map(|chunk| {
let agg = chunk.into_iter().map(|(att, _)| att).fold::<Option<
Attestation<MainnetEthSpec>,
>, _>(
None,
|att, new_att| {
if let Some(mut a) = att {
a.aggregate(new_att);
Some(a)
} else {
Some(new_att.clone())
}
},
);
agg.unwrap()
})
.collect::<Vec<_>>();
for att in aggs1.into_iter().chain(aggs2.into_iter()) {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let attesting_indices = get_attesting_indices_from_state(&state, &att).unwrap();
op_pool.insert_attestation(att, attesting_indices).unwrap();
2019-03-26 07:29:02 +00:00
}
}
// The attestations should get aggregated into two attestations that comprise all
// validators.
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let stats = op_pool.attestation_stats();
assert_eq!(stats.num_attestation_data, committees.len());
assert_eq!(stats.num_attestations, 2 * committees.len());
assert_eq!(stats.max_aggregates_per_data, 2);
}
2019-03-26 07:29:02 +00:00
/// Create a bunch of attestations signed by a small number of validators, and another
/// bunch signed by a larger number, such that there are at least `max_attestations`
/// signed by the larger number. Then, check that `get_attestations` only returns the
/// high-quality attestations. To ensure that no aggregation occurs, ALL attestations
/// are also signed by the 0th member of the committee.
#[test]
fn attestation_get_max() {
let small_step_size = 2;
let big_step_size = 4;
let num_committees = big_step_size;
let (harness, ref spec) = attestation_test_state::<MainnetEthSpec>(num_committees);
2019-03-26 07:29:02 +00:00
let mut state = harness.get_current_state();
2019-05-09 08:56:41 +00:00
let op_pool = OperationPool::<MainnetEthSpec>::new();
2019-03-26 07:29:02 +00:00
let slot = state.slot();
let committees = state
.get_beacon_committees_at_slot(slot)
.unwrap()
.into_iter()
.map(BeaconCommittee::into_owned)
.collect::<Vec<_>>();
2019-03-26 07:29:02 +00:00
let max_attestations = <MainnetEthSpec as EthSpec>::MaxAttestations::to_usize();
let target_committee_size = spec.target_committee_size as usize;
let num_validators = num_committees
* MainnetEthSpec::slots_per_epoch() as usize
* spec.target_committee_size;
let attestations = harness.make_attestations(
(0..num_validators).collect::<Vec<_>>().as_slice(),
&state,
Hash256::zero(),
SignedBeaconBlockHash::from(Hash256::zero()),
slot,
);
let insert_attestations = |attestations: Vec<(Attestation<MainnetEthSpec>, SubnetId)>,
step_size| {
let att_0 = attestations.get(0).unwrap().0.clone();
let aggs = attestations
.chunks_exact(step_size)
.map(|chunk| {
chunk
.into_iter()
.map(|(att, _)| att)
.fold::<Attestation<MainnetEthSpec>, _>(
att_0.clone(),
|mut att, new_att| {
att.aggregate(new_att);
att
},
)
})
.collect::<Vec<_>>();
for att in aggs {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let attesting_indices = get_attesting_indices_from_state(&state, &att).unwrap();
op_pool.insert_attestation(att, attesting_indices).unwrap();
2019-05-09 08:56:41 +00:00
}
};
2019-03-26 07:29:02 +00:00
for (atts, _) in attestations {
assert_eq!(atts.len(), target_committee_size);
// Attestations signed by only 2-3 validators
insert_attestations(atts.clone(), small_step_size);
// Attestations signed by 4+ validators
insert_attestations(atts, big_step_size);
}
2019-03-26 07:29:02 +00:00
let num_small = target_committee_size / small_step_size;
let num_big = target_committee_size / big_step_size;
2019-03-26 07:29:02 +00:00
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let stats = op_pool.attestation_stats();
assert_eq!(stats.num_attestation_data, committees.len());
assert_eq!(
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
stats.num_attestations,
(num_small + num_big) * committees.len()
);
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
assert!(stats.num_attestations > max_attestations);
2019-05-09 08:56:41 +00:00
*state.slot_mut() += spec.min_attestation_inclusion_delay;
let best_attestations = op_pool
.get_attestations(&state, |_| true, |_| true, spec)
.expect("should have best attestations");
assert_eq!(best_attestations.len(), max_attestations);
// All the best attestations should be signed by at least `big_step_size` (4) validators.
for att in &best_attestations {
assert!(att.aggregation_bits.num_set_bits() >= big_step_size);
2019-03-26 07:29:02 +00:00
}
}
#[test]
fn attestation_rewards() {
let small_step_size = 2;
let big_step_size = 4;
let num_committees = big_step_size;
let (harness, ref spec) = attestation_test_state::<MainnetEthSpec>(num_committees);
let mut state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slot = state.slot();
let committees = state
.get_beacon_committees_at_slot(slot)
.unwrap()
.into_iter()
.map(BeaconCommittee::into_owned)
.collect::<Vec<_>>();
let max_attestations = <MainnetEthSpec as EthSpec>::MaxAttestations::to_usize();
let target_committee_size = spec.target_committee_size as usize;
// Each validator will have a multiple of 1_000_000_000 wei.
// Safe from overflow unless there are about 18B validators (2^64 / 1_000_000_000).
for i in 0..state.validators().len() {
state.validators_mut()[i].effective_balance = 1_000_000_000 * i as u64;
}
let num_validators = num_committees
* MainnetEthSpec::slots_per_epoch() as usize
* spec.target_committee_size;
let attestations = harness.make_attestations(
(0..num_validators).collect::<Vec<_>>().as_slice(),
&state,
Hash256::zero(),
SignedBeaconBlockHash::from(Hash256::zero()),
slot,
);
let insert_attestations = |attestations: Vec<(Attestation<MainnetEthSpec>, SubnetId)>,
step_size| {
let att_0 = attestations.get(0).unwrap().0.clone();
let aggs = attestations
.chunks_exact(step_size)
.map(|chunk| {
chunk
.into_iter()
.map(|(att, _)| att)
.fold::<Attestation<MainnetEthSpec>, _>(
att_0.clone(),
|mut att, new_att| {
att.aggregate(new_att);
att
},
)
})
.collect::<Vec<_>>();
for att in aggs {
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let attesting_indices = get_attesting_indices_from_state(&state, &att).unwrap();
op_pool.insert_attestation(att, attesting_indices).unwrap();
}
};
for (atts, _) in attestations {
assert_eq!(atts.len(), target_committee_size);
// Attestations signed by only 2-3 validators
insert_attestations(atts.clone(), small_step_size);
// Attestations signed by 4+ validators
insert_attestations(atts, big_step_size);
}
let num_small = target_committee_size / small_step_size;
let num_big = target_committee_size / big_step_size;
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
assert_eq!(
op_pool.attestation_stats().num_attestation_data,
committees.len()
);
assert_eq!(
op_pool.num_attestations(),
(num_small + num_big) * committees.len()
);
assert!(op_pool.num_attestations() > max_attestations);
*state.slot_mut() += spec.min_attestation_inclusion_delay;
let best_attestations = op_pool
.get_attestations(&state, |_| true, |_| true, spec)
.expect("should have valid best attestations");
assert_eq!(best_attestations.len(), max_attestations);
let total_active_balance = state.get_total_active_balance().unwrap();
// Set of indices covered by previous attestations in `best_attestations`.
let mut seen_indices = BTreeSet::<u64>::new();
// Used for asserting that rewards are in decreasing order.
let mut prev_reward = u64::max_value();
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
let mut reward_cache = RewardCache::default();
reward_cache.update(&state).unwrap();
for att in best_attestations {
let attesting_indices = get_attesting_indices_from_state(&state, &att).unwrap();
let split_attestation = SplitAttestation::new(att, attesting_indices);
let mut fresh_validators_rewards = AttMaxCover::new(
split_attestation.as_ref(),
&state,
&reward_cache,
total_active_balance,
spec,
)
.unwrap()
.fresh_validators_rewards;
// Remove validators covered by previous attestations.
fresh_validators_rewards
.retain(|validator_index, _| !seen_indices.contains(validator_index));
// Check that rewards are in decreasing order
let rewards = fresh_validators_rewards.values().sum();
assert!(prev_reward >= rewards);
prev_reward = rewards;
seen_indices.extend(fresh_validators_rewards.keys());
}
}
/// Insert two slashings for the same proposer and ensure only one is returned.
#[test]
fn duplicate_proposer_slashing() {
let harness = get_harness(32, None);
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let proposer_index = 0;
let slashing1 = harness.make_proposer_slashing(proposer_index);
let slashing2 = ProposerSlashing {
signed_header_1: slashing1.signed_header_2.clone(),
signed_header_2: slashing1.signed_header_1.clone(),
};
// Both slashings should be valid and accepted by the pool.
op_pool
.insert_proposer_slashing(slashing1.clone().validate(&state, &harness.spec).unwrap());
op_pool
.insert_proposer_slashing(slashing2.clone().validate(&state, &harness.spec).unwrap());
// Should only get the second slashing back.
assert_eq!(
op_pool.get_slashings_and_exits(&state, &harness.spec).0,
vec![slashing2]
);
}
// Sanity check on the pruning of proposer slashings
#[test]
fn prune_proposer_slashing_noop() {
let harness = get_harness(32, None);
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slashing = harness.make_proposer_slashing(0);
op_pool.insert_proposer_slashing(slashing.clone().validate(&state, &harness.spec).unwrap());
op_pool.prune_proposer_slashings(&state);
assert_eq!(
op_pool.get_slashings_and_exits(&state, &harness.spec).0,
vec![slashing]
);
}
// Sanity check on the pruning of attester slashings
#[test]
fn prune_attester_slashing_noop() {
let harness = get_harness(32, None);
let spec = &harness.spec;
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slashing = harness.make_attester_slashing(vec![1, 3, 5, 7, 9]);
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attester_slashing(slashing.clone().validate(&state, spec).unwrap());
op_pool.prune_attester_slashings(&state);
assert_eq!(
op_pool.get_slashings_and_exits(&state, &harness.spec).1,
vec![slashing]
);
}
// Check that we get maximum coverage for attester slashings (highest qty of validators slashed)
#[test]
fn simple_max_cover_attester_slashing() {
let harness = get_harness(32, None);
let spec = &harness.spec;
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slashing_1 = harness.make_attester_slashing(vec![1]);
let slashing_2 = harness.make_attester_slashing(vec![2, 3]);
let slashing_3 = harness.make_attester_slashing(vec![4, 5, 6]);
let slashing_4 = harness.make_attester_slashing(vec![7, 8, 9, 10]);
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attester_slashing(slashing_1.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_2.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_3.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_4.clone().validate(&state, spec).unwrap());
let best_slashings = op_pool.get_slashings_and_exits(&state, &harness.spec);
assert_eq!(best_slashings.1, vec![slashing_4, slashing_3]);
}
// Check that we get maximum coverage for attester slashings with overlapping indices
#[test]
fn overlapping_max_cover_attester_slashing() {
let harness = get_harness(32, None);
let spec = &harness.spec;
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slashing_1 = harness.make_attester_slashing(vec![1, 2, 3, 4]);
let slashing_2 = harness.make_attester_slashing(vec![1, 2, 5]);
let slashing_3 = harness.make_attester_slashing(vec![5, 6]);
let slashing_4 = harness.make_attester_slashing(vec![6]);
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attester_slashing(slashing_1.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_2.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_3.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_4.clone().validate(&state, spec).unwrap());
let best_slashings = op_pool.get_slashings_and_exits(&state, &harness.spec);
assert_eq!(best_slashings.1, vec![slashing_1, slashing_3]);
}
// Max coverage of attester slashings taking into account proposer slashings
#[test]
fn max_coverage_attester_proposer_slashings() {
let harness = get_harness(32, None);
let spec = &harness.spec;
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let p_slashing = harness.make_proposer_slashing(1);
let a_slashing_1 = harness.make_attester_slashing(vec![1, 2, 3, 4]);
let a_slashing_2 = harness.make_attester_slashing(vec![1, 3, 4]);
let a_slashing_3 = harness.make_attester_slashing(vec![5, 6]);
op_pool.insert_proposer_slashing(p_slashing.clone().validate(&state, spec).unwrap());
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attester_slashing(a_slashing_1.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(a_slashing_2.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(a_slashing_3.clone().validate(&state, spec).unwrap());
let best_slashings = op_pool.get_slashings_and_exits(&state, &harness.spec);
assert_eq!(best_slashings.1, vec![a_slashing_1, a_slashing_3]);
}
//Max coverage checking that non overlapping indices are still recognized for their value
#[test]
fn max_coverage_different_indices_set() {
let harness = get_harness(32, None);
let spec = &harness.spec;
let state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
let slashing_1 = harness.make_attester_slashing_different_indices(
vec![1, 2, 3, 4, 5, 6],
vec![3, 4, 5, 6, 7, 8],
);
let slashing_2 = harness.make_attester_slashing(vec![5, 6]);
let slashing_3 = harness.make_attester_slashing(vec![1, 2, 3]);
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attester_slashing(slashing_1.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_2.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_3.clone().validate(&state, spec).unwrap());
let best_slashings = op_pool.get_slashings_and_exits(&state, &harness.spec);
assert_eq!(best_slashings.1, vec![slashing_1, slashing_3]);
}
// Max coverage should be affected by the overall effective balances
#[test]
fn max_coverage_effective_balances() {
let harness = get_harness(32, None);
let spec = &harness.spec;
let mut state = harness.get_current_state();
let op_pool = OperationPool::<MainnetEthSpec>::new();
state.validators_mut()[1].effective_balance = 17_000_000_000;
state.validators_mut()[2].effective_balance = 17_000_000_000;
state.validators_mut()[3].effective_balance = 17_000_000_000;
let slashing_1 = harness.make_attester_slashing(vec![1, 2, 3]);
let slashing_2 = harness.make_attester_slashing(vec![4, 5, 6]);
let slashing_3 = harness.make_attester_slashing(vec![7, 8]);
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
op_pool.insert_attester_slashing(slashing_1.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_2.clone().validate(&state, spec).unwrap());
op_pool.insert_attester_slashing(slashing_3.clone().validate(&state, spec).unwrap());
let best_slashings = op_pool.get_slashings_and_exits(&state, &harness.spec);
assert_eq!(best_slashings.1, vec![slashing_2, slashing_3]);
}
/// End-to-end test of basic sync contribution handling.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn sync_contribution_aggregation_insert_get_prune() {
let (harness, _) = sync_contribution_test_state::<MainnetEthSpec>(1).await;
let op_pool = OperationPool::<MainnetEthSpec>::new();
let state = harness.get_current_state();
let block_root = *state
.get_block_root(state.slot() - Slot::new(1))
.ok()
.expect("block root should exist at slot");
let contributions = harness.make_sync_contributions(
&state,
block_root,
state.slot() - Slot::new(1),
RelativeSyncCommittee::Current,
);
for (_, contribution_and_proof) in contributions {
let contribution = contribution_and_proof
.expect("contribution exists for committee")
.message
.contribution;
op_pool.insert_sync_contribution(contribution).unwrap();
}
assert_eq!(op_pool.sync_contributions.read().len(), 1);
assert_eq!(
op_pool.num_sync_contributions(),
SYNC_COMMITTEE_SUBNET_COUNT as usize
);
let sync_aggregate = op_pool
.get_sync_aggregate(&state)
.expect("Should calculate the sync aggregate")
.expect("Should have block sync aggregate");
assert_eq!(
sync_aggregate.sync_committee_bits.num_set_bits(),
MainnetEthSpec::sync_committee_size()
);
// Prune sync contributions shouldn't do anything at this point.
op_pool.prune_sync_contributions(state.slot() - Slot::new(1));
assert_eq!(
op_pool.num_sync_contributions(),
SYNC_COMMITTEE_SUBNET_COUNT as usize
);
op_pool.prune_sync_contributions(state.slot());
assert_eq!(
op_pool.num_sync_contributions(),
SYNC_COMMITTEE_SUBNET_COUNT as usize
);
// But once we advance to more than one slot after the contribution, it should prune it
// out of existence.
op_pool.prune_sync_contributions(state.slot() + Slot::new(1));
assert_eq!(op_pool.num_sync_contributions(), 0);
}
/// Adding a sync contribution already in the pool should not increase the size of the pool.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn sync_contribution_duplicate() {
let (harness, _) = sync_contribution_test_state::<MainnetEthSpec>(1).await;
let op_pool = OperationPool::<MainnetEthSpec>::new();
let state = harness.get_current_state();
let block_root = *state
.get_block_root(state.slot() - Slot::new(1))
.ok()
.expect("block root should exist at slot");
let contributions = harness.make_sync_contributions(
&state,
block_root,
state.slot() - Slot::new(1),
RelativeSyncCommittee::Current,
);
for (_, contribution_and_proof) in contributions {
let contribution = contribution_and_proof
.expect("contribution exists for committee")
.message
.contribution;
op_pool
.insert_sync_contribution(contribution.clone())
.unwrap();
op_pool.insert_sync_contribution(contribution).unwrap();
}
assert_eq!(op_pool.sync_contributions.read().len(), 1);
assert_eq!(
op_pool.num_sync_contributions(),
SYNC_COMMITTEE_SUBNET_COUNT as usize
);
}
/// Adding a sync contribution already in the pool with more bits set should increase the
/// number of bits set in the aggregate.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn sync_contribution_with_more_bits() {
let (harness, _) = sync_contribution_test_state::<MainnetEthSpec>(1).await;
let op_pool = OperationPool::<MainnetEthSpec>::new();
let state = harness.get_current_state();
let block_root = *state
.get_block_root(state.slot() - Slot::new(1))
.ok()
.expect("block root should exist at slot");
let contributions = harness.make_sync_contributions(
&state,
block_root,
state.slot() - Slot::new(1),
RelativeSyncCommittee::Current,
);
let expected_bits = MainnetEthSpec::sync_committee_size() - (2 * contributions.len());
let mut first_contribution = contributions[0]
.1
.as_ref()
.unwrap()
.message
.contribution
.clone();
// Add all contributions, but unset the first two bits of each.
for (_, contribution_and_proof) in contributions {
let mut contribution_fewer_bits = contribution_and_proof
.expect("contribution exists for committee")
.message
.contribution;
// Unset the first two bits of each contribution.
contribution_fewer_bits
.aggregation_bits
.set(0, false)
.expect("set bit");
contribution_fewer_bits
.aggregation_bits
.set(1, false)
.expect("set bit");
op_pool
.insert_sync_contribution(contribution_fewer_bits)
.unwrap();
}
let sync_aggregate = op_pool
.get_sync_aggregate(&state)
.expect("Should calculate the sync aggregate")
.expect("Should have block sync aggregate");
assert_eq!(
sync_aggregate.sync_committee_bits.num_set_bits(),
expected_bits
);
// Unset the first bit of the first contribution and re-insert it. This should increase the
// number of bits set in the sync aggregate by one.
first_contribution
.aggregation_bits
.set(0, false)
.expect("set bit");
op_pool
.insert_sync_contribution(first_contribution)
.unwrap();
// The sync aggregate should now include the additional set bit.
let sync_aggregate = op_pool
.get_sync_aggregate(&state)
.expect("Should calculate the sync aggregate")
.expect("Should have block sync aggregate");
assert_eq!(
sync_aggregate.sync_committee_bits.num_set_bits(),
expected_bits + 1
);
}
/// Adding a sync contribution already in the pool with fewer bits set should not increase the
/// number of bits set in the aggregate.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn sync_contribution_with_fewer_bits() {
let (harness, _) = sync_contribution_test_state::<MainnetEthSpec>(1).await;
let op_pool = OperationPool::<MainnetEthSpec>::new();
let state = harness.get_current_state();
let block_root = *state
.get_block_root(state.slot() - Slot::new(1))
.ok()
.expect("block root should exist at slot");
let contributions = harness.make_sync_contributions(
&state,
block_root,
state.slot() - Slot::new(1),
RelativeSyncCommittee::Current,
);
let expected_bits = MainnetEthSpec::sync_committee_size() - (2 * contributions.len());
let mut first_contribution = contributions[0]
.1
.as_ref()
.unwrap()
.message
.contribution
.clone();
// Add all contributions, but unset the first two bits of each.
for (_, contribution_and_proof) in contributions {
let mut contribution_fewer_bits = contribution_and_proof
.expect("contribution exists for committee")
.message
.contribution;
// Unset the first two bits of each contribution.
contribution_fewer_bits
.aggregation_bits
.set(0, false)
.expect("set bit");
contribution_fewer_bits
.aggregation_bits
.set(1, false)
.expect("set bit");
op_pool
.insert_sync_contribution(contribution_fewer_bits)
.unwrap();
}
let sync_aggregate = op_pool
.get_sync_aggregate(&state)
.expect("Should calculate the sync aggregate")
.expect("Should have block sync aggregate");
assert_eq!(
sync_aggregate.sync_committee_bits.num_set_bits(),
expected_bits
);
// Unset the first three bits of the first contribution and re-insert it. This should
// not affect the number of bits set in the sync aggregate.
first_contribution
.aggregation_bits
.set(0, false)
.expect("set bit");
first_contribution
.aggregation_bits
.set(1, false)
.expect("set bit");
first_contribution
.aggregation_bits
.set(2, false)
.expect("set bit");
op_pool
.insert_sync_contribution(first_contribution)
.unwrap();
// The sync aggregate should still have the same number of set bits.
let sync_aggregate = op_pool
.get_sync_aggregate(&state)
.expect("Should calculate the sync aggregate")
.expect("Should have block sync aggregate");
assert_eq!(
sync_aggregate.sync_committee_bits.num_set_bits(),
expected_bits
);
}
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
fn cross_fork_harness<E: EthSpec>() -> (BeaconChainHarness<EphemeralHarnessType<E>>, ChainSpec)
{
let mut spec = test_spec::<E>();
// Give some room to sign surround slashings.
spec.altair_fork_epoch = Some(Epoch::new(3));
spec.bellatrix_fork_epoch = Some(Epoch::new(6));
// To make exits immediately valid.
spec.shard_committee_period = 0;
let num_validators = 32;
let harness = get_harness::<E>(num_validators, Some(spec.clone()));
(harness, spec)
}
/// Test several cross-fork voluntary exits:
///
/// - phase0 exit (not valid after Bellatrix)
/// - phase0 exit signed with Altair fork version (only valid after Bellatrix)
#[tokio::test]
async fn cross_fork_exits() {
let (harness, spec) = cross_fork_harness::<MainnetEthSpec>();
let altair_fork_epoch = spec.altair_fork_epoch.unwrap();
let bellatrix_fork_epoch = spec.bellatrix_fork_epoch.unwrap();
let slots_per_epoch = MainnetEthSpec::slots_per_epoch();
let op_pool = OperationPool::<MainnetEthSpec>::new();
// Sign an exit in phase0 with a phase0 epoch.
let exit1 = harness.make_voluntary_exit(0, Epoch::new(0));
// Advance to Altair.
harness
.extend_to_slot(altair_fork_epoch.start_slot(slots_per_epoch))
.await;
let altair_head = harness.chain.canonical_head.cached_head().snapshot;
assert_eq!(altair_head.beacon_state.current_epoch(), altair_fork_epoch);
// Add exit 1 to the op pool during Altair. It's still valid at this point and should be
// returned.
let verified_exit1 = exit1
.clone()
.validate(&altair_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_voluntary_exit(verified_exit1);
let exits =
op_pool.get_voluntary_exits(&altair_head.beacon_state, |_| true, &harness.chain.spec);
assert!(exits.contains(&exit1));
assert_eq!(exits.len(), 1);
// Advance to Bellatrix.
harness
.extend_to_slot(bellatrix_fork_epoch.start_slot(slots_per_epoch))
.await;
let bellatrix_head = harness.chain.canonical_head.cached_head().snapshot;
assert_eq!(
bellatrix_head.beacon_state.current_epoch(),
bellatrix_fork_epoch
);
// Sign an exit with the Altair domain and a phase0 epoch. This is a weird type of exit
// that is valid because after the Bellatrix fork we'll use the Altair fork domain to verify
// all prior epochs.
let exit2 = harness.make_voluntary_exit(2, Epoch::new(0));
let verified_exit2 = exit2
.clone()
.validate(&bellatrix_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_voluntary_exit(verified_exit2);
// Attempting to fetch exit1 now should fail, despite it still being in the pool.
// exit2 should still be valid, because it was signed with the Altair fork domain.
assert_eq!(op_pool.voluntary_exits.read().len(), 2);
let exits =
op_pool.get_voluntary_exits(&bellatrix_head.beacon_state, |_| true, &harness.spec);
assert_eq!(&exits, &[exit2]);
}
/// Test several cross-fork proposer slashings:
///
/// - phase0 slashing (not valid after Bellatrix)
/// - Bellatrix signed with Altair fork version (not valid after Bellatrix)
/// - phase0 exit signed with Altair fork version (only valid after Bellatrix)
#[tokio::test]
async fn cross_fork_proposer_slashings() {
let (harness, spec) = cross_fork_harness::<MainnetEthSpec>();
let slots_per_epoch = MainnetEthSpec::slots_per_epoch();
let altair_fork_epoch = spec.altair_fork_epoch.unwrap();
let bellatrix_fork_epoch = spec.bellatrix_fork_epoch.unwrap();
let bellatrix_fork_slot = bellatrix_fork_epoch.start_slot(slots_per_epoch);
let op_pool = OperationPool::<MainnetEthSpec>::new();
// Sign a proposer slashing in phase0 with a phase0 epoch.
let slashing1 = harness.make_proposer_slashing_at_slot(0, Some(Slot::new(1)));
// Advance to Altair.
harness
.extend_to_slot(altair_fork_epoch.start_slot(slots_per_epoch))
.await;
let altair_head = harness.chain.canonical_head.cached_head().snapshot;
assert_eq!(altair_head.beacon_state.current_epoch(), altair_fork_epoch);
// Add slashing1 to the op pool during Altair. It's still valid at this point and should be
// returned.
let verified_slashing1 = slashing1
.clone()
.validate(&altair_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_proposer_slashing(verified_slashing1);
let (proposer_slashings, _, _) =
op_pool.get_slashings_and_exits(&altair_head.beacon_state, &harness.chain.spec);
assert!(proposer_slashings.contains(&slashing1));
assert_eq!(proposer_slashings.len(), 1);
// Sign a proposer slashing with a Bellatrix slot using the Altair fork domain.
//
// This slashing is valid only before the Bellatrix fork epoch.
let slashing2 = harness.make_proposer_slashing_at_slot(1, Some(bellatrix_fork_slot));
let verified_slashing2 = slashing2
.clone()
.validate(&altair_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_proposer_slashing(verified_slashing2);
let (proposer_slashings, _, _) =
op_pool.get_slashings_and_exits(&altair_head.beacon_state, &harness.chain.spec);
assert!(proposer_slashings.contains(&slashing1));
assert!(proposer_slashings.contains(&slashing2));
assert_eq!(proposer_slashings.len(), 2);
// Advance to Bellatrix.
harness.extend_to_slot(bellatrix_fork_slot).await;
let bellatrix_head = harness.chain.canonical_head.cached_head().snapshot;
assert_eq!(
bellatrix_head.beacon_state.current_epoch(),
bellatrix_fork_epoch
);
// Sign a proposer slashing with the Altair domain and a phase0 slot. This is a weird type
// of slashing that is only valid after the Bellatrix fork because we'll use the Altair fork
// domain to verify all prior epochs.
let slashing3 = harness.make_proposer_slashing_at_slot(2, Some(Slot::new(1)));
let verified_slashing3 = slashing3
.clone()
.validate(&bellatrix_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_proposer_slashing(verified_slashing3);
// Attempting to fetch slashing1 now should fail, despite it still being in the pool.
// Likewise slashing2 is also invalid now because it should be signed with the
// Bellatrix fork version.
// slashing3 should still be valid, because it was signed with the Altair fork domain.
assert_eq!(op_pool.proposer_slashings.read().len(), 3);
let (proposer_slashings, _, _) =
op_pool.get_slashings_and_exits(&bellatrix_head.beacon_state, &harness.spec);
assert!(proposer_slashings.contains(&slashing3));
assert_eq!(proposer_slashings.len(), 1);
}
/// Test several cross-fork attester slashings:
///
/// - both target epochs in phase0 (not valid after Bellatrix)
/// - both target epochs in Bellatrix but signed with Altair domain (not valid after Bellatrix)
/// - Altair attestation that surrounds a phase0 attestation (not valid after Bellatrix)
/// - both target epochs in phase0 but signed with Altair domain (only valid after Bellatrix)
#[tokio::test]
async fn cross_fork_attester_slashings() {
let (harness, spec) = cross_fork_harness::<MainnetEthSpec>();
let slots_per_epoch = MainnetEthSpec::slots_per_epoch();
let zero_epoch = Epoch::new(0);
let altair_fork_epoch = spec.altair_fork_epoch.unwrap();
let bellatrix_fork_epoch = spec.bellatrix_fork_epoch.unwrap();
let bellatrix_fork_slot = bellatrix_fork_epoch.start_slot(slots_per_epoch);
let op_pool = OperationPool::<MainnetEthSpec>::new();
// Sign an attester slashing with the phase0 fork version, with both target epochs in phase0.
let slashing1 = harness.make_attester_slashing_with_epochs(
vec![0],
None,
Some(zero_epoch),
None,
Some(zero_epoch),
);
// Advance to Altair.
harness
.extend_to_slot(altair_fork_epoch.start_slot(slots_per_epoch))
.await;
let altair_head = harness.chain.canonical_head.cached_head().snapshot;
assert_eq!(altair_head.beacon_state.current_epoch(), altair_fork_epoch);
// Add slashing1 to the op pool during Altair. It's still valid at this point and should be
// returned.
let verified_slashing1 = slashing1
.clone()
.validate(&altair_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_attester_slashing(verified_slashing1);
// Sign an attester slashing with two Bellatrix epochs using the Altair fork domain.
//
// This slashing is valid only before the Bellatrix fork epoch.
let slashing2 = harness.make_attester_slashing_with_epochs(
vec![1],
None,
Some(bellatrix_fork_epoch),
None,
Some(bellatrix_fork_epoch),
);
let verified_slashing2 = slashing2
.clone()
.validate(&altair_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_attester_slashing(verified_slashing2);
let (_, attester_slashings, _) =
op_pool.get_slashings_and_exits(&altair_head.beacon_state, &harness.chain.spec);
assert!(attester_slashings.contains(&slashing1));
assert!(attester_slashings.contains(&slashing2));
assert_eq!(attester_slashings.len(), 2);
// Sign an attester slashing where an Altair attestation surrounds a phase0 one.
//
// This slashing is valid only before the Bellatrix fork epoch.
let slashing3 = harness.make_attester_slashing_with_epochs(
vec![2],
Some(Epoch::new(0)),
Some(altair_fork_epoch),
Some(Epoch::new(1)),
Some(altair_fork_epoch - 1),
);
let verified_slashing3 = slashing3
.clone()
.validate(&altair_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_attester_slashing(verified_slashing3);
// All three slashings should be valid and returned from the pool at this point.
// Seeing as we can only extract 2 at time we'll just pretend that validator 0 is already
// slashed.
let mut to_be_slashed = hashset! {0};
let attester_slashings =
op_pool.get_attester_slashings(&altair_head.beacon_state, &mut to_be_slashed);
assert!(attester_slashings.contains(&slashing2));
assert!(attester_slashings.contains(&slashing3));
assert_eq!(attester_slashings.len(), 2);
// Advance to Bellatrix.
harness.extend_to_slot(bellatrix_fork_slot).await;
let bellatrix_head = harness.chain.canonical_head.cached_head().snapshot;
assert_eq!(
bellatrix_head.beacon_state.current_epoch(),
bellatrix_fork_epoch
);
// Sign an attester slashing with the Altair domain and phase0 epochs. This is a weird type
// of slashing that is only valid after the Bellatrix fork because we'll use the Altair fork
// domain to verify all prior epochs.
let slashing4 = harness.make_attester_slashing_with_epochs(
vec![3],
Some(Epoch::new(0)),
Some(altair_fork_epoch - 1),
Some(Epoch::new(0)),
Some(altair_fork_epoch - 1),
);
let verified_slashing4 = slashing4
.clone()
.validate(&bellatrix_head.beacon_state, &harness.chain.spec)
.unwrap();
op_pool.insert_attester_slashing(verified_slashing4);
// All slashings except slashing4 are now invalid (despite being present in the pool).
assert_eq!(op_pool.attester_slashings.read().len(), 4);
let (_, attester_slashings, _) =
op_pool.get_slashings_and_exits(&bellatrix_head.beacon_state, &harness.spec);
assert!(attester_slashings.contains(&slashing4));
assert_eq!(attester_slashings.len(), 1);
// Pruning the attester slashings should remove all but slashing4.
op_pool.prune_attester_slashings(&bellatrix_head.beacon_state);
assert_eq!(op_pool.attester_slashings.read().len(), 1);
}
2019-03-06 03:46:12 +00:00
}