mirror of
https://gitlab.com/pulsechaincom/prysm-pulse.git
synced 2024-12-25 21:07:18 +00:00
0326be86b5
* starting on patch * finish determining all required patches * properly redefine the patch rules * new patch * rem double semicolon * fix patch file * Merge branch 'master' of github.com:prysmaticlabs/prysm into deprecate-eth-protos * building the deps * test target passes using ethereumapis * compile gateway * attempting to build everything * e2e use ethereumapis * more fixes for slasher * other item * getting closer to compiling slasher * build slasher package * Merge branch 'master' into deprecate-eth-protos * Merge branch 'master' into deprecate-eth-protos * fix benches * lint gazelle * Merge branch 'deprecate-eth-protos' of github.com:prysmaticlabs/prysm into deprecate-eth-protos * proper gateway * lint * Merge branch 'master' into deprecate-eth-protos * fix build * Merge branch 'deprecate-eth-protos' of github.com:prysmaticlabs/prysm into deprecate-eth-protos * use swag * resolve * ignore change * include new patch changes * fix test * builds * fix e2e * gaz
709 lines
31 KiB
Go
709 lines
31 KiB
Go
// Package epoch contains epoch processing libraries. These libraries
|
|
// process new balance for the validators, justify and finalize new
|
|
// check points, shuffle and reassign validators to different slots and
|
|
// shards.
|
|
package epoch
|
|
|
|
import (
|
|
"bytes"
|
|
"fmt"
|
|
"sort"
|
|
|
|
"github.com/pkg/errors"
|
|
ethpb "github.com/prysmaticlabs/ethereumapis/eth/v1alpha1"
|
|
"github.com/prysmaticlabs/go-ssz"
|
|
"github.com/prysmaticlabs/prysm/beacon-chain/core/helpers"
|
|
"github.com/prysmaticlabs/prysm/beacon-chain/core/validators"
|
|
pb "github.com/prysmaticlabs/prysm/proto/beacon/p2p/v1"
|
|
"github.com/prysmaticlabs/prysm/shared/mathutil"
|
|
"github.com/prysmaticlabs/prysm/shared/params"
|
|
)
|
|
|
|
// MatchedAttestations is an object that contains the correctly
|
|
// voted attestations based on source, target and head criteria.
|
|
type MatchedAttestations struct {
|
|
source []*pb.PendingAttestation
|
|
Target []*pb.PendingAttestation
|
|
head []*pb.PendingAttestation
|
|
}
|
|
|
|
// MatchAttestations matches the attestations gathered in a span of an epoch
|
|
// and categorize them whether they correctly voted for source, target and head.
|
|
// We combined the individual helpers from spec for efficiency and to achieve O(N) run time.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def get_matching_source_attestations(state: BeaconState, epoch: Epoch) -> List[PendingAttestation]:
|
|
// assert epoch in (get_current_epoch(state), get_previous_epoch(state))
|
|
// return state.current_epoch_attestations if epoch == get_current_epoch(state) else state.previous_epoch_attestations
|
|
//
|
|
// def get_matching_target_attestations(state: BeaconState, epoch: Epoch) -> List[PendingAttestation]:
|
|
// return [
|
|
// a for a in get_matching_source_attestations(state, epoch)
|
|
// if a.data.target_root == get_block_root(state, epoch)
|
|
// ]
|
|
//
|
|
// def get_matching_head_attestations(state: BeaconState, epoch: Epoch) -> List[PendingAttestation]:
|
|
// return [
|
|
// a for a in get_matching_source_attestations(state, epoch)
|
|
// if a.data.beacon_block_root == get_block_root_at_slot(state, get_attestation_data_slot(state, a.data))
|
|
// ]
|
|
func MatchAttestations(state *pb.BeaconState, epoch uint64) (*MatchedAttestations, error) {
|
|
currentEpoch := helpers.CurrentEpoch(state)
|
|
previousEpoch := helpers.PrevEpoch(state)
|
|
|
|
// Input epoch for matching the source attestations has to be within range
|
|
// of current epoch & previous epoch.
|
|
if epoch != currentEpoch && epoch != previousEpoch {
|
|
return nil, fmt.Errorf("input epoch: %d != current epoch: %d or previous epoch: %d",
|
|
epoch, currentEpoch, previousEpoch)
|
|
}
|
|
|
|
// Decide if the source attestations are coming from current or previous epoch.
|
|
var srcAtts []*pb.PendingAttestation
|
|
if epoch == currentEpoch {
|
|
srcAtts = state.CurrentEpochAttestations
|
|
} else {
|
|
srcAtts = state.PreviousEpochAttestations
|
|
}
|
|
targetRoot, err := helpers.BlockRoot(state, epoch)
|
|
if err != nil {
|
|
return nil, errors.Wrapf(err, "could not get block root for epoch %d", epoch)
|
|
}
|
|
|
|
tgtAtts := make([]*pb.PendingAttestation, 0, len(srcAtts))
|
|
headAtts := make([]*pb.PendingAttestation, 0, len(srcAtts))
|
|
for _, srcAtt := range srcAtts {
|
|
// If the target root matches attestation's target root,
|
|
// then we know this attestation has correctly voted for target.
|
|
if bytes.Equal(srcAtt.Data.Target.Root, targetRoot) {
|
|
tgtAtts = append(tgtAtts, srcAtt)
|
|
}
|
|
headRoot, err := helpers.BlockRootAtSlot(state, srcAtt.Data.Slot)
|
|
if err != nil {
|
|
return nil, errors.Wrapf(err, "could not get block root for slot %d", srcAtt.Data.Slot)
|
|
}
|
|
if bytes.Equal(srcAtt.Data.BeaconBlockRoot, headRoot) {
|
|
headAtts = append(headAtts, srcAtt)
|
|
}
|
|
}
|
|
|
|
return &MatchedAttestations{
|
|
source: srcAtts,
|
|
Target: tgtAtts,
|
|
head: headAtts,
|
|
}, nil
|
|
}
|
|
|
|
// AttestingBalance returns the total balance from all the attesting indices.
|
|
//
|
|
// WARNING: This method allocates a new copy of the attesting validator indices set and is
|
|
// considered to be very memory expensive. Avoid using this unless you really
|
|
// need to get attesting balance from attestations.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def get_attesting_balance(state: BeaconState, attestations: List[PendingAttestation]) -> Gwei:
|
|
// return get_total_balance(state, get_unslashed_attesting_indices(state, attestations))
|
|
func AttestingBalance(state *pb.BeaconState, atts []*pb.PendingAttestation) (uint64, error) {
|
|
indices, err := unslashedAttestingIndices(state, atts)
|
|
if err != nil {
|
|
return 0, errors.Wrap(err, "could not get attesting indices")
|
|
}
|
|
return helpers.TotalBalance(state, indices), nil
|
|
}
|
|
|
|
// ProcessJustificationAndFinalization processes justification and finalization during
|
|
// epoch processing. This is where a beacon node can justify and finalize a new epoch.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def process_justification_and_finalization(state: BeaconState) -> None:
|
|
// if get_current_epoch(state) <= GENESIS_EPOCH + 1:
|
|
// return
|
|
//
|
|
// previous_epoch = get_previous_epoch(state)
|
|
// current_epoch = get_current_epoch(state)
|
|
// old_previous_justified_checkpoint = state.previous_justified_checkpoint
|
|
// old_current_justified_checkpoint = state.current_justified_checkpoint
|
|
//
|
|
// # Process justifications
|
|
// state.previous_justified_checkpoint = state.current_justified_checkpoint
|
|
// state.justification_bits[1:] = state.justification_bits[:-1]
|
|
// state.justification_bits[0] = 0b0
|
|
// matching_target_attestations = get_matching_target_attestations(state, previous_epoch) # Previous epoch
|
|
// if get_attesting_balance(state, matching_target_attestations) * 3 >= get_total_active_balance(state) * 2:
|
|
// state.current_justified_checkpoint = Checkpoint(epoch=previous_epoch,
|
|
// root=get_block_root(state, previous_epoch))
|
|
// state.justification_bits[1] = 0b1
|
|
// matching_target_attestations = get_matching_target_attestations(state, current_epoch) # Current epoch
|
|
// if get_attesting_balance(state, matching_target_attestations) * 3 >= get_total_active_balance(state) * 2:
|
|
// state.current_justified_checkpoint = Checkpoint(epoch=current_epoch,
|
|
// root=get_block_root(state, current_epoch))
|
|
// state.justification_bits[0] = 0b1
|
|
//
|
|
// # Process finalizations
|
|
// bits = state.justification_bits
|
|
// # The 2nd/3rd/4th most recent epochs are justified, the 2nd using the 4th as source
|
|
// if all(bits[1:4]) and old_previous_justified_checkpoint.epoch + 3 == current_epoch:
|
|
// state.finalized_checkpoint = old_previous_justified_checkpoint
|
|
// # The 2nd/3rd most recent epochs are justified, the 2nd using the 3rd as source
|
|
// if all(bits[1:3]) and old_previous_justified_checkpoint.epoch + 2 == current_epoch:
|
|
// state.finalized_checkpoint = old_previous_justified_checkpoint
|
|
// # The 1st/2nd/3rd most recent epochs are justified, the 1st using the 3rd as source
|
|
// if all(bits[0:3]) and old_current_justified_checkpoint.epoch + 2 == current_epoch:
|
|
// state.finalized_checkpoint = old_current_justified_checkpoint
|
|
// # The 1st/2nd most recent epochs are justified, the 1st using the 2nd as source
|
|
// if all(bits[0:2]) and old_current_justified_checkpoint.epoch + 1 == current_epoch:
|
|
// state.finalized_checkpoint = old_current_justified_checkpoint
|
|
func ProcessJustificationAndFinalization(state *pb.BeaconState, prevAttestedBal uint64, currAttestedBal uint64) (*pb.BeaconState, error) {
|
|
if state.Slot <= helpers.StartSlot(2) {
|
|
return state, nil
|
|
}
|
|
|
|
prevEpoch := helpers.PrevEpoch(state)
|
|
currentEpoch := helpers.CurrentEpoch(state)
|
|
oldPrevJustifiedCheckpoint := state.PreviousJustifiedCheckpoint
|
|
oldCurrJustifiedCheckpoint := state.CurrentJustifiedCheckpoint
|
|
|
|
totalBal, err := helpers.TotalActiveBalance(state)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not get total balance")
|
|
}
|
|
|
|
// Process justifications
|
|
state.PreviousJustifiedCheckpoint = state.CurrentJustifiedCheckpoint
|
|
state.JustificationBits.Shift(1)
|
|
|
|
// Note: the spec refers to the bit index position starting at 1 instead of starting at zero.
|
|
// We will use that paradigm here for consistency with the godoc spec definition.
|
|
|
|
// If 2/3 or more of total balance attested in the previous epoch.
|
|
if 3*prevAttestedBal >= 2*totalBal {
|
|
blockRoot, err := helpers.BlockRoot(state, prevEpoch)
|
|
if err != nil {
|
|
return nil, errors.Wrapf(err, "could not get block root for previous epoch %d", prevEpoch)
|
|
}
|
|
state.CurrentJustifiedCheckpoint = ðpb.Checkpoint{Epoch: prevEpoch, Root: blockRoot}
|
|
state.JustificationBits.SetBitAt(1, true)
|
|
}
|
|
|
|
// If 2/3 or more of the total balance attested in the current epoch.
|
|
if 3*currAttestedBal >= 2*totalBal {
|
|
blockRoot, err := helpers.BlockRoot(state, currentEpoch)
|
|
if err != nil {
|
|
return nil, errors.Wrapf(err, "could not get block root for current epoch %d", prevEpoch)
|
|
}
|
|
state.CurrentJustifiedCheckpoint = ðpb.Checkpoint{Epoch: currentEpoch, Root: blockRoot}
|
|
state.JustificationBits.SetBitAt(0, true)
|
|
}
|
|
|
|
// Process finalization according to ETH2.0 specifications.
|
|
justification := state.JustificationBits.Bytes()[0]
|
|
|
|
// 2nd/3rd/4th (0b1110) most recent epochs are justified, the 2nd using the 4th as source.
|
|
if justification&0x0E == 0x0E && (oldPrevJustifiedCheckpoint.Epoch+3) == currentEpoch {
|
|
state.FinalizedCheckpoint = oldPrevJustifiedCheckpoint
|
|
}
|
|
|
|
// 2nd/3rd (0b0110) most recent epochs are justified, the 2nd using the 3rd as source.
|
|
if justification&0x06 == 0x06 && (oldPrevJustifiedCheckpoint.Epoch+2) == currentEpoch {
|
|
state.FinalizedCheckpoint = oldPrevJustifiedCheckpoint
|
|
}
|
|
|
|
// 1st/2nd/3rd (0b0111) most recent epochs are justified, the 1st using the 3rd as source.
|
|
if justification&0x07 == 0x07 && (oldCurrJustifiedCheckpoint.Epoch+2) == currentEpoch {
|
|
state.FinalizedCheckpoint = oldCurrJustifiedCheckpoint
|
|
}
|
|
|
|
// The 1st/2nd (0b0011) most recent epochs are justified, the 1st using the 2nd as source
|
|
if justification&0x03 == 0x03 && (oldCurrJustifiedCheckpoint.Epoch+1) == currentEpoch {
|
|
state.FinalizedCheckpoint = oldCurrJustifiedCheckpoint
|
|
}
|
|
|
|
return state, nil
|
|
}
|
|
|
|
// ProcessRewardsAndPenalties processes the rewards and penalties of individual validator.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def process_rewards_and_penalties(state: BeaconState) -> None:
|
|
// if get_current_epoch(state) == GENESIS_EPOCH:
|
|
// return
|
|
//
|
|
// rewards1, penalties1 = get_attestation_deltas(state)
|
|
// rewards2, penalties2 = get_crosslink_deltas(state)
|
|
// for i in range(len(state.validator_registry)):
|
|
// increase_balance(state, i, rewards1[i] + rewards2[i])
|
|
// decrease_balance(state, i, penalties1[i] + penalties2[i])
|
|
func ProcessRewardsAndPenalties(state *pb.BeaconState) (*pb.BeaconState, error) {
|
|
// Can't process rewards and penalties in genesis epoch.
|
|
if helpers.CurrentEpoch(state) == 0 {
|
|
return state, nil
|
|
}
|
|
attsRewards, attsPenalties, err := attestationDelta(state)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not get attestation delta")
|
|
}
|
|
|
|
for i := 0; i < len(state.Validators); i++ {
|
|
state = helpers.IncreaseBalance(state, uint64(i), attsRewards[i])
|
|
state = helpers.DecreaseBalance(state, uint64(i), attsPenalties[i])
|
|
}
|
|
return state, nil
|
|
}
|
|
|
|
// ProcessRegistryUpdates rotates validators in and out of active pool.
|
|
// the amount to rotate is determined churn limit.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def process_registry_updates(state: BeaconState) -> None:
|
|
// # Process activation eligibility and ejections
|
|
// for index, validator in enumerate(state.validator_registry):
|
|
// if (
|
|
// validator.activation_eligibility_epoch == FAR_FUTURE_EPOCH and
|
|
// validator.effective_balance >= MAX_EFFECTIVE_BALANCE
|
|
// ):
|
|
// validator.activation_eligibility_epoch = get_current_epoch(state)
|
|
//
|
|
// if is_active_validator(validator, get_current_epoch(state)) and validator.effective_balance <= EJECTION_BALANCE:
|
|
// initiate_validator_exit(state, index)
|
|
//
|
|
// # Queue validators eligible for activation and not dequeued for activation prior to finalized epoch
|
|
// activation_queue = sorted([
|
|
// index for index, validator in enumerate(state.validator_registry) if
|
|
// validator.activation_eligibility_epoch != FAR_FUTURE_EPOCH and
|
|
// validator.activation_epoch >= get_delayed_activation_exit_epoch(state.finalized_epoch)
|
|
// ], key=lambda index: state.validator_registry[index].activation_eligibility_epoch)
|
|
// # Dequeued validators for activation up to churn limit (without resetting activation epoch)
|
|
// for index in activation_queue[:get_churn_limit(state)]:
|
|
// validator = state.validator_registry[index]
|
|
// if validator.activation_epoch == FAR_FUTURE_EPOCH:
|
|
// validator.activation_epoch = get_delayed_activation_exit_epoch(get_current_epoch(state))
|
|
func ProcessRegistryUpdates(state *pb.BeaconState) (*pb.BeaconState, error) {
|
|
currentEpoch := helpers.CurrentEpoch(state)
|
|
|
|
var err error
|
|
for idx, validator := range state.Validators {
|
|
// Process the validators for activation eligibility.
|
|
eligibleToActivate := validator.ActivationEligibilityEpoch == params.BeaconConfig().FarFutureEpoch
|
|
properBalance := validator.EffectiveBalance >= params.BeaconConfig().MaxEffectiveBalance
|
|
if eligibleToActivate && properBalance {
|
|
validator.ActivationEligibilityEpoch = currentEpoch
|
|
}
|
|
// Process the validators for ejection.
|
|
isActive := helpers.IsActiveValidator(validator, currentEpoch)
|
|
belowEjectionBalance := validator.EffectiveBalance <= params.BeaconConfig().EjectionBalance
|
|
if isActive && belowEjectionBalance {
|
|
state, err = validators.InitiateValidatorExit(state, uint64(idx))
|
|
if err != nil {
|
|
return nil, errors.Wrapf(err, "could not initiate exit for validator %d", idx)
|
|
}
|
|
}
|
|
}
|
|
|
|
// Queue the validators whose eligible to activate and sort them by activation eligibility epoch number
|
|
var activationQ []uint64
|
|
for idx, validator := range state.Validators {
|
|
eligibleActivated := validator.ActivationEligibilityEpoch != params.BeaconConfig().FarFutureEpoch
|
|
canBeActive := validator.ActivationEpoch >= helpers.DelayedActivationExitEpoch(state.FinalizedCheckpoint.Epoch)
|
|
if eligibleActivated && canBeActive {
|
|
activationQ = append(activationQ, uint64(idx))
|
|
}
|
|
}
|
|
sort.Slice(activationQ, func(i, j int) bool {
|
|
return state.Validators[i].ActivationEligibilityEpoch < state.Validators[j].ActivationEligibilityEpoch
|
|
})
|
|
|
|
// Only activate just enough validators according to the activation churn limit.
|
|
limit := len(activationQ)
|
|
activeValidatorCount, err := helpers.ActiveValidatorCount(state, currentEpoch)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not get active validator count")
|
|
}
|
|
|
|
churnLimit, err := helpers.ValidatorChurnLimit(activeValidatorCount)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not get churn limit")
|
|
}
|
|
|
|
// Prevent churn limit cause index out of bound.
|
|
if int(churnLimit) < limit {
|
|
limit = int(churnLimit)
|
|
}
|
|
for _, index := range activationQ[:limit] {
|
|
validator := state.Validators[index]
|
|
if validator.ActivationEpoch == params.BeaconConfig().FarFutureEpoch {
|
|
validator.ActivationEpoch = helpers.DelayedActivationExitEpoch(currentEpoch)
|
|
}
|
|
}
|
|
return state, nil
|
|
}
|
|
|
|
// ProcessSlashings processes the slashed validators during epoch processing,
|
|
//
|
|
// def process_slashings(state: BeaconState) -> None:
|
|
// epoch = get_current_epoch(state)
|
|
// total_balance = get_total_active_balance(state)
|
|
// for index, validator in enumerate(state.validators):
|
|
// if validator.slashed and epoch + EPOCHS_PER_SLASHINGS_VECTOR // 2 == validator.withdrawable_epoch:
|
|
// increment = EFFECTIVE_BALANCE_INCREMENT # Factored out from penalty numerator to avoid uint64 overflow
|
|
// penalty_numerator = validator.effective_balance // increment * min(sum(state.slashings) * 3, total_balance)
|
|
// penalty = penalty_numerator // total_balance * increment
|
|
// decrease_balance(state, ValidatorIndex(index), penalty)
|
|
func ProcessSlashings(state *pb.BeaconState) (*pb.BeaconState, error) {
|
|
currentEpoch := helpers.CurrentEpoch(state)
|
|
totalBalance, err := helpers.TotalActiveBalance(state)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not get total active balance")
|
|
}
|
|
|
|
// Compute slashed balances in the current epoch
|
|
exitLength := params.BeaconConfig().EpochsPerSlashingsVector
|
|
|
|
// Compute the sum of state slashings
|
|
totalSlashing := uint64(0)
|
|
for _, slashing := range state.Slashings {
|
|
totalSlashing += slashing
|
|
}
|
|
|
|
// Compute slashing for each validator.
|
|
for index, validator := range state.Validators {
|
|
correctEpoch := (currentEpoch + exitLength/2) == validator.WithdrawableEpoch
|
|
if validator.Slashed && correctEpoch {
|
|
minSlashing := mathutil.Min(totalSlashing*3, totalBalance)
|
|
increment := params.BeaconConfig().EffectiveBalanceIncrement
|
|
penaltyNumerator := validator.EffectiveBalance / increment * minSlashing
|
|
penalty := penaltyNumerator / totalBalance * increment
|
|
state = helpers.DecreaseBalance(state, uint64(index), penalty)
|
|
}
|
|
}
|
|
return state, err
|
|
}
|
|
|
|
// ProcessFinalUpdates processes the final updates during epoch processing.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def process_final_updates(state: BeaconState) -> None:
|
|
// current_epoch = get_current_epoch(state)
|
|
// next_epoch = Epoch(current_epoch + 1)
|
|
// # Reset eth1 data votes
|
|
// if (state.slot + 1) % SLOTS_PER_ETH1_VOTING_PERIOD == 0:
|
|
// state.eth1_data_votes = []
|
|
// # Update effective balances with hysteresis
|
|
// for index, validator in enumerate(state.validators):
|
|
// balance = state.balances[index]
|
|
// HALF_INCREMENT = EFFECTIVE_BALANCE_INCREMENT // 2
|
|
// if balance < validator.effective_balance or validator.effective_balance + 3 * HALF_INCREMENT < balance:
|
|
// validator.effective_balance = min(balance - balance % EFFECTIVE_BALANCE_INCREMENT, MAX_EFFECTIVE_BALANCE)
|
|
// # Set active index root
|
|
// index_epoch = Epoch(next_epoch + ACTIVATION_EXIT_DELAY)
|
|
// index_root_position = index_epoch % EPOCHS_PER_HISTORICAL_VECTOR
|
|
// indices_list = List[ValidatorIndex, VALIDATOR_REGISTRY_LIMIT](get_active_validator_indices(state, index_epoch))
|
|
// state.active_index_roots[index_root_position] = hash_tree_root(indices_list)
|
|
// # Set committees root
|
|
// committee_root_position = next_epoch % EPOCHS_PER_HISTORICAL_VECTOR
|
|
// state.compact_committees_roots[committee_root_position] = get_compact_committees_root(state, next_epoch)
|
|
// # Reset slashings
|
|
// state.slashings[next_epoch % EPOCHS_PER_SLASHINGS_VECTOR] = Gwei(0)
|
|
// # Set randao mix
|
|
// state.randao_mixes[next_epoch % EPOCHS_PER_HISTORICAL_VECTOR] = get_randao_mix(state, current_epoch)
|
|
// # Set historical root accumulator
|
|
// if next_epoch % (SLOTS_PER_HISTORICAL_ROOT // SLOTS_PER_EPOCH) == 0:
|
|
// historical_batch = HistoricalBatch(block_roots=state.block_roots, state_roots=state.state_roots)
|
|
// state.historical_roots.append(hash_tree_root(historical_batch))
|
|
// # Update start shard
|
|
// state.start_shard = Shard((state.start_shard + get_shard_delta(state, current_epoch)) % SHARD_COUNT)
|
|
// # Rotate current/previous epoch attestations
|
|
// state.previous_epoch_attestations = state.current_epoch_attestations
|
|
// state.current_epoch_attestations = []
|
|
func ProcessFinalUpdates(state *pb.BeaconState) (*pb.BeaconState, error) {
|
|
currentEpoch := helpers.CurrentEpoch(state)
|
|
nextEpoch := currentEpoch + 1
|
|
|
|
// Reset ETH1 data votes.
|
|
if (state.Slot+1)%params.BeaconConfig().SlotsPerEth1VotingPeriod == 0 {
|
|
state.Eth1DataVotes = []*ethpb.Eth1Data{}
|
|
}
|
|
|
|
// Update effective balances with hysteresis.
|
|
for i, v := range state.Validators {
|
|
balance := state.Balances[i]
|
|
halfInc := params.BeaconConfig().EffectiveBalanceIncrement / 2
|
|
if balance < v.EffectiveBalance || v.EffectiveBalance+3*halfInc < balance {
|
|
v.EffectiveBalance = params.BeaconConfig().MaxEffectiveBalance
|
|
if v.EffectiveBalance > balance-balance%params.BeaconConfig().EffectiveBalanceIncrement {
|
|
v.EffectiveBalance = balance - balance%params.BeaconConfig().EffectiveBalanceIncrement
|
|
}
|
|
}
|
|
}
|
|
|
|
// Set total slashed balances.
|
|
slashedExitLength := params.BeaconConfig().EpochsPerSlashingsVector
|
|
state.Slashings[nextEpoch%slashedExitLength] = 0
|
|
|
|
// Set RANDAO mix.
|
|
randaoMixLength := params.BeaconConfig().EpochsPerHistoricalVector
|
|
mix := helpers.RandaoMix(state, currentEpoch)
|
|
state.RandaoMixes[nextEpoch%randaoMixLength] = mix
|
|
|
|
// Set historical root accumulator.
|
|
epochsPerHistoricalRoot := params.BeaconConfig().SlotsPerHistoricalRoot / params.BeaconConfig().SlotsPerEpoch
|
|
if nextEpoch%epochsPerHistoricalRoot == 0 {
|
|
historicalBatch := &pb.HistoricalBatch{
|
|
BlockRoots: state.BlockRoots,
|
|
StateRoots: state.StateRoots,
|
|
}
|
|
batchRoot, err := ssz.HashTreeRoot(historicalBatch)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not hash historical batch")
|
|
}
|
|
state.HistoricalRoots = append(state.HistoricalRoots, batchRoot[:])
|
|
}
|
|
|
|
// Rotate current and previous epoch attestations.
|
|
state.PreviousEpochAttestations = state.CurrentEpochAttestations
|
|
state.CurrentEpochAttestations = []*pb.PendingAttestation{}
|
|
|
|
return state, nil
|
|
}
|
|
|
|
// unslashedAttestingIndices returns all the attesting indices from a list of attestations,
|
|
// it sorts the indices and filters out the slashed ones.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def get_unslashed_attesting_indices(state: BeaconState,
|
|
// attestations: Sequence[PendingAttestation]) -> Set[ValidatorIndex]:
|
|
// output = set() # type: Set[ValidatorIndex]
|
|
// for a in attestations:
|
|
// output = output.union(get_attesting_indices(state, a.data, a.aggregation_bits))
|
|
// return set(filter(lambda index: not state.validators[index].slashed, output))
|
|
func unslashedAttestingIndices(state *pb.BeaconState, atts []*pb.PendingAttestation) ([]uint64, error) {
|
|
var setIndices []uint64
|
|
seen := make(map[uint64]bool)
|
|
for _, att := range atts {
|
|
attestingIndices, err := helpers.AttestingIndices(state, att.Data, att.AggregationBits)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "could not get attester indices")
|
|
}
|
|
// Create a set for attesting indices
|
|
set := make([]uint64, 0, len(attestingIndices))
|
|
for _, index := range attestingIndices {
|
|
if !seen[index] {
|
|
set = append(set, index)
|
|
}
|
|
seen[index] = true
|
|
}
|
|
setIndices = append(setIndices, set...)
|
|
}
|
|
// Sort the attesting set indices by increasing order.
|
|
sort.Slice(setIndices, func(i, j int) bool { return setIndices[i] < setIndices[j] })
|
|
// Remove the slashed validator indices.
|
|
for i := 0; i < len(setIndices); i++ {
|
|
if state.Validators[setIndices[i]].Slashed {
|
|
setIndices = append(setIndices[:i], setIndices[i+1:]...)
|
|
}
|
|
}
|
|
|
|
return setIndices, nil
|
|
}
|
|
|
|
// BaseReward takes state and validator index and calculate
|
|
// individual validator's base reward quotient.
|
|
//
|
|
// Note: Adjusted quotient is calculated of base reward because it's too inefficient
|
|
// to repeat the same calculation for every validator versus just doing it once.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def get_base_reward(state: BeaconState, index: ValidatorIndex) -> Gwei:
|
|
// total_balance = get_total_active_balance(state)
|
|
// effective_balance = state.validator_registry[index].effective_balance
|
|
// return effective_balance * BASE_REWARD_FACTOR // integer_squareroot(total_balance) // BASE_REWARDS_PER_EPOCH
|
|
func BaseReward(state *pb.BeaconState, index uint64) (uint64, error) {
|
|
totalBalance, err := helpers.TotalActiveBalance(state)
|
|
if err != nil {
|
|
return 0, errors.Wrap(err, "could not calculate active balance")
|
|
}
|
|
effectiveBalance := state.Validators[index].EffectiveBalance
|
|
baseReward := effectiveBalance * params.BeaconConfig().BaseRewardFactor /
|
|
mathutil.IntegerSquareRoot(totalBalance) / params.BeaconConfig().BaseRewardsPerEpoch
|
|
return baseReward, nil
|
|
}
|
|
|
|
// attestationDelta calculates the rewards and penalties of individual
|
|
// validator for voting the correct FFG source, FFG target, and head. It
|
|
// also calculates proposer delay inclusion and inactivity rewards
|
|
// and penalties. Individual rewards and penalties are returned in list.
|
|
//
|
|
// Note: we calculated adjusted quotient outside of base reward because it's too inefficient
|
|
// to repeat the same calculation for every validator versus just doing it once.
|
|
//
|
|
// Spec pseudocode definition:
|
|
// def get_attestation_deltas(state: BeaconState) -> Tuple[Sequence[Gwei], Sequence[Gwei]]:
|
|
// previous_epoch = get_previous_epoch(state)
|
|
// total_balance = get_total_active_balance(state)
|
|
// rewards = [Gwei(0) for _ in range(len(state.validators))]
|
|
// penalties = [Gwei(0) for _ in range(len(state.validators))]
|
|
// eligible_validator_indices = [
|
|
// ValidatorIndex(index) for index, v in enumerate(state.validators)
|
|
// if is_active_validator(v, previous_epoch) or (v.slashed and previous_epoch + 1 < v.withdrawable_epoch)
|
|
// ]
|
|
//
|
|
// # Micro-incentives for matching FFG source, FFG target, and head
|
|
// matching_source_attestations = get_matching_source_attestations(state, previous_epoch)
|
|
// matching_target_attestations = get_matching_target_attestations(state, previous_epoch)
|
|
// matching_head_attestations = get_matching_head_attestations(state, previous_epoch)
|
|
// for attestations in (matching_source_attestations, matching_target_attestations, matching_head_attestations):
|
|
// unslashed_attesting_indices = get_unslashed_attesting_indices(state, attestations)
|
|
// attesting_balance = get_total_balance(state, unslashed_attesting_indices)
|
|
// for index in eligible_validator_indices:
|
|
// if index in unslashed_attesting_indices:
|
|
// rewards[index] += get_base_reward(state, index) * attesting_balance // total_balance
|
|
// else:
|
|
// penalties[index] += get_base_reward(state, index)
|
|
//
|
|
// # Proposer and inclusion delay micro-rewards
|
|
// for index in get_unslashed_attesting_indices(state, matching_source_attestations):
|
|
// index = ValidatorIndex(index)
|
|
// attestation = min([
|
|
// a for a in matching_source_attestations
|
|
// if index in get_attesting_indices(state, a.data, a.aggregation_bits)
|
|
// ], key=lambda a: a.inclusion_delay)
|
|
// proposer_reward = Gwei(get_base_reward(state, index) // PROPOSER_REWARD_QUOTIENT)
|
|
// rewards[attestation.proposer_index] += proposer_reward
|
|
// max_attester_reward = get_base_reward(state, index) - proposer_reward
|
|
// rewards[index] += Gwei(max_attester_reward // attestation.inclusion_delay)
|
|
//
|
|
// # Inactivity penalty
|
|
// finality_delay = previous_epoch - state.finalized_checkpoint.epoch
|
|
// if finality_delay > MIN_EPOCHS_TO_INACTIVITY_PENALTY:
|
|
// matching_target_attesting_indices = get_unslashed_attesting_indices(state, matching_target_attestations)
|
|
// for index in eligible_validator_indices:
|
|
// index = ValidatorIndex(index)
|
|
// penalties[index] += Gwei(BASE_REWARDS_PER_EPOCH * get_base_reward(state, index))
|
|
// if index not in matching_target_attesting_indices:
|
|
// penalties[index] += Gwei(
|
|
// state.validators[index].effective_balance * finality_delay // INACTIVITY_PENALTY_QUOTIENT
|
|
// )
|
|
//
|
|
// return rewards, penalties
|
|
func attestationDelta(state *pb.BeaconState) ([]uint64, []uint64, error) {
|
|
prevEpoch := helpers.PrevEpoch(state)
|
|
totalBalance, err := helpers.TotalActiveBalance(state)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get total active balance")
|
|
}
|
|
|
|
rewards := make([]uint64, len(state.Validators))
|
|
penalties := make([]uint64, len(state.Validators))
|
|
|
|
// Filter out the list of eligible validator indices. The eligible validator
|
|
// has to be active or slashed but before withdrawn.
|
|
var eligible []uint64
|
|
for i, v := range state.Validators {
|
|
isActive := helpers.IsActiveValidator(v, prevEpoch)
|
|
isSlashed := v.Slashed && (prevEpoch+1 < v.WithdrawableEpoch)
|
|
if isActive || isSlashed {
|
|
eligible = append(eligible, uint64(i))
|
|
}
|
|
}
|
|
|
|
// Apply rewards and penalties for voting correct source target and head.
|
|
// Construct a attestations list contains source, target and head attestations.
|
|
atts, err := MatchAttestations(state, prevEpoch)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get source, target and head attestations")
|
|
}
|
|
var attsPackage [][]*pb.PendingAttestation
|
|
attsPackage = append(attsPackage, atts.source)
|
|
attsPackage = append(attsPackage, atts.Target)
|
|
attsPackage = append(attsPackage, atts.head)
|
|
|
|
// Cache the validators who voted correctly for source in a map
|
|
// to calculate earliest attestation rewards later.
|
|
attestersVotedSource := make(map[uint64]*pb.PendingAttestation)
|
|
// Compute rewards / penalties for each attestation in the list and update
|
|
// the rewards and penalties lists.
|
|
for i, matchAtt := range attsPackage {
|
|
indices, err := unslashedAttestingIndices(state, matchAtt)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get attestation indices")
|
|
}
|
|
|
|
attested := make(map[uint64]bool)
|
|
// Construct a map to look up validators that voted for source, target or head.
|
|
for _, index := range indices {
|
|
if i == 0 {
|
|
attestersVotedSource[index] = &pb.PendingAttestation{InclusionDelay: params.BeaconConfig().FarFutureEpoch}
|
|
}
|
|
attested[index] = true
|
|
}
|
|
attestedBalance := helpers.TotalBalance(state, indices)
|
|
|
|
// Update rewards and penalties to each eligible validator index.
|
|
for _, index := range eligible {
|
|
base, err := BaseReward(state, index)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get base reward")
|
|
}
|
|
if _, ok := attested[index]; ok {
|
|
rewards[index] += base * attestedBalance / totalBalance
|
|
} else {
|
|
penalties[index] += base
|
|
}
|
|
}
|
|
}
|
|
|
|
// For every index, filter the matching source attestation that correspond to the index,
|
|
// sort by inclusion delay and get the one that was included on chain first.
|
|
for _, att := range atts.source {
|
|
indices, err := helpers.AttestingIndices(state, att.Data, att.AggregationBits)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get attester indices")
|
|
}
|
|
for _, i := range indices {
|
|
if _, ok := attestersVotedSource[i]; ok {
|
|
if attestersVotedSource[i].InclusionDelay > att.InclusionDelay {
|
|
attestersVotedSource[i] = att
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
for i, a := range attestersVotedSource {
|
|
baseReward, err := BaseReward(state, i)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get proposer reward")
|
|
}
|
|
proposerReward := baseReward / params.BeaconConfig().ProposerRewardQuotient
|
|
rewards[a.ProposerIndex] += proposerReward
|
|
attesterReward := baseReward - proposerReward
|
|
rewards[i] += attesterReward / a.InclusionDelay
|
|
}
|
|
|
|
// Apply penalties for quadratic leaks.
|
|
// When epoch since finality exceeds inactivity penalty constant, the penalty gets increased
|
|
// based on the finality delay.
|
|
finalityDelay := prevEpoch - state.FinalizedCheckpoint.Epoch
|
|
if finalityDelay > params.BeaconConfig().MinEpochsToInactivityPenalty {
|
|
targetIndices, err := unslashedAttestingIndices(state, atts.Target)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get attestation indices")
|
|
}
|
|
attestedTarget := make(map[uint64]bool)
|
|
for _, index := range targetIndices {
|
|
attestedTarget[index] = true
|
|
}
|
|
for _, index := range eligible {
|
|
base, err := BaseReward(state, index)
|
|
if err != nil {
|
|
return nil, nil, errors.Wrap(err, "could not get base reward")
|
|
}
|
|
penalties[index] += params.BeaconConfig().BaseRewardsPerEpoch * base
|
|
if _, ok := attestedTarget[index]; !ok {
|
|
penalties[index] += state.Validators[index].EffectiveBalance * finalityDelay /
|
|
params.BeaconConfig().InactivityPenaltyQuotient
|
|
}
|
|
}
|
|
}
|
|
|
|
return rewards, penalties, nil
|
|
}
|