nimbus-eth2/beacon_chain/spec/state_transition_block.nim

899 lines
38 KiB
Nim

# beacon_chain
# Copyright (c) 2018-2023 Status Research & Development GmbH
# Licensed and distributed under either of
# * MIT license (license terms in the root directory or at https://opensource.org/licenses/MIT).
# * Apache v2 license (license terms in the root directory or at https://www.apache.org/licenses/LICENSE-2.0).
# at your option. This file may not be copied, modified, or distributed except according to those terms.
# State transition - block processing, as described in
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#block-processing
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/altair/beacon-chain.md#block-processing
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/bellatrix/beacon-chain.md#block-processing
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/capella/beacon-chain.md#block-processing
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.1/specs/deneb/beacon-chain.md#block-processing
#
# The entry point is `process_block` which is at the bottom of this file.
#
# General notes about the code:
# * Weird styling - the sections taken from the spec use python styling while
# the others use NEP-1 - helps grepping identifiers in spec
# * When updating the code, add TODO sections to mark where there are clear
# improvements to be made - other than that, keep things similar to spec unless
# motivated by security or performance considerations
{.push raises: [].}
import
chronicles, metrics,
../extras,
./datatypes/[phase0, altair, bellatrix, deneb],
"."/[beaconstate, eth2_merkleization, helpers, validator, signatures],
kzg4844/kzg_abi, kzg4844/kzg_ex
from std/algorithm import fill, sorted
from std/sequtils import count, filterIt, mapIt
from ./datatypes/capella import
BeaconState, MAX_WITHDRAWALS_PER_PAYLOAD, SignedBLSToExecutionChange,
Withdrawal
export extras, phase0, altair
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#block-header
func process_block_header*(
state: var ForkyBeaconState, blck: SomeForkyBeaconBlock,
flags: UpdateFlags, cache: var StateCache): Result[void, cstring] =
# Verify that the slots match
if not (blck.slot == state.slot):
return err("process_block_header: slot mismatch")
# Verify that the block is newer than latest block header
if not (blck.slot > state.latest_block_header.slot):
return err("process_block_header: block not newer than latest block header")
# Verify that proposer index is the correct index
let proposer_index = get_beacon_proposer_index(state, cache).valueOr:
return err("process_block_header: proposer missing")
if not (blck.proposer_index == proposer_index):
return err("process_block_header: proposer index incorrect")
# Verify that the parent matches
if not (blck.parent_root == hash_tree_root(state.latest_block_header)):
return err("process_block_header: previous block root mismatch")
# Verify proposer is not slashed
if state.validators.item(blck.proposer_index).slashed:
return err("process_block_header: proposer slashed")
# Cache current block as the new latest block
state.latest_block_header = BeaconBlockHeader(
slot: blck.slot,
proposer_index: blck.proposer_index,
parent_root: blck.parent_root,
# state_root: zeroed, overwritten in the next `process_slot` call
body_root: hash_tree_root(blck.body),
)
ok()
func `xor`[T: array](a, b: T): T =
for i in 0..<result.len:
result[i] = a[i] xor b[i]
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#randao
proc process_randao(
state: var ForkyBeaconState, body: SomeForkyBeaconBlockBody,
flags: UpdateFlags, cache: var StateCache): Result[void, cstring] =
let
proposer_index = get_beacon_proposer_index(state, cache).valueOr:
return err("process_randao: proposer index missing, probably along with any active validators")
# Verify RANDAO reveal
let epoch = state.get_current_epoch()
if skipBlsValidation notin flags and body.randao_reveal isnot TrustedSig:
let proposer_pubkey = state.validators.item(proposer_index).pubkey
# `state_transition.makeBeaconBlock` ensures this is run with a trusted
# signature, but unless the full skipBlsValidation is specified, RANDAO
# epoch signatures still have to be verified.
if not verify_epoch_signature(
state.fork, state.genesis_validators_root, epoch, proposer_pubkey,
body.randao_reveal):
return err("process_randao: invalid epoch signature")
# Mix in RANDAO reveal
let
mix = get_randao_mix(state, epoch)
rr = eth2digest(body.randao_reveal.toRaw()).data
state.randao_mixes.mitem(epoch mod EPOCHS_PER_HISTORICAL_VECTOR).data =
mix.data xor rr
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#eth1-data
func process_eth1_data(
state: var ForkyBeaconState,
body: SomeForkyBeaconBlockBody): Result[void, cstring] =
if not state.eth1_data_votes.add body.eth1_data:
# Count is reset in process_final_updates, so this should never happen
return err("process_eth1_data: no more room for eth1 data")
if state.eth1_data_votes.asSeq.count(body.eth1_data).uint64 * 2 >
SLOTS_PER_ETH1_VOTING_PERIOD:
state.eth1_data = body.eth1_data
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#is_slashable_validator
func is_slashable_validator(validator: Validator, epoch: Epoch): bool =
# Check if ``validator`` is slashable.
(not validator.slashed) and
(validator.activation_epoch <= epoch) and
(epoch < validator.withdrawable_epoch)
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#proposer-slashings
proc check_proposer_slashing*(
state: ForkyBeaconState, proposer_slashing: SomeProposerSlashing,
flags: UpdateFlags):
Result[ValidatorIndex, cstring] =
let
header_1 = proposer_slashing.signed_header_1.message
header_2 = proposer_slashing.signed_header_2.message
# Verify header slots match
if not (header_1.slot == header_2.slot):
return err("check_proposer_slashing: slot mismatch")
# Verify header proposer indices match
if not (header_1.proposer_index == header_2.proposer_index):
return err("check_proposer_slashing: proposer indices mismatch")
# Verify the headers are different
if not (header_1 != header_2):
return err("check_proposer_slashing: headers not different")
# Verify the proposer is slashable
if header_1.proposer_index >= state.validators.lenu64:
return err("check_proposer_slashing: invalid proposer index")
let proposer = unsafeAddr state.validators[header_1.proposer_index]
if not is_slashable_validator(proposer[], get_current_epoch(state)):
return err("check_proposer_slashing: slashed proposer")
# Verify signatures
if skipBlsValidation notin flags:
for signed_header in [proposer_slashing.signed_header_1,
proposer_slashing.signed_header_2]:
if not verify_block_signature(
state.fork, state.genesis_validators_root, signed_header.message.slot,
signed_header.message, proposer[].pubkey,
signed_header.signature):
return err("check_proposer_slashing: invalid signature")
# Verified above against state.validators
ValidatorIndex.init(header_1.proposer_index)
proc check_proposer_slashing*(
state: var ForkedHashedBeaconState, proposer_slashing: SomeProposerSlashing,
flags: UpdateFlags): Result[ValidatorIndex, cstring] =
withState(state):
check_proposer_slashing(forkyState.data, proposer_slashing, flags)
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#proposer-slashings
proc process_proposer_slashing*(
cfg: RuntimeConfig, state: var ForkyBeaconState,
proposer_slashing: SomeProposerSlashing, flags: UpdateFlags,
cache: var StateCache):
Result[void, cstring] =
let proposer_index = ? check_proposer_slashing(state, proposer_slashing, flags)
? slash_validator(cfg, state, proposer_index, cache)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#is_slashable_attestation_data
func is_slashable_attestation_data(
data_1: AttestationData, data_2: AttestationData): bool =
## Check if ``data_1`` and ``data_2`` are slashable according to Casper FFG
## rules.
# Double vote
(data_1 != data_2 and data_1.target.epoch == data_2.target.epoch) or
# Surround vote
(data_1.source.epoch < data_2.source.epoch and
data_2.target.epoch < data_1.target.epoch)
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#attester-slashings
proc check_attester_slashing*(
state: ForkyBeaconState,
attester_slashing: SomeAttesterSlashing,
flags: UpdateFlags
): Result[seq[ValidatorIndex], cstring] =
let
attestation_1 = attester_slashing.attestation_1
attestation_2 = attester_slashing.attestation_2
if not is_slashable_attestation_data(
attestation_1.data, attestation_2.data):
return err("Attester slashing: surround or double vote check failed")
if not is_valid_indexed_attestation(state, attestation_1, flags).isOk():
return err("Attester slashing: invalid attestation 1")
if not is_valid_indexed_attestation(state, attestation_2, flags).isOk():
return err("Attester slashing: invalid attestation 2")
var slashed_indices: seq[ValidatorIndex]
let attesting_indices_2 = toHashSet(attestation_2.attesting_indices.asSeq)
for index in sorted(filterIt(
attestation_1.attesting_indices.asSeq, it in attesting_indices_2),
system.cmp):
if is_slashable_validator(
state.validators[index], get_current_epoch(state)):
slashed_indices.add ValidatorIndex.init(index).expect(
"checked by is_valid_indexed_attestation")
if slashed_indices.len == 0:
return err("Attester slashing: Trying to slash participant(s) twice")
ok slashed_indices
proc check_attester_slashing*(
state: var ForkedHashedBeaconState, attester_slashing: SomeAttesterSlashing,
flags: UpdateFlags): Result[seq[ValidatorIndex], cstring] =
withState(state):
check_attester_slashing(forkyState.data, attester_slashing, flags)
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#attester-slashings
proc process_attester_slashing*(
cfg: RuntimeConfig,
state: var ForkyBeaconState,
attester_slashing: SomeAttesterSlashing,
flags: UpdateFlags,
cache: var StateCache
): Result[void, cstring] =
let slashed_attesters =
? check_attester_slashing(state, attester_slashing, flags)
for index in slashed_attesters:
? slash_validator(cfg, state, index, cache)
ok()
func findValidatorIndex*(state: ForkyBeaconState, pubkey: ValidatorPubKey):
Opt[ValidatorIndex] =
# This linear scan is unfortunate, but should be fairly fast as we do a simple
# byte comparison of the key. The alternative would be to build a Table, but
# given that each block can hold no more than 16 deposits, it's slower to
# build the table and use it for lookups than to scan it like this.
# Once we have a reusable, long-lived cache, this should be revisited
for vidx in state.validators.vindices:
if state.validators.asSeq[vidx].pubkey == pubkey:
return Opt[ValidatorIndex].ok(vidx)
proc process_deposit*(cfg: RuntimeConfig,
state: var ForkyBeaconState,
deposit: Deposit,
flags: UpdateFlags): Result[void, cstring] =
## Process an Eth1 deposit, registering a validator or increasing its balance.
# Verify the Merkle branch
if not is_valid_merkle_branch(
hash_tree_root(deposit.data),
deposit.proof,
DEPOSIT_CONTRACT_TREE_DEPTH + 1, # Add 1 for the `List` length mix-in
state.eth1_deposit_index,
state.eth1_data.deposit_root,
):
return err("process_deposit: deposit Merkle validation failed")
# Deposits must be processed in order
state.eth1_deposit_index += 1
let
pubkey = deposit.data.pubkey
amount = deposit.data.amount
index = findValidatorIndex(state, pubkey)
if index.isSome():
# Increase balance by deposit amount
increase_balance(state, index.get(), amount)
else:
# Verify the deposit signature (proof of possession) which is not checked
# by the deposit contract
if verify_deposit_signature(cfg, deposit.data):
# New validator! Add validator and balance entries
if not state.validators.add(get_validator_from_deposit(deposit.data)):
return err("process_deposit: too many validators")
if not state.balances.add(amount):
static: doAssert state.balances.maxLen == state.validators.maxLen
raiseAssert "adding validator succeeded, so should balances"
when state is altair.BeaconState or state is bellatrix.BeaconState or
state is capella.BeaconState or state is deneb.BeaconState:
if not state.previous_epoch_participation.add(ParticipationFlags(0)):
return err("process_deposit: too many validators (previous_epoch_participation)")
if not state.current_epoch_participation.add(ParticipationFlags(0)):
return err("process_deposit: too many validators (current_epoch_participation)")
if not state.inactivity_scores.add(0'u64):
return err("process_deposit: too many validators (inactivity_scores)")
doAssert state.validators.len == state.balances.len
else:
# Deposits may come with invalid signatures - in that case, they are not
# turned into a validator but still get processed to keep the deposit
# index correct
trace "Skipping deposit with invalid signature",
deposit = shortLog(deposit.data)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#voluntary-exits
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.1/specs/deneb/beacon-chain.md#modified-process_voluntary_exit
proc check_voluntary_exit*(
cfg: RuntimeConfig,
state: ForkyBeaconState,
signed_voluntary_exit: SomeSignedVoluntaryExit,
flags: UpdateFlags): Result[ValidatorIndex, cstring] =
let voluntary_exit = signed_voluntary_exit.message
if voluntary_exit.validator_index >= state.validators.lenu64:
return err("Exit: invalid validator index")
let validator = unsafeAddr state.validators[voluntary_exit.validator_index]
# Verify the validator is active
if not is_active_validator(validator[], get_current_epoch(state)):
return err("Exit: validator not active")
# Verify exit has not been initiated
if validator[].exit_epoch != FAR_FUTURE_EPOCH:
return err("Exit: validator has exited")
# Exits must specify an epoch when they become valid; they are not valid
# before then
if not (get_current_epoch(state) >= voluntary_exit.epoch):
return err("Exit: exit epoch not passed")
# Verify the validator has been active long enough
if not (get_current_epoch(state) >= validator[].activation_epoch +
cfg.SHARD_COMMITTEE_PERIOD):
return err("Exit: not in validator set long enough")
# Verify signature
if skipBlsValidation notin flags:
let exitSignatureFork =
when typeof(state).kind >= ConsensusFork.Deneb:
Fork(
previous_version: cfg.CAPELLA_FORK_VERSION,
current_version: cfg.CAPELLA_FORK_VERSION,
epoch: cfg.CAPELLA_FORK_EPOCH)
else:
state.fork
if not verify_voluntary_exit_signature(
exitSignatureFork, state.genesis_validators_root, voluntary_exit,
validator[].pubkey, signed_voluntary_exit.signature):
return err("Exit: invalid signature")
# Checked above
ValidatorIndex.init(voluntary_exit.validator_index)
proc check_voluntary_exit*(
cfg: RuntimeConfig, state: ForkedHashedBeaconState,
signed_voluntary_exit: SomeSignedVoluntaryExit,
flags: UpdateFlags): Result[ValidatorIndex, cstring] =
withState(state):
check_voluntary_exit(cfg, forkyState.data, signed_voluntary_exit, flags)
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#voluntary-exits
proc process_voluntary_exit*(
cfg: RuntimeConfig,
state: var ForkyBeaconState,
signed_voluntary_exit: SomeSignedVoluntaryExit,
flags: UpdateFlags,
cache: var StateCache): Result[void, cstring] =
let exited_validator =
? check_voluntary_exit(cfg, state, signed_voluntary_exit, flags)
? initiate_validator_exit(cfg, state, exited_validator, cache)
ok()
proc process_bls_to_execution_change*(
cfg: RuntimeConfig, state: var (capella.BeaconState | deneb.BeaconState),
signed_address_change: SignedBLSToExecutionChange): Result[void, cstring] =
? check_bls_to_execution_change(
cfg.genesisFork, state, signed_address_change, {})
let address_change = signed_address_change.message
var withdrawal_credentials =
state.validators.item(address_change.validator_index).withdrawal_credentials
withdrawal_credentials.data[0] = ETH1_ADDRESS_WITHDRAWAL_PREFIX
withdrawal_credentials.data.fill(1, 11, 0)
withdrawal_credentials.data[12..31] =
address_change.to_execution_address.data
state.validators.mitem(address_change.validator_index).withdrawal_credentials =
withdrawal_credentials
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/phase0/beacon-chain.md#operations
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/capella/beacon-chain.md#modified-process_operations
proc process_operations(cfg: RuntimeConfig,
state: var ForkyBeaconState,
body: SomeForkyBeaconBlockBody,
base_reward_per_increment: Gwei,
flags: UpdateFlags,
cache: var StateCache): Result[void, cstring] =
# Verify that outstanding deposits are processed up to the maximum number of
# deposits
let
req_deposits = min(MAX_DEPOSITS,
state.eth1_data.deposit_count - state.eth1_deposit_index)
if state.eth1_data.deposit_count < state.eth1_deposit_index or
body.deposits.lenu64 != req_deposits:
return err("incorrect number of deposits")
for op in body.proposer_slashings:
? process_proposer_slashing(cfg, state, op, flags, cache)
for op in body.attester_slashings:
? process_attester_slashing(cfg, state, op, flags, cache)
for op in body.attestations:
? process_attestation(state, op, flags, base_reward_per_increment, cache)
for op in body.deposits:
? process_deposit(cfg, state, op, flags)
for op in body.voluntary_exits:
? process_voluntary_exit(cfg, state, op, flags, cache)
when typeof(body).kind >= ConsensusFork.Capella:
for op in body.bls_to_execution_changes:
? process_bls_to_execution_change(cfg, state, op)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/altair/beacon-chain.md#sync-aggregate-processing
func get_participant_reward*(total_active_balance: Gwei): Gwei =
let
total_active_increments =
total_active_balance div EFFECTIVE_BALANCE_INCREMENT
total_base_rewards =
get_base_reward_per_increment(total_active_balance) *
total_active_increments
max_participant_rewards =
total_base_rewards * SYNC_REWARD_WEIGHT div
WEIGHT_DENOMINATOR div SLOTS_PER_EPOCH
max_participant_rewards div SYNC_COMMITTEE_SIZE
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/altair/beacon-chain.md#sync-aggregate-processing
func get_proposer_reward*(participant_reward: Gwei): Gwei =
participant_reward * PROPOSER_WEIGHT div (WEIGHT_DENOMINATOR - PROPOSER_WEIGHT)
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/altair/beacon-chain.md#sync-aggregate-processing
proc process_sync_aggregate*(
state: var (altair.BeaconState | bellatrix.BeaconState |
capella.BeaconState | deneb.BeaconState),
sync_aggregate: SomeSyncAggregate, total_active_balance: Gwei,
flags: UpdateFlags,
cache: var StateCache):
Result[void, cstring] =
if strictVerification in flags and state.slot > 1.Slot:
template sync_committee_bits(): auto = sync_aggregate.sync_committee_bits
let num_active_participants = countOnes(sync_committee_bits).uint64
if num_active_participants * 3 < static(sync_committee_bits.len * 2):
fatal "Low sync committee participation",
slot = state.slot, num_active_participants
quit 1
# Verify sync committee aggregate signature signing over the previous slot
# block root
when sync_aggregate.sync_committee_signature isnot TrustedSig:
var participant_pubkeys: seq[ValidatorPubKey]
for i in 0 ..< state.current_sync_committee.pubkeys.len:
if sync_aggregate.sync_committee_bits[i]:
participant_pubkeys.add state.current_sync_committee.pubkeys.data[i]
# p2p-interface message validators check for empty sync committees, so it
# shouldn't run except as part of test suite.
if participant_pubkeys.len == 0:
if sync_aggregate.sync_committee_signature != ValidatorSig.infinity():
return err("process_sync_aggregate: empty sync aggregates need signature of point at infinity")
else:
# Empty participants allowed
let
previous_slot = max(state.slot, Slot(1)) - 1
beacon_block_root = get_block_root_at_slot(state, previous_slot)
if not verify_sync_committee_signature(
state.fork, state.genesis_validators_root, previous_slot,
beacon_block_root, participant_pubkeys,
sync_aggregate.sync_committee_signature):
return err("process_sync_aggregate: invalid signature")
# Compute participant and proposer rewards
let
participant_reward = get_participant_reward(total_active_balance)
proposer_reward = state_transition_block.get_proposer_reward(participant_reward)
proposer_index = get_beacon_proposer_index(state, cache).valueOr:
# We're processing a block, so this can't happen, in theory (!)
return err("process_sync_aggregate: no proposer")
# Apply participant and proposer rewards
let indices = get_sync_committee_cache(state, cache).current_sync_committee
# TODO could use a sequtils2 zipIt
for i in 0 ..< min(
state.current_sync_committee.pubkeys.len,
sync_aggregate.sync_committee_bits.len):
let participant_index = indices[i]
if sync_aggregate.sync_committee_bits[i]:
increase_balance(state, participant_index, participant_reward)
increase_balance(state, proposer_index, proposer_reward)
else:
decrease_balance(state, participant_index, participant_reward)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/bellatrix/beacon-chain.md#process_execution_payload
proc process_execution_payload*(
state: var bellatrix.BeaconState, payload: bellatrix.ExecutionPayload,
notify_new_payload: bellatrix.ExecutePayload): Result[void, cstring] =
# Verify consistency of the parent hash with respect to the previous
# execution payload header
if is_merge_transition_complete(state):
if not (payload.parent_hash ==
state.latest_execution_payload_header.block_hash):
return err("process_execution_payload: payload and state parent hash mismatch")
# Verify prev_randao
if not (payload.prev_randao == get_randao_mix(state, get_current_epoch(state))):
return err("process_execution_payload: payload and state randomness mismatch")
# Verify timestamp
if not (payload.timestamp == compute_timestamp_at_slot(state, state.slot)):
return err("process_execution_payload: invalid timestamp")
# Verify the execution payload is valid
if not notify_new_payload(payload):
return err("process_execution_payload: execution payload invalid")
# Cache execution payload header
state.latest_execution_payload_header = bellatrix.ExecutionPayloadHeader(
parent_hash: payload.parent_hash,
fee_recipient: payload.fee_recipient,
state_root: payload.state_root,
receipts_root: payload.receipts_root,
logs_bloom: payload.logs_bloom,
prev_randao: payload.prev_randao,
block_number: payload.block_number,
gas_limit: payload.gas_limit,
gas_used: payload.gas_used,
timestamp: payload.timestamp,
base_fee_per_gas: payload.base_fee_per_gas,
block_hash: payload.block_hash,
extra_data: payload.extra_data,
transactions_root: hash_tree_root(payload.transactions))
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/capella/beacon-chain.md#modified-process_execution_payload
proc process_execution_payload*(
state: var capella.BeaconState, payload: capella.ExecutionPayload,
notify_new_payload: capella.ExecutePayload): Result[void, cstring] =
# Verify consistency of the parent hash with respect to the previous
# execution payload header
if not (payload.parent_hash ==
state.latest_execution_payload_header.block_hash):
return err("process_execution_payload: payload and state parent hash mismatch")
# Verify prev_randao
if not (payload.prev_randao == get_randao_mix(state, get_current_epoch(state))):
return err("process_execution_payload: payload and state randomness mismatch")
# Verify timestamp
if not (payload.timestamp == compute_timestamp_at_slot(state, state.slot)):
return err("process_execution_payload: invalid timestamp")
# Verify the execution payload is valid
if not notify_new_payload(payload):
return err("process_execution_payload: execution payload invalid")
# Cache execution payload header
state.latest_execution_payload_header = capella.ExecutionPayloadHeader(
parent_hash: payload.parent_hash,
fee_recipient: payload.fee_recipient,
state_root: payload.state_root,
receipts_root: payload.receipts_root,
logs_bloom: payload.logs_bloom,
prev_randao: payload.prev_randao,
block_number: payload.block_number,
gas_limit: payload.gas_limit,
gas_used: payload.gas_used,
timestamp: payload.timestamp,
base_fee_per_gas: payload.base_fee_per_gas,
block_hash: payload.block_hash,
extra_data: payload.extra_data,
transactions_root: hash_tree_root(payload.transactions),
withdrawals_root: hash_tree_root(payload.withdrawals)) # [New in Capella]
ok()
# TODO workaround for https://github.com/nim-lang/Nim/issues/18095
# copy of datatypes/deneb.nim
type SomeDenebBeaconBlockBody =
deneb.BeaconBlockBody | deneb.SigVerifiedBeaconBlockBody |
deneb.TrustedBeaconBlockBody
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/deneb/beacon-chain.md#process_execution_payload
proc process_execution_payload*(
state: var deneb.BeaconState, body: SomeDenebBeaconBlockBody,
notify_new_payload: deneb.ExecutePayload): Result[void, cstring] =
template payload: auto = body.execution_payload
# Verify consistency of the parent hash with respect to the previous
# execution payload header
if not (payload.parent_hash ==
state.latest_execution_payload_header.block_hash):
return err("process_execution_payload: payload and state parent hash mismatch")
# Verify prev_randao
if not (payload.prev_randao == get_randao_mix(state, get_current_epoch(state))):
return err("process_execution_payload: payload and state randomness mismatch")
# Verify timestamp
if not (payload.timestamp == compute_timestamp_at_slot(state, state.slot)):
return err("process_execution_payload: invalid timestamp")
# [New in Deneb] Verify commitments are under limit
if not (lenu64(body.blob_kzg_commitments) <= MAX_BLOBS_PER_BLOCK):
return err("process_execution_payload: too many KZG commitments")
# Verify the execution payload is valid
if not notify_new_payload(payload):
return err("process_execution_payload: execution payload invalid")
# Cache execution payload header
state.latest_execution_payload_header = deneb.ExecutionPayloadHeader(
parent_hash: payload.parent_hash,
fee_recipient: payload.fee_recipient,
state_root: payload.state_root,
receipts_root: payload.receipts_root,
logs_bloom: payload.logs_bloom,
prev_randao: payload.prev_randao,
block_number: payload.block_number,
gas_limit: payload.gas_limit,
gas_used: payload.gas_used,
timestamp: payload.timestamp,
base_fee_per_gas: payload.base_fee_per_gas,
block_hash: payload.block_hash,
extra_data: payload.extra_data,
transactions_root: hash_tree_root(payload.transactions),
withdrawals_root: hash_tree_root(payload.withdrawals),
blob_gas_used: payload.blob_gas_used, # [New in Deneb]
excess_blob_gas: payload.excess_blob_gas) # [New in Deneb]
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/capella/beacon-chain.md#new-process_withdrawals
func process_withdrawals*(
state: var (capella.BeaconState | deneb.BeaconState),
payload: capella.ExecutionPayload | deneb.ExecutionPayload):
Result[void, cstring] =
let expected_withdrawals = get_expected_withdrawals(state)
if not (len(payload.withdrawals) == len(expected_withdrawals)):
return err("process_withdrawals: different numbers of payload and expected withdrawals")
for i in 0 ..< len(expected_withdrawals):
if expected_withdrawals[i] != payload.withdrawals[i]:
return err("process_withdrawals: mismatched expected and payload withdrawal")
let validator_index =
ValidatorIndex.init(expected_withdrawals[i].validator_index).valueOr:
return err("process_withdrawals: invalid validator index")
decrease_balance(
state, validator_index, expected_withdrawals[i].amount)
# Update the next withdrawal index if this block contained withdrawals
if len(expected_withdrawals) != 0:
let latest_withdrawal = expected_withdrawals[^1]
state.next_withdrawal_index = WithdrawalIndex(latest_withdrawal.index + 1)
# Update the next validator index to start the next withdrawal sweep
if len(expected_withdrawals) == MAX_WITHDRAWALS_PER_PAYLOAD:
# Next sweep starts after the latest withdrawal's validator index
let next_validator_index =
(expected_withdrawals[^1].validator_index + 1) mod
lenu64(state.validators)
state.next_withdrawal_validator_index = next_validator_index
else:
# Advance sweep by the max length of the sweep if there was not a full set
# of withdrawals
let next_index =
state.next_withdrawal_validator_index +
MAX_VALIDATORS_PER_WITHDRAWALS_SWEEP
let next_validator_index = next_index mod lenu64(state.validators)
state.next_withdrawal_validator_index = next_validator_index
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.1/specs/deneb/beacon-chain.md#kzg_commitment_to_versioned_hash
func kzg_commitment_to_versioned_hash*(
kzg_commitment: KzgCommitment): VersionedHash =
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.3/specs/deneb/beacon-chain.md#blob
const VERSIONED_HASH_VERSION_KZG = 0x01'u8
var res: VersionedHash
res[0] = VERSIONED_HASH_VERSION_KZG
res[1 .. 31] = eth2digest(kzg_commitment).data.toOpenArray(1, 31)
res
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/deneb/fork-choice.md#validate_blobs
proc validate_blobs*(expected_kzg_commitments: seq[KzgCommitment],
blobs: seq[KzgBlob],
proofs: seq[KzgProof]):
Result[void, cstring] =
if expected_kzg_commitments.len != blobs.len:
return err("validate_blobs: different commitment and blob lengths")
if proofs.len != blobs.len:
return err("validate_blobs: different proof and blob lengths")
let res = verifyProofs(blobs, expected_kzg_commitments, proofs).valueOr:
return err("validate_blobs: proof verification error")
if not res:
return err("validate_blobs: proof verification failed")
ok()
# TODO workaround for https://github.com/nim-lang/Nim/issues/18095
# copy of datatypes/phase0.nim
type SomePhase0Block =
phase0.BeaconBlock | phase0.SigVerifiedBeaconBlock | phase0.TrustedBeaconBlock
proc process_block*(
cfg: RuntimeConfig,
state: var phase0.BeaconState, blck: SomePhase0Block, flags: UpdateFlags,
cache: var StateCache): Result[void, cstring]=
## When there's a new block, we need to verify that the block is sane and
## update the state accordingly - the state is left in an unknown state when
## block application fails (!)
? process_block_header(state, blck, flags, cache)
? process_randao(state, blck.body, flags, cache)
? process_eth1_data(state, blck.body)
? process_operations(cfg, state, blck.body, 0.Gwei, flags, cache)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/altair/beacon-chain.md#block-processing
# TODO workaround for https://github.com/nim-lang/Nim/issues/18095
# copy of datatypes/altair.nim
type SomeAltairBlock =
altair.BeaconBlock | altair.SigVerifiedBeaconBlock | altair.TrustedBeaconBlock
proc process_block*(
cfg: RuntimeConfig,
state: var altair.BeaconState, blck: SomeAltairBlock, flags: UpdateFlags,
cache: var StateCache): Result[void, cstring]=
## When there's a new block, we need to verify that the block is sane and
## update the state accordingly - the state is left in an unknown state when
## block application fails (!)
? process_block_header(state, blck, flags, cache)
? process_randao(state, blck.body, flags, cache)
? process_eth1_data(state, blck.body)
let
total_active_balance = get_total_active_balance(state, cache)
base_reward_per_increment =
get_base_reward_per_increment(total_active_balance)
? process_operations(
cfg, state, blck.body, base_reward_per_increment, flags, cache)
? process_sync_aggregate(
state, blck.body.sync_aggregate, total_active_balance,
flags, cache) # [New in Altair]
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/specs/bellatrix/beacon-chain.md#block-processing
# TODO workaround for https://github.com/nim-lang/Nim/issues/18095
type SomeBellatrixBlock =
bellatrix.BeaconBlock | bellatrix.SigVerifiedBeaconBlock | bellatrix.TrustedBeaconBlock
proc process_block*(
cfg: RuntimeConfig,
state: var bellatrix.BeaconState, blck: SomeBellatrixBlock,
flags: UpdateFlags, cache: var StateCache): Result[void, cstring]=
## When there's a new block, we need to verify that the block is sane and
## update the state accordingly - the state is left in an unknown state when
## block application fails (!)
? process_block_header(state, blck, flags, cache)
if is_execution_enabled(state, blck.body):
? process_execution_payload(
state, blck.body.execution_payload,
func(_: bellatrix.ExecutionPayload): bool = true) # [New in Bellatrix]
? process_randao(state, blck.body, flags, cache)
? process_eth1_data(state, blck.body)
let
total_active_balance = get_total_active_balance(state, cache)
base_reward_per_increment =
get_base_reward_per_increment(total_active_balance)
? process_operations(
cfg, state, blck.body, base_reward_per_increment, flags, cache)
? process_sync_aggregate(
state, blck.body.sync_aggregate, total_active_balance, flags, cache)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/capella/beacon-chain.md#block-processing
# TODO workaround for https://github.com/nim-lang/Nim/issues/18095
type SomeCapellaBlock =
capella.BeaconBlock | capella.SigVerifiedBeaconBlock | capella.TrustedBeaconBlock
proc process_block*(
cfg: RuntimeConfig,
state: var capella.BeaconState, blck: SomeCapellaBlock,
flags: UpdateFlags, cache: var StateCache): Result[void, cstring]=
## When there's a new block, we need to verify that the block is sane and
## update the state accordingly - the state is left in an unknown state when
## block application fails (!)
? process_block_header(state, blck, flags, cache)
# Consensus specs v1.4.0 unconditionally assume is_execution_enabled is
# true, but intentionally keep such a check.
if is_execution_enabled(state, blck.body):
? process_withdrawals(
state, blck.body.execution_payload) # [New in Capella]
? process_execution_payload(
state, blck.body.execution_payload,
func(_: capella.ExecutionPayload): bool = true) # [Modified in Capella]
? process_randao(state, blck.body, flags, cache)
? process_eth1_data(state, blck.body)
let
total_active_balance = get_total_active_balance(state, cache)
base_reward_per_increment =
get_base_reward_per_increment(total_active_balance)
? process_operations(
cfg, state, blck.body, base_reward_per_increment,
flags, cache) # [Modified in Capella]
? process_sync_aggregate(
state, blck.body.sync_aggregate, total_active_balance, flags, cache)
ok()
# https://github.com/ethereum/consensus-specs/blob/v1.3.0/specs/deneb/beacon-chain.md#block-processing
# TODO workaround for https://github.com/nim-lang/Nim/issues/18095
type SomeDenebBlock =
deneb.BeaconBlock | deneb.SigVerifiedBeaconBlock | deneb.TrustedBeaconBlock
proc process_block*(
cfg: RuntimeConfig,
state: var deneb.BeaconState, blck: SomeDenebBlock,
flags: UpdateFlags, cache: var StateCache): Result[void, cstring]=
## When there's a new block, we need to verify that the block is sane and
## update the state accordingly - the state is left in an unknown state when
## block application fails (!)
? process_block_header(state, blck, flags, cache)
# Consensus specs v1.4.0 unconditionally assume is_execution_enabled is
# true, but intentionally keep such a check.
if is_execution_enabled(state, blck.body):
? process_withdrawals(state, blck.body.execution_payload)
? process_execution_payload(
state, blck.body,
func(_: deneb.ExecutionPayload): bool = true) # [Modified in Deneb]
? process_randao(state, blck.body, flags, cache)
? process_eth1_data(state, blck.body)
let
total_active_balance = get_total_active_balance(state, cache)
base_reward_per_increment =
get_base_reward_per_increment(total_active_balance)
? process_operations(
cfg, state, blck.body, base_reward_per_increment, flags, cache)
? process_sync_aggregate(
state, blck.body.sync_aggregate, total_active_balance, flags, cache)
ok()