The core protocol of WoopChain
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
woop/consensus/view_change.go

716 lines
24 KiB

package consensus
import (
"bytes"
"encoding/binary"
"encoding/hex"
"sync"
"time"
"github.com/harmony-one/harmony/crypto/bls"
"github.com/ethereum/go-ethereum/rlp"
"github.com/harmony-one/harmony/core/types"
"github.com/ethereum/go-ethereum/common"
bls_core "github.com/harmony-one/bls/ffi/go/bls"
msg_pb "github.com/harmony-one/harmony/api/proto/message"
"github.com/harmony-one/harmony/consensus/quorum"
"github.com/harmony-one/harmony/consensus/signature"
bls_cosi "github.com/harmony-one/harmony/crypto/bls"
nodeconfig "github.com/harmony-one/harmony/internal/configs/node"
"github.com/harmony-one/harmony/internal/utils"
"github.com/harmony-one/harmony/p2p"
)
// MaxViewIDDiff limits the received view ID to only 100 further from the current view ID
const MaxViewIDDiff = 100
// State contains current mode and current viewID
type State struct {
mode Mode
modeMux sync.RWMutex
// current view id in normal mode
// it changes per successful consensus
curViewID uint64
cViewMux sync.RWMutex
// view changing id is used during view change mode
// it is the next view id
viewChangingID uint64
viewMux sync.RWMutex
}
// Mode return the current node mode
func (pm *State) Mode() Mode {
pm.modeMux.RLock()
defer pm.modeMux.RUnlock()
return pm.mode
}
// SetMode set the node mode as required
func (pm *State) SetMode(s Mode) {
pm.modeMux.Lock()
defer pm.modeMux.Unlock()
pm.mode = s
}
// GetCurViewID return the current view id
func (pm *State) GetCurViewID() uint64 {
pm.cViewMux.RLock()
defer pm.cViewMux.RUnlock()
return pm.curViewID
}
// SetCurViewID sets the current view id
func (pm *State) SetCurViewID(viewID uint64) {
pm.cViewMux.Lock()
defer pm.cViewMux.Unlock()
pm.curViewID = viewID
}
// GetViewChangingID return the current view changing id
// It is meaningful during view change mode
func (pm *State) GetViewChangingID() uint64 {
pm.viewMux.RLock()
defer pm.viewMux.RUnlock()
return pm.viewChangingID
}
// SetViewChangingID set the current view changing id
// It is meaningful during view change mode
func (pm *State) SetViewChangingID(id uint64) {
pm.viewMux.Lock()
defer pm.viewMux.Unlock()
pm.viewChangingID = id
}
// GetViewChangeDuraion return the duration of the current view change
// It increase in the power of difference betweeen view changing ID and current view ID
func (pm *State) GetViewChangeDuraion() time.Duration {
pm.viewMux.RLock()
pm.cViewMux.RLock()
defer pm.viewMux.RUnlock()
defer pm.cViewMux.RUnlock()
diff := int64(pm.viewChangingID - pm.curViewID)
return time.Duration(diff * diff * int64(viewChangeDuration))
}
// switchPhase will switch FBFTPhase to nextPhase if the desirePhase equals the nextPhase
func (consensus *Consensus) switchPhase(desired FBFTPhase, override bool) {
if override {
consensus.phase = desired
return
}
var nextPhase FBFTPhase
switch consensus.phase {
case FBFTAnnounce:
nextPhase = FBFTPrepare
case FBFTPrepare:
nextPhase = FBFTCommit
case FBFTCommit:
nextPhase = FBFTAnnounce
}
if nextPhase == desired {
consensus.phase = nextPhase
}
}
// GetNextLeaderKey uniquely determine who is the leader for given viewID
func (consensus *Consensus) GetNextLeaderKey() *bls.PublicKeyWrapper {
wasFound, next := consensus.Decider.NextAfter(consensus.LeaderPubKey)
if !wasFound {
consensus.getLogger().Warn().
Str("key", consensus.LeaderPubKey.Bytes.Hex()).
Msg("GetNextLeaderKey: currentLeaderKey not found")
}
return next
}
// ResetViewChangeState reset the state for viewchange
func (consensus *Consensus) ResetViewChangeState() {
consensus.getLogger().Debug().
Str("Phase", consensus.phase.String()).
Msg("[ResetViewChangeState] Resetting view change state")
consensus.current.SetMode(Normal)
consensus.m1Payload = []byte{}
consensus.bhpSigs = map[uint64]map[string]*bls_core.Sign{}
consensus.nilSigs = map[uint64]map[string]*bls_core.Sign{}
consensus.viewIDSigs = map[uint64]map[string]*bls_core.Sign{}
5 years ago
consensus.bhpBitmap = map[uint64]*bls_cosi.Mask{}
consensus.nilBitmap = map[uint64]*bls_cosi.Mask{}
consensus.viewIDBitmap = map[uint64]*bls_cosi.Mask{}
consensus.Decider.ResetViewChangeVotes()
}
func createTimeout() map[TimeoutType]*utils.Timeout {
timeouts := make(map[TimeoutType]*utils.Timeout)
timeouts[timeoutConsensus] = utils.NewTimeout(phaseDuration)
timeouts[timeoutViewChange] = utils.NewTimeout(viewChangeDuration)
timeouts[timeoutBootstrap] = utils.NewTimeout(bootstrapDuration)
return timeouts
}
// startViewChange send a new view change
func (consensus *Consensus) startViewChange(viewID uint64) {
if consensus.disableViewChange {
return
}
consensus.consensusTimeout[timeoutConsensus].Stop()
consensus.consensusTimeout[timeoutBootstrap].Stop()
consensus.current.SetMode(ViewChanging)
consensus.SetViewChangingID(viewID)
consensus.LeaderPubKey = consensus.GetNextLeaderKey()
duration := consensus.current.GetViewChangeDuraion()
consensus.getLogger().Warn().
Uint64("viewChangingID", consensus.GetViewChangingID()).
Dur("timeoutDuration", duration).
Str("NextLeader", consensus.LeaderPubKey.Bytes.Hex()).
Msg("[startViewChange]")
for _, key := range consensus.priKey {
if !consensus.IsValidatorInCommittee(key.Pub.Bytes) {
continue
}
msgToSend := consensus.constructViewChangeMessage(&key)
consensus.host.SendMessageToGroups([]nodeconfig.GroupID{
nodeconfig.NewGroupIDByShardID(nodeconfig.ShardID(consensus.ShardID)),
},
p2p.ConstructMessage(msgToSend),
)
}
consensus.consensusTimeout[timeoutViewChange].SetDuration(duration)
consensus.consensusTimeout[timeoutViewChange].Start()
}
func (consensus *Consensus) onViewChange(msg *msg_pb.Message) {
recvMsg, err := ParseViewChangeMessage(msg)
if err != nil {
consensus.getLogger().Warn().Msg("[onViewChange] Unable To Parse Viewchange Message")
return
}
// if not leader, noop
newLeaderKey := recvMsg.LeaderPubkey
newLeaderPriKey, err := consensus.GetLeaderPrivateKey(newLeaderKey.Object)
if err != nil {
return
}
if consensus.Decider.IsQuorumAchieved(quorum.ViewChange) {
consensus.getLogger().Info().
Int64("have", consensus.Decider.SignersCount(quorum.ViewChange)).
Int64("need", consensus.Decider.TwoThirdsSignersCount()).
Interface("validatorPubKeys", recvMsg.SenderPubkeys).
Msg("[onViewChange] Received Enough View Change Messages")
return
}
if !consensus.onViewChangeSanityCheck(recvMsg) {
return
}
// already checked the length of SenderPubkeys in onViewChangeSanityCheck
senderKey := recvMsg.SenderPubkeys[0]
consensus.vcLock.Lock()
defer consensus.vcLock.Unlock()
5 years ago
// update the dictionary key if the viewID is first time received
consensus.addViewIDKeyIfNotExist(recvMsg.ViewID)
// TODO: remove NIL type message
// add self m1 or m2 type message signature and bitmap
_, ok1 := consensus.nilSigs[recvMsg.ViewID][newLeaderKey.Bytes.Hex()]
_, ok2 := consensus.bhpSigs[recvMsg.ViewID][newLeaderKey.Bytes.Hex()]
if !(ok1 || ok2) {
// add own signature for newview message
preparedMsgs := consensus.FBFTLog.GetMessagesByTypeSeq(
msg_pb.MessageType_PREPARED, recvMsg.BlockNum,
)
preparedMsg := consensus.FBFTLog.FindMessageByMaxViewID(preparedMsgs)
hasBlock := false
if preparedMsg != nil {
if preparedBlock := consensus.FBFTLog.GetBlockByHash(
preparedMsg.BlockHash,
); preparedBlock != nil {
if consensus.BlockVerifier(preparedBlock); err != nil {
consensus.getLogger().Error().Err(err).Msg("[onViewChange] My own prepared block verification failed")
} else {
hasBlock = true
}
}
}
if hasBlock {
consensus.getLogger().Info().Msg("[onViewChange] add my M1 type messaage")
msgToSign := append(preparedMsg.BlockHash[:], preparedMsg.Payload...)
for i, key := range consensus.priKey {
if err := consensus.bhpBitmap[recvMsg.ViewID].SetKey(key.Pub.Bytes, true); err != nil {
consensus.getLogger().Warn().Msgf("[onViewChange] bhpBitmap setkey failed for key at index %d", i)
continue
}
consensus.bhpSigs[recvMsg.ViewID][key.Pub.Bytes.Hex()] = key.Pri.SignHash(msgToSign)
}
// if m1Payload is empty, we just add one
if len(consensus.m1Payload) == 0 {
consensus.m1Payload = append(preparedMsg.BlockHash[:], preparedMsg.Payload...)
}
} else {
consensus.getLogger().Info().Msg("[onViewChange] add my M2(NIL) type messaage")
for i, key := range consensus.priKey {
if err := consensus.nilBitmap[recvMsg.ViewID].SetKey(key.Pub.Bytes, true); err != nil {
consensus.getLogger().Warn().Msgf("[onViewChange] nilBitmap setkey failed for key at index %d", i)
continue
}
consensus.nilSigs[recvMsg.ViewID][key.Pub.Bytes.Hex()] = key.Pri.SignHash(NIL)
}
}
}
// add self m3 type message signature and bitmap
_, ok3 := consensus.viewIDSigs[recvMsg.ViewID][newLeaderKey.Bytes.Hex()]
5 years ago
if !ok3 {
viewIDBytes := make([]byte, 8)
binary.LittleEndian.PutUint64(viewIDBytes, recvMsg.ViewID)
for i, key := range consensus.priKey {
if err := consensus.viewIDBitmap[recvMsg.ViewID].SetKey(key.Pub.Bytes, true); err != nil {
consensus.getLogger().Warn().Msgf("[onViewChange] viewIDBitmap setkey failed for key at index %d", i)
continue
}
consensus.viewIDSigs[recvMsg.ViewID][key.Pub.Bytes.Hex()] = key.Pri.SignHash(viewIDBytes)
}
}
preparedBlock := &types.Block{}
hasBlock := false
if len(recvMsg.Payload) != 0 && len(recvMsg.Block) != 0 {
if err := rlp.DecodeBytes(recvMsg.Block, preparedBlock); err != nil {
consensus.getLogger().Warn().
Err(err).
Uint64("MsgBlockNum", recvMsg.BlockNum).
Msg("[onViewChange] Unparseable prepared block data")
return
}
hasBlock = true
}
// m2 type message
if !hasBlock {
_, ok := consensus.nilSigs[recvMsg.ViewID][senderKey.Bytes.Hex()]
if ok {
consensus.getLogger().Debug().
Str("validatorPubKey", senderKey.Bytes.Hex()).
Msg("[onViewChange] Already Received M2 message from validator")
return
}
if !recvMsg.ViewchangeSig.VerifyHash(senderKey.Object, NIL) {
consensus.getLogger().Warn().Msg("[onViewChange] Failed To Verify Signature For M2 Type Viewchange Message")
return
}
consensus.getLogger().Info().
Str("validatorPubKey", senderKey.Bytes.Hex()).
Msg("[onViewChange] Add M2 (NIL) type message")
consensus.nilSigs[recvMsg.ViewID][senderKey.Bytes.Hex()] = recvMsg.ViewchangeSig
consensus.nilBitmap[recvMsg.ViewID].SetKey(senderKey.Bytes, true) // Set the bitmap indicating that this validator signed.
} else { // m1 type message
if consensus.BlockVerifier(preparedBlock); err != nil {
consensus.getLogger().Error().Err(err).Msg("[onViewChange] Prepared block verification failed")
return
}
_, ok := consensus.bhpSigs[recvMsg.ViewID][senderKey.Bytes.Hex()]
if ok {
consensus.getLogger().Debug().
Str("validatorPubKey", senderKey.Bytes.Hex()).
Msg("[onViewChange] Already Received M1 Message From the Validator")
return
}
if !recvMsg.ViewchangeSig.VerifyHash(senderKey.Object, recvMsg.Payload) {
consensus.getLogger().Warn().Msg("[onViewChange] Failed to Verify Signature for M1 Type Viewchange Message")
return
}
// first time receive m1 type message, need verify validity of prepared message
if len(consensus.m1Payload) == 0 || !bytes.Equal(consensus.m1Payload, recvMsg.Payload) {
if len(recvMsg.Payload) <= 32 {
consensus.getLogger().Warn().
Int("len", len(recvMsg.Payload)).
Msg("[onViewChange] M1 RecvMsg Payload Not Enough Length")
return
}
blockHash := recvMsg.Payload[:32]
aggSig, mask, err := consensus.ReadSignatureBitmapPayload(recvMsg.Payload, 32)
if err != nil {
consensus.getLogger().Error().Err(err).Msg("[onViewChange] M1 RecvMsg Payload Read Error")
return
}
if !consensus.Decider.IsQuorumAchievedByMask(mask) {
consensus.getLogger().Warn().
Msgf("[onViewChange] Quorum Not achieved")
return
}
// Verify the multi-sig for prepare phase
if !aggSig.VerifyHash(mask.AggregatePublic, blockHash[:]) {
consensus.getLogger().Warn().
Hex("blockHash", blockHash).
Msg("[onViewChange] failed to verify multi signature for m1 prepared payload")
return
}
// if m1Payload is empty, we just add one
if len(consensus.m1Payload) == 0 {
consensus.m1Payload = append(recvMsg.Payload[:0:0], recvMsg.Payload...)
// create prepared message for new leader
preparedMsg := FBFTMessage{
MessageType: msg_pb.MessageType_PREPARED,
ViewID: recvMsg.ViewID,
BlockNum: recvMsg.BlockNum,
}
preparedMsg.BlockHash = common.Hash{}
copy(preparedMsg.BlockHash[:], recvMsg.Payload[:32])
preparedMsg.Payload = make([]byte, len(recvMsg.Payload)-32)
copy(preparedMsg.Payload[:], recvMsg.Payload[32:])
preparedMsg.SenderPubkeys = []*bls.PublicKeyWrapper{newLeaderKey}
consensus.getLogger().Info().Msg("[onViewChange] New Leader Prepared Message Added")
consensus.FBFTLog.AddMessage(&preparedMsg)
consensus.FBFTLog.AddBlock(preparedBlock)
}
}
consensus.getLogger().Info().
Str("validatorPubKey", senderKey.Bytes.Hex()).
Msg("[onViewChange] Add M1 (prepared) type message")
consensus.bhpSigs[recvMsg.ViewID][senderKey.Bytes.Hex()] = recvMsg.ViewchangeSig
consensus.bhpBitmap[recvMsg.ViewID].SetKey(senderKey.Bytes, true) // Set the bitmap indicating that this validator signed.
}
// check and add viewID (m3 type) message signature
if _, ok := consensus.viewIDSigs[recvMsg.ViewID][senderKey.Bytes.Hex()]; ok {
consensus.getLogger().Debug().
Str("validatorPubKey", senderKey.Bytes.Hex()).
Msg("[onViewChange] Already Received M3(ViewID) message from the validator")
return
}
viewIDHash := make([]byte, 8)
binary.LittleEndian.PutUint64(viewIDHash, recvMsg.ViewID)
if !recvMsg.ViewidSig.VerifyHash(senderKey.Object, viewIDHash) {
consensus.getLogger().Warn().
Uint64("MsgViewID", recvMsg.ViewID).
Msg("[onViewChange] Failed to Verify M3 Message Signature")
return
}
consensus.getLogger().Info().
Str("validatorPubKey", senderKey.Bytes.Hex()).
Msg("[onViewChange] Add M3 (ViewID) type message")
5 years ago
consensus.viewIDSigs[recvMsg.ViewID][senderKey.Bytes.Hex()] = recvMsg.ViewidSig
// Set the bitmap indicating that this validator signed.
consensus.viewIDBitmap[recvMsg.ViewID].SetKey(senderKey.Bytes, true)
consensus.getLogger().Info().
5 years ago
Int("have", len(consensus.viewIDSigs[recvMsg.ViewID])).
Int64("total", consensus.Decider.ParticipantsCount()).
Msg("[onViewChange]")
// received enough view change messages, change state to normal consensus
if consensus.Decider.IsQuorumAchievedByMask(consensus.viewIDBitmap[recvMsg.ViewID]) {
consensus.current.SetMode(Normal)
consensus.LeaderPubKey = newLeaderKey
consensus.ResetState()
if len(consensus.m1Payload) == 0 {
// TODO(Chao): explain why ReadySignal is sent only in this case but not the other case.
// Make sure the newly proposed block have the correct view ID
consensus.SetCurViewID(recvMsg.ViewID)
go func() {
consensus.ReadySignal <- struct{}{}
}()
} else {
consensus.getLogger().Info().
Str("From", consensus.phase.String()).
Str("To", FBFTCommit.String()).
Msg("[OnViewChange] Switching phase")
consensus.switchPhase(FBFTCommit, true)
copy(consensus.blockHash[:], consensus.m1Payload[:32])
aggSig, mask, err := consensus.ReadSignatureBitmapPayload(consensus.m1Payload, 32)
if err != nil {
consensus.getLogger().Error().Err(err).
Msg("[onViewChange] ReadSignatureBitmapPayload Fail")
return
}
consensus.aggregatedPrepareSig = aggSig
consensus.prepareBitmap = mask
// Leader sign and add commit message
block := consensus.FBFTLog.GetBlockByHash(consensus.blockHash)
if block == nil {
consensus.getLogger().Warn().Msg("[onViewChange] failed to get prepared block for self commit")
return
}
commitPayload := signature.ConstructCommitPayload(consensus.ChainReader,
block.Epoch(), block.Hash(), block.NumberU64(), block.Header().ViewID().Uint64())
for i, key := range consensus.priKey {
if err := consensus.commitBitmap.SetKey(key.Pub.Bytes, true); err != nil {
consensus.getLogger().Warn().
Msgf("[OnViewChange] New Leader commit bitmap set failed for key at index %d", i)
continue
}
if _, err := consensus.Decider.SubmitVote(
quorum.Commit,
[]bls.SerializedPublicKey{key.Pub.Bytes},
key.Pri.SignHash(commitPayload),
common.BytesToHash(consensus.blockHash[:]),
block.NumberU64(),
block.Header().ViewID().Uint64(),
); err != nil {
consensus.getLogger().Warn().Msg("submit vote on viewchange commit failed")
return
}
}
}
consensus.SetCurViewID(recvMsg.ViewID)
[staking][validation][protocol] (#2396) * [staking][validation][protocol] Limit max bls keys * [staking-era] Fold banned and active into single field * [slash][effective] Remove LRU cache for slash, change .Active to enumeration * [slash] Remove leftover wrong usage of Logger * [slash][offchain] Only Decode if len > 0 * [offchain] cosmetic * [slash] Remove some logs in proposal * [webhook] Move webhook with call for when cannot commit block * [shard] Finally make finding subcommittee by shardID an explicit error * [node] Whitespace, prefer literal * [webhook] Report bad block to webhook * [slash] Expand verify, remove bad log usage, explicit error handle * [slash] Check on key size * [slash] Explicit upper bound of pending slashes * [slash] Use right epoch snapshot, fail to verify if epoch wrong on beaconchain * [multibls] Make max count allowed be 1/3 of external slots * [quorum] Remove bad API of ShardIDProvider, factor out committee key as method of committee * [verify] Begin factor out of common verification approach * [project] Further remove RawJSON log, use proper epoch for snapshot * [slash] Implement verification * [slash] Implement BLS key verification of ballots * [rpc] Keep validator information as meaningful as possible * [staking] Never can stop being banned * [slash] Comments and default Unknown case of eligibility * [slash] Be explicit on what input values allowed when want to change EPOSStatus * [consensus] Remove unneeded TODO * [verify] Add proper error message * [rpc] Give back to caller their wrong chain id * [chain] Add extra map dump of delegation sizing for downstream analysis * [engine] Less code, more methods * [offchain] More leniency in handling slash bytes and delete from pending * [validator] Remove errors on bad input for edit
5 years ago
msgToSend := consensus.constructNewViewMessage(
recvMsg.ViewID, newLeaderPriKey,
[staking][validation][protocol] (#2396) * [staking][validation][protocol] Limit max bls keys * [staking-era] Fold banned and active into single field * [slash][effective] Remove LRU cache for slash, change .Active to enumeration * [slash] Remove leftover wrong usage of Logger * [slash][offchain] Only Decode if len > 0 * [offchain] cosmetic * [slash] Remove some logs in proposal * [webhook] Move webhook with call for when cannot commit block * [shard] Finally make finding subcommittee by shardID an explicit error * [node] Whitespace, prefer literal * [webhook] Report bad block to webhook * [slash] Expand verify, remove bad log usage, explicit error handle * [slash] Check on key size * [slash] Explicit upper bound of pending slashes * [slash] Use right epoch snapshot, fail to verify if epoch wrong on beaconchain * [multibls] Make max count allowed be 1/3 of external slots * [quorum] Remove bad API of ShardIDProvider, factor out committee key as method of committee * [verify] Begin factor out of common verification approach * [project] Further remove RawJSON log, use proper epoch for snapshot * [slash] Implement verification * [slash] Implement BLS key verification of ballots * [rpc] Keep validator information as meaningful as possible * [staking] Never can stop being banned * [slash] Comments and default Unknown case of eligibility * [slash] Be explicit on what input values allowed when want to change EPOSStatus * [consensus] Remove unneeded TODO * [verify] Add proper error message * [rpc] Give back to caller their wrong chain id * [chain] Add extra map dump of delegation sizing for downstream analysis * [engine] Less code, more methods * [offchain] More leniency in handling slash bytes and delete from pending * [validator] Remove errors on bad input for edit
5 years ago
)
5 years ago
consensus.getLogger().Warn().
Int("payloadSize", len(consensus.m1Payload)).
Hex("M1Payload", consensus.m1Payload).
Msg("[onViewChange] Sent NewView Message")
[double-sign] Provide proof of double sign in slash record sent to beaconchain (#2253) * [double-sign] Commit changes in consensus needed for double-sign * [double-sign] Leader captures when valdator double signs, broadcasts to beaconchain * [slash] Add quick iteration tool for testing double-signing * [slash] Add webhook example * [slash] Add http server for hook to trigger double sign behavior * [double-sign] Use bin/trigger-double-sign to cause a double-sign * [double-sign] Full feedback loop working * [slash] Thread through the slash records in the block proposal step * [slash] Compute the slashing rate * [double-sign] Generalize yaml malicious for many keys * [double-sign][slash] Modify data structures, verify via webhook handler * [slash][double-sign] Find one address of bls public key signer, seemingly settle on data structures * [slash] Apply to state slashing for double signing * [slash][double-sign] Checkpoint for working code that slashes on beaconchain * [slash] Keep track of the total slash and total reporters reward * [slash] Dump account state before and after the slash * [slash] Satisfy Travis * [slash][state] Apply slash to the snapshot at beginning of epoch, now need to capture also the new delegates * [slash] Capture the unique new delegations since snapshot as well * [slash] Filter undelegation by epoch of double sign * [slash] Add TODO of correctness needed in slash needs on off-chain data * [rpc] Fix closure issue on shardID * [slash] Add delegator to double-sign testing script * [slash] Expand crt-validator.sh with commenting printfs and make delegation * [slash] Finish track payment of leftover slash debt after undelegation runs out * [slash] Now be explicit about error wrt delegatorSlashApply * [slash] Capture specific sanity check on slash paidoff * [slash] Track slash from undelegation piecemeal * [slash][delegation] Named slice types, .String() * [slash] Do no RLP encode twice, once is enough * [slash] Remove special case of validators own delegation * [slash] Refactor approach to slash state application * [slash] Begin expanding out Verify * [slash] Slash on snapshot delegations, not current * [slash] Fix Epoch Cmp * [slash] Third iteration on slash logic * [slash] Use full slash amount * [slash] More log, whitespace * [slash] Remove Println, add log * [slash] Remove debug Println * [slash] Add record in unit test * [slash] Build Validator snapshot, current. Fill out slash record * [slash] Need to get RLP dump of a header to use in test * [slash] Factor out double sign test constants * [slash] Factor out common for validator, stub out slash application, finish out deserialization setup * [slash] Factor out data structure creation because of var lexical scoping * [slash] Seem to have pipeline of unit test e2e executing * [slash] Add expected snitch, slash amounts * [slash] Checkpoint * [slash] Unit test correctly checks case of validator own stake which could drop below 1 ONE in slashing * [config] add double-sign testnet config (#1) Signed-off-by: Leo Chen <leo@harmony.one> * [slash] Commit for as is code & data of current dump.json * [slash] Order of state operation not correct in test, hence bad results, thank you dlv * [slash] Add snapshot state dump * [slash] Pay off slash of validator own delegation correctly * [slash] Pay off slash debt with special case for min-self * [slash] Pass first scenario conclusively * [slash] 2% slash passes unit test for own delegation and external * [slash] Parameterize unit test to easily test .02 vs .80 slash * [slash] Handle own delegation correctly at 80% slash * [slash] Have 80% slash working with external delegator * [slash] Remove debug code from slash * [slash] Adjust Apply signature, test again for 2% slash * [slash] Factor out scenario in testing so can test 2% and 80% at same time * [slash] Correct balance deduction on plan delegation * [slash] Mock out ChainReader for TestVerify * [slash] Small surface area interface, now feedback loop for verify * [slash] Remove development json * [slash] trigger-double-sign consumes yaml * [slash] Remove dead code * [slash][test] Factor ValidatorWrapper into scenario * [slash][test] Add example from local-testing dump - caution might be off * [slash] Factor out mutation of slashDebt * [slash][test] Factor out tests so can easily load test-case from bytes * [slash] Fix payment mistake in validator own delegation wrt min-self-delgation respected * [slash] Satisfy Travis * [slash] Begin cleanup of PR * [slash] Apply slash from header to Finalize via state processor * [slash] Productionize code, Println => logs; adjust slash picked in newblock * [slash] Need pointer for rlp.Decode * [slash] ValidatorInformation use full wrapper * Fix median stake * [staking] Adjust MarshalJSON for Validator, Wrapper * Refactor offchain data commit; Make block onchain/offchain commit atomic (#2279) * Refactor offchain data; Add epoch to ValidatorSnapshot * Make block onchain/offchain data commit atomically * [slash][committee] Set .Active to false on double sign, do not consider banned or inactive for committee assignment * [effective] VC eligible.go * [consensus] Redundant field in printf * [docker] import-ks for a dev account * [slash] Create BLS key for dockerfile and crt-validator.sh * [slash][docker] Easy deployment of double-sign testing * [docker] Have slash work as single docker command * [rpc] Fix median-stake RPC * [slash] Update webhook with default docker BLS key * [docker][slash] Fresh yaml copy for docker build, remove dev code in main.go * [slash] Remove helper binary, commented out code, change to local config * [params] Factor out test genesis value * Add shard checking to Tx-Pool & correct blacklist (#2301) * [core] Fix blacklist & add shardID check * [staking + node + cmd] Fix blacklist & add shardID check * [slash] Adjust to PR comments part 1 * [docker] Use different throw away funded account * [docker] Create easier testing for delegation with private keys * [docker] Update yaml * [slash] Remove special case for slashing validator own delegation wrt min-self-delegate * [docker] Install nano as well * [slash] Early error if banned * [quorum] Expose earning account in decider marshal json * Revert "Refactor offchain data commit; Make block onchain/offchain commit atomic (#2279)" This reverts commit 9ffbf682c075b49188923c65a0bbf39ac188be00. * [slash] Add non-sanity check way to update validator * [reward] Increase percision on percentage in schedule * [slash] Adjust logs * [committee] Check eligibility of validator before doing sanity check * [slash] Update docker * [slash] Move create validator script to test * [slash] More log * [param] Make things faster * [slash][off-chain] Clear out slashes from pending in writeblockwithstate * [cross-link] Log is not error, just info * [blockchain] Not necessary to guard DeletePendingSlashingCandidates * [slash][consensus] Use plain []byte for signature b/c bls.Sign has private impl fields, rlp does not encode that * [slash][test] Use faucet as sender, assume user imported * [slash] Test setup * [slash] reserve error for real error in logs * [slash][availability] Apply availability correct, bump signing count each block * [slash][staking] Consider banned field in sanity check, pay snitch only half of what was actually slashed * [slash] Pay as much as can * [slash] use right nowAmt * [slash] Take away from rewards as well * [slash] iterate faster * [slash] Remove dev based timing * [slash] Add more log, sanity check incoming slash records, only count external for slash rate * [availability][state] Adjust signature of ValidatorWrapper wrt state, filter out for staked validators, correct availaibility measure on running counters * [availability] More log * [slash] Simply pre slash erra slashing * [slash] Remove development code * [slash] Use height from recvMsg, todo on epoch * [staking] Not necessary to touch LastEpochInCommittee in staking_verifier * [slash] Undo ds in endpoint pattern config * [slash] Add TODO and log when delegation becomes 0 b/c slash debt payment * [slash] Abstract staked validators from shard.State into type, set slash rate based BLSKey count Co-authored-by: Leo Chen <leo@harmony.one> Co-authored-by: flicker-harmony <52401354+flicker-harmony@users.noreply.github.com> Co-authored-by: Rongjian Lan <rongjian@harmony.one> Co-authored-by: Daniel Van Der Maden <daniel@harmony.one>
5 years ago
if err := consensus.msgSender.SendWithRetry(
consensus.blockNum,
msg_pb.MessageType_NEWVIEW,
[]nodeconfig.GroupID{
nodeconfig.NewGroupIDByShardID(nodeconfig.ShardID(consensus.ShardID))},
p2p.ConstructMessage(msgToSend),
[double-sign] Provide proof of double sign in slash record sent to beaconchain (#2253) * [double-sign] Commit changes in consensus needed for double-sign * [double-sign] Leader captures when valdator double signs, broadcasts to beaconchain * [slash] Add quick iteration tool for testing double-signing * [slash] Add webhook example * [slash] Add http server for hook to trigger double sign behavior * [double-sign] Use bin/trigger-double-sign to cause a double-sign * [double-sign] Full feedback loop working * [slash] Thread through the slash records in the block proposal step * [slash] Compute the slashing rate * [double-sign] Generalize yaml malicious for many keys * [double-sign][slash] Modify data structures, verify via webhook handler * [slash][double-sign] Find one address of bls public key signer, seemingly settle on data structures * [slash] Apply to state slashing for double signing * [slash][double-sign] Checkpoint for working code that slashes on beaconchain * [slash] Keep track of the total slash and total reporters reward * [slash] Dump account state before and after the slash * [slash] Satisfy Travis * [slash][state] Apply slash to the snapshot at beginning of epoch, now need to capture also the new delegates * [slash] Capture the unique new delegations since snapshot as well * [slash] Filter undelegation by epoch of double sign * [slash] Add TODO of correctness needed in slash needs on off-chain data * [rpc] Fix closure issue on shardID * [slash] Add delegator to double-sign testing script * [slash] Expand crt-validator.sh with commenting printfs and make delegation * [slash] Finish track payment of leftover slash debt after undelegation runs out * [slash] Now be explicit about error wrt delegatorSlashApply * [slash] Capture specific sanity check on slash paidoff * [slash] Track slash from undelegation piecemeal * [slash][delegation] Named slice types, .String() * [slash] Do no RLP encode twice, once is enough * [slash] Remove special case of validators own delegation * [slash] Refactor approach to slash state application * [slash] Begin expanding out Verify * [slash] Slash on snapshot delegations, not current * [slash] Fix Epoch Cmp * [slash] Third iteration on slash logic * [slash] Use full slash amount * [slash] More log, whitespace * [slash] Remove Println, add log * [slash] Remove debug Println * [slash] Add record in unit test * [slash] Build Validator snapshot, current. Fill out slash record * [slash] Need to get RLP dump of a header to use in test * [slash] Factor out double sign test constants * [slash] Factor out common for validator, stub out slash application, finish out deserialization setup * [slash] Factor out data structure creation because of var lexical scoping * [slash] Seem to have pipeline of unit test e2e executing * [slash] Add expected snitch, slash amounts * [slash] Checkpoint * [slash] Unit test correctly checks case of validator own stake which could drop below 1 ONE in slashing * [config] add double-sign testnet config (#1) Signed-off-by: Leo Chen <leo@harmony.one> * [slash] Commit for as is code & data of current dump.json * [slash] Order of state operation not correct in test, hence bad results, thank you dlv * [slash] Add snapshot state dump * [slash] Pay off slash of validator own delegation correctly * [slash] Pay off slash debt with special case for min-self * [slash] Pass first scenario conclusively * [slash] 2% slash passes unit test for own delegation and external * [slash] Parameterize unit test to easily test .02 vs .80 slash * [slash] Handle own delegation correctly at 80% slash * [slash] Have 80% slash working with external delegator * [slash] Remove debug code from slash * [slash] Adjust Apply signature, test again for 2% slash * [slash] Factor out scenario in testing so can test 2% and 80% at same time * [slash] Correct balance deduction on plan delegation * [slash] Mock out ChainReader for TestVerify * [slash] Small surface area interface, now feedback loop for verify * [slash] Remove development json * [slash] trigger-double-sign consumes yaml * [slash] Remove dead code * [slash][test] Factor ValidatorWrapper into scenario * [slash][test] Add example from local-testing dump - caution might be off * [slash] Factor out mutation of slashDebt * [slash][test] Factor out tests so can easily load test-case from bytes * [slash] Fix payment mistake in validator own delegation wrt min-self-delgation respected * [slash] Satisfy Travis * [slash] Begin cleanup of PR * [slash] Apply slash from header to Finalize via state processor * [slash] Productionize code, Println => logs; adjust slash picked in newblock * [slash] Need pointer for rlp.Decode * [slash] ValidatorInformation use full wrapper * Fix median stake * [staking] Adjust MarshalJSON for Validator, Wrapper * Refactor offchain data commit; Make block onchain/offchain commit atomic (#2279) * Refactor offchain data; Add epoch to ValidatorSnapshot * Make block onchain/offchain data commit atomically * [slash][committee] Set .Active to false on double sign, do not consider banned or inactive for committee assignment * [effective] VC eligible.go * [consensus] Redundant field in printf * [docker] import-ks for a dev account * [slash] Create BLS key for dockerfile and crt-validator.sh * [slash][docker] Easy deployment of double-sign testing * [docker] Have slash work as single docker command * [rpc] Fix median-stake RPC * [slash] Update webhook with default docker BLS key * [docker][slash] Fresh yaml copy for docker build, remove dev code in main.go * [slash] Remove helper binary, commented out code, change to local config * [params] Factor out test genesis value * Add shard checking to Tx-Pool & correct blacklist (#2301) * [core] Fix blacklist & add shardID check * [staking + node + cmd] Fix blacklist & add shardID check * [slash] Adjust to PR comments part 1 * [docker] Use different throw away funded account * [docker] Create easier testing for delegation with private keys * [docker] Update yaml * [slash] Remove special case for slashing validator own delegation wrt min-self-delegate * [docker] Install nano as well * [slash] Early error if banned * [quorum] Expose earning account in decider marshal json * Revert "Refactor offchain data commit; Make block onchain/offchain commit atomic (#2279)" This reverts commit 9ffbf682c075b49188923c65a0bbf39ac188be00. * [slash] Add non-sanity check way to update validator * [reward] Increase percision on percentage in schedule * [slash] Adjust logs * [committee] Check eligibility of validator before doing sanity check * [slash] Update docker * [slash] Move create validator script to test * [slash] More log * [param] Make things faster * [slash][off-chain] Clear out slashes from pending in writeblockwithstate * [cross-link] Log is not error, just info * [blockchain] Not necessary to guard DeletePendingSlashingCandidates * [slash][consensus] Use plain []byte for signature b/c bls.Sign has private impl fields, rlp does not encode that * [slash][test] Use faucet as sender, assume user imported * [slash] Test setup * [slash] reserve error for real error in logs * [slash][availability] Apply availability correct, bump signing count each block * [slash][staking] Consider banned field in sanity check, pay snitch only half of what was actually slashed * [slash] Pay as much as can * [slash] use right nowAmt * [slash] Take away from rewards as well * [slash] iterate faster * [slash] Remove dev based timing * [slash] Add more log, sanity check incoming slash records, only count external for slash rate * [availability][state] Adjust signature of ValidatorWrapper wrt state, filter out for staked validators, correct availaibility measure on running counters * [availability] More log * [slash] Simply pre slash erra slashing * [slash] Remove development code * [slash] Use height from recvMsg, todo on epoch * [staking] Not necessary to touch LastEpochInCommittee in staking_verifier * [slash] Undo ds in endpoint pattern config * [slash] Add TODO and log when delegation becomes 0 b/c slash debt payment * [slash] Abstract staked validators from shard.State into type, set slash rate based BLSKey count Co-authored-by: Leo Chen <leo@harmony.one> Co-authored-by: flicker-harmony <52401354+flicker-harmony@users.noreply.github.com> Co-authored-by: Rongjian Lan <rongjian@harmony.one> Co-authored-by: Daniel Van Der Maden <daniel@harmony.one>
5 years ago
); err != nil {
consensus.getLogger().Err(err).
Msg("could not send out the NEWVIEW message")
}
consensus.SetCurViewID(recvMsg.ViewID)
consensus.ResetViewChangeState()
consensus.consensusTimeout[timeoutViewChange].Stop()
consensus.consensusTimeout[timeoutConsensus].Start()
consensus.getLogger().Info().Str("myKey", newLeaderKey.Bytes.Hex()).Msg("[onViewChange] I am the New Leader")
}
}
// TODO: move to consensus_leader.go later
func (consensus *Consensus) onNewView(msg *msg_pb.Message) {
consensus.getLogger().Info().Msg("[onNewView] Received NewView Message")
recvMsg, err := consensus.ParseNewViewMessage(msg)
if err != nil {
consensus.getLogger().Warn().Err(err).Msg("[onNewView] Unable to Parse NewView Message")
return
}
if !consensus.onNewViewSanityCheck(recvMsg) {
return
}
if len(recvMsg.SenderPubkeys) != 1 {
consensus.getLogger().Error().Msg("[onNewView] multiple signers in view change message.")
return
}
senderKey := recvMsg.SenderPubkeys[0]
consensus.vcLock.Lock()
defer consensus.vcLock.Unlock()
if recvMsg.M3AggSig == nil || recvMsg.M3Bitmap == nil {
consensus.getLogger().Error().Msg("[onNewView] M3AggSig or M3Bitmap is nil")
return
}
m3Sig := recvMsg.M3AggSig
m3Mask := recvMsg.M3Bitmap
viewIDBytes := make([]byte, 8)
binary.LittleEndian.PutUint64(viewIDBytes, recvMsg.ViewID)
if !consensus.Decider.IsQuorumAchievedByMask(m3Mask) {
consensus.getLogger().Warn().
Msgf("[onNewView] Quorum Not achieved")
return
}
if !m3Sig.VerifyHash(m3Mask.AggregatePublic, viewIDBytes) {
consensus.getLogger().Warn().
Str("m3Sig", m3Sig.SerializeToHexStr()).
Hex("m3Mask", m3Mask.Bitmap).
Uint64("MsgViewID", recvMsg.ViewID).
Msg("[onNewView] Unable to Verify Aggregated Signature of M3 (ViewID) payload")
return
}
m2Mask := recvMsg.M2Bitmap
if recvMsg.M2AggSig != nil {
consensus.getLogger().Info().Msg("[onNewView] M2AggSig (NIL) is Not Empty")
m2Sig := recvMsg.M2AggSig
if !m2Sig.VerifyHash(m2Mask.AggregatePublic, NIL) {
consensus.getLogger().Warn().
Msg("[onNewView] Unable to Verify Aggregated Signature of M2 (NIL) payload")
return
}
}
// check when M3 sigs > M2 sigs, then M1 (recvMsg.Payload) should not be empty
preparedBlock := &types.Block{}
hasBlock := false
if len(recvMsg.Payload) != 0 && len(recvMsg.Block) != 0 {
if err := rlp.DecodeBytes(recvMsg.Block, preparedBlock); err != nil {
consensus.getLogger().Warn().
Err(err).
Uint64("MsgBlockNum", recvMsg.BlockNum).
Msg("[onNewView] Unparseable prepared block data")
return
}
blockHash := recvMsg.Payload[:32]
preparedBlockHash := preparedBlock.Hash()
if !bytes.Equal(preparedBlockHash[:], blockHash) {
consensus.getLogger().Warn().
Err(err).
Str("blockHash", preparedBlock.Hash().Hex()).
Str("payloadBlockHash", hex.EncodeToString(blockHash)).
Msg("[onNewView] Prepared block hash doesn't match msg block hash.")
return
}
hasBlock = true
if consensus.BlockVerifier(preparedBlock); err != nil {
consensus.getLogger().Error().Err(err).Msg("[onNewView] Prepared block verification failed")
return
}
}
if m2Mask == nil || m2Mask.Bitmap == nil ||
(m2Mask != nil && m2Mask.Bitmap != nil &&
utils.CountOneBits(m3Mask.Bitmap) > utils.CountOneBits(m2Mask.Bitmap)) {
if len(recvMsg.Payload) <= 32 {
consensus.getLogger().Info().
Msg("[onNewView] M1 (prepared) Type Payload Not Have Enough Length")
return
}
// m1 is not empty, check it's valid
blockHash := recvMsg.Payload[:32]
aggSig, mask, err := consensus.ReadSignatureBitmapPayload(recvMsg.Payload, 32)
if err != nil {
consensus.getLogger().Error().Err(err).
Msg("[onNewView] ReadSignatureBitmapPayload Failed")
return
}
if !aggSig.VerifyHash(mask.AggregatePublic, blockHash) {
consensus.getLogger().Warn().
Msg("[onNewView] Failed to Verify Signature for M1 (prepare) message")
return
}
copy(consensus.blockHash[:], blockHash)
consensus.aggregatedPrepareSig = aggSig
consensus.prepareBitmap = mask
// create prepared message from newview
preparedMsg := FBFTMessage{
MessageType: msg_pb.MessageType_PREPARED,
ViewID: recvMsg.ViewID,
BlockNum: recvMsg.BlockNum,
}
preparedMsg.BlockHash = common.Hash{}
copy(preparedMsg.BlockHash[:], blockHash[:])
preparedMsg.Payload = make([]byte, len(recvMsg.Payload)-32)
copy(preparedMsg.Payload[:], recvMsg.Payload[32:])
preparedMsg.SenderPubkeys = []*bls.PublicKeyWrapper{senderKey}
consensus.FBFTLog.AddMessage(&preparedMsg)
if hasBlock {
consensus.FBFTLog.AddBlock(preparedBlock)
}
}
// newView message verified success, override my state
consensus.SetCurViewID(recvMsg.ViewID)
consensus.LeaderPubKey = senderKey
consensus.ResetViewChangeState()
// change view and leaderKey to keep in sync with network
if consensus.blockNum != recvMsg.BlockNum {
consensus.getLogger().Info().
Str("newLeaderKey", consensus.LeaderPubKey.Bytes.Hex()).
Uint64("MsgBlockNum", recvMsg.BlockNum).
Msg("[onNewView] New Leader Changed")
return
}
// NewView message is verified, change state to normal consensus
if hasBlock {
// Construct and send the commit message
commitPayload := signature.ConstructCommitPayload(consensus.ChainReader,
preparedBlock.Epoch(), preparedBlock.Hash(), preparedBlock.NumberU64(), preparedBlock.Header().ViewID().Uint64())
groupID := []nodeconfig.GroupID{
nodeconfig.NewGroupIDByShardID(nodeconfig.ShardID(consensus.ShardID))}
priKeys := []*bls.PrivateKeyWrapper{}
p2pMsgs := []*NetworkMessage{}
for i, key := range consensus.priKey {
if !consensus.IsValidatorInCommittee(key.Pub.Bytes) {
continue
}
priKeys = append(priKeys, &consensus.priKey[i])
if !consensus.MultiSig {
network, err := consensus.construct(
msg_pb.MessageType_COMMIT,
commitPayload,
[]*bls.PrivateKeyWrapper{&key},
)
if err != nil {
consensus.getLogger().Err(err).Msg("could not create commit message")
return
}
p2pMsgs = append(p2pMsgs, network)
}
}
if consensus.MultiSig {
network, err := consensus.construct(
msg_pb.MessageType_COMMIT,
commitPayload,
priKeys,
)
if err != nil {
consensus.getLogger().Err(err).Msg("could not create commit message")
return
}
p2pMsgs = append(p2pMsgs, network)
}
for _, p2pMsg := range p2pMsgs {
consensus.getLogger().Info().Msg("onNewView === commit")
consensus.host.SendMessageToGroups(
groupID,
p2p.ConstructMessage(p2pMsg.Bytes),
)
[slash][consensus] Notice double sign & broadcast, factor out tech debt of consensus (#2152) * [slash] Remove dead interface, associated piping * [slash] Expand out structs * [consensus] Write to a chan when find a case of double-signing, remove dead code * [slash] Broadcast the noticing of a double signing * [rawdb] CRUD for slashing candidates * [slashing][node][proto] Broadcast the slash record after receive from consensus, handle received proto message, persist in off-chain db while pending * [slash][node][propose-block] Add verified slashes proposed into the header in block proposal * [slash][shard] Factor out external validator as method on shard state, add double-signature field * [slash][engine] Apply slash, name boolean expression for sorts, use stable sort * [slash] Abstract Ballot results so keep track of both pre and post double sign event * [slash] Fix type errors on test code * [slash] Read from correct rawdb * [slash] Add epoch based guards in CRUD of slashing * [slash] Write to correct cache for slashing candidates * [shard] Use explicit named type of BLS Signature, use convention * [slash] Fix mistake done in refactor, improper header used. Factor out fromSlice to set * [slash][node] Restore newblock to master, try again minimial change * [cx-receipts] Break up one-liner, use SliceStable, not Slice * [network] Finish refactor that makes network message headers once * [network] Simplify creation further of headers write * [slash] Adjust data structure of slash after offline discussion with RJ, Chao * [slash] Still did need signature of the double signature * [consensus] Prepare message does not have block header * [consensus] Soft reset three files to 968517d~1 * [consensus] Begin factor consensus network intended message out with prepare first * [consensus] Factor out Prepared message * [consensus] Factor out announce message creation * [consensus] Committed Message, branch on verify sender key for clearer log * [consensus] Committed Message Factor out * [consensus] Do jenkins MVP of signatures adjustment * [main][slash] Provide YAML config as webhook config for double sign event * [consensus] Adjust signatures, whitespace, lessen GC pressure * [consensus] Remove dead code * [consensus] Factor out commit overloaded message, give commit payload override in construct * [consensus] Fix travis tests * [consensus] Provide block bytes in SubmitVote(quorum.Commit) * [consensus] Factor out noisy sanity checks in BFT, move existing commit check earlier as was before * [quorum] Adjust signatures in quorum * [staking] Adjust after merge from master * [consensus] Finish refactor of consensus * [node] Fix import * [consensus] Fix travis * [consensus] Use origin/master copy of block, fix mistake of pointer to empty byte * [consensus] Less verbose bools * [consensus] Remove unused trailing mutation hook in message construct * [consensus] Address some TODOs on err, comment out double sign
5 years ago
}
consensus.getLogger().Info().
Str("From", consensus.phase.String()).
Str("To", FBFTCommit.String()).
Msg("[OnViewChange] Switching phase")
consensus.switchPhase(FBFTCommit, true)
} else {
consensus.ResetState()
consensus.getLogger().Info().Msg("onNewView === announce")
}
consensus.getLogger().Info().
Str("newLeaderKey", consensus.LeaderPubKey.Bytes.Hex()).
Msg("new leader changed")
consensus.getLogger().Info().
Msg("validator start consensus timer and stop view change timer")
consensus.consensusTimeout[timeoutConsensus].Start()
consensus.consensusTimeout[timeoutViewChange].Stop()
}