1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
|
// Copyright 2018 The dexon-consensus Authors
// This file is part of the dexon-consensus library.
//
// The dexon-consensus library is free software: you can redistribute it
// and/or modify it under the terms of the GNU Lesser General Public License as
// published by the Free Software Foundation, either version 3 of the License,
// or (at your option) any later version.
//
// The dexon-consensus library is distributed in the hope that it will be
// useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser
// General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the dexon-consensus library. If not, see
// <http://www.gnu.org/licenses/>.
package syncer
import (
"context"
"fmt"
"sort"
"sync"
"time"
"github.com/dexon-foundation/dexon-consensus/common"
"github.com/dexon-foundation/dexon-consensus/core"
"github.com/dexon-foundation/dexon-consensus/core/crypto"
"github.com/dexon-foundation/dexon-consensus/core/db"
"github.com/dexon-foundation/dexon-consensus/core/types"
"github.com/dexon-foundation/dexon-consensus/core/utils"
)
var (
// ErrAlreadySynced is reported when syncer is synced.
ErrAlreadySynced = fmt.Errorf("already synced")
// ErrNotSynced is reported when syncer is not synced yet.
ErrNotSynced = fmt.Errorf("not synced yet")
// ErrGenesisBlockReached is reported when genesis block reached.
ErrGenesisBlockReached = fmt.Errorf("genesis block reached")
// ErrInvalidBlockOrder is reported when SyncBlocks receives unordered
// blocks.
ErrInvalidBlockOrder = fmt.Errorf("invalid block order")
// ErrMismatchBlockHashSequence means the delivering sequence is not
// correct, compared to finalized blocks.
ErrMismatchBlockHashSequence = fmt.Errorf("mismatch block hash sequence")
// ErrInvalidSyncingFinalizationHeight raised when the blocks to sync is
// not following the compaction chain tip in database.
ErrInvalidSyncingFinalizationHeight = fmt.Errorf(
"invalid syncing finalization height")
)
// Consensus is for syncing consensus module.
type Consensus struct {
db db.Database
gov core.Governance
dMoment time.Time
logger common.Logger
app core.Application
prv crypto.PrivateKey
network core.Network
nodeSetCache *utils.NodeSetCache
tsigVerifier *core.TSigVerifierCache
randomnessResults map[common.Hash]*types.BlockRandomnessResult
blocks types.BlocksByPosition
agreementModule *agreement
agreementRoundCut uint64
heightEvt *common.Event
roundEvt *utils.RoundEvent
// lock for accessing all fields.
lock sync.RWMutex
duringBuffering bool
latestCRSRound uint64
waitGroup sync.WaitGroup
agreementWaitGroup sync.WaitGroup
pullChan chan common.Hash
receiveChan chan *types.Block
ctx context.Context
ctxCancel context.CancelFunc
syncedLastBlock *types.Block
syncedConsensus *core.Consensus
syncedSkipNext bool
dummyCancel context.CancelFunc
dummyFinished <-chan struct{}
dummyMsgBuffer []interface{}
initChainTipHeight uint64
}
// NewConsensus creates an instance for Consensus (syncer consensus).
func NewConsensus(
dMoment time.Time,
app core.Application,
gov core.Governance,
db db.Database,
network core.Network,
prv crypto.PrivateKey,
logger common.Logger) *Consensus {
con := &Consensus{
dMoment: dMoment,
app: app,
gov: gov,
db: db,
network: network,
nodeSetCache: utils.NewNodeSetCache(gov),
tsigVerifier: core.NewTSigVerifierCache(gov, 7),
prv: prv,
logger: logger,
receiveChan: make(chan *types.Block, 1000),
pullChan: make(chan common.Hash, 1000),
randomnessResults: make(map[common.Hash]*types.BlockRandomnessResult),
heightEvt: common.NewEvent(),
}
con.ctx, con.ctxCancel = context.WithCancel(context.Background())
_, con.initChainTipHeight = db.GetCompactionChainTipInfo()
con.agreementModule = newAgreement(
con.receiveChan, con.pullChan, con.nodeSetCache, con.logger)
con.agreementWaitGroup.Add(1)
go func() {
defer con.agreementWaitGroup.Done()
con.agreementModule.run()
}()
return con
}
func (con *Consensus) assureBuffering() {
if func() bool {
con.lock.RLock()
defer con.lock.RUnlock()
return con.duringBuffering
}() {
return
}
con.lock.Lock()
defer con.lock.Unlock()
if con.duringBuffering {
return
}
con.duringBuffering = true
// Get latest block to prepare utils.RoundEvent.
var (
err error
blockHash, height = con.db.GetCompactionChainTipInfo()
)
if height == 0 {
con.roundEvt, err = utils.NewRoundEvent(con.ctx, con.gov, con.logger,
0, 0, core.ConfigRoundShift)
} else {
var b types.Block
if b, err = con.db.GetBlock(blockHash); err == nil {
con.roundEvt, err = utils.NewRoundEvent(con.ctx, con.gov,
con.logger, b.Position.Round, b.Finalization.Height,
core.ConfigRoundShift)
}
}
if err != nil {
panic(err)
}
// Make sure con.roundEvt stopped before stopping con.agreementModule.
con.waitGroup.Add(1)
// Register a round event handler to reset node set cache, this handler
// should be the highest priority.
con.roundEvt.Register(func(evts []utils.RoundEventParam) {
for _, e := range evts {
if e.Reset == 0 {
continue
}
con.nodeSetCache.Purge(e.Round + 1)
con.tsigVerifier.Purge(e.Round + 1)
}
})
// Register a round event handler to notify CRS to agreementModule.
con.roundEvt.Register(func(evts []utils.RoundEventParam) {
con.waitGroup.Add(1)
go func() {
defer con.waitGroup.Done()
for _, e := range evts {
select {
case <-con.ctx.Done():
return
default:
}
for func() bool {
select {
case <-con.ctx.Done():
return false
case con.agreementModule.inputChan <- e.Round:
return false
case <-time.After(500 * time.Millisecond):
con.logger.Warn(
"agreement input channel is full when notifying new round",
"round", e.Round,
)
return true
}
}() {
}
}
}()
})
// Register a round event handler to validate next round.
con.roundEvt.Register(func(evts []utils.RoundEventParam) {
con.heightEvt.RegisterHeight(
evts[len(evts)-1].NextRoundValidationHeight(),
utils.RoundEventRetryHandlerGenerator(con.roundEvt, con.heightEvt),
)
})
con.roundEvt.TriggerInitEvent()
con.startAgreement()
con.startNetwork()
}
func (con *Consensus) checkIfSynced(blocks []*types.Block) (synced bool) {
con.lock.RLock()
defer con.lock.RUnlock()
defer func() {
con.logger.Debug("syncer synced status",
"last-block", blocks[len(blocks)-1],
"synced", synced,
)
}()
if len(con.blocks) == 0 || len(blocks) == 0 {
return
}
synced = !blocks[len(blocks)-1].Position.Older(con.blocks[0].Position)
return
}
func (con *Consensus) buildAllEmptyBlocks() {
con.lock.Lock()
defer con.lock.Unlock()
// Clean empty blocks on tips of chains.
for len(con.blocks) > 0 && con.isEmptyBlock(con.blocks[0]) {
con.blocks = con.blocks[1:]
}
// Build empty blocks.
for i, b := range con.blocks {
if con.isEmptyBlock(b) {
if con.blocks[i-1].Position.Height+1 == b.Position.Height {
con.buildEmptyBlock(b, con.blocks[i-1])
}
}
}
}
// ForceSync forces syncer to become synced.
func (con *Consensus) ForceSync(skip bool) {
if con.syncedLastBlock != nil {
return
}
hash, _ := con.db.GetCompactionChainTipInfo()
var block types.Block
block, err := con.db.GetBlock(hash)
if err != nil {
panic(err)
}
con.logger.Info("Force Sync", "block", &block)
con.syncedLastBlock = &block
con.stopBuffering()
// We might call stopBuffering without calling assureBuffering.
if con.dummyCancel == nil {
con.dummyCancel, con.dummyFinished = utils.LaunchDummyReceiver(
context.Background(), con.network.ReceiveChan(),
func(msg interface{}) {
con.dummyMsgBuffer = append(con.dummyMsgBuffer, msg)
})
}
con.syncedSkipNext = skip
}
// SyncBlocks syncs blocks from compaction chain, latest is true if the caller
// regards the blocks are the latest ones. Notice that latest can be true for
// many times.
// NOTICE: parameter "blocks" should be consecutive in compaction height.
// NOTICE: this method is not expected to be called concurrently.
func (con *Consensus) SyncBlocks(
blocks []*types.Block, latest bool) (synced bool, err error) {
defer func() {
con.logger.Debug("SyncBlocks returned",
"synced", synced,
"error", err,
"last-block", con.syncedLastBlock,
)
}()
if con.syncedLastBlock != nil {
synced, err = true, ErrAlreadySynced
return
}
if len(blocks) == 0 {
return
}
// Check if blocks are consecutive.
for i := 1; i < len(blocks); i++ {
if blocks[i].Finalization.Height != blocks[i-1].Finalization.Height+1 {
err = ErrInvalidBlockOrder
return
}
}
// Make sure the first block is the next block of current compaction chain
// tip in DB.
_, tipHeight := con.db.GetCompactionChainTipInfo()
if blocks[0].Finalization.Height != tipHeight+1 {
con.logger.Error("mismatched finalization height",
"now", blocks[0].Finalization.Height,
"expected", tipHeight+1,
)
err = ErrInvalidSyncingFinalizationHeight
return
}
con.logger.Trace("syncBlocks",
"position", &blocks[0].Position,
"final height", blocks[0].Finalization.Height,
"len", len(blocks),
"latest", latest,
)
for _, b := range blocks {
if err = con.db.PutBlock(*b); err != nil {
// A block might be put into db when confirmed by BA, but not
// finalized yet.
if err == db.ErrBlockExists {
err = con.db.UpdateBlock(*b)
}
if err != nil {
return
}
}
if err = con.db.PutCompactionChainTipInfo(
b.Hash, b.Finalization.Height); err != nil {
return
}
con.heightEvt.NotifyHeight(b.Finalization.Height)
}
if latest {
con.assureBuffering()
con.buildAllEmptyBlocks()
// Check if compaction and agreements' blocks are overlapped. The
// overlapping of compaction chain and BA's oldest blocks means the
// syncing is done.
if con.checkIfSynced(blocks) {
con.stopBuffering()
con.syncedLastBlock = blocks[len(blocks)-1]
synced = true
}
}
return
}
// GetSyncedConsensus returns the core.Consensus instance after synced.
func (con *Consensus) GetSyncedConsensus() (*core.Consensus, error) {
con.lock.Lock()
defer con.lock.Unlock()
if con.syncedConsensus != nil {
return con.syncedConsensus, nil
}
if con.syncedLastBlock == nil {
return nil, ErrNotSynced
}
// flush all blocks in con.blocks into core.Consensus, and build
// core.Consensus from syncer.
randomnessResults := []*types.BlockRandomnessResult{}
for _, r := range con.randomnessResults {
randomnessResults = append(randomnessResults, r)
}
con.dummyCancel()
<-con.dummyFinished
var err error
con.syncedConsensus, err = core.NewConsensusFromSyncer(
con.syncedLastBlock,
con.syncedSkipNext,
con.dMoment,
con.app,
con.gov,
con.db,
con.network,
con.prv,
con.blocks,
randomnessResults,
con.dummyMsgBuffer,
con.logger)
return con.syncedConsensus, err
}
// stopBuffering stops the syncer buffering routines.
//
// This method is mainly for caller to stop the syncer before synced, the syncer
// would call this method automatically after being synced.
func (con *Consensus) stopBuffering() {
if func() (notBuffering bool) {
con.lock.RLock()
defer con.lock.RUnlock()
notBuffering = !con.duringBuffering
return
}() {
return
}
if func() (alreadyCanceled bool) {
con.lock.Lock()
defer con.lock.Unlock()
if !con.duringBuffering {
alreadyCanceled = true
return
}
con.duringBuffering = false
con.logger.Trace("syncer is about to stop")
// Stop network and CRS routines, wait until they are all stoped.
con.ctxCancel()
return
}() {
return
}
con.logger.Trace("stop syncer modules")
con.roundEvt.Stop()
con.waitGroup.Done()
// Wait for all routines depends on con.agreementModule stopped.
con.waitGroup.Wait()
// Since there is no one waiting for the receive channel of fullnode, we
// need to launch a dummy receiver right away.
con.dummyCancel, con.dummyFinished = utils.LaunchDummyReceiver(
context.Background(), con.network.ReceiveChan(),
func(msg interface{}) {
con.dummyMsgBuffer = append(con.dummyMsgBuffer, msg)
})
// Stop agreements.
con.logger.Trace("stop syncer agreement modules")
con.stopAgreement()
con.logger.Trace("syncer stopped")
return
}
// isEmptyBlock checks if a block is an empty block by both its hash and parent
// hash are empty.
func (con *Consensus) isEmptyBlock(b *types.Block) bool {
return b.Hash == common.Hash{} && b.ParentHash == common.Hash{}
}
// buildEmptyBlock builds an empty block in agreement.
func (con *Consensus) buildEmptyBlock(b *types.Block, parent *types.Block) {
cfg := utils.GetConfigWithPanic(con.gov, b.Position.Round, con.logger)
b.Timestamp = parent.Timestamp.Add(cfg.MinBlockInterval)
b.Witness.Height = parent.Witness.Height
b.Witness.Data = make([]byte, len(parent.Witness.Data))
copy(b.Witness.Data, parent.Witness.Data)
}
// startAgreement starts agreements for receiving votes and agreements.
func (con *Consensus) startAgreement() {
// Start a routine for listening receive channel and pull block channel.
go func() {
for {
select {
case b, ok := <-con.receiveChan:
if !ok {
return
}
func() {
con.lock.Lock()
defer con.lock.Unlock()
if len(con.blocks) > 0 &&
!b.Position.Newer(con.blocks[0].Position) {
return
}
con.blocks = append(con.blocks, b)
sort.Sort(con.blocks)
}()
case h, ok := <-con.pullChan:
if !ok {
return
}
con.network.PullBlocks(common.Hashes{h})
}
}
}()
}
func (con *Consensus) cacheRandomnessResult(r *types.BlockRandomnessResult) {
// There is no block randomness at round-0.
if r.Position.Round == 0 {
return
}
// We only have to cache randomness result after cutting round.
if func() bool {
con.lock.RLock()
defer con.lock.RUnlock()
if len(con.blocks) > 0 && r.Position.Older(con.blocks[0].Position) {
return true
}
if r.Position.Round > con.latestCRSRound {
// We can't process randomness from rounds that its CRS is still
// unknown.
return true
}
_, exists := con.randomnessResults[r.BlockHash]
return exists
}() {
return
}
v, ok, err := con.tsigVerifier.UpdateAndGet(r.Position.Round)
if err != nil {
con.logger.Error("Unable to get tsig verifier",
"hash", r.BlockHash.String()[:6],
"position", r.Position,
"error", err,
)
return
}
if !ok {
con.logger.Error("Tsig is not ready", "position", &r.Position)
return
}
if !v.VerifySignature(r.BlockHash, crypto.Signature{
Type: "bls",
Signature: r.Randomness}) {
con.logger.Info("Block randomness is not valid",
"position", r.Position,
"hash", r.BlockHash.String()[:6],
)
return
}
con.lock.Lock()
defer con.lock.Unlock()
con.randomnessResults[r.BlockHash] = r
}
// startNetwork starts network for receiving blocks and agreement results.
func (con *Consensus) startNetwork() {
con.waitGroup.Add(1)
go func() {
defer con.waitGroup.Done()
loop:
for {
select {
case val := <-con.network.ReceiveChan():
switch v := val.(type) {
case *types.Block:
case *types.AgreementResult:
// Avoid byzantine nodes attack by broadcasting older
// agreement results. Normal nodes might report 'synced'
// while still fall behind other nodes.
if v.Position.Height <= con.initChainTipHeight {
continue loop
}
case *types.BlockRandomnessResult:
con.cacheRandomnessResult(v)
continue loop
default:
continue loop
}
con.agreementModule.inputChan <- val
case <-con.ctx.Done():
break loop
}
}
}()
}
func (con *Consensus) stopAgreement() {
if con.agreementModule.inputChan != nil {
close(con.agreementModule.inputChan)
}
con.agreementWaitGroup.Wait()
con.agreementModule.inputChan = nil
close(con.receiveChan)
close(con.pullChan)
}
|