/
state.go
523 lines (468 loc) · 17.8 KB
/
state.go
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
package driver
import (
"bytes"
"context"
"encoding/json"
"errors"
"fmt"
"io"
gosync "sync"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum-optimism/optimism/op-node/eth"
"github.com/ethereum-optimism/optimism/op-node/rollup"
"github.com/ethereum-optimism/optimism/op-node/rollup/derive"
"github.com/ethereum-optimism/optimism/op-service/backoff"
)
// Deprecated: use eth.SyncStatus instead.
type SyncStatus = eth.SyncStatus
// sealingDuration defines the expected time it takes to seal the block
const sealingDuration = time.Millisecond * 50
type Driver struct {
l1State L1StateIface
// The derivation pipeline is reset whenever we reorg.
// The derivation pipeline determines the new l2Safe.
derivation DerivationPipeline
// Requests to block the event loop for synchronous execution to avoid reading an inconsistent state
stateReq chan chan struct{}
// Upon receiving a channel in this channel, the derivation pipeline is forced to be reset.
// It tells the caller that the reset occurred by closing the passed in channel.
forceReset chan chan struct{}
// Upon receiving a hash in this channel, the sequencer is started at the given hash.
// It tells the caller that the sequencer started by closing the passed in channel (or returning an error).
startSequencer chan hashAndErrorChannel
// Upon receiving a channel in this channel, the sequencer is stopped.
// It tells the caller that the sequencer stopped by returning the latest sequenced L2 block hash.
stopSequencer chan chan hashAndError
// Rollup config: rollup chain configuration
config *rollup.Config
// Driver config: verifier and sequencer settings
driverConfig *Config
// L1 Signals:
//
// Not all L1 blocks, or all changes, have to be signalled:
// the derivation process traverses the chain and handles reorgs as necessary,
// the driver just needs to be aware of the *latest* signals enough so to not
// lag behind actionable data.
l1HeadSig chan eth.L1BlockRef
l1SafeSig chan eth.L1BlockRef
l1FinalizedSig chan eth.L1BlockRef
// Interface to signal the L2 block range to sync.
altSync AltSync
// L2 Signals:
unsafeL2Payloads chan *eth.ExecutionPayload
l1 L1Chain
l2 L2Chain
sequencer SequencerIface
network Network // may be nil, network for is optional
metrics Metrics
log log.Logger
snapshotLog log.Logger
done chan struct{}
wg gosync.WaitGroup
}
// Start starts up the state loop.
// The loop will have been started iff err is not nil.
func (s *Driver) Start() error {
s.derivation.Reset()
s.wg.Add(1)
go s.eventLoop()
return nil
}
func (s *Driver) Close() error {
s.done <- struct{}{}
s.wg.Wait()
return nil
}
// OnL1Head signals the driver that the L1 chain changed the "unsafe" block,
// also known as head of the chain, or "latest".
func (s *Driver) OnL1Head(ctx context.Context, unsafe eth.L1BlockRef) error {
select {
case <-ctx.Done():
return ctx.Err()
case s.l1HeadSig <- unsafe:
return nil
}
}
// OnL1Safe signals the driver that the L1 chain changed the "safe",
// also known as the justified checkpoint (as seen on L1 beacon-chain).
func (s *Driver) OnL1Safe(ctx context.Context, safe eth.L1BlockRef) error {
select {
case <-ctx.Done():
return ctx.Err()
case s.l1SafeSig <- safe:
return nil
}
}
func (s *Driver) OnL1Finalized(ctx context.Context, finalized eth.L1BlockRef) error {
select {
case <-ctx.Done():
return ctx.Err()
case s.l1FinalizedSig <- finalized:
return nil
}
}
func (s *Driver) OnUnsafeL2Payload(ctx context.Context, payload *eth.ExecutionPayload) error {
s.log.Debug("On unsafeL2Payloads channel buffer size", "length", len(s.unsafeL2Payloads))
select {
case <-ctx.Done():
return ctx.Err()
case s.unsafeL2Payloads <- payload:
return nil
}
}
// the eventLoop responds to L1 changes and internal timers to produce L2 blocks.
func (s *Driver) eventLoop() {
defer s.wg.Done()
s.log.Info("State loop started")
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
// stepReqCh is used to request that the driver attempts to step forward by one L1 block.
stepReqCh := make(chan struct{}, 1)
// channel, nil by default (not firing), but used to schedule re-attempts with delay
var delayedStepReq <-chan time.Time
// keep track of consecutive failed attempts, to adjust the backoff time accordingly
bOffStrategy := backoff.Exponential()
stepAttempts := 0
// step requests a derivation step to be taken. Won't deadlock if the channel is full.
step := func() {
select {
case stepReqCh <- struct{}{}:
// Don't deadlock if the channel is already full
default:
}
}
// reqStep requests a derivation step nicely, with a delay if this is a reattempt, or not at all if we already scheduled a reattempt.
reqStep := func() {
if stepAttempts > 0 {
// if this is not the first attempt, we re-schedule with a backoff, *without blocking other events*
if delayedStepReq == nil {
delay := bOffStrategy.Duration(stepAttempts)
s.log.Debug("scheduling re-attempt with delay", "attempts", stepAttempts, "delay", delay)
delayedStepReq = time.After(delay)
} else {
s.log.Debug("ignoring step request, already scheduled re-attempt after previous failure", "attempts", stepAttempts)
}
} else {
step()
}
}
// We call reqStep right away to finish syncing to the tip of the chain if we're behind.
// reqStep will also be triggered when the L1 head moves forward or if there was a reorg on the
// L1 chain that we need to handle.
reqStep()
sequencerTimer := time.NewTimer(0)
var sequencerCh <-chan time.Time
planSequencerAction := func() {
delay := s.sequencer.PlanNextSequencerAction()
sequencerCh = sequencerTimer.C
if len(sequencerCh) > 0 { // empty if not already drained before resetting
<-sequencerCh
}
sequencerTimer.Reset(delay)
// Without this sleep, even if delay is 0, sequencerCh will not be ready in the following select judge.
// As a result, during deriving a batch blocks, producing new block will be stopped.
// After sleep for 0.1 ms, both sequencerCh and stepReqCh might be ready.
// And select judge will random choose to proceed derivation or produce new blocks
time.Sleep(100 * time.Microsecond)
}
// Create a ticker to check if there is a gap in the engine queue. Whenever
// there is, we send requests to sync source to retrieve the missing payloads.
syncCheckInterval := time.Duration(s.config.BlockTime) * time.Second * 2
altSyncTicker := time.NewTicker(syncCheckInterval)
defer altSyncTicker.Stop()
lastUnsafeL2 := s.derivation.UnsafeL2Head()
for {
// If we are sequencing, and the L1 state is ready, update the trigger for the next sequencer action.
// This may adjust at any time based on fork-choice changes or previous errors.
// And avoid sequencing if the derivation pipeline indicates the engine is not ready.
if s.driverConfig.SequencerEnabled && !s.driverConfig.SequencerStopped &&
s.l1State.L1Head() != (eth.L1BlockRef{}) && s.derivation.EngineReady() {
if s.driverConfig.SequencerMaxSafeLag > 0 && s.derivation.SafeL2Head().Number+s.driverConfig.SequencerMaxSafeLag <= s.derivation.UnsafeL2Head().Number {
// If the safe head has fallen behind by a significant number of blocks, delay creating new blocks
// until the safe lag is below SequencerMaxSafeLag.
if sequencerCh != nil {
s.log.Warn(
"Delay creating new block since safe lag exceeds limit",
"safe_l2", s.derivation.SafeL2Head(),
"unsafe_l2", s.derivation.UnsafeL2Head(),
)
sequencerCh = nil
}
} else if s.sequencer.BuildingOnto().ID() != s.derivation.UnsafeL2Head().ID() {
// If we are sequencing, and the L1 state is ready, update the trigger for the next sequencer action.
// This may adjust at any time based on fork-choice changes or previous errors.
//
// update sequencer time if the head changed
planSequencerAction()
}
} else {
sequencerCh = nil
}
// If the engine is not ready, or if the L2 head is actively changing, then reset the alt-sync:
// there is no need to request L2 blocks when we are syncing already.
if head := s.derivation.UnsafeL2Head(); head != lastUnsafeL2 || !s.derivation.EngineReady() {
lastUnsafeL2 = head
altSyncTicker.Reset(syncCheckInterval)
}
select {
case <-sequencerCh:
payload, err := s.sequencer.RunNextSequencerAction(ctx)
if err != nil {
s.log.Error("Sequencer critical error", "err", err)
return
}
if s.network != nil && payload != nil {
// Publishing of unsafe data via p2p is optional.
// Errors are not severe enough to change/halt sequencing but should be logged and metered.
if err := s.network.PublishL2Payload(ctx, payload); err != nil {
s.log.Warn("failed to publish newly created block", "id", payload.ID(), "err", err)
s.metrics.RecordPublishingError()
}
}
planSequencerAction() // schedule the next sequencer action to keep the sequencing looping
case <-altSyncTicker.C:
// Check if there is a gap in the current unsafe payload queue.
ctx, cancel := context.WithTimeout(ctx, time.Second*2)
err := s.checkForGapInUnsafeQueue(ctx)
cancel()
if err != nil {
s.log.Warn("failed to check for unsafe L2 blocks to sync", "err", err)
}
case payload := <-s.unsafeL2Payloads:
s.snapshot("New unsafe payload")
s.log.Info("Optimistically queueing unsafe L2 execution payload", "id", payload.ID())
s.derivation.AddUnsafePayload(payload)
s.metrics.RecordReceivedUnsafePayload(payload)
for len(s.unsafeL2Payloads) > 0 {
payload = <-s.unsafeL2Payloads
s.log.Info("Optimistically queueing unsafe L2 execution payload", "id", payload.ID())
s.derivation.AddUnsafePayload(payload)
s.metrics.RecordReceivedUnsafePayload(payload)
}
reqStep()
case newL1Head := <-s.l1HeadSig:
s.l1State.HandleNewL1HeadBlock(newL1Head)
reqStep() // a new L1 head may mean we have the data to not get an EOF again.
case newL1Safe := <-s.l1SafeSig:
s.l1State.HandleNewL1SafeBlock(newL1Safe)
// no step, justified L1 information does not do anything for L2 derivation or status
case newL1Finalized := <-s.l1FinalizedSig:
s.l1State.HandleNewL1FinalizedBlock(newL1Finalized)
s.derivation.Finalize(newL1Finalized)
reqStep() // we may be able to mark more L2 data as finalized now
case <-delayedStepReq:
delayedStepReq = nil
step()
case <-stepReqCh:
s.metrics.SetDerivationIdle(false)
s.log.Debug("Derivation process step", "onto_origin", s.derivation.Origin(), "attempts", stepAttempts)
err := s.derivation.Step(context.Background())
stepAttempts += 1 // count as attempt by default. We reset to 0 if we are making healthy progress.
if err == io.EOF {
s.log.Debug("Derivation process went idle", "progress", s.derivation.Origin(), "err", err)
stepAttempts = 0
s.metrics.SetDerivationIdle(true)
continue
} else if err != nil && errors.Is(err, derive.EngineP2PSyncing) {
s.log.Debug("Derivation process went idle because the engine is syncing", "progress", s.derivation.Origin(), "sync_target", s.derivation.EngineSyncTarget(), "err", err)
stepAttempts = 0
s.metrics.SetDerivationIdle(true)
continue
} else if err != nil && errors.Is(err, derive.ErrReset) {
// If the pipeline corrupts, e.g. due to a reorg, simply reset it
s.log.Warn("Derivation pipeline is reset", "err", err)
s.derivation.Reset()
s.metrics.RecordPipelineReset()
continue
} else if err != nil && errors.Is(err, derive.ErrTemporary) {
s.log.Warn("Derivation process temporary error", "attempts", stepAttempts, "err", err)
reqStep()
continue
} else if err != nil && errors.Is(err, derive.ErrCritical) {
s.log.Error("Derivation process critical error", "err", err)
return
} else if err != nil && errors.Is(err, derive.NotEnoughData) {
stepAttempts = 0 // don't do a backoff for this error
reqStep()
continue
} else if err != nil {
s.log.Error("Derivation process error", "attempts", stepAttempts, "err", err)
reqStep()
continue
} else {
stepAttempts = 0
reqStep() // continue with the next step if we can
}
case respCh := <-s.stateReq:
respCh <- struct{}{}
case respCh := <-s.forceReset:
s.log.Warn("Derivation pipeline is manually reset")
s.derivation.Reset()
s.metrics.RecordPipelineReset()
close(respCh)
case resp := <-s.startSequencer:
unsafeHead := s.derivation.UnsafeL2Head().Hash
if !s.driverConfig.SequencerStopped {
resp.err <- errors.New("sequencer already running")
} else if !bytes.Equal(unsafeHead[:], resp.hash[:]) {
resp.err <- fmt.Errorf("block hash does not match: head %s, received %s", unsafeHead.String(), resp.hash.String())
} else {
s.log.Info("Sequencer has been started")
s.driverConfig.SequencerStopped = false
close(resp.err)
planSequencerAction() // resume sequencing
}
case respCh := <-s.stopSequencer:
if s.driverConfig.SequencerStopped {
respCh <- hashAndError{err: errors.New("sequencer not running")}
} else {
s.log.Warn("Sequencer has been stopped")
s.driverConfig.SequencerStopped = true
respCh <- hashAndError{hash: s.derivation.UnsafeL2Head().Hash}
}
case <-s.done:
return
}
}
}
// ResetDerivationPipeline forces a reset of the derivation pipeline.
// It waits for the reset to occur. It simply unblocks the caller rather
// than fully cancelling the reset request upon a context cancellation.
func (s *Driver) ResetDerivationPipeline(ctx context.Context) error {
respCh := make(chan struct{}, 1)
select {
case <-ctx.Done():
return ctx.Err()
case s.forceReset <- respCh:
select {
case <-ctx.Done():
return ctx.Err()
case <-respCh:
return nil
}
}
}
func (s *Driver) StartSequencer(ctx context.Context, blockHash common.Hash) error {
if !s.driverConfig.SequencerEnabled {
return errors.New("sequencer is not enabled")
}
h := hashAndErrorChannel{
hash: blockHash,
err: make(chan error, 1),
}
select {
case <-ctx.Done():
return ctx.Err()
case s.startSequencer <- h:
select {
case <-ctx.Done():
return ctx.Err()
case e := <-h.err:
return e
}
}
}
func (s *Driver) StopSequencer(ctx context.Context) (common.Hash, error) {
if !s.driverConfig.SequencerEnabled {
return common.Hash{}, errors.New("sequencer is not enabled")
}
respCh := make(chan hashAndError, 1)
select {
case <-ctx.Done():
return common.Hash{}, ctx.Err()
case s.stopSequencer <- respCh:
select {
case <-ctx.Done():
return common.Hash{}, ctx.Err()
case he := <-respCh:
return he.hash, he.err
}
}
}
// syncStatus returns the current sync status, and should only be called synchronously with
// the driver event loop to avoid retrieval of an inconsistent status.
func (s *Driver) syncStatus() *eth.SyncStatus {
return ð.SyncStatus{
CurrentL1: s.derivation.Origin(),
CurrentL1Finalized: s.derivation.FinalizedL1(),
HeadL1: s.l1State.L1Head(),
SafeL1: s.l1State.L1Safe(),
FinalizedL1: s.l1State.L1Finalized(),
UnsafeL2: s.derivation.UnsafeL2Head(),
SafeL2: s.derivation.SafeL2Head(),
FinalizedL2: s.derivation.Finalized(),
UnsafeL2SyncTarget: s.derivation.UnsafeL2SyncTarget(),
EngineSyncTarget: s.derivation.EngineSyncTarget(),
}
}
// SyncStatus blocks the driver event loop and captures the syncing status.
// If the event loop is too busy and the context expires, a context error is returned.
func (s *Driver) SyncStatus(ctx context.Context) (*eth.SyncStatus, error) {
wait := make(chan struct{})
select {
case s.stateReq <- wait:
resp := s.syncStatus()
<-wait
return resp, nil
case <-ctx.Done():
return nil, ctx.Err()
}
}
// BlockRefWithStatus blocks the driver event loop and captures the syncing status,
// along with an L2 block reference by number consistent with that same status.
// If the event loop is too busy and the context expires, a context error is returned.
func (s *Driver) BlockRefWithStatus(ctx context.Context, num uint64) (eth.L2BlockRef, *eth.SyncStatus, error) {
wait := make(chan struct{})
select {
case s.stateReq <- wait:
resp := s.syncStatus()
ref, err := s.l2.L2BlockRefByNumber(ctx, num)
<-wait
return ref, resp, err
case <-ctx.Done():
return eth.L2BlockRef{}, nil, ctx.Err()
}
}
// deferJSONString helps avoid a JSON-encoding performance hit if the snapshot logger does not run
type deferJSONString struct {
x any
}
func (v deferJSONString) String() string {
out, _ := json.Marshal(v.x)
return string(out)
}
func (s *Driver) snapshot(event string) {
s.snapshotLog.Info("Rollup State Snapshot",
"event", event,
"l1Head", deferJSONString{s.l1State.L1Head()},
"l1Current", deferJSONString{s.derivation.Origin()},
"l2Head", deferJSONString{s.derivation.UnsafeL2Head()},
"l2Safe", deferJSONString{s.derivation.SafeL2Head()},
"l2FinalizedHead", deferJSONString{s.derivation.Finalized()})
}
type hashAndError struct {
hash common.Hash
err error
}
type hashAndErrorChannel struct {
hash common.Hash
err chan error
}
// checkForGapInUnsafeQueue checks if there is a gap in the unsafe queue and attempts to retrieve the missing payloads from an alt-sync method.
// WARNING: This is only an outgoing signal, the blocks are not guaranteed to be retrieved.
// Results are received through OnUnsafeL2Payload.
func (s *Driver) checkForGapInUnsafeQueue(ctx context.Context) error {
start := s.derivation.UnsafeL2Head()
end := s.derivation.UnsafeL2SyncTarget()
// Check if we have missing blocks between the start and end. Request them if we do.
if end == (eth.L2BlockRef{}) {
s.log.Debug("requesting sync with open-end range", "start", start)
return s.altSync.RequestL2Range(ctx, start, eth.L2BlockRef{})
} else if end.Number > start.Number+1 {
s.log.Debug("requesting missing unsafe L2 block range", "start", start, "end", end, "size", end.Number-start.Number)
return s.altSync.RequestL2Range(ctx, start, end)
}
return nil
}