aboutsummaryrefslogtreecommitdiffstats
path: root/core/chain_indexer.go
blob: 0b927116d066730c7a4a52ed4562451d1f549b98 (plain) (blame)
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
// Copyright 2017 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum 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 go-ethereum 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 go-ethereum library. If not, see <http://www.gnu.org/licenses/>.

package core

import (
    "encoding/binary"
    "fmt"
    "sync"
    "sync/atomic"
    "time"

    "github.com/ethereum/go-ethereum/common"
    "github.com/ethereum/go-ethereum/core/rawdb"
    "github.com/ethereum/go-ethereum/core/types"
    "github.com/ethereum/go-ethereum/ethdb"
    "github.com/ethereum/go-ethereum/event"
    "github.com/ethereum/go-ethereum/log"
)

// ChainIndexerBackend defines the methods needed to process chain segments in
// the background and write the segment results into the database. These can be
// used to create filter blooms or CHTs.
type ChainIndexerBackend interface {
    // Reset initiates the processing of a new chain segment, potentially terminating
    // any partially completed operations (in case of a reorg).
    Reset(section uint64, prevHead common.Hash) error

    // Process crunches through the next header in the chain segment. The caller
    // will ensure a sequential order of headers.
    Process(header *types.Header)

    // Commit finalizes the section metadata and stores it into the database.
    Commit() error
}

// ChainIndexerChain interface is used for connecting the indexer to a blockchain
type ChainIndexerChain interface {
    // CurrentHeader retrieves the latest locally known header.
    CurrentHeader() *types.Header

    // SubscribeChainEvent subscribes to new head header notifications.
    SubscribeChainEvent(ch chan<- ChainEvent) event.Subscription
}

// ChainIndexer does a post-processing job for equally sized sections of the
// canonical chain (like BlooomBits and CHT structures). A ChainIndexer is
// connected to the blockchain through the event system by starting a
// ChainEventLoop in a goroutine.
//
// Further child ChainIndexers can be added which use the output of the parent
// section indexer. These child indexers receive new head notifications only
// after an entire section has been finished or in case of rollbacks that might
// affect already finished sections.
type ChainIndexer struct {
    chainDb  ethdb.Database      // Chain database to index the data from
    indexDb  ethdb.Database      // Prefixed table-view of the db to write index metadata into
    backend  ChainIndexerBackend // Background processor generating the index data content
    children []*ChainIndexer     // Child indexers to cascade chain updates to

    active uint32          // Flag whether the event loop was started
    update chan struct{}   // Notification channel that headers should be processed
    quit   chan chan error // Quit channel to tear down running goroutines

    sectionSize uint64 // Number of blocks in a single chain segment to process
    confirmsReq uint64 // Number of confirmations before processing a completed segment

    storedSections uint64 // Number of sections successfully indexed into the database
    knownSections  uint64 // Number of sections known to be complete (block wise)
    cascadedHead   uint64 // Block number of the last completed section cascaded to subindexers

    throttling time.Duration // Disk throttling to prevent a heavy upgrade from hogging resources

    log  log.Logger
    lock sync.RWMutex
}

// NewChainIndexer creates a new chain indexer to do background processing on
// chain segments of a given size after certain number of confirmations passed.
// The throttling parameter might be used to prevent database thrashing.
func NewChainIndexer(chainDb, indexDb ethdb.Database, backend ChainIndexerBackend, section, confirm uint64, throttling time.Duration, kind string) *ChainIndexer {
    c := &ChainIndexer{
        chainDb:     chainDb,
        indexDb:     indexDb,
        backend:     backend,
        update:      make(chan struct{}, 1),
        quit:        make(chan chan error),
        sectionSize: section,
        confirmsReq: confirm,
        throttling:  throttling,
        log:         log.New("type", kind),
    }
    // Initialize database dependent fields and start the updater
    c.loadValidSections()
    go c.updateLoop()

    return c
}

// AddKnownSectionHead marks a new section head as known/processed if it is newer
// than the already known best section head
func (c *ChainIndexer) AddKnownSectionHead(section uint64, shead common.Hash) {
    c.lock.Lock()
    defer c.lock.Unlock()

    if section < c.storedSections {
        return
    }
    c.setSectionHead(section, shead)
    c.setValidSections(section + 1)
}

// Start creates a goroutine to feed chain head events into the indexer for
// cascading background processing. Children do not need to be started, they
// are notified about new events by their parents.
func (c *ChainIndexer) Start(chain ChainIndexerChain) {
    events := make(chan ChainEvent, 10)
    sub := chain.SubscribeChainEvent(events)

    go c.eventLoop(chain.CurrentHeader(), events, sub)
}

// Close tears down all goroutines belonging to the indexer and returns any error
// that might have occurred internally.
func (c *ChainIndexer) Close() error {
    var errs []error

    // Tear down the primary update loop
    errc := make(chan error)
    c.quit <- errc
    if err := <-errc; err != nil {
        errs = append(errs, err)
    }
    // If needed, tear down the secondary event loop
    if atomic.LoadUint32(&c.active) != 0 {
        c.quit <- errc
        if err := <-errc; err != nil {
            errs = append(errs, err)
        }
    }
    // Close all children
    for _, child := range c.children {
        if err := child.Close(); err != nil {
            errs = append(errs, err)
        }
    }
    // Return any failures
    switch {
    case len(errs) == 0:
        return nil

    case len(errs) == 1:
        return errs[0]

    default:
        return fmt.Errorf("%v", errs)
    }
}

// eventLoop is a secondary - optional - event loop of the indexer which is only
// started for the outermost indexer to push chain head events into a processing
// queue.
func (c *ChainIndexer) eventLoop(currentHeader *types.Header, events chan ChainEvent, sub event.Subscription) {
    // Mark the chain indexer as active, requiring an additional teardown
    atomic.StoreUint32(&c.active, 1)

    defer sub.Unsubscribe()

    // Fire the initial new head event to start any outstanding processing
    c.newHead(currentHeader.Number.Uint64(), false)

    var (
        prevHeader = currentHeader
        prevHash   = currentHeader.Hash()
    )
    for {
        select {
        case errc := <-c.quit:
            // Chain indexer terminating, report no failure and abort
            errc <- nil
            return

        case ev, ok := <-events:
            // Received a new event, ensure it's not nil (closing) and update
            if !ok {
                errc := <-c.quit
                errc <- nil
                return
            }
            header := ev.Block.Header()
            if header.ParentHash != prevHash {
                // Reorg to the common ancestor (might not exist in light sync mode, skip reorg then)
                // TODO(karalabe, zsfelfoldi): This seems a bit brittle, can we detect this case explicitly?

                // TODO(karalabe): This operation is expensive and might block, causing the event system to
                // potentially also lock up. We need to do with on a different thread somehow.
                if h := rawdb.FindCommonAncestor(c.chainDb, prevHeader, header); h != nil {
                    c.newHead(h.Number.Uint64(), true)
                }
            }
            c.newHead(header.Number.Uint64(), false)

            prevHeader, prevHash = header, header.Hash()
        }
    }
}

// newHead notifies the indexer about new chain heads and/or reorgs.
func (c *ChainIndexer) newHead(head uint64, reorg bool) {
    c.lock.Lock()
    defer c.lock.Unlock()

    // If a reorg happened, invalidate all sections until that point
    if reorg {
        // Revert the known section number to the reorg point
        changed := head / c.sectionSize
        if changed < c.knownSections {
            c.knownSections = changed
        }
        // Revert the stored sections from the database to the reorg point
        if changed < c.storedSections {
            c.setValidSections(changed)
        }
        // Update the new head number to the finalized section end and notify children
        head = changed * c.sectionSize

        if head < c.cascadedHead {
            c.cascadedHead = head
            for _, child := range c.children {
                child.newHead(c.cascadedHead, true)
            }
        }
        return
    }
    // No reorg, calculate the number of newly known sections and update if high enough
    var sections uint64
    if head >= c.confirmsReq {
        sections = (head + 1 - c.confirmsReq) / c.sectionSize
        if sections > c.knownSections {
            c.knownSections = sections

            select {
            case c.update <- struct{}{}:
            default:
            }
        }
    }
}

// updateLoop is the main event loop of the indexer which pushes chain segments
// down into the processing backend.
func (c *ChainIndexer) updateLoop() {
    var (
        updating bool
        updated  time.Time
    )

    for {
        select {
        case errc := <-c.quit:
            // Chain indexer terminating, report no failure and abort
            errc <- nil
            return

        case <-c.update:
            // Section headers completed (or rolled back), update the index
            c.lock.Lock()
            if c.knownSections > c.storedSections {
                // Periodically print an upgrade log message to the user
                if time.Since(updated) > 8*time.Second {
                    if c.knownSections > c.storedSections+1 {
                        updating = true
                        c.log.Info("Upgrading chain index", "percentage", c.storedSections*100/c.knownSections)
                    }
                    updated = time.Now()
                }
                // Cache the current section count and head to allow unlocking the mutex
                section := c.storedSections
                var oldHead common.Hash
                if section > 0 {
                    oldHead = c.SectionHead(section - 1)
                }
                // Process the newly defined section in the background
                c.lock.Unlock()
                newHead, err := c.processSection(section, oldHead)
                if err != nil {
                    c.log.Error("Section processing failed", "error", err)
                }
                c.lock.Lock()

                // If processing succeeded and no reorgs occcurred, mark the section completed
                if err == nil && oldHead == c.SectionHead(section-1) {
                    c.setSectionHead(section, newHead)
                    c.setValidSections(section + 1)
                    if c.storedSections == c.knownSections && updating {
                        updating = false
                        c.log.Info("Finished upgrading chain index")
                    }

                    c.cascadedHead = c.storedSections*c.sectionSize - 1
                    for _, child := range c.children {
                        c.log.Trace("Cascading chain index update", "head", c.cascadedHead)
                        child.newHead(c.cascadedHead, false)
                    }
                } else {
                    // If processing failed, don't retry until further notification
                    c.log.Debug("Chain index processing failed", "section", section, "err", err)
                    c.knownSections = c.storedSections
                }
            }
            // If there are still further sections to process, reschedule
            if c.knownSections > c.storedSections {
                time.AfterFunc(c.throttling, func() {
                    select {
                    case c.update <- struct{}{}:
                    default:
                    }
                })
            }
            c.lock.Unlock()
        }
    }
}

// processSection processes an entire section by calling backend functions while
// ensuring the continuity of the passed headers. Since the chain mutex is not
// held while processing, the continuity can be broken by a long reorg, in which
// case the function returns with an error.
func (c *ChainIndexer) processSection(section uint64, lastHead common.Hash) (common.Hash, error) {
    c.log.Trace("Processing new chain section", "section", section)

    // Reset and partial processing

    if err := c.backend.Reset(section, lastHead); err != nil {
        c.setValidSections(0)
        return common.Hash{}, err
    }

    for number := section * c.sectionSize; number < (section+1)*c.sectionSize; number++ {
        hash := rawdb.ReadCanonicalHash(c.chainDb, number)
        if hash == (common.Hash{}) {
            return common.Hash{}, fmt.Errorf("canonical block #%d unknown", number)
        }
        header := rawdb.ReadHeader(c.chainDb, hash, number)
        if header == nil {
            return common.Hash{}, fmt.Errorf("block #%d [%x…] not found", number, hash[:4])
        } else if header.ParentHash != lastHead {
            return common.Hash{}, fmt.Errorf("chain reorged during section processing")
        }
        c.backend.Process(header)
        lastHead = header.Hash()
    }
    if err := c.backend.Commit(); err != nil {
        c.log.Error("Section commit failed", "error", err)
        return common.Hash{}, err
    }
    return lastHead, nil
}

// Sections returns the number of processed sections maintained by the indexer
// and also the information about the last header indexed for potential canonical
// verifications.
func (c *ChainIndexer) Sections() (uint64, uint64, common.Hash) {
    c.lock.Lock()
    defer c.lock.Unlock()

    return c.storedSections, c.storedSections*c.sectionSize - 1, c.SectionHead(c.storedSections - 1)
}

// AddChildIndexer adds a child ChainIndexer that can use the output of this one
func (c *ChainIndexer) AddChildIndexer(indexer *ChainIndexer) {
    c.lock.Lock()
    defer c.lock.Unlock()

    c.children = append(c.children, indexer)

    // Cascade any pending updates to new children too
    if c.storedSections > 0 {
        indexer.newHead(c.storedSections*c.sectionSize-1, false)
    }
}

// loadValidSections reads the number of valid sections from the index database
// and caches is into the local state.
func (c *ChainIndexer) loadValidSections() {
    data, _ := c.indexDb.Get([]byte("count"))
    if len(data) == 8 {
        c.storedSections = binary.BigEndian.Uint64(data[:])
    }
}

// setValidSections writes the number of valid sections to the index database
func (c *ChainIndexer) setValidSections(sections uint64) {
    // Set the current number of valid sections in the database
    var data [8]byte
    binary.BigEndian.PutUint64(data[:], sections)
    c.indexDb.Put([]byte("count"), data[:])

    // Remove any reorged sections, caching the valids in the mean time
    for c.storedSections > sections {
        c.storedSections--
        c.removeSectionHead(c.storedSections)
    }
    c.storedSections = sections // needed if new > old
}

// SectionHead retrieves the last block hash of a processed section from the
// index database.
func (c *ChainIndexer) SectionHead(section uint64) common.Hash {
    var data [8]byte
    binary.BigEndian.PutUint64(data[:], section)

    hash, _ := c.indexDb.Get(append([]byte("shead"), data[:]...))
    if len(hash) == len(common.Hash{}) {
        return common.BytesToHash(hash)
    }
    return common.Hash{}
}

// setSectionHead writes the last block hash of a processed section to the index
// database.
func (c *ChainIndexer) setSectionHead(section uint64, hash common.Hash) {
    var data [8]byte
    binary.BigEndian.PutUint64(data[:], section)

    c.indexDb.Put(append([]byte("shead"), data[:]...), hash.Bytes())
}

// removeSectionHead removes the reference to a processed section from the index
// database.
func (c *ChainIndexer) removeSectionHead(section uint64) {
    var data [8]byte
    binary.BigEndian.PutUint64(data[:], section)

    c.indexDb.Delete(append([]byte("shead"), data[:]...))
}