-
Notifications
You must be signed in to change notification settings - Fork 337
/
mode_put.go
540 lines (484 loc) · 15.5 KB
/
mode_put.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
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
// Copyright 2018 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 localstore
import (
"context"
"encoding/binary"
"encoding/hex"
"errors"
"fmt"
"time"
"github.com/ethersphere/bee/pkg/sharky"
"github.com/ethersphere/bee/pkg/shed"
"github.com/ethersphere/bee/pkg/storage"
"github.com/ethersphere/bee/pkg/swarm"
"github.com/syndtr/goleveldb/leveldb"
)
var (
ErrOverwriteImmutable = errors.New("index already exists - double issuance on immutable batch")
ErrOverwrite = errors.New("index already exists with newer timestamp - double issuance on batch")
)
// Put stores Chunks to database and depending
// on the Putter mode, it updates required indexes.
// Put is required to implement storage.Store
// interface.
func (db *DB) Put(ctx context.Context, mode storage.ModePut, chs ...swarm.Chunk) (exist []bool, err error) {
db.metrics.ModePut.Inc()
defer totalTimeMetric(db.metrics.TotalTimePut, time.Now())
exist, err = db.put(ctx, mode, chs...)
if err != nil {
db.metrics.ModePutFailure.Inc()
}
return exist, err
}
type releaseLocations []sharky.Location
func (r *releaseLocations) add(loc sharky.Location) {
*r = append(*r, loc)
}
// put stores Chunks to database and updates other indexes. It acquires batchMu
// to protect two calls of this function for the same address in parallel. Item
// fields Address and Data must not be with their nil values. If chunks with the
// same address are passed in arguments, only the first chunk will be stored,
// and following ones will have exist set to true for their index in exist
// slice. This is the same behaviour as if the same chunks are passed one by one
// in multiple put method calls.
func (db *DB) put(ctx context.Context, mode storage.ModePut, chs ...swarm.Chunk) (exist []bool, retErr error) {
// protect parallel updates
db.lock.Lock(lockKeyGC)
if db.gcRunning {
for _, ch := range chs {
db.dirtyAddresses = append(db.dirtyAddresses, ch.Address())
}
}
db.lock.Unlock(lockKeyGC)
batch := new(leveldb.Batch)
// variables that provide information for operations
// to be done after write batch function successfully executes
var (
gcSizeChange int64 // number to add or subtract from gcSize
)
var triggerPushFeed bool // signal push feed subscriptions to iterate
triggerPullFeed := make(map[uint8]struct{}) // signal pull feed subscriptions to iterate
exist = make([]bool, len(chs))
// A lazy populated map of bin ids to properly set
// BinID values for new chunks based on initial value from database
// and incrementing them.
// Values from this map are stored with the batch
binIDs := make(map[uint8]uint64)
var (
// this is the list of locations that need to be released if the batch is
// successfully committed due to postageIndex collisions
releaseLocs = new(releaseLocations)
// this is the list of locations that need to be released if the batch is NOT
// successfully committed as they have already been committed to sharky
committedLocations []sharky.Location
)
putChunk := func(ch swarm.Chunk, index int, putOp func(shed.Item, bool) (int64, error)) (bool, int64, error) {
if containsChunk(ch.Address(), chs[:index]...) {
return true, 0, nil
}
item := chunkToItem(ch)
storedItem, err := db.retrievalDataIndex.Get(item)
if err != nil && !errors.Is(err, leveldb.ErrNotFound) {
return false, 0, fmt.Errorf("failed reading retrievalIndex: %w", err)
}
if errors.Is(err, leveldb.ErrNotFound) {
// This is a new chunk so add to sharky. Also check for double issuance.
gcChange, err := db.checkAndRemoveStampIndex(item, batch, releaseLocs)
if err != nil {
if errors.Is(err, ErrOverwrite) && mode == storage.ModePutSync {
// if the chunk is overwriting a newer valid chunk for the
// same postage index, ignore it and dont return error so that
// syncing can continue
return false, 0, nil
}
return false, 0, err
}
l, err := db.sharky.Write(ctx, item.Data)
if err != nil {
return false, 0, fmt.Errorf("failed writing to sharky: %w", err)
}
committedLocations = append(committedLocations, l)
item.Location, err = l.MarshalBinary()
if err != nil {
return false, 0, fmt.Errorf("failed serializing sharky location: %w", err)
}
gcChangeNew, err := putOp(item, false)
return false, gcChangeNew + gcChange, err
}
gcChange, err := putOp(storedItem, true)
if err != nil {
return false, 0, err
}
return true, gcChange, nil
}
// If for whatever reason we fail to commit the batch, we should release all
// the chunks that have been committed to sharky
defer func() {
if retErr != nil {
for _, l := range committedLocations {
// the passed in context could be expired or cancelled, causing a leak by not relesing the
// already committed chunks, so we use an empty context
err := db.sharky.Release(context.Background(), l)
if err != nil {
db.logger.Warning("failed releasing sharky location on error", "error", err)
}
}
}
}()
switch mode {
case storage.ModePutRequest, storage.ModePutRequestPin, storage.ModePutRequestCache:
for i, ch := range chs {
pin := mode == storage.ModePutRequestPin // force pin in this mode
cache := mode == storage.ModePutRequestCache // force cache
exists, c, err := putChunk(ch, i, func(item shed.Item, exists bool) (int64, error) {
return db.putRequest(ctx, batch, binIDs, item, pin, cache, exists)
})
if err != nil {
return nil, fmt.Errorf("put request: %w", err)
}
exist[i] = exists
gcSizeChange += c
}
db.lock.Lock(lockKeyGC)
defer db.lock.Unlock(lockKeyGC)
case storage.ModePutUpload, storage.ModePutUploadPin:
for i, ch := range chs {
pin := mode == storage.ModePutUploadPin
exists, c, err := putChunk(ch, i, func(item shed.Item, exists bool) (int64, error) {
return db.putUpload(batch, binIDs, item, pin, exists)
})
if err != nil {
return nil, fmt.Errorf("put upload: %w", err)
}
exist[i] = exists
if !exists {
// chunk is new so, trigger subscription feeds
// after the batch is successfully written
triggerPushFeed = true
}
gcSizeChange += c
}
db.lock.Lock(lockKeyUpload)
defer db.lock.Unlock(lockKeyUpload)
case storage.ModePutSync:
for i, ch := range chs {
exists, c, err := putChunk(ch, i, func(item shed.Item, exists bool) (int64, error) {
return db.putSync(batch, binIDs, item, exists)
})
if err != nil {
return nil, fmt.Errorf("put sync: %w", err)
}
exist[i] = exists
if !exists {
// chunk is new so, trigger pull subscription feed
// after the batch is successfully written
triggerPullFeed[db.po(ch.Address())] = struct{}{}
}
gcSizeChange += c
}
db.lock.Lock(lockKeyGC)
defer db.lock.Unlock(lockKeyGC)
default:
return nil, ErrInvalidMode
}
for po, id := range binIDs {
db.binIDs.PutInBatch(batch, uint64(po), id)
}
err := db.incGCSizeInBatch(batch, gcSizeChange)
if err != nil {
return nil, fmt.Errorf("inc gc: %w", err)
}
err = db.shed.WriteBatch(batch)
if err != nil {
return nil, fmt.Errorf("write batch: %w", err)
}
for _, v := range *releaseLocs {
err = db.sharky.Release(ctx, v)
if err != nil {
db.logger.Warning("failed releasing sharky location", "location", v)
}
}
for po := range triggerPullFeed {
db.triggerPullSubscriptions(po)
}
if triggerPushFeed {
db.triggerPushSubscriptions()
}
return exist, nil
}
// checkAndRemoveStampIndex will check if we have the postageIndexIndex already taken
// for a particular {BatchID, BatchIndex}. If yes and the batch is immutable, we
// return error, if the batch is not immutable we replace the index to point to the
// new chunk if the timestamp of the new chunk is later.
// If the index is not taken, we do nothing. This is done to guard against
// overissuance of batches.
func (db *DB) checkAndRemoveStampIndex(
item shed.Item,
batch *leveldb.Batch,
loc *releaseLocations,
) (int64, error) {
previous, err := db.postageIndexIndex.Get(item)
if errors.Is(err, leveldb.ErrNotFound) {
return 0, nil
}
if err != nil {
return 0, fmt.Errorf("failed reading postageIndexIndex: %w", err)
}
if item.Immutable {
return 0, ErrOverwriteImmutable
}
// if a chunk is found with the same postage stamp index,
// replace it with the new one only if timestamp is later
if prev, cur := timestamps(previous, item); prev >= cur {
db.logger.Warning("postage stamp index exists", "prev", prev, "cur", cur, "chunk_address", hex.EncodeToString(item.Address))
return 0, ErrOverwrite
}
// remove older chunk
previousIdx, err := db.retrievalDataIndex.Get(previous)
if err != nil {
if errors.Is(err, leveldb.ErrNotFound) {
// currently there are stale postageIndexIndex entries in the localstore
// due to a bug found recently. This error is mainly ignored as the
// chunk is already gone and the index is overwritten.
db.logger.Debug("old postage stamp index missing", "Address", swarm.NewAddress(previous.Address))
return 0, nil
}
return 0, fmt.Errorf("could not fetch previous item: %w", err)
}
gcSizeChange, err := db.setRemove(batch, previousIdx, true)
if err != nil {
return 0, fmt.Errorf("setRemove on double issuance: %w", err)
}
l, err := sharky.LocationFromBinary(previousIdx.Location)
if err != nil {
return 0, fmt.Errorf("failed getting location: %w", err)
}
loc.add(l)
return gcSizeChange, nil
}
// putRequest adds an Item to the batch by updating required indexes:
// - put to indexes: retrieve, gc
// - it does not enter the syncpool
//
// The batch can be written to the database.
// Provided batch and binID map are updated.
func (db *DB) putRequest(
ctx context.Context,
batch *leveldb.Batch,
binIDs map[uint8]uint64,
item shed.Item,
forcePin, forceCache, exists bool,
) (int64, error) {
var err error
if !exists {
item.StoreTimestamp = now()
item.BinID, err = db.incBinID(binIDs, db.po(swarm.NewAddress(item.Address)))
if err != nil {
return 0, err
}
err = db.retrievalDataIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.postageChunksIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.postageIndexIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
item.AccessTimestamp = now()
err = db.retrievalAccessIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
}
// If forceCache is set, the stamp is invalid and we are storing it just for
// bandwidth incentives so we add to cache.
// If the request doesnt explicitly want to pin the chunk and it is not within
// our radius, we add it to cache. The 'within radius' part is a little debatable,
// but this is mainly done to opportunistically make the chunk available for pullSyncing.
if forceCache || (!forcePin && !withinRadiusFn(db, item)) {
return db.addToCache(batch, item)
}
if withinRadiusFn(db, item) {
found, err := db.pullIndex.Has(item)
if err != nil {
return 0, err
}
if found {
// this means it could be a duplicate put request. Dont update the
// pin counter.
return 0, nil
}
err = db.pullIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
}
return db.setPin(batch, item)
}
// putUpload adds an Item to the batch by updating required indexes:
// - put to indexes: retrieve, push
//
// The batch can be written to the database.
// Provided batch and binID map are updated.
func (db *DB) putUpload(
batch *leveldb.Batch,
binIDs map[uint8]uint64,
item shed.Item,
pin bool,
exists bool,
) (int64, error) {
var err error
if !exists {
item.StoreTimestamp = now()
item.BinID, err = db.incBinID(binIDs, db.po(swarm.NewAddress(item.Address)))
if err != nil {
return 0, fmt.Errorf("inc bin id: %w", err)
}
err = db.retrievalDataIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.postageIndexIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.postageChunksIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
}
err = db.pushIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
if pin {
return db.setPin(batch, item)
}
return 0, nil
}
// putSync adds an Item to the batch by updating required indexes:
// - put to indexes: retrieve, pull, gc
//
// The batch can be written to the database.
// Provided batch and binID map are updated.
func (db *DB) putSync(
batch *leveldb.Batch,
binIDs map[uint8]uint64,
item shed.Item,
exists bool,
) (gcSizeChange int64, err error) {
if !exists {
item.StoreTimestamp = now()
item.BinID, err = db.incBinID(binIDs, db.po(swarm.NewAddress(item.Address)))
if err != nil {
return 0, err
}
err = db.retrievalDataIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.postageChunksIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.postageIndexIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
item.AccessTimestamp = now()
err = db.retrievalAccessIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
}
found, err := db.pullIndex.Has(item)
if err != nil {
return 0, err
}
if found {
// this means it could be a duplicate put request. Dont update the
// pin counter.
return 0, nil
}
// if this is an existing chunk being Put with ModeSync, we just need to add
// the pullIndex and pin it
err = db.pullIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
return db.setPin(batch, item)
}
func (db *DB) addToCache(
batch *leveldb.Batch,
item shed.Item,
) (int64, error) {
// add new entry to gc index ONLY if it is not present in pinIndex
ok, err := db.pinIndex.Has(item)
if err != nil {
return 0, fmt.Errorf("failed checking pinIndex: %w", err)
}
if ok {
// if the chunk is pinned we dont add it to cache
return 0, nil
}
exists, err := db.gcIndex.Has(item)
if err != nil && !errors.Is(err, leveldb.ErrNotFound) {
return 0, err
}
if exists {
return 0, nil
}
err = db.gcIndex.PutInBatch(batch, item)
if err != nil {
return 0, err
}
err = db.pullIndex.DeleteInBatch(batch, item)
if err != nil {
return 0, err
}
return 1, nil
}
// incBinID is a helper function for db.put* methods that increments bin id
// based on the current value in the database. This function must be called under
// a db.batchMu lock. Provided binID map is updated.
func (db *DB) incBinID(binIDs map[uint8]uint64, po uint8) (id uint64, err error) {
if _, ok := binIDs[po]; !ok {
binIDs[po], err = db.binIDs.Get(uint64(po))
if err != nil {
return 0, err
}
}
binIDs[po]++
return binIDs[po], nil
}
// containsChunk returns true if the chunk with a specific address
// is present in the provided chunk slice.
func containsChunk(addr swarm.Address, chs ...swarm.Chunk) bool {
for _, c := range chs {
if addr.Equal(c.Address()) {
return true
}
}
return false
}
func timestamps(previous, current shed.Item) (uint64, uint64) {
return binary.BigEndian.Uint64(previous.Timestamp), binary.BigEndian.Uint64(current.Timestamp)
}