forked from cosmos/iavl
-
Notifications
You must be signed in to change notification settings - Fork 2
/
nodedb.go
478 lines (393 loc) · 12.4 KB
/
nodedb.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
package iavl
import (
"bytes"
"container/list"
"fmt"
"sort"
"sync"
"github.com/tendermint/tendermint/crypto/tmhash"
dbm "github.com/tendermint/tm-db"
)
const (
int64Size = 8
hashSize = tmhash.Size
)
var (
// All node keys are prefixed with the byte 'n'. This ensures no collision is
// possible with the other keys, and makes them easier to traverse. They are indexed by the node hash.
nodeKeyFormat = NewKeyFormat('n', hashSize) // n<hash>
// Orphans are keyed in the database by their expected lifetime.
// The first number represents the *last* version at which the orphan needs
// to exist, while the second number represents the *earliest* version at
// which it is expected to exist - which starts out by being the version
// of the node being orphaned.
orphanKeyFormat = NewKeyFormat('o', int64Size, int64Size, hashSize) // o<last-version><first-version><hash>
// Root nodes are indexed separately by their version
rootKeyFormat = NewKeyFormat('r', int64Size) // r<version>
)
type nodeDB struct {
mtx sync.Mutex // Read/write lock.
db dbm.DB // Persistent node storage.
batch dbm.Batch // Batched writing buffer.
latestVersion int64
nodeCache map[string]*list.Element // Node cache.
nodeCacheSize int // Node cache size limit in elements.
nodeCacheQueue *list.List // LRU queue of cache elements. Used for deletion.
}
func newNodeDB(db dbm.DB, cacheSize int) *nodeDB {
ndb := &nodeDB{
db: db,
batch: db.NewBatch(),
latestVersion: 0, // initially invalid
nodeCache: make(map[string]*list.Element),
nodeCacheSize: cacheSize,
nodeCacheQueue: list.New(),
}
return ndb
}
// GetNode gets a node from cache or disk. If it is an inner node, it does not
// load its children.
func (ndb *nodeDB) GetNode(hash []byte) *Node {
ndb.mtx.Lock()
defer ndb.mtx.Unlock()
if len(hash) == 0 {
panic("nodeDB.GetNode() requires hash")
}
// Check the cache.
if elem, ok := ndb.nodeCache[string(hash)]; ok {
// Already exists. Move to back of nodeCacheQueue.
ndb.nodeCacheQueue.MoveToBack(elem)
return elem.Value.(*Node)
}
// Doesn't exist, load.
buf := ndb.db.Get(ndb.nodeKey(hash))
if buf == nil {
panic(fmt.Sprintf("Value missing for hash %x corresponding to nodeKey %s", hash, ndb.nodeKey(hash)))
}
node, err := MakeNode(buf)
if err != nil {
panic(fmt.Sprintf("Error reading Node. bytes: %x, error: %v", buf, err))
}
node.hash = hash
node.persisted = true
ndb.cacheNode(node)
return node
}
// SaveNode saves a node to disk.
func (ndb *nodeDB) SaveNode(node *Node) {
ndb.mtx.Lock()
defer ndb.mtx.Unlock()
if node.hash == nil {
panic("Expected to find node.hash, but none found.")
}
if node.persisted {
panic("Shouldn't be calling save on an already persisted node.")
}
// Save node bytes to db.
buf := new(bytes.Buffer)
if err := node.writeBytes(buf); err != nil {
panic(err)
}
ndb.batch.Set(ndb.nodeKey(node.hash), buf.Bytes())
debug("BATCH SAVE %X %p\n", node.hash, node)
node.persisted = true
ndb.cacheNode(node)
}
// Has checks if a hash exists in the database.
func (ndb *nodeDB) Has(hash []byte) bool {
key := ndb.nodeKey(hash)
if ldb, ok := ndb.db.(*dbm.GoLevelDB); ok {
exists, err := ldb.DB().Has(key, nil)
if err != nil {
panic("Got error from leveldb: " + err.Error())
}
return exists
}
return ndb.db.Get(key) != nil
}
// SaveBranch saves the given node and all of its descendants.
// NOTE: This function clears leftNode/rigthNode recursively and
// calls _hash() on the given node.
// TODO refactor, maybe use hashWithCount() but provide a callback.
func (ndb *nodeDB) SaveBranch(node *Node) []byte {
if node.persisted {
return node.hash
}
if node.leftNode != nil {
node.leftHash = ndb.SaveBranch(node.leftNode)
}
if node.rightNode != nil {
node.rightHash = ndb.SaveBranch(node.rightNode)
}
node._hash()
ndb.SaveNode(node)
node.leftNode = nil
node.rightNode = nil
return node.hash
}
// DeleteVersion deletes a tree version from disk.
func (ndb *nodeDB) DeleteVersion(version int64, checkLatestVersion bool) {
ndb.mtx.Lock()
defer ndb.mtx.Unlock()
ndb.deleteOrphans(version)
ndb.deleteRoot(version, checkLatestVersion)
}
// Saves orphaned nodes to disk under a special prefix.
// version: the new version being saved.
// orphans: the orphan nodes created since version-1
func (ndb *nodeDB) SaveOrphans(version int64, orphans map[string]int64) {
ndb.mtx.Lock()
defer ndb.mtx.Unlock()
toVersion := ndb.getPreviousVersion(version)
for hash, fromVersion := range orphans {
debug("SAVEORPHAN %v-%v %X\n", fromVersion, toVersion, hash)
ndb.saveOrphan([]byte(hash), fromVersion, toVersion)
}
}
// Saves a single orphan to disk.
func (ndb *nodeDB) saveOrphan(hash []byte, fromVersion, toVersion int64) {
if fromVersion > toVersion {
panic(fmt.Sprintf("Orphan expires before it comes alive. %d > %d", fromVersion, toVersion))
}
key := ndb.orphanKey(fromVersion, toVersion, hash)
ndb.batch.Set(key, hash)
}
// deleteOrphans deletes orphaned nodes from disk, and the associated orphan
// entries.
func (ndb *nodeDB) deleteOrphans(version int64) {
// Will be zero if there is no previous version.
predecessor := ndb.getPreviousVersion(version)
// Traverse orphans with a lifetime ending at the version specified.
// TODO optimize.
ndb.traverseOrphansVersion(version, func(key, hash []byte) {
var fromVersion, toVersion int64
// See comment on `orphanKeyFmt`. Note that here, `version` and
// `toVersion` are always equal.
orphanKeyFormat.Scan(key, &toVersion, &fromVersion)
// Delete orphan key and reverse-lookup key.
ndb.batch.Delete(key)
// If there is no predecessor, or the predecessor is earlier than the
// beginning of the lifetime (ie: negative lifetime), or the lifetime
// spans a single version and that version is the one being deleted, we
// can delete the orphan. Otherwise, we shorten its lifetime, by
// moving its endpoint to the previous version.
if predecessor < fromVersion || fromVersion == toVersion {
debug("DELETE predecessor:%v fromVersion:%v toVersion:%v %X\n", predecessor, fromVersion, toVersion, hash)
ndb.batch.Delete(ndb.nodeKey(hash))
ndb.uncacheNode(hash)
} else {
debug("MOVE predecessor:%v fromVersion:%v toVersion:%v %X\n", predecessor, fromVersion, toVersion, hash)
ndb.saveOrphan(hash, fromVersion, predecessor)
}
})
}
func (ndb *nodeDB) nodeKey(hash []byte) []byte {
return nodeKeyFormat.KeyBytes(hash)
}
func (ndb *nodeDB) orphanKey(fromVersion, toVersion int64, hash []byte) []byte {
return orphanKeyFormat.Key(toVersion, fromVersion, hash)
}
func (ndb *nodeDB) rootKey(version int64) []byte {
return rootKeyFormat.Key(version)
}
func (ndb *nodeDB) getLatestVersion() int64 {
if ndb.latestVersion == 0 {
ndb.latestVersion = ndb.getPreviousVersion(1<<63 - 1)
}
return ndb.latestVersion
}
func (ndb *nodeDB) updateLatestVersion(version int64) {
if ndb.latestVersion < version {
ndb.latestVersion = version
}
}
func (ndb *nodeDB) resetLatestVersion(version int64) {
ndb.latestVersion = version
}
func (ndb *nodeDB) getPreviousVersion(version int64) int64 {
itr := ndb.db.ReverseIterator(
rootKeyFormat.Key(1),
rootKeyFormat.Key(version),
)
defer itr.Close()
pversion := int64(-1)
for ; itr.Valid(); itr.Next() {
k := itr.Key()
rootKeyFormat.Scan(k, &pversion)
return pversion
}
return 0
}
// deleteRoot deletes the root entry from disk, but not the node it points to.
func (ndb *nodeDB) deleteRoot(version int64, checkLatestVersion bool) {
if checkLatestVersion && version == ndb.getLatestVersion() {
panic("Tried to delete latest version")
}
key := ndb.rootKey(version)
ndb.batch.Delete(key)
}
func (ndb *nodeDB) traverseOrphans(fn func(k, v []byte)) {
ndb.traversePrefix(orphanKeyFormat.Key(), fn)
}
// Traverse orphans ending at a certain version.
func (ndb *nodeDB) traverseOrphansVersion(version int64, fn func(k, v []byte)) {
ndb.traversePrefix(orphanKeyFormat.Key(version), fn)
}
// Traverse all keys.
func (ndb *nodeDB) traverse(fn func(key, value []byte)) {
itr := ndb.db.Iterator(nil, nil)
defer itr.Close()
for ; itr.Valid(); itr.Next() {
fn(itr.Key(), itr.Value())
}
}
// Traverse all keys with a certain prefix.
func (ndb *nodeDB) traversePrefix(prefix []byte, fn func(k, v []byte)) {
itr := dbm.IteratePrefix(ndb.db, prefix)
defer itr.Close()
for ; itr.Valid(); itr.Next() {
fn(itr.Key(), itr.Value())
}
}
func (ndb *nodeDB) uncacheNode(hash []byte) {
if elem, ok := ndb.nodeCache[string(hash)]; ok {
ndb.nodeCacheQueue.Remove(elem)
delete(ndb.nodeCache, string(hash))
}
}
// Add a node to the cache and pop the least recently used node if we've
// reached the cache size limit.
func (ndb *nodeDB) cacheNode(node *Node) {
elem := ndb.nodeCacheQueue.PushBack(node)
ndb.nodeCache[string(node.hash)] = elem
if ndb.nodeCacheQueue.Len() > ndb.nodeCacheSize {
oldest := ndb.nodeCacheQueue.Front()
hash := ndb.nodeCacheQueue.Remove(oldest).(*Node).hash
delete(ndb.nodeCache, string(hash))
}
}
// Write to disk.
func (ndb *nodeDB) Commit() {
ndb.mtx.Lock()
defer ndb.mtx.Unlock()
ndb.batch.Write()
ndb.batch.Close()
ndb.batch = ndb.db.NewBatch()
}
func (ndb *nodeDB) getRoot(version int64) []byte {
return ndb.db.Get(ndb.rootKey(version))
}
func (ndb *nodeDB) getRoots() (map[int64][]byte, error) {
roots := map[int64][]byte{}
ndb.traversePrefix(rootKeyFormat.Key(), func(k, v []byte) {
var version int64
rootKeyFormat.Scan(k, &version)
roots[version] = v
})
return roots, nil
}
// SaveRoot creates an entry on disk for the given root, so that it can be
// loaded later.
func (ndb *nodeDB) SaveRoot(root *Node, version int64) error {
if len(root.hash) == 0 {
panic("Hash should not be empty")
}
return ndb.saveRoot(root.hash, version)
}
// SaveEmptyRoot creates an entry on disk for an empty root.
func (ndb *nodeDB) SaveEmptyRoot(version int64) error {
return ndb.saveRoot([]byte{}, version)
}
func (ndb *nodeDB) saveRoot(hash []byte, version int64) error {
ndb.mtx.Lock()
defer ndb.mtx.Unlock()
if version != ndb.getLatestVersion()+1 {
return fmt.Errorf("Must save consecutive versions. Expected %d, got %d", ndb.getLatestVersion()+1, version)
}
key := ndb.rootKey(version)
ndb.batch.Set(key, hash)
ndb.updateLatestVersion(version)
return nil
}
////////////////// Utility and test functions /////////////////////////////////
func (ndb *nodeDB) leafNodes() []*Node {
leaves := []*Node{}
ndb.traverseNodes(func(hash []byte, node *Node) {
if node.isLeaf() {
leaves = append(leaves, node)
}
})
return leaves
}
func (ndb *nodeDB) nodes() []*Node {
nodes := []*Node{}
ndb.traverseNodes(func(hash []byte, node *Node) {
nodes = append(nodes, node)
})
return nodes
}
func (ndb *nodeDB) orphans() [][]byte {
orphans := [][]byte{}
ndb.traverseOrphans(func(k, v []byte) {
orphans = append(orphans, v)
})
return orphans
}
func (ndb *nodeDB) roots() map[int64][]byte {
roots, _ := ndb.getRoots()
return roots
}
// Not efficient.
// NOTE: DB cannot implement Size() because
// mutations are not always synchronous.
func (ndb *nodeDB) size() int {
size := 0
ndb.traverse(func(k, v []byte) {
size++
})
return size
}
func (ndb *nodeDB) traverseNodes(fn func(hash []byte, node *Node)) {
nodes := []*Node{}
ndb.traversePrefix(nodeKeyFormat.Key(), func(key, value []byte) {
node, err := MakeNode(value)
if err != nil {
panic(fmt.Sprintf("Couldn't decode node from database: %v", err))
}
nodeKeyFormat.Scan(key, &node.hash)
nodes = append(nodes, node)
})
sort.Slice(nodes, func(i, j int) bool {
return bytes.Compare(nodes[i].key, nodes[j].key) < 0
})
for _, n := range nodes {
fn(n.hash, n)
}
}
func (ndb *nodeDB) String() string {
var str string
index := 0
ndb.traversePrefix(rootKeyFormat.Key(), func(key, value []byte) {
str += fmt.Sprintf("%s: %x\n", string(key), value)
})
str += "\n"
ndb.traverseOrphans(func(key, value []byte) {
str += fmt.Sprintf("%s: %x\n", string(key), value)
})
str += "\n"
ndb.traverseNodes(func(hash []byte, node *Node) {
if len(hash) == 0 {
str += fmt.Sprintf("<nil>\n")
} else if node == nil {
str += fmt.Sprintf("%s%40x: <nil>\n", nodeKeyFormat.Prefix(), hash)
} else if node.value == nil && node.height > 0 {
str += fmt.Sprintf("%s%40x: %s %-16s h=%d version=%d\n",
nodeKeyFormat.Prefix(), hash, node.key, "", node.height, node.version)
} else {
str += fmt.Sprintf("%s%40x: %s = %-16s h=%d version=%d\n",
nodeKeyFormat.Prefix(), hash, node.key, node.value, node.height, node.version)
}
index++
})
return "-" + "\n" + str + "-"
}