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
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
|
// Copyright (C) 2022 Luke Shumaker <lukeshu@lukeshu.com>
//
// SPDX-License-Identifier: GPL-2.0-or-later
package rebuildnodes
import (
"context"
"fmt"
"sort"
"time"
"github.com/datawire/dlib/dlog"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfs"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfs/btrfsitem"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfs/btrfsprim"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfs/btrfstree"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfs/btrfsvol"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfsprogs/btrfsinspect"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfsprogs/btrfsinspect/rebuildnodes/btrees"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfsprogs/btrfsinspect/rebuildnodes/graph"
"git.lukeshu.com/btrfs-progs-ng/lib/btrfsprogs/btrfsinspect/rebuildnodes/keyio"
"git.lukeshu.com/btrfs-progs-ng/lib/containers"
"git.lukeshu.com/btrfs-progs-ng/lib/maps"
"git.lukeshu.com/btrfs-progs-ng/lib/textui"
)
type keyAndTree struct {
btrfsprim.Key
TreeID btrfsprim.ObjID
}
func (a keyAndTree) Cmp(b keyAndTree) int {
if d := a.Key.Cmp(b.Key); d != 0 {
return d
}
return containers.NativeCmp(a.TreeID, b.TreeID)
}
type rebuilder struct {
sb btrfstree.Superblock
rebuilt *btrees.RebuiltTrees
graph graph.Graph
keyIO *keyio.Handle
curKey keyAndTree
queue []keyAndTree
pendingAugments map[btrfsprim.ObjID][]map[btrfsvol.LogicalAddr]int
}
func RebuildNodes(ctx context.Context, fs *btrfs.FS, nodeScanResults btrfsinspect.ScanDevicesResult) (map[btrfsprim.ObjID]containers.Set[btrfsvol.LogicalAddr], error) {
nodeGraph, err := ScanDevices(ctx, fs, nodeScanResults) // ScanDevices does its own logging
if err != nil {
return nil, err
}
dlog.Info(ctx, "Reading superblock...")
sb, err := fs.Superblock()
if err != nil {
return nil, err
}
dlog.Info(ctx, "Rebuilding node tree...")
keyIO := keyio.NewHandle(fs, *sb, nodeGraph)
o := &rebuilder{
sb: *sb,
graph: nodeGraph,
keyIO: keyIO,
}
o.rebuilt = btrees.NewRebuiltTrees(*sb, nodeGraph, keyIO,
o.cbAddedItem, o.cbLookupRoot, o.cbLookupUUID)
if err := o.rebuild(ctx); err != nil {
return nil, err
}
return o.rebuilt.ListRoots(), nil
}
func (o *rebuilder) ioErr(ctx context.Context, err error) {
err = fmt.Errorf("should not happen: i/o error: %w", err)
dlog.Error(ctx, err)
panic(err)
}
type rebuildStats struct {
PassNum int
Task string
N, D int
}
func (s rebuildStats) String() string {
pct := 100
if s.D > 0 {
pct = int(100 * float64(s.N) / float64(s.D))
}
return fmt.Sprintf("... pass %d: %s %v%% (%v/%v)",
s.PassNum, s.Task, pct, s.N, s.D)
}
func (o *rebuilder) rebuild(ctx context.Context) error {
passNum := 0
dlog.Infof(ctx, "... pass %d: scanning for implied items", passNum)
// Seed the queue
o.pendingAugments = make(map[btrfsprim.ObjID][]map[btrfsvol.LogicalAddr]int)
o.rebuilt.AddTree(ctx, btrfsprim.ROOT_TREE_OBJECTID)
o.rebuilt.AddTree(ctx, btrfsprim.CHUNK_TREE_OBJECTID)
//o.rebuilt.AddTree(ctx, btrfsprim.TREE_LOG_OBJECTID) // TODO(lukeshu): Special LOG_TREE handling
o.rebuilt.AddTree(ctx, btrfsprim.BLOCK_GROUP_TREE_OBJECTID)
for {
// Handle items in the queue
queue := o.queue
o.queue = nil
progressWriter := textui.NewProgress[rebuildStats](ctx, dlog.LogLevelInfo, 1*time.Second)
queueProgress := func(done int) {
progressWriter.Set(rebuildStats{
PassNum: passNum,
Task: "processing item queue",
N: done,
D: len(queue),
})
}
for i, key := range queue {
queueProgress(i)
o.curKey = key
itemBody, ok := o.rebuilt.Load(ctx, key.TreeID, key.Key)
if !ok {
o.ioErr(ctx, fmt.Errorf("could not read previously read item: %v", key))
}
handleItem(o, ctx, key.TreeID, btrfstree.Item{
Key: key.Key,
Body: itemBody,
})
if key.ItemType == btrfsitem.ROOT_ITEM_KEY {
o.rebuilt.AddTree(ctx, key.ObjectID)
}
}
queueProgress(len(queue))
progressWriter.Done()
// Check if we can bail
if len(o.queue) == 0 && len(o.pendingAugments) == 0 {
break
}
// Apply augments that were requested while handling items from the queue
resolvedAugments := make(map[btrfsprim.ObjID]containers.Set[btrfsvol.LogicalAddr], len(o.pendingAugments))
numAugments := 0
for _, treeID := range maps.SortedKeys(o.pendingAugments) {
dlog.Infof(ctx, "... ... augments for tree %v:", treeID)
resolvedAugments[treeID] = o.resolveTreeAugments(ctx, o.pendingAugments[treeID])
numAugments += len(resolvedAugments[treeID])
}
o.pendingAugments = make(map[btrfsprim.ObjID][]map[btrfsvol.LogicalAddr]int)
progressWriter = textui.NewProgress[rebuildStats](ctx, dlog.LogLevelInfo, 1*time.Second)
numAugmented := 0
augmentProgress := func() {
progressWriter.Set(rebuildStats{
PassNum: passNum,
Task: "applying augments",
N: numAugmented,
D: numAugments,
})
}
for _, treeID := range maps.SortedKeys(resolvedAugments) {
for _, nodeAddr := range maps.SortedKeys(resolvedAugments[treeID]) {
augmentProgress()
o.rebuilt.AddRoot(ctx, treeID, nodeAddr)
numAugmented++
}
}
augmentProgress()
progressWriter.Done()
passNum++
dlog.Infof(ctx, "... pass %d: scanning for implied items", passNum)
}
return nil
}
func (o *rebuilder) cbAddedItem(ctx context.Context, tree btrfsprim.ObjID, key btrfsprim.Key) {
o.queue = append(o.queue, keyAndTree{
TreeID: tree,
Key: key,
})
}
func (o *rebuilder) cbLookupRoot(ctx context.Context, tree btrfsprim.ObjID) (offset btrfsprim.Generation, item btrfsitem.Root, ok bool) {
key, ok := o._want(ctx, btrfsprim.ROOT_TREE_OBJECTID, tree, btrfsitem.ROOT_ITEM_KEY)
if !ok {
o.queue = append(o.queue, o.curKey)
return 0, btrfsitem.Root{}, false
}
itemBody, ok := o.rebuilt.Load(ctx, btrfsprim.ROOT_TREE_OBJECTID, key)
if !ok {
o.ioErr(ctx, fmt.Errorf("could not read previously read item: %v", key))
}
switch itemBody := itemBody.(type) {
case btrfsitem.Root:
return btrfsprim.Generation(key.Offset), itemBody, true
case btrfsitem.Error:
o.fsErr(ctx, fmt.Errorf("error decoding item: tree=%v key=%v: %w", btrfsprim.ROOT_TREE_OBJECTID, key, itemBody.Err))
return 0, btrfsitem.Root{}, false
default:
// This is a panic because the item decoder should not emit ROOT_ITEM items as anything but
// btrfsitem.Root or btrfsitem.Error without this code also being updated.
panic(fmt.Errorf("should not happen: ROOT_ITEM item has unexpected type: %T", itemBody))
}
}
func (o *rebuilder) cbLookupUUID(ctx context.Context, uuid btrfsprim.UUID) (id btrfsprim.ObjID, ok bool) {
key := btrfsitem.UUIDToKey(uuid)
if ok := o._wantOff(ctx, btrfsprim.UUID_TREE_OBJECTID, key.ObjectID, key.ItemType, key.Offset); !ok {
o.queue = append(o.queue, o.curKey)
return 0, false
}
itemBody, ok := o.rebuilt.Load(ctx, btrfsprim.UUID_TREE_OBJECTID, key)
if !ok {
o.ioErr(ctx, fmt.Errorf("could not read previously read item: %v", key))
}
switch itemBody := itemBody.(type) {
case btrfsitem.UUIDMap:
return itemBody.ObjID, true
case btrfsitem.Error:
o.fsErr(ctx, fmt.Errorf("error decoding item: tree=%v key=%v: %w", btrfsprim.UUID_TREE_OBJECTID, key, itemBody.Err))
return 0, false
default:
// This is a panic because the item decoder should not emit UUID_SUBVOL items as anything but
// btrfsitem.UUIDMap or btrfsitem.Error without this code also being updated.
panic(fmt.Errorf("should not happen: UUID_SUBVOL item has unexpected type: %T", itemBody))
}
}
func (o *rebuilder) resolveTreeAugments(ctx context.Context, listsWithDistances []map[btrfsvol.LogicalAddr]int) containers.Set[btrfsvol.LogicalAddr] {
distances := make(map[btrfsvol.LogicalAddr]int)
generations := make(map[btrfsvol.LogicalAddr]btrfsprim.Generation)
lists := make([]containers.Set[btrfsvol.LogicalAddr], len(listsWithDistances))
for i, listWithDistances := range listsWithDistances {
lists[i] = make(containers.Set[btrfsvol.LogicalAddr], len(listWithDistances))
for item, dist := range listWithDistances {
lists[i].Insert(item)
distances[item] = dist
generations[item] = o.graph.Nodes[item].Generation
}
}
// Define an algorithm that takes several lists of items, and returns a
// set of those items such that each input list contains zero or one of
// the items from your return set. The same item may appear in multiple
// of the input lists.
//
// > Example 1: Given the input lists
// >
// > 0: [A, B]
// > 2: [A, C]
// >
// > legal solutions would be `[]`, `[A]`, `[B]`, `[C]`, or `[B, C]`. It
// > would not be legal to return `[A, B]` or `[A, C]`.
//
// > Example 2: Given the input lists
// >
// > 1: [A, B]
// > 2: [A]
// > 3: [B]
// >
// > legal solution woudl be `[]`, `[A]` or `[B]`. It would not be legal
// > to return `[A, B]`.
//
// The algorithm should optimize for the following goals:
//
// - We prefer that each input list have an item in the return set.
//
// > In Example 1, while `[]`, `[B]`, and `[C]` are permissable
// > solutions, they are not optimal, because one or both of the input
// > lists are not represented.
// >
// > It may be the case that it is not possible to represent all lists
// > in the result; in Example 2, either list 2 or list 3 must be
// > unrepresented.
//
// - Each item has a non-negative scalar "distance" score, we prefer
// lower distances. Distance scores are comparable; 0 is preferred,
// and a distance of 4 is twice as bad as a distance of 2.
//
// - Each item has a "generation" score, we prefer higher generations.
// Generation scores should not be treated as a linear scale; the
// magnitude of deltas is meaningless; only the sign of a delta is
// meaningful.
//
// > So it would be wrong to say something like
// >
// > desirability = (-a*distance) + (b*generation) // for some constants `a` and `b`
// >
// > because `generation` can't be used that way
//
// - We prefer items that appear in more lists over items that appear in
// fewer lists.
//
// The relative priority of these 4 goals is undefined; preferrably the
// algorithm should be defined in a way that makes it easy to adjust the
// relative priorities.
ret := make(containers.Set[btrfsvol.LogicalAddr])
illegal := make(containers.Set[btrfsvol.LogicalAddr]) // cannot-be-accepted and already-accepted
accept := func(item btrfsvol.LogicalAddr) {
ret.Insert(item)
for _, list := range lists {
if list.Has(item) {
illegal.InsertFrom(list)
}
}
}
counts := make(map[btrfsvol.LogicalAddr]int)
for _, list := range lists {
for item := range list {
counts[item] = counts[item] + 1
}
}
sortedItems := maps.Keys(distances)
sort.Slice(sortedItems, func(i, j int) bool {
iItem, jItem := sortedItems[i], sortedItems[j]
if counts[iItem] != counts[jItem] {
return counts[iItem] > counts[jItem] // reverse this check; higher counts should sort lower
}
if distances[iItem] != distances[jItem] {
return distances[iItem] < distances[jItem]
}
if generations[iItem] != generations[jItem] {
return generations[iItem] > generations[jItem] // reverse this check; higher generations should sort lower
}
return iItem < jItem // laddr is as good a tiebreaker as anything
})
for _, item := range sortedItems {
if !illegal.Has(item) {
accept(item)
}
}
for i, list := range lists {
chose := list.Intersection(ret)
if len(chose) == 0 {
dlog.Infof(ctx, "... ... ... lists[%d]: chose (none) from %v", i, maps.SortedKeys(list))
} else {
dlog.Infof(ctx, "... ... ... lists[%d]: chose %v from %v", i, chose.TakeOne(), maps.SortedKeys(list))
}
}
return ret
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
func (o *rebuilder) wantAugment(ctx context.Context, treeID btrfsprim.ObjID, choices containers.Set[btrfsvol.LogicalAddr]) {
if len(choices) == 0 {
dlog.Errorf(ctx, "augment(tree=%v): could not find wanted item", treeID)
return
}
choicesWithDist := make(map[btrfsvol.LogicalAddr]int, len(choices))
for choice := range choices {
dist, ok := o.rebuilt.COWDistance(ctx, treeID, o.graph.Nodes[choice].Owner)
if !ok {
panic(fmt.Errorf("should not happen: .wantAugment called for tree=%v with invalid choice=%v", treeID, choice))
}
choicesWithDist[choice] = dist
}
dlog.Infof(ctx, "augment(tree=%v): %v", treeID, maps.SortedKeys(choicesWithDist))
o.pendingAugments[treeID] = append(o.pendingAugments[treeID], choicesWithDist)
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// fsErr implements rebuildCallbacks.
func (o *rebuilder) fsErr(ctx context.Context, e error) {
dlog.Errorf(ctx, "filesystem error: %v", e)
}
// want implements rebuildCallbacks.
func (o *rebuilder) want(ctx context.Context, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType) {
o._want(ctx, treeID, objID, typ)
}
func (o *rebuilder) _want(ctx context.Context, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType) (key btrfsprim.Key, ok bool) {
if !o.rebuilt.AddTree(ctx, treeID) {
o.queue = append(o.queue, o.curKey)
return btrfsprim.Key{}, false
}
// check if we already have it
tgt := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
}
if key, ok := o.rebuilt.Search(ctx, treeID, func(key btrfsprim.Key) int {
key.Offset = 0
return tgt.Cmp(key)
}); ok {
return key, true
}
// OK, we need to insert it
ctx = dlog.WithField(ctx, "want_key", fmt.Sprintf("tree=%v key={%v %v ?}", treeID, objID, typ))
wants := make(containers.Set[btrfsvol.LogicalAddr])
o.rebuilt.Keys(treeID).Subrange(
func(k btrfsprim.Key, _ keyio.ItemPtr) int { k.Offset = 0; return tgt.Cmp(k) },
func(_ btrfsprim.Key, v keyio.ItemPtr) bool {
wants.InsertFrom(o.rebuilt.LeafToRoots(ctx, treeID, v.Node))
return true
})
o.wantAugment(ctx, treeID, wants)
return btrfsprim.Key{}, false
}
// wantOff implements rebuildCallbacks.
func (o *rebuilder) wantOff(ctx context.Context, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType, off uint64) {
o._wantOff(ctx, treeID, objID, typ, off)
}
func (o *rebuilder) _wantOff(ctx context.Context, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType, off uint64) (ok bool) {
if !o.rebuilt.AddTree(ctx, treeID) {
o.queue = append(o.queue, o.curKey)
return false
}
// check if we already have it
tgt := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
Offset: off,
}
if _, ok := o.rebuilt.Load(ctx, treeID, tgt); ok {
return true
}
// OK, we need to insert it
ctx = dlog.WithField(ctx, "want_key", fmt.Sprintf("tree=%v key=%v", treeID, tgt))
wants := make(containers.Set[btrfsvol.LogicalAddr])
o.rebuilt.Keys(treeID).Subrange(
func(k btrfsprim.Key, _ keyio.ItemPtr) int { return tgt.Cmp(k) },
func(_ btrfsprim.Key, v keyio.ItemPtr) bool {
wants.InsertFrom(o.rebuilt.LeafToRoots(ctx, treeID, v.Node))
return true
})
o.wantAugment(ctx, treeID, wants)
return false
}
// wantFunc implements rebuildCallbacks.
func (o *rebuilder) wantFunc(ctx context.Context, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType, fn func(btrfsitem.Item) bool) {
if !o.rebuilt.AddTree(ctx, treeID) {
o.queue = append(o.queue, o.curKey)
return
}
// check if we already have it
tgt := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
}
keys := o.rebuilt.SearchAll(ctx, treeID, func(key btrfsprim.Key) int {
key.Offset = 0
return tgt.Cmp(key)
})
for _, itemKey := range keys {
itemBody, ok := o.rebuilt.Load(ctx, treeID, itemKey)
if !ok {
o.ioErr(ctx, fmt.Errorf("could not read previously read item: %v", itemKey))
}
if fn(itemBody) {
return
}
}
// OK, we need to insert it
ctx = dlog.WithField(ctx, "want_key", fmt.Sprintf("tree=%v key=%v +func", treeID, tgt))
wants := make(containers.Set[btrfsvol.LogicalAddr])
o.rebuilt.Keys(treeID).Subrange(
func(k btrfsprim.Key, _ keyio.ItemPtr) int { k.Offset = 0; return tgt.Cmp(k) },
func(k btrfsprim.Key, v keyio.ItemPtr) bool {
itemBody, ok := o.keyIO.ReadItem(v)
if !ok {
o.ioErr(ctx, fmt.Errorf("could not read previously read item: %v at %v", k, v))
}
if fn(itemBody) {
wants.InsertFrom(o.rebuilt.LeafToRoots(ctx, treeID, v.Node))
}
return true
})
o.wantAugment(ctx, treeID, wants)
}
func (o *rebuilder) _wantRange(
ctx context.Context,
treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType,
sizeFn func(btrfsprim.Key) (uint64, error),
beg, end uint64,
) {
if !o.rebuilt.AddTree(ctx, treeID) {
o.queue = append(o.queue, o.curKey)
return
}
// Step 1: Build a listing of the runs that we do have.
runMin := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
Offset: 0, // *NOT* `beg`
}
runMax := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
Offset: end - 1,
}
runKeys := o.rebuilt.SearchAll(ctx, treeID, func(key btrfsprim.Key) int {
switch {
case runMin.Cmp(key) < 0:
return 1
case runMax.Cmp(key) > 0:
return -1
default:
return 0
}
})
// Step 2: Build a listing of the gaps.
//
// Start with a gap of the whole range, then subtract each run
// from it.
type gap struct {
// range is [Beg,End)
Beg, End uint64
}
gaps := &containers.RBTree[containers.NativeOrdered[uint64], gap]{
KeyFn: func(gap gap) containers.NativeOrdered[uint64] {
return containers.NativeOrdered[uint64]{Val: gap.Beg}
},
}
gaps.Insert(gap{
Beg: beg,
End: end,
})
for _, runKey := range runKeys {
runSize, err := sizeFn(runKey)
if err != nil {
o.fsErr(ctx, fmt.Errorf("tree=%v key=%v: %w", treeID, runKey, err))
}
if runSize == 0 {
continue
}
runBeg := runKey.Offset
runEnd := runBeg + runSize
if runEnd <= beg {
continue
}
overlappingGaps := gaps.SearchRange(func(gap gap) int {
switch {
case gap.End <= runBeg:
return 1
case runEnd <= gap.Beg:
return -1
default:
return 0
}
})
if len(overlappingGaps) == 0 {
continue
}
gapsBeg := overlappingGaps[0].Beg
gapsEnd := overlappingGaps[len(overlappingGaps)-1].End
for _, gap := range overlappingGaps {
gaps.Delete(containers.NativeOrdered[uint64]{Val: gap.Beg})
}
if gapsBeg < runBeg {
gaps.Insert(gap{
Beg: gapsBeg,
End: runBeg,
})
}
if gapsEnd > runEnd {
gaps.Insert(gap{
Beg: runEnd,
End: gapsEnd,
})
}
}
// Step 2: Fill each gap.
_ = gaps.Walk(func(rbNode *containers.RBNode[gap]) error {
gap := rbNode.Value
last := gap.Beg
runMin := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
Offset: 0, // *NOT* `gap.Beg`
}
runMax := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
Offset: gap.End - 1,
}
o.rebuilt.Keys(treeID).Subrange(
func(key btrfsprim.Key, _ keyio.ItemPtr) int {
switch {
case runMin.Cmp(key) < 0:
return 1
case runMax.Cmp(key) > 0:
return -1
default:
return 0
}
},
func(k btrfsprim.Key, v keyio.ItemPtr) bool {
runSize, err := sizeFn(k)
if err != nil {
o.fsErr(ctx, fmt.Errorf("tree=%v key=%v: %w", treeID, k, err))
return true
}
if runSize == 0 {
return true
}
runBeg := k.Offset
runEnd := runBeg + runSize
if runEnd <= gap.Beg {
return true
}
// TODO: This is dumb and greedy.
if last < runBeg {
// log an error
o.wantAugment(dlog.WithField(ctx, "want_key", fmt.Sprintf("tree=%v key={%v, %v, %v-%v}",
treeID, objID, typ, last, runBeg)),
treeID, nil)
}
o.wantAugment(dlog.WithField(ctx, "want_key", fmt.Sprintf("tree=%v key={%v, %v, %v-%v}",
treeID, objID, typ, gap.Beg, gap.End)),
treeID, o.rebuilt.LeafToRoots(ctx, treeID, v.Node))
last = runEnd
return true
})
if last < gap.End {
// log an error
o.wantAugment(dlog.WithField(ctx, "want_key", fmt.Sprintf("tree=%v key={%v, %v, %v-%v}",
treeID, objID, typ, last, gap.End)),
treeID, nil)
}
return nil
})
}
// func implements rebuildCallbacks.
//
// interval is [beg, end)
func (o *rebuilder) wantCSum(ctx context.Context, beg, end btrfsvol.LogicalAddr) {
const treeID = btrfsprim.CSUM_TREE_OBJECTID
o._wantRange(ctx, treeID, btrfsprim.EXTENT_CSUM_OBJECTID, btrfsprim.EXTENT_CSUM_KEY,
func(key btrfsprim.Key) (uint64, error) {
ptr, ok := o.rebuilt.Keys(treeID).Load(key)
if !ok {
panic(fmt.Errorf("should not happen: could not load key: %v", key))
}
body, ok := o.keyIO.ReadItem(ptr)
if !ok {
panic(fmt.Errorf("should not happen: could not read item: %v", key))
}
switch body := body.(type) {
case btrfsitem.ExtentCSum:
return uint64(body.Size()), nil
case btrfsitem.Error:
return 0, fmt.Errorf("error decoding item: tree=%v key=%v: %w", treeID, key, body.Err)
default:
// This is a panic because the item decoder should not emit EXTENT_CSUM
// items as anything but btrfsitem.ExtentCSum or btrfsitem.Error without
// this code also being updated.
panic(fmt.Errorf("should not happen: EXTENT_CSUM item has unexpected type: %T", body))
}
},
uint64(beg), uint64(end))
}
// wantFileExt implements rebuildCallbacks.
func (o *rebuilder) wantFileExt(ctx context.Context, treeID btrfsprim.ObjID, ino btrfsprim.ObjID, size int64) {
o._wantRange(ctx, treeID, ino, btrfsprim.EXTENT_DATA_KEY,
func(key btrfsprim.Key) (uint64, error) {
ptr, ok := o.rebuilt.Keys(treeID).Load(key)
if !ok {
panic(fmt.Errorf("should not happen: could not load key: %v", key))
}
body, ok := o.keyIO.ReadItem(ptr)
if !ok {
panic(fmt.Errorf("should not happen: could not read item: %v", key))
}
switch body := body.(type) {
case btrfsitem.FileExtent:
size, err := body.Size()
return uint64(size), err
case btrfsitem.Error:
return 0, fmt.Errorf("error decoding item: tree=%v key=%v: %w", treeID, key, body.Err)
default:
// This is a panic because the item decoder should not emit EXTENT_DATA
// items as anything but btrfsitem.FileExtent or btrfsitem.Error without
// this code also being updated.
panic(fmt.Errorf("should not happen: EXTENT_DATA item has unexpected type: %T", body))
}
},
0, uint64(size))
}
|