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
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
|
// Copyright (C) 2022-2023 Luke Shumaker <lukeshu@lukeshu.com>
//
// SPDX-License-Identifier: GPL-2.0-or-later
package rebuildnodes
import (
"bytes"
"context"
"fmt"
"runtime"
"sort"
"time"
"github.com/datawire/dlib/dgroup"
"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/btrfssum"
"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 := containers.NativeCmp(a.TreeID, b.TreeID); d != 0 {
return d
}
return a.Key.Cmp(b.Key)
}
func (o keyAndTree) String() string {
return fmt.Sprintf("tree=%v key=%v", o.TreeID, o.Key)
}
type rebuilder struct {
sb btrfstree.Superblock
graph graph.Graph
keyIO *keyio.Handle
rebuilt *btrees.RebuiltForrest
curKey keyAndTree
treeQueue containers.Set[btrfsprim.ObjID]
itemQueue containers.Set[keyAndTree]
augmentQueue map[btrfsprim.ObjID]*treeAugmentQueue
numAugments int
}
type treeAugmentQueue struct {
single map[string]btrfsvol.LogicalAddr
multi map[string]containers.Set[btrfsvol.LogicalAddr]
}
type Rebuilder interface {
Rebuild(context.Context) error
ListRoots() map[btrfsprim.ObjID]containers.Set[btrfsvol.LogicalAddr]
}
func NewRebuilder(ctx context.Context, fs *btrfs.FS, nodeScanResults btrfsinspect.ScanDevicesResult) (Rebuilder, error) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.step", "read-fs-data")
sb, nodeGraph, keyIO, err := ScanDevices(ctx, fs, nodeScanResults) // ScanDevices does its own logging
if err != nil {
return nil, err
}
o := &rebuilder{
sb: sb,
graph: nodeGraph,
keyIO: keyIO,
}
o.rebuilt = btrees.NewRebuiltForrest(sb, nodeGraph, keyIO,
o.cbAddedItem, o.cbLookupRoot, o.cbLookupUUID)
return o, 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)
}
func (o *rebuilder) ListRoots() map[btrfsprim.ObjID]containers.Set[btrfsvol.LogicalAddr] {
return o.rebuilt.ListRoots()
}
type itemStats struct {
textui.Portion[int]
NumAugments int
NumAugmentTrees int
}
func (s itemStats) String() string {
return textui.Sprintf("%v (queued %v augments across %v trees)",
s.Portion, s.NumAugments, s.NumAugmentTrees)
}
func (o *rebuilder) Rebuild(_ctx context.Context) error {
_ctx = dlog.WithField(_ctx, "btrfsinspect.rebuild-nodes.step", "rebuild")
// Initialize
o.itemQueue = make(containers.Set[keyAndTree])
o.augmentQueue = make(map[btrfsprim.ObjID]*treeAugmentQueue)
// Seed the queue
o.treeQueue = containers.NewSet[btrfsprim.ObjID](
btrfsprim.ROOT_TREE_OBJECTID,
btrfsprim.CHUNK_TREE_OBJECTID,
// btrfsprim.TREE_LOG_OBJECTID, // TODO(lukeshu): Special LOG_TREE handling
btrfsprim.BLOCK_GROUP_TREE_OBJECTID,
)
for passNum := 0; len(o.treeQueue) > 0 || len(o.itemQueue) > 0 || len(o.augmentQueue) > 0; passNum++ {
passCtx := dlog.WithField(_ctx, "btrfsinspect.rebuild-nodes.rebuild.pass", passNum)
// Add items to the queue (drain o.treeQueue, fill o.itemQueue)
if true {
stepCtx := dlog.WithField(passCtx, "btrfsinspect.rebuild-nodes.rebuild.substep", "collect-items")
treeQueue := o.treeQueue
o.treeQueue = make(containers.Set[btrfsprim.ObjID])
// Because trees can be wildly different sizes, it's impossible to have a meaningful
// progress percentage here.
for _, treeID := range maps.SortedKeys(treeQueue) {
if err := _ctx.Err(); err != nil {
return err
}
o.curKey = keyAndTree{TreeID: treeID}
o.rebuilt.Tree(stepCtx, treeID)
}
}
runtime.GC()
// Handle items in the queue (drain o.itemQueue, fill o.augmentQueue and o.treeQueue)
if true {
stepCtx := dlog.WithField(passCtx, "btrfsinspect.rebuild-nodes.rebuild.substep", "process-items")
itemQueue := maps.Keys(o.itemQueue)
o.itemQueue = make(containers.Set[keyAndTree])
sort.Slice(itemQueue, func(i, j int) bool {
return itemQueue[i].Cmp(itemQueue[j]) < 0
})
var progress itemStats
progress.D = len(itemQueue)
progressWriter := textui.NewProgress[itemStats](stepCtx, dlog.LogLevelInfo, textui.Tunable(1*time.Second))
stepCtx = dlog.WithField(stepCtx, "btrfsinspect.rebuild-nodes.rebuild.substep.progress", &progress)
type keyAndBody struct {
keyAndTree
Body btrfsitem.Item
}
itemChan := make(chan keyAndBody, textui.Tunable(300)) // average items-per-node≈100; let's have a buffer of ~3 nodes
grp := dgroup.NewGroup(stepCtx, dgroup.GroupConfig{})
grp.Go("io", func(stepCtx context.Context) error {
defer close(itemChan)
for _, key := range itemQueue {
if err := stepCtx.Err(); err != nil {
return err
}
itemCtx := dlog.WithField(stepCtx, "btrfsinspect.rebuild-nodes.rebuild.process.item", key)
itemBody, ok := o.rebuilt.Tree(itemCtx, key.TreeID).ReadItem(itemCtx, key.Key)
if !ok {
o.ioErr(itemCtx, fmt.Errorf("could not read previously read item: %v", key))
}
itemChan <- keyAndBody{
keyAndTree: key,
Body: itemBody,
}
}
return nil
})
grp.Go("cpu", func(stepCtx context.Context) error {
defer progressWriter.Done()
for item := range itemChan {
itemCtx := dlog.WithField(stepCtx, "btrfsinspect.rebuild-nodes.rebuild.process.item", item.keyAndTree)
o.curKey = item.keyAndTree
handleItem(o, itemCtx, item.TreeID, btrfstree.Item{
Key: item.Key,
Body: item.Body,
})
if item.ItemType == btrfsitem.ROOT_ITEM_KEY {
o.treeQueue.Insert(item.ObjectID)
}
progress.N++
progress.NumAugments = o.numAugments
progress.NumAugmentTrees = len(o.augmentQueue)
progressWriter.Set(progress)
}
return nil
})
if err := grp.Wait(); err != nil {
return err
}
}
runtime.GC()
// Apply augments (drain o.augmentQueue, fill o.itemQueue)
if true {
stepCtx := dlog.WithField(passCtx, "btrfsinspect.rebuild-nodes.rebuild.substep", "apply-augments")
resolvedAugments := make(map[btrfsprim.ObjID]containers.Set[btrfsvol.LogicalAddr], len(o.augmentQueue))
var progress textui.Portion[int]
for _, treeID := range maps.SortedKeys(o.augmentQueue) {
if err := _ctx.Err(); err != nil {
return err
}
treeCtx := dlog.WithField(stepCtx, "btrfsinspect.rebuild-nodes.rebuild.augment.tree", treeID)
resolvedAugments[treeID] = o.resolveTreeAugments(treeCtx, treeID)
progress.D += len(resolvedAugments[treeID])
}
o.augmentQueue = make(map[btrfsprim.ObjID]*treeAugmentQueue)
o.numAugments = 0
runtime.GC()
progressWriter := textui.NewProgress[textui.Portion[int]](stepCtx, dlog.LogLevelInfo, textui.Tunable(1*time.Second))
stepCtx = dlog.WithField(stepCtx, "btrfsinspect.rebuild-nodes.rebuild.substep.progress", &progress)
for _, treeID := range maps.SortedKeys(resolvedAugments) {
treeCtx := dlog.WithField(stepCtx, "btrfsinspect.rebuild-nodes.rebuild.augment.tree", treeID)
for _, nodeAddr := range maps.SortedKeys(resolvedAugments[treeID]) {
if err := _ctx.Err(); err != nil {
progressWriter.Set(progress)
progressWriter.Done()
return err
}
progressWriter.Set(progress)
o.rebuilt.Tree(treeCtx, treeID).AddRoot(treeCtx, nodeAddr)
progress.N++
}
}
progressWriter.Set(progress)
progressWriter.Done()
}
runtime.GC()
}
return nil
}
func (o *rebuilder) enqueueRetry() {
if o.curKey.Key == (btrfsprim.Key{}) {
o.treeQueue.Insert(o.curKey.TreeID)
} else {
o.itemQueue.Insert(o.curKey)
}
}
func (o *rebuilder) cbAddedItem(ctx context.Context, tree btrfsprim.ObjID, key btrfsprim.Key) {
o.itemQueue.Insert(keyAndTree{
TreeID: tree,
Key: key,
})
}
func (o *rebuilder) cbLookupRoot(ctx context.Context, tree btrfsprim.ObjID) (offset btrfsprim.Generation, item btrfsitem.Root, ok bool) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.add-tree.want.reason", "tree Root")
key := keyAndTree{
TreeID: btrfsprim.ROOT_TREE_OBJECTID,
Key: btrfsprim.Key{
ObjectID: tree,
ItemType: btrfsitem.ROOT_ITEM_KEY,
},
}
wantKey := fmt.Sprintf("tree=%v key={%v %v ?}", key.TreeID, key.ObjectID, key.ItemType)
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.add-tree.want.key", wantKey)
key.Key, ok = o._want(ctx, key.TreeID, wantKey, key.ObjectID, key.ItemType)
if !ok {
o.enqueueRetry()
return 0, btrfsitem.Root{}, false
}
itemBody, ok := o.rebuilt.Tree(ctx, key.TreeID).ReadItem(ctx, key.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: %v: %w", 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) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.add-tree.want.reason", "resolve parent UUID")
key := keyAndTree{TreeID: btrfsprim.UUID_TREE_OBJECTID, Key: btrfsitem.UUIDToKey(uuid)}
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.add-tree.want.key", key.String())
if !o._wantOff(ctx, key.TreeID, key.String(), key.Key) {
o.enqueueRetry()
return 0, false
}
itemBody, ok := o.rebuilt.Tree(ctx, key.TreeID).ReadItem(ctx, key.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: %v: %w", 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, treeID btrfsprim.ObjID) containers.Set[btrfsvol.LogicalAddr] {
// 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.
type ChoiceInfo struct {
Count int
Distance int
Generation btrfsprim.Generation
}
choices := make(map[btrfsvol.LogicalAddr]ChoiceInfo)
// o.augmentQueue[treeID].single is optimized storage for
// lists with exactly 1 item.
for _, choice := range o.augmentQueue[treeID].single {
if old, ok := choices[choice]; ok {
old.Count++
choices[choice] = old
} else {
choices[choice] = ChoiceInfo{
Count: 1,
Distance: discardOK(o.rebuilt.Tree(ctx, treeID).COWDistance(o.graph.Nodes[choice].Owner)),
Generation: o.graph.Nodes[choice].Generation,
}
}
}
// o.augmentQueue[treeID].multi is the main list storage.
for _, list := range o.augmentQueue[treeID].multi {
for choice := range list {
if old, ok := choices[choice]; ok {
old.Count++
choices[choice] = old
} else {
choices[choice] = ChoiceInfo{
Count: 1,
Distance: discardOK(o.rebuilt.Tree(ctx, treeID).COWDistance(o.graph.Nodes[choice].Owner)),
Generation: o.graph.Nodes[choice].Generation,
}
}
}
}
// > 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 would 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 permissible
// > 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; preferably 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 o.augmentQueue[treeID].multi {
if list.Has(item) {
illegal.InsertFrom(list)
}
}
}
sortedItems := maps.Keys(choices)
sort.Slice(sortedItems, func(i, j int) bool {
iItem, jItem := sortedItems[i], sortedItems[j]
if choices[iItem].Count != choices[jItem].Count {
return choices[iItem].Count > choices[jItem].Count // reverse this check; higher counts should sort lower
}
if choices[iItem].Distance != choices[jItem].Distance {
return choices[iItem].Distance < choices[jItem].Distance
}
if choices[iItem].Generation != choices[jItem].Generation {
return choices[iItem].Generation > choices[jItem].Generation // 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)
}
}
// Log our result
wantKeys := append(
maps.Keys(o.augmentQueue[treeID].single),
maps.Keys(o.augmentQueue[treeID].multi)...)
sort.Strings(wantKeys)
for _, wantKey := range wantKeys {
list, ok := o.augmentQueue[treeID].multi[wantKey]
if !ok {
list = containers.NewSet[btrfsvol.LogicalAddr](o.augmentQueue[treeID].single[wantKey])
}
chose := list.Intersection(ret)
if len(chose) == 0 {
dlog.Infof(ctx, "lists[%q]: chose (none) from %v", wantKey, maps.SortedKeys(list))
} else {
dlog.Infof(ctx, "lists[%q]: chose %v from %v", wantKey, chose.TakeOne(), maps.SortedKeys(list))
}
}
// Free some memory
o.augmentQueue[treeID].single = nil
o.augmentQueue[treeID].multi = nil
return ret
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
func (o *rebuilder) hasAugment(treeID btrfsprim.ObjID, wantKey string) bool {
if treeQueue, ok := o.augmentQueue[treeID]; ok {
if treeQueue.single != nil {
if _, ok := treeQueue.single[wantKey]; ok {
return true
}
}
if treeQueue.multi != nil {
if _, ok := treeQueue.multi[wantKey]; ok {
return true
}
}
}
return false
}
func (o *rebuilder) wantAugment(ctx context.Context, treeID btrfsprim.ObjID, wantKey string, choices containers.Set[btrfsvol.LogicalAddr]) {
if len(choices) == 0 {
choices = nil
dlog.Error(ctx, "could not find wanted item")
} else {
dlog.Infof(ctx, "choices=%v", maps.SortedKeys(choices))
}
if o.augmentQueue[treeID] == nil {
o.augmentQueue[treeID] = new(treeAugmentQueue)
}
if len(choices) == 1 {
if o.augmentQueue[treeID].single == nil {
o.augmentQueue[treeID].single = make(map[string]btrfsvol.LogicalAddr)
}
o.augmentQueue[treeID].single[wantKey] = choices.TakeOne()
} else {
if o.augmentQueue[treeID].multi == nil {
o.augmentQueue[treeID].multi = make(map[string]containers.Set[btrfsvol.LogicalAddr])
}
o.augmentQueue[treeID].multi[wantKey] = choices
}
o.numAugments++
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// 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, reason string, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.reason", reason)
wantKey := fmt.Sprintf("tree=%v key={%v %v ?}", treeID, objID, typ)
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", wantKey)
o._want(ctx, treeID, wantKey, objID, typ)
}
func (o *rebuilder) _want(ctx context.Context, treeID btrfsprim.ObjID, wantKey string, objID btrfsprim.ObjID, typ btrfsprim.ItemType) (key btrfsprim.Key, ok bool) {
if o.rebuilt.Tree(ctx, treeID) == nil {
o.enqueueRetry()
return btrfsprim.Key{}, false
}
// check if we already have it
tgt := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
}
if key, _, ok := o.rebuilt.Tree(ctx, treeID).Items(ctx).Search(func(key btrfsprim.Key, _ keyio.ItemPtr) int {
key.Offset = 0
return tgt.Cmp(key)
}); ok {
return key, true
}
// OK, we need to insert it
if o.hasAugment(treeID, wantKey) {
return btrfsprim.Key{}, false
}
wants := make(containers.Set[btrfsvol.LogicalAddr])
o.rebuilt.Tree(ctx, treeID).PotentialItems(ctx).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.Tree(ctx, treeID).LeafToRoots(ctx, v.Node))
return true
})
o.wantAugment(ctx, treeID, wantKey, wants)
return btrfsprim.Key{}, false
}
// wantOff implements rebuildCallbacks.
func (o *rebuilder) wantOff(ctx context.Context, reason string, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType, off uint64) {
key := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
Offset: off,
}
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.reason", reason)
wantKey := keyAndTree{TreeID: treeID, Key: key}.String()
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", wantKey)
o._wantOff(ctx, treeID, wantKey, key)
}
func (o *rebuilder) _wantOff(ctx context.Context, treeID btrfsprim.ObjID, wantKey string, tgt btrfsprim.Key) (ok bool) {
if o.rebuilt.Tree(ctx, treeID) == nil {
o.enqueueRetry()
return false
}
// check if we already have it
if _, ok := o.rebuilt.Tree(ctx, treeID).Items(ctx).Load(tgt); ok {
return true
}
// OK, we need to insert it
if o.hasAugment(treeID, wantKey) {
return false
}
wants := make(containers.Set[btrfsvol.LogicalAddr])
o.rebuilt.Tree(ctx, treeID).PotentialItems(ctx).Subrange(
func(k btrfsprim.Key, _ keyio.ItemPtr) int { return tgt.Cmp(k) },
func(_ btrfsprim.Key, v keyio.ItemPtr) bool {
wants.InsertFrom(o.rebuilt.Tree(ctx, treeID).LeafToRoots(ctx, v.Node))
return true
})
o.wantAugment(ctx, treeID, wantKey, wants)
return false
}
func (o *rebuilder) _wantFunc(ctx context.Context, treeID btrfsprim.ObjID, wantKey string, objID btrfsprim.ObjID, typ btrfsprim.ItemType, fn func(keyio.ItemPtr) bool) {
if o.rebuilt.Tree(ctx, treeID) == nil {
o.enqueueRetry()
return
}
// check if we already have it
tgt := btrfsprim.Key{
ObjectID: objID,
ItemType: typ,
}
found := false
o.rebuilt.Tree(ctx, treeID).Items(ctx).Subrange(
func(key btrfsprim.Key, _ keyio.ItemPtr) int {
key.Offset = 0
return tgt.Cmp(key)
},
func(_ btrfsprim.Key, itemPtr keyio.ItemPtr) bool {
if fn(itemPtr) {
found = true
}
return !found
})
if found {
return
}
// OK, we need to insert it
if o.hasAugment(treeID, wantKey) {
return
}
wants := make(containers.Set[btrfsvol.LogicalAddr])
o.rebuilt.Tree(ctx, treeID).PotentialItems(ctx).Subrange(
func(k btrfsprim.Key, _ keyio.ItemPtr) int { k.Offset = 0; return tgt.Cmp(k) },
func(k btrfsprim.Key, v keyio.ItemPtr) bool {
if fn(v) {
wants.InsertFrom(o.rebuilt.Tree(ctx, treeID).LeafToRoots(ctx, v.Node))
}
return true
})
o.wantAugment(ctx, treeID, wantKey, wants)
}
// wantDirIndex implements rebuildCallbacks.
func (o *rebuilder) wantDirIndex(ctx context.Context, reason string, treeID btrfsprim.ObjID, objID btrfsprim.ObjID, name []byte) {
typ := btrfsitem.DIR_INDEX_KEY
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.reason", reason)
wantKey := fmt.Sprintf("tree=%v key={%v %v ?} name=%q", treeID, objID, typ, name)
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", wantKey)
o._wantFunc(ctx, treeID, wantKey, objID, typ, func(ptr keyio.ItemPtr) bool {
itemName, ok := o.keyIO.Names[ptr]
return ok && bytes.Equal(itemName, name)
})
}
func (o *rebuilder) _wantRange(
ctx context.Context,
treeID btrfsprim.ObjID, objID btrfsprim.ObjID, typ btrfsprim.ItemType,
beg, end uint64,
) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key",
fmt.Sprintf("tree=%v key={%v %v ?}", treeID, objID, typ))
if o.rebuilt.Tree(ctx, treeID) == nil {
o.enqueueRetry()
return
}
sizeFn := func(key btrfsprim.Key) (uint64, error) {
ptr, ok := o.rebuilt.Tree(ctx, treeID).PotentialItems(ctx).Load(key)
if !ok {
panic(fmt.Errorf("should not happen: could not load key: %v", keyAndTree{TreeID: treeID, Key: key}))
}
sizeAndErr, ok := o.keyIO.Sizes[ptr]
if !ok {
panic(fmt.Errorf("should not happen: %v item did not have a size recorded", typ))
}
return sizeAndErr.Size, sizeAndErr.Err
}
// 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,
}
// 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,
})
o.rebuilt.Tree(ctx, treeID).Items(ctx).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(runKey btrfsprim.Key, _ keyio.ItemPtr) bool {
runSize, err := sizeFn(runKey)
if err != nil {
o.fsErr(ctx, fmt.Errorf("get size: %v: %w", keyAndTree{TreeID: treeID, Key: runKey}, err))
return true
}
if runSize == 0 {
return true
}
runBeg := runKey.Offset
runEnd := runBeg + runSize
if runEnd <= beg {
return true
}
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 {
return true
}
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,
})
}
return true
})
// 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.Tree(ctx, treeID).PotentialItems(ctx).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("get size: %v: %w", keyAndTree{TreeID: treeID, Key: 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
wantKey := fmt.Sprintf("tree=%v key={%v %v %v-%v}", treeID, objID, typ, last, runBeg)
if !o.hasAugment(treeID, wantKey) {
wantCtx := dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", wantKey)
o.wantAugment(wantCtx, treeID, wantKey, nil)
}
}
wantKey := fmt.Sprintf("tree=%v key={%v %v %v-%v}", treeID, objID, typ, gap.Beg, gap.End)
if !o.hasAugment(treeID, wantKey) {
wantCtx := dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", wantKey)
o.wantAugment(wantCtx, treeID, wantKey, o.rebuilt.Tree(wantCtx, treeID).LeafToRoots(wantCtx, v.Node))
}
last = runEnd
return true
})
if last < gap.End {
// log an error
wantKey := fmt.Sprintf("tree=%v key={%v, %v, %v-%v}", treeID, objID, typ, last, gap.End)
if !o.hasAugment(treeID, wantKey) {
wantCtx := dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", wantKey)
o.wantAugment(wantCtx, treeID, wantKey, nil)
}
}
return nil
})
}
// func implements rebuildCallbacks.
//
// interval is [beg, end)
func (o *rebuilder) wantCSum(ctx context.Context, reason string, inodeTree, inode btrfsprim.ObjID, beg, end btrfsvol.LogicalAddr) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.reason", reason)
inodeKey := keyAndTree{
TreeID: inodeTree,
Key: btrfsprim.Key{
ObjectID: inode,
ItemType: btrfsitem.INODE_ITEM_KEY,
Offset: 0,
},
}
inodeCtx := dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.key", inodeKey.String())
if !o._wantOff(inodeCtx, inodeKey.TreeID, inodeKey.String(), inodeKey.Key) {
o.enqueueRetry()
return
}
inodeBody, ok := o.rebuilt.Tree(inodeCtx, inodeKey.TreeID).ReadItem(inodeCtx, inodeKey.Key)
if !ok {
o.ioErr(inodeCtx, fmt.Errorf("could not read previously read item: %v", inodeKey))
}
switch inodeBody := inodeBody.(type) {
case btrfsitem.Inode:
if inodeBody.Flags.Has(btrfsitem.INODE_NODATASUM) {
return
}
beg = roundDown(beg, btrfssum.BlockSize)
end = roundUp(end, btrfssum.BlockSize)
const treeID = btrfsprim.CSUM_TREE_OBJECTID
o._wantRange(ctx, treeID, btrfsprim.EXTENT_CSUM_OBJECTID, btrfsprim.EXTENT_CSUM_KEY,
uint64(beg), uint64(end))
case btrfsitem.Error:
o.fsErr(inodeCtx, fmt.Errorf("error decoding item: %v: %w", inodeKey, inodeBody.Err))
default:
// This is a panic because the item decoder should not emit INODE_ITEM items as anything but
// btrfsitem.Inode or btrfsitem.Error without this code also being updated.
panic(fmt.Errorf("should not happen: INODE_ITEM item has unexpected type: %T", inodeBody))
}
}
// wantFileExt implements rebuildCallbacks.
func (o *rebuilder) wantFileExt(ctx context.Context, reason string, treeID btrfsprim.ObjID, ino btrfsprim.ObjID, size int64) {
ctx = dlog.WithField(ctx, "btrfsinspect.rebuild-nodes.rebuild.want.reason", reason)
o._wantRange(ctx, treeID, ino, btrfsprim.EXTENT_DATA_KEY,
0, uint64(size))
}
|