summaryrefslogtreecommitdiff
path: root/pkg/btrfs/io4_fs.go
blob: 18ae42acee6dc5d3eba72f297dd65e57a242e104 (plain)
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
package btrfs

import (
	"fmt"
	"io"
	"path/filepath"
	"reflect"
	"sort"
	"sync"

	"github.com/datawire/dlib/derror"

	"lukeshu.com/btrfs-tools/pkg/btrfs/btrfsitem"
	"lukeshu.com/btrfs-tools/pkg/btrfs/btrfsvol"
	"lukeshu.com/btrfs-tools/pkg/util"
)

type BareInode struct {
	Inode     ObjID
	InodeItem *btrfsitem.Inode
	Errs      derror.MultiError
}

type FullInode struct {
	BareInode
	OtherItems []Item
}

type InodeRef struct {
	Inode ObjID
	btrfsitem.InodeRef
}

type Dir struct {
	FullInode
	DotDot          *InodeRef
	ChildrenByName  map[string]btrfsitem.DirEntry
	ChildrenByIndex map[uint64]btrfsitem.DirEntry
	SV              *Subvolume
}

type FileExtent struct {
	OffsetWithinFile int64
	btrfsitem.FileExtent
}

type File struct {
	FullInode
	Extents []FileExtent
	SV      *Subvolume
}

type Subvolume struct {
	FS     *FS
	TreeID ObjID

	rootOnce sync.Once
	rootVal  btrfsitem.Root
	rootErr  error

	bareInodeCache util.LRUCache[ObjID, *BareInode]
	fullInodeCache util.LRUCache[ObjID, *FullInode]
	dirCache       util.LRUCache[ObjID, *Dir]
	fileCache      util.LRUCache[ObjID, *File]
}

func (sv *Subvolume) init() {
	sv.rootOnce.Do(func() {
		root, err := sv.FS.TreeLookup(ROOT_TREE_OBJECTID, Key{
			ObjectID: sv.TreeID,
			ItemType: btrfsitem.ROOT_ITEM_KEY,
			Offset:   0,
		})
		if err != nil {
			sv.rootErr = err
			return
		}

		rootBody, ok := root.Body.(btrfsitem.Root)
		if !ok {
			sv.rootErr = fmt.Errorf("FS_TREE ROOT_ITEM has malformed body")
			return
		}

		sv.rootVal = rootBody
	})
}

func (sv *Subvolume) GetRootInode() (ObjID, error) {
	sv.init()
	return sv.rootVal.RootDirID, sv.rootErr
}

func (sv *Subvolume) LoadBareInode(inode ObjID) (*BareInode, error) {
	val := sv.bareInodeCache.GetOrElse(inode, func() (val *BareInode) {
		val = &BareInode{
			Inode: inode,
		}
		item, err := sv.FS.TreeLookup(sv.TreeID, Key{
			ObjectID: inode,
			ItemType: btrfsitem.INODE_ITEM_KEY,
			Offset:   0,
		})
		if err != nil {
			val.Errs = append(val.Errs, err)
			return
		}

		itemBody, ok := item.Body.(btrfsitem.Inode)
		if !ok {
			val.Errs = append(val.Errs, fmt.Errorf("malformed inode"))
			return
		}
		val.InodeItem = &itemBody

		return
	})
	if val.InodeItem == nil {
		return nil, val.Errs
	}
	return val, nil
}

func (sv *Subvolume) LoadFullInode(inode ObjID) (*FullInode, error) {
	val := sv.fullInodeCache.GetOrElse(inode, func() (val *FullInode) {
		val = &FullInode{
			BareInode: BareInode{
				Inode: inode,
			},
		}
		items, err := sv.FS.TreeSearchAll(sv.TreeID, func(key Key) int {
			return util.CmpUint(inode, key.ObjectID)
		})
		if err != nil {
			val.Errs = append(val.Errs, err)
			if len(items) == 0 {
				return
			}
		}
		for _, item := range items {
			switch item.Head.Key.ItemType {
			case btrfsitem.INODE_ITEM_KEY:
				itemBody := item.Body.(btrfsitem.Inode)
				if val.InodeItem != nil {
					if !reflect.DeepEqual(itemBody, *val.InodeItem) {
						val.Errs = append(val.Errs, fmt.Errorf("multiple inodes"))
					}
					continue
				}
				val.InodeItem = &itemBody
			default:
				val.OtherItems = append(val.OtherItems, item)
			}
		}
		return
	})
	if val.InodeItem == nil && val.OtherItems == nil {
		return nil, val.Errs
	}
	return val, nil
}

func (sv *Subvolume) LoadDir(inode ObjID) (*Dir, error) {
	val := sv.dirCache.GetOrElse(inode, func() (val *Dir) {
		val = new(Dir)
		fullInode, err := sv.LoadFullInode(inode)
		if err != nil {
			val.Errs = append(val.Errs, err)
			return
		}
		val.FullInode = *fullInode
		val.SV = sv
		val.populate()
		return
	})
	if val.Inode == 0 {
		return nil, val.Errs
	}
	return val, nil
}

func (ret *Dir) populate() {
	ret.ChildrenByName = make(map[string]btrfsitem.DirEntry)
	ret.ChildrenByIndex = make(map[uint64]btrfsitem.DirEntry)
	for _, item := range ret.OtherItems {
		switch item.Head.Key.ItemType {
		case btrfsitem.INODE_REF_KEY:
			ref := InodeRef{
				Inode:    ObjID(item.Head.Key.Offset),
				InodeRef: item.Body.(btrfsitem.InodeRef),
			}
			if ret.DotDot != nil {
				if !reflect.DeepEqual(ref, *ret.DotDot) {
					ret.Errs = append(ret.Errs, fmt.Errorf("multiple INODE_REF items on a directory"))
				}
				continue
			}
			ret.DotDot = &ref
		case btrfsitem.DIR_ITEM_KEY:
			body := item.Body.(btrfsitem.DirEntries)
			if len(body) != 1 {
				ret.Errs = append(ret.Errs, fmt.Errorf("multiple direntries in single DIR_ITEM?"))
				continue
			}
			for _, entry := range body {
				namehash := btrfsitem.NameHash(entry.Name)
				if namehash != item.Head.Key.Offset {
					ret.Errs = append(ret.Errs, fmt.Errorf("direntry crc32c mismatch: key=%#x crc32c(%q)=%#x",
						item.Head.Key.Offset, entry.Name, namehash))
					continue
				}
				if other, exists := ret.ChildrenByName[string(entry.Name)]; exists {
					if !reflect.DeepEqual(entry, other) {
						ret.Errs = append(ret.Errs, fmt.Errorf("multiple instances of direntry name %q", entry.Name))
					}
					continue
				}
				ret.ChildrenByName[string(entry.Name)] = entry
			}
		case btrfsitem.DIR_INDEX_KEY:
			index := item.Head.Key.Offset
			body := item.Body.(btrfsitem.DirEntries)
			if len(body) != 1 {
				ret.Errs = append(ret.Errs, fmt.Errorf("multiple direntries in single DIR_INDEX?"))
				continue
			}
			for _, entry := range body {
				if other, exists := ret.ChildrenByIndex[index]; exists {
					if !reflect.DeepEqual(entry, other) {
						ret.Errs = append(ret.Errs, fmt.Errorf("multiple instances of direntry index %v", index))
					}
					continue
				}
				ret.ChildrenByIndex[index] = entry
			}
		//case btrfsitem.XATTR_ITEM_KEY:
		default:
			panic(fmt.Errorf("TODO: handle item type %v", item.Head.Key.ItemType))
		}
	}
	entriesWithIndexes := make(map[string]struct{})
	nextIndex := uint64(2)
	for index, entry := range ret.ChildrenByIndex {
		if index+1 > nextIndex {
			nextIndex = index + 1
		}
		entriesWithIndexes[string(entry.Name)] = struct{}{}
		if other, exists := ret.ChildrenByName[string(entry.Name)]; !exists {
			ret.Errs = append(ret.Errs, fmt.Errorf("missing by-name direntry for %q", entry.Name))
			ret.ChildrenByName[string(entry.Name)] = entry
		} else if !reflect.DeepEqual(entry, other) {
			ret.Errs = append(ret.Errs, fmt.Errorf("direntry index %v and direntry name %q disagree", index, entry.Name))
			ret.ChildrenByName[string(entry.Name)] = entry
		}
	}
	for _, name := range util.SortedMapKeys(ret.ChildrenByName) {
		if _, exists := entriesWithIndexes[name]; !exists {
			ret.Errs = append(ret.Errs, fmt.Errorf("missing by-index direntry for %q", name))
			ret.ChildrenByIndex[nextIndex] = ret.ChildrenByName[name]
			nextIndex++
		}
	}
}

func (dir *Dir) AbsPath() (string, error) {
	rootInode, err := dir.SV.GetRootInode()
	if err != nil {
		return "", err
	}
	if rootInode == dir.Inode {
		return "/", nil
	}
	if dir.DotDot == nil {
		return "", fmt.Errorf("missing .. entry in dir inode %v", dir.Inode)
	}
	parent, err := dir.SV.LoadDir(dir.DotDot.Inode)
	if err != nil {
		return "", err
	}
	parentName, err := parent.AbsPath()
	if err != nil {
		return "", err
	}
	return filepath.Join(parentName, string(dir.DotDot.Name)), nil
}

func (sv *Subvolume) LoadFile(inode ObjID) (*File, error) {
	val := sv.fileCache.GetOrElse(inode, func() (val *File) {
		val = new(File)
		fullInode, err := sv.LoadFullInode(inode)
		if err != nil {
			val.Errs = append(val.Errs, err)
			return
		}
		val.FullInode = *fullInode
		val.SV = sv
		val.populate()
		return
	})
	if val.Inode == 0 {
		return nil, val.Errs
	}
	return val, nil
}

func (ret *File) populate() {
	for _, item := range ret.OtherItems {
		switch item.Head.Key.ItemType {
		case btrfsitem.INODE_REF_KEY:
			// TODO
		case btrfsitem.EXTENT_DATA_KEY:
			ret.Extents = append(ret.Extents, FileExtent{
				OffsetWithinFile: int64(item.Head.Key.Offset),
				FileExtent:       item.Body.(btrfsitem.FileExtent),
			})
		default:
			panic(fmt.Errorf("TODO: handle item type %v", item.Head.Key.ItemType))
		}
	}

	// These should already be sorted, because of the nature of
	// the btree; but this is a recovery tool for corrupt
	// filesystems, so go ahead and ensure that it's sorted.
	sort.Slice(ret.Extents, func(i, j int) bool {
		return ret.Extents[i].OffsetWithinFile < ret.Extents[j].OffsetWithinFile
	})

	pos := int64(0)
	for _, extent := range ret.Extents {
		if extent.OffsetWithinFile != pos {
			if extent.OffsetWithinFile > pos {
				ret.Errs = append(ret.Errs, fmt.Errorf("extent gap from %v to %v",
					pos, extent.OffsetWithinFile))
			} else {
				ret.Errs = append(ret.Errs, fmt.Errorf("extent overlap from %v to %v",
					extent.OffsetWithinFile, pos))
			}
		}
		size, err := extent.Size()
		if err != nil {
			ret.Errs = append(ret.Errs, fmt.Errorf("extent %v: %w", extent.OffsetWithinFile, err))
		}
		pos += size
	}
	if ret.InodeItem != nil && pos != ret.InodeItem.Size {
		if ret.InodeItem.Size > pos {
			ret.Errs = append(ret.Errs, fmt.Errorf("extent gap from %v to %v",
				pos, ret.InodeItem.Size))
		} else {
			ret.Errs = append(ret.Errs, fmt.Errorf("extent mapped past end of file from %v to %v",
				ret.InodeItem.Size, pos))
		}
	}
}

func (file *File) ReadAt(dat []byte, off int64) (int, error) {
	// These stateles maybe-short-reads each do an O(n) extent
	// lookup, so reading a file is O(n^2), but we expect n to be
	// small, so whatev.  Turn file.Extents it in to an rbtree if
	// it becomes a problem.
	done := 0
	for done < len(dat) {
		n, err := file.maybeShortReadAt(dat[done:], off+int64(done))
		done += n
		if err != nil {
			return done, err
		}
	}
	return done, nil
}

func (file *File) maybeShortReadAt(dat []byte, off int64) (int, error) {
	for _, extent := range file.Extents {
		extBeg := extent.OffsetWithinFile
		if extBeg > off {
			break
		}
		extLen, err := extent.Size()
		if err != nil {
			continue
		}
		extEnd := extBeg + extLen
		if extEnd <= off {
			continue
		}
		offsetWithinExt := off - extent.OffsetWithinFile
		readSize := util.Min(int64(len(dat)), extLen-offsetWithinExt)
		switch extent.Type {
		case btrfsitem.FILE_EXTENT_INLINE:
			return copy(dat, extent.BodyInline[offsetWithinExt:offsetWithinExt+readSize]), nil
		case btrfsitem.FILE_EXTENT_REG, btrfsitem.FILE_EXTENT_PREALLOC:
			return file.SV.FS.ReadAt(dat[:readSize],
				extent.BodyExtent.DiskByteNr.
					Add(extent.BodyExtent.Offset).
					Add(btrfsvol.AddrDelta(offsetWithinExt)))
		}
	}
	if file.InodeItem != nil && off >= file.InodeItem.Size {
		return 0, io.EOF
	}
	return 0, fmt.Errorf("read: could not map position %v", off)
}

var _ io.ReaderAt = (*File)(nil)