/
fs.go
650 lines (583 loc) · 17.4 KB
/
fs.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
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
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package store
import (
"bytes"
"context"
"encoding/base64"
"encoding/json"
"errors"
"fmt"
"io"
"os/exec"
"sync"
"syscall"
"time"
"github.com/containerd/containerd/reference"
"github.com/containerd/log"
"github.com/containerd/stargz-snapshotter/cache"
"github.com/containerd/stargz-snapshotter/fs/layer"
"github.com/containerd/stargz-snapshotter/fs/remote"
fusefs "github.com/hanwen/go-fuse/v2/fs"
"github.com/hanwen/go-fuse/v2/fuse"
digest "github.com/opencontainers/go-digest"
"golang.org/x/sync/singleflight"
)
const (
defaultLinkMode = syscall.S_IFLNK | 0400 // -r--------
defaultDirMode = syscall.S_IFDIR | 0500 // dr-x------
defaultFileMode = 0400 // -r--------
layerFileMode = 0400 // -r--------
blockSize = 4096
poolLink = "pool"
layerLink = "diff"
blobLink = "blob"
layerInfoLink = "info"
layerUseFile = "use"
fusermountBin = "fusermount"
)
func Mount(ctx context.Context, mountpoint string, layerManager *LayerManager, debug bool) error {
timeSec := time.Second
rawFS := fusefs.NewNodeFS(&rootnode{
fs: &fs{
layerManager: layerManager,
nodeMap: new(idMap),
layerMap: new(idMap),
},
}, &fusefs.Options{
AttrTimeout: &timeSec,
EntryTimeout: &timeSec,
NullPermissions: true,
})
mountOpts := &fuse.MountOptions{
AllowOther: true, // allow users other than root&mounter to access fs
FsName: "stargzstore",
Debug: debug,
}
if _, err := exec.LookPath(fusermountBin); err == nil {
mountOpts.Options = []string{"suid"} // option for fusermount; allow setuid inside container
} else {
log.G(ctx).WithError(err).Debugf("%s not installed; trying direct mount", fusermountBin)
mountOpts.DirectMount = true
}
server, err := fuse.NewServer(rawFS, mountpoint, mountOpts)
if err != nil {
return err
}
go server.Serve()
return server.WaitMount()
}
type fs struct {
layerManager *LayerManager
// nodeMap manages inode numbers for nodes other than nodes in layers
// (i.e. nodes other than ones inside `diff` directories).
// - inode number = [ 0 ][ uint32 ID ]
nodeMap *idMap
// layerMap manages upper bits of inode numbers for nodes inside layers.
// - inode number = [ uint32 layer ID ][ uint32 number (unique inside `diff` directory) ]
// inodes numbers of noeds inside each `diff` directory are prefixed by an unique uint32
// so that they don't conflict with nodes outside `diff` directories.
layerMap *idMap
knownNode map[string]map[string]*layerReleasable
knownNodeMu sync.Mutex
}
type layerReleasable struct {
n fusefs.InodeEmbedder
released bool
mu sync.Mutex
}
func (lh *layerReleasable) releasable() bool {
lh.mu.Lock()
released := lh.released
lh.mu.Unlock()
return released && isForgotten(lh.n.EmbeddedInode())
}
func (lh *layerReleasable) release() {
lh.mu.Lock()
lh.released = true
lh.mu.Unlock()
}
func isForgotten(n *fusefs.Inode) bool {
if !n.Forgotten() {
return false
}
for _, cn := range n.Children() {
if !isForgotten(cn) {
return false
}
}
return true
}
type inoReleasable struct {
n fusefs.InodeEmbedder
}
func (r *inoReleasable) releasable() bool {
return r.n.EmbeddedInode().Forgotten()
}
func (fs *fs) newInodeWithID(ctx context.Context, p func(uint32) fusefs.InodeEmbedder) (*fusefs.Inode, syscall.Errno) {
var ino fusefs.InodeEmbedder
if err := fs.nodeMap.add(func(id uint32) (releasable, error) {
ino = p(id)
return &inoReleasable{ino}, nil
}); err != nil || ino == nil {
log.G(ctx).WithError(err).Debug("cannot generate ID")
return nil, syscall.EIO
}
return ino.EmbeddedInode(), 0
}
// rootnode is the mountpoint node of stargz-store.
type rootnode struct {
fusefs.Inode
fs *fs
}
var _ = (fusefs.InodeEmbedder)((*rootnode)(nil))
var _ = (fusefs.NodeLookuper)((*rootnode)(nil))
// Lookup loads manifest and config of specified name (image reference)
// and returns refnode of the specified name
func (n *rootnode) Lookup(ctx context.Context, name string, out *fuse.EntryOut) (*fusefs.Inode, syscall.Errno) {
// lookup on memory nodes
if cn := n.GetChild(name); cn != nil {
switch tn := cn.Operations().(type) {
case *fusefs.MemSymlink:
copyAttr(&out.Attr, &tn.Attr)
case *refnode:
copyAttr(&out.Attr, &tn.attr)
default:
log.G(ctx).Warn("rootnode.Lookup: uknown node type detected")
return nil, syscall.EIO
}
out.Attr.Ino = cn.StableAttr().Ino
return cn, 0
}
switch name {
case poolLink:
sAttr := defaultLinkAttr(&out.Attr)
cn := &fusefs.MemSymlink{Data: []byte(n.fs.layerManager.refPool.root())}
copyAttr(&cn.Attr, &out.Attr)
return n.fs.newInodeWithID(ctx, func(ino uint32) fusefs.InodeEmbedder {
out.Attr.Ino = uint64(ino)
cn.Attr.Ino = uint64(ino)
sAttr.Ino = uint64(ino)
return n.NewInode(ctx, cn, sAttr)
})
}
refBytes, err := base64.StdEncoding.DecodeString(name)
if err != nil {
log.G(ctx).WithError(err).Debugf("failed to decode ref base64 %q", name)
return nil, syscall.EINVAL
}
ref := string(refBytes)
refspec, err := reference.Parse(ref)
if err != nil {
log.G(ctx).WithError(err).Warnf("invalid reference %q for %q", ref, name)
return nil, syscall.EINVAL
}
sAttr := defaultDirAttr(&out.Attr)
cn := &refnode{
fs: n.fs,
ref: refspec,
}
copyAttr(&cn.attr, &out.Attr)
return n.fs.newInodeWithID(ctx, func(ino uint32) fusefs.InodeEmbedder {
out.Attr.Ino = uint64(ino)
cn.attr.Ino = uint64(ino)
sAttr.Ino = uint64(ino)
return n.NewInode(ctx, cn, sAttr)
})
}
// refnode is the node at <mountpoint>/<imageref>.
type refnode struct {
fusefs.Inode
fs *fs
attr fuse.Attr
ref reference.Spec
}
var _ = (fusefs.InodeEmbedder)((*refnode)(nil))
var _ = (fusefs.NodeLookuper)((*refnode)(nil))
// Lookup returns layernode of the specified name
func (n *refnode) Lookup(ctx context.Context, name string, out *fuse.EntryOut) (*fusefs.Inode, syscall.Errno) {
// lookup on memory nodes
if cn := n.GetChild(name); cn != nil {
switch tn := cn.Operations().(type) {
case *layernode:
copyAttr(&out.Attr, &tn.attr)
default:
log.G(ctx).Warn("refnode.Lookup: uknown node type detected")
return nil, syscall.EIO
}
out.Attr.Ino = cn.StableAttr().Ino
return cn, 0
}
targetDigest, err := digest.Parse(name)
if err != nil {
log.G(ctx).WithError(err).Warnf("invalid digest for %q", name)
return nil, syscall.EINVAL
}
sAttr := defaultDirAttr(&out.Attr)
cn := &layernode{
fs: n.fs,
digest: targetDigest,
refnode: n,
}
copyAttr(&cn.attr, &out.Attr)
return n.fs.newInodeWithID(ctx, func(ino uint32) fusefs.InodeEmbedder {
out.Attr.Ino = uint64(ino)
cn.attr.Ino = uint64(ino)
sAttr.Ino = uint64(ino)
return n.NewInode(ctx, cn, sAttr)
})
}
var _ = (fusefs.NodeRmdirer)((*refnode)(nil))
// Rmdir marks this layer as "release".
// We don't use layernode.Unlink because Unlink event doesn't reach here when "use" file isn't visible
// to the filesystem client.
func (n *refnode) Rmdir(ctx context.Context, name string) syscall.Errno {
targetDigest, err := digest.Parse(name)
if err != nil {
log.G(ctx).WithError(err).Warnf("invalid digest for %q during release", name)
return syscall.EINVAL
}
current, err := n.fs.layerManager.release(ctx, n.ref, targetDigest)
if err != nil {
log.G(ctx).WithError(err).Warnf("failed to release layer %v / %v", n.ref, targetDigest)
return syscall.EIO
}
if current == 0 {
n.fs.knownNodeMu.Lock()
lh, ok := n.fs.knownNode[n.ref.String()][targetDigest.String()]
if !ok {
n.fs.knownNodeMu.Unlock()
log.G(ctx).WithError(err).Warnf("node of layer %v/%v is not registered", n.ref, targetDigest)
return syscall.EIO
}
lh.release()
delete(n.fs.knownNode[n.ref.String()], targetDigest.String())
if len(n.fs.knownNode[n.ref.String()]) == 0 {
delete(n.fs.knownNode, n.ref.String())
}
n.fs.knownNodeMu.Unlock()
}
log.G(ctx).WithField("refcounter", current).Infof("layer %v/%v is marked as RELEASE", n.ref, targetDigest)
return syscall.ENOENT
}
// layernode is the node at <mountpoint>/<imageref>/<layerdigest>.
type layernode struct {
fusefs.Inode
attr fuse.Attr
fs *fs
refnode *refnode
digest digest.Digest
}
var _ = (fusefs.InodeEmbedder)((*layernode)(nil))
var _ = (fusefs.NodeCreater)((*layernode)(nil))
// Create marks this layer as "using".
// We don't use refnode.Mkdir because Mkdir event doesn't reach here if layernode already exists.
func (n *layernode) Create(ctx context.Context, name string, flags uint32, mode uint32, out *fuse.EntryOut) (node *fusefs.Inode, fh fusefs.FileHandle, fuseFlags uint32, errno syscall.Errno) {
if name == layerUseFile {
current := n.fs.layerManager.use(n.refnode.ref, n.digest)
log.G(ctx).WithField("refcounter", current).Infof("layer %v / %v is marked as USING", n.refnode.ref, n.digest)
}
return nil, nil, 0, syscall.ENOENT
}
var _ = (fusefs.NodeLookuper)((*layernode)(nil))
// Lookup routes to the target file stored in the pool, based on the specified file name.
func (n *layernode) Lookup(ctx context.Context, name string, out *fuse.EntryOut) (*fusefs.Inode, syscall.Errno) {
switch name {
case layerInfoLink:
info, err := n.fs.layerManager.getLayerInfo(ctx, n.refnode.ref, n.digest)
if err != nil {
log.G(ctx).WithError(err).Warnf("failed to get layer info for %q: %q", name, n.digest)
return nil, syscall.EIO
}
buf := new(bytes.Buffer)
if err := json.NewEncoder(buf).Encode(&info); err != nil {
log.G(ctx).WithError(err).Warnf("failed to encode layer info for %q: %q", name, n.digest)
return nil, syscall.EIO
}
infoData := buf.Bytes()
sAttr := defaultFileAttr(uint64(len(infoData)), &out.Attr)
cn := &fusefs.MemRegularFile{Data: infoData}
copyAttr(&cn.Attr, &out.Attr)
return n.fs.newInodeWithID(ctx, func(ino uint32) fusefs.InodeEmbedder {
out.Attr.Ino = uint64(ino)
cn.Attr.Ino = uint64(ino)
sAttr.Ino = uint64(ino)
return n.NewInode(ctx, cn, sAttr)
})
case layerLink, blobLink:
// Check if layer is already known
if name == layerLink {
n.fs.knownNodeMu.Lock()
if lh, ok := n.fs.knownNode[n.refnode.ref.String()][n.digest.String()]; ok {
var ao fuse.AttrOut
if errno := lh.n.(fusefs.NodeGetattrer).Getattr(ctx, nil, &ao); errno != 0 {
return nil, errno
}
copyAttr(&out.Attr, &ao.Attr)
n.fs.knownNodeMu.Unlock()
return n.NewInode(ctx, lh.n, fusefs.StableAttr{
Mode: out.Attr.Mode,
Ino: out.Attr.Ino,
}), 0
}
n.fs.knownNodeMu.Unlock()
}
// Resolve layer
l, err := n.fs.layerManager.getLayer(ctx, n.refnode.ref, n.digest)
if err != nil {
cErr := ctx.Err()
if errors.Is(cErr, context.Canceled) || errors.Is(err, context.Canceled) {
// When filesystem client canceled to lookup this layer,
// do not log this as "preparation failure" because it's
// intensional.
log.G(ctx).WithError(err).Debugf("error resolving layer (context error: %v)", cErr)
return nil, syscall.EIO
}
log.G(ctx).WithField(remoteSnapshotLogKey, prepareFailed).
WithField("layerdigest", n.digest).
WithError(err).
Debugf("error resolving layer (context error: %v)", cErr)
log.G(ctx).WithError(err).Warnf("failed to mount layer %q: %q", name, n.digest)
return nil, syscall.EIO
}
if name == blobLink {
sAttr := layerToAttr(l, &out.Attr)
cn := &blobnode{l: l}
copyAttr(&cn.attr, &out.Attr)
return n.fs.newInodeWithID(ctx, func(ino uint32) fusefs.InodeEmbedder {
out.Attr.Ino = uint64(ino)
cn.attr.Ino = uint64(ino)
sAttr.Ino = uint64(ino)
return n.NewInode(ctx, cn, sAttr)
})
}
var cn *fusefs.Inode
var errno syscall.Errno
err = n.fs.layerMap.add(func(id uint32) (releasable, error) {
root, err := l.RootNode(id)
if err != nil {
return nil, err
}
var ao fuse.AttrOut
errno = root.(fusefs.NodeGetattrer).Getattr(ctx, nil, &ao)
if errno != 0 {
return nil, fmt.Errorf("failed to get root node: %v", errno)
}
copyAttr(&out.Attr, &ao.Attr)
cn = n.NewInode(ctx, root, fusefs.StableAttr{
Mode: out.Attr.Mode,
Ino: out.Attr.Ino,
})
rr := &layerReleasable{n: root}
n.fs.knownNodeMu.Lock()
if n.fs.knownNode == nil {
n.fs.knownNode = make(map[string]map[string]*layerReleasable)
}
if n.fs.knownNode[n.refnode.ref.String()] == nil {
n.fs.knownNode[n.refnode.ref.String()] = make(map[string]*layerReleasable)
}
n.fs.knownNode[n.refnode.ref.String()][n.digest.String()] = rr
n.fs.knownNodeMu.Unlock()
return rr, nil
})
if err != nil || errno != 0 {
log.G(ctx).WithField(remoteSnapshotLogKey, prepareFailed).
WithField("layerdigest", n.digest).
WithError(err).
WithField("errno", errno).
Debugf("failed to get root node")
if errno == 0 {
errno = syscall.EIO
}
return nil, errno
}
return cn, 0
case layerUseFile:
log.G(ctx).Debugf("\"use\" file is referred but return ENOENT for reference management")
return nil, syscall.ENOENT
default:
log.G(ctx).Warnf("unknown filename %q", name)
return nil, syscall.ENOENT
}
}
// blobnode is a regular file node that contains raw blob data
type blobnode struct {
fusefs.Inode
l layer.Layer
attr fuse.Attr
}
var _ = (fusefs.InodeEmbedder)((*blobnode)(nil))
var _ = (fusefs.NodeOpener)((*blobnode)(nil))
func (n *blobnode) Open(ctx context.Context, flags uint32) (fh fusefs.FileHandle, fuseFlags uint32, errno syscall.Errno) {
return &blobfile{l: n.l}, 0, 0
}
// blob file is the file handle of blob contents.
type blobfile struct {
l layer.Layer
}
var _ = (fusefs.FileReader)((*blobfile)(nil))
func (f *blobfile) Read(ctx context.Context, dest []byte, off int64) (fuse.ReadResult, syscall.Errno) {
s, err := f.l.ReadAt(dest, off,
remote.WithContext(ctx), // Make cancellable
remote.WithCacheOpts(cache.Direct()), // Do not pollute mem cache
)
if err != nil && err != io.EOF {
return nil, syscall.EIO
}
return fuse.ReadResultData(dest[:s]), 0
}
var _ = (fusefs.FileGetattrer)((*blobfile)(nil))
func (f *blobfile) Getattr(ctx context.Context, out *fuse.AttrOut) syscall.Errno {
layerToAttr(f.l, &out.Attr)
return 0
}
func copyAttr(dest, src *fuse.Attr) {
dest.Ino = src.Ino
dest.Size = src.Size
dest.Blocks = src.Blocks
dest.Atime = src.Atime
dest.Mtime = src.Mtime
dest.Ctime = src.Ctime
dest.Atimensec = src.Atimensec
dest.Mtimensec = src.Mtimensec
dest.Ctimensec = src.Ctimensec
dest.Mode = src.Mode
dest.Nlink = src.Nlink
dest.Owner = src.Owner
dest.Rdev = src.Rdev
dest.Blksize = src.Blksize
dest.Padding = src.Padding
}
func layerToAttr(l layer.Layer, out *fuse.Attr) fusefs.StableAttr {
// out.Ino
out.Size = uint64(l.Info().Size)
out.Blksize = blockSize
out.Blocks = out.Size / uint64(out.Blksize)
if out.Size%uint64(out.Blksize) > 0 {
out.Blocks++
}
out.Nlink = 1
out.Mode = layerFileMode
out.Owner = fuse.Owner{Uid: 0, Gid: 0}
// out.Mtime
// out.Mtimensec
// out.Rdev
// out.Padding
return fusefs.StableAttr{
Mode: out.Mode,
}
}
func defaultFileAttr(size uint64, out *fuse.Attr) fusefs.StableAttr {
// out.Ino
out.Size = size
out.Blksize = blockSize
out.Blocks = out.Size / uint64(out.Blksize)
if out.Size%uint64(out.Blksize) > 0 {
out.Blocks++
}
out.Nlink = 1
out.Mode = defaultFileMode
out.Owner = fuse.Owner{Uid: 0, Gid: 0}
// out.Mtime
// out.Mtimensec
// out.Rdev
// out.Padding
return fusefs.StableAttr{
Mode: out.Mode,
}
}
func defaultDirAttr(out *fuse.Attr) fusefs.StableAttr {
// out.Ino
out.Size = 0
// out.Blksize
// out.Blocks
// out.Nlink
out.Mode = defaultDirMode
out.Owner = fuse.Owner{Uid: 0, Gid: 0}
// out.Mtime
// out.Mtimensec
// out.Rdev
// out.Padding
return fusefs.StableAttr{
Mode: out.Mode,
}
}
func defaultLinkAttr(out *fuse.Attr) fusefs.StableAttr {
// out.Ino
out.Size = 0
// out.Blksize
// out.Blocks
// out.Nlink
out.Mode = defaultLinkMode
out.Owner = fuse.Owner{Uid: 0, Gid: 0}
// out.Mtime
// out.Mtimensec
// out.Rdev
// out.Padding
return fusefs.StableAttr{
Mode: out.Mode,
}
}
// idMap manages uint32 IDs with automatic GC for releasable objects.
type idMap struct {
m map[uint32]releasable
max uint32
mu sync.Mutex
cleanupG singleflight.Group
}
type releasable interface {
releasable() bool
}
// add reserves an unique uint32 object for the provided releasable object.
// when that object become releasable, that ID will be reused for other objects.
func (m *idMap) add(p func(uint32) (releasable, error)) error {
m.cleanupG.Do("cleanup", func() (interface{}, error) {
m.mu.Lock()
defer m.mu.Unlock()
max := uint32(0)
for i := uint32(0); i <= m.max; i++ {
if e, ok := m.m[i]; ok {
if e.releasable() {
delete(m.m, i)
} else {
max = i
}
}
}
m.max = max
return nil, nil
})
m.mu.Lock()
defer m.mu.Unlock()
if m.m == nil {
m.m = make(map[uint32]releasable)
}
for i := uint32(0); i <= ^uint32(0); i++ {
if i == 0 {
continue
}
e, ok := m.m[i]
if !ok || e.releasable() {
r, err := p(i)
if err != nil {
return err
}
if m.max < i {
m.max = i
}
m.m[i] = r
return nil
}
}
return fmt.Errorf("no ID is usable")
}