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
|
// Copyright 2018 Google LLC
//
// 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 ramfs
import (
"sync"
"syscall"
"gvisor.googlesource.com/gvisor/pkg/abi/linux"
"gvisor.googlesource.com/gvisor/pkg/sentry/context"
"gvisor.googlesource.com/gvisor/pkg/sentry/fs"
"gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil"
"gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport"
"gvisor.googlesource.com/gvisor/pkg/syserror"
)
// CreateOps represents operations to create different file types.
type CreateOps struct {
// NewDir creates a new directory.
NewDir func(ctx context.Context, dir *fs.Inode, perms fs.FilePermissions) (*fs.Inode, error)
// NewFile creates a new file.
NewFile func(ctx context.Context, dir *fs.Inode, perms fs.FilePermissions) (*fs.Inode, error)
// NewSymlink creates a new symlink with permissions 0777.
NewSymlink func(ctx context.Context, dir *fs.Inode, target string) (*fs.Inode, error)
// NewBoundEndpoint creates a new socket.
NewBoundEndpoint func(ctx context.Context, dir *fs.Inode, ep transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Inode, error)
// NewFifo creates a new fifo.
NewFifo func(ctx context.Context, dir *fs.Inode, perm fs.FilePermissions) (*fs.Inode, error)
}
// Dir represents a single directory in the filesystem.
//
// +stateify savable
type Dir struct {
fsutil.InodeGenericChecker `state:"nosave"`
fsutil.InodeIsDirTruncate `state:"nosave"`
fsutil.InodeNoopRelease `state:"nosave"`
fsutil.InodeNoopWriteOut `state:"nosave"`
fsutil.InodeNotMappable `state:"nosave"`
fsutil.InodeNotSocket `state:"nosave"`
fsutil.InodeNotSymlink `state:"nosave"`
fsutil.InodeVirtual `state:"nosave"`
fsutil.InodeSimpleAttributes
fsutil.InodeSimpleExtendedAttributes
// CreateOps may be provided.
//
// These may only be modified during initialization (while the application
// is not running). No sychronization is performed when accessing these
// operations during syscalls.
*CreateOps `state:"nosave"`
// mu protects the fields below.
mu sync.Mutex `state:"nosave"`
// children are inodes that are in this directory. A reference is held
// on each inode while it is in the map.
children map[string]*fs.Inode
// dentryMap is a sortedDentryMap containing entries for all children.
// Its entries are kept up-to-date with d.children.
dentryMap *fs.SortedDentryMap
}
var _ fs.InodeOperations = (*Dir)(nil)
// NewDir returns a new Dir with the given contents and attributes.
func NewDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwner, perms fs.FilePermissions) *Dir {
d := &Dir{
InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, perms, linux.RAMFS_MAGIC),
}
if contents == nil {
contents = make(map[string]*fs.Inode)
}
d.children = contents
// Build the entries map ourselves, rather than calling addChildLocked,
// because it will be faster.
entries := make(map[string]fs.DentAttr, len(contents))
for name, inode := range contents {
entries[name] = fs.DentAttr{
Type: inode.StableAttr.Type,
InodeID: inode.StableAttr.InodeID,
}
}
d.dentryMap = fs.NewSortedDentryMap(entries)
// Directories have an extra link, corresponding to '.'.
d.AddLink()
return d
}
// addChildLocked add the child inode, inheriting its reference.
func (d *Dir) addChildLocked(name string, inode *fs.Inode) {
d.children[name] = inode
d.dentryMap.Add(name, fs.DentAttr{
Type: inode.StableAttr.Type,
InodeID: inode.StableAttr.InodeID,
})
// If the child is a directory, increment this dir's link count,
// corresponding to '..' from the subdirectory.
if fs.IsDir(inode.StableAttr) {
d.AddLink()
}
// Given we're now adding this inode to the directory we must also
// increase its link count. Similarly we decremented it in removeChildLocked.
inode.AddLink()
}
// AddChild adds a child to this dir.
func (d *Dir) AddChild(ctx context.Context, name string, inode *fs.Inode) {
d.mu.Lock()
defer d.mu.Unlock()
d.addChildLocked(name, inode)
}
// FindChild returns (child, true) if the directory contains name.
func (d *Dir) FindChild(name string) (*fs.Inode, bool) {
d.mu.Lock()
defer d.mu.Unlock()
child, ok := d.children[name]
return child, ok
}
// Children returns the names and DentAttrs of all children. It can be used to
// implement Readdir for types that embed ramfs.Dir.
func (d *Dir) Children() ([]string, map[string]fs.DentAttr) {
d.mu.Lock()
defer d.mu.Unlock()
// Return a copy to prevent callers from modifying our children.
names, entries := d.dentryMap.GetAll()
namesCopy := make([]string, len(names))
copy(namesCopy, names)
entriesCopy := make(map[string]fs.DentAttr)
for k, v := range entries {
entriesCopy[k] = v
}
return namesCopy, entriesCopy
}
// removeChildLocked attempts to remove an entry from this directory.
func (d *Dir) removeChildLocked(ctx context.Context, name string) (*fs.Inode, error) {
inode, ok := d.children[name]
if !ok {
return nil, syserror.EACCES
}
delete(d.children, name)
d.dentryMap.Remove(name)
d.NotifyModification(ctx)
// If the child was a subdirectory, then we must decrement this dir's
// link count which was the child's ".." directory entry.
if fs.IsDir(inode.StableAttr) {
d.DropLink()
}
// Update ctime.
inode.InodeOperations.NotifyStatusChange(ctx)
// Given we're now removing this inode to the directory we must also
// decrease its link count. Similarly it is increased in addChildLocked.
inode.DropLink()
return inode, nil
}
// Remove removes the named non-directory.
func (d *Dir) Remove(ctx context.Context, _ *fs.Inode, name string) error {
d.mu.Lock()
defer d.mu.Unlock()
inode, err := d.removeChildLocked(ctx, name)
if err != nil {
return err
}
// Remove our reference on the inode.
inode.DecRef()
return nil
}
// RemoveDirectory removes the named directory.
func (d *Dir) RemoveDirectory(ctx context.Context, _ *fs.Inode, name string) error {
d.mu.Lock()
defer d.mu.Unlock()
// Get the child and make sure it is not empty.
childInode, err := d.walkLocked(ctx, name)
if err != nil {
return err
}
if ok, err := hasChildren(ctx, childInode); err != nil {
return err
} else if ok {
return syserror.ENOTEMPTY
}
// Child was empty. Proceed with removal.
inode, err := d.removeChildLocked(ctx, name)
if err != nil {
return err
}
// Remove our reference on the inode.
inode.DecRef()
return nil
}
// Lookup loads an inode at p into a Dirent.
func (d *Dir) Lookup(ctx context.Context, _ *fs.Inode, p string) (*fs.Dirent, error) {
d.mu.Lock()
defer d.mu.Unlock()
inode, err := d.walkLocked(ctx, p)
if err != nil {
return nil, err
}
// Take a reference on the inode before returning it. This reference
// is owned by the dirent we are about to create.
inode.IncRef()
return fs.NewDirent(inode, p), nil
}
// walkLocked must be called with d.mu held.
func (d *Dir) walkLocked(ctx context.Context, p string) (*fs.Inode, error) {
d.NotifyAccess(ctx)
// Lookup a child node.
if inode, ok := d.children[p]; ok {
return inode, nil
}
// fs.InodeOperations.Lookup returns syserror.ENOENT if p
// does not exist.
return nil, syserror.ENOENT
}
// createInodeOperationsCommon creates a new child node at this dir by calling
// makeInodeOperations. It is the common logic for creating a new child.
func (d *Dir) createInodeOperationsCommon(ctx context.Context, name string, makeInodeOperations func() (*fs.Inode, error)) (*fs.Inode, error) {
d.mu.Lock()
defer d.mu.Unlock()
if _, ok := d.children[name]; ok {
return nil, syscall.EEXIST
}
inode, err := makeInodeOperations()
if err != nil {
return nil, err
}
d.addChildLocked(name, inode)
d.NotifyModification(ctx)
return inode, nil
}
// Create creates a new Inode with the given name and returns its File.
func (d *Dir) Create(ctx context.Context, dir *fs.Inode, name string, flags fs.FileFlags, perms fs.FilePermissions) (*fs.File, error) {
if d.CreateOps == nil || d.CreateOps.NewFile == nil {
return nil, syserror.EACCES
}
inode, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) {
return d.NewFile(ctx, dir, perms)
})
if err != nil {
return nil, err
}
// Take an extra ref on inode, which will be owned by the dirent.
inode.IncRef()
// Create the Dirent and corresponding file.
created := fs.NewDirent(inode, name)
defer created.DecRef()
return created.Inode.GetFile(ctx, created, flags)
}
// CreateLink returns a new link.
func (d *Dir) CreateLink(ctx context.Context, dir *fs.Inode, oldname, newname string) error {
if d.CreateOps == nil || d.CreateOps.NewSymlink == nil {
return syserror.EACCES
}
_, err := d.createInodeOperationsCommon(ctx, newname, func() (*fs.Inode, error) {
return d.NewSymlink(ctx, dir, oldname)
})
return err
}
// CreateHardLink creates a new hard link.
func (d *Dir) CreateHardLink(ctx context.Context, dir *fs.Inode, target *fs.Inode, name string) error {
d.mu.Lock()
defer d.mu.Unlock()
// Take an extra reference on the inode and add it to our children.
target.IncRef()
// The link count will be incremented in addChildLocked.
d.addChildLocked(name, target)
d.NotifyModification(ctx)
// Update ctime.
target.InodeOperations.NotifyStatusChange(ctx)
return nil
}
// CreateDirectory returns a new subdirectory.
func (d *Dir) CreateDirectory(ctx context.Context, dir *fs.Inode, name string, perms fs.FilePermissions) error {
if d.CreateOps == nil || d.CreateOps.NewDir == nil {
return syserror.EACCES
}
_, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) {
return d.NewDir(ctx, dir, perms)
})
// TODO: Support updating status times, as those should be
// updated by links.
return err
}
// Bind implements fs.InodeOperations.Bind.
func (d *Dir) Bind(ctx context.Context, dir *fs.Inode, name string, ep transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Dirent, error) {
if d.CreateOps == nil || d.CreateOps.NewBoundEndpoint == nil {
return nil, syserror.EACCES
}
inode, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) {
return d.NewBoundEndpoint(ctx, dir, ep, perms)
})
if err == syscall.EEXIST {
return nil, syscall.EADDRINUSE
}
if err != nil {
return nil, err
}
// Take another ref on inode which will be donated to the new dirent.
inode.IncRef()
return fs.NewDirent(inode, name), nil
}
// CreateFifo implements fs.InodeOperations.CreateFifo.
func (d *Dir) CreateFifo(ctx context.Context, dir *fs.Inode, name string, perms fs.FilePermissions) error {
if d.CreateOps == nil || d.CreateOps.NewFifo == nil {
return syserror.EACCES
}
_, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) {
return d.NewFifo(ctx, dir, perms)
})
return err
}
// GetFile implements fs.InodeOperations.GetFile.
func (d *Dir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) {
flags.Pread = true
return fs.NewFile(ctx, dirent, flags, &dirFileOperations{dir: d}), nil
}
// Rename implements fs.InodeOperations.Rename.
func (*Dir) Rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent *fs.Inode, newName string) error {
return Rename(ctx, oldParent.InodeOperations, oldName, newParent.InodeOperations, newName)
}
// dirFileOperations implements fs.FileOperations for a ramfs directory.
//
// +stateify savable
type dirFileOperations struct {
fsutil.DirFileOperations `state:"nosave"`
// dirCursor contains the name of the last directory entry that was
// serialized.
dirCursor string
// dir is the ramfs dir that this file corresponds to.
dir *Dir
}
var _ fs.FileOperations = (*dirFileOperations)(nil)
// Seek implements fs.FileOperations.Seek.
func (dfo *dirFileOperations) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) {
return fsutil.SeekWithDirCursor(ctx, file, whence, offset, &dfo.dirCursor)
}
// IterateDir implements DirIterator.IterateDir.
func (dfo *dirFileOperations) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) {
dfo.dir.mu.Lock()
defer dfo.dir.mu.Unlock()
n, err := fs.GenericReaddir(dirCtx, dfo.dir.dentryMap)
return offset + n, err
}
// Readdir implements FileOperations.Readdir.
func (dfo *dirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) {
root := fs.RootFromContext(ctx)
defer root.DecRef()
dirCtx := &fs.DirCtx{
Serializer: serializer,
DirCursor: &dfo.dirCursor,
}
dfo.dir.InodeSimpleAttributes.NotifyAccess(ctx)
return fs.DirentReaddir(ctx, file.Dirent, dfo, root, dirCtx, file.Offset())
}
// hasChildren is a helper method that determines whether an arbitrary inode
// (not necessarily ramfs) has any children.
func hasChildren(ctx context.Context, inode *fs.Inode) (bool, error) {
// Take an extra ref on inode which will be given to the dirent and
// dropped when that dirent is destroyed.
inode.IncRef()
d := fs.NewTransientDirent(inode)
defer d.DecRef()
file, err := inode.GetFile(ctx, d, fs.FileFlags{Read: true})
if err != nil {
return false, err
}
defer file.DecRef()
ser := &fs.CollectEntriesSerializer{}
if err := file.Readdir(ctx, ser); err != nil {
return false, err
}
// We will always write "." and "..", so ignore those two.
if ser.Written() > 2 {
return true, nil
}
return false, nil
}
// Rename renames from a *ramfs.Dir to another *ramfs.Dir.
func Rename(ctx context.Context, oldParent fs.InodeOperations, oldName string, newParent fs.InodeOperations, newName string) error {
op, ok := oldParent.(*Dir)
if !ok {
return syserror.EXDEV
}
np, ok := newParent.(*Dir)
if !ok {
return syserror.EXDEV
}
np.mu.Lock()
defer np.mu.Unlock()
// Check whether the ramfs entry to be replaced is a non-empty directory.
if replaced, ok := np.children[newName]; ok {
if fs.IsDir(replaced.StableAttr) {
if ok, err := hasChildren(ctx, replaced); err != nil {
return err
} else if ok {
return syserror.ENOTEMPTY
}
}
}
// Be careful, we may have already grabbed this mutex above.
if op != np {
op.mu.Lock()
defer op.mu.Unlock()
}
// Do the swap.
n := op.children[oldName]
op.removeChildLocked(ctx, oldName)
np.addChildLocked(newName, n)
// Update ctime.
n.InodeOperations.NotifyStatusChange(ctx)
return nil
}
|