-
Notifications
You must be signed in to change notification settings - Fork 0
/
fuse.go
679 lines (564 loc) · 15.9 KB
/
fuse.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
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
package hafs
import (
"errors"
"io"
iofs "io/fs"
"log"
mathrand "math/rand"
"os"
"sync"
"sync/atomic"
"time"
"github.com/hanwen/go-fuse/v2/fuse"
"golang.org/x/sys/unix"
)
type openFile struct {
maybeHasPosixLock atomicBool
di *DirIter
f HafsFile
}
type HafsFuseOptions struct {
CacheDentries time.Duration
CacheAttributes time.Duration
Logf func(string, ...interface{})
}
type FuseFs struct {
server *fuse.Server
cacheDentries time.Duration
cacheAttributes time.Duration
logf func(string, ...interface{})
fs *Fs
fileHandleCounter uint64
fh2OpenFile sync.Map
}
func NewFuseFs(fs *Fs, opts HafsFuseOptions) *FuseFs {
if opts.Logf == nil {
opts.Logf = log.Printf
}
return &FuseFs{
cacheDentries: opts.CacheDentries,
cacheAttributes: opts.CacheAttributes,
logf: opts.Logf,
fs: fs,
fh2OpenFile: sync.Map{},
}
}
func (fs *FuseFs) newFileHandle(f *openFile) uint64 {
fh := atomic.AddUint64(&fs.fileHandleCounter, 1)
fs.fh2OpenFile.Store(fh, f)
return fh
}
func (fs *FuseFs) getFileFromHandle(handle uint64) (*openFile, bool) {
f, ok := fs.fh2OpenFile.Load(handle)
if !ok {
return nil, false
}
return f.(*openFile), true
}
func (fs *FuseFs) releaseFileHandle(handle uint64) (*openFile, bool) {
f, ok := fs.fh2OpenFile.LoadAndDelete(handle)
if !ok {
return nil, false
}
return f.(*openFile), true
}
func (fs *FuseFs) errToFuseStatus(err error) fuse.Status {
if err == nil {
return fuse.OK
}
if errno, ok := err.(unix.Errno); ok {
return fuse.Status(errno)
}
if errors.Is(err, iofs.ErrNotExist) {
return fuse.Status(unix.ENOENT)
} else if errors.Is(err, iofs.ErrPermission) {
return fuse.Status(unix.EPERM)
} else if errors.Is(err, iofs.ErrExist) {
return fuse.Status(unix.EEXIST)
} else if errors.Is(err, iofs.ErrInvalid) {
return fuse.Status(unix.EINVAL)
}
// Log all io errors that don't have a clear cause.
fs.logf("io error: %s", err)
return fuse.Status(fuse.EIO)
}
func (fs *FuseFs) fillFuseAttrFromStat(stat *Stat, out *fuse.Attr) {
out.Ino = stat.Ino
out.Size = stat.Size
out.Blocks = stat.Size / 512
out.Blksize = CHUNK_SIZE
out.Atime = stat.Atimesec
out.Atimensec = stat.Atimensec
out.Mtime = stat.Mtimesec
out.Mtimensec = stat.Mtimensec
out.Ctime = stat.Ctimesec
out.Ctimensec = stat.Ctimensec
out.Mode = stat.Mode
out.Nlink = stat.Nlink
out.Owner.Uid = stat.Uid
out.Owner.Gid = stat.Gid
out.Rdev = stat.Rdev
}
func (fs *FuseFs) fillFuseAttrOutFromStat(stat *Stat, out *fuse.AttrOut) {
fs.fillFuseAttrFromStat(stat, &out.Attr)
out.AttrValid = uint64(fs.cacheAttributes.Nanoseconds() / 1_000_000_000)
out.AttrValidNsec = uint32(uint64(fs.cacheAttributes.Nanoseconds()) - out.AttrValid*1_000_000_000)
}
func (fs *FuseFs) fillFuseEntryOutFromStat(stat *Stat, out *fuse.EntryOut) {
out.Generation = 0
out.NodeId = stat.Ino
fs.fillFuseAttrFromStat(stat, &out.Attr)
out.AttrValid = uint64(fs.cacheAttributes.Nanoseconds() / 1_000_000_000)
out.AttrValidNsec = uint32(uint64(fs.cacheAttributes.Nanoseconds()) - out.AttrValid*1_000_000_000)
out.EntryValid = uint64(fs.cacheDentries.Nanoseconds() / 1_000_000_000)
out.EntryValidNsec = uint32(uint64(fs.cacheDentries.Nanoseconds()) - out.EntryValid*1_000_000_000)
}
func (fs *FuseFs) Init(server *fuse.Server) {
fs.server = server
}
func (fs *FuseFs) Lookup(cancel <-chan struct{}, header *fuse.InHeader, name string, out *fuse.EntryOut) fuse.Status {
stat, err := fs.fs.Lookup(header.NodeId, name)
if err != nil {
return fs.errToFuseStatus(err)
}
fs.fillFuseEntryOutFromStat(&stat, out)
return fuse.OK
}
func (fs *FuseFs) Forget(nodeId, nlookup uint64) {
}
func (fs *FuseFs) GetAttr(cancel <-chan struct{}, in *fuse.GetAttrIn, out *fuse.AttrOut) fuse.Status {
stat, err := fs.fs.GetStat(in.NodeId)
if err != nil {
return fs.errToFuseStatus(err)
}
fs.fillFuseAttrOutFromStat(&stat, out)
return fuse.OK
}
func (fs *FuseFs) SetAttr(cancel <-chan struct{}, in *fuse.SetAttrIn, out *fuse.AttrOut) fuse.Status {
modStat := ModStatOpts{}
if mtime, ok := in.GetMTime(); ok {
modStat.SetMtime(mtime)
}
if atime, ok := in.GetATime(); ok {
modStat.SetAtime(atime)
}
if ctime, ok := in.GetCTime(); ok {
modStat.SetCtime(ctime)
}
if size, ok := in.GetSize(); ok {
modStat.Valid |= MODSTAT_SIZE
modStat.SetSize(size)
}
if mode, ok := in.GetMode(); ok {
modStat.SetMode(mode)
}
if uid, ok := in.GetUID(); ok {
modStat.SetUid(uid)
}
if gid, ok := in.GetGID(); ok {
modStat.SetGid(gid)
}
stat, err := fs.fs.ModStat(in.NodeId, modStat)
if err != nil {
return fs.errToFuseStatus(err)
}
fs.fillFuseAttrOutFromStat(&stat, out)
return fuse.OK
}
func (fs *FuseFs) Open(cancel <-chan struct{}, in *fuse.OpenIn, out *fuse.OpenOut) fuse.Status {
f, _, err := fs.fs.OpenFile(in.NodeId, OpenFileOpts{
Truncate: in.Flags&unix.O_TRUNC != 0,
})
if err != nil {
return fs.errToFuseStatus(err)
}
switch f.(type) {
case *objectStoreSmallReadOnlyFile:
// For now only the small read only files support readahead,
// and are immutable plus cacheable. The streaming files
// can't do this because the kernel readahead makes the
// read offset bounce around.
out.OpenFlags |= fuse.FOPEN_KEEP_CACHE
default:
out.OpenFlags |= fuse.FOPEN_DIRECT_IO
}
out.Fh = fs.newFileHandle(&openFile{f: f})
return fuse.OK
}
func (fs *FuseFs) Create(cancel <-chan struct{}, in *fuse.CreateIn, name string, out *fuse.CreateOut) fuse.Status {
f, stat, err := fs.fs.CreateFile(in.NodeId, name, CreateFileOpts{
Truncate: in.Flags&unix.O_TRUNC != 0,
Mode: in.Mode,
Uid: in.Owner.Uid,
Gid: in.Owner.Gid,
})
if err != nil {
return fs.errToFuseStatus(err)
}
out.OpenFlags |= fuse.FOPEN_DIRECT_IO
fs.fillFuseEntryOutFromStat(&stat, &out.EntryOut)
out.Fh = fs.newFileHandle(&openFile{f: f})
return fuse.OK
}
func (fs *FuseFs) Rename(cancel <-chan struct{}, in *fuse.RenameIn, fromName string, toName string) fuse.Status {
fromDir := in.NodeId
toDir := in.Newdir
err := fs.fs.Rename(fromDir, toDir, fromName, toName)
if err != nil {
return fs.errToFuseStatus(err)
}
return fuse.OK
}
func (fs *FuseFs) Read(cancel <-chan struct{}, in *fuse.ReadIn, buf []byte) (fuse.ReadResult, fuse.Status) {
f, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return nil, fuse.EBADF
}
nTotal := uint32(0)
for nTotal != uint32(len(buf)) {
n, err := f.f.ReadData(buf[nTotal:], uint64(in.Offset)+uint64(nTotal))
nTotal += n
if err == io.EOF {
break
}
if err != nil {
return nil, fs.errToFuseStatus(err)
}
}
return fuse.ReadResultData(buf[:nTotal]), fuse.OK
}
func (fs *FuseFs) Write(cancel <-chan struct{}, in *fuse.WriteIn, buf []byte) (uint32, fuse.Status) {
nTotal := uint32(0)
f, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return nTotal, fuse.EBADF
}
for nTotal != uint32(len(buf)) {
n, err := f.f.WriteData(buf[nTotal:], uint64(in.Offset)+uint64(nTotal))
nTotal += uint32(n)
if err != nil {
return nTotal, fs.errToFuseStatus(err)
}
}
return nTotal, fuse.OK
}
func (fs *FuseFs) Lseek(cancel <-chan struct{}, in *fuse.LseekIn, out *fuse.LseekOut) fuse.Status {
// XXX We do support sparse files, so this could be implemented.
// It's worth noting that it seems like fuse only uses Lseek for SEEK_DATA and SEEK_HOLE but
// we could be wrong on that.
return fuse.ENOSYS
}
func (fs *FuseFs) Fsync(cancel <-chan struct{}, in *fuse.FsyncIn) fuse.Status {
f, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return fuse.EBADF
}
err := f.f.Fsync()
if err != nil {
return fs.errToFuseStatus(err)
}
return fuse.OK
}
func (fs *FuseFs) Flush(cancel <-chan struct{}, in *fuse.FlushIn) fuse.Status {
f, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return fuse.EBADF
}
fsyncErr := f.f.Fsync()
if f.maybeHasPosixLock.Load() {
// Note, this this behavior intentionally violates posix lock semantics:
//
// Normally posix locks are associated with a process id, so any file descriptor
// that opens and closes a file will release the lock for all files, but here we only
// are only releasing it if the file that created the lock is closed.
//
// We *could* implement the full semantics, but at increased complexity, reduced
// performance, and mainly to support potentially questionable use cases. For now
// we will instead document our semantics and keep it simple.
fs.releaseLocks(in.NodeId, in.LockOwner)
}
return fs.errToFuseStatus(fsyncErr)
}
func (fs *FuseFs) Release(cancel <-chan struct{}, in *fuse.ReleaseIn) {
f, ok := fs.releaseFileHandle(in.Fh)
if ok && f.f != nil {
_ = f.f.Close()
}
const FUSE_RELEASE_FLOCK_UNLOCK = (1 << 1) // XXX remove once constant is in upstream go-fuse.
if in.ReleaseFlags&FUSE_RELEASE_FLOCK_UNLOCK != 0 {
fs.releaseLocks(in.NodeId, in.LockOwner)
}
}
func (fs *FuseFs) Unlink(cancel <-chan struct{}, in *fuse.InHeader, name string) fuse.Status {
err := fs.fs.Unlink(in.NodeId, name)
return fs.errToFuseStatus(err)
}
func (fs *FuseFs) Rmdir(cancel <-chan struct{}, in *fuse.InHeader, name string) fuse.Status {
err := fs.fs.Unlink(in.NodeId, name)
return fs.errToFuseStatus(err)
}
func (fs *FuseFs) Link(cancel <-chan struct{}, in *fuse.LinkIn, name string, out *fuse.EntryOut) fuse.Status {
stat, err := fs.fs.HardLink(in.NodeId, in.Oldnodeid, name)
if err != nil {
return fs.errToFuseStatus(err)
}
fs.fillFuseEntryOutFromStat(&stat, out)
return fuse.OK
}
func (fs *FuseFs) Symlink(cancel <-chan struct{}, in *fuse.InHeader, pointedTo string, linkName string, out *fuse.EntryOut) fuse.Status {
stat, err := fs.fs.Mknod(in.NodeId, linkName, MknodOpts{
Mode: S_IFLNK | 0o777,
Uid: in.Owner.Uid,
Gid: in.Owner.Gid,
LinkTarget: []byte(pointedTo),
})
if err != nil {
return fs.errToFuseStatus(err)
}
fs.fillFuseEntryOutFromStat(&stat, out)
return fuse.OK
}
func (fs *FuseFs) Readlink(cancel <-chan struct{}, in *fuse.InHeader) ([]byte, fuse.Status) {
l, err := fs.fs.ReadSymlink(in.NodeId)
if err != nil {
return nil, fs.errToFuseStatus(err)
}
return l, fuse.OK
}
func (fs *FuseFs) Mkdir(cancel <-chan struct{}, in *fuse.MkdirIn, name string, out *fuse.EntryOut) fuse.Status {
stat, err := fs.fs.Mknod(in.NodeId, name, MknodOpts{
Mode: (^S_IFMT & in.Mode) | S_IFDIR,
Uid: in.Owner.Uid,
Gid: in.Owner.Gid,
})
if err != nil {
return fs.errToFuseStatus(err)
}
fs.fillFuseEntryOutFromStat(&stat, out)
return fuse.OK
}
func (fs *FuseFs) OpenDir(cancel <-chan struct{}, in *fuse.OpenIn, out *fuse.OpenOut) fuse.Status {
dirIter, err := fs.fs.IterDirEnts(in.NodeId)
if err != nil {
return fs.errToFuseStatus(err)
}
out.OpenFlags |= fuse.FOPEN_DIRECT_IO
out.Fh = fs.newFileHandle(&openFile{
di: dirIter,
f: &invalidFile{},
})
return fuse.OK
}
func (fs *FuseFs) ReadDir(cancel <-chan struct{}, in *fuse.ReadIn, out *fuse.DirEntryList) fuse.Status {
d, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return fuse.EBADF
}
if d.di == nil {
return fuse.Status(unix.EBADF)
}
// XXX TODO verify offset is correct.
for {
ent, err := d.di.Next()
if err != nil {
if errors.Is(err, io.EOF) {
break
}
return fs.errToFuseStatus(err)
}
fuseDirEnt := fuse.DirEntry{
Name: ent.Name,
Mode: ent.Mode,
Ino: ent.Ino,
}
if !out.AddDirEntry(fuseDirEnt) {
d.di.Unget(ent)
break
}
}
return fuse.OK
}
func (fs *FuseFs) ReadDirPlus(cancel <-chan struct{}, in *fuse.ReadIn, out *fuse.DirEntryList) fuse.Status {
d, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return fuse.EBADF
}
if d.di == nil {
return fuse.Status(unix.EBADF)
}
// XXX TODO verify offset is correct.
for {
ent, stat, err := d.di.NextPlus()
if err != nil {
if errors.Is(err, io.EOF) {
break
}
return fs.errToFuseStatus(err)
}
fuseDirEnt := fuse.DirEntry{
Name: ent.Name,
Mode: ent.Mode,
Ino: ent.Ino,
}
entryOut := out.AddDirLookupEntry(fuseDirEnt)
if entryOut != nil {
fs.fillFuseEntryOutFromStat(&stat, entryOut)
} else {
d.di.UngetPlus(ent, stat)
break
}
}
return fuse.OK
}
func (fs *FuseFs) FsyncDir(cancel <-chan struct{}, in *fuse.FsyncIn) fuse.Status {
return fuse.OK
}
func (fs *FuseFs) ReleaseDir(in *fuse.ReleaseIn) {
f, ok := fs.releaseFileHandle(in.Fh)
if ok && f.f != nil {
_ = f.f.Close()
}
}
func (fs *FuseFs) GetXAttr(cancel <-chan struct{}, in *fuse.InHeader, attr string, dest []byte) (uint32, fuse.Status) {
x, err := fs.fs.GetXAttr(in.NodeId, attr)
if err != nil {
return 0, fs.errToFuseStatus(err)
}
if len(dest) < len(x) {
return uint32(len(x)), fuse.ERANGE
}
copy(dest, x)
return uint32(len(x)), fuse.OK
}
func (fs *FuseFs) ListXAttr(cancel <-chan struct{}, in *fuse.InHeader, dest []byte) (uint32, fuse.Status) {
xattrs, err := fs.fs.ListXAttr(in.NodeId)
if err != nil {
return 0, fs.errToFuseStatus(err)
}
nNeeded := uint32(0)
for _, x := range xattrs {
nNeeded += uint32(len(x)) + 1
}
if uint32(len(dest)) < nNeeded {
return nNeeded, fuse.ERANGE
}
for _, x := range xattrs {
copy(dest[:len(x)], x)
dest[len(x)] = 0
dest = dest[len(x)+1:]
}
return nNeeded, fuse.OK
}
func (fs *FuseFs) SetXAttr(cancel <-chan struct{}, in *fuse.SetXAttrIn, attr string, data []byte) fuse.Status {
err := fs.fs.SetXAttr(in.NodeId, attr, data)
return fs.errToFuseStatus(err)
}
func (fs *FuseFs) RemoveXAttr(cancel <-chan struct{}, in *fuse.InHeader, attr string) fuse.Status {
err := fs.fs.RemoveXAttr(in.NodeId, attr)
return fs.errToFuseStatus(err)
}
func (fs *FuseFs) releaseLocks(ino uint64, lockOwner uint64) {
for {
_, err := fs.fs.TrySetLock(ino, SetLockOpts{
Typ: LOCK_NONE,
Owner: lockOwner,
})
if err == nil {
break
}
fs.logf("unable to release lock ino=%d owner=%d: %s", ino, lockOwner, err)
time.Sleep(1 * time.Second)
}
}
func (fs *FuseFs) GetLk(cancel <-chan struct{}, in *fuse.LkIn, out *fuse.LkOut) fuse.Status {
return fuse.ENOSYS
}
func (fs *FuseFs) SetLk(cancel <-chan struct{}, in *fuse.LkIn) fuse.Status {
f, ok := fs.getFileFromHandle(in.Fh)
if !ok {
return fuse.EBADF
}
var lockType LockType
switch in.Lk.Typ {
case unix.F_RDLCK:
lockType = LOCK_SHARED
case unix.F_WRLCK:
lockType = LOCK_EXCLUSIVE
case unix.F_UNLCK:
lockType = LOCK_NONE
default:
return fuse.ENOTSUP
}
if in.Lk.Start != 0 {
return fuse.ENOTSUP
}
if in.Lk.End != 0x7fffffffffffffff {
return fuse.ENOTSUP
}
ok, err := fs.fs.TrySetLock(in.NodeId, SetLockOpts{
Typ: lockType,
Owner: in.Owner,
})
if in.LkFlags&fuse.FUSE_LK_FLOCK == 0 {
f.maybeHasPosixLock.Store(true)
}
if err != nil {
return fs.errToFuseStatus(err)
}
if !ok {
return fuse.EAGAIN
}
return fuse.OK
}
func (fs *FuseFs) SetLkw(cancel <-chan struct{}, in *fuse.LkIn) fuse.Status {
nAttempts := uint64(0)
for {
status := fs.SetLk(cancel, in)
if status != fuse.EAGAIN {
return status
}
select {
case <-cancel:
return fuse.EINTR
default:
}
if nAttempts >= 2 {
// Random delay to partially mitigate thundering herd on contended lock.
time.Sleep(time.Duration(mathrand.Int()%5_000) * time.Millisecond)
}
err := fs.fs.AwaitExclusiveLockRelease(cancel, in.NodeId)
if err != nil {
return fs.errToFuseStatus(err)
}
nAttempts += 1
}
}
func (fs *FuseFs) StatFs(cancel <-chan struct{}, in *fuse.InHeader, out *fuse.StatfsOut) fuse.Status {
stats, err := fs.fs.FsStats()
if err != nil {
return fs.errToFuseStatus(err)
}
out.Bsize = CHUNK_SIZE
out.Blocks = stats.UsedBytes / CHUNK_SIZE
out.Bfree = stats.FreeBytes / CHUNK_SIZE
out.Bavail = out.Bfree
out.NameLen = 4096
return fuse.OK
}
func (fs *FuseFs) Access(cancel <-chan struct{}, input *fuse.AccessIn) fuse.Status {
return fuse.ENOSYS
}
func (fs *FuseFs) CopyFileRange(cancel <-chan struct{}, input *fuse.CopyFileRangeIn) (uint32, fuse.Status) {
return 0, fuse.ENOSYS
}
func (fs *FuseFs) Fallocate(cancel <-chan struct{}, in *fuse.FallocateIn) fuse.Status {
return fuse.ENOSYS
}
func (fs *FuseFs) Mknod(cancel <-chan struct{}, input *fuse.MknodIn, name string, out *fuse.EntryOut) fuse.Status {
return fuse.ENOSYS
}
func (fs *FuseFs) SetDebug(dbg bool) {
}
func (fs *FuseFs) String() string {
return os.Args[0]
}