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
|
// Copyright 2019 The gVisor 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 gofer
import (
"fmt"
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/atomicbitops"
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/errors/linuxerr"
"gvisor.dev/gvisor/pkg/fdnotifier"
"gvisor.dev/gvisor/pkg/hostarch"
"gvisor.dev/gvisor/pkg/metric"
"gvisor.dev/gvisor/pkg/safemem"
"gvisor.dev/gvisor/pkg/sentry/fsmetric"
"gvisor.dev/gvisor/pkg/sentry/fsutil"
"gvisor.dev/gvisor/pkg/sentry/memmap"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
)
// specialFileFD implements vfs.FileDescriptionImpl for pipes, sockets, device
// special files, and (when filesystemOptions.regularFilesUseSpecialFileFD is
// in effect) regular files. specialFileFD differs from regularFileFD by using
// per-FD handles instead of shared per-dentry handles, and never buffering I/O.
//
// +stateify savable
type specialFileFD struct {
fileDescription
specialFDEntry
// releaseMu synchronizes the closing of fd.handle with fd.sync(). It's safe
// to access fd.handle without locking for operations that require a ref to
// be held by the caller, e.g. vfs.FileDescriptionImpl implementations.
releaseMu sync.RWMutex `state:"nosave"`
// handle is used for file I/O. handle is immutable.
handle handle `state:"nosave"`
// isRegularFile is true if this FD represents a regular file which is only
// possible when filesystemOptions.regularFilesUseSpecialFileFD is in
// effect. isRegularFile is immutable.
isRegularFile bool
// seekable is true if this file description represents a file for which
// file offset is significant, i.e. a regular file, character device or
// block device. seekable is immutable.
seekable bool
// haveQueue is true if this file description represents a file for which
// queue may send I/O readiness events. haveQueue is immutable.
haveQueue bool `state:"nosave"`
queue waiter.Queue
// If seekable is true, off is the file offset. off is protected by mu.
mu sync.Mutex `state:"nosave"`
off int64
// If haveBuf is non-zero, this FD represents a pipe, and buf contains data
// read from the pipe from previous calls to specialFileFD.savePipeData().
// haveBuf and buf are protected by bufMu.
bufMu sync.Mutex `state:"nosave"`
haveBuf atomicbitops.Uint32
buf []byte
// If handle.fd >= 0, hostFileMapper caches mappings of handle.fd, and
// hostFileMapperInitOnce is used to initialize it on first use.
hostFileMapperInitOnce sync.Once `state:"nosave"`
hostFileMapper fsutil.HostFileMapper
// If handle.fd >= 0, fileRefs counts references on memmap.File offsets.
// fileRefs is protected by fileRefsMu.
fileRefsMu sync.Mutex `state:"nosave"`
fileRefs fsutil.FrameRefSet
}
func newSpecialFileFD(h handle, mnt *vfs.Mount, d *dentry, flags uint32) (*specialFileFD, error) {
ftype := d.fileType()
seekable := ftype == linux.S_IFREG || ftype == linux.S_IFCHR || ftype == linux.S_IFBLK
haveQueue := (ftype == linux.S_IFIFO || ftype == linux.S_IFSOCK || ftype == linux.S_IFCHR) && h.fd >= 0
fd := &specialFileFD{
handle: h,
isRegularFile: ftype == linux.S_IFREG,
seekable: seekable,
haveQueue: haveQueue,
}
fd.LockFD.Init(&d.locks)
if haveQueue {
if err := fdnotifier.AddFD(h.fd, &fd.queue); err != nil {
return nil, err
}
}
if err := fd.vfsfd.Init(fd, flags, mnt, &d.vfsd, &vfs.FileDescriptionOptions{
AllowDirectIO: true,
DenyPRead: !seekable,
DenyPWrite: !seekable,
}); err != nil {
if haveQueue {
fdnotifier.RemoveFD(h.fd)
}
return nil, err
}
d.fs.syncMu.Lock()
d.fs.specialFileFDs.PushBack(fd)
d.fs.syncMu.Unlock()
if fd.vfsfd.IsWritable() && (d.mode.Load()&0111 != 0) {
metric.SuspiciousOperationsMetric.Increment("opened_write_execute_file")
}
if h.fd >= 0 {
fsmetric.GoferOpensHost.Increment()
} else {
fsmetric.GoferOpens9P.Increment()
}
return fd, nil
}
// Release implements vfs.FileDescriptionImpl.Release.
func (fd *specialFileFD) Release(ctx context.Context) {
if fd.haveQueue {
fdnotifier.RemoveFD(fd.handle.fd)
}
fd.releaseMu.Lock()
fd.handle.close(ctx)
fd.releaseMu.Unlock()
fs := fd.vfsfd.Mount().Filesystem().Impl().(*filesystem)
fs.syncMu.Lock()
fs.specialFileFDs.Remove(fd)
fs.syncMu.Unlock()
}
// OnClose implements vfs.FileDescriptionImpl.OnClose.
func (fd *specialFileFD) OnClose(ctx context.Context) error {
if !fd.vfsfd.IsWritable() {
return nil
}
return fd.handle.fdLisa.Flush(ctx)
}
// Readiness implements waiter.Waitable.Readiness.
func (fd *specialFileFD) Readiness(mask waiter.EventMask) waiter.EventMask {
if fd.haveQueue {
return fdnotifier.NonBlockingPoll(fd.handle.fd, mask)
}
return fd.fileDescription.Readiness(mask)
}
// EventRegister implements waiter.Waitable.EventRegister.
func (fd *specialFileFD) EventRegister(e *waiter.Entry) error {
if fd.haveQueue {
fd.queue.EventRegister(e)
if err := fdnotifier.UpdateFD(fd.handle.fd); err != nil {
fd.queue.EventUnregister(e)
return err
}
return nil
}
return fd.fileDescription.EventRegister(e)
}
// EventUnregister implements waiter.Waitable.EventUnregister.
func (fd *specialFileFD) EventUnregister(e *waiter.Entry) {
if fd.haveQueue {
fd.queue.EventUnregister(e)
if err := fdnotifier.UpdateFD(fd.handle.fd); err != nil {
panic(fmt.Sprint("UpdateFD:", err))
}
return
}
fd.fileDescription.EventUnregister(e)
}
// Epollable implements FileDescriptionImpl.Epollable.
func (fd *specialFileFD) Epollable() bool {
if fd.haveQueue {
return true
}
return fd.fileDescription.Epollable()
}
func (fd *specialFileFD) Allocate(ctx context.Context, mode, offset, length uint64) error {
if fd.isRegularFile {
d := fd.dentry()
return d.doAllocate(ctx, offset, length, func() error {
return fd.handle.fdLisa.Allocate(ctx, mode, offset, length)
})
}
return fd.FileDescriptionDefaultImpl.Allocate(ctx, mode, offset, length)
}
// PRead implements vfs.FileDescriptionImpl.PRead.
func (fd *specialFileFD) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error) {
start := fsmetric.StartReadWait()
defer func() {
if fd.handle.fd >= 0 {
fsmetric.GoferReadsHost.Increment()
fsmetric.FinishReadWait(fsmetric.GoferReadWaitHost, start)
} else {
fsmetric.GoferReads9P.Increment()
fsmetric.FinishReadWait(fsmetric.GoferReadWait9P, start)
}
}()
if fd.seekable && offset < 0 {
return 0, linuxerr.EINVAL
}
// Check that flags are supported.
//
// TODO(gvisor.dev/issue/2601): Support select preadv2 flags.
if opts.Flags&^linux.RWF_HIPRI != 0 {
return 0, linuxerr.EOPNOTSUPP
}
if d := fd.dentry(); d.cachedMetadataAuthoritative() {
d.touchAtime(fd.vfsfd.Mount())
}
bufN := int64(0)
if fd.haveBuf.Load() != 0 {
var err error
fd.bufMu.Lock()
if len(fd.buf) != 0 {
var n int
n, err = dst.CopyOut(ctx, fd.buf)
dst = dst.DropFirst(n)
fd.buf = fd.buf[n:]
if len(fd.buf) == 0 {
fd.haveBuf.Store(0)
fd.buf = nil
}
bufN = int64(n)
if offset >= 0 {
offset += bufN
}
}
fd.bufMu.Unlock()
if err != nil {
return bufN, err
}
}
rw := getHandleReadWriter(ctx, &fd.handle, offset)
n, err := dst.CopyOutFrom(ctx, rw)
putHandleReadWriter(rw)
if linuxerr.Equals(linuxerr.EAGAIN, err) {
err = linuxerr.ErrWouldBlock
}
return bufN + n, err
}
// Read implements vfs.FileDescriptionImpl.Read.
func (fd *specialFileFD) Read(ctx context.Context, dst usermem.IOSequence, opts vfs.ReadOptions) (int64, error) {
if !fd.seekable {
return fd.PRead(ctx, dst, -1, opts)
}
fd.mu.Lock()
n, err := fd.PRead(ctx, dst, fd.off, opts)
fd.off += n
fd.mu.Unlock()
return n, err
}
// PWrite implements vfs.FileDescriptionImpl.PWrite.
func (fd *specialFileFD) PWrite(ctx context.Context, src usermem.IOSequence, offset int64, opts vfs.WriteOptions) (int64, error) {
n, _, err := fd.pwrite(ctx, src, offset, opts)
return n, err
}
// pwrite returns the number of bytes written, final offset, error. The final
// offset should be ignored by PWrite.
func (fd *specialFileFD) pwrite(ctx context.Context, src usermem.IOSequence, offset int64, opts vfs.WriteOptions) (written, finalOff int64, err error) {
if fd.seekable && offset < 0 {
return 0, offset, linuxerr.EINVAL
}
// Check that flags are supported.
//
// TODO(gvisor.dev/issue/2601): Support select pwritev2 flags.
if opts.Flags&^linux.RWF_HIPRI != 0 {
return 0, offset, linuxerr.EOPNOTSUPP
}
d := fd.dentry()
if fd.isRegularFile {
// If the regular file fd was opened with O_APPEND, make sure the file
// size is updated. There is a possible race here if size is modified
// externally after metadata cache is updated.
if fd.vfsfd.StatusFlags()&linux.O_APPEND != 0 && !d.cachedMetadataAuthoritative() {
if err := d.updateMetadata(ctx, nil); err != nil {
return 0, offset, err
}
}
// We need to hold the metadataMu *while* writing to a regular file.
d.metadataMu.Lock()
defer d.metadataMu.Unlock()
// Set offset to file size if the regular file was opened with O_APPEND.
if fd.vfsfd.StatusFlags()&linux.O_APPEND != 0 {
// Holding d.metadataMu is sufficient for reading d.size.
offset = int64(d.size.RacyLoad())
}
limit, err := vfs.CheckLimit(ctx, offset, src.NumBytes())
if err != nil {
return 0, offset, err
}
src = src.TakeFirst64(limit)
}
if d.cachedMetadataAuthoritative() {
if fd.isRegularFile {
d.touchCMtimeLocked()
} else {
d.touchCMtime()
}
}
// handleReadWriter always writes to the remote file. So O_DIRECT is
// effectively always set. Invalidate pages in d.mappings that have been
// written to.
pgstart := hostarch.PageRoundDown(uint64(offset))
pgend, ok := hostarch.PageRoundUp(uint64(offset + src.NumBytes()))
if !ok {
return 0, offset, linuxerr.EINVAL
}
mr := memmap.MappableRange{pgstart, pgend}
d.mapsMu.Lock()
d.mappings.Invalidate(mr, memmap.InvalidateOpts{})
d.mapsMu.Unlock()
rw := getHandleReadWriter(ctx, &fd.handle, offset)
n, err := src.CopyInTo(ctx, rw)
putHandleReadWriter(rw)
if n > 0 && fd.vfsfd.StatusFlags()&(linux.O_DSYNC|linux.O_SYNC) != 0 {
// Note that if syncing the remote file fails, then we can't guarantee that
// any data was actually written with the semantics of O_DSYNC or
// O_SYNC, so we return zero bytes written. Compare Linux's
// mm/filemap.c:generic_file_write_iter() =>
// include/linux/fs.h:generic_write_sync().
if err := fd.sync(ctx, false /* forFilesystemSync */); err != nil {
return 0, offset, err
}
}
if linuxerr.Equals(linuxerr.EAGAIN, err) {
err = linuxerr.ErrWouldBlock
}
// Update offset if the offset is valid.
if offset >= 0 {
offset += n
}
// Update file size for regular files.
if fd.isRegularFile {
// d.metadataMu is already locked at this point.
if uint64(offset) > d.size.RacyLoad() {
d.dataMu.Lock()
defer d.dataMu.Unlock()
d.size.Store(uint64(offset))
}
}
return int64(n), offset, err
}
// Write implements vfs.FileDescriptionImpl.Write.
func (fd *specialFileFD) Write(ctx context.Context, src usermem.IOSequence, opts vfs.WriteOptions) (int64, error) {
if !fd.seekable {
return fd.PWrite(ctx, src, -1, opts)
}
fd.mu.Lock()
n, off, err := fd.pwrite(ctx, src, fd.off, opts)
fd.off = off
fd.mu.Unlock()
return n, err
}
// Seek implements vfs.FileDescriptionImpl.Seek.
func (fd *specialFileFD) Seek(ctx context.Context, offset int64, whence int32) (int64, error) {
if !fd.seekable {
return 0, linuxerr.ESPIPE
}
fd.mu.Lock()
defer fd.mu.Unlock()
newOffset, err := regularFileSeekLocked(ctx, fd.dentry(), fd.off, offset, whence)
if err != nil {
return 0, err
}
fd.off = newOffset
return newOffset, nil
}
// Sync implements vfs.FileDescriptionImpl.Sync.
func (fd *specialFileFD) Sync(ctx context.Context) error {
return fd.sync(ctx, false /* forFilesystemSync */)
}
func (fd *specialFileFD) sync(ctx context.Context, forFilesystemSync bool) error {
// Locks to ensure it didn't race with fd.Release().
fd.releaseMu.RLock()
defer fd.releaseMu.RUnlock()
if !fd.handle.isOpen() {
return nil
}
err := func() error {
// If we have a host FD, fsyncing it is likely to be faster than an fsync
// RPC.
if fd.handle.fd >= 0 {
ctx.UninterruptibleSleepStart(false)
err := unix.Fsync(int(fd.handle.fd))
ctx.UninterruptibleSleepFinish(false)
return err
}
return fd.handle.fdLisa.Sync(ctx)
}()
if err != nil {
if !forFilesystemSync {
return err
}
// Only return err if we can reasonably have expected sync to succeed
// (fd represents a regular file that was opened for writing).
if fd.isRegularFile && fd.vfsfd.IsWritable() {
return err
}
ctx.Debugf("gofer.specialFileFD.sync: syncing non-writable or non-regular-file FD failed: %v", err)
}
return nil
}
// ConfigureMMap implements vfs.FileDescriptionImpl.ConfigureMMap.
func (fd *specialFileFD) ConfigureMMap(ctx context.Context, opts *memmap.MMapOpts) error {
if fd.handle.fd < 0 || fd.filesystem().opts.forcePageCache {
return linuxerr.ENODEV
}
// After this point, fd may be used as a memmap.Mappable and memmap.File.
fd.hostFileMapperInitOnce.Do(fd.hostFileMapper.Init)
return vfs.GenericConfigureMMap(&fd.vfsfd, fd, opts)
}
// AddMapping implements memmap.Mappable.AddMapping.
func (fd *specialFileFD) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar hostarch.AddrRange, offset uint64, writable bool) error {
d := fd.dentry()
d.mapsMu.Lock()
defer d.mapsMu.Unlock()
d.mappings.AddMapping(ms, ar, offset, writable)
fd.hostFileMapper.IncRefOn(memmap.MappableRange{offset, offset + uint64(ar.Length())})
return nil
}
// RemoveMapping implements memmap.Mappable.RemoveMapping.
func (fd *specialFileFD) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar hostarch.AddrRange, offset uint64, writable bool) {
d := fd.dentry()
d.mapsMu.Lock()
defer d.mapsMu.Unlock()
d.mappings.RemoveMapping(ms, ar, offset, writable)
fd.hostFileMapper.DecRefOn(memmap.MappableRange{offset, offset + uint64(ar.Length())})
}
// CopyMapping implements memmap.Mappable.CopyMapping.
func (fd *specialFileFD) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR hostarch.AddrRange, offset uint64, writable bool) error {
return fd.AddMapping(ctx, ms, dstAR, offset, writable)
}
// Translate implements memmap.Mappable.Translate.
func (fd *specialFileFD) Translate(ctx context.Context, required, optional memmap.MappableRange, at hostarch.AccessType) ([]memmap.Translation, error) {
mr := optional
if fd.filesystem().opts.limitHostFDTranslation {
mr = maxFillRange(required, optional)
}
return []memmap.Translation{
{
Source: mr,
File: fd,
Offset: mr.Start,
Perms: hostarch.AnyAccess,
},
}, nil
}
// InvalidateUnsavable implements memmap.Mappable.InvalidateUnsavable.
func (fd *specialFileFD) InvalidateUnsavable(ctx context.Context) error {
return nil
}
// IncRef implements memmap.File.IncRef.
func (fd *specialFileFD) IncRef(fr memmap.FileRange) {
fd.fileRefsMu.Lock()
defer fd.fileRefsMu.Unlock()
fd.fileRefs.IncRefAndAccount(fr)
}
// DecRef implements memmap.File.DecRef.
func (fd *specialFileFD) DecRef(fr memmap.FileRange) {
fd.fileRefsMu.Lock()
defer fd.fileRefsMu.Unlock()
fd.fileRefs.DecRefAndAccount(fr)
}
// MapInternal implements memmap.File.MapInternal.
func (fd *specialFileFD) MapInternal(fr memmap.FileRange, at hostarch.AccessType) (safemem.BlockSeq, error) {
fd.requireHostFD()
return fd.hostFileMapper.MapInternal(fr, int(fd.handle.fd), at.Write)
}
// FD implements memmap.File.FD.
func (fd *specialFileFD) FD() int {
fd.requireHostFD()
return int(fd.handle.fd)
}
func (fd *specialFileFD) requireHostFD() {
if fd.handle.fd < 0 {
// This is possible if fd was successfully mmapped before saving, then
// was restored without a host FD. This is unrecoverable: without a
// host FD, we can't mmap this file post-restore.
panic("gofer.specialFileFD can no longer be memory-mapped without a host FD")
}
}
|