summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry/syscalls/linux/vfs2/splice.go
blob: 16f59fce9c36647104398db7c4baec127e035f73 (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
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
// Copyright 2020 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 vfs2

import (
	"io"

	"gvisor.dev/gvisor/pkg/abi/linux"
	"gvisor.dev/gvisor/pkg/sentry/arch"
	"gvisor.dev/gvisor/pkg/sentry/kernel"
	"gvisor.dev/gvisor/pkg/sentry/kernel/pipe"
	"gvisor.dev/gvisor/pkg/sentry/vfs"
	"gvisor.dev/gvisor/pkg/syserror"
	"gvisor.dev/gvisor/pkg/usermem"
	"gvisor.dev/gvisor/pkg/waiter"
)

// Splice implements Linux syscall splice(2).
func Splice(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
	inFD := args[0].Int()
	inOffsetPtr := args[1].Pointer()
	outFD := args[2].Int()
	outOffsetPtr := args[3].Pointer()
	count := int64(args[4].SizeT())
	flags := args[5].Int()

	if count == 0 {
		return 0, nil, nil
	}
	if count > int64(kernel.MAX_RW_COUNT) {
		count = int64(kernel.MAX_RW_COUNT)
	}

	// Check for invalid flags.
	if flags&^(linux.SPLICE_F_MOVE|linux.SPLICE_F_NONBLOCK|linux.SPLICE_F_MORE|linux.SPLICE_F_GIFT) != 0 {
		return 0, nil, syserror.EINVAL
	}

	// Get file descriptions.
	inFile := t.GetFileVFS2(inFD)
	if inFile == nil {
		return 0, nil, syserror.EBADF
	}
	defer inFile.DecRef(t)
	outFile := t.GetFileVFS2(outFD)
	if outFile == nil {
		return 0, nil, syserror.EBADF
	}
	defer outFile.DecRef(t)

	// Check that both files support the required directionality.
	if !inFile.IsReadable() || !outFile.IsWritable() {
		return 0, nil, syserror.EBADF
	}

	// The operation is non-blocking if anything is non-blocking.
	//
	// N.B. This is a rather simplistic heuristic that avoids some
	// poor edge case behavior since the exact semantics here are
	// underspecified and vary between versions of Linux itself.
	nonBlock := ((inFile.StatusFlags()|outFile.StatusFlags())&linux.O_NONBLOCK != 0) || (flags&linux.SPLICE_F_NONBLOCK != 0)

	// At least one file description must represent a pipe.
	inPipeFD, inIsPipe := inFile.Impl().(*pipe.VFSPipeFD)
	outPipeFD, outIsPipe := outFile.Impl().(*pipe.VFSPipeFD)
	if !inIsPipe && !outIsPipe {
		return 0, nil, syserror.EINVAL
	}

	// Copy in offsets.
	inOffset := int64(-1)
	if inOffsetPtr != 0 {
		if inIsPipe {
			return 0, nil, syserror.ESPIPE
		}
		if inFile.Options().DenyPRead {
			return 0, nil, syserror.EINVAL
		}
		if _, err := t.CopyIn(inOffsetPtr, &inOffset); err != nil {
			return 0, nil, err
		}
		if inOffset < 0 {
			return 0, nil, syserror.EINVAL
		}
	}
	outOffset := int64(-1)
	if outOffsetPtr != 0 {
		if outIsPipe {
			return 0, nil, syserror.ESPIPE
		}
		if outFile.Options().DenyPWrite {
			return 0, nil, syserror.EINVAL
		}
		if _, err := t.CopyIn(outOffsetPtr, &outOffset); err != nil {
			return 0, nil, err
		}
		if outOffset < 0 {
			return 0, nil, syserror.EINVAL
		}
	}

	// Move data.
	var (
		n   int64
		err error
	)
	dw := dualWaiter{
		inFile:  inFile,
		outFile: outFile,
	}
	defer dw.destroy()
	for {
		// If both input and output are pipes, delegate to the pipe
		// implementation. Otherwise, exactly one end is a pipe, which
		// we ensure is consistently ordered after the non-pipe FD's
		// locks by passing the pipe FD as usermem.IO to the non-pipe
		// end.
		switch {
		case inIsPipe && outIsPipe:
			n, err = pipe.Splice(t, outPipeFD, inPipeFD, count)
		case inIsPipe:
			if outOffset != -1 {
				n, err = outFile.PWrite(t, inPipeFD.IOSequence(count), outOffset, vfs.WriteOptions{})
				outOffset += n
			} else {
				n, err = outFile.Write(t, inPipeFD.IOSequence(count), vfs.WriteOptions{})
			}
		case outIsPipe:
			if inOffset != -1 {
				n, err = inFile.PRead(t, outPipeFD.IOSequence(count), inOffset, vfs.ReadOptions{})
				inOffset += n
			} else {
				n, err = inFile.Read(t, outPipeFD.IOSequence(count), vfs.ReadOptions{})
			}
		default:
			panic("not possible")
		}

		if n != 0 || err != syserror.ErrWouldBlock || nonBlock {
			break
		}
		if err = dw.waitForBoth(t); err != nil {
			break
		}
	}

	// Copy updated offsets out.
	if inOffsetPtr != 0 {
		if _, err := t.CopyOut(inOffsetPtr, &inOffset); err != nil {
			return 0, nil, err
		}
	}
	if outOffsetPtr != 0 {
		if _, err := t.CopyOut(outOffsetPtr, &outOffset); err != nil {
			return 0, nil, err
		}
	}

	if n == 0 {
		return 0, nil, err
	}

	// On Linux, inotify behavior is not very consistent with splice(2). We try
	// our best to emulate Linux for very basic calls to splice, where for some
	// reason, events are generated for output files, but not input files.
	outFile.Dentry().InotifyWithParent(t, linux.IN_MODIFY, 0, vfs.PathEvent)
	return uintptr(n), nil, nil
}

// Tee implements Linux syscall tee(2).
func Tee(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
	inFD := args[0].Int()
	outFD := args[1].Int()
	count := int64(args[2].SizeT())
	flags := args[3].Int()

	if count == 0 {
		return 0, nil, nil
	}
	if count > int64(kernel.MAX_RW_COUNT) {
		count = int64(kernel.MAX_RW_COUNT)
	}

	// Check for invalid flags.
	if flags&^(linux.SPLICE_F_MOVE|linux.SPLICE_F_NONBLOCK|linux.SPLICE_F_MORE|linux.SPLICE_F_GIFT) != 0 {
		return 0, nil, syserror.EINVAL
	}

	// Get file descriptions.
	inFile := t.GetFileVFS2(inFD)
	if inFile == nil {
		return 0, nil, syserror.EBADF
	}
	defer inFile.DecRef(t)
	outFile := t.GetFileVFS2(outFD)
	if outFile == nil {
		return 0, nil, syserror.EBADF
	}
	defer outFile.DecRef(t)

	// Check that both files support the required directionality.
	if !inFile.IsReadable() || !outFile.IsWritable() {
		return 0, nil, syserror.EBADF
	}

	// The operation is non-blocking if anything is non-blocking.
	//
	// N.B. This is a rather simplistic heuristic that avoids some
	// poor edge case behavior since the exact semantics here are
	// underspecified and vary between versions of Linux itself.
	nonBlock := ((inFile.StatusFlags()|outFile.StatusFlags())&linux.O_NONBLOCK != 0) || (flags&linux.SPLICE_F_NONBLOCK != 0)

	// Both file descriptions must represent pipes.
	inPipeFD, inIsPipe := inFile.Impl().(*pipe.VFSPipeFD)
	outPipeFD, outIsPipe := outFile.Impl().(*pipe.VFSPipeFD)
	if !inIsPipe || !outIsPipe {
		return 0, nil, syserror.EINVAL
	}

	// Copy data.
	var (
		n   int64
		err error
	)
	dw := dualWaiter{
		inFile:  inFile,
		outFile: outFile,
	}
	defer dw.destroy()
	for {
		n, err = pipe.Tee(t, outPipeFD, inPipeFD, count)
		if n != 0 || err != syserror.ErrWouldBlock || nonBlock {
			break
		}
		if err = dw.waitForBoth(t); err != nil {
			break
		}
	}
	if n == 0 {
		return 0, nil, err
	}
	outFile.Dentry().InotifyWithParent(t, linux.IN_MODIFY, 0, vfs.PathEvent)
	return uintptr(n), nil, nil
}

// Sendfile implements linux system call sendfile(2).
func Sendfile(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
	outFD := args[0].Int()
	inFD := args[1].Int()
	offsetAddr := args[2].Pointer()
	count := int64(args[3].SizeT())

	inFile := t.GetFileVFS2(inFD)
	if inFile == nil {
		return 0, nil, syserror.EBADF
	}
	defer inFile.DecRef(t)
	if !inFile.IsReadable() {
		return 0, nil, syserror.EBADF
	}

	outFile := t.GetFileVFS2(outFD)
	if outFile == nil {
		return 0, nil, syserror.EBADF
	}
	defer outFile.DecRef(t)
	if !outFile.IsWritable() {
		return 0, nil, syserror.EBADF
	}

	// Verify that the outFile Append flag is not set.
	if outFile.StatusFlags()&linux.O_APPEND != 0 {
		return 0, nil, syserror.EINVAL
	}

	// Verify that inFile is a regular file or block device. This is a
	// requirement; the same check appears in Linux
	// (fs/splice.c:splice_direct_to_actor).
	if stat, err := inFile.Stat(t, vfs.StatOptions{Mask: linux.STATX_TYPE}); err != nil {
		return 0, nil, err
	} else if stat.Mask&linux.STATX_TYPE == 0 ||
		(stat.Mode&linux.S_IFMT != linux.S_IFREG && stat.Mode&linux.S_IFMT != linux.S_IFBLK) {
		return 0, nil, syserror.EINVAL
	}

	// Copy offset if it exists.
	offset := int64(-1)
	if offsetAddr != 0 {
		if inFile.Options().DenyPRead {
			return 0, nil, syserror.ESPIPE
		}
		if _, err := t.CopyIn(offsetAddr, &offset); err != nil {
			return 0, nil, err
		}
		if offset < 0 {
			return 0, nil, syserror.EINVAL
		}
		if offset+count < 0 {
			return 0, nil, syserror.EINVAL
		}
	}

	// Validate count. This must come after offset checks.
	if count < 0 {
		return 0, nil, syserror.EINVAL
	}
	if count == 0 {
		return 0, nil, nil
	}
	if count > int64(kernel.MAX_RW_COUNT) {
		count = int64(kernel.MAX_RW_COUNT)
	}

	// Copy data.
	var (
		n   int64
		err error
	)
	dw := dualWaiter{
		inFile:  inFile,
		outFile: outFile,
	}
	defer dw.destroy()
	outPipeFD, outIsPipe := outFile.Impl().(*pipe.VFSPipeFD)
	// Reading from input file should never block, since it is regular or
	// block device. We only need to check if writing to the output file
	// can block.
	nonBlock := outFile.StatusFlags()&linux.O_NONBLOCK != 0
	if outIsPipe {
		for n < count {
			var spliceN int64
			if offset != -1 {
				spliceN, err = inFile.PRead(t, outPipeFD.IOSequence(count), offset, vfs.ReadOptions{})
				offset += spliceN
			} else {
				spliceN, err = inFile.Read(t, outPipeFD.IOSequence(count), vfs.ReadOptions{})
			}
			n += spliceN
			if err == syserror.ErrWouldBlock && !nonBlock {
				err = dw.waitForBoth(t)
			}
			if err != nil {
				break
			}
		}
	} else {
		// Read inFile to buffer, then write the contents to outFile.
		buf := make([]byte, count)
		for n < count {
			var readN int64
			if offset != -1 {
				readN, err = inFile.PRead(t, usermem.BytesIOSequence(buf), offset, vfs.ReadOptions{})
				offset += readN
			} else {
				readN, err = inFile.Read(t, usermem.BytesIOSequence(buf), vfs.ReadOptions{})
			}
			if readN == 0 && err == io.EOF {
				// We reached the end of the file. Eat the
				// error and exit the loop.
				err = nil
				break
			}
			n += readN
			if err != nil {
				break
			}

			// Write all of the bytes that we read. This may need
			// multiple write calls to complete.
			wbuf := buf[:n]
			for len(wbuf) > 0 {
				var writeN int64
				writeN, err = outFile.Write(t, usermem.BytesIOSequence(wbuf), vfs.WriteOptions{})
				wbuf = wbuf[writeN:]
				if err == syserror.ErrWouldBlock && !nonBlock {
					err = dw.waitForOut(t)
				}
				if err != nil {
					// We didn't complete the write. Only
					// report the bytes that were actually
					// written, and rewind the offset.
					notWritten := int64(len(wbuf))
					n -= notWritten
					if offset != -1 {
						offset -= notWritten
					}
					break
				}
			}
			if err == syserror.ErrWouldBlock && !nonBlock {
				err = dw.waitForBoth(t)
			}
			if err != nil {
				break
			}
		}
	}

	if offsetAddr != 0 {
		// Copy out the new offset.
		if _, err := t.CopyOut(offsetAddr, offset); err != nil {
			return 0, nil, err
		}
	}

	if n == 0 {
		return 0, nil, err
	}

	inFile.Dentry().InotifyWithParent(t, linux.IN_ACCESS, 0, vfs.PathEvent)
	outFile.Dentry().InotifyWithParent(t, linux.IN_MODIFY, 0, vfs.PathEvent)
	return uintptr(n), nil, nil
}

// dualWaiter is used to wait on one or both vfs.FileDescriptions. It is not
// thread-safe, and does not take a reference on the vfs.FileDescriptions.
//
// Users must call destroy() when finished.
type dualWaiter struct {
	inFile  *vfs.FileDescription
	outFile *vfs.FileDescription

	inW   waiter.Entry
	inCh  chan struct{}
	outW  waiter.Entry
	outCh chan struct{}
}

// waitForBoth waits for both dw.inFile and dw.outFile to be ready.
func (dw *dualWaiter) waitForBoth(t *kernel.Task) error {
	if dw.inFile.Readiness(eventMaskRead)&eventMaskRead == 0 {
		if dw.inCh == nil {
			dw.inW, dw.inCh = waiter.NewChannelEntry(nil)
			dw.inFile.EventRegister(&dw.inW, eventMaskRead)
			// We might be ready now. Try again before blocking.
			return nil
		}
		if err := t.Block(dw.inCh); err != nil {
			return err
		}
	}
	return dw.waitForOut(t)
}

// waitForOut waits for dw.outfile to be read.
func (dw *dualWaiter) waitForOut(t *kernel.Task) error {
	if dw.outFile.Readiness(eventMaskWrite)&eventMaskWrite == 0 {
		if dw.outCh == nil {
			dw.outW, dw.outCh = waiter.NewChannelEntry(nil)
			dw.outFile.EventRegister(&dw.outW, eventMaskWrite)
			// We might be ready now. Try again before blocking.
			return nil
		}
		if err := t.Block(dw.outCh); err != nil {
			return err
		}
	}
	return nil
}

// destroy cleans up resources help by dw. No more calls to wait* can occur
// after destroy is called.
func (dw *dualWaiter) destroy() {
	if dw.inCh != nil {
		dw.inFile.EventUnregister(&dw.inW)
		dw.inCh = nil
	}
	if dw.outCh != nil {
		dw.outFile.EventUnregister(&dw.outW)
		dw.outCh = nil
	}
	dw.inFile = nil
	dw.outFile = nil
}