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
|
// 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 (
"sync"
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/fdnotifier"
"gvisor.dev/gvisor/pkg/safemem"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
)
// specialFileFD implements vfs.FileDescriptionImpl for pipes, sockets, device
// special files, and (when filesystemOptions.specialRegularFiles 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.
type specialFileFD struct {
fileDescription
// handle is used for file I/O. handle is immutable.
handle handle
// seekable is true if this file description represents a file for which
// file offset is significant, i.e. a regular file. seekable is immutable.
seekable bool
// mayBlock is true if this file description represents a file for which
// queue may send I/O readiness events. mayBlock is immutable.
mayBlock bool
queue waiter.Queue
// If seekable is true, off is the file offset. off is protected by mu.
mu sync.Mutex
off int64
}
func newSpecialFileFD(h handle, mnt *vfs.Mount, d *dentry, locks *vfs.FileLocks, flags uint32) (*specialFileFD, error) {
ftype := d.fileType()
seekable := ftype == linux.S_IFREG
mayBlock := ftype == linux.S_IFIFO || ftype == linux.S_IFSOCK
fd := &specialFileFD{
handle: h,
seekable: seekable,
mayBlock: mayBlock,
}
fd.LockFD.Init(locks)
if mayBlock && h.fd >= 0 {
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{
DenyPRead: !seekable,
DenyPWrite: !seekable,
}); err != nil {
if mayBlock && h.fd >= 0 {
fdnotifier.RemoveFD(h.fd)
}
return nil, err
}
return fd, nil
}
// Release implements vfs.FileDescriptionImpl.Release.
func (fd *specialFileFD) Release() {
if fd.mayBlock && fd.handle.fd >= 0 {
fdnotifier.RemoveFD(fd.handle.fd)
}
fd.handle.close(context.Background())
fs := fd.vfsfd.Mount().Filesystem().Impl().(*filesystem)
fs.syncMu.Lock()
delete(fs.specialFileFDs, 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.file.flush(ctx)
}
// Readiness implements waiter.Waitable.Readiness.
func (fd *specialFileFD) Readiness(mask waiter.EventMask) waiter.EventMask {
if fd.mayBlock {
return fdnotifier.NonBlockingPoll(fd.handle.fd, mask)
}
return fd.fileDescription.Readiness(mask)
}
// EventRegister implements waiter.Waitable.EventRegister.
func (fd *specialFileFD) EventRegister(e *waiter.Entry, mask waiter.EventMask) {
if fd.mayBlock {
fd.queue.EventRegister(e, mask)
return
}
fd.fileDescription.EventRegister(e, mask)
}
// EventUnregister implements waiter.Waitable.EventUnregister.
func (fd *specialFileFD) EventUnregister(e *waiter.Entry) {
if fd.mayBlock {
fd.queue.EventUnregister(e)
return
}
fd.fileDescription.EventUnregister(e)
}
// PRead implements vfs.FileDescriptionImpl.PRead.
func (fd *specialFileFD) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error) {
if fd.seekable && offset < 0 {
return 0, syserror.EINVAL
}
// Check that flags are supported. Silently ignore RWF_HIPRI.
if opts.Flags&^linux.RWF_HIPRI != 0 {
return 0, syserror.EOPNOTSUPP
}
// Going through dst.CopyOutFrom() holds MM locks around file operations of
// unknown duration. For regularFileFD, doing so is necessary to support
// mmap due to lock ordering; MM locks precede dentry.dataMu. That doesn't
// hold here since specialFileFD doesn't client-cache data. Just buffer the
// read instead.
if d := fd.dentry(); d.fs.opts.interop != InteropModeShared {
d.touchAtime(fd.vfsfd.Mount())
}
buf := make([]byte, dst.NumBytes())
n, err := fd.handle.readToBlocksAt(ctx, safemem.BlockSeqOf(safemem.BlockFromSafeSlice(buf)), uint64(offset))
if err == syserror.EAGAIN {
err = syserror.ErrWouldBlock
}
if n == 0 {
return 0, err
}
if cp, cperr := dst.CopyOut(ctx, buf[:n]); cperr != nil {
return int64(cp), cperr
}
return int64(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) {
if fd.seekable && offset < 0 {
return 0, syserror.EINVAL
}
// Check that flags are supported. Silently ignore RWF_HIPRI.
if opts.Flags&^linux.RWF_HIPRI != 0 {
return 0, syserror.EOPNOTSUPP
}
if fd.seekable {
limit, err := vfs.CheckLimit(ctx, offset, src.NumBytes())
if err != nil {
return 0, err
}
src = src.TakeFirst64(limit)
}
// Do a buffered write. See rationale in PRead.
if d := fd.dentry(); d.fs.opts.interop != InteropModeShared {
d.touchCMtime()
}
buf := make([]byte, src.NumBytes())
// Don't do partial writes if we get a partial read from src.
if _, err := src.CopyIn(ctx, buf); err != nil {
return 0, err
}
n, err := fd.handle.writeFromBlocksAt(ctx, safemem.BlockSeqOf(safemem.BlockFromSafeSlice(buf)), uint64(offset))
if err == syserror.EAGAIN {
err = syserror.ErrWouldBlock
}
return int64(n), 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, err := fd.PWrite(ctx, src, fd.off, opts)
fd.off += n
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, syserror.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 {
if !fd.vfsfd.IsWritable() {
return nil
}
return fd.handle.sync(ctx)
}
|
