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
|
// 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 vfs
import (
"gvisor.dev/gvisor/pkg/atomicbitops"
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/errors/linuxerr"
"gvisor.dev/gvisor/pkg/sync"
)
// Dentry represents a node in a Filesystem tree at which a file exists.
//
// Dentries are reference-counted. Unless otherwise specified, all Dentry
// methods require that a reference is held.
//
// Dentry is loosely analogous to Linux's struct dentry, but:
//
// - VFS does not associate Dentries with inodes. gVisor interacts primarily
// with filesystems that are accessed through filesystem APIs (as opposed to
// raw block devices); many such APIs support only paths and file descriptors,
// and not inodes. Furthermore, when parties outside the scope of VFS can
// rename inodes on such filesystems, VFS generally cannot "follow" the rename,
// both due to synchronization issues and because it may not even be able to
// name the destination path; this implies that it would in fact be incorrect
// for Dentries to be associated with inodes on such filesystems. Consequently,
// operations that are inode operations in Linux are FilesystemImpl methods
// and/or FileDescriptionImpl methods in gVisor's VFS. Filesystems that do
// support inodes may store appropriate state in implementations of DentryImpl.
//
// - VFS does not require that Dentries are instantiated for all paths accessed
// through VFS, only those that are tracked beyond the scope of a single
// Filesystem operation. This includes file descriptions, mount points, mount
// roots, process working directories, and chroots. This avoids instantiation
// of Dentries for operations on mutable remote filesystems that can't actually
// cache any state in the Dentry.
//
// - VFS does not track filesystem structure (i.e. relationships between
// Dentries), since both the relevant state and synchronization are
// filesystem-specific.
//
// - For the reasons above, VFS is not directly responsible for managing Dentry
// lifetime. Dentry reference counts only indicate the extent to which VFS
// requires Dentries to exist; Filesystems may elect to cache or discard
// Dentries with zero references.
//
// +stateify savable
type Dentry struct {
// mu synchronizes deletion/invalidation and mounting over this Dentry.
mu sync.Mutex `state:"nosave"`
// dead is true if the file represented by this Dentry has been deleted (by
// CommitDeleteDentry or CommitRenameReplaceDentry) or invalidated (by
// InvalidateDentry). dead is protected by mu.
dead bool
// evictable is set by the VFS layer or filesystems like overlayfs as a hint
// that this dentry will not be accessed hence forth. So filesystems that
// cache dentries locally can use this hint to release the dentry when all
// references are dropped. evictable is protected by mu.
evictable bool
// mounts is the number of Mounts for which this Dentry is Mount.point.
mounts atomicbitops.Uint32
// impl is the DentryImpl associated with this Dentry. impl is immutable.
// This should be the last field in Dentry.
impl DentryImpl
}
// Init must be called before first use of d.
func (d *Dentry) Init(impl DentryImpl) {
d.impl = impl
}
// Impl returns the DentryImpl associated with d.
func (d *Dentry) Impl() DentryImpl {
return d.impl
}
// DentryImpl contains implementation details for a Dentry. Implementations of
// DentryImpl should contain their associated Dentry by value as their first
// field.
//
// +stateify savable
type DentryImpl interface {
// IncRef increments the Dentry's reference count. A Dentry with a non-zero
// reference count must remain coherent with the state of the filesystem.
IncRef()
// TryIncRef increments the Dentry's reference count and returns true. If
// the Dentry's reference count is zero, TryIncRef may do nothing and
// return false. (It is also permitted to succeed if it can restore the
// guarantee that the Dentry is coherent with the state of the filesystem.)
//
// TryIncRef does not require that a reference is held on the Dentry.
TryIncRef() bool
// DecRef decrements the Dentry's reference count.
DecRef(ctx context.Context)
// InotifyWithParent notifies all watches on the targets represented by this
// dentry and its parent. The parent's watches are notified first, followed
// by this dentry's.
//
// InotifyWithParent automatically adds the IN_ISDIR flag for dentries
// representing directories.
//
// Note that the events may not actually propagate up to the user, depending
// on the event masks.
InotifyWithParent(ctx context.Context, events, cookie uint32, et EventType)
// Watches returns the set of inotify watches for the file corresponding to
// the Dentry. Dentries that are hard links to the same underlying file
// share the same watches.
//
// The caller does not need to hold a reference on the dentry.
Watches() *Watches
// OnZeroWatches is called whenever the number of watches on a dentry drops
// to zero. This is needed by some FilesystemImpls (e.g. gofer) to manage
// dentry lifetime.
//
// The caller does not need to hold a reference on the dentry. OnZeroWatches
// may acquire inotify locks, so to prevent deadlock, no inotify locks should
// be held by the caller.
OnZeroWatches(ctx context.Context)
}
// IncRef increments d's reference count.
func (d *Dentry) IncRef() {
d.impl.IncRef()
}
// TryIncRef increments d's reference count and returns true. If d's reference
// count is zero, TryIncRef may instead do nothing and return false.
func (d *Dentry) TryIncRef() bool {
return d.impl.TryIncRef()
}
// DecRef decrements d's reference count.
func (d *Dentry) DecRef(ctx context.Context) {
d.impl.DecRef(ctx)
}
// IsDead returns true if d has been deleted or invalidated by its owning
// filesystem.
func (d *Dentry) IsDead() bool {
d.mu.Lock()
defer d.mu.Unlock()
return d.dead
}
// IsEvictable returns true if d is evictable from filesystem dentry cache.
func (d *Dentry) IsEvictable() bool {
d.mu.Lock()
defer d.mu.Unlock()
return d.evictable
}
// MarkEvictable marks d as evictable.
func (d *Dentry) MarkEvictable() {
d.mu.Lock()
defer d.mu.Unlock()
d.evictable = true
}
func (d *Dentry) isMounted() bool {
return d.mounts.Load() != 0
}
// InotifyWithParent notifies all watches on the targets represented by d and
// its parent of events.
func (d *Dentry) InotifyWithParent(ctx context.Context, events, cookie uint32, et EventType) {
d.impl.InotifyWithParent(ctx, events, cookie, et)
}
// Watches returns the set of inotify watches associated with d.
func (d *Dentry) Watches() *Watches {
return d.impl.Watches()
}
// OnZeroWatches performs cleanup tasks whenever the number of watches on a
// dentry drops to zero.
func (d *Dentry) OnZeroWatches(ctx context.Context) {
d.impl.OnZeroWatches(ctx)
}
// The following functions are exported so that filesystem implementations can
// use them. The vfs package, and users of VFS, should not call these
// functions.
// PrepareDeleteDentry must be called before attempting to delete the file
// represented by d. If PrepareDeleteDentry succeeds, the caller must call
// AbortDeleteDentry or CommitDeleteDentry depending on the deletion's outcome.
// +checklocksacquire:d.mu
func (vfs *VirtualFilesystem) PrepareDeleteDentry(mntns *MountNamespace, d *Dentry) error {
vfs.mountMu.Lock()
if mntns.mountpoints[d] != 0 {
vfs.mountMu.Unlock()
return linuxerr.EBUSY // +checklocksforce: inconsistent return.
}
d.mu.Lock()
vfs.mountMu.Unlock()
// Return with d.mu locked to block attempts to mount over it; it will be
// unlocked by AbortDeleteDentry or CommitDeleteDentry.
return nil
}
// AbortDeleteDentry must be called after PrepareDeleteDentry if the deletion
// fails.
// +checklocksrelease:d.mu
func (vfs *VirtualFilesystem) AbortDeleteDentry(d *Dentry) {
d.mu.Unlock()
}
// CommitDeleteDentry must be called after PrepareDeleteDentry if the deletion
// succeeds.
// +checklocksrelease:d.mu
func (vfs *VirtualFilesystem) CommitDeleteDentry(ctx context.Context, d *Dentry) {
d.dead = true
d.mu.Unlock()
if d.isMounted() {
vfs.forgetDeadMountpoint(ctx, d)
}
}
// InvalidateDentry is called when d ceases to represent the file it formerly
// did for reasons outside of VFS' control (e.g. d represents the local state
// of a file on a remote filesystem on which the file has already been
// deleted).
func (vfs *VirtualFilesystem) InvalidateDentry(ctx context.Context, d *Dentry) {
d.mu.Lock()
d.dead = true
d.mu.Unlock()
if d.isMounted() {
vfs.forgetDeadMountpoint(ctx, d)
}
}
// PrepareRenameDentry must be called before attempting to rename the file
// represented by from. If to is not nil, it represents the file that will be
// replaced or exchanged by the rename. If PrepareRenameDentry succeeds, the
// caller must call AbortRenameDentry, CommitRenameReplaceDentry, or
// CommitRenameExchangeDentry depending on the rename's outcome.
//
// Preconditions:
// - If to is not nil, it must be a child Dentry from the same Filesystem.
// - from != to.
//
// +checklocksacquire:from.mu
// +checklocksacquire:to.mu
func (vfs *VirtualFilesystem) PrepareRenameDentry(mntns *MountNamespace, from, to *Dentry) error {
vfs.mountMu.Lock()
if mntns.mountpoints[from] != 0 {
vfs.mountMu.Unlock()
return linuxerr.EBUSY // +checklocksforce: no locks acquired.
}
if to != nil {
if mntns.mountpoints[to] != 0 {
vfs.mountMu.Unlock()
return linuxerr.EBUSY // +checklocksforce: no locks acquired.
}
to.mu.Lock()
}
from.mu.Lock()
vfs.mountMu.Unlock()
// Return with from.mu and to.mu locked, which will be unlocked by
// AbortRenameDentry, CommitRenameReplaceDentry, or
// CommitRenameExchangeDentry.
return nil // +checklocksforce: to may not be acquired.
}
// AbortRenameDentry must be called after PrepareRenameDentry if the rename
// fails.
// +checklocksrelease:from.mu
// +checklocksrelease:to.mu
func (vfs *VirtualFilesystem) AbortRenameDentry(from, to *Dentry) {
from.mu.Unlock()
if to != nil {
to.mu.Unlock()
}
}
// CommitRenameReplaceDentry must be called after the file represented by from
// is renamed without RENAME_EXCHANGE. If to is not nil, it represents the file
// that was replaced by from.
//
// Preconditions: PrepareRenameDentry was previously called on from and to.
// +checklocksrelease:from.mu
// +checklocksrelease:to.mu
func (vfs *VirtualFilesystem) CommitRenameReplaceDentry(ctx context.Context, from, to *Dentry) {
from.mu.Unlock()
if to != nil {
to.dead = true
to.mu.Unlock()
if to.isMounted() {
vfs.forgetDeadMountpoint(ctx, to)
}
}
}
// CommitRenameExchangeDentry must be called after the files represented by
// from and to are exchanged by rename(RENAME_EXCHANGE).
//
// Preconditions: PrepareRenameDentry was previously called on from and to.
// +checklocksrelease:from.mu
// +checklocksrelease:to.mu
func (vfs *VirtualFilesystem) CommitRenameExchangeDentry(from, to *Dentry) {
from.mu.Unlock()
to.mu.Unlock()
}
// forgetDeadMountpoint is called when a mount point is deleted or invalidated
// to umount all mounts using it in all other mount namespaces.
//
// forgetDeadMountpoint is analogous to Linux's
// fs/namespace.c:__detach_mounts().
func (vfs *VirtualFilesystem) forgetDeadMountpoint(ctx context.Context, d *Dentry) {
var (
vdsToDecRef []VirtualDentry
mountsToDecRef []*Mount
)
vfs.mountMu.Lock()
vfs.mounts.seq.BeginWrite()
for mnt := range vfs.mountpoints[d] {
vdsToDecRef, mountsToDecRef = vfs.umountRecursiveLocked(mnt, &umountRecursiveOptions{}, vdsToDecRef, mountsToDecRef)
}
vfs.mounts.seq.EndWrite()
vfs.mountMu.Unlock()
for _, vd := range vdsToDecRef {
vd.DecRef(ctx)
}
for _, mnt := range mountsToDecRef {
mnt.DecRef(ctx)
}
}
|