|
| 1 | +/* |
| 2 | + Copyright The containerd Authors. |
| 3 | +
|
| 4 | + Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | + you may not use this file except in compliance with the License. |
| 6 | + You may obtain a copy of the License at |
| 7 | +
|
| 8 | + http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | +
|
| 10 | + Unless required by applicable law or agreed to in writing, software |
| 11 | + distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | + See the License for the specific language governing permissions and |
| 14 | + limitations under the License. |
| 15 | +*/ |
| 16 | + |
| 17 | +package blockfile |
| 18 | + |
| 19 | +import ( |
| 20 | + "context" |
| 21 | + "fmt" |
| 22 | + "os" |
| 23 | + "path/filepath" |
| 24 | + |
| 25 | + "github.com/containerd/containerd/log" |
| 26 | + "github.com/containerd/containerd/mount" |
| 27 | + "github.com/containerd/containerd/plugin" |
| 28 | + "github.com/containerd/containerd/snapshots" |
| 29 | + "github.com/containerd/containerd/snapshots/storage" |
| 30 | + |
| 31 | + "github.com/containerd/continuity/fs" |
| 32 | +) |
| 33 | + |
| 34 | +// SnapshotterConfig holds the configurable properties for the blockfile snapshotter |
| 35 | +type SnapshotterConfig struct { |
| 36 | + // recreateScratch is whether scratch should be recreated even |
| 37 | + // if already exists |
| 38 | + recreateScratch bool |
| 39 | + |
| 40 | + scratchGenerator func(string) error |
| 41 | + |
| 42 | + // fsType is the filesystem type for the mount (defaults to ext4) |
| 43 | + fsType string |
| 44 | +} |
| 45 | + |
| 46 | +// Opt is an option to configure the overlay snapshotter |
| 47 | +type Opt func(string, *SnapshotterConfig) |
| 48 | + |
| 49 | +// WithScratchFile provides a scratch file which will get copied on startup |
| 50 | +// if the scratch file needs to be generated. |
| 51 | +func WithScratchFile(src string) Opt { |
| 52 | + return func(root string, config *SnapshotterConfig) { |
| 53 | + config.scratchGenerator = func(dst string) error { |
| 54 | + // Copy src to dst |
| 55 | + scratch := filepath.Join(root, "scratch") |
| 56 | + if err := fs.CopyFile(scratch, src); err != nil { |
| 57 | + return fmt.Errorf("failed to copy scratch: %w", err) |
| 58 | + } |
| 59 | + return nil |
| 60 | + } |
| 61 | + } |
| 62 | +} |
| 63 | + |
| 64 | +// WithFSType defines the filesystem type to apply to mounts of the blockfile |
| 65 | +func WithFSType(fsType string) Opt { |
| 66 | + return func(root string, config *SnapshotterConfig) { |
| 67 | + config.fsType = fsType |
| 68 | + } |
| 69 | +} |
| 70 | + |
| 71 | +type snapshotter struct { |
| 72 | + root string |
| 73 | + scratch string |
| 74 | + fsType string |
| 75 | + ms *storage.MetaStore |
| 76 | +} |
| 77 | + |
| 78 | +// NewSnapshotter returns a Snapshotter which copies layers on the underlying |
| 79 | +// file system. A metadata file is stored under the root. |
| 80 | +func NewSnapshotter(root string, opts ...Opt) (snapshots.Snapshotter, error) { |
| 81 | + var config SnapshotterConfig |
| 82 | + if err := os.MkdirAll(root, 0700); err != nil { |
| 83 | + return nil, err |
| 84 | + } |
| 85 | + |
| 86 | + for _, opt := range opts { |
| 87 | + opt(root, &config) |
| 88 | + } |
| 89 | + |
| 90 | + scratch := filepath.Join(root, "scratch") |
| 91 | + createScratch := config.recreateScratch |
| 92 | + if !createScratch { |
| 93 | + if _, err := os.Stat(scratch); err != nil { |
| 94 | + if !os.IsNotExist(err) { |
| 95 | + return nil, fmt.Errorf("unable to stat scratch file: %w", err) |
| 96 | + } |
| 97 | + createScratch = true |
| 98 | + } |
| 99 | + } |
| 100 | + if createScratch { |
| 101 | + if config.scratchGenerator == nil { |
| 102 | + return nil, fmt.Errorf("no scratch file generator: %w", plugin.ErrSkipPlugin) |
| 103 | + } |
| 104 | + if err := config.scratchGenerator(scratch); err != nil { |
| 105 | + return nil, fmt.Errorf("failed to generate scratch file: %w", err) |
| 106 | + } |
| 107 | + } |
| 108 | + |
| 109 | + if config.fsType == "" { |
| 110 | + config.fsType = "ext4" |
| 111 | + } |
| 112 | + |
| 113 | + ms, err := storage.NewMetaStore(filepath.Join(root, "metadata.db")) |
| 114 | + if err != nil { |
| 115 | + return nil, err |
| 116 | + } |
| 117 | + |
| 118 | + if err := os.Mkdir(filepath.Join(root, "snapshots"), 0700); err != nil && !os.IsExist(err) { |
| 119 | + return nil, err |
| 120 | + } |
| 121 | + |
| 122 | + return &snapshotter{ |
| 123 | + root: root, |
| 124 | + scratch: scratch, |
| 125 | + fsType: config.fsType, |
| 126 | + ms: ms, |
| 127 | + }, nil |
| 128 | +} |
| 129 | + |
| 130 | +// Stat returns the info for an active or committed snapshot by name or |
| 131 | +// key. |
| 132 | +// |
| 133 | +// Should be used for parent resolution, existence checks and to discern |
| 134 | +// the kind of snapshot. |
| 135 | +func (o *snapshotter) Stat(ctx context.Context, key string) (info snapshots.Info, err error) { |
| 136 | + err = o.ms.WithTransaction(ctx, false, func(ctx context.Context) error { |
| 137 | + _, info, _, err = storage.GetInfo(ctx, key) |
| 138 | + return err |
| 139 | + }) |
| 140 | + if err != nil { |
| 141 | + return snapshots.Info{}, err |
| 142 | + } |
| 143 | + |
| 144 | + return info, nil |
| 145 | +} |
| 146 | + |
| 147 | +func (o *snapshotter) Update(ctx context.Context, info snapshots.Info, fieldpaths ...string) (_ snapshots.Info, err error) { |
| 148 | + err = o.ms.WithTransaction(ctx, true, func(ctx context.Context) error { |
| 149 | + info, err = storage.UpdateInfo(ctx, info, fieldpaths...) |
| 150 | + return err |
| 151 | + }) |
| 152 | + if err != nil { |
| 153 | + return snapshots.Info{}, err |
| 154 | + } |
| 155 | + |
| 156 | + return info, nil |
| 157 | +} |
| 158 | + |
| 159 | +func (o *snapshotter) Usage(ctx context.Context, key string) (usage snapshots.Usage, err error) { |
| 160 | + var ( |
| 161 | + id string |
| 162 | + info snapshots.Info |
| 163 | + ) |
| 164 | + |
| 165 | + err = o.ms.WithTransaction(ctx, false, func(ctx context.Context) error { |
| 166 | + id, info, usage, err = storage.GetInfo(ctx, key) |
| 167 | + if err != nil { |
| 168 | + return err |
| 169 | + } |
| 170 | + |
| 171 | + // Current usage calculation is an approximation based on the size |
| 172 | + // of the block file - the size of its parent. This does not consider |
| 173 | + // that the filesystem may not support shared extents between the block |
| 174 | + // file and its parents, in which case the accurate calculation would just |
| 175 | + // be the size of the block file. Additionally, this does not take into |
| 176 | + // consideration that file may have been removed before being adding, |
| 177 | + // making the number of shared extents between the parent and the block |
| 178 | + // file smaller than the parent, under reporting actual usage. |
| 179 | + // |
| 180 | + // A more ideal calculation would look like: |
| 181 | + // size(block) - usage(extent_intersection(block,parent)) |
| 182 | + // OR |
| 183 | + // usage(extent_union(block,parent)) - size(parent) |
| 184 | + |
| 185 | + if info.Kind == snapshots.KindActive { |
| 186 | + // TODO: Use size calculator from fs package |
| 187 | + st, err := os.Stat(o.getBlockFile(id)) |
| 188 | + if err != nil { |
| 189 | + return err |
| 190 | + } |
| 191 | + usage.Size = st.Size() |
| 192 | + usage.Inodes = 1 |
| 193 | + } |
| 194 | + |
| 195 | + if info.Parent != "" { |
| 196 | + // GetInfo returns total number of bytes used by a snapshot (including parent). |
| 197 | + // So subtract parent usage in order to get delta consumed by layer itself. |
| 198 | + _, _, parentUsage, err := storage.GetInfo(ctx, info.Parent) |
| 199 | + if err != nil { |
| 200 | + return err |
| 201 | + } |
| 202 | + |
| 203 | + usage.Size -= parentUsage.Size |
| 204 | + } |
| 205 | + |
| 206 | + return err |
| 207 | + }) |
| 208 | + if err != nil { |
| 209 | + return snapshots.Usage{}, err |
| 210 | + } |
| 211 | + |
| 212 | + return usage, nil |
| 213 | +} |
| 214 | + |
| 215 | +func (o *snapshotter) Prepare(ctx context.Context, key, parent string, opts ...snapshots.Opt) ([]mount.Mount, error) { |
| 216 | + return o.createSnapshot(ctx, snapshots.KindActive, key, parent, opts) |
| 217 | +} |
| 218 | + |
| 219 | +func (o *snapshotter) View(ctx context.Context, key, parent string, opts ...snapshots.Opt) ([]mount.Mount, error) { |
| 220 | + return o.createSnapshot(ctx, snapshots.KindView, key, parent, opts) |
| 221 | +} |
| 222 | + |
| 223 | +// Mounts returns the mounts for the transaction identified by key. Can be |
| 224 | +// called on an read-write or readonly transaction. |
| 225 | +// |
| 226 | +// This can be used to recover mounts after calling View or Prepare. |
| 227 | +func (o *snapshotter) Mounts(ctx context.Context, key string) (_ []mount.Mount, err error) { |
| 228 | + var s storage.Snapshot |
| 229 | + err = o.ms.WithTransaction(ctx, false, func(ctx context.Context) error { |
| 230 | + s, err = storage.GetSnapshot(ctx, key) |
| 231 | + if err != nil { |
| 232 | + return fmt.Errorf("failed to get snapshot mount: %w", err) |
| 233 | + } |
| 234 | + |
| 235 | + return nil |
| 236 | + }) |
| 237 | + if err != nil { |
| 238 | + return nil, err |
| 239 | + } |
| 240 | + |
| 241 | + return o.mounts(s), nil |
| 242 | +} |
| 243 | + |
| 244 | +func (o *snapshotter) Commit(ctx context.Context, name, key string, opts ...snapshots.Opt) error { |
| 245 | + return o.ms.WithTransaction(ctx, true, func(ctx context.Context) error { |
| 246 | + id, _, _, err := storage.GetInfo(ctx, key) |
| 247 | + if err != nil { |
| 248 | + return err |
| 249 | + } |
| 250 | + |
| 251 | + st, err := os.Stat(o.getBlockFile(id)) |
| 252 | + if err != nil { |
| 253 | + return err |
| 254 | + } |
| 255 | + |
| 256 | + usage := snapshots.Usage{ |
| 257 | + Size: st.Size(), |
| 258 | + Inodes: 1, |
| 259 | + } |
| 260 | + |
| 261 | + if _, err = storage.CommitActive(ctx, key, name, usage, opts...); err != nil { |
| 262 | + return fmt.Errorf("failed to commit snapshot: %w", err) |
| 263 | + } |
| 264 | + return nil |
| 265 | + }) |
| 266 | +} |
| 267 | + |
| 268 | +// Remove abandons the transaction identified by key. All resources |
| 269 | +// associated with the key will be removed. |
| 270 | +func (o *snapshotter) Remove(ctx context.Context, key string) (err error) { |
| 271 | + var ( |
| 272 | + renamed, path string |
| 273 | + restore bool |
| 274 | + ) |
| 275 | + |
| 276 | + err = o.ms.WithTransaction(ctx, true, func(ctx context.Context) error { |
| 277 | + id, _, err := storage.Remove(ctx, key) |
| 278 | + if err != nil { |
| 279 | + return fmt.Errorf("failed to remove: %w", err) |
| 280 | + } |
| 281 | + |
| 282 | + path = o.getBlockFile(id) |
| 283 | + renamed = filepath.Join(o.root, "snapshots", "rm-"+id) |
| 284 | + if err = os.Rename(path, renamed); err != nil { |
| 285 | + if !os.IsNotExist(err) { |
| 286 | + return fmt.Errorf("failed to rename: %w", err) |
| 287 | + } |
| 288 | + renamed = "" |
| 289 | + } |
| 290 | + |
| 291 | + restore = true |
| 292 | + return nil |
| 293 | + }) |
| 294 | + |
| 295 | + if err != nil { |
| 296 | + if renamed != "" && restore { |
| 297 | + if err1 := os.Rename(renamed, path); err1 != nil { |
| 298 | + // May cause inconsistent data on disk |
| 299 | + log.G(ctx).WithError(err1).WithField("path", renamed).Error("failed to rename after failed commit") |
| 300 | + } |
| 301 | + } |
| 302 | + return err |
| 303 | + } |
| 304 | + if renamed != "" { |
| 305 | + if err := os.Remove(renamed); err != nil { |
| 306 | + // Must be cleaned up, any "rm-*" could be removed if no active transactions |
| 307 | + log.G(ctx).WithError(err).WithField("path", renamed).Warnf("failed to remove root filesystem") |
| 308 | + } |
| 309 | + } |
| 310 | + |
| 311 | + return nil |
| 312 | +} |
| 313 | + |
| 314 | +// Walk the committed snapshots. |
| 315 | +func (o *snapshotter) Walk(ctx context.Context, fn snapshots.WalkFunc, fs ...string) error { |
| 316 | + return o.ms.WithTransaction(ctx, false, func(ctx context.Context) error { |
| 317 | + return storage.WalkInfo(ctx, fn, fs...) |
| 318 | + }) |
| 319 | +} |
| 320 | + |
| 321 | +func (o *snapshotter) createSnapshot(ctx context.Context, kind snapshots.Kind, key, parent string, opts []snapshots.Opt) (_ []mount.Mount, err error) { |
| 322 | + var s storage.Snapshot |
| 323 | + |
| 324 | + err = o.ms.WithTransaction(ctx, true, func(ctx context.Context) error { |
| 325 | + s, err = storage.CreateSnapshot(ctx, kind, key, parent, opts...) |
| 326 | + if err != nil { |
| 327 | + return fmt.Errorf("failed to create snapshot: %w", err) |
| 328 | + } |
| 329 | + |
| 330 | + if len(s.ParentIDs) == 0 || s.Kind == snapshots.KindActive { |
| 331 | + path := o.getBlockFile(s.ID) |
| 332 | + |
| 333 | + if len(s.ParentIDs) > 0 { |
| 334 | + if err = fs.CopyFile(path, o.getBlockFile(s.ParentIDs[0])); err != nil { |
| 335 | + return fmt.Errorf("copying of parent failed: %w", err) |
| 336 | + } |
| 337 | + } else { |
| 338 | + if err = fs.CopyFile(path, o.scratch); err != nil { |
| 339 | + return fmt.Errorf("copying of scratch failed: %w", err) |
| 340 | + } |
| 341 | + } |
| 342 | + } |
| 343 | + |
| 344 | + return nil |
| 345 | + }) |
| 346 | + if err != nil { |
| 347 | + return nil, err |
| 348 | + } |
| 349 | + |
| 350 | + return o.mounts(s), nil |
| 351 | +} |
| 352 | + |
| 353 | +func (o *snapshotter) getBlockFile(id string) string { |
| 354 | + return filepath.Join(o.root, "snapshots", id) |
| 355 | +} |
| 356 | + |
| 357 | +func (o *snapshotter) mounts(s storage.Snapshot) []mount.Mount { |
| 358 | + var ( |
| 359 | + mountOptions = []string{ |
| 360 | + "loop", |
| 361 | + } |
| 362 | + source string |
| 363 | + ) |
| 364 | + |
| 365 | + if s.Kind == snapshots.KindView { |
| 366 | + mountOptions = append(mountOptions, "ro") |
| 367 | + } else { |
| 368 | + mountOptions = append(mountOptions, "rw") |
| 369 | + } |
| 370 | + |
| 371 | + if len(s.ParentIDs) == 0 || s.Kind == snapshots.KindActive { |
| 372 | + source = o.getBlockFile(s.ID) |
| 373 | + } else { |
| 374 | + source = o.getBlockFile(s.ParentIDs[0]) |
| 375 | + } |
| 376 | + |
| 377 | + return []mount.Mount{ |
| 378 | + { |
| 379 | + Source: source, |
| 380 | + Type: o.fsType, |
| 381 | + Options: mountOptions, |
| 382 | + }, |
| 383 | + } |
| 384 | +} |
| 385 | + |
| 386 | +// Close closes the snapshotter |
| 387 | +func (o *snapshotter) Close() error { |
| 388 | + return o.ms.Close() |
| 389 | +} |
0 commit comments