| 1 | package hostfiles |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "context" |
| 6 | "crypto/sha256" |
| 7 | "encoding/hex" |
| 8 | "encoding/json" |
| 9 | "errors" |
| 10 | "fmt" |
| 11 | "slices" |
| 12 | "sort" |
| 13 | "strings" |
| 14 | "sync" |
| 15 | |
| 16 | "github.com/carlosframework/platform/internal/release" |
| 17 | ) |
| 18 | |
| 19 | // ErrSuperseded means another writer replaced the manifest. The caller |
| 20 | // stops publishing: it is no longer this host's executor, or two |
| 21 | // publishers were started for one host. |
| 22 | var ErrSuperseded = errors.New("hostfiles: another writer published this manifest") |
| 23 | |
| 24 | // Publisher stages changes and commits them as one manifest revision. It |
| 25 | // is safe for concurrent use. |
| 26 | type Publisher struct { |
| 27 | s release.Store |
| 28 | prefix string |
| 29 | |
| 30 | mu sync.Mutex |
| 31 | man Manifest |
| 32 | etag string |
| 33 | next int64 // the last sequence number handed out; man.Seq is the last committed |
| 34 | dirty bool |
| 35 | // inflight holds, by sha256, every manifest body a failed Put may have |
| 36 | // applied since the last successful commit, with the sequence number it |
| 37 | // carried. Every one of those Puts named the same etag, so at most one |
| 38 | // of them can be in storage. |
| 39 | inflight map[string]int64 |
| 40 | } |
| 41 | |
| 42 | func Open(ctx context.Context, s release.Store, prefix string) (*Publisher, error) { |
| 43 | if err := checkPrefix(prefix); err != nil { |
| 44 | return nil, err |
| 45 | } |
| 46 | m, etag, _, err := load(ctx, s, prefix) |
| 47 | if err != nil { |
| 48 | return nil, err |
| 49 | } |
| 50 | // A manifest written before the path cap may hold a path past it. No |
| 51 | // reader accepts it (checkManifest), and no Commit could ever write it |
| 52 | // again, so it is refused here, terminally, rather than continued into |
| 53 | // a publisher whose every commit fails. An operator replaces it. |
| 54 | if err := manifestWithinCap(m); err != nil { |
| 55 | return nil, fmt.Errorf("hostfiles: open %s: the manifest cannot be continued, an operator must replace it: %w", prefix, err) |
| 56 | } |
| 57 | return &Publisher{s: s, prefix: prefix, man: m, etag: etag, next: m.Seq}, nil |
| 58 | } |
| 59 | |
| 60 | // RestorableWithinCap is the check a caller makes before it starts work |
| 61 | // that will need m restored (a move's cut, the claim's files backfill): |
| 62 | // every entry path of m, live, tombstoned or skipped, within the tree's |
| 63 | // length cap. A manifest that fails it is ErrTooDeep: no reader accepts |
| 64 | // it, so an operator replaces it before anything that needs it starts |
| 65 | // (#236, Paul's ruling of 2026-10-15). |
| 66 | func RestorableWithinCap(m Manifest) error { return manifestWithinCap(m) } |
| 67 | |
| 68 | // manifestWithinCap holds every entry path of m to withinTreeCap, the cap |
| 69 | // the readers apply: Open refuses to continue a manifest that fails it, |
| 70 | // and Commit refuses to write one. |
| 71 | func manifestWithinCap(m Manifest) error { |
| 72 | for path := range m.Entries { |
| 73 | if err := withinTreeCap(path); err != nil { |
| 74 | return err |
| 75 | } |
| 76 | } |
| 77 | return nil |
| 78 | } |
| 79 | |
| 80 | // OpenAt is a publisher over the manifest bytes a caller already read and |
| 81 | // checked, with the token the store answered for them: the publisher's |
| 82 | // baseline and its compare-and-swap token are exactly those, never a |
| 83 | // second read (Task 6 fix round 2, finding 2). A manifest replaced at the |
| 84 | // key after that read makes the first Commit fail (ErrSuperseded) rather |
| 85 | // than publish over bytes nobody checked. |
| 86 | func OpenAt(s release.Store, prefix string, body []byte, etag string) (*Publisher, error) { |
| 87 | if s == nil { |
| 88 | return nil, errors.New("hostfiles: open: no store") |
| 89 | } |
| 90 | if err := checkPrefix(prefix); err != nil { |
| 91 | return nil, err |
| 92 | } |
| 93 | if etag == "" { |
| 94 | return nil, fmt.Errorf("hostfiles: open %s at its read bytes: no token", prefix) |
| 95 | } |
| 96 | m, err := decode(prefix, body) |
| 97 | if err != nil { |
| 98 | return nil, err |
| 99 | } |
| 100 | // The same terminal refusal as Open's: a manifest past the path cap |
| 101 | // cannot be continued. |
| 102 | if err := manifestWithinCap(m); err != nil { |
| 103 | return nil, fmt.Errorf("hostfiles: open %s: the manifest cannot be continued, an operator must replace it: %w", prefix, err) |
| 104 | } |
| 105 | return &Publisher{s: s, prefix: prefix, man: m, etag: etag, next: m.Seq}, nil |
| 106 | } |
| 107 | |
| 108 | func sum(body []byte) string { |
| 109 | h := sha256.Sum256(body) |
| 110 | return hex.EncodeToString(h[:]) |
| 111 | } |
| 112 | |
| 113 | // put writes body as the next immutable object under this publisher's own |
| 114 | // prefix and returns its entry, whose Object is the full key. Callers hold |
| 115 | // p.mu. |
| 116 | func (p *Publisher) put(ctx context.Context, body []byte, mode uint32, torn bool) (Entry, error) { |
| 117 | sha := sum(body) |
| 118 | seq := p.next + 1 |
| 119 | key := p.prefix + objectKey(seq, sha) |
| 120 | if _, err := p.s.Put(ctx, key, body, ""); err != nil && !errors.Is(err, release.ErrCASConflict) { |
| 121 | // A conflict means this exact key exists: the same sequence number |
| 122 | // and the same hash, left by an attempt that crashed before its |
| 123 | // commit. The bytes are the ones we were about to write. |
| 124 | return Entry{}, err |
| 125 | } |
| 126 | p.next = seq |
| 127 | return Entry{Object: key, SHA256: sha, Size: int64(len(body)), Mode: mode & 0o777, Seq: seq, Torn: torn}, nil |
| 128 | } |
| 129 | |
| 130 | // Stage records path's content. Content identical to what is already |
| 131 | // staged or committed takes no sequence number and writes nothing. |
| 132 | func (p *Publisher) Stage(ctx context.Context, path string, body []byte, mode uint32, torn bool) error { |
| 133 | clean, err := CleanPath(path) |
| 134 | if err != nil { |
| 135 | return err |
| 136 | } |
| 137 | if err := withinTreeCap(clean); err != nil { |
| 138 | return err |
| 139 | } |
| 140 | p.mu.Lock() |
| 141 | defer p.mu.Unlock() |
| 142 | if cur, ok := p.man.Entries[clean]; ok && !cur.Deleted && cur.Skipped == "" && |
| 143 | cur.SHA256 == sum(body) && cur.Mode == mode&0o777 && cur.Torn == torn { |
| 144 | return nil |
| 145 | } |
| 146 | e, err := p.put(ctx, body, mode, torn) |
| 147 | if err != nil { |
| 148 | return err |
| 149 | } |
| 150 | p.man.Entries[clean], p.dirty = e, true |
| 151 | return nil |
| 152 | } |
| 153 | |
| 154 | // StageSkipped records that path exists but was not published, and why. |
| 155 | func (p *Publisher) StageSkipped(path, why string) error { |
| 156 | clean, err := CleanPath(path) |
| 157 | if err != nil { |
| 158 | return err |
| 159 | } |
| 160 | if err := withinTreeCap(clean); err != nil { |
| 161 | return err |
| 162 | } |
| 163 | p.mu.Lock() |
| 164 | defer p.mu.Unlock() |
| 165 | if cur, ok := p.man.Entries[clean]; ok && cur.Skipped == why { |
| 166 | return nil |
| 167 | } |
| 168 | p.next++ |
| 169 | p.man.Entries[clean], p.dirty = Entry{Seq: p.next, Skipped: why}, true |
| 170 | return nil |
| 171 | } |
| 172 | |
| 173 | // StageDelete records that path is gone. Deleting a path the manifest never |
| 174 | // had, or already has as deleted, is a no-op. |
| 175 | func (p *Publisher) StageDelete(path string) error { |
| 176 | clean, err := CleanPath(path) |
| 177 | if err != nil { |
| 178 | return err |
| 179 | } |
| 180 | if err := withinTreeCap(clean); err != nil { |
| 181 | return err |
| 182 | } |
| 183 | p.mu.Lock() |
| 184 | defer p.mu.Unlock() |
| 185 | if cur, ok := p.man.Entries[clean]; !ok || cur.Deleted { |
| 186 | return nil |
| 187 | } |
| 188 | p.next++ |
| 189 | p.man.Entries[clean], p.dirty = Entry{Seq: p.next, Deleted: true}, true |
| 190 | return nil |
| 191 | } |
| 192 | |
| 193 | // StagePlatform records a platform-managed item (wake state, an env file). |
| 194 | // It shares the manifest's ordering but is never written into the files |
| 195 | // directory by Restore. |
| 196 | func (p *Publisher) StagePlatform(ctx context.Context, name string, body []byte) error { |
| 197 | if name == "" || strings.ContainsAny(name, "/\x00") { |
| 198 | return fmt.Errorf("hostfiles: platform item %q is not a plain name", name) |
| 199 | } |
| 200 | p.mu.Lock() |
| 201 | defer p.mu.Unlock() |
| 202 | if cur, ok := p.man.Platform[name]; ok && cur.SHA256 == sum(body) { |
| 203 | return nil |
| 204 | } |
| 205 | e, err := p.put(ctx, body, 0o600, false) |
| 206 | if err != nil { |
| 207 | return err |
| 208 | } |
| 209 | p.man.Platform[name], p.dirty = e, true |
| 210 | return nil |
| 211 | } |
| 212 | |
| 213 | // Commit publishes everything staged as one manifest revision and returns |
| 214 | // its sequence number. With nothing staged it returns the current one. A |
| 215 | // failed commit leaves the staged changes in place, so the caller retries; |
| 216 | // ErrSuperseded is final. |
| 217 | func (p *Publisher) Commit(ctx context.Context) (int64, error) { |
| 218 | p.mu.Lock() |
| 219 | defer p.mu.Unlock() |
| 220 | if !p.dirty { |
| 221 | return p.man.Seq, nil |
| 222 | } |
| 223 | // The readers refuse a manifest with a path past the cap |
| 224 | // (checkManifest), so none is written: Open refuses to continue one and |
| 225 | // Stage refuses to add one; this is the last line before the store. |
| 226 | if err := manifestWithinCap(p.man); err != nil { |
| 227 | return 0, err |
| 228 | } |
| 229 | out := p.man |
| 230 | out.Seq = p.next |
| 231 | body, err := json.Marshal(out) |
| 232 | if err != nil { |
| 233 | return 0, fmt.Errorf("hostfiles: marshal manifest: %w", err) |
| 234 | } |
| 235 | etag, err := p.s.Put(ctx, ManifestKey(p.prefix), body, p.etag) |
| 236 | if errors.Is(err, release.ErrCASConflict) { |
| 237 | // A lost response looks like this: the store applied an earlier Put, |
| 238 | // the reply never arrived, and this retry's etag is now stale. If |
| 239 | // the manifest holds exactly the bytes this Commit is writing, or |
| 240 | // the bytes an earlier failed Commit wrote, it is ours. |
| 241 | cur, curEtag, gerr := p.s.Get(ctx, ManifestKey(p.prefix)) |
| 242 | if gerr != nil { |
| 243 | // Not knowing is not the same as being superseded: leave the |
| 244 | // staged state and etag alone so the next Commit repeats the check. |
| 245 | return 0, gerr |
| 246 | } |
| 247 | if bytes.Equal(cur, body) { |
| 248 | err, etag = nil, curEtag |
| 249 | } else if seq, ok := p.inflight[sum(cur)]; ok { |
| 250 | // An earlier attempt landed and later stages changed the body. |
| 251 | // Adopt it, then write the current body once on top of it. |
| 252 | p.man.Seq, p.etag, p.inflight = seq, curEtag, nil |
| 253 | etag, err = p.s.Put(ctx, ManifestKey(p.prefix), body, p.etag) |
| 254 | if errors.Is(err, release.ErrCASConflict) { |
| 255 | return 0, fmt.Errorf("%w: %s", ErrSuperseded, p.prefix) |
| 256 | } |
| 257 | } else { |
| 258 | return 0, fmt.Errorf("%w: %s", ErrSuperseded, p.prefix) |
| 259 | } |
| 260 | } |
| 261 | if err != nil { |
| 262 | if !errors.Is(err, release.ErrCASConflict) { |
| 263 | if p.inflight == nil { |
| 264 | p.inflight = map[string]int64{} |
| 265 | } |
| 266 | p.inflight[sum(body)] = out.Seq |
| 267 | } |
| 268 | return 0, err |
| 269 | } |
| 270 | p.man.Seq, p.etag, p.dirty, p.inflight = out.Seq, etag, false, nil |
| 271 | return out.Seq, nil |
| 272 | } |
| 273 | |
| 274 | // SetOrigins sets the manifest's carried origin set (Manifest.Origins), |
| 275 | // sorted and deduplicated. The same set is not a change; a different one |
| 276 | // goes out with the next Commit. |
| 277 | func (p *Publisher) SetOrigins(origins []string) { |
| 278 | set := map[string]bool{} |
| 279 | for _, o := range origins { |
| 280 | if o != "" { |
| 281 | set[o] = true |
| 282 | } |
| 283 | } |
| 284 | out := make([]string, 0, len(set)) |
| 285 | for o := range set { |
| 286 | out = append(out, o) |
| 287 | } |
| 288 | sort.Strings(out) |
| 289 | p.mu.Lock() |
| 290 | defer p.mu.Unlock() |
| 291 | if slices.Equal(out, p.man.Origins) { |
| 292 | return |
| 293 | } |
| 294 | p.man.Origins, p.dirty = out, true |
| 295 | } |
| 296 | |
| 297 | func (p *Publisher) Dirty() bool { p.mu.Lock(); defer p.mu.Unlock(); return p.dirty } |
| 298 | |
| 299 | func (p *Publisher) Seq() int64 { p.mu.Lock(); defer p.mu.Unlock(); return p.man.Seq } |
| 300 | |
| 301 | // Paths lists every path the manifest holds as present: published or |
| 302 | // skipped, not deleted. |
| 303 | func (p *Publisher) Paths() []string { |
| 304 | p.mu.Lock() |
| 305 | defer p.mu.Unlock() |
| 306 | var out []string |
| 307 | for path, e := range p.man.Entries { |
| 308 | if !e.Deleted { |
| 309 | out = append(out, path) |
| 310 | } |
| 311 | } |
| 312 | sort.Strings(out) |
| 313 | return out |
| 314 | } |
| 315 | |
| 316 | // pending is the last sequence number handed out, committed or not. |
| 317 | func (p *Publisher) pending() int64 { |
| 318 | p.mu.Lock() |
| 319 | defer p.mu.Unlock() |
| 320 | return p.next |
| 321 | } |
| 322 | |
amadan