| 1 | package activator |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "context" |
| 6 | "encoding/base64" |
| 7 | "errors" |
| 8 | "log" |
| 9 | "os" |
| 10 | "path/filepath" |
| 11 | "strings" |
| 12 | "sync" |
| 13 | "sync/atomic" |
| 14 | "syscall" |
| 15 | "testing" |
| 16 | "time" |
| 17 | |
| 18 | "github.com/carlosframework/platform/internal/provision" |
| 19 | "github.com/carlosframework/platform/internal/registry" |
| 20 | ) |
| 21 | |
| 22 | // keyLines returns every CARLOS_INSTANCE_KEY value in raw, in order. |
| 23 | func keyLines(raw []byte) []string { |
| 24 | var out []string |
| 25 | for _, line := range strings.Split(string(raw), "\n") { |
| 26 | if v, ok := strings.CutPrefix(line, InstanceKeyVar+"="); ok { |
| 27 | out = append(out, v) |
| 28 | } |
| 29 | } |
| 30 | return out |
| 31 | } |
| 32 | |
| 33 | // syncProbe counts the key's syncs and fails them on demand. Set the |
| 34 | // error fields only between calls. |
| 35 | type syncProbe struct { |
| 36 | file, dir int |
| 37 | fileErr, dirErr error |
| 38 | } |
| 39 | |
| 40 | func probeInstanceKeySync(t *testing.T) *syncProbe { |
| 41 | t.Helper() |
| 42 | p := &syncProbe{} |
| 43 | adminTokenMu.Lock() |
| 44 | old := instanceKeySync |
| 45 | instanceKeySync.file = func(f *os.File) error { |
| 46 | p.file++ |
| 47 | if p.fileErr != nil { |
| 48 | return p.fileErr |
| 49 | } |
| 50 | return f.Sync() |
| 51 | } |
| 52 | instanceKeySync.dir = func(d string) error { |
| 53 | p.dir++ |
| 54 | if p.dirErr != nil { |
| 55 | return p.dirErr |
| 56 | } |
| 57 | return syncDir(d) |
| 58 | } |
| 59 | adminTokenMu.Unlock() |
| 60 | t.Cleanup(func() { |
| 61 | adminTokenMu.Lock() |
| 62 | instanceKeySync = old |
| 63 | adminTokenMu.Unlock() |
| 64 | }) |
| 65 | return p |
| 66 | } |
| 67 | |
| 68 | func TestEnsureInstanceKeyMintsOnceInTheAgreedFormat(t *testing.T) { |
| 69 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 70 | got, err := ensureInstanceKeyNoConfig(p) |
| 71 | if err != nil || got != InstanceKeyMinted { |
| 72 | t.Fatalf("first ensure = %v, %v; want minted", got, err) |
| 73 | } |
| 74 | raw, _ := os.ReadFile(p) |
| 75 | keys := keyLines(raw) |
| 76 | if len(keys) != 1 { |
| 77 | t.Fatalf("file holds %d keys, want 1", len(keys)) |
| 78 | } |
| 79 | dec, err := base64.StdEncoding.Strict().DecodeString(keys[0]) |
| 80 | if err != nil || len(dec) != 32 || len(keys[0]) != 44 { |
| 81 | t.Fatalf("key is not 32 bytes in padded standard base64 (len %d, decode err %v)", len(keys[0]), err) |
| 82 | } |
| 83 | if !bytes.HasSuffix(raw, []byte("\n")) { |
| 84 | t.Fatal("the key line has no newline: the next append would join it") |
| 85 | } |
| 86 | if st, _ := os.Stat(p); st.Mode().Perm() != 0o600 { |
| 87 | t.Fatalf("mode = %o, want 600", st.Mode().Perm()) |
| 88 | } |
| 89 | // The parser every launch path uses yields the exact characters minted. |
| 90 | env, _ := readEnvFile(p) |
| 91 | if env[InstanceKeyVar] != keys[0] { |
| 92 | t.Fatal("readEnvFile does not yield the minted key verbatim") |
| 93 | } |
| 94 | again, err := ensureInstanceKeyNoConfig(p) |
| 95 | if err != nil || again != InstanceKeyPresent { |
| 96 | t.Fatalf("second ensure = %v, %v; want present", again, err) |
| 97 | } |
| 98 | raw2, _ := os.ReadFile(p) |
| 99 | if !bytes.Equal(raw, raw2) { |
| 100 | t.Fatal("second ensure changed the file") |
| 101 | } |
| 102 | } |
| 103 | |
| 104 | // Never overwrite: a key already there, of any shape, is kept byte for byte |
| 105 | // and only reported. |
| 106 | func TestEnsureInstanceKeyNeverRewritesAnExistingKey(t *testing.T) { |
| 107 | for _, existing := range []string{ |
| 108 | "CARLOS_INSTANCE_KEY=hand-written-not-base64\n", |
| 109 | "A=1\nCARLOS_INSTANCE_KEY=" + strings.Repeat("A", 43) + "=", // well formed, no trailing newline |
| 110 | } { |
| 111 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 112 | if err := os.WriteFile(p, []byte(existing), 0o600); err != nil { |
| 113 | t.Fatal(err) |
| 114 | } |
| 115 | var buf bytes.Buffer |
| 116 | log.SetOutput(&buf) |
| 117 | got, err := ensureInstanceKeyNoConfig(p) |
| 118 | log.SetOutput(os.Stderr) |
| 119 | if err != nil || got != InstanceKeyPresent { |
| 120 | t.Fatalf("ensure over %q = %v, %v; want present", existing, got, err) |
| 121 | } |
| 122 | if raw, _ := os.ReadFile(p); string(raw) != existing { |
| 123 | t.Fatalf("file changed: %q -> %q", existing, raw) |
| 124 | } |
| 125 | if strings.Contains(buf.String(), "hand-written-not-base64") { |
| 126 | t.Fatal("the shape warning logged the key's value") |
| 127 | } |
| 128 | } |
| 129 | } |
| 130 | |
| 131 | func TestEnsureInstanceKeyAppendsOnAFreshLineAndKeepsEveryOtherLine(t *testing.T) { |
| 132 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 133 | before := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("a", 64) + "\n# comment\nDATABASE_URL=sqlite" // no trailing newline |
| 134 | if err := os.WriteFile(p, []byte(before), 0o644); err != nil { |
| 135 | t.Fatal(err) |
| 136 | } |
| 137 | if got, err := ensureInstanceKeyNoConfig(p); err != nil || got != InstanceKeyMinted { |
| 138 | t.Fatalf("ensure = %v, %v", got, err) |
| 139 | } |
| 140 | raw, _ := os.ReadFile(p) |
| 141 | if !strings.HasPrefix(string(raw), before+"\n"+InstanceKeyVar+"=") { |
| 142 | t.Fatalf("file = %q: the key must land on its own line after the untouched original", raw) |
| 143 | } |
| 144 | env, _ := readEnvFile(p) |
| 145 | if env["DATABASE_URL"] != "sqlite" || len(env["CARLOS_ADMIN_TOKEN"]) != 64 { |
| 146 | t.Fatal("an existing line was lost") |
| 147 | } |
| 148 | if st, _ := os.Stat(p); st.Mode().Perm() != 0o600 { |
| 149 | t.Fatalf("mode = %o, want the 0644 file tightened to 600", st.Mode().Perm()) |
| 150 | } |
| 151 | } |
| 152 | |
| 153 | // A key the app's configuration supplies is the app's key: minting one into |
| 154 | // <host>.env would override it (mergeEnv) and change the key under its data. |
| 155 | // The decision is the effective env's, so an empty CARLOS_INSTANCE_KEY= in |
| 156 | // the own file does not hide the configured key (finding 4). |
| 157 | func TestEnsureInstanceKeyDefersToTheRenderedConfiguration(t *testing.T) { |
| 158 | for _, own := range []string{"A=1\n", "A=1\n" + InstanceKeyVar + "=\n"} { |
| 159 | dir := t.TempDir() |
| 160 | p := filepath.Join(dir, "shop.example.com.env") |
| 161 | if err := os.WriteFile(filepath.Join(dir, "shop.example.com.config.env"), []byte(InstanceKeyVar+"=from-config\n"), 0o600); err != nil { |
| 162 | t.Fatal(err) |
| 163 | } |
| 164 | if err := os.WriteFile(p, []byte(own), 0o600); err != nil { |
| 165 | t.Fatal(err) |
| 166 | } |
| 167 | got, err := ensureInstanceKeyNoConfig(p) |
| 168 | if err != nil || got != InstanceKeyFromConfig { |
| 169 | t.Fatalf("own %q: ensure = %v, %v; want from-config", own, got, err) |
| 170 | } |
| 171 | if raw, _ := os.ReadFile(p); string(raw) != own { |
| 172 | t.Fatalf("own file changed: %q", raw) |
| 173 | } |
| 174 | } |
| 175 | } |
| 176 | |
| 177 | func TestBackfillInstanceKeyAppendsOnlyToAnExistingEnvFile(t *testing.T) { |
| 178 | b := &ExecBackend{DataDir: t.TempDir()} |
| 179 | fresh := registry.Route{Host: "new.example.com", Socket: "/run/carlos/new.sock"} |
| 180 | got, err := b.BackfillInstanceKey(t.Context(), fresh) |
| 181 | if err != nil || got != InstanceKeyNoEnvFile { |
| 182 | t.Fatalf("backfill with no env file = %v, %v; want no-env-file", got, err) |
| 183 | } |
| 184 | if _, err := os.Lstat(filepath.Join(b.DataDir, fresh.Host+".env")); !os.IsNotExist(err) { |
| 185 | t.Fatalf("backfill created an env file: %v", err) |
| 186 | } |
| 187 | old := registry.Route{Host: "shop.example.com", Socket: "/run/carlos/shop.sock"} |
| 188 | envPath := filepath.Join(b.DataDir, old.Host+".env") |
| 189 | tok := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("b", 64) + "\n" |
| 190 | if err := os.WriteFile(envPath, []byte(tok), 0o600); err != nil { |
| 191 | t.Fatal(err) |
| 192 | } |
| 193 | if got, err := b.BackfillInstanceKey(t.Context(), old); err != nil || got != InstanceKeyMinted { |
| 194 | t.Fatalf("backfill = %v, %v; want minted", got, err) |
| 195 | } |
| 196 | raw, _ := os.ReadFile(envPath) |
| 197 | if !strings.HasPrefix(string(raw), tok) || len(keyLines(raw)) != 1 { |
| 198 | t.Fatalf("file = %q", raw) |
| 199 | } |
| 200 | // Idempotent: a second backfill (the next agent start) writes nothing. |
| 201 | if got, err := b.BackfillInstanceKey(t.Context(), old); err != nil || got != InstanceKeyPresent { |
| 202 | t.Fatalf("second backfill = %v, %v; want present", got, err) |
| 203 | } |
| 204 | if raw2, _ := os.ReadFile(envPath); !bytes.Equal(raw, raw2) { |
| 205 | t.Fatal("second backfill changed the file") |
| 206 | } |
| 207 | } |
| 208 | |
| 209 | // Review focus 1: a hand-edited file with two keys. Readers take the last |
| 210 | // (parseEnvBytes), and the ensure writes nothing. |
| 211 | func TestEnsureInstanceKeyLeavesAFileWithTwoKeysAlone(t *testing.T) { |
| 212 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 213 | body := InstanceKeyVar + "=first\n" + InstanceKeyVar + "=second\n" |
| 214 | if err := os.WriteFile(p, []byte(body), 0o600); err != nil { |
| 215 | t.Fatal(err) |
| 216 | } |
| 217 | if got, err := ensureInstanceKeyNoConfig(p); err != nil || got != InstanceKeyPresent { |
| 218 | t.Fatalf("ensure = %v, %v; want present", got, err) |
| 219 | } |
| 220 | if raw, _ := os.ReadFile(p); string(raw) != body { |
| 221 | t.Fatalf("file changed: %q", raw) |
| 222 | } |
| 223 | if env, _ := readEnvFile(p); env[InstanceKeyVar] != "second" { |
| 224 | t.Fatalf("readers see %q, want the last line's value", env[InstanceKeyVar]) |
| 225 | } |
| 226 | } |
| 227 | |
| 228 | // Review focus 2: an empty value is no key. One is appended after it, and |
| 229 | // readers take the appended one. |
| 230 | func TestEnsureInstanceKeyTreatsAnEmptyValueAsAbsent(t *testing.T) { |
| 231 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 232 | if err := os.WriteFile(p, []byte(InstanceKeyVar+"=\n"), 0o600); err != nil { |
| 233 | t.Fatal(err) |
| 234 | } |
| 235 | if got, err := ensureInstanceKeyNoConfig(p); err != nil || got != InstanceKeyMinted { |
| 236 | t.Fatalf("ensure = %v, %v; want minted", got, err) |
| 237 | } |
| 238 | env, _ := readEnvFile(p) |
| 239 | if !wellFormedInstanceKey(env[InstanceKeyVar]) { |
| 240 | t.Fatal("readers do not see the appended key") |
| 241 | } |
| 242 | } |
| 243 | |
| 244 | // Review focus 3: a seed, a visitor's wake and the backfill can ask at |
| 245 | // once. Exactly one key is written. |
| 246 | func TestEnsureInstanceKeyMintsOneKeyUnderConcurrentCallers(t *testing.T) { |
| 247 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 248 | var wg sync.WaitGroup |
| 249 | for i := 0; i < 8; i++ { |
| 250 | wg.Add(1) |
| 251 | go func() { |
| 252 | defer wg.Done() |
| 253 | if _, err := ensureInstanceKeyNoConfig(p); err != nil { |
| 254 | t.Error(err) |
| 255 | } |
| 256 | }() |
| 257 | } |
| 258 | wg.Wait() |
| 259 | if raw, _ := os.ReadFile(p); len(keyLines(raw)) != 1 { |
| 260 | t.Fatalf("%d keys written", len(keyLines(raw))) |
| 261 | } |
| 262 | } |
| 263 | |
| 264 | // Review focus 5: an empty CARLOS_INSTANCE_KEY in the rendered |
| 265 | // configuration supplies nothing, so the platform mints one. |
| 266 | func TestEnsureInstanceKeyMintsWhenTheConfigurationValueIsEmpty(t *testing.T) { |
| 267 | dir := t.TempDir() |
| 268 | if err := os.WriteFile(filepath.Join(dir, "shop.example.com.config.env"), []byte(InstanceKeyVar+"=\n"), 0o600); err != nil { |
| 269 | t.Fatal(err) |
| 270 | } |
| 271 | if got, err := ensureInstanceKeyNoConfig(filepath.Join(dir, "shop.example.com.env")); err != nil || got != InstanceKeyMinted { |
| 272 | t.Fatalf("ensure = %v, %v; want minted", got, err) |
| 273 | } |
| 274 | } |
| 275 | |
| 276 | // Finding 2: a mint whose file or directory sync fails reports failure, |
| 277 | // and so does every retry until a sync succeeds. The retry finds the key |
| 278 | // present, keeps its bytes, and still completes the chmod and both syncs |
| 279 | // before it reports success. |
| 280 | func TestEnsureInstanceKeyRedoesAFailedSyncBeforeReportingSuccess(t *testing.T) { |
| 281 | for _, tc := range []struct { |
| 282 | name string |
| 283 | fail func(*syncProbe, error) |
| 284 | }{ |
| 285 | {"file sync", func(p *syncProbe, err error) { p.fileErr = err }}, |
| 286 | {"directory sync", func(p *syncProbe, err error) { p.dirErr = err }}, |
| 287 | } { |
| 288 | t.Run(tc.name, func(t *testing.T) { |
| 289 | probe := probeInstanceKeySync(t) |
| 290 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 291 | tc.fail(probe, errors.New("injected: EIO")) |
| 292 | if _, err := ensureInstanceKeyNoConfig(p); err == nil { |
| 293 | t.Fatal("a mint whose sync failed reported success") |
| 294 | } |
| 295 | raw, _ := os.ReadFile(p) |
| 296 | if len(keyLines(raw)) != 1 { |
| 297 | t.Fatalf("the failed mint left %d keys, want its 1", len(keyLines(raw))) |
| 298 | } |
| 299 | if _, err := ensureInstanceKeyNoConfig(p); err == nil { |
| 300 | t.Fatal("a retry reported success while the sync still fails") |
| 301 | } |
| 302 | if err := os.Chmod(p, 0o644); err != nil { |
| 303 | t.Fatal(err) |
| 304 | } |
| 305 | tc.fail(probe, nil) |
| 306 | before := *probe |
| 307 | got, err := ensureInstanceKeyNoConfig(p) |
| 308 | if err != nil || got != InstanceKeyPresent { |
| 309 | t.Fatalf("retry = %v, %v; want present", got, err) |
| 310 | } |
| 311 | if probe.file != before.file+1 || probe.dir != before.dir+1 { |
| 312 | t.Fatalf("the retry synced the file %d and the directory %d times, want once each", probe.file-before.file, probe.dir-before.dir) |
| 313 | } |
| 314 | if raw2, _ := os.ReadFile(p); !bytes.Equal(raw, raw2) { |
| 315 | t.Fatal("a retry rewrote the key") |
| 316 | } |
| 317 | if st, _ := os.Stat(p); st.Mode().Perm() != 0o600 { |
| 318 | t.Fatalf("mode = %o, want the present path to set 600", st.Mode().Perm()) |
| 319 | } |
| 320 | }) |
| 321 | } |
| 322 | } |
| 323 | |
| 324 | // Ruling 26: a replica row is never minted a key, on any path; the only |
| 325 | // key it may hold is its home's (replicaKeyGate). |
| 326 | func TestReplicaRowsGetNoPlatformKey(t *testing.T) { |
| 327 | b := &ExecBackend{DataDir: t.TempDir()} |
| 328 | rt := registry.Route{Host: "shop-au.example.com", Kind: "instance", Socket: "/run/carlos/shop-au.sock", Replica: true} |
| 329 | envPath := filepath.Join(b.DataDir, rt.Host+".env") |
| 330 | if err := os.WriteFile(envPath, []byte("CARLOS_ADMIN_TOKEN=t\n"), 0o600); err != nil { |
| 331 | t.Fatal(err) |
| 332 | } |
| 333 | if got, err := b.ensureInstanceKeyFor(t.Context(), rt, b.pathsFor(rt)); err != nil || got != InstanceKeyReplica { |
| 334 | t.Fatalf("launch ensure = %v, %v; want replica", got, err) |
| 335 | } |
| 336 | if got, err := b.BackfillInstanceKey(t.Context(), rt); err != nil || got != InstanceKeyReplica { |
| 337 | t.Fatalf("backfill = %v, %v; want replica", got, err) |
| 338 | } |
| 339 | if raw, _ := os.ReadFile(envPath); len(keyLines(raw)) != 0 { |
| 340 | t.Fatal("a replica row was given a platform key") |
| 341 | } |
| 342 | } |
| 343 | |
| 344 | // The flat ensure never follows a symlink planted at <host>.env (spec §7.2, |
| 345 | // the hazard TestEnsureAdminTokenRefusesToChmodThroughASymlink guards). On |
| 346 | // every path, mint, existing key, backfill and the claim settle, it refuses |
| 347 | // and leaves the link's target's mode and bytes exactly as they were. |
| 348 | func TestEnsureInstanceKeyRefusesToFollowASymlink(t *testing.T) { |
| 349 | wellFormed := InstanceKeyVar + "=" + strings.Repeat("A", 43) + "=\n" |
| 350 | for _, tc := range []struct { |
| 351 | name string |
| 352 | target string // the target's contents |
| 353 | mode os.FileMode |
| 354 | call func(b *ExecBackend, rt registry.Route, link string) error |
| 355 | }{ |
| 356 | {"mint", "A=1\n", 0o644, func(_ *ExecBackend, _ registry.Route, link string) error { |
| 357 | _, err := ensureInstanceKeyNoConfig(link) |
| 358 | return err |
| 359 | }}, |
| 360 | {"existing key", wellFormed, 0o644, func(_ *ExecBackend, _ registry.Route, link string) error { |
| 361 | _, err := ensureInstanceKeyNoConfig(link) |
| 362 | return err |
| 363 | }}, |
| 364 | {"backfill mint", "A=1\n", 0o644, func(b *ExecBackend, rt registry.Route, _ string) error { |
| 365 | _, err := b.BackfillInstanceKey(t.Context(), rt) |
| 366 | return err |
| 367 | }}, |
| 368 | {"backfill existing key", wellFormed, 0o644, func(b *ExecBackend, rt registry.Route, _ string) error { |
| 369 | _, err := b.BackfillInstanceKey(t.Context(), rt) |
| 370 | return err |
| 371 | }}, |
| 372 | {"claim settle", wellFormed, 0o600, func(b *ExecBackend, rt registry.Route, _ string) error { |
| 373 | return b.settleClaimInstanceEnv(rt) |
| 374 | }}, |
| 375 | } { |
| 376 | t.Run(tc.name, func(t *testing.T) { |
| 377 | b := &ExecBackend{DataDir: t.TempDir()} |
| 378 | rt := registry.Route{Host: "shop.example.com", Socket: "/run/carlos/shop.sock"} |
| 379 | target := filepath.Join(t.TempDir(), "someone-elses-file") |
| 380 | if err := os.WriteFile(target, []byte(tc.target), tc.mode); err != nil { |
| 381 | t.Fatal(err) |
| 382 | } |
| 383 | if err := os.Chmod(target, tc.mode); err != nil { |
| 384 | t.Fatal(err) |
| 385 | } |
| 386 | link := filepath.Join(b.DataDir, rt.Host+".env") |
| 387 | if err := os.Symlink(target, link); err != nil { |
| 388 | t.Skipf("this filesystem will not make symlinks: %v", err) |
| 389 | } |
| 390 | err := tc.call(b, rt, link) |
| 391 | if !errors.Is(err, errOwnEnvSymlink) { |
| 392 | t.Fatalf("err = %v, want errOwnEnvSymlink", err) |
| 393 | } |
| 394 | if strings.Contains(err.Error(), strings.Repeat("A", 43)) { |
| 395 | t.Fatal("the refusal put the key in its error") |
| 396 | } |
| 397 | st, err := os.Stat(target) |
| 398 | if err != nil { |
| 399 | t.Fatal(err) |
| 400 | } |
| 401 | if st.Mode().Perm() != tc.mode { |
| 402 | t.Fatalf("mode of the symlink's TARGET = %o, want the untouched %o", st.Mode().Perm(), tc.mode) |
| 403 | } |
| 404 | if raw, _ := os.ReadFile(target); string(raw) != tc.target { |
| 405 | t.Fatalf("the symlink's TARGET changed: %q -> %q", tc.target, raw) |
| 406 | } |
| 407 | }) |
| 408 | } |
| 409 | } |
| 410 | |
| 411 | // A dangling link is refused too: the mint's O_CREATE must not create the |
| 412 | // file it points at. |
| 413 | func TestEnsureInstanceKeyRefusesADanglingSymlink(t *testing.T) { |
| 414 | dir := t.TempDir() |
| 415 | target := filepath.Join(t.TempDir(), "not-yet") |
| 416 | link := filepath.Join(dir, "shop.example.com.env") |
| 417 | if err := os.Symlink(target, link); err != nil { |
| 418 | t.Skipf("this filesystem will not make symlinks: %v", err) |
| 419 | } |
| 420 | if _, err := ensureInstanceKeyNoConfig(link); !errors.Is(err, errOwnEnvSymlink) { |
| 421 | t.Fatalf("err = %v, want errOwnEnvSymlink", err) |
| 422 | } |
| 423 | if _, err := os.Lstat(target); !os.IsNotExist(err) { |
| 424 | t.Fatalf("the mint created the link's target: %v", err) |
| 425 | } |
| 426 | } |
| 427 | |
| 428 | // The refusal is the open's own, not a check before it: the helper refuses |
| 429 | // a link for every flag set the ensure uses, so no Lstat-to-open window |
| 430 | // exists on unix. |
| 431 | func TestOpenOwnEnvNoFollowRefusesALinkForEveryFlagSet(t *testing.T) { |
| 432 | dir := t.TempDir() |
| 433 | target := filepath.Join(dir, "target") |
| 434 | if err := os.WriteFile(target, []byte("A=1\n"), 0o644); err != nil { |
| 435 | t.Fatal(err) |
| 436 | } |
| 437 | link := filepath.Join(dir, "h.env") |
| 438 | if err := os.Symlink(target, link); err != nil { |
| 439 | t.Skipf("this filesystem will not make symlinks: %v", err) |
| 440 | } |
| 441 | for _, flags := range []int{os.O_RDONLY, os.O_APPEND | os.O_WRONLY, os.O_APPEND | os.O_WRONLY | os.O_CREATE} { |
| 442 | perm := os.FileMode(0) |
| 443 | if flags&os.O_CREATE != 0 { |
| 444 | perm = 0o600 |
| 445 | } |
| 446 | f, err := openOwnEnvNoFollow(link, flags, perm) |
| 447 | if err == nil { |
| 448 | f.Close() |
| 449 | t.Fatalf("flags %#x: opened through a symlink", flags) |
| 450 | } |
| 451 | if !errors.Is(err, errOwnEnvSymlink) { |
| 452 | t.Fatalf("flags %#x: err = %v, want errOwnEnvSymlink", flags, err) |
| 453 | } |
| 454 | } |
| 455 | // An ordinary file still opens. |
| 456 | f, err := openOwnEnvNoFollow(target, os.O_RDONLY, 0) |
| 457 | if err != nil { |
| 458 | t.Fatal(err) |
| 459 | } |
| 460 | f.Close() |
| 461 | } |
| 462 | |
| 463 | // setInstanceKeySeams swaps the append and the before-open hook for one |
| 464 | // test, under adminTokenMu as their readers require. A nil argument |
| 465 | // leaves that seam as it is. |
| 466 | func setInstanceKeySeams(t *testing.T, write func(*os.File, []byte) (int, error), beforeOpen func(string)) { |
| 467 | t.Helper() |
| 468 | adminTokenMu.Lock() |
| 469 | oldWrite, oldHook := instanceKeyWrite, instanceKeyHooks.beforeOpen |
| 470 | if write != nil { |
| 471 | instanceKeyWrite = write |
| 472 | } |
| 473 | if beforeOpen != nil { |
| 474 | instanceKeyHooks.beforeOpen = beforeOpen |
| 475 | } |
| 476 | adminTokenMu.Unlock() |
| 477 | t.Cleanup(func() { |
| 478 | adminTokenMu.Lock() |
| 479 | instanceKeyWrite, instanceKeyHooks.beforeOpen = oldWrite, oldHook |
| 480 | adminTokenMu.Unlock() |
| 481 | }) |
| 482 | } |
| 483 | |
| 484 | // A mint whose append fails part way (a full disk) is undone: the file is |
| 485 | // byte for byte what it was, so no short key line survives to be read as |
| 486 | // present, and the next call mints a full key. |
| 487 | func TestEnsureInstanceKeyUndoesAPartialAppend(t *testing.T) { |
| 488 | for _, tc := range []struct { |
| 489 | name string |
| 490 | before *string // nil: no file before the call |
| 491 | write func(f *os.File, b []byte) (int, error) |
| 492 | }{ |
| 493 | {"nothing written, ENOSPC", ptr("A=1\n"), func(*os.File, []byte) (int, error) { return 0, syscall.ENOSPC }}, |
| 494 | {"short key line, ENOSPC", ptr("A=1\n"), func(f *os.File, b []byte) (int, error) { |
| 495 | n, _ := f.Write(b[:len(InstanceKeyVar)+5]) |
| 496 | return n, syscall.ENOSPC |
| 497 | }}, |
| 498 | {"short after the fresh-line prefix, ENOSPC", ptr("A=1"), func(f *os.File, b []byte) (int, error) { |
| 499 | n, _ := f.Write(b[:3]) |
| 500 | return n, syscall.ENOSPC |
| 501 | }}, |
| 502 | {"short with no error", ptr("A=1\n"), func(f *os.File, b []byte) (int, error) { |
| 503 | return f.Write(b[:len(b)-1]) |
| 504 | }}, |
| 505 | {"new file, short key line", nil, func(f *os.File, b []byte) (int, error) { |
| 506 | n, _ := f.Write(b[:10]) |
| 507 | return n, syscall.ENOSPC |
| 508 | }}, |
| 509 | } { |
| 510 | t.Run(tc.name, func(t *testing.T) { |
| 511 | probe := probeInstanceKeySync(t) |
| 512 | p := filepath.Join(t.TempDir(), "shop.example.com.env") |
| 513 | if tc.before != nil { |
| 514 | if err := os.WriteFile(p, []byte(*tc.before), 0o600); err != nil { |
| 515 | t.Fatal(err) |
| 516 | } |
| 517 | } |
| 518 | setInstanceKeySeams(t, tc.write, nil) |
| 519 | if _, err := ensureInstanceKeyNoConfig(p); err == nil { |
| 520 | t.Fatal("a mint whose append failed reported success") |
| 521 | } |
| 522 | raw, err := os.ReadFile(p) |
| 523 | if err != nil { |
| 524 | t.Fatal(err) |
| 525 | } |
| 526 | want := "" |
| 527 | if tc.before != nil { |
| 528 | want = *tc.before |
| 529 | } |
| 530 | if string(raw) != want { |
| 531 | t.Fatalf("after the failed append the file is %q, want it undone to %q", raw, want) |
| 532 | } |
| 533 | if probe.file != 1 { |
| 534 | t.Fatalf("the undo synced the file %d times, want once", probe.file) |
| 535 | } |
| 536 | setInstanceKeySeams(t, func(f *os.File, b []byte) (int, error) { return f.Write(b) }, nil) |
| 537 | got, err := ensureInstanceKeyNoConfig(p) |
| 538 | if err != nil || got != InstanceKeyMinted { |
| 539 | t.Fatalf("next ensure = %v, %v; want minted", got, err) |
| 540 | } |
| 541 | raw, _ = os.ReadFile(p) |
| 542 | keys := keyLines(raw) |
| 543 | if len(keys) != 1 || !wellFormedInstanceKey(keys[0]) { |
| 544 | t.Fatalf("file = %q: want one full key", raw) |
| 545 | } |
| 546 | if !strings.HasPrefix(string(raw), want) { |
| 547 | t.Fatalf("file = %q: the original bytes were lost", raw) |
| 548 | } |
| 549 | }) |
| 550 | } |
| 551 | } |
| 552 | |
| 553 | func ptr(s string) *string { return &s } |
| 554 | |
| 555 | // The append open refuses a link on its own, not only because the read |
| 556 | // before it did: a link swapped in after a clean read (the hook) is |
| 557 | // refused on the mint, the existing-key and the backfill paths, and its |
| 558 | // target is left exactly as it was. |
| 559 | func TestEnsureInstanceKeyOpenRefusesALinkSwappedInAfterTheRead(t *testing.T) { |
| 560 | wellFormed := InstanceKeyVar + "=" + strings.Repeat("A", 43) + "=\n" |
| 561 | for _, tc := range []struct { |
| 562 | name string |
| 563 | own string |
| 564 | call func(b *ExecBackend, rt registry.Route, path string) error |
| 565 | }{ |
| 566 | {"mint", "A=1\n", func(_ *ExecBackend, _ registry.Route, path string) error { |
| 567 | _, err := ensureInstanceKeyNoConfig(path) |
| 568 | return err |
| 569 | }}, |
| 570 | {"existing key", wellFormed, func(_ *ExecBackend, _ registry.Route, path string) error { |
| 571 | _, err := ensureInstanceKeyNoConfig(path) |
| 572 | return err |
| 573 | }}, |
| 574 | {"backfill mint", "A=1\n", func(b *ExecBackend, rt registry.Route, _ string) error { |
| 575 | _, err := b.BackfillInstanceKey(t.Context(), rt) |
| 576 | return err |
| 577 | }}, |
| 578 | } { |
| 579 | t.Run(tc.name, func(t *testing.T) { |
| 580 | b := &ExecBackend{DataDir: t.TempDir()} |
| 581 | rt := registry.Route{Host: "shop.example.com", Socket: "/run/carlos/shop.sock"} |
| 582 | path := filepath.Join(b.DataDir, rt.Host+".env") |
| 583 | if err := os.WriteFile(path, []byte(tc.own), 0o600); err != nil { |
| 584 | t.Fatal(err) |
| 585 | } |
| 586 | target := filepath.Join(t.TempDir(), "someone-elses-file") |
| 587 | const targetBody = "B=2\n" |
| 588 | if err := os.WriteFile(target, []byte(targetBody), 0o644); err != nil { |
| 589 | t.Fatal(err) |
| 590 | } |
| 591 | if err := os.Chmod(target, 0o644); err != nil { |
| 592 | t.Fatal(err) |
| 593 | } |
| 594 | swapped := 0 |
| 595 | setInstanceKeySeams(t, nil, func(p string) { |
| 596 | if p != path { |
| 597 | return |
| 598 | } |
| 599 | swapped++ |
| 600 | if err := os.Remove(p); err != nil { |
| 601 | t.Error(err) |
| 602 | } |
| 603 | if err := os.Symlink(target, p); err != nil { |
| 604 | t.Error(err) |
| 605 | } |
| 606 | }) |
| 607 | err := tc.call(b, rt, path) |
| 608 | if swapped != 1 { |
| 609 | t.Fatalf("the hook ran %d times, want once", swapped) |
| 610 | } |
| 611 | if !errors.Is(err, errOwnEnvSymlink) { |
| 612 | t.Fatalf("err = %v, want errOwnEnvSymlink", err) |
| 613 | } |
| 614 | st, err := os.Stat(target) |
| 615 | if err != nil { |
| 616 | t.Fatal(err) |
| 617 | } |
| 618 | if st.Mode().Perm() != 0o644 { |
| 619 | t.Fatalf("mode of the symlink's TARGET = %o, want the untouched 644", st.Mode().Perm()) |
| 620 | } |
| 621 | if raw, _ := os.ReadFile(target); string(raw) != targetBody { |
| 622 | t.Fatalf("the symlink's TARGET changed: %q", raw) |
| 623 | } |
| 624 | }) |
| 625 | } |
| 626 | } |
| 627 | |
| 628 | // A refusal names the path once. |
| 629 | func TestEnsureInstanceKeySymlinkErrorNamesThePathOnce(t *testing.T) { |
| 630 | dir := t.TempDir() |
| 631 | target := filepath.Join(dir, "target") |
| 632 | if err := os.WriteFile(target, []byte("A=1\n"), 0o644); err != nil { |
| 633 | t.Fatal(err) |
| 634 | } |
| 635 | link := filepath.Join(dir, "shop.example.com.env") |
| 636 | if err := os.Symlink(target, link); err != nil { |
| 637 | t.Skipf("this filesystem will not make symlinks: %v", err) |
| 638 | } |
| 639 | _, err := ensureInstanceKeyNoConfig(link) |
| 640 | if !errors.Is(err, errOwnEnvSymlink) { |
| 641 | t.Fatalf("err = %v, want errOwnEnvSymlink", err) |
| 642 | } |
| 643 | if n := strings.Count(err.Error(), link); n != 1 { |
| 644 | t.Fatalf("error %q names the path %d times, want once", err, n) |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | // startRecorded starts host on b with the recording binary and returns the |
| 649 | // environment the child saw, or nil when no child ran (none recorded within a |
| 650 | // second, or none is left running that could). The route's socket is a temp |
| 651 | // path, so the row classifies as foreign; startRecordedRoute takes the row as |
| 652 | // given. |
| 653 | func startRecorded(t *testing.T, b *ExecBackend, host string) (map[string]string, error) { |
| 654 | t.Helper() |
| 655 | return startRecordedRoute(t, b, registry.Route{Host: host, Kind: "instance", Socket: tmpSock(t)}) |
| 656 | } |
| 657 | |
| 658 | func startRecordedRoute(t *testing.T, b *ExecBackend, rt registry.Route) (map[string]string, error) { |
| 659 | t.Helper() |
| 660 | _, envPath := stageRecordingLiveFullEnv(t, b, rt.Host, "v1") |
| 661 | ctx, cancel := recordedCtx(2*time.Second, envPath) |
| 662 | defer cancel() |
| 663 | proc, _, err := b.Start(ctx, rt) // the fake binary never binds: the health wait fails, as in the sibling tests |
| 664 | if proc != nil { |
| 665 | defer proc.Stop(context.Background()) |
| 666 | } |
| 667 | raw := awaitChildFile(envPath, time.Second) |
| 668 | if raw == "" { |
| 669 | return nil, err |
| 670 | } |
| 671 | child := map[string]string{} |
| 672 | for _, line := range strings.Split(raw, "\n") { |
| 673 | if k, v, ok := strings.Cut(line, "="); ok { |
| 674 | child[k] = v |
| 675 | } |
| 676 | } |
| 677 | return child, err |
| 678 | } |
| 679 | |
| 680 | // Start hands the child the key it minted, and a second Start (a restart, a |
| 681 | // redeploy) hands it the same one. |
| 682 | func TestExecBackendStartHandsTheChildTheMintedInstanceKey(t *testing.T) { |
| 683 | b := testExecBackend(t) |
| 684 | host := "shop.example.com" |
| 685 | first, _ := startRecorded(t, b, host) |
| 686 | minted, err := readEnvFile(filepath.Join(b.DataDir, host+".env")) |
| 687 | if err != nil { |
| 688 | t.Fatal(err) |
| 689 | } |
| 690 | if first == nil || !wellFormedInstanceKey(minted[InstanceKeyVar]) || first[InstanceKeyVar] != minted[InstanceKeyVar] { |
| 691 | t.Fatal("the child was not handed the key minted into <host>.env") |
| 692 | } |
| 693 | if second, _ := startRecorded(t, b, host); second == nil || second[InstanceKeyVar] != minted[InstanceKeyVar] { |
| 694 | t.Fatal("a second start handed the child a different key") |
| 695 | } |
| 696 | } |
| 697 | |
| 698 | // Task 3 review (M4): the tests above launch through the foreign arm, since |
| 699 | // a temp socket is neither default. A row that classifies as flat goes |
| 700 | // through the same spawn body, and this pins it end to end: Start mints |
| 701 | // into <DataDir>/<host>.env and hands the child that key. |
| 702 | func TestExecBackendStartHandsAFlatRowsChildTheMintedInstanceKey(t *testing.T) { |
| 703 | b := testExecBackend(t) |
| 704 | b.RunDirOverride = shortRunDir(t) |
| 705 | host := "shop.example.com" |
| 706 | rt := flatGenRoute(b, host) |
| 707 | if l, why := b.Layout(rt); l != LayoutFlat { |
| 708 | t.Fatalf("fixture classifies %q (%s), want flat", l, why) |
| 709 | } |
| 710 | child, _ := startRecordedRoute(t, b, rt) |
| 711 | minted, err := readEnvFile(filepath.Join(b.DataDir, host+".env")) |
| 712 | if err != nil { |
| 713 | t.Fatal(err) |
| 714 | } |
| 715 | if child == nil || !wellFormedInstanceKey(minted[InstanceKeyVar]) || child[InstanceKeyVar] != minted[InstanceKeyVar] { |
| 716 | t.Fatal("a flat row's child was not handed the key minted into <host>.env") |
| 717 | } |
| 718 | } |
| 719 | |
| 720 | // Finding 2 through a launch: while the key's sync fails, Start fails and |
| 721 | // no child runs. Once the sync succeeds, the child is handed the key the |
| 722 | // failed attempt wrote, never a second one. |
| 723 | func TestExecBackendStartStartsNoChildUntilTheKeyIsSynced(t *testing.T) { |
| 724 | for _, tc := range []struct { |
| 725 | name string |
| 726 | fail func(*syncProbe, error) |
| 727 | }{ |
| 728 | {"file sync", func(p *syncProbe, err error) { p.fileErr = err }}, |
| 729 | {"directory sync", func(p *syncProbe, err error) { p.dirErr = err }}, |
| 730 | } { |
| 731 | t.Run(tc.name, func(t *testing.T) { |
| 732 | probe := probeInstanceKeySync(t) |
| 733 | b := testExecBackend(t) |
| 734 | host := "shop.example.com" |
| 735 | tc.fail(probe, errors.New("injected: EIO")) |
| 736 | for i := 0; i < 2; i++ { |
| 737 | child, err := startRecorded(t, b, host) |
| 738 | if err == nil || !strings.Contains(err.Error(), "instance key for "+host) { |
| 739 | t.Fatalf("start %d with a failing sync: err = %v, want the instance key's error", i, err) |
| 740 | } |
| 741 | if child != nil { |
| 742 | t.Fatalf("start %d ran a child before the key was durable", i) |
| 743 | } |
| 744 | } |
| 745 | written, _ := readEnvFile(filepath.Join(b.DataDir, host+".env")) |
| 746 | tc.fail(probe, nil) |
| 747 | child, _ := startRecorded(t, b, host) |
| 748 | if child == nil || child[InstanceKeyVar] == "" || child[InstanceKeyVar] != written[InstanceKeyVar] { |
| 749 | t.Fatal("after the sync succeeded the child was not handed the key the failed attempt wrote") |
| 750 | } |
| 751 | }) |
| 752 | } |
| 753 | } |
| 754 | |
| 755 | // Finding 4 through a launch: <host>.env and <host>.config.env together. |
| 756 | // An empty override never hands the child an empty key while the |
| 757 | // configuration supplies one, and no key is minted over the configured one. |
| 758 | // With neither supplying one, the child gets the minted key. |
| 759 | func TestExecBackendStartNeverHandsTheChildAnEmptyKeyWhileOneExists(t *testing.T) { |
| 760 | tok := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("c", 64) + "\n" |
| 761 | for _, tc := range []struct { |
| 762 | name, own, config string |
| 763 | configured bool |
| 764 | }{ |
| 765 | {"empty override, configured key", tok + InstanceKeyVar + "=\n", InstanceKeyVar + "=from-config\n", true}, |
| 766 | {"no override, configured key", tok, InstanceKeyVar + "=from-config\n", true}, |
| 767 | {"empty override, empty configuration", tok + InstanceKeyVar + "=\n", InstanceKeyVar + "=\n", false}, |
| 768 | } { |
| 769 | t.Run(tc.name, func(t *testing.T) { |
| 770 | b := testExecBackend(t) |
| 771 | host := "shop.example.com" |
| 772 | own := filepath.Join(b.DataDir, host+".env") |
| 773 | if err := os.WriteFile(own, []byte(tc.own), 0o600); err != nil { |
| 774 | t.Fatal(err) |
| 775 | } |
| 776 | if err := os.WriteFile(filepath.Join(b.DataDir, host+".config.env"), []byte(tc.config), 0o600); err != nil { |
| 777 | t.Fatal(err) |
| 778 | } |
| 779 | child, _ := startRecorded(t, b, host) |
| 780 | if child == nil { |
| 781 | t.Fatal("no child ran") |
| 782 | } |
| 783 | raw, _ := os.ReadFile(own) |
| 784 | if tc.configured { |
| 785 | if child[InstanceKeyVar] != "from-config" { |
| 786 | t.Fatalf("the child was handed %q, want the configured key", child[InstanceKeyVar]) |
| 787 | } |
| 788 | if string(raw) != tc.own { |
| 789 | t.Fatal("a key was minted over the configured one") |
| 790 | } |
| 791 | return |
| 792 | } |
| 793 | env, _ := readEnvFile(own) |
| 794 | if !wellFormedInstanceKey(child[InstanceKeyVar]) || child[InstanceKeyVar] != env[InstanceKeyVar] { |
| 795 | t.Fatal("the child was not handed the minted key") |
| 796 | } |
| 797 | }) |
| 798 | } |
| 799 | } |
| 800 | |
| 801 | // Coverage (Astra round 1): the first process of a fresh flat instance is |
| 802 | // often its seed. The seed mints the key, sees it, and the first start |
| 803 | // hands the instance the same one. |
| 804 | func TestExecBackendSeedMintsTheKeyForAFreshFlatInstance(t *testing.T) { |
| 805 | b := testExecBackend(t) |
| 806 | host := "seeded.example.com" |
| 807 | _, envPath := stageRecordingLiveFullEnv(t, b, host, "v1") |
| 808 | if err := b.Seed(context.Background(), registry.Route{Host: host, Kind: "instance", Seed: `["seed"]`}); err != nil { |
| 809 | t.Fatalf("Seed: %v", err) |
| 810 | } |
| 811 | seed := map[string]string{} |
| 812 | for _, line := range strings.Split(waitForFile(t, envPath, time.Second), "\n") { |
| 813 | if k, v, ok := strings.Cut(line, "="); ok { |
| 814 | seed[k] = v |
| 815 | } |
| 816 | } |
| 817 | own, _ := readEnvFile(filepath.Join(b.DataDir, host+".env")) |
| 818 | if !wellFormedInstanceKey(own[InstanceKeyVar]) || seed[InstanceKeyVar] != own[InstanceKeyVar] { |
| 819 | t.Fatal("the seed did not mint the key and see it") |
| 820 | } |
| 821 | if child, _ := startRecorded(t, b, host); child == nil || child[InstanceKeyVar] != own[InstanceKeyVar] { |
| 822 | t.Fatal("the first start handed the instance a different key from its seed's") |
| 823 | } |
| 824 | } |
| 825 | |
| 826 | // Coverage (Astra round 1): a hibernation wake restores the database and |
| 827 | // leaves <host>.env alone, so the start after the restore hands the |
| 828 | // instance the key it had. |
| 829 | func TestHibernationRestoreKeepsTheInstanceKey(t *testing.T) { |
| 830 | b := testExecBackend(t) |
| 831 | host := "sleepy.example.com" |
| 832 | if child, _ := startRecorded(t, b, host); child == nil { |
| 833 | t.Fatal("no child ran") |
| 834 | } |
| 835 | envPath := filepath.Join(b.DataDir, host+".env") |
| 836 | before, _ := os.ReadFile(envPath) |
| 837 | key := parseEnvBytes(before)[InstanceKeyVar] |
| 838 | if !wellFormedInstanceKey(key) { |
| 839 | t.Fatal("the first start minted no key") |
| 840 | } |
| 841 | old := restoreReplica |
| 842 | restoreReplica = func(_ context.Context, dbPath, bucket, prefix, region string) (bool, error) { |
| 843 | return true, writeMinimalSQLite(dbPath) |
| 844 | } |
| 845 | t.Cleanup(func() { restoreReplica = old }) |
| 846 | _ = os.Remove(filepath.Join(b.DataDir, host+".db")) // hibernated: the database is gone |
| 847 | rt := registry.Route{Host: host, Kind: "instance", Socket: tmpSock(t)} |
| 848 | if fresh, err := b.Restore(context.Background(), rt); err != nil || fresh { |
| 849 | t.Fatalf("Restore = fresh %v, %v; want a restored database", fresh, err) |
| 850 | } |
| 851 | if after, _ := os.ReadFile(envPath); !bytes.Equal(before, after) { |
| 852 | t.Fatal("the restore changed <host>.env") |
| 853 | } |
| 854 | if child, _ := startRecorded(t, b, host); child == nil || child[InstanceKeyVar] != key { |
| 855 | t.Fatal("the start after a restore handed the instance a different key") |
| 856 | } |
| 857 | } |
| 858 | |
| 859 | // refuseAfterFirstClassification makes b classify a row as it does today |
| 860 | // the first time it is asked and as layout (fenced or conflict) every time |
| 861 | // after: a launch dispatcher's own check passes, and the launch body's |
| 862 | // pathsFor sees a layout journal or a colliding peer that appeared since. |
| 863 | func refuseAfterFirstClassification(b *ExecBackend, layout Layout) { |
| 864 | var calls atomic.Int32 |
| 865 | switch layout { |
| 866 | case LayoutFenced: |
| 867 | b.LayoutFacts.Journal = func(string) bool { return calls.Add(1) > 1 } |
| 868 | case LayoutConflict: |
| 869 | b.LayoutFacts.SameID = func(a, b string) bool { return true } |
| 870 | b.LayoutFacts.Peers = func() ([]registry.Route, error) { |
| 871 | if calls.Add(1) == 1 { |
| 872 | return nil, nil |
| 873 | } |
| 874 | return []registry.Route{{Host: "other.example.com", Kind: "instance", Hibernate: true, Socket: "/nonexistent/other.sock"}}, nil |
| 875 | } |
| 876 | } |
| 877 | } |
| 878 | |
| 879 | // Review (Task 2): the mint refuses a fenced or conflicting row on the |
| 880 | // paths the launch body hands it, not only on the dispatcher's earlier |
| 881 | // classification. A row that becomes fenced or conflicting between Start's |
| 882 | // or Seed's check and the body's own pathsFor fails the launch with the |
| 883 | // key's error, gets no key, and runs no child. |
| 884 | func TestLaunchBodiesMintNoKeyForARowRefusedSinceTheDispatch(t *testing.T) { |
| 885 | for _, layout := range []Layout{LayoutFenced, LayoutConflict} { |
| 886 | for _, launch := range []string{"start", "seed"} { |
| 887 | t.Run(string(layout)+"/"+launch, func(t *testing.T) { |
| 888 | b := testExecBackend(t) |
| 889 | host := "shop.example.com" |
| 890 | _, envPath := stageRecordingLiveFullEnv(t, b, host, "v1") |
| 891 | refuseAfterFirstClassification(b, layout) |
| 892 | rt := registry.Route{Host: host, Kind: "instance", Socket: tmpSock(t), Seed: `["seed"]`} |
| 893 | ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second) |
| 894 | defer cancel() |
| 895 | var err error |
| 896 | if launch == "start" { |
| 897 | var proc Proc |
| 898 | proc, _, err = b.Start(ctx, rt) |
| 899 | if proc != nil { |
| 900 | defer proc.Stop(context.Background()) |
| 901 | } |
| 902 | } else { |
| 903 | err = b.Seed(ctx, rt) |
| 904 | } |
| 905 | if err == nil || !strings.Contains(err.Error(), "instance key for "+host) || !strings.Contains(err.Error(), "not minted") { |
| 906 | t.Fatalf("%s of a row %s since the dispatch: err = %v, want the key's refusal", launch, layout, err) |
| 907 | } |
| 908 | // The host is named once, by the wrapper. (A fenced reason |
| 909 | // also names it inside the journal path, so the count is |
| 910 | // of the "<host>: " prefix.) |
| 911 | if n := strings.Count(err.Error(), host+": "); n != 1 { |
| 912 | t.Fatalf("the refusal names the host %d times: %v", n, err) |
| 913 | } |
| 914 | raw, _ := os.ReadFile(filepath.Join(b.DataDir, host+".env")) |
| 915 | if n := len(keyLines(raw)); n != 0 { |
| 916 | t.Fatalf("a %s row was minted %d key(s)", layout, n) |
| 917 | } |
| 918 | if _, serr := os.Stat(envPath); !errors.Is(serr, os.ErrNotExist) { |
| 919 | t.Fatalf("a child ran for a %s row (stat %v)", layout, serr) |
| 920 | } |
| 921 | }) |
| 922 | } |
| 923 | } |
| 924 | } |
| 925 | |
| 926 | // The refusal itself, for every layout pathsFor can report it on: a flat |
| 927 | // row and a nested one, each fenced and conflicting. Nothing is written in |
| 928 | // the data directory or the instance directory. (A nested row's prepare |
| 929 | // cannot reach the mint in a test once fenced: its fenced paths name |
| 930 | // /run/carlos, which prepare creates first. prepareAt hands the mint its |
| 931 | // paths unchanged, so this is the refusal it gets.) |
| 932 | func TestEnsureInstanceKeyForRefusesFencedAndConflictRows(t *testing.T) { |
| 933 | for _, nested := range []bool{false, true} { |
| 934 | for _, layout := range []Layout{LayoutFenced, LayoutConflict} { |
| 935 | name := "flat" |
| 936 | if nested { |
| 937 | name = "nested" |
| 938 | } |
| 939 | t.Run(name+"/"+string(layout), func(t *testing.T) { |
| 940 | var b *ExecBackend |
| 941 | var rt registry.Route |
| 942 | if nested { |
| 943 | b, rt = nestedBackend(t, &fakeControl{inv: "inv1"}, &fakeUnit{}) |
| 944 | } else { |
| 945 | b = testExecBackend(t) |
| 946 | rt = registry.Route{Host: "shop.example.com", Kind: "instance", Hibernate: true, |
| 947 | DBPath: provision.DefaultDBPath(b.DataDir, "shop.example.com"), Socket: filepath.Join(provision.SocketDir, "shop.example.com.sock")} |
| 948 | } |
| 949 | if l, _ := b.Layout(rt); (nested && l != LayoutNested) || (!nested && l != LayoutFlat) { |
| 950 | t.Fatalf("fixture classifies as %s", l) |
| 951 | } |
| 952 | refuseAfterFirstClassification(b, layout) |
| 953 | _, _ = b.Layout(rt) // the dispatcher's check |
| 954 | p := b.pathsFor(rt) |
| 955 | if p.layout != layout { |
| 956 | t.Fatalf("paths classify as %s, want %s", p.layout, layout) |
| 957 | } |
| 958 | if _, err := b.ensureInstanceKeyFor(t.Context(), rt, p); err == nil || !strings.Contains(err.Error(), "not minted") { |
| 959 | t.Fatalf("ensure for a %s row: err = %v, want a refusal", layout, err) |
| 960 | } else if strings.HasPrefix(err.Error(), rt.Host) { |
| 961 | t.Fatalf("the refusal names the host itself (%v); its callers do", err) |
| 962 | } |
| 963 | for _, f := range []string{ |
| 964 | filepath.Join(b.DataDir, rt.Host+".env"), |
| 965 | filepath.Join(provision.InstanceDir(b.DataDir, rt.Host), rt.Host+".env"), |
| 966 | } { |
| 967 | if _, err := os.Lstat(f); !errors.Is(err, os.ErrNotExist) { |
| 968 | t.Fatalf("%s exists after a refused ensure (lstat %v)", f, err) |
| 969 | } |
| 970 | } |
| 971 | }) |
| 972 | } |
| 973 | } |
| 974 | } |
| 975 | |
| 976 | // The backfill refuses the same rows (instancekey.go BackfillInstanceKey): |
| 977 | // an existing env file without a key is left byte for byte. |
| 978 | func TestBackfillInstanceKeyRefusesFencedAndConflictRows(t *testing.T) { |
| 979 | for _, layout := range []Layout{LayoutFenced, LayoutConflict} { |
| 980 | t.Run(string(layout), func(t *testing.T) { |
| 981 | b := testExecBackend(t) |
| 982 | host := "shop.example.com" |
| 983 | rt := registry.Route{Host: host, Kind: "instance", Hibernate: true, |
| 984 | DBPath: provision.DefaultDBPath(b.DataDir, host), Socket: filepath.Join(provision.SocketDir, host+".sock")} |
| 985 | own := filepath.Join(b.DataDir, host+".env") |
| 986 | tok := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("d", 64) + "\n" |
| 987 | if err := os.WriteFile(own, []byte(tok), 0o600); err != nil { |
| 988 | t.Fatal(err) |
| 989 | } |
| 990 | refuseAfterFirstClassification(b, layout) |
| 991 | _, _ = b.Layout(rt) |
| 992 | if _, err := b.BackfillInstanceKey(t.Context(), rt); err == nil || !strings.Contains(err.Error(), "not backfilled") { |
| 993 | t.Fatalf("backfill of a %s row: err = %v, want a refusal", layout, err) |
| 994 | } else if strings.HasPrefix(err.Error(), host) { |
| 995 | t.Fatalf("the refusal names the host itself (%v); the agent's backfill does", err) |
| 996 | } |
| 997 | if raw, _ := os.ReadFile(own); string(raw) != tok { |
| 998 | t.Fatalf("the backfill changed a %s row's env file", layout) |
| 999 | } |
| 1000 | }) |
| 1001 | } |
| 1002 | } |
| 1003 | |
| 1004 | // v0.26.1 hotfix: the agent's backfill now runs beside the activator |
| 1005 | // rather than before it, so a backfill and a start's own mint can reach |
| 1006 | // the same host at once. Both read the files and ask a slow bucket outside |
| 1007 | // adminTokenMu, and both are held there until each has done so, so each |
| 1008 | // decides from a first read that found no key. They still decide and |
| 1009 | // append under the lock, create-only, so the host ends with exactly one |
| 1010 | // key, appended whole after the bytes that were there, and exactly one of |
| 1011 | // the two reports it minted. |
| 1012 | func TestBackfillAndAStartsMintRacingOnOneHostWriteOneKey(t *testing.T) { |
| 1013 | const rounds = 40 |
| 1014 | tok := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("b", 64) + "\n" |
| 1015 | for i := range rounds { |
| 1016 | b := &ExecBackend{DataDir: t.TempDir()} |
| 1017 | b.ConfiguredKey = func(ctx context.Context, _ registry.Route) (string, error) { |
| 1018 | select { |
| 1019 | case <-time.After(5 * time.Millisecond): // a slow bucket that sets no key |
| 1020 | return "", nil |
| 1021 | case <-ctx.Done(): |
| 1022 | return "", ctx.Err() |
| 1023 | } |
| 1024 | } |
| 1025 | rt := registry.Route{Host: "shop.example.com", Kind: "instance", Status: "ready", Socket: "/run/carlos/shop.sock"} |
| 1026 | own := filepath.Join(b.DataDir, rt.Host+".env") |
| 1027 | if err := os.WriteFile(own, []byte(tok), 0o600); err != nil { |
| 1028 | t.Fatal(err) |
| 1029 | } |
| 1030 | // Both callers pass phase one before either takes the lock. |
| 1031 | var phase sync.WaitGroup |
| 1032 | phase.Add(2) |
| 1033 | instanceKeyHooks.betweenPhases = func(string) { phase.Done(); phase.Wait() } |
| 1034 | var wg sync.WaitGroup |
| 1035 | var backfilled, started InstanceKeyOutcome |
| 1036 | var berr, serr error |
| 1037 | wg.Add(2) |
| 1038 | go func() { |
| 1039 | defer wg.Done() |
| 1040 | backfilled, berr = b.BackfillInstanceKey(t.Context(), rt) |
| 1041 | }() |
| 1042 | go func() { |
| 1043 | defer wg.Done() |
| 1044 | started, serr = b.ensureInstanceKeyFor(t.Context(), rt, b.pathsFor(rt)) |
| 1045 | }() |
| 1046 | wg.Wait() |
| 1047 | instanceKeyHooks.betweenPhases = nil |
| 1048 | if berr != nil || serr != nil { |
| 1049 | t.Fatalf("round %d: backfill %v, %v; start %v, %v", i, backfilled, berr, started, serr) |
| 1050 | } |
| 1051 | minted := 0 |
| 1052 | for _, o := range []InstanceKeyOutcome{backfilled, started} { |
| 1053 | switch o { |
| 1054 | case InstanceKeyMinted: |
| 1055 | minted++ |
| 1056 | case InstanceKeyPresent: |
| 1057 | default: |
| 1058 | t.Fatalf("round %d: outcome %v, want minted or present", i, o) |
| 1059 | } |
| 1060 | } |
| 1061 | raw, _ := os.ReadFile(own) |
| 1062 | keys := keyLines(raw) |
| 1063 | if minted != 1 || len(keys) != 1 || !wellFormedInstanceKey(keys[0]) || string(raw) != tok+InstanceKeyVar+"="+keys[0]+"\n" { |
| 1064 | t.Fatalf("round %d: %d minted, %d keys, file %d bytes; want one whole key appended once", i, minted, len(keys), len(raw)) |
| 1065 | } |
| 1066 | } |
| 1067 | } |
| 1068 | |