| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "context" |
| 6 | "crypto" |
| 7 | "crypto/rand" |
| 8 | "crypto/rsa" |
| 9 | "crypto/sha256" |
| 10 | "crypto/x509" |
| 11 | "crypto/x509/pkix" |
| 12 | "encoding/base64" |
| 13 | "encoding/json" |
| 14 | "errors" |
| 15 | "math/big" |
| 16 | "net/http" |
| 17 | "net/http/httptest" |
| 18 | "os" |
| 19 | "strings" |
| 20 | "sync" |
| 21 | "testing" |
| 22 | "time" |
| 23 | |
| 24 | "github.com/carlosframework/platform/internal/boxid" |
| 25 | "github.com/carlosframework/platform/internal/hostactors" |
| 26 | "github.com/carlosframework/platform/internal/lease/leasetest" |
| 27 | "github.com/carlosframework/platform/internal/release" |
| 28 | "github.com/carlosframework/platform/internal/status" |
| 29 | ) |
| 30 | |
| 31 | // carlos system box publish-record, through the production wiring: the |
| 32 | // real S3Store (openDeployStoreEnv over CARLOS_DEPLOYMENT_BUCKET and |
| 33 | // CARLOS_S3_ENDPOINT) against leasetest's S3 server, which enforces |
| 34 | // If-None-Match: * itself, behind the dev bucket's box_records policy; |
| 35 | // the real IMDS reader over fakeIMDS (boxrecord_test.go); and enrollment's |
| 36 | // own verifier over documents a test AWS signs. |
| 37 | |
| 38 | const ( |
| 39 | publishMachineID = "0123456789abcdef0123456789abcdef" |
| 40 | publishBucket = "deploy" |
| 41 | publishAgentUID = 999 |
| 42 | publishAccount = "111122223333" |
| 43 | publishRegion = "eu-west-1" |
| 44 | publishInstance = "i-0abc" |
| 45 | ) |
| 46 | |
| 47 | // testAWS stands in for AWS's instance-identity signer the way boxid's |
| 48 | // own tests do (boxid/internal/signtest, which this package cannot |
| 49 | // import): one RSA key, a self-signed certificate for it, and PKCS #1 |
| 50 | // v1.5 SHA-256 signatures over identity documents. |
| 51 | type testAWS struct { |
| 52 | key *rsa.PrivateKey |
| 53 | cert *x509.Certificate |
| 54 | } |
| 55 | |
| 56 | func newTestAWS(t testing.TB) *testAWS { |
| 57 | t.Helper() |
| 58 | key, err := rsa.GenerateKey(rand.Reader, 2048) |
| 59 | if err != nil { |
| 60 | t.Fatal(err) |
| 61 | } |
| 62 | tmpl := &x509.Certificate{SerialNumber: big.NewInt(1), Subject: pkix.Name{CommonName: "test"}, |
| 63 | NotBefore: time.Now().Add(-time.Hour), NotAfter: time.Now().Add(time.Hour)} |
| 64 | der, err := x509.CreateCertificate(rand.Reader, tmpl, tmpl, &key.PublicKey, key) |
| 65 | if err != nil { |
| 66 | t.Fatal(err) |
| 67 | } |
| 68 | cert, err := x509.ParseCertificate(der) |
| 69 | if err != nil { |
| 70 | t.Fatal(err) |
| 71 | } |
| 72 | return &testAWS{key: key, cert: cert} |
| 73 | } |
| 74 | |
| 75 | // identityDoc is an identity document as IMDS serves one, with the two |
| 76 | // fields that change over an instance's life. |
| 77 | func identityDoc(account, region, instance, pendingTime, instanceType string) string { |
| 78 | return "{\n \"accountId\" : \"" + account + "\",\n \"instanceId\" : \"" + instance + "\",\n \"instanceType\" : \"" + instanceType + |
| 79 | "\",\n \"pendingTime\" : \"" + pendingTime + "\",\n \"region\" : \"" + region + "\"\n}\n" |
| 80 | } |
| 81 | |
| 82 | func (a *testAWS) sign(t testing.TB, doc string) string { |
| 83 | t.Helper() |
| 84 | sum := sha256.Sum256([]byte(doc)) |
| 85 | sig, err := rsa.SignPKCS1v15(rand.Reader, a.key, crypto.SHA256, sum[:]) |
| 86 | if err != nil { |
| 87 | t.Fatal(err) |
| 88 | } |
| 89 | return base64.StdEncoding.EncodeToString(sig) |
| 90 | } |
| 91 | |
| 92 | // certs certifies eu-west-1, this box's region, and eu-west-2, so a record |
| 93 | // signed in another supported region verifies and is still refused. |
| 94 | func (a *testAWS) certs() (map[string]*x509.Certificate, error) { |
| 95 | return map[string]*x509.Certificate{"eu-west-1": a.cert, "eu-west-2": a.cert}, nil |
| 96 | } |
| 97 | |
| 98 | func (a *testAWS) record(t testing.TB, box, doc, writtenAt string) boxid.Record { |
| 99 | return boxid.Record{V: 1, Box: box, IdentityDocument: base64.StdEncoding.EncodeToString([]byte(doc)), |
| 100 | IdentitySignature: a.sign(t, doc), WrittenAt: writtenAt} |
| 101 | } |
| 102 | |
| 103 | // s3Request is one request the fake S3 saw. |
| 104 | type s3Request struct{ Method, Key, IfNoneMatch, IfMatch string } |
| 105 | |
| 106 | // boxRecordsBucket is the fake deployment bucket: leasetest's versioned S3 |
| 107 | // server, behind the dev bucket's deny-only box_records policy. Under |
| 108 | // control/boxes/ a PUT without If-None-Match: * is AccessDenied |
| 109 | // (BoxRecordsAreCreateOnly) and so is any DELETE |
| 110 | // (BoxRecordsAreNeverDeleted). Every request is recorded. aws and doc |
| 111 | // are the box's AWS and the identity document IMDS serves it now. |
| 112 | // |
| 113 | // Hooks, each used once: afterGet runs after the first GET has been |
| 114 | // answered (a writer arriving between the verb's read and |
| 115 | // publishBoxRecord's); beforePut runs before the first PUT reaches the |
| 116 | // server (a writer arriving just before the create). commitThenFail lists |
| 117 | // statuses for successive PUTs: each such PUT is applied and then answered |
| 118 | // with that status instead, as when S3 commits a write whose answer is |
| 119 | // lost. |
| 120 | type boxRecordsBucket struct { |
| 121 | model *leasetest.VersionedBucket |
| 122 | mu sync.Mutex |
| 123 | seen []s3Request |
| 124 | aws *testAWS |
| 125 | doc string |
| 126 | |
| 127 | afterGet func() |
| 128 | beforePut func() |
| 129 | commitThenFail []int |
| 130 | } |
| 131 | |
| 132 | func (b *boxRecordsBucket) requests() []s3Request { |
| 133 | b.mu.Lock() |
| 134 | defer b.mu.Unlock() |
| 135 | return append([]s3Request(nil), b.seen...) |
| 136 | } |
| 137 | |
| 138 | func (b *boxRecordsBucket) handler() http.Handler { |
| 139 | next, _ := leasetest.NewS3Server(b.model) |
| 140 | return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { |
| 141 | _, key, _ := strings.Cut(strings.TrimPrefix(r.URL.Path, "/"), "/") |
| 142 | b.mu.Lock() |
| 143 | b.seen = append(b.seen, s3Request{r.Method, key, r.Header.Get("If-None-Match"), r.Header.Get("If-Match")}) |
| 144 | var afterGet, beforePut func() |
| 145 | fail := 0 |
| 146 | switch r.Method { |
| 147 | case http.MethodGet: |
| 148 | afterGet, b.afterGet = b.afterGet, nil |
| 149 | case http.MethodPut: |
| 150 | beforePut, b.beforePut = b.beforePut, nil |
| 151 | if len(b.commitThenFail) > 0 { |
| 152 | fail, b.commitThenFail = b.commitThenFail[0], b.commitThenFail[1:] |
| 153 | } |
| 154 | } |
| 155 | b.mu.Unlock() |
| 156 | if strings.HasPrefix(key, boxid.Prefix) { |
| 157 | if r.Method == http.MethodDelete || (r.Method == http.MethodPut && r.Header.Get("If-None-Match") != "*") { |
| 158 | s3Error(w, http.StatusForbidden, "AccessDenied") |
| 159 | return |
| 160 | } |
| 161 | } |
| 162 | if beforePut != nil { |
| 163 | beforePut() |
| 164 | } |
| 165 | if fail != 0 { |
| 166 | next.ServeHTTP(httptest.NewRecorder(), r) |
| 167 | s3Error(w, fail, http.StatusText(fail)) |
| 168 | return |
| 169 | } |
| 170 | next.ServeHTTP(w, r) |
| 171 | if afterGet != nil { |
| 172 | afterGet() |
| 173 | } |
| 174 | }) |
| 175 | } |
| 176 | |
| 177 | func s3Error(w http.ResponseWriter, status int, code string) { |
| 178 | w.Header().Set("Content-Type", "application/xml") |
| 179 | w.WriteHeader(status) |
| 180 | _, _ = w.Write([]byte("<Error><Code>" + strings.ReplaceAll(code, " ", "") + "</Code><Message>" + code + "</Message></Error>")) |
| 181 | } |
| 182 | |
| 183 | // publishFixture points the process's AWS and CARLOS_* environment at a |
| 184 | // fresh fake bucket, as host.env would on a box, and makes the box's AWS. |
| 185 | func publishFixture(t *testing.T) *boxRecordsBucket { |
| 186 | t.Helper() |
| 187 | b := &boxRecordsBucket{model: leasetest.NewVersionedBucket(), aws: newTestAWS(t), |
| 188 | doc: identityDoc(publishAccount, publishRegion, publishInstance, "2026-10-06T08:00:00Z", "t4g.small")} |
| 189 | b.model.AddBucket(publishBucket) |
| 190 | server := httptest.NewServer(b.handler()) |
| 191 | t.Cleanup(server.Close) |
| 192 | for k, v := range map[string]string{ |
| 193 | "AWS_ACCESS_KEY_ID": "test-key", "AWS_SECRET_ACCESS_KEY": "test-secret", "AWS_SESSION_TOKEN": "", |
| 194 | "AWS_REGION": "eu-west-1", "AWS_EC2_METADATA_DISABLED": "true", |
| 195 | "AWS_CONFIG_FILE": "/nonexistent", "AWS_SHARED_CREDENTIALS_FILE": "/nonexistent", |
| 196 | "CARLOS_S3_ENDPOINT": server.URL, "CARLOS_DEPLOYMENT_BUCKET": publishBucket, |
| 197 | "CARLOS_DEPLOYMENT_DIR": "", "CARLOS_FLEET": "", "CARLOS_FLEET_BOX": "", "CARLOS_BOX_REFERENCE": "", |
| 198 | "CARLOS_REGION_HA": "", |
| 199 | } { |
| 200 | t.Setenv(k, v) |
| 201 | } |
| 202 | return b |
| 203 | } |
| 204 | |
| 205 | // imdsRead is the real IMDS reader over fakeIMDS serving b's current |
| 206 | // document and AWS's signature over it. |
| 207 | func (b *boxRecordsBucket) imdsRead(t *testing.T, written time.Time) func(string) (boxid.Record, error) { |
| 208 | client := imdsClient(&fakeIMDS{doc: b.doc, sig: b.aws.sign(t, b.doc)}) |
| 209 | return func(box string) (boxid.Record, error) { |
| 210 | return readIdentityRecord(client, "http://imds.test", box, written) |
| 211 | } |
| 212 | } |
| 213 | |
| 214 | // publishRun is the verb's production wiring with the test's seams: the |
| 215 | // agent user's uid, the fixed machine id, the fake IMDS and the test |
| 216 | // AWS's certificates. |
| 217 | func (b *boxRecordsBucket) publishRun(t *testing.T) boxPublishRun { |
| 218 | return boxPublishRun{getenv: os.Getenv, geteuid: func() int { return publishAgentUID }, |
| 219 | machineID: func() string { return publishMachineID }, open: openDeployStoreEnv, read: b.imdsRead(t, recordNow), |
| 220 | certs: b.aws.certs} |
| 221 | } |
| 222 | |
| 223 | func (b *boxRecordsBucket) run(t *testing.T) (string, error) { |
| 224 | t.Helper() |
| 225 | return b.publishRun(t).publish(context.Background()) |
| 226 | } |
| 227 | |
| 228 | // thisBoxRecord is the record this box would publish now, written at |
| 229 | // writtenAt. |
| 230 | func (b *boxRecordsBucket) thisBoxRecord(t *testing.T, writtenAt string) boxid.Record { |
| 231 | return b.aws.record(t, status.Label(publishMachineID), b.doc, writtenAt) |
| 232 | } |
| 233 | |
| 234 | // assertCreateOnly: no DELETE, no request off this box's key, and every |
| 235 | // PUT create-only. Overwrites would also have been denied by the fake's |
| 236 | // policy, but the verb must never even ask. |
| 237 | func assertCreateOnly(t *testing.T, reqs []s3Request) (puts int) { |
| 238 | t.Helper() |
| 239 | key := boxid.Key(status.Label(publishMachineID)) |
| 240 | for _, r := range reqs { |
| 241 | if r.Key != key { |
| 242 | t.Errorf("%s %s: publish-record touched another key", r.Method, r.Key) |
| 243 | } |
| 244 | switch r.Method { |
| 245 | case http.MethodPut: |
| 246 | puts++ |
| 247 | if r.IfNoneMatch != "*" || r.IfMatch != "" { |
| 248 | t.Errorf("PUT %s with If-None-Match %q If-Match %q; want create-only", r.Key, r.IfNoneMatch, r.IfMatch) |
| 249 | } |
| 250 | case http.MethodGet: |
| 251 | default: |
| 252 | t.Errorf("publish-record sent %s %s", r.Method, r.Key) |
| 253 | } |
| 254 | } |
| 255 | return puts |
| 256 | } |
| 257 | |
| 258 | func inPlaceLine(box string) string { |
| 259 | return "the record at control/boxes/" + box + ".json is valid AWS-signed evidence for " + box + |
| 260 | " (instance i-0abc), so nothing was written. enrollment still makes its own checks before it accepts the box" |
| 261 | } |
| 262 | |
| 263 | func TestPublishRecordWritesAFreshBoxsRecord(t *testing.T) { |
| 264 | b := publishFixture(t) |
| 265 | box := status.Label(publishMachineID) |
| 266 | line, err := b.run(t) |
| 267 | if err != nil { |
| 268 | t.Fatal(err) |
| 269 | } |
| 270 | if want := "published the identity record for " + box + " (instance i-0abc) to control/boxes/" + box + ".json"; line != want { |
| 271 | t.Fatalf("line = %q\nwant %q", line, want) |
| 272 | } |
| 273 | raw, ok := b.model.Current(publishBucket, "control/boxes/"+box+".json") |
| 274 | if !ok { |
| 275 | t.Fatal("no record at control/boxes/<box>.json") |
| 276 | } |
| 277 | var rec map[string]any |
| 278 | if err := json.Unmarshal(raw, &rec); err != nil { |
| 279 | t.Fatal(err) |
| 280 | } |
| 281 | want := b.thisBoxRecord(t, "2026-10-06T09:00:00Z") |
| 282 | if len(rec) != 5 || rec["v"] != float64(1) || rec["box"] != box || rec["identity_document"] != want.IdentityDocument || |
| 283 | rec["written_at"] != "2026-10-06T09:00:00Z" { |
| 284 | t.Fatalf("record = %v", rec) |
| 285 | } |
| 286 | // The record the verb wrote is evidence enrollment accepts. |
| 287 | certs, _ := b.aws.certs() |
| 288 | if _, err := (boxid.Verifier{Account: publishAccount, Certs: certs}).Enroll(raw, boxid.Expect{Box: box, AWSRegion: publishRegion}); err != nil { |
| 289 | t.Fatalf("the published record does not enroll: %v", err) |
| 290 | } |
| 291 | if puts := assertCreateOnly(t, b.requests()); puts != 1 { |
| 292 | t.Fatalf("%d PUTs, want one", puts) |
| 293 | } |
| 294 | } |
| 295 | |
| 296 | // R3: there is no region HA gate. Run after the loop has published (its |
| 297 | // publishBoxRecord, store, label and IMDS read, an hour earlier), the |
| 298 | // verb keeps that record and changes nothing. |
| 299 | func TestPublishRecordAfterTheLoopIsANoOp(t *testing.T) { |
| 300 | b := publishFixture(t) |
| 301 | ctx := context.Background() |
| 302 | store, err := openDeployStoreEnv(ctx, os.Getenv) |
| 303 | if err != nil { |
| 304 | t.Fatal(err) |
| 305 | } |
| 306 | box := metricsBoxLabel(os.Getenv, publishMachineID) |
| 307 | loopRead := b.imdsRead(t, recordNow.Add(-time.Hour)) |
| 308 | if done, err := publishBoxRecord(ctx, store, box, func() (boxid.Record, error) { return loopRead(box) }); !done || err != nil { |
| 309 | t.Fatalf("the loop's publish: done=%v err=%v", done, err) |
| 310 | } |
| 311 | before, _ := b.model.Current(publishBucket, boxid.Key(box)) |
| 312 | seenBefore := len(b.requests()) |
| 313 | line, err := b.run(t) |
| 314 | if err != nil { |
| 315 | t.Fatal(err) |
| 316 | } |
| 317 | if want := inPlaceLine(box); line != want { |
| 318 | t.Fatalf("line = %q\nwant %q", line, want) |
| 319 | } |
| 320 | if after, _ := b.model.Current(publishBucket, boxid.Key(box)); string(after) != string(before) { |
| 321 | t.Fatalf("the loop's record was rewritten:\n%s\n%s", before, after) |
| 322 | } |
| 323 | if puts := assertCreateOnly(t, b.requests()[seenBefore:]); puts != 0 { |
| 324 | t.Fatalf("%d PUTs, want none", puts) |
| 325 | } |
| 326 | } |
| 327 | |
| 328 | // Astra, round 2: an older record AWS signed for this same instance, in |
| 329 | // this account and region, is evidence enrollment accepts, although its |
| 330 | // document's mutable fields have changed since (a stop and start, a |
| 331 | // resize). It is kept byte for byte, never replaced, and nothing is |
| 332 | // written. |
| 333 | func TestPublishRecordKeepsAnOlderValidRecord(t *testing.T) { |
| 334 | b := publishFixture(t) |
| 335 | box := status.Label(publishMachineID) |
| 336 | older := identityDoc(publishAccount, publishRegion, publishInstance, "2026-09-01T00:00:00Z", "t4g.micro") |
| 337 | if older == b.doc { |
| 338 | t.Fatal("the two documents are the same") |
| 339 | } |
| 340 | before := mustJSON(t, b.aws.record(t, box, older, "2026-09-01T00:05:00Z")) |
| 341 | b.model.Put(publishBucket, boxid.Key(box), before) |
| 342 | line, err := b.run(t) |
| 343 | if err != nil { |
| 344 | t.Fatal(err) |
| 345 | } |
| 346 | if want := inPlaceLine(box); line != want { |
| 347 | t.Fatalf("line = %q\nwant %q", line, want) |
| 348 | } |
| 349 | if after, _ := b.model.Current(publishBucket, boxid.Key(box)); string(after) != string(before) { |
| 350 | t.Fatalf("the older record was replaced:\n%s\n%s", before, after) |
| 351 | } |
| 352 | if puts := assertCreateOnly(t, b.requests()); puts != 0 { |
| 353 | t.Fatalf("%d PUTs, want none", puts) |
| 354 | } |
| 355 | } |
| 356 | |
| 357 | // Anything enrollment's verifier refuses is refused, named, and left |
| 358 | // alone. |
| 359 | func TestPublishRecordRefusesWhatEnrollmentWouldRefuse(t *testing.T) { |
| 360 | box := status.Label(publishMachineID) |
| 361 | prefix := "refused: the record already under " + box + " is not valid evidence for this box: " |
| 362 | type held struct { |
| 363 | box, doc string |
| 364 | forger bool // signed by another key |
| 365 | edit func(*boxid.Record) |
| 366 | raw func([]byte) []byte |
| 367 | } |
| 368 | cases := map[string]struct { |
| 369 | h held |
| 370 | want string |
| 371 | }{ |
| 372 | "different account": {held{doc: identityDoc("999999999999", publishRegion, publishInstance, "2026-09-01T00:00:00Z", "t4g.small")}, "it is signed for account 999999999999, and this box is in account 111122223333"}, |
| 373 | "different region": {held{doc: identityDoc(publishAccount, "eu-west-2", publishInstance, "2026-09-01T00:00:00Z", "t4g.small")}, "it is signed in region eu-west-2, and this box is in eu-west-1"}, |
| 374 | "another instance": {held{doc: identityDoc(publishAccount, publishRegion, "i-0zzz", "2026-09-01T00:00:00Z", "t4g.small")}, "it is signed for instance i-0zzz, and this box is i-0abc"}, |
| 375 | "bad signature": {held{forger: true}, "it does not verify as AWS-signed evidence: signature: crypto/rsa: verification error"}, |
| 376 | "missing signature": {held{edit: func(r *boxid.Record) { r.IdentitySignature = "" }}, "it does not verify as AWS-signed evidence: signature: crypto/rsa: verification error"}, |
| 377 | "altered document": {held{edit: func(r *boxid.Record) { |
| 378 | r.IdentityDocument = base64.StdEncoding.EncodeToString([]byte(strings.Replace(identityDocFor(r), "t4g.small", "m7g.large", 1))) |
| 379 | }}, "it does not verify as AWS-signed evidence: signature: crypto/rsa: verification error"}, |
| 380 | "different box": {held{box: "bert-sawing"}, `it is for box "bert-sawing"`}, |
| 381 | "version 2": {held{edit: func(r *boxid.Record) { r.V = 2 }}, "its version is 2, not 1"}, |
| 382 | "unknown field": {held{raw: func(b []byte) []byte { return append(bytes.TrimSuffix(b, []byte("}")), []byte(`,"extra":1}`)...) }}, "it cannot be read as an identity record"}, |
| 383 | "trailing data": {held{raw: func(b []byte) []byte { return append(b, []byte(`{}`)...) }}, "it cannot be read as an identity record"}, |
| 384 | "not json": {held{raw: func([]byte) []byte { return []byte("not json") }}, "it cannot be read as an identity record"}, |
| 385 | "document not json": {held{edit: func(r *boxid.Record) { r.IdentityDocument = base64.StdEncoding.EncodeToString([]byte("nope")) }}, "it cannot be read as an identity record"}, |
| 386 | } |
| 387 | for name, tc := range cases { |
| 388 | t.Run(name, func(t *testing.T) { |
| 389 | b := publishFixture(t) |
| 390 | signer, doc, hbox := b.aws, b.doc, box |
| 391 | if tc.h.forger { |
| 392 | signer = newTestAWS(t) |
| 393 | } |
| 394 | if tc.h.doc != "" { |
| 395 | doc = tc.h.doc |
| 396 | } |
| 397 | if tc.h.box != "" { |
| 398 | hbox = tc.h.box |
| 399 | } |
| 400 | rec := signer.record(t, hbox, doc, "2026-09-01T00:05:00Z") |
| 401 | if tc.h.edit != nil { |
| 402 | tc.h.edit(&rec) |
| 403 | } |
| 404 | before := mustJSON(t, rec) |
| 405 | if tc.h.raw != nil { |
| 406 | before = tc.h.raw(before) |
| 407 | } |
| 408 | b.model.Put(publishBucket, boxid.Key(box), before) |
| 409 | line, err := b.run(t) |
| 410 | if want := prefix + tc.want + ". nothing was changed"; err == nil || err.Error() != want { |
| 411 | t.Fatalf("line=%q err=%v\nwant %q", line, err, want) |
| 412 | } |
| 413 | if after, _ := b.model.Current(publishBucket, boxid.Key(box)); string(after) != string(before) { |
| 414 | t.Fatalf("the record was changed:\n%s\n%s", before, after) |
| 415 | } |
| 416 | if puts := assertCreateOnly(t, b.requests()); puts != 0 { |
| 417 | t.Fatalf("%d PUTs, want none", puts) |
| 418 | } |
| 419 | }) |
| 420 | } |
| 421 | } |
| 422 | |
| 423 | // identityDocFor is r's decoded identity document. |
| 424 | func identityDocFor(r *boxid.Record) string { |
| 425 | doc, _ := base64.StdEncoding.DecodeString(r.IdentityDocument) |
| 426 | return string(doc) |
| 427 | } |
| 428 | |
| 429 | // A record that arrives after the verb's read and before |
| 430 | // publishBoxRecord's is held to the same verifier: publishBoxRecord alone |
| 431 | // would take any record naming this instance. |
| 432 | func TestPublishRecordHoldsALateRecordToTheSameVerifier(t *testing.T) { |
| 433 | box := status.Label(publishMachineID) |
| 434 | for name, tc := range map[string]struct { |
| 435 | forged bool |
| 436 | want string |
| 437 | }{ |
| 438 | "valid, older": {false, inPlaceLine(box)}, |
| 439 | "same instance, forged signature": {true, "refused: the record already under " + box + " is not valid evidence for this box: it does not verify as AWS-signed evidence: signature: crypto/rsa: verification error. nothing was changed"}, |
| 440 | } { |
| 441 | t.Run(name, func(t *testing.T) { |
| 442 | b := publishFixture(t) |
| 443 | signer := b.aws |
| 444 | if tc.forged { |
| 445 | signer = newTestAWS(t) |
| 446 | } |
| 447 | late := mustJSON(t, signer.record(t, box, identityDoc(publishAccount, publishRegion, publishInstance, "2026-09-01T00:00:00Z", "t4g.micro"), "2026-09-01T00:05:00Z")) |
| 448 | b.afterGet = func() { b.model.Put(publishBucket, boxid.Key(box), late) } |
| 449 | line, err := b.run(t) |
| 450 | got := line |
| 451 | if err != nil { |
| 452 | got = err.Error() |
| 453 | } |
| 454 | if got != tc.want || (err != nil) != tc.forged { |
| 455 | t.Fatalf("line=%q err=%v\nwant %q", line, err, tc.want) |
| 456 | } |
| 457 | if after, _ := b.model.Current(publishBucket, boxid.Key(box)); string(after) != string(late) { |
| 458 | t.Fatalf("the late record was changed:\n%s", after) |
| 459 | } |
| 460 | if puts := assertCreateOnly(t, b.requests()); puts != 0 { |
| 461 | t.Fatalf("%d PUTs, want none", puts) |
| 462 | } |
| 463 | }) |
| 464 | } |
| 465 | } |
| 466 | |
| 467 | // The conditional create is the bucket's: a record another instance |
| 468 | // creates just before the verb's write is not overwritten, because the |
| 469 | // fake answers the If-None-Match: * PUT with 412. |
| 470 | func TestPublishRecordLosesARaceWithoutOverwriting(t *testing.T) { |
| 471 | b := publishFixture(t) |
| 472 | box := status.Label(publishMachineID) |
| 473 | racer := mustJSON(t, b.aws.record(t, box, identityDoc(publishAccount, publishRegion, "i-0zzz", "2026-10-06T08:00:00Z", "t4g.small"), "2026-10-01T00:00:00Z")) |
| 474 | b.beforePut = func() { b.model.Put(publishBucket, boxid.Key(box), racer) } |
| 475 | want := "refused: another record already holds " + box + ": boxid: " + box + " already has a different record: store: compare-and-swap conflict. nothing was changed" |
| 476 | if line, err := b.run(t); err == nil || err.Error() != want { |
| 477 | t.Fatalf("line=%q err=%v\nwant %q", line, err, want) |
| 478 | } |
| 479 | if after, _ := b.model.Current(publishBucket, boxid.Key(box)); string(after) != string(racer) { |
| 480 | t.Fatalf("the racing record was overwritten:\n%s", after) |
| 481 | } |
| 482 | if puts := assertCreateOnly(t, b.requests()); puts != 1 { |
| 483 | t.Fatalf("%d PUTs, want the one conditional create", puts) |
| 484 | } |
| 485 | } |
| 486 | |
| 487 | // S3 commits the create but its answer is lost (500). The SDK retries, |
| 488 | // the retry is answered 412, and boxid.Publish finds exactly this run's |
| 489 | // bytes. That is this run's record: published, although the store never |
| 490 | // saw a successful Put. |
| 491 | func TestPublishRecordReportsARetriedCreateAsPublished(t *testing.T) { |
| 492 | b := publishFixture(t) |
| 493 | box := status.Label(publishMachineID) |
| 494 | b.commitThenFail = []int{http.StatusInternalServerError} |
| 495 | line, err := b.run(t) |
| 496 | if err != nil { |
| 497 | t.Fatal(err) |
| 498 | } |
| 499 | if want := "published the identity record for " + box + " (instance i-0abc) to control/boxes/" + box + ".json"; line != want { |
| 500 | t.Fatalf("line = %q\nwant %q", line, want) |
| 501 | } |
| 502 | got, _ := b.model.Current(publishBucket, boxid.Key(box)) |
| 503 | var rec boxid.Record |
| 504 | if json.Unmarshal(got, &rec) != nil || rec.WrittenAt != "2026-10-06T09:00:00Z" || rec.IdentityDocument != base64.StdEncoding.EncodeToString([]byte(b.doc)) { |
| 505 | t.Fatalf("stored %s", got) |
| 506 | } |
| 507 | if puts := assertCreateOnly(t, b.requests()); puts != 2 { |
| 508 | t.Fatalf("%d PUTs, want the create and its retry", puts) |
| 509 | } |
| 510 | } |
| 511 | |
| 512 | // Once a create has reached the bucket, a failure does not claim nothing |
| 513 | // was written. Here S3 committed it and answered 400. |
| 514 | func TestPublishRecordSaysAnUncertainWriteMayHaveHappened(t *testing.T) { |
| 515 | b := publishFixture(t) |
| 516 | box := status.Label(publishMachineID) |
| 517 | b.commitThenFail = []int{http.StatusBadRequest} |
| 518 | _, err := b.run(t) |
| 519 | if err == nil || !strings.HasPrefix(err.Error(), "the identity record for "+box+" may or may not have been written: ") || |
| 520 | !strings.HasSuffix(err.Error(), ". it is safe to run again, which only ever creates the record and says what is there") { |
| 521 | t.Fatalf("err = %v", err) |
| 522 | } |
| 523 | if _, ok := b.model.Current(publishBucket, boxid.Key(box)); !ok { |
| 524 | t.Fatal("the fake did not commit the create") |
| 525 | } |
| 526 | // The second run finds valid evidence and says so. |
| 527 | if line, err := b.run(t); err != nil || line != inPlaceLine(box) { |
| 528 | t.Fatalf("second run: line=%q err=%v", line, err) |
| 529 | } |
| 530 | } |
| 531 | |
| 532 | // Failures before any create reached the bucket say nothing was written. |
| 533 | func TestPublishRecordReportsFailuresBeforeAnyWriteAsRetryable(t *testing.T) { |
| 534 | box := status.Label(publishMachineID) |
| 535 | for name, tc := range map[string]struct { |
| 536 | imdsDown bool |
| 537 | doc string // the document IMDS serves, when not the usual one |
| 538 | reason string |
| 539 | }{ |
| 540 | "IMDS down": {imdsDown: true, reason: "IMDS down"}, |
| 541 | "own identity incomplete": {doc: "{\"accountId\" : \"111122223333\", \"region\" : \"eu-west-1\"}\n", reason: "the identity document AWS served this box names no account, region or instance"}, |
| 542 | } { |
| 543 | t.Run(name, func(t *testing.T) { |
| 544 | b := publishFixture(t) |
| 545 | if tc.doc != "" { |
| 546 | b.doc = tc.doc |
| 547 | } |
| 548 | run := b.publishRun(t) |
| 549 | if tc.imdsDown { |
| 550 | run.read = func(string) (boxid.Record, error) { return boxid.Record{}, errors.New("IMDS down") } |
| 551 | } |
| 552 | _, err := run.publish(context.Background()) |
| 553 | want := "the identity record for " + box + " was not published: " + tc.reason + ". nothing was written, so it is safe to run again" |
| 554 | if err == nil || err.Error() != want { |
| 555 | t.Fatalf("err = %v\nwant %q", err, want) |
| 556 | } |
| 557 | if puts := assertCreateOnly(t, b.requests()); puts != 0 { |
| 558 | t.Fatalf("%d PUTs, want none", puts) |
| 559 | } |
| 560 | }) |
| 561 | } |
| 562 | } |
| 563 | |
| 564 | // refusedEarly runs the verb with every later seam tripwired: it must |
| 565 | // refuse before the machine id, the store, IMDS or the certificates are |
| 566 | // touched. |
| 567 | func refusedEarly(t *testing.T, b *boxRecordsBucket, run boxPublishRun, want string) { |
| 568 | t.Helper() |
| 569 | touched := []string{} |
| 570 | run.machineID = func() string { touched = append(touched, "machine id"); return publishMachineID } |
| 571 | run.open = func(context.Context, func(string) string) (release.Store, error) { |
| 572 | touched = append(touched, "store") |
| 573 | return nil, errors.New("opened") |
| 574 | } |
| 575 | run.read = func(string) (boxid.Record, error) { touched = append(touched, "IMDS"); return boxid.Record{}, nil } |
| 576 | run.certs = func() (map[string]*x509.Certificate, error) { touched = append(touched, "certs"); return nil, nil } |
| 577 | _, err := run.publish(context.Background()) |
| 578 | if err == nil || err.Error() != want { |
| 579 | t.Fatalf("err = %v\nwant %q", err, want) |
| 580 | } |
| 581 | if len(touched) != 0 || len(b.requests()) != 0 { |
| 582 | t.Fatalf("read %v and sent %d requests before refusing", touched, len(b.requests())) |
| 583 | } |
| 584 | } |
| 585 | |
| 586 | // Census ruling R1: root is refused before any store, IMDS or host read. |
| 587 | func TestPublishRecordRefusesRoot(t *testing.T) { |
| 588 | b := publishFixture(t) |
| 589 | run := b.publishRun(t) |
| 590 | run.geteuid = func() int { return 0 } |
| 591 | refusedEarly(t, b, run, "carlos system box publish-record runs as the agent user, not root. run it with systemd-run -p User=carlos.") |
| 592 | } |
| 593 | |
| 594 | func TestPublishRecordRefusesAFleetBox(t *testing.T) { |
| 595 | b := publishFixture(t) |
| 596 | t.Setenv("CARLOS_FLEET", "acme") |
| 597 | t.Setenv("CARLOS_FLEET_BOX", "kass-1") |
| 598 | refusedEarly(t, b, b.publishRun(t), boxRecordFleetRefusal) |
| 599 | } |
| 600 | |
| 601 | // A deployment directory cannot refuse an overwrite (DirStore.Put checks, |
| 602 | // then renames), so only an S3 bucket will do. |
| 603 | func TestPublishRecordRequiresAnS3Bucket(t *testing.T) { |
| 604 | b := publishFixture(t) |
| 605 | dir := t.TempDir() |
| 606 | t.Setenv("CARLOS_DEPLOYMENT_BUCKET", "") |
| 607 | t.Setenv("CARLOS_DEPLOYMENT_DIR", dir) |
| 608 | refusedEarly(t, b, b.publishRun(t), boxRecordNoBucket) |
| 609 | if entries, _ := os.ReadDir(dir); len(entries) != 0 { |
| 610 | t.Fatalf("the deployment directory was written: %v", entries) |
| 611 | } |
| 612 | // Belt and braces: a store that is not S3 is refused even with a |
| 613 | // bucket named. |
| 614 | t.Setenv("CARLOS_DEPLOYMENT_BUCKET", publishBucket) |
| 615 | run := b.publishRun(t) |
| 616 | run.open = func(context.Context, func(string) string) (release.Store, error) { |
| 617 | return release.NewDirStore(dir), nil |
| 618 | } |
| 619 | if _, err := run.publish(context.Background()); err == nil || err.Error() != boxRecordNoBucket { |
| 620 | t.Fatalf("err = %v", err) |
| 621 | } |
| 622 | if entries, _ := os.ReadDir(dir); len(entries) != 0 { |
| 623 | t.Fatalf("the deployment directory was written: %v", entries) |
| 624 | } |
| 625 | } |
| 626 | |
| 627 | // A box its region's group launched publishes under its group reference, |
| 628 | // the label boxRecordLoop takes (metricsBoxLabel). |
| 629 | func TestPublishRecordUsesTheGroupReference(t *testing.T) { |
| 630 | b := publishFixture(t) |
| 631 | t.Setenv("CARLOS_BOX_REFERENCE", "group") |
| 632 | withGroupIdentity(t, publishAccount, publishInstance, nil) |
| 633 | ref := hostactors.GroupBoxReference(publishAccount, publishInstance) |
| 634 | if loop := metricsBoxLabel(os.Getenv, publishMachineID); loop != ref { |
| 635 | t.Fatalf("the loop's label is %q, want the group reference %q", loop, ref) |
| 636 | } |
| 637 | line, err := b.run(t) |
| 638 | if err != nil { |
| 639 | t.Fatal(err) |
| 640 | } |
| 641 | if want := "published the identity record for " + ref + " (instance i-0abc) to control/boxes/" + ref + ".json"; line != want { |
| 642 | t.Fatalf("line = %q\nwant %q", line, want) |
| 643 | } |
| 644 | raw, ok := b.model.Current(publishBucket, boxid.Key(ref)) |
| 645 | var rec boxid.Record |
| 646 | if !ok || json.Unmarshal(raw, &rec) != nil || rec.Box != ref { |
| 647 | t.Fatalf("record under the group reference: %s", raw) |
| 648 | } |
| 649 | if _, ok := b.model.Current(publishBucket, boxid.Key(status.Label(publishMachineID))); ok { |
| 650 | t.Fatal("a group box published under its machine-id label") |
| 651 | } |
| 652 | } |
| 653 | |
| 654 | // The verb's certificates are enrollment's: boxid.EmbeddedCerts covers |
| 655 | // the dev region, so a real box's record can verify. |
| 656 | func TestPublishRecordProductionCertsCoverTheDevRegion(t *testing.T) { |
| 657 | certs, err := boxid.EmbeddedCerts() |
| 658 | if err != nil || certs[publishRegion] == nil { |
| 659 | t.Fatalf("embedded certs: %v (eu-west-1 present: %v)", err, certs[publishRegion] != nil) |
| 660 | } |
| 661 | } |
| 662 | |
| 663 | // Census ruling R4: the store publish-record hands publishBoxRecord sends |
| 664 | // the bucket nothing for a delete, a list, an overwrite or any other key. |
| 665 | func TestBoxRecordStoreOnlyReadsAndCreatesItsOwnKey(t *testing.T) { |
| 666 | b := publishFixture(t) |
| 667 | ctx := context.Background() |
| 668 | inner, err := openDeployStoreEnv(ctx, os.Getenv) |
| 669 | if err != nil { |
| 670 | t.Fatal(err) |
| 671 | } |
| 672 | if _, ok := inner.(*release.S3Store); !ok { |
| 673 | t.Fatalf("store is %T, want the S3Store", inner) |
| 674 | } |
| 675 | s := &boxRecordStore{inner: inner, key: boxid.Key("annie-welding")} |
| 676 | if _, err := s.Put(ctx, boxid.Key("annie-welding"), []byte("{}"), `"etag"`); !errors.Is(err, errBoxRecordStore) { |
| 677 | t.Errorf("overwrite (replace-only Put): %v", err) |
| 678 | } |
| 679 | if _, err := s.Put(ctx, boxid.Key("bert-sawing"), []byte("{}"), ""); !errors.Is(err, errBoxRecordStore) { |
| 680 | t.Errorf("Put on another box's key: %v", err) |
| 681 | } |
| 682 | if _, err := s.Put(ctx, "control/serial", []byte("{}"), ""); !errors.Is(err, errBoxRecordStore) { |
| 683 | t.Errorf("Put on another key: %v", err) |
| 684 | } |
| 685 | if _, _, err := s.Get(ctx, boxid.Key("bert-sawing")); !errors.Is(err, errBoxRecordStore) { |
| 686 | t.Errorf("Get on another box's key: %v", err) |
| 687 | } |
| 688 | if err := s.Delete(ctx, boxid.Key("annie-welding")); !errors.Is(err, errBoxRecordStore) { |
| 689 | t.Errorf("Delete: %v", err) |
| 690 | } |
| 691 | if _, err := s.List(ctx, boxid.Prefix); !errors.Is(err, errBoxRecordStore) { |
| 692 | t.Errorf("List: %v", err) |
| 693 | } |
| 694 | if _, err := s.ListDirs(ctx, "control/"); !errors.Is(err, errBoxRecordStore) { |
| 695 | t.Errorf("ListDirs: %v", err) |
| 696 | } |
| 697 | if reqs := b.requests(); len(reqs) != 0 || s.attempted || s.wrote { |
| 698 | t.Fatalf("refused calls reached S3: %v (attempted=%v wrote=%v)", reqs, s.attempted, s.wrote) |
| 699 | } |
| 700 | if _, err := s.Put(ctx, boxid.Key("annie-welding"), []byte("{}"), ""); err != nil || !s.wrote { |
| 701 | t.Fatalf("create: err=%v wrote=%v", err, s.wrote) |
| 702 | } |
| 703 | if _, err := s.Put(ctx, boxid.Key("annie-welding"), []byte("{}"), ""); !errors.Is(err, release.ErrCASConflict) { |
| 704 | t.Fatalf("a second create: %v, want the bucket's 412 as ErrCASConflict", err) |
| 705 | } |
| 706 | if reqs := b.requests(); len(reqs) != 2 || reqs[0].IfNoneMatch != "*" || reqs[1].IfNoneMatch != "*" { |
| 707 | t.Fatalf("requests %v, want two create-only PUTs", reqs) |
| 708 | } |
| 709 | } |
| 710 | |
| 711 | // boxRecordLoop's gate is unchanged: with CARLOS_REGION_HA off, or on a |
| 712 | // fleet box, it returns at once and never reaches the bucket. Past the |
| 713 | // gate it would retry each minute, so a broken gate shows as a timeout. |
| 714 | func TestBoxRecordLoopGateIsUnchanged(t *testing.T) { |
| 715 | for name, env := range map[string]map[string]string{ |
| 716 | "region HA off": {"CARLOS_REGION_HA": ""}, |
| 717 | "region HA unparseable": {"CARLOS_REGION_HA": "yes"}, |
| 718 | "fleet box, region HA on": {"CARLOS_REGION_HA": "1", "CARLOS_FLEET": "acme", "CARLOS_FLEET_BOX": "kass-1"}, |
| 719 | } { |
| 720 | t.Run(name, func(t *testing.T) { |
| 721 | b := publishFixture(t) |
| 722 | for k, v := range env { |
| 723 | t.Setenv(k, v) |
| 724 | } |
| 725 | done := make(chan struct{}) |
| 726 | go func() { boxRecordLoop(); close(done) }() |
| 727 | select { |
| 728 | case <-done: |
| 729 | case <-time.After(10 * time.Second): |
| 730 | t.Fatal("boxRecordLoop did not return: its region HA gate let it run") |
| 731 | } |
| 732 | if n := len(b.requests()); n != 0 { |
| 733 | t.Fatalf("boxRecordLoop sent %d requests with its gate shut", n) |
| 734 | } |
| 735 | }) |
| 736 | } |
| 737 | } |
| 738 | |
| 739 | // Box publication reaches the deployment bucket through openDeployStoreEnv |
| 740 | // and requires the *release.S3Store it returns (boxpublish.go's |
| 741 | // opened.(*release.S3Store)). The change index must not cost it that type: |
| 742 | // the opener hands back the same concrete store with the index enabled, |
| 743 | // against the real S3 wiring, and publication still writes only its own |
| 744 | // control/boxes/ key (a key outside every change scope, so no index traffic). |
| 745 | // Not parallel: publishFixture sets the process environment. |
| 746 | func TestBoxPublishThroughDeploymentOpener(t *testing.T) { |
| 747 | b := publishFixture(t) |
| 748 | ctx := context.Background() |
| 749 | opened, err := openDeployStoreEnv(ctx, os.Getenv) |
| 750 | if err != nil { |
| 751 | t.Fatal(err) |
| 752 | } |
| 753 | if _, ok := opened.(*release.S3Store); !ok { |
| 754 | t.Fatalf("openDeployStoreEnv returned %T; box publication needs the *release.S3Store", opened) |
| 755 | } |
| 756 | // The opener enabled the index: a scoped write bumps its generation. |
| 757 | if _, err := opened.Put(ctx, "accounts/shop/manifest.json", []byte("{}"), ""); err != nil { |
| 758 | t.Fatal(err) |
| 759 | } |
| 760 | idx, _, err := release.ReadChangeIndex(ctx, opened) |
| 761 | if err != nil { |
| 762 | t.Fatal(err) |
| 763 | } |
| 764 | if got := idx.Entries["accounts/shop"]; got != 1 { |
| 765 | t.Fatalf("accounts/shop generation = %d after one scoped Put, want 1 (entries %v)", got, idx.Entries) |
| 766 | } |
| 767 | // Publication itself is unchanged: one create-only PUT on its own key. |
| 768 | before := len(b.requests()) |
| 769 | if _, err := b.run(t); err != nil { |
| 770 | t.Fatal(err) |
| 771 | } |
| 772 | if puts := assertCreateOnly(t, b.requests()[before:]); puts != 1 { |
| 773 | t.Fatalf("%d PUTs, want one", puts) |
| 774 | } |
| 775 | } |
| 776 | |
amadan