| 1 | package edge |
| 2 | |
| 3 | import ( |
| 4 | "bufio" |
| 5 | "context" |
| 6 | "crypto/tls" |
| 7 | "crypto/x509" |
| 8 | "errors" |
| 9 | "fmt" |
| 10 | "io" |
| 11 | "net" |
| 12 | "net/http" |
| 13 | "net/http/httptest" |
| 14 | "os" |
| 15 | "strings" |
| 16 | "sync" |
| 17 | "testing" |
| 18 | "time" |
| 19 | |
| 20 | "github.com/carlosframework/platform/internal/edgeproof" |
| 21 | "github.com/carlosframework/platform/internal/lbhealth" |
| 22 | "github.com/carlosframework/platform/internal/registry" |
| 23 | "github.com/carlosframework/platform/internal/release" |
| 24 | ) |
| 25 | |
| 26 | // putJSON writes a control-plane object body verbatim, replacing whatever is |
| 27 | // there. release.Store.Put with an empty ifMatch is create-only, so a test |
| 28 | // that rewrites a key (the last-good and stale-file cases below) has to carry |
| 29 | // the etag — the same putReplacing shape certsync.go uses in production. |
| 30 | func putJSON(t *testing.T, s release.Store, key, body string) { |
| 31 | t.Helper() |
| 32 | _, etag, err := s.Get(context.Background(), key) |
| 33 | switch { |
| 34 | case errors.Is(err, os.ErrNotExist): |
| 35 | etag = "" |
| 36 | case err != nil: |
| 37 | t.Fatalf("get %s: %v", key, err) |
| 38 | } |
| 39 | if _, err := s.Put(context.Background(), key, []byte(body), etag); err != nil { |
| 40 | t.Fatalf("put %s: %v", key, err) |
| 41 | } |
| 42 | } |
| 43 | |
| 44 | func refreshHomeMap(t *testing.T, m *homeMap) { |
| 45 | t.Helper() |
| 46 | if err := m.Refresh(context.Background()); err != nil { |
| 47 | t.Fatalf("refresh: %v", err) |
| 48 | } |
| 49 | } |
| 50 | |
| 51 | func TestHomeMapFoldsAndExcludesOwnBox(t *testing.T) { |
| 52 | s := meshTestStore(t) |
| 53 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"a.test":{"role":"home"}}}`) |
| 54 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 55 | m := newHomeMap(s, "annie") |
| 56 | refreshHomeMap(t, m) |
| 57 | if eip, box, ok := m.Home("b.test"); !ok || eip != "2.2.2.2" || box != "betty" { |
| 58 | t.Fatalf("b.test home = %s/%s/%v", eip, box, ok) |
| 59 | } |
| 60 | if _, _, ok := m.Home("a.test"); ok { |
| 61 | t.Fatal("own-box hosts must not offer a backhaul target") |
| 62 | } |
| 63 | } |
| 64 | |
| 65 | // Task 1's writer publishes registry row hosts VERBATIM, catch-all rows |
| 66 | // included, so the reader owes lookupUpstream's one-label wildcard rule: an |
| 67 | // exact match wins, then "*.<parent>" — and only ONE label, so a two-label |
| 68 | // host misses exactly as it does locally. |
| 69 | func TestHomeMapAppliesOneLabelWildcardRule(t *testing.T) { |
| 70 | s := meshTestStore(t) |
| 71 | putJSON(t, s, "control/edges/betty.json", |
| 72 | `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"*.apps.test":{"role":"home"},"exact.apps.test":{"role":"home"}}}`) |
| 73 | m := newHomeMap(s, "annie") |
| 74 | refreshHomeMap(t, m) |
| 75 | |
| 76 | // Exact form. |
| 77 | if eip, _, ok := m.Home("exact.apps.test"); !ok || eip != "2.2.2.2" { |
| 78 | t.Fatalf("exact host = %s/%v, want the home edge", eip, ok) |
| 79 | } |
| 80 | // Wildcard form: one label under the catch-all parent. |
| 81 | if eip, box, ok := m.Home("paul.apps.test"); !ok || eip != "2.2.2.2" || box != "betty" { |
| 82 | t.Fatalf("one-label wildcard host = %s/%s/%v", eip, box, ok) |
| 83 | } |
| 84 | // Two labels: a miss here exactly as in lookupUpstream, which composes |
| 85 | // "*.b.apps.test" and finds nothing. |
| 86 | if _, _, ok := m.Home("a.b.apps.test"); ok { |
| 87 | t.Fatal("a two-label host must not match a one-label wildcard") |
| 88 | } |
| 89 | // The wildcard key itself is not a request host and must not resolve as |
| 90 | // one via the fallback ("*.apps.test" composes "*.apps.test" again). |
| 91 | if _, _, ok := m.Home("apps.test"); ok { |
| 92 | t.Fatal("the catch-all parent must not match its own wildcard") |
| 93 | } |
| 94 | } |
| 95 | |
| 96 | // The role object is read-local anticipation (Task 1). Only "home" routes; |
| 97 | // anything else is a role this build does not implement, and an unknown role |
| 98 | // must neither route nor fail the poll. |
| 99 | func TestHomeMapIgnoresUnknownRoles(t *testing.T) { |
| 100 | s := meshTestStore(t) |
| 101 | putJSON(t, s, "control/edges/betty.json", |
| 102 | `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"read.test":{"role":"read"},"home.test":{"role":"home"},"blank.test":{}}}`) |
| 103 | m := newHomeMap(s, "annie") |
| 104 | refreshHomeMap(t, m) // an unknown role is not an error |
| 105 | if _, _, ok := m.Home("read.test"); ok { |
| 106 | t.Fatal("role:\"read\" is not a home — it must not offer a backhaul target") |
| 107 | } |
| 108 | if _, _, ok := m.Home("blank.test"); ok { |
| 109 | t.Fatal("an absent role is not a home") |
| 110 | } |
| 111 | if _, _, ok := m.Home("home.test"); !ok { |
| 112 | t.Fatal("role:\"home\" must still route alongside roles this build ignores") |
| 113 | } |
| 114 | } |
| 115 | |
| 116 | // A box with no discoverable public IP publishes its hosts list with eip:"" |
| 117 | // (Task 1's DR-runbook case). There is nothing to dial, so it must never be |
| 118 | // offered as a target — a "" EIP would otherwise become a dial to port 443 of |
| 119 | // nothing. |
| 120 | func TestHomeMapSkipsBoxesWithNoEIP(t *testing.T) { |
| 121 | s := meshTestStore(t) |
| 122 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"","hosts":{"b.test":{"role":"home"}}}`) |
| 123 | m := newHomeMap(s, "annie") |
| 124 | refreshHomeMap(t, m) |
| 125 | if _, _, ok := m.Home("b.test"); ok { |
| 126 | t.Fatal("a box with no EIP has no dial address and must not be a backhaul target") |
| 127 | } |
| 128 | } |
| 129 | |
| 130 | // Two boxes claiming the same host is a placement bug, but every edge must |
| 131 | // still pick the SAME one — a coin flip per box would split traffic across a |
| 132 | // move instead of converging on it. |
| 133 | func TestHomeMapPicksTheSameHomeOnADuplicateClaim(t *testing.T) { |
| 134 | s := meshTestStore(t) |
| 135 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 136 | putJSON(t, s, "control/edges/carol.json", `{"v":1,"box":"carol","eip":"3.3.3.3","hosts":{"b.test":{"role":"home"}}}`) |
| 137 | for i := 0; i < 5; i++ { |
| 138 | m := newHomeMap(s, "annie") |
| 139 | refreshHomeMap(t, m) |
| 140 | if _, box, _ := m.Home("b.test"); box != "betty" { |
| 141 | t.Fatalf("duplicate claim resolved to %q, want the deterministic first (betty)", box) |
| 142 | } |
| 143 | } |
| 144 | } |
| 145 | |
| 146 | // A box that is retired (its file deleted) must LEAVE the map — the fold is a |
| 147 | // full rebuild from each list, never a merge into the previous one. |
| 148 | func TestHomeMapForgetsADeletedEdgesFile(t *testing.T) { |
| 149 | s := meshTestStore(t) |
| 150 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 151 | m := newHomeMap(s, "annie") |
| 152 | refreshHomeMap(t, m) |
| 153 | if _, _, ok := m.Home("b.test"); !ok { |
| 154 | t.Fatal("precondition: b.test should be homed on betty") |
| 155 | } |
| 156 | if err := s.Delete(context.Background(), "control/edges/betty.json"); err != nil { |
| 157 | t.Fatalf("delete: %v", err) |
| 158 | } |
| 159 | refreshHomeMap(t, m) |
| 160 | if _, _, ok := m.Home("b.test"); ok { |
| 161 | t.Fatal("a retired box's hosts must leave the map on the next poll") |
| 162 | } |
| 163 | } |
| 164 | |
| 165 | // One unreadable object is not a bad poll: the previous copy of that box's |
| 166 | // file carries forward and every other box's file still lands, exactly as the |
| 167 | // cert mirror does it. |
| 168 | func TestHomeMapCarriesForwardOneUnreadableFile(t *testing.T) { |
| 169 | inner := meshTestStore(t) |
| 170 | s := &oneKeyFailingStore{Store: inner} |
| 171 | putJSON(t, inner, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 172 | m := newHomeMap(s, "annie") |
| 173 | refreshHomeMap(t, m) |
| 174 | |
| 175 | // betty's file goes unreadable; carol appears in the same poll. |
| 176 | putJSON(t, inner, "control/edges/carol.json", `{"v":1,"box":"carol","eip":"3.3.3.3","hosts":{"c.test":{"role":"home"}}}`) |
| 177 | s.failKey("control/edges/betty.json") |
| 178 | refreshHomeMap(t, m) |
| 179 | |
| 180 | if eip, _, ok := m.Home("b.test"); !ok || eip != "2.2.2.2" { |
| 181 | t.Fatalf("b.test = %s/%v — the previous copy of an unreadable file must carry forward", eip, ok) |
| 182 | } |
| 183 | if _, _, ok := m.Home("c.test"); !ok { |
| 184 | t.Fatal("one bad object must not discard the rest of the rebuild") |
| 185 | } |
| 186 | } |
| 187 | |
| 188 | // A corrupt body is the same class of fault as an unreadable one: carry the |
| 189 | // last good parse forward rather than dropping the box off the map. |
| 190 | func TestHomeMapCarriesForwardACorruptFile(t *testing.T) { |
| 191 | s := meshTestStore(t) |
| 192 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 193 | m := newHomeMap(s, "annie") |
| 194 | refreshHomeMap(t, m) |
| 195 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty",`) |
| 196 | refreshHomeMap(t, m) |
| 197 | if eip, _, ok := m.Home("b.test"); !ok || eip != "2.2.2.2" { |
| 198 | t.Fatalf("b.test = %s/%v — a corrupt file must leave the last good parse standing", eip, ok) |
| 199 | } |
| 200 | } |
| 201 | |
| 202 | // A failed List is a bucket fault: last-good stands and the error is returned |
| 203 | // for the loop to log. An empty map here would silently stop every backhaul |
| 204 | // in the deployment on one blip. |
| 205 | func TestHomeMapKeepsLastGoodWhenTheBucketIsDown(t *testing.T) { |
| 206 | inner := meshTestStore(t) |
| 207 | s := &listFailingStore{Store: inner} |
| 208 | putJSON(t, inner, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 209 | m := newHomeMap(s, "annie") |
| 210 | refreshHomeMap(t, m) |
| 211 | |
| 212 | s.setFail(true) |
| 213 | if err := m.Refresh(context.Background()); err == nil { |
| 214 | t.Fatal("a failed List must be reported, not swallowed") |
| 215 | } |
| 216 | if _, _, ok := m.Home("b.test"); !ok { |
| 217 | t.Fatal("last-good must stand through a bucket blip") |
| 218 | } |
| 219 | } |
| 220 | |
| 221 | // An edge with no box identity of its own cannot exclude itself, so it must |
| 222 | // never offer a target at all — fail closed rather than backhaul to itself. |
| 223 | func TestHomeMapWithoutASelfLabelNeverOffersATarget(t *testing.T) { |
| 224 | s := meshTestStore(t) |
| 225 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 226 | m := newHomeMap(s, "") |
| 227 | refreshHomeMap(t, m) |
| 228 | if _, _, ok := m.Home("b.test"); ok { |
| 229 | t.Fatal("an edge that does not know its own box label must not backhaul") |
| 230 | } |
| 231 | } |
| 232 | |
| 233 | // A nil map (no deployment store, so no mesh at all) answers like an empty |
| 234 | // one rather than panicking on the request path. |
| 235 | func TestNilHomeMapIsInert(t *testing.T) { |
| 236 | var m *homeMap |
| 237 | if _, _, ok := m.Home("b.test"); ok { |
| 238 | t.Fatal("a nil home map must offer nothing") |
| 239 | } |
| 240 | if m.KnownHost("b.test") { |
| 241 | t.Fatal("a nil home map knows no host") |
| 242 | } |
| 243 | if cur, next := m.meshKeys(); cur != nil || next != nil { |
| 244 | t.Fatal("a nil home map holds no mesh keys") |
| 245 | } |
| 246 | } |
| 247 | |
| 248 | // KnownHost is the membership question the three cert gates ask through |
| 249 | // Edge.deploymentHost (#277), and it is EXACT-only where Home is not. Home's |
| 250 | // one-label fallback is what makes an invented subdomain under a remote |
| 251 | // catch-all DIALABLE; letting the same fallback into a cert gate would let |
| 252 | // that invented label drive an ACME order and spend a read-through budget, |
| 253 | // which is precisely what PR #180's Allowed rule refuses. The two answers |
| 254 | // diverging on the same map, in the same test, is the pin. |
| 255 | func TestKnownHostExactOnly(t *testing.T) { |
| 256 | s := meshTestStore(t) |
| 257 | putJSON(t, s, "control/edges/annie.json", |
| 258 | `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"own.test":{"role":"home"}}}`) |
| 259 | putJSON(t, s, "control/edges/betty.json", |
| 260 | `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"*.apps.test":{"role":"home"},"exact.apps.test":{"role":"home"}}}`) |
| 261 | m := newHomeMap(s, "annie") |
| 262 | refreshHomeMap(t, m) |
| 263 | |
| 264 | if !m.KnownHost("exact.apps.test") { |
| 265 | t.Fatal("an exactly-named remote-homed host is the whole point of the predicate") |
| 266 | } |
| 267 | // Precondition: the fallback Home applies is live on this very map — so |
| 268 | // the refusal below is KnownHost's own, not an empty map's. |
| 269 | if _, _, ok := m.Home("invented.apps.test"); !ok { |
| 270 | t.Fatal("test bug: the one-label wildcard rule did not apply") |
| 271 | } |
| 272 | if m.KnownHost("invented.apps.test") { |
| 273 | t.Fatal("a label under a mirrored catch-all must never be admitted — that is #180's rule") |
| 274 | } |
| 275 | if m.KnownHost("*.apps.test") { |
| 276 | t.Fatal("the catch-all key itself is not a certifiable name") |
| 277 | } |
| 278 | if m.KnownHost("stranger.example") { |
| 279 | t.Fatal("a host no box in the deployment homes is a stranger") |
| 280 | } |
| 281 | // Own-box rows are excluded by the fold and stay excluded here. Harmless: |
| 282 | // deploymentHost's other half (Allowed) is what admits this box's routes. |
| 283 | if m.KnownHost("own.test") { |
| 284 | t.Fatal("the fold excludes this box's own file, so KnownHost inherits that exclusion") |
| 285 | } |
| 286 | // Nothing polled yet: no membership, and no panic on the gate's path. |
| 287 | if newHomeMap(s, "annie").KnownHost("exact.apps.test") { |
| 288 | t.Fatal("a map that has never polled knows no host") |
| 289 | } |
| 290 | } |
| 291 | |
| 292 | // A box publishing eip:"" is in the fold's no-dial exclusion, and KnownHost |
| 293 | // inherits it: there is no DNS answer pointing at such a box, so a validation |
| 294 | // for its hosts has nowhere legitimate to land. |
| 295 | func TestKnownHostSkipsBoxesWithNoEIP(t *testing.T) { |
| 296 | s := meshTestStore(t) |
| 297 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"","hosts":{"b.test":{"role":"home"}}}`) |
| 298 | m := newHomeMap(s, "annie") |
| 299 | refreshHomeMap(t, m) |
| 300 | if m.KnownHost("b.test") { |
| 301 | t.Fatal("a box with no public IP offers no validation target") |
| 302 | } |
| 303 | } |
| 304 | |
| 305 | // ---- hop acceptance --------------------------------------------------------- |
| 306 | |
| 307 | // forwardEcho is an upstream that reports back exactly the trust-bearing |
| 308 | // headers it was handed, so a test can assert both what survived and what was |
| 309 | // stripped on the way in. |
| 310 | func forwardEcho(t *testing.T) *httptest.Server { |
| 311 | t.Helper() |
| 312 | ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { |
| 313 | fmt.Fprintf(w, "xff=%s|xfh=%s|hop=%s|auth=%s", |
| 314 | r.Header.Get("X-Forwarded-For"), r.Header.Get("X-Forwarded-Host"), |
| 315 | r.Header.Get(hopHeader), r.Header.Get(hopAuthHeader)) |
| 316 | })) |
| 317 | t.Cleanup(ts.Close) |
| 318 | return ts |
| 319 | } |
| 320 | |
| 321 | // hopEdge builds an edge serving hosts from one echo backend, with a mesh key |
| 322 | // minted and a home map armed over the same store — the receiving box of a |
| 323 | // backhaul hop. |
| 324 | func hopEdge(t *testing.T, self string, hosts ...string) (*Edge, *homeMap, []byte) { |
| 325 | t.Helper() |
| 326 | s := meshTestStore(t) |
| 327 | if err := mintMeshKey(context.Background(), s); err != nil { |
| 328 | t.Fatalf("mint: %v", err) |
| 329 | } |
| 330 | backend := forwardEcho(t) |
| 331 | reg := openReg(t) |
| 332 | for _, h := range hosts { |
| 333 | if err := reg.Upsert(registry.Route{Host: h, Kind: "service", Addr: backend.Listener.Addr().String(), Status: "ready"}); err != nil { |
| 334 | t.Fatalf("upsert %s: %v", h, err) |
| 335 | } |
| 336 | } |
| 337 | e, err := New(reg, nil) |
| 338 | if err != nil { |
| 339 | t.Fatalf("New: %v", err) |
| 340 | } |
| 341 | t.Cleanup(e.probers.Stop) |
| 342 | m := newHomeMap(s, self) |
| 343 | refreshHomeMap(t, m) |
| 344 | e.homes.Store(m) |
| 345 | key, _ := m.meshKeys() |
| 346 | if len(key) == 0 { |
| 347 | t.Fatal("precondition: the mesh key must be mirrored") |
| 348 | } |
| 349 | return e, m, key |
| 350 | } |
| 351 | |
| 352 | // hopRequest drives one request at the edge with the given headers. |
| 353 | func hopRequest(t *testing.T, e *Edge, host string, headers map[string]string) *http.Response { |
| 354 | t.Helper() |
| 355 | srv := httptest.NewServer(e) |
| 356 | t.Cleanup(srv.Close) |
| 357 | req, err := http.NewRequest("GET", srv.URL+"/", nil) |
| 358 | if err != nil { |
| 359 | t.Fatalf("new request: %v", err) |
| 360 | } |
| 361 | req.Host = host |
| 362 | for k, v := range headers { |
| 363 | req.Header.Set(k, v) |
| 364 | } |
| 365 | resp, err := http.DefaultClient.Do(req) |
| 366 | if err != nil { |
| 367 | t.Fatalf("do: %v", err) |
| 368 | } |
| 369 | t.Cleanup(func() { resp.Body.Close() }) |
| 370 | return resp |
| 371 | } |
| 372 | |
| 373 | func readBody(t *testing.T, resp *http.Response) string { |
| 374 | t.Helper() |
| 375 | b, err := io.ReadAll(resp.Body) |
| 376 | if err != nil { |
| 377 | t.Fatalf("read body: %v", err) |
| 378 | } |
| 379 | return string(b) |
| 380 | } |
| 381 | |
| 382 | // A verified hop is served locally, and the chain the sending edge appended |
| 383 | // is the one thing an ordinary client is never trusted for. |
| 384 | func TestVerifiedHopIsServedLocallyAndHonoursTheInboundXFF(t *testing.T) { |
| 385 | e, _, key := hopEdge(t, "annie", "app.test") |
| 386 | resp := hopRequest(t, e, "app.test", map[string]string{ |
| 387 | hopHeader: "1", |
| 388 | hopAuthHeader: signHop(key, "app.test", time.Now()), |
| 389 | "X-Forwarded-For": "203.0.113.9", |
| 390 | }) |
| 391 | if resp.StatusCode != http.StatusOK { |
| 392 | t.Fatalf("status = %d, want 200", resp.StatusCode) |
| 393 | } |
| 394 | body := readBody(t, resp) |
| 395 | // Verbatim, with NO element appended: the last element must still be the |
| 396 | // client. internal/console/authcli.go's clientIP takes the last element as |
| 397 | // the caller and every tenant framework's rightmost-public heuristic does |
| 398 | // the same, so appending this edge's peer (the sibling edge) would make a |
| 399 | // backhauled request look as though infrastructure called it. |
| 400 | if !strings.HasPrefix(body, "xff=203.0.113.9|") { |
| 401 | t.Fatalf("body = %q — a verified hop's inbound XFF must pass through verbatim, last element still the client", body) |
| 402 | } |
| 403 | // The hop's own trust headers are the edge's vocabulary, never the app's. |
| 404 | if !strings.Contains(body, "|hop=|auth=") { |
| 405 | t.Fatalf("body = %q — the hop headers must not reach the app", body) |
| 406 | } |
| 407 | } |
| 408 | |
| 409 | // The MAC binds the CONCRETE request host, and a catch-all route serves it |
| 410 | // through lookupUpstream's wildcard fallback — the two halves of carry C1 on |
| 411 | // the receiving side. |
| 412 | func TestVerifiedHopServesAWildcardRoutedHost(t *testing.T) { |
| 413 | e, _, key := hopEdge(t, "annie", "*.apps.test") |
| 414 | resp := hopRequest(t, e, "paul.apps.test", map[string]string{ |
| 415 | hopHeader: "1", |
| 416 | hopAuthHeader: signHop(key, "paul.apps.test", time.Now()), |
| 417 | }) |
| 418 | if resp.StatusCode != http.StatusOK { |
| 419 | t.Fatalf("status = %d, want 200 — a catch-all route serves a hop like any other client", resp.StatusCode) |
| 420 | } |
| 421 | } |
| 422 | |
| 423 | // The stale-map refusal: the sender's map named us for a host we do not |
| 424 | // serve. 503 + Retry-After: 2, and structurally never a re-forward — the home |
| 425 | // map here is POISONED to name another box for exactly that host, so anything |
| 426 | // that consulted it would leave through a dial instead. |
| 427 | func TestVerifiedHopForAnUnservedHostRefusesWithoutForwarding(t *testing.T) { |
| 428 | e, m, key := hopEdge(t, "annie", "app.test") |
| 429 | putJSON(t, m.s, "control/edges/betty.json", |
| 430 | `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"moved.test":{"role":"home"}}}`) |
| 431 | refreshHomeMap(t, m) |
| 432 | if _, _, ok := m.Home("moved.test"); !ok { |
| 433 | t.Fatal("precondition: the map must name a remote home for moved.test") |
| 434 | } |
| 435 | |
| 436 | resp := hopRequest(t, e, "moved.test", map[string]string{ |
| 437 | hopHeader: "1", |
| 438 | hopAuthHeader: signHop(key, "moved.test", time.Now()), |
| 439 | }) |
| 440 | if resp.StatusCode != http.StatusServiceUnavailable { |
| 441 | t.Fatalf("status = %d, want 503 for a verified hop naming a host we do not serve", resp.StatusCode) |
| 442 | } |
| 443 | if got := resp.Header.Get("Retry-After"); got != "2" { |
| 444 | t.Fatalf("Retry-After = %q, want 2", got) |
| 445 | } |
| 446 | if body := readBody(t, resp); strings.Contains(body, "moved.test") || strings.Contains(body, "betty") { |
| 447 | t.Fatalf("body = %q — the refusal must not name the host or where it went", body) |
| 448 | } |
| 449 | } |
| 450 | |
| 451 | // Task 1: the refusal used to be silent. A home move stuck mid-way can leave |
| 452 | // a sender retrying into this refusal for minutes (the loop cadences, |
| 453 | // edgesfile.go/homemap.go's own docs), so it now logs — but first-and- |
| 454 | // every-tenth, logPartialPoll's own cadence, or a stuck move would fill the |
| 455 | // journal with one line per hop. |
| 456 | func TestStaleHopRefusalLogsFirstAndEveryTenth(t *testing.T) { |
| 457 | e, m, key := hopEdge(t, "annie", "app.test") |
| 458 | putJSON(t, m.s, "control/edges/betty.json", |
| 459 | `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"moved.test":{"role":"home"}}}`) |
| 460 | refreshHomeMap(t, m) |
| 461 | |
| 462 | buf := captureLog(t) |
| 463 | for i := 0; i < 10; i++ { |
| 464 | resp := hopRequest(t, e, "moved.test", map[string]string{ |
| 465 | hopHeader: "1", |
| 466 | hopAuthHeader: signHop(key, "moved.test", time.Now()), |
| 467 | }) |
| 468 | if resp.StatusCode != http.StatusServiceUnavailable { |
| 469 | t.Fatalf("refusal #%d: status = %d, want 503", i+1, resp.StatusCode) |
| 470 | } |
| 471 | } |
| 472 | got := buf.String() |
| 473 | if n := strings.Count(got, "moved.test"); n != 2 { |
| 474 | t.Fatalf("log mentions the stale host %d time(s), want exactly 2 (occurrence #1 and #10, not one line per refusal): %s", n, got) |
| 475 | } |
| 476 | if !strings.Contains(got, "(#1)") { |
| 477 | t.Fatalf("log output = %q, want the first occurrence logged immediately", got) |
| 478 | } |
| 479 | if !strings.Contains(got, "(#10)") { |
| 480 | t.Fatalf("log output = %q, want the tenth occurrence logged", got) |
| 481 | } |
| 482 | if !strings.Contains(got, "box annie") { |
| 483 | t.Fatalf("log output = %q, want this box's own label named in the line", got) |
| 484 | } |
| 485 | } |
| 486 | |
| 487 | // Review catch on this task: host is client-controlled (a wildcard-homed |
| 488 | // route lets a sending box mint a signed hop for any invented subdomain, |
| 489 | // staleHopN's own doc), so staleHopN must not grow without bound. Drives |
| 490 | // logStaleHop directly rather than through HTTP, so cap-plus-a-few distinct |
| 491 | // hosts stays a cheap, focused test of the invariant: the map's key count |
| 492 | // stops at cap+1 (every tracked key plus the one shared overflow bucket), |
| 493 | // and hosts past the cap accumulate under that bucket instead of growing |
| 494 | // the map further. |
| 495 | func TestStaleHopNIsCappedWithAnOverflowBucket(t *testing.T) { |
| 496 | _, m, _ := hopEdge(t, "annie", "local.test") |
| 497 | |
| 498 | for i := 0; i < staleHopKeyCap; i++ { |
| 499 | m.logStaleHop(fmt.Sprintf("host-%d.example.com", i)) |
| 500 | } |
| 501 | if got := len(m.staleHopN); got != staleHopKeyCap { |
| 502 | t.Fatalf("staleHopN has %d key(s) at the cap, want exactly %d", got, staleHopKeyCap) |
| 503 | } |
| 504 | |
| 505 | // Three MORE distinct hosts, past the cap. |
| 506 | for i := 0; i < 3; i++ { |
| 507 | m.logStaleHop(fmt.Sprintf("overflow-%d.example.com", i)) |
| 508 | } |
| 509 | if got, want := len(m.staleHopN), staleHopKeyCap+1; got != want { |
| 510 | t.Fatalf("staleHopN has %d key(s) past the cap, want %d (cap distinct keys + one overflow bucket, not %d+3)", got, want, staleHopKeyCap) |
| 511 | } |
| 512 | const overflowKey = "annie (other)" |
| 513 | if n := m.staleHopN[overflowKey]; n != 3 { |
| 514 | t.Fatalf("overflow key %q count = %d, want 3 — each new host past the cap must fold into the shared bucket, not mint its own", overflowKey, n) |
| 515 | } |
| 516 | |
| 517 | // A key that was already tracked BEFORE the cap keeps its own identity |
| 518 | // and cadence — the cap only refuses NEW keys, never demotes an |
| 519 | // existing one. |
| 520 | m.logStaleHop("host-0.example.com") |
| 521 | if n := m.staleHopN["annie host-0.example.com"]; n != 2 { |
| 522 | t.Fatalf("a pre-cap key's count = %d, want 2 — it must keep incrementing under its own key, not fold into overflow", n) |
| 523 | } |
| 524 | } |
| 525 | |
| 526 | // Verification failure is not an error path: it is "you are an ordinary |
| 527 | // client", with every forwarding claim you made discarded. |
| 528 | func TestUnverifiedHopIsServedAsAnOrdinaryClientWithForwardedHeadersStripped(t *testing.T) { |
| 529 | e, _, key := hopEdge(t, "annie", "app.test") |
| 530 | tampered := []byte(signHop(key, "app.test", time.Now())) |
| 531 | tampered[len(tampered)-1] ^= 'f' ^ '0' // flip the last hex nibble |
| 532 | |
| 533 | resp := hopRequest(t, e, "app.test", map[string]string{ |
| 534 | hopHeader: "1", |
| 535 | hopAuthHeader: string(tampered), |
| 536 | "X-Forwarded-For": "203.0.113.9", |
| 537 | "X-Forwarded-Host": "evil.test", |
| 538 | "X-Forwarded-Proto": "https", |
| 539 | }) |
| 540 | if resp.StatusCode != http.StatusOK { |
| 541 | t.Fatalf("status = %d, want 200 — an unverified hop is an ordinary client, not a rejection", resp.StatusCode) |
| 542 | } |
| 543 | body := readBody(t, resp) |
| 544 | if strings.Contains(body, "203.0.113.9") { |
| 545 | t.Fatalf("body = %q — an unverified sender's X-Forwarded-For must not be honoured", body) |
| 546 | } |
| 547 | if strings.Contains(body, "evil.test") { |
| 548 | t.Fatalf("body = %q — an unverified sender's X-Forwarded-Host must not be honoured", body) |
| 549 | } |
| 550 | if !strings.Contains(body, "|hop=|auth=") { |
| 551 | t.Fatalf("body = %q — the hop headers must be stripped either way", body) |
| 552 | } |
| 553 | } |
| 554 | |
| 555 | // The strip is at the header level, not just a proxy side effect: every route |
| 556 | // class behind the edge (static trees, the status host, the activator) reads |
| 557 | // the same request. |
| 558 | func TestAcceptHopStripsEveryForwardedHeaderOnFailedVerification(t *testing.T) { |
| 559 | e, _, _ := hopEdge(t, "annie", "app.test") |
| 560 | r := httptest.NewRequest("GET", "http://app.test/", nil) |
| 561 | r.Header.Set(hopHeader, "1") |
| 562 | r.Header.Set(hopAuthHeader, "0:deadbeef") |
| 563 | r.Header.Set("Forwarded", "for=203.0.113.9") |
| 564 | r.Header.Set("X-Forwarded-For", "203.0.113.9") |
| 565 | r.Header.Set("X-Forwarded-Host", "evil.test") |
| 566 | r.Header.Set("X-Forwarded-Proto", "https") |
| 567 | |
| 568 | r, verified := e.acceptHop(r, "app.test") |
| 569 | if verified { |
| 570 | t.Fatal("a garbage auth header must not verify") |
| 571 | } |
| 572 | for _, h := range []string{"Forwarded", "X-Forwarded-For", "X-Forwarded-Host", "X-Forwarded-Proto", hopHeader, hopAuthHeader} { |
| 573 | if got := r.Header.Get(h); got != "" { |
| 574 | t.Fatalf("%s survived an unverified hop as %q", h, got) |
| 575 | } |
| 576 | } |
| 577 | if hopVerified(r.Context()) { |
| 578 | t.Fatal("an unverified request must not carry the hop mark") |
| 579 | } |
| 580 | |
| 581 | // The half-formed case: a signature with no X-Carlos-Hop beside it. It is |
| 582 | // not a hop and must not be treated as one — but the header is still this |
| 583 | // edge's vocabulary, and leaving it on the request would let an app that |
| 584 | // reflects its request headers launder it back onto the wire (the |
| 585 | // scrubInboundPurgeHeader precedent, purge_header.go). |
| 586 | stray := httptest.NewRequest("GET", "http://app.test/", nil) |
| 587 | stray.Header.Set(hopAuthHeader, "0:deadbeef") |
| 588 | stray, verified = e.acceptHop(stray, "app.test") |
| 589 | if verified { |
| 590 | t.Fatal("an auth header with no hop marker is not a hop") |
| 591 | } |
| 592 | if got := stray.Header.Get(hopAuthHeader); got != "" { |
| 593 | t.Fatalf("%s survived without a hop marker as %q — hop vocabulary must never reach the app", hopAuthHeader, got) |
| 594 | } |
| 595 | } |
| 596 | |
| 597 | // The MAC is over the host, so a signature minted for one hop cannot be |
| 598 | // replayed against another — and the timestamp bounds it in time. |
| 599 | func TestHopVerificationIsBoundToTheHostAndTheClock(t *testing.T) { |
| 600 | e, _, key := hopEdge(t, "annie", "app.test", "other.test") |
| 601 | |
| 602 | wrongHost := hopRequest(t, e, "app.test", map[string]string{ |
| 603 | hopHeader: "1", |
| 604 | hopAuthHeader: signHop(key, "other.test", time.Now()), |
| 605 | "X-Forwarded-For": "203.0.113.9", |
| 606 | }) |
| 607 | if body := readBody(t, wrongHost); strings.Contains(body, "203.0.113.9") { |
| 608 | t.Fatalf("body = %q — a signature minted for another host must not verify", body) |
| 609 | } |
| 610 | |
| 611 | stale := hopRequest(t, e, "app.test", map[string]string{ |
| 612 | hopHeader: "1", |
| 613 | hopAuthHeader: signHop(key, "app.test", time.Now().Add(-2*hopSkew)), |
| 614 | "X-Forwarded-For": "203.0.113.9", |
| 615 | }) |
| 616 | if body := readBody(t, stale); strings.Contains(body, "203.0.113.9") { |
| 617 | t.Fatalf("body = %q — a signature outside the skew bound must not verify", body) |
| 618 | } |
| 619 | } |
| 620 | |
| 621 | // Fail closed: with no mesh key mirrored (a brand new deployment, or a box |
| 622 | // whose bucket is unreachable at boot) nothing verifies. |
| 623 | func TestHopWithoutAMirroredMeshKeyIsAnOrdinaryClient(t *testing.T) { |
| 624 | e, _, key := hopEdge(t, "annie", "app.test") |
| 625 | e.homes.Store(nil) // no mesh at all — the no-deployment-store case |
| 626 | resp := hopRequest(t, e, "app.test", map[string]string{ |
| 627 | hopHeader: "1", |
| 628 | hopAuthHeader: signHop(key, "app.test", time.Now()), |
| 629 | "X-Forwarded-For": "203.0.113.9", |
| 630 | }) |
| 631 | if resp.StatusCode != http.StatusOK { |
| 632 | t.Fatalf("status = %d, want 200", resp.StatusCode) |
| 633 | } |
| 634 | if body := readBody(t, resp); strings.Contains(body, "203.0.113.9") { |
| 635 | t.Fatalf("body = %q — with no key to verify against, the sender is an ordinary client", body) |
| 636 | } |
| 637 | } |
| 638 | |
| 639 | // An unserved host that NO box homes keeps today's answer exactly: 404, with |
| 640 | // no Retry-After, and nothing said about the rest of the deployment. |
| 641 | // |
| 642 | // The home map here is deliberately non-empty and deliberately about a |
| 643 | // DIFFERENT host: the map is consulted (Task 4's backhaul) and misses, which |
| 644 | // is the only route to this answer now that a homed host is dialled instead. |
| 645 | // The stale-map 503 stays reachable only by a party holding the mesh key, so |
| 646 | // the two answers still cannot be told apart by a stranger. |
| 647 | func TestUnknownHostWithNoHomeIsStillThe404(t *testing.T) { |
| 648 | e, m, _ := hopEdge(t, "annie", "app.test") |
| 649 | putJSON(t, m.s, "control/edges/betty.json", |
| 650 | `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"moved.test":{"role":"home"}}}`) |
| 651 | refreshHomeMap(t, m) |
| 652 | resp := hopRequest(t, e, "nowhere.test", nil) |
| 653 | if resp.StatusCode != http.StatusNotFound { |
| 654 | t.Fatalf("status = %d, want the unchanged 404 for a host nothing in the deployment homes", resp.StatusCode) |
| 655 | } |
| 656 | if got := resp.Header.Get("Retry-After"); got != "" { |
| 657 | t.Fatalf("Retry-After = %q — a stranger must not be told to come back", got) |
| 658 | } |
| 659 | } |
| 660 | |
| 661 | // Regression pin: an ordinary client's forged X-Forwarded-For is no more |
| 662 | // trusted after this change than before it. |
| 663 | func TestOrdinaryClientForgedXFFIsStillDiscarded(t *testing.T) { |
| 664 | e, _, _ := hopEdge(t, "annie", "app.test") |
| 665 | resp := hopRequest(t, e, "app.test", map[string]string{"X-Forwarded-For": "203.0.113.9"}) |
| 666 | if resp.StatusCode != http.StatusOK { |
| 667 | t.Fatalf("status = %d, want 200", resp.StatusCode) |
| 668 | } |
| 669 | if body := readBody(t, resp); strings.Contains(body, "203.0.113.9") { |
| 670 | t.Fatalf("body = %q — a client-supplied XFF is never honoured", body) |
| 671 | } |
| 672 | } |
| 673 | |
| 674 | // ---- backhaul dialing (the sending half) ------------------------------------ |
| 675 | |
| 676 | // homeEdgeSeen is what a fake home edge was handed on the wire. |
| 677 | type homeEdgeSeen struct { |
| 678 | n int |
| 679 | sni, host, hop, auth string |
| 680 | proof string |
| 681 | xff []string |
| 682 | } |
| 683 | |
| 684 | type homeEdgeRecord struct { |
| 685 | mu sync.Mutex |
| 686 | seen homeEdgeSeen |
| 687 | } |
| 688 | |
| 689 | func (h *homeEdgeRecord) get() homeEdgeSeen { |
| 690 | h.mu.Lock() |
| 691 | defer h.mu.Unlock() |
| 692 | return h.seen |
| 693 | } |
| 694 | |
| 695 | const homeEdgeBody = "served by the home edge" |
| 696 | |
| 697 | // fakeHomeEdge stands in for the box that homes a host: a TLS server that |
| 698 | // records the ClientHello's SNI and the trust-bearing headers of whatever |
| 699 | // arrives, then runs h (a plain body writer unless the test needs a protocol |
| 700 | // switch). |
| 701 | // |
| 702 | // It also arms the two dial seams for the test's lifetime — backhaulPort, so |
| 703 | // the dial reaches an ephemeral httptest port instead of 443, and |
| 704 | // backhaulRootCAs, so this server's self-signed certificate is trusted the way |
| 705 | // a real home edge's publicly-trusted one is. Nothing about verification is |
| 706 | // weakened: httptest's certificate names example.com, so every test below asks |
| 707 | // for exactly that host and a name mismatch would fail the handshake here just |
| 708 | // as it would in production. |
| 709 | func fakeHomeEdge(t *testing.T, h http.Handler) (*homeEdgeRecord, string) { |
| 710 | t.Helper() |
| 711 | rec := &homeEdgeRecord{} |
| 712 | if h == nil { |
| 713 | h = http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { fmt.Fprint(w, homeEdgeBody) }) |
| 714 | } |
| 715 | ts := httptest.NewUnstartedServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { |
| 716 | rec.mu.Lock() |
| 717 | rec.seen.n++ |
| 718 | rec.seen.host = r.Host |
| 719 | rec.seen.hop = r.Header.Get(hopHeader) |
| 720 | rec.seen.auth = r.Header.Get(hopAuthHeader) |
| 721 | rec.seen.xff = r.Header.Values("X-Forwarded-For") |
| 722 | rec.seen.proof = r.Header.Get(edgeproof.Header) |
| 723 | rec.mu.Unlock() |
| 724 | h.ServeHTTP(w, r) |
| 725 | })) |
| 726 | ts.TLS = &tls.Config{GetConfigForClient: func(chi *tls.ClientHelloInfo) (*tls.Config, error) { |
| 727 | rec.mu.Lock() |
| 728 | rec.seen.sni = chi.ServerName |
| 729 | rec.mu.Unlock() |
| 730 | return nil, nil // the server's own config, unchanged — this hook only watches |
| 731 | }} |
| 732 | ts.StartTLS() |
| 733 | t.Cleanup(ts.Close) |
| 734 | |
| 735 | eip, port, err := net.SplitHostPort(ts.Listener.Addr().String()) |
| 736 | if err != nil { |
| 737 | t.Fatalf("split home edge addr: %v", err) |
| 738 | } |
| 739 | pool := x509.NewCertPool() |
| 740 | pool.AddCert(ts.Certificate()) |
| 741 | oldPort, oldCAs := backhaulPort, backhaulRootCAs |
| 742 | backhaulPort, backhaulRootCAs = port, pool |
| 743 | t.Cleanup(func() { backhaulPort, backhaulRootCAs = oldPort, oldCAs }) |
| 744 | return rec, eip |
| 745 | } |
| 746 | |
| 747 | // homedOn publishes one box's edges file naming hosts as its own, and lands it |
| 748 | // in the map — the second box a single-box deployment never has. |
| 749 | func homedOn(t *testing.T, m *homeMap, box, eip string, hosts ...string) { |
| 750 | t.Helper() |
| 751 | entries := make([]string, 0, len(hosts)) |
| 752 | for _, h := range hosts { |
| 753 | entries = append(entries, fmt.Sprintf("%q:{\"role\":\"home\"}", h)) |
| 754 | } |
| 755 | putJSON(t, m.s, "control/edges/"+box+".json", fmt.Sprintf( |
| 756 | `{"v":1,"box":%q,"eip":%q,"hosts":{%s}}`, box, eip, strings.Join(entries, ","))) |
| 757 | refreshHomeMap(t, m) |
| 758 | } |
| 759 | |
| 760 | // The whole of the sending half in one request: a host this box does not serve |
| 761 | // but another box homes reaches that box over real TLS, with the SNI it will |
| 762 | // need to present a certificate for, the real Host header, a hop signature the |
| 763 | // home edge can verify, and an X-Forwarded-For that names only the client. |
| 764 | func TestBackhaulDialsTheHomeEdgeOverVerifiedTLS(t *testing.T) { |
| 765 | e, m, key := hopEdge(t, "annie", "local.test") |
| 766 | rec, eip := fakeHomeEdge(t, nil) |
| 767 | homedOn(t, m, "betty", eip, "example.com") |
| 768 | |
| 769 | resp := hopRequest(t, e, "example.com", nil) |
| 770 | if resp.StatusCode != http.StatusOK { |
| 771 | t.Fatalf("status = %d, want 200 — a remote-homed host must be served through its home edge", resp.StatusCode) |
| 772 | } |
| 773 | if body := readBody(t, resp); body != homeEdgeBody { |
| 774 | t.Fatalf("body = %q, want %q", body, homeEdgeBody) |
| 775 | } |
| 776 | seen := rec.get() |
| 777 | if seen.n != 1 { |
| 778 | t.Fatalf("home edge saw %d request(s), want exactly one — one hop, ever", seen.n) |
| 779 | } |
| 780 | if seen.sni != "example.com" { |
| 781 | t.Fatalf("SNI = %q, want the request host — the home edge answers with the cert for the host being asked for", seen.sni) |
| 782 | } |
| 783 | if seen.host != "example.com" { |
| 784 | t.Fatalf("Host = %q, want the real hostname — the home edge routes on it", seen.host) |
| 785 | } |
| 786 | if seen.hop != "1" { |
| 787 | t.Fatalf("%s = %q, want 1", hopHeader, seen.hop) |
| 788 | } |
| 789 | if !verifyHopEitherKey(key, nil, "example.com", seen.auth, time.Now()) { |
| 790 | t.Fatalf("%s = %q did not verify under the deployment's mesh key", hopAuthHeader, seen.auth) |
| 791 | } |
| 792 | // One element, the client. Neither this edge's own address nor the home EIP |
| 793 | // may be appended: the home edge restores this chain verbatim on a verified |
| 794 | // hop, and every consumer takes the LAST element as the caller |
| 795 | // (setForwarded, homemap.go). |
| 796 | if len(seen.xff) != 1 || strings.Contains(seen.xff[0], ",") { |
| 797 | t.Fatalf("X-Forwarded-For = %q, want a single-element chain naming only the client", seen.xff) |
| 798 | } |
| 799 | } |
| 800 | |
| 801 | // Precedence, first rule: a host served HERE is served here, whatever the home |
| 802 | // map says about it. The map is poisoned to name a reachable home for exactly |
| 803 | // that host, so a lookup in the wrong order would show up as a dial. |
| 804 | func TestLocalHitIgnoresTheHomeMap(t *testing.T) { |
| 805 | e, m, _ := hopEdge(t, "annie", "example.com") |
| 806 | rec, eip := fakeHomeEdge(t, nil) |
| 807 | homedOn(t, m, "betty", eip, "example.com") |
| 808 | |
| 809 | resp := hopRequest(t, e, "example.com", nil) |
| 810 | if resp.StatusCode != http.StatusOK { |
| 811 | t.Fatalf("status = %d, want 200", resp.StatusCode) |
| 812 | } |
| 813 | if body := readBody(t, resp); !strings.HasPrefix(body, "xff=") { |
| 814 | t.Fatalf("body = %q, want the LOCAL backend's answer — a local hit must never consult the home map", body) |
| 815 | } |
| 816 | if n := rec.get().n; n != 0 { |
| 817 | t.Fatalf("home edge saw %d request(s), want none", n) |
| 818 | } |
| 819 | } |
| 820 | |
| 821 | // One hop, ever: a request that ARRIVED as a verified hop is refused, never |
| 822 | // forwarded onward, even when the map offers a live target for it. The fake |
| 823 | // home edge is the zero-call witness — a re-forward would answer 200 here. |
| 824 | func TestArrivedHopNeverDialsTheHomeEdge(t *testing.T) { |
| 825 | e, m, key := hopEdge(t, "annie", "local.test") |
| 826 | rec, eip := fakeHomeEdge(t, nil) |
| 827 | homedOn(t, m, "betty", eip, "example.com") |
| 828 | |
| 829 | resp := hopRequest(t, e, "example.com", map[string]string{ |
| 830 | hopHeader: "1", |
| 831 | hopAuthHeader: signHop(key, "example.com", time.Now()), |
| 832 | }) |
| 833 | if resp.StatusCode != http.StatusServiceUnavailable { |
| 834 | t.Fatalf("status = %d, want 503 — an arrived hop is refused, not re-forwarded", resp.StatusCode) |
| 835 | } |
| 836 | if got := resp.Header.Get("Retry-After"); got != "2" { |
| 837 | t.Fatalf("Retry-After = %q, want 2", got) |
| 838 | } |
| 839 | if n := rec.get().n; n != 0 { |
| 840 | t.Fatalf("home edge saw %d request(s), want none — a hop must never become a second hop", n) |
| 841 | } |
| 842 | } |
| 843 | |
| 844 | // The backhaul's own guard, exercised directly rather than through ServeHTTP's |
| 845 | // branch above it: if this call ever moves, an arrived hop must still fail |
| 846 | // closed here instead of being forwarded. |
| 847 | func TestBackhaulHelperRefusesAnArrivedHopOnItsOwn(t *testing.T) { |
| 848 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 849 | rec, eip := fakeHomeEdge(t, nil) |
| 850 | homedOn(t, m, "betty", eip, "example.com") |
| 851 | |
| 852 | r := httptest.NewRequest("GET", "http://example.com/", nil) |
| 853 | r = r.WithContext(context.WithValue(r.Context(), hopVerifiedKey{}, true)) |
| 854 | w := httptest.NewRecorder() |
| 855 | if !e.backhaul(w, r, "example.com") { |
| 856 | t.Fatal("backhaul must ANSWER an arrived hop, not fall through to the unknown-host path") |
| 857 | } |
| 858 | if w.Code != http.StatusServiceUnavailable { |
| 859 | t.Fatalf("status = %d, want the stale-map refusal", w.Code) |
| 860 | } |
| 861 | if n := rec.get().n; n != 0 { |
| 862 | t.Fatalf("home edge saw %d request(s), want none", n) |
| 863 | } |
| 864 | } |
| 865 | |
| 866 | // Fail closed on the SENDING side too: with no mesh key mirrored there is |
| 867 | // nothing to prove the hop with, and signing under an empty key would cross a |
| 868 | // region only to be served as an anonymous client. The request gets today's |
| 869 | // unknown-host answer instead. |
| 870 | func TestBackhaulRefusesToDialWithoutAMeshKey(t *testing.T) { |
| 871 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 872 | rec, eip := fakeHomeEdge(t, nil) |
| 873 | homedOn(t, m, "betty", eip, "example.com") |
| 874 | if err := m.s.Delete(context.Background(), meshKeyKey); err != nil { |
| 875 | t.Fatalf("delete mesh key: %v", err) |
| 876 | } |
| 877 | refreshHomeMap(t, m) |
| 878 | if cur, next := m.meshKeys(); len(cur) != 0 || len(next) != 0 { |
| 879 | t.Fatal("precondition: the mirror must hold no key") |
| 880 | } |
| 881 | |
| 882 | resp := hopRequest(t, e, "example.com", nil) |
| 883 | if resp.StatusCode != http.StatusNotFound { |
| 884 | t.Fatalf("status = %d, want the unchanged 404 — an unprovable hop is not dialled", resp.StatusCode) |
| 885 | } |
| 886 | if n := rec.get().n; n != 0 { |
| 887 | t.Fatalf("home edge saw %d request(s), want none — never sign with an empty key", n) |
| 888 | } |
| 889 | } |
| 890 | |
| 891 | // A stranger forging the hop vocabulary for a remote-homed host is backhauled |
| 892 | // like any other client, and gains exactly nothing by having tried. |
| 893 | // |
| 894 | // Two halves have to hold at once and they meet only here. acceptHop deletes |
| 895 | // both hop headers and — verification having failed — strips every forwarding |
| 896 | // claim the forger made, so the request reaching this point is an ordinary |
| 897 | // one; backhaul then mints a FRESH signature under the deployment's own key, |
| 898 | // so the home edge sees a hop it can verify and an X-Forwarded-For naming the |
| 899 | // forger. What the forger must NOT get is the stale-map 503 (which would tell |
| 900 | // them the host lives elsewhere), a chain of their own choosing, or their |
| 901 | // garbage signature carried onward under this edge's name. |
| 902 | func TestForgedHopForARemoteHostIsBackhauledAsAnOrdinaryClient(t *testing.T) { |
| 903 | e, m, key := hopEdge(t, "annie", "local.test") |
| 904 | rec, eip := fakeHomeEdge(t, nil) |
| 905 | homedOn(t, m, "betty", eip, "example.com") |
| 906 | |
| 907 | const garbage = "0:deadbeef" |
| 908 | resp := hopRequest(t, e, "example.com", map[string]string{ |
| 909 | hopHeader: "1", |
| 910 | hopAuthHeader: garbage, |
| 911 | "X-Forwarded-For": "203.0.113.9", |
| 912 | }) |
| 913 | if resp.StatusCode != http.StatusOK { |
| 914 | t.Fatalf("status = %d, want 200 — a forged hop is an ordinary client, not a refusal", resp.StatusCode) |
| 915 | } |
| 916 | if got := resp.Header.Get("Retry-After"); got != "" { |
| 917 | t.Fatalf("Retry-After = %q — the stale-map answer must stay unreachable without the mesh key", got) |
| 918 | } |
| 919 | if body := readBody(t, resp); body != homeEdgeBody { |
| 920 | t.Fatalf("body = %q, want the home edge's answer %q", body, homeEdgeBody) |
| 921 | } |
| 922 | seen := rec.get() |
| 923 | if seen.n != 1 { |
| 924 | t.Fatalf("home edge saw %d request(s), want exactly one", seen.n) |
| 925 | } |
| 926 | if seen.auth == garbage { |
| 927 | t.Fatalf("%s = %q — the forger's signature must never be forwarded under this edge's name", hopAuthHeader, seen.auth) |
| 928 | } |
| 929 | if seen.hop != "1" || !verifyHopEitherKey(key, nil, "example.com", seen.auth, time.Now()) { |
| 930 | t.Fatalf("hop=%q auth=%q — the hop must carry a FRESH signature this edge minted", seen.hop, seen.auth) |
| 931 | } |
| 932 | // The forger's claimed chain is gone and the one that arrives names the |
| 933 | // forger's own address, once. Anything else would let a stranger choose |
| 934 | // what the app (and the console's rate limiter) believes about who called. |
| 935 | if len(seen.xff) != 1 || strings.Contains(seen.xff[0], ",") { |
| 936 | t.Fatalf("X-Forwarded-For = %q, want a single-element chain", seen.xff) |
| 937 | } |
| 938 | if strings.Contains(seen.xff[0], "203.0.113.9") { |
| 939 | t.Fatalf("X-Forwarded-For = %q — a forged chain must be discarded, not carried", seen.xff) |
| 940 | } |
| 941 | if ip := net.ParseIP(seen.xff[0]); ip == nil || !ip.IsLoopback() { |
| 942 | t.Fatalf("X-Forwarded-For = %q, want the forger's own (loopback) address", seen.xff) |
| 943 | } |
| 944 | } |
| 945 | |
| 946 | // A catch-all row in another box's edges file resolves through the same |
| 947 | // one-label wildcard rule lookupUpstream uses — but the signature must be over |
| 948 | // the CONCRETE request host, because that is what the home edge verifies |
| 949 | // against. Signing the map KEY would produce a MAC for "*.com", which is not a |
| 950 | // request host under any canonicalization. |
| 951 | func TestBackhaulSignsTheConcreteHostNotTheWildcardKey(t *testing.T) { |
| 952 | e, m, key := hopEdge(t, "annie", "local.test") |
| 953 | rec, eip := fakeHomeEdge(t, nil) |
| 954 | homedOn(t, m, "betty", eip, "*.com") |
| 955 | if _, _, ok := m.Home("example.com"); !ok { |
| 956 | t.Fatal("precondition: the wildcard row must offer a home for example.com") |
| 957 | } |
| 958 | |
| 959 | if resp := hopRequest(t, e, "example.com", nil); resp.StatusCode != http.StatusOK { |
| 960 | t.Fatalf("status = %d, want 200", resp.StatusCode) |
| 961 | } |
| 962 | seen := rec.get() |
| 963 | if !verifyHop(key, "example.com", seen.auth, time.Now()) { |
| 964 | t.Fatalf("%s = %q must verify against the concrete request host", hopAuthHeader, seen.auth) |
| 965 | } |
| 966 | if verifyHop(key, "*.com", seen.auth, time.Now()) { |
| 967 | t.Fatalf("%s = %q verifies against the wildcard map key — the MAC must bind the concrete host", hopAuthHeader, seen.auth) |
| 968 | } |
| 969 | } |
| 970 | |
| 971 | // A WebSocket must compose through the hop with no code of its own: the dial |
| 972 | // branch sits above ServeHTTP's cache/compression wrapping, so the client's |
| 973 | // hijackable writer reaches httputil.ReverseProxy's protocol-switch path, and |
| 974 | // the transport speaks HTTP/1.1 (no ForceAttemptHTTP2) so that path works. |
| 975 | // Bytes must then flow both ways with no whole-body deadline cutting them off. |
| 976 | func TestBackhaulCarriesAWebSocketUpgrade(t *testing.T) { |
| 977 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 978 | rec, eip := fakeHomeEdge(t, http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { |
| 979 | hj, ok := w.(http.Hijacker) |
| 980 | if !ok { |
| 981 | http.Error(w, "not hijackable", http.StatusInternalServerError) |
| 982 | return |
| 983 | } |
| 984 | conn, buf, err := hj.Hijack() |
| 985 | if err != nil { |
| 986 | return |
| 987 | } |
| 988 | defer conn.Close() |
| 989 | buf.WriteString("HTTP/1.1 101 Switching Protocols\r\nUpgrade: websocket\r\nConnection: Upgrade\r\n\r\n") |
| 990 | if err := buf.Flush(); err != nil { |
| 991 | return |
| 992 | } |
| 993 | line, err := buf.ReadString('\n') |
| 994 | if err != nil { |
| 995 | return |
| 996 | } |
| 997 | buf.WriteString("home:" + line) |
| 998 | buf.Flush() |
| 999 | })) |
| 1000 | homedOn(t, m, "betty", eip, "example.com") |
| 1001 | |
| 1002 | srv := httptest.NewServer(e) |
| 1003 | t.Cleanup(srv.Close) |
| 1004 | c, err := net.Dial("tcp", strings.TrimPrefix(srv.URL, "http://")) |
| 1005 | if err != nil { |
| 1006 | t.Fatalf("dial edge: %v", err) |
| 1007 | } |
| 1008 | defer c.Close() |
| 1009 | if err := c.SetDeadline(time.Now().Add(10 * time.Second)); err != nil { |
| 1010 | t.Fatalf("set deadline: %v", err) |
| 1011 | } |
| 1012 | if _, err := fmt.Fprint(c, "GET /socket HTTP/1.1\r\nHost: example.com\r\nConnection: Upgrade\r\nUpgrade: websocket\r\n\r\n"); err != nil { |
| 1013 | t.Fatalf("write handshake: %v", err) |
| 1014 | } |
| 1015 | br := bufio.NewReader(c) |
| 1016 | resp, err := http.ReadResponse(br, nil) |
| 1017 | if err != nil { |
| 1018 | t.Fatalf("read handshake response: %v", err) |
| 1019 | } |
| 1020 | if resp.StatusCode != http.StatusSwitchingProtocols { |
| 1021 | t.Fatalf("handshake status = %d, want 101 — an upgrade must switch protocols through the hop", resp.StatusCode) |
| 1022 | } |
| 1023 | if _, err := fmt.Fprint(c, "ping\n"); err != nil { |
| 1024 | t.Fatalf("write post-upgrade frame: %v", err) |
| 1025 | } |
| 1026 | line, err := br.ReadString('\n') |
| 1027 | if err != nil { |
| 1028 | t.Fatalf("read post-upgrade frame: %v", err) |
| 1029 | } |
| 1030 | if line != "home:ping\n" { |
| 1031 | t.Fatalf("post-upgrade echo = %q, want %q — bytes must flow both ways across the backhaul", line, "home:ping\n") |
| 1032 | } |
| 1033 | if seen := rec.get(); seen.hop != "1" { |
| 1034 | t.Fatalf("%s = %q — an upgrade is a hop like any other", hopHeader, seen.hop) |
| 1035 | } |
| 1036 | } |
| 1037 | |
| 1038 | // A home whose edge cannot be reached is a bad gateway, not a 503: this box |
| 1039 | // believed a specific box was there. The unreachable target here is the |
| 1040 | // documented cert-less bootstrap window's shape too — a dial or handshake that |
| 1041 | // does not complete, healed by the next mirror poll. |
| 1042 | func TestBackhaulToAnUnreachableHomeIsA502(t *testing.T) { |
| 1043 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1044 | // A port nothing listens on, reached through the same seam the fake home |
| 1045 | // edge uses. 127.0.0.1 refuses immediately rather than hanging for the dial |
| 1046 | // budget. |
| 1047 | ln, err := net.Listen("tcp", "127.0.0.1:0") |
| 1048 | if err != nil { |
| 1049 | t.Fatalf("listen: %v", err) |
| 1050 | } |
| 1051 | eip, port, err := net.SplitHostPort(ln.Addr().String()) |
| 1052 | if err != nil { |
| 1053 | t.Fatalf("split: %v", err) |
| 1054 | } |
| 1055 | ln.Close() |
| 1056 | old := backhaulPort |
| 1057 | backhaulPort = port |
| 1058 | t.Cleanup(func() { backhaulPort = old }) |
| 1059 | homedOn(t, m, "betty", eip, "example.com") |
| 1060 | |
| 1061 | resp := hopRequest(t, e, "example.com", nil) |
| 1062 | if resp.StatusCode != http.StatusBadGateway { |
| 1063 | t.Fatalf("status = %d, want 502 for a home edge that could not be reached", resp.StatusCode) |
| 1064 | } |
| 1065 | if got := resp.Header.Get("Retry-After"); got != "" { |
| 1066 | t.Fatalf("Retry-After = %q — 503 + Retry-After is reserved for knowing nobody is home", got) |
| 1067 | } |
| 1068 | } |
| 1069 | |
| 1070 | // Task 1: an unreachable home used to log unthrottled, once per failed dial — |
| 1071 | // a home box down for any real stretch would fill the journal with one line |
| 1072 | // per request. It now goes through e.logThrottled, keyed "backhaul-error |
| 1073 | // "+box, so a run of failures against the SAME home box produces one line |
| 1074 | // (occurrence #1), not one per failure. |
| 1075 | func TestBackhaulErrorHandlerRoutesThroughLogThrottled(t *testing.T) { |
| 1076 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1077 | ln, err := net.Listen("tcp", "127.0.0.1:0") |
| 1078 | if err != nil { |
| 1079 | t.Fatalf("listen: %v", err) |
| 1080 | } |
| 1081 | eip, port, err := net.SplitHostPort(ln.Addr().String()) |
| 1082 | if err != nil { |
| 1083 | t.Fatalf("split: %v", err) |
| 1084 | } |
| 1085 | ln.Close() |
| 1086 | old := backhaulPort |
| 1087 | backhaulPort = port |
| 1088 | t.Cleanup(func() { backhaulPort = old }) |
| 1089 | homedOn(t, m, "betty", eip, "example.com") |
| 1090 | |
| 1091 | buf := captureLog(t) |
| 1092 | for i := 0; i < 3; i++ { |
| 1093 | resp := hopRequest(t, e, "example.com", nil) |
| 1094 | if resp.StatusCode != http.StatusBadGateway { |
| 1095 | t.Fatalf("attempt #%d: status = %d, want 502", i+1, resp.StatusCode) |
| 1096 | } |
| 1097 | } |
| 1098 | got := buf.String() |
| 1099 | // logThrottled's own format is "edge: <msg> (occurrence #N)" — it prints |
| 1100 | // no key of its own, so the key's effect is observed as throttling: three |
| 1101 | // failed dials against the SAME "backhaul-error "+box key must produce |
| 1102 | // exactly one logged line (#1), not three. |
| 1103 | if n := strings.Count(got, "backhaul"); n != 1 { |
| 1104 | t.Fatalf("log mentions the backhaul failure %d time(s), want exactly 1 (logThrottled's first-and-every-300th, not one line per failed dial): %s", n, got) |
| 1105 | } |
| 1106 | if !strings.Contains(got, "occurrence #1") { |
| 1107 | t.Fatalf("log output = %q, want the first occurrence logged (logThrottled's own format)", got) |
| 1108 | } |
| 1109 | if strings.Contains(got, "occurrence #2") || strings.Contains(got, "occurrence #3") { |
| 1110 | t.Fatalf("log output = %q, want the second and third failures throttled, not logged again", got) |
| 1111 | } |
| 1112 | if !strings.Contains(got, "example.com") || !strings.Contains(got, "betty") { |
| 1113 | t.Fatalf("log output = %q, want the host and the home box named", got) |
| 1114 | } |
| 1115 | } |
| 1116 | |
| 1117 | // One transport per home BOX. The pool is what makes a backhaul cost one |
| 1118 | // handshake rather than one per request; keying it per box rather than per |
| 1119 | // host is what keeps a catch-all home — where every invented subdomain is a |
| 1120 | // backhaul target — from growing the map without limit. Per-host identity |
| 1121 | // still exists, one layer down: the outbound authority is the hostname, so SNI |
| 1122 | // and pooled connections are per host inside the box's transport (the SNI |
| 1123 | // assertion in TestBackhaulDialsTheHomeEdgeOverVerifiedTLS is that half). |
| 1124 | func TestBackhaulTransportsAreCachedPerHomeBox(t *testing.T) { |
| 1125 | e, _, _ := hopEdge(t, "annie", "local.test") |
| 1126 | a := e.backhaulTransport("1.1.1.1") |
| 1127 | if b := e.backhaulTransport("1.1.1.1"); b != a { |
| 1128 | t.Fatal("the same home box must reuse one transport, or every request pays a fresh TLS handshake") |
| 1129 | } |
| 1130 | if b := e.backhaulTransport("2.2.2.2"); b == a { |
| 1131 | t.Fatal("a different home box must get its own transport — the dial address is baked into it") |
| 1132 | } |
| 1133 | if sn := a.TLSClientConfig.ServerName; sn != "" { |
| 1134 | t.Fatalf("ServerName = %q, want it unset — pinning one host per transport is what a wildcard home would multiply", sn) |
| 1135 | } |
| 1136 | if a.TLSClientConfig.InsecureSkipVerify { |
| 1137 | t.Fatal("the hop verifies the home edge's certificate — always") |
| 1138 | } |
| 1139 | } |
| 1140 | |
| 1141 | // Two hosts homed on one box share that box's transport, and each still |
| 1142 | // reaches the home edge under its OWN name: the shared transport is the memory |
| 1143 | // property, the per-host SNI is the correctness one. |
| 1144 | func TestBackhaulSharesOneTransportAcrossAWildcardHome(t *testing.T) { |
| 1145 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1146 | rec, eip := fakeHomeEdge(t, nil) |
| 1147 | homedOn(t, m, "betty", eip, "*.com") |
| 1148 | |
| 1149 | if resp := hopRequest(t, e, "example.com", nil); resp.StatusCode != http.StatusOK { |
| 1150 | t.Fatalf("status = %d, want 200", resp.StatusCode) |
| 1151 | } |
| 1152 | if sni := rec.get().sni; sni != "example.com" { |
| 1153 | t.Fatalf("SNI = %q, want the concrete request host", sni) |
| 1154 | } |
| 1155 | n := 0 |
| 1156 | e.backhaulTransports.Range(func(_, _ any) bool { n++; return true }) |
| 1157 | if n != 1 { |
| 1158 | t.Fatalf("%d transports cached for one home box, want 1 — a catch-all home must not mint one per hostname", n) |
| 1159 | } |
| 1160 | if _, ok := e.backhaulTransports.Load(eip); !ok { |
| 1161 | t.Fatalf("the cached transport must be keyed by the home box's EIP (%s)", eip) |
| 1162 | } |
| 1163 | } |
| 1164 | |
| 1165 | // The :80 half of the backhaul, carried from Task 4's review. A host homed on |
| 1166 | // another box is answered on 443 (the hop), so plain http:// must redirect it |
| 1167 | // there like any other routable host — otherwise a first-time visitor typing a |
| 1168 | // bare hostname in the wrong region gets a 404 for an app that is up. |
| 1169 | // |
| 1170 | // The redirect grants nothing that the 443 answer does not: it is a 302 to the |
| 1171 | // SAME host, never an ACME order (httpFallbackHandler's own doc records why |
| 1172 | // this gate is wider than Allowed's). And a verified hop can never arrive |
| 1173 | // here — the sending edge dials 443 — so there is no hop logic on this path. |
| 1174 | func TestHTTPFallbackRedirectsARemoteHomedHost(t *testing.T) { |
| 1175 | e, m, _ := hopEdge(t, "annie") // serves nothing itself |
| 1176 | homedOn(t, m, "betty", "2.2.2.2", "remote.test", "*.wild.test") |
| 1177 | h := e.httpFallbackHandler() |
| 1178 | |
| 1179 | for _, host := range []string{"remote.test", "paul.wild.test"} { |
| 1180 | req := httptest.NewRequest(http.MethodGet, "http://"+host+"/deep?q=1", nil) |
| 1181 | req.Host = host |
| 1182 | rec := httptest.NewRecorder() |
| 1183 | h.ServeHTTP(rec, req) |
| 1184 | if rec.Code != http.StatusFound { |
| 1185 | t.Fatalf("%s: status = %d, want 302 — the host is homed on another box and 443 backhauls it", host, rec.Code) |
| 1186 | } |
| 1187 | if loc := rec.Header().Get("Location"); loc != "https://"+host+"/deep?q=1" { |
| 1188 | t.Fatalf("%s: Location = %q", host, loc) |
| 1189 | } |
| 1190 | } |
| 1191 | |
| 1192 | // A host nothing in the deployment homes keeps today's answer exactly. |
| 1193 | req := httptest.NewRequest(http.MethodGet, "http://nowhere.test/", nil) |
| 1194 | req.Host = "nowhere.test" |
| 1195 | rec := httptest.NewRecorder() |
| 1196 | h.ServeHTTP(rec, req) |
| 1197 | if rec.Code != http.StatusNotFound { |
| 1198 | t.Fatalf("nowhere.test: status = %d, want the unchanged 404", rec.Code) |
| 1199 | } |
| 1200 | if loc := rec.Header().Get("Location"); loc != "" { |
| 1201 | t.Fatalf("nowhere.test: Location = %q, want none", loc) |
| 1202 | } |
| 1203 | } |
| 1204 | |
| 1205 | // ---- containment: the per-home-box in-flight cap -------------------------- |
| 1206 | |
| 1207 | // backhaulCap lowers the in-flight cap for one test and restores it after — |
| 1208 | // the package-seam pattern backhaulPort and homeMapInterval already follow. |
| 1209 | func backhaulCap(t *testing.T, n int64) { |
| 1210 | t.Helper() |
| 1211 | old := backhaulMaxInflight.Load() |
| 1212 | backhaulMaxInflight.Store(n) |
| 1213 | t.Cleanup(func() { backhaulMaxInflight.Store(old) }) |
| 1214 | } |
| 1215 | |
| 1216 | // backhaulGetter drives requests for host at one edge over a single reused |
| 1217 | // server, so several can be in flight at once (hopRequest stands a server per |
| 1218 | // call, which is fine for one-shot tests and useless for a concurrency one). |
| 1219 | func backhaulGetter(t *testing.T, e *Edge) func(host string) (*http.Response, error) { |
| 1220 | t.Helper() |
| 1221 | srv := httptest.NewServer(e) |
| 1222 | t.Cleanup(srv.Close) |
| 1223 | return func(host string) (*http.Response, error) { |
| 1224 | req, err := http.NewRequest(http.MethodGet, srv.URL+"/", nil) |
| 1225 | if err != nil { |
| 1226 | return nil, err |
| 1227 | } |
| 1228 | req.Host = host |
| 1229 | return http.DefaultClient.Do(req) |
| 1230 | } |
| 1231 | } |
| 1232 | |
| 1233 | // The cap itself: with two hops held open toward one home box, a third is |
| 1234 | // answered 503 + Retry-After: 2 HERE — no dial, no cross-region bytes — and |
| 1235 | // the slot a finished hop gives back is immediately usable again. |
| 1236 | // |
| 1237 | // This is what bounds a scanner's forged-Host flood: egress is (concurrent |
| 1238 | // streams × per-stream bandwidth), and this caps the first factor per box. |
| 1239 | func TestBackhaulRefusesAtInflightCap(t *testing.T) { |
| 1240 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1241 | backhaulCap(t, 2) |
| 1242 | |
| 1243 | started, hold := make(chan struct{}, 4), make(chan struct{}) |
| 1244 | rec, eip := fakeHomeEdge(t, http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { |
| 1245 | started <- struct{}{} |
| 1246 | <-hold |
| 1247 | fmt.Fprint(w, homeEdgeBody) |
| 1248 | })) |
| 1249 | homedOn(t, m, "betty", eip, "example.com") |
| 1250 | get := backhaulGetter(t, e) |
| 1251 | |
| 1252 | done := make(chan error, 2) |
| 1253 | for i := 0; i < 2; i++ { |
| 1254 | go func() { |
| 1255 | resp, err := get("example.com") |
| 1256 | if err != nil { |
| 1257 | done <- err |
| 1258 | return |
| 1259 | } |
| 1260 | defer resp.Body.Close() |
| 1261 | if resp.StatusCode != http.StatusOK { |
| 1262 | done <- fmt.Errorf("held hop: status = %d, want 200", resp.StatusCode) |
| 1263 | return |
| 1264 | } |
| 1265 | body, err := io.ReadAll(resp.Body) |
| 1266 | if err != nil { |
| 1267 | done <- err |
| 1268 | return |
| 1269 | } |
| 1270 | if string(body) != homeEdgeBody { |
| 1271 | done <- fmt.Errorf("held hop: body = %q", body) |
| 1272 | return |
| 1273 | } |
| 1274 | done <- nil |
| 1275 | }() |
| 1276 | } |
| 1277 | for i := 0; i < 2; i++ { |
| 1278 | select { |
| 1279 | case <-started: |
| 1280 | case err := <-done: |
| 1281 | t.Fatalf("a hop finished before the cap was full: %v", err) |
| 1282 | case <-time.After(15 * time.Second): |
| 1283 | t.Fatal("the two hops that fill the cap never reached the home edge") |
| 1284 | } |
| 1285 | } |
| 1286 | |
| 1287 | // Both slots are held. The third request must be refused by THIS box. |
| 1288 | dialled := rec.get().n |
| 1289 | resp, err := get("example.com") |
| 1290 | if err != nil { |
| 1291 | t.Fatalf("third request: %v", err) |
| 1292 | } |
| 1293 | defer resp.Body.Close() |
| 1294 | if resp.StatusCode != http.StatusServiceUnavailable { |
| 1295 | t.Fatalf("status = %d, want 503 at the in-flight cap", resp.StatusCode) |
| 1296 | } |
| 1297 | if got := resp.Header.Get("Retry-After"); got != "2" { |
| 1298 | t.Fatalf("Retry-After = %q, want 2 — the cap clears as hops finish", got) |
| 1299 | } |
| 1300 | if got := rec.get().n; got != dialled { |
| 1301 | t.Fatalf("the home edge saw %d request(s), want %d — a refusal must cost no dial and no cross-region bytes", got, dialled) |
| 1302 | } |
| 1303 | |
| 1304 | // Released slots are usable again: the cap is a bound on CONCURRENCY, not |
| 1305 | // a quota that a box burns through and never gets back. |
| 1306 | close(hold) |
| 1307 | for i := 0; i < 2; i++ { |
| 1308 | select { |
| 1309 | case err := <-done: |
| 1310 | if err != nil { |
| 1311 | t.Fatal(err) |
| 1312 | } |
| 1313 | case <-time.After(15 * time.Second): |
| 1314 | t.Fatal("a held hop never completed after its home edge was released") |
| 1315 | } |
| 1316 | } |
| 1317 | // Retried for a bounded window rather than asserted on the first attempt: |
| 1318 | // a held hop's client-side goroutine reports completion when the BODY is |
| 1319 | // read, which can be a hair before the edge's own deferred release runs. |
| 1320 | // The property under test is that the slots come back, not that they come |
| 1321 | // back within one scheduling quantum. |
| 1322 | var ( |
| 1323 | after *http.Response |
| 1324 | lastErr error |
| 1325 | ) |
| 1326 | deadline := time.Now().Add(10 * time.Second) |
| 1327 | for { |
| 1328 | after, lastErr = get("example.com") |
| 1329 | if lastErr != nil { |
| 1330 | t.Fatalf("fourth request: %v", lastErr) |
| 1331 | } |
| 1332 | if after.StatusCode == http.StatusOK { |
| 1333 | break |
| 1334 | } |
| 1335 | after.Body.Close() |
| 1336 | if after.StatusCode != http.StatusServiceUnavailable { |
| 1337 | t.Fatalf("status = %d after the held hops finished, want 200", after.StatusCode) |
| 1338 | } |
| 1339 | if time.Now().After(deadline) { |
| 1340 | t.Fatal("the cap never cleared after both held hops finished — a released slot was not given back") |
| 1341 | } |
| 1342 | time.Sleep(10 * time.Millisecond) |
| 1343 | } |
| 1344 | defer after.Body.Close() |
| 1345 | if got := rec.get().n; got < dialled+1 { |
| 1346 | t.Fatalf("the home edge saw %d request(s), want at least %d — the fourth request must dial normally", got, dialled+1) |
| 1347 | } |
| 1348 | } |
| 1349 | |
| 1350 | // The cap is per HOME BOX: one saturated box must not stop this edge serving |
| 1351 | // hosts homed on any other. A deployment-wide counter would let one busy (or |
| 1352 | // one attacked) box take the whole mesh down. |
| 1353 | func TestBackhaulInflightCapIsPerHomeBox(t *testing.T) { |
| 1354 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1355 | backhaulCap(t, 1) |
| 1356 | rec, eip := fakeHomeEdge(t, nil) |
| 1357 | homedOn(t, m, "carol", eip, "example.com") |
| 1358 | // Betty's EIP is never dialled in this test — her slot is filled directly |
| 1359 | // below, so the refusal happens before any dial. (TEST-NET-1: a dial |
| 1360 | // would hang for the whole 5s budget and be visible as a failure.) |
| 1361 | homedOn(t, m, "betty", "192.0.2.5", "saturated.test") |
| 1362 | |
| 1363 | release, ok := e.backhaulAcquire("betty") |
| 1364 | if !ok { |
| 1365 | t.Fatal("the first slot toward a box must always be available") |
| 1366 | } |
| 1367 | t.Cleanup(release) |
| 1368 | |
| 1369 | if resp := hopRequest(t, e, "saturated.test", nil); resp.StatusCode != http.StatusServiceUnavailable { |
| 1370 | t.Fatalf("status = %d, want 503 — betty's only slot is held", resp.StatusCode) |
| 1371 | } |
| 1372 | resp := hopRequest(t, e, "example.com", nil) |
| 1373 | if resp.StatusCode != http.StatusOK { |
| 1374 | t.Fatalf("status = %d, want 200 — carol's cap is her own", resp.StatusCode) |
| 1375 | } |
| 1376 | if body := readBody(t, resp); body != homeEdgeBody { |
| 1377 | t.Fatalf("body = %q, want the home edge's own answer", body) |
| 1378 | } |
| 1379 | if n := rec.get().n; n != 1 { |
| 1380 | t.Fatalf("carol saw %d request(s), want exactly one", n) |
| 1381 | } |
| 1382 | } |
| 1383 | |
| 1384 | // A slot is held for the LIFE of the stream, not until the response headers |
| 1385 | // arrive: a long-lived stream (a WebSocket, an SSE feed, a large download) is |
| 1386 | // precisely the sustained cross-region egress the cap exists to bound, so it |
| 1387 | // must count against the cap for as long as it flows. |
| 1388 | func TestBackhaulSlotHeldForStreamLife(t *testing.T) { |
| 1389 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1390 | backhaulCap(t, 1) |
| 1391 | |
| 1392 | streaming, finish := make(chan struct{}), make(chan struct{}) |
| 1393 | _, eip := fakeHomeEdge(t, http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { |
| 1394 | fmt.Fprint(w, "first ") |
| 1395 | w.(http.Flusher).Flush() // headers and a first chunk are on the wire |
| 1396 | close(streaming) |
| 1397 | <-finish |
| 1398 | fmt.Fprint(w, "last") |
| 1399 | })) |
| 1400 | homedOn(t, m, "betty", eip, "example.com") |
| 1401 | get := backhaulGetter(t, e) |
| 1402 | |
| 1403 | stream, err := get("example.com") |
| 1404 | if err != nil { |
| 1405 | t.Fatalf("stream request: %v", err) |
| 1406 | } |
| 1407 | defer stream.Body.Close() |
| 1408 | if stream.StatusCode != http.StatusOK { |
| 1409 | t.Fatalf("stream status = %d, want 200", stream.StatusCode) |
| 1410 | } |
| 1411 | select { |
| 1412 | case <-streaming: |
| 1413 | case <-time.After(15 * time.Second): |
| 1414 | t.Fatal("the streaming hop never reached the home edge") |
| 1415 | } |
| 1416 | |
| 1417 | // Headers have arrived and the body is still flowing: the slot is held. |
| 1418 | busy, err := get("example.com") |
| 1419 | if err != nil { |
| 1420 | t.Fatalf("second request: %v", err) |
| 1421 | } |
| 1422 | defer busy.Body.Close() |
| 1423 | if busy.StatusCode != http.StatusServiceUnavailable { |
| 1424 | t.Fatalf("status = %d while a hop is mid-stream, want 503 — the slot is held for the stream's life", busy.StatusCode) |
| 1425 | } |
| 1426 | |
| 1427 | close(finish) |
| 1428 | body, err := io.ReadAll(stream.Body) |
| 1429 | if err != nil { |
| 1430 | t.Fatalf("read stream: %v", err) |
| 1431 | } |
| 1432 | if string(body) != "first last" { |
| 1433 | t.Fatalf("stream body = %q, want both chunks", body) |
| 1434 | } |
| 1435 | stream.Body.Close() |
| 1436 | |
| 1437 | // And the moment it ends, the slot is back. |
| 1438 | after, err := get("example.com") |
| 1439 | if err != nil { |
| 1440 | t.Fatalf("third request: %v", err) |
| 1441 | } |
| 1442 | defer after.Body.Close() |
| 1443 | if after.StatusCode != http.StatusOK { |
| 1444 | t.Fatalf("status = %d after the stream ended, want 200 — the slot must have been released", after.StatusCode) |
| 1445 | } |
| 1446 | } |
| 1447 | |
| 1448 | // Saturation is one throttled journal line per home box, never silence and |
| 1449 | // never one line per refused request: a flood is a STANDING condition, and |
| 1450 | // the key is the box (bounded by the deployment) and never the host (which a |
| 1451 | // stranger picks, and which a catch-all home would multiply without limit). |
| 1452 | func TestBackhaulSaturationLogsThrottledPerBox(t *testing.T) { |
| 1453 | e, m, _ := hopEdge(t, "annie", "local.test") |
| 1454 | backhaulCap(t, 1) |
| 1455 | // Neither EIP is ever dialled: both boxes' only slot is held below, so |
| 1456 | // every request in this test is refused before the dial. |
| 1457 | homedOn(t, m, "betty", "192.0.2.5", "betty.test") |
| 1458 | homedOn(t, m, "carol", "192.0.2.6", "carol.test") |
| 1459 | for _, box := range []string{"betty", "carol"} { |
| 1460 | release, ok := e.backhaulAcquire(box) |
| 1461 | if !ok { |
| 1462 | t.Fatalf("%s: the first slot toward a box must always be available", box) |
| 1463 | } |
| 1464 | t.Cleanup(release) |
| 1465 | } |
| 1466 | |
| 1467 | buf := captureLog(t) |
| 1468 | for _, host := range []string{"betty.test", "betty.test", "carol.test"} { |
| 1469 | if resp := hopRequest(t, e, host, nil); resp.StatusCode != http.StatusServiceUnavailable { |
| 1470 | t.Fatalf("%s: status = %d, want 503", host, resp.StatusCode) |
| 1471 | } |
| 1472 | } |
| 1473 | |
| 1474 | got := buf.String() |
| 1475 | // logThrottled prints no key of its own, so key isolation is observed as |
| 1476 | // TWO first occurrences (one bucket per box) with betty's second refusal |
| 1477 | // throttled inside her own bucket. |
| 1478 | if n := strings.Count(got, "occurrence #1"); n != 2 { |
| 1479 | t.Fatalf("%d first-occurrence line(s), want one per home box (two buckets): %s", n, got) |
| 1480 | } |
| 1481 | if strings.Contains(got, "occurrence #2") { |
| 1482 | t.Fatalf("log output = %q, want the second refusal against the same box throttled", got) |
| 1483 | } |
| 1484 | if n := strings.Count(got, "in-flight cap"); n != 2 { |
| 1485 | t.Fatalf("%d saturation line(s), want exactly two: %s", n, got) |
| 1486 | } |
| 1487 | if !strings.Contains(got, "betty") || !strings.Contains(got, "carol") { |
| 1488 | t.Fatalf("log output = %q, want each saturated home box named", got) |
| 1489 | } |
| 1490 | } |
| 1491 | |
| 1492 | // The knob: unset, unparseable and non-positive all mean the default. There |
| 1493 | // is deliberately no "off" — an unbounded cross-region path is what the cap |
| 1494 | // closes — and a typo must not silently zero-cap a deployment either. |
| 1495 | func TestBackhaulMaxInflightFromEnv(t *testing.T) { |
| 1496 | // The ambient value is captured and restored by hand: t.Setenv restores |
| 1497 | // what IT set, but the unset case below has to clear the variable, and an |
| 1498 | // operator's own CARLOS_BACKHAUL_MAX_INFLIGHT must survive this test. |
| 1499 | ambient, had := os.LookupEnv("CARLOS_BACKHAUL_MAX_INFLIGHT") |
| 1500 | t.Cleanup(func() { |
| 1501 | if had { |
| 1502 | os.Setenv("CARLOS_BACKHAUL_MAX_INFLIGHT", ambient) |
| 1503 | return |
| 1504 | } |
| 1505 | os.Unsetenv("CARLOS_BACKHAUL_MAX_INFLIGHT") |
| 1506 | }) |
| 1507 | |
| 1508 | for _, tc := range []struct { |
| 1509 | set bool |
| 1510 | raw string |
| 1511 | want int |
| 1512 | }{ |
| 1513 | {set: false, want: defaultBackhaulMaxInflight}, |
| 1514 | {set: true, raw: "", want: defaultBackhaulMaxInflight}, |
| 1515 | {set: true, raw: " ", want: defaultBackhaulMaxInflight}, |
| 1516 | {set: true, raw: "sixty-four", want: defaultBackhaulMaxInflight}, |
| 1517 | {set: true, raw: "0", want: defaultBackhaulMaxInflight}, |
| 1518 | {set: true, raw: "-5", want: defaultBackhaulMaxInflight}, |
| 1519 | {set: true, raw: "128", want: 128}, |
| 1520 | {set: true, raw: " 8 ", want: 8}, |
| 1521 | } { |
| 1522 | if tc.set { |
| 1523 | t.Setenv("CARLOS_BACKHAUL_MAX_INFLIGHT", tc.raw) |
| 1524 | } else { |
| 1525 | os.Unsetenv("CARLOS_BACKHAUL_MAX_INFLIGHT") |
| 1526 | } |
| 1527 | if got := backhaulInflightFromEnv(); got != tc.want { |
| 1528 | t.Fatalf("CARLOS_BACKHAUL_MAX_INFLIGHT=%q (set=%v) → %d, want %d", tc.raw, tc.set, got, tc.want) |
| 1529 | } |
| 1530 | } |
| 1531 | } |
| 1532 | |
| 1533 | // oneKeyFailingStore fails Get for one named key — a single unreadable object |
| 1534 | // mid-poll, the case the carry-forward exists for. |
| 1535 | type oneKeyFailingStore struct { |
| 1536 | release.Store |
| 1537 | mu sync.Mutex |
| 1538 | bad string |
| 1539 | } |
| 1540 | |
| 1541 | func (o *oneKeyFailingStore) failKey(key string) { |
| 1542 | o.mu.Lock() |
| 1543 | o.bad = key |
| 1544 | o.mu.Unlock() |
| 1545 | } |
| 1546 | |
| 1547 | func (o *oneKeyFailingStore) Get(ctx context.Context, key string) ([]byte, string, error) { |
| 1548 | o.mu.Lock() |
| 1549 | bad := o.bad |
| 1550 | o.mu.Unlock() |
| 1551 | if key == bad { |
| 1552 | return nil, "", errors.New("bucket down for this object") |
| 1553 | } |
| 1554 | return o.Store.Get(ctx, key) |
| 1555 | } |
| 1556 | |
| 1557 | // ---- owners (#175 P7, the edge-local door) --------------------------------- |
| 1558 | |
| 1559 | // Owner answers through the same lookup as Home, exact and one-label |
| 1560 | // wildcard, and its dial target is Home's for the same host. |
| 1561 | func TestOwnerFoldsThroughExactAndWildcardLookup(t *testing.T) { |
| 1562 | s := meshTestStore(t) |
| 1563 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{`+ |
| 1564 | `"acme.go.tito.io":{"role":"home","account":"bae","app":"titogo"},`+ |
| 1565 | `"*.classic.tito.io":{"role":"home","account":"bae","app":"classic"}}}`) |
| 1566 | m := newHomeMap(s, "annie") |
| 1567 | refreshHomeMap(t, m) |
| 1568 | for _, c := range []struct{ host, app string }{ |
| 1569 | {"acme.go.tito.io", "titogo"}, |
| 1570 | {"ti.classic.tito.io", "classic"}, |
| 1571 | } { |
| 1572 | acct, app, eip, box, ok := m.Owner(c.host) |
| 1573 | if !ok || acct != "bae" || app != c.app { |
| 1574 | t.Fatalf("Owner(%s) = %s/%s %v, want bae/%s", c.host, acct, app, ok, c.app) |
| 1575 | } |
| 1576 | heip, hbox, hok := m.Home(c.host) |
| 1577 | if !hok || eip != heip || box != hbox { |
| 1578 | t.Fatalf("Owner(%s) dials %s/%s, Home dials %s/%s", c.host, eip, box, heip, hbox) |
| 1579 | } |
| 1580 | } |
| 1581 | if _, _, _, _, ok := m.Owner("a.b.classic.tito.io"); ok { |
| 1582 | t.Fatal("a two-label host must miss the one-label wildcard, as Home does") |
| 1583 | } |
| 1584 | } |
| 1585 | |
| 1586 | // No published owner (a mixed hostname, or a publisher that predates the |
| 1587 | // field), a cert-only entry, an own-box entry, or a nil map: ok is false and |
| 1588 | // nothing else is answered. |
| 1589 | func TestOwnerIsFalseWithoutAnOwnerOrADialTarget(t *testing.T) { |
| 1590 | s := meshTestStore(t) |
| 1591 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"own.test":{"role":"home","account":"bae","app":"titogo"}}}`) |
| 1592 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{`+ |
| 1593 | `"mixed.test":{"role":"home"},`+ |
| 1594 | `"halfowned.test":{"role":"home","account":"bae"},`+ |
| 1595 | `"*.certonly.test":{"role":"wildcard","wildcard":true,"account":"bae","app":"titogo"}}}`) |
| 1596 | m := newHomeMap(s, "annie") |
| 1597 | refreshHomeMap(t, m) |
| 1598 | for _, host := range []string{"mixed.test", "halfowned.test", "x.certonly.test", "own.test", "nowhere.test"} { |
| 1599 | if acct, app, eip, box, ok := m.Owner(host); ok || acct != "" || app != "" || eip != "" || box != "" { |
| 1600 | t.Errorf("Owner(%s) = %q/%q %q/%q %v, want zeros and false", host, acct, app, eip, box, ok) |
| 1601 | } |
| 1602 | } |
| 1603 | // An exact entry with no owner is the winner: the lookup does not fall |
| 1604 | // through to a wildcard that would have named one. |
| 1605 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{`+ |
| 1606 | `"exact.wild.test":{"role":"home"},"*.wild.test":{"role":"home","account":"bae","app":"titogo"}}}`) |
| 1607 | refreshHomeMap(t, m) |
| 1608 | if _, _, _, _, ok := m.Owner("exact.wild.test"); ok { |
| 1609 | t.Error("an ownerless exact entry borrowed the wildcard's owner") |
| 1610 | } |
| 1611 | var nilMap *homeMap |
| 1612 | if acct, app, eip, box, ok := nilMap.Owner("acme.go.tito.io"); ok || acct != "" || app != "" || eip != "" || box != "" { |
| 1613 | t.Error("a nil map answered an owner") |
| 1614 | } |
| 1615 | } |
| 1616 | |
| 1617 | // A route replacing an alias on a name collision carries the ROUTE's owner, |
| 1618 | // whatever box sorted first. The owner is never merged across entries. |
| 1619 | func TestOwnerOfARouteReplacingAnAliasIsTheRoutes(t *testing.T) { |
| 1620 | s := meshTestStore(t) |
| 1621 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"shared.test":{"role":"home","alias":true,"account":"bae","app":"squatter"},`+ |
| 1622 | `"ownerless.test":{"role":"home","alias":true,"account":"bae","app":"titogo"}}}`) |
| 1623 | putJSON(t, s, "control/edges/carol.json", `{"v":1,"box":"carol","eip":"3.3.3.3","hosts":{"shared.test":{"role":"home","account":"bae","app":"titogo"},`+ |
| 1624 | `"ownerless.test":{"role":"home"}}}`) |
| 1625 | m := newHomeMap(s, "annie") |
| 1626 | refreshHomeMap(t, m) |
| 1627 | acct, app, eip, box, ok := m.Owner("shared.test") |
| 1628 | if !ok || acct != "bae" || app != "titogo" || eip != "3.3.3.3" || box != "carol" { |
| 1629 | t.Fatalf("Owner(shared.test) = %s/%s %s/%s %v, want the route's bae/titogo on carol", acct, app, eip, box, ok) |
| 1630 | } |
| 1631 | // The ownerless route on carol replaces betty's owned alias: its lack of |
| 1632 | // an owner is not filled in from the alias it beat. |
| 1633 | if _, _, _, box, ok := m.Owner("ownerless.test"); ok { |
| 1634 | t.Fatalf("Owner(ownerless.test) = %s, true; the losing alias's owner leaked into the winner", box) |
| 1635 | } |
| 1636 | if _, box, _ := m.Home("ownerless.test"); box != "carol" { |
| 1637 | t.Fatalf("Home(ownerless.test) = %s, want carol", box) |
| 1638 | } |
| 1639 | } |
| 1640 | |
| 1641 | func TestHomeMapHigherStampWinsWhateverTheSortOrder(t *testing.T) { |
| 1642 | s := meshTestStore(t) |
| 1643 | // "annie" sorts first and is the dead box; "betty" holds the newer installation. |
| 1644 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home","stamp":41}}}`) |
| 1645 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"shop.test":{"role":"home","stamp":57}}}`) |
| 1646 | m := newHomeMap(s, "carol") |
| 1647 | refreshHomeMap(t, m) |
| 1648 | if eip, box, ok := m.Home("shop.test"); !ok || box != "betty" || eip != "2.2.2.2" { |
| 1649 | t.Fatalf("Home = %q %q %v, want betty", eip, box, ok) |
| 1650 | } |
| 1651 | // And the other way round, so it is the stamp and not the name. |
| 1652 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home","stamp":90}}}`) |
| 1653 | refreshHomeMap(t, m) |
| 1654 | if _, box, _ := m.Home("shop.test"); box != "annie" { |
| 1655 | t.Fatalf("Home box = %q, want annie", box) |
| 1656 | } |
| 1657 | } |
| 1658 | |
| 1659 | func TestHomeMapStampedBeatsUnstamped(t *testing.T) { |
| 1660 | s := meshTestStore(t) |
| 1661 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home"}}}`) |
| 1662 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"shop.test":{"role":"home","stamp":3}}}`) |
| 1663 | m := newHomeMap(s, "carol") |
| 1664 | refreshHomeMap(t, m) |
| 1665 | if _, box, _ := m.Home("shop.test"); box != "betty" { |
| 1666 | t.Fatalf("Home box = %q, want betty", box) |
| 1667 | } |
| 1668 | } |
| 1669 | |
| 1670 | func TestHomeMapUnstampedKeepsTodaysRule(t *testing.T) { |
| 1671 | s := meshTestStore(t) |
| 1672 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home","alias":true}}}`) |
| 1673 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"shop.test":{"role":"home"}}}`) |
| 1674 | m := newHomeMap(s, "carol") |
| 1675 | refreshHomeMap(t, m) |
| 1676 | if _, box, _ := m.Home("shop.test"); box != "betty" { |
| 1677 | t.Fatalf("a route still beats an alias: got %q", box) |
| 1678 | } |
| 1679 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home"}}}`) |
| 1680 | refreshHomeMap(t, m) |
| 1681 | if _, box, _ := m.Home("shop.test"); box != "annie" { |
| 1682 | t.Fatalf("two plain routes: first box in sorted order; got %q", box) |
| 1683 | } |
| 1684 | } |
| 1685 | |
| 1686 | func TestHomeMapAStampedAliasDoesNotBeatAnUnstampedRoute(t *testing.T) { |
| 1687 | s := meshTestStore(t) |
| 1688 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home"}}}`) |
| 1689 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"shop.test":{"role":"home","alias":true,"stamp":9}}}`) |
| 1690 | m := newHomeMap(s, "carol") |
| 1691 | refreshHomeMap(t, m) |
| 1692 | if _, box, _ := m.Home("shop.test"); box != "annie" { |
| 1693 | t.Fatalf("got %q: #287's rule (a route beats an alias) comes before stamps", box) |
| 1694 | } |
| 1695 | } |
| 1696 | |
| 1697 | func TestHomeMapAStampedEntryWithNoDialAddressDoesNotWin(t *testing.T) { |
| 1698 | s := meshTestStore(t) |
| 1699 | putJSON(t, s, "control/edges/annie.json", `{"v":1,"box":"annie","eip":"1.1.1.1","hosts":{"shop.test":{"role":"home","stamp":2}}}`) |
| 1700 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"","hosts":{"shop.test":{"role":"home","stamp":99}}}`) |
| 1701 | m := newHomeMap(s, "carol") |
| 1702 | refreshHomeMap(t, m) |
| 1703 | if eip, box, ok := m.Home("shop.test"); !ok || box != "annie" || eip != "1.1.1.1" { |
| 1704 | t.Fatalf("Home = %q %q %v, want annie: a box nothing can dial is skipped today and stays skipped", eip, box, ok) |
| 1705 | } |
| 1706 | } |
| 1707 | |
| 1708 | func file(box, eip string, hosts map[string]edgesHostView) edgesFileView { |
| 1709 | return edgesFileView{V: 1, Box: box, EIP: eip, Hosts: hosts} |
| 1710 | } |
| 1711 | |
| 1712 | func TestHomeMapReportsTwoBoxesAtTheSameStamp(t *testing.T) { |
| 1713 | files := map[string]edgesFileView{ |
| 1714 | "control/edges/annie.json": file("annie", "1.1.1.1", map[string]edgesHostView{ |
| 1715 | "tied.test": {Role: "home", Stamp: 7}, "fine.test": {Role: "home", Stamp: 7}, "plain.test": {Role: "home"}}), |
| 1716 | "control/edges/betty.json": file("betty", "2.2.2.2", map[string]edgesHostView{ |
| 1717 | "tied.test": {Role: "home", Stamp: 7}, "fine.test": {Role: "home", Stamp: 9}, "plain.test": {Role: "home"}}), |
| 1718 | } |
| 1719 | hosts, tied := foldHomes(files, "carol", false) |
| 1720 | if len(tied) != 1 || tied[0] != "tied.test" { |
| 1721 | t.Fatalf("tied = %v, want [tied.test]: unstamped duplicates are normal (static sites run on every box) and are not reported", tied) |
| 1722 | } |
| 1723 | if hosts["tied.test"].box != "annie" { |
| 1724 | t.Fatalf("a tie keeps the first box in sorted order; got %q", hosts["tied.test"].box) |
| 1725 | } |
| 1726 | if hosts["fine.test"].box != "betty" { |
| 1727 | t.Fatalf("fine.test box = %q", hosts["fine.test"].box) |
| 1728 | } |
| 1729 | } |
| 1730 | |
| 1731 | // A tie between two boxes is reported even when a third box holds a higher |
| 1732 | // stamp and wins the dial: the report does not depend on the winner. |
| 1733 | func TestHomeMapReportsATieBelowTheWinner(t *testing.T) { |
| 1734 | files := map[string]edgesFileView{ |
| 1735 | "control/edges/annie.json": file("annie", "1.1.1.1", map[string]edgesHostView{"h.test": {Role: "home", Stamp: 9}}), |
| 1736 | "control/edges/betty.json": file("betty", "2.2.2.2", map[string]edgesHostView{"h.test": {Role: "home", Stamp: 7}}), |
| 1737 | "control/edges/carol.json": file("carol", "3.3.3.3", map[string]edgesHostView{"h.test": {Role: "home", Stamp: 7}}), |
| 1738 | } |
| 1739 | hosts, tied := foldHomes(files, "dave", false) |
| 1740 | if len(tied) != 1 || tied[0] != "h.test" { |
| 1741 | t.Fatalf("tied = %v, want [h.test]", tied) |
| 1742 | } |
| 1743 | if hosts["h.test"].box != "annie" { |
| 1744 | t.Fatalf("the highest stamp still wins the dial; got %q", hosts["h.test"].box) |
| 1745 | } |
| 1746 | } |
| 1747 | |
| 1748 | // In a two-box region each box excludes itself from routing. It must |
| 1749 | // still see that it and its neighbour publish one host at one stamp. |
| 1750 | func TestHomeMapReportsATieWithThisBoxItself(t *testing.T) { |
| 1751 | files := map[string]edgesFileView{ |
| 1752 | "control/edges/annie.json": file("annie", "1.1.1.1", map[string]edgesHostView{"h.test": {Role: "home", Stamp: 7}}), |
| 1753 | "control/edges/betty.json": file("betty", "2.2.2.2", map[string]edgesHostView{"h.test": {Role: "home", Stamp: 7}}), |
| 1754 | } |
| 1755 | hosts, tied := foldHomes(files, "annie", false) |
| 1756 | if len(tied) != 1 || tied[0] != "h.test" { |
| 1757 | t.Fatalf("tied = %v, want [h.test]", tied) |
| 1758 | } |
| 1759 | if hosts["h.test"].box != "betty" { |
| 1760 | t.Fatalf("this box is still excluded from its own routing; got %q", hosts["h.test"].box) |
| 1761 | } |
| 1762 | } |
| 1763 | |
| 1764 | // A route catch-all here does not beat another box's EXACT host: the |
| 1765 | // precedence one box keeps (exact, then "*.<parent>") holds across boxes. |
| 1766 | // A label nobody serves exactly still falls to the local catch-all. |
| 1767 | func TestLocalWildcardYieldsToARemoteExactHost(t *testing.T) { |
| 1768 | e, m, _ := hopEdge(t, "annie", "*.example.com") |
| 1769 | rec, eip := fakeHomeEdge(t, nil) |
| 1770 | homedOn(t, m, "betty", eip, "acme.example.com") |
| 1771 | |
| 1772 | resp := hopRequest(t, e, "acme.example.com", nil) |
| 1773 | if body := readBody(t, resp); resp.StatusCode != http.StatusOK || body != homeEdgeBody { |
| 1774 | t.Fatalf("acme.example.com = %d %q, want the home edge's answer — another box serves it exactly", resp.StatusCode, body) |
| 1775 | } |
| 1776 | if n := rec.get().n; n != 1 { |
| 1777 | t.Fatalf("home edge saw %d request(s), want 1", n) |
| 1778 | } |
| 1779 | |
| 1780 | resp = hopRequest(t, e, "other.example.com", nil) |
| 1781 | if body := readBody(t, resp); resp.StatusCode != http.StatusOK || !strings.HasPrefix(body, "xff=") { |
| 1782 | t.Fatalf("other.example.com = %d %q, want the LOCAL catch-all — nobody serves it exactly", resp.StatusCode, body) |
| 1783 | } |
| 1784 | if n := rec.get().n; n != 1 { |
| 1785 | t.Fatalf("home edge saw %d request(s), want still 1", n) |
| 1786 | } |
| 1787 | } |
| 1788 | |
| 1789 | // ---- #336: only the hosts half is gated on the change index ------------------ |
| 1790 | |
| 1791 | // armedHomeMap is a home map over s with the change index armed (one watch over |
| 1792 | // the same store, the agent's in production) and a clock that starts now, so |
| 1793 | // the health state's real-time reads agree with it. |
| 1794 | func armedHomeMap(t *testing.T, s *changeIndexStore, self string) (*homeMap, *testClock) { |
| 1795 | t.Helper() |
| 1796 | w := release.NewChangeWatch(s) |
| 1797 | clk := &testClock{t: time.Now()} |
| 1798 | m := newHomeMap(s, self) |
| 1799 | m.watch = func() *release.ChangeWatch { return w } |
| 1800 | m.now = clk.now |
| 1801 | return m, clk |
| 1802 | } |
| 1803 | |
| 1804 | func homePoll(t *testing.T, m *homeMap) { |
| 1805 | t.Helper() |
| 1806 | if err := m.poll(context.Background()); err != nil { |
| 1807 | t.Fatalf("poll: %v", err) |
| 1808 | } |
| 1809 | } |
| 1810 | |
| 1811 | func edgesReads(s *changeIndexStore) int { |
| 1812 | return s.count("List", edgesPrefix) + s.count("ListETags", edgesPrefix) + s.count("Get", edgesPrefix) |
| 1813 | } |
| 1814 | |
| 1815 | // codex 8: a mesh-key rotation writes control/edge-mesh-key, which bumps no |
| 1816 | // scope, so the keys must not wait on control/edges. |
| 1817 | func TestHomeMapKeysPolledIndependently(t *testing.T) { |
| 1818 | t.Parallel() |
| 1819 | s := newChangeIndexStore(t) |
| 1820 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 1821 | putJSON(t, s, meshKeyKey, strings.Repeat("aa", 32)) |
| 1822 | m, clk := armedHomeMap(t, s, "annie") |
| 1823 | homePoll(t, m) |
| 1824 | if _, _, ok := m.Home("b.test"); !ok { |
| 1825 | t.Fatal("precondition: the first poll folds the edges files") |
| 1826 | } |
| 1827 | |
| 1828 | putJSON(t, s, meshKeyKey, strings.Repeat("bb", 32)) |
| 1829 | s.reset() |
| 1830 | clk.advance(homeMapInterval) |
| 1831 | homePoll(t, m) |
| 1832 | if cur, _ := m.meshKeys(); string(cur) != strings.Repeat("\xbb", 32) { |
| 1833 | t.Fatalf("the rotated mesh key must load within one poll, got %x", cur) |
| 1834 | } |
| 1835 | if n := edgesReads(s); n != 0 { |
| 1836 | t.Fatalf("control/edges did not move, yet the poll made %d read(s) under it", n) |
| 1837 | } |
| 1838 | if n := s.count("Get", meshKeyKey); n < 1 { |
| 1839 | t.Fatal("the mesh key must be read on every poll") |
| 1840 | } |
| 1841 | if _, _, ok := m.Home("b.test"); !ok { |
| 1842 | t.Fatal("a skipped hosts read must keep the fold") |
| 1843 | } |
| 1844 | } |
| 1845 | |
| 1846 | // codex 8: the five-minute refresh heartbeat holds across polls whose hosts |
| 1847 | // read is skipped, and the forwarding latch stays set. |
| 1848 | func TestHomeMapHeartbeatOnUnchangedPoll(t *testing.T) { |
| 1849 | t.Parallel() |
| 1850 | s := newChangeIndexStore(t) |
| 1851 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 1852 | putJSON(t, s, meshKeyKey, strings.Repeat("aa", 32)) |
| 1853 | m, clk := armedHomeMap(t, s, "annie") |
| 1854 | health := &lbhealth.State{} |
| 1855 | m.health = health |
| 1856 | homePoll(t, m) |
| 1857 | for i := 0; i < 4; i++ { |
| 1858 | s.reset() |
| 1859 | clk.advance(lbhealth.RefreshWindow - time.Minute) |
| 1860 | homePoll(t, m) |
| 1861 | if n := edgesReads(s); n != 0 { |
| 1862 | t.Fatalf("unchanged poll %d made %d read(s) under control/edges", i, n) |
| 1863 | } |
| 1864 | for _, c := range health.Failing(clk.now()) { |
| 1865 | if c == lbhealth.CondRefresh || c == lbhealth.CondForwarding { |
| 1866 | t.Fatalf("unchanged poll %d: health fails %q — a skipped hosts read must still beat", i, c) |
| 1867 | } |
| 1868 | } |
| 1869 | } |
| 1870 | } |
| 1871 | |
| 1872 | // The hosts half re-reads when control/edges moves, and at the backstop with |
| 1873 | // the index unmoved. |
| 1874 | func TestHomeMapHostsReadOnMoveAndAtBackstop(t *testing.T) { |
| 1875 | t.Parallel() |
| 1876 | s := newChangeIndexStore(t) |
| 1877 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 1878 | m, clk := armedHomeMap(t, s, "annie") |
| 1879 | homePoll(t, m) |
| 1880 | |
| 1881 | putJSON(t, s, "control/edges/carol.json", `{"v":1,"box":"carol","eip":"3.3.3.3","hosts":{"c.test":{"role":"home"}}}`) |
| 1882 | s.reset() |
| 1883 | clk.advance(homeMapInterval) |
| 1884 | homePoll(t, m) |
| 1885 | if _, _, ok := m.Home("c.test"); !ok { |
| 1886 | t.Fatal("a moved control/edges must be read on the next poll") |
| 1887 | } |
| 1888 | if n := s.count("List", edgesPrefix); n != 1 { |
| 1889 | t.Fatalf("moved: %d list(s) of control/edges, want 1", n) |
| 1890 | } |
| 1891 | |
| 1892 | s.reset() |
| 1893 | clk.advance(release.ChangeIndexBackstop*9/10 - time.Minute) |
| 1894 | homePoll(t, m) |
| 1895 | if n := edgesReads(s); n != 0 { |
| 1896 | t.Fatalf("young and unmoved: %d read(s), want none", n) |
| 1897 | } |
| 1898 | s.reset() |
| 1899 | clk.advance(release.ChangeIndexBackstop*2/10 + 2*time.Minute) |
| 1900 | homePoll(t, m) |
| 1901 | if n := s.count("List", edgesPrefix); n != 1 { |
| 1902 | t.Fatalf("at the backstop: %d list(s) of control/edges, want 1", n) |
| 1903 | } |
| 1904 | } |
| 1905 | |
| 1906 | // Disarmed, every poll reads the hosts as today. |
| 1907 | func TestHomeMapDisarmedPollReadsHosts(t *testing.T) { |
| 1908 | t.Parallel() |
| 1909 | s := newChangeIndexStore(t) |
| 1910 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 1911 | m := newHomeMap(s, "annie") |
| 1912 | m.watch = func() *release.ChangeWatch { return nil } |
| 1913 | for i := 0; i < 2; i++ { |
| 1914 | s.reset() |
| 1915 | homePoll(t, m) |
| 1916 | if n := s.count("List", edgesPrefix); n != 1 { |
| 1917 | t.Fatalf("disarmed poll %d: %d list(s), want 1", i, n) |
| 1918 | } |
| 1919 | if n := s.count("GetIfChanged", ""); n != 0 { |
| 1920 | t.Fatalf("disarmed poll %d read the index", i) |
| 1921 | } |
| 1922 | } |
| 1923 | } |
| 1924 | |
| 1925 | // codex 8 again: a hung index GET elsewhere in the process holds the shared |
| 1926 | // watch's poll lock, and the home map must still finish its poll and stamp |
| 1927 | // RefreshAttempted, or the load balancer pulls a working box after five |
| 1928 | // minutes. |
| 1929 | func TestHomeMapStampsRefreshWhileTheWatchIsStuck(t *testing.T) { |
| 1930 | t.Parallel() |
| 1931 | s := newChangeIndexStore(t) |
| 1932 | putJSON(t, s, "control/edges/betty.json", `{"v":1,"box":"betty","eip":"2.2.2.2","hosts":{"b.test":{"role":"home"}}}`) |
| 1933 | putJSON(t, s, meshKeyKey, strings.Repeat("aa", 32)) |
| 1934 | m, clk := armedHomeMap(t, s, "annie") |
| 1935 | m.watchTimeout = 50 * time.Millisecond |
| 1936 | health := &lbhealth.State{} |
| 1937 | m.health = health |
| 1938 | homePoll(t, m) |
| 1939 | |
| 1940 | entered, release := s.blockIndex(t) |
| 1941 | go func() { _, _ = m.watch().Poll(context.Background()) }() |
| 1942 | <-entered |
| 1943 | for i := 0; i < 2; i++ { |
| 1944 | clk.advance(lbhealth.RefreshWindow - time.Minute) |
| 1945 | done := make(chan error, 1) |
| 1946 | go func() { done <- m.poll(context.Background()) }() |
| 1947 | select { |
| 1948 | case err := <-done: |
| 1949 | if err != nil { |
| 1950 | t.Fatalf("poll %d: %v", i, err) |
| 1951 | } |
| 1952 | case <-time.After(5 * time.Second): |
| 1953 | release() |
| 1954 | t.Fatalf("poll %d waited on the stuck watch", i) |
| 1955 | } |
| 1956 | for _, c := range health.Failing(clk.now()) { |
| 1957 | if c == lbhealth.CondRefresh { |
| 1958 | t.Fatalf("poll %d: the refresh heartbeat lapsed behind a stuck watch", i) |
| 1959 | } |
| 1960 | } |
| 1961 | } |
| 1962 | } |
| 1963 | |
amadan