| 1 | //go:build linux |
| 2 | |
| 3 | package host |
| 4 | |
| 5 | import ( |
| 6 | "bytes" |
| 7 | "context" |
| 8 | "crypto/ed25519" |
| 9 | "log" |
| 10 | "os" |
| 11 | "path/filepath" |
| 12 | "slices" |
| 13 | "strings" |
| 14 | "testing" |
| 15 | "time" |
| 16 | |
| 17 | "github.com/carlosframework/platform/internal/genselect" |
| 18 | "github.com/carlosframework/platform/internal/hostactors" |
| 19 | "github.com/carlosframework/platform/internal/hostbox" |
| 20 | "github.com/carlosframework/platform/internal/hostclaim" |
| 21 | "github.com/carlosframework/platform/internal/hostops" |
| 22 | "github.com/carlosframework/platform/internal/hostwrite" |
| 23 | "github.com/carlosframework/platform/internal/lease" |
| 24 | "github.com/carlosframework/platform/internal/park" |
| 25 | "github.com/carlosframework/platform/internal/registry" |
| 26 | "github.com/carlosframework/platform/internal/release" |
| 27 | "github.com/carlosframework/platform/internal/rootstate" |
| 28 | ) |
| 29 | |
| 30 | type fix4Owners struct{ ClaimOwnerBindings } |
| 31 | |
| 32 | func (o fix4Owners) OpenClaimOwners(context.Context, release.Store, ClaimRuntimeResources) (ClaimOwnerBindings, error) { |
| 33 | return o.ClaimOwnerBindings, nil |
| 34 | } |
| 35 | |
| 36 | type fix4UnusedDependencies struct { |
| 37 | hostbox.LocalAuthority |
| 38 | StopControllerReader |
| 39 | lease.VersionedStore |
| 40 | ClaimWakeInputs |
| 41 | hostbox.RegionalEvidence |
| 42 | } |
| 43 | |
| 44 | func fix4JournalFixture(t *testing.T) (release.Store, ClaimRuntimeResources, fix4Owners, string) { |
| 45 | t.Helper() |
| 46 | root := ed25519.NewKeyFromSeed(bytes.Repeat([]byte{1}, ed25519.SeedSize)) |
| 47 | evidence := ed25519.NewKeyFromSeed(bytes.Repeat([]byte{2}, ed25519.SeedSize)) |
| 48 | store := release.NewDirStore(t.TempDir()) |
| 49 | identity := handoverRecoveryIdentity(t, hostbox.Actor{Box: "box", Enrollment: strings.Repeat("a", 32), Revision: 1, Binding: "1", Region: "ie", ControlBucket: "deployment"}) |
| 50 | actor, err := identity.Verify(t.Context()) |
| 51 | if err != nil { |
| 52 | t.Fatal(err) |
| 53 | } |
| 54 | row := hostactors.Box{BoxID: actor.Box, Enrollment: actor.Enrollment, Namespace: actor.Namespace, Label: actor.Box, Region: actor.Region, Pool: "default", Cloud: hostactors.Cloud{Provider: actor.Cloud.Provider, Account: actor.Cloud.Account, Region: actor.Cloud.Region, InstanceID: actor.Cloud.InstanceID}, EvidenceSHA256: strings.Repeat("c", 64), CensusPublicKey: evidence.Public().(ed25519.PublicKey), State: "active"} |
| 55 | roster := hostactors.Roster{V: 1, DeploymentBucket: actor.ControlBucket, Revision: 1, RequestLimitNS: int64(hostbox.RequestLimit), Boxes: map[string]hostactors.Box{actor.Box: row}, Consoles: map[string]hostactors.Console{}, Holds: map[string]hostactors.Reservation{}, Activations: map[string]hostactors.Reservation{}} |
| 56 | body, err := release.SignRootEnvelope(release.RootSigner(root), roster) |
| 57 | if err != nil { |
| 58 | t.Fatal(err) |
| 59 | } |
| 60 | if _, err = store.Put(t.Context(), hostactors.Key, body, ""); err != nil { |
| 61 | t.Fatal(err) |
| 62 | } |
| 63 | directory := hostactors.Directory{Store: store, PublicKey: root.Public().(ed25519.PublicKey), Bucket: actor.ControlBucket} |
| 64 | unused := fix4UnusedDependencies{} |
| 65 | owners := fix4Owners{ClaimOwnerBindings{ |
| 66 | Identity: identity, RetirementAuthority: &hostops.ProductionAuthority{Store: store, Directory: directory, Now: time.Now}, EvidenceKey: evidence, |
| 67 | Local: unused, Controllers: unused, Data: unused, Inputs: unused, Evidence: unused, |
| 68 | InputPolicy: func(context.Context, *hostclaim.Claim, hostbox.Selection, hostwrite.Identity) error { return nil }, |
| 69 | MovesPolicy: func(context.Context, *hostclaim.Claim, hostbox.Selection, hostwrite.Identity) error { return nil }, |
| 70 | LeaseSet: func(context.Context, string, string) ([]string, error) { return nil, nil }, |
| 71 | View: func(context.Context, hostclaim.Observed) (park.View, error) { return park.View{}, nil }, |
| 72 | Authorize: func(context.Context, hostclaim.Observed, hostclaim.Obligation) error { return nil }, |
| 73 | AdmissionEvidence: func(context.Context, hostclaim.Observed, hostclaim.Admission, *genselect.Plan) error { return nil }, |
| 74 | PublicationEvidence: func(context.Context, hostclaim.Observed) error { return nil }, |
| 75 | Close: func(context.Context) error { return nil }, |
| 76 | }} |
| 77 | reg, err := registry.Open(filepath.Join(t.TempDir(), "registry.db")) |
| 78 | if err != nil { |
| 79 | t.Fatal(err) |
| 80 | } |
| 81 | t.Cleanup(func() { reg.Close() }) |
| 82 | resources := ClaimRuntimeResources{Mode: "instance-control", Registry: reg, DataDir: t.TempDir(), OptDir: t.TempDir(), UnitsDir: t.TempDir()} |
| 83 | legacy := filepath.Join(resources.DataDir, ".hostclaims") |
| 84 | if err = os.Mkdir(legacy, 0700); err != nil { |
| 85 | t.Fatal(err) |
| 86 | } |
| 87 | journalPath := filepath.Join(t.TempDir(), "journals.db") |
| 88 | authorityDir := t.TempDir() |
| 89 | originalAuthority := openClaimRootAuthority |
| 90 | t.Cleanup(func() { openClaimRootAuthority = originalAuthority }) |
| 91 | openClaimRootAuthority = func() (rootstate.Store, error) { return &rollbackDisk{dir: authorityDir}, nil } |
| 92 | original := openClaimRootJournal |
| 93 | t.Cleanup(func() { openClaimRootJournal = original }) |
| 94 | openClaimRootJournal = func() (*hostbox.LaunchJournal, error) { return hostbox.OpenLaunchJournal(journalPath) } |
| 95 | return store, resources, owners, legacy |
| 96 | } |
| 97 | |
| 98 | func TestPlan9Fix4RootAssemblyIgnoresReplacedLegacyDirectory(t *testing.T) { |
| 99 | fixtureWithoutRootMigrationCensus(t) |
| 100 | for _, mode := range []string{"instance-control", "adopt"} { |
| 101 | for _, replacement := range []string{"missing", "renamed", "symlink", "file"} { |
| 102 | t.Run(mode+"/"+replacement, func(t *testing.T) { |
| 103 | store, resources, owners, legacy := fix4JournalFixture(t) |
| 104 | resources.Mode = mode |
| 105 | if mode == "adopt" { |
| 106 | owners.EvidenceKey = nil |
| 107 | } |
| 108 | journal, err := openClaimRootJournal() |
| 109 | if err != nil { |
| 110 | t.Fatal(err) |
| 111 | } |
| 112 | service, err := hostbox.NewJournalService(journal) |
| 113 | if err != nil { |
| 114 | t.Fatal(err) |
| 115 | } |
| 116 | if alert, err := service.CheckLegacyMigration(t.Context(), legacy); alert != nil || err != nil { |
| 117 | t.Fatalf("marker: %v %v", alert, err) |
| 118 | } |
| 119 | if err = journal.Close(); err != nil { |
| 120 | t.Fatal(err) |
| 121 | } |
| 122 | target := t.TempDir() |
| 123 | if replacement == "renamed" { |
| 124 | err = os.Rename(legacy, legacy+"-moved") |
| 125 | } else { |
| 126 | err = os.Remove(legacy) |
| 127 | } |
| 128 | if err != nil { |
| 129 | t.Fatal(err) |
| 130 | } |
| 131 | switch replacement { |
| 132 | case "symlink": |
| 133 | err = os.Symlink(target, legacy) |
| 134 | case "file": |
| 135 | err = os.WriteFile(legacy, []byte("agent replacement"), 0640) |
| 136 | } |
| 137 | if err != nil { |
| 138 | t.Fatal(err) |
| 139 | } |
| 140 | var logs bytes.Buffer |
| 141 | previous := log.Writer() |
| 142 | log.SetOutput(&logs) |
| 143 | defer log.SetOutput(previous) |
| 144 | assembly, err := OpenProductionClaimAssembly(t.Context(), store, resources, owners) |
| 145 | if err != nil { |
| 146 | t.Fatalf("agent-controlled legacy directory prevented root startup: %v", err) |
| 147 | } |
| 148 | if err = assembly.Close(t.Context()); err != nil { |
| 149 | t.Fatal(err) |
| 150 | } |
| 151 | if !strings.Contains(logs.String(), "ignored after migration") { |
| 152 | t.Fatalf("missing legacy directory alert: %q", logs.String()) |
| 153 | } |
| 154 | if replacement == "missing" || replacement == "renamed" { |
| 155 | if _, err = os.Lstat(legacy); !os.IsNotExist(err) { |
| 156 | t.Fatalf("root recreated agent directory: %v", err) |
| 157 | } |
| 158 | } |
| 159 | if replacement == "symlink" { |
| 160 | if got, err := os.Readlink(legacy); err != nil || got != target { |
| 161 | t.Fatalf("root changed symlink: %q %v", got, err) |
| 162 | } |
| 163 | } |
| 164 | if replacement == "file" { |
| 165 | if got, err := os.ReadFile(legacy); err != nil || string(got) != "agent replacement" { |
| 166 | t.Fatalf("root changed file: %q %v", got, err) |
| 167 | } |
| 168 | } |
| 169 | }) |
| 170 | } |
| 171 | } |
| 172 | } |
| 173 | |
| 174 | func TestPlan9Fix4MigrationRetryAfterCommitRecoversOwnedMask(t *testing.T) { |
| 175 | for _, removeLegacy := range []bool{false, true} { |
| 176 | t.Run(map[bool]string{false: "unchanged", true: "removed"}[removeLegacy], func(t *testing.T) { |
| 177 | store, resources, owners, legacy := fix4JournalFixture(t) |
| 178 | var calls []string |
| 179 | plan9Fix2Systemctl(t, "enabled", &calls) |
| 180 | owner := plan9Fix2MaskOwner(t) |
| 181 | launch, err := hostbox.OpenLaunchJournal(filepath.Join(legacy, "launches.db")) |
| 182 | if err != nil { |
| 183 | t.Fatal(err) |
| 184 | } |
| 185 | if err = launch.Close(); err != nil { |
| 186 | t.Fatal(err) |
| 187 | } |
| 188 | inventory, err := openLegacyInventory(t, filepath.Join(legacy, "inventory.db")) |
| 189 | if err != nil { |
| 190 | t.Fatal(err) |
| 191 | } |
| 192 | if err = inventory.Close(); err != nil { |
| 193 | t.Fatal(err) |
| 194 | } |
| 195 | journal, err := openClaimRootJournal() |
| 196 | if err != nil { |
| 197 | t.Fatal(err) |
| 198 | } |
| 199 | service, err := hostbox.NewJournalService(journal) |
| 200 | if err != nil { |
| 201 | t.Fatal(err) |
| 202 | } |
| 203 | err = service.MigrateLegacy(t.Context(), legacy, "box", strings.Repeat("a", 32), func(ctx context.Context) (func(), error) { |
| 204 | _, err := quiesceLegacyClaimJournals(ctx) |
| 205 | return func() {}, err |
| 206 | }, nil) |
| 207 | if err != nil { |
| 208 | t.Fatal(err) |
| 209 | } |
| 210 | if err = journal.Close(); err != nil { |
| 211 | t.Fatal(err) |
| 212 | } |
| 213 | raw, err := os.ReadFile(owner) |
| 214 | record, recorded := decodeClaimEdgeMaskRecord(raw, plan9Fix2Boot(t)) |
| 215 | if err != nil || !recorded { |
| 216 | t.Fatalf("commit lost crash mask identity: %q %v", raw, err) |
| 217 | } |
| 218 | if ours, identityErr := claimEdgeMaskStillOwned(record); identityErr != nil || !ours { |
| 219 | t.Fatalf("commit changed crash mask identity: %v %v", ours, identityErr) |
| 220 | } |
| 221 | if removeLegacy { |
| 222 | if err = os.RemoveAll(legacy); err != nil { |
| 223 | t.Fatal(err) |
| 224 | } |
| 225 | } |
| 226 | calls = nil |
| 227 | plan9Fix2Systemctl(t, "masked-runtime", &calls) |
| 228 | if err = migrateClaimJournals(t.Context(), store, resources, owners); err != nil { |
| 229 | t.Fatal(err) |
| 230 | } |
| 231 | if !slices.Contains(calls, "unmask --runtime carlos-edge.service") { |
| 232 | t.Fatalf("committed retry left edge masked: %v", calls) |
| 233 | } |
| 234 | if _, err = os.Stat(owner); !os.IsNotExist(err) { |
| 235 | t.Fatalf("migration mask record remains: %v", err) |
| 236 | } |
| 237 | for _, call := range calls { |
| 238 | if strings.HasPrefix(call, "mask ") { |
| 239 | t.Fatalf("completed retry masked edge again: %v", calls) |
| 240 | } |
| 241 | } |
| 242 | }) |
| 243 | } |
| 244 | } |
| 245 | |
| 246 | func TestPlan9Fix4CompletedMigrationKeepsOperatorMask(t *testing.T) { |
| 247 | for _, test := range []struct{ name, state, record string }{ |
| 248 | {"unrecorded-runtime", "masked-runtime", ""}, |
| 249 | {"different-boot", "masked-runtime", "another-boot"}, |
| 250 | {"persistent", "masked", "same-boot"}, |
| 251 | {"unmasked", "enabled", "same-boot"}, |
| 252 | } { |
| 253 | t.Run(test.name, func(t *testing.T) { |
| 254 | store, resources, owners, _ := fix4JournalFixture(t) |
| 255 | owner := plan9Fix2MaskOwner(t) |
| 256 | var calls []string |
| 257 | plan9Fix2Systemctl(t, test.state, &calls) |
| 258 | if err := migrateClaimJournals(t.Context(), store, resources, owners); err != nil { |
| 259 | t.Fatal(err) |
| 260 | } |
| 261 | record := test.record |
| 262 | if record == "same-boot" { |
| 263 | record = plan9Fix2Boot(t) |
| 264 | } |
| 265 | if record != "" { |
| 266 | if err := os.WriteFile(owner, []byte(record), 0600); err != nil { |
| 267 | t.Fatal(err) |
| 268 | } |
| 269 | } |
| 270 | calls = nil |
| 271 | if err := migrateClaimJournals(t.Context(), store, resources, owners); err != nil { |
| 272 | t.Fatal(err) |
| 273 | } |
| 274 | for _, call := range calls { |
| 275 | if strings.HasPrefix(call, "unmask") || strings.HasPrefix(call, "mask") { |
| 276 | t.Fatalf("operator state changed: %v", calls) |
| 277 | } |
| 278 | } |
| 279 | }) |
| 280 | } |
| 281 | } |
| 282 | |
| 283 | func TestPlan9Fix4CompletedMigrationRetainsRecordOnUnmaskFailure(t *testing.T) { |
| 284 | store, resources, owners, _ := fix4JournalFixture(t) |
| 285 | owner := plan9Fix2MaskOwner(t) |
| 286 | var calls []string |
| 287 | plan9Fix2Systemctl(t, "masked-runtime", &calls) |
| 288 | if err := migrateClaimJournals(t.Context(), store, resources, owners); err != nil { |
| 289 | t.Fatal(err) |
| 290 | } |
| 291 | record := plan11Rev4WriteOwnedMask(t, owner) |
| 292 | systemctl := claimSystemctl |
| 293 | claimSystemctl = func(ctx context.Context, args ...string) ([]byte, error) { |
| 294 | if args[0] == "unmask" { |
| 295 | return nil, os.ErrPermission |
| 296 | } |
| 297 | return systemctl(ctx, args...) |
| 298 | } |
| 299 | if err := migrateClaimJournals(t.Context(), store, resources, owners); err == nil { |
| 300 | t.Fatal("unmask failure reported success") |
| 301 | } |
| 302 | if raw, err := os.ReadFile(owner); err != nil { |
| 303 | t.Fatalf("failed recovery discarded mask ownership: %q %v", raw, err) |
| 304 | } else if got, ok := decodeClaimEdgeMaskRecord(raw, plan9Fix2Boot(t)); !ok || got != record { |
| 305 | t.Fatalf("failed recovery changed mask ownership: %q", raw) |
| 306 | } |
| 307 | claimSystemctl = systemctl |
| 308 | if err := migrateClaimJournals(t.Context(), store, resources, owners); err != nil { |
| 309 | t.Fatal(err) |
| 310 | } |
| 311 | if _, err := os.Stat(owner); !os.IsNotExist(err) { |
| 312 | t.Fatalf("successful recovery kept ownership record: %v", err) |
| 313 | } |
| 314 | } |
| 315 | |