| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "crypto/ed25519" |
| 6 | "encoding/json" |
| 7 | "errors" |
| 8 | "flag" |
| 9 | "fmt" |
| 10 | "github.com/carlosframework/platform/internal/hostactors" |
| 11 | "github.com/carlosframework/platform/internal/hostfloor" |
| 12 | "github.com/carlosframework/platform/internal/release" |
| 13 | "github.com/carlosframework/platform/internal/version" |
| 14 | "log" |
| 15 | "os" |
| 16 | "os/user" |
| 17 | "path/filepath" |
| 18 | "runtime" |
| 19 | "strconv" |
| 20 | "strings" |
| 21 | ) |
| 22 | |
| 23 | func RunSystemGuardArm(args []string) { |
| 24 | fs := flag.NewFlagSet("system guard-arm", flag.ExitOnError) |
| 25 | consolePath := fs.String("console", "", "") |
| 26 | policyPath := fs.String("policy", "", "") |
| 27 | identityPath := fs.String("identity", "", "") |
| 28 | enrollmentPath := fs.String("enrollment", "", "") |
| 29 | _ = fs.Parse(args) |
| 30 | if os.Geteuid() != 0 || *policyPath == "" || *identityPath == "" || *enrollmentPath == "" { |
| 31 | log.Fatal(hostfloor.ErrFloor) |
| 32 | } |
| 33 | lockDir := systemRootDir(release.OptDir()) |
| 34 | if err := prepareGuardArmLockDir(release.OptDir(), lockDir); err != nil { |
| 35 | log.Fatal(err) |
| 36 | } |
| 37 | lock, err := os.OpenFile(filepath.Join(lockDir, systemLockFile), os.O_CREATE|os.O_RDWR, 0600) |
| 38 | if err != nil { |
| 39 | log.Fatal(err) |
| 40 | } |
| 41 | defer lock.Close() |
| 42 | if err = flockExclusiveNonblock(lock); err != nil { |
| 43 | log.Fatal(err) |
| 44 | } |
| 45 | d, err := floorDisk(release.OptDir()) |
| 46 | if err != nil { |
| 47 | log.Fatal(err) |
| 48 | } |
| 49 | raw, err := os.ReadFile(*policyPath) |
| 50 | if err != nil { |
| 51 | log.Fatal(err) |
| 52 | } |
| 53 | identity, err := os.ReadFile(*identityPath) |
| 54 | if err != nil { |
| 55 | log.Fatal(err) |
| 56 | } |
| 57 | hash, err := hostfloor.ExecutingHash() |
| 58 | if err != nil { |
| 59 | log.Fatal(err) |
| 60 | } |
| 61 | platform := runtime.GOOS + "-" + runtime.GOARCH |
| 62 | p, f, err := guardArmInputs(raw, identity, d.PublicKey, d.Bucket, version.Build, platform, hash) |
| 63 | if err != nil { |
| 64 | log.Fatal(err) |
| 65 | } |
| 66 | if err = guardArmIdle(d.Dir, systemRootDir(release.OptDir())); err != nil { |
| 67 | log.Fatal(err) |
| 68 | } |
| 69 | for _, path := range []string{systemBinaryPath(release.OptDir()), systemPrevPath(release.OptDir())} { |
| 70 | candidate, err := hostfloor.OpenProtected(path) |
| 71 | if err != nil { |
| 72 | log.Fatal(err) |
| 73 | } |
| 74 | b, err := os.ReadFile(candidate.Name()) |
| 75 | candidate.Close() |
| 76 | if err != nil || hostfloor.CheckImage(d.PublicKey, f, identity, version.Build, platform, hostfloor.Digest(b)) != nil { |
| 77 | log.Fatal(hostfloor.ErrFloor) |
| 78 | } |
| 79 | } |
| 80 | // Enrollment input is reporting configuration, verified against the signed |
| 81 | // directory before root persists it. It cannot grant roster membership. |
| 82 | enrollment, err := hostfloor.ReadEnrollment(*enrollmentPath) |
| 83 | if err != nil { |
| 84 | log.Fatal(err) |
| 85 | } |
| 86 | store, pub, err := openSystemStore(context.Background()) |
| 87 | if err != nil { |
| 88 | log.Fatal(err) |
| 89 | } |
| 90 | roster, err := readGuardRoster(context.Background(), store, pub, d.Bucket) |
| 91 | if err != nil { |
| 92 | log.Fatal(err) |
| 93 | } |
| 94 | b, ok := roster.Boxes[enrollment.Box] |
| 95 | if ok { |
| 96 | if b.Enrollment != enrollment.Enrollment || b.Cloud.InstanceID != enrollment.InstanceID { |
| 97 | log.Fatal(hostfloor.ErrFloor) |
| 98 | } |
| 99 | } else { |
| 100 | // A post-floor join must arm and report before it becomes an actor. |
| 101 | // The invitation grants a reporting nonce, never execution authority. |
| 102 | wire, _, e := store.Get(context.Background(), "control/host-actor-invitations/"+enrollment.Box+".json") |
| 103 | if e != nil { |
| 104 | log.Fatal(e) |
| 105 | } |
| 106 | var invitation hostactors.Invitation |
| 107 | if release.VerifyRootEnvelope(pub, wire, &invitation) != nil || invitation.Bucket != d.Bucket || invitation.Reference != enrollment.Box || invitation.Enrollment != enrollment.Enrollment { |
| 108 | log.Fatal(hostfloor.ErrFloor) |
| 109 | } |
| 110 | } |
| 111 | if err = guardArmControllers(dataDir()); err != nil { |
| 112 | log.Fatal(err) |
| 113 | } |
| 114 | // The guard directories are traversed by non-root units (the console |
| 115 | // and the edge exec the guard binary through them), so their mode |
| 116 | // cannot depend on the caller's umask: a umask 077 shell once left |
| 117 | // them 0700 and the guarded console failed with exit 203. |
| 118 | for _, dir := range []string{d.Dir, filepath.Join(d.Dir, "identities")} { |
| 119 | if err = mkdirAllExact(dir, 0755); err != nil { |
| 120 | log.Fatal(err) |
| 121 | } |
| 122 | } |
| 123 | path := filepath.Join(d.Dir, "policy.json") |
| 124 | if _, err = os.Stat(path); os.IsNotExist(err) { |
| 125 | initial, _ := json.Marshal(hostfloor.Policy{Schema: 1, DeploymentBucket: d.Bucket, Mode: "unpublished"}) |
| 126 | if err = hostfloor.Durable(path, initial, 0644); err != nil { |
| 127 | log.Fatal(err) |
| 128 | } |
| 129 | } |
| 130 | // A migrated box retains its recovery checker until the root floor and |
| 131 | // selector are durable. No-census arming keeps its original ordering. |
| 132 | if id, required, e := rootMovesFloorIdentity(); e != nil { |
| 133 | log.Fatal(e) |
| 134 | } else if required { |
| 135 | projection, e := readRootFloorProjection(release.OptDir()) |
| 136 | if e != nil || projection.Identity != id { |
| 137 | log.Fatal(errRootFloorHistory) |
| 138 | } |
| 139 | signed, e := release.VerifySystemIdentity(d.PublicKey, identity) |
| 140 | if e != nil { |
| 141 | log.Fatal(e) |
| 142 | } |
| 143 | image := rootFloorBinary{Version: signed.Current.ReleaseTag, Protocol: signed.Current.MovesProtocol, Digest: hash} |
| 144 | if !image.supports(projection.Floor) { |
| 145 | log.Fatal(errRootFloorHistory) |
| 146 | } |
| 147 | } |
| 148 | // The checker is outside system/router and outside every restore set. |
| 149 | if err = copyFileDurable("/proc/self/exe", filepath.Join(d.Dir, "carlos"), 0755); err != nil { |
| 150 | log.Fatal(err) |
| 151 | } |
| 152 | snapshot, err := release.VerifySystemIdentity(d.PublicKey, identity) |
| 153 | if err != nil { |
| 154 | log.Fatal(err) |
| 155 | } |
| 156 | if err = hostfloor.Durable(filepath.Join(d.Dir, "identities", snapshot.Current.ReleaseTag+".json"), identity, 0644); err != nil { |
| 157 | log.Fatal(err) |
| 158 | } |
| 159 | enrollmentRaw, _ := json.Marshal(enrollment) |
| 160 | if err = hostfloor.Durable("/etc/carlos/host-actor.json", enrollmentRaw, 0644); err != nil { |
| 161 | log.Fatal(err) |
| 162 | } |
| 163 | dropins := map[string]string{ |
| 164 | "carlos-system-update.service": "[Service]\nExecStart=\nExecStart=/opt/carlos/system/guard/carlos system guard-exec apply\n", |
| 165 | "carlos-system-recover.service": "[Service]\nExecStart=\nExecStart=/opt/carlos/system/guard/carlos system guard-exec recover\n", |
| 166 | "carlos-edge.service": "[Unit]\nRequires=carlos-system-recover.service\nAfter=carlos-system-recover.service\n[Service]\nExecStart=\nExecStart=/opt/carlos/system/guard/carlos system guard-exec edge\n", |
| 167 | } |
| 168 | if *consolePath != "" { |
| 169 | if !filepath.IsAbs(*consolePath) || filepath.Clean(*consolePath) != *consolePath { |
| 170 | log.Fatal(hostfloor.ErrFloor) |
| 171 | } |
| 172 | for _, path := range []string{*consolePath, *consolePath + ".prev"} { |
| 173 | file, e := hostfloor.OpenProtected(path) |
| 174 | if e != nil { |
| 175 | log.Fatal(e) |
| 176 | } |
| 177 | body, e := os.ReadFile(file.Name()) |
| 178 | file.Close() |
| 179 | if e != nil || hostfloor.CheckComponent(d.PublicKey, f, identity, version.Build, platform, hostfloor.Digest(body), "console") != nil { |
| 180 | log.Fatal(hostfloor.ErrFloor) |
| 181 | } |
| 182 | } |
| 183 | dropins["console.service"] = hostfloor.ConsoleGuardDropIn(*consolePath) |
| 184 | } |
| 185 | for unit, body := range dropins { |
| 186 | dir := filepath.Join("/etc/systemd/system", unit+".d") |
| 187 | if err = os.MkdirAll(dir, 0755); err != nil { |
| 188 | log.Fatal(err) |
| 189 | } |
| 190 | if err = hostfloor.Durable(filepath.Join(dir, "90-host-floor.conf"), []byte(body), 0644); err != nil { |
| 191 | log.Fatal(err) |
| 192 | } |
| 193 | } |
| 194 | if _, err = productionSysOps().systemctl("daemon-reload"); err != nil { |
| 195 | log.Fatal(err) |
| 196 | } |
| 197 | if *consolePath != "" { |
| 198 | if err = hostfloor.Durable(filepath.Join(d.Dir, "console-ready.json"), []byte(*consolePath), 0644); err != nil { |
| 199 | log.Fatal(err) |
| 200 | } |
| 201 | } |
| 202 | if err = d.Persist(p, map[string][]byte{snapshot.Current.ReleaseTag: identity}); err != nil { |
| 203 | log.Fatal(err) |
| 204 | } |
| 205 | } |
| 206 | |
| 207 | // guardArmInputs is guard-arm's parse of its --policy and --identity files, |
| 208 | // byte for byte as read: an armed policy for this deployment, and an identity |
| 209 | // that names the running build (build, platform and the executing image's |
| 210 | // hash) at or above the policy's floor. The claims phases sign-policy and |
| 211 | // publish-identity write exactly these files (--out). |
| 212 | func guardArmInputs(policy, identity []byte, pub ed25519.PublicKey, bucket, build, platform, hash string) (hostfloor.Policy, hostfloor.Verified, error) { |
| 213 | var p hostfloor.Policy |
| 214 | if json.Unmarshal(policy, &p) != nil || p.Mode != "armed" { |
| 215 | return hostfloor.Policy{}, hostfloor.Verified{}, hostfloor.ErrFloor |
| 216 | } |
| 217 | f, err := p.Verify(pub, bucket) |
| 218 | if err != nil { |
| 219 | return hostfloor.Policy{}, hostfloor.Verified{}, err |
| 220 | } |
| 221 | if err = hostfloor.CheckImage(pub, f, identity, build, platform, hash); err != nil { |
| 222 | return hostfloor.Policy{}, hostfloor.Verified{}, err |
| 223 | } |
| 224 | return p, f, nil |
| 225 | } |
| 226 | |
| 227 | // guardArmIdle refuses to arm while root's update state is on the box: an |
| 228 | // open journal, or any update-begun.json* in the guard directory. Main's |
| 229 | // region marker, its temporary file or a stale copy beside an armed policy |
| 230 | // would wedge every guarded update and recovery (Opus re-review M1-3). |
| 231 | // guard-arm calls it holding the apply lock, so no update writes either |
| 232 | // while it looks. |
| 233 | func guardArmIdle(guardDir, rootDir string) error { |
| 234 | journal := filepath.Join(rootDir, "journal.json") |
| 235 | if _, err := os.Lstat(journal); err == nil { |
| 236 | return fmt.Errorf("%w: %s is open; let the update finish or recover first", hostfloor.ErrFloor, journal) |
| 237 | } else if !errors.Is(err, os.ErrNotExist) { |
| 238 | return err |
| 239 | } |
| 240 | entries, err := os.ReadDir(guardDir) |
| 241 | if errors.Is(err, os.ErrNotExist) { |
| 242 | return nil |
| 243 | } |
| 244 | if err != nil { |
| 245 | return err |
| 246 | } |
| 247 | for _, e := range entries { |
| 248 | if strings.HasPrefix(e.Name(), updateBegunFile) { |
| 249 | return fmt.Errorf("%w: %s is present; let the update finish or recover first", hostfloor.ErrFloor, filepath.Join(guardDir, e.Name())) |
| 250 | } |
| 251 | } |
| 252 | return nil |
| 253 | } |
| 254 | |
| 255 | // lookupServiceUser resolves the controllers' account; a test seam. |
| 256 | var lookupServiceUser = user.Lookup |
| 257 | |
| 258 | // guardArmControllers creates <DataDir>/.hostexec for the service account |
| 259 | // before any armed state is written. The armed update path refuses while it |
| 260 | // is missing (Opus re-review M1-2). |
| 261 | func guardArmControllers(dataDir string) error { |
| 262 | u, err := lookupServiceUser(serviceUser) |
| 263 | if err != nil { |
| 264 | return err |
| 265 | } |
| 266 | uid, err := strconv.Atoi(u.Uid) |
| 267 | if err != nil { |
| 268 | return fmt.Errorf("%w: %s has uid %q", hostfloor.ErrFloor, serviceUser, u.Uid) |
| 269 | } |
| 270 | gid, err := strconv.Atoi(u.Gid) |
| 271 | if err != nil { |
| 272 | return fmt.Errorf("%w: %s has gid %q", hostfloor.ErrFloor, serviceUser, u.Gid) |
| 273 | } |
| 274 | return hostfloor.EnsureControllerDir(dataDir, uid, gid) |
| 275 | } |
| 276 | |
| 277 | // mkdirAllExact is os.MkdirAll with the caller's umask taken out: every |
| 278 | // directory it creates, and dir itself whether or not it was created, ends |
| 279 | // with exactly mode. Existing parents keep their modes. |
| 280 | func mkdirAllExact(dir string, mode os.FileMode) error { |
| 281 | var created []string |
| 282 | for p := filepath.Clean(dir); ; p = filepath.Dir(p) { |
| 283 | if _, err := os.Lstat(p); err == nil { |
| 284 | break |
| 285 | } else if !os.IsNotExist(err) { |
| 286 | return err |
| 287 | } |
| 288 | created = append(created, p) |
| 289 | if parent := filepath.Dir(p); parent == p { |
| 290 | break |
| 291 | } |
| 292 | } |
| 293 | if err := os.MkdirAll(dir, mode); err != nil { |
| 294 | return err |
| 295 | } |
| 296 | for _, p := range append(created, filepath.Clean(dir)) { |
| 297 | if err := os.Chmod(p, mode); err != nil { |
| 298 | return err |
| 299 | } |
| 300 | } |
| 301 | return nil |
| 302 | } |
| 303 | |
| 304 | // prepareGuardArmLockDir makes guard-arm's lock directory under |
| 305 | // <opt>/system without the caller's umask, and repairs <opt>/system itself |
| 306 | // to 0755: the guard directories sit under the same parent and non-root |
| 307 | // units exec the guard binary through it, so a umask 077 run must not |
| 308 | // leave it closed. |
| 309 | func prepareGuardArmLockDir(optDir, lockDir string) error { |
| 310 | if err := mkdirAllExact(filepath.Join(optDir, "system"), 0755); err != nil { |
| 311 | return err |
| 312 | } |
| 313 | return mkdirAllExact(lockDir, 0755) |
| 314 | } |
| 315 | |
amadan