| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "github.com/carlosframework/platform/internal/hostfloor" |
| 7 | "github.com/carlosframework/platform/internal/release" |
| 8 | "log" |
| 9 | "os" |
| 10 | "path/filepath" |
| 11 | "runtime" |
| 12 | "syscall" |
| 13 | "time" |
| 14 | ) |
| 15 | |
| 16 | func floorDisk(optDir string) (hostfloor.Disk, error) { |
| 17 | pub, err := release.ParsePublicKey(os.Getenv("CARLOS_RELEASE_PUBKEY"), "CARLOS_RELEASE_PUBKEY") |
| 18 | if err != nil { |
| 19 | return hostfloor.Disk{}, err |
| 20 | } |
| 21 | bucket := os.Getenv("CARLOS_DEPLOYMENT_BUCKET") |
| 22 | if bucket == "" { |
| 23 | return hostfloor.Disk{}, hostfloor.ErrFloor |
| 24 | } |
| 25 | return hostfloor.Disk{Dir: filepath.Join(optDir, "system", "guard"), Bucket: bucket, PublicKey: pub}, nil |
| 26 | } |
| 27 | func RunSystemGuardExec(args []string) { |
| 28 | if len(args) < 1 { |
| 29 | log.Fatal(hostfloor.ErrFloor) |
| 30 | } |
| 31 | d, err := floorDisk(release.OptDir()) |
| 32 | if err != nil { |
| 33 | log.Fatal(err) |
| 34 | } |
| 35 | policy, local, err := d.Load() |
| 36 | if err != nil { |
| 37 | log.Fatal(err) |
| 38 | } |
| 39 | if args[0] == "recover" { |
| 40 | held, err := recoverLockHeld(release.OptDir()) |
| 41 | if err != nil { |
| 42 | log.Fatal(err) |
| 43 | } |
| 44 | if held { |
| 45 | log.Print("system guard-exec recover: an apply holds the lock; it recovers at its own start") |
| 46 | return |
| 47 | } |
| 48 | } |
| 49 | networkFree := false |
| 50 | if args[0] == "recover" { |
| 51 | _, networkFree, err = rootMovesFloorIdentity() |
| 52 | if err != nil { |
| 53 | log.Fatal(err) |
| 54 | } |
| 55 | } |
| 56 | if policy.Mode != "unpublished" && !networkFree { |
| 57 | ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second) |
| 58 | store, pub, e := openSystemStore(ctx) |
| 59 | if e == nil { |
| 60 | e = hostfloor.CheckRemotePolicy(ctx, store, pub, d.Bucket, policy, local) |
| 61 | } |
| 62 | cancel() |
| 63 | if e != nil && !hostfloor.Offline(e) { |
| 64 | log.Fatal(e) |
| 65 | } |
| 66 | } |
| 67 | current := systemBinaryPath(release.OptDir()) |
| 68 | candidate := current |
| 69 | component := "router" |
| 70 | switch args[0] { |
| 71 | case "apply", "recover": |
| 72 | _, err := os.Stat(filepath.Join(systemRootDir(release.OptDir()), "journal.json")) |
| 73 | if err == nil { |
| 74 | candidate = systemPrevPath(release.OptDir()) |
| 75 | } else if !errors.Is(err, os.ErrNotExist) { |
| 76 | log.Fatal(err) |
| 77 | } |
| 78 | mode := args[0] |
| 79 | args = []string{"system", "apply"} |
| 80 | if mode == "recover" { |
| 81 | args = append(args, "--recover") |
| 82 | } |
| 83 | case "console": |
| 84 | if len(args) < 2 { |
| 85 | log.Fatal(hostfloor.ErrFloor) |
| 86 | } |
| 87 | component = "console" |
| 88 | candidate = args[1] |
| 89 | args = args[2:] |
| 90 | case "edge": |
| 91 | args = []string{"agent"} |
| 92 | default: |
| 93 | log.Fatal(hostfloor.ErrFloor) |
| 94 | } |
| 95 | if component == "router" && len(args) > 1 && args[0] == "system" { |
| 96 | id, required, e := rootMovesFloorIdentity() |
| 97 | if e != nil { |
| 98 | log.Fatal(e) |
| 99 | } |
| 100 | if required { |
| 101 | journal, e := rootFloorJournalPresent(release.OptDir()) |
| 102 | if e != nil { |
| 103 | log.Fatal(e) |
| 104 | } |
| 105 | e = withRootFloorGuardCandidate(context.Background(), release.OptDir(), id, journal && len(args) > 1 && args[0] == "system", func(selected string) error { |
| 106 | return d.Exec(selected, runtime.GOOS+"-"+runtime.GOARCH, args, os.Environ(), component) |
| 107 | }) |
| 108 | if e != nil { |
| 109 | log.Fatal(e) |
| 110 | } |
| 111 | return |
| 112 | } |
| 113 | } |
| 114 | if err = d.Exec(candidate, runtime.GOOS+"-"+runtime.GOARCH, args, os.Environ(), component); err != nil { |
| 115 | log.Fatal(err) |
| 116 | } |
| 117 | } |
| 118 | |
| 119 | // recoverLockHeld reports whether another process holds the apply lock. |
| 120 | // Only root's apply, recover and guard-arm take it, and an apply recovers |
| 121 | // any journal itself at its start, so guard-exec recover leaves the box to |
| 122 | // the holder at once. The edge drop-in Requires= the recover unit, so every |
| 123 | // edge restart inside an apply starts a recover; waiting out |
| 124 | // systemRecoverLockWait there kept the restarted edge down for 30 s (Opus 8b |
| 125 | // review I-2). A missing lock file is a free lock and is not created. |
| 126 | func recoverLockHeld(optDir string) (bool, error) { |
| 127 | lock, err := os.OpenFile(filepath.Join(systemRootDir(optDir), systemLockFile), os.O_RDWR, 0) |
| 128 | if errors.Is(err, os.ErrNotExist) { |
| 129 | return false, nil |
| 130 | } |
| 131 | if err != nil { |
| 132 | return false, err |
| 133 | } |
| 134 | defer lock.Close() |
| 135 | err = flockExclusiveNonblock(lock) |
| 136 | if errors.Is(err, syscall.EWOULDBLOCK) { |
| 137 | return true, nil |
| 138 | } |
| 139 | return false, err |
| 140 | } |
| 141 | |
amadan