| 1 | //go:build linux && rootrunner |
| 2 | |
| 3 | package host |
| 4 | |
| 5 | import ( |
| 6 | "encoding/json" |
| 7 | "errors" |
| 8 | "os" |
| 9 | "os/exec" |
| 10 | "strings" |
| 11 | "testing" |
| 12 | |
| 13 | "github.com/carlosframework/platform/internal/controlsock" |
| 14 | "github.com/carlosframework/platform/internal/hostguard" |
| 15 | "github.com/carlosframework/platform/internal/rootstate" |
| 16 | ) |
| 17 | |
| 18 | const lifecycleCrashPath = "/var/lib/carlos-claims/lifecycle-crash" |
| 19 | |
| 20 | func lifecycleCrash(t *testing.T, point string) { |
| 21 | t.Helper() |
| 22 | if os.Getenv("PLAN14_LIFECYCLE_CRASH") != point { |
| 23 | return |
| 24 | } |
| 25 | if _, err := os.Stat(lifecycleCrashPath); err == nil { |
| 26 | return |
| 27 | } else if !errors.Is(err, os.ErrNotExist) { |
| 28 | t.Fatal(err) |
| 29 | } |
| 30 | state, err := rootstate.Open("/var/lib/carlos-claims", false, false) |
| 31 | lifecycleCheck(t, "crash checkpoint store", err) |
| 32 | lifecycleCheck(t, "durable crash checkpoint", state.Write("lifecycle-crash", []byte(point))) |
| 33 | os.Exit(86) |
| 34 | } |
| 35 | |
| 36 | type lifecycleCrashStore struct { |
| 37 | rootstate.Store |
| 38 | t *testing.T |
| 39 | } |
| 40 | |
| 41 | func (s lifecycleCrashStore) Write(name string, body []byte) error { |
| 42 | if err := s.Store.Write(name, body); err != nil { |
| 43 | return err |
| 44 | } |
| 45 | if name != lifecycleHost+".json" { |
| 46 | return nil |
| 47 | } |
| 48 | var r struct { |
| 49 | Installation hostguard.RootInstallation |
| 50 | Retired []hostguard.Captured |
| 51 | Barrier *hostguard.FenceTarget |
| 52 | State struct { |
| 53 | ActiveStamp uint64 `json:"active_stamp"` |
| 54 | } |
| 55 | } |
| 56 | if err := json.Unmarshal(body, &r); err != nil { |
| 57 | return err |
| 58 | } |
| 59 | if r.Barrier != nil { |
| 60 | lifecycleCrash(s.t, "handover") |
| 61 | } |
| 62 | if r.State.ActiveStamp == 0 && r.Installation.Stamp != 0 && len(r.Retired) != 0 { |
| 63 | lifecycleCrash(s.t, "retirement") |
| 64 | } |
| 65 | return nil |
| 66 | } |
| 67 | |
| 68 | func lifecycleControlFaults(t *testing.T) { |
| 69 | original := openClaimRootAuthority |
| 70 | openClaimRootAuthority = func() (rootstate.Store, error) { |
| 71 | s, err := original() |
| 72 | if err != nil { |
| 73 | return nil, err |
| 74 | } |
| 75 | return lifecycleCrashStore{s, t}, nil |
| 76 | } |
| 77 | t.Cleanup(func() { openClaimRootAuthority = original }) |
| 78 | } |
| 79 | |
| 80 | func lifecycleStartControl(t *testing.T, endpoint string) func() { |
| 81 | t.Helper() |
| 82 | raw, err := unitsFS.ReadFile("units/carlos-instance-control.service") |
| 83 | lifecycleCheck(t, "controller template", err) |
| 84 | var lines []string |
| 85 | for _, line := range strings.Split(string(raw), "\n") { |
| 86 | key, _, _ := strings.Cut(line, "=") |
| 87 | if key == "ExecStart" || key == "Requires" || key == "After" { |
| 88 | continue |
| 89 | } |
| 90 | lines = append(lines, line) |
| 91 | } |
| 92 | lines = append(lines, "[Service]", "ExecStart=/opt/carlos/carlos -test.run=^TestPlan14LifecycleControlProcessPayload$ -test.v -test.timeout=12m", "Environment=PLAN14_LIFECYCLE_CONTROL=1", "Environment=PLAN14_LIFECYCLE_CRASH="+os.Getenv("PLAN14_LIFECYCLE_CRASH"), "Environment=CARLOS_S3_ENDPOINT="+endpoint, "ReadWritePaths=/data /opt/carlos /var/lib/carlos-claims /var/lib/carlos-release /sys/fs/cgroup "+lifecycleHome, "RestrictAddressFamilies=", "RestrictAddressFamilies=AF_UNIX AF_INET", "MemoryMax=512M", "StandardOutput=append:"+lifecycleHome+"/controller.log", "StandardError=append:"+lifecycleHome+"/controller.log") |
| 93 | path := "/run/systemd/system/carlos-instance-control.service" |
| 94 | f, err := os.OpenFile(path, os.O_CREATE|os.O_EXCL|os.O_WRONLY, 0644) |
| 95 | lifecycleCheck(t, "exclusive controller unit", err) |
| 96 | _, err = f.WriteString(strings.Join(lines, "\n")) |
| 97 | f.Close() |
| 98 | lifecycleCheck(t, "controller unit", err) |
| 99 | out, err := exec.Command("systemctl", "daemon-reload").CombinedOutput() |
| 100 | lifecycleCheck(t, string(out), err) |
| 101 | start := func() { |
| 102 | exec.Command("systemctl", "stop", "carlos-instance-control.service").Run() |
| 103 | os.Remove("/run/carlos-control/lifecycle-ready") |
| 104 | os.Remove(controlsock.Path) |
| 105 | out, err := exec.Command("systemctl", "start", "carlos-instance-control.service").CombinedOutput() |
| 106 | lifecycleCheck(t, string(out), err) |
| 107 | } |
| 108 | t.Cleanup(func() { |
| 109 | exec.Command("systemctl", "stop", "carlos-instance-control.service").Run() |
| 110 | os.Remove(path) |
| 111 | exec.Command("systemctl", "daemon-reload").Run() |
| 112 | body, _ := os.ReadFile(lifecycleHome + "/controller.log") |
| 113 | t.Logf("controller log: %s", body) |
| 114 | }) |
| 115 | start() |
| 116 | return start |
| 117 | } |
| 118 | |
amadan