//go:build linux && rootrunner package host import ( "context" "encoding/json" "errors" "net" "os" "os/exec" "path/filepath" "strings" "testing" "time" "github.com/carlosframework/platform/internal/rootstate" "golang.org/x/sys/unix" ) const unitMigrationRebootDir = "/var/lib/carlos-e52-migration" const migrationRebootUnit = "plan14-e52-migration.service" const migrationRebootSocket = "/run/carlos/plan14-e52.sock" func migrationRebootMaterial(t *testing.T) claimUnitMaterializer { t.Helper() store, err := rootstate.Open(filepath.Join(unitMigrationRebootDir, "state"), true, false) if err != nil { t.Fatal(err) } t.Cleanup(func() { store.Close() }) files := claimUnitFiles{dir: "/etc/systemd/system"} k := claimPredecessorKernel{ctl: rebootPredecessorCtl, group: claimPredecessorGroup, boot: claimUnitBoot} return claimUnitMaterializer{state: store, read: files.read, create: files.create, replace: files.replace, predecessor: k.capture, stopPredecessor: k.stop, socketCapture: captureProductionClaimUnitSocket, socketPark: parkProductionClaimUnitSocket, idle: func(string) error { return nil }} } func TestRootRebootClaimUnitMigrationPhase1(t *testing.T) { for _, dir := range []string{unitMigrationRebootDir, "/run/carlos", "/run/carlos-control"} { if err := os.Mkdir(dir, 0755); err != nil { t.Fatal(err) } } binary, err := os.Executable() if err != nil { t.Fatal(err) } body, err := os.ReadFile(binary) if err != nil { t.Fatal(err) } if err := os.WriteFile(filepath.Join(unitMigrationRebootDir, "payload"), body, 0755); err != nil { t.Fatal(err) } legacy := []byte("[Service]\nType=simple\nEnvironment=PLAN14_TENANT_SOCKET=" + migrationRebootSocket + "\nExecStart=" + unitMigrationRebootDir + "/payload plan14-tenant\n") path := "/etc/systemd/system/" + migrationRebootUnit f, err := os.OpenFile(path, os.O_CREATE|os.O_EXCL|os.O_WRONLY, 0644) if err != nil { t.Fatal(err) } _, err = f.Write(legacy) if err = errors.Join(err, f.Close()); err != nil { t.Fatal(err) } if out, err := rebootPredecessorCtl(t.Context(), "daemon-reload"); err != nil { t.Fatal(out, err) } if out, err := rebootPredecessorCtl(t.Context(), "start", migrationRebootUnit); err != nil { t.Fatal(out, err) } for until := time.Now().Add(5 * time.Second); ; { conn, err := net.DialTimeout("unix", migrationRebootSocket, time.Second) if err == nil { conn.Close() break } if time.Now().After(until) { t.Fatal(err) } time.Sleep(20 * time.Millisecond) } _, in, s := plan14UnitBindingFixture(t, migrationRebootUnit, "", migrationRebootSocket) s.Program, s.PredecessorDigest = strings.Repeat("a", 64), claimUnitDigest(legacy) b := claimUnitBinding{Schema: 1, Owner: unitBindingOwner(in), Shape: s} m := migrationRebootMaterial(t) m.stopPredecessor = func(context.Context, claimUnitPredecessor) error { return os.ErrClosed } if err := m.install(t.Context(), b); !errors.Is(err, os.ErrClosed) { t.Fatal("migration intent boundary", err) } r, err := m.load(b.Owner.Host) if err != nil || r.Migration == nil || r.Migration.Socket.Boot == "" || r.Identity != (rootstate.FileIdentity{}) { t.Fatal("lost durable migration", r, err) } raw, _ := json.Marshal(b) if err := os.WriteFile(filepath.Join(unitMigrationRebootDir, "binding.json"), raw, 0600); err != nil { t.Fatal(err) } old := claimEdgeMaskOwner defer func() { claimEdgeMaskOwner = old }() claimEdgeMaskOwner = filepath.Join(unitMigrationRebootDir, "edge-mask") if _, err := createClaimEdgeMask(t.Context(), r.Migration.Socket.Boot); err != nil { t.Fatal(err) } t.Logf("migration saved boot=%s device=%d mount=%d socket=%d:%d", r.Migration.Definition.Boot, r.Migration.Definition.Device, r.Migration.Definition.Mount, r.Migration.Socket.Device, r.Migration.Socket.Inode) } func TestRootRebootClaimUnitMigrationPhase2(t *testing.T) { t.Cleanup(func() { rebootPredecessorCtl(context.Background(), "stop", migrationRebootUnit) os.Remove("/etc/systemd/system/" + migrationRebootUnit) os.Remove("/run/systemd/system/carlos-edge.service") rebootPredecessorCtl(context.Background(), "daemon-reload") for _, dir := range []string{unitMigrationRebootDir, "/run/carlos", "/run/carlos-control"} { os.RemoveAll(dir) } }) for _, dir := range []string{"/run/carlos", "/run/carlos-control"} { if err := os.Mkdir(dir, 0755); err != nil { t.Fatal(err) } } var b claimUnitBinding raw, err := os.ReadFile(filepath.Join(unitMigrationRebootDir, "binding.json")) if err != nil || json.Unmarshal(raw, &b) != nil { t.Fatal(err) } m := migrationRebootMaterial(t) r, err := m.load(b.Owner.Host) if err != nil { t.Fatal(err) } boot, err := claimUnitBoot() if err != nil || boot == r.Migration.Definition.Boot { t.Fatal("actual reboot required", err) } oldMaskOwner := claimEdgeMaskOwner defer func() { claimEdgeMaskOwner = oldMaskOwner }() claimEdgeMaskOwner = filepath.Join(unitMigrationRebootDir, "edge-mask") if err := os.Symlink("/dev/null", "/run/systemd/system/carlos-edge.service"); err != nil { t.Fatal(err) } if out, err := rebootPredecessorCtl(t.Context(), "daemon-reload"); err != nil { t.Fatal(out, err) } if err := recoverLegacyClaimEdgeMask(t.Context()); err != nil { t.Fatal(err) } if target, err := os.Readlink("/run/systemd/system/carlos-edge.service"); err != nil || target != "/dev/null" { t.Fatal("old boot removed replacement mask", err) } if _, err := os.Lstat(claimEdgeMaskOwner); !os.IsNotExist(err) { t.Fatal("old mask record retained", err) } path := "/etc/systemd/system/" + migrationRebootUnit if err := unix.Mount(path, path, "", unix.MS_BIND, ""); err != nil { t.Fatal(err) } refused := m.install(t.Context(), b) if err := unix.Unmount(path, 0); err != nil { t.Fatal(err) } if refused == nil { t.Fatal("substituted definition mount accepted") } listener, err := net.ListenUnix("unix", &net.UnixAddr{Name: migrationRebootSocket, Net: "unix"}) if err != nil { t.Fatal(err) } if err = m.install(t.Context(), b); err == nil { t.Fatal("post-boot replacement socket removed") } if _, err := os.Lstat(migrationRebootSocket); err != nil { t.Fatal("replacement socket lost", err) } listener.Close() if err := m.install(t.Context(), b); err != nil { t.Fatal("resume migration after reboot", err) } if err := m.verify(b); err != nil { t.Fatal(err) } if err := projectClaimBoundSocket(b.Shape); err != nil { t.Fatal(err) } if target, err := os.Readlink(migrationRebootSocket); err != nil || target != claimSocketTarget(b.Shape) { t.Fatal("socket projection", target, err) } if out, err := exec.Command("systemctl", "daemon-reload").CombinedOutput(); err != nil { t.Fatal(string(out), err) } if _, err := rebootPredecessorCtl(t.Context(), "start", migrationRebootUnit); err == nil { time.Sleep(100 * time.Millisecond) pid, _ := rebootPredecessorCtl(t.Context(), "show", "-p", "MainPID", "--value", migrationRebootUnit) if pid != "0" { t.Fatal("migration started without reservation", pid) } } }