| 1 | package console |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/json" |
| 6 | "errors" |
| 7 | "fmt" |
| 8 | "log" |
| 9 | "log/slog" |
| 10 | "os" |
| 11 | "path/filepath" |
| 12 | "runtime" |
| 13 | "strings" |
| 14 | "time" |
| 15 | |
| 16 | "github.com/carlosframework/platform/internal/hostactors" |
| 17 | "github.com/carlosframework/platform/internal/hostfloor" |
| 18 | "github.com/carlosframework/platform/internal/release" |
| 19 | "github.com/carlosframework/platform/internal/version" |
| 20 | ) |
| 21 | |
| 22 | // hostRunningImage is the census's seam for what this process is running. It |
| 23 | // is nil in production, where the console reads its guard files off the box |
| 24 | // (/opt/carlos/system/guard and the systemd drop-in); a test that has no box |
| 25 | // installs one, so the worker's real write path can be exercised. |
| 26 | var hostRunningImage func() (hostfloor.Running, error) |
| 27 | |
| 28 | func (s *Server) localHostReport(ctx context.Context) (hostactors.Report, error) { |
| 29 | if hostRunningImage != nil { |
| 30 | return s.hostReport(ctx, hostRunningImage) |
| 31 | } |
| 32 | disk := hostfloor.Disk{Dir: filepath.Join(release.OptDir(), "system", "guard"), Bucket: s.cfg.Bucket, PublicKey: s.cfg.ReleasePub} |
| 33 | return s.hostReport(ctx, func() (hostfloor.Running, error) { |
| 34 | return disk.Running(version.Build, runtime.GOOS+"-"+runtime.GOARCH, "console") |
| 35 | }) |
| 36 | } |
| 37 | |
| 38 | func (s *Server) hostReport(ctx context.Context, runningImage func() (hostfloor.Running, error)) (hostactors.Report, error) { |
| 39 | id := s.cfg.HostConsoleID |
| 40 | if id == "" { |
| 41 | return hostactors.Report{}, hostactors.ErrRoster |
| 42 | } |
| 43 | d := hostactors.Directory{Store: s.cfg.Releases, PublicKey: s.cfg.ReleasePub, Bucket: s.cfg.Bucket} |
| 44 | roster, err := d.Read(ctx) |
| 45 | if err != nil { |
| 46 | return hostactors.Report{}, err |
| 47 | } |
| 48 | c, ok := roster.Roster.Consoles[id] |
| 49 | if !ok { |
| 50 | c, err = d.InvitedConsole(ctx, roster, id) |
| 51 | if err != nil { |
| 52 | return hostactors.Report{}, err |
| 53 | } |
| 54 | } |
| 55 | if c.Compute != nil { |
| 56 | local, err := hostfloor.ReadEnrollment("/etc/carlos/host-actor.json") |
| 57 | box, exists := roster.Roster.Boxes[c.Compute.Box] |
| 58 | if err != nil || !exists || local.Box != c.Compute.Box || local.Enrollment != c.Compute.Enrollment || box.Enrollment != local.Enrollment || local.InstanceID != box.Cloud.InstanceID { |
| 59 | return hostactors.Report{}, hostactors.ErrRoster |
| 60 | } |
| 61 | } |
| 62 | running, err := runningImage() |
| 63 | if err != nil { |
| 64 | return hostactors.Report{}, err |
| 65 | } |
| 66 | return hostactors.Report{Compute: c.Compute, Component: "console", GeneratedAt: time.Now().UTC(), Actor: id, Enrollment: c.Enrollment, InstanceID: id, ProcessID: hostReportProcessID(running.BootID), Release: version.Build, ReleaseTag: running.ReleaseTag, Platform: runtime.GOOS + "-" + runtime.GOARCH, BinarySHA256: running.Hash, MovesProtocol: running.Protocol, BootID: running.BootID, FloorSHA256: running.FloorSHA256, FloorGuardReady: running.GuardReady, HostClaimEnabled: s.cfg.HostClaim}, nil |
| 67 | } |
| 68 | func (s *Server) StartHostReports(ctx context.Context) { |
| 69 | // Reporting before the flag is enabled is harmless and needed for arming. |
| 70 | if s.cfg.HostConsoleID == "" || strings.ContainsAny(s.cfg.HostConsoleID, "/\\") || s.cfg.Releases == nil { |
| 71 | return |
| 72 | } |
| 73 | ticker := time.NewTicker(5 * time.Minute) |
| 74 | defer ticker.Stop() |
| 75 | s.runHostReports(ctx, ticker.C) |
| 76 | } |
| 77 | |
| 78 | func (s *Server) runHostReports(ctx context.Context, ticks <-chan time.Time) { |
| 79 | collisionLogged := false |
| 80 | for { |
| 81 | report, err := s.localHostReport(ctx) |
| 82 | if err == nil { |
| 83 | key := "status/consoles/" + s.cfg.HostConsoleID + ".json" |
| 84 | raw, _ := json.Marshal(report) |
| 85 | err = putHostReport(ctx, s.cfg.Releases, key, raw) |
| 86 | } |
| 87 | if len(s.cfg.Key) != 0 { |
| 88 | d := hostactors.Directory{Store: s.cfg.Releases, PublicKey: s.cfg.ReleasePub, Bucket: s.cfg.Bucket} |
| 89 | if o, e := d.Read(ctx); e == nil { |
| 90 | for id := range o.Roster.Activations { |
| 91 | _ = d.ReleaseActivation(ctx, release.RootSigner(s.cfg.Key), id, time.Now()) |
| 92 | } |
| 93 | } |
| 94 | } |
| 95 | if errors.Is(err, errHostReportCollision) { |
| 96 | if !collisionLogged { |
| 97 | slog.Info("console: host census report", "collision", err) |
| 98 | collisionLogged = true |
| 99 | } |
| 100 | } else if err != nil { |
| 101 | log.Printf("console: host census report: %v", err) |
| 102 | } else { |
| 103 | collisionLogged = false |
| 104 | } |
| 105 | select { |
| 106 | case <-ctx.Done(): |
| 107 | return |
| 108 | case <-ticks: |
| 109 | } |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | // putHostReport replaces the console's report, creating it when absent. |
| 114 | // |
| 115 | // A fresh report refuses replacement unless its recorded process is provably |
| 116 | // gone on this boot. The replacement uses the checked ETag. |
| 117 | func putHostReport(ctx context.Context, store release.Store, key string, raw []byte) error { |
| 118 | old, etag, err := store.Get(ctx, key) |
| 119 | if err != nil && !errors.Is(err, os.ErrNotExist) { // S3Store wraps it |
| 120 | return err |
| 121 | } |
| 122 | if err == nil { |
| 123 | if err := hostReportCollision(old, raw, time.Now()); err != nil { |
| 124 | return err |
| 125 | } |
| 126 | } |
| 127 | _, err = store.Put(ctx, key, raw, etag) |
| 128 | return err |
| 129 | } |
| 130 | |
| 131 | var errHostReportCollision = errors.New("another console process is using the same CARLOS_CONSOLE_ID") |
| 132 | |
| 133 | func hostReportCollision(old, raw []byte, now time.Time) error { |
| 134 | var prev, next hostactors.Report |
| 135 | if json.Unmarshal(old, &prev) != nil || json.Unmarshal(raw, &next) != nil { |
| 136 | return nil |
| 137 | } |
| 138 | if prev.ProcessID == "" || prev.ProcessID == next.ProcessID || prev.GeneratedAt.IsZero() { |
| 139 | return nil |
| 140 | } |
| 141 | age := now.Sub(prev.GeneratedAt) |
| 142 | if age > 15*time.Minute || age < -5*time.Second { |
| 143 | return nil |
| 144 | } |
| 145 | if hostReportProcessGone(prev.ProcessID, next.ProcessID) { |
| 146 | return nil |
| 147 | } |
| 148 | return fmt.Errorf("console report: process %s published this console's report at %s; %w", prev.ProcessID, prev.GeneratedAt.Format(time.RFC3339), errHostReportCollision) |
| 149 | } |
| 150 | |
amadan