| 1 | package hostfence |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/json" |
| 6 | "errors" |
| 7 | "github.com/carlosframework/platform/internal/awsinv" |
| 8 | "github.com/carlosframework/platform/internal/hostactors" |
| 9 | "github.com/carlosframework/platform/internal/hostfloor" |
| 10 | "github.com/carlosframework/platform/internal/release" |
| 11 | "github.com/carlosframework/platform/internal/termfence" |
| 12 | "time" |
| 13 | ) |
| 14 | |
| 15 | type Worker = termfence.WrittenBy |
| 16 | type Termination = termfence.Record |
| 17 | type EvidenceRef struct { |
| 18 | Enrollment string `json:"enrollment"` |
| 19 | Key string `json:"key"` |
| 20 | SHA256 string `json:"sha256"` |
| 21 | } |
| 22 | type TerminationWorker struct { |
| 23 | Fences release.Store |
| 24 | Enabled bool |
| 25 | Directory hostactors.Directory |
| 26 | Self Worker |
| 27 | Reader func(context.Context, string) (awsinv.ExactEC2API, error) |
| 28 | Now func() time.Time |
| 29 | } |
| 30 | |
| 31 | func FenceKey(enrollment string) string { return "control/box-fences/" + enrollment + ".json" } |
| 32 | func (w TerminationWorker) binding(ctx context.Context, box string) (hostactors.Box, error) { |
| 33 | if !w.Enabled || w.Now == nil || w.Fences == nil { |
| 34 | return hostactors.Box{}, ErrObservation |
| 35 | } |
| 36 | o, err := w.Directory.Read(ctx) |
| 37 | if err != nil { |
| 38 | return hostactors.Box{}, err |
| 39 | } |
| 40 | self, ok := o.Roster.Boxes[w.Self.Box] |
| 41 | if !ok || self.State != "active" || self.Enrollment != w.Self.Enrollment { |
| 42 | return hostactors.Box{}, ErrObservation |
| 43 | } |
| 44 | eligible, err := w.Directory.Lookup(ctx, w.Self.Box) |
| 45 | if err != nil || eligible.State != "active" || eligible.Enrollment != w.Self.Enrollment { |
| 46 | return hostactors.Box{}, ErrObservation |
| 47 | } |
| 48 | target, ok := o.Roster.Boxes[box] |
| 49 | if !ok || target.Region != self.Region || target.Cloud.Region != self.Cloud.Region || target.Cloud.Account != self.Cloud.Account { |
| 50 | return hostactors.Box{}, ErrObservation |
| 51 | } |
| 52 | return target, nil |
| 53 | } |
| 54 | func (w TerminationWorker) Describe(ctx context.Context, box string) (EvidenceRef, error) { |
| 55 | b, err := w.binding(ctx, box) |
| 56 | if err != nil { |
| 57 | return EvidenceRef{}, err |
| 58 | } |
| 59 | if w.Reader == nil { |
| 60 | return EvidenceRef{}, ErrObservation |
| 61 | } |
| 62 | api, err := w.Reader(ctx, b.Cloud.Region) |
| 63 | if err != nil { |
| 64 | return EvidenceRef{}, err |
| 65 | } |
| 66 | out, err := termfence.DescribeInstance(ctx, api, b.Cloud) |
| 67 | if err != nil { |
| 68 | return EvidenceRef{}, err |
| 69 | } |
| 70 | return w.verifyAndPersist(ctx, box, b, termfence.FromDescribe(out, b.Cloud.InstanceID, b.Cloud.Region)) |
| 71 | } |
| 72 | |
| 73 | func (w TerminationWorker) verifyAndPersist(ctx context.Context, box string, b hostactors.Box, observation termfence.Observation) (EvidenceRef, error) { |
| 74 | o, err := w.Directory.Read(ctx) |
| 75 | if err != nil { |
| 76 | return EvidenceRef{}, err |
| 77 | } |
| 78 | self, ok := o.Roster.Boxes[w.Self.Box] |
| 79 | if !ok || self.State != "active" || self.Enrollment != w.Self.Enrollment { |
| 80 | return EvidenceRef{}, ErrObservation |
| 81 | } |
| 82 | f, err := termfence.Verify(termfence.Box{Ref: box, Enrollment: b.Enrollment, Cloud: b.Cloud}, termfence.Box{Ref: w.Self.Box, Enrollment: w.Self.Enrollment, Cloud: self.Cloud}, observation, w.Now()) |
| 83 | if err != nil { |
| 84 | return EvidenceRef{}, err |
| 85 | } |
| 86 | f.ObservedAt = w.Now().UTC().Format(time.RFC3339Nano) |
| 87 | if err := hostactors.ValidateTerminationFence(o, box, f, w.Now()); err != nil { |
| 88 | return EvidenceRef{}, err |
| 89 | } |
| 90 | return w.persist(ctx, f) |
| 91 | } |
| 92 | |
| 93 | // Event is called only by the regional queue consumer, after the existing |
| 94 | // queue URL and rule-ARN policy validation. It is never an HTTP request body. |
| 95 | func (w TerminationWorker) Event(ctx context.Context, box string, raw []byte) (EvidenceRef, error) { |
| 96 | b, err := w.binding(ctx, box) |
| 97 | if err != nil { |
| 98 | return EvidenceRef{}, err |
| 99 | } |
| 100 | observation, err := termfence.FromEvent(raw) |
| 101 | if err != nil { |
| 102 | return EvidenceRef{}, err |
| 103 | } |
| 104 | var event struct { |
| 105 | Time time.Time `json:"time"` |
| 106 | } |
| 107 | if json.Unmarshal(raw, &event) != nil || event.Time.IsZero() || event.Time.After(w.Now().Add(5*time.Second)) { |
| 108 | return EvidenceRef{}, ErrObservation |
| 109 | } |
| 110 | return w.verifyAndPersist(ctx, box, b, observation) |
| 111 | } |
| 112 | |
| 113 | func (w TerminationWorker) persist(ctx context.Context, f Termination) (EvidenceRef, error) { |
| 114 | raw, err := json.Marshal(f) |
| 115 | if err != nil { |
| 116 | return EvidenceRef{}, err |
| 117 | } |
| 118 | key := FenceKey(f.Enrollment) |
| 119 | _, err = w.Fences.Put(ctx, key, raw, "") |
| 120 | if err != nil { |
| 121 | saved, _, readErr := w.Fences.Get(ctx, key) |
| 122 | if readErr != nil { |
| 123 | return EvidenceRef{}, errors.Join(err, readErr) |
| 124 | } |
| 125 | var old Termination |
| 126 | o, readErr := w.Directory.Read(ctx) |
| 127 | if readErr != nil { |
| 128 | return EvidenceRef{}, readErr |
| 129 | } |
| 130 | if json.Unmarshal(saved, &old) != nil || old.Enrollment != f.Enrollment || old.Cloud != f.Cloud || w.Directory.ValidateTerminationFenceAuthority(ctx, o, f.Box, old, w.Now()) != nil { |
| 131 | return EvidenceRef{}, ErrObservation |
| 132 | } |
| 133 | raw = saved |
| 134 | } |
| 135 | return EvidenceRef{Enrollment: f.Enrollment, Key: key, SHA256: hostfloor.Digest(raw)}, nil |
| 136 | } |
| 137 | |