| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "context" |
| 6 | "crypto/sha256" |
| 7 | "encoding/hex" |
| 8 | "encoding/json" |
| 9 | "fmt" |
| 10 | "io" |
| 11 | "net" |
| 12 | "net/http" |
| 13 | "reflect" |
| 14 | "slices" |
| 15 | "strconv" |
| 16 | "time" |
| 17 | |
| 18 | "github.com/carlosframework/platform/internal/hostbox" |
| 19 | "github.com/carlosframework/platform/internal/hostclaim" |
| 20 | "github.com/carlosframework/platform/internal/hostguard" |
| 21 | ) |
| 22 | |
| 23 | type recoveryRebindRequest struct { |
| 24 | V int |
| 25 | Prior hostguard.Captured |
| 26 | Revision uint64 |
| 27 | RecordKey, RecordETag string |
| 28 | Next hostbox.ColdSuccessor |
| 29 | } |
| 30 | |
| 31 | type recoveryRebindIntent struct{ Recovery *recoveryRebindRequest } |
| 32 | |
| 33 | func recoveryInventory(in hostbox.StopInventory, next hostbox.ColdSuccessor) hostbox.StopInventory { |
| 34 | in.Entries = append([]hostbox.StopInventoryEntry(nil), in.Entries...) |
| 35 | in.InstallationStamp, in.DirectoryRevision = next.Stamp, next.DirectoryRevision |
| 36 | sum := sha256.Sum256([]byte(fmt.Sprintf("%s\x00%s\x00%d\x00%s", next.Host, next.Incarnation, next.Stamp, next.Box))) |
| 37 | owner := hex.EncodeToString(sum[:]) |
| 38 | for i := range in.Entries { |
| 39 | e := &in.Entries[i] |
| 40 | e.Ownership = owner |
| 41 | if e.Kind == "route" || e.Kind == "timer" { |
| 42 | name := e.RelativePath |
| 43 | if e.Kind == "timer" { |
| 44 | name = name[len(next.Host)+1:] |
| 45 | } |
| 46 | sum := sha256.Sum256([]byte(owner + "\x00" + e.Kind + "\x00" + name)) |
| 47 | e.RowIdentity = hex.EncodeToString(sum[:]) |
| 48 | } |
| 49 | } |
| 50 | return in |
| 51 | } |
| 52 | |
| 53 | func (h ClaimRootHTTP) rebindRecoveryInventory(ctx context.Context, q recoveryRebindRequest) error { |
| 54 | guard := h.Stop.Coordinator.Guard |
| 55 | if guard.Root == nil || h.Reconcile.Inventory == nil || h.InventoryActor == nil || q.V != 1 || q.Revision == 0 || q.RecordKey == "" || q.RecordETag == "" { |
| 56 | return hostbox.ErrUnsupported |
| 57 | } |
| 58 | if h.RequireNewWork != nil { |
| 59 | if err := h.RequireNewWork(ctx); err != nil { |
| 60 | return err |
| 61 | } |
| 62 | } |
| 63 | actor, err := h.InventoryActor(ctx) |
| 64 | if err != nil { |
| 65 | return err |
| 66 | } |
| 67 | n := q.Next |
| 68 | if n.Box != actor.Box || n.Enrollment != actor.Enrollment || actor.Binding == "" || n.DirectoryRevision != actor.Binding || n.Incarnation != q.Prior.Incarnation || n.Stamp <= q.Prior.Stamp || n.Session != n.Stamp || n.Generation == "" || n.ETag == "" { |
| 69 | return hostbox.ErrChanged |
| 70 | } |
| 71 | source, revision, err := h.Reconcile.Inventory.Snapshot(ctx, n.Host, q.Prior) |
| 72 | if err != nil { |
| 73 | return err |
| 74 | } |
| 75 | // The predecessor is reused only under today's binding (roster binding |
| 76 | // R3): an installation stamped under another binding (before a |
| 77 | // quarantine, or a numeral) is refused before any reuse, so the rebound |
| 78 | // inventory copies a verified value. |
| 79 | if revision != q.Revision || !source.Complete || source.Host != n.Host || source.Incarnation != q.Prior.Incarnation || source.InstallationStamp != q.Prior.Stamp || source.Box != actor.Box || source.Enrollment != actor.Enrollment || source.DirectoryRevision != actor.Binding || len(source.Entries) == 0 { |
| 80 | return hostbox.ErrTypedInventory |
| 81 | } |
| 82 | observed, err := hostclaim.Read(ctx, h.Stop.Coordinator.Evidence, n.Host) |
| 83 | if err != nil { |
| 84 | return err |
| 85 | } |
| 86 | if observed.Claim == nil || observed.Claim.Holder == nil { |
| 87 | return hostbox.ErrChanged |
| 88 | } |
| 89 | holder := observed.Claim.Holder |
| 90 | next := hostguard.RootInstallation{Host: n.Host, Incarnation: n.Incarnation, Stamp: n.Stamp, Account: holder.Account, App: holder.App, Record: q.RecordKey, Box: n.Box, Enrollment: n.Enrollment, Migration: 1, InventoryRevision: 1} |
| 91 | var rebound hostbox.StopInventory |
| 92 | check := func(c *hostclaim.Claim) error { |
| 93 | if slices.Contains(c.Lost, n.Box) { |
| 94 | return hostbox.ErrChanged |
| 95 | } |
| 96 | fresh, err := hostclaim.Read(ctx, h.Stop.Coordinator.Evidence, n.Host) |
| 97 | if err != nil { |
| 98 | return err |
| 99 | } |
| 100 | if fresh.Absent || fresh.ETag != n.ETag || !reflect.DeepEqual(c, fresh.Claim) || c.Holder == nil { |
| 101 | return hostbox.ErrChanged |
| 102 | } |
| 103 | holder := c.Holder |
| 104 | stamp, err := strconv.ParseUint(c.Op.Journal[hostclaim.JournalSourceStamp], 10, 64) |
| 105 | if err != nil || stamp != q.Prior.Stamp || !claimRecoveryKind(c.Op.Kind) || c.Op.Phase != hostclaim.PhaseProbing || c.Op.ID != n.Op || c.Op.Phase != n.Phase || c.Op.Exec != n.Box || c.Op.Intent["source"] != n.Box || holder.Key != q.RecordKey || holder.Incarnation != n.Incarnation || holder.Stamp != n.Stamp || holder.Session != n.Session || holder.Generation != n.Generation || holder.WriterBox != n.Box { |
| 106 | return hostbox.ErrChanged |
| 107 | } |
| 108 | if len(holder.Admissions) == 0 { |
| 109 | return hostbox.ErrChanged |
| 110 | } |
| 111 | a := holder.Admissions[len(holder.Admissions)-1] |
| 112 | if a.Mode != "cold" || a.Session != n.Session || a.Generation != n.Generation || a.Box != n.Box { |
| 113 | return hostbox.ErrChanged |
| 114 | } |
| 115 | canonical, etag, err := rootClaimRecord(ctx, h.Stop.Coordinator.Evidence, c, h.Roots["data"]) |
| 116 | if err != nil { |
| 117 | return err |
| 118 | } |
| 119 | if etag != q.RecordETag || canonical.Host != n.Host || canonical.Account != next.Account || canonical.App != next.App { |
| 120 | return hostbox.ErrChanged |
| 121 | } |
| 122 | canonical.ClaimIncarnation, canonical.ClaimStamp = q.Prior.Incarnation, q.Prior.Stamp |
| 123 | routes := 0 |
| 124 | for _, e := range source.Entries { |
| 125 | if e.Kind == "route" { |
| 126 | routes++ |
| 127 | canonical.ClaimOwner, canonical.ClaimRowID = e.Ownership, e.RowIdentity |
| 128 | } |
| 129 | } |
| 130 | if routes != 1 || hostbox.ValidateAgentInventory(source, &canonical, h.Roots["data"]) != nil { |
| 131 | return hostbox.ErrTypedInventory |
| 132 | } |
| 133 | current, rev, err := h.Reconcile.Inventory.Snapshot(ctx, n.Host, q.Prior) |
| 134 | if err != nil { |
| 135 | return err |
| 136 | } |
| 137 | if rev != q.Revision || !reflect.DeepEqual(source, current) { |
| 138 | return hostbox.ErrChanged |
| 139 | } |
| 140 | rebound = recoveryInventory(source, n) |
| 141 | return nil |
| 142 | } |
| 143 | raw, err := json.Marshal(recoveryRebindIntent{Recovery: &q}) |
| 144 | if err != nil { |
| 145 | return err |
| 146 | } |
| 147 | return guard.RebindRootInventory(ctx, q.Prior, q.Revision, next, raw, check, func() error { |
| 148 | current, rev, err := h.Reconcile.Inventory.Snapshot(ctx, n.Host, hostguard.Captured{Incarnation: n.Incarnation, Stamp: n.Stamp}) |
| 149 | if err != nil { |
| 150 | return err |
| 151 | } |
| 152 | if rev != 0 { |
| 153 | if rev != 1 || !reflect.DeepEqual(current, rebound) { |
| 154 | return hostbox.ErrChanged |
| 155 | } |
| 156 | return nil |
| 157 | } |
| 158 | return h.Reconcile.Inventory.WriteStopInventory(ctx, rebound) |
| 159 | }) |
| 160 | } |
| 161 | |
| 162 | func (h ClaimRootHTTP) serveRecoveryRebind(w http.ResponseWriter, r *http.Request) { |
| 163 | var q recoveryRebindRequest |
| 164 | dec := json.NewDecoder(io.LimitReader(r.Body, 32<<10)) |
| 165 | dec.DisallowUnknownFields() |
| 166 | if dec.Decode(&q) != nil || dec.Decode(new(any)) != io.EOF { |
| 167 | w.WriteHeader(http.StatusBadRequest) |
| 168 | return |
| 169 | } |
| 170 | if err := h.rebindRecoveryInventory(r.Context(), q); err != nil { |
| 171 | w.WriteHeader(http.StatusConflict) |
| 172 | return |
| 173 | } |
| 174 | w.WriteHeader(http.StatusNoContent) |
| 175 | } |
| 176 | func recoveryRebindClient(socket string) func(context.Context, recoveryRebindRequest) error { |
| 177 | return func(ctx context.Context, q recoveryRebindRequest) error { |
| 178 | body, err := json.Marshal(q) |
| 179 | if err != nil { |
| 180 | return err |
| 181 | } |
| 182 | transport := &http.Transport{DialContext: func(ctx context.Context, _, _ string) (net.Conn, error) { |
| 183 | return (&net.Dialer{}).DialContext(ctx, "unix", socket) |
| 184 | }} |
| 185 | defer transport.CloseIdleConnections() |
| 186 | client := &http.Client{Transport: transport, Timeout: 30 * time.Second} |
| 187 | for attempt := 0; attempt < 2; attempt++ { |
| 188 | req, e := http.NewRequestWithContext(ctx, http.MethodPost, "http://instance-control/claim/recovery-rebind", bytes.NewReader(body)) |
| 189 | if e != nil { |
| 190 | return e |
| 191 | } |
| 192 | response, e := client.Do(req) |
| 193 | if e != nil { |
| 194 | err = e |
| 195 | continue |
| 196 | } |
| 197 | response.Body.Close() |
| 198 | if response.StatusCode != http.StatusNoContent { |
| 199 | return hostbox.ErrChanged |
| 200 | } |
| 201 | return nil |
| 202 | } |
| 203 | return err |
| 204 | } |
| 205 | } |
| 206 | |