ipn,types/persist: store disallowed TKA's in prefs, lock local-disable

Signed-off-by: Tom DNetto <tom@tailscale.com>

client/tailscale/localclient.go

@@ -853,6 +853,21 @@ func (lc *LocalClient) NetworkLockLog(ctx context.Context, maxEntries int) ([]ip
 	return decodeJSON[[]ipnstate.NetworkLockUpdate](body)
 }
 
+// NetworkLockForceLocalDisable forcibly shuts down network lock on this node.
+func (lc *LocalClient) NetworkLockForceLocalDisable(ctx context.Context) error {
+	// This endpoint expects an empty JSON stanza as the payload.
+	var b bytes.Buffer
+	if err := json.NewEncoder(&b).Encode(struct{}{}); err != nil {
+		return err
+	}
+
+	if _, err := lc.send(ctx, "POST", "/localapi/v0/tka/force-local-disable", 200, &b); err != nil {
+		return fmt.Errorf("error: %w", err)
+	}
+	return nil
+}
+
+
 // SetServeConfig sets or replaces the serving settings.
 // If config is nil, settings are cleared and serving is disabled.
 func (lc *LocalClient) SetServeConfig(ctx context.Context, config *ipn.ServeConfig) error {

cmd/tailscale/cli/network-lock.go

@@ -37,6 +37,7 @@
 		nlDisableCmd,
 		nlDisablementKDFCmd,
 		nlLogCmd,
+		nlLocalDisableCmd,
 	},
 	Exec: runNetworkLockStatus,
 }
@@ -348,6 +349,17 @@ func runNetworkLockDisable(ctx context.Context, args []string) error {
 	return localClient.NetworkLockDisable(ctx, secrets[0])
 }
 
+var nlLocalDisableCmd = &ffcli.Command{
+	Name:       "local-disable",
+	ShortUsage: "local-disable",
+	ShortHelp:  "Disables the currently-active tailnet lock for this node",
+	Exec:       runNetworkLockLocalDisable,
+}
+
+func runNetworkLockLocalDisable(ctx context.Context, args []string) error {
+	return localClient.NetworkLockForceLocalDisable(ctx)
+}
+
 var nlDisablementKDFCmd = &ffcli.Command{
 	Name:       "disablement-kdf",
 	ShortUsage: "disablement-kdf <hex-encoded-disablement-secret>",

ipn/ipnlocal/network-lock.go

@@ -25,6 +25,7 @@
 	"tailscale.com/tka"
 	"tailscale.com/types/key"
 	"tailscale.com/types/netmap"
+	"tailscale.com/types/persist"
 	"tailscale.com/types/tkatype"
 	"tailscale.com/util/mak"
 )
@@ -134,7 +135,7 @@ func (b *LocalBackend) tkaSyncIfNeeded(nm *netmap.NetworkMap, prefs ipn.PrefsVie
 		}
 
 		if wantEnabled && !isEnabled {
-			if err := b.tkaBootstrapFromGenesisLocked(bs.GenesisAUM); err != nil {
+			if err := b.tkaBootstrapFromGenesisLocked(bs.GenesisAUM, prefs.Persist()); err != nil {
 				return fmt.Errorf("bootstrap: %w", err)
 			}
 			isEnabled = true
@@ -278,7 +279,7 @@ func (b *LocalBackend) chonkPathLocked() string {
 // tailnet key authority, based on the given genesis AUM.
 //
 // b.mu must be held.
-func (b *LocalBackend) tkaBootstrapFromGenesisLocked(g tkatype.MarshaledAUM) error {
+func (b *LocalBackend) tkaBootstrapFromGenesisLocked(g tkatype.MarshaledAUM, persist *persist.Persist) error {
 	if err := b.CanSupportNetworkLock(); err != nil {
 		return err
 	}
@@ -288,6 +289,19 @@ func (b *LocalBackend) tkaBootstrapFromGenesisLocked(g tkatype.MarshaledAUM) err
 		return fmt.Errorf("reading genesis: %v", err)
 	}
 
+	if persist != nil && len(persist.DisallowedTKAStateIDs) > 0 {
+		if genesis.State == nil {
+			return errors.New("invalid genesis: missing State")
+		}
+		bootstrapStateID := fmt.Sprintf("%d:%d", genesis.State.StateID1, genesis.State.StateID2)
+
+		for _, stateID := range persist.DisallowedTKAStateIDs {
+			if stateID == bootstrapStateID {
+				return fmt.Errorf("TKA with stateID of %q is disallowed on this node", stateID)
+			}
+		}
+	}
+
 	chonkDir := b.chonkPathLocked()
 	if err := os.Mkdir(filepath.Dir(chonkDir), 0755); err != nil && !os.IsExist(err) {
 		return fmt.Errorf("creating chonk root dir: %v", err)
@@ -495,6 +509,31 @@ func (b *LocalBackend) NetworkLockKeyTrustedForTest(keyID tkatype.KeyID) bool {
 	return b.tka.authority.KeyTrusted(keyID)
 }
 
+// NetworkLockForceLocalDisable shuts down TKA locally, and denylists the current
+// TKA from being initialized locally in future.
+func (b *LocalBackend) NetworkLockForceLocalDisable() error {
+	b.mu.Lock()
+	defer b.mu.Unlock()
+	if b.tka == nil {
+		return errNetworkLockNotActive
+	}
+
+	id1, id2 := b.tka.authority.StateIDs()
+	stateID := fmt.Sprintf("%d:%d", id1, id2)
+
+	newPrefs := b.pm.CurrentPrefs().AsStruct().Clone() // .Persist should always be initialized here.
+	newPrefs.Persist.DisallowedTKAStateIDs = append(newPrefs.Persist.DisallowedTKAStateIDs, stateID)
+	if err := b.pm.SetPrefs(newPrefs.View()); err != nil {
+		return fmt.Errorf("saving prefs: %w", err)
+	}
+
+	if err := os.RemoveAll(b.chonkPathLocked()); err != nil {
+		return fmt.Errorf("deleting TKA state: %w", err)
+	}
+	b.tka = nil
+	return nil
+}
+
 // NetworkLockSign signs the given node-key and submits it to the control plane.
 // rotationPublic, if specified, must be an ed25519 public key.
 func (b *LocalBackend) NetworkLockSign(nodeKey key.NodePublic, rotationPublic []byte) error {

ipn/ipnlocal/network-lock_test.go

@@ -778,3 +778,103 @@ func TestTKASign(t *testing.T) {
 		t.Errorf("NetworkLockSign() failed: %v", err)
 	}
 }
+
+func TestTKAForceDisable(t *testing.T) {
+	envknob.Setenv("TAILSCALE_USE_WIP_CODE", "1")
+	defer envknob.Setenv("TAILSCALE_USE_WIP_CODE", "")
+	nodePriv := key.NewNode()
+
+	// Make a fake TKA authority, to seed local state.
+	disablementSecret := bytes.Repeat([]byte{0xa5}, 32)
+	nlPriv := key.NewNLPrivate()
+	key := tka.Key{Kind: tka.Key25519, Public: nlPriv.Public().Verifier(), Votes: 2}
+
+	pm := must.Get(newProfileManager(new(mem.Store), t.Logf, ""))
+	must.Do(pm.SetPrefs((&ipn.Prefs{
+		Persist: &persist.Persist{
+			PrivateNodeKey: nodePriv,
+			NetworkLockKey: nlPriv,
+		},
+	}).View()))
+
+	temp := t.TempDir()
+	tkaPath := filepath.Join(temp, "tka-profile", string(pm.CurrentProfile().ID))
+	os.Mkdir(tkaPath, 0755)
+	chonk, err := tka.ChonkDir(tkaPath)
+	if err != nil {
+		t.Fatal(err)
+	}
+	authority, genesis, err := tka.Create(chonk, tka.State{
+		Keys:               []tka.Key{key},
+		DisablementSecrets: [][]byte{tka.DisablementKDF(disablementSecret)},
+	}, nlPriv)
+	if err != nil {
+		t.Fatalf("tka.Create() failed: %v", err)
+	}
+
+	ts, client := fakeNoiseServer(t, http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		defer r.Body.Close()
+		switch r.URL.Path {
+		case "/machine/tka/bootstrap":
+			body := new(tailcfg.TKABootstrapRequest)
+			if err := json.NewDecoder(r.Body).Decode(body); err != nil {
+				t.Fatal(err)
+			}
+			if body.Version != tailcfg.CurrentCapabilityVersion {
+				t.Errorf("bootstrap CapVer = %v, want %v", body.Version, tailcfg.CurrentCapabilityVersion)
+			}
+			if body.NodeKey != nodePriv.Public() {
+				t.Errorf("nodeKey=%v, want %v", body.NodeKey, nodePriv.Public())
+			}
+
+			w.WriteHeader(200)
+			out := tailcfg.TKABootstrapResponse{
+				GenesisAUM: genesis.Serialize(),
+			}
+			if err := json.NewEncoder(w).Encode(out); err != nil {
+				t.Fatal(err)
+			}
+
+		default:
+			t.Errorf("unhandled endpoint path: %v", r.URL.Path)
+			w.WriteHeader(404)
+		}
+	}))
+	defer ts.Close()
+
+	cc := fakeControlClient(t, client)
+	b := LocalBackend{
+		varRoot: temp,
+		cc:      cc,
+		ccAuto:  cc,
+		logf:    t.Logf,
+		tka: &tkaState{
+			authority: authority,
+			storage:   chonk,
+		},
+		pm:    pm,
+		store: pm.Store(),
+	}
+
+	if err := b.NetworkLockForceLocalDisable(); err != nil {
+		t.Fatalf("NetworkLockForceLocalDisable() failed: %v", err)
+	}
+	if b.tka != nil {
+		t.Fatal("tka was not shut down")
+	}
+	if _, err := os.Stat(b.chonkPathLocked()); err == nil || !os.IsNotExist(err) {
+		t.Errorf("os.Stat(chonkDir) = %v, want ErrNotExist", err)
+	}
+
+	err = b.tkaSyncIfNeeded(&netmap.NetworkMap{
+		TKAEnabled: true,
+		TKAHead:    authority.Head(),
+	}, pm.CurrentPrefs())
+	if err != nil && err.Error() != "bootstrap: TKA with stateID of \"0:0\" is disallowed on this node" {
+		t.Errorf("tkaSyncIfNeededLocked() failed: %v", err)
+	}
+
+	if b.tka != nil {
+		t.Fatal("tka was re-initalized")
+	}
+}

ipn/localapi/localapi.go

@@ -89,6 +89,7 @@
 	"tka/sign":                (*Handler).serveTKASign,
 	"tka/status":              (*Handler).serveTKAStatus,
 	"tka/disable":             (*Handler).serveTKADisable,
+	"tka/force-local-disable": (*Handler).serveTKALocalDisable,
 	"upload-client-metrics":   (*Handler).serveUploadClientMetrics,
 	"watch-ipn-bus":           (*Handler).serveWatchIPNBus,
 	"whois":                   (*Handler).serveWhoIs,
@@ -1243,6 +1244,30 @@ func (h *Handler) serveTKADisable(w http.ResponseWriter, r *http.Request) {
 	w.WriteHeader(200)
 }
 
+func (h *Handler) serveTKALocalDisable(w http.ResponseWriter, r *http.Request) {
+	if !h.PermitWrite {
+		http.Error(w, "network-lock modify access denied", http.StatusForbidden)
+		return
+	}
+	if r.Method != http.MethodPost {
+		http.Error(w, "use POST", http.StatusMethodNotAllowed)
+		return
+	}
+
+	// Require a JSON stanza for the body as an additional CSRF protection.
+	var req struct{}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		http.Error(w, "invalid JSON body", 400)
+		return
+	}
+
+	if err := h.b.NetworkLockForceLocalDisable(); err != nil {
+		http.Error(w, "network-lock local disable failed: "+err.Error(), http.StatusBadRequest)
+		return
+	}
+	w.WriteHeader(200)
+}
+
 func (h *Handler) serveTKALog(w http.ResponseWriter, r *http.Request) {
 	if r.Method != http.MethodGet {
 		http.Error(w, "use GET", http.StatusMethodNotAllowed)

tka/tka.go

@@ -714,3 +714,8 @@ func (a *Authority) Keys() []Key {
 	}
 	return out
 }
+
+// StateIDs returns the stateIDs for this tailnet key authority.
+func (a *Authority) StateIDs() (uint64, uint64) {
+	return a.state.StateID1, a.state.StateID2
+}

types/persist/persist.go

@@ -7,6 +7,7 @@
 
 import (
 	"fmt"
+	"reflect"
 
 	"tailscale.com/tailcfg"
 	"tailscale.com/types/key"
@@ -39,6 +40,12 @@ type Persist struct {
 	UserProfile       tailcfg.UserProfile
 	NetworkLockKey    key.NLPrivate
 	NodeID            tailcfg.StableNodeID
+
+	// DisallowedTKAStateIDs stores the tka.State.StateID values which
+	// this node will not operate network lock on. This is used to
+	// prevent bootstrapping TKA onto a key authority which was forcibly
+	// disabled.
+	DisallowedTKAStateIDs []string
 }
 
 // PublicNodeKey returns the public key for the node key.
@@ -70,7 +77,8 @@ func (p *Persist) Equals(p2 *Persist) bool {
 		p.LoginName == p2.LoginName &&
 		p.UserProfile == p2.UserProfile &&
 		p.NetworkLockKey.Equal(p2.NetworkLockKey) &&
-		p.NodeID == p2.NodeID
+		p.NodeID == p2.NodeID &&
+		reflect.DeepEqual(p.DisallowedTKAStateIDs, p2.DisallowedTKAStateIDs)
 }
 
 func (p *Persist) Pretty() string {

types/persist/persist_clone.go

@@ -20,6 +20,7 @@ func (src *Persist) Clone() *Persist {
 	}
 	dst := new(Persist)
 	*dst = *src
+	dst.DisallowedTKAStateIDs = append(src.DisallowedTKAStateIDs[:0:0], src.DisallowedTKAStateIDs...)
 	return dst
 }
 
@@ -34,4 +35,5 @@ func (src *Persist) Clone() *Persist {
 	UserProfile                     tailcfg.UserProfile
 	NetworkLockKey                  key.NLPrivate
 	NodeID                          tailcfg.StableNodeID
+	DisallowedTKAStateIDs           []string
 }{})

types/persist/persist_test.go

@@ -22,7 +22,7 @@ func fieldsOf(t reflect.Type) (fields []string) {
 }
 
 func TestPersistEqual(t *testing.T) {
-	persistHandles := []string{"LegacyFrontendPrivateMachineKey", "PrivateNodeKey", "OldPrivateNodeKey", "Provider", "LoginName", "UserProfile", "NetworkLockKey", "NodeID"}
+	persistHandles := []string{"LegacyFrontendPrivateMachineKey", "PrivateNodeKey", "OldPrivateNodeKey", "Provider", "LoginName", "UserProfile", "NetworkLockKey", "NodeID", "DisallowedTKAStateIDs"}
 	if have := fieldsOf(reflect.TypeOf(Persist{})); !reflect.DeepEqual(have, persistHandles) {
 		t.Errorf("Persist.Equal check might be out of sync\nfields: %q\nhandled: %q\n",
 			have, persistHandles)
@@ -133,6 +133,11 @@ func TestPersistEqual(t *testing.T) {
 			&Persist{NodeID: "abc"},
 			false,
 		},
+		{
+			&Persist{DisallowedTKAStateIDs: nil},
+			&Persist{DisallowedTKAStateIDs: []string{"0:0"}},
+			false,
+		},
 	}
 	for i, test := range tests {
 		if got := test.a.Equals(test.b); got != test.want {

types/persist/persist_view.go

@@ -13,6 +13,7 @@
 	"tailscale.com/tailcfg"
 	"tailscale.com/types/key"
 	"tailscale.com/types/structs"
+	"tailscale.com/types/views"
 )
 
 //go:generate go run tailscale.com/cmd/cloner  -clonefunc=false -type=Persist
@@ -72,6 +73,9 @@ func (v PersistView) LoginName() string                  { return v.ж.LoginName
 func (v PersistView) UserProfile() tailcfg.UserProfile   { return v.ж.UserProfile }
 func (v PersistView) NetworkLockKey() key.NLPrivate      { return v.ж.NetworkLockKey }
 func (v PersistView) NodeID() tailcfg.StableNodeID       { return v.ж.NodeID }
+func (v PersistView) DisallowedTKAStateIDs() views.Slice[string] {
+	return views.SliceOf(v.ж.DisallowedTKAStateIDs)
+}
 
 // A compilation failure here means this code must be regenerated, with the command at the top of this file.
 var _PersistViewNeedsRegeneration = Persist(struct {
@@ -84,4 +88,5 @@ func (v PersistView) NodeID() tailcfg.StableNodeID       { return v.ж.NodeID }
 	UserProfile                     tailcfg.UserProfile
 	NetworkLockKey                  key.NLPrivate
 	NodeID                          tailcfg.StableNodeID
+	DisallowedTKAStateIDs           []string
 }{})