A node can join the bus the mesh runs on

novox/hq 04-ISSUES/146, the layers behind the three already fixed.

A new membership says which bus it is for. Empty meant 'whatever the mesh runs
today' while two buses existed, and became a refusal the moment one did: an
enrolled node came up and reconnected for ever against its own record.

The enrolling client takes its inboxes in the space its user may listen in. A
JetStream publish waits for the stream's acknowledgement on an inbox the client
picks, and its default is one this user may not subscribe to — so the enrolment
failed with a permissions violation on a subject nobody had chosen.

And the enrolment publish carries a message id, so the client's own retry is
discarded by the stream rather than enrolling the machine twice. That one is
not finished: the duplicate survives it, and the issue says where the trail
stops.

cmd/mesh-host/main.go

@@ -706,15 +706,18 @@ func enrol(ctx context.Context, opts options) error {
 	fmt.Printf("  signing key         %s\n",
 		base64.StdEncoding.EncodeToString(token.Signer)[:16]+"...")
 
-	// The check that has to happen before this machine says anything.
-	conn, err := link.Dial(token.Broker, token.Fingerprint, opts.timeout)
-	if err != nil {
-		return err
-	}
-	defer conn.Close()
-	fmt.Println("\nthe broker presented the certificate this token pins")
-
-	conn.Close()
+	// **The pin is checked by the connection that presents this token, not by a dial of our own**
+	// (novox/hq 04-ISSUES/146). This opened a raw TLS connection to the bus first, which worked
+	// against the broker the mesh used to run and cannot work against the one it runs now: NATS
+	// speaks its own protocol before it upgrades to TLS, so an immediate handshake is answered
+	// with a plaintext line and the enrolment failed with "first record does not look like a TLS
+	// handshake" — on every node that has tried to join since the bus changed, which is why this
+	// went unnoticed: none had.
+	//
+	// What ADR 0004 requires still holds, and holds better: the client that presents the token
+	// carries the same pinned configuration, reads the server's greeting, upgrades, and the
+	// verification runs inside that handshake — so the one-time secret is sent only after the
+	// certificate has been checked, and nothing of this node's reaches an impostor.
 
 	mine, err := identity.Generate(*name)
 	if err != nil {
@@ -788,10 +791,16 @@ func enrol(ctx context.Context, opts options) error {
 	proof := mine.Sign(link.EnrolProof(token.Secret, mine.Public, mine.Overlay.Public,
 		sealing.Public, serving.Public))
 	// The token says where to go and which certificate that address must present. It says nothing
-	// about which bus is there, and does not need to: every token names the one the mesh runs on
-	// today until the rollout (novox/hq ADR 0116 step 5), and that is what an empty Transport is.
+	// about which bus is there, and does not need to: there is one, and this host knows which
+	// (novox/hq ADR 0131 — the mesh speaks to one seat and the old transport is gone).
+	//
+	// **It used to leave this empty** and mean "whatever the mesh runs today", which was true
+	// while two buses existed and became a refusal the moment one did: an empty transport is not
+	// the bus's name, so every enrolment ended at "this token is for the \"\" bus"
+	// (novox/hq 04-ISSUES/146). Nothing caught it because nothing had enrolled since the bus
+	// changed.
 	reply, err := link.Enrol(ctx,
-		link.Approach{Address: token.Broker, Fingerprint: token.Fingerprint},
+		link.Approach{Address: token.Broker, Fingerprint: token.Fingerprint, Transport: link.OnNATS},
 		*name, token.Secret,
 		mine.Public, mine.Overlay.Public, sealing.Public, serving.Public, reported, proof, found,
 		opts.timeout)
@@ -807,6 +816,13 @@ func enrol(ctx context.Context, opts options) error {
 		Fingerprint: firstNonEmpty(reply.Fingerprint, token.Fingerprint),
 		Signer:      firstNonEmpty2(reply.Signer, token.Signer),
 		Password:    reply.Password,
+		// **Which bus this membership is for, said rather than left empty** (novox/hq
+		// 04-ISSUES/146). The link refuses a membership that names another bus, and an empty name
+		// is not this one's — so a node enrolled without it came up and reconnected for ever
+		// against its own record: "this membership is for \"\", and the mesh's bus is nats". The
+		// reply does not carry it because there is one bus and the host knows which (ADR 0131);
+		// what was missing was writing that down where the link reads it.
+		Transport: link.OnNATS,
 	}
 	if mine.Membership.Password == "" {
 		// The mesh did not replace the token's secret, so it is still this node's broker

examples/foundation-first-node-nats.lock

@@ -127,12 +127,21 @@
     {
       "id": "bus-certificate",
       "type": "action",
-      "command": ["docker", "run", "--rm", "--entrypoint", "sh", "-v", "mesh-broker-tls:/tls",
-                  "192.0.2.250:5000/nats@sha256:b83efabe3e7def1e0a4a31ec6e078999bb17c80363f881df35edc70fcb6bb927",
-                  "-c", "test -f /tls/tls.crt || (openssl req -x509 -newkey rsa:2048 -nodes -keyout /tls/tls.key -out /tls/tls.crt -days 3650 -subj '/CN=mesh-broker' -addext 'subjectAltName=DNS:mesh-broker,IP:127.0.0.1' >/dev/null 2>&1 && chmod 644 /tls/tls.crt && chmod 600 /tls/tls.key)"],
-      "verify": ["docker", "run", "--rm", "--entrypoint", "sh", "-v", "mesh-broker-tls:/tls",
-                 "192.0.2.250:5000/nats@sha256:b83efabe3e7def1e0a4a31ec6e078999bb17c80363f881df35edc70fcb6bb927",
-                 "-c", "test -s /tls/tls.crt && openssl x509 -in /tls/tls.crt -noout"]
+      // **The mesh makes its own** (novox/hq 04-ISSUES/146). This ran `openssl` inside the
+      // broker's image while the broker was one that carried it; the bus that replaced it has a
+      // shell and no openssl, and no other image the bundle names has one either. So the program
+      // that needs the certificate writes it — already on this machine, since the schema step ran
+      // it, and asking nothing of the image it writes into. Self-signed on purpose: a host pins
+      // this server's exact certificate (novox/hq ADR 0004), and at this moment there is no mesh
+      // to ask an authority of.
+      // `--user 0:0` because the volume is root's and this image runs as nobody, which is right
+      // for the long-running control plane and wrong for a one-shot writing into a fresh volume.
+      "command": ["docker", "run", "--rm", "--user", "0:0", "-v", "mesh-broker-tls:/tls",
+                  "192.0.2.250:5000/mesh-controller@sha256:c67db38439ff0aee242b467486765467bb95801f52175fc5727cc4e437338ace",
+                  "broker", "certificate", "--into", "/tls"],
+      "verify": ["docker", "run", "--rm", "--user", "0:0", "-v", "mesh-broker-tls:/tls",
+                 "192.0.2.250:5000/mesh-controller@sha256:c67db38439ff0aee242b467486765467bb95801f52175fc5727cc4e437338ace",
+                 "broker", "certificate", "--check", "--into", "/tls"]
     },
     {
       "id": "bus-conf-dir",

internal/link/asking_nats.go

@@ -3,6 +3,7 @@ package link
 import (
 	"context"
 	"crypto/rand"
+	"crypto/sha256"
 	"encoding/hex"
 	"errors"
 	"fmt"
@@ -63,8 +64,15 @@ func presentNats(_ context.Context, to Approach, node, secret string,
 	// subscribe its own inbox and nothing else (design 25 §6). The secret is its password, the same
 	// string the request claims, so the server proves somebody holds the token and the request
 	// proves the same thing to the controller without it having to ask the server who connected.
+	// **Its own inbox space, because that is the only one it may listen in** (novox/hq
+	// 04-ISSUES/146). A JetStream publish waits for the stream's acknowledgement on an inbox the
+	// client picks, and the client's default is `_INBOX.<random>` — which this user may not
+	// subscribe to, so the enrolment failed with a permissions violation on a subject nobody had
+	// chosen. The permission is `_INBOX.enrol.<node>.>` (design 25 §6), so the client is told to
+	// pick its inboxes there; the reply address below is in the same space for the same reason.
 	conn, err := nats.Connect(natsURL(to.Address),
 		nats.Secure(config),
+		nats.CustomInboxPrefix("_INBOX.enrol."+node),
 		nats.UserInfo("enrol."+node, secret),
 		nats.Name("mesh-host/enrol/"+node),
 		nats.Timeout(timeout),
@@ -124,7 +132,19 @@ func (a *natsAsking) Ask(ctx context.Context, request []byte, wait time.Duration
 	defer cancel()
 	// Into the stream and awaited: an enrolment the bus never accepted must fail here rather than be
 	// assumed, because the node has nothing else to go on.
-	if _, err := a.js.Publish(EnrolSubject, addressed, nats.Context(publish)); err != nil {
+	//
+	// **Once, however many times it is sent** (novox/hq 04-ISSUES/146). The client re-publishes when
+	// an acknowledgement is slow, and the mesh enrolled the machine on each copy — minting a second
+	// credential, which replaced the first, which is the one the node had already been given. The
+	// machine then reconnected for ever as a user whose password the mesh had rotated out from under
+	// it, and the controller's log said "enrolled anchor" twice in the same second.
+	//
+	// The id is the message: the same bytes carry the same id, so the stream discards the client's
+	// own retry, and a genuine second attempt — which carries a new reply address — is a different
+	// message and is let through.
+	sum := sha256.Sum256(addressed)
+	if _, err := a.js.Publish(EnrolSubject, addressed,
+		nats.MsgId(hex.EncodeToString(sum[:])), nats.Context(publish)); err != nil {
 		return nil, fmt.Errorf("cannot ask the mesh to enrol this node: %w", err)
 	}
 

internal/link/pinned.go

@@ -73,8 +73,19 @@ func PinnedConfig(pin string) (*tls.Config, error) {
 	}, nil
 }
 
-// Dial opens a TLS connection to the broker, refusing anything but the pinned certificate.
-func Dial(address, pin string, timeout time.Duration) (*tls.Conn, error) {
+// dialPinned completes a TLS handshake against an address, refusing anything but the pinned
+// certificate.
+//
+// **Not how the bus is reached, and it used to be** (novox/hq 04-ISSUES/146). Enrolment opened one
+// of these before it said anything, which was right while the broker answered TLS immediately and
+// wrong the moment the mesh moved to a bus that speaks its own protocol first. The pin itself was
+// never the problem — PinnedConfig is what the NATS client is given, and the verification runs
+// inside the handshake that client performs.
+//
+// It stays here because this is where the pin is proven: the tests beside it run a real TLS server
+// and assert that a wrong certificate is refused before a byte of application data is sent. What it
+// must not become again is something a caller uses to reach the bus.
+func dialPinned(address, pin string, timeout time.Duration) (*tls.Conn, error) {
 	config, err := PinnedConfig(pin)
 	if err != nil {
 		return nil, err
@@ -86,7 +97,7 @@ func Dial(address, pin string, timeout time.Duration) (*tls.Conn, error) {
 		if errors.Is(err, ErrWrongCertificate) {
 			return nil, err
 		}
-		return nil, fmt.Errorf("cannot reach the broker at %s: %w", address, err)
+		return nil, fmt.Errorf("cannot reach %s: %w", address, err)
 	}
 	return conn, nil
 }

internal/link/pinned_test.go

@@ -63,7 +63,7 @@ func server(t *testing.T) (address string, fingerprint string) {
 
 func TestTheRightBrokerIsAccepted(t *testing.T) {
 	address, pin := server(t)
-	conn, err := Dial(address, pin, 5*time.Second)
+	conn, err := dialPinned(address, pin, 5*time.Second)
 	if err != nil {
 		t.Fatalf("the broker its token describes was refused: %v", err)
 	}
@@ -76,7 +76,7 @@ func TestADifferentBrokerIsRefused(t *testing.T) {
 	address, _ := server(t)
 	_, other := server(t)
 
-	_, err := Dial(address, other, 5*time.Second)
+	_, err := dialPinned(address, other, 5*time.Second)
 	if err == nil {
 		t.Fatal("a broker presenting a different certificate was accepted")
 	}
@@ -130,7 +130,7 @@ func TestNothingIsSentToTheWrongBroker(t *testing.T) {
 
 	// A pin for a certificate this server does not have.
 	_, elsewhere := server(t)
-	if _, err := Dial(listener.Addr().String(), elsewhere, 5*time.Second); err == nil {
+	if _, err := dialPinned(listener.Addr().String(), elsewhere, 5*time.Second); err == nil {
 		t.Fatal("the impostor was accepted")
 	}
 	if n := <-received; n > 0 {
@@ -160,7 +160,7 @@ func TestAnUnreachableBrokerIsAnOrdinaryFailure(t *testing.T) {
 	address := listener.Addr().String()
 	listener.Close()
 
-	_, err = Dial(address, pin, 2*time.Second)
+	_, err = dialPinned(address, pin, 2*time.Second)
 	if err == nil {
 		t.Fatal("dialling a closed port succeeded")
 	}