Merge pull request 'A changed jail filter restarts fail2ban' (#231) from jschoubben/jail-filter-restart into main

fail2ban restarts when the composed jail file changes, and each filter is
a file of its own, so a module that changed only its failregex left the
running jail on the old pattern. The jail file now names each filter's
digest.

cmd/mesh-controller/foundation_scope_test.go

@@ -1,33 +0,0 @@
-package main
-
-// The broker opening belongs only on the node that listens on it (novox/hq: it leaked onto
-// every enrolled node's declaration, opening a from-anywhere hole for a port nothing there
-// serves). foundationPortsFor is the scope.
-
-import (
-	"testing"
-
-	"github.com/novox/mesh-controller/internal/catalogue"
-)
-
-func TestTheBrokerHostGetsTheFoundationOpening(t *testing.T) {
-	broker := catalogue.Manifest{Module: "lavinmq", Listens: []catalogue.Listening{
-		{Port: 5671, Protocol: "tcp", From: "mesh"},
-		{Port: 5672, Protocol: "tcp", From: "mesh"},
-	}}
-	got := foundationPortsFor(5671, []catalogue.Manifest{broker})
-	if len(got) != 1 || got[0] != 5671 {
-		t.Fatalf("the node that listens on the broker port keeps it; got %v", got)
-	}
-}
-
-func TestANodeThatOnlyDialsTheBrokerGetsNoOpening(t *testing.T) {
-	// ace's set: things that reach the broker as a client, none listening on 5671.
-	ace := []catalogue.Manifest{
-		{Module: "plex", Listens: []catalogue.Listening{{Port: 32400, Protocol: "tcp", From: "anywhere"}}},
-		{Module: "postgres", Listens: []catalogue.Listening{{Port: 5432, Protocol: "tcp", From: "mesh"}}},
-	}
-	if got := foundationPortsFor(5671, ace); got != nil {
-		t.Fatalf("a node that only dials out opens nothing for the broker; got %v", got)
-	}
-}

cmd/mesh-controller/plan.go

@@ -16,8 +16,6 @@ import (
 	"github.com/novox/mesh-controller/internal/inventory"
 	"github.com/novox/mesh-controller/internal/licences"
 	"github.com/novox/mesh-controller/internal/overlay"
-	"net"
-	"strconv"
 )
 
 // working out what one machine should be.
@@ -648,25 +646,12 @@ func renderingFor(ctx context.Context, open *stores, node string,
 		names[name] = at
 	}
 
-	// The ports the mesh itself needs open, which no module declares. Read from the broker this
-	// control plane was told about rather than written down twice: the address a node is handed in
-	// its token and the port its machine must accept on are the same fact.
-	//
-	// **Only on the node that listens on it** (novox/hq issue: the broker opening leaked onto
-	// every node). The opening exists to WIDEN the broker's port to from-anywhere — a machine
-	// enrolling is not on the mesh yet, so the broker's own `from: mesh` listen would refuse its
-	// first dial. That widening belongs on the broker's host and nowhere else: a node that only
-	// dials out needs no incoming rule, and an opening for a port nothing here listens on is a
-	// from-anywhere hole for a dead port. So the foundation port is kept only when a module
-	// resolved onto THIS node actually listens on it.
+	// **The bus is never public** (novox/hq ADR 0169). It was a foundation port — widened from the
+	// broker's own `from: mesh` to from-anywhere on the broker's host, so a machine could enrol
+	// before it had an address on the private network. A machine joins through the tunnel now, and
+	// every link to the bus crosses it, so its reach is what the `nats` module declares: the mesh.
+	// Nothing the mesh itself needs is opened beyond what a module declares.
 	var foundation []int
-	if b, err := broker.FromEnvironment(); err == nil {
-		if _, port, err := net.SplitHostPort(b.Address); err == nil {
-			if n, err := strconv.Atoi(port); err == nil {
-				foundation = foundationPortsFor(n, plan.Modules)
-			}
-		}
-	}
 
 	// And, for a module that keeps them, every operator-sealed secret in the mesh — the vault's
 	// copy, outside the store (novox/hq ADR 0085, amended). Read only; nothing here mints. The
@@ -1380,23 +1365,6 @@ func composeBusUsers(ctx context.Context, inv *inventory.Inventory,
 	return broker.ComposeAccounts(filled)
 }
 
-// foundationPortsFor is the broker port, kept only when a module resolved onto this node listens
-// on it (novox/hq issue: the broker opening leaked onto every node). The foundation opening
-// exists to WIDEN the broker's `from: mesh` port to from-anywhere, because a machine enrolling is
-// not on the mesh yet and its first dial would be refused. That widening belongs on the broker's
-// host alone: a node that only dials out needs no incoming rule, and an opening for a port
-// nothing here listens on is a from-anywhere hole for a dead port.
-func foundationPortsFor(brokerPort int, modules []catalogue.Manifest) []int {
-	for _, m := range modules {
-		for _, l := range m.Listens {
-			if l.Port == brokerPort {
-				return []int{brokerPort}
-			}
-		}
-	}
-	return nil
-}
-
 // providerModuleOf is which module answers a need on the providing node: the one in this node's
 // own set when the provider is here, else the one the catalogue says offers it.
 func providerModuleOf(resolved catalogue.Resolution, open *stores, ctx context.Context, n catalogue.Needed) string {

internal/catalogue/build.go

@@ -84,8 +84,13 @@ func (m Manifest) Resolve(built []Built) (Manifest, error) {
 			if a.Loads == nil && len(m.Tools) > 0 {
 				loads = append([]string(nil), a.Entrypoints...)
 			}
+			// **Kept, never routed** (ADR 0155): the builder publishes to the store at the address
+			// it reached it by, and a manifest carrying that address names an installation —
+			// registration refused node-tools for exactly this on 2026-10-02. The build record
+			// already keeps the store-relative form; the resolved manifest keeps the same, and
+			// composition routes it through the store a machine reaches (Routed).
 			out.Bundles = append(out.Bundles, Bundle{
-				Name: a.Name, Source: made.Reference, Digest: made.Digest,
+				Name: a.Name, Source: Recorded(made.Reference), Digest: made.Digest,
 				Language: a.Language, Entrypoints: append([]string(nil), a.Entrypoints...),
 				Loads: loads,
 			})
@@ -135,7 +140,8 @@ func (m Manifest) Resolve(built []Built) (Manifest, error) {
 			// The same on the wire: both are bytes fetched by digest and unpacked. They differ in
 			// how they were made — one packed as it stood, the other compiled first — and a
 			// machine has no reason to care which.
-			filled["source"] = artifact.Reference
+			// Kept, not routed, for the reason the bundles above are (ADR 0155).
+			filled["source"] = Recorded(artifact.Reference)
 			filled["digest"] = artifact.Digest
 			// **And `${version}`, so a resource can name a place that is this build's alone**
 			// (novox/hq ADR 0141, 04-ISSUES/142). A component is unpacked into a directory named

internal/catalogue/jails_into.go

@@ -1,6 +1,8 @@
 package catalogue
 
 import (
+	"crypto/sha256"
+	"encoding/hex"
 	"fmt"
 	"sort"
 	"strings"
@@ -43,8 +45,16 @@ func jailsInto(modules []Manifest, j *Jailing) []map[string]any {
 
 	out := make([]map[string]any, 0, len(jails)+1)
 	for _, d := range jails {
-		fmt.Fprintf(&composed, "\n# from %s\n[%s]\nenabled = true\nfilter = %s\n%s\n",
-			d.module, d.jail.Name, d.jail.Name, strings.TrimRight(d.jail.Jail, "\n"))
+		// **The filter's digest rides in the jail file.** fail2ban is restarted when this file
+		// changes, and the filter is a file of its own: a module that changed only what a failure
+		// looks like rewrote the filter on disk and left the running jail on the old pattern, with
+		// nothing said (novox/hq issue 191's rollout found it on gitea's sshd). Naming the filter's
+		// digest here makes a changed pattern a changed jail file, so the restart the service
+		// already takes on it covers the filter too.
+		sum := sha256.Sum256([]byte(d.jail.Failregex))
+		fmt.Fprintf(&composed, "\n# from %s, filter %s\n[%s]\nenabled = true\nfilter = %s\n%s\n",
+			d.module, hex.EncodeToString(sum[:])[:12], d.jail.Name, d.jail.Name,
+			strings.TrimRight(d.jail.Jail, "\n"))
 		// The filter is a file of its own, named as the jail's filter= references it.
 		out = append(out, map[string]any{
 			"id":   "filter-" + d.jail.Name,

internal/catalogue/jails_into_test.go

@@ -44,3 +44,29 @@ func TestTheComposedJailFileIsWrittenEvenWhenEmpty(t *testing.T) {
 		t.Fatalf("the empty composed jail file was not written alone: %v", files)
 	}
 }
+
+// A changed pattern restarts fail2ban (novox/hq issue 191's rollout): the service restarts when the
+// composed jail file changes, and the filter is a file of its own, so the jail file names the
+// filter's digest. Changing only the failregex must change the jail file; the same pattern must not.
+func TestAChangedFilterChangesTheJailFile(t *testing.T) {
+	jailFile := func(failregex string) string {
+		modules := []Manifest{
+			{Module: "fail2ban", Jailing: &Jailing{Into: "/etc/fail2ban/jail.d/mesh.conf", FilterInto: "/etc/fail2ban/filter.d"}},
+			{Module: "gitea", Jails: []Jail{{Name: "gitea", Failregex: failregex, Jail: "port = 222"}}},
+		}
+		for _, f := range jailsInto(modules, modules[0].Jailing) {
+			if f["id"] == ComposedJailsID() {
+				return f["content"].(string)
+			}
+		}
+		t.Fatal("no composed jail file")
+		return ""
+	}
+	before := jailFile("web login failed from <HOST>")
+	if again := jailFile("web login failed from <HOST>"); again != before {
+		t.Errorf("the same pattern composed a different jail file, which would restart fail2ban for nothing")
+	}
+	if after := jailFile("web login failed from <HOST>\n            Invalid user .* from <HOST>"); after == before {
+		t.Errorf("a changed pattern left the jail file as it was, so fail2ban keeps the old filter:\n%s", after)
+	}
+}

internal/catalogue/resolve_keeps_references_test.go

@@ -0,0 +1,42 @@
+package catalogue
+
+import (
+	"strings"
+	"testing"
+)
+
+// A resolved manifest keeps a built artifact's reference in the store-relative form, never the
+// address the builder reached the store by (novox/hq ADR 0155): on 2026-10-02 the first bundle
+// resolved on the mesh carried the store's host in bundles[0].source and registration refused it
+// as naming an installation. Archive resources are the same kind of reference and get the same.
+func TestAResolvedReferenceIsKeptNotRouted(t *testing.T) {
+	m, err := ParseManifest([]byte(`{
+	  "module": "sample", "version": "1",
+	  "tools": ["one"],
+	  "build": {"artifacts": [
+	    {"name": "code", "kind": "bundle", "language": "typescript", "entrypoints": ["tools/index.js"]},
+	    {"name": "files", "kind": "archive", "from": "files"}
+	  ]},
+	  "resources": [{"id": "packed", "type": "archive", "path": "/opt/sample", "artifact": "files"}]
+	}`))
+	if err != nil {
+		t.Fatal(err)
+	}
+	digest := "sha256:" + strings.Repeat("ab", 32)
+	resolved, err := m.Resolve([]Built{
+		{Name: "code", Kind: ArtifactBundle, Reference: "http://store.example:5100/v2/sample/code/blobs/" + digest, Digest: digest},
+		{Name: "files", Kind: ArtifactArchive, Reference: "http://store.example:5100/v2/sample/files/blobs/" + digest, Digest: digest},
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got, want := resolved.Bundles[0].Source, ArtifactStoreScheme+"sample/code/blobs/"+digest; got != want {
+		t.Errorf("bundle source %q, want the kept form %q", got, want)
+	}
+	if got, want := resolved.Resources[0]["source"], ArtifactStoreScheme+"sample/files/blobs/"+digest; got != want {
+		t.Errorf("archive source %q, want the kept form %q", got, want)
+	}
+	if problems := InstallationProblems(resolved); len(problems) != 0 {
+		t.Errorf("a resolved manifest names an installation: %v", problems)
+	}
+}

internal/catalogue/the_bus_is_never_public_test.go

@@ -0,0 +1,23 @@
+package catalogue
+
+import (
+	"regexp"
+	"testing"
+)
+
+// The bus is never public (novox/hq ADR 0169). Its port is what the bus module declares, the mesh,
+// and the control plane adds no opening of its own: a machine joins through the tunnel, so the
+// broker's host is filtered like any other. Before this, the broker port was a foundation port and
+// rendered from anywhere beside its from-the-mesh rule.
+func TestTheBusPortIsReachedFromTheMeshAlone(t *testing.T) {
+	rules := []Rule{{Port: 4222, Protocol: "tcp", From: FromMesh, Because: []string{"nats"},
+		Why: []string{"the mesh bus"}}}
+	out := AsNftables(rules, []string{"10.10.0.1", "10.10.0.2"}, true, nil, []string{"eth0"}, "mesh0")
+
+	if !regexp.MustCompile(`ip saddr \{ 10\.10\.0\.1, 10\.10\.0\.2 \} tcp dport 4222 accept`).MatchString(out) {
+		t.Fatalf("the bus is not reachable from the mesh:\n%s", out)
+	}
+	if regexp.MustCompile(`(?m)^\s*tcp dport 4222 accept`).MatchString(out) {
+		t.Fatalf("the bus is reachable from anywhere:\n%s", out)
+	}
+}