Add built-in TLS (self-signed autogen, CA pinning) and optional closed registration

coordinator/internal/agent/auth.go

@@ -75,7 +75,7 @@ func (p *WorkerKeyToken) exchangeLocked() error {
 		return err
 	}
 	request.Header.Set("Content-Type", "application/json")
-	client := &http.Client{Timeout: p.timeout}
+	client := &http.Client{Timeout: p.timeout, Transport: tlsTransport(nil)}
 	response, err := client.Do(request)
 	if err != nil {
 		return fmt.Errorf("worker key exchange request failed")

coordinator/internal/agent/check.go

@@ -42,7 +42,7 @@ func checkHTTP(ctx context.Context, url string, timeout time.Duration) (CheckIte
 	if err != nil {
 		return CheckItem{Name: "coordinator", OK: false, Detail: "invalid URL"}, ""
 	}
-	resp, err := http.DefaultClient.Do(req)
+	resp, err := (&http.Client{Timeout: timeout, Transport: tlsTransport(nil)}).Do(req)
 	if err != nil {
 		detail := err.Error()
 		if strings.Contains(detail, "connection refused") {
@@ -148,7 +148,7 @@ func CheckAuth(ctx context.Context, url, token, workerKey, userserviceURL string
 		item.Detail = "no credential configured — will be checked at registration"
 		return item
 	}
-	client := &http.Client{Timeout: 30 * time.Second}
+	client := &http.Client{Timeout: 30 * time.Second, Transport: tlsTransport(nil)}
 	if workerKey != "" && userserviceURL != "" {
 		payload, _ := json.Marshal(map[string]string{"key": workerKey})
 		req, err := http.NewRequestWithContext(ctx, http.MethodPost, strings.TrimRight(userserviceURL, "/")+"/worker-tokens/exchange", strings.NewReader(string(payload)))

coordinator/internal/agent/client.go

@@ -60,10 +60,12 @@ func NewClient(baseURL string, tokens TokenProvider, timeout time.Duration) *Cli
 		timeout: timeout,
 		apiClient: &http.Client{
 			Timeout:       timeout,
+			Transport:     tlsTransport(nil),
 			CheckRedirect: func(*http.Request, []*http.Request) error { return http.ErrUseLastResponse },
 		},
 		dlClient: &http.Client{
-			Timeout: transferTimeout,
+			Timeout:   transferTimeout,
+			Transport: tlsTransport(nil),
 			CheckRedirect: func(req *http.Request, via []*http.Request) error {
 				if len(via) >= 10 {
 					return fmt.Errorf("too many redirects")

coordinator/internal/agent/client_test.go

@@ -1,8 +1,10 @@
 package agent
 
 import (
+	"context"
 	"crypto/sha256"
 	"encoding/json"
+	"encoding/pem"
 	"errors"
 	"fmt"
 	"net/http"
@@ -237,3 +239,58 @@ func TestNewClientTransferTimeoutExceedsAPITimeout(t *testing.T) {
 		t.Errorf("transfer timeout = %v, want 4x the api timeout", short.dlClient.Timeout)
 	}
 }
+
+func TestTLSClientHonoursSkipVerify(t *testing.T) {
+	t.Setenv("SCIMESH_INSECURE_SKIP_VERIFY", "1")
+	server := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		_, _ = w.Write([]byte(`{"status":"ok"}`))
+	}))
+	defer server.Close()
+	client := tlsClient(5 * time.Second)
+	req, _ := http.NewRequestWithContext(context.Background(), http.MethodGet, server.URL+"/health", nil)
+	resp, err := client.Do(req)
+	if err != nil {
+		t.Fatalf("TLS server must be reachable with skip-verify: %v", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode != http.StatusOK {
+		t.Errorf("status = %d", resp.StatusCode)
+	}
+}
+
+func TestTLSClientFailsWithoutTrust(t *testing.T) {
+	t.Setenv("SCIMESH_INSECURE_SKIP_VERIFY", "")
+	t.Setenv("SCIMESH_CA_CERT", "")
+	server := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {}))
+	defer server.Close()
+	client := tlsClient(5 * time.Second)
+	req, _ := http.NewRequestWithContext(context.Background(), http.MethodGet, server.URL+"/health", nil)
+	if resp, err := client.Do(req); err == nil {
+		_ = resp.Body.Close()
+		t.Error("untrusted TLS server must fail verification")
+	}
+}
+
+func TestTLSClientTrustsCAPool(t *testing.T) {
+	server := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		_, _ = w.Write([]byte("ok"))
+	}))
+	defer server.Close()
+	ca := server.Certificate()
+	path := filepath.Join(t.TempDir(), "ca.pem")
+	if err := os.WriteFile(path, pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: ca.Raw}), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("SCIMESH_CA_CERT", path)
+	t.Setenv("SCIMESH_INSECURE_SKIP_VERIFY", "")
+	client := tlsClient(5 * time.Second)
+	req, _ := http.NewRequestWithContext(context.Background(), http.MethodGet, server.URL+"/health", nil)
+	resp, err := client.Do(req)
+	if err != nil {
+		t.Fatalf("CA-trusted TLS server must verify: %v", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode != http.StatusOK {
+		t.Errorf("status = %d", resp.StatusCode)
+	}
+}

coordinator/internal/agent/tls.go

@@ -0,0 +1,69 @@
+package agent
+
+import (
+	"crypto/tls"
+	"crypto/x509"
+	"log/slog"
+	"net"
+	"net/http"
+	"os"
+	"time"
+)
+
+// tlsClient builds an HTTP client whose transport trusts the coordinator's
+// TLS certificate:
+//
+//   - SCIMESH_CA_CERT=/path/to/ca.pem adds a root CA (for self-signed or
+//     private-CA coordinators);
+//   - SCIMESH_INSECURE_SKIP_VERIFY=1 disables verification entirely — only
+//     for trusted LANs where a self-signed certificate was auto-generated.
+//
+// Both settings are deliberately opt-in and noisy: a coordinator without them
+// fails to verify, never silently downgrades.
+func tlsClient(timeout time.Duration) *http.Client {
+	return &http.Client{Timeout: timeout, Transport: tlsTransport(nil)}
+}
+
+// tlsTransport configures a transport honouring the trust environment.
+func tlsTransport(base *http.Transport) *http.Transport {
+	if base == nil {
+		base = &http.Transport{
+			Proxy:               http.ProxyFromEnvironment,
+			DialContext:         (&net.Dialer{Timeout: 30 * time.Second, KeepAlive: 30 * time.Second}).DialContext,
+			MaxIdleConns:        100,
+			IdleConnTimeout:     90 * time.Second,
+			TLSHandshakeTimeout: 10 * time.Second,
+		}
+	}
+	caPath := os.Getenv("SCIMESH_CA_CERT")
+	skip := os.Getenv("SCIMESH_INSECURE_SKIP_VERIFY") == "1"
+	if caPath == "" && !skip {
+		return base
+	}
+	tlsConfig := &tls.Config{MinVersion: tls.VersionTLS12} //nolint:gosec // G402: min TLS 1.2 by default
+	if caPath != "" {
+		//nolint:gosec // G304: SCIMESH_CA_CERT is operator-configured
+		pem, err := os.ReadFile(caPath)
+		if err != nil {
+			slog.Warn("could not read SCIMESH_CA_CERT", "path", caPath, "err", err)
+			return base
+		}
+		pool, err := x509.SystemCertPool()
+		if err != nil {
+			pool = x509.NewCertPool()
+		}
+		if !pool.AppendCertsFromPEM(pem) {
+			slog.Warn("SCIMESH_CA_CERT contained no usable certificates", "path", caPath)
+			return base
+		}
+		tlsConfig.RootCAs = pool
+	}
+	if skip {
+		// G402 is about production code paths; here the operator explicitly
+		// opts into an unverified LAN trust root, so the bypass is intended.
+		tlsConfig.InsecureSkipVerify = true //nolint:gosec // G402: operator opt-in for self-signed LAN certs
+		slog.Warn("SCIMESH_INSECURE_SKIP_VERIFY=1: TLS certificate verification is disabled")
+	}
+	base.TLSClientConfig = tlsConfig
+	return base
+}

coordinator/internal/infra/config.go

@@ -55,6 +55,14 @@ type Config struct {
 	// Directory of the built MkDocs site (site/) served at /ui/docs/. Empty
 	// disables the docs route; the UI shows a hint page instead.
 	DocsDir string
+	// TLSCertFile and TLSKeyFile enable HTTPS when both are set. Self-signed
+	// certificates are fine for a trusted LAN; workers then need
+	// SCIMESH_CA_CERT or SCIMESH_INSECURE_SKIP_VERIFY to connect.
+	TLSCertFile string
+	TLSKeyFile  string
+	// DisableRegistration forbids new UI accounts; the bootstrap admin still
+	// works. Existing accounts and worker keys are unaffected.
+	DisableRegistration bool
 	// Upper bound on an uploaded dataset or artifact body, in bytes.
 	MaxUploadBytes int64
 
@@ -123,6 +131,9 @@ func LoadConfig() (Config, error) {
 		LogFile:              os.Getenv("LOG_FILE"),
 		StorageDir:           getEnv("COORDINATOR_STORAGE_DIR", "./data"),
 		DocsDir:              os.Getenv("SCIMESH_DOCS_DIR"),
+		TLSCertFile:          os.Getenv("SCIMESH_TLS_CERT"),
+		TLSKeyFile:           os.Getenv("SCIMESH_TLS_KEY"),
+		DisableRegistration:  os.Getenv("SCIMESH_DISABLE_REGISTRATION") == "1",
 		MaxUploadBytes:       1 << 30, // 1 GiB
 		DBMaxConns:           10,
 		DBConnectTimeout:     30 * time.Second,

coordinator/internal/infra/server.go

@@ -13,17 +13,25 @@ import (
 const shutdownGrace = 15 * time.Second
 
 // Run serves handler until ctx is cancelled, then drains in-flight requests.
-func RunServer(ctx context.Context, log *slog.Logger, addr string, handler http.Handler) error {
+func RunServer(ctx context.Context, log *slog.Logger, addr string, handler http.Handler, tls ...TLSConfig) error {
 	srv := &http.Server{
 		Addr:              addr,
 		Handler:           handler,
 		ReadHeaderTimeout: 5 * time.Second,
 	}
+	secure := len(tls) > 0 && tls[0].CertFile != "" && tls[0].KeyFile != ""
 
 	// Buffered so this goroutine can exit even when nobody reads the channel
 	// (the ctx.Done branch below) — an unbuffered send would leak it forever.
 	errCh := make(chan error, 1)
 	go func() {
+		if secure {
+			log.Info("coordinator listening (https)", "addr", addr)
+			if err := srv.ListenAndServeTLS(tls[0].CertFile, tls[0].KeyFile); err != nil && !errors.Is(err, http.ErrServerClosed) {
+				errCh <- err
+			}
+			return
+		}
 		log.Info("coordinator listening", "addr", addr)
 		if err := srv.ListenAndServe(); err != nil && !errors.Is(err, http.ErrServerClosed) {
 			errCh <- err
@@ -72,3 +80,9 @@ func RunPeriodic(ctx context.Context, log *slog.Logger, name string, interval ti
 		}
 	}
 }
+
+// TLSConfig enables HTTPS for the coordinator server.
+type TLSConfig struct {
+	CertFile string
+	KeyFile  string
+}

coordinator/internal/transport/http/server.go

@@ -51,6 +51,8 @@ type Server struct {
 	// userserviceURL is the base URL the UI proxies login/registration to. Empty
 	// keeps the static basic-auth UI.
 	userserviceURL string
+	// disableRegistration forbids new accounts; login keeps working.
+	disableRegistration bool
 	// publicCoordinatorURL / publicUserserviceURL are the browser-facing URLs
 	// rendered into the worker-enrollment command. Either may be empty; the
 	// template falls back (own origin / userserviceURL respectively).
@@ -70,6 +72,19 @@ type Server struct {
 func NewServer(uc UseCases, log *slog.Logger, requestTimeout, heartbeatInterval time.Duration,
 	maxUploadBytes int64, jwtSecret, userserviceURL string, m *metrics.Metrics, ready func(context.Context) error,
 	publicURLs ...string) *Server {
+	return NewServerWithOptions(uc, log, requestTimeout, heartbeatInterval, maxUploadBytes, jwtSecret, userserviceURL, m, ready, ServerOptions{}, publicURLs...)
+}
+
+// ServerOptions configures non-positional behaviour of the operator UI.
+type ServerOptions struct {
+	DisableRegistration bool
+}
+
+// NewServerWithOptions is NewServer plus explicit options; the option-less
+// variant exists so existing call sites and tests need no change.
+func NewServerWithOptions(uc UseCases, log *slog.Logger, requestTimeout, heartbeatInterval time.Duration,
+	maxUploadBytes int64, jwtSecret, userserviceURL string, m *metrics.Metrics, ready func(context.Context) error,
+	opts ServerOptions, publicURLs ...string) *Server {
 	if m == nil {
 		m = metrics.New()
 	}
@@ -88,6 +103,7 @@ func NewServer(uc UseCases, log *slog.Logger, requestTimeout, heartbeatInterval
 	}
 	return &Server{
 		uc:                   uc,
+		disableRegistration:  opts.DisableRegistration,
 		log:                  log,
 		requestTimeout:       requestTimeout,
 		heartbeatInterval:    heartbeatInterval,

coordinator/internal/transport/http/ui_auth.go

@@ -108,6 +108,10 @@ func (s *Server) handleUILogin(w http.ResponseWriter, r *http.Request) {
 // user to the login page. The new account is a plain user until an admin
 // promotes or verifies it.
 func (s *Server) handleUIRegister(w http.ResponseWriter, r *http.Request) {
+	if s.disableRegistration {
+		http.Redirect(w, r, "/ui/register?error=registration+disabled", http.StatusSeeOther)
+		return
+	}
 	email, password := r.FormValue("email"), r.FormValue("password")
 
 	status, _, err := s.callUserservice(r.Context(), "/register", email, password)

coordinator/internal/transport/http/ui_auth_internal_test.go

@@ -226,3 +226,18 @@ func TestLoginFormRendersNext(t *testing.T) {
 		t.Error("login form must not render next when absent")
 	}
 }
+
+func TestRegistrationDisabledRejectsNewAccounts(t *testing.T) {
+	stub := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		t.Fatal("userservice must not be called when registration is disabled")
+	}))
+	defer stub.Close()
+	s := newLoginServer(stub)
+	s.disableRegistration = true
+
+	rec := httptest.NewRecorder()
+	s.handleUIRegister(rec, postForm("/ui/register", url.Values{"email": {"a@b.io"}, "password": {"pw"}}))
+	if rec.Code != http.StatusSeeOther || !strings.Contains(rec.Header().Get("Location"), "registration+disabled") {
+		t.Errorf("got %d -> %q, want 303 to the registration-disabled error", rec.Code, rec.Header().Get("Location"))
+	}
+}

coordinator/internal/userservice/transport/http/handlers.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"log/slog"
 	"net/http"
+	"os"
 
 	"github.com/google/uuid"
 
@@ -38,6 +39,12 @@ func (h *Handlers) handleHealth(w http.ResponseWriter, _ *http.Request) {
 // handleRegister creates an account. It returns 201 with the public user view,
 // 409 if the email is taken, or 400 on a malformed body / weak password.
 func (h *Handlers) handleRegister(w http.ResponseWriter, r *http.Request) {
+	// Standalone deployments can close self-service registration while keeping
+	// the bootstrap admin and existing accounts (USERSERVICE_DISABLE_REGISTRATION=1).
+	if os.Getenv("USERSERVICE_DISABLE_REGISTRATION") == "1" {
+		writeJSON(w, http.StatusForbidden, errorResponse{Error: "registration disabled", RequestID: requestIDFrom(r.Context())})
+		return
+	}
 	var req registerRequest
 	if !decodeJSON(w, r, &req) {
 		return

coordinator/internal/userservice/transport/http/server_test.go

@@ -479,3 +479,12 @@ func TestAdminListsUsersAndKeys(t *testing.T) {
 		t.Errorf("admin revoke unknown key: got %d, want 404", rec.Code)
 	}
 }
+
+func TestRegistrationDisabledEnv(t *testing.T) {
+	t.Setenv("USERSERVICE_DISABLE_REGISTRATION", "1")
+	h := newTestServer()
+	rec := do(t, h, http.MethodPost, "/register", "", map[string]string{"email": "blocked@x.io", "password": "pw"})
+	if rec.Code != http.StatusForbidden {
+		t.Errorf("register when disabled: got %d, want 403", rec.Code)
+	}
+}