postfix: no reload for a map that did not change; DNS re-checks limited (audit W6/M9)

Security audit 2026-10-17, findings W6/M9 (low), stage 4. Every
application, upstream, recipient or report-address save reloaded Postfix
through supervisord, and POST .../dns-recheck bypassed the five-minute
cache for up to ~14 resolver queries - both at whatever rate an
authenticated user pressed.

- writeChanged: each generated map is written only when its content
  differs from the file on disk, and Postfix is reloaded only when one was
  (sender map, inbound maps, DMARC maps). Health's Reload configuration
  ends with an explicit reload (app.Service.ReloadPostfix, SenderMaps
  gains Reload), so it still re-reads whatever is on disk. No debounce:
  with unchanged saves gone, a burst is real changes Postfix must see.
- auth.AllowDNSRecheck: twenty forced re-checks in 15 minutes per user;
  past that the outbound, inbound and settings re-check show the cached
  result without the "re-checked" flash, and the log says why.

Tests: TestARebuildThatChangesNothingDoesNotReload,
TestDNSRechecksArePerUser; the fakes gain Reload. Docs: guide § Overview
and Health; CHANGELOG; plan.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -32,6 +32,12 @@ Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); version
   a monitor next to it; the internet gets `404`.
 - **Every container has a process ceiling** (audit I3): `pids_limit` 1024
   for SelfPost, 256 for the proxies and rspamd, 64 for certbot.
+- **A save that changes no map no longer reloads Postfix, and DNS
+  re-checks are limited** (audit W6/M9): every application, upstream,
+  recipient or report-address save reloaded Postfix, and *Re-check DNS*
+  bypassed the cache without limit. Now only a changed file is written and
+  reloaded (Health's *Reload configuration* always re-reads), and a user
+  re-checks twenty times in 15 minutes.
 - **The TLS private key no longer has to be readable by Postfix** (audit
   I1): the guide gave it to group 10998 for the panel's backup, and
   `postfix` is in that group, so every Postfix daemon - an smtpd talking to

docs/guide.md

@@ -516,7 +516,8 @@ readings, numbered in the order of what stops mail first:
 - **Hostname and reverse DNS** — compares `SELFPOST_HOSTNAME` against the PTR
   record the internet publishes for this server's IP (forward-confirmed
   reverse DNS), see [Server-level DNS](#server-level-dns-ptrrdns); use
-  *Re-check DNS* after changing DNS. *Greeting* is the `220` line Postfix
+  *Re-check DNS* after changing DNS (twenty re-checks in 15 minutes per user;
+past that the page shows the last result). *Greeting* is the `220` line Postfix
   opens every connection with; it must name the hostname, and another name
   there is a warning.
 - **Notification mail** — whether the DNS records the panel's own mail is
@@ -538,7 +539,8 @@ readings, numbered in the order of what stops mail first:
   selector table (`/data/dkim/selectors`), the Postfix sender map, the
   inbound relay maps and the filter lists (with `INBOUND_RELAY_ENABLE=true`),
   the DMARC report maps (with `DMARC_REPORTS_ENABLE=true`) - removes the keys
-  of deleted domains, and has Postfix re-read them; the signer needs no
+  of deleted domains, and has Postfix re-read them (a save elsewhere that
+  changes no file asks no reload; this button always does); the signer needs no
   reload, it reads the table for every message. This is
   what the first start after a restore does (see [Restore](#restore)). Use it
   if the daemons drifted from what the panel shows after manual edits under

docs/security-audit-2026-10-17.md

@@ -804,9 +804,14 @@ the argument.
   SelfPost (Postfix's default_process_limit of 100 per service, three smtpd
   services), 256 for the proxies and rspamd, 64 for certbot; e2e reads
   `pids.max` back (checkPidsLimit).
-- [ ] **W6/M9.** Coalesce Postfix reloads (reload only when the rendered map
+- [x] **W6/M9.** Coalesce Postfix reloads (reload only when the rendered map
   changed; a 2 s debounce for bursts); a per-user limiter on `dns-recheck`.
-  — **Opus**, 1.5 h.
+  — **Opus**, 1.5 h. *Done 2026-10-17:* a rebuild that changes no file
+  writes nothing and asks no reload (`writeChanged`), Health's button
+  re-reads in any case; no debounce - with unchanged saves gone, a burst is
+  real changes, each of which Postfix must see; twenty re-checks in 15
+  minutes per user, then the cached result. TestARebuildThatChangesNothingDoesNotReload,
+  TestDNSRechecksArePerUser.
 - [ ] **W4/W5/W8.** Key size 1024-4096 on import; ≤ 500 applications per
   export; an import named as the hostname refused with a sentence. —
   **Sonnet**, 1.5 h.

internal/app/service.go

@@ -17,6 +17,15 @@ import (
 // *postfix.Postfix satisfies it; tests substitute a fake.
 type SenderMaps interface {
 	RebuildSenderLoginMaps(bindings []postfix.Binding) error
+	// Reload has Postfix re-read its configuration whatever changed.
+	Reload() error
+}
+
+// ReloadPostfix has Postfix re-read its configuration. A rebuild that
+// changes no file asks no reload; Health's Reload configuration ends with
+// this one, so that it re-reads whatever is on disk.
+func (s *Service) ReloadPostfix() error {
+	return s.pf.Reload()
 }
 
 // Service coordinates application state across SQLite, sasldb2 and the Postfix

internal/app/service_test.go

@@ -20,6 +20,8 @@ type fakeMaps struct {
 	failNext bool
 }
 
+func (f *fakeMaps) Reload() error { return nil }
+
 func (f *fakeMaps) RebuildSenderLoginMaps(b []postfix.Binding) error {
 	f.calls++
 	if f.failNext {
@@ -514,6 +516,8 @@ type overlapMaps struct {
 	inside, most atomic.Int32
 }
 
+func (o *overlapMaps) Reload() error { return nil }
+
 func (o *overlapMaps) RebuildSenderLoginMaps([]postfix.Binding) error {
 	n := o.inside.Add(1)
 	for {

internal/postfix/dmarc.go

@@ -29,13 +29,8 @@ func (p *Postfix) RebuildDMARCMaps(cfg DMARCMapsConfig) error {
 		return err
 	}
 	rc, tr, rd := p.dmarcMapPaths()
-	if err := writeFileAtomic(rc, recip, 0o640); err != nil {
+	changed, err := writeChanged(mapFile{rc, recip}, mapFile{tr, transport}, mapFile{rd, relay})
-		return err
+	if err != nil || !changed {
-	}
-	if err := writeFileAtomic(tr, transport, 0o640); err != nil {
-		return err
-	}
-	if err := writeFileAtomic(rd, relay, 0o640); err != nil {
 		return err
 	}
 	return p.reload()

internal/postfix/inbound.go

@@ -40,16 +40,8 @@ func (p *Postfix) RebuildInboundMaps(routes []InboundRoute) error {
 		return err
 	}
 	rd, tr, rc, tl := p.inboundMapPaths()
-	if err := writeFileAtomic(rd, relay, 0o640); err != nil {
+	changed, err := writeChanged(mapFile{rd, relay}, mapFile{tr, transport}, mapFile{rc, recipients}, mapFile{tl, tlsPolicy})
-		return err
+	if err != nil || !changed {
-	}
-	if err := writeFileAtomic(tr, transport, 0o640); err != nil {
-		return err
-	}
-	if err := writeFileAtomic(rc, recipients, 0o640); err != nil {
-		return err
-	}
-	if err := writeFileAtomic(tl, tlsPolicy, 0o640); err != nil {
 		return err
 	}
 	return p.reload()

internal/postfix/postfix.go

@@ -69,7 +69,8 @@ func (p *Postfix) RebuildSenderLoginMaps(bindings []Binding) error {
 	if err != nil {
 		return err
 	}
-	if err := writeFileAtomic(p.senderLoginMapsPath, content, 0o640); err != nil {
+	changed, err := writeChanged(mapFile{p.senderLoginMapsPath, content})
+	if err != nil || !changed {
 		return err
 	}
 	return p.reload()

internal/postfix/postfix_test.go

@@ -132,3 +132,33 @@ func TestRebuildRejectsUnsafeWithoutWriting(t *testing.T) {
 		t.Errorf("reload called %d times, want 1 (no reload on failure)", *reloads)
 	}
 }
+
+// A rebuild that changes no file writes nothing and asks no reload; a change
+// in any one file reloads once (security audit 2026-10-17, W6).
+func TestARebuildThatChangesNothingDoesNotReload(t *testing.T) {
+	p, reloads := newTestPostfix(t)
+	b := []Binding{{"@example.com", "app1"}}
+	for i := 0; i < 3; i++ {
+		if err := p.RebuildSenderLoginMaps(b); err != nil {
+			t.Fatal(err)
+		}
+	}
+	if *reloads != 1 {
+		t.Errorf("three identical sender-map rebuilds reloaded %d times, want 1", *reloads)
+	}
+	routes := []InboundRoute{{Domain: "in.example.com", Host: "mx.example.net", Port: 25, TLSMode: "may", RecipientMode: "any"}}
+	for i := 0; i < 2; i++ {
+		if err := p.RebuildInboundMaps(routes); err != nil {
+			t.Fatal(err)
+		}
+	}
+	if *reloads != 2 {
+		t.Errorf("two identical inbound rebuilds after it: %d reloads in all, want 2", *reloads)
+	}
+	if err := p.RebuildSenderLoginMaps(append(b, Binding{"@other.example", "app2"})); err != nil {
+		t.Fatal(err)
+	}
+	if *reloads != 3 {
+		t.Errorf("a changed map: %d reloads in all, want 3", *reloads)
+	}
+}

internal/postfix/write.go

@@ -1,6 +1,7 @@
 package postfix
 
 import (
+	"bytes"
 	"os"
 
 	"github.com/mixeme/selfpost/internal/atomicfile"
@@ -12,3 +13,28 @@ import (
 func writeFileAtomic(path string, data []byte, perm os.FileMode) error {
 	return atomicfile.Write(path, data, perm)
 }
+
+// mapFile is one generated table and what it should hold.
+type mapFile struct {
+	path string
+	data []byte
+}
+
+// writeChanged writes each file whose content differs from what is on disk
+// and says whether any did. A save that changes nothing Postfix reads - a
+// domain user saving a form as it was, or changing a field no map holds -
+// then writes nothing and asks no reload: every save used to reload Postfix,
+// at whatever rate a user cared to press (security audit 2026-10-17, W6).
+func writeChanged(files ...mapFile) (bool, error) {
+	changed := false
+	for _, f := range files {
+		if old, err := os.ReadFile(f.path); err == nil && bytes.Equal(old, f.data) {
+			continue
+		}
+		if err := writeFileAtomic(f.path, f.data, 0o640); err != nil {
+			return changed, err
+		}
+		changed = true
+	}
+	return changed, nil
+}

internal/web/auth/auth.go

@@ -54,6 +54,8 @@ type Module struct {
 	// ignored (links.go, sendReset).
 	resetDayLimiter    *rateLimiter
 	resetNoticeLimiter *rateLimiter
+	// recheckLimiter counts the DNS re-checks a user forces (AllowDNSRecheck).
+	recheckLimiter *rateLimiter
 
 	proxyWatch proxyWatch
 	refusals   *refusalLog
@@ -98,6 +100,7 @@ func New(st *store.Store, cfg Config, v *view.Engine, setupTokenPath string) *Mo
 		// silence for the rest of the day (security audit 2026-10-17, A3).
 		resetDayLimiter:    newRateLimiter(resetsPerDay, 24*time.Hour),
 		resetNoticeLimiter: newRateLimiter(1, 24*time.Hour),
+		recheckLimiter:     newRateLimiter(20, 15*time.Minute),
 		trustedProxies:     cfg.TrustedProxyCIDRs,
 	}
 	m.setupLimiter.startSweeper()
@@ -112,6 +115,7 @@ func New(st *store.Store, cfg Config, v *view.Engine, setupTokenPath string) *Mo
 	m.mailLimiter.startSweeper()
 	m.resetDayLimiter.startSweeper()
 	m.resetNoticeLimiter.startSweeper()
+	m.recheckLimiter.startSweeper()
 	// Made now rather than at the first unknown name, which would otherwise
 	// pay for two bcrypt runs (review of audit stage 4).
 	go noSuchUserHash()
@@ -131,6 +135,14 @@ func (m *Module) AllowPasswordChange(userID int64) bool {
 	return m.passwordLimiter.Allow("user:" + strconv.FormatInt(userID, 10))
 }
 
+// AllowDNSRecheck is whether a user may make the DNS checks again, past the
+// five-minute cache: each re-check is up to fourteen resolver queries, and a
+// domain user could press it without end (security audit 2026-10-17, M9).
+// Twenty in fifteen minutes per user.
+func (m *Module) AllowDNSRecheck(userID int64) bool {
+	return m.recheckLimiter.Allow("user:" + strconv.FormatInt(userID, 10))
+}
+
 // PasswordCheckPassed gives back the attempt AllowPasswordChange took for a
 // check that found the right password: only wrong guesses count, so an
 // administrator who adds ten users in a quarter of an hour is not locked out

internal/web/auth/limitkeys_test.go

@@ -114,3 +114,21 @@ func TestProxyWatchAndRefusalsAreLogged(t *testing.T) {
 		t.Errorf("refusals logged %d times (want 5), or the name unescaped:\n%s", n, got)
 	}
 }
+
+// A user forces the DNS checks twenty times in fifteen minutes, then sees
+// the cached result (security audit 2026-10-17, M9); another user is not
+// affected.
+func TestDNSRechecksArePerUser(t *testing.T) {
+	m := testModule(t, false)
+	for i := 0; i < 20; i++ {
+		if !m.AllowDNSRecheck(1) {
+			t.Fatalf("re-check %d refused", i+1)
+		}
+	}
+	if m.AllowDNSRecheck(1) {
+		t.Error("the 21st re-check was allowed")
+	}
+	if !m.AllowDNSRecheck(2) {
+		t.Error("another user's first re-check was refused")
+	}
+}

internal/web/handlers/handlers_apps.go

@@ -273,10 +273,26 @@ func (h *Handlers) HandleDomainDNSRecheck(w http.ResponseWriter, r *http.Request
 		http.Error(w, "internal error", http.StatusInternalServerError)
 		return
 	}
+	if !h.allowRecheck(r) {
+		http.Redirect(w, r, fmt.Sprintf("/outbound/domains/%d", d.ID), http.StatusSeeOther)
+		return
+	}
 	h.domainDNS(d, record, d.DMARCRua, true)
 	http.Redirect(w, r, fmt.Sprintf("/outbound/domains/%d?rechecked=1", d.ID), http.StatusSeeOther)
 }
 
+// allowRecheck is whether the signed-in user may force the DNS checks now
+// (auth.AllowDNSRecheck). Past the limit the page shows the cached result,
+// without saying it re-checked; the log says why.
+func (h *Handlers) allowRecheck(r *http.Request) bool {
+	p, ok := h.principal(r)
+	if !ok || h.auth == nil || h.auth.AllowDNSRecheck(p.ID) {
+		return true
+	}
+	logf("panel: user %d: DNS re-checks asked for too often; the cached result is shown", p.ID)
+	return false
+}
+
 // limitChoice names which of the form's three limit choices an application has:
 // the domain limit (none of its own), manual, or auto.
 func limitChoice(rl store.RateLimit, configured bool) string {

internal/web/handlers/handlers_apps_test.go

@@ -20,6 +20,8 @@ var offlineDNS = dnscheck.New([]string{"127.0.0.1:9"})
 
 type nopSenderMaps struct{}
 
+func (nopSenderMaps) Reload() error { return nil }
+
 func (nopSenderMaps) RebuildSenderLoginMaps([]postfix.Binding) error { return nil }
 
 // appStand is a panel with one sending domain and a working application

internal/web/handlers/handlers_domains.go

@@ -206,6 +206,9 @@ func (h *Handlers) HandleReload(w http.ResponseWriter, r *http.Request) {
 	if h.dmarc != nil && h.dmarc.Enabled() {
 		steps = append(steps, step{"the DMARC report maps", h.dmarc.Resync})
 	}
+	// A rebuild that changed nothing asked no reload (security audit
+	// 2026-10-17, W6): this button re-reads whatever is on disk.
+	steps = append(steps, step{"Postfix's reload", h.apps.ReloadPostfix})
 	for _, s := range steps {
 		if err := s.run(); err != nil {
 			logf("panel: manual reload (%s): %v", s.name, err)

internal/web/handlers/handlers_errors_test.go

@@ -26,6 +26,8 @@ const savedElsewhere = "Check the logs and try again."
 
 type failingSenderMaps struct{}
 
+func (failingSenderMaps) Reload() error { return nil }
+
 func (failingSenderMaps) RebuildSenderLoginMaps([]postfix.Binding) error {
 	return errors.New("write /srv/" + leak + "/sender_login_maps: permission denied")
 }

internal/web/handlers/handlers_inbound.go

@@ -275,6 +275,10 @@ func (h *Handlers) HandleInboundDNSRecheck(w http.ResponseWriter, r *http.Reques
 	if !ok {
 		return
 	}
+	if !h.allowRecheck(r) {
+		http.Redirect(w, r, fmt.Sprintf("/inbound/domains/%d", d.ID), http.StatusSeeOther)
+		return
+	}
 	if h.dns != nil && h.cfg.Hostname != "" {
 		h.dns.InboundMX(d.Name, h.cfg.Hostname, true)
 	}

internal/web/handlers/handlers_settings.go

@@ -203,6 +203,10 @@ func (h *Handlers) HandleSettingsDNSRecheck(w http.ResponseWriter, r *http.Reque
 	if _, ok := h.requireGlobal(w, r); !ok {
 		return
 	}
+	if !h.allowRecheck(r) {
+		http.Redirect(w, r, "/server/settings", http.StatusSeeOther)
+		return
+	}
 	h.dns.Server(h.cfg.Hostname, true)
 	h.serverMail(true)
 	http.Redirect(w, r, "/server/settings?rechecked=1", http.StatusSeeOther)