Configure timeout for HTTP2 transport

Otherwise, a stale TCP connection may still alive for too long, causing
unexpected failed to connect upstream error when network changed.

README.md

@@ -4,6 +4,8 @@
 [![Go Reference](https://pkg.go.dev/badge/github.com/Control-D-Inc/ctrld.svg)](https://pkg.go.dev/github.com/Control-D-Inc/ctrld)
 [![Go Report Card](https://goreportcard.com/badge/github.com/Control-D-Inc/ctrld)](https://goreportcard.com/report/github.com/Control-D-Inc/ctrld)
 
+![ctrld spash image](/docs/ctrldsplash.png)
+
 A highly configurable DNS forwarding proxy with support for:
 - Multiple listeners for incoming queries
 - Multiple upstreams with fallbacks
@@ -103,9 +105,11 @@ Available Commands:
   start       Quick start service and configure DNS on interface
   stop        Quick stop service and remove DNS from interface
   restart     Restart the ctrld service
+  reload      Reload the ctrld service
   status      Show status of the ctrld service
   uninstall   Stop and uninstall the ctrld service
   clients     Manage clients
+  upgrade     Upgrading ctrld to latest version
 
 Flags:
   -h, --help            help for ctrld

cmd/cli/cli_test.go

@@ -16,8 +16,31 @@ func Test_writeConfigFile(t *testing.T) {
 	_, err := os.Stat(configPath)
 	assert.True(t, os.IsNotExist(err))
 
-	assert.NoError(t, writeConfigFile())
+	assert.NoError(t, writeConfigFile(&cfg))
 
 	_, err = os.Stat(configPath)
 	require.NoError(t, err)
 }
+
+func Test_isStableVersion(t *testing.T) {
+	tests := []struct {
+		name     string
+		ver      string
+		isStable bool
+	}{
+		{"stable", "v1.3.5", true},
+		{"pre", "v1.3.5-next", false},
+		{"pre with commit hash", "v1.3.5-next-asdf", false},
+		{"dev", "dev", false},
+		{"empty", "dev", false},
+	}
+	for _, tc := range tests {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			if got := isStableVersion(tc.ver); got != tc.isStable {
+				t.Errorf("unexpected result for %s, want: %v, got: %v", tc.ver, tc.isStable, got)
+			}
+		})
+	}
+}

cmd/cli/control_server.go

@@ -10,6 +10,8 @@ import (
 	"sort"
 	"time"
 
+	"github.com/kardianos/service"
+
 	dto "github.com/prometheus/client_model/go"
 
 	"github.com/Control-D-Inc/ctrld"
@@ -21,6 +23,8 @@ const (
 	startedPath      = "/started"
 	reloadPath       = "/reload"
 	deactivationPath = "/deactivation"
+	cdPath           = "/cd"
+	ifacePath        = "/iface"
 )
 
 type controlServer struct {
@@ -69,7 +73,7 @@ func (p *prog) registerControlServerHandler() {
 		sort.Slice(clients, func(i, j int) bool {
 			return clients[i].IP.Less(clients[j].IP)
 		})
-		if p.cfg.Service.MetricsQueryStats {
+		if p.metricsQueryStats.Load() {
 			for _, client := range clients {
 				client.IncludeQueryCount = true
 				dm := &dto.Metric{}
@@ -171,6 +175,25 @@ func (p *prog) registerControlServerHandler() {
 		}
 		w.WriteHeader(code)
 	}))
+	p.cs.register(cdPath, http.HandlerFunc(func(w http.ResponseWriter, request *http.Request) {
+		if cdUID != "" {
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte(cdUID))
+			return
+		}
+		w.WriteHeader(http.StatusBadRequest)
+	}))
+	p.cs.register(ifacePath, http.HandlerFunc(func(w http.ResponseWriter, request *http.Request) {
+		// p.setDNS is only called when running as a service
+		if !service.Interactive() {
+			<-p.csSetDnsDone
+			if p.csSetDnsOk {
+				w.Write([]byte(iface))
+				return
+			}
+		}
+		w.WriteHeader(http.StatusBadRequest)
+	}))
 }
 
 func jsonResponse(next http.Handler) http.Handler {

cmd/cli/dns_proxy.go

@@ -21,6 +21,7 @@ import (
 	"tailscale.com/net/tsaddr"
 
 	"github.com/Control-D-Inc/ctrld"
+	"github.com/Control-D-Inc/ctrld/internal/controld"
 	"github.com/Control-D-Inc/ctrld/internal/dnscache"
 	ctrldnet "github.com/Control-D-Inc/ctrld/internal/net"
 )
@@ -32,6 +33,9 @@ const (
 	// https://thekelleys.org.uk/gitweb/?p=dnsmasq.git;a=blob;f=src/dns-protocol.h;h=76ac66a8c28317e9c121a74ab5fd0e20f6237dc8;hb=HEAD#l81
 	// This is also dns.EDNS0LOCALSTART, but define our own constant here for clarification.
 	EDNS0_OPTION_MAC = 0xFDE9
+
+	// selfUninstallMaxQueries is number of REFUSED queries seen before checking for self-uninstallation.
+	selfUninstallMaxQueries = 32
 )
 
 var osUpstreamConfig = &ctrld.UpstreamConfig{
@@ -89,6 +93,7 @@ func (p *prog) serveDNS(listenerNum string) error {
 			_ = w.WriteMsg(answer)
 			return
 		}
+		listenerConfig := p.cfg.Listener[listenerNum]
 		reqId := requestID()
 		ctx := context.WithValue(context.Background(), ctrld.ReqIdCtxKey{}, reqId)
 		if !listenerConfig.AllowWanClients && isWanClient(w.RemoteAddr()) {
@@ -101,6 +106,15 @@ func (p *prog) serveDNS(listenerNum string) error {
 		go p.detectLoop(m)
 		q := m.Question[0]
 		domain := canonicalName(q.Name)
+		if domain == selfCheckInternalTestDomain {
+			answer := resolveInternalDomainTestQuery(ctx, domain, m)
+			_ = w.WriteMsg(answer)
+			return
+		}
+		if _, ok := p.cacheFlushDomainsMap[domain]; ok && p.cache != nil {
+			p.cache.Purge()
+			ctrld.Log(ctx, mainLog.Load().Debug(), "received query %q, local cache is purged", domain)
+		}
 		remoteIP, _, _ := net.SplitHostPort(w.RemoteAddr().String())
 		ci := p.getClientInfo(remoteIP, m)
 		ci.ClientIDPref = p.cfg.Service.ClientIDPref
@@ -134,6 +148,7 @@ func (p *prog) serveDNS(listenerNum string) error {
 				failoverRcodes: failoverRcode,
 				ufr:            ur,
 			})
+			go p.doSelfUninstall(pr.answer)
 			answer = pr.answer
 			rtt := time.Since(t)
 			ctrld.Log(ctx, mainLog.Load().Debug(), "received response of %d bytes in %s", answer.Len(), rtt)
@@ -201,12 +216,9 @@ func (p *prog) serveDNS(listenerNum string) error {
 			addr := net.JoinHostPort(listenerConfig.IP, strconv.Itoa(listenerConfig.Port))
 			s, errCh := runDNSServer(addr, proto, handler)
 			defer s.Shutdown()
-			select {
-			case err := <-errCh:
-				return err
-			case <-time.After(5 * time.Second):
-				p.started <- struct{}{}
-			}
+
+			p.started <- struct{}{}
+
 			select {
 			case <-p.stopCh:
 			case <-ctx.Done():
@@ -282,7 +294,7 @@ networkRules:
 macRules:
 	for _, rule := range lc.Policy.Macs {
 		for source, targets := range rule {
-			if source != "" && strings.EqualFold(source, srcMac) {
+			if source != "" && (strings.EqualFold(source, srcMac) || wildcardMatches(strings.ToLower(source), strings.ToLower(srcMac))) {
 				matchedPolicy = lc.Policy.Name
 				matchedNetwork = source
 				networkTargets = targets
@@ -590,7 +602,8 @@ func canonicalName(fqdn string) string {
 	return q
 }
 
-func wildcardMatches(wildcard, domain string) bool {
+// wildcardMatches reports whether string str matches the wildcard pattern.
+func wildcardMatches(wildcard, str string) bool {
 	// Wildcard match.
 	wildCardParts := strings.Split(wildcard, "*")
 	if len(wildCardParts) != 2 {
@@ -600,15 +613,15 @@ func wildcardMatches(wildcard, domain string) bool {
 	switch {
 	case len(wildCardParts[0]) > 0 && len(wildCardParts[1]) > 0:
 		// Domain must match both prefix and suffix.
-		return strings.HasPrefix(domain, wildCardParts[0]) && strings.HasSuffix(domain, wildCardParts[1])
+		return strings.HasPrefix(str, wildCardParts[0]) && strings.HasSuffix(str, wildCardParts[1])
 
 	case len(wildCardParts[1]) > 0:
 		// Only suffix must match.
-		return strings.HasSuffix(domain, wildCardParts[1])
+		return strings.HasSuffix(str, wildCardParts[1])
 
 	case len(wildCardParts[0]) > 0:
 		// Only prefix must match.
-		return strings.HasPrefix(domain, wildCardParts[0])
+		return strings.HasPrefix(str, wildCardParts[0])
 	}
 
 	return false
@@ -742,20 +755,19 @@ func runDNSServer(addr, network string, handler dns.Handler) (*dns.Server, <-cha
 		Handler: handler,
 	}
 
-	waitLock := sync.Mutex{}
-	waitLock.Lock()
-	s.NotifyStartedFunc = waitLock.Unlock
+	startedCh := make(chan struct{})
+	s.NotifyStartedFunc = func() { sync.OnceFunc(func() { close(startedCh) })() }
 
 	errCh := make(chan error)
 	go func() {
 		defer close(errCh)
 		if err := s.ListenAndServe(); err != nil {
-			waitLock.Unlock()
+			s.NotifyStartedFunc()
 			mainLog.Load().Error().Err(err).Msgf("could not listen and serve on: %s", s.Addr)
 			errCh <- err
 		}
 	}()
-	waitLock.Lock()
+	<-startedCh
 	return s, errCh
 }
 
@@ -806,6 +818,13 @@ func (p *prog) getClientInfo(remoteIP string, msg *dns.Msg) *ctrld.ClientInfo {
 		ci.Hostname = p.ciTable.LookupHostname(ci.IP, ci.Mac)
 	}
 	ci.Self = queryFromSelf(ci.IP)
+	// If this is a query from self, but ci.IP is not loopback IP,
+	// try using hostname mapping for lookback IP if presents.
+	if ci.Self {
+		if name := p.ciTable.LocalHostname(); name != "" {
+			ci.Hostname = name
+		}
+	}
 	p.spoofLoopbackIpInClientInfo(ci)
 	return ci
 }
@@ -823,6 +842,51 @@ func (p *prog) spoofLoopbackIpInClientInfo(ci *ctrld.ClientInfo) {
 	}
 }
 
+// doSelfUninstall performs self-uninstall if these condition met:
+//
+// - There is only 1 ControlD upstream in-use.
+// - Number of refused queries seen so far equals to selfUninstallMaxQueries.
+// - The cdUID is deleted.
+func (p *prog) doSelfUninstall(answer *dns.Msg) {
+	if !p.canSelfUninstall.Load() || answer == nil || answer.Rcode != dns.RcodeRefused {
+		return
+	}
+
+	p.selfUninstallMu.Lock()
+	defer p.selfUninstallMu.Unlock()
+	if p.checkingSelfUninstall {
+		return
+	}
+
+	logger := mainLog.Load().With().Str("mode", "self-uninstall").Logger()
+	if p.refusedQueryCount > selfUninstallMaxQueries {
+		p.checkingSelfUninstall = true
+		_, err := controld.FetchResolverConfig(cdUID, rootCmd.Version, cdDev)
+		logger.Debug().Msg("maximum number of refused queries reached, checking device status")
+		selfUninstallCheck(err, p, logger)
+
+		if err != nil {
+			logger.Warn().Err(err).Msg("could not fetch resolver config")
+		}
+		// Cool-of period to prevent abusing the API.
+		go p.selfUninstallCoolOfPeriod()
+		return
+	}
+	p.refusedQueryCount++
+}
+
+// selfUninstallCoolOfPeriod waits for 30 minutes before
+// calling API again for checking ControlD device status.
+func (p *prog) selfUninstallCoolOfPeriod() {
+	t := time.NewTimer(time.Minute * 30)
+	defer t.Stop()
+	<-t.C
+	p.selfUninstallMu.Lock()
+	p.checkingSelfUninstall = false
+	p.refusedQueryCount = 0
+	p.selfUninstallMu.Unlock()
+}
+
 // queryFromSelf reports whether the input IP is from device running ctrld.
 func queryFromSelf(ip string) bool {
 	netIP := netip.MustParseAddr(ip)
@@ -936,3 +1000,21 @@ func isWanClient(na net.Addr) bool {
 		!ip.IsLinkLocalMulticast() &&
 		!tsaddr.CGNATRange().Contains(ip)
 }
+
+// resolveInternalDomainTestQuery resolves internal test domain query, returning the answer to the caller.
+func resolveInternalDomainTestQuery(ctx context.Context, domain string, m *dns.Msg) *dns.Msg {
+	ctrld.Log(ctx, mainLog.Load().Debug(), "internal domain test query")
+
+	q := m.Question[0]
+	answer := new(dns.Msg)
+	rrStr := fmt.Sprintf("%s A %s", domain, net.IPv4zero)
+	if q.Qtype == dns.TypeAAAA {
+		rrStr = fmt.Sprintf("%s AAAA %s", domain, net.IPv6zero)
+	}
+	rr, err := dns.NewRR(rrStr)
+	if err == nil {
+		answer.Answer = append(answer.Answer, rr)
+	}
+	answer.SetReply(m)
+	return answer
+}

cmd/cli/dns_proxy_test.go

@@ -22,14 +22,21 @@ func Test_wildcardMatches(t *testing.T) {
 		domain   string
 		match    bool
 	}{
-		{"prefix parent should not match", "*.windscribe.com", "windscribe.com", false},
-		{"prefix", "*.windscribe.com", "anything.windscribe.com", true},
-		{"prefix not match other domain", "*.windscribe.com", "example.com", false},
-		{"prefix not match domain in name", "*.windscribe.com", "wwindscribe.com", false},
-		{"suffix", "suffix.*", "suffix.windscribe.com", true},
-		{"suffix not match other", "suffix.*", "suffix1.windscribe.com", false},
-		{"both", "suffix.*.windscribe.com", "suffix.anything.windscribe.com", true},
-		{"both not match", "suffix.*.windscribe.com", "suffix1.suffix.windscribe.com", false},
+		{"domain - prefix parent should not match", "*.windscribe.com", "windscribe.com", false},
+		{"domain - prefix", "*.windscribe.com", "anything.windscribe.com", true},
+		{"domain - prefix not match other s", "*.windscribe.com", "example.com", false},
+		{"domain - prefix not match s in name", "*.windscribe.com", "wwindscribe.com", false},
+		{"domain - suffix", "suffix.*", "suffix.windscribe.com", true},
+		{"domain - suffix not match other", "suffix.*", "suffix1.windscribe.com", false},
+		{"domain - both", "suffix.*.windscribe.com", "suffix.anything.windscribe.com", true},
+		{"domain - both not match", "suffix.*.windscribe.com", "suffix1.suffix.windscribe.com", false},
+		{"mac - prefix", "*:98:05:b4:2b", "d4:67:98:05:b4:2b", true},
+		{"mac - prefix not match other s", "*:98:05:b4:2b", "0d:ba:54:09:94:2c", false},
+		{"mac - prefix not match s in name", "*:98:05:b4:2b", "e4:67:97:05:b4:2b", false},
+		{"mac - suffix", "d4:67:98:*", "d4:67:98:05:b4:2b", true},
+		{"mac - suffix not match other", "d4:67:98:*", "d4:67:97:15:b4:2b", false},
+		{"mac - both", "d4:67:98:*:b4:2b", "d4:67:98:05:b4:2b", true},
+		{"mac - both not match", "d4:67:98:*:b4:2b", "d4:67:97:05:c4:2b", false},
 	}
 
 	for _, tc := range tests {

cmd/cli/loop.go

@@ -105,6 +105,10 @@ func (p *prog) checkDnsLoop() {
 	for uid := range p.loop {
 		msg := loopTestMsg(uid)
 		uc := upstream[uid]
+		// Skipping upstream which is being marked as down.
+		if uc == nil {
+			continue
+		}
 		resolver, err := ctrld.NewResolver(uc)
 		if err != nil {
 			mainLog.Load().Warn().Err(err).Msgf("could not perform loop check for upstream: %q, endpoint: %q", uc.Name, uc.Endpoint)

cmd/cli/main.go

@@ -35,6 +35,9 @@ var (
 	nextdns           string
 	cdUpstreamProto   string
 	deactivationPin   int64
+	skipSelfChecks    bool
+	cleanup           bool
+	startOnly         bool
 
 	mainLog       atomic.Pointer[zerolog.Logger]
 	consoleWriter zerolog.ConsoleWriter
@@ -62,8 +65,11 @@ func Main() {
 }
 
 func normalizeLogFilePath(logFilePath string) string {
-	if logFilePath == "" || filepath.IsAbs(logFilePath) || service.Interactive() {
-		return logFilePath
+	// In cleanup mode, we always want the full log file path.
+	if !cleanup {
+		if logFilePath == "" || filepath.IsAbs(logFilePath) || service.Interactive() {
+			return logFilePath
+		}
 	}
 	if homedir != "" {
 		return filepath.Join(homedir, logFilePath)
@@ -120,14 +126,14 @@ func initLoggingWithBackup(doBackup bool) {
 		flags := os.O_CREATE | os.O_RDWR | os.O_APPEND
 		if doBackup {
 			// Backup old log file with .1 suffix.
-			if err := os.Rename(logFilePath, logFilePath+".1"); err != nil && !os.IsNotExist(err) {
+			if err := os.Rename(logFilePath, logFilePath+oldLogSuffix); err != nil && !os.IsNotExist(err) {
 				mainLog.Load().Error().Msgf("could not backup old log file: %v", err)
 			} else {
 				// Backup was created, set flags for truncating old log file.
 				flags = os.O_CREATE | os.O_RDWR
 			}
 		}
-		logFile, err := os.OpenFile(logFilePath, flags, os.FileMode(0o600))
+		logFile, err := openLogFile(logFilePath, flags)
 		if err != nil {
 			mainLog.Load().Error().Msgf("failed to create log file: %v", err)
 			os.Exit(1)

cmd/cli/metrics.go

@@ -107,7 +107,7 @@ func (p *prog) runMetricsServer(ctx context.Context, reloadCh chan struct{}) {
 
 	reg := prometheus.NewRegistry()
 	// Register queries count stats if enabled.
-	if cfg.Service.MetricsQueryStats {
+	if p.metricsQueryStats.Load() {
 		reg.MustRegister(statsQueriesCount)
 		reg.MustRegister(statsClientQueriesCount)
 	}

cmd/cli/net_darwin.go

@@ -43,20 +43,32 @@ func networkServiceName(ifaceName string, r io.Reader) string {
 	return ""
 }
 
-// validInterface reports whether the *net.Interface is a valid one, which includes:
-//
-// - en0: physical wireless
-// - en1: Thunderbolt 1
-// - en2: Thunderbolt 2
-// - en3: Thunderbolt 3
-// - en4: Thunderbolt 4
-//
-// For full list, see: https://unix.stackexchange.com/questions/603506/what-are-these-ifconfig-interfaces-on-macos
-func validInterface(iface *net.Interface) bool {
-	switch iface.Name {
-	case "en0", "en1", "en2", "en3", "en4":
-		return true
-	default:
-		return false
-	}
+// validInterface reports whether the *net.Interface is a valid one.
+func validInterface(iface *net.Interface, validIfacesMap map[string]struct{}) bool {
+	_, ok := validIfacesMap[iface.Name]
+	return ok
+}
+
+func validInterfacesMap() map[string]struct{} {
+	b, err := exec.Command("networksetup", "-listallhardwareports").Output()
+	if err != nil {
+		return nil
+	}
+	return parseListAllHardwarePorts(bytes.NewReader(b))
+}
+
+// parseListAllHardwarePorts parses output of "networksetup -listallhardwareports"
+// and returns map presents all hardware ports.
+func parseListAllHardwarePorts(r io.Reader) map[string]struct{} {
+	m := make(map[string]struct{})
+	scanner := bufio.NewScanner(r)
+	for scanner.Scan() {
+		line := scanner.Text()
+		after, ok := strings.CutPrefix(line, "Device: ")
+		if !ok {
+			continue
+		}
+		m[after] = struct{}{}
+	}
+	return m
 }

cmd/cli/net_darwin_test.go

@@ -1,6 +1,7 @@
 package cli
 
 import (
+	"maps"
 	"strings"
 	"testing"
 
@@ -57,3 +58,47 @@ func Test_networkServiceName(t *testing.T) {
 		})
 	}
 }
+
+const listallhardwareportsOutput = `
+Hardware Port: Ethernet Adapter (en6)
+Device: en6
+Ethernet Address: 3a:3e:fc:1e:ab:41
+
+Hardware Port: Ethernet Adapter (en7)
+Device: en7
+Ethernet Address: 3a:3e:fc:1e:ab:42
+
+Hardware Port: Thunderbolt Bridge
+Device: bridge0
+Ethernet Address: 36:21:bb:3a:7a:40
+
+Hardware Port: Wi-Fi
+Device: en0
+Ethernet Address: a0:78:17:68:56:3f
+
+Hardware Port: Thunderbolt 1
+Device: en1
+Ethernet Address: 36:21:bb:3a:7a:40
+
+Hardware Port: Thunderbolt 2
+Device: en2
+Ethernet Address: 36:21:bb:3a:7a:44
+
+VLAN Configurations
+===================
+`
+
+func Test_parseListAllHardwarePorts(t *testing.T) {
+	expected := map[string]struct{}{
+		"en0":     {},
+		"en1":     {},
+		"en2":     {},
+		"en6":     {},
+		"en7":     {},
+		"bridge0": {},
+	}
+	m := parseListAllHardwarePorts(strings.NewReader(listallhardwareportsOutput))
+	if !maps.Equal(m, expected) {
+		t.Errorf("unexpected output, want: %v, got: %v", expected, m)
+	}
+}

cmd/cli/net_others.go

@@ -6,4 +6,6 @@ import "net"
 
 func patchNetIfaceName(iface *net.Interface) error { return nil }
 
-func validInterface(iface *net.Interface) bool { return true }
+func validInterface(iface *net.Interface, validIfacesMap map[string]struct{}) bool { return true }
+
+func validInterfacesMap() map[string]struct{} { return nil }

cmd/cli/net_windows.go

@@ -10,7 +10,7 @@ func patchNetIfaceName(iface *net.Interface) error {
 
 // validInterface reports whether the *net.Interface is a valid one.
 // On Windows, only physical interfaces are considered valid.
-func validInterface(iface *net.Interface) bool {
+func validInterface(iface *net.Interface, validIfacesMap map[string]struct{}) bool {
 	if iface == nil {
 		return false
 	}
@@ -19,3 +19,5 @@ func validInterface(iface *net.Interface) bool {
 	}
 	return false
 }
+
+func validInterfacesMap() map[string]struct{} { return nil }

cmd/cli/os_darwin.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net"
 	"os/exec"
+	"strings"
 
 	"github.com/Control-D-Inc/ctrld/internal/resolvconffile"
 )
@@ -30,6 +31,18 @@ func deAllocateIP(ip string) error {
 	return nil
 }
 
+// setDnsIgnoreUnusableInterface likes setDNS, but return a nil error if the interface is not usable.
+func setDnsIgnoreUnusableInterface(iface *net.Interface, nameservers []string) error {
+	if err := setDNS(iface, nameservers); err != nil {
+		// TODO: investiate whether we can detect this without relying on error message.
+		if strings.Contains(err.Error(), " is not a recognized network service") {
+			return nil
+		}
+		return err
+	}
+	return nil
+}
+
 // set the dns server for the provided network interface
 // networksetup -setdnsservers Wi-Fi 8.8.8.8 1.1.1.1
 // TODO(cuonglm): use system API
@@ -43,6 +56,18 @@ func setDNS(iface *net.Interface, nameservers []string) error {
 	return nil
 }
 
+// resetDnsIgnoreUnusableInterface likes resetDNS, but return a nil error if the interface is not usable.
+func resetDnsIgnoreUnusableInterface(iface *net.Interface) error {
+	if err := resetDNS(iface); err != nil {
+		// TODO: investiate whether we can detect this without relying on error message.
+		if strings.Contains(err.Error(), " is not a recognized network service") {
+			return nil
+		}
+		return err
+	}
+	return nil
+}
+
 // TODO(cuonglm): use system API
 func resetDNS(iface *net.Interface) error {
 	if ns := savedStaticNameservers(iface); len(ns) > 0 {

cmd/cli/os_freebsd.go

@@ -29,6 +29,11 @@ func deAllocateIP(ip string) error {
 	return nil
 }
 
+// setDnsIgnoreUnusableInterface likes setDNS, but return a nil error if the interface is not usable.
+func setDnsIgnoreUnusableInterface(iface *net.Interface, nameservers []string) error {
+	return setDNS(iface, nameservers)
+}
+
 // set the dns server for the provided network interface
 func setDNS(iface *net.Interface, nameservers []string) error {
 	r, err := dns.NewOSConfigurator(logf, iface.Name)
@@ -49,6 +54,11 @@ func setDNS(iface *net.Interface, nameservers []string) error {
 	return nil
 }
 
+// resetDnsIgnoreUnusableInterface likes resetDNS, but return a nil error if the interface is not usable.
+func resetDnsIgnoreUnusableInterface(iface *net.Interface) error {
+	return resetDNS(iface)
+}
+
 func resetDNS(iface *net.Interface) error {
 	r, err := dns.NewOSConfigurator(logf, iface.Name)
 	if err != nil {

cmd/cli/os_linux.go

@@ -9,6 +9,7 @@ import (
 	"net"
 	"net/netip"
 	"os/exec"
+	"slices"
 	"strings"
 	"syscall"
 	"time"
@@ -23,6 +24,8 @@ import (
 	"github.com/Control-D-Inc/ctrld/internal/resolvconffile"
 )
 
+const resolvConfBackupFailedMsg = "open /etc/resolv.pre-ctrld-backup.conf: read-only file system"
+
 // allocate loopback ip
 // sudo ip a add 127.0.0.2/24 dev lo
 func allocateIP(ip string) error {
@@ -45,7 +48,11 @@ func deAllocateIP(ip string) error {
 
 const maxSetDNSAttempts = 5
 
-// set the dns server for the provided network interface
+// setDnsIgnoreUnusableInterface likes setDNS, but return a nil error if the interface is not usable.
+func setDnsIgnoreUnusableInterface(iface *net.Interface, nameservers []string) error {
+	return setDNS(iface, nameservers)
+}
+
 func setDNS(iface *net.Interface, nameservers []string) error {
 	r, err := dns.NewOSConfigurator(logf, iface.Name)
 	if err != nil {
@@ -115,6 +122,11 @@ func setDNS(iface *net.Interface, nameservers []string) error {
 	return nil
 }
 
+// resetDnsIgnoreUnusableInterface likes resetDNS, but return a nil error if the interface is not usable.
+func resetDnsIgnoreUnusableInterface(iface *net.Interface) error {
+	return resetDNS(iface)
+}
+
 func resetDNS(iface *net.Interface) (err error) {
 	defer func() {
 		if err == nil {
@@ -276,8 +288,7 @@ func ignoringEINTR(fn func() error) error {
 func isSubSet(s1, s2 []string) bool {
 	ok := true
 	for _, ns := range s1 {
-		// TODO(cuonglm): use slices.Contains once upgrading to go1.21
-		if sliceContains(s2, ns) {
+		if slices.Contains(s2, ns) {
 			continue
 		}
 		ok = false
@@ -285,19 +296,3 @@ func isSubSet(s1, s2 []string) bool {
 	}
 	return ok
 }
-
-// sliceContains reports whether v is present in s.
-func sliceContains[S ~[]E, E comparable](s S, v E) bool {
-	return sliceIndex(s, v) >= 0
-}
-
-// sliceIndex returns the index of the first occurrence of v in s,
-// or -1 if not present.
-func sliceIndex[S ~[]E, E comparable](s S, v E) int {
-	for i := range s {
-		if v == s[i] {
-			return i
-		}
-	}
-	return -1
-}

cmd/cli/os_windows.go

@@ -16,7 +16,6 @@ import (
 )
 
 const (
-	forwardersFilename       = ".forwarders.txt"
 	v4InterfaceKeyPathFormat = `HKLM:\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters\Interfaces\`
 	v6InterfaceKeyPathFormat = `HKLM:\SYSTEM\CurrentControlSet\Services\Tcpip6\Parameters\Interfaces\`
 )
@@ -26,6 +25,12 @@ var (
 	resetDNSOnce sync.Once
 )
 
+// setDnsIgnoreUnusableInterface likes setDNS, but return a nil error if the interface is not usable.
+func setDnsIgnoreUnusableInterface(iface *net.Interface, nameservers []string) error {
+	return setDNS(iface, nameservers)
+}
+
+// setDNS sets the dns server for the provided network interface
 func setDNS(iface *net.Interface, nameservers []string) error {
 	if len(nameservers) == 0 {
 		return errors.New("empty DNS nameservers")
@@ -34,18 +39,13 @@ func setDNS(iface *net.Interface, nameservers []string) error {
 		// If there's a Dns server running, that means we are on AD with Dns feature enabled.
 		// Configuring the Dns server to forward queries to ctrld instead.
 		if windowsHasLocalDnsServerRunning() {
-			file := absHomeDir(forwardersFilename)
-			if data, _ := os.ReadFile(file); len(data) > 0 {
-				if err := removeDnsServerForwarders(strings.Split(string(data), ",")); err != nil {
-					mainLog.Load().Error().Err(err).Msg("could not remove current forwarders settings")
-				} else {
-					mainLog.Load().Debug().Msg("removed current forwarders settings.")
-				}
-			}
+			file := absHomeDir(windowsForwardersFilename)
+			oldForwardersContent, _ := os.ReadFile(file)
 			if err := os.WriteFile(file, []byte(strings.Join(nameservers, ",")), 0600); err != nil {
 				mainLog.Load().Warn().Err(err).Msg("could not save forwarders settings")
 			}
-			if err := addDnsServerForwarders(nameservers); err != nil {
+			oldForwarders := strings.Split(string(oldForwardersContent), ",")
+			if err := addDnsServerForwarders(nameservers, oldForwarders); err != nil {
 				mainLog.Load().Warn().Err(err).Msg("could not set forwarders settings")
 			}
 		}
@@ -61,12 +61,17 @@ func setDNS(iface *net.Interface, nameservers []string) error {
 	return nil
 }
 
+// resetDnsIgnoreUnusableInterface likes resetDNS, but return a nil error if the interface is not usable.
+func resetDnsIgnoreUnusableInterface(iface *net.Interface) error {
+	return resetDNS(iface)
+}
+
 // TODO(cuonglm): should we use system API?
 func resetDNS(iface *net.Interface) error {
 	resetDNSOnce.Do(func() {
 		// See corresponding comment in setDNS.
 		if windowsHasLocalDnsServerRunning() {
-			file := absHomeDir(forwardersFilename)
+			file := absHomeDir(windowsForwardersFilename)
 			content, err := os.ReadFile(file)
 			if err != nil {
 				mainLog.Load().Error().Err(err).Msg("could not read forwarders settings")
@@ -202,14 +207,32 @@ func currentStaticDNS(iface *net.Interface) ([]string, error) {
 	return ns, nil
 }
 
-// addDnsServerForwarders adds given nameservers to DNS server forwarders list.
-func addDnsServerForwarders(nameservers []string) error {
-	for _, ns := range nameservers {
-		cmd := fmt.Sprintf("Add-DnsServerForwarder -IPAddress %s", ns)
-		if out, err := powershell(cmd); err != nil {
-			return fmt.Errorf("%w: %s", err, string(out))
+// addDnsServerForwarders adds given nameservers to DNS server forwarders list,
+// and also removing old forwarders if provided.
+func addDnsServerForwarders(nameservers, old []string) error {
+	newForwardersMap := make(map[string]struct{})
+	newForwarders := make([]string, len(nameservers))
+	for i := range nameservers {
+		newForwardersMap[nameservers[i]] = struct{}{}
+		newForwarders[i] = fmt.Sprintf("%q", nameservers[i])
+	}
+	oldForwarders := old[:0]
+	for _, fwd := range old {
+		if _, ok := newForwardersMap[fwd]; !ok {
+			oldForwarders = append(oldForwarders, fwd)
 		}
 	}
+	// NOTE: It is important to add new forwarder before removing old one.
+	//       Testing on Windows Server 2022 shows that removing forwarder1
+	//       then adding forwarder2 sometimes ends up adding both of them
+	//       to the forwarders list.
+	cmd := fmt.Sprintf("Add-DnsServerForwarder -IPAddress %s", strings.Join(newForwarders, ","))
+	if len(oldForwarders) > 0 {
+		cmd = fmt.Sprintf("%s ; Remove-DnsServerForwarder -IPAddress %s -Force", cmd, strings.Join(oldForwarders, ","))
+	}
+	if out, err := powershell(cmd); err != nil {
+		return fmt.Errorf("%w: %s", err, string(out))
+	}
 	return nil
 }
 

cmd/cli/prog.go

@@ -12,19 +12,24 @@ import (
 	"net/url"
 	"os"
 	"runtime"
+	"slices"
 	"sort"
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"syscall"
+	"time"
 
 	"github.com/kardianos/service"
+	"github.com/rs/zerolog"
 	"github.com/spf13/viper"
 	"tailscale.com/net/interfaces"
 	"tailscale.com/net/tsaddr"
 
 	"github.com/Control-D-Inc/ctrld"
 	"github.com/Control-D-Inc/ctrld/internal/clientinfo"
+	"github.com/Control-D-Inc/ctrld/internal/controld"
 	"github.com/Control-D-Inc/ctrld/internal/dnscache"
 	"github.com/Control-D-Inc/ctrld/internal/router"
 )
@@ -33,11 +38,23 @@ const (
 	defaultSemaphoreCap  = 256
 	ctrldLogUnixSock     = "ctrld_start.sock"
 	ctrldControlUnixSock = "ctrld_control.sock"
-	upstreamPrefix       = "upstream."
-	upstreamOS           = upstreamPrefix + "os"
-	upstreamPrivate      = upstreamPrefix + "private"
+	// iOS unix socket name max length is 11.
+	ctrldControlUnixSockMobile = "cd.sock"
+	upstreamPrefix             = "upstream."
+	upstreamOS                 = upstreamPrefix + "os"
+	upstreamPrivate            = upstreamPrefix + "private"
+	dnsWatchdogDefaultInterval = 20 * time.Second
 )
 
+// ControlSocketName returns name for control unix socket.
+func ControlSocketName() string {
+	if isMobile() {
+		return ctrldControlUnixSockMobile
+	} else {
+		return ctrldControlUnixSock
+	}
+}
+
 var logf = func(format string, args ...any) {
 	mainLog.Load().Debug().Msgf(format, args...)
 }
@@ -51,25 +68,38 @@ var svcConfig = &service.Config{
 var useSystemdResolved = false
 
 type prog struct {
-	mu           sync.Mutex
-	waitCh       chan struct{}
-	stopCh       chan struct{}
-	reloadCh     chan struct{} // For Windows.
-	reloadDoneCh chan struct{}
-	logConn      net.Conn
-	cs           *controlServer
+	mu               sync.Mutex
+	waitCh           chan struct{}
+	stopCh           chan struct{}
+	reloadCh         chan struct{} // For Windows.
+	reloadDoneCh     chan struct{}
+	apiReloadCh      chan *ctrld.Config
+	logConn          net.Conn
+	cs               *controlServer
+	csSetDnsDone     chan struct{}
+	csSetDnsOk       bool
+	dnsWatchDogOnce  sync.Once
+	dnsWg            sync.WaitGroup
+	dnsWatcherStopCh chan struct{}
 
-	cfg            *ctrld.Config
-	localUpstreams []string
-	ptrNameservers []string
-	appCallback    *AppCallback
-	cache          dnscache.Cacher
-	sema           semaphore
-	ciTable        *clientinfo.Table
-	um             *upstreamMonitor
-	router         router.Router
-	ptrLoopGuard   *loopGuard
-	lanLoopGuard   *loopGuard
+	cfg                  *ctrld.Config
+	localUpstreams       []string
+	ptrNameservers       []string
+	appCallback          *AppCallback
+	cache                dnscache.Cacher
+	cacheFlushDomainsMap map[string]struct{}
+	sema                 semaphore
+	ciTable              *clientinfo.Table
+	um                   *upstreamMonitor
+	router               router.Router
+	ptrLoopGuard         *loopGuard
+	lanLoopGuard         *loopGuard
+	metricsQueryStats    atomic.Bool
+
+	selfUninstallMu       sync.Mutex
+	refusedQueryCount     int
+	canSelfUninstall      atomic.Bool
+	checkingSelfUninstall bool
 
 	loopMu sync.Mutex
 	loop   map[string]bool
@@ -103,11 +133,15 @@ func (p *prog) runWait() {
 			p.run(reload, reloadCh)
 			reload = true
 		}()
+
+		var newCfg *ctrld.Config
 		select {
 		case sig := <-reloadSigCh:
 			logger.Notice().Msgf("got signal: %s, reloading...", sig.String())
 		case <-p.reloadCh:
 			logger.Notice().Msg("reloading...")
+		case apiCfg := <-p.apiReloadCh:
+			newCfg = apiCfg
 		case <-p.stopCh:
 			close(reloadCh)
 			return
@@ -117,28 +151,31 @@ func (p *prog) runWait() {
 			close(reloadCh)
 			<-done
 		}
-		newCfg := &ctrld.Config{}
-		v := viper.NewWithOptions(viper.KeyDelimiter("::"))
-		ctrld.InitConfig(v, "ctrld")
-		if configPath != "" {
-			v.SetConfigFile(configPath)
-		}
-		if err := v.ReadInConfig(); err != nil {
-			logger.Err(err).Msg("could not read new config")
-			waitOldRunDone()
-			continue
-		}
-		if err := v.Unmarshal(&newCfg); err != nil {
-			logger.Err(err).Msg("could not unmarshal new config")
-			waitOldRunDone()
-			continue
-		}
-		if cdUID != "" {
-			if err := processCDFlags(newCfg); err != nil {
-				logger.Err(err).Msg("could not fetch ControlD config")
+
+		if newCfg == nil {
+			newCfg = &ctrld.Config{}
+			v := viper.NewWithOptions(viper.KeyDelimiter("::"))
+			ctrld.InitConfig(v, "ctrld")
+			if configPath != "" {
+				v.SetConfigFile(configPath)
+			}
+			if err := v.ReadInConfig(); err != nil {
+				logger.Err(err).Msg("could not read new config")
 				waitOldRunDone()
 				continue
 			}
+			if err := v.Unmarshal(&newCfg); err != nil {
+				logger.Err(err).Msg("could not unmarshal new config")
+				waitOldRunDone()
+				continue
+			}
+			if cdUID != "" {
+				if err := processCDFlags(newCfg); err != nil {
+					logger.Err(err).Msg("could not fetch ControlD config")
+					waitOldRunDone()
+					continue
+				}
+			}
 		}
 
 		waitOldRunDone()
@@ -164,6 +201,10 @@ func (p *prog) runWait() {
 			continue
 		}
 
+		if err := writeConfigFile(newCfg); err != nil {
+			logger.Err(err).Msg("could not write new config")
+		}
+
 		// This needs to be done here, otherwise, the DNS handler may observe an invalid
 		// upstream config because its initialization function have not been called yet.
 		mainLog.Load().Debug().Msg("setup upstream with new config")
@@ -174,6 +215,7 @@ func (p *prog) runWait() {
 		p.mu.Unlock()
 
 		logger.Notice().Msg("reloading config successfully")
+
 		select {
 		case p.reloadDoneCh <- struct{}{}:
 		default:
@@ -182,9 +224,6 @@ func (p *prog) runWait() {
 }
 
 func (p *prog) preRun() {
-	if !service.Interactive() {
-		p.setDNS()
-	}
 	if runtime.GOOS == "darwin" {
 		p.onStopped = append(p.onStopped, func() {
 			if !service.Interactive() {
@@ -194,12 +233,76 @@ func (p *prog) preRun() {
 	}
 }
 
+func (p *prog) postRun() {
+	if !service.Interactive() {
+		p.resetDNS()
+		ns := ctrld.InitializeOsResolver()
+		mainLog.Load().Debug().Msgf("initialized OS resolver with nameservers: %v", ns)
+		p.setDNS()
+	}
+}
+
+// apiConfigReload calls API to check for latest config update then reload ctrld if necessary.
+func (p *prog) apiConfigReload() {
+	if cdUID == "" {
+		return
+	}
+
+	secs := 3600
+	if p.cfg.Service.RefetchTime != nil && *p.cfg.Service.RefetchTime > 0 {
+		secs = *p.cfg.Service.RefetchTime
+	}
+
+	ticker := time.NewTicker(time.Duration(secs) * time.Second)
+	defer ticker.Stop()
+
+	logger := mainLog.Load().With().Str("mode", "api-reload").Logger()
+	logger.Debug().Msg("starting custom config reload timer")
+	lastUpdated := time.Now().Unix()
+	for {
+		select {
+		case <-ticker.C:
+			resolverConfig, err := controld.FetchResolverConfig(cdUID, rootCmd.Version, cdDev)
+			selfUninstallCheck(err, p, logger)
+			if err != nil {
+				logger.Warn().Err(err).Msg("could not fetch resolver config")
+				continue
+			}
+
+			if resolverConfig.Ctrld.CustomConfig == "" {
+				continue
+			}
+
+			if resolverConfig.Ctrld.CustomLastUpdate > lastUpdated {
+				lastUpdated = time.Now().Unix()
+				cfg := &ctrld.Config{}
+				if err := validateCdRemoteConfig(resolverConfig, cfg); err != nil {
+					logger.Warn().Err(err).Msg("skipping invalid custom config")
+					if _, err := controld.UpdateCustomLastFailed(cdUID, rootCmd.Version, cdDev, true); err != nil {
+						logger.Error().Err(err).Msg("could not mark custom last update failed")
+					}
+					break
+				}
+				setListenerDefaultValue(cfg)
+				logger.Debug().Msg("custom config changes detected, reloading...")
+				p.apiReloadCh <- cfg
+			} else {
+				logger.Debug().Msg("custom config does not change")
+			}
+		case <-p.stopCh:
+			return
+		}
+	}
+}
+
 func (p *prog) setupUpstream(cfg *ctrld.Config) {
 	localUpstreams := make([]string, 0, len(cfg.Upstream))
 	ptrNameservers := make([]string, 0, len(cfg.Upstream))
+	isControlDUpstream := false
 	for n := range cfg.Upstream {
 		uc := cfg.Upstream[n]
 		uc.Init()
+		isControlDUpstream = isControlDUpstream || uc.IsControlD()
 		if uc.BootstrapIP == "" {
 			uc.SetupBootstrapIP()
 			mainLog.Load().Info().Msgf("bootstrap IPs for upstream.%s: %q", n, uc.BootstrapIPs())
@@ -216,6 +319,10 @@ func (p *prog) setupUpstream(cfg *ctrld.Config) {
 			ptrNameservers = append(ptrNameservers, uc.Endpoint)
 		}
 	}
+	// Self-uninstallation is ok If there is only 1 ControlD upstream, and no remote config.
+	if len(cfg.Upstream) == 1 && isControlDUpstream {
+		p.canSelfUninstall.Store(true)
+	}
 	p.localUpstreams = localUpstreams
 	p.ptrNameservers = ptrNameservers
 }
@@ -237,17 +344,31 @@ func (p *prog) run(reload bool, reloadCh chan struct{}) {
 	numListeners := len(p.cfg.Listener)
 	if !reload {
 		p.started = make(chan struct{}, numListeners)
+		if p.cs != nil {
+			p.csSetDnsDone = make(chan struct{}, 1)
+			p.registerControlServerHandler()
+			if err := p.cs.start(); err != nil {
+				mainLog.Load().Warn().Err(err).Msg("could not start control server")
+			}
+			mainLog.Load().Debug().Msgf("control server started: %s", p.cs.addr)
+		}
 	}
 	p.onStartedDone = make(chan struct{})
 	p.loop = make(map[string]bool)
 	p.lanLoopGuard = newLoopGuard()
 	p.ptrLoopGuard = newLoopGuard()
+	p.cacheFlushDomainsMap = nil
+	p.metricsQueryStats.Store(p.cfg.Service.MetricsQueryStats)
 	if p.cfg.Service.CacheEnable {
 		cacher, err := dnscache.NewLRUCache(p.cfg.Service.CacheSize)
 		if err != nil {
 			mainLog.Load().Error().Err(err).Msg("failed to create cacher, caching is disabled")
 		} else {
 			p.cache = cacher
+			p.cacheFlushDomainsMap = make(map[string]struct{}, 256)
+			for _, domain := range p.cfg.Service.CacheFlushDomains {
+				p.cacheFlushDomainsMap[canonicalName(domain)] = struct{}{}
+			}
 		}
 	}
 
@@ -364,12 +485,8 @@ func (p *prog) run(reload bool, reloadCh chan struct{}) {
 		if p.logConn != nil {
 			_ = p.logConn.Close()
 		}
-		if p.cs != nil {
-			p.registerControlServerHandler()
-			if err := p.cs.start(); err != nil {
-				mainLog.Load().Warn().Err(err).Msg("could not start control server")
-			}
-		}
+		go p.apiConfigReload()
+		p.postRun()
 	}
 	wg.Wait()
 }
@@ -413,17 +530,25 @@ func (p *prog) deAllocateIP() error {
 }
 
 func (p *prog) setDNS() {
+	setDnsOK := false
+	defer func() {
+		p.csSetDnsOk = setDnsOK
+		p.csSetDnsDone <- struct{}{}
+		close(p.csSetDnsDone)
+	}()
+
 	if cfg.Listener == nil {
 		return
 	}
 	if iface == "" {
 		return
 	}
+	runningIface := iface
 	// allIfaces tracks whether we should set DNS for all physical interfaces.
 	allIfaces := false
-	if iface == "auto" {
-		iface = defaultIfaceName()
-		// If iface is "auto", it means user does not specify "--iface" flag.
+	if runningIface == "auto" {
+		runningIface = defaultIfaceName()
+		// If runningIface is "auto", it means user does not specify "--iface" flag.
 		// In this case, ctrld has to set DNS for all physical interfaces, so
 		// thing will still work when user switch from one to the other.
 		allIfaces = requiredMultiNICsConfig()
@@ -432,8 +557,8 @@ func (p *prog) setDNS() {
 	if lc == nil {
 		return
 	}
-	logger := mainLog.Load().With().Str("iface", iface).Logger()
-	netIface, err := netInterface(iface)
+	logger := mainLog.Load().With().Str("iface", runningIface).Logger()
+	netIface, err := netInterface(runningIface)
 	if err != nil {
 		logger.Error().Err(err).Msg("could not get interface")
 		return
@@ -467,33 +592,108 @@ func (p *prog) setDNS() {
 		logger.Error().Err(err).Msgf("could not set DNS for interface")
 		return
 	}
+	setDnsOK = true
 	logger.Debug().Msg("setting DNS successfully")
 	if shouldWatchResolvconf() {
 		servers := make([]netip.Addr, len(nameservers))
 		for i := range nameservers {
 			servers[i] = netip.MustParseAddr(nameservers[i])
 		}
-		go watchResolvConf(netIface, servers, setResolvConf)
+		p.dnsWg.Add(1)
+		go func() {
+			defer p.dnsWg.Done()
+			p.watchResolvConf(netIface, servers, setResolvConf)
+		}()
 	}
 	if allIfaces {
 		withEachPhysicalInterfaces(netIface.Name, "set DNS", func(i *net.Interface) error {
-			return setDNS(i, nameservers)
+			return setDnsIgnoreUnusableInterface(i, nameservers)
 		})
 	}
+	if p.dnsWatchdogEnabled() {
+		p.dnsWg.Add(1)
+		go func() {
+			defer p.dnsWg.Done()
+			p.dnsWatchdog(netIface, nameservers, allIfaces)
+		}()
+	}
+}
+
+// dnsWatchdogEnabled reports whether DNS watchdog is enabled.
+func (p *prog) dnsWatchdogEnabled() bool {
+	if ptr := p.cfg.Service.DnsWatchdogEnabled; ptr != nil {
+		return *ptr
+	}
+	return true
+}
+
+// dnsWatchdogDuration returns the time duration between each DNS watchdog loop.
+func (p *prog) dnsWatchdogDuration() time.Duration {
+	if ptr := p.cfg.Service.DnsWatchdogInvterval; ptr != nil {
+		if (*ptr).Seconds() > 0 {
+			return *ptr
+		}
+	}
+	return dnsWatchdogDefaultInterval
+}
+
+// dnsWatchdog watches for DNS changes on Darwin and Windows then re-applying ctrld's settings.
+// This is only works when deactivation pin set.
+func (p *prog) dnsWatchdog(iface *net.Interface, nameservers []string, allIfaces bool) {
+	if !requiredMultiNICsConfig() {
+		return
+	}
+
+	p.dnsWatchDogOnce.Do(func() {
+		mainLog.Load().Debug().Msg("start DNS settings watchdog")
+		ns := nameservers
+		slices.Sort(ns)
+		ticker := time.NewTicker(p.dnsWatchdogDuration())
+		logger := mainLog.Load().With().Str("iface", iface.Name).Logger()
+		for {
+			select {
+			case <-p.dnsWatcherStopCh:
+				return
+			case <-p.stopCh:
+				mainLog.Load().Debug().Msg("stop dns watchdog")
+				return
+			case <-ticker.C:
+				if dnsChanged(iface, ns) {
+					logger.Debug().Msg("DNS settings were changed, re-applying settings")
+					if err := setDNS(iface, ns); err != nil {
+						mainLog.Load().Error().Err(err).Str("iface", iface.Name).Msgf("could not re-apply DNS settings")
+					}
+				}
+				if allIfaces {
+					withEachPhysicalInterfaces(iface.Name, "re-applying DNS", func(i *net.Interface) error {
+						if dnsChanged(i, ns) {
+							if err := setDnsIgnoreUnusableInterface(i, nameservers); err != nil {
+								mainLog.Load().Error().Err(err).Str("iface", i.Name).Msgf("could not re-apply DNS settings")
+							} else {
+								mainLog.Load().Debug().Msgf("re-applying DNS for interface %q successfully", i.Name)
+							}
+						}
+						return nil
+					})
+				}
+			}
+		}
+	})
 }
 
 func (p *prog) resetDNS() {
 	if iface == "" {
 		return
 	}
+	runningIface := iface
 	allIfaces := false
-	if iface == "auto" {
-		iface = defaultIfaceName()
+	if runningIface == "auto" {
+		runningIface = defaultIfaceName()
 		// See corresponding comments in (*prog).setDNS function.
 		allIfaces = requiredMultiNICsConfig()
 	}
-	logger := mainLog.Load().With().Str("iface", iface).Logger()
-	netIface, err := netInterface(iface)
+	logger := mainLog.Load().With().Str("iface", runningIface).Logger()
+	netIface, err := netInterface(runningIface)
 	if err != nil {
 		logger.Error().Err(err).Msg("could not get interface")
 		return
@@ -509,7 +709,7 @@ func (p *prog) resetDNS() {
 	}
 	logger.Debug().Msg("Restoring DNS successfully")
 	if allIfaces {
-		withEachPhysicalInterfaces(netIface.Name, "reset DNS", resetDNS)
+		withEachPhysicalInterfaces(netIface.Name, "reset DNS", resetDnsIgnoreUnusableInterface)
 	}
 }
 
@@ -689,13 +889,14 @@ func canBeLocalUpstream(addr string) bool {
 // the interface that matches excludeIfaceName. The context is used to clarify the
 // log message when error happens.
 func withEachPhysicalInterfaces(excludeIfaceName, context string, f func(i *net.Interface) error) {
+	validIfacesMap := validInterfacesMap()
 	interfaces.ForeachInterface(func(i interfaces.Interface, prefixes []netip.Prefix) {
 		// Skip loopback/virtual interface.
 		if i.IsLoopback() || len(i.HardwareAddr) == 0 {
 			return
 		}
 		// Skip invalid interface.
-		if !validInterface(i.Interface) {
+		if !validInterface(i.Interface, validIfacesMap) {
 			return
 		}
 		netIface := i.Interface
@@ -709,7 +910,9 @@ func withEachPhysicalInterfaces(excludeIfaceName, context string, f func(i *net.
 		}
 		// TODO: investigate whether we should report this error?
 		if err := f(netIface); err == nil {
-			mainLog.Load().Debug().Msgf("%s for interface %q successfully", context, i.Name)
+			if context != "" {
+				mainLog.Load().Debug().Msgf("%s for interface %q successfully", context, i.Name)
+			}
 		} else if !errors.Is(err, errSaveCurrentStaticDNSNotSupported) {
 			mainLog.Load().Err(err).Msgf("%s for interface %q failed", context, i.Name)
 		}
@@ -768,3 +971,24 @@ func savedStaticNameservers(iface *net.Interface) []string {
 	}
 	return nil
 }
+
+// dnsChanged reports whether DNS settings for given interface was changed.
+// The caller must sort the nameservers before calling this function.
+func dnsChanged(iface *net.Interface, nameservers []string) bool {
+	curNameservers, _ := currentStaticDNS(iface)
+	slices.Sort(curNameservers)
+	return !slices.Equal(curNameservers, nameservers)
+}
+
+// selfUninstallCheck checks if the error dues to controld.InvalidConfigCode, perform self-uninstall then.
+func selfUninstallCheck(uninstallErr error, p *prog, logger zerolog.Logger) {
+	var uer *controld.UtilityErrorResponse
+	if errors.As(uninstallErr, &uer) && uer.ErrorField.Code == controld.InvalidConfigCode {
+		// Ensure all DNS watchers goroutine are terminated, so it won't mess up with self-uninstall.
+		close(p.dnsWatcherStopCh)
+		p.dnsWg.Wait()
+
+		// Perform self-uninstall now.
+		selfUninstall(p, logger)
+	}
+}

cmd/cli/prog_linux.go

@@ -1,6 +1,13 @@
 package cli
 
 import (
+	"bufio"
+	"bytes"
+	"io"
+	"os"
+	"os/exec"
+	"strings"
+
 	"github.com/kardianos/service"
 
 	"github.com/Control-D-Inc/ctrld/internal/dns"
@@ -10,6 +17,10 @@ func init() {
 	if r, err := dns.NewOSConfigurator(func(format string, args ...any) {}, "lo"); err == nil {
 		useSystemdResolved = r.Mode() == "systemd-resolved"
 	}
+	// Disable quic-go's ECN support by default, see https://github.com/quic-go/quic-go/issues/3911
+	if os.Getenv("QUIC_GO_DISABLE_ECN") == "" {
+		os.Setenv("QUIC_GO_DISABLE_ECN", "true")
+	}
 }
 
 func setDependencies(svc *service.Config) {
@@ -18,12 +29,34 @@ func setDependencies(svc *service.Config) {
 		"After=network-online.target",
 		"Wants=NetworkManager-wait-online.service",
 		"After=NetworkManager-wait-online.service",
-		"Wants=systemd-networkd-wait-online.service",
 		"Wants=nss-lookup.target",
 		"After=nss-lookup.target",
 	}
+	if out, _ := exec.Command("networkctl", "--no-pager").CombinedOutput(); len(out) > 0 {
+		if wantsSystemDNetworkdWaitOnline(bytes.NewReader(out)) {
+			svc.Dependencies = append(svc.Dependencies, "Wants=systemd-networkd-wait-online.service")
+		}
+	}
 }
 
 func setWorkingDirectory(svc *service.Config, dir string) {
 	svc.WorkingDirectory = dir
 }
+
+// wantsSystemDNetworkdWaitOnline reports whether "systemd-networkd-wait-online" service
+// is required to be added to ctrld dependencies services.
+// The input reader r is the output of "networkctl --no-pager" command.
+func wantsSystemDNetworkdWaitOnline(r io.Reader) bool {
+	scanner := bufio.NewScanner(r)
+	// Skip header
+	scanner.Scan()
+	configured := false
+	for scanner.Scan() {
+		fields := strings.Fields(scanner.Text())
+		if len(fields) > 0 && fields[len(fields)-1] == "configured" {
+			configured = true
+			break
+		}
+	}
+	return configured
+}

cmd/cli/prog_linux_test.go

@@ -0,0 +1,48 @@
+package cli
+
+import (
+	"io"
+	"strings"
+	"testing"
+)
+
+const (
+	networkctlUnmanagedOutput = `IDX LINK            TYPE     OPERATIONAL SETUP    
+  1 lo              loopback carrier     unmanaged
+  2 wlp0s20f3       wlan     routable    unmanaged
+  3 tailscale0      none     routable    unmanaged
+  4 br-9ac33145e060 bridge   no-carrier  unmanaged
+  5 docker0         bridge   no-carrier  unmanaged
+
+5 links listed.
+`
+	networkctlManagedOutput = `IDX LINK            TYPE     OPERATIONAL SETUP    
+  1 lo              loopback carrier     unmanaged
+  2 wlp0s20f3       wlan     routable    configured
+  3 tailscale0      none     routable    unmanaged
+  4 br-9ac33145e060 bridge   no-carrier  unmanaged
+  5 docker0         bridge   no-carrier  unmanaged
+
+5 links listed.
+`
+)
+
+func Test_wantsSystemDNetworkdWaitOnline(t *testing.T) {
+	tests := []struct {
+		name     string
+		r        io.Reader
+		required bool
+	}{
+		{"unmanaged", strings.NewReader(networkctlUnmanagedOutput), false},
+		{"managed", strings.NewReader(networkctlManagedOutput), true},
+		{"empty", strings.NewReader(""), false},
+	}
+	for _, tc := range tests {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			if required := wantsSystemDNetworkdWaitOnline(tc.r); required != tc.required {
+				t.Errorf("wants %v got %v", tc.required, required)
+			}
+		})
+	}
+}

cmd/cli/prog_test.go

@@ -0,0 +1,57 @@
+package cli
+
+import (
+	"testing"
+	"time"
+
+	"github.com/Control-D-Inc/ctrld"
+	"github.com/stretchr/testify/assert"
+)
+
+func Test_prog_dnsWatchdogEnabled(t *testing.T) {
+	p := &prog{cfg: &ctrld.Config{}}
+
+	// Default value is true.
+	assert.True(t, p.dnsWatchdogEnabled())
+
+	tests := []struct {
+		name    string
+		enabled bool
+	}{
+		{"enabled", true},
+		{"disabled", false},
+	}
+
+	for _, tc := range tests {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			p.cfg.Service.DnsWatchdogEnabled = &tc.enabled
+			assert.Equal(t, tc.enabled, p.dnsWatchdogEnabled())
+		})
+	}
+}
+
+func Test_prog_dnsWatchdogInterval(t *testing.T) {
+	p := &prog{cfg: &ctrld.Config{}}
+
+	// Default value is 20s.
+	assert.Equal(t, dnsWatchdogDefaultInterval, p.dnsWatchdogDuration())
+
+	tests := []struct {
+		name     string
+		duration time.Duration
+		expected time.Duration
+	}{
+		{"valid", time.Minute, time.Minute},
+		{"zero", 0, dnsWatchdogDefaultInterval},
+		{"nagative", time.Duration(-1 * time.Minute), dnsWatchdogDefaultInterval},
+	}
+
+	for _, tc := range tests {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			p.cfg.Service.DnsWatchdogInvterval = &tc.duration
+			assert.Equal(t, tc.expected, p.dnsWatchdogDuration())
+		})
+	}
+}

cmd/cli/prometheus.go

@@ -51,7 +51,7 @@ var statsClientQueriesCount = prometheus.NewCounterVec(prometheus.CounterOpts{
 
 // WithLabelValuesInc increases prometheus counter by 1 if query stats is enabled.
 func (p *prog) WithLabelValuesInc(c *prometheus.CounterVec, lvs ...string) {
-	if p.cfg.Service.MetricsQueryStats {
+	if p.metricsQueryStats.Load() {
 		c.WithLabelValues(lvs...).Inc()
 	}
 }

cmd/cli/resolvconf.go

@@ -8,15 +8,15 @@ import (
 	"github.com/fsnotify/fsnotify"
 )
 
-const (
-	resolvConfPath            = "/etc/resolv.conf"
-	resolvConfBackupFailedMsg = "open /etc/resolv.pre-ctrld-backup.conf: read-only file system"
-)
-
 // watchResolvConf watches any changes to /etc/resolv.conf file,
 // and reverting to the original config set by ctrld.
-func watchResolvConf(iface *net.Interface, ns []netip.Addr, setDnsFn func(iface *net.Interface, ns []netip.Addr) error) {
-	mainLog.Load().Debug().Msg("start watching /etc/resolv.conf file")
+func (p *prog) watchResolvConf(iface *net.Interface, ns []netip.Addr, setDnsFn func(iface *net.Interface, ns []netip.Addr) error) {
+	resolvConfPath := "/etc/resolv.conf"
+	// Evaluating symbolics link to watch the target file that /etc/resolv.conf point to.
+	if rp, _ := filepath.EvalSymlinks(resolvConfPath); rp != "" {
+		resolvConfPath = rp
+	}
+	mainLog.Load().Debug().Msgf("start watching %s file", resolvConfPath)
 	watcher, err := fsnotify.NewWatcher()
 	if err != nil {
 		mainLog.Load().Warn().Err(err).Msg("could not create watcher for /etc/resolv.conf")
@@ -28,12 +28,17 @@ func watchResolvConf(iface *net.Interface, ns []netip.Addr, setDnsFn func(iface
 	// see: https://github.com/fsnotify/fsnotify#watching-a-file-doesnt-work-well
 	watchDir := filepath.Dir(resolvConfPath)
 	if err := watcher.Add(watchDir); err != nil {
-		mainLog.Load().Warn().Err(err).Msg("could not add /etc/resolv.conf to watcher list")
+		mainLog.Load().Warn().Err(err).Msgf("could not add %s to watcher list", watchDir)
 		return
 	}
 
 	for {
 		select {
+		case <-p.dnsWatcherStopCh:
+			return
+		case <-p.stopCh:
+			mainLog.Load().Debug().Msgf("stopping watcher for %s", resolvConfPath)
+			return
 		case event, ok := <-watcher.Events:
 			if !ok {
 				return

cmd/cli/resolvconf_darwin.go

@@ -3,15 +3,44 @@ package cli
 import (
 	"net"
 	"net/netip"
+	"os"
+	"slices"
+
+	"github.com/Control-D-Inc/ctrld/internal/dns/resolvconffile"
 )
 
+const resolvConfPath = "/etc/resolv.conf"
+
 // setResolvConf sets the content of resolv.conf file using the given nameservers list.
 func setResolvConf(iface *net.Interface, ns []netip.Addr) error {
 	servers := make([]string, len(ns))
 	for i := range ns {
 		servers[i] = ns[i].String()
 	}
-	return setDNS(iface, servers)
+	if err := setDNS(iface, servers); err != nil {
+		return err
+	}
+	slices.Sort(servers)
+	curNs := currentDNS(iface)
+	slices.Sort(curNs)
+	if !slices.Equal(curNs, servers) {
+		c, err := resolvconffile.ParseFile(resolvConfPath)
+		if err != nil {
+			return err
+		}
+		c.Nameservers = ns
+		f, err := os.Create(resolvConfPath)
+		if err != nil {
+			return err
+		}
+		defer f.Close()
+
+		if err := c.Write(f); err != nil {
+			return err
+		}
+		return f.Close()
+	}
+	return nil
 }
 
 // shouldWatchResolvconf reports whether ctrld should watch changes to resolv.conf file with given OS configurator.

cmd/cli/self_delete_others.go

@@ -0,0 +1,7 @@
+//go:build !windows
+
+package cli
+
+var supportedSelfDelete = true
+
+func selfDeleteExe() error { return nil }

cmd/cli/self_delete_windows.go

@@ -0,0 +1,134 @@
+// Copied from https://github.com/secur30nly/go-self-delete
+// with modification to suitable for ctrld usage.
+
+/*
+   License: MIT Licence
+
+   References:
+       - https://github.com/LloydLabs/delete-self-poc
+       - https://twitter.com/jonasLyk/status/1350401461985955840
+*/
+
+package cli
+
+import (
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var supportedSelfDelete = false
+
+type FILE_RENAME_INFO struct {
+	Union struct {
+		ReplaceIfExists bool
+		Flags           uint32
+	}
+	RootDirectory  windows.Handle
+	FileNameLength uint32
+	FileName       [1]uint16
+}
+
+type FILE_DISPOSITION_INFO struct {
+	DeleteFile bool
+}
+
+func dsOpenHandle(pwPath *uint16) (windows.Handle, error) {
+	handle, err := windows.CreateFile(
+		pwPath,
+		windows.DELETE,
+		0,
+		nil,
+		windows.OPEN_EXISTING,
+		windows.FILE_ATTRIBUTE_NORMAL,
+		0,
+	)
+
+	if err != nil {
+		return 0, err
+	}
+
+	return handle, nil
+}
+
+func dsRenameHandle(hHandle windows.Handle) error {
+	var fRename FILE_RENAME_INFO
+	DS_STREAM_RENAME, err := windows.UTF16FromString(":deadbeef")
+
+	if err != nil {
+		return err
+	}
+
+	lpwStream := &DS_STREAM_RENAME[0]
+	fRename.FileNameLength = uint32(unsafe.Sizeof(lpwStream))
+
+	windows.NewLazyDLL("kernel32.dll").NewProc("RtlCopyMemory").Call(
+		uintptr(unsafe.Pointer(&fRename.FileName[0])),
+		uintptr(unsafe.Pointer(lpwStream)),
+		unsafe.Sizeof(lpwStream),
+	)
+
+	err = windows.SetFileInformationByHandle(
+		hHandle,
+		windows.FileRenameInfo,
+		(*byte)(unsafe.Pointer(&fRename)),
+		uint32(unsafe.Sizeof(fRename)+unsafe.Sizeof(lpwStream)),
+	)
+
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func dsDepositeHandle(hHandle windows.Handle) error {
+	var fDelete FILE_DISPOSITION_INFO
+	fDelete.DeleteFile = true
+
+	err := windows.SetFileInformationByHandle(
+		hHandle,
+		windows.FileDispositionInfo,
+		(*byte)(unsafe.Pointer(&fDelete)),
+		uint32(unsafe.Sizeof(fDelete)),
+	)
+
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func selfDeleteExe() error {
+	var wcPath [windows.MAX_PATH + 1]uint16
+	var hCurrent windows.Handle
+
+	_, err := windows.GetModuleFileName(0, &wcPath[0], windows.MAX_PATH)
+	if err != nil {
+		return err
+	}
+
+	hCurrent, err = dsOpenHandle(&wcPath[0])
+	if err != nil || hCurrent == windows.InvalidHandle {
+		return err
+	}
+
+	if err := dsRenameHandle(hCurrent); err != nil {
+		_ = windows.CloseHandle(hCurrent)
+		return err
+	}
+	_ = windows.CloseHandle(hCurrent)
+
+	hCurrent, err = dsOpenHandle(&wcPath[0])
+	if err != nil || hCurrent == windows.InvalidHandle {
+		return err
+	}
+
+	if err := dsDepositeHandle(hCurrent); err != nil {
+		_ = windows.CloseHandle(hCurrent)
+		return err
+	}
+
+	return windows.CloseHandle(hCurrent)
+}

cmd/cli/self_kill_others.go

@@ -0,0 +1,16 @@
+//go:build !unix
+
+package cli
+
+import (
+	"os"
+
+	"github.com/rs/zerolog"
+)
+
+func selfUninstall(p *prog, logger zerolog.Logger) {
+	if uninstallInvalidCdUID(p, logger, false) {
+		logger.Warn().Msgf("service was uninstalled because device %q does not exist", cdUID)
+		os.Exit(0)
+	}
+}

cmd/cli/self_kill_unix.go

@@ -0,0 +1,45 @@
+//go:build unix
+
+package cli
+
+import (
+	"fmt"
+	"os"
+	"os/exec"
+	"runtime"
+	"syscall"
+
+	"github.com/rs/zerolog"
+)
+
+func selfUninstall(p *prog, logger zerolog.Logger) {
+	if runtime.GOOS == "linux" {
+		selfUninstallLinux(p, logger)
+	}
+
+	bin, err := os.Executable()
+	if err != nil {
+		logger.Fatal().Err(err).Msg("could not determine executable")
+	}
+	args := []string{"uninstall"}
+	if !deactivationPinNotSet() {
+		args = append(args, fmt.Sprintf("--pin=%d", cdDeactivationPin))
+	}
+	cmd := exec.Command(bin, args...)
+	cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
+	if err := cmd.Start(); err != nil {
+		logger.Fatal().Err(err).Msg("could not start self uninstall command")
+	}
+	cmd.Stdout = os.Stdout
+	cmd.Stderr = os.Stderr
+	logger.Warn().Msgf("service was uninstalled because device %q does not exist", cdUID)
+	_ = cmd.Wait()
+	os.Exit(0)
+}
+
+func selfUninstallLinux(p *prog, logger zerolog.Logger) {
+	if uninstallInvalidCdUID(p, logger, true) {
+		logger.Warn().Msgf("service was uninstalled because device %q does not exist", cdUID)
+		os.Exit(0)
+	}
+}

cmd/cli/service.go

@@ -20,14 +20,17 @@ func newService(i service.Interface, c *service.Config) (service.Service, error)
 		return nil, err
 	}
 	switch {
-	case router.IsOldOpenwrt():
-		return &procd{&sysV{s}}, nil
+	case router.IsOldOpenwrt(), router.IsNetGearOrbi():
+		return &procd{sysV: &sysV{s}, svcConfig: c}, nil
 	case router.IsGLiNet():
 		return &sysV{s}, nil
 	case s.Platform() == "unix-systemv":
 		return &sysV{s}, nil
 	case s.Platform() == "linux-systemd":
 		return &systemd{s}, nil
+	case s.Platform() == "darwin-launchd":
+		return newLaunchd(s), nil
+
 	}
 	return s, nil
 }
@@ -89,25 +92,31 @@ func (s *sysV) Status() (service.Status, error) {
 // like old GL.iNET Opal router.
 type procd struct {
 	*sysV
+	svcConfig *service.Config
 }
 
 func (s *procd) Status() (service.Status, error) {
 	if !s.installed() {
 		return service.StatusUnknown, service.ErrNotInstalled
 	}
-	exe, err := os.Executable()
-	if err != nil {
-		return service.StatusUnknown, nil
+	bin := s.svcConfig.Executable
+	if bin == "" {
+		exe, err := os.Executable()
+		if err != nil {
+			return service.StatusUnknown, nil
+		}
+		bin = exe
 	}
+
 	// Looking for something like "/sbin/ctrld run ".
-	shellCmd := fmt.Sprintf("ps | grep -q %q", exe+" [r]un ")
+	shellCmd := fmt.Sprintf("ps | grep -q %q", bin+" [r]un ")
 	if err := exec.Command("sh", "-c", shellCmd).Run(); err != nil {
 		return service.StatusStopped, nil
 	}
 	return service.StatusRunning, nil
 }
 
-// procd wraps a service.Service, and provide status command to
+// systemd wraps a service.Service, and provide status command to
 // report the status correctly.
 type systemd struct {
 	service.Service
@@ -121,6 +130,29 @@ func (s *systemd) Status() (service.Status, error) {
 	return s.Service.Status()
 }
 
+func newLaunchd(s service.Service) *launchd {
+	return &launchd{
+		Service:      s,
+		statusErrMsg: "Permission denied",
+	}
+}
+
+// launchd wraps a service.Service, and provide status command to
+// report the status correctly when not running as root on Darwin.
+//
+// TODO: remove this wrapper once https://github.com/kardianos/service/issues/400 fixed.
+type launchd struct {
+	service.Service
+	statusErrMsg string
+}
+
+func (l *launchd) Status() (service.Status, error) {
+	if os.Geteuid() != 0 {
+		return service.StatusUnknown, errors.New(l.statusErrMsg)
+	}
+	return l.Service.Status()
+}
+
 type task struct {
 	f            func() error
 	abortOnError bool

cmd/cli/service_others.go

@@ -9,3 +9,7 @@ import (
 func hasElevatedPrivilege() (bool, error) {
 	return os.Geteuid() == 0, nil
 }
+
+func openLogFile(path string, flags int) (*os.File, error) {
+	return os.OpenFile(path, flags, os.FileMode(0o600))
+}

cmd/cli/service_windows.go

@@ -1,6 +1,11 @@
 package cli
 
-import "golang.org/x/sys/windows"
+import (
+	"os"
+	"syscall"
+
+	"golang.org/x/sys/windows"
+)
 
 func hasElevatedPrivilege() (bool, error) {
 	var sid *windows.SID
@@ -22,3 +27,55 @@ func hasElevatedPrivilege() (bool, error) {
 	token := windows.Token(0)
 	return token.IsMember(sid)
 }
+
+func openLogFile(path string, mode int) (*os.File, error) {
+	if len(path) == 0 {
+		return nil, &os.PathError{Path: path, Op: "open", Err: syscall.ERROR_FILE_NOT_FOUND}
+	}
+
+	pathP, err := syscall.UTF16PtrFromString(path)
+	if err != nil {
+		return nil, err
+	}
+	var access uint32
+	switch mode & (os.O_RDONLY | os.O_WRONLY | os.O_RDWR) {
+	case os.O_RDONLY:
+		access = windows.GENERIC_READ
+	case os.O_WRONLY:
+		access = windows.GENERIC_WRITE
+	case os.O_RDWR:
+		access = windows.GENERIC_READ | windows.GENERIC_WRITE
+	}
+	if mode&os.O_CREATE != 0 {
+		access |= windows.GENERIC_WRITE
+	}
+	if mode&os.O_APPEND != 0 {
+		access &^= windows.GENERIC_WRITE
+		access |= windows.FILE_APPEND_DATA
+	}
+
+	shareMode := uint32(syscall.FILE_SHARE_READ | syscall.FILE_SHARE_WRITE | syscall.FILE_SHARE_DELETE)
+
+	var sa *syscall.SecurityAttributes
+
+	var createMode uint32
+	switch {
+	case mode&(os.O_CREATE|os.O_EXCL) == (os.O_CREATE | os.O_EXCL):
+		createMode = windows.CREATE_NEW
+	case mode&(os.O_CREATE|os.O_TRUNC) == (os.O_CREATE | os.O_TRUNC):
+		createMode = windows.CREATE_ALWAYS
+	case mode&os.O_CREATE == os.O_CREATE:
+		createMode = windows.OPEN_ALWAYS
+	case mode&os.O_TRUNC == os.O_TRUNC:
+		createMode = windows.TRUNCATE_EXISTING
+	default:
+		createMode = windows.OPEN_EXISTING
+	}
+
+	handle, err := syscall.CreateFile(pathP, access, shareMode, sa, createMode, syscall.FILE_ATTRIBUTE_NORMAL, 0)
+	if err != nil {
+		return nil, &os.PathError{Path: path, Op: "open", Err: err}
+	}
+
+	return os.NewFile(uintptr(handle), path), nil
+}

cmd/ctrld_library/main.go

@@ -61,8 +61,13 @@ func mapCallback(callback AppCallback) cli.AppCallback {
 	}
 }
 
-func (c *Controller) Stop(Pin int64) int {
-	errorCode := cli.CheckDeactivationPin(Pin)
+func (c *Controller) Stop(restart bool, pin int64) int {
+	var errorCode = 0
+	// Force disconnect without checking pin.
+	// In iOS restart is required if vpn detects no connectivity after network change.
+	if !restart {
+		errorCode = cli.CheckDeactivationPin(pin, c.stopCh)
+	}
 	if errorCode == 0 && c.stopCh != nil {
 		close(c.stopCh)
 		c.stopCh = nil

config.go

@@ -25,6 +25,7 @@ import (
 	"github.com/go-playground/validator/v10"
 	"github.com/miekg/dns"
 	"github.com/spf13/viper"
+	"golang.org/x/net/http2"
 	"golang.org/x/sync/singleflight"
 	"tailscale.com/logtail/backoff"
 	"tailscale.com/net/tsaddr"
@@ -46,6 +47,15 @@ const (
 	// depending on the record type of the DNS query.
 	IpStackSplit = "split"
 
+	// FreeDnsDomain is the domain name of free ControlD service.
+	FreeDnsDomain = "freedns.controld.com"
+	// FreeDNSBoostrapIP is the IP address of freedns.controld.com.
+	FreeDNSBoostrapIP = "76.76.2.11"
+	// PremiumDnsDomain is the domain name of premium ControlD service.
+	PremiumDnsDomain = "dns.controld.com"
+	// PremiumDNSBoostrapIP is the IP address of dns.controld.com.
+	PremiumDNSBoostrapIP = "76.76.2.22"
+
 	controlDComDomain = "controld.com"
 	controlDNetDomain = "controld.net"
 	controlDDevDomain = "controld.dev"
@@ -104,14 +114,14 @@ func InitConfig(v *viper.Viper, name string) {
 	})
 	v.SetDefault("upstream", map[string]*UpstreamConfig{
 		"0": {
-			BootstrapIP: "76.76.2.11",
+			BootstrapIP: FreeDNSBoostrapIP,
 			Name:        "Control D - Anti-Malware",
 			Type:        ResolverTypeDOH,
 			Endpoint:    "https://freedns.controld.com/p1",
 			Timeout:     5000,
 		},
 		"1": {
-			BootstrapIP: "76.76.2.11",
+			BootstrapIP: FreeDNSBoostrapIP,
 			Name:        "Control D - No Ads",
 			Type:        ResolverTypeDOQ,
 			Endpoint:    "p2.freedns.controld.com",
@@ -179,26 +189,30 @@ func (c *Config) FirstUpstream() *UpstreamConfig {
 
 // ServiceConfig specifies the general ctrld config.
 type ServiceConfig struct {
-	LogLevel                string `mapstructure:"log_level" toml:"log_level,omitempty"`
-	LogPath                 string `mapstructure:"log_path" toml:"log_path,omitempty"`
-	CacheEnable             bool   `mapstructure:"cache_enable" toml:"cache_enable,omitempty"`
-	CacheSize               int    `mapstructure:"cache_size" toml:"cache_size,omitempty"`
-	CacheTTLOverride        int    `mapstructure:"cache_ttl_override" toml:"cache_ttl_override,omitempty"`
-	CacheServeStale         bool   `mapstructure:"cache_serve_stale" toml:"cache_serve_stale,omitempty"`
-	MaxConcurrentRequests   *int   `mapstructure:"max_concurrent_requests" toml:"max_concurrent_requests,omitempty" validate:"omitempty,gte=0"`
-	DHCPLeaseFile           string `mapstructure:"dhcp_lease_file_path" toml:"dhcp_lease_file_path" validate:"omitempty,file"`
-	DHCPLeaseFileFormat     string `mapstructure:"dhcp_lease_file_format" toml:"dhcp_lease_file_format" validate:"required_unless=DHCPLeaseFile '',omitempty,oneof=dnsmasq isc-dhcp"`
-	DiscoverMDNS            *bool  `mapstructure:"discover_mdns" toml:"discover_mdns,omitempty"`
-	DiscoverARP             *bool  `mapstructure:"discover_arp" toml:"discover_arp,omitempty"`
-	DiscoverDHCP            *bool  `mapstructure:"discover_dhcp" toml:"discover_dhcp,omitempty"`
-	DiscoverPtr             *bool  `mapstructure:"discover_ptr" toml:"discover_ptr,omitempty"`
-	DiscoverHosts           *bool  `mapstructure:"discover_hosts" toml:"discover_hosts,omitempty"`
-	DiscoverRefreshInterval int    `mapstructure:"discover_refresh_interval" toml:"discover_refresh_interval,omitempty"`
-	ClientIDPref            string `mapstructure:"client_id_preference" toml:"client_id_preference,omitempty" validate:"omitempty,oneof=host mac"`
-	MetricsQueryStats       bool   `mapstructure:"metrics_query_stats" toml:"metrics_query_stats,omitempty"`
-	MetricsListener         string `mapstructure:"metrics_listener" toml:"metrics_listener,omitempty"`
-	Daemon                  bool   `mapstructure:"-" toml:"-"`
-	AllocateIP              bool   `mapstructure:"-" toml:"-"`
+	LogLevel                string         `mapstructure:"log_level" toml:"log_level,omitempty"`
+	LogPath                 string         `mapstructure:"log_path" toml:"log_path,omitempty"`
+	CacheEnable             bool           `mapstructure:"cache_enable" toml:"cache_enable,omitempty"`
+	CacheSize               int            `mapstructure:"cache_size" toml:"cache_size,omitempty"`
+	CacheTTLOverride        int            `mapstructure:"cache_ttl_override" toml:"cache_ttl_override,omitempty"`
+	CacheServeStale         bool           `mapstructure:"cache_serve_stale" toml:"cache_serve_stale,omitempty"`
+	CacheFlushDomains       []string       `mapstructure:"cache_flush_domains" toml:"cache_flush_domains" validate:"max=256"`
+	MaxConcurrentRequests   *int           `mapstructure:"max_concurrent_requests" toml:"max_concurrent_requests,omitempty" validate:"omitempty,gte=0"`
+	DHCPLeaseFile           string         `mapstructure:"dhcp_lease_file_path" toml:"dhcp_lease_file_path" validate:"omitempty,file"`
+	DHCPLeaseFileFormat     string         `mapstructure:"dhcp_lease_file_format" toml:"dhcp_lease_file_format" validate:"required_unless=DHCPLeaseFile '',omitempty,oneof=dnsmasq isc-dhcp"`
+	DiscoverMDNS            *bool          `mapstructure:"discover_mdns" toml:"discover_mdns,omitempty"`
+	DiscoverARP             *bool          `mapstructure:"discover_arp" toml:"discover_arp,omitempty"`
+	DiscoverDHCP            *bool          `mapstructure:"discover_dhcp" toml:"discover_dhcp,omitempty"`
+	DiscoverPtr             *bool          `mapstructure:"discover_ptr" toml:"discover_ptr,omitempty"`
+	DiscoverHosts           *bool          `mapstructure:"discover_hosts" toml:"discover_hosts,omitempty"`
+	DiscoverRefreshInterval int            `mapstructure:"discover_refresh_interval" toml:"discover_refresh_interval,omitempty"`
+	ClientIDPref            string         `mapstructure:"client_id_preference" toml:"client_id_preference,omitempty" validate:"omitempty,oneof=host mac"`
+	MetricsQueryStats       bool           `mapstructure:"metrics_query_stats" toml:"metrics_query_stats,omitempty"`
+	MetricsListener         string         `mapstructure:"metrics_listener" toml:"metrics_listener,omitempty"`
+	DnsWatchdogEnabled      *bool          `mapstructure:"dns_watchdog_enabled" toml:"dns_watchdog_enabled,omitempty"`
+	DnsWatchdogInvterval    *time.Duration `mapstructure:"dns_watchdog_interval" toml:"dns_watchdog_interval,omitempty"`
+	RefetchTime             *int           `mapstructure:"refetch_time" toml:"refetch_time,omitempty"`
+	Daemon                  bool           `mapstructure:"-" toml:"-"`
+	AllocateIP              bool           `mapstructure:"-" toml:"-"`
 }
 
 // NetworkConfig specifies configuration for networks where ctrld will handle requests.
@@ -285,6 +299,7 @@ type Rule map[string][]string
 
 // Init initialized necessary values for an UpstreamConfig.
 func (uc *UpstreamConfig) Init() {
+	uc.initDoHScheme()
 	uc.uid = upstreamUID()
 	if u, err := url.Parse(uc.Endpoint); err == nil {
 		uc.Domain = u.Host
@@ -305,7 +320,7 @@ func (uc *UpstreamConfig) Init() {
 		}
 	}
 	if uc.IPStack == "" {
-		if uc.isControlD() {
+		if uc.IsControlD() {
 			uc.IPStack = IpStackSplit
 		} else {
 			uc.IPStack = IpStackBoth
@@ -343,7 +358,7 @@ func (uc *UpstreamConfig) UpstreamSendClientInfo() bool {
 	}
 	switch uc.Type {
 	case ResolverTypeDOH, ResolverTypeDOH3:
-		if uc.isControlD() || uc.isNextDNS() {
+		if uc.IsControlD() || uc.isNextDNS() {
 			return true
 		}
 	}
@@ -390,7 +405,7 @@ func (uc *UpstreamConfig) UID() string {
 // The first usable IP will be used as bootstrap IP of the upstream.
 func (uc *UpstreamConfig) setupBootstrapIP(withBootstrapDNS bool) {
 	b := backoff.NewBackoff("setupBootstrapIP", func(format string, args ...any) {}, 10*time.Second)
-	isControlD := uc.isControlD()
+	isControlD := uc.IsControlD()
 	for {
 		uc.bootstrapIPs = lookupIP(uc.Domain, uc.Timeout, withBootstrapDNS)
 		// For ControlD upstream, the bootstrap IPs could not be RFC 1918 addresses,
@@ -475,6 +490,13 @@ func (uc *UpstreamConfig) newDOHTransport(addrs []string) *http.Transport {
 		ClientSessionCache: tls.NewLRUClientSessionCache(0),
 	}
 
+	// Prevent bad tcp connection hanging the requests for too long.
+	// See: https://github.com/golang/go/issues/36026
+	if t2, err := http2.ConfigureTransports(transport); err == nil {
+		t2.ReadIdleTimeout = 10 * time.Second
+		t2.PingTimeout = 5 * time.Second
+	}
+
 	dialerTimeoutMs := 2000
 	if uc.Timeout > 0 && uc.Timeout < dialerTimeoutMs {
 		dialerTimeoutMs = uc.Timeout
@@ -510,38 +532,59 @@ func (uc *UpstreamConfig) newDOHTransport(addrs []string) *http.Transport {
 
 // Ping warms up the connection to DoH/DoH3 upstream.
 func (uc *UpstreamConfig) Ping() {
+	_ = uc.ping()
+}
+
+// ErrorPing is like Ping, but return an error if any.
+func (uc *UpstreamConfig) ErrorPing() error {
+	return uc.ping()
+}
+
+func (uc *UpstreamConfig) ping() error {
 	switch uc.Type {
 	case ResolverTypeDOH, ResolverTypeDOH3:
 	default:
-		return
+		return nil
 	}
 
-	ping := func(t http.RoundTripper) {
+	ping := func(t http.RoundTripper) error {
 		if t == nil {
-			return
+			return nil
 		}
 		ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
 		defer cancel()
-		req, _ := http.NewRequestWithContext(ctx, "HEAD", uc.Endpoint, nil)
-		resp, _ := t.RoundTrip(req)
-		if resp == nil {
-			return
+		req, err := http.NewRequestWithContext(ctx, "HEAD", uc.Endpoint, nil)
+		if err != nil {
+			return err
+		}
+		resp, err := t.RoundTrip(req)
+		if err != nil {
+			return err
 		}
 		defer resp.Body.Close()
 		_, _ = io.Copy(io.Discard, resp.Body)
+		return nil
 	}
 
 	for _, typ := range []uint16{dns.TypeA, dns.TypeAAAA} {
 		switch uc.Type {
 		case ResolverTypeDOH:
-			ping(uc.dohTransport(typ))
+
+			if err := ping(uc.dohTransport(typ)); err != nil {
+				return err
+			}
 		case ResolverTypeDOH3:
-			ping(uc.doh3Transport(typ))
+			if err := ping(uc.doh3Transport(typ)); err != nil {
+				return err
+			}
 		}
 	}
+
+	return nil
 }
 
-func (uc *UpstreamConfig) isControlD() bool {
+// IsControlD reports whether this is a ControlD upstream.
+func (uc *UpstreamConfig) IsControlD() bool {
 	domain := uc.Domain
 	if domain == "" {
 		if u, err := url.Parse(uc.Endpoint); err == nil {
@@ -631,6 +674,18 @@ func (uc *UpstreamConfig) netForDNSType(dnsType uint16) (string, string) {
 	return "tcp-tls", "udp"
 }
 
+// initDoHScheme initializes the endpoint scheme for DoH/DoH3 upstream if not present.
+func (uc *UpstreamConfig) initDoHScheme() {
+	switch uc.Type {
+	case ResolverTypeDOH, ResolverTypeDOH3:
+	default:
+		return
+	}
+	if !strings.HasPrefix(uc.Endpoint, "https://") {
+		uc.Endpoint = "https://" + uc.Endpoint
+	}
+}
+
 // Init initialized necessary values for an ListenerConfig.
 func (lc *ListenerConfig) Init() {
 	if lc.Policy != nil {
@@ -683,6 +738,7 @@ func upstreamConfigStructLevelValidation(sl validator.StructLevel) {
 		return
 	}
 
+	uc.initDoHScheme()
 	// DoH/DoH3 requires endpoint is an HTTP url.
 	if uc.Type == ResolverTypeDOH || uc.Type == ResolverTypeDOH3 {
 		u, err := url.Parse(uc.Endpoint)
@@ -690,10 +746,6 @@ func upstreamConfigStructLevelValidation(sl validator.StructLevel) {
 			sl.ReportError(uc.Endpoint, "endpoint", "Endpoint", "http_url", "")
 			return
 		}
-		if u.Scheme != "http" && u.Scheme != "https" {
-			sl.ReportError(uc.Endpoint, "endpoint", "Endpoint", "http_url", "")
-			return
-		}
 	}
 }
 

config_test.go

@@ -1,9 +1,11 @@
 package ctrld_test
 
 import (
+	"fmt"
 	"os"
 	"strings"
 	"testing"
+	"time"
 
 	"github.com/go-playground/validator/v10"
 	"github.com/spf13/viper"
@@ -21,6 +23,8 @@ func TestLoadConfig(t *testing.T) {
 
 	assert.Equal(t, "info", cfg.Service.LogLevel)
 	assert.Equal(t, "/path/to/log.log", cfg.Service.LogPath)
+	assert.Equal(t, false, *cfg.Service.DnsWatchdogEnabled)
+	assert.Equal(t, time.Duration(20*time.Second), *cfg.Service.DnsWatchdogInvterval)
 
 	assert.Len(t, cfg.Network, 2)
 	assert.Contains(t, cfg.Network, "0")
@@ -102,6 +106,8 @@ func TestConfigValidation(t *testing.T) {
 		{"invalid lease file format", configWithInvalidLeaseFileFormat(t), true},
 		{"invalid doh/doh3 endpoint", configWithInvalidDoHEndpoint(t), true},
 		{"invalid client id pref", configWithInvalidClientIDPref(t), true},
+		{"doh endpoint without scheme", dohUpstreamEndpointWithoutScheme(t), false},
+		{"maximum number of flush cache domains", configWithInvalidFlushCacheDomain(t), true},
 	}
 
 	for _, tc := range tests {
@@ -167,6 +173,12 @@ func invalidUpstreamType(t *testing.T) *ctrld.Config {
 	return cfg
 }
 
+func dohUpstreamEndpointWithoutScheme(t *testing.T) *ctrld.Config {
+	cfg := defaultConfig(t)
+	cfg.Upstream["0"].Endpoint = "freedns.controld.com/p1"
+	return cfg
+}
+
 func invalidUpstreamTimeout(t *testing.T) *ctrld.Config {
 	cfg := defaultConfig(t)
 	cfg.Upstream["0"].Timeout = -1
@@ -258,7 +270,7 @@ func configWithInvalidLeaseFileFormat(t *testing.T) *ctrld.Config {
 
 func configWithInvalidDoHEndpoint(t *testing.T) *ctrld.Config {
 	cfg := defaultConfig(t)
-	cfg.Upstream["0"].Endpoint = "1.1.1.1"
+	cfg.Upstream["0"].Endpoint = "/1.1.1.1"
 	cfg.Upstream["0"].Type = ctrld.ResolverTypeDOH
 	return cfg
 }
@@ -268,3 +280,12 @@ func configWithInvalidClientIDPref(t *testing.T) *ctrld.Config {
 	cfg.Service.ClientIDPref = "foo"
 	return cfg
 }
+
+func configWithInvalidFlushCacheDomain(t *testing.T) *ctrld.Config {
+	cfg := defaultConfig(t)
+	cfg.Service.CacheFlushDomains = make([]string, 257)
+	for i := range cfg.Service.CacheFlushDomains {
+		cfg.Service.CacheFlushDomains[i] = fmt.Sprintf("%d.com", i)
+	}
+	return cfg
+}

docker/Dockerfile.debug

@@ -8,7 +8,7 @@
 # - Non-cgo ctrld binary.
 #
 # CI_COMMIT_TAG is used to set the version of ctrld binary.
-FROM golang:1.20-bullseye as base
+FROM golang:bullseye as base
 
 WORKDIR /app
 

docs/config.md

@@ -14,7 +14,7 @@ The config file allows for advanced configuration of the `ctrld` utility to cove
 
 
 ## Config Location
-`ctrld` uses [TOML](toml_link) format for its configuration file. Default configuration file is `ctrld.toml` found in following order:
+`ctrld` uses [TOML][toml_link] format for its configuration file. Default configuration file is `ctrld.toml` found in following order:
 
  - `/etc/controld` on *nix.
  - User's home directory on Windows.
@@ -157,6 +157,13 @@ stale cached records (regardless of their TTLs) until upstream comes online.
 - Required: no
 - Default: false
 
+### cache_flush_domains
+When `ctrld` receives query with domain name in `cache_flush_domains`, the local cache will be discarded
+before serving the query.
+
+- Type: array of strings
+- Required: no
+
 ### max_concurrent_requests
 The number of concurrent requests that will be handled, must be a non-negative integer. 
 Tweaking this value depends on the capacity of your system.
@@ -220,7 +227,7 @@ DHCP leases file format.
 
 - Type: string
 - Required: no
-- Valid values: `dnsmasq`, `isc-dhcp`
+- Valid values: `dnsmasq`, `isc-dhcp`, `kea-dhcp4`
 - Default: ""
 
 ### client_id_preference
@@ -245,6 +252,35 @@ Specifying the `ip` and `port` of the Prometheus metrics server. The Prometheus
 - Required: no
 - Default: ""
 
+### dns_watchdog_enabled
+Checking DNS changes to network interfaces and reverting to ctrld's own settings.
+
+The DNS watchdog process only runs on Windows and MacOS.
+
+- Type: boolean
+- Required: no
+- Default: true
+
+### dns_watchdog_interval
+Time duration between each DNS watchdog iteration.
+
+A duration string is a possibly signed sequence of decimal numbers, each with optional fraction and a unit suffix,
+such as "300ms", "-1.5h" or "2h45m". Valid time units are "ns", "us" (or "µs"), "ms", "s", "m", "h".
+
+If the time duration is non-positive, default value will be used.
+
+- Type: time duration string
+- Required: no
+- Default: 20s
+
+### refetch_time
+Time in seconds between each iteration that reloads custom config if changed.
+
+The value must be a positive number, any invalid value will be ignored and default value will be used.
+- Type: number
+- Required: no
+- Default: 3600
+
 ## Upstream
 The `[upstream]` section specifies the DNS upstream servers that `ctrld` will forward DNS requests to.
 
@@ -329,7 +365,7 @@ The protocol that `ctrld` will use to send DNS requests to upstream.
 
  - Type: string
  - Required: yes
- - Valid values: `doh`, `doh3`, `dot`, `doq`, `legacy`, `os`
+ - Valid values: `doh`, `doh3`, `dot`, `doq`, `legacy`
 
 ### ip_stack
 Specifying what kind of ip stack that `ctrld` will use to connect to upstream.
@@ -531,7 +567,7 @@ And within each policy, the rules are processed from top to bottom.
 ### failover_rcodes
 For non success response, `failover_rcodes` allows the request to be forwarded to next upstream, if the response `RCODE` matches any value defined in `failover_rcodes`.
 
-- Type: array of string
+- Type: array of strings
 - Required: no
 - Default: []
 - 
@@ -548,7 +584,7 @@ networks = [
 
 If `upstream.0` returns a NXDOMAIN response, the request will be forwarded to `upstream.1` instead of returning immediately to the client.
 
-See all available DNS Rcodes value [here](rcode_link).
+See all available DNS Rcodes value [here][rcode_link].
 
 [toml_link]: https://toml.io/en
 [rcode_link]: https://www.iana.org/assignments/dns-parameters/dns-parameters.xhtml#dns-parameters-6

doh.go

@@ -60,17 +60,10 @@ func init() {
 	}
 }
 
-// TODO: use sync.OnceValue when upgrading to go1.21
-var xCdOsValueOnce sync.Once
-var xCdOsValue string
-
-func dohOsHeaderValue() string {
-	xCdOsValueOnce.Do(func() {
-		oi := osinfo.New()
-		xCdOsValue = strings.Join([]string{EncodeOsNameMap[runtime.GOOS], EncodeArchNameMap[runtime.GOARCH], oi.Dist}, "-")
-	})
-	return xCdOsValue
-}
+var dohOsHeaderValue = sync.OnceValue(func() string {
+	oi := osinfo.New()
+	return strings.Join([]string{EncodeOsNameMap[runtime.GOOS], EncodeArchNameMap[runtime.GOARCH], oi.Dist}, "-")
+})()
 
 func newDohResolver(uc *UpstreamConfig) *dohResolver {
 	r := &dohResolver{
@@ -154,7 +147,7 @@ func addHeader(ctx context.Context, req *http.Request, uc *UpstreamConfig) {
 		if ci, ok := ctx.Value(ClientInfoCtxKey{}).(*ClientInfo); ok && ci != nil {
 			printed = ci.Mac != "" || ci.IP != "" || ci.Hostname != ""
 			switch {
-			case uc.isControlD():
+			case uc.IsControlD():
 				dohHeader = newControlDHeaders(ci)
 			case uc.isNextDNS():
 				dohHeader = newNextDNSHeaders(ci)
@@ -172,7 +165,6 @@ func addHeader(ctx context.Context, req *http.Request, uc *UpstreamConfig) {
 // newControlDHeaders returns DoH/Doh3 HTTP request headers for ControlD upstream.
 func newControlDHeaders(ci *ClientInfo) http.Header {
 	header := make(http.Header)
-	header.Set(dohOsHeader, dohOsHeaderValue())
 	if ci.Mac != "" {
 		header.Set(dohMacHeader, ci.Mac)
 	}
@@ -183,7 +175,7 @@ func newControlDHeaders(ci *ClientInfo) http.Header {
 		header.Set(dohHostHeader, ci.Hostname)
 	}
 	if ci.Self {
-		header.Set(dohOsHeader, dohOsHeaderValue())
+		header.Set(dohOsHeader, dohOsHeaderValue)
 	}
 	switch ci.ClientIDPref {
 	case "mac":

doh_test.go

@@ -6,7 +6,7 @@ import (
 )
 
 func Test_dohOsHeaderValue(t *testing.T) {
-	val := dohOsHeaderValue()
+	val := dohOsHeaderValue
 	if val == "" {
 		t.Fatalf("empty %s", dohOsHeader)
 	}

doq_quic_free.go

@@ -1,18 +0,0 @@
-//go:build qf
-
-package ctrld
-
-import (
-	"context"
-	"errors"
-
-	"github.com/miekg/dns"
-)
-
-type doqResolver struct {
-	uc *UpstreamConfig
-}
-
-func (r *doqResolver) Resolve(ctx context.Context, msg *dns.Msg) (*dns.Msg, error) {
-	return nil, errors.New("DoQ is not supported")
-}

dot.go

@@ -18,7 +18,7 @@ func (r *dotResolver) Resolve(ctx context.Context, msg *dns.Msg) (*dns.Msg, erro
 	// dns.controld.dev first. By using a dialer with custom resolver,
 	// we ensure that we can always resolve the bootstrap domain
 	// regardless of the machine DNS status.
-	dialer := newDialer(net.JoinHostPort(bootstrapDNS, "53"))
+	dialer := newDialer(net.JoinHostPort(controldBootstrapDns, "53"))
 	dnsTyp := uint16(0)
 	if msg != nil && len(msg.Question) > 0 {
 		dnsTyp = msg.Question[0].Qtype

go.mod

@@ -3,6 +3,7 @@ module github.com/Control-D-Inc/ctrld
 go 1.21
 
 require (
+	github.com/Masterminds/semver v1.5.0
 	github.com/coreos/go-systemd/v22 v22.5.0
 	github.com/cuonglm/osinfo v0.0.0-20230921071424-e0e1b1e0bbbf
 	github.com/frankban/quicktest v1.14.5
@@ -17,26 +18,28 @@ require (
 	github.com/kardianos/service v1.2.1
 	github.com/mdlayher/ndp v1.0.1
 	github.com/miekg/dns v1.1.55
+	github.com/minio/selfupdate v0.6.0
 	github.com/olekukonko/tablewriter v0.0.5
 	github.com/pelletier/go-toml/v2 v2.0.8
 	github.com/prometheus/client_golang v1.15.1
 	github.com/prometheus/client_model v0.4.0
 	github.com/prometheus/prom2json v1.3.3
-	github.com/quic-go/quic-go v0.41.0
+	github.com/quic-go/quic-go v0.42.0
 	github.com/rs/zerolog v1.28.0
 	github.com/spf13/cobra v1.7.0
 	github.com/spf13/pflag v1.0.5
 	github.com/spf13/viper v1.16.0
 	github.com/stretchr/testify v1.8.3
 	github.com/vishvananda/netlink v1.2.1-beta.2
-	golang.org/x/net v0.17.0
+	golang.org/x/net v0.23.0
 	golang.org/x/sync v0.2.0
-	golang.org/x/sys v0.13.0
+	golang.org/x/sys v0.18.0
 	golang.zx2c4.com/wireguard/windows v0.5.3
 	tailscale.com v1.44.0
 )
 
 require (
+	aead.dev/minisign v0.2.0 // indirect
 	github.com/alexbrainman/sspi v0.0.0-20210105120005-909beea2cc74 // indirect
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/cespare/xxhash/v2 v2.2.0 // indirect
@@ -77,14 +80,14 @@ require (
 	github.com/subosito/gotenv v1.4.2 // indirect
 	github.com/u-root/uio v0.0.0-20230305220412-3e8cd9d6bf63 // indirect
 	github.com/vishvananda/netns v0.0.4 // indirect
-	go.uber.org/mock v0.3.0 // indirect
+	go.uber.org/mock v0.4.0 // indirect
 	go4.org/mem v0.0.0-20220726221520-4f986261bf13 // indirect
-	golang.org/x/crypto v0.14.0 // indirect
+	golang.org/x/crypto v0.21.0 // indirect
 	golang.org/x/exp v0.0.0-20230425010034-47ecfdc1ba53 // indirect
 	golang.org/x/mod v0.11.0 // indirect
-	golang.org/x/text v0.13.0 // indirect
+	golang.org/x/text v0.14.0 // indirect
 	golang.org/x/tools v0.9.1 // indirect
-	google.golang.org/protobuf v1.30.0 // indirect
+	google.golang.org/protobuf v1.33.0 // indirect
 	gopkg.in/ini.v1 v1.67.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -1,3 +1,5 @@
+aead.dev/minisign v0.2.0 h1:kAWrq/hBRu4AARY6AlciO83xhNnW9UaC8YipS2uhLPk=
+aead.dev/minisign v0.2.0/go.mod h1:zdq6LdSd9TbuSxchxwhpA9zEb9YXcVGoE8JakuiGaIQ=
 cloud.google.com/go v0.26.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
 cloud.google.com/go v0.34.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
 cloud.google.com/go v0.38.0/go.mod h1:990N+gfupTy94rShfmMCWGDn0LpTmnzTp2qbd1dvSRU=
@@ -38,6 +40,8 @@ cloud.google.com/go/storage v1.14.0/go.mod h1:GrKmX003DSIwi9o29oFT7YDnHYwZoctc3f
 dmitri.shuralyov.com/gpu/mtl v0.0.0-20190408044501-666a987793e9/go.mod h1:H6x//7gZCb22OMCxBHrMx7a5I7Hp++hsVxbQ4BYO7hU=
 github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
 github.com/BurntSushi/xgb v0.0.0-20160522181843-27f122750802/go.mod h1:IVnqGOEym/WlBOVXweHU+Q+/VP0lqqI8lqeDx9IjBqo=
+github.com/Masterminds/semver v1.5.0 h1:H65muMkzWKEuNDnfl9d70GUjFniHKHRbFPGBuZ3QEww=
+github.com/Masterminds/semver v1.5.0/go.mod h1:MB6lktGJrhw8PrUyiEoblNEGEQ+RzHPF078ddwwvV3Y=
 github.com/Windscribe/zerolog v0.0.0-20230503170159-e6aa153233be h1:qBKVRi7Mom5heOkyZ+NCIu9HZBiNCsRqrRe5t9pooik=
 github.com/Windscribe/zerolog v0.0.0-20230503170159-e6aa153233be/go.mod h1:/tk+P47gFdPXq4QYjvCmT5/Gsug2nagsFWBWhAiSi1w=
 github.com/alexbrainman/sspi v0.0.0-20210105120005-909beea2cc74 h1:Kk6a4nehpJ3UuJRqlA3JxYxBZEqCeOmATOvrbT4p9RA=
@@ -222,6 +226,8 @@ github.com/mdlayher/socket v0.4.1 h1:eM9y2/jlbs1M615oshPQOHZzj6R6wMT7bX5NPiQvn2U
 github.com/mdlayher/socket v0.4.1/go.mod h1:cAqeGjoufqdxWkD7DkpyS+wcefOtmu5OQ8KuoJGIReA=
 github.com/miekg/dns v1.1.55 h1:GoQ4hpsj0nFLYe+bWiCToyrBEJXkQfOOIvFGFy0lEgo=
 github.com/miekg/dns v1.1.55/go.mod h1:uInx36IzPl7FYnDcMeVWxj9byh7DutNykX4G9Sj60FY=
+github.com/minio/selfupdate v0.6.0 h1:i76PgT0K5xO9+hjzKcacQtO7+MjJ4JKA8Ak8XQ9DDwU=
+github.com/minio/selfupdate v0.6.0/go.mod h1:bO02GTIPCMQFTEvE5h4DjYB58bCoZ35XLeBf0buTDdM=
 github.com/mitchellh/mapstructure v1.5.0 h1:jeMsZIYE/09sWLaz43PL7Gy6RuMjD2eJVyuac5Z2hdY=
 github.com/mitchellh/mapstructure v1.5.0/go.mod h1:bFUtVrKA4DC2yAKiSyO/QUcy7e+RRV2QTWOzhPopBRo=
 github.com/olekukonko/tablewriter v0.0.5 h1:P2Ga83D34wi1o9J6Wh1mRuqd4mF/x/lgBS7N7AbDhec=
@@ -253,8 +259,8 @@ github.com/prometheus/prom2json v1.3.3 h1:IYfSMiZ7sSOfliBoo89PcufjWO4eAR0gznGcET
 github.com/prometheus/prom2json v1.3.3/go.mod h1:Pv4yIPktEkK7btWsrUTWDDDrnpUrAELaOCj+oFwlgmc=
 github.com/quic-go/qpack v0.4.0 h1:Cr9BXA1sQS2SmDUWjSofMPNKmvF6IiIfDRmgU0w1ZCo=
 github.com/quic-go/qpack v0.4.0/go.mod h1:UZVnYIfi5GRk+zI9UMaCPsmZ2xKJP7XBUvVyT1Knj9A=
-github.com/quic-go/quic-go v0.41.0 h1:aD8MmHfgqTURWNJy48IYFg2OnxwHT3JL7ahGs73lb4k=
-github.com/quic-go/quic-go v0.41.0/go.mod h1:qCkNjqczPEvgsOnxZ0eCD14lv+B2LHlFAB++CNOh9hA=
+github.com/quic-go/quic-go v0.42.0 h1:uSfdap0eveIl8KXnipv9K7nlwZ5IqLlYOpJ58u5utpM=
+github.com/quic-go/quic-go v0.42.0/go.mod h1:132kz4kL3F9vxhW3CtQJLDVwcFe5wdWeJXXijhsO57M=
 github.com/rivo/uniseg v0.2.0/go.mod h1:J6wj4VEh+S6ZtnVlnTBMWIodfgj8LQOQFoIToxlJtxc=
 github.com/rivo/uniseg v0.4.4 h1:8TfxU8dW6PdqD27gjM8MVNuicgxIjxpm4K7x4jp8sis=
 github.com/rivo/uniseg v0.4.4/go.mod h1:FN3SvrM+Zdj16jyLfmOkMNblXMcoc8DfTHruCPUcx88=
@@ -310,8 +316,8 @@ go.opencensus.io v0.22.2/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
 go.opencensus.io v0.22.3/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
 go.opencensus.io v0.22.4/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
 go.opencensus.io v0.22.5/go.mod h1:5pWMHQbX5EPX2/62yrJeAkowc+lfs/XD7Uxpq3pI6kk=
-go.uber.org/mock v0.3.0 h1:3mUxI1No2/60yUYax92Pt8eNOEecx2D3lcXZh2NEZJo=
-go.uber.org/mock v0.3.0/go.mod h1:a6FSlNadKUHUa9IP5Vyt1zh4fC7uAwxMutEAscFbkZc=
+go.uber.org/mock v0.4.0 h1:VcM4ZOtdbR4f6VXfiOpwpVJDL6lCReaZ6mw31wqh7KU=
+go.uber.org/mock v0.4.0/go.mod h1:a6FSlNadKUHUa9IP5Vyt1zh4fC7uAwxMutEAscFbkZc=
 go4.org/mem v0.0.0-20220726221520-4f986261bf13 h1:CbZeCBZ0aZj8EfVgnqQcYZgf0lpZ3H9rmp5nkDTAst8=
 go4.org/mem v0.0.0-20220726221520-4f986261bf13/go.mod h1:reUoABIJ9ikfM5sgtSF3Wushcza7+WeD01VB9Lirh3g=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
@@ -319,11 +325,13 @@ golang.org/x/crypto v0.0.0-20190510104115-cbcb75029529/go.mod h1:yigFU9vqHzYiE8U
 golang.org/x/crypto v0.0.0-20190605123033-f99c8df09eb5/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
+golang.org/x/crypto v0.0.0-20210220033148-5ea612d1eb83/go.mod h1:jdWPYTVW3xRLrWPugEBEK3UY2ZEsg3UU495nc5E+M+I=
 golang.org/x/crypto v0.0.0-20210421170649-83a5a9bb288b/go.mod h1:T9bdIzuCu7OtxOm1hfPfRQxPLYneinmdGuTeoZ9dtd4=
+golang.org/x/crypto v0.0.0-20211209193657-4570a0811e8b/go.mod h1:IxCIyHEi3zRg3s0A5j5BB6A9Jmi73HwBIUl50j+osU4=
 golang.org/x/crypto v0.0.0-20211215153901-e495a2d5b3d3/go.mod h1:IxCIyHEi3zRg3s0A5j5BB6A9Jmi73HwBIUl50j+osU4=
 golang.org/x/crypto v0.0.0-20220722155217-630584e8d5aa/go.mod h1:IxCIyHEi3zRg3s0A5j5BB6A9Jmi73HwBIUl50j+osU4=
-golang.org/x/crypto v0.14.0 h1:wBqGXzWJW6m1XrIKlAH0Hs1JJ7+9KBwnIO8v66Q9cHc=
-golang.org/x/crypto v0.14.0/go.mod h1:MVFd36DqK4CsrnJYDkBA3VC4m2GkXAM0PvzMCn4JQf4=
+golang.org/x/crypto v0.21.0 h1:X31++rzVUdKhX5sWmSOFZxx8UW/ldWx55cbf08iNAMA=
+golang.org/x/crypto v0.21.0/go.mod h1:0BP7YvVV9gBbVKyeTG0Gyn+gZm94bibOW5BjDEYAOMs=
 golang.org/x/exp v0.0.0-20190121172915-509febef88a4/go.mod h1:CJ0aWSM057203Lf6IL+f9T1iT9GByDxfZKAQTCR3kQA=
 golang.org/x/exp v0.0.0-20190306152737-a1d7652674e8/go.mod h1:CJ0aWSM057203Lf6IL+f9T1iT9GByDxfZKAQTCR3kQA=
 golang.org/x/exp v0.0.0-20190510132918-efd6b22b2522/go.mod h1:ZjyILWgesfNpC6sMxTJOJm9Kp84zZh5NQWvqDGG3Qr8=
@@ -394,8 +402,8 @@ golang.org/x/net v0.0.0-20201209123823-ac852fbbde11/go.mod h1:m0MpNAwzfU5UDzcl9v
 golang.org/x/net v0.0.0-20201224014010-6772e930b67b/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
 golang.org/x/net v0.0.0-20211112202133-69e39bad7dc2/go.mod h1:9nx3DQGgdP8bBQD5qxJ1jj9UTztislL4KSBs9R2vV5Y=
-golang.org/x/net v0.17.0 h1:pVaXccu2ozPjCXewfr1S7xza/zcXTity9cCdXQYSjIM=
-golang.org/x/net v0.17.0/go.mod h1:NxSsAGuq816PNPmqtQdLE42eU2Fs7NoRIZrHJAlaCOE=
+golang.org/x/net v0.23.0 h1:7EYJ93RZ9vYSZAIb2x3lnuvqO5zneoD6IvWjuhfxjTs=
+golang.org/x/net v0.23.0/go.mod h1:JKghWKKOSdJwpW2GEx0Ja7fmaKnMsbu+MWVZTokSYmg=
 golang.org/x/oauth2 v0.0.0-20180821212333-d2e6202438be/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
 golang.org/x/oauth2 v0.0.0-20190226205417-e64efc72b421/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/oauth2 v0.0.0-20190604053449-0f29369cfe45/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
@@ -429,6 +437,7 @@ golang.org/x/sys v0.0.0-20190606165138-5da285871e9c/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20190624142023-c5567b49c5d0/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20190726091711-fc99dfbffb4e/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191001151750-bb3f8db39f24/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20191026070338-33540a1f6037/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191204072324-ce4227a45e2e/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191228213918-04cbcbbfeed8/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200113162924-86b910548bc1/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
@@ -454,6 +463,7 @@ golang.org/x/sys v0.0.0-20201201145000-ef89a241ccb3/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20210104204734-6f8348627aad/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210119212857-b64e53b001e4/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210225134936-a50acf3fe073/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20210228012217-479acdf4ea46/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210423082822-04245dca01da/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210423185535-09eb48e85fd7/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210615035016-665e8c7367d1/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
@@ -465,8 +475,9 @@ golang.org/x/sys v0.0.0-20220811171246-fbc7d0a398ab/go.mod h1:oPkhp1MJrh7nUepCBc
 golang.org/x/sys v0.0.0-20220908164124-27713097b956/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.4.1-0.20230131160137-e7d7f63158de/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.13.0 h1:Af8nKPmuFypiUBjVoU9V20FiaFXOcuZI21p0ycVYYGE=
-golang.org/x/sys v0.13.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
+golang.org/x/sys v0.18.0 h1:DBdB3niSjOA/O0blCZBqDefyWNYveAYMNF1Wum0DYQ4=
+golang.org/x/sys v0.18.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/term v0.0.0-20201117132131-f5c789dd3221/go.mod h1:Nr5EML6q2oocZ2LXRh80K7BxOlk5/8JxuGnuhpl+muw=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
 golang.org/x/text v0.0.0-20170915032832-14c0d48ead0c/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
@@ -476,11 +487,13 @@ golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/text v0.3.4/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/text v0.3.6/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/text v0.3.7/go.mod h1:u+2+/6zg+i71rQMx5EYifcz6MCKuco9NR6JIITiCfzQ=
-golang.org/x/text v0.13.0 h1:ablQoSUd0tRdKxZewP80B+BaqeKJuVhuRxj/dkrun3k=
-golang.org/x/text v0.13.0/go.mod h1:TvPlkZtksWOMsz7fbANvkp4WM8x/WCo/om8BMLbz+aE=
+golang.org/x/text v0.14.0 h1:ScX5w1eTa3QqT8oi6+ziP7dTV1S2+ALU0bI+0zXKWiQ=
+golang.org/x/text v0.14.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
 golang.org/x/time v0.0.0-20181108054448-85acf8d2951c/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/time v0.0.0-20190308202827-9d24e82272b4/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/time v0.0.0-20191024005414-555d28b269f0/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
+golang.org/x/time v0.5.0 h1:o7cqy6amK/52YcAKIPlM3a+Fpj35zvRj2TP+e1xFSfk=
+golang.org/x/time v0.5.0/go.mod h1:3BpzKBy/shNhVucY/MWOyx10tF3SFh9QdLuxbVysPQM=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190114222345-bf090417da8b/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190226205152-f727befe758c/go.mod h1:9Yl7xja0Znq3iFh3HoIrodX9oNMXvdceNzlUR8zjMvY=
@@ -626,8 +639,8 @@ google.golang.org/protobuf v1.24.0/go.mod h1:r/3tXBNzIEhYS9I1OUVjXDlt8tc493IdKGj
 google.golang.org/protobuf v1.25.0/go.mod h1:9JNX74DMeImyA3h4bdi1ymwjUzf21/xIlbajtzgsN7c=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
 google.golang.org/protobuf v1.26.0/go.mod h1:9q0QmTI4eRPtz6boOQmLYwt+qCgq0jsYwAQnmE0givc=
-google.golang.org/protobuf v1.30.0 h1:kPPoIgf3TsEvrm0PFe15JQ+570QVxYzEvvHqChK+cng=
-google.golang.org/protobuf v1.30.0/go.mod h1:HV8QOd/L58Z+nl8r43ehVNZIU/HEI6OcFqwMG9pJV4I=
+google.golang.org/protobuf v1.33.0 h1:uNO2rsAINq/JlFpSdYEKIZ0uKD/R9cpdv0T+yoGwGmI=
+google.golang.org/protobuf v1.33.0/go.mod h1:c6P6GXX6sHbq/GpV6MGZEdwhWPcYBgnhAHhKbcUYpos=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20180628173108-788fd7840127/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c h1:Hei/4ADfdWqJk1ZMxUNpqntNwaWcugrBjAiHlqqRiVk=

internal/clientinfo/client_info.go

@@ -14,6 +14,11 @@ import (
 	"github.com/Control-D-Inc/ctrld/internal/controld"
 )
 
+const (
+	ipV4Loopback = "127.0.0.1"
+	ipv6Loopback = "::1"
+)
+
 // IpResolver is the interface for retrieving IP from Mac.
 type IpResolver interface {
 	fmt.Stringer
@@ -224,6 +229,7 @@ func (t *Table) init() {
 			cancel()
 		}()
 		go t.ndp.listen(ctx)
+		go t.ndp.subscribe(ctx)
 	}
 	// PTR lookup.
 	if t.discoverPTR() {
@@ -321,6 +327,16 @@ func (t *Table) LookupRFC1918IPv4(mac string) string {
 	return ""
 }
 
+// LocalHostname returns the localhost hostname associated with loopback IP.
+func (t *Table) LocalHostname() string {
+	for _, ip := range []string{ipV4Loopback, ipv6Loopback} {
+		if name := t.LookupHostname(ip, ""); name != "" {
+			return name
+		}
+	}
+	return ""
+}
+
 type macEntry struct {
 	mac string
 	src string

internal/clientinfo/dhcp.go

@@ -353,8 +353,8 @@ func (d *dhcp) addSelf() {
 		return
 	}
 	hostname = normalizeHostname(hostname)
-	d.ip2name.Store("127.0.0.1", hostname)
-	d.ip2name.Store("::1", hostname)
+	d.ip2name.Store(ipV4Loopback, hostname)
+	d.ip2name.Store(ipv6Loopback, hostname)
 	found := false
 	interfaces.ForeachInterface(func(i interfaces.Interface, prefixes []netip.Prefix) {
 		mac := i.HardwareAddr.String()
@@ -375,15 +375,17 @@ func (d *dhcp) addSelf() {
 			d.mac.Store(ip.String(), mac)
 			d.ip.Store(mac, ip.String())
 			if ip.To4() != nil {
-				d.mac.Store("127.0.0.1", mac)
+				d.mac.Store(ipV4Loopback, mac)
 			} else {
-				d.mac.Store("::1", mac)
+				d.mac.Store(ipv6Loopback, mac)
 			}
 			d.mac2name.Store(mac, hostname)
 			d.ip2name.Store(ip.String(), hostname)
 			// If we have self IP set, and this IP is it, use this IP only.
 			if ip.String() == d.selfIP {
 				found = true
+				d.mac.Store(ipV4Loopback, mac)
+				d.mac.Store(ipv6Loopback, mac)
 			}
 		}
 	})

internal/clientinfo/hostsfile.go

@@ -95,7 +95,7 @@ func (hf *hostsFile) LookupHostnameByIP(ip string) string {
 	hf.mu.Lock()
 	defer hf.mu.Unlock()
 	if names := hf.m[ip]; len(names) > 0 {
-		isLoopback := ip == "127.0.0.1" || ip == "::1"
+		isLoopback := ip == ipV4Loopback || ip == ipv6Loopback
 		for _, hostname := range names {
 			name := normalizeHostname(hostname)
 			// Ignoring ipv4/ipv6 loopback entry.

internal/clientinfo/mdns.go

@@ -122,8 +122,8 @@ func (m *mdns) probeLoop(conns []*net.UDPConn, remoteAddr net.Addr, quitCh chan
 	bo := backoff.NewBackoff("mdns probe", func(format string, args ...any) {}, time.Second*30)
 	for {
 		err := m.probe(conns, remoteAddr)
-		if isErrNetUnreachableOrInvalid(err) {
-			ctrld.ProxyLogger.Load().Warn().Msgf("stop probing %q: network unreachable or invalid", remoteAddr)
+		if shouldStopProbing(err) {
+			ctrld.ProxyLogger.Load().Warn().Msgf("stop probing %q: %v", remoteAddr, err)
 			break
 		}
 		if err != nil {
@@ -165,7 +165,7 @@ func (m *mdns) readLoop(conn *net.UDPConn) {
 		}
 
 		var ip, name string
-		rrs := make([]dns.RR, 0, len(msg.Answer)+len(msg.Extra))
+		var rrs []dns.RR
 		rrs = append(rrs, msg.Answer...)
 		rrs = append(rrs, msg.Extra...)
 		for _, rr := range rrs {
@@ -273,10 +273,14 @@ func multicastInterfaces() ([]net.Interface, error) {
 	return interfaces, nil
 }
 
-func isErrNetUnreachableOrInvalid(err error) bool {
+// shouldStopProbing reports whether ctrld should stop probing mdns.
+func shouldStopProbing(err error) bool {
 	var se *os.SyscallError
 	if errors.As(err, &se) {
-		return se.Err == syscall.ENETUNREACH || se.Err == syscall.EINVAL
+		switch se.Err {
+		case syscall.ENETUNREACH, syscall.EINVAL, syscall.EPERM:
+			return true
+		}
 	}
 	return false
 }

internal/clientinfo/ndp.go

@@ -15,6 +15,7 @@ import (
 	"github.com/mdlayher/ndp"
 
 	"github.com/Control-D-Inc/ctrld"
+	ctrldnet "github.com/Control-D-Inc/ctrld/internal/net"
 )
 
 // ndpDiscover provides client discovery functionality using NDP protocol.
@@ -69,15 +70,45 @@ func (nd *ndpDiscover) List() []string {
 	return ips
 }
 
+// saveInfo saves ip and mac info to mapping table.
+func (nd *ndpDiscover) saveInfo(ip, mac string) {
+	ip = normalizeIP(ip)
+	// Store ip => map mapping,
+	nd.mac.Store(ip, mac)
+
+	// Do not store mac => ip mapping if new ip is a link local unicast.
+	if ctrldnet.IsLinkLocalUnicastIPv6(ip) {
+		return
+	}
+
+	// If there is old ip => mac mapping, delete it.
+	if old, existed := nd.ip.Load(mac); existed {
+		oldIP := old.(string)
+		if oldIP != ip {
+			nd.mac.Delete(oldIP)
+		}
+	}
+	// Store mac => ip mapping.
+	nd.ip.Store(mac, ip)
+}
+
 // listen listens on ipv6 link local for Neighbor Solicitation message
 // to update new neighbors information to ndp table.
 func (nd *ndpDiscover) listen(ctx context.Context) {
-	ifi, err := firstInterfaceWithV6LinkLocal()
+	ifis, err := allInterfacesWithV6LinkLocal()
 	if err != nil {
-		ctrld.ProxyLogger.Load().Debug().Err(err).Msg("failed to find valid ipv6")
+		ctrld.ProxyLogger.Load().Debug().Err(err).Msg("failed to find valid ipv6 interfaces")
 		return
 	}
-	c, ip, err := ndp.Listen(ifi, ndp.LinkLocal)
+	for _, ifi := range ifis {
+		go func(ifi *net.Interface) {
+			nd.listenOnInterface(ctx, ifi)
+		}(ifi)
+	}
+}
+
+func (nd *ndpDiscover) listenOnInterface(ctx context.Context, ifi *net.Interface) {
+	c, ip, err := ndp.Listen(ifi, ndp.Unspecified)
 	if err != nil {
 		ctrld.ProxyLogger.Load().Debug().Err(err).Msg("ndp listen failed")
 		return
@@ -111,8 +142,7 @@ func (nd *ndpDiscover) listen(ctx context.Context) {
 		for _, opt := range am.Options {
 			if lla, ok := opt.(*ndp.LinkLayerAddress); ok {
 				mac := lla.Addr.String()
-				nd.mac.Store(fromIP, mac)
-				nd.ip.Store(mac, fromIP)
+				nd.saveInfo(fromIP, mac)
 			}
 		}
 	}
@@ -127,8 +157,7 @@ func (nd *ndpDiscover) scanWindows(r io.Reader) {
 			continue
 		}
 		if mac := parseMAC(fields[1]); mac != "" {
-			nd.mac.Store(fields[0], mac)
-			nd.ip.Store(mac, fields[0])
+			nd.saveInfo(fields[0], mac)
 		}
 	}
 }
@@ -147,8 +176,7 @@ func (nd *ndpDiscover) scanUnix(r io.Reader) {
 			if idx := strings.IndexByte(ip, '%'); idx != -1 {
 				ip = ip[:idx]
 			}
-			nd.mac.Store(ip, mac)
-			nd.ip.Store(mac, ip)
+			nd.saveInfo(ip, mac)
 		}
 	}
 }
@@ -183,14 +211,15 @@ func parseMAC(mac string) string {
 	return hw.String()
 }
 
-// firstInterfaceWithV6LinkLocal returns the first interface which is capable of using NDP.
-func firstInterfaceWithV6LinkLocal() (*net.Interface, error) {
+// allInterfacesWithV6LinkLocal returns all interfaces which is capable of using NDP.
+func allInterfacesWithV6LinkLocal() ([]*net.Interface, error) {
 	ifis, err := net.Interfaces()
 	if err != nil {
 		return nil, err
 	}
-
+	res := make([]*net.Interface, 0, len(ifis))
 	for _, ifi := range ifis {
+		ifi := ifi
 		// Skip if iface is down/loopback/non-multicast.
 		if ifi.Flags&net.FlagUp == 0 || ifi.Flags&net.FlagLoopback != 0 || ifi.Flags&net.FlagMulticast == 0 {
 			continue
@@ -211,9 +240,10 @@ func firstInterfaceWithV6LinkLocal() (*net.Interface, error) {
 				return nil, fmt.Errorf("invalid ip address: %s", ipNet.String())
 			}
 			if ip.Is6() && !ip.Is4In6() {
-				return &ifi, nil
+				res = append(res, &ifi)
+				break
 			}
 		}
 	}
-	return nil, errors.New("no interface can be used")
+	return res, nil
 }

internal/clientinfo/ndp_linux.go

@@ -1,7 +1,10 @@
 package clientinfo
 
 import (
+	"context"
+
 	"github.com/vishvananda/netlink"
+	"golang.org/x/sys/unix"
 
 	"github.com/Control-D-Inc/ctrld"
 )
@@ -15,10 +18,47 @@ func (nd *ndpDiscover) scan() {
 	}
 
 	for _, n := range neighs {
+		// Skipping non-reachable neighbors.
+		if n.State&netlink.NUD_REACHABLE == 0 {
+			continue
+		}
 		ip := n.IP.String()
 		mac := n.HardwareAddr.String()
-		nd.mac.Store(ip, mac)
-		nd.ip.Store(mac, ip)
+		nd.saveInfo(ip, mac)
+	}
+}
+
+// subscribe watches NDP table changes and update new information to local table.
+func (nd *ndpDiscover) subscribe(ctx context.Context) {
+	ch := make(chan netlink.NeighUpdate)
+	done := make(chan struct{})
+	defer close(done)
+	if err := netlink.NeighSubscribe(ch, done); err != nil {
+		ctrld.ProxyLogger.Load().Err(err).Msg("could not perform neighbor subscribing")
+		return
+	}
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case nu := <-ch:
+			if nu.Family != netlink.FAMILY_V6 {
+				continue
+			}
+			ip := normalizeIP(nu.IP.String())
+			if nu.Type == unix.RTM_DELNEIGH {
+				ctrld.ProxyLogger.Load().Debug().Msgf("removing NDP neighbor: %s", ip)
+				nd.mac.Delete(ip)
+				continue
+			}
+			mac := nu.HardwareAddr.String()
+			switch nu.State {
+			case netlink.NUD_REACHABLE:
+				nd.saveInfo(ip, mac)
+			case netlink.NUD_FAILED:
+				ctrld.ProxyLogger.Load().Debug().Msgf("removing NDP neighbor with failed state: %s", ip)
+				nd.mac.Delete(ip)
+			}
+		}
 	}
-
 }

internal/clientinfo/ndp_others.go

@@ -4,6 +4,7 @@ package clientinfo
 
 import (
 	"bytes"
+	"context"
 	"os/exec"
 	"runtime"
 
@@ -29,3 +30,7 @@ func (nd *ndpDiscover) scan() {
 		nd.scanUnix(bytes.NewReader(data))
 	}
 }
+
+// subscribe watches NDP table changes and update new information to local table.
+// This is a stub method, and only works on Linux at this moment.
+func (nd *ndpDiscover) subscribe(ctx context.Context) {}

internal/clientinfo/ndp_test.go

@@ -45,20 +45,22 @@ ff02::c                                       33-33-00-00-00-0c  Permanent
 	nd.scanWindows(r)
 
 	count := 0
+	expectedCount := 6
 	nd.mac.Range(func(key, value any) bool {
 		count++
 		return true
 	})
-	if count != 6 {
-		t.Errorf("unexpected count, want 6, got: %d", count)
+	if count != expectedCount {
+		t.Errorf("unexpected count, want %d, got: %d", expectedCount, count)
 	}
 
 	count = 0
+	expectedCount = 4
 	nd.ip.Range(func(key, value any) bool {
 		count++
 		return true
 	})
-	if count != 5 {
-		t.Errorf("unexpected count, want 5, got: %d", count)
+	if count != expectedCount {
+		t.Errorf("unexpected count, want %d, got: %d", expectedCount, count)
 	}
 }

internal/controld/config.go

@@ -26,14 +26,15 @@ const (
 	apiDomainDev       = "api.controld.dev"
 	resolverDataURLCom = "https://api.controld.com/utility"
 	resolverDataURLDev = "https://api.controld.dev/utility"
-	InvalidConfigCode  = 40401
+	InvalidConfigCode  = 40402
 )
 
 // ResolverConfig represents Control D resolver data.
 type ResolverConfig struct {
 	DOH   string `json:"doh"`
 	Ctrld struct {
-		CustomConfig string `json:"custom_config"`
+		CustomConfig     string `json:"custom_config"`
+		CustomLastUpdate int64  `json:"custom_last_update"`
 	} `json:"ctrld"`
 	Exclude         []string `json:"exclude"`
 	UID             string   `json:"uid"`
@@ -76,17 +77,28 @@ func FetchResolverConfig(rawUID, version string, cdDev bool) (*ResolverConfig, e
 		req.ClientID = clientID
 	}
 	body, _ := json.Marshal(req)
-	return postUtilityAPI(version, cdDev, bytes.NewReader(body))
+	return postUtilityAPI(version, cdDev, false, bytes.NewReader(body))
 }
 
 // FetchResolverUID fetch resolver uid from provision token.
 func FetchResolverUID(pt, version string, cdDev bool) (*ResolverConfig, error) {
 	hostname, _ := os.Hostname()
 	body, _ := json.Marshal(utilityOrgRequest{ProvToken: pt, Hostname: hostname})
-	return postUtilityAPI(version, cdDev, bytes.NewReader(body))
+	return postUtilityAPI(version, cdDev, false, bytes.NewReader(body))
 }
 
-func postUtilityAPI(version string, cdDev bool, body io.Reader) (*ResolverConfig, error) {
+// UpdateCustomLastFailed calls API to mark custom config is bad.
+func UpdateCustomLastFailed(rawUID, version string, cdDev, lastUpdatedFailed bool) (*ResolverConfig, error) {
+	uid, clientID := ParseRawUID(rawUID)
+	req := utilityRequest{UID: uid}
+	if clientID != "" {
+		req.ClientID = clientID
+	}
+	body, _ := json.Marshal(req)
+	return postUtilityAPI(version, cdDev, true, bytes.NewReader(body))
+}
+
+func postUtilityAPI(version string, cdDev, lastUpdatedFailed bool, body io.Reader) (*ResolverConfig, error) {
 	apiUrl := resolverDataURLCom
 	if cdDev {
 		apiUrl = resolverDataURLDev
@@ -98,6 +110,9 @@ func postUtilityAPI(version string, cdDev bool, body io.Reader) (*ResolverConfig
 	q := req.URL.Query()
 	q.Set("platform", "ctrld")
 	q.Set("version", version)
+	if lastUpdatedFailed {
+		q.Set("custom_last_failed", "1")
+	}
 	req.URL.RawQuery = q.Encode()
 	req.Header.Add("Content-Type", "application/json")
 	transport := http.DefaultTransport.(*http.Transport).Clone()

internal/dnscache/cache.go

@@ -12,6 +12,7 @@ import (
 type Cacher interface {
 	Get(Key) *Value
 	Add(Key, *Value)
+	Purge()
 }
 
 // Key is the caching key for DNS message.
@@ -34,15 +35,22 @@ type LRUCache struct {
 	cacher *lru.ARCCache[Key, *Value]
 }
 
+// Get looks up key's value from cache.
 func (l *LRUCache) Get(key Key) *Value {
 	v, _ := l.cacher.Get(key)
 	return v
 }
 
+// Add adds a value to cache.
 func (l *LRUCache) Add(key Key, value *Value) {
 	l.cacher.Add(key, value)
 }
 
+// Purge clears the cache.
+func (l *LRUCache) Purge() {
+	l.cacher.Purge()
+}
+
 // NewLRUCache creates a new LRUCache instance with given size.
 func NewLRUCache(size int) (*LRUCache, error) {
 	cacher, err := lru.NewARC[Key, *Value](size)

internal/net/net.go

@@ -115,6 +115,15 @@ func IsIPv6(ip string) bool {
 	return parsedIP != nil && parsedIP.To4() == nil && parsedIP.To16() != nil
 }
 
+// IsLinkLocalUnicastIPv6 checks if the provided IP is a link local unicast v6 address.
+func IsLinkLocalUnicastIPv6(ip string) bool {
+	parsedIP := net.ParseIP(ip)
+	if parsedIP == nil || parsedIP.To4() != nil || parsedIP.To16() == nil {
+		return false
+	}
+	return parsedIP.To16().IsLinkLocalUnicast()
+}
+
 type parallelDialerResult struct {
 	conn net.Conn
 	err  error

internal/router/dnsmasq/dnsmasq.go

@@ -10,6 +10,8 @@ import (
 	"github.com/Control-D-Inc/ctrld"
 )
 
+const CtrldMarker = `# GENERATED BY ctrld - DO NOT MODIFY`
+
 const ConfigContentTmpl = `# GENERATED BY ctrld - DO NOT MODIFY
 no-resolv
 {{- range .Upstreams}}

internal/router/netgear_orbi_voxel/procd.go

@@ -0,0 +1,22 @@
+package netgear
+
+const openWrtScript = `#!/bin/sh /etc/rc.common
+USE_PROCD=1
+# After dnsmasq starts
+START=61
+# Before network stops
+STOP=89
+cmd="{{.Path}}{{range .Arguments}} {{.|cmd}}{{end}}"
+name="{{.Name}}"
+pid_file="/var/run/${name}.pid"
+
+start_service() {
+    echo "Starting ${name}"
+    procd_open_instance
+    procd_set_param command ${cmd}
+    procd_set_param respawn             # respawn automatically if something died
+    procd_set_param pidfile ${pid_file} # write a pid file on instance start and remove it on stop
+    procd_close_instance
+    echo "${name} has been started"
+}
+`

internal/router/netgear_orbi_voxel/voxel.go

@@ -0,0 +1,220 @@
+package netgear
+
+import (
+	"bufio"
+	"bytes"
+	"fmt"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strings"
+
+	"github.com/kardianos/service"
+
+	"github.com/Control-D-Inc/ctrld"
+	"github.com/Control-D-Inc/ctrld/internal/router/dnsmasq"
+	"github.com/Control-D-Inc/ctrld/internal/router/nvram"
+)
+
+const (
+	Name                                = "netgear_orbi_voxel"
+	netgearOrbiVoxelDNSMasqConfigPath   = "/etc/dnsmasq.conf"
+	netgearOrbiVoxelHomedir             = "/mnt/bitdefender"
+	netgearOrbiVoxelStartupScript       = "/mnt/bitdefender/rc.user"
+	netgearOrbiVoxelStartupScriptBackup = "/mnt/bitdefender/rc.user.bak"
+	netgearOrbiVoxelStartupScriptMarker = "\n# GENERATED BY ctrld"
+)
+
+var nvramKvMap = map[string]string{
+	"dns_hijack": "0", // Disable dns hijacking
+}
+
+type NetgearOrbiVoxel struct {
+	cfg *ctrld.Config
+}
+
+// New returns a router.Router for configuring/setup/run ctrld on ddwrt routers.
+func New(cfg *ctrld.Config) *NetgearOrbiVoxel {
+	return &NetgearOrbiVoxel{cfg: cfg}
+}
+
+func (d *NetgearOrbiVoxel) ConfigureService(svc *service.Config) error {
+	if err := d.checkInstalledDir(); err != nil {
+		return err
+	}
+	svc.Option["SysvScript"] = openWrtScript
+	return nil
+}
+
+func (d *NetgearOrbiVoxel) Install(_ *service.Config) error {
+	// Ignoring error here at this moment is ok, since everything will be wiped out on reboot.
+	_ = exec.Command("/etc/init.d/ctrld", "enable").Run()
+	if err := d.checkInstalledDir(); err != nil {
+		return err
+	}
+	if err := backupVoxelStartupScript(); err != nil {
+		return fmt.Errorf("backup startup script: %w", err)
+	}
+	if err := writeVoxelStartupScript(); err != nil {
+		return fmt.Errorf("writing startup script: %w", err)
+	}
+	return nil
+}
+
+func (d *NetgearOrbiVoxel) Uninstall(_ *service.Config) error {
+	if err := os.Remove(netgearOrbiVoxelStartupScript); err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	err := os.Rename(netgearOrbiVoxelStartupScriptBackup, netgearOrbiVoxelStartupScript)
+	if err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	return nil
+}
+
+func (d *NetgearOrbiVoxel) PreRun() error {
+	return nil
+}
+
+func (d *NetgearOrbiVoxel) Setup() error {
+	if d.cfg.FirstListener().IsDirectDnsListener() {
+		return nil
+	}
+	// Already setup.
+	if val, _ := nvram.Run("get", nvram.CtrldSetupKey); val == "1" {
+		return nil
+	}
+
+	data, err := dnsmasq.ConfTmplWithCacheDisabled(dnsmasq.ConfigContentTmpl, d.cfg, false)
+	if err != nil {
+		return err
+	}
+	currentConfig, _ := os.ReadFile(netgearOrbiVoxelDNSMasqConfigPath)
+	configContent := append(currentConfig, data...)
+	if err := os.WriteFile(netgearOrbiVoxelDNSMasqConfigPath, configContent, 0600); err != nil {
+		return err
+	}
+	// Restart dnsmasq service.
+	if err := restartDNSMasq(); err != nil {
+		return err
+	}
+
+	if err := nvram.SetKV(nvramKvMap, nvram.CtrldSetupKey); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func (d *NetgearOrbiVoxel) Cleanup() error {
+	if d.cfg.FirstListener().IsDirectDnsListener() {
+		return nil
+	}
+	if val, _ := nvram.Run("get", nvram.CtrldSetupKey); val != "1" {
+		return nil // was restored, nothing to do.
+	}
+
+	// Restore old configs.
+	if err := nvram.Restore(nvramKvMap, nvram.CtrldSetupKey); err != nil {
+		return err
+	}
+
+	// Restore dnsmasq config.
+	if err := restoreDnsmasqConf(); err != nil {
+		return err
+	}
+
+	// Restart dnsmasq service.
+	if err := restartDNSMasq(); err != nil {
+		return err
+	}
+	return nil
+}
+
+// checkInstalledDir checks that ctrld binary was installed in the correct directory.
+func (d *NetgearOrbiVoxel) checkInstalledDir() error {
+	exePath, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("checkHomeDir: failed to get binary path %w", err)
+	}
+	if !strings.HasSuffix(filepath.Dir(exePath), netgearOrbiVoxelHomedir) {
+		return fmt.Errorf("checkHomeDir: could not install service outside %s", netgearOrbiVoxelHomedir)
+	}
+	return nil
+}
+
+// backupVoxelStartupScript creates a backup of original startup script if existed.
+func backupVoxelStartupScript() error {
+	// Do nothing if the startup script was modified by ctrld.
+	script, _ := os.ReadFile(netgearOrbiVoxelStartupScript)
+	if bytes.Contains(script, []byte(netgearOrbiVoxelStartupScriptMarker)) {
+		return nil
+	}
+	err := os.Rename(netgearOrbiVoxelStartupScript, netgearOrbiVoxelStartupScriptBackup)
+	if err != nil && !os.IsNotExist(err) {
+		return fmt.Errorf("backupVoxelStartupScript: %w", err)
+	}
+	return nil
+}
+
+// writeVoxelStartupScript writes startup script to re-install ctrld upon reboot.
+// See: https://github.com/SVoxel/ORBI-RBK50/pull/7
+func writeVoxelStartupScript() error {
+	exe, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("configure service: failed to get binary path %w", err)
+	}
+	// This is called when "ctrld start ..." runs, so recording
+	// the same command line arguments to use in startup script.
+	argStr := strings.Join(os.Args[1:], " ")
+	script, _ := os.ReadFile(netgearOrbiVoxelStartupScriptBackup)
+	script = append(script, fmt.Sprintf("%s\n%q %s\n", netgearOrbiVoxelStartupScriptMarker, exe, argStr)...)
+	f, err := os.Create(netgearOrbiVoxelStartupScript)
+	if err != nil {
+		return fmt.Errorf("failed to create startup script: %w", err)
+	}
+	defer f.Close()
+
+	if _, err := f.Write(script); err != nil {
+		return fmt.Errorf("failed to write startup script: %w", err)
+	}
+	if err := f.Close(); err != nil {
+		return fmt.Errorf("failed to save startup script: %w", err)
+	}
+	return nil
+}
+
+// restoreDnsmasqConf restores original dnsmasq configuration.
+func restoreDnsmasqConf() error {
+	f, err := os.Open(netgearOrbiVoxelDNSMasqConfigPath)
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+
+	var bs []byte
+	buf := bytes.NewBuffer(bs)
+
+	removed := false
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if line == dnsmasq.CtrldMarker {
+			removed = true
+		}
+		if !removed {
+			_, err := buf.WriteString(line + "\n")
+			if err != nil {
+				return err
+			}
+		}
+	}
+	return os.WriteFile(netgearOrbiVoxelDNSMasqConfigPath, buf.Bytes(), 0644)
+}
+
+func restartDNSMasq() error {
+	if out, err := exec.Command("/etc/init.d/dnsmasq", "restart").CombinedOutput(); err != nil {
+		return fmt.Errorf("restartDNSMasq: %s, %w", string(out), err)
+	}
+	return nil
+}

internal/router/openwrt/procd.go

@@ -18,6 +18,7 @@ start_service() {
     procd_set_param stdout 1            # forward stdout of the command to logd
     procd_set_param stderr 1            # same for stderr
     procd_set_param pidfile ${pid_file} # write a pid file on instance start and remove it on stop
+    procd_set_param term_timeout 10
     procd_close_instance
     echo "${name} has been started"
 }

internal/router/os_config_freebsd.go

@@ -0,0 +1,40 @@
+package router
+
+import (
+	"encoding/xml"
+	"os"
+)
+
+// Config represents /conf/config.xml file found on pfsense/opnsense.
+type Config struct {
+	PfsenseUnbound  *string `xml:"unbound>enable,omitempty"`
+	OPNsenseUnbound *string `xml:"OPNsense>unboundplus>general>enabled,omitempty"`
+	Dnsmasq         *string `xml:"dnsmasq>enable,omitempty"`
+}
+
+// DnsmasqEnabled reports whether dnsmasq is enabled.
+func (c *Config) DnsmasqEnabled() bool {
+	if isPfsense() { // pfsense only set the attribute if dnsmasq is enabled.
+		return c.Dnsmasq != nil
+	}
+	return c.Dnsmasq != nil && *c.Dnsmasq == "1"
+}
+
+// UnboundEnabled reports whether unbound is enabled.
+func (c *Config) UnboundEnabled() bool {
+	if isPfsense() { // pfsense only set the attribute if unbound is enabled.
+		return c.PfsenseUnbound != nil
+	}
+	return c.OPNsenseUnbound != nil && *c.OPNsenseUnbound == "1"
+}
+
+// currentConfig does unmarshalling /conf/config.xml file,
+// return the corresponding *Config represent it.
+func currentConfig() (*Config, error) {
+	buf, _ := os.ReadFile("/conf/config.xml")
+	c := Config{}
+	if err := xml.Unmarshal(buf, &c); err != nil {
+		return nil, err
+	}
+	return &c, nil
+}

internal/router/os_freebsd.go

@@ -111,8 +111,16 @@ func (or *osRouter) Setup() error {
 
 func (or *osRouter) Cleanup() error {
 	if or.cdMode {
-		_ = exec.Command(unboundRcPath, "onerestart").Run()
-		_ = exec.Command(dnsmasqRcPath, "onerestart").Run()
+		c, err := currentConfig()
+		if err != nil {
+			return err
+		}
+		if c.UnboundEnabled() {
+			_ = exec.Command(unboundRcPath, "onerestart").Run()
+		}
+		if c.DnsmasqEnabled() {
+			_ = exec.Command(dnsmasqRcPath, "onerestart").Run()
+		}
 	}
 	return nil
 }

internal/router/router.go

@@ -7,6 +7,7 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
+	"strings"
 	"sync/atomic"
 
 	"github.com/kardianos/service"
@@ -18,6 +19,7 @@ import (
 	"github.com/Control-D-Inc/ctrld/internal/router/edgeos"
 	"github.com/Control-D-Inc/ctrld/internal/router/firewalla"
 	"github.com/Control-D-Inc/ctrld/internal/router/merlin"
+	netgear "github.com/Control-D-Inc/ctrld/internal/router/netgear_orbi_voxel"
 	"github.com/Control-D-Inc/ctrld/internal/router/openwrt"
 	"github.com/Control-D-Inc/ctrld/internal/router/synology"
 	"github.com/Control-D-Inc/ctrld/internal/router/tomato"
@@ -66,10 +68,17 @@ func New(cfg *ctrld.Config, cdMode bool) Router {
 		return tomato.New(cfg)
 	case firewalla.Name:
 		return firewalla.New(cfg)
+	case netgear.Name:
+		return netgear.New(cfg)
 	}
 	return newOsRouter(cfg, cdMode)
 }
 
+// IsNetGearOrbi reports whether the router is a Netgear Orbi router.
+func IsNetGearOrbi() bool {
+	return Name() == netgear.Name
+}
+
 // IsGLiNet reports whether the router is an GL.iNet router.
 func IsGLiNet() bool {
 	if Name() != openwrt.Name {
@@ -90,6 +99,11 @@ func IsOldOpenwrt() bool {
 	return cmd == ""
 }
 
+// WaitProcessExited reports whether the "ctrld stop" command have to wait until ctrld process exited.
+func WaitProcessExited() bool {
+	return Name() == openwrt.Name
+}
+
 var routerPlatform atomic.Pointer[router]
 
 type router struct {
@@ -145,12 +159,22 @@ func LocalResolverIP() string {
 // HomeDir returns the home directory of ctrld on current router.
 func HomeDir() (string, error) {
 	switch Name() {
-	case ddwrt.Name, merlin.Name, tomato.Name:
+	case ddwrt.Name, firewalla.Name, merlin.Name, netgear.Name, tomato.Name:
 		exe, err := os.Executable()
 		if err != nil {
 			return "", err
 		}
 		return filepath.Dir(exe), nil
+	case edgeos.Name:
+		exe, err := os.Executable()
+		if err != nil {
+			return "", err
+		}
+		// Using binary directory as home dir if it is located in /config.
+		// Otherwise, fallback to old behavior for compatibility.
+		if strings.HasPrefix(exe, "/config/") {
+			return filepath.Dir(exe), nil
+		}
 	}
 	return "", nil
 }
@@ -198,6 +222,9 @@ func distroName() string {
 	case bytes.HasPrefix(unameO(), []byte("ASUSWRT-Merlin")):
 		return merlin.Name
 	case haveFile("/etc/openwrt_version"):
+		if haveFile("/bin/config") { // TODO: is there any more reliable way?
+			return netgear.Name
+		}
 		return openwrt.Name
 	case isUbios():
 		return ubios.Name

internal/router/service_merlin.go

@@ -49,11 +49,15 @@ func (s *merlinSvc) Platform() string {
 }
 
 func (s *merlinSvc) configPath() string {
-	path, err := os.Executable()
-	if err != nil {
-		return ""
+	bin := s.Config.Executable
+	if bin == "" {
+		path, err := os.Executable()
+		if err != nil {
+			return ""
+		}
+		bin = path
 	}
-	return path + ".startup"
+	return bin + ".startup"
 }
 
 func (s *merlinSvc) template() *template.Template {

nameservers_windows.go

@@ -1,12 +1,9 @@
 package ctrld
 
 import (
-	"net"
 	"syscall"
 
 	"golang.zx2c4.com/wireguard/windows/tunnel/winipcfg"
-
-	"golang.org/x/sys/windows"
 )
 
 func dnsFns() []dnsFn {
@@ -20,40 +17,23 @@ func dnsFromAdapter() []string {
 	}
 	ns := make([]string, 0, len(aas)*2)
 	seen := make(map[string]bool)
-	do := func(addr windows.SocketAddress) {
-		sa, err := addr.Sockaddr.Sockaddr()
-		if err != nil {
-			return
+	addressMap := make(map[string]struct{})
+	for _, aa := range aas {
+		for a := aa.FirstUnicastAddress; a != nil; a = a.Next {
+			addressMap[a.Address.IP().String()] = struct{}{}
 		}
-		var ip net.IP
-		switch sa := sa.(type) {
-		case *syscall.SockaddrInet4:
-			ip = net.IPv4(sa.Addr[0], sa.Addr[1], sa.Addr[2], sa.Addr[3])
-		case *syscall.SockaddrInet6:
-			ip = make(net.IP, net.IPv6len)
-			copy(ip, sa.Addr[:])
-			if ip[0] == 0xfe && ip[1] == 0xc0 {
-				// Ignore these fec0/10 ones. Windows seems to
-				// populate them as defaults on its misc rando
-				// interfaces.
-				return
-			}
-		default:
-			return
-
-		}
-		if ip.IsLoopback() || seen[ip.String()] {
-			return
-		}
-		seen[ip.String()] = true
-		ns = append(ns, ip.String())
 	}
 	for _, aa := range aas {
 		for dns := aa.FirstDNSServerAddress; dns != nil; dns = dns.Next {
-			do(dns.Address)
-		}
-		for gw := aa.FirstGatewayAddress; gw != nil; gw = gw.Next {
-			do(gw.Address)
+			ip := dns.Address.IP()
+			if ip == nil || ip.IsLoopback() || seen[ip.String()] {
+				continue
+			}
+			if _, ok := addressMap[ip.String()]; ok {
+				continue
+			}
+			seen[ip.String()] = true
+			ns = append(ns, ip.String())
 		}
 	}
 	return ns

resolver.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net"
 	"net/netip"
+	"slices"
 	"sync"
 	"time"
 
@@ -30,18 +31,49 @@ const (
 	ResolverTypePrivate = "private"
 )
 
-const bootstrapDNS = "76.76.2.22"
+const (
+	controldBootstrapDns = "76.76.2.22"
+	controldPublicDns    = "76.76.2.0"
+)
+
+var controldPublicDnsWithPort = net.JoinHostPort(controldPublicDns, "53")
 
 // or is the Resolver used for ResolverTypeOS.
 var or = &osResolver{nameservers: defaultNameservers()}
 
-// defaultNameservers returns OS nameservers plus ctrld bootstrap nameserver.
+// defaultNameservers returns OS nameservers plus ControlD public DNS.
 func defaultNameservers() []string {
 	ns := nameservers()
-	ns = append(ns, net.JoinHostPort(bootstrapDNS, "53"))
 	return ns
 }
 
+// InitializeOsResolver initializes OS resolver using the current system DNS settings.
+// It returns the nameservers that is going to be used by the OS resolver.
+//
+// It's the caller's responsibility to ensure the system DNS is in a clean state before
+// calling this function.
+func InitializeOsResolver() []string {
+	or.nameservers = or.nameservers[:0]
+	for _, ns := range defaultNameservers() {
+		if testNameserver(ns) {
+			or.nameservers = append(or.nameservers, ns)
+		}
+	}
+	or.nameservers = append(or.nameservers, controldPublicDnsWithPort)
+	return or.nameservers
+}
+
+// testPlainDnsNameserver sends a test query to DNS nameserver to check if the server is available.
+func testNameserver(addr string) bool {
+	msg := new(dns.Msg)
+	msg.SetQuestion(".", dns.TypeNS)
+	client := new(dns.Client)
+	ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+	defer cancel()
+	_, _, err := client.ExchangeContext(ctx, msg, addr)
+	return err == nil
+}
+
 // Resolver is the interface that wraps the basic DNS operations.
 //
 // Resolve resolves the DNS query, return the result and the corresponding error.
@@ -76,8 +108,9 @@ type osResolver struct {
 }
 
 type osResolverResult struct {
-	answer *dns.Msg
-	err    error
+	answer              *dns.Msg
+	err                 error
+	isControlDPublicDNS bool
 }
 
 // Resolve resolves DNS queries using pre-configured nameservers.
@@ -103,19 +136,34 @@ func (o *osResolver) Resolve(ctx context.Context, msg *dns.Msg) (*dns.Msg, error
 		go func(server string) {
 			defer wg.Done()
 			answer, _, err := dnsClient.ExchangeContext(ctx, msg.Copy(), server)
-			ch <- &osResolverResult{answer: answer, err: err}
+			ch <- &osResolverResult{answer: answer, err: err, isControlDPublicDNS: server == controldPublicDnsWithPort}
 		}(server)
 	}
 
+	var (
+		nonSuccessAnswer      *dns.Msg
+		controldSuccessAnswer *dns.Msg
+	)
 	errs := make([]error, 0, numServers)
 	for res := range ch {
-		if res.err == nil {
-			cancel()
-			return res.answer, res.err
+		switch {
+		case res.answer != nil && res.answer.Rcode == dns.RcodeSuccess:
+			if res.isControlDPublicDNS {
+				controldSuccessAnswer = res.answer // only use ControlD answer as last one.
+			} else {
+				cancel()
+				return res.answer, nil
+			}
+		case res.answer != nil:
+			nonSuccessAnswer = res.answer
 		}
 		errs = append(errs, res.err)
 	}
-
+	for _, answer := range []*dns.Msg{controldSuccessAnswer, nonSuccessAnswer} {
+		if answer != nil {
+			return answer, nil
+		}
+	}
 	return nil, errors.Join(errs...)
 }
 
@@ -125,7 +173,7 @@ type legacyResolver struct {
 
 func (r *legacyResolver) Resolve(ctx context.Context, msg *dns.Msg) (*dns.Msg, error) {
 	// See comment in (*dotResolver).resolve method.
-	dialer := newDialer(net.JoinHostPort(bootstrapDNS, "53"))
+	dialer := newDialer(net.JoinHostPort(controldBootstrapDns, "53"))
 	dnsTyp := uint16(0)
 	if msg != nil && len(msg.Question) > 0 {
 		dnsTyp = msg.Question[0].Qtype
@@ -163,7 +211,7 @@ func LookupIP(domain string) []string {
 func lookupIP(domain string, timeout int, withBootstrapDNS bool) (ips []string) {
 	resolver := &osResolver{nameservers: nameservers()}
 	if withBootstrapDNS {
-		resolver.nameservers = append([]string{net.JoinHostPort(bootstrapDNS, "53")}, resolver.nameservers...)
+		resolver.nameservers = append([]string{net.JoinHostPort(controldBootstrapDns, "53")}, resolver.nameservers...)
 	}
 	ProxyLogger.Load().Debug().Msgf("resolving %q using bootstrap DNS %q", domain, resolver.nameservers)
 	timeoutMs := 2000
@@ -239,7 +287,7 @@ func lookupIP(domain string, timeout int, withBootstrapDNS bool) (ips []string)
 //   - Input servers.
 func NewBootstrapResolver(servers ...string) Resolver {
 	resolver := &osResolver{nameservers: nameservers()}
-	resolver.nameservers = append([]string{net.JoinHostPort(bootstrapDNS, "53")}, resolver.nameservers...)
+	resolver.nameservers = append([]string{controldPublicDnsWithPort}, resolver.nameservers...)
 	for _, ns := range servers {
 		resolver.nameservers = append([]string{net.JoinHostPort(ns, "53")}, resolver.nameservers...)
 	}
@@ -266,11 +314,11 @@ func NewPrivateResolver() Resolver {
 		//  - Direct listener that has ctrld as an upstream (e.g: dnsmasq).
 		//
 		// causing the query always succeed.
-		if sliceContains(resolveConfNss, host) {
+		if slices.Contains(resolveConfNss, host) {
 			continue
 		}
 		// Ignoring local RFC 1918 addresses.
-		if sliceContains(localRfc1918Addrs, host) {
+		if slices.Contains(localRfc1918Addrs, host) {
 			continue
 		}
 		ip := net.ParseIP(host)
@@ -322,20 +370,3 @@ func newDialer(dnsAddress string) *net.Dialer {
 		},
 	}
 }
-
-// TODO(cuonglm): use slices.Contains once upgrading to go1.21
-// sliceContains reports whether v is present in s.
-func sliceContains[S ~[]E, E comparable](s S, v E) bool {
-	return sliceIndex(s, v) >= 0
-}
-
-// sliceIndex returns the index of the first occurrence of v in s,
-// or -1 if not present.
-func sliceIndex[S ~[]E, E comparable](s S, v E) int {
-	for i := range s {
-		if v == s[i] {
-			return i
-		}
-	}
-	return -1
-}

resolver_test.go

@@ -2,6 +2,8 @@ package ctrld
 
 import (
 	"context"
+	"net"
+	"sync"
 	"testing"
 	"time"
 
@@ -28,6 +30,57 @@ func Test_osResolver_Resolve(t *testing.T) {
 	}
 }
 
+func Test_osResolver_ResolveWithNonSuccessAnswer(t *testing.T) {
+	ns := make([]string, 0, 2)
+	servers := make([]*dns.Server, 0, 2)
+	successHandler := dns.HandlerFunc(func(w dns.ResponseWriter, msg *dns.Msg) {
+		m := new(dns.Msg)
+		m.SetRcode(msg, dns.RcodeSuccess)
+		w.WriteMsg(m)
+	})
+	nonSuccessHandlerWithRcode := func(rcode int) dns.HandlerFunc {
+		return dns.HandlerFunc(func(w dns.ResponseWriter, msg *dns.Msg) {
+			m := new(dns.Msg)
+			m.SetRcode(msg, rcode)
+			w.WriteMsg(m)
+		})
+	}
+
+	handlers := []dns.Handler{
+		nonSuccessHandlerWithRcode(dns.RcodeRefused),
+		nonSuccessHandlerWithRcode(dns.RcodeNameError),
+		successHandler,
+	}
+	for i := range handlers {
+		pc, err := net.ListenPacket("udp", ":0")
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+
+		s, addr, err := runLocalPacketConnTestServer(t, pc, handlers[i])
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+		ns = append(ns, addr)
+		servers = append(servers, s)
+	}
+	defer func() {
+		for _, server := range servers {
+			server.Shutdown()
+		}
+	}()
+	resolver := &osResolver{nameservers: ns}
+	msg := new(dns.Msg)
+	msg.SetQuestion(".", dns.TypeNS)
+	answer, err := resolver.Resolve(context.Background(), msg)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if answer.Rcode != dns.RcodeSuccess {
+		t.Errorf("unexpected return code: %s", dns.RcodeToString[answer.Rcode])
+	}
+}
+
 func Test_upstreamTypeFromEndpoint(t *testing.T) {
 	tests := []struct {
 		name         string
@@ -51,3 +104,33 @@ func Test_upstreamTypeFromEndpoint(t *testing.T) {
 		})
 	}
 }
+
+func runLocalPacketConnTestServer(t *testing.T, pc net.PacketConn, handler dns.Handler, opts ...func(*dns.Server)) (*dns.Server, string, error) {
+	t.Helper()
+
+	server := &dns.Server{
+		PacketConn:   pc,
+		ReadTimeout:  time.Hour,
+		WriteTimeout: time.Hour,
+		Handler:      handler,
+	}
+
+	waitLock := sync.Mutex{}
+	waitLock.Lock()
+	server.NotifyStartedFunc = waitLock.Unlock
+
+	for _, opt := range opts {
+		opt(server)
+	}
+
+	addr, closer := pc.LocalAddr().String(), pc
+	go func() {
+		if err := server.ActivateAndServe(); err != nil {
+			t.Error(err)
+		}
+		closer.Close()
+	}()
+
+	waitLock.Lock()
+	return server, addr, nil
+}

testhelper/config.go

@@ -27,6 +27,8 @@ var sampleConfigContent = `
 [service]
 log_level = "info"
 log_path = "/path/to/log.log"
+dns_watchdog_enabled = false
+dns_watchdog_interval = "20s"
 
 [network.0]
 name = "Home Wifi"