feat(dns): add optimistic caching

app/dns/cache_controller.go

@@ -3,32 +3,37 @@ package dns
 import (
 	"context"
 	go_errors "errors"
+	"sync"
+	"time"
+
 	"github.com/xtls/xray-core/common"
 	"github.com/xtls/xray-core/common/errors"
 	"github.com/xtls/xray-core/common/net"
 	"github.com/xtls/xray-core/common/signal/pubsub"
 	"github.com/xtls/xray-core/common/task"
 	dns_feature "github.com/xtls/xray-core/features/dns"
 	"golang.org/x/net/dns/dnsmessage"
-	"sync"
-	"time"
 )
 
 type CacheController struct {
 	sync.RWMutex
-	ips          map[string]*record
-	pub          *pubsub.Service
-	cacheCleanup *task.Periodic
-	name         string
-	disableCache bool
+	ips             map[string]*record
+	pub             *pubsub.Service
+	cacheCleanup    *task.Periodic
+	name            string
+	disableCache    bool
+	serveStale      bool
+	serveExpiredTTL int32
 }
 
-func NewCacheController(name string, disableCache bool) *CacheController {
+func NewCacheController(name string, disableCache bool, serveStale bool, serveExpiredTTL uint32) *CacheController {
 	c := &CacheController{
-		name:         name,
-		disableCache: disableCache,
-		ips:          make(map[string]*record),
-		pub:          pubsub.NewService(),
+		name:            name,
+		disableCache:    disableCache,
+		serveStale:      serveStale,
+		serveExpiredTTL: -int32(serveExpiredTTL),
+		ips:             make(map[string]*record),
+		pub:             pubsub.NewService(),
 	}
 
 	c.cacheCleanup = &task.Periodic{
@@ -41,6 +46,10 @@ func NewCacheController(name string, disableCache bool) *CacheController {
 // CacheCleanup clears expired items from cache
 func (c *CacheController) CacheCleanup() error {
 	now := time.Now()
+	if c.serveStale && c.serveExpiredTTL != 0 {
+		now = now.Add(time.Duration(c.serveExpiredTTL) * time.Second)
+	}
+
 	c.Lock()
 	defer c.Unlock()
 
@@ -94,26 +103,28 @@ func (c *CacheController) updateIP(req *dnsRequest, ipRec *IPRecord) {
 	case dnsmessage.TypeA:
 		c.pub.Publish(req.domain+"4", nil)
 		if !c.disableCache {
-			_, _, err := rec.AAAA.getIPs()
-			if !go_errors.Is(err, errRecordNotFound) {
+			_, ttl, err := rec.AAAA.getIPs()
+			if ttl > 0 && !go_errors.Is(err, errRecordNotFound) {
 				c.pub.Publish(req.domain+"6", nil)
 			}
 		}
 	case dnsmessage.TypeAAAA:
 		c.pub.Publish(req.domain+"6", nil)
 		if !c.disableCache {
-			_, _, err := rec.A.getIPs()
-			if !go_errors.Is(err, errRecordNotFound) {
+			_, ttl, err := rec.A.getIPs()
+			if ttl > 0 && !go_errors.Is(err, errRecordNotFound) {
 				c.pub.Publish(req.domain+"4", nil)
 			}
 		}
 	}
 
 	c.Unlock()
-	common.Must(c.cacheCleanup.Start())
+	if !c.serveStale || c.serveExpiredTTL != 0 {
+		common.Must(c.cacheCleanup.Start())
+	}
 }
 
-func (c *CacheController) findIPsForDomain(domain string, option dns_feature.IPOption) ([]net.IP, uint32, error) {
+func (c *CacheController) findIPsForDomain(domain string, option dns_feature.IPOption) ([]net.IP, int32, error) {
 	c.RLock()
 	record, found := c.ips[domain]
 	c.RUnlock()
@@ -124,7 +135,7 @@ func (c *CacheController) findIPsForDomain(domain string, option dns_feature.IPO
 
 	var errs []error
 	var allIPs []net.IP
-	var rTTL uint32 = dns_feature.DefaultTTL
+	var rTTL int32 = dns_feature.DefaultTTL
 
 	mergeReq := option.IPv4Enable && option.IPv6Enable
 

app/dns/config.proto

@@ -32,6 +32,8 @@ message NameServer {
   string tag = 9;
   uint64 timeoutMs = 10;
   bool disableCache = 11;
+  bool serveStale = 15;
+  optional uint32 serveExpiredTTL = 16;
   bool finalQuery = 12;
   repeated xray.app.router.GeoIP unexpected_geoip = 13;
   bool actUnprior = 14;
@@ -80,6 +82,8 @@ message Config {
 
   // DisableCache disables DNS cache
   bool disableCache = 8;
+  bool serveStale = 12;
+  uint32 serveExpiredTTL = 13;
 
   QueryStrategy query_strategy = 9;
 

app/dns/dns.go

@@ -118,6 +118,11 @@ func New(ctx context.Context, config *Config) (*DNS, error) {
 		}
 
 		disableCache := config.DisableCache || ns.DisableCache
+		serveStale := config.ServeStale || ns.ServeStale
+		serveExpiredTTL := config.ServeExpiredTTL
+		if ns.ServeExpiredTTL != nil {
+			serveExpiredTTL = *ns.ServeExpiredTTL
+		}
 
 		var tag = defaultTag
 		if len(ns.Tag) > 0 {
@@ -128,7 +133,7 @@ func New(ctx context.Context, config *Config) (*DNS, error) {
 			return nil, errors.New("no QueryStrategy available for ", ns.Address)
 		}
 
-		client, err := NewClient(ctx, ns, myClientIP, disableCache, tag, clientIPOption, &matcherInfos, updateDomain)
+		client, err := NewClient(ctx, ns, myClientIP, disableCache, serveStale, serveExpiredTTL, tag, clientIPOption, &matcherInfos, updateDomain)
 		if err != nil {
 			return nil, errors.New("failed to create client").Base(err)
 		}

app/dns/dnscommon.go

@@ -3,6 +3,7 @@ package dns
 import (
 	"context"
 	"encoding/binary"
+	"math"
 	"strings"
 	"time"
 
@@ -38,19 +39,14 @@ type IPRecord struct {
 	RawHeader *dnsmessage.Header
 }
 
-func (r *IPRecord) getIPs() ([]net.IP, uint32, error) {
+func (r *IPRecord) getIPs() ([]net.IP, int32, error) {
 	if r == nil {
 		return nil, 0, errRecordNotFound
 	}
+
 	untilExpire := time.Until(r.Expire).Seconds()
-	if untilExpire <= 0 {
-		return nil, 0, errRecordNotFound
-	}
+	ttl := int32(math.Ceil(untilExpire))
 
-	ttl := uint32(untilExpire) + 1
-	if ttl == 1 {
-		r.Expire = time.Now().Add(time.Second) // To ensure that two consecutive requests get the same result
-	}
 	if r.RCode != dnsmessage.RCodeSuccess {
 		return nil, ttl, dns_feature.RCodeError(r.RCode)
 	}

app/dns/nameserver.go

@@ -41,7 +41,7 @@ type Client struct {
 }
 
 // NewServer creates a name server object according to the network destination url.
-func NewServer(ctx context.Context, dest net.Destination, dispatcher routing.Dispatcher, disableCache bool, clientIP net.IP) (Server, error) {
+func NewServer(ctx context.Context, dest net.Destination, dispatcher routing.Dispatcher, disableCache bool, serveStale bool, serveExpiredTTL uint32, clientIP net.IP) (Server, error) {
 	if address := dest.Address; address.Family().IsDomain() {
 		u, err := url.Parse(address.Domain())
 		if err != nil {
@@ -51,19 +51,19 @@ func NewServer(ctx context.Context, dest net.Destination, dispatcher routing.Dis
 		case strings.EqualFold(u.String(), "localhost"):
 			return NewLocalNameServer(), nil
 		case strings.EqualFold(u.Scheme, "https"): // DNS-over-HTTPS Remote mode
-			return NewDoHNameServer(u, dispatcher, false, disableCache, clientIP), nil
+			return NewDoHNameServer(u, dispatcher, false, disableCache, serveStale, serveExpiredTTL, clientIP), nil
 		case strings.EqualFold(u.Scheme, "h2c"): // DNS-over-HTTPS h2c Remote mode
-			return NewDoHNameServer(u, dispatcher, true, disableCache, clientIP), nil
+			return NewDoHNameServer(u, dispatcher, true, disableCache, serveStale, serveExpiredTTL, clientIP), nil
 		case strings.EqualFold(u.Scheme, "https+local"): // DNS-over-HTTPS Local mode
-			return NewDoHNameServer(u, nil, false, disableCache, clientIP), nil
+			return NewDoHNameServer(u, nil, false, disableCache, serveStale, serveExpiredTTL, clientIP), nil
 		case strings.EqualFold(u.Scheme, "h2c+local"): // DNS-over-HTTPS h2c Local mode
-			return NewDoHNameServer(u, nil, true, disableCache, clientIP), nil
+			return NewDoHNameServer(u, nil, true, disableCache, serveStale, serveExpiredTTL, clientIP), nil
 		case strings.EqualFold(u.Scheme, "quic+local"): // DNS-over-QUIC Local mode
-			return NewQUICNameServer(u, disableCache, clientIP)
+			return NewQUICNameServer(u, disableCache, serveStale, serveExpiredTTL, clientIP)
 		case strings.EqualFold(u.Scheme, "tcp"): // DNS-over-TCP Remote mode
-			return NewTCPNameServer(u, dispatcher, disableCache, clientIP)
+			return NewTCPNameServer(u, dispatcher, disableCache, serveStale, serveExpiredTTL, clientIP)
 		case strings.EqualFold(u.Scheme, "tcp+local"): // DNS-over-TCP Local mode
-			return NewTCPLocalNameServer(u, disableCache, clientIP)
+			return NewTCPLocalNameServer(u, disableCache, serveStale, serveExpiredTTL, clientIP)
 		case strings.EqualFold(u.String(), "fakedns"):
 			var fd dns.FakeDNSEngine
 			err = core.RequireFeatures(ctx, func(fdns dns.FakeDNSEngine) {
@@ -79,7 +79,7 @@ func NewServer(ctx context.Context, dest net.Destination, dispatcher routing.Dis
 		dest.Network = net.Network_UDP
 	}
 	if dest.Network == net.Network_UDP { // UDP classic DNS mode
-		return NewClassicNameServer(dest, dispatcher, disableCache, clientIP), nil
+		return NewClassicNameServer(dest, dispatcher, disableCache, serveStale, serveExpiredTTL, clientIP), nil
 	}
 	return nil, errors.New("No available name server could be created from ", dest).AtWarning()
 }
@@ -89,7 +89,7 @@ func NewClient(
 	ctx context.Context,
 	ns *NameServer,
 	clientIP net.IP,
-	disableCache bool,
+	disableCache bool, serveStale bool, serveExpiredTTL uint32,
 	tag string,
 	ipOption dns.IPOption,
 	matcherInfos *[]*DomainMatcherInfo,
@@ -99,7 +99,7 @@ func NewClient(
 
 	err := core.RequireFeatures(ctx, func(dispatcher routing.Dispatcher) error {
 		// Create a new server for each client for now
-		server, err := NewServer(ctx, ns.Address.AsDestination(), dispatcher, disableCache, clientIP)
+		server, err := NewServer(ctx, ns.Address.AsDestination(), dispatcher, disableCache, serveStale, serveExpiredTTL, clientIP)
 		if err != nil {
 			return errors.New("failed to create nameserver").Base(err).AtWarning()
 		}

app/dns/nameserver_doh.go

@@ -38,15 +38,15 @@ type DoHNameServer struct {
 }
 
 // NewDoHNameServer creates DOH/DOHL client object for remote/local resolving.
-func NewDoHNameServer(url *url.URL, dispatcher routing.Dispatcher, h2c bool, disableCache bool, clientIP net.IP) *DoHNameServer {
+func NewDoHNameServer(url *url.URL, dispatcher routing.Dispatcher, h2c bool, disableCache bool, serveStale bool, serveExpiredTTL uint32, clientIP net.IP) *DoHNameServer {
 	url.Scheme = "https"
 	mode := "DOH"
 	if dispatcher == nil {
 		mode = "DOHL"
 	}
 	errors.LogInfo(context.Background(), "DNS: created ", mode, " client for ", url.String(), ", with h2c ", h2c)
 	s := &DoHNameServer{
-		cacheController: NewCacheController(mode+"//"+url.Host, disableCache),
+		cacheController: NewCacheController(mode+"//"+url.Host, disableCache, serveStale, serveExpiredTTL),
 		dohURL:          url.String(),
 		clientIP:        clientIP,
 	}
@@ -126,7 +126,9 @@ func (s *DoHNameServer) sendQuery(ctx context.Context, noResponseErrCh chan<- er
 
 	if s.Name()+"." == "DOH//"+domain {
 		errors.LogError(ctx, s.Name(), " tries to resolve itself! Use IP or set \"hosts\" instead.")
-		noResponseErrCh <- errors.New("tries to resolve itself!", s.Name())
+		if noResponseErrCh != nil {
+			noResponseErrCh <- errors.New("tries to resolve itself!", s.Name())
+		}
 		return
 	}
 
@@ -167,19 +169,25 @@ func (s *DoHNameServer) sendQuery(ctx context.Context, noResponseErrCh chan<- er
 			b, err := dns.PackMessage(r.msg)
 			if err != nil {
 				errors.LogErrorInner(ctx, err, "failed to pack dns query for ", domain)
-				noResponseErrCh <- err
+				if noResponseErrCh != nil {
+					noResponseErrCh <- err
+				}
 				return
 			}
 			resp, err := s.dohHTTPSContext(dnsCtx, b.Bytes())
 			if err != nil {
 				errors.LogErrorInner(ctx, err, "failed to retrieve response for ", domain)
-				noResponseErrCh <- err
+				if noResponseErrCh != nil {
+					noResponseErrCh <- err
+				}
 				return
 			}
 			rec, err := parseResponse(resp)
 			if err != nil {
 				errors.LogErrorInner(ctx, err, "failed to handle DOH response for ", domain)
-				noResponseErrCh <- err
+				if noResponseErrCh != nil {
+					noResponseErrCh <- err
+				}
 				return
 			}
 			s.cacheController.updateIP(r, rec)
@@ -218,20 +226,28 @@ func (s *DoHNameServer) dohHTTPSContext(ctx context.Context, b []byte) ([]byte,
 // QueryIP implements Server.
 func (s *DoHNameServer) QueryIP(ctx context.Context, domain string, option dns_feature.IPOption) ([]net.IP, uint32, error) { // nolint: dupl
 	fqdn := Fqdn(domain)
-	sub4, sub6 := s.cacheController.registerSubscribers(fqdn, option)
-	defer closeSubscribers(sub4, sub6)
 
 	if s.cacheController.disableCache {
 		errors.LogDebug(ctx, "DNS cache is disabled. Querying IP for ", domain, " at ", s.Name())
 	} else {
 		ips, ttl, err := s.cacheController.findIPsForDomain(fqdn, option)
 		if !go_errors.Is(err, errRecordNotFound) {
-			errors.LogDebugInner(ctx, err, s.Name(), " cache HIT ", domain, " -> ", ips)
-			log.Record(&log.DNSLog{Server: s.Name(), Domain: domain, Result: ips, Status: log.DNSCacheHit, Elapsed: 0, Error: err})
-			return ips, ttl, err
+			if ttl > 0 {
+				errors.LogDebugInner(ctx, err, s.Name(), " cache HIT ", domain, " -> ", ips)
+				log.Record(&log.DNSLog{Server: s.Name(), Domain: domain, Result: ips, Status: log.DNSCacheHit, Elapsed: 0, Error: err})
+				return ips, uint32(ttl), err
+			}
+			if s.cacheController.serveStale && (s.cacheController.serveExpiredTTL == 0 || s.cacheController.serveExpiredTTL < ttl) {
+				errors.LogDebugInner(ctx, err, s.Name(), " cache OPTIMISTE ", domain, " -> ", ips)
+				s.sendQuery(ctx, nil, fqdn, option)
+				return ips, 1, err
+			}
 		}
 	}
 
+	sub4, sub6 := s.cacheController.registerSubscribers(fqdn, option)
+	defer closeSubscribers(sub4, sub6)
+
 	noResponseErrCh := make(chan error, 2)
 	s.sendQuery(ctx, noResponseErrCh, fqdn, option)
 	start := time.Now()
@@ -259,6 +275,11 @@ func (s *DoHNameServer) QueryIP(ctx context.Context, domain string, option dns_f
 
 	ips, ttl, err := s.cacheController.findIPsForDomain(fqdn, option)
 	log.Record(&log.DNSLog{Server: s.Name(), Domain: domain, Result: ips, Status: log.DNSQueried, Elapsed: time.Since(start), Error: err})
-	return ips, ttl, err
-
+	var rTTL uint32
+	if ttl <= 0 {
+		rTTL = 1
+	} else {
+		rTTL = uint32(ttl)
+	}
+	return ips, rTTL, err
 }

app/dns/nameserver_doh_test.go

@@ -17,7 +17,7 @@ func TestDOHNameServer(t *testing.T) {
 	url, err := url.Parse("https+local://1.1.1.1/dns-query")
 	common.Must(err)
 
-	s := NewDoHNameServer(url, nil, false, false, net.IP(nil))
+	s := NewDoHNameServer(url, nil, false, false, false, 0, net.IP(nil))
 	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
 	ips, _, err := s.QueryIP(ctx, "google.com", dns_feature.IPOption{
 		IPv4Enable: true,
@@ -34,7 +34,7 @@ func TestDOHNameServerWithCache(t *testing.T) {
 	url, err := url.Parse("https+local://1.1.1.1/dns-query")
 	common.Must(err)
 
-	s := NewDoHNameServer(url, nil, false, false, net.IP(nil))
+	s := NewDoHNameServer(url, nil, false, false, false, 0, net.IP(nil))
 	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
 	ips, _, err := s.QueryIP(ctx, "google.com", dns_feature.IPOption{
 		IPv4Enable: true,
@@ -62,7 +62,7 @@ func TestDOHNameServerWithIPv4Override(t *testing.T) {
 	url, err := url.Parse("https+local://1.1.1.1/dns-query")
 	common.Must(err)
 
-	s := NewDoHNameServer(url, nil, false, false, net.IP(nil))
+	s := NewDoHNameServer(url, nil, false, false, false, 0, net.IP(nil))
 	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
 	ips, _, err := s.QueryIP(ctx, "google.com", dns_feature.IPOption{
 		IPv4Enable: true,
@@ -85,7 +85,7 @@ func TestDOHNameServerWithIPv6Override(t *testing.T) {
 	url, err := url.Parse("https+local://1.1.1.1/dns-query")
 	common.Must(err)
 
-	s := NewDoHNameServer(url, nil, false, false, net.IP(nil))
+	s := NewDoHNameServer(url, nil, false, false, false, 0, net.IP(nil))
 	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
 	ips, _, err := s.QueryIP(ctx, "google.com", dns_feature.IPOption{
 		IPv4Enable: false,