Add mdns package and Client implementation (#55)

* add mdns

* define mdns.Client

* fix some mdns stuff

* refine dns package for use with mdns

* remove Querier and Responder and replace with Client

* add record setting methods on dns.record to reuse record buffer

* protect against unbounded client answer growth

* round off sharp mdns edges; reduce allocs

* add mutlicast to stacks; add xnet tests for mdns

* fix documentation on acceptmulticast field

* mdns working

dns/client.go

@@ -34,7 +34,7 @@ func (sudp *Client) ConnectionID() *uint64 { return &sudp.connID }
 func (c *Client) StartResolve(localPort, txid uint16, cfg ResolveConfig) error {
 	nd := len(cfg.Questions)
 	if nd > math.MaxUint16 {
-		return lneto.ErrBufferFull
+		return lneto.ErrInvalidConfig
 	}
 	c.reset(localPort, txid, dnsSendQuery, cfg.EnableRecursion)
 	c.msg.LimitResourceDecoding(uint16(nd), uint16(nd), 0, 0)

dns/dns.go

@@ -57,6 +57,13 @@ type Name struct {
 	data []byte
 }
 
+// NamesEqual reports whether two DNS names are equal by comparing
+// their wire-format representations directly. This is case-sensitive;
+// for case-insensitive comparison use [NamesEqualFold].
+func NamesEqual(a, b Name) bool {
+	return internal.BytesEqual(a.data, b.data)
+}
+
 type ZFlags uint16
 
 func NewResource(name Name, typ Type, class Class, ttl uint32, data []byte) Resource {
@@ -86,83 +93,91 @@ func (r *Resource) SetEDNS0(UDPlength uint16, rcode RCode, zflags ZFlags, data [
 	r.data = append(r.data[:0], data...)
 }
 
-// Decode decodes the DNS message in b into m. It returns the number of bytes
+// DecodeMessage decodes the DNS message into question, answer, authority and additional resources.
+// It returns the number of bytes
 // consumed from b (0 if no bytes were consumed) and any error encountered.
 // If the message was not completely parsed due to LimitResourceDecoding,
 // incompleteButOK is true and an error is returned, though the message is still usable.
-func (m *Message) Decode(msg []byte) (_ uint16, incompleteButOK bool, err error) {
-	if len(msg) < SizeHeader {
-		return 0, false, errBaseLen
-	} else if len(msg) > math.MaxUint16 {
-		return 0, false, errResTooLong
-	}
-	m.Reset()
+//
+// The slice memory is overwritten and capacity used as the limit of encoding.
+// If the argument slice is nil it is skipped for decoding but does not prevent further decoding
+// of other answers, authorities or additionals from being decoded.
+func DecodeMessage(q *[]Question, answers, authorities, additionals *[]Resource, msg []byte) (_ uint16, incompleteButOK bool, err error) {
 	hdr, err := NewFrame(msg)
 	if err != nil {
 		return 0, false, err
 	}
-	nq := int(hdr.QDCount())
+	qd := hdr.QDCount()
+	nq := int(qd)
 	off := uint16(SizeHeader)
 	// Return tooManyErr if found to flag to the caller that the message was
 	// decoded but contained too many resources to decode completely.
-
 	var tooManyErr error
 	switch {
-	case nq > cap(m.Questions):
+	case nq > caporzero(q):
 		tooManyErr = errTooManyQuestions
-	case hdr.ANCount() > uint16(cap(m.Answers)):
+	case int(hdr.ANCount()) > caporzero(answers):
 		tooManyErr = errTooManyAnswers
-	case hdr.NSCount() > uint16(cap(m.Authorities)):
+	case int(hdr.NSCount()) > caporzero(authorities):
 		tooManyErr = errTooManyAuthorities
-	case hdr.ARCount() > uint16(cap(m.Additionals)):
+	case int(hdr.ARCount()) > caporzero(additionals):
 		tooManyErr = errTooManyAdditionals
 	}
-	if nq > cap(m.Questions) {
-		nq = cap(m.Questions)
-	}
-	m.Questions = m.Questions[:nq]
-	for i := 0; i < nq; i++ {
-		off, err = m.Questions[i].Decode(msg, off)
-		if err != nil {
-			m.Questions = m.Questions[:i] // Trim non-decoded/failed questions.
-			return off, false, err
+	if q != nil {
+		if nq > cap(*q) {
+			nq = cap(*q)
 		}
+		*q = (*q)[:nq]
+		for i := 0; i < nq; i++ {
+			off, err = (*q)[i].Decode(msg, off)
+			if err != nil {
+				*q = (*q)[:i] // Trim non-decoded/failed questions.
+				return off, false, err
+			}
+		}
+	} else {
+		nq = 0 // No question slice provided, skip all questions below.
 	}
 	// Skip undecoded questions.
-	qd := hdr.QDCount()
 	for i := 0; i < int(qd)-nq; i++ {
 		off, err = skipQuestion(msg, off)
 		if err != nil {
 			return off, false, err
 		}
 	}
-
-	off, err = decodeToCapResources(&m.Answers, msg, hdr.ANCount(), off)
+	off, err = decodeToCapResources(answers, msg, hdr.ANCount(), off)
 	if err != nil {
 		return off, false, err
 	}
-	off, err = decodeToCapResources(&m.Authorities, msg, hdr.NSCount(), off)
+	off, err = decodeToCapResources(authorities, msg, hdr.NSCount(), off)
 	if err != nil {
 		return off, false, err
 	}
-	off, err = decodeToCapResources(&m.Additionals, msg, hdr.ARCount(), off)
+	off, err = decodeToCapResources(additionals, msg, hdr.ARCount(), off)
 	if err != nil {
 		return off, false, err
 	}
 	return off, tooManyErr != nil, tooManyErr
 }
 
+// Decode decodes the DNS message in b into m. It is a convenience wrapper for [DecodeMessage].
+func (m *Message) Decode(msg []byte) (_ uint16, incompleteButOK bool, err error) {
+	return DecodeMessage(&m.Questions, &m.Answers, &m.Authorities, &m.Additionals, msg)
+}
+
 func decodeToCapResources(dst *[]Resource, msg []byte, nrec, off uint16) (_ uint16, err error) {
 	originalRec := nrec
-	if nrec > uint16(cap(*dst)) {
-		nrec = uint16(cap(*dst)) // Decode up to cap. Caller will return an error flag.
-	}
-	*dst = (*dst)[:nrec]
-	for i := uint16(0); i < nrec; i++ {
-		off, err = (*dst)[i].Decode(msg, off)
-		if err != nil {
-			*dst = (*dst)[:i] // Trim non-decoded/failed resources.
-			return off, err
+	if dst != nil {
+		if nrec > uint16(cap(*dst)) {
+			nrec = uint16(cap(*dst)) // Decode up to cap. Caller will return an error flag.
+		}
+		*dst = (*dst)[:nrec]
+		for i := uint16(0); i < nrec; i++ {
+			off, err = (*dst)[i].Decode(msg, off)
+			if err != nil {
+				*dst = (*dst)[:i] // Trim non-decoded/failed resources.
+				return off, err
+			}
 		}
 	}
 	// Parse undecoded resources, effectively skipping them.
@@ -468,6 +483,20 @@ func NewName(domain string) (Name, error) {
 	return name, nil
 }
 
+// TrimLabels returns a Name sharing the same backing data with the first n labels removed.
+// For example, trimming 1 label from "My Web._http._tcp.local" yields "_http._tcp.local".
+// Returns an empty Name if n exceeds the number of labels.
+func (n Name) TrimLabels(skip int) Name {
+	off := 0
+	for i := 0; i < skip; i++ {
+		if off >= len(n.data) {
+			return Name{}
+		}
+		off += 1 + int(n.data[off])
+	}
+	return Name{data: n.data[off:]}
+}
+
 // Len returns the length over-the-wire of the encoded Name.
 func (n *Name) Len() uint16 {
 	if len(n.data) > math.MaxUint16 {
@@ -666,6 +695,48 @@ func (dst *Resource) CopyFrom(r Resource) {
 	dst.data = append(dst.data[:0], r.data...)
 }
 
+// SetA sets an A (IPv4 address) resource record, reusing internal buffers.
+func (r *Resource) SetA(name Name, class Class, ttl uint32, addr []byte) {
+	r.setHeader(name, TypeA, class, ttl)
+	r.data = append(r.data[:0], addr...)
+	r.header.Length = uint16(len(r.data))
+}
+
+// SetPTR sets a PTR (pointer) resource record, reusing internal buffers.
+func (r *Resource) SetPTR(name Name, class Class, ttl uint32, target Name) {
+	r.setHeader(name, TypePTR, class, ttl)
+	r.data, _ = target.AppendTo(r.data[:0])
+	r.header.Length = uint16(len(r.data))
+}
+
+// SetSRV sets a SRV (service locator) resource record, reusing internal buffers.
+func (r *Resource) SetSRV(name Name, class Class, ttl uint32, priority, weight, port uint16, target Name) {
+	r.setHeader(name, TypeSRV, class, ttl)
+	r.data = binary.BigEndian.AppendUint16(r.data[:0], priority)
+	r.data = binary.BigEndian.AppendUint16(r.data, weight)
+	r.data = binary.BigEndian.AppendUint16(r.data, port)
+	r.data, _ = target.AppendTo(r.data)
+	r.header.Length = uint16(len(r.data))
+}
+
+// SetTXT sets a TXT resource record, reusing internal buffers.
+func (r *Resource) SetTXT(name Name, class Class, ttl uint32, txt []byte) {
+	r.setHeader(name, TypeTXT, class, ttl)
+	if len(txt) == 0 {
+		r.data = append(r.data[:0], 0)
+	} else {
+		r.data = append(r.data[:0], txt...)
+	}
+	r.header.Length = uint16(len(r.data))
+}
+
+func (r *Resource) setHeader(name Name, typ Type, class Class, ttl uint32) {
+	r.header.Name.CopyFrom(name)
+	r.header.Type = typ
+	r.header.Class = class
+	r.header.TTL = ttl
+}
+
 func (dst *ResourceHeader) CopyFrom(rh ResourceHeader) {
 	dst.Name.CopyFrom(rh.Name)
 	dst.Type = rh.Type
@@ -673,3 +744,10 @@ func (dst *ResourceHeader) CopyFrom(rh ResourceHeader) {
 	dst.TTL = rh.TTL
 	dst.Length = rh.Length
 }
+
+func caporzero[T any](v *[]T) int {
+	if v == nil {
+		return 0
+	}
+	return cap(*v)
+}

examples/xcurl/main.go

@@ -117,6 +117,7 @@ func run() (err error) {
 	}
 	fmt.Println("NIC hardware address:", net.HardwareAddr(nicHW[:]).String(), "bridgeHW:", net.HardwareAddr(brHW[:]).String(), "mtu:", mtu, "addr:", nicAddr.String())
 	var stack xnet.StackAsync
+
 	err = stack.Reset(xnet.StackConfig{
 		Hostname:        "xnet-test",
 		RandSeed:        softRand,
@@ -221,6 +222,7 @@ func run() (err error) {
 		return fmt.Errorf("DHCP failed: %w", err)
 	}
 	timeDHCP()
+
 	err = stack.AssimilateDHCPResults(results)
 	if err != nil {
 		return fmt.Errorf("assimilating DHCP results: %w", err)

http/httpraw/parse.go

@@ -295,12 +295,14 @@ func (h *Header) reuseOrAppend(tok headerSlice, value string) headerSlice {
 }
 
 func (h *Header) appendSlice(value string) headerSlice {
+	debuglog("http:appendslice:start")
 	free := h.hbuf.free()
 	if len(value) > free {
 		if h.flags.hasAny(flagNoBufferGrow) {
 			h.flags |= flagOOMReached
 			return headerSlice{}
 		}
+		debuglog("http:appendslice:grow-buf")
 		h.hbuf.buf = slices.Grow(h.hbuf.buf, len(value)+1) // Grow 1 beyond due to slice validity.
 	}
 	h.flags |= flagMangledBuffer

internal/ip.go

@@ -52,3 +52,41 @@ func SetIPAddrs(buf []byte, id uint16, src, dst []byte) (err error) {
 	copy(dstaddr, dst)
 	return nil
 }
+
+// SetMulticast sets the IP destination to multicastAddr and derives the
+// Ethernet destination MAC from it. It supports IPv4 (RFC 1112 §6.4) and
+// IPv6 (RFC 2464 §7) multicast MAC mapping.
+func SetMulticast(ethernetCarrier []byte, ipOff int, multicastAddr []byte) (err error) {
+	ip := ethernetCarrier[ipOff:]
+	mac := ethernetCarrier[0:6]
+	version := ip[0] >> 4
+	switch version {
+	case 4:
+		if len(multicastAddr) != 4 {
+			return lneto.ErrMismatchLen
+		}
+		copy(ip[16:20], multicastAddr)
+		// IPv4 multicast MAC: 01:00:5e + low 23 bits of IP destination.
+		mac[0] = 0x01
+		mac[1] = 0x00
+		mac[2] = 0x5e
+		mac[3] = multicastAddr[1] & 0x7f
+		mac[4] = multicastAddr[2]
+		mac[5] = multicastAddr[3]
+	case 6:
+		if len(multicastAddr) != 16 {
+			return lneto.ErrMismatchLen
+		}
+		copy(ip[24:40], multicastAddr)
+		// IPv6 multicast MAC: 33:33 + last 4 bytes of IP destination.
+		mac[0] = 0x33
+		mac[1] = 0x33
+		mac[2] = multicastAddr[12]
+		mac[3] = multicastAddr[13]
+		mac[4] = multicastAddr[14]
+		mac[5] = multicastAddr[15]
+	default:
+		return lneto.ErrUnsupported
+	}
+	return nil
+}

internet/pcap/capture.go

@@ -19,6 +19,7 @@ import (
 	"github.com/soypat/lneto/ipv4"
 	"github.com/soypat/lneto/ipv4/icmpv4"
 	"github.com/soypat/lneto/ipv6"
+	"github.com/soypat/lneto/mdns"
 	"github.com/soypat/lneto/ntp"
 	"github.com/soypat/lneto/tcp"
 	"github.com/soypat/lneto/udp"
@@ -359,7 +360,7 @@ func (pc *PacketBreakdown) CaptureUDP(dst []Frame, pkt []byte, bitOffset int) ([
 	srcport := ufrm.SourcePort()
 	if dhcpv4.PayloadIsDHCPv4(payload) {
 		dst, err = pc.CaptureDHCPv4(dst, pkt, end)
-	} else if dstport == dns.ServerPort || srcport == dns.ServerPort {
+	} else if dstport == dns.ServerPort || srcport == dns.ServerPort || dstport == mdns.Port || srcport == mdns.Port {
 		dst, err = pc.CaptureDNS(dst, pkt, end)
 	} else if dstport == ntp.ServerPort || srcport == ntp.ServerPort {
 		dst, err = pc.CaptureNTP(dst, pkt, end)
@@ -437,7 +438,7 @@ func (pc *PacketBreakdown) CaptureDNS(dst []Frame, pkt []byte, bitOffset int) ([
 		return dst, errNotByteAligned
 	}
 	dnsData := pkt[bitOffset/8:]
-	pc.dmsg.LimitResourceDecoding(20, 20, 20, 20)
+	pc.dmsg.LimitResourceDecoding(4, 4, 4, 4)
 	off, incomplete, err := pc.dmsg.Decode(dnsData)
 	if err != nil && !incomplete {
 		return dst, err
@@ -1265,6 +1266,43 @@ var baseDHCPv4Fields = [...]FrameField{
 	},
 }
 
+var baseDNSFields = [...]FrameField{
+	{
+		Class:          FieldClassID,
+		FrameBitOffset: 0,
+		BitLength:      2 * octet,
+	},
+	{
+		Class:          FieldClassFlags,
+		FrameBitOffset: 2 * octet,
+		BitLength:      2 * octet,
+	},
+	{
+		Name:           "Questions",
+		Class:          FieldClassSize,
+		FrameBitOffset: 4 * octet,
+		BitLength:      2 * octet,
+	},
+	{
+		Name:           "Answers",
+		Class:          FieldClassSize,
+		FrameBitOffset: 6 * octet,
+		BitLength:      2 * octet,
+	},
+	{
+		Name:           "Authorities",
+		Class:          FieldClassSize,
+		FrameBitOffset: 8 * octet,
+		BitLength:      2 * octet,
+	},
+	{
+		Name:           "Additionals",
+		Class:          FieldClassSize,
+		FrameBitOffset: 10 * octet,
+		BitLength:      2 * octet,
+	},
+}
+
 var baseNTPFields = [...]FrameField{
 	{
 		Name:           "Mode",

internet/stack-ethernet.go

@@ -33,11 +33,12 @@ type StackEthernetConfig struct {
 }
 
 type StackEthernet struct {
-	connID   uint64
-	handlers handlers
-	mac      [6]byte
-	gwmac    [6]byte
-	mtu      uint16
+	connID          uint64
+	handlers        handlers
+	mac             [6]byte
+	gwmac           [6]byte
+	mtu             uint16
+	acceptMulticast bool
 	// crcupdate set when crc32 has been configured to be appended.
 	crcupdate func(crc uint32, p []byte) uint32
 }
@@ -50,6 +51,10 @@ func (ls *StackEthernet) Gateway6() (gw [6]byte) {
 	return ls.gwmac
 }
 
+func (ls *StackEthernet) SetAcceptMulticast(accept bool) {
+	ls.acceptMulticast = accept
+}
+
 func (ls *StackEthernet) SetHardwareAddr6(mac [6]byte) {
 	ls.mac = mac
 }
@@ -83,11 +88,12 @@ func (ls *StackEthernet) Configure(cfg StackEthernetConfig) error {
 	}
 	ls.handlers.reset("StackEthernet", cfg.MaxNodes)
 	*ls = StackEthernet{
-		connID:   ls.connID + 1,
-		handlers: ls.handlers,
-		mac:      cfg.MAC,
-		gwmac:    cfg.Gateway,
-		mtu:      uint16(cfg.MTU),
+		connID:          ls.connID + 1,
+		handlers:        ls.handlers,
+		mac:             cfg.MAC,
+		gwmac:           cfg.Gateway,
+		mtu:             uint16(cfg.MTU),
+		acceptMulticast: ls.acceptMulticast,
 	}
 	if cfg.AppendCRC32 {
 		ls.crcupdate = cfg.CRC32Update
@@ -121,7 +127,9 @@ func (ls *StackEthernet) Demux(carrierData []byte, frameOffset int) (err error)
 	dstaddr := efrm.DestinationHardwareAddr()
 	var vld lneto.Validator
 	if !efrm.IsBroadcast() && ls.mac != *dstaddr {
-		goto DROP
+		if !ls.acceptMulticast || dstaddr[0]&1 == 0 {
+			goto DROP
+		}
 	}
 	efrm.ValidateSize(&vld)
 	if vld.HasError() {

internet/stack-ip.go

@@ -16,11 +16,12 @@ import (
 var _ StackNode = (*StackIP)(nil)
 
 type StackIP struct {
-	connID    uint64
-	ipID      uint16
-	ip        [4]byte
-	validator lneto.Validator
-	handlers  handlers
+	connID          uint64
+	ipID            uint16
+	ip              [4]byte
+	acceptMulticast bool
+	validator       lneto.Validator
+	handlers        handlers
 }
 
 func (sb *StackIP) Reset(addr netip.Addr, maxNodes int) error {
@@ -33,10 +34,11 @@ func (sb *StackIP) Reset(addr netip.Addr, maxNodes int) error {
 	}
 	sb.handlers.reset("StackIP", maxNodes)
 	*sb = StackIP{
-		connID:    sb.connID + 1,
-		validator: sb.validator,
-		handlers:  sb.handlers,
-		ip:        sb.ip,
+		connID:          sb.connID + 1,
+		validator:       sb.validator,
+		handlers:        sb.handlers,
+		ip:              sb.ip,
+		acceptMulticast: sb.acceptMulticast,
 	}
 	return nil
 }
@@ -65,6 +67,10 @@ func (sb *StackIP) Addr() netip.Addr {
 	return netip.AddrFrom4(sb.ip)
 }
 
+func (sb *StackIP) SetAcceptMulticast(accept bool) {
+	sb.acceptMulticast = accept
+}
+
 func (sb *StackIP) SetLogger(logger *slog.Logger) {
 	sb.handlers.log = logger
 }
@@ -79,8 +85,10 @@ func (sb *StackIP) Demux(carrierData []byte, offset int) error {
 	}
 	dst := ifrm.DestinationAddr()
 	if sb.ip != ([4]byte{}) && *dst != sb.ip {
-		sb.handlers.debug("ip:not-for-us")
-		return lneto.ErrPacketDrop // Not meant for us.
+		if !sb.acceptMulticast || dst[0]&0xF0 != 0xE0 {
+			sb.handlers.debug("ip:not-for-us")
+			return lneto.ErrPacketDrop // Not meant for us.
+		}
 	}
 
 	sb.validator.ResetErr()

ipv4/definitions.go

@@ -6,6 +6,14 @@ const (
 	sizeHeader = 20
 )
 
+// IsMulticast reports whether addr is an IPv4 multicast address (224.0.0.0/4),
+// i.e. the most significant nibble is 0xE (1110 in binary) as defined in [RFC1112].
+//
+// [RFC1112]: https://datatracker.ietf.org/doc/html/rfc1112
+func IsMulticast(addr [4]byte) bool {
+	return addr[0]&0xf0 == 0xe0
+}
+
 // ToS represents the Traffic Class (a.k.a Type of Service). It is 8 bits long. 6 MSB are Differentiated Services; 2 LSB are Explicit Congenstion Notification.
 type ToS uint8
 

mdns/client.go

@@ -0,0 +1,263 @@
+package mdns
+
+import (
+	"math"
+	"net"
+
+	"github.com/soypat/lneto"
+	"github.com/soypat/lneto/dns"
+	"github.com/soypat/lneto/internal"
+)
+
+const (
+	// Port is the mDNS UDP port (RFC 6762 §1).
+	Port = 5353
+
+	// Default TTL for mDNS records (RFC 6762 §11).
+	DefaultTTL uint32 = 120
+	// classCacheFlush is bit 15 of the Class field, indicating the record
+	// is from a unique source and should replace cached entries (RFC 6762 §10.2).
+	classCacheFlush uint16 = 1 << 15
+	mdnsTxID               = 0
+	mdnsFlags              = 0
+)
+
+type querierState uint8
+
+const (
+	querierIdle          querierState = iota
+	querierSendQuery                  // Query ready to be sent.
+	querierAwaitResponse              // Waiting for answers.
+	querierFailed                     // failed query
+	querierDone                       // Answers collected.
+)
+
+// Client provides both querying and service multicast DNS functionality
+// once configured and attached to MDNS port 5353.
+//
+// Clients are attached to MDNS ports and function until manual detachment
+// due to their dual design: they double as a querier and service discovery.
+type Client struct {
+	connID uint64
+	closed bool
+	lport  uint16
+	ip     []byte
+	// Query State:
+	qstate querierState
+	qcode  dns.RCode
+	qerr   error
+	// qmsg is used for queries to marshal/unmarshal our
+	// outgoing queries and responses to our queries.
+	qans []dns.Resource
+	qqst []dns.Question
+	// Response state:
+	services []Service // Stores services we'd broadcast.
+	rans     []dns.Resource
+	rqst     []dns.Question
+}
+
+type ClientConfig struct {
+	LocalPort     uint16
+	Services      []Service
+	MulticastAddr []byte
+}
+
+func (c *Client) Configure(cfg ClientConfig) error {
+	if cfg.LocalPort == 0 {
+		return lneto.ErrZeroSource
+	}
+	c.reset(cfg.LocalPort)
+	c.services = append(c.services[:0], cfg.Services...)
+	c.ip = append(c.ip[:0], cfg.MulticastAddr...)
+	internal.SliceReuse(&c.rqst, len(cfg.Services))
+	// Each service can produce up to 4 answer records (PTR+SRV+TXT+A).
+	nrans := 2 * len(cfg.Services)
+	if nrans > 0 {
+		nrans = max(4, nrans)
+	}
+	internal.SliceReuse(&c.rans, nrans)
+	return nil
+}
+
+func (c *Client) Protocol() uint64 { return uint64(lneto.IPProtoUDP) }
+
+func (c *Client) LocalPort() uint16 { return c.lport }
+
+func (c *Client) ConnectionID() *uint64 { return &c.connID }
+
+type ResolveConfig struct {
+	Questions          []dns.Question
+	MaxResponseAnswers uint16
+}
+
+func (c *Client) StartResolve(cfg ResolveConfig) error {
+	nq := len(cfg.Questions)
+	if nq > math.MaxUint16 || nq == 0 || cfg.MaxResponseAnswers == 0 {
+		return lneto.ErrInvalidConfig
+	}
+	c.qreset(querierSendQuery)
+	internal.SliceReuse(&c.qans, int(cfg.MaxResponseAnswers))
+	internal.SliceReuse(&c.qqst, nq)
+	c.qqst = c.qqst[:nq]
+	for i := range c.qqst {
+		c.qqst[i].CopyFrom(cfg.Questions[i])
+	}
+	return nil
+}
+
+func (c *Client) reset(localport uint16) {
+	*c = Client{
+		connID: c.connID + 1,
+		lport:  localport,
+		// Ensure memory reused:
+		qqst:     c.qqst[:0],
+		qans:     c.qans[:0],
+		services: c.services[:0],
+		rans:     c.rans[:0],
+		ip:       c.ip[:0],
+	}
+}
+
+// qreset resets the current query state. It is only a partial reset of a Client.
+func (c *Client) qreset(state querierState) {
+	c.qstate = state
+}
+
+// Encapsulate writes a pending mDNS packet into carrierData[offsetToFrame:].
+// Pending responses take priority over outgoing queries.
+func (c *Client) Encapsulate(carrierData []byte, offsetToIP, offsetToFrame int) (n int, err error) {
+	if c.isClosed() {
+		return 0, net.ErrClosed
+	}
+	if len(c.rans) > 0 {
+		// Pending response to an incoming query.
+		n, err = c.encapsResponse(carrierData[offsetToFrame:])
+	} else if c.qstate == querierSendQuery {
+		n, err = c.encapsQuery(carrierData[offsetToFrame:])
+	}
+	if n > 0 && offsetToIP >= 0 {
+		// Set Multicast IP destination and Ethernet MAC.
+		internal.SetMulticast(carrierData, offsetToIP, c.ip)
+	}
+	return n, err
+}
+
+// Demux processes an incoming mDNS response packet. Answers are accumulated
+// into the internal message. Once sufficient answers are collected or a
+// timeout occurs the querier transitions to querierDone.
+func (c *Client) Demux(carrierData []byte, frameOffset int) error {
+	if c.isClosed() {
+		return net.ErrClosed
+	}
+	frame := carrierData[frameOffset:]
+	f, err := dns.NewFrame(frame)
+	if err != nil {
+		return err
+	} else if f.TxID() != 0 {
+		return lneto.ErrPacketDrop
+	}
+	flags := f.Flags()
+	isresponse := flags.IsResponse()
+	if isresponse && c.qstate == querierAwaitResponse {
+		c.qcode = flags.ResponseCode()
+		// Decode response into our message, collecting answers.
+		_, _, c.qerr = dns.DecodeMessage(nil, &c.qans, nil, nil, frame)
+		if c.qerr != nil {
+			c.qstate = querierFailed
+			return c.qerr
+		}
+		c.qstate = querierDone
+		return nil // success.
+	}
+	freeAns := cap(c.rans) - len(c.rans)
+	if !isresponse && len(c.services) > 0 && freeAns > 0 {
+		// Incoming query — match against our services.
+		var query dns.Message
+		query.LimitResourceDecoding(f.QDCount(), 0, 0, 0)
+		_, _, err = query.Decode(frame)
+		if err != nil {
+			return err
+		}
+		for i := range query.Questions {
+			q := &query.Questions[i]
+			for j := range c.services {
+				if matchQuestion(q, &c.services[j]) {
+					addServiceAnswers(&c.rans, q, &c.services[j])
+					break
+				}
+			}
+		}
+	}
+	return nil
+}
+
+func (c *Client) encapsQuery(frame []byte) (int, error) {
+	msg := dns.Message{
+		Questions: c.qqst,
+	}
+	msglen := msg.Len()
+	if int(msglen) > len(frame) {
+		c.qerr = lneto.ErrShortBuffer
+		c.qstate = querierFailed
+		return 0, c.qerr
+	}
+	// mDNS queries use txid=0 and no flags (RFC 6762 §18.1).
+	data, err := msg.AppendTo(frame[:0], mdnsTxID, mdnsFlags)
+	if err != nil {
+		c.qerr = err
+		c.qstate = querierFailed
+		return 0, err
+	} else if len(data) != int(msglen) {
+		panic("bad dns length calculation") // panic since this is a big bug in lneto.
+	}
+	c.qstate = querierAwaitResponse
+	return len(data), nil
+}
+
+func (c *Client) encapsResponse(frame []byte) (int, error) {
+	var msg dns.Message
+	msg.Answers = c.rans
+	msglen := msg.Len()
+	if int(msglen) > len(frame) {
+		return 0, lneto.ErrShortBuffer
+	}
+	// mDNS responses: txid=0, QR=1, AA=1 (RFC 6762 §18.4, §6).
+	flags := dns.HeaderFlags(1<<15 | 1<<10)
+	data, err := msg.AppendTo(frame[:0], 0, flags)
+	if err != nil {
+		return 0, err
+	}
+	return len(data), nil
+}
+
+// Abort closes the client, causing all subsequent Encapsulate/Demux calls to return [net.ErrClosed].
+func (c *Client) Abort() {
+	c.closed = true
+}
+
+func (c *Client) isClosed() bool {
+	return c.closed
+}
+
+// AnswersCopyTo checks if [Client.StartResolve] ended succesfully before
+// doing a deep copy of answers received to the argument buffer using [dns.Resource.CopyFrom].
+func (c *Client) AnswersCopyTo(dst []dns.Resource) (n int, done bool, err error) {
+	if len(dst) == 0 {
+		return 0, false, lneto.ErrShortBuffer
+	} else if c.qstate == querierIdle {
+		return 0, false, net.ErrClosed
+	} else if c.qstate == querierFailed {
+		return 0, false, c.qerr
+	} else if c.qstate != querierDone {
+		return 0, false, nil
+	}
+	for i := range min(len(dst), len(c.qans)) {
+		dst[i].CopyFrom(c.qans[i])
+		n++
+	}
+	rcode := c.qcode
+	if rcode != 0 {
+		return n, true, rcode
+	}
+	return n, true, nil
+}

mdns/definitions.go

@@ -0,0 +1,122 @@
+package mdns
+
+import (
+	"github.com/soypat/lneto/dns"
+)
+
+// Service describes a service to advertise via mDNS.
+// A single Service produces PTR, SRV, TXT, and A resource records.
+//
+// i.e: To generate a hostname styled A record like the one
+// linux machines provide to reach them at hostname.local:
+//
+//	s := Service{
+//		Host: dns.NewName("yourhostname.local"),
+//		Addr: ipAddressSlice,
+//	}
+type Service struct {
+	// Name is the fully-qualified service instance name in wire format,
+	//  e.g. "My Web Server._http._tcp.local".
+	Name dns.Name
+	// Host is the hostname in wire format, e.g. "mydevice.local".
+	Host dns.Name
+	// TXTData is raw TXT record data (length-prefixed strings).
+	TXTData []byte
+	// Addr is the IP address for the A record.
+	Addr []byte
+	// TTL is the record TTL in seconds. Zero uses DefaultTTL.
+	TTL uint32
+	// Port is the TCP/UDP port for the SRV record.
+	Port uint16
+}
+
+func (s *Service) ttl() uint32 {
+	if s.TTL == 0 {
+		return DefaultTTL
+	}
+	return s.TTL
+}
+
+// serviceType extracts the service type portion of the instance name.
+// For "_http._tcp.local" it returns the same; for "My Web._http._tcp.local"
+// it returns "_http._tcp.local" by trimming the first label.
+// Returns a view into the original Name data — zero allocation.
+func (s *Service) serviceType() dns.Name {
+	var totalLabels int
+	s.Name.VisitLabels(func(label []byte) {
+		totalLabels++
+	})
+	if totalLabels <= 3 {
+		return s.Name
+	}
+	return s.Name.TrimLabels(1)
+}
+
+// matchQuestion reports whether the question matches the given service.
+func matchQuestion(q *dns.Question, svc *Service) bool {
+	switch q.Type {
+	case dns.TypePTR:
+		svcType := svc.serviceType()
+		return dns.NamesEqual(q.Name, svcType)
+	case dns.TypeSRV, dns.TypeTXT:
+		return dns.NamesEqual(q.Name, svc.Name)
+	case dns.TypeA:
+		return dns.NamesEqual(q.Name, svc.Host)
+	case dns.TypeALL:
+		svcType := svc.serviceType()
+		return dns.NamesEqual(q.Name, svc.Name) || dns.NamesEqual(q.Name, svc.Host) || dns.NamesEqual(q.Name, svcType)
+	}
+	return false
+}
+
+// addServiceAnswers adds the appropriate answer records for a matched question.
+// It grows ans in-place, reusing existing Resource buffers when available.
+func addServiceAnswers(dst *[]dns.Resource, q *dns.Question, svc *Service) {
+	cacheFlush := dns.Class(uint16(dns.ClassINET) | classCacheFlush)
+	ttl := svc.ttl()
+	txtData := svc.TXTData
+	avail := cap(*dst) - len(*dst)
+	switch q.Type {
+	case dns.TypePTR:
+		if avail < 1 {
+			return
+		}
+		setPTR(growSlice(dst), svc)
+	case dns.TypeSRV:
+		if avail < 2 {
+			return
+		}
+		growSlice(dst).SetSRV(svc.Name, cacheFlush, ttl, 0, 0, svc.Port, svc.Host)
+		growSlice(dst).SetA(svc.Host, cacheFlush, ttl, svc.Addr)
+	case dns.TypeTXT:
+		if avail < 1 {
+			return
+		}
+		growSlice(dst).SetTXT(svc.Name, cacheFlush, ttl, txtData)
+	case dns.TypeA:
+		if avail < 1 {
+			return
+		}
+		growSlice(dst).SetA(svc.Host, cacheFlush, ttl, svc.Addr)
+	case dns.TypeALL:
+		if avail < 4 {
+			return
+		}
+		setPTR(growSlice(dst), svc)
+		growSlice(dst).SetSRV(svc.Name, cacheFlush, ttl, 0, 0, svc.Port, svc.Host)
+		growSlice(dst).SetTXT(svc.Name, cacheFlush, ttl, txtData)
+		growSlice(dst).SetA(svc.Host, cacheFlush, ttl, svc.Addr)
+	}
+}
+
+// growSlice grows the slice by one element and returns a pointer to the new last element.
+// Panics if at capacity — callers must check available space before calling.
+func growSlice[T any](s *[]T) *T {
+	*s = (*s)[:len(*s)+1]
+	return &(*s)[len(*s)-1]
+}
+
+func setPTR(ans *dns.Resource, svc *Service) {
+	svcType := svc.serviceType()
+	ans.SetPTR(svcType, dns.ClassINET, svc.ttl(), svc.Name)
+}

mdns/mdns_test.go

@@ -0,0 +1,409 @@
+package mdns
+
+import (
+	"encoding/binary"
+	"net"
+	"testing"
+
+	"github.com/soypat/lneto/dns"
+)
+
+func mustNewName(s string) dns.Name {
+	n, err := dns.NewName(s)
+	if err != nil {
+		panic(err)
+	}
+	return n
+}
+
+func testService() Service {
+	return Service{
+		Name: mustNewName("My Web._http._tcp.local"),
+		Host: mustNewName("mydevice.local"),
+		Addr: []byte{192, 168, 1, 50},
+		Port: 80,
+	}
+}
+
+func newQuerier(t *testing.T, questions []dns.Question, maxAnswers uint16) *Client {
+	t.Helper()
+	var c Client
+	err := c.Configure(ClientConfig{LocalPort: Port})
+	if err != nil {
+		t.Fatal(err)
+	}
+	err = c.StartResolve(ResolveConfig{
+		Questions:          questions,
+		MaxResponseAnswers: maxAnswers,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	return &c
+}
+
+func newResponder(t *testing.T, services []Service) *Client {
+	t.Helper()
+	var c Client
+	err := c.Configure(ClientConfig{
+		LocalPort: Port,
+		Services:  services,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	return &c
+}
+
+// queryRespond performs the full query->demux->encapsulate->demux cycle
+// between a querier and responder, returning the response length.
+func queryRespond(t *testing.T, querier, responder *Client, buf []byte) int {
+	t.Helper()
+	n, err := querier.Encapsulate(buf, -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("querier encapsulate:", err, n)
+	}
+	if err = responder.Demux(buf[:n], 0); err != nil {
+		t.Fatal("responder demux:", err)
+	}
+	n, err = responder.Encapsulate(buf, -1, 0)
+	if err != nil {
+		t.Fatal("responder encapsulate:", err)
+	}
+	if n > 0 {
+		if err = querier.Demux(buf[:n], 0); err != nil {
+			t.Fatal("querier demux:", err)
+		}
+	}
+	return n
+}
+
+func TestClientQueryPTR(t *testing.T) {
+	svc := testService()
+	responder := newResponder(t, []Service{svc})
+	querier := newQuerier(t, []dns.Question{{
+		Name:  mustNewName("_http._tcp.local"),
+		Type:  dns.TypePTR,
+		Class: dns.ClassINET,
+	}}, 8)
+
+	var buf [1024]byte
+
+	// Querier encapsulates query.
+	n, err := querier.Encapsulate(buf[:], -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("querier encapsulate:", err, n)
+	}
+
+	// Verify query wire format: txid=0, flags=0.
+	f, _ := dns.NewFrame(buf[:n])
+	if f.TxID() != 0 {
+		t.Errorf("mDNS query txid=%d, want 0", f.TxID())
+	}
+	if f.Flags() != 0 {
+		t.Errorf("mDNS query flags=%d, want 0", f.Flags())
+	}
+	if f.QDCount() != 1 {
+		t.Errorf("mDNS query QDCount=%d, want 1", f.QDCount())
+	}
+
+	// Responder demuxes and encapsulates response.
+	if err = responder.Demux(buf[:n], 0); err != nil {
+		t.Fatal("responder demux:", err)
+	}
+	n, err = responder.Encapsulate(buf[:], -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("responder encapsulate:", err, n)
+	}
+
+	// Verify response wire format: QR=1, AA=1.
+	f, _ = dns.NewFrame(buf[:n])
+	flags := f.Flags()
+	if !flags.IsResponse() {
+		t.Error("mDNS response missing QR bit")
+	}
+	if !flags.IsAuthorativeAnswer() {
+		t.Error("mDNS response missing AA bit")
+	}
+	if f.ANCount() == 0 {
+		t.Fatal("mDNS response has 0 answers")
+	}
+
+	// Querier demuxes response and reads answers.
+	if err = querier.Demux(buf[:n], 0); err != nil {
+		t.Fatal("querier demux:", err)
+	}
+	var answers [8]dns.Resource
+	nans, done, err := querier.AnswersCopyTo(answers[:])
+	if err != nil {
+		t.Fatal(err)
+	} else if !done {
+		t.Fatal("expected done")
+	} else if nans == 0 {
+		t.Fatal("querier got 0 answers")
+	}
+
+	// The PTR answer data should contain the instance name in wire format.
+	ptrData := answers[0].RawData()
+	if len(ptrData) == 0 {
+		t.Fatal("PTR answer has no data")
+	}
+	var ptrName dns.Name
+	_, err = ptrName.Decode(ptrData, 0)
+	if err != nil {
+		t.Fatal("decode PTR name:", err)
+	}
+	if ptrName.String() != svc.Name.String() {
+		t.Errorf("PTR target=%q, want %q", ptrName.String(), svc.Name.String())
+	}
+}
+
+func TestClientQuerySRV(t *testing.T) {
+	svc := testService()
+	responder := newResponder(t, []Service{svc})
+	querier := newQuerier(t, []dns.Question{{
+		Name:  mustNewName("My Web._http._tcp.local"),
+		Type:  dns.TypeSRV,
+		Class: dns.ClassINET,
+	}}, 8)
+
+	var buf [1024]byte
+	queryRespond(t, querier, responder, buf[:])
+
+	var answers [8]dns.Resource
+	nans, done, err := querier.AnswersCopyTo(answers[:])
+	if err != nil || !done {
+		t.Fatal("expected done without error:", err)
+	}
+	// SRV query should return SRV + A record.
+	if nans < 2 {
+		t.Fatalf("expected at least 2 answers (SRV+A), got %d", nans)
+	}
+
+	// First answer should be SRV. Parse priority(2)+weight(2)+port(2)+target.
+	srvData := answers[0].RawData()
+	if len(srvData) < 6 {
+		t.Fatalf("SRV data too short: %d bytes", len(srvData))
+	}
+	gotPort := binary.BigEndian.Uint16(srvData[4:6])
+	if gotPort != svc.Port {
+		t.Errorf("SRV port=%d, want %d", gotPort, svc.Port)
+	}
+	var srvTarget dns.Name
+	_, err = srvTarget.Decode(srvData, 6)
+	if err != nil {
+		t.Fatal("decode SRV target:", err)
+	}
+	if srvTarget.String() != svc.Host.String() {
+		t.Errorf("SRV target=%q, want %q", srvTarget.String(), svc.Host.String())
+	}
+
+	// Second answer should be A record with 4-byte IP.
+	aData := answers[1].RawData()
+	if len(aData) != 4 {
+		t.Fatalf("A record data length=%d, want 4", len(aData))
+	}
+	if [4]byte(aData) != [4]byte(svc.Addr) {
+		t.Errorf("A record addr=%v, want %v", aData, svc.Addr)
+	}
+}
+
+func TestClientQueryARecord(t *testing.T) {
+	svc := testService()
+	responder := newResponder(t, []Service{svc})
+	querier := newQuerier(t, []dns.Question{{
+		Name:  mustNewName("mydevice.local"),
+		Type:  dns.TypeA,
+		Class: dns.ClassINET,
+	}}, 4)
+
+	var buf [1024]byte
+	queryRespond(t, querier, responder, buf[:])
+
+	var answers [4]dns.Resource
+	nans, done, err := querier.AnswersCopyTo(answers[:])
+	if err != nil || !done {
+		t.Fatal("expected done without error:", err)
+	}
+	if nans != 1 {
+		t.Fatalf("expected 1 answer, got %d", nans)
+	}
+	aData := answers[0].RawData()
+	if [4]byte(aData) != [4]byte(svc.Addr) {
+		t.Errorf("A record addr=%v, want %v", aData, svc.Addr)
+	}
+}
+
+// TODO: TestClientAnnounce — test unsolicited announcement of all registered services.
+// func TestClientAnnounce(t *testing.T) { ... }
+
+// TODO: TestClientProbeFinish — test probing sequence for name uniqueness (RFC 6762 §8.1).
+// func TestClientProbeFinish(t *testing.T) { ... }
+
+func TestClientIgnoresQueriesWithoutServices(t *testing.T) {
+	// A querier-only client should ignore incoming queries.
+	querier := newQuerier(t, []dns.Question{{
+		Name:  mustNewName("_http._tcp.local"),
+		Type:  dns.TypePTR,
+		Class: dns.ClassINET,
+	}}, 4)
+
+	// Encapsulate query, then feed it back — should be ignored (not a response).
+	var buf [512]byte
+	n, _ := querier.Encapsulate(buf[:], -1, 0)
+	err := querier.Demux(buf[:n], 0)
+	if err != nil {
+		t.Fatal("demux query:", err)
+	}
+	// No response should be pending.
+	n, _ = querier.Encapsulate(buf[:], -1, 0)
+	if n != 0 {
+		t.Error("querier without services should not respond to queries")
+	}
+}
+
+func TestClientResponderIgnoresResponses(t *testing.T) {
+	svc := testService()
+	responder := newResponder(t, []Service{svc})
+
+	// Build a response packet (QR=1).
+	var msg dns.Message
+	var buf [512]byte
+	const responseFlags = dns.HeaderFlags(1 << 15)
+	data, err := msg.AppendTo(buf[:0], 0, responseFlags)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Feed a response to the responder — should be ignored.
+	err = responder.Demux(data, 0)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Nothing should be pending.
+	n, _ := responder.Encapsulate(buf[:], -1, 0)
+	if n != 0 {
+		t.Error("responder should not respond to responses")
+	}
+}
+
+func TestClientUnmatchedQuery(t *testing.T) {
+	svc := testService()
+	responder := newResponder(t, []Service{svc})
+
+	// Query for a name the responder doesn't know.
+	querier := newQuerier(t, []dns.Question{{
+		Name:  mustNewName("_ftp._tcp.local"),
+		Type:  dns.TypePTR,
+		Class: dns.ClassINET,
+	}}, 4)
+
+	var buf [1024]byte
+	n, _ := querier.Encapsulate(buf[:], -1, 0)
+	responder.Demux(buf[:n], 0)
+
+	// Responder should have nothing pending.
+	n, _ = responder.Encapsulate(buf[:], -1, 0)
+	if n != 0 {
+		t.Error("responder should not respond to unmatched query")
+	}
+}
+
+func TestClientMultipleResponders(t *testing.T) {
+	// Two responder clients advertising different instances of the same service type.
+	svc1 := Service{
+		Name: mustNewName("Device A._http._tcp.local"),
+		Host: mustNewName("device-a.local"),
+		Addr: []byte{192, 168, 1, 10},
+		Port: 80,
+	}
+	svc2 := Service{
+		Name: mustNewName("Device B._http._tcp.local"),
+		Host: mustNewName("device-b.local"),
+		Addr: []byte{192, 168, 1, 11},
+		Port: 8080,
+	}
+
+	responder1 := newResponder(t, []Service{svc1})
+	responder2 := newResponder(t, []Service{svc2})
+	querier := newQuerier(t, []dns.Question{{
+		Name:  mustNewName("_http._tcp.local"),
+		Type:  dns.TypePTR,
+		Class: dns.ClassINET,
+	}}, 8)
+
+	var buf [1024]byte
+
+	// Querier sends query.
+	n, _ := querier.Encapsulate(buf[:], -1, 0)
+	query := make([]byte, n)
+	copy(query, buf[:n])
+
+	// Both responders process the query.
+	responder1.Demux(query, 0)
+	responder2.Demux(query, 0)
+
+	// Querier receives response from responder 1.
+	n, _ = responder1.Encapsulate(buf[:], -1, 0)
+	querier.Demux(buf[:n], 0)
+
+	var answers [8]dns.Resource
+	nans, _, err := querier.AnswersCopyTo(answers[:])
+	if err != nil {
+		t.Fatal(err)
+	}
+	if nans < 1 {
+		t.Fatalf("expected at least 1 answer from first responder, got %d", nans)
+	}
+
+	// Start a new resolve to receive from responder 2.
+	querier.StartResolve(ResolveConfig{
+		Questions: []dns.Question{{
+			Name:  mustNewName("_http._tcp.local"),
+			Type:  dns.TypePTR,
+			Class: dns.ClassINET,
+		}},
+		MaxResponseAnswers: 8,
+	})
+	// Re-send query for second responder.
+	n, _ = querier.Encapsulate(buf[:], -1, 0)
+
+	n, _ = responder2.Encapsulate(buf[:], -1, 0)
+	querier.Demux(buf[:n], 0)
+
+	nans, _, err = querier.AnswersCopyTo(answers[:])
+	if err != nil {
+		t.Fatal(err)
+	}
+	if nans < 1 {
+		t.Fatalf("expected at least 1 answer from second responder, got %d", nans)
+	}
+}
+
+func TestClientAbort(t *testing.T) {
+	svc := testService()
+	responder := newResponder(t, []Service{svc})
+
+	responder.Abort()
+
+	// Demux on aborted client should return ErrClosed.
+	var buf [512]byte
+	var msg dns.Message
+	msg.AddQuestions([]dns.Question{{
+		Name:  mustNewName("_http._tcp.local"),
+		Type:  dns.TypePTR,
+		Class: dns.ClassINET,
+	}})
+	data, _ := msg.AppendTo(buf[:0], 0, 0)
+	err := responder.Demux(data, 0)
+	if err != net.ErrClosed {
+		t.Errorf("expected net.ErrClosed, got %v", err)
+	}
+
+	// Encapsulate on aborted client should return ErrClosed.
+	_, err = responder.Encapsulate(buf[:], -1, 0)
+	if err != net.ErrClosed {
+		t.Errorf("expected net.ErrClosed, got %v", err)
+	}
+}

x/xnet/stack-async.go

@@ -46,6 +46,8 @@ type StackAsync struct {
 	ntpUDP internet.StackUDPPort
 	ntp    ntp.Client
 
+	userUDPs []internet.StackUDPPort
+
 	sysprec int8 // NTP system precision.
 
 	prng uint32
@@ -60,12 +62,16 @@ type StackConfig struct {
 	StaticAddress         netip.Addr
 	DNSServer             netip.Addr
 	NTPServer             netip.Addr
+	RandSeed              int64
 	Hostname              string
 	MaxTCPConns           int
-	RandSeed              int64
-	HardwareAddress       [6]byte
-	MTU                   uint16
+	MaxUDPConns           int
 	EthernetTxCRC32Update func(crc uint32, b []byte) uint32
+
+	HardwareAddress [6]byte
+	MTU             uint16
+	// Accept multicast ethernet and IP packets. Needed for MDNS.
+	AcceptMulticast bool
 }
 
 func (s *StackAsync) Hostname() string {
@@ -120,18 +126,24 @@ func (s *StackAsync) Reset(cfg StackConfig) error {
 	if err != nil {
 		return err
 	}
+	s.link.SetAcceptMulticast(cfg.AcceptMulticast)
 	const ipNodes = 2 // UDP, TCP ports.
 	err = s.ip.Reset(addr, ipNodes)
 	if err != nil {
 		return err
 	}
+	s.ip.SetAcceptMulticast(cfg.AcceptMulticast)
 	//
 	err = s.resetARP()
 	if err != nil {
 		return err
 	}
-	const udpMaintenanceConns = 3 // DHCP, DNS, NTP.
-	err = s.udps.ResetUDP(udpMaintenanceConns)
+	udpConns := 3 + cfg.MaxUDPConns // DHCP, DNS, NTP + user-registered.
+	err = s.udps.ResetUDP(udpConns)
+	if err != nil {
+		return err
+	}
+	internal.SliceReuse(&s.userUDPs, cfg.MaxUDPConns)
 	if err != nil {
 		return err
 	}
@@ -322,6 +334,21 @@ func (s *StackAsync) RegisterListener(listener *tcp.Listener) (err error) {
 	return s.tcps.Register(listener, nil)
 }
 
+// RegisterUDP registers a StackNode on a UDP port with the given remote address and port.
+// The StackUDPPort wrapping is handled internally. The number of user-registered UDP ports
+// is limited by [StackConfig.MaxUDPConns].
+func (s *StackAsync) RegisterUDP(node internet.StackNode, remoteAddr []byte, remotePort uint16) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	idx := len(s.userUDPs)
+	if idx >= cap(s.userUDPs) {
+		return lneto.ErrBufferFull
+	}
+	s.userUDPs = s.userUDPs[:idx+1]
+	s.userUDPs[idx].SetStackNode(node, remoteAddr, remotePort)
+	return s.udps.Register(&s.userUDPs[idx])
+}
+
 var errNoDNSServer = errors.New("no DNS server- did DHCP complete? You can set a predetermined DNS server in Stack configuration")
 
 func (s *StackAsync) StartLookupIP(host string) error {

x/xnet/xnet_mdns_test.go

@@ -0,0 +1,332 @@
+package xnet
+
+import (
+	"encoding/binary"
+	"net/netip"
+	"testing"
+
+	"github.com/soypat/lneto/dns"
+	"github.com/soypat/lneto/ethernet"
+	"github.com/soypat/lneto/mdns"
+)
+
+func TestMDNS_QueryResponse(t *testing.T) {
+	const MTU = 1500
+	svcName, err := dns.NewName("My Web._http._tcp.local")
+	if err != nil {
+		t.Fatal(err)
+	}
+	hostName, err := dns.NewName("mydevice.local")
+	if err != nil {
+		t.Fatal(err)
+	}
+	svcType, err := dns.NewName("_http._tcp.local")
+	if err != nil {
+		t.Fatal(err)
+	}
+	svc := mdns.Service{
+		Name: svcName,
+		Host: hostName,
+		Addr: []byte{192, 168, 1, 50},
+		Port: 80,
+	}
+
+	responderAddr := netip.AddrFrom4([4]byte{192, 168, 1, 50})
+	responderMAC := [6]byte{0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0x01}
+	querierAddr := netip.AddrFrom4([4]byte{192, 168, 1, 100})
+	querierMAC := [6]byte{0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0x02}
+	mcastAddr := []byte{224, 0, 0, 251}
+
+	// Setup responder stack with mDNS service.
+	responderStack := new(StackAsync)
+	err = responderStack.Reset(StackConfig{
+		Hostname:        "responder",
+		RandSeed:        1234,
+		StaticAddress:   responderAddr,
+		HardwareAddress: responderMAC,
+		MTU:             MTU,
+		MaxUDPConns:     1,
+		AcceptMulticast: true,
+	})
+	if err != nil {
+		t.Fatal("responder reset:", err)
+	}
+	responderStack.SetGateway6(querierMAC)
+
+	var responderClient mdns.Client
+	err = responderClient.Configure(mdns.ClientConfig{
+		LocalPort:     mdns.Port,
+		Services:      []mdns.Service{svc},
+		MulticastAddr: mcastAddr,
+	})
+	if err != nil {
+		t.Fatal("responder configure:", err)
+	}
+	err = responderStack.RegisterUDP(&responderClient, mcastAddr, mdns.Port)
+	if err != nil {
+		t.Fatal("responder register:", err)
+	}
+
+	// Setup querier stack.
+	querierStack := new(StackAsync)
+	err = querierStack.Reset(StackConfig{
+		Hostname:        "querier",
+		RandSeed:        5678,
+		StaticAddress:   querierAddr,
+		HardwareAddress: querierMAC,
+		MTU:             MTU,
+		MaxUDPConns:     1,
+		AcceptMulticast: true,
+	})
+	if err != nil {
+		t.Fatal("querier reset:", err)
+	}
+	querierStack.SetGateway6(responderMAC)
+
+	var querierClient mdns.Client
+	err = querierClient.Configure(mdns.ClientConfig{
+		LocalPort:     mdns.Port,
+		MulticastAddr: mcastAddr,
+	})
+	if err != nil {
+		t.Fatal("querier configure:", err)
+	}
+	err = querierClient.StartResolve(mdns.ResolveConfig{
+		Questions: []dns.Question{{
+			Name:  svcType,
+			Type:  dns.TypePTR,
+			Class: dns.ClassINET,
+		}},
+		MaxResponseAnswers: 4,
+	})
+	if err != nil {
+		t.Fatal("start resolve:", err)
+	}
+	err = querierStack.RegisterUDP(&querierClient, mcastAddr, mdns.Port)
+	if err != nil {
+		t.Fatal("querier register:", err)
+	}
+
+	const carrierDataSize = MTU + ethernet.MaxOverheadSize
+	var buf [carrierDataSize]byte
+
+	// Querier encapsulates query through full stack (Ethernet+IP+UDP+mDNS).
+	n, err := querierStack.Encapsulate(buf[:], -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("querier encapsulate:", err, n)
+	}
+
+	// Verify mDNS query wire format at DNS layer.
+	const ethHdrLen = 14
+	ipIHL := int(buf[ethHdrLen]&0x0f) * 4
+	dnsStart := ethHdrLen + ipIHL + 8
+	dnsFrame, err := dns.NewFrame(buf[dnsStart:n])
+	if err != nil {
+		t.Fatal("parse query dns frame:", err)
+	}
+	if dnsFrame.TxID() != 0 {
+		t.Errorf("mDNS query txid=%d, want 0", dnsFrame.TxID())
+	}
+	if dnsFrame.Flags() != 0 {
+		t.Errorf("mDNS query flags=%d, want 0", dnsFrame.Flags())
+	}
+
+	// Responder demuxes the query (multicast MAC+IP accepted via AcceptMulticast).
+	err = responderStack.Demux(buf[:n], 0)
+	if err != nil {
+		t.Fatal("responder demux:", err)
+	}
+
+	// Responder encapsulates response.
+	n, err = responderStack.Encapsulate(buf[:], -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("responder encapsulate:", err, n)
+	}
+
+	// Verify response DNS flags.
+	ipIHL = int(buf[ethHdrLen]&0x0f) * 4
+	dnsStart = ethHdrLen + ipIHL + 8
+	dnsFrame, err = dns.NewFrame(buf[dnsStart:n])
+	if err != nil {
+		t.Fatal("parse response dns frame:", err)
+	}
+	flags := dnsFrame.Flags()
+	if !flags.IsResponse() {
+		t.Error("mDNS response missing QR bit")
+	}
+	if !flags.IsAuthorativeAnswer() {
+		t.Error("mDNS response missing AA bit")
+	}
+	if dnsFrame.ANCount() == 0 {
+		t.Fatal("mDNS response has 0 answers")
+	}
+
+	// Querier demuxes response.
+	err = querierStack.Demux(buf[:n], 0)
+	if err != nil {
+		t.Fatal("querier demux:", err)
+	}
+
+	// Read answers.
+	var answers [4]dns.Resource
+	nans, done, err := querierClient.AnswersCopyTo(answers[:])
+	if err != nil {
+		t.Fatal("answers:", err)
+	}
+	if !done {
+		t.Fatal("expected done")
+	}
+	if nans == 0 {
+		t.Fatal("got 0 answers")
+	}
+
+	// Verify PTR answer points to our service instance name.
+	ptrData := answers[0].RawData()
+	var ptrTarget dns.Name
+	_, err = ptrTarget.Decode(ptrData, 0)
+	if err != nil {
+		t.Fatal("decode PTR target:", err)
+	}
+	if !dns.NamesEqual(ptrTarget, svcName) {
+		t.Errorf("PTR target=%q, want %q", ptrTarget.String(), svcName.String())
+	}
+}
+
+func TestMDNS_SRVThroughStack(t *testing.T) {
+	const MTU = 1500
+	svcName, err := dns.NewName("My Web._http._tcp.local")
+	if err != nil {
+		t.Fatal(err)
+	}
+	hostName, err := dns.NewName("mydevice.local")
+	if err != nil {
+		t.Fatal(err)
+	}
+	svc := mdns.Service{
+		Name: svcName,
+		Host: hostName,
+		Addr: []byte{192, 168, 1, 50},
+		Port: 80,
+	}
+	mcastAddr := []byte{224, 0, 0, 251}
+
+	responderMAC := [6]byte{0x01, 0x02, 0x03, 0x04, 0x05, 0x01}
+	querierMAC := [6]byte{0x01, 0x02, 0x03, 0x04, 0x05, 0x02}
+
+	// Create responder.
+	responderStack, _ := newMDNSStack(t, "responder", 1111,
+		netip.AddrFrom4([4]byte{192, 168, 1, 50}), responderMAC, querierMAC,
+		mdns.ClientConfig{LocalPort: mdns.Port, Services: []mdns.Service{svc}, MulticastAddr: mcastAddr},
+	)
+
+	// Create querier.
+	querierStack, querierClient := newMDNSStack(t, "querier", 2222,
+		netip.AddrFrom4([4]byte{192, 168, 1, 100}), querierMAC, responderMAC,
+		mdns.ClientConfig{LocalPort: mdns.Port, MulticastAddr: mcastAddr},
+	)
+	err = querierClient.StartResolve(mdns.ResolveConfig{
+		Questions: []dns.Question{{
+			Name:  svcName,
+			Type:  dns.TypeSRV,
+			Class: dns.ClassINET,
+		}},
+		MaxResponseAnswers: 4,
+	})
+	if err != nil {
+		t.Fatal("start resolve:", err)
+	}
+
+	// Full round-trip through both stacks.
+	var buf [MTU + ethernet.MaxOverheadSize]byte
+	mdnsQueryRespond(t, querierStack, responderStack, buf[:])
+
+	var answers [4]dns.Resource
+	nans, done, err := querierClient.AnswersCopyTo(answers[:])
+	if err != nil || !done {
+		t.Fatal("expected done:", err)
+	}
+	if nans < 2 {
+		t.Fatalf("expected at least 2 answers (SRV+A), got %d", nans)
+	}
+
+	// Verify SRV port.
+	srvData := answers[0].RawData()
+	if len(srvData) < 6 {
+		t.Fatalf("SRV data too short: %d", len(srvData))
+	}
+	gotPort := binary.BigEndian.Uint16(srvData[4:6])
+	if gotPort != svc.Port {
+		t.Errorf("SRV port=%d, want %d", gotPort, svc.Port)
+	}
+
+	// Verify A record.
+	aData := answers[1].RawData()
+	if [4]byte(aData) != [4]byte(svc.Addr) {
+		t.Errorf("A record addr=%v, want %v", aData, svc.Addr)
+	}
+}
+
+// newMDNSStack creates a StackAsync with an mDNS client registered on its UDP ports.
+func newMDNSStack(t *testing.T, hostname string, seed int64,
+	addr netip.Addr, mac, gatewayMAC [6]byte,
+	mdnsCfg mdns.ClientConfig,
+) (*StackAsync, *mdns.Client) {
+	t.Helper()
+	const MTU = 1500
+	stack := new(StackAsync)
+	err := stack.Reset(StackConfig{
+		Hostname:        hostname,
+		RandSeed:        seed,
+		StaticAddress:   addr,
+		HardwareAddress: mac,
+		MTU:             MTU,
+		MaxUDPConns:     1,
+		AcceptMulticast: true,
+	})
+	if err != nil {
+		t.Fatal(hostname, "reset:", err)
+	}
+	stack.SetGateway6(gatewayMAC)
+
+	var client mdns.Client
+	err = client.Configure(mdnsCfg)
+	if err != nil {
+		t.Fatal(hostname, "mdns configure:", err)
+	}
+
+	err = stack.RegisterUDP(&client, mdnsCfg.MulticastAddr, mdns.Port)
+	if err != nil {
+		t.Fatal(hostname, "register udp:", err)
+	}
+	return stack, &client
+}
+
+// mdnsQueryRespond performs a full Ethernet+IP+UDP+mDNS query→response cycle
+// between two stacks with AcceptMulticast enabled.
+func mdnsQueryRespond(t *testing.T, querier, responder *StackAsync, buf []byte) {
+	t.Helper()
+
+	// Querier encapsulates query.
+	n, err := querier.Encapsulate(buf, -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("querier encapsulate:", err, n)
+	}
+
+	// Responder demuxes multicast query directly.
+	err = responder.Demux(buf[:n], 0)
+	if err != nil {
+		t.Fatal("responder demux:", err)
+	}
+
+	// Responder encapsulates response.
+	n, err = responder.Encapsulate(buf, -1, 0)
+	if err != nil || n == 0 {
+		t.Fatal("responder encapsulate:", err, n)
+	}
+
+	// Querier demuxes multicast response.
+	err = querier.Demux(buf[:n], 0)
+	if err != nil {
+		t.Fatal("querier demux:", err)
+	}
+}