fix(udp): preserve wildcard reply sources and isolate destination sessions (#565)

* fix(udp): preserve local destinations and unicast reply sources on Linux

* fix(udp): select unicast sources for IPv6 multicast replies

* fix(udp): retain interface zones for link-local multicast sessions

* fix(udp): isolate interface-local multicast sessions

* fix(udp): include ingress interfaces in session identity

* fix(udp): restrict destination zones to IPv6

* ci(udp): follow rc5 patched Go toolchain requirement

---------

Co-authored-by: sagit <36596628+Sagit-chu@users.noreply.github.com>

.github/workflows/udp-destination.yml

@@ -0,0 +1,87 @@
+name: UDP Destination Regression
+
+on:
+  push:
+    branches: ['main', 'fix/udp-wildcard-source', 'pr/udp-wildcard-source']
+  pull_request:
+    paths:
+      - 'go-gost/**'
+      - '.github/workflows/udp-destination.yml'
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: udp-destination-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  linux:
+    runs-on: ubuntu-24.04
+    timeout-minutes: 20
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: 'go-gost/go.mod'
+          cache-dependency-path: |
+            go-gost/go.sum
+            go-gost/x/go.sum
+
+      - name: Test agent module
+        working-directory: go-gost
+        run: go test ./...
+
+      - name: Test protocol module
+        working-directory: go-gost/x
+        run: go test ./...
+
+      - name: Check UDP races
+        working-directory: go-gost/x
+        run: go test -race ./internal/net/udp ./listener/udp
+
+      - name: Vet UDP implementation
+        working-directory: go-gost/x
+        run: go vet ./internal/net/udp ./listener/udp
+
+      - name: Build destination regression executable
+        working-directory: go-gost/x
+        run: |
+          go test -c -race -o "$RUNNER_TEMP/udp-destination.test" ./listener/udp
+          go test -c -race -o "$RUNNER_TEMP/udp-packet.test" ./internal/net/udp
+
+      - name: Test multiple destinations in isolated network namespace
+        run: |
+          sudo unshare --net -- sh -eu -c '
+            sysctl -qw net.ipv6.conf.all.disable_ipv6=0
+            ip link set lo up
+            ip addr add 198.18.0.1/32 dev lo
+            ip addr add 169.254.1.2/32 dev lo
+            ip -6 addr add fd00:1::1/128 dev lo nodad
+            ip -6 addr add fd00:2::1/128 dev lo nodad
+            ip -6 addr add fd00:3::1/128 dev lo nodad
+            for suffix in "" 2; do
+              ip link add "mc-tx$suffix" type veth peer name "mc-rx$suffix"
+              ip link set "mc-tx$suffix" addrgenmode none up
+              ip link set "mc-rx$suffix" addrgenmode none up
+              ip -6 addr add fe80::1/64 dev "mc-tx$suffix" nodad
+              ip -6 addr add fe80::2/64 dev "mc-rx$suffix" nodad
+            done
+            ip addr add 198.18.1.1/24 dev mc-tx
+            ip addr add 198.18.2.1/24 dev mc-tx2
+            export FLUX_TEST_IPV6_ADDRS=fd00:1::1,fd00:2::1,fd00:3::1
+            export FLUX_TEST_IPV4_LINK_LOCAL=169.254.1.2
+            export FLUX_TEST_MULTICAST_RX=mc-rx FLUX_TEST_MULTICAST_TX=mc-tx
+            export FLUX_TEST_MULTICAST_RX2=mc-rx2 FLUX_TEST_MULTICAST_TX2=mc-tx2
+            "$1" -test.v -test.timeout=60s -test.count=5
+            exec "$2" -test.v -test.timeout=60s -test.count=5
+          ' sh "$RUNNER_TEMP/udp-destination.test" "$RUNNER_TEMP/udp-packet.test"
+
+      - name: Build Linux agent
+        working-directory: go-gost
+        env:
+          CGO_ENABLED: '0'
+          GOOS: linux
+          GOARCH: amd64
+        run: go build -trimpath -o "$RUNNER_TEMP/flux-agent" .

docs/udp-wildcard-source.md

@@ -0,0 +1,133 @@
+# Preserve UDP reply source addresses on Linux
+
+The Linux UDP listener preserves the destination address of each incoming
+datagram so that replies use the correct local source. This change was developed
+against FLVX `3.0.1-rc4`, commit
+`d9303032cad09b33cd0b7c91704fd671fbfdd078`.
+It also includes the `3.0.1-rc5` runtime update at
+`c794586cc7773e5f967783895d3071538442bbf7`.
+
+## Problem
+
+A wildcard UDP socket bound to `0.0.0.0:port` or `[::]:port` can receive datagrams
+addressed to several local IP addresses. Previously, the listener retained only
+the client address. A reply sent through `WriteTo` could therefore use a source
+IP chosen by the kernel instead of the IP the client contacted. DNS clients, for
+example, can reject such replies as coming from an unexpected source.
+
+The UDP session pool was also keyed only by the client address and port. When
+one client socket contacted multiple local addresses on the same listener port,
+these requests could share a session despite having different destinations.
+
+## Changes
+
+- Linux wildcard UDP sockets enable IPv4/IPv6 packet information. `recvmsg`
+  preserves the datagram's local destination and interface; `sendmsg` uses the
+  corresponding source address and interface when replying.
+- IPv4 keeps the original destination (`ipi_addr`) separate from the local
+  reply address (`ipi_spec_dst`), so broadcast requests receive unicast replies
+  rather than attempting to send with a broadcast source address.
+- IPv6 multicast destinations remain part of session identity, but replies
+  leave the source unspecified and retain the received interface index so the
+  kernel can select a valid local unicast source.
+- The destination metadata travels through existing metrics, statistics,
+  admission and traffic-limiter wrappers. The exposed remote address remains
+  the real client address.
+- Sessions are keyed by the local destination endpoint, remote client endpoint
+  and received interface index. Identical IPv4 broadcast or multicast endpoints
+  on different interfaces therefore retain separate reply control information.
+  An accepted session reports its actual local destination.
+  Link-local IPv6 unicast and link-local or interface-local multicast
+  destinations retain their interface zone, keeping identical multicast groups
+  on different interfaces separate even when the client uses the same
+  non-link-local IP and source port.
+  IPv4 destinations, including link-local unicast and multicast, never receive
+  an IPv6 zone; their interface isolation uses the separate index in the key.
+- Explicitly bound UDP sockets retain their existing behavior. Non-Linux
+  platforms retain their existing implementation. No configuration keys or
+  panel API changes are required.
+
+The fix affects the shared UDP listener, including services dynamically created
+after startup. It does not configure listening addresses, NAT, routing or
+firewalls. Multi-homed hosts still need correct source policy routing for their
+uplinks; this patch does not fix SSH/TCP routing or packet loss before ingress.
+
+## Build
+
+Use Go 1.26.10 or a compatible newer toolchain. From the repository root:
+
+```sh
+cd go-gost
+CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -trimpath \
+  -ldflags '-s -w' -o flux_agent .
+```
+
+Use the agent with the normal panel integration and forwarding configuration.
+Selecting a wildcard listener enables destination preservation automatically
+on Linux. The patch does not select a listener address or change the upstream
+connection's outbound interface.
+
+## Regression tests
+
+The Linux destination tests cover IPv4 wildcard and dual-stack wildcard sockets,
+explicit IPv4/IPv6 bindings, one client source port contacting multiple local
+destinations, repeated requests, session expiry, and active metrics, statistics,
+admission and traffic-limit wrappers. A separate IPv6 case binds the client to
+`::1` and checks the exact remote IP and port for every destination. Broadcast
+tests exercise both IPv4 and dual-stack listeners, retaining the broadcast
+destination while checking that replies use a local unicast address.
+An IPv6 multicast regression sends across an isolated veth pair and checks the
+multicast session destination and the unicast reply, including interface zones.
+A two-interface case uses one fixed ULA client endpoint to contact the same
+link-local or interface-local multicast group on both interfaces and checks
+distinct sessions and their individual reply-interface metadata.
+IPv4 and dual-stack cases likewise use a single fixed client endpoint to send
+to one multicast group or the limited broadcast address on two interfaces,
+checking separate sessions and their reply-interface indices.
+Link-local IPv4 unicast and multicast regressions also check that accepted
+local endpoints have no IPv6 zone, on both IPv4 and dual-stack listeners.
+
+Run the standard suites on Linux:
+
+```sh
+(cd go-gost && go test ./...)
+(cd go-gost/x && go test ./...)
+```
+
+To exercise multiple IPv6 addresses without modifying the host's live network,
+compile the tests first and run them inside a disposable network namespace.
+The following commands require Linux, root access for `unshare`, and `iproute2`:
+
+```sh
+cd go-gost/x
+go test -c -race -o /tmp/flvx-udp-destination.test ./listener/udp
+go test -c -race -o /tmp/flvx-udp-packet.test ./internal/net/udp
+sudo unshare --net -- sh -eu -c '
+  sysctl -qw net.ipv6.conf.all.disable_ipv6=0
+  ip link set lo up
+  ip addr add 198.18.0.1/32 dev lo
+  ip addr add 169.254.1.2/32 dev lo
+  ip -6 addr add fd00:1::1/128 dev lo nodad
+  ip -6 addr add fd00:2::1/128 dev lo nodad
+  ip -6 addr add fd00:3::1/128 dev lo nodad
+  for suffix in "" 2; do
+    ip link add "mc-tx$suffix" type veth peer name "mc-rx$suffix"
+    ip link set "mc-tx$suffix" addrgenmode none up
+    ip link set "mc-rx$suffix" addrgenmode none up
+    ip -6 addr add fe80::1/64 dev "mc-tx$suffix" nodad
+    ip -6 addr add fe80::2/64 dev "mc-rx$suffix" nodad
+  done
+  ip addr add 198.18.1.1/24 dev mc-tx
+  ip addr add 198.18.2.1/24 dev mc-tx2
+  export FLUX_TEST_IPV6_ADDRS=fd00:1::1,fd00:2::1,fd00:3::1
+  export FLUX_TEST_IPV4_LINK_LOCAL=169.254.1.2
+  export FLUX_TEST_MULTICAST_RX=mc-rx FLUX_TEST_MULTICAST_TX=mc-tx
+  export FLUX_TEST_MULTICAST_RX2=mc-rx2 FLUX_TEST_MULTICAST_TX2=mc-tx2
+  /tmp/flvx-udp-destination.test -test.v -test.timeout=60s -test.count=5
+  exec /tmp/flvx-udp-packet.test -test.v -test.timeout=60s -test.count=5
+'
+```
+
+The addresses above are isolated test addresses, not deployment settings.
+The `UDP Destination Regression` GitHub Actions workflow runs these namespace
+tests, the race detector, both agent module suites, and a Linux agent build.

go-gost/x/internal/net/udp/listener.go

@@ -3,6 +3,7 @@ package udp
 import (
 	"errors"
 	"net"
+	"strconv"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -73,6 +74,7 @@ func (ln *listener) listenLoop() {
 
 		n, raddr, err := ln.conn.ReadFrom(b)
 		if err != nil {
+			bufpool.Put(b)
 			ln.errChan <- err
 			close(ln.errChan)
 			return
@@ -110,15 +112,27 @@ func (ln *listener) Close() error {
 }
 
 func (ln *listener) getConn(raddr net.Addr) *conn {
-	c, ok := ln.connPool.Get(raddr.String())
+	laddr := ln.Addr()
+	peer := raddr
+	var packet *packetAddr
+	if received, ok := raddr.(*packetAddr); ok {
+		packet = received
+		laddr, peer = received.local, received.Addr
+	}
+	key := laddr.String() + "|" + peer.String()
+	if packet != nil {
+		key += "|" + strconv.Itoa(packet.ifIndex)
+	}
+	c, ok := ln.connPool.Get(key)
 	if ok && !c.isClosed() {
 		return c
 	}
 
-	c = newConn(ln.conn, ln.Addr(), raddr, ln.config.ReadQueueSize, ln.config.Keepalive)
+	c = newConn(ln.conn, laddr, peer, ln.config.ReadQueueSize, ln.config.Keepalive)
+	c.packet = packet
 	select {
 	case ln.cqueue <- c:
-		ln.connPool.Set(raddr.String(), c)
+		ln.connPool.Set(key, c)
 		return c
 	default:
 		c.Close()
@@ -132,6 +146,7 @@ type conn struct {
 	net.PacketConn
 	localAddr  net.Addr
 	remoteAddr net.Addr
+	packet     *packetAddr
 	rc         chan []byte // data receive queue
 	idle       int32       // indicate the connection is idle
 	closed     chan struct{}
@@ -176,6 +191,9 @@ func (c *conn) WriteTo(b []byte, addr net.Addr) (n int, err error) {
 	if !c.keepalive {
 		defer c.Close()
 	}
+	if c.packet != nil {
+		addr = &packetAddr{Addr: addr, local: c.packet.local, control: c.packet.control, ifIndex: c.packet.ifIndex}
+	}
 	return c.PacketConn.WriteTo(b, addr)
 }
 

go-gost/x/internal/net/udp/packet.go

@@ -0,0 +1,12 @@
+package udp
+
+import "net"
+
+// packetAddr carries the received destination through the PacketConn wrappers.
+// Its public address remains the real peer for admission, metrics and limits.
+type packetAddr struct {
+	net.Addr
+	local   *net.UDPAddr
+	control []byte
+	ifIndex int
+}

go-gost/x/internal/net/udp/packet_linux.go

@@ -0,0 +1,116 @@
+package udp
+
+import (
+	"encoding/binary"
+	"errors"
+	"net"
+
+	"golang.org/x/net/ipv6"
+	"golang.org/x/sys/unix"
+)
+
+// PreserveDestination pins wildcard UDP replies to the address they arrived on.
+func PreserveDestination(c *net.UDPConn) (net.PacketConn, error) {
+	local := c.LocalAddr().(*net.UDPAddr)
+	if !local.IP.IsUnspecified() {
+		return c, nil
+	}
+	raw, err := c.SyscallConn()
+	if err != nil {
+		return nil, err
+	}
+	var sockErr error
+	if err = raw.Control(func(fd uintptr) {
+		if local.IP.To4() != nil {
+			sockErr = unix.SetsockoptInt(int(fd), unix.IPPROTO_IP, unix.IP_PKTINFO, 1)
+			return
+		}
+		if sockErr = unix.SetsockoptInt(int(fd), unix.IPPROTO_IPV6, unix.IPV6_RECVPKTINFO, 1); sockErr != nil {
+			return
+		}
+		var onlyV6 int
+		onlyV6, sockErr = unix.GetsockoptInt(int(fd), unix.IPPROTO_IPV6, unix.IPV6_V6ONLY)
+		if sockErr == nil && onlyV6 == 0 {
+			sockErr = unix.SetsockoptInt(int(fd), unix.IPPROTO_IP, unix.IP_PKTINFO, 1)
+		}
+	}); err != nil {
+		return nil, err
+	}
+	if sockErr != nil {
+		return nil, sockErr
+	}
+	return &destinationConn{UDPConn: c}, nil
+}
+
+type destinationConn struct {
+	*net.UDPConn
+}
+
+func (c *destinationConn) ReadFrom(b []byte) (int, net.Addr, error) {
+	var oob [128]byte
+	n, oobn, flags, remote, err := c.ReadMsgUDP(b, oob[:])
+	if err != nil {
+		return 0, remote, err
+	}
+	if flags&unix.MSG_CTRUNC != 0 {
+		return 0, nil, errors.New("UDP destination control message truncated")
+	}
+	var dst net.IP
+	var control []byte
+	var ifIndex int
+	if remote.IP.To4() != nil {
+		messages, err := unix.ParseSocketControlMessage(oob[:oobn])
+		if err != nil {
+			return 0, nil, err
+		}
+		for _, message := range messages {
+			if message.Header.Level != unix.IPPROTO_IP || message.Header.Type != unix.IP_PKTINFO {
+				continue
+			}
+			var info unix.Inet4Pktinfo
+			if _, err := binary.Decode(message.Data, binary.NativeEndian, &info); err != nil {
+				return 0, nil, err
+			}
+			dst, ifIndex = net.IP(info.Addr[:]), int(info.Ifindex)
+			// Spec_dst remains a local unicast address for broadcast destinations.
+			control = unix.PktInfo4(&unix.Inet4Pktinfo{Ifindex: info.Ifindex, Spec_dst: info.Spec_dst})
+			break
+		}
+	} else {
+		var cm ipv6.ControlMessage
+		if err = cm.Parse(oob[:oobn]); err != nil {
+			return 0, nil, err
+		}
+		dst, ifIndex = cm.Dst, cm.IfIndex
+		source := dst
+		if dst.IsMulticast() {
+			source = nil
+		}
+		control = (&ipv6.ControlMessage{Src: source, IfIndex: ifIndex}).Marshal()
+	}
+	if len(dst) == 0 || dst.IsUnspecified() {
+		return 0, nil, errors.New("UDP destination address missing")
+	}
+	local := &net.UDPAddr{IP: append(net.IP(nil), dst...), Port: c.LocalAddr().(*net.UDPAddr).Port}
+	if dst.To4() == nil && (dst.IsLinkLocalUnicast() || dst.IsLinkLocalMulticast() || dst.IsInterfaceLocalMulticast()) {
+		if iface, lookupErr := net.InterfaceByIndex(ifIndex); lookupErr == nil {
+			local.Zone = iface.Name
+		}
+	}
+	return n, &packetAddr{Addr: remote, local: local, control: control, ifIndex: ifIndex}, nil
+}
+
+func (c *destinationConn) WriteTo(b []byte, addr net.Addr) (int, error) {
+	if packet, ok := addr.(*packetAddr); ok {
+		remote, ok := packet.Addr.(*net.UDPAddr)
+		if !ok {
+			return 0, errors.New("invalid UDP peer address")
+		}
+		n, _, err := c.WriteMsgUDP(b, packet.control, remote)
+		if n < 0 {
+			n = 0
+		}
+		return n, err
+	}
+	return c.UDPConn.WriteTo(b, addr)
+}

go-gost/x/internal/net/udp/packet_linux_test.go

@@ -0,0 +1,304 @@
+package udp
+
+import (
+	"encoding/binary"
+	"fmt"
+	"io"
+	"net"
+	"os"
+	"testing"
+	"time"
+
+	"golang.org/x/net/ipv4"
+	"golang.org/x/net/ipv6"
+	"golang.org/x/sys/unix"
+
+	xlogger "github.com/go-gost/x/logger"
+)
+
+func TestMulticastDestinationRepliesFromUnicast(t *testing.T) {
+	rxName, txName := os.Getenv("FLUX_TEST_MULTICAST_RX"), os.Getenv("FLUX_TEST_MULTICAST_TX")
+	if rxName == "" || txName == "" {
+		t.Skip("requires isolated multicast veth interfaces")
+	}
+	rx, err := net.InterfaceByName(rxName)
+	if err != nil {
+		t.Fatal(err)
+	}
+	tx, err := net.InterfaceByName(txName)
+	if err != nil {
+		t.Fatal(err)
+	}
+	group := net.ParseIP("ff02::114")
+	socket, err := net.ListenMulticastUDP("udp6", rx, &net.UDPAddr{IP: group})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer socket.Close()
+	pc, err := PreserveDestination(socket)
+	if err != nil {
+		t.Fatal(err)
+	}
+	ln := NewListener(pc, &ListenConfig{
+		Backlog: 1, ReadQueueSize: 1, ReadBufferSize: 128, TTL: time.Second, Keepalive: true,
+		Logger: xlogger.NewLogger(xlogger.OutputOption(io.Discard)),
+	})
+	defer ln.Close()
+	client, err := net.ListenUDP("udp6", &net.UDPAddr{IP: net.ParseIP("fe80::1"), Zone: txName})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer client.Close()
+	if err := ipv6.NewPacketConn(client).SetMulticastInterface(tx); err != nil {
+		t.Fatal(err)
+	}
+	client.SetDeadline(time.Now().Add(3 * time.Second))
+	port := socket.LocalAddr().(*net.UDPAddr).Port
+	serverResult := make(chan error, 1)
+	go func() {
+		c, err := ln.Accept()
+		if err != nil {
+			serverResult <- err
+			return
+		}
+		defer c.Close()
+		local, peer := c.LocalAddr().(*net.UDPAddr), c.RemoteAddr().(*net.UDPAddr)
+		if !local.IP.Equal(group) || local.Port != port || local.Zone != rxName || !peer.IP.Equal(net.ParseIP("fe80::1")) ||
+			peer.Port != client.LocalAddr().(*net.UDPAddr).Port || peer.Zone != rxName {
+			serverResult <- fmt.Errorf("unexpected multicast session: local=%s peer=%s", local, peer)
+			return
+		}
+		var buf [128]byte
+		n, err := c.Read(buf[:])
+		if err == nil {
+			_, err = c.Write(buf[:n])
+		}
+		serverResult <- err
+	}()
+	if _, err := client.WriteToUDP([]byte("multicast"), &net.UDPAddr{IP: group, Port: port, Zone: txName}); err != nil {
+		t.Fatal(err)
+	}
+	select {
+	case err := <-serverResult:
+		if err != nil {
+			t.Fatal(err)
+		}
+	case <-time.After(3 * time.Second):
+		t.Fatal("multicast session timed out")
+	}
+	var buf [128]byte
+	n, from, err := client.ReadFromUDP(buf[:])
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !from.IP.Equal(net.ParseIP("fe80::2")) || from.Port != port || from.Zone != txName || string(buf[:n]) != "multicast" {
+		t.Fatalf("unexpected multicast reply %q from %s", buf[:n], from)
+	}
+}
+
+func TestLinkLocalMulticastSessionsRetainZones(t *testing.T) {
+	testScopedMulticastSessions(t, "ff02::114", false)
+}
+
+func TestIPv4DestinationSessionsRetainInterfaces(t *testing.T) {
+	ifNames := []string{os.Getenv("FLUX_TEST_MULTICAST_TX"), os.Getenv("FLUX_TEST_MULTICAST_TX2")}
+	if ifNames[0] == "" || ifNames[1] == "" {
+		t.Skip("requires two isolated IPv4 multicast interfaces")
+	}
+	for _, network := range []string{"udp4", "udp"} {
+		for _, destination := range []string{"239.255.0.114", "224.0.0.114", "255.255.255.255"} {
+			t.Run(network+"/"+destination, func(t *testing.T) {
+				bind := net.IPv4zero
+				if network == "udp" {
+					bind = net.IPv6unspecified
+				}
+				server, err := net.ListenUDP(network, &net.UDPAddr{IP: bind})
+				if err != nil {
+					t.Fatal(err)
+				}
+				defer server.Close()
+				pc, err := PreserveDestination(server)
+				if err != nil {
+					t.Fatal(err)
+				}
+				client, err := net.ListenUDP("udp4", &net.UDPAddr{IP: net.ParseIP("198.18.0.1")})
+				if err != nil {
+					t.Fatal(err)
+				}
+				defer client.Close()
+				raw, err := client.SyscallConn()
+				if err != nil {
+					t.Fatal(err)
+				}
+				var sockErr error
+				if err := raw.Control(func(fd uintptr) {
+					sockErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, unix.SO_BROADCAST, 1)
+				}); err != nil {
+					t.Fatal(err)
+				}
+				if sockErr != nil {
+					t.Fatal(sockErr)
+				}
+				target := &net.UDPAddr{IP: net.ParseIP(destination), Port: server.LocalAddr().(*net.UDPAddr).Port}
+				for _, name := range ifNames {
+					iface, err := net.InterfaceByName(name)
+					if err != nil {
+						t.Fatal(err)
+					}
+					if target.IP.IsMulticast() {
+						if err := ipv4.NewPacketConn(server).JoinGroup(iface, target); err != nil {
+							t.Fatal(err)
+						}
+					}
+					if _, err := ipv4.NewPacketConn(client).WriteTo([]byte(name), &ipv4.ControlMessage{
+						Src: client.LocalAddr().(*net.UDPAddr).IP, IfIndex: iface.Index,
+					}, target); err != nil {
+						t.Fatal(err)
+					}
+				}
+				server.SetReadDeadline(time.Now().Add(3 * time.Second))
+				ln := NewListener(pc, &ListenConfig{
+					Backlog: 2, ReadQueueSize: 2, ReadBufferSize: 128, TTL: 10 * time.Second, Keepalive: true,
+					Logger: xlogger.NewLogger(xlogger.OutputOption(io.Discard)),
+				})
+				defer ln.Close()
+				seen := make(map[string]bool)
+				for range ifNames {
+					c, err := ln.Accept()
+					if err != nil {
+						t.Fatalf("interface did not get a separate session: %v", err)
+					}
+					defer c.Close()
+					var buf [128]byte
+					n, err := c.Read(buf[:])
+					if err != nil {
+						t.Fatal(err)
+					}
+					name := string(buf[:n])
+					if (name != ifNames[0] && name != ifNames[1]) || seen[name] ||
+						c.LocalAddr().String() != target.String() || c.RemoteAddr().String() != client.LocalAddr().String() {
+						t.Fatalf("unexpected interface session: data=%q local=%s peer=%s", buf[:n], c.LocalAddr(), c.RemoteAddr())
+					}
+					iface, err := net.InterfaceByName(name)
+					if err != nil {
+						t.Fatal(err)
+					}
+					messages, err := unix.ParseSocketControlMessage(c.(*conn).packet.control)
+					if err != nil || len(messages) != 1 {
+						t.Fatalf("invalid reply control: %v", err)
+					}
+					var info unix.Inet4Pktinfo
+					if _, err := binary.Decode(messages[0].Data, binary.NativeEndian, &info); err != nil {
+						t.Fatal(err)
+					}
+					if int(info.Ifindex) != iface.Index {
+						t.Fatalf("reply interface = %d, want %d", info.Ifindex, iface.Index)
+					}
+					seen[name] = true
+				}
+			})
+		}
+	}
+}
+
+func TestInterfaceLocalMulticastSessionsRetainZones(t *testing.T) {
+	testScopedMulticastSessions(t, "ff01::114", true)
+}
+
+func testScopedMulticastSessions(t *testing.T, destination string, interfaceLocal bool) {
+	t.Helper()
+	rxNames := []string{os.Getenv("FLUX_TEST_MULTICAST_RX"), os.Getenv("FLUX_TEST_MULTICAST_RX2")}
+	txNames := []string{os.Getenv("FLUX_TEST_MULTICAST_TX"), os.Getenv("FLUX_TEST_MULTICAST_TX2")}
+	for i := range rxNames {
+		if rxNames[i] == "" || txNames[i] == "" {
+			t.Skip("requires two isolated multicast veth pairs")
+		}
+	}
+	if interfaceLocal {
+		// Interface-local multicast is looped back on the sending interface.
+		rxNames = txNames
+	}
+	group := net.ParseIP(destination)
+	server, err := net.ListenUDP("udp6", &net.UDPAddr{IP: net.IPv6unspecified})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer server.Close()
+	clientIP := net.ParseIP("fd00:1::1")
+	client, err := net.ListenUDP("udp6", &net.UDPAddr{IP: clientIP})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer client.Close()
+	pc, err := PreserveDestination(server)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for i, rxName := range rxNames {
+		rx, err := net.InterfaceByName(rxName)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if err := ipv6.NewPacketConn(server).JoinGroup(rx, &net.UDPAddr{IP: group}); err != nil {
+			t.Fatal(err)
+		}
+		tx, err := net.InterfaceByName(txNames[i])
+		if err != nil {
+			t.Fatal(err)
+		}
+		if err := ipv6.NewPacketConn(client).SetMulticastInterface(tx); err != nil {
+			t.Fatal(err)
+		}
+		if _, err := client.WriteToUDP([]byte(rxName), &net.UDPAddr{
+			IP: group, Port: server.LocalAddr().(*net.UDPAddr).Port, Zone: txNames[i],
+		}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	server.SetReadDeadline(time.Now().Add(3 * time.Second))
+	ln := NewListener(pc, &ListenConfig{
+		Backlog: 2, ReadQueueSize: 2, ReadBufferSize: 128, TTL: 10 * time.Second, Keepalive: true,
+		Logger: xlogger.NewLogger(xlogger.OutputOption(io.Discard)),
+	})
+	defer ln.Close()
+	first, err := ln.Accept()
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer first.Close()
+	second, err := ln.Accept()
+	if err != nil {
+		t.Fatalf("second interface did not get a separate session: %v", err)
+	}
+	defer second.Close()
+	if first == second {
+		t.Fatal("interface-scoped multicast sessions were shared")
+	}
+	seen := make(map[string]bool)
+	for _, c := range []net.Conn{first, second} {
+		var buf [128]byte
+		n, err := c.Read(buf[:])
+		if err != nil {
+			t.Fatal(err)
+		}
+		zone := string(buf[:n])
+		local, peer := c.LocalAddr().(*net.UDPAddr), c.RemoteAddr().(*net.UDPAddr)
+		if (zone != rxNames[0] && zone != rxNames[1]) || seen[zone] || !local.IP.Equal(group) ||
+			local.Port != server.LocalAddr().(*net.UDPAddr).Port || local.Zone != zone ||
+			!peer.IP.Equal(clientIP) || peer.Port != client.LocalAddr().(*net.UDPAddr).Port || peer.Zone != "" {
+			t.Fatalf("unexpected scoped session: data=%q local=%s peer=%s", buf[:n], local, peer)
+		}
+		var reply ipv6.ControlMessage
+		if err := reply.Parse(c.(*conn).packet.control); err != nil {
+			t.Fatal(err)
+		}
+		iface, err := net.InterfaceByName(zone)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if reply.IfIndex != iface.Index {
+			t.Fatalf("reply interface = %d, want %d", reply.IfIndex, iface.Index)
+		}
+		seen[zone] = true
+	}
+}

go-gost/x/internal/net/udp/packet_other.go

@@ -0,0 +1,9 @@
+//go:build !linux
+
+package udp
+
+import "net"
+
+func PreserveDestination(c *net.UDPConn) (net.PacketConn, error) {
+	return c, nil
+}

go-gost/x/listener/udp/destination_linux_test.go

@@ -0,0 +1,289 @@
+package udp
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"net"
+	"os"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"golang.org/x/sys/unix"
+
+	corelimiter "github.com/go-gost/core/limiter"
+	coretraffic "github.com/go-gost/core/limiter/traffic"
+	corelistener "github.com/go-gost/core/listener"
+	corestats "github.com/go-gost/core/observer/stats"
+	xadmission "github.com/go-gost/x/admission"
+	xtraffic "github.com/go-gost/x/limiter/traffic"
+	xlogger "github.com/go-gost/x/logger"
+	xmetadata "github.com/go-gost/x/metadata"
+	"github.com/go-gost/x/metrics"
+	xstats "github.com/go-gost/x/observer/stats"
+)
+
+type countedTraffic struct {
+	coretraffic.TrafficLimiter
+	in, out atomic.Int64
+}
+
+func (l *countedTraffic) In(ctx context.Context, key string, opts ...corelimiter.Option) coretraffic.Limiter {
+	l.in.Add(1)
+	return l.TrafficLimiter.In(ctx, key, opts...)
+}
+
+func (l *countedTraffic) Out(ctx context.Context, key string, opts ...corelimiter.Option) coretraffic.Limiter {
+	l.out.Add(1)
+	return l.TrafficLimiter.Out(ctx, key, opts...)
+}
+
+func TestDestinationPreservedThroughWrappers(t *testing.T) {
+	v4 := []string{"127.0.0.1", "127.0.0.2"}
+	if address := os.Getenv("FLUX_TEST_IPV4_LINK_LOCAL"); address != "" {
+		v4 = append(v4, address)
+	}
+	v6 := strings.FieldsFunc(os.Getenv("FLUX_TEST_IPV6_ADDRS"), func(r rune) bool { return r == ',' })
+	if len(v6) == 0 {
+		v6 = []string{"::1"}
+	}
+	for _, tc := range []struct {
+		name, bind string
+		targets    []string
+		clientBind string
+	}{
+		{"ipv4-wildcard", "0.0.0.0:0", v4, ""},
+		{"dual-stack-wildcard", "[::]:0", append(v4, v6...), ""},
+		{"fixed-ipv6-source", "[::]:0", v6, "::1"},
+		{"explicit-ipv4", "127.0.0.1:0", []string{"127.0.0.1"}, ""},
+		{"explicit-ipv6", "[::1]:0", []string{"::1"}, ""},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			metrics.Enable(true)
+			defer metrics.Enable(false)
+			log := xlogger.NewLogger(xlogger.OutputOption(io.Discard))
+			stats := xstats.NewStats(false)
+			lim := &countedTraffic{TrafficLimiter: xtraffic.NewTrafficLimiter(
+				xtraffic.LimitsOption("$ 1MB 1MB", "$$ 1MB 1MB"), xtraffic.LoggerOption(log),
+			)}
+			ln := NewListener(corelistener.AddrOption(tc.bind), corelistener.LoggerOption(log),
+				corelistener.ServiceOption(t.Name()), corelistener.StatsOption(stats),
+				corelistener.TrafficLimiterOption(lim),
+				corelistener.AdmissionOption(xadmission.NewAdmission(
+					xadmission.WhitelistOption(true),
+					xadmission.MatchersOption([]string{"127.0.0.0/8", "169.254.0.0/16", "::1", "fd00::/8"}),
+					xadmission.LoggerOption(log),
+				)),
+			)
+			if err := ln.Init(xmetadata.NewMetadata(map[string]any{"keepAlive": true})); err != nil {
+				t.Fatal(err)
+			}
+			defer ln.Close()
+			clientIP := net.IPv6unspecified
+			if tc.clientBind != "" {
+				clientIP = net.ParseIP(tc.clientBind)
+			}
+			client, err := net.ListenUDP("udp", &net.UDPAddr{IP: clientIP})
+			if err != nil {
+				t.Fatal(err)
+			}
+			defer client.Close()
+			client.SetDeadline(time.Now().Add(5 * time.Second))
+			clientAddr := client.LocalAddr().(*net.UDPAddr)
+			var accepted atomic.Int64
+			errors := make(chan error, 128)
+			go func() {
+				for {
+					c, err := ln.Accept()
+					if err != nil {
+						return
+					}
+					accepted.Add(1)
+					go func() {
+						defer c.Close()
+						buf := make([]byte, 128)
+						for {
+							n, err := c.Read(buf)
+							if err != nil {
+								return
+							}
+							var round, target int
+							if _, err := fmt.Sscanf(string(buf[:n]), "%d/%d", &round, &target); err != nil {
+								errors <- err
+								return
+							}
+							local := c.LocalAddr().(*net.UDPAddr)
+							if local.IP.To4() != nil && local.Zone != "" {
+								errors <- fmt.Errorf("IPv4 destination has a zone: %s", local)
+								return
+							}
+							if !local.IP.Equal(net.ParseIP(tc.targets[target])) {
+								errors <- fmt.Errorf("local destination = %s, want %s", local.IP, tc.targets[target])
+								return
+							}
+							peer, ok := c.RemoteAddr().(*net.UDPAddr)
+							if !ok {
+								errors <- fmt.Errorf("peer type changed: %T", c.RemoteAddr())
+								return
+							}
+							if tc.clientBind != "" && (!peer.IP.Equal(clientAddr.IP) || peer.Port != clientAddr.Port) {
+								errors <- fmt.Errorf("peer endpoint = %s, want %s", peer, clientAddr)
+								return
+							}
+							if _, err := c.Write(buf[:n]); err != nil {
+								errors <- err
+								return
+							}
+						}
+					}()
+				}
+			}()
+			port := ln.Addr().(*net.UDPAddr).Port
+			var bytes int
+			for round := 0; round < 3; round++ {
+				for target, ip := range tc.targets {
+					payload := []byte(fmt.Sprintf("%d/%d", round, target))
+					bytes += len(payload)
+					if _, err := client.WriteToUDP(payload, &net.UDPAddr{IP: net.ParseIP(ip), Port: port}); err != nil {
+						t.Fatal(err)
+					}
+				}
+				seen := make(map[int]bool)
+				for range tc.targets {
+					buf := make([]byte, 128)
+					n, from, err := client.ReadFromUDP(buf)
+					if err != nil {
+						select {
+						case serverErr := <-errors:
+							t.Fatal(serverErr)
+						default:
+							t.Fatal(err)
+						}
+					}
+					var gotRound, target int
+					if _, err := fmt.Sscanf(string(buf[:n]), "%d/%d", &gotRound, &target); err != nil {
+						t.Fatal(err)
+					}
+					if gotRound != round || seen[target] || from.Port != port || !from.IP.Equal(net.ParseIP(tc.targets[target])) {
+						t.Fatalf("unexpected reply %q from %s", buf[:n], from)
+					}
+					seen[target] = true
+				}
+			}
+			if accepted.Load() != int64(len(tc.targets)) {
+				t.Fatalf("accepted %d sessions, want %d", accepted.Load(), len(tc.targets))
+			}
+			deadline := time.Now().Add(time.Second)
+			for stats.Get(corestats.KindOutputBytes) < uint64(bytes) && time.Now().Before(deadline) {
+				time.Sleep(time.Millisecond)
+			}
+			if stats.Get(corestats.KindInputBytes) != uint64(bytes) || stats.Get(corestats.KindOutputBytes) != uint64(bytes) {
+				t.Fatalf("traffic stats lost: in=%d out=%d want=%d", stats.Get(corestats.KindInputBytes), stats.Get(corestats.KindOutputBytes), bytes)
+			}
+			if lim.in.Load() == 0 || lim.out.Load() == 0 {
+				t.Fatal("traffic limiter bypassed")
+			}
+		})
+	}
+}
+
+func TestBroadcastDestinationRepliesFromUnicast(t *testing.T) {
+	for _, bind := range []string{"0.0.0.0:0", "[::]:0"} {
+		t.Run(bind, func(t *testing.T) {
+			ln := NewListener(corelistener.AddrOption(bind),
+				corelistener.LoggerOption(xlogger.NewLogger(xlogger.OutputOption(io.Discard))))
+			if err := ln.Init(xmetadata.NewMetadata(map[string]any{"keepAlive": true})); err != nil {
+				t.Fatal(err)
+			}
+			defer ln.Close()
+			client, err := net.ListenUDP("udp4", &net.UDPAddr{IP: net.IPv4(127, 0, 0, 1)})
+			if err != nil {
+				t.Fatal(err)
+			}
+			defer client.Close()
+			raw, err := client.SyscallConn()
+			if err != nil {
+				t.Fatal(err)
+			}
+			var sockErr error
+			if err := raw.Control(func(fd uintptr) {
+				sockErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, unix.SO_BROADCAST, 1)
+			}); err != nil {
+				t.Fatal(err)
+			}
+			if sockErr != nil {
+				t.Fatal(sockErr)
+			}
+			client.SetDeadline(time.Now().Add(5 * time.Second))
+			target := &net.UDPAddr{IP: net.IPv4(127, 255, 255, 255), Port: ln.Addr().(*net.UDPAddr).Port}
+			if _, err := client.WriteToUDP([]byte("broadcast"), target); err != nil {
+				t.Fatal(err)
+			}
+			conn, err := acceptWithTimeout(t, ln, time.Second)
+			if err != nil {
+				t.Fatal(err)
+			}
+			defer conn.Close()
+			if conn.LocalAddr().String() != target.String() {
+				t.Fatalf("local destination = %s, want %s", conn.LocalAddr(), target)
+			}
+			if conn.RemoteAddr().String() != client.LocalAddr().String() {
+				t.Fatalf("peer endpoint = %s, want %s", conn.RemoteAddr(), client.LocalAddr())
+			}
+			var buf [128]byte
+			n, err := conn.Read(buf[:])
+			if err != nil {
+				t.Fatal(err)
+			}
+			if _, err := conn.Write(buf[:n]); err != nil {
+				t.Fatalf("reply to broadcast request: %v", err)
+			}
+			n, from, err := client.ReadFromUDP(buf[:])
+			if err != nil {
+				t.Fatal(err)
+			}
+			if !from.IP.Equal(net.IPv4(127, 0, 0, 1)) || from.Port != target.Port || string(buf[:n]) != "broadcast" {
+				t.Fatalf("unexpected broadcast reply %q from %s", buf[:n], from)
+			}
+		})
+	}
+}
+
+func TestDestinationSessionExpires(t *testing.T) {
+	ln := NewListener(corelistener.AddrOption("0.0.0.0:0"),
+		corelistener.LoggerOption(xlogger.NewLogger(xlogger.OutputOption(io.Discard))))
+	if err := ln.Init(xmetadata.NewMetadata(map[string]any{"keepAlive": true, "ttl": "20ms"})); err != nil {
+		t.Fatal(err)
+	}
+	defer ln.Close()
+	client, err := net.DialUDP("udp4", nil, &net.UDPAddr{IP: net.IPv4(127, 0, 0, 2), Port: ln.Addr().(*net.UDPAddr).Port})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer client.Close()
+	client.Write([]byte("first"))
+	first, err := acceptWithTimeout(t, ln, time.Second)
+	if err != nil {
+		t.Fatal(err)
+	}
+	var b [16]byte
+	first.Read(b[:])
+	time.Sleep(80 * time.Millisecond)
+	if _, err := first.Read(b[:]); err == nil {
+		t.Fatal("idle session did not expire")
+	}
+	client.Write([]byte("second"))
+	second, err := acceptWithTimeout(t, ln, time.Second)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer second.Close()
+	if first == second {
+		t.Fatal("expired session reused")
+	}
+	if second.LocalAddr().(*net.UDPAddr).IP.String() != "127.0.0.2" {
+		t.Fatalf("new session lost destination: %s", second.LocalAddr())
+	}
+}

go-gost/x/listener/udp/listener.go

@@ -56,11 +56,15 @@ func (l *udpListener) Init(md md.Metadata) (err error) {
 		return
 	}
 
-	var conn net.PacketConn
+	socket, err := net.ListenUDP(network, laddr)
-	conn, err = net.ListenUDP(network, laddr)
 	if err != nil {
 		return
 	}
+	conn, err := udp.PreserveDestination(socket)
+	if err != nil {
+		socket.Close()
+		return err
+	}
 	conn = metrics.WrapPacketConn(l.options.Service, conn)
 	conn = stats.WrapPacketConn(conn, l.options.Stats)
 	conn = admission.WrapPacketConn(l.options.Admission, conn)