Merge remote-tracking branch 'origin/master' into multiport

We need to compare just the IPs here and not the IP+Port

This is a regression with the merge of v1.9.4 and the change to
netip.AddrPort

- https://github.com/slackhq/nebula/pull/768/commits/dabce8a1b416e1c0a70ba1254597a31b0d555c79

.github/actions/code-sign/action.yml

@@ -43,15 +43,8 @@ runs:
       with:
         role-to-assume: ${{ inputs.role }}
         aws-region: ${{ inputs.region }}
-        # An STS secret key with special characters does not survive the
-        # pwsh -> make -> MSYS sh -> aws.exe chain, and SigV4 then signs with a
-        # key that no longer matches, so the first S3 upload fails with
-        # SignatureDoesNotMatch. Retries the assume until it comes back clean.
-        # Same fix as DefinedNet/dnclient#867.
-        special-characters-workaround: true
-        # Overridden by the workaround above and kept for whenever that goes:
-        # the default 12 rides out IAM trust-policy propagation, and once the
-        # role is stable a real misconfiguration should fail fast.
+        # Default is 12 retries to ride out IAM trust-policy propagation; once
+        # the role is stable we want a real misconfiguration to fail fast.
         retry-max-attempts: 5
 
     - name: Sign .exe files

.github/workflows/release.yml

@@ -73,11 +73,8 @@ jobs:
   build-darwin:
     name: Build Universal Darwin
     env:
-      HAS_SIGNING_CREDS: ${{ secrets.APPLE_SIGNING_ROLE_ARN != '' }}
+      HAS_SIGNING_CREDS: ${{ secrets.AC_USERNAME != '' }}
     runs-on: macos-latest
-    permissions:
-      id-token: write
-      contents: read
     steps:
       - uses: actions/checkout@v7
 
@@ -86,68 +83,17 @@ jobs:
           go-version: '1.26'
           check-latest: true
 
-      # GitHub holds ARNs, not credentials, and ARNs outlive a rotation
-      - name: Configure AWS credentials
-        if: env.HAS_SIGNING_CREDS == 'true'
-        uses: aws-actions/configure-aws-credentials@v6
-        with:
-          role-to-assume: ${{ secrets.APPLE_SIGNING_ROLE_ARN }}
-          aws-region: us-east-2
-
-      # parse-json-secrets unpacks into SIGNING_* and ASC_*, masked on the way in
-      - name: Fetch signing credentials
-        if: env.HAS_SIGNING_CREDS == 'true'
-        uses: aws-actions/aws-secretsmanager-get-secrets@v3
-        with:
-          parse-json-secrets: true
-          secret-ids: |
-            SIGNING,${{ secrets.APPLE_SIGNING_DEVELOPER_ID_ARN }}
-            ASC,${{ secrets.APPLE_NOTARY_KEY_ARN }}
-
       - name: Import certificates
         if: env.HAS_SIGNING_CREDS == 'true'
         uses: Apple-Actions/import-codesign-certs@v7
         with:
-          p12-file-base64: ${{ env.SIGNING_P12_BASE64 }}
-          p12-password: ${{ env.SIGNING_PASSWORD }}
-
-      # The action imports but does not check the chain validates, which is how a p12
-      # missing its intermediate reaches a failing codesign
-      - name: Check the identity is usable
-        if: env.HAS_SIGNING_CREDS == 'true'
-        run: |
-          : "${SIGNING_IDENTITY_SHA1:?empty, so the secret has no identity_sha1}"
-          identities=$(security find-identity -v -p codesigning signing_temp.keychain)
-          case "$identities" in
-          *"$SIGNING_IDENTITY_SHA1"*) ;;
-          *) printf '%s\n' "$identities" >&2; exit 1 ;;
-          esac
-
-      # notarytool wants the key as a file
-      - name: Write the App Store Connect key
-        if: env.HAS_SIGNING_CREDS == 'true'
-        run: |
-          mkdir -p ~/private_keys
-          chmod 700 ~/private_keys
-          key_path="$HOME/private_keys/AuthKey_${ASC_KEY_ID}.p8"
-          (umask 077; printf '%s\n' "$ASC_PRIVATE_KEY" > "$key_path")
-          echo "ASC_P8=$key_path" >> "$GITHUB_ENV"
-
-      - name: Drop the credentials from the environment
-        if: env.HAS_SIGNING_CREDS == 'true'
-        run: |
-          # The action's own inventory, so a new field in a secret is covered
-          python3 -c '
-          import json, os
-          raw = os.environ.get("SECRETS_LIST_CLEAN_UP")
-          if raw is None and os.environ.get("SIGNING_P12_BASE64"):
-              raise SystemExit("SECRETS_LIST_CLEAN_UP is gone, fetched secrets are not being scrubbed")
-          keep = {"SIGNING_IDENTITY_SHA1", "ASC_KEY_ID", "ASC_ISSUER_ID"}
-          names = [n for n in json.loads(raw or "[]") if n not in keep]
-          print("\n".join(f"{n}=" for n in dict.fromkeys(names)))
-          ' >> "$GITHUB_ENV"
+          p12-file-base64: ${{ secrets.APPLE_DEVELOPER_CERTIFICATE_P12_BASE64 }}
+          p12-password: ${{ secrets.APPLE_DEVELOPER_CERTIFICATE_PASSWORD }}
 
       - name: Build, sign, and notarize
+        env:
+          AC_USERNAME: ${{ secrets.AC_USERNAME }}
+          AC_PASSWORD: ${{ secrets.AC_PASSWORD }}
         run: |
           rm -rf release
           mkdir release
@@ -156,34 +102,17 @@ jobs:
           lipo -create -output ./release/nebula ./build/darwin-amd64/nebula ./build/darwin-arm64/nebula
           lipo -create -output ./release/nebula-cert ./build/darwin-amd64/nebula-cert ./build/darwin-arm64/nebula-cert
 
-          # Unset in a fork, which has no credentials to sign with
-          if [ -n "$SIGNING_IDENTITY_SHA1" ]; then
-            codesign -s "$SIGNING_IDENTITY_SHA1" -f -v --timestamp --options=runtime -i "net.defined.nebula" ./release/nebula
-            codesign -s "$SIGNING_IDENTITY_SHA1" -f -v --timestamp --options=runtime -i "net.defined.nebula-cert" ./release/nebula-cert
+          if [ -n "$AC_USERNAME" ]; then
+            codesign -s "10BC1FDDEB6CE753550156C0669109FAC49E4D1E" -f -v --timestamp --options=runtime -i "net.defined.nebula" ./release/nebula
+            codesign -s "10BC1FDDEB6CE753550156C0669109FAC49E4D1E" -f -v --timestamp --options=runtime -i "net.defined.nebula-cert" ./release/nebula-cert
           fi
 
           zip -j release/nebula-darwin.zip release/nebula-cert release/nebula
 
-          if [ -n "$ASC_P8" ]; then
-            xcrun notarytool submit ./release/nebula-darwin.zip --key "$ASC_P8" --key-id "$ASC_KEY_ID" --issuer "$ASC_ISSUER_ID" --wait
+          if [ -n "$AC_USERNAME" ]; then
+            xcrun notarytool submit ./release/nebula-darwin.zip --team-id "576H3XS7FP" --apple-id "$AC_USERNAME" --password "$AC_PASSWORD" --wait
           fi
 
-      - name: Drop the signing key
-        if: always() && env.HAS_SIGNING_CREDS == 'true'
-        run: |
-          # Locked, not deleted: import-codesign-certs deletes it in its own post
-          # step and fails the job if it is already gone. Locked is unusable.
-          security lock-keychain signing_temp.keychain || true
-          rm -f "$ASC_P8"
-          # Nothing later in this job needs AWS
-          python3 -c '
-          import json, os
-          names = json.loads(os.environ.get("SECRETS_LIST_CLEAN_UP") or "[]")
-          names += ["ASC_P8", "SIGNING_IDENTITY_SHA1", "ASC_KEY_ID", "ASC_ISSUER_ID",
-                    "AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY", "AWS_SESSION_TOKEN"]
-          print("\n".join(f"{n}=" for n in dict.fromkeys(names)))
-          ' >> "$GITHUB_ENV"
-
       - name: Upload artifacts
         uses: actions/upload-artifact@v7
         with:

.github/workflows/smoke.yml

@@ -60,4 +60,12 @@ jobs:
       working-directory: ./.github/workflows/smoke
       run: NAME="smoke-p256" ./smoke.sh
 
+    - name: setup docker image for multiport
+      working-directory: ./.github/workflows/smoke
+      run: NAME="smoke-multiport" MULTIPORT_TX=true MULTIPORT_RX=true MULTIPORT_HANDSHAKE=true ./build.sh
+
+    - name: run smoke
+      working-directory: ./.github/workflows/smoke
+      run: NAME="smoke-multiport" ./smoke.sh
+
     timeout-minutes: 10

.github/workflows/smoke/genconfig.sh

@@ -48,6 +48,10 @@ listen:
 
 tun:
   dev: ${TUN_DEV:-tun0}
+  multiport:
+    tx_enabled: ${MULTIPORT_TX:-false}
+    rx_enabled: ${MULTIPORT_RX:-false}
+    tx_handshake: ${MULTIPORT_HANDSHAKE:-false}
 
 firewall:
   inbound_action: reject

CHANGELOG.md

@@ -7,31 +7,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-## [1.11.1] - 2026-08-21
-
-See the [v1.11.1](https://github.com/slackhq/nebula/milestone/30?closed=1) milestone for a complete list of changes.
-
-### Changed
-
-- IPv6 packets whose next header is a protocol Nebula does not parse (SCTP, GRE, IP-in-IP, etc.) are now
-  classified as that protocol with no ports, closing a firewall bypass where a crafted payload could steer
-  the classifier into reading one as TCP/UDP and matching a TCP/UDP rule. These packets are now matched as
-  their true protocol, so only a `proto: any` rule allows them. If you carry one of these protocols over the
-  overlay, confirm a `proto: any` rule covers it before upgrading, it may have been passing only through this
-  bypass. (#1840)
-- Drop the dependency on `github.com/cyberdelia/go-metrics-graphite`, which has been unmaintained for over ten
-  years, by inlining the small amount of code Nebula used. (#1832)
-
-### Fixed
-
-- The ICMPv6 type was read from the wrong byte when classifying IPv6 packets, so the echo identifier used
-  for conntrack was never picked up. (#1840)
-- Enforce outbound message counter limits so a tunnel is rehandshaked before the counter can wrap, preventing
-  nonce reuse. This is unreachable in practice, but is enforced as a defense-in-depth measure. (#1841)
-- Prevent `nebula-cert ca` from running out of memory on 32bit systems when generating encrypted private keys. (#1834)
-- Tolerate `ErrDumpInterrupted` when listing tun addresses on Linux, so a transient interrupted netlink dump
-  no longer aborts startup. (#1835)
-
 ## [1.11.0] - 2026-07-23
 
 See the [v1.11.0](https://github.com/slackhq/nebula/milestone/25?closed=1) milestone for a complete list of changes.
@@ -895,9 +870,7 @@ created.)
 
 - Initial public release.
 
-[Unreleased]: https://github.com/slackhq/nebula/compare/v1.11.1...HEAD
-[1.11.1]: https://github.com/slackhq/nebula/releases/tag/v1.11.1
-[1.11.0]: https://github.com/slackhq/nebula/releases/tag/v1.11.0
+[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.3...HEAD
 [1.10.3]: https://github.com/slackhq/nebula/releases/tag/v1.10.3
 [1.10.2]: https://github.com/slackhq/nebula/releases/tag/v1.10.2
 [1.10.1]: https://github.com/slackhq/nebula/releases/tag/v1.10.1

Makefile

@@ -268,6 +268,10 @@ smoke-relay-docker: bin-docker
 	cd .github/workflows/smoke/ && ./build-relay.sh
 	cd .github/workflows/smoke/ && ./smoke-relay.sh
 
+smoke-multiport-docker: bin-docker
+	cd .github/workflows/smoke/ && NAME="smoke-multiport" MULTIPORT_TX=true MULTIPORT_RX=true MULTIPORT_HANDSHAKE=true ./build.sh
+	cd .github/workflows/smoke/ && NAME="smoke-multiport" ./smoke.sh
+
 smoke-docker-ipv6: export SMOKE_OVERLAY_IPV6 = 1
 smoke-docker-ipv6: smoke-docker
 

cmd/nebula-cert/ca.go

@@ -8,7 +8,6 @@ import (
 	"fmt"
 	"io"
 	"math"
-	"math/bits"
 	"net/netip"
 	"os"
 	"strings"
@@ -45,20 +44,6 @@ type caFlags struct {
 }
 
 func newCaFlags() *caFlags {
-	// prevent running out of memory on 32-bit systems by defaulting to
-	// RFC9106's recommendation for memory-constrained environments
-	var (
-		defaultArgonMemory     uint
-		defaultArgonIterations uint
-	)
-	if bits.UintSize == 32 {
-		defaultArgonMemory = 64 * 1024
-		defaultArgonIterations = 3
-	} else {
-		defaultArgonMemory = 2 * 1024 * 1024
-		defaultArgonIterations = 1
-	}
-
 	cf := caFlags{set: flag.NewFlagSet("ca", flag.ContinueOnError)}
 	cf.set.Usage = func() {}
 	cf.name = cf.set.String("name", "", "Required: name of the certificate authority")
@@ -70,9 +55,9 @@ func newCaFlags() *caFlags {
 	cf.groups = cf.set.String("groups", "", "Optional: comma separated list of groups. This will limit which groups subordinate certs can use")
 	cf.networks = cf.set.String("networks", "", "Optional: comma separated list of ip address and network in CIDR notation. This will limit which ip addresses and networks subordinate certs can use in networks")
 	cf.unsafeNetworks = cf.set.String("unsafe-networks", "", "Optional: comma separated list of ip address and network in CIDR notation. This will limit which ip addresses and networks subordinate certs can use in unsafe networks")
-	cf.argonMemory = cf.set.Uint("argon-memory", defaultArgonMemory, "Optional: Argon2 memory parameter (in KiB) used for encrypted private key passphrase")
+	cf.argonMemory = cf.set.Uint("argon-memory", 2*1024*1024, "Optional: Argon2 memory parameter (in KiB) used for encrypted private key passphrase")
 	cf.argonParallelism = cf.set.Uint("argon-parallelism", 4, "Optional: Argon2 parallelism parameter used for encrypted private key passphrase")
-	cf.argonIterations = cf.set.Uint("argon-iterations", defaultArgonIterations, "Optional: Argon2 iterations parameter used for encrypted private key passphrase")
+	cf.argonIterations = cf.set.Uint("argon-iterations", 1, "Optional: Argon2 iterations parameter used for encrypted private key passphrase")
 	cf.encryption = cf.set.Bool("encrypt", false, "Optional: prompt for passphrase and write out-key in an encrypted format")
 	cf.curve = cf.set.String("curve", "25519", "EdDSA/ECDSA Curve (25519, P256)")
 	cf.p11url = p11Flag(cf.set)

cmd/nebula-cert/ca_test.go

@@ -7,9 +7,7 @@ import (
 	"bytes"
 	"encoding/pem"
 	"errors"
-	"math/bits"
 	"os"
-	"strconv"
 	"strings"
 	"testing"
 	"time"
@@ -24,18 +22,6 @@ func Test_caSummary(t *testing.T) {
 }
 
 func Test_caHelp(t *testing.T) {
-	var (
-		defaultArgonMemory     string
-		defaultArgonIterations string
-	)
-	if bits.UintSize == 32 {
-		defaultArgonMemory = strconv.Itoa(64 * 1024)
-		defaultArgonIterations = strconv.Itoa(3)
-	} else {
-		defaultArgonMemory = strconv.Itoa(2 * 1024 * 1024)
-		defaultArgonIterations = strconv.Itoa(1)
-	}
-
 	ob := &bytes.Buffer{}
 	caHelp(ob)
 	assert.Equal(
@@ -43,9 +29,9 @@ func Test_caHelp(t *testing.T) {
 		"Usage of "+os.Args[0]+" ca <flags>: create a self signed certificate authority\n"+
 			"  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"+
 			"  -argon-iterations uint\n"+
-			"    \tOptional: Argon2 iterations parameter used for encrypted private key passphrase (default "+defaultArgonIterations+")\n"+
+			"    \tOptional: Argon2 iterations parameter used for encrypted private key passphrase (default 1)\n"+
 			"  -argon-memory uint\n"+
-			"    \tOptional: Argon2 memory parameter (in KiB) used for encrypted private key passphrase (default "+defaultArgonMemory+")\n"+
+			"    \tOptional: Argon2 memory parameter (in KiB) used for encrypted private key passphrase (default 2097152)\n"+
 			"  -argon-parallelism uint\n"+
 			"    \tOptional: Argon2 parallelism parameter used for encrypted private key passphrase (default 4)\n"+
 			"  -curve string\n"+
@@ -202,16 +188,10 @@ func Test_ca(t *testing.T) {
 	k, _ := pem.Decode(rb)
 	ned, err := cert.UnmarshalNebulaEncryptedData(k.Bytes)
 	require.NoError(t, err)
-
-	if bits.UintSize == 32 {
-		assert.Equal(t, uint32(64*1024), ned.EncryptionMetadata.Argon2Parameters.Memory)
-		assert.Equal(t, uint32(3), ned.EncryptionMetadata.Argon2Parameters.Iterations)
-	} else {
-		assert.Equal(t, uint32(2*1024*1024), ned.EncryptionMetadata.Argon2Parameters.Memory)
-		assert.Equal(t, uint32(1), ned.EncryptionMetadata.Argon2Parameters.Iterations)
-	}
-
+	// we won't know salt in advance, so just check start of string
+	assert.Equal(t, uint32(2*1024*1024), ned.EncryptionMetadata.Argon2Parameters.Memory)
 	assert.Equal(t, uint8(4), ned.EncryptionMetadata.Argon2Parameters.Parallelism)
+	assert.Equal(t, uint32(1), ned.EncryptionMetadata.Argon2Parameters.Iterations)
 
 	// verify the key is valid and decrypt-able
 	var curve cert.Curve

config/multiport.go

@@ -0,0 +1,10 @@
+package config
+
+type MultiPortConfig struct {
+	Tx               bool
+	Rx               bool
+	TxBasePort       uint16
+	TxPorts          int
+	TxHandshake      bool
+	TxHandshakeDelay int64
+}

connection_manager.go

@@ -323,12 +323,6 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 		return closeTunnel, hostinfo, nil
 	}
 
-	if hostinfo.ConnectionState != nil && hostinfo.ConnectionState.messageCounter.Load() >= RejectAfterMessages {
-		// Send path can't encrypt a CloseTunnel notify, so just delete locally; the peer recovers via recv_error.
-		hostinfo.logger(cm.l).Error("Dropping tunnel, message counter is exhausted")
-		return deleteTunnel, hostinfo, nil
-	}
-
 	primary := cm.hostMap.Hosts[hostinfo.vpnAddrs[0]]
 	mainHostInfo := true
 	if primary != nil && primary != hostinfo {
@@ -454,11 +448,6 @@ func (cm *connectionManager) shouldSwapPrimary(current *HostInfo) bool {
 		return false
 	}
 
-	if current.ConnectionState.messageCounter.Load() >= RehandshakeAfterMessages {
-		// This tunnel is being rolled for counter exhaustion, never swap back onto its spent key.
-		return false
-	}
-
 	crt := cm.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
 	if crt == nil {
 		//my cert was reloaded away. We should definitely swap from this tunnel
@@ -555,15 +544,6 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 			"reason", "current cert version < pki.initiatingVersion",
 		)
 
-		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
-		return
-	}
-	if hostinfo.ConnectionState.messageCounter.Load() >= RehandshakeAfterMessages {
-		cm.l.Info("Re-handshaking with remote",
-			"vpnAddrs", hostinfo.vpnAddrs,
-			"reason", "message counter rehandshake threshold reached",
-		)
-
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return
 	}

connection_manager_test.go

@@ -199,79 +199,6 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 }
 
-func Test_NewConnectionManager_CounterLimits(t *testing.T) {
-	l := test.NewLogger()
-	localrange := netip.MustParsePrefix("10.1.1.1/24")
-	vpnIp := netip.MustParseAddr("172.1.1.2")
-	preferredRanges := []netip.Prefix{localrange}
-
-	// Very incomplete mock objects
-	hostMap := newHostMap(l)
-	hostMap.preferredRanges.Store(&preferredRanges)
-
-	cs := &CertState{
-		initiatingVersion: cert.Version1,
-		privateKey:        []byte{},
-		v1Cert:            &dummyCert{version: cert.Version1},
-		v1Credential:      nil,
-	}
-
-	lh := newTestLighthouse()
-	ifce := &Interface{
-		hostMap:          hostMap,
-		inside:           &overlaytest.NoopTun{},
-		outside:          &udp.NoopConn{},
-		firewall:         &Firewall{},
-		lightHouse:       lh,
-		pki:              &PKI{},
-		myVpnAddrs:       []netip.Addr{netip.MustParseAddr("172.1.1.1")}, // sorts below vpnIp so shouldSwapPrimary can proceed
-		handshakeManager: NewHandshakeManager(l, hostMap, lh, &udp.NoopConn{}, defaultHandshakeConfig),
-		l:                l,
-	}
-	ifce.pki.cs.Store(cs)
-
-	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(test.NewLogger(), conf, nil)
-	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
-	nc.intf = ifce
-
-	hostinfo := &HostInfo{
-		vpnAddrs:      []netip.Addr{vpnIp},
-		localIndexId:  1099,
-		remoteIndexId: 9901,
-	}
-	hostinfo.ConnectionState = &ConnectionState{
-		myCert: &dummyCert{version: cert.Version1},
-	}
-	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
-
-	// Below the rehandshake threshold, no handshake is started
-	hostinfo.ConnectionState.messageCounter.Store(RehandshakeAfterMessages - 1)
-	nc.tryRehandshake(hostinfo)
-	assert.Nil(t, ifce.handshakeManager.QueryVpnAddr(vpnIp))
-
-	// A tunnel on its current cert would normally swap to primary
-	assert.True(t, nc.shouldSwapPrimary(hostinfo))
-
-	// At the rehandshake threshold, a new handshake is started
-	hostinfo.ConnectionState.messageCounter.Store(RehandshakeAfterMessages)
-	nc.tryRehandshake(hostinfo)
-	assert.NotNil(t, ifce.handshakeManager.QueryVpnAddr(vpnIp))
-
-	// An exhausted tunnel being rolled must never swap back to primary onto its spent key
-	assert.False(t, nc.shouldSwapPrimary(hostinfo))
-
-	// Still below the reject limit, the tunnel stays up
-	nc.In(hostinfo)
-	decision, _, _ := nc.makeTrafficDecision(hostinfo.localIndexId, time.Now())
-	assert.Equal(t, tryRehandshake, decision)
-
-	// At the reject limit, the tunnel is deleted locally without a doomed CloseTunnel notify
-	hostinfo.ConnectionState.messageCounter.Store(RejectAfterMessages)
-	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, time.Now())
-	assert.Equal(t, deleteTunnel, decision)
-}
-
 func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
 	l := test.NewLogger()
 	localrange := netip.MustParsePrefix("10.1.1.1/24")

connection_state.go

@@ -2,7 +2,6 @@ package nebula
 
 import (
 	"encoding/json"
-	"fmt"
 	"log/slog"
 	"sync"
 	"sync/atomic"
@@ -13,18 +12,7 @@ import (
 	"github.com/slackhq/nebula/noiseutil"
 )
 
-const (
-	ReplayWindow = 1024
-
-	// RehandshakeAfterMessages rolls keys inside the AES-GCM data-volume margin (~2^-36 advantage at 64KB frames).
-	RehandshakeAfterMessages = uint64(1) << 34
-
-	// RejectAfterMessages is the nonce ceiling enforced by noiseutil; a tunnel here is deleted locally, not notified.
-	RejectAfterMessages = noiseutil.RejectAfterMessages
-)
-
-// RehandshakeAfterMessages must stay below RejectAfterMessages so tunnels roll before the hard send stop.
-const _ = RejectAfterMessages - RehandshakeAfterMessages
+const ReplayWindow = 1024
 
 type ConnectionState struct {
 	eKey           noiseutil.CipherState
@@ -42,12 +30,7 @@ type ConnectionState struct {
 // completed handshake.Result. It seeds messageCounter and the replay window so
 // that the post-handshake message indices already used on the wire don't count
 // as missed traffic in the data plane.
-func newConnectionStateFromResult(r *handshake.Result) (*ConnectionState, error) {
-	// Refuse a MessageIndex too big for the replay window: it can only be a bug, and would spin the seed loop below.
-	if r.MessageIndex >= ReplayWindow {
-		return nil, fmt.Errorf("handshake message index %d exceeds replay window", r.MessageIndex)
-	}
-
+func newConnectionStateFromResult(r *handshake.Result) *ConnectionState {
 	ci := &ConnectionState{
 		myCert:    r.MyCert,
 		initiator: r.Initiator,
@@ -60,7 +43,7 @@ func newConnectionStateFromResult(r *handshake.Result) (*ConnectionState, error)
 	for i := uint64(1); i <= r.MessageIndex; i++ {
 		ci.window.Update(nil, i)
 	}
-	return ci, nil
+	return ci
 }
 
 func (cs *ConnectionState) MarshalJSON() ([]byte, error) {
@@ -71,16 +54,6 @@ func (cs *ConnectionState) MarshalJSON() ([]byte, error) {
 	})
 }
 
-// NextMessageCounter reserves the next 1-based counter; RejectAfterMessages is the first we refuse, pinned to not wrap.
-func (cs *ConnectionState) NextMessageCounter() (uint64, bool) {
-	c := cs.messageCounter.Add(1)
-	if c >= RejectAfterMessages {
-		cs.messageCounter.Store(RejectAfterMessages)
-		return c, false
-	}
-	return c, true
-}
-
 func (cs *ConnectionState) Curve() cert.Curve {
 	return cs.myCert.Curve()
 }

connection_state_test.go

@@ -6,12 +6,11 @@ import (
 	"time"
 
 	"github.com/flynn/noise"
-	"github.com/rcrowley/go-metrics"
 	"github.com/slackhq/nebula/cert"
 	ct "github.com/slackhq/nebula/cert_test"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/handshake"
 	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -57,6 +56,7 @@ func runTestHandshake(t *testing.T) (initR, respR *handshake.Result) {
 		cert.Version2, initCreds, verifier,
 		func() (uint32, error) { return 1000, nil },
 		true, header.HandshakeIXPSK0,
+		config.MultiPortConfig{},
 	)
 	require.NoError(t, err)
 
@@ -64,6 +64,7 @@ func runTestHandshake(t *testing.T) (initR, respR *handshake.Result) {
 		cert.Version2, respCreds, verifier,
 		func() (uint32, error) { return 2000, nil },
 		false, header.HandshakeIXPSK0,
+		config.MultiPortConfig{},
 	)
 	require.NoError(t, err)
 
@@ -81,51 +82,11 @@ func runTestHandshake(t *testing.T) (initR, respR *handshake.Result) {
 	return initR, respR
 }
 
-func TestConnectionState_NextMessageCounter(t *testing.T) {
-	cs := &ConnectionState{}
-	cs.messageCounter.Store(RejectAfterMessages - 2)
-
-	c, ok := cs.NextMessageCounter()
-	assert.True(t, ok)
-	assert.Equal(t, RejectAfterMessages-1, c)
-
-	// Hitting the limit refuses and pins the counter there
-	c, ok = cs.NextMessageCounter()
-	assert.False(t, ok)
-	assert.Equal(t, RejectAfterMessages, c)
-	assert.Equal(t, RejectAfterMessages, cs.messageCounter.Load())
-
-	// Continued send attempts stay refused and the counter never wraps
-	for i := 0; i < 10; i++ {
-		_, ok = cs.NextMessageCounter()
-		assert.False(t, ok)
-	}
-	assert.Equal(t, RejectAfterMessages, cs.messageCounter.Load())
-}
-
-// TestSendNoMetricsDropsExhausted drives the send path to the exhausted drop; metric and out flag prove it.
-func TestSendNoMetricsDropsExhausted(t *testing.T) {
-	initR, _ := runTestHandshake(t)
-	ci, err := newConnectionStateFromResult(initR)
-	require.NoError(t, err)
-	ci.messageCounter.Store(RejectAfterMessages - 1)
-
-	f := &Interface{l: test.NewLogger(), messageMetrics: &MessageMetrics{txExhausted: metrics.NewCounter()}}
-	hostinfo := &HostInfo{vpnAddrs: []netip.Addr{netip.MustParseAddr("10.0.0.1")}, ConnectionState: ci}
-
-	f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, []byte{}, make([]byte, 12), make([]byte, mtu), 0)
-
-	// The crossing send is refused: it records an exhaustion drop and never reaches connectionManager.Out.
-	assert.Equal(t, int64(1), f.messageMetrics.txExhausted.Count())
-	assert.False(t, hostinfo.out.Load())
-}
-
 func TestNewConnectionStateFromResult(t *testing.T) {
 	initR, respR := runTestHandshake(t)
 
 	t.Run("initiator", func(t *testing.T) {
-		ci, err := newConnectionStateFromResult(initR)
-		require.NoError(t, err)
+		ci := newConnectionStateFromResult(initR)
 		assert.True(t, ci.initiator)
 		assert.Equal(t, initR.MyCert, ci.myCert)
 		assert.Equal(t, initR.RemoteCert, ci.peerCert)
@@ -144,17 +105,8 @@ func TestNewConnectionStateFromResult(t *testing.T) {
 		assert.True(t, ci.window.Check(nil, 3), "counter 3 must not be pre-seeded")
 	})
 
-	t.Run("message index too large is refused", func(t *testing.T) {
-		bad := *initR
-		bad.MessageIndex = ReplayWindow
-		ci, err := newConnectionStateFromResult(&bad)
-		require.Error(t, err)
-		assert.Nil(t, ci)
-	})
-
 	t.Run("responder", func(t *testing.T) {
-		ci, err := newConnectionStateFromResult(respR)
-		require.NoError(t, err)
+		ci := newConnectionStateFromResult(respR)
 		assert.False(t, ci.initiator)
 		assert.Equal(t, respR.MyCert, ci.myCert)
 		assert.Equal(t, respR.RemoteCert, ci.peerCert)

examples/config.yml

@@ -328,6 +328,47 @@ tun:
   # SO_RCVBUFFORCE is used to avoid having to raise the system wide max
   #use_system_route_table_buffer_size: 0
 
+  # EXPERIMENTAL: This option may change or disappear in the future.
+  # Multiport spreads outgoing UDP packets across multiple UDP send ports,
+  # which allows nebula to work around any issues on the underlay network.
+  # Some example issues this could work around:
+  #  - UDP rate limits on a per flow basis.
+  #  - Partial underlay network failure in which some flows work and some don't
+  # Agreement is done during the handshake to decide if multiport mode will
+  # be used for a given tunnel (one side must have tx_enabled set, the other
+  # side must have rx_enabled set)
+  #
+  # NOTE: you cannot use multiport on a host if you are relying on UDP hole
+  # punching to get through a NAT or firewall.
+  #
+  # NOTE: Linux only (uses raw sockets to send). Also currently only works
+  # with IPv4 underlay network remotes.
+  #
+  # The default values are listed below:
+  #multiport:
+    # This host support sending via multiple UDP ports.
+    #tx_enabled: false
+    #
+    # This host supports receiving packets sent from multiple UDP ports.
+    #rx_enabled: false
+    #
+    # How many UDP ports to use when sending. The lowest source port will be
+    # listen.port and go up to (but not including) listen.port + tx_ports.
+    #tx_ports: 100
+    #
+    # NOTE: All of your hosts must be running a version of Nebula that supports
+    # multiport if you want to enable this feature. Older versions of Nebula
+    # will be confused by these multiport handshakes.
+    #
+    # If handshakes are not getting a response, attempt to transmit handshakes
+    # using random UDP source ports (to get around partial underlay network
+    # failures).
+    #tx_handshake: false
+    #
+    # How many unresponded handshakes we should send before we attempt to
+    # send multiport handshakes.
+    #tx_handshake_delay: 2
+
 # Configure logging level
 logging:
   # trace, debug, info, warn, or error. Default is info and is reloadable.

firewall/packet.go

@@ -3,6 +3,7 @@ package firewall
 import (
 	"encoding/json"
 	"fmt"
+	mathrand "math/rand"
 	"net/netip"
 )
 
@@ -65,3 +66,30 @@ func (fp Packet) MarshalJSON() ([]byte, error) {
 		"Fragment":   fp.Fragment,
 	})
 }
+
+// UDPSendPort calculates the UDP port to send from when using multiport mode.
+// The result will be from [0, numBuckets)
+func (fp Packet) UDPSendPort(numBuckets int) uint16 {
+	if numBuckets <= 1 {
+		return 0
+	}
+
+	// If there is no port (like an ICMP packet), pick a random UDP send port
+	if fp.LocalPort == 0 {
+		return uint16(mathrand.Intn(numBuckets))
+	}
+
+	// A decent enough 32bit hash function
+	// Prospecting for Hash Functions
+	// - https://nullprogram.com/blog/2018/07/31/
+	// - https://github.com/skeeto/hash-prospector
+	//   [16 21f0aaad 15 d35a2d97 15] = 0.10760229515479501
+	x := (uint32(fp.LocalPort) << 16) | uint32(fp.RemotePort)
+	x ^= x >> 16
+	x *= 0x21f0aaad
+	x ^= x >> 15
+	x *= 0xd35a2d97
+	x ^= x >> 15
+
+	return uint16(x) % uint16(numBuckets)
+}

go.mod

@@ -7,6 +7,7 @@ require (
 	filippo.io/bigmod v0.1.0
 	github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be
 	github.com/armon/go-radix v1.0.0
+	github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
 	github.com/flynn/noise v1.1.0
 	github.com/gaissmai/bart v0.28.0
 	github.com/gogo/protobuf v1.3.2

go.sum

@@ -19,6 +19,8 @@ github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6r
 github.com/cespare/xxhash/v2 v2.1.1/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
+github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432 h1:M5QgkYacWj0Xs8MhpIK/5uwU02icXpEoSo9sM2aRCps=
+github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432/go.mod h1:xwIwAxMvYnVrGJPe2FKx5prTrnAjGOD8zvDOnxnrrkM=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=

graphite.go

@@ -1,117 +0,0 @@
-package nebula
-
-// This file is a trimmed, inlined copy of the graphite exporter from
-// github.com/cyberdelia/go-metrics-graphite, retaining only the Config type and
-// the Once entrypoint that Nebula uses. The upstream package has been
-// unmaintained for 10+ years, so it was vendored here to drop the dependency.
-// See https://github.com/slackhq/nebula/issues/1831.
-//
-// Copyright 2015 Timothée Peignier. All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are met:
-//
-//  1. Redistributions of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//  2. Redistributions in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
-// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
-// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
-// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
-// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
-// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
-// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
-// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
-// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-import (
-	"bufio"
-	"fmt"
-	"net"
-	"strconv"
-	"strings"
-	"time"
-
-	"github.com/rcrowley/go-metrics"
-)
-
-// graphiteConfigExport provides a container with configuration parameters for
-// the Graphite exporter.
-type graphiteConfigExport struct {
-	Addr          *net.TCPAddr     // Network address to connect to
-	Registry      metrics.Registry // Registry to be exported
-	FlushInterval time.Duration    // Flush interval
-	DurationUnit  time.Duration    // Time conversion unit for durations
-	Prefix        string           // Prefix to be prepended to metric names
-	Percentiles   []float64        // Percentiles to export from timers and histograms
-}
-
-// graphiteOnce performs a single submission to Graphite, returning a non-nil
-// error on failed connections.
-func graphiteOnce(c graphiteConfigExport) error {
-	now := time.Now().Unix()
-	du := float64(c.DurationUnit)
-	flushSeconds := float64(c.FlushInterval) / float64(time.Second)
-	conn, err := net.DialTCP("tcp", nil, c.Addr)
-	if err != nil {
-		return err
-	}
-	defer conn.Close()
-	w := bufio.NewWriter(conn)
-	c.Registry.Each(func(name string, i any) {
-		switch metric := i.(type) {
-		case metrics.Counter:
-			count := metric.Count()
-			fmt.Fprintf(w, "%s.%s.count %d %d\n", c.Prefix, name, count, now)
-			fmt.Fprintf(w, "%s.%s.count_ps %.2f %d\n", c.Prefix, name, float64(count)/flushSeconds, now)
-		case metrics.Gauge:
-			fmt.Fprintf(w, "%s.%s.value %d %d\n", c.Prefix, name, metric.Value(), now)
-		case metrics.GaugeFloat64:
-			fmt.Fprintf(w, "%s.%s.value %f %d\n", c.Prefix, name, metric.Value(), now)
-		case metrics.Histogram:
-			h := metric.Snapshot()
-			ps := h.Percentiles(c.Percentiles)
-			fmt.Fprintf(w, "%s.%s.count %d %d\n", c.Prefix, name, h.Count(), now)
-			fmt.Fprintf(w, "%s.%s.min %d %d\n", c.Prefix, name, h.Min(), now)
-			fmt.Fprintf(w, "%s.%s.max %d %d\n", c.Prefix, name, h.Max(), now)
-			fmt.Fprintf(w, "%s.%s.mean %.2f %d\n", c.Prefix, name, h.Mean(), now)
-			fmt.Fprintf(w, "%s.%s.std-dev %.2f %d\n", c.Prefix, name, h.StdDev(), now)
-			for psIdx, psKey := range c.Percentiles {
-				key := strings.Replace(strconv.FormatFloat(psKey*100.0, 'f', -1, 64), ".", "", 1)
-				fmt.Fprintf(w, "%s.%s.%s-percentile %.2f %d\n", c.Prefix, name, key, ps[psIdx], now)
-			}
-		case metrics.Meter:
-			m := metric.Snapshot()
-			fmt.Fprintf(w, "%s.%s.count %d %d\n", c.Prefix, name, m.Count(), now)
-			fmt.Fprintf(w, "%s.%s.one-minute %.2f %d\n", c.Prefix, name, m.Rate1(), now)
-			fmt.Fprintf(w, "%s.%s.five-minute %.2f %d\n", c.Prefix, name, m.Rate5(), now)
-			fmt.Fprintf(w, "%s.%s.fifteen-minute %.2f %d\n", c.Prefix, name, m.Rate15(), now)
-			fmt.Fprintf(w, "%s.%s.mean %.2f %d\n", c.Prefix, name, m.RateMean(), now)
-		case metrics.Timer:
-			t := metric.Snapshot()
-			ps := t.Percentiles(c.Percentiles)
-			count := t.Count()
-			fmt.Fprintf(w, "%s.%s.count %d %d\n", c.Prefix, name, count, now)
-			fmt.Fprintf(w, "%s.%s.count_ps %.2f %d\n", c.Prefix, name, float64(count)/flushSeconds, now)
-			fmt.Fprintf(w, "%s.%s.min %d %d\n", c.Prefix, name, t.Min()/int64(du), now)
-			fmt.Fprintf(w, "%s.%s.max %d %d\n", c.Prefix, name, t.Max()/int64(du), now)
-			fmt.Fprintf(w, "%s.%s.mean %.2f %d\n", c.Prefix, name, t.Mean()/du, now)
-			fmt.Fprintf(w, "%s.%s.std-dev %.2f %d\n", c.Prefix, name, t.StdDev()/du, now)
-			for psIdx, psKey := range c.Percentiles {
-				key := strings.Replace(strconv.FormatFloat(psKey*100.0, 'f', -1, 64), ".", "", 1)
-				fmt.Fprintf(w, "%s.%s.%s-percentile %.2f %d\n", c.Prefix, name, key, ps[psIdx]/du, now)
-			}
-			fmt.Fprintf(w, "%s.%s.one-minute %.2f %d\n", c.Prefix, name, t.Rate1(), now)
-			fmt.Fprintf(w, "%s.%s.five-minute %.2f %d\n", c.Prefix, name, t.Rate5(), now)
-			fmt.Fprintf(w, "%s.%s.fifteen-minute %.2f %d\n", c.Prefix, name, t.Rate15(), now)
-			fmt.Fprintf(w, "%s.%s.mean-rate %.2f %d\n", c.Prefix, name, t.RateMean(), now)
-		}
-		w.Flush()
-	})
-	return nil
-}

handshake/handshake.proto

@@ -24,6 +24,14 @@ message NebulaHandshakeDetails {
   uint64 Cookie = 4 [deprecated = true];
   uint64 Time = 5;
   uint32 CertVersion = 8;
-  // reserved for WIP multiport
-  reserved 6, 7;
+
+  MultiPortDetails InitiatorMultiPort = 6;
+  MultiPortDetails ResponderMultiPort = 7;
+}
+
+message MultiPortDetails {
+  bool RxSupported = 1;
+  bool TxSupported = 2;
+  uint32 BasePort = 3;
+  uint32 TotalPorts = 4;
 }

handshake/helpers_test.go

@@ -8,6 +8,7 @@ import (
 	"github.com/flynn/noise"
 	"github.com/slackhq/nebula/cert"
 	ct "github.com/slackhq/nebula/cert_test"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 	"github.com/stretchr/testify/require"
 )
@@ -71,6 +72,7 @@ func newTestMachine(
 		cs.version, cs.getCredential,
 		verifier, func() (uint32, error) { return localIndex, nil },
 		initiator, header.HandshakeIXPSK0,
+		config.MultiPortConfig{},
 	)
 	require.NoError(t, err)
 	return m

handshake/machine.go

@@ -3,11 +3,13 @@ package handshake
 import (
 	"bytes"
 	"fmt"
+	"math"
 	"slices"
 	"time"
 
 	"github.com/flynn/noise"
 	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 )
 
@@ -39,6 +41,10 @@ type Result struct {
 	HandshakeTime uint64
 	MessageIndex  uint64 // number of messages exchanged during the handshake
 	Initiator     bool
+
+	MultiportRx       bool
+	MultiportTx       bool
+	MultiportBasePort uint16
 }
 
 // Machine drives a Noise handshake through N messages. It handles Noise
@@ -67,6 +73,8 @@ type Machine struct {
 	remoteCertSet  bool
 	payloadSet     bool
 	failed         bool
+
+	multiport config.MultiPortConfig
 }
 
 // NewMachine creates a handshake state machine. The subtype determines both
@@ -80,6 +88,7 @@ func NewMachine(
 	allocIndex IndexAllocator,
 	initiator bool,
 	subtype header.MessageSubType,
+	multiport config.MultiPortConfig,
 ) (*Machine, error) {
 	info, err := subtypeInfoFor(subtype)
 	if err != nil {
@@ -108,6 +117,8 @@ func NewMachine(
 			Initiator: initiator,
 			Cipher:    cred.cipherSuite,
 		},
+
+		multiport: multiport,
 	}, nil
 }
 
@@ -298,7 +309,7 @@ func (m *Machine) processPayload(msg []byte, flags msgFlags) error {
 	}
 
 	// Assert the payload contains exactly what we expect
-	hasPayloadData := payload.InitiatorIndex != 0 || payload.ResponderIndex != 0 || payload.Time != 0
+	hasPayloadData := payload.InitiatorIndex != 0 || payload.ResponderIndex != 0 || payload.Time != 0 || payload.InitiatorMultiPort != nil || payload.ResponderMultiPort != nil
 	if hasPayloadData != flags.expectsPayload {
 		m.failed = true
 		return ErrUnexpectedContent
@@ -315,8 +326,22 @@ func (m *Machine) processPayload(msg []byte, flags msgFlags) error {
 		var remoteIndex uint32
 		if m.result.Initiator {
 			remoteIndex = payload.ResponderIndex
+			if payload.ResponderMultiPort != nil {
+				m.result.MultiportRx = payload.ResponderMultiPort.RxSupported
+				m.result.MultiportTx = payload.ResponderMultiPort.TxSupported
+				if payload.ResponderMultiPort.BasePort <= math.MaxUint16 {
+					m.result.MultiportBasePort = uint16(payload.ResponderMultiPort.BasePort)
+				}
+			}
 		} else {
 			remoteIndex = payload.InitiatorIndex
+			if payload.InitiatorMultiPort != nil {
+				m.result.MultiportRx = payload.InitiatorMultiPort.RxSupported
+				m.result.MultiportTx = payload.InitiatorMultiPort.TxSupported
+				if payload.InitiatorMultiPort.BasePort <= math.MaxUint16 {
+					m.result.MultiportBasePort = uint16(payload.InitiatorMultiPort.BasePort)
+				}
+			}
 		}
 		// The payload presence check above can be satisfied by Time alone, so a payload
 		// could still carry a zero index here. We need to reject it.
@@ -397,11 +422,28 @@ func (m *Machine) marshalOutgoing(flags msgFlags) ([]byte, error) {
 
 		if m.result.Initiator {
 			p.InitiatorIndex = m.result.LocalIndex
+			if m.multiport.Rx || m.multiport.Tx {
+				p.InitiatorMultiPort = &PayloadMultiPortDetails{
+					RxSupported: m.multiport.Rx,
+					TxSupported: m.multiport.Tx,
+					BasePort:    uint32(m.multiport.TxBasePort),
+					TotalPorts:  uint32(m.multiport.TxPorts),
+				}
+			}
 		} else {
 			p.ResponderIndex = m.result.LocalIndex
 			p.InitiatorIndex = m.result.RemoteIndex
+			if m.multiport.Rx || m.multiport.Tx {
+				p.ResponderMultiPort = &PayloadMultiPortDetails{
+					RxSupported: m.multiport.Rx,
+					TxSupported: m.multiport.Tx,
+					BasePort:    uint32(m.multiport.TxBasePort),
+					TotalPorts:  uint32(m.multiport.TxPorts),
+				}
+			}
 		}
 		p.Time = uint64(time.Now().UnixNano())
+
 	}
 	if flags.expectsCert {
 		cred := m.getCred(m.myVersion)

handshake/machine_test.go

@@ -8,6 +8,7 @@ import (
 	"github.com/flynn/noise"
 	"github.com/slackhq/nebula/cert"
 	ct "github.com/slackhq/nebula/cert_test"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/noiseutil"
 	"github.com/stretchr/testify/assert"
@@ -444,6 +445,7 @@ func TestMachineThreeMessagePattern(t *testing.T) {
 		initCS.getCredential, v,
 		func() (uint32, error) { return 1000, nil },
 		true, header.HandshakeXXPSK0,
+		config.MultiPortConfig{},
 	)
 	require.NoError(t, err)
 
@@ -452,6 +454,7 @@ func TestMachineThreeMessagePattern(t *testing.T) {
 		respCS.getCredential, v,
 		func() (uint32, error) { return 2000, nil },
 		false, header.HandshakeXXPSK0,
+		config.MultiPortConfig{},
 	)
 	require.NoError(t, err)
 

handshake/payload.go

@@ -20,6 +20,16 @@ type Payload struct {
 	ResponderIndex uint32
 	Time           uint64
 	CertVersion    uint32
+
+	InitiatorMultiPort *PayloadMultiPortDetails
+	ResponderMultiPort *PayloadMultiPortDetails
+}
+
+type PayloadMultiPortDetails struct {
+	RxSupported bool
+	TxSupported bool
+	BasePort    uint32
+	TotalPorts  uint32
 }
 
 // Proto field numbers for NebulaHandshakeDetails
@@ -29,6 +39,17 @@ const (
 	fieldResponderIndex = 3 // uint32
 	fieldTime           = 5 // uint64
 	fieldCertVersion    = 8 // uint32
+
+	fieldInitiatorMultiPort = 6 // MultiPortDetails
+	fieldResponderMultiPort = 7 // MultiPortDetails
+)
+
+// Proto field numbers for MultiPortDetails
+const (
+	fieldMultiportRxSupported = 1 // bool
+	fieldMultiportTxSupported = 2 // bool
+	fieldMultiportBasePort    = 3 // uint32
+	fieldMultiportTotalPorts  = 4 // uint32
 )
 
 // MarshalPayload encodes a handshake payload in protobuf wire format compatible
@@ -57,6 +78,16 @@ func MarshalPayload(out []byte, p Payload) []byte {
 		details = protowire.AppendTag(details, fieldCertVersion, protowire.VarintType)
 		details = protowire.AppendVarint(details, uint64(p.CertVersion))
 	}
+	if p.InitiatorMultiPort != nil {
+		details = protowire.AppendTag(details, fieldInitiatorMultiPort, protowire.BytesType)
+		details = protowire.AppendVarint(details, uint64(p.InitiatorMultiPort.size()))
+		details = p.InitiatorMultiPort.marshal(details)
+	}
+	if p.ResponderMultiPort != nil {
+		details = protowire.AppendTag(details, fieldResponderMultiPort, protowire.BytesType)
+		details = protowire.AppendVarint(details, uint64(p.ResponderMultiPort.size()))
+		details = p.ResponderMultiPort.marshal(details)
+	}
 
 	out = protowire.AppendTag(out, 1, protowire.BytesType)
 	out = protowire.AppendBytes(out, details)
@@ -64,6 +95,23 @@ func MarshalPayload(out []byte, p Payload) []byte {
 	return out
 }
 
+func (p PayloadMultiPortDetails) marshal(details []byte) []byte {
+	details = protowire.AppendTag(details, fieldMultiportRxSupported, protowire.VarintType)
+	details = protowire.AppendVarint(details, protowire.EncodeBool(p.RxSupported))
+	details = protowire.AppendTag(details, fieldMultiportTxSupported, protowire.VarintType)
+	details = protowire.AppendVarint(details, protowire.EncodeBool(p.TxSupported))
+	details = protowire.AppendTag(details, fieldMultiportBasePort, protowire.VarintType)
+	details = protowire.AppendVarint(details, uint64(p.BasePort))
+	details = protowire.AppendTag(details, fieldMultiportTotalPorts, protowire.VarintType)
+	details = protowire.AppendVarint(details, uint64(p.TotalPorts))
+
+	return details
+}
+
+func (p PayloadMultiPortDetails) size() int {
+	return 4 + 2 + protowire.SizeVarint(uint64(p.BasePort)) + protowire.SizeVarint(uint64(p.TotalPorts))
+}
+
 // UnmarshalPayload decodes a protobuf-encoded NebulaHandshake message.
 func UnmarshalPayload(b []byte) (Payload, error) {
 	var p Payload
@@ -161,6 +209,97 @@ func unmarshalPayloadDetails(p *Payload, b []byte) error {
 			}
 			p.CertVersion = uint32(v)
 			b = b[n:]
+		case fieldInitiatorMultiPort:
+			if typ != protowire.BytesType {
+				return errInvalidHandshakeDetails
+			}
+			d, n := protowire.ConsumeBytes(b)
+			if n < 0 {
+				return errInvalidHandshakeMessage
+			}
+			b = b[n:]
+			p.InitiatorMultiPort = new(PayloadMultiPortDetails)
+			if err := unmarshalPayloadMultiPortDetails(p.InitiatorMultiPort, d); err != nil {
+				return err
+			}
+		case fieldResponderMultiPort:
+			if typ != protowire.BytesType {
+				return errInvalidHandshakeDetails
+			}
+			d, n := protowire.ConsumeBytes(b)
+			if n < 0 {
+				return errInvalidHandshakeMessage
+			}
+			b = b[n:]
+			p.ResponderMultiPort = new(PayloadMultiPortDetails)
+			if err := unmarshalPayloadMultiPortDetails(p.ResponderMultiPort, d); err != nil {
+				return err
+			}
+		default:
+			n := protowire.ConsumeFieldValue(num, typ, b)
+			if n < 0 {
+				return errInvalidHandshakeDetails
+			}
+			b = b[n:]
+		}
+	}
+	return nil
+}
+
+func unmarshalPayloadMultiPortDetails(p *PayloadMultiPortDetails, b []byte) error {
+	for len(b) > 0 {
+		num, typ, n := protowire.ConsumeTag(b)
+		if n < 0 {
+			return errInvalidHandshakeDetails
+		}
+		b = b[n:]
+
+		// For known field numbers, reject any non-matching wire type as a
+		// hard error rather than silently skipping. The caller will catch
+		// missing-field cases downstream, but a wire-type mismatch on a tag
+		// we know is a peer protocol violation worth flagging here.
+		// Repeated occurrences of a singular field follow proto3 last-wins.
+		switch num {
+		case fieldMultiportRxSupported:
+			if typ != protowire.VarintType {
+				return errInvalidHandshakeDetails
+			}
+			v, n := protowire.ConsumeVarint(b)
+			if n < 0 || v > math.MaxUint32 {
+				return errInvalidHandshakeDetails
+			}
+			p.RxSupported = protowire.DecodeBool(v)
+			b = b[n:]
+		case fieldMultiportTxSupported:
+			if typ != protowire.VarintType {
+				return errInvalidHandshakeDetails
+			}
+			v, n := protowire.ConsumeVarint(b)
+			if n < 0 || v > math.MaxUint32 {
+				return errInvalidHandshakeDetails
+			}
+			p.TxSupported = protowire.DecodeBool(v)
+			b = b[n:]
+		case fieldMultiportBasePort:
+			if typ != protowire.VarintType {
+				return errInvalidHandshakeDetails
+			}
+			v, n := protowire.ConsumeVarint(b)
+			if n < 0 || v > math.MaxUint32 {
+				return errInvalidHandshakeDetails
+			}
+			p.BasePort = uint32(v)
+			b = b[n:]
+		case fieldMultiportTotalPorts:
+			if typ != protowire.VarintType {
+				return errInvalidHandshakeDetails
+			}
+			v, n := protowire.ConsumeVarint(b)
+			if n < 0 || v > math.MaxUint32 {
+				return errInvalidHandshakeDetails
+			}
+			p.TotalPorts = uint32(v)
+			b = b[n:]
 		default:
 			n := protowire.ConsumeFieldValue(num, typ, b)
 			if n < 0 {

handshake/payload_test.go

@@ -117,24 +117,24 @@ func TestPayloadUnknownFields(t *testing.T) {
 		assert.Equal(t, uint32(88), got.ResponderIndex)
 	})
 
-	t.Run("reserved fields 6 and 7 are skipped", func(t *testing.T) {
-		// Fields 6 and 7 are reserved in the proto definition
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, 100)
-		details = protowire.AppendTag(details, 6, protowire.VarintType)
-		details = protowire.AppendVarint(details, 1)
-		details = protowire.AppendTag(details, 7, protowire.VarintType)
-		details = protowire.AppendVarint(details, 2)
+	// t.Run("reserved fields 6 and 7 are skipped", func(t *testing.T) {
+	// 	// Fields 6 and 7 are reserved in the proto definition
+	// 	var details []byte
+	// 	details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
+	// 	details = protowire.AppendVarint(details, 100)
+	// 	details = protowire.AppendTag(details, 6, protowire.VarintType)
+	// 	details = protowire.AppendVarint(details, 1)
+	// 	details = protowire.AppendTag(details, 7, protowire.VarintType)
+	// 	details = protowire.AppendVarint(details, 2)
 
-		var data []byte
-		data = protowire.AppendTag(data, 1, protowire.BytesType)
-		data = protowire.AppendBytes(data, details)
+	// 	var data []byte
+	// 	data = protowire.AppendTag(data, 1, protowire.BytesType)
+	// 	data = protowire.AppendBytes(data, details)
 
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-		assert.Equal(t, uint32(100), got.InitiatorIndex)
-	})
+	// 	got, err := UnmarshalPayload(data)
+	// 	require.NoError(t, err)
+	// 	assert.Equal(t, uint32(100), got.InitiatorIndex)
+	// })
 }
 
 func TestPayloadBytesConsumed(t *testing.T) {

handshake_manager.go

@@ -14,6 +14,7 @@ import (
 
 	"github.com/rcrowley/go-metrics"
 	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/handshake"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/udp"
@@ -71,6 +72,9 @@ type HandshakeManager struct {
 	f                      *Interface
 	l                      *slog.Logger
 
+	multiPort config.MultiPortConfig
+	udpRaw    *udp.RawConn
+
 	// can be used to trigger outbound handshake for the given vpnIp
 	trigger chan netip.Addr
 }
@@ -291,6 +295,7 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 
 	// Send the handshake to all known ips, stage 2 takes care of assigning the hostinfo.remote based on the first to reply
 	var sentTo []netip.AddrPort
+	var sentMultiport bool
 	hostinfo.remotes.ForEach(hm.mainHostMap.GetPreferredRanges(), func(addr netip.AddrPort, _ bool) {
 		hm.messageMetrics.Tx(header.Handshake, hh.machine.Subtype(), 1)
 		err := hm.outside.WriteTo(stage0, addr)
@@ -311,6 +316,29 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 		} else {
 			sentTo = append(sentTo, addr)
 		}
+
+		// Attempt a multiport handshake if we are past the TxHandshakeDelay attempts
+		if hm.multiPort.TxHandshake && hm.udpRaw != nil && hh.counter >= hm.multiPort.TxHandshakeDelay {
+			sentMultiport = true
+			// We need to re-allocate with 8 bytes at the start of SOCK_RAW
+			raw := hostinfo.HandshakePacket[0x80]
+			if raw == nil {
+				raw = make([]byte, len(hostinfo.HandshakePacket[0])+udp.RawOverhead)
+				copy(raw[udp.RawOverhead:], hostinfo.HandshakePacket[0])
+				hostinfo.HandshakePacket[0x80] = raw
+			}
+
+			hm.messageMetrics.Tx(header.Handshake, header.MessageSubType(hostinfo.HandshakePacket[0][1]), 1)
+			err = hm.udpRaw.WriteTo(raw, udp.RandomSendPort.UDPSendPort(hm.multiPort.TxPorts), addr)
+			if err != nil {
+				hostinfo.logger(hm.l).Error("Failed to send handshake message",
+					"error", err,
+					"udpAddr", addr,
+					"initiatorIndex", hostinfo.localIndexId,
+					"handshake", hsFields,
+				)
+			}
+		}
 	})
 
 	// Don't be too noisy or confusing if we fail to send a handshake - if we don't get through we'll eventually log a timeout,
@@ -320,6 +348,7 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 			"udpAddrs", sentTo,
 			"initiatorIndex", hostinfo.localIndexId,
 			"handshake", hsFields,
+			"multiportHandshake", sentMultiport,
 		)
 	} else if hm.l.Enabled(context.Background(), slog.LevelDebug) {
 		hostinfo.logger(hm.l).Debug("Handshake message sent",
@@ -672,6 +701,7 @@ func (hm *HandshakeManager) buildStage0Packet(hh *HandshakeHostInfo) bool {
 		v, cs.GetCredential,
 		hm.certVerifier(), func() (uint32, error) { return hm.allocateIndex(hh) },
 		true, header.HandshakeIXPSK0,
+		hm.multiPort,
 	)
 	if err != nil {
 		hm.f.l.Error("Failed to create handshake machine",
@@ -713,6 +743,7 @@ func (hm *HandshakeManager) beginHandshake(via ViaSender, packet []byte, h *head
 		v, cs.GetCredential,
 		hm.certVerifier(), func() (uint32, error) { return generateIndex(f.l) },
 		false, header.HandshakeIXPSK0,
+		hm.multiPort,
 	)
 	if err != nil {
 		f.l.Error("Failed to create handshake machine", "from", via, "error", err)
@@ -737,6 +768,12 @@ func (hm *HandshakeManager) beginHandshake(via ViaSender, packet []byte, h *head
 		return
 	}
 
+	if !via.IsRelayed && result.MultiportTx && result.MultiportBasePort != via.UdpAddr.Port() {
+		// The other side sent us a handshake from a different port, make sure
+		// we send responses back to the BasePort
+		via.UdpAddr = netip.AddrPortFrom(via.UdpAddr.Addr(), result.MultiportBasePort)
+	}
+
 	remoteCert := result.RemoteCert
 	if remoteCert == nil {
 		f.l.Error("Handshake did not produce a peer certificate", "from", via)
@@ -749,14 +786,8 @@ func (hm *HandshakeManager) beginHandshake(via ViaSender, packet []byte, h *head
 		return
 	}
 
-	connState, err := newConnectionStateFromResult(result)
-	if err != nil {
-		f.l.Error("Discarding handshake with an invalid message index", "error", err, "vpnAddrs", vpnAddrs)
-		return
-	}
-
 	hostinfo := &HostInfo{
-		ConnectionState:   connState,
+		ConnectionState:   newConnectionStateFromResult(result),
 		localIndexId:      result.LocalIndex,
 		remoteIndexId:     result.RemoteIndex,
 		vpnAddrs:          vpnAddrs,
@@ -767,6 +798,8 @@ func (hm *HandshakeManager) beginHandshake(via ViaSender, packet []byte, h *head
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},
+		multiportTx: hm.multiPort.Tx && result.MultiportRx,
+		multiportRx: hm.multiPort.Rx && result.MultiportTx,
 	}
 
 	msg := "Handshake message received"
@@ -783,6 +816,8 @@ func (hm *HandshakeManager) beginHandshake(via ViaSender, packet []byte, h *head
 		"initiatorIndex", result.RemoteIndex,
 		"responderIndex", result.LocalIndex,
 		"handshake", m{"stage": uint64(machine.MessageIndex()), "style": header.SubTypeName(header.Handshake, machine.Subtype())},
+		"multiportTx", hostinfo.multiportTx,
+		"multiportRx", hostinfo.multiportRx,
 	)
 
 	// packet aliases the listener's incoming buffer, so this copy must stay.
@@ -873,14 +908,16 @@ func (hm *HandshakeManager) continueHandshake(via ViaSender, hh *HandshakeHostIn
 		return
 	}
 
-	// Handshake complete; build the ConnectionState now that we have keys and a verified peer cert.
-	cs, err := newConnectionStateFromResult(result)
-	if err != nil {
-		f.l.Error("Discarding handshake with an invalid message index", "error", err, "vpnAddrs", hostinfo.vpnAddrs)
-		hm.DeleteHostInfo(hostinfo)
-		return
+	if !via.IsRelayed && result.MultiportTx && result.MultiportBasePort != via.UdpAddr.Port() {
+		// The other side sent us a handshake from a different port, make sure
+		// we send responses back to the BasePort
+		via.UdpAddr = netip.AddrPortFrom(via.UdpAddr.Addr(), result.MultiportBasePort)
 	}
-	hostinfo.ConnectionState = cs
+	hostinfo.multiportTx = hm.multiPort.Tx && result.MultiportRx
+	hostinfo.multiportRx = hm.multiPort.Rx && result.MultiportTx
+
+	// Handshake complete; build the ConnectionState now that we have keys and a verified peer cert.
+	hostinfo.ConnectionState = newConnectionStateFromResult(result)
 
 	remoteCert := result.RemoteCert
 	if remoteCert == nil {
@@ -973,6 +1010,8 @@ func (hm *HandshakeManager) continueHandshake(via ViaSender, hh *HandshakeHostIn
 		"handshake", m{"stage": uint64(machine.MessageIndex()), "style": header.SubTypeName(header.Handshake, machine.Subtype())},
 		"durationNs", duration,
 		"sentCachedPackets", len(hh.packetStore),
+		"multiportTx", hostinfo.multiportTx,
+		"multiportRx", hostinfo.multiportRx,
 	)
 
 	hostinfo.vpnAddrs = vpnAddrs
@@ -1114,6 +1153,10 @@ func (hm *HandshakeManager) handleCheckAndCompleteError(err error, existing, hos
 
 	switch err {
 	case ErrAlreadySeen:
+		if hostinfo.multiportRx {
+			// The other host is sending to us with multiport, so only grab the IP
+			via.UdpAddr = netip.AddrPortFrom(via.UdpAddr.Addr(), hostinfo.GetRemote().Port())
+		}
 		if existing.SetRemoteIfPreferred(f.hostMap, via) {
 			f.SendMessageToVpnAddr(header.Test, header.TestRequest, hostinfo.vpnAddrs[0], []byte(""), make([]byte, 12, 12), make([]byte, mtu))
 		}

hostmap.go

@@ -254,6 +254,12 @@ type HostInfo struct {
 	networks   *bart.Table[NetworkType]
 	relayState RelayState
 
+	// If true, we should send to this remote using multiport
+	multiportTx bool
+
+	// If true, we will receive from this remote using multiport
+	multiportRx bool
+
 	// HandshakePacket records the packets used to create this hostinfo
 	// We need these to avoid replayed handshake packets creating new hostinfos which causes churn
 	HandshakePacket map[uint8][]byte

inside.go

@@ -10,6 +10,7 @@ import (
 	"github.com/slackhq/nebula/iputil"
 	"github.com/slackhq/nebula/noiseutil"
 	"github.com/slackhq/nebula/routing"
+	"github.com/slackhq/nebula/udp"
 )
 
 func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache) {
@@ -73,7 +74,7 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 
 	dropReason := f.firewall.Drop(*fwPacket, false, hostinfo, f.pki.GetCAPool(), localCache)
 	if dropReason == nil {
-		f.sendNoMetrics(header.Message, 0, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, packet, nb, out, q)
+		f.sendNoMetrics(header.Message, 0, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, packet, nb, out, q, fwPacket)
 
 	} else {
 		f.rejectInside(packet, out, q)
@@ -122,7 +123,7 @@ func (f *Interface) rejectOutside(packet []byte, ci *ConnectionState, hostinfo *
 		return
 	}
 
-	f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, out, nb, packet, q)
+	f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, out, nb, packet, q, nil)
 }
 
 // Handshake will attempt to initiate a tunnel with the provided vpn address. This is a no-op if the tunnel is already established or being established
@@ -235,7 +236,7 @@ func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubTyp
 		return
 	}
 
-	f.sendNoMetrics(header.Message, st, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, p, nb, out, 0)
+	f.sendNoMetrics(header.Message, st, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, p, nb, out, 0, nil)
 }
 
 // SendMessageToVpnAddr handles real addr:port lookup and sends to the current best known address for vpnAddr.
@@ -267,20 +268,12 @@ func (f *Interface) SendMessageToHostInfo(t header.MessageType, st header.Messag
 
 func (f *Interface) send(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, p, nb, out []byte) {
 	f.messageMetrics.Tx(t, st, 1)
-	f.sendNoMetrics(t, st, ci, hostinfo, netip.AddrPort{}, p, nb, out, 0)
+	f.sendNoMetrics(t, st, ci, hostinfo, netip.AddrPort{}, p, nb, out, 0, nil)
 }
 
 func (f *Interface) sendTo(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte) {
 	f.messageMetrics.Tx(t, st, 1)
-	f.sendNoMetrics(t, st, ci, hostinfo, remote, p, nb, out, 0)
-}
-
-// dropExhausted records an exhaustion drop and logs once, on the crossing send, for a spent tunnel.
-func (f *Interface) dropExhausted(hostinfo *HostInfo, c uint64, msg string) {
-	f.messageMetrics.TxExhausted(1)
-	if c == RejectAfterMessages {
-		hostinfo.logger(f.l).Error(msg)
-	}
+	f.sendNoMetrics(t, st, ci, hostinfo, remote, p, nb, out, 0, nil)
 }
 
 // SendVia sends a payload through a Relay tunnel. No authentication or encryption is done
@@ -302,14 +295,7 @@ func (f *Interface) SendVia(via *HostInfo,
 		// NOTE: for goboring AESGCMTLS we need to lock because of the nonce check
 		via.ConnectionState.writeLock.Lock()
 	}
-	c, ok := via.ConnectionState.NextMessageCounter()
-	if !ok {
-		if noiseutil.EncryptLockNeeded {
-			via.ConnectionState.writeLock.Unlock()
-		}
-		f.dropExhausted(via, c, "Dropping outbound relay packets, tunnel message counter is exhausted")
-		return
-	}
+	c := via.ConnectionState.messageCounter.Add(1)
 
 	out = header.Encode(out, header.Version, header.Message, header.MessageRelay, relay.RemoteIndex, c)
 	f.connectionManager.Out(via)
@@ -355,10 +341,27 @@ func (f *Interface) SendVia(via *HostInfo,
 	f.connectionManager.RelayUsed(relay.LocalIndex)
 }
 
-func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte, q int) {
+func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte, q int, udpPortGetter udp.SendPortGetter) {
 	if ci.eKey == nil {
 		return
 	}
+
+	multiport := f.multiPort.Tx && hostinfo.multiportTx
+	rawOut := out
+	if multiport {
+		if len(out) < udp.RawOverhead {
+			// NOTE: This is because some spots in the code send us `out[:0]`, so
+			// we need to expand the slice back out to get our 8 bytes back.
+			out = out[:udp.RawOverhead]
+		}
+		// Preserve bytes needed for the raw socket
+		out = out[udp.RawOverhead:]
+
+		if udpPortGetter == nil {
+			udpPortGetter = udp.RandomSendPort
+		}
+	}
+
 	useRelay := !remote.IsValid() && !hostinfo.GetRemote().IsValid()
 	fullOut := out
 
@@ -376,14 +379,7 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 		// NOTE: for goboring AESGCMTLS we need to lock because of the nonce check
 		ci.writeLock.Lock()
 	}
-	c, ok := ci.NextMessageCounter()
-	if !ok {
-		if noiseutil.EncryptLockNeeded {
-			ci.writeLock.Unlock()
-		}
-		f.dropExhausted(hostinfo, c, "Dropping outbound packets, tunnel message counter is exhausted")
-		return
-	}
+	c := ci.messageCounter.Add(1)
 
 	//l.WithField("trace", string(debug.Stack())).Error("out Header ", &Header{Version, t, st, 0, hostinfo.remoteIndexId, c}, p)
 	out = header.Encode(out, header.Version, t, st, hostinfo.remoteIndexId, c)
@@ -418,7 +414,13 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 	}
 
 	if remote.IsValid() {
-		err = f.writers[q].WriteTo(out, remote)
+		if multiport {
+			rawOut = rawOut[:len(out)+udp.RawOverhead]
+			port := udpPortGetter.UDPSendPort(f.multiPort.TxPorts)
+			err = f.udpRaw.WriteTo(rawOut, port, remote)
+		} else {
+			err = f.writers[q].WriteTo(out, remote)
+		}
 		if err != nil {
 			hostinfo.logger(f.l).Error("Failed to write outgoing packet",
 				"error", err,
@@ -426,7 +428,13 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 			)
 		}
 	} else if hr := hostinfo.GetRemote(); hr.IsValid() {
-		err = f.writers[q].WriteTo(out, hr)
+		if multiport {
+			rawOut = rawOut[:len(out)+udp.RawOverhead]
+			port := udpPortGetter.UDPSendPort(f.multiPort.TxPorts)
+			err = f.udpRaw.WriteTo(rawOut, port, hr)
+		} else {
+			err = f.writers[q].WriteTo(out, hr)
+		}
 		if err != nil {
 			hostinfo.logger(f.l).Error("Failed to write outgoing packet",
 				"error", err,

interface.go

@@ -99,6 +99,9 @@ type Interface struct {
 	// triggerShutdown is a function that will be run exactly once, when onFatal swaps something non-nil into fatalErr
 	triggerShutdown func()
 
+	udpRaw    *udp.RawConn
+	multiPort config.MultiPortConfig
+
 	metricHandshakes    metrics.Histogram
 	messageMetrics      *MessageMetrics
 	cachedPacketMetrics *cachedPacketMetrics
@@ -106,6 +109,15 @@ type Interface struct {
 	l *slog.Logger
 }
 
+type MultiPortConfig struct {
+	Tx               bool
+	Rx               bool
+	TxBasePort       uint16
+	TxPorts          int
+	TxHandshake      bool
+	TxHandshakeDelay int64
+}
+
 type EncWriter interface {
 	SendVia(via *HostInfo,
 		relay *Relay,
@@ -249,6 +261,8 @@ func (f *Interface) activate() error {
 
 	metrics.GetOrRegisterGauge("routines", nil).Update(int64(f.routines))
 
+	metrics.GetOrRegisterGauge("multiport.tx_ports", nil).Update(int64(f.multiPort.TxPorts))
+
 	// Prepare n tun queues
 	var reader io.ReadWriteCloser = f.inside
 	for i := 0; i < f.routines; i++ {
@@ -505,6 +519,8 @@ func (f *Interface) emitStats(ctx context.Context, i time.Duration) {
 
 	udpStats := udp.NewUDPStatsEmitter(f.writers)
 
+	var rawStats func()
+
 	certExpirationGauge := metrics.GetOrRegisterGauge("certificate.ttl_seconds", nil)
 	certInitiatingVersion := metrics.GetOrRegisterGauge("certificate.initiating_version", nil)
 	certMaxVersion := metrics.GetOrRegisterGauge("certificate.max_version", nil)
@@ -519,6 +535,13 @@ func (f *Interface) emitStats(ctx context.Context, i time.Duration) {
 		certExpirationGauge.Update(int64(defaultCrt.NotAfter().Sub(time.Now()) / time.Second))
 		certInitiatingVersion.Update(int64(defaultCrt.Version()))
 
+		if f.udpRaw != nil {
+			if rawStats == nil {
+				rawStats = udp.NewRawStatsEmitter(f.udpRaw)
+			}
+			rawStats()
+		}
+
 		// Report the max certificate version we are capable of using
 		if certState.v2Cert != nil {
 			certMaxVersion.Update(int64(certState.v2Cert.Version()))

iputil/packet.go

@@ -2,16 +2,11 @@ package iputil
 
 import (
 	"encoding/binary"
-	"errors"
 
 	"golang.org/x/net/ipv4"
 	"golang.org/x/net/ipv6"
 )
 
-// ErrIPv6CouldNotFindPayload is returned when the ipv6 extension header chain is truncated before a terminal
-// upper layer protocol is reached.
-var ErrIPv6CouldNotFindPayload = errors.New("could not find payload in ipv6 packet")
-
 const (
 	// MaxIPv4RejectPacketSize is the largest IPv4 reject packet:
 	// - 20 byte ipv4 header
@@ -204,8 +199,8 @@ func ipv4CreateRejectTCPPacket(packet []byte, out []byte) []byte {
 }
 
 func ipv6CreateRejectPacket(packet []byte, out []byte) []byte {
-	proto, offset, isFragment, err := IPv6FindUpperProtocol(packet)
-	if err != nil || isFragment {
+	proto, offset, isFragment := ipv6FindUpperProtocol(packet)
+	if isFragment {
 		return nil
 	}
 	switch proto {
@@ -338,18 +333,7 @@ func ipv6CreateRejectTCPPacket(packet []byte, out []byte, offset int) []byte {
 	return out
 }
 
-// IPv6FindUpperProtocol walks the ipv6 extension header chain and returns the upper layer protocol, the
-// offset it begins at, and whether the packet is a non-first fragment. Only the RFC 8200 and IANA extension
-// headers below are walked. Everything else, including Mobility (135), HIP (139), Shim6 (140), experimental
-// 253/254, and real upper layer protocols like SCTP or GRE, is terminal. Walking those as extension headers
-// is a firewall bypass, so they fail closed. For a non-first fragment the returned protocol is the fragmented
-// protocol and offset points at the fragment header, there is no transport header to locate. Returns
-// ErrIPv6CouldNotFindPayload if packet is smaller than an ipv6 header or the chain is truncated before a
-// terminal protocol is reached.
-func IPv6FindUpperProtocol(packet []byte) (nextHeader uint8, offset int, isFragment bool, err error) {
-	if len(packet) < ipv6.HeaderLen {
-		return 0, 0, false, ErrIPv6CouldNotFindPayload
-	}
+func ipv6FindUpperProtocol(packet []byte) (nextHeader uint8, offset int, isFragment bool) {
 	nextHeader = packet[6]
 	offset = ipv6.HeaderLen
 
@@ -357,36 +341,30 @@ func IPv6FindUpperProtocol(packet []byte) (nextHeader uint8, offset int, isFragm
 		switch nextHeader {
 		case 0, 43, 60: // Hop-by-Hop, Routing, Destination
 			if len(packet) < offset+2 {
-				return nextHeader, offset, isFragment, ErrIPv6CouldNotFindPayload
+				return nextHeader, offset, isFragment
 			}
 			nextHeader = packet[offset]
 			offset += (int(packet[offset+1]) + 1) << 3
 
 		case 44: // Fragment
 			if len(packet) < offset+8 {
-				return nextHeader, offset, isFragment, ErrIPv6CouldNotFindPayload
+				return nextHeader, offset, isFragment
 			}
-			// Non-first fragments carry no transport header, report the fragmented protocol and stop
 			if packet[offset+2] != 0 || packet[offset+3]&0xf8 != 0 {
-				return packet[offset], offset, true, nil
+				isFragment = true
 			}
 			nextHeader = packet[offset]
 			offset += 8
 
 		case 51: // AH
 			if len(packet) < offset+2 {
-				return nextHeader, offset, isFragment, ErrIPv6CouldNotFindPayload
+				return nextHeader, offset, isFragment
 			}
 			nextHeader = packet[offset]
 			offset += (int(packet[offset+1]) + 2) << 2
 
 		default:
-			// A prior extension header can declare a length that advances offset past the packet. The terminal
-			// protocol's header isn't actually here, so treat the chain as truncated rather than classifying it.
-			if offset > len(packet) {
-				return nextHeader, offset, isFragment, ErrIPv6CouldNotFindPayload
-			}
-			return nextHeader, offset, isFragment, nil
+			return nextHeader, offset, isFragment
 		}
 	}
 }

iputil/packet_test.go

@@ -6,7 +6,6 @@ import (
 	"testing"
 
 	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
 	"golang.org/x/net/ipv4"
 	"golang.org/x/net/ipv6"
 )
@@ -475,61 +474,3 @@ func TestCreateICMPEchoResponse_IPv6_NotICMPv6(t *testing.T) {
 	result := CreateICMPEchoResponse(packet, out)
 	assert.Nil(t, result)
 }
-
-func Test_IPv6FindUpperProtocol(t *testing.T) {
-	src := net.ParseIP("fd00::1")
-	dst := net.ParseIP("fd00::2")
-
-	// 8 byte extension/transport stand-ins, first byte is the next header, second is the length field
-	extToTCP := []byte{6, 0, 0, 0, 0, 0, 0, 0}        // len 0 -> 8 bytes, next = TCP
-	extToUDP := []byte{17, 0, 0, 0, 0, 0, 0, 0}       // len 0 -> 8 bytes, next = UDP
-	extToRouting := []byte{43, 0, 0, 0, 0, 0, 0, 0}   // len 0 -> 8 bytes, next = Routing
-	ahToUDP := []byte{17, 0, 0, 0, 0, 0, 0, 0}        // AH len 0 -> (0+2)<<2 = 8 bytes, next = UDP
-	firstFragToUDP := []byte{17, 0, 0, 1, 0, 0, 0, 1} // frag offset 0, M=1, next = UDP
-	nonFirstFrag := []byte{17, 0, 0, 9, 0, 0, 0, 1}   // frag offset non-zero, next = UDP
-	transport := []byte{0, 80, 1, 187, 0, 0, 0, 0}    // stand-in bytes, IPv6FindUpperProtocol never reads ports
-
-	tests := []struct {
-		name         string
-		nextHeader   uint8
-		payload      []byte
-		wantProto    uint8
-		wantOffset   int
-		wantFragment bool
-		wantErr      error
-	}{
-		{"plain udp", 17, transport, 17, ipv6.HeaderLen, false, nil},
-		{"hop-by-hop then tcp", 0, append(extToTCP, transport...), 6, ipv6.HeaderLen + 8, false, nil},
-		{"routing then tcp", 43, append(extToTCP, transport...), 6, ipv6.HeaderLen + 8, false, nil},
-		{"destination then udp", 60, append(extToUDP, transport...), 17, ipv6.HeaderLen + 8, false, nil},
-		{"hop-by-hop, routing, then tcp", 0, append(append(extToRouting, extToTCP...), transport...), 6, ipv6.HeaderLen + 16, false, nil},
-		{"ah then udp", 51, append(ahToUDP, transport...), 17, ipv6.HeaderLen + 8, false, nil},
-		{"first fragment walks to transport", 44, append(firstFragToUDP, transport...), 17, ipv6.HeaderLen + 8, false, nil},
-		{"non-first fragment stops", 44, append(nonFirstFrag, transport...), 17, ipv6.HeaderLen, true, nil},
-		{"unknown protocol is terminal", 132, transport, 132, ipv6.HeaderLen, false, nil}, // SCTP
-		{"truncated extension header", 0, nil, 0, ipv6.HeaderLen, false, ErrIPv6CouldNotFindPayload},
-		// Destination Options with a declared length (255+1)*8 = 2048 that runs past the 48 byte buffer, next = SCTP
-		{"extension length past buffer", 60, []byte{132, 255, 0, 0, 0, 0, 0, 0}, 132, ipv6.HeaderLen + 2048, false, ErrIPv6CouldNotFindPayload},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			packet := makeIPv6Packet(src, dst, tt.nextHeader, tt.payload)
-			proto, offset, isFragment, err := IPv6FindUpperProtocol(packet)
-			if tt.wantErr != nil {
-				assert.ErrorIs(t, err, tt.wantErr)
-				return
-			}
-			require.NoError(t, err)
-			assert.Equal(t, tt.wantProto, proto)
-			assert.Equal(t, tt.wantOffset, offset)
-			assert.Equal(t, tt.wantFragment, isFragment)
-		})
-	}
-
-	// A packet smaller than an ipv6 header must error rather than panic reading byte 6
-	t.Run("shorter than ipv6 header", func(t *testing.T) {
-		_, _, _, err := IPv6FindUpperProtocol(make([]byte, 6))
-		assert.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
-	})
-}

main.go

@@ -11,7 +11,6 @@ import (
 	"time"
 
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/noiseutil"
 	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/sshd"
 	"github.com/slackhq/nebula/udp"
@@ -21,12 +20,6 @@ import (
 
 type m = map[string]any
 
-// maxRoutines caps routines below the RejectHeadroom nonce gap so concurrent senders can't race the counter past wrap.
-const maxRoutines = 1 << 16
-
-// The reject headroom must exceed every sender that can be mid-reservation at once, about two per routine.
-const _ = noiseutil.RejectHeadroom - 4*maxRoutines
-
 func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, deviceFactory overlay.DeviceFactory) (retcon *Control, reterr error) {
 	ctx, cancel := context.WithCancel(context.Background())
 	// Automatically cancel the context if Main returns an error, to signal all created goroutines to quit.
@@ -88,6 +81,9 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		if routines < 1 {
 			routines = 1
 		}
+		if routines > 1 {
+			l.Info("Using multiple routines", "routines", routines)
+		}
 	} else {
 		// deprecated and undocumented
 		tunQueues := c.GetInt("tun.routines", 1)
@@ -97,12 +93,6 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 			l.Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead", "routines", routines)
 		}
 	}
-	if routines > maxRoutines {
-		l.Warn("Using multiple routines", "routines", maxRoutines, "clamped", true, "requestedRoutines", routines)
-		routines = maxRoutines
-	} else if routines > 1 {
-		l.Info("Using multiple routines", "routines", routines)
-	}
 
 	// EXPERIMENTAL
 	// Intentionally not documented yet while we do more testing and determine
@@ -254,6 +244,39 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		ifce.writers = udpConns
 		lightHouse.ifce = ifce
 
+		loadMultiPortConfig := func(c *config.C) {
+			ifce.multiPort.Rx = c.GetBool("tun.multiport.rx_enabled", false)
+
+			tx := c.GetBool("tun.multiport.tx_enabled", false)
+
+			if tx && ifce.udpRaw == nil {
+				ifce.udpRaw, err = udp.NewRawConn(l, c.GetString("listen.host", "0.0.0.0"), port, uint16(port))
+				if err != nil {
+					l.Error("Failed to get raw socket for tun.multiport.tx_enabled", "error", err)
+					ifce.udpRaw = nil
+					tx = false
+				}
+			}
+
+			if tx {
+				ifce.multiPort.TxBasePort = uint16(port)
+				ifce.multiPort.TxPorts = c.GetInt("tun.multiport.tx_ports", 100)
+				ifce.multiPort.TxHandshake = c.GetBool("tun.multiport.tx_handshake", false)
+				ifce.multiPort.TxHandshakeDelay = int64(c.GetInt("tun.multiport.tx_handshake_delay", 2))
+				ifce.udpRaw.ReloadConfig(c)
+			}
+			ifce.multiPort.Tx = tx
+
+			// TODO: if we upstream this, make this cleaner
+			handshakeManager.udpRaw = ifce.udpRaw
+			handshakeManager.multiPort = ifce.multiPort
+
+			l.Info("Multiport configured", "multiPort", ifce.multiPort)
+		}
+
+		loadMultiPortConfig(c)
+		c.RegisterReloadCallback(loadMultiPortConfig)
+
 		ifce.RegisterConfigChangeCallbacks(c)
 		ifce.reloadDisconnectInvalid(c)
 		ifce.reloadSendRecvError(c)

message_metrics.go

@@ -14,8 +14,7 @@ type MessageMetrics struct {
 	rxUnknown metrics.Counter
 	txUnknown metrics.Counter
 
-	rxInvalid   metrics.Counter
-	txExhausted metrics.Counter
+	rxInvalid metrics.Counter
 }
 
 func (m *MessageMetrics) Rx(t header.MessageType, s header.MessageSubType, i int64) {
@@ -42,13 +41,6 @@ func (m *MessageMetrics) RxInvalid(i int64) {
 	}
 }
 
-// TxExhausted counts outbound packets dropped because the tunnel's message counter is spent.
-func (m *MessageMetrics) TxExhausted(i int64) {
-	if m != nil && m.txExhausted != nil {
-		m.txExhausted.Inc(i)
-	}
-}
-
 func newMessageMetrics() *MessageMetrics {
 	gen := func(t string) [][]metrics.Counter {
 		return [][]metrics.Counter{
@@ -69,10 +61,9 @@ func newMessageMetrics() *MessageMetrics {
 		rx: gen("rx"),
 		tx: gen("tx"),
 
-		rxUnknown:   metrics.GetOrRegisterCounter("messages.rx.other", nil),
-		txUnknown:   metrics.GetOrRegisterCounter("messages.tx.other", nil),
-		rxInvalid:   metrics.GetOrRegisterCounter("messages.rx.invalid", nil),
-		txExhausted: metrics.GetOrRegisterCounter("messages.tx.exhausted", nil),
+		rxUnknown: metrics.GetOrRegisterCounter("messages.rx.other", nil),
+		txUnknown: metrics.GetOrRegisterCounter("messages.tx.other", nil),
+		rxInvalid: metrics.GetOrRegisterCounter("messages.rx.invalid", nil),
 	}
 }
 

noiseutil/aesgcm.go

@@ -25,9 +25,6 @@ func (s *CipherStateAESGCM) EncryptDanger(out, ad, plaintext []byte, n uint64, n
 	if s == nil {
 		return nil, errors.New("no cipher state available to encrypt")
 	}
-	if n >= RejectAfterMessages {
-		return nil, ErrMessageCounterExhausted
-	}
 	nb[0] = 0
 	nb[1] = 0
 	nb[2] = 0

noiseutil/chachapoly.go

@@ -24,9 +24,6 @@ func (s *CipherStateChaChaPoly) EncryptDanger(out, ad, plaintext []byte, n uint6
 	if s == nil {
 		return nil, errors.New("no cipher state available to encrypt")
 	}
-	if n >= RejectAfterMessages {
-		return nil, ErrMessageCounterExhausted
-	}
 	nb[0] = 0
 	nb[1] = 0
 	nb[2] = 0

noiseutil/cipher_state.go

@@ -1,22 +1,11 @@
 package noiseutil
 
 import (
-	"errors"
 	"fmt"
-	"math"
 
 	"github.com/flynn/noise"
 )
 
-// RejectHeadroom is the wrap gap for senders racing the counter, sized large enough for any routine count.
-const RejectHeadroom = uint64(1) << 40
-
-// RejectAfterMessages is the nonce ceiling: encrypting stops RejectHeadroom short of the wrap.
-const RejectAfterMessages = math.MaxUint64 - RejectHeadroom
-
-// ErrMessageCounterExhausted is returned by EncryptDanger once the nonce reaches RejectAfterMessages.
-var ErrMessageCounterExhausted = errors.New("message counter exhausted")
-
 // CipherState is the post-handshake AEAD cipher used for the data plane.
 // Each supported cipher has its own concrete implementation in this package with the nonce endianness hardcoded,
 // so the encrypt/decrypt fast path avoids interface dispatch on the byte order.

noiseutil/cipher_state_test.go

@@ -1,7 +1,6 @@
 package noiseutil
 
 import (
-	"math"
 	"testing"
 
 	"github.com/flynn/noise"
@@ -90,24 +89,6 @@ func roundtrip(t *testing.T, enc, dec CipherState) {
 	assert.Equal(t, 16, enc.Overhead())
 }
 
-func TestEncryptRejectsExhaustedCounter(t *testing.T) {
-	// Pin the headroom below the uint64 wrap so a typo can't silently move the ceiling.
-	require.Equal(t, uint64(1)<<40, RejectHeadroom)
-	require.Equal(t, math.MaxUint64-RejectHeadroom, RejectAfterMessages)
-
-	encA, _ := buildCipherStates(t, CipherAESGCM)
-	encC, _ := buildCipherStates(t, noise.CipherChaChaPoly)
-	nb := make([]byte, 12)
-
-	for _, cs := range []CipherState{NewCipherStateAESGCM(encA), NewCipherStateChaChaPoly(encC)} {
-		_, err := cs.EncryptDanger(nil, nil, []byte("x"), RejectAfterMessages-1, nb)
-		require.NoError(t, err)
-
-		_, err = cs.EncryptDanger(nil, nil, []byte("x"), RejectAfterMessages, nb)
-		require.ErrorIs(t, err, ErrMessageCounterExhausted)
-	}
-}
-
 func BenchmarkCipherStateEncryptAESGCM(b *testing.B) {
 	enc, _ := buildCipherStatesB(b, CipherAESGCM)
 	benchEncryptCipherState(b, NewCipherState(enc, CipherAESGCM))

outside.go

@@ -13,7 +13,6 @@ import (
 
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/iputil"
 	"golang.org/x/net/ipv4"
 )
 
@@ -265,6 +264,15 @@ func (f *Interface) sendCloseTunnel(h *HostInfo) {
 func (f *Interface) handleHostRoaming(hostinfo *HostInfo, via ViaSender) {
 	curRemote := hostinfo.GetRemote()
 	if !via.IsRelayed && curRemote != via.UdpAddr {
+		if hostinfo.multiportRx {
+			// If the remote is sending with multiport, we aren't roaming unless
+			// the IP has changed
+			if curRemote.Addr().Compare(via.UdpAddr.Addr()) == 0 {
+				return
+			}
+			// Keep the port from the original hostinfo, because the remote is transmitting from multiport ports
+			via.UdpAddr = netip.AddrPortFrom(via.UdpAddr.Addr(), curRemote.Port())
+		}
 		if !f.lightHouse.GetRemoteAllowList().AllowAll(hostinfo.vpnAddrs, via.UdpAddr.Addr()) {
 			if f.l.Enabled(context.Background(), slog.LevelDebug) {
 				hostinfo.logger(f.l).Debug("lighthouse.remote_allow_list denied roaming", "newAddr", via.UdpAddr)
@@ -300,6 +308,7 @@ var (
 	ErrIPv4InvalidHeaderLength = errors.New("invalid ipv4 header length")
 	ErrIPv4PacketTooShort      = errors.New("ipv4 packet is too short")
 	ErrIPv6PacketTooShort      = errors.New("ipv6 packet is too short")
+	ErrIPv6CouldNotFindPayload = errors.New("could not find payload in ipv6 packet")
 )
 
 // newPacket validates and parses the interesting bits for the firewall out of the ip and sub protocol headers
@@ -332,59 +341,101 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 		fp.RemoteAddr, _ = netip.AddrFromSlice(data[24:40])
 	}
 
-	// Walk the extension header chain to the upper layer protocol. iputil.IPv6FindUpperProtocol is the single
-	// source of truth for which headers are extension headers, so this stays in lockstep with the reject path
-	// and cannot drift into misreading an unknown protocol (SCTP, GRE, etc.) as a forged transport.
-	proto, offset, isFragment, err := iputil.IPv6FindUpperProtocol(data)
-	if err != nil {
-		return ErrIPv6PacketTooShort
-	}
-
-	fp.Protocol = proto
-	fp.Fragment = isFragment
-	if isFragment {
-		// Non-first fragments carry no transport header, so we have no ports to read
-		fp.RemotePort = 0
-		fp.LocalPort = 0
-		return nil
-	}
-
-	switch layers.IPProtocol(proto) {
-	case layers.IPProtocolICMPv6:
-		// An ICMPv6 message is at least type, code and checksum, 4 bytes. Only echo carries more than we read.
-		if dataLen < offset+4 {
-			return ErrIPv6PacketTooShort
+	protoAt := 6             // NextHeader is at 6 bytes into the ipv6 header
+	offset := ipv6.HeaderLen // Start at the end of the ipv6 header
+	next := 0
+	for {
+		if protoAt >= dataLen {
+			break
 		}
-		fp.LocalPort = 0      //incoming vs outgoing doesn't matter for icmpv6
-		switch data[offset] { //icmp type
-		case layers.ICMPv6TypeEchoRequest, layers.ICMPv6TypeEchoReply:
+		proto := layers.IPProtocol(data[protoAt])
+
+		switch proto {
+		case layers.IPProtocolESP, layers.IPProtocolNoNextHeader:
+			fp.Protocol = uint8(proto)
+			fp.RemotePort = 0
+			fp.LocalPort = 0
+			fp.Fragment = false
+			return nil
+
+		case layers.IPProtocolICMPv6:
 			if dataLen < offset+6 {
 				return ErrIPv6PacketTooShort
 			}
-			fp.RemotePort = binary.BigEndian.Uint16(data[offset+4 : offset+6]) //identifier
+			fp.Protocol = uint8(proto)
+			fp.LocalPort = 0 //incoming vs outgoing doesn't matter for icmpv6
+			icmptype := data[offset+1]
+			switch icmptype {
+			case layers.ICMPv6TypeEchoRequest, layers.ICMPv6TypeEchoReply:
+				fp.RemotePort = binary.BigEndian.Uint16(data[offset+4 : offset+6]) //identifier
+			default:
+				fp.RemotePort = 0
+			}
+			fp.Fragment = false
+			return nil
+
+		case layers.IPProtocolTCP, layers.IPProtocolUDP:
+			if dataLen < offset+4 {
+				return ErrIPv6PacketTooShort
+			}
+
+			fp.Protocol = uint8(proto)
+			if incoming {
+				fp.RemotePort = binary.BigEndian.Uint16(data[offset : offset+2])
+				fp.LocalPort = binary.BigEndian.Uint16(data[offset+2 : offset+4])
+			} else {
+				fp.LocalPort = binary.BigEndian.Uint16(data[offset : offset+2])
+				fp.RemotePort = binary.BigEndian.Uint16(data[offset+2 : offset+4])
+			}
+
+			fp.Fragment = false
+			return nil
+
+		case layers.IPProtocolIPv6Fragment:
+			// Fragment header is 8 bytes, need at least offset+4 to read the offset field
+			if dataLen < offset+8 {
+				return ErrIPv6PacketTooShort
+			}
+
+			// Check if this is the first fragment
+			fragmentOffset := binary.BigEndian.Uint16(data[offset+2:offset+4]) &^ uint16(0x7) // Remove the reserved and M flag bits
+			if fragmentOffset != 0 {
+				// Non-first fragment, use what we have now and stop processing
+				fp.Protocol = data[offset]
+				fp.Fragment = true
+				fp.RemotePort = 0
+				fp.LocalPort = 0
+				return nil
+			}
+
+			// The next loop should be the transport layer since we are the first fragment
+			next = 8 // Fragment headers are always 8 bytes
+
+		case layers.IPProtocolAH:
+			// Auth headers, used by IPSec, have a different meaning for header length
+			if dataLen <= offset+1 {
+				break
+			}
+			next = (int(data[offset+1]) + 2) << 2
+
 		default:
-			fp.RemotePort = 0
+			// Normal ipv6 header length processing
+			if dataLen <= offset+1 {
+				break
+			}
+			next = (int(data[offset+1]) + 1) << 3
 		}
 
-	case layers.IPProtocolTCP, layers.IPProtocolUDP:
-		if dataLen < offset+4 {
-			return ErrIPv6PacketTooShort
-		}
-		if incoming {
-			fp.RemotePort = binary.BigEndian.Uint16(data[offset : offset+2])
-			fp.LocalPort = binary.BigEndian.Uint16(data[offset+2 : offset+4])
-		} else {
-			fp.LocalPort = binary.BigEndian.Uint16(data[offset : offset+2])
-			fp.RemotePort = binary.BigEndian.Uint16(data[offset+2 : offset+4])
+		if next <= 0 {
+			// Safety check, each ipv6 header has to be at least 8 bytes
+			next = 8
 		}
 
-	default:
-		// don't set ports for protocols Nebula doesn't inspect
-		fp.RemotePort = 0
-		fp.LocalPort = 0
+		protoAt = offset
+		offset = offset + next
 	}
 
-	return nil
+	return ErrIPv6CouldNotFindPayload
 }
 
 func parseV4(data []byte, incoming bool, fp *firewall.Packet) error {

outside_test.go

@@ -14,7 +14,6 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"golang.org/x/net/ipv4"
-	"golang.org/x/net/ipv6"
 )
 
 func Test_newPacket(t *testing.T) {
@@ -116,12 +115,12 @@ func Test_newPacket_v6(t *testing.T) {
 	require.NoError(t, err)
 
 	err = newPacket(buffer.Bytes(), true, p)
-	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
 	// A v6 packet with a hop-by-hop extension
 	// ICMPv6 Payload (Echo Request)
 	icmpLayer := layers.ICMPv6{
-		TypeCode: layers.CreateICMPv6TypeCode(layers.ICMPv6TypeEchoRequest, 0),
+		TypeCode: layers.ICMPv6TypeEchoRequest,
 	}
 	// Hop-by-Hop Extension Header
 	hopOption := layers.IPv6HopByHopOption{}
@@ -150,12 +149,12 @@ func Test_newPacket_v6(t *testing.T) {
 	// A full IPv6 header and 1 byte in the first extension, but missing
 	// the length byte.
 	err = newPacket(buffer.Bytes()[:41], true, p)
-	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
 	// A full IPv6 header plus 1 full extension, but only 1 byte of the
 	// next layer, missing length byte
 	err = newPacket(buffer.Bytes()[:49], true, p)
-	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 	err = nil
 
 	// A good ICMP packet
@@ -168,7 +167,7 @@ func Test_newPacket_v6(t *testing.T) {
 	}
 
 	icmp := layers.ICMPv6{
-		TypeCode: layers.CreateICMPv6TypeCode(layers.ICMPv6TypeEchoRequest, 0),
+		TypeCode: layers.ICMPv6TypeEchoRequest,
 		Checksum: 0x1234,
 	}
 
@@ -190,18 +189,6 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.Equal(t, uint16(0), p.LocalPort)
 	assert.False(t, p.Fragment)
 
-	// A minimal 4 byte non-echo ICMPv6 message (type, code, checksum), no identifier to read
-	icmpMin := make([]byte, ipv6.HeaderLen+4)
-	copy(icmpMin, buffer.Bytes()[:ipv6.HeaderLen])
-	icmpMin[6] = byte(layers.IPProtocolICMPv6)
-	icmpMin[ipv6.HeaderLen] = 1 // type 1, destination unreachable, not echo
-	err = newPacket(icmpMin, true, p)
-	require.NoError(t, err)
-	assert.Equal(t, uint8(layers.IPProtocolICMPv6), p.Protocol)
-	assert.Equal(t, uint16(0), p.RemotePort)
-	assert.Equal(t, uint16(0), p.LocalPort)
-	assert.False(t, p.Fragment)
-
 	// A good ESP packet
 	b := buffer.Bytes()
 	b[6] = byte(layers.IPProtocolESP)
@@ -226,15 +213,11 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.Equal(t, uint16(0), p.LocalPort)
 	assert.False(t, p.Fragment)
 
-	// An unknown protocol packet, we don't dissect it so we fail closed on its true protocol with no ports
+	// An unknown protocol packet
 	b = buffer.Bytes()
 	b[6] = 255 // 255 is a reserved protocol number
 	err = newPacket(b, true, p)
-	require.NoError(t, err)
-	assert.Equal(t, uint8(255), p.Protocol)
-	assert.Equal(t, uint16(0), p.RemotePort)
-	assert.Equal(t, uint16(0), p.LocalPort)
-	assert.False(t, p.Fragment)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
 	// A good UDP packet
 	ip = layers.IPv6{
@@ -351,14 +334,14 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.Equal(t, uint16(22), p.LocalPort)
 	assert.False(t, p.Fragment)
 
-	// Ensure buffer bounds checking during processing, a truncated AH header can't reach the payload
+	// Ensure buffer bounds checking during processing
 	err = newPacket(b[:41], true, p)
 	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
 
 	// Invalid AH header
 	b = buffer.Bytes()
 	err = newPacket(b, true, p)
-	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 }
 
 func Test_newPacket_ipv6Fragment(t *testing.T) {
@@ -692,66 +675,3 @@ func Test_newPacket_v6ExtHeaderOverflow(t *testing.T) {
 	// the host delivers to, not the forged 443 at the overflowed offset.
 	assert.Equal(t, uint16(22), p.LocalPort, "firewall must parse the real transport header, not the overflowed offset")
 }
-
-// Test_newPacket_v6ExtHeaderPastBuffer is a regression test for an extension header whose declared length
-// advances the walk past the end of the packet. The upper layer protocol's header isn't actually present,
-// so parseV6 must drop the packet rather than classify it as the terminal protocol with no ports.
-func Test_newPacket_v6ExtHeaderPastBuffer(t *testing.T) {
-	p := &firewall.Packet{}
-
-	pkt := make([]byte, 48)
-	pkt[0] = 0x60
-	pkt[6] = byte(layers.IPProtocolIPv6Destination) // Destination Options
-	pkt[7] = 64                                     // hop limit
-	pkt[40] = byte(layers.IPProtocolSCTP)           // Dest Options next header = SCTP
-	pkt[41] = 255                                   // declared length (255+1)*8 = 2048, past the 48 byte buffer
-
-	require.ErrorIs(t, newPacket(pkt, true, p), ErrIPv6PacketTooShort)
-}
-
-// Test_newPacket_v6ExtHeaderConfusion is a regression test for parseV6 walking any unrecognized
-// Next Header as if it were an ipv6 extension header. A real upper layer protocol Nebula doesn't
-// dissect (SCTP here) is not walkable, so applying the (len+1)*8 formula marched into the SCTP
-// payload and landed on a byte that looked like UDP, forging a protocol/port pair the firewall
-// would trust while the host delivered the real SCTP datagram. The fix fails closed: the packet
-// is classified as its true protocol with no ports, so it only matches an `any` rule.
-func Test_newPacket_v6ExtHeaderConfusion(t *testing.T) {
-	p := &firewall.Packet{}
-
-	pkt := make([]byte, 52)
-	pkt[0] = 0x60                        // version 6
-	pkt[6] = byte(layers.IPProtocolSCTP) // NextHeader = SCTP, a real protocol, not an extension header
-	pkt[7] = 64                          // hop limit
-
-	// Real SCTP header at offset 40. Pre-fix parseV6 walked SCTP as an extension header: byte 41 (0x00, the
-	// low byte of the src port below) was read as the header length, giving next=(0+1)*8=8, which landed the
-	// walk on byte 40 (0x11), misread as NextHeader=UDP, then bytes 48-51 as ports.
-	binary.BigEndian.PutUint16(pkt[40:42], 0x1100) // SCTP src port; byte 40=0x11, byte 41=0x00
-	binary.BigEndian.PutUint16(pkt[42:44], 445)    // SCTP dst port, never read by parseV6
-	binary.BigEndian.PutUint16(pkt[48:50], 53)     // SCTP checksum bytes, pre-fix forged RemotePort
-	binary.BigEndian.PutUint16(pkt[50:52], 53)     // pre-fix forged LocalPort
-
-	require.NoError(t, newPacket(pkt, true, p))
-	assert.Equal(t, uint8(layers.IPProtocolSCTP), p.Protocol, "must classify as the true protocol, not the forged UDP")
-	assert.Equal(t, uint16(0), p.RemotePort)
-	assert.Equal(t, uint16(0), p.LocalPort)
-	assert.False(t, p.Fragment)
-
-	// Same confusion, but the unknown protocol sits after a real extension header. The HopByHop is walked
-	// correctly, then SCTP must still fail closed instead of being walked into its own payload. Protocol is
-	// the only assertion that discriminates the fix here, a regression that walked SCTP would misclassify it.
-	chained := make([]byte, 60)
-	chained[0] = 0x60                                  // version 6
-	chained[6] = byte(layers.IPProtocolIPv6HopByHop)   // NextHeader = HopByHop extension
-	chained[7] = 64                                    // hop limit
-	chained[40] = byte(layers.IPProtocolSCTP)          // HopByHop NextHeader = SCTP
-	chained[41] = 0                                    // HopByHop length 0 -> 8 bytes, SCTP begins at offset 48
-	binary.BigEndian.PutUint16(chained[48:50], 0x1100) // SCTP src port, pre-fix forged NextHeader/length bait
-	binary.BigEndian.PutUint16(chained[50:52], 445)    // SCTP dst port, never read by parseV6
-
-	require.NoError(t, newPacket(chained, true, p))
-	assert.Equal(t, uint8(layers.IPProtocolSCTP), p.Protocol, "must fail closed on the unknown protocol after the extension header")
-	assert.Equal(t, uint16(0), p.RemotePort)
-	assert.Equal(t, uint16(0), p.LocalPort)
-	assert.False(t, p.Fragment)
-}

overlay/tun_linux.go

@@ -5,7 +5,6 @@ package overlay
 
 import (
 	"encoding/binary"
-	"errors"
 	"fmt"
 	"io"
 	"log/slog"
@@ -484,16 +483,7 @@ func (t *tun) addIPs(link netlink.Link) error {
 	//iterate over remainder, remove whoever shouldn't be there
 	al, err := netlink.AddrList(link, netlink.FAMILY_ALL)
 	if err != nil {
-		//RTM_GETADDR dumps the whole system, so any concurrent address change
-		//interrupts it - including the kernel's async tentative->preferred
-		//flip of an IPv6 address the AddrReplace calls above just added,
-		//which makes this a race against our own setup. Partial results are
-		//still returned; the worst case is a stale address surviving until
-		//the next config reload, which beats failing startup over it.
-		if !errors.Is(err, netlink.ErrDumpInterrupted) {
-			return fmt.Errorf("failed to get tun address list: %s", err)
-		}
-		t.l.Warn("tun address list dump was interrupted, stale addresses may remain")
+		return fmt.Errorf("failed to get tun address list: %s", err)
 	}
 
 	for i := range al {

stats.go

@@ -13,6 +13,7 @@ import (
 	"sync/atomic"
 	"time"
 
+	graphite "github.com/cyberdelia/go-metrics-graphite"
 	mp "github.com/nbrownus/go-metrics-prometheus"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
@@ -252,7 +253,7 @@ func (s *statsServer) buildRuntime(cfg statsConfig) ([]func(), *http.Server) {
 		// loadStatsConfig already resolved and validated the address; re-parse
 		// the resolved form (no DNS lookup) to get a *net.TCPAddr.
 		addr, _ := net.ResolveTCPAddr(cfg.graphite.protocol, cfg.graphite.resolvedAddr)
-		gcfg := graphiteConfigExport{
+		gcfg := graphite.Config{
 			Addr:          addr,
 			Registry:      metrics.DefaultRegistry,
 			FlushInterval: cfg.interval,
@@ -261,7 +262,7 @@ func (s *statsServer) buildRuntime(cfg statsConfig) ([]func(), *http.Server) {
 			Percentiles:   []float64{0.5, 0.75, 0.95, 0.99, 0.999},
 		}
 		captureFns = append(captureFns, func() {
-			if err := graphiteOnce(gcfg); err != nil {
+			if err := graphite.Once(gcfg); err != nil {
 				s.l.Error("Graphite export failed", "error", err)
 			}
 		})

stats_test.go

@@ -371,7 +371,7 @@ func waitForListening(t *testing.T, addr string) {
 	})
 }
 
-// graphiteSink is a minimal TCP accept-and-discard server so graphiteOnce
+// graphiteSink is a minimal TCP accept-and-discard server so graphite.Once
 // calls in tests don't spam error logs or wedge on connection refused.
 type graphiteSink struct {
 	ln net.Listener

udp/udp_raw.go

@@ -0,0 +1,16 @@
+package udp
+
+import mathrand "math/rand"
+
+type SendPortGetter interface {
+	// UDPSendPort returns the port to use
+	UDPSendPort(maxPort int) uint16
+}
+
+type randomSendPort struct{}
+
+func (randomSendPort) UDPSendPort(maxPort int) uint16 {
+	return uint16(mathrand.Intn(maxPort))
+}
+
+var RandomSendPort = randomSendPort{}

udp/udp_raw_linux.go

@@ -0,0 +1,191 @@
+//go:build !android && !e2e_testing
+// +build !android,!e2e_testing
+
+package udp
+
+import (
+	"encoding/binary"
+	"fmt"
+	"log/slog"
+	"net"
+	"net/netip"
+	"syscall"
+	"unsafe"
+
+	"github.com/rcrowley/go-metrics"
+	"github.com/slackhq/nebula/config"
+	"golang.org/x/net/ipv4"
+	"golang.org/x/sys/unix"
+)
+
+// RawOverhead is the number of bytes that need to be reserved at the start of
+// the raw bytes passed to (*RawConn).WriteTo. This is used by WriteTo to prefix
+// the IP and UDP headers.
+const RawOverhead = 28
+
+type RawConn struct {
+	sysFd    int
+	basePort uint16
+	l        *slog.Logger
+}
+
+func NewRawConn(l *slog.Logger, ip string, port int, basePort uint16) (*RawConn, error) {
+	syscall.ForkLock.RLock()
+	// With IPPROTO_UDP, the linux kernel tries to deliver every UDP packet
+	// received in the system to our socket. This constantly overflows our
+	// buffer and marks our socket as having dropped packets. This makes the
+	// stats on the socket useless.
+	//
+	// In contrast, IPPROTO_RAW is not delivered any packets and thus our read
+	// buffer will not fill up and mark as having dropped packets. The only
+	// difference is that we have to assemble the IP header as well, but this
+	// is fairly easy since Linux does the checksum for us.
+	//
+	// TODO: How to get this working with Inet6 correctly? I was having issues
+	// with the source address when testing before, probably need to `bind(2)`?
+	fd, err := unix.Socket(unix.AF_INET, unix.SOCK_RAW, unix.IPPROTO_RAW)
+	if err == nil {
+		unix.CloseOnExec(fd)
+	}
+	syscall.ForkLock.RUnlock()
+	if err != nil {
+		return nil, err
+	}
+
+	// We only want to send, not recv. This will hopefully help the kernel avoid
+	// wasting time on us
+	if err = unix.SetsockoptInt(fd, unix.SOL_SOCKET, unix.SO_RCVBUF, 0); err != nil {
+		return nil, fmt.Errorf("unable to set SO_RCVBUF: %s", err)
+	}
+
+	var lip [16]byte
+	copy(lip[:], net.ParseIP(ip))
+
+	// TODO do we need to `bind(2)` so that we send from the correct address/interface?
+	if err = unix.Bind(fd, &unix.SockaddrInet6{Addr: lip, Port: port}); err != nil {
+		return nil, fmt.Errorf("unable to bind to socket: %s", err)
+	}
+
+	return &RawConn{
+		sysFd:    fd,
+		basePort: basePort,
+		l:        l,
+	}, nil
+}
+
+// WriteTo must be called with raw leaving the first `udp.RawOverhead` bytes empty,
+// for the IP/UDP headers.
+func (u *RawConn) WriteTo(raw []byte, fromPort uint16, ip netip.AddrPort) error {
+	var rsa unix.RawSockaddrInet4
+	rsa.Family = unix.AF_INET
+	rsa.Addr = ip.Addr().As4()
+
+	totalLen := len(raw)
+	udpLen := totalLen - ipv4.HeaderLen
+
+	// IP header
+	raw[0] = byte(ipv4.Version<<4 | (ipv4.HeaderLen >> 2 & 0x0f))
+	raw[1] = 0 // tos
+	binary.BigEndian.PutUint16(raw[2:4], uint16(totalLen))
+	binary.BigEndian.PutUint16(raw[4:6], 0)   // id (linux does it for us)
+	binary.BigEndian.PutUint16(raw[6:8], 0)   // frag options
+	raw[8] = byte(64)                         // ttl
+	raw[9] = byte(17)                         // protocol
+	binary.BigEndian.PutUint16(raw[10:12], 0) // checksum (linux does it for us)
+	binary.BigEndian.PutUint32(raw[12:16], 0) // src (linux does it for us)
+	copy(raw[16:20], rsa.Addr[:])             // dst
+
+	// UDP header
+	fromPort = u.basePort + fromPort
+	binary.BigEndian.PutUint16(raw[20:22], uint16(fromPort))  // src port
+	binary.BigEndian.PutUint16(raw[22:24], uint16(ip.Port())) // dst port
+	binary.BigEndian.PutUint16(raw[24:26], uint16(udpLen))    // UDP length
+	binary.BigEndian.PutUint16(raw[26:28], 0)                 // checksum (optional)
+
+	for {
+		_, _, err := unix.Syscall6(
+			unix.SYS_SENDTO,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&raw[0])),
+			uintptr(len(raw)),
+			uintptr(0),
+			uintptr(unsafe.Pointer(&rsa)),
+			uintptr(unix.SizeofSockaddrInet4),
+		)
+
+		if err != 0 {
+			return &net.OpError{Op: "sendto", Err: err}
+		}
+
+		//TODO: handle incomplete writes
+
+		return nil
+	}
+}
+
+func (u *RawConn) ReloadConfig(c *config.C) {
+	b := c.GetInt("listen.write_buffer", 0)
+	if b <= 0 {
+		return
+	}
+
+	if err := u.SetSendBuffer(b); err != nil {
+		u.l.Error("Failed to set listen.write_buffer", "error", err)
+		return
+	}
+
+	s, err := u.GetSendBuffer()
+	if err != nil {
+		u.l.Warn("Failed to get listen.write_buffer", "error", err)
+		return
+	}
+
+	u.l.Info("listen.write_buffer was set", "size", s)
+}
+
+func (u *RawConn) SetSendBuffer(n int) error {
+	return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_SNDBUFFORCE, n)
+}
+
+func (u *RawConn) GetSendBuffer() (int, error) {
+	return unix.GetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_SNDBUF)
+}
+
+func (u *RawConn) getMemInfo(meminfo *[unix.SK_MEMINFO_VARS]uint32) error {
+	var vallen uint32 = 4 * unix.SK_MEMINFO_VARS
+	_, _, err := unix.Syscall6(unix.SYS_GETSOCKOPT, uintptr(u.sysFd), uintptr(unix.SOL_SOCKET), uintptr(unix.SO_MEMINFO), uintptr(unsafe.Pointer(meminfo)), uintptr(unsafe.Pointer(&vallen)), 0)
+	if err != 0 {
+		return err
+	}
+	return nil
+}
+
+func NewRawStatsEmitter(rawConn *RawConn) func() {
+	// Check if our kernel supports SO_MEMINFO before registering the gauges
+	var gauges [unix.SK_MEMINFO_VARS]metrics.Gauge
+	var meminfo [unix.SK_MEMINFO_VARS]uint32
+	if err := rawConn.getMemInfo(&meminfo); err == nil {
+		gauges = [unix.SK_MEMINFO_VARS]metrics.Gauge{
+			metrics.GetOrRegisterGauge("raw.rmem_alloc", nil),
+			metrics.GetOrRegisterGauge("raw.rcvbuf", nil),
+			metrics.GetOrRegisterGauge("raw.wmem_alloc", nil),
+			metrics.GetOrRegisterGauge("raw.sndbuf", nil),
+			metrics.GetOrRegisterGauge("raw.fwd_alloc", nil),
+			metrics.GetOrRegisterGauge("raw.wmem_queued", nil),
+			metrics.GetOrRegisterGauge("raw.optmem", nil),
+			metrics.GetOrRegisterGauge("raw.backlog", nil),
+			metrics.GetOrRegisterGauge("raw.drops", nil),
+		}
+	} else {
+		// return no-op because we don't support SO_MEMINFO
+		return func() {}
+	}
+
+	return func() {
+		if err := rawConn.getMemInfo(&meminfo); err == nil {
+			for j := 0; j < unix.SK_MEMINFO_VARS; j++ {
+				gauges[j].Update(int64(meminfo[j]))
+			}
+		}
+	}
+}

udp/udp_raw_unsupported.go

@@ -0,0 +1,29 @@
+//go:build !linux || android || e2e_testing
+// +build !linux android e2e_testing
+
+package udp
+
+import (
+	"fmt"
+	"log/slog"
+	"net/netip"
+	"runtime"
+
+	"github.com/slackhq/nebula/config"
+)
+
+const RawOverhead = 0
+
+type RawConn struct{}
+
+func NewRawConn(l *slog.Logger, ip string, port int, basePort uint16) (*RawConn, error) {
+	return nil, fmt.Errorf("multiport tx is not supported on %s", runtime.GOOS)
+}
+
+func (u *RawConn) WriteTo(raw []byte, fromPort uint16, addr netip.AddrPort) error {
+	return fmt.Errorf("multiport tx is not supported on %s", runtime.GOOS)
+}
+
+func (u *RawConn) ReloadConfig(c *config.C) {}
+
+func NewRawStatsEmitter(rawConn *RawConn) func() { return func() {} }