Address PR feedback: remove outbound rate limit, improve config docs

Remove rate limiting from StartHandshake (outbound) since DoS
protection only needs to limit inbound handshakes. This also avoids
returning nil from StartHandshake which historically always returned
non-nil. Update config comment to note openssl speed is single-core
and suggest scaling by routines.

Co-Authored-By: Claude <svc-devxp-claude@slack-corp.com>

.github/workflows/release.yml

@@ -209,11 +209,10 @@ jobs:
         id: create_release
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          GITHUB_REF_NAME: ${{ github.ref_name }}
         run: |
           cd artifacts
           gh release create \
             --verify-tag \
-            --title "Release ${GITHUB_REF_NAME}" \
-            "${GITHUB_REF_NAME}" \
+            --title "Release ${{ github.ref_name }}" \
+            "${{ github.ref_name }}" \
             SHASUM256.txt *-latest/*.zip *-latest/*.tar.gz

.github/workflows/smoke/smoke-vagrant.sh

@@ -29,17 +29,6 @@ docker run --name lighthouse1 --rm "$CONTAINER" -config lighthouse1.yml -test
 docker run --name host2 --rm "$CONTAINER" -config host2.yml -test
 
 vagrant up
-
-# OpenBSD: synced folders are disabled because Vagrant's rsync installer
-# uses ftp.openbsd.org which no longer hosts packages for older releases.
-# Copy build artifacts in via scp instead.
-case "$1" in
-    openbsd-*)
-        vagrant ssh -c "sudo mkdir -p /nebula" -- -T
-        tar -cf - -C build . | vagrant ssh -c "sudo tar -xf - -C /nebula && sudo chmod -R a+r /nebula" -- -T
-        ;;
-esac
-
 vagrant ssh -c "cd /nebula && /nebula/$1-nebula -config host3.yml -test" -- -T
 
 docker run --name lighthouse1 --device /dev/net/tun:/dev/net/tun --cap-add NET_ADMIN --rm "$CONTAINER" -config lighthouse1.yml 2>&1 | tee logs/lighthouse1 | sed -u 's/^/  [lighthouse1]  /' &

.github/workflows/smoke/smoke.sh

@@ -124,7 +124,6 @@ set -x
 # host2 speaking to host4 on UDP 4000 should allow it to reply, when firewall rules would normally not permit this
 docker exec host2 sh -c "/usr/bin/echo host2 | ncat -nuv 192.168.100.4 4000"
 docker exec host2 ncat -e '/usr/bin/echo helloagainfromhost2' -nkluv 0.0.0.0 4000 &
-sleep 1
 docker exec host4 sh -c "/usr/bin/echo host4 | ncat -nuv 192.168.100.2 4000"
 
 docker exec host4 sh -c 'kill 1'

.github/workflows/smoke/vagrant-openbsd-amd64/Vagrantfile

@@ -3,5 +3,5 @@
 Vagrant.configure("2") do |config|
   config.vm.box = "generic/openbsd7"
 
-  config.vm.synced_folder ".", "/vagrant", disabled: true
+  config.vm.synced_folder "../build", "/nebula", type: "rsync"
 end

README.md

@@ -57,7 +57,7 @@ Check the [releases](https://github.com/slackhq/nebula/releases/latest) page for
     docker pull nebulaoss/nebula
     ```
 
-#### Mobile ([source code](https://github.com/DefinedNet/mobile_nebula))
+#### Mobile
 
 - [iOS](https://apps.apple.com/us/app/mobile-nebula/id1509587936?itsct=apps_box&itscg=30200)
 - [Android](https://play.google.com/store/apps/details?id=net.defined.mobile_nebula&pcampaignid=pcampaignidMKT-Other-global-all-co-prtnr-py-PartBadge-Mar2515-1)
@@ -76,8 +76,6 @@ Nebula was created to provide a mechanism for groups of hosts to communicate sec
 
 ## Getting started (quickly)
 
-**Don't want to manage your own PKI and lighthouses?** [Managed Nebula](https://www.defined.net/) from Defined Networking handles all of this for you.
-
 To set up a Nebula network, you'll need:
 
 #### 1. The [Nebula binaries](https://github.com/slackhq/nebula/releases) or [Distribution Packages](https://github.com/slackhq/nebula#distribution-packages) for your specific platform. Specifically you'll need `nebula-cert` and the specific nebula binary for each platform you use.

cert/ca_pool.go

@@ -1,14 +1,11 @@
 package cert
 
 import (
-	"bufio"
-	"bytes"
-	"encoding/pem"
 	"errors"
 	"fmt"
-	"io"
 	"net/netip"
 	"slices"
+	"strings"
 	"time"
 )
 
@@ -32,46 +29,22 @@ func NewCAPool() *CAPool {
 // If the pool contains any expired certificates, an ErrExpired will be
 // returned along with the pool. The caller must handle any such errors.
 func NewCAPoolFromPEM(caPEMs []byte) (*CAPool, error) {
-	return NewCAPoolFromPEMReader(bytes.NewReader(caPEMs))
-}
-
-// NewCAPoolFromPEMReader will create a new CA pool from the provided reader.
-// The reader must contain a PEM-encoded set of nebula certificates.
-func NewCAPoolFromPEMReader(r io.Reader) (*CAPool, error) {
 	pool := NewCAPool()
-
+	var err error
 	var expired bool
-
-	scanner := bufio.NewScanner(r)
-	scanner.Split(SplitPEM)
-
-	for scanner.Scan() {
-		pemBytes := scanner.Bytes()
-
-		block, rest := pem.Decode(pemBytes)
-		if len(bytes.TrimSpace(rest)) > 0 {
-			return nil, ErrInvalidPEMBlock
+	for {
+		caPEMs, err = pool.AddCAFromPEM(caPEMs)
+		if errors.Is(err, ErrExpired) {
+			expired = true
+			err = nil
 		}
-		if block == nil {
-			return nil, ErrInvalidPEMBlock
-		}
-
-		c, err := unmarshalCertificateBlock(block)
 		if err != nil {
 			return nil, err
 		}
-
-		err = pool.AddCA(c)
-		if errors.Is(err, ErrExpired) {
-			expired = true
-			continue
-		} else if err != nil {
-			return nil, err
+		if len(caPEMs) == 0 || strings.TrimSpace(string(caPEMs)) == "" {
+			break
 		}
 	}
-	if err := scanner.Err(); err != nil {
-		return nil, ErrInvalidPEMBlock
-	}
 
 	if expired {
 		return pool, ErrExpired

cert/ca_pool_test.go

@@ -1,10 +1,7 @@
 package cert
 
 import (
-	"bytes"
-	"io"
 	"net/netip"
-	"strings"
 	"testing"
 	"time"
 
@@ -115,60 +112,6 @@ k+coOv04r+zh33ISyhbsafnYduN17p2eD7CmHvHuerguXD9f32gcxo/KsFCKEjMe
 	assert.Len(t, ppppp.CAs, 1)
 }
 
-// oneByteReader wraps a reader to return at most 1 byte per Read call,
-// exercising the streaming accumulation logic in NewCAPoolFromPEMReader.
-type oneByteReader struct {
-	r io.Reader
-}
-
-func (o *oneByteReader) Read(p []byte) (int, error) {
-	if len(p) == 0 {
-		return 0, nil
-	}
-	return o.r.Read(p[:1])
-}
-
-func TestNewCAPoolFromPEMReader_EmptyReader(t *testing.T) {
-	pool, err := NewCAPoolFromPEMReader(bytes.NewReader(nil))
-	require.NoError(t, err)
-	assert.Empty(t, pool.CAs)
-
-	pool, err = NewCAPoolFromPEMReader(strings.NewReader("   \n\t\n  "))
-	require.NoError(t, err)
-	assert.Empty(t, pool.CAs)
-}
-
-func TestNewCAPoolFromPEMReader_OneByteReads(t *testing.T) {
-	ca1, _, _, pem1 := NewTestCaCert(Version2, Curve_CURVE25519, time.Now(), time.Now().Add(time.Hour), nil, nil, nil)
-	ca2, _, _, pem2 := NewTestCaCert(Version2, Curve_CURVE25519, time.Now(), time.Now().Add(time.Hour), nil, nil, nil)
-
-	bundle := append(pem1, pem2...)
-	pool, err := NewCAPoolFromPEMReader(&oneByteReader{r: bytes.NewReader(bundle)})
-	require.NoError(t, err)
-	assert.Len(t, pool.CAs, 2)
-
-	fp1, err := ca1.Fingerprint()
-	require.NoError(t, err)
-	fp2, err := ca2.Fingerprint()
-	require.NoError(t, err)
-
-	assert.Contains(t, pool.CAs, fp1)
-	assert.Contains(t, pool.CAs, fp2)
-}
-
-func TestNewCAPoolFromPEMReader_TruncatedPEM(t *testing.T) {
-	_, err := NewCAPoolFromPEMReader(strings.NewReader("-----BEGIN NEBULA CERTIFICATE-----\npartialdata"))
-	assert.ErrorIs(t, err, ErrInvalidPEMBlock)
-}
-
-func TestNewCAPoolFromPEMReader_TrailingGarbage(t *testing.T) {
-	_, _, _, pem1 := NewTestCaCert(Version2, Curve_CURVE25519, time.Now(), time.Now().Add(time.Hour), nil, nil, nil)
-
-	bundle := append(pem1, []byte("some trailing garbage")...)
-	_, err := NewCAPoolFromPEMReader(bytes.NewReader(bundle))
-	assert.ErrorIs(t, err, ErrInvalidPEMBlock)
-}
-
 func TestCertificateV1_Verify(t *testing.T) {
 	ca, _, caKey, _ := NewTestCaCert(Version1, Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, nil)
 	c, _, _, _ := NewTestCert(Version1, Curve_CURVE25519, ca, caKey, "test cert", time.Now(), time.Now().Add(5*time.Minute), nil, nil, nil)

cert/p256/p256.go

@@ -44,12 +44,7 @@ func swap(r, s []byte) ([]byte, []byte, error) {
 	}
 	sNormalized := nMod.Nat().Sub(bigS, nMod)
 
-	result := sNormalized.Bytes(nMod)
-	for len(result) > 1 && result[0] == 0 {
-		result = result[1:]
-	}
-
-	return r, result, nil
+	return r, sNormalized.Bytes(nMod), nil
 }
 
 func Normalize(sig []byte) ([]byte, error) {

cert/pem.go

@@ -1,66 +1,12 @@
 package cert
 
 import (
-	"bytes"
 	"encoding/pem"
-	"errors"
 	"fmt"
 
 	"golang.org/x/crypto/ed25519"
 )
 
-var ErrTruncatedPEMBlock = errors.New("truncated PEM block")
-
-// SplitPEM is a split function for bufio.Scanner that returns each PEM block.
-func SplitPEM(data []byte, atEOF bool) (advance int, token []byte, err error) {
-	// Look for the start of a PEM block
-	start := bytes.Index(data, []byte("-----BEGIN "))
-	if start == -1 {
-		if atEOF && len(bytes.TrimSpace(data)) > 0 {
-			// Non-whitespace content with no PEM block
-			return 0, nil, ErrTruncatedPEMBlock
-		}
-		if atEOF {
-			return len(data), nil, nil
-		}
-		// Request more data
-		return 0, nil, nil
-	}
-
-	// Look for the end marker
-	endMarkerStart := bytes.Index(data[start:], []byte("-----END "))
-	if endMarkerStart == -1 {
-		if atEOF {
-			// Incomplete PEM block at EOF
-			return 0, nil, ErrTruncatedPEMBlock
-		}
-		// Need more data to find the end
-		return 0, nil, nil
-	}
-
-	// Find the actual end of the END line (after the newline)
-	endMarkerStart += start
-	endLineEnd := bytes.IndexByte(data[endMarkerStart:], '\n')
-	var end int
-	if endLineEnd == -1 {
-		if atEOF {
-			// END marker without newline at EOF - take it anyway
-			end = len(data)
-		} else {
-			// Need more data
-			return 0, nil, nil
-		}
-	} else {
-		end = endMarkerStart + endLineEnd + 1
-	}
-
-	// Extract the PEM block
-	pemBlock := data[start:end]
-
-	// Return the valid PEM block
-	return end, pemBlock, nil
-}
-
 const ( //cert banners
 	CertificateBanner   = "NEBULA CERTIFICATE"
 	CertificateV2Banner = "NEBULA CERTIFICATE V2"
@@ -91,7 +37,19 @@ func UnmarshalCertificateFromPEM(b []byte) (Certificate, []byte, error) {
 		return nil, r, ErrInvalidPEMBlock
 	}
 
-	c, err := unmarshalCertificateBlock(p)
+	var c Certificate
+	var err error
+
+	switch p.Type {
+	// Implementations must validate the resulting certificate contains valid information
+	case CertificateBanner:
+		c, err = unmarshalCertificateV1(p.Bytes, nil)
+	case CertificateV2Banner:
+		c, err = unmarshalCertificateV2(p.Bytes, nil, Curve_CURVE25519)
+	default:
+		return nil, r, ErrInvalidPEMCertificateBanner
+	}
+
 	if err != nil {
 		return nil, r, err
 	}
@@ -100,20 +58,6 @@ func UnmarshalCertificateFromPEM(b []byte) (Certificate, []byte, error) {
 
 }
 
-// unmarshalCertificateBlock decodes a single PEM block into a certificate.
-// It expects a Nebula certificate banner and returns ErrInvalidPEMCertificateBanner otherwise.
-func unmarshalCertificateBlock(block *pem.Block) (Certificate, error) {
-	switch block.Type {
-	// Implementations must validate the resulting certificate contains valid information
-	case CertificateBanner:
-		return unmarshalCertificateV1(block.Bytes, nil)
-	case CertificateV2Banner:
-		return unmarshalCertificateV2(block.Bytes, nil, Curve_CURVE25519)
-	default:
-		return nil, ErrInvalidPEMCertificateBanner
-	}
-}
-
 func marshalCertPublicKeyToPEM(c Certificate) []byte {
 	if c.IsCA() {
 		return MarshalSigningPublicKeyToPEM(c.Curve(), c.PublicKey())

cert/pem_test.go

@@ -1,88 +1,12 @@
 package cert
 
 import (
-	"bufio"
-	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
 
-func scanAll(t *testing.T, input string) ([]string, error) {
-	t.Helper()
-	scanner := bufio.NewScanner(strings.NewReader(input))
-	scanner.Split(SplitPEM)
-	var blocks []string
-	for scanner.Scan() {
-		blocks = append(blocks, scanner.Text())
-	}
-	return blocks, scanner.Err()
-}
-
-func TestSplitPEM_Single(t *testing.T) {
-	input := "-----BEGIN TEST-----\ndata\n-----END TEST-----\n"
-	blocks, err := scanAll(t, input)
-	require.NoError(t, err)
-	require.Len(t, blocks, 1)
-	require.Equal(t, input, blocks[0])
-}
-
-func TestSplitPEM_Multiple(t *testing.T) {
-	block1 := "-----BEGIN TEST-----\naaa\n-----END TEST-----\n"
-	block2 := "-----BEGIN TEST-----\nbbb\n-----END TEST-----\n"
-	blocks, err := scanAll(t, block1+block2)
-	require.NoError(t, err)
-	require.Len(t, blocks, 2)
-	require.Equal(t, block1, blocks[0])
-	require.Equal(t, block2, blocks[1])
-}
-
-func TestSplitPEM_CommentsAndWhitespaceBetweenBlocks(t *testing.T) {
-	input := "# comment\n\n-----BEGIN TEST-----\naaa\n-----END TEST-----\n\n# another comment\n\n-----BEGIN TEST-----\nbbb\n-----END TEST-----\n"
-	blocks, err := scanAll(t, input)
-	require.NoError(t, err)
-	require.Len(t, blocks, 2)
-}
-
-func TestSplitPEM_Empty(t *testing.T) {
-	blocks, err := scanAll(t, "")
-	require.NoError(t, err)
-	require.Empty(t, blocks)
-}
-
-func TestSplitPEM_WhitespaceOnly(t *testing.T) {
-	blocks, err := scanAll(t, "  \n\t\n  ")
-	require.NoError(t, err)
-	require.Empty(t, blocks)
-}
-
-func TestSplitPEM_TrailingGarbage(t *testing.T) {
-	input := "-----BEGIN TEST-----\ndata\n-----END TEST-----\ngarbage"
-	blocks, err := scanAll(t, input)
-	require.ErrorIs(t, err, ErrTruncatedPEMBlock)
-	require.Len(t, blocks, 1)
-}
-
-func TestSplitPEM_TruncatedBlock(t *testing.T) {
-	input := "-----BEGIN TEST-----\npartial data with no end"
-	_, err := scanAll(t, input)
-	require.ErrorIs(t, err, ErrTruncatedPEMBlock)
-}
-
-func TestSplitPEM_NoEndNewline(t *testing.T) {
-	input := "-----BEGIN TEST-----\ndata\n-----END TEST-----"
-	blocks, err := scanAll(t, input)
-	require.NoError(t, err)
-	require.Len(t, blocks, 1)
-	require.Equal(t, input, blocks[0])
-}
-
-func TestSplitPEM_GarbageOnly(t *testing.T) {
-	_, err := scanAll(t, "this is not PEM data")
-	require.ErrorIs(t, err, ErrTruncatedPEMBlock)
-}
-
 func TestUnmarshalCertificateFromPEM(t *testing.T) {
 	goodCert := []byte(`
 # A good cert

cmd/nebula-cert/verify.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"io"
 	"os"
+	"strings"
 	"time"
 
 	"github.com/slackhq/nebula/cert"
@@ -39,15 +40,21 @@ func verify(args []string, out io.Writer, errOut io.Writer) error {
 		return err
 	}
 
-	caFile, err := os.Open(*vf.caPath)
+	rawCACert, err := os.ReadFile(*vf.caPath)
 	if err != nil {
 		return fmt.Errorf("error while reading ca: %w", err)
 	}
-	defer caFile.Close()
 
-	caPool, err := cert.NewCAPoolFromPEMReader(caFile)
-	if err != nil && !errors.Is(err, cert.ErrExpired) {
-		return fmt.Errorf("error while adding ca cert to pool: %w", err)
+	caPool := cert.NewCAPool()
+	for {
+		rawCACert, err = caPool.AddCAFromPEM(rawCACert)
+		if err != nil {
+			return fmt.Errorf("error while adding ca cert to pool: %w", err)
+		}
+
+		if rawCACert == nil || len(rawCACert) == 0 || strings.TrimSpace(string(rawCACert)) == "" {
+			break
+		}
 	}
 
 	rawCert, err := os.ReadFile(*vf.certPath)

cmd/nebula-cert/verify_test.go

@@ -64,7 +64,7 @@ func Test_verify(t *testing.T) {
 	err = verify([]string{"-ca", caFile.Name(), "-crt", "does_not_exist"}, ob, eb)
 	assert.Empty(t, ob.String())
 	assert.Empty(t, eb.String())
-	require.ErrorIs(t, err, cert.ErrInvalidPEMBlock)
+	require.EqualError(t, err, "error while adding ca cert to pool: input did not contain a valid PEM encoded block")
 
 	// make a ca for later
 	caPub, caPriv, _ := ed25519.GenerateKey(rand.Reader)

e2e/handshake_manager_test.go

@@ -1,565 +0,0 @@
-//go:build e2e_testing
-// +build e2e_testing
-
-package e2e
-
-import (
-	"net/netip"
-	"testing"
-	"time"
-
-	"github.com/slackhq/nebula"
-	"github.com/slackhq/nebula/cert"
-	"github.com/slackhq/nebula/cert_test"
-	"github.com/slackhq/nebula/e2e/router"
-	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/udp"
-	"github.com/stretchr/testify/assert"
-)
-
-// makeHandshakePacket creates a handshake packet with the given parameters.
-func makeHandshakePacket(from, to netip.AddrPort, subtype header.MessageSubType, remoteIndex uint32, counter uint64) *udp.Packet {
-	data := make([]byte, 200)
-	header.Encode(data, header.Version, header.Handshake, subtype, remoteIndex, counter)
-	for i := header.Len; i < len(data); i++ {
-		data[i] = byte(i)
-	}
-	return &udp.Packet{To: to, From: from, Data: data}
-}
-
-func TestHandshakeRetransmitDuplicate(t *testing.T) {
-	// Verify the responder correctly handles receiving the same msg1 multiple times
-	// (retransmission). The duplicate goes through CheckAndComplete -> ErrAlreadySeen
-	// and the cached response is resent.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	t.Log("Trigger handshake from me to them")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
-
-	t.Log("Grab my msg1")
-	msg1 := myControl.GetFromUDP(true)
-
-	t.Log("Inject msg1 into them, first time")
-	theirControl.InjectUDPPacket(msg1)
-	_ = theirControl.GetFromUDP(true)
-
-	t.Log("Inject the SAME msg1 again, tests ErrAlreadySeen path")
-	theirControl.InjectUDPPacket(msg1)
-	resp2 := theirControl.GetFromUDP(true)
-	assert.NotNil(t, resp2, "should get cached response on duplicate msg1")
-
-	t.Log("Complete handshake with cached response")
-	myControl.InjectUDPPacket(resp2)
-	myControl.WaitForType(1, 0, theirControl)
-
-	t.Log("Drain cached packet and verify tunnel works")
-	cachedPacket := theirControl.GetFromTun(true)
-	assertUdpPacket(t, []byte("Hi"), cachedPacket, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	t.Log("Verify only one tunnel exists on each side")
-	assert.Len(t, myControl.ListHostmapHosts(false), 1)
-	assert.Len(t, theirControl.ListHostmapHosts(false), 1)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeTruncatedPacketRecovery(t *testing.T) {
-	// Verify that a truncated handshake packet is ignored and the real
-	// packet can still complete the handshake.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	t.Log("Trigger handshake")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
-
-	t.Log("Get msg1 and deliver to responder")
-	msg1 := myControl.GetFromUDP(true)
-	theirControl.InjectUDPPacket(msg1)
-
-	t.Log("Get the real response")
-	realResp := theirControl.GetFromUDP(true)
-
-	t.Log("Truncate the response and inject, should be ignored")
-	truncResp := realResp.Copy()
-	truncResp.Data = truncResp.Data[:header.Len]
-	myControl.InjectUDPPacket(truncResp)
-
-	t.Log("Verify pending handshake survived the truncated packet")
-	assert.NotEmpty(t, myControl.ListHostmapHosts(true), "pending handshake should still exist")
-
-	t.Log("Inject real response, should complete handshake")
-	myControl.InjectUDPPacket(realResp)
-	myControl.WaitForType(1, 0, theirControl)
-
-	t.Log("Drain and verify tunnel")
-	cachedPacket := theirControl.GetFromTun(true)
-	assertUdpPacket(t, []byte("Hi"), cachedPacket, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeOrphanedMsg2Dropped(t *testing.T) {
-	// A msg2 arriving with no matching pending index should be silently dropped
-	// with no response sent and no state changes.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	t.Log("Complete a normal handshake")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
-	r.RouteForAllUntilTxTun(theirControl)
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	t.Log("Record hostmap state")
-	myIndexes := len(myControl.ListHostmapIndexes(false))
-
-	t.Log("Inject a fake msg2 with unknown RemoteIndex")
-	myControl.InjectUDPPacket(makeHandshakePacket(theirUdpAddr, myUdpAddr, header.HandshakeIXPSK0, 0xDEADBEEF, 2))
-
-	t.Log("Verify no new indexes created")
-	assert.Equal(t, myIndexes, len(myControl.ListHostmapIndexes(false)))
-
-	t.Log("Verify no UDP response was sent")
-	time.Sleep(100 * time.Millisecond)
-	assert.Nil(t, myControl.GetFromUDP(false), "should not send a response to orphaned msg2")
-
-	t.Log("Verify existing tunnel still works")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeUnknownMessageCounter(t *testing.T) {
-	// A handshake packet with an unexpected message counter should be silently
-	// dropped with no side effects and no UDP response.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, _, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	t.Log("Inject handshake with MessageCounter=3")
-	myControl.InjectUDPPacket(makeHandshakePacket(theirUdpAddr, myUdpAddr, header.HandshakeIXPSK0, 0, 3))
-
-	t.Log("Inject handshake with MessageCounter=99")
-	myControl.InjectUDPPacket(makeHandshakePacket(theirUdpAddr, myUdpAddr, header.HandshakeIXPSK0, 0, 99))
-
-	t.Log("Verify no tunnels or pending handshakes")
-	assert.Empty(t, myControl.ListHostmapHosts(false))
-	assert.Empty(t, myControl.ListHostmapHosts(true))
-
-	t.Log("Verify no UDP response was sent")
-	time.Sleep(100 * time.Millisecond)
-	assert.Nil(t, myControl.GetFromUDP(false))
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeUnknownSubtype(t *testing.T) {
-	// A handshake packet with an unknown subtype should be silently dropped.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, _, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, _, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.Start()
-	theirControl.Start()
-
-	t.Log("Inject handshake with unknown subtype 99")
-	myControl.InjectUDPPacket(makeHandshakePacket(theirUdpAddr, myUdpAddr, header.MessageSubType(99), 0, 1))
-
-	t.Log("Verify no tunnels or pending handshakes")
-	assert.Empty(t, myControl.ListHostmapHosts(false))
-	assert.Empty(t, myControl.ListHostmapHosts(true))
-
-	t.Log("Verify no UDP response was sent")
-	time.Sleep(100 * time.Millisecond)
-	assert.Nil(t, myControl.GetFromUDP(false))
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeLateResponse(t *testing.T) {
-	// After a handshake times out, a late response should be silently ignored
-	// with no new tunnels created.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", m{
-		"handshakes": m{
-			"try_interval": "200ms",
-			"retries":      2,
-		},
-	})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	t.Log("Trigger handshake from me")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
-
-	t.Log("Grab msg1 but don't deliver")
-	msg1 := myControl.GetFromUDP(true)
-
-	t.Log("Wait for handshake to time out")
-	for i := 0; i < 5; i++ {
-		time.Sleep(300 * time.Millisecond)
-		myControl.GetFromUDP(false)
-	}
-
-	t.Log("Confirm no pending handshakes remain")
-	assert.Empty(t, myControl.ListHostmapHosts(true))
-
-	t.Log("Deliver old msg1 to them, they create a tunnel")
-	theirControl.InjectUDPPacket(msg1)
-	resp := theirControl.GetFromUDP(true)
-	assert.NotNil(t, resp)
-
-	t.Log("Inject late response into me, should be ignored")
-	myControl.InjectUDPPacket(resp)
-
-	t.Log("No tunnel should exist on my side")
-	assert.Empty(t, myControl.ListHostmapHosts(false))
-	assert.Empty(t, myControl.ListHostmapHosts(true))
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeSelfConnectionRejected(t *testing.T) {
-	// Verify that a node rejects a handshake containing its own VPN IP in the
-	// peer cert. We do this by sending the initiator's own msg1 back to itself.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-
-	// Need a lighthouse entry to trigger a handshake
-	myControl.InjectLightHouseAddr(netip.MustParseAddr("10.128.0.2"), netip.MustParseAddrPort("10.0.0.2:4242"))
-
-	myControl.Start()
-
-	t.Log("Trigger handshake from me")
-	myControl.InjectTunUDPPacket(netip.MustParseAddr("10.128.0.2"), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
-	msg1 := myControl.GetFromUDP(true)
-
-	t.Log("Drain any handshake retransmits before injecting")
-	time.Sleep(100 * time.Millisecond)
-	for myControl.GetFromUDP(false) != nil {
-	}
-
-	t.Log("Feed my own msg1 back to me as if it came from someone else")
-	selfMsg := msg1.Copy()
-	selfMsg.From = netip.MustParseAddrPort("10.0.0.99:4242")
-	selfMsg.To = myUdpAddr
-	myControl.InjectUDPPacket(selfMsg)
-
-	t.Log("Verify no response was sent (self-connection rejected)")
-	time.Sleep(100 * time.Millisecond)
-	// Drain any further retransmits from the original handshake, then check
-	// that none of them are a handshake response (MessageCounter=2)
-	h := &header.H{}
-	for {
-		p := myControl.GetFromUDP(false)
-		if p == nil {
-			break
-		}
-		_ = h.Parse(p.Data)
-		assert.NotEqual(t, uint64(2), h.MessageCounter,
-			"should not send a stage 2 response to self-connection")
-	}
-
-	t.Log("Verify no tunnel to myself was created")
-	assert.Nil(t, myControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false))
-
-	myControl.Stop()
-}
-
-func TestHandshakeMessageCounter0Dropped(t *testing.T) {
-	// MessageCounter=0 is not a valid handshake message and should be dropped.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, _, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	_, _, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.Start()
-
-	t.Log("Inject handshake with MessageCounter=0")
-	myControl.InjectUDPPacket(makeHandshakePacket(theirUdpAddr, myUdpAddr, header.HandshakeIXPSK0, 0, 0))
-
-	time.Sleep(100 * time.Millisecond)
-	assert.Empty(t, myControl.ListHostmapHosts(false))
-	assert.Empty(t, myControl.ListHostmapHosts(true))
-	assert.Nil(t, myControl.GetFromUDP(false))
-
-	myControl.Stop()
-}
-
-func TestHandshakeRemoteAllowList(t *testing.T) {
-	// Verify that a handshake from a blocked underlay IP is dropped with no
-	// response and no state changes. Then verify the same packet from an
-	// allowed IP succeeds.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", m{
-		"lighthouse": m{
-			"remote_allow_list": m{
-				"10.0.0.0/8": true,
-				"0.0.0.0/0":  false,
-			},
-		},
-	})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	t.Log("Trigger handshake from them")
-	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi"))
-	msg1 := theirControl.GetFromUDP(true)
-
-	t.Log("Rewrite the source to a blocked IP and inject")
-	blockedMsg := msg1.Copy()
-	blockedMsg.From = netip.MustParseAddrPort("192.168.1.1:4242")
-	myControl.InjectUDPPacket(blockedMsg)
-
-	t.Log("Verify no tunnel, no pending, no response from blocked source")
-	time.Sleep(100 * time.Millisecond)
-	assert.Empty(t, myControl.ListHostmapHosts(false))
-	assert.Empty(t, myControl.ListHostmapHosts(true))
-	assert.Nil(t, myControl.GetFromUDP(false), "should not respond to blocked source")
-
-	t.Log("Now inject the real packet from the allowed source")
-	myControl.InjectUDPPacket(msg1)
-
-	t.Log("Verify handshake completes from allowed source")
-	resp := myControl.GetFromUDP(true)
-	assert.NotNil(t, resp)
-	theirControl.InjectUDPPacket(resp)
-	theirControl.WaitForType(1, 0, myControl)
-
-	t.Log("Drain cached packet and verify tunnel works")
-	cachedPacket := myControl.GetFromTun(true)
-	assertUdpPacket(t, []byte("Hi"), cachedPacket, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), 80, 80)
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeAlreadySeenPreferredRemote(t *testing.T) {
-	// When a duplicate msg1 arrives via ErrAlreadySeen, verify the tunnel
-	// remains functional and hostmap index count is stable.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	t.Log("Complete a normal handshake via the router")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
-	r.RouteForAllUntilTxTun(theirControl)
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	t.Log("Record hostmap state")
-	theirIndexes := len(theirControl.ListHostmapIndexes(false))
-	hi := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-	assert.NotNil(t, hi)
-	originalRemote := hi.CurrentRemote
-
-	t.Log("Re-trigger traffic to cause a new handshake attempt (ErrAlreadySeen)")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("roam"))
-	r.RouteForAllUntilTxTun(theirControl)
-
-	t.Log("Verify tunnel still works")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	t.Log("Verify remote is still valid and index count is stable")
-	hi2 := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-	assert.NotNil(t, hi2)
-	assert.Equal(t, originalRemote, hi2.CurrentRemote)
-	assert.Equal(t, theirIndexes, len(theirControl.ListHostmapIndexes(false)),
-		"no extra indexes should be created from ErrAlreadySeen")
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestHandshakeWrongResponderPacketStore(t *testing.T) {
-	// Verify that when the wrong host responds, the cached packets are
-	// transferred to the new handshake, the evil tunnel is closed, evil's
-	// address is blocked, and the correct tunnel is eventually established.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.100/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.99/24", nil)
-	evilControl, evilVpnIpNet, evilUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "evil", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), evilUdpAddr)
-
-	r := router.NewR(t, myControl, theirControl, evilControl)
-	defer r.RenderFlow()
-
-	myControl.Start()
-	theirControl.Start()
-	evilControl.Start()
-
-	t.Log("Send multiple packets to them (cached during handshake)")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("packet1"))
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("packet2"))
-
-	t.Log("Route until evil tunnel is closed")
-	h := &header.H{}
-	r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
-		if err := h.Parse(p.Data); err != nil {
-			panic(err)
-		}
-		if h.Type == header.CloseTunnel && p.To == evilUdpAddr {
-			return router.RouteAndExit
-		}
-		return router.KeepRouting
-	})
-
-	t.Log("Verify evil's address is blocked in the new pending handshake")
-	pendingHI := myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), true)
-	if pendingHI != nil {
-		assert.NotContains(t, pendingHI.RemoteAddrs, evilUdpAddr,
-			"evil's address should be blocked")
-	}
-
-	t.Log("Inject correct lighthouse addr for them")
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	t.Log("Route until cached packets arrive at the real them")
-	p := r.RouteForAllUntilTxTun(theirControl)
-	assert.NotNil(t, p, "a cached packet should be delivered to the correct host")
-
-	t.Log("Verify the correct host has a tunnel")
-	assertHostInfoPair(t, myUdpAddr, theirUdpAddr, myVpnIpNet, theirVpnIpNet, myControl, theirControl)
-
-	t.Log("Verify no hostinfo artifacts from evil remain")
-	assert.Nil(t, myControl.GetHostInfoByVpnAddr(evilVpnIpNet[0].Addr(), true),
-		"no pending hostinfo for evil")
-	assert.Nil(t, myControl.GetHostInfoByVpnAddr(evilVpnIpNet[0].Addr(), false),
-		"no main hostinfo for evil")
-
-	myControl.Stop()
-	theirControl.Stop()
-	evilControl.Stop()
-}
-
-func TestHandshakeRelayComplete(t *testing.T) {
-	// Verify that a relay handshake completes correctly and relay state is
-	// properly maintained on all three nodes.
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
-	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "relay", "10.128.0.128/24", m{"relay": m{"am_relay": true}})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", m{"relay": m{"use_relays": true}})
-
-	myControl.InjectLightHouseAddr(relayVpnIpNet[0].Addr(), relayUdpAddr)
-	myControl.InjectRelays(theirVpnIpNet[0].Addr(), []netip.Addr{relayVpnIpNet[0].Addr()})
-	relayControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	r := router.NewR(t, myControl, relayControl, theirControl)
-	defer r.RenderFlow()
-
-	myControl.Start()
-	relayControl.Start()
-	theirControl.Start()
-
-	t.Log("Trigger handshake via relay")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi via relay"))
-
-	p := r.RouteForAllUntilTxTun(theirControl)
-	assertUdpPacket(t, []byte("Hi via relay"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
-
-	t.Log("Verify bidirectional tunnel via relay")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	t.Log("Verify relay state on my side shows relay-to-me")
-	myHI := myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), false)
-	assert.NotNil(t, myHI)
-	assert.NotEmpty(t, myHI.CurrentRelaysToMe, "should have relay-to-me for them")
-
-	t.Log("Verify relay state on their side shows relay-to-me")
-	theirHI := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-	assert.NotNil(t, theirHI)
-	assert.NotEmpty(t, theirHI.CurrentRelaysToMe, "should have relay-to-me for me")
-
-	t.Log("Verify relay node shows through-me relays")
-	relayHI := relayControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-	assert.NotNil(t, relayHI)
-
-	myControl.Stop()
-	relayControl.Stop()
-	theirControl.Stop()
-}
-
-// NOTE: Relay V1 cert + IPv6 rejection is not tested here because
-// InjectTunUDPPacket from a V4 node to a V6 address panics in the test
-// framework. The check is in handshake_manager.go handleOutbound relay
-// logic (lines ~304-313): if the relay host has a V1 cert and either
-// address is IPv6, the relay is skipped.
-
-// NOTE: Relay reestablishment (Disestablished state transition) is covered
-// by the existing TestReestablishRelays in handshakes_test.go.

examples/config.yml

@@ -204,12 +204,6 @@ punchy:
   # Trusted SSH CA public keys. These are the public keys of the CAs that are allowed to sign SSH keys for access.
   #trusted_cas:
     #- "ssh public key string"
-  # sandbox_dir restricts file paths for profiling commands (start-cpu-profile, save-heap-profile,
-  # save-mutex-profile) to the specified directory. Relative paths will be resolved within this directory,
-  # and absolute paths outside of it will be rejected. Default is $TMP/nebula-debug.
-  # The directory is NOT automatically created.
-  # Overriding this to "" is the same as "/" and will allow overwriting any path on the host.
-  #sandbox_dir: /var/tmp/nebula-debug
 
 # EXPERIMENTAL: relay support for networks that can't establish direct connections.
 relay:
@@ -348,6 +342,14 @@ logging:
   # after receiving the response for lighthouse queries
   #trigger_buffer: 64
 
+  # max_rate limits the number of new inbound handshakes per second. Once the limit is reached,
+  # new handshakes are dropped until the next second. A value of 0 means unlimited (default).
+  # This is useful for preventing DoS attacks that attempt to exhaust CPU with handshake crypto.
+  # Running `openssl speed ecdhp256` on your hardware can be a good rule of thumb for choosing
+  # a max, as each handshake performs similar DH operations. Note that this benchmarks a single
+  # core, so you may wish to scale the value by the number of `routines` configured.
+  #max_rate: 0
+
 # Tunnel manager settings
 #tunnels:
   # drop_inactive controls whether inactive tunnels are maintained or dropped after the inactive_timeout period has

go.mod

@@ -13,8 +13,8 @@ require (
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
 	github.com/kardianos/service v1.2.4
-	github.com/miekg/dns v1.1.72
-	github.com/miekg/pkcs11 v1.1.2
+	github.com/miekg/dns v1.1.70
+	github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b
 	github.com/nbrownus/go-metrics-prometheus v0.0.0-20210712211119-974a6260965f
 	github.com/prometheus/client_golang v1.23.2
 	github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475

go.sum

@@ -85,10 +85,10 @@ github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0SNc=
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
 github.com/matttproud/golang_protobuf_extensions v1.0.1/go.mod h1:D8He9yQNgCq6Z5Ld7szi9bcBfOoFv/3dc6xSMkL2PC0=
-github.com/miekg/dns v1.1.72 h1:vhmr+TF2A3tuoGNkLDFK9zi36F2LS+hKTRW0Uf8kbzI=
-github.com/miekg/dns v1.1.72/go.mod h1:+EuEPhdHOsfk6Wk5TT2CzssZdqkmFhf8r+aVyDEToIs=
-github.com/miekg/pkcs11 v1.1.2 h1:/VxmeAX5qU6Q3EwafypogwWbYryHFmF2RpkJmw3m4MQ=
-github.com/miekg/pkcs11 v1.1.2/go.mod h1:XsNlhZGX73bx86s2hdc/FuaLm2CPZJemRLMA+WTFxgs=
+github.com/miekg/dns v1.1.70 h1:DZ4u2AV35VJxdD9Fo9fIWm119BsQL5cZU1cQ9s0LkqA=
+github.com/miekg/dns v1.1.70/go.mod h1:+EuEPhdHOsfk6Wk5TT2CzssZdqkmFhf8r+aVyDEToIs=
+github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b h1:J/AzCvg5z0Hn1rqZUJjpbzALUmkKX0Zwbc/i4fw7Sfk=
+github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b/go.mod h1:XsNlhZGX73bx86s2hdc/FuaLm2CPZJemRLMA+WTFxgs=
 github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
 github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
 github.com/modern-go/reflect2 v0.0.0-20180701023420-4b7aa43c6742/go.mod h1:bx2lNnkwVCuqBIxFjflWJWanXIb3RllmbCylyMrvgv0=

handshake_manager.go

@@ -23,22 +23,25 @@ const (
 	DefaultHandshakeRetries       = 10
 	DefaultHandshakeTriggerBuffer = 64
 	DefaultUseRelays              = true
+	DefaultMaxHandshakeRate       = 0 // 0 means unlimited
 )
 
 var (
 	defaultHandshakeConfig = HandshakeConfig{
-		tryInterval:   DefaultHandshakeTryInterval,
-		retries:       DefaultHandshakeRetries,
-		triggerBuffer: DefaultHandshakeTriggerBuffer,
-		useRelays:     DefaultUseRelays,
+		tryInterval:      DefaultHandshakeTryInterval,
+		retries:          DefaultHandshakeRetries,
+		triggerBuffer:    DefaultHandshakeTriggerBuffer,
+		useRelays:        DefaultUseRelays,
+		maxHandshakeRate: DefaultMaxHandshakeRate,
 	}
 )
 
 type HandshakeConfig struct {
-	tryInterval   time.Duration
-	retries       int64
-	triggerBuffer int
-	useRelays     bool
+	tryInterval      time.Duration
+	retries          int64
+	triggerBuffer    int
+	useRelays        bool
+	maxHandshakeRate int
 
 	messageMetrics *MessageMetrics
 }
@@ -58,9 +61,15 @@ type HandshakeManager struct {
 	messageMetrics         *MessageMetrics
 	metricInitiated        metrics.Counter
 	metricTimedOut         metrics.Counter
+	metricRateLimited      metrics.Counter
 	f                      *Interface
 	l                      *logrus.Logger
 
+	// Rate limiting for new handshakes (token bucket)
+	rateBucket   int // tokens currently available
+	rateMax      int // max tokens (== max handshakes per second), 0 means unlimited
+	rateLastTick time.Time
+
 	// can be used to trigger outbound handshake for the given vpnIp
 	trigger chan netip.Addr
 }
@@ -116,10 +125,41 @@ func NewHandshakeManager(l *logrus.Logger, mainHostMap *HostMap, lightHouse *Lig
 		messageMetrics:         config.messageMetrics,
 		metricInitiated:        metrics.GetOrRegisterCounter("handshake_manager.initiated", nil),
 		metricTimedOut:         metrics.GetOrRegisterCounter("handshake_manager.timed_out", nil),
+		metricRateLimited:      metrics.GetOrRegisterCounter("handshake_manager.rate_limited", nil),
+		rateBucket:             config.maxHandshakeRate,
+		rateMax:                config.maxHandshakeRate,
+		rateLastTick:           time.Now(),
 		l:                      l,
 	}
 }
 
+// handshakeRateAllow checks the token bucket rate limiter and returns true if a
+// new handshake is allowed. Must be called with hm.Lock held.
+func (hm *HandshakeManager) handshakeRateAllow(now time.Time) bool {
+	if hm.rateMax == 0 {
+		return true
+	}
+
+	// Refill tokens based on elapsed time
+	elapsed := now.Sub(hm.rateLastTick)
+	if elapsed >= time.Second {
+		// Add tokens for full seconds elapsed
+		tokens := int(elapsed/time.Second) * hm.rateMax
+		hm.rateBucket += tokens
+		if hm.rateBucket > hm.rateMax {
+			hm.rateBucket = hm.rateMax
+		}
+		hm.rateLastTick = now
+	}
+
+	if hm.rateBucket > 0 {
+		hm.rateBucket--
+		return true
+	}
+
+	return false
+}
+
 func (hm *HandshakeManager) Run(ctx context.Context) {
 	clockSource := time.NewTicker(hm.config.tryInterval)
 	defer clockSource.Stop()
@@ -149,6 +189,15 @@ func (hm *HandshakeManager) HandleIncoming(via ViaSender, packet []byte, h *head
 	case header.HandshakeIXPSK0:
 		switch h.MessageCounter {
 		case 1:
+			// Check rate limit for new incoming handshakes
+			hm.Lock()
+			allowed := hm.handshakeRateAllow(time.Now())
+			hm.Unlock()
+			if !allowed {
+				hm.metricRateLimited.Inc(1)
+				hm.l.WithField("from", via).Debug("Handshake rate limit reached, dropping incoming handshake")
+				return
+			}
 			ixHandshakeStage1(hm.f, via, packet, h)
 
 		case 2:

handshake_manager_test.go

@@ -65,6 +65,68 @@ func Test_NewHandshakeManagerVpnIp(t *testing.T) {
 	assert.NotContains(t, blah.vpnIps, ip)
 }
 
+func Test_HandshakeManagerRateLimit(t *testing.T) {
+	l := test.NewLogger()
+	localrange := netip.MustParsePrefix("10.1.1.1/24")
+	preferredRanges := []netip.Prefix{localrange}
+	mainHM := newHostMap(l)
+	mainHM.preferredRanges.Store(&preferredRanges)
+
+	lh := newTestLighthouse()
+
+	config := defaultHandshakeConfig
+	config.maxHandshakeRate = 2
+
+	hm := NewHandshakeManager(l, mainHM, lh, &udp.NoopConn{}, config)
+	hm.f = &Interface{handshakeManager: hm, pki: &PKI{}, l: l}
+
+	now := time.Now()
+
+	// Should allow up to maxHandshakeRate handshakes
+	hm.Lock()
+	assert.True(t, hm.handshakeRateAllow(now), "first handshake should be allowed")
+	assert.True(t, hm.handshakeRateAllow(now), "second handshake should be allowed")
+	assert.False(t, hm.handshakeRateAllow(now), "third handshake should be rate limited")
+	hm.Unlock()
+
+	// After advancing time by 1 second, tokens should refill
+	hm.Lock()
+	assert.True(t, hm.handshakeRateAllow(now.Add(time.Second)), "handshake should be allowed after token refill")
+	hm.Unlock()
+}
+
+func Test_HandshakeManagerRateLimitUnlimited(t *testing.T) {
+	l := test.NewLogger()
+	localrange := netip.MustParsePrefix("10.1.1.1/24")
+	preferredRanges := []netip.Prefix{localrange}
+	mainHM := newHostMap(l)
+	mainHM.preferredRanges.Store(&preferredRanges)
+
+	lh := newTestLighthouse()
+
+	cs := &CertState{
+		initiatingVersion: cert.Version1,
+		privateKey:        []byte{},
+		v1Cert:            &dummyCert{version: cert.Version1},
+		v1HandshakeBytes:  []byte{},
+	}
+
+	// Default config has maxHandshakeRate=0 (unlimited)
+	hm := NewHandshakeManager(l, mainHM, lh, &udp.NoopConn{}, defaultHandshakeConfig)
+	hm.f = &Interface{handshakeManager: hm, pki: &PKI{}, l: l}
+	hm.f.pki.cs.Store(cs)
+
+	// Should allow many handshakes with no limit
+	// Limited to 10 due to test lighthouse query channel buffer
+	for i := 0; i < 10; i++ {
+		ip := netip.MustParseAddr("172.1.1.1").As16()
+		ip[15] = byte(i + 1)
+		addr := netip.AddrFrom16(ip)
+		h := hm.StartHandshake(addr, nil)
+		assert.NotNil(t, h, "handshake %d should be allowed with unlimited rate", i)
+	}
+}
+
 func testCountTimerWheelEntries(tw *LockingTimerWheel[netip.Addr]) (c int) {
 	for _, i := range tw.t.wheel {
 		n := i.Head

interface.go

@@ -7,6 +7,7 @@ import (
 	"io"
 	"net/netip"
 	"os"
+	"runtime"
 	"sync/atomic"
 	"time"
 
@@ -262,6 +263,8 @@ func (f *Interface) run() {
 }
 
 func (f *Interface) listenOut(i int) {
+	runtime.LockOSThread()
+
 	var li udp.Conn
 	if i > 0 {
 		li = f.writers[i]
@@ -282,6 +285,8 @@ func (f *Interface) listenOut(i int) {
 }
 
 func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
+	runtime.LockOSThread()
+
 	packet := make([]byte, mtu)
 	out := make([]byte, mtu)
 	fwPacket := &firewall.Packet{}

main.go

@@ -204,10 +204,11 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	useRelays := c.GetBool("relay.use_relays", DefaultUseRelays) && !c.GetBool("relay.am_relay", false)
 
 	handshakeConfig := HandshakeConfig{
-		tryInterval:   c.GetDuration("handshakes.try_interval", DefaultHandshakeTryInterval),
-		retries:       int64(c.GetInt("handshakes.retries", DefaultHandshakeRetries)),
-		triggerBuffer: c.GetInt("handshakes.trigger_buffer", DefaultHandshakeTriggerBuffer),
-		useRelays:     useRelays,
+		tryInterval:      c.GetDuration("handshakes.try_interval", DefaultHandshakeTryInterval),
+		retries:          int64(c.GetInt("handshakes.retries", DefaultHandshakeRetries)),
+		triggerBuffer:    c.GetInt("handshakes.trigger_buffer", DefaultHandshakeTriggerBuffer),
+		useRelays:        useRelays,
+		maxHandshakeRate: c.GetInt("handshakes.max_rate", DefaultMaxHandshakeRate),
 
 		messageMetrics: messageMetrics,
 	}

overlay/tun_linux.go

@@ -26,6 +26,7 @@ import (
 
 type tun struct {
 	io.ReadWriteCloser
+	fd          int
 	Device      string
 	vpnNetworks []netip.Prefix
 	MaxMTU      int
@@ -72,11 +73,6 @@ type ifreqQLEN struct {
 
 func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
 	file := os.NewFile(uintptr(deviceFd), "/dev/net/tun")
-	err := unix.SetNonblock(deviceFd, true)
-	if err != nil {
-		_ = file.Close()
-		return nil, err
-	}
 
 	t, err := newTunGeneric(c, l, file, vpnNetworks)
 	if err != nil {
@@ -89,7 +85,7 @@ func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []net
 }
 
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueue bool) (*tun, error) {
-	fd, err := unix.Open("/dev/net/tun", os.O_RDWR|unix.O_NONBLOCK, 0)
+	fd, err := unix.Open("/dev/net/tun", os.O_RDWR, 0)
 	if err != nil {
 		// If /dev/net/tun doesn't exist, try to create it (will happen in docker)
 		if os.IsNotExist(err) {
@@ -102,7 +98,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 				return nil, fmt.Errorf("failed to create /dev/net/tun: %w", err)
 			}
 
-			fd, err = unix.Open("/dev/net/tun", os.O_RDWR|unix.O_NONBLOCK, 0)
+			fd, err = unix.Open("/dev/net/tun", os.O_RDWR, 0)
 			if err != nil {
 				return nil, fmt.Errorf("created /dev/net/tun, but still failed: %w", err)
 			}
@@ -140,6 +136,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 func newTunGeneric(c *config.C, l *logrus.Logger, file *os.File, vpnNetworks []netip.Prefix) (*tun, error) {
 	t := &tun{
 		ReadWriteCloser:           file,
+		fd:                        int(file.Fd()),
 		vpnNetworks:               vpnNetworks,
 		TXQueueLen:                c.GetInt("tun.tx_queue", 500),
 		useSystemRoutes:           c.GetBool("tun.use_system_route_table", false),
@@ -242,7 +239,7 @@ func (t *tun) SupportsMultiqueue() bool {
 }
 
 func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
-	fd, err := unix.Open("/dev/net/tun", os.O_RDWR|unix.O_NONBLOCK, 0)
+	fd, err := unix.Open("/dev/net/tun", os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
@@ -264,6 +261,29 @@ func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	return r
 }
 
+func (t *tun) Write(b []byte) (int, error) {
+	var nn int
+	maximum := len(b)
+
+	for {
+		n, err := unix.Write(t.fd, b[nn:maximum])
+		if n > 0 {
+			nn += n
+		}
+		if nn == len(b) {
+			return nn, err
+		}
+
+		if err != nil {
+			return nn, err
+		}
+
+		if n == 0 {
+			return nn, io.ErrUnexpectedEOF
+		}
+	}
+}
+
 func (t *tun) deviceBytes() (o [16]byte) {
 	for i, c := range t.Device {
 		o[i] = byte(c)

pki.go

@@ -5,7 +5,6 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"io"
 	"net"
 	"net/netip"
 	"os"
@@ -488,25 +487,25 @@ func loadCertificate(b []byte) (cert.Certificate, []byte, error) {
 }
 
 func loadCAPoolFromConfig(l *logrus.Logger, c *config.C) (*cert.CAPool, error) {
+	var rawCA []byte
+	var err error
+
 	caPathOrPEM := c.GetString("pki.ca", "")
 	if caPathOrPEM == "" {
 		return nil, errors.New("no pki.ca path or PEM data provided")
 	}
 
-	var caReader io.ReadCloser
-	var err error
-
 	if strings.Contains(caPathOrPEM, "-----BEGIN") {
-		caReader = io.NopCloser(strings.NewReader(caPathOrPEM))
+		rawCA = []byte(caPathOrPEM)
+
 	} else {
-		caReader, err = os.Open(caPathOrPEM)
+		rawCA, err = os.ReadFile(caPathOrPEM)
 		if err != nil {
 			return nil, fmt.Errorf("unable to read pki.ca file %s: %s", caPathOrPEM, err)
 		}
 	}
-	defer caReader.Close()
 
-	caPool, err := cert.NewCAPoolFromPEMReader(caReader)
+	caPool, err := cert.NewCAPoolFromPEM(rawCA)
 	if errors.Is(err, cert.ErrExpired) {
 		var expired int
 		for _, crt := range caPool.CAs {

pki_hup_benchmark_test.go

@@ -1,121 +0,0 @@
-package nebula
-
-import (
-	"bytes"
-	"fmt"
-	"net/netip"
-	"os"
-	"path/filepath"
-	"runtime"
-	"testing"
-	"time"
-
-	"github.com/slackhq/nebula/cert"
-	cert_test "github.com/slackhq/nebula/cert_test"
-	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/test"
-	"github.com/stretchr/testify/require"
-)
-
-func BenchmarkReloadConfigWithCAs(b *testing.B) {
-	prevProcs := runtime.GOMAXPROCS(1)
-	b.Cleanup(func() { runtime.GOMAXPROCS(prevProcs) })
-
-	for _, size := range []int{100, 250, 500, 1000, 5000} {
-		b.Run(fmt.Sprintf("%dCAs", size), func(b *testing.B) {
-			l := test.NewLogger()
-			dir := b.TempDir()
-
-			ca, caKey, caBundle := buildCABundle(b, size)
-			caPath, certPath, keyPath := writePKIFiles(b, dir, ca, caKey, caBundle)
-
-			configBody := fmt.Sprintf(`pki:
-  ca: %s
-  cert: %s
-  key: %s
-`, caPath, certPath, keyPath)
-
-			configPath := filepath.Join(dir, "config.yml")
-			require.NoError(b, os.WriteFile(configPath, []byte(configBody), 0o600))
-
-			c := config.NewC(l)
-			require.NoError(b, c.Load(dir))
-
-			_, err := NewPKIFromConfig(l, c)
-			require.NoError(b, err)
-
-			b.ReportAllocs()
-			b.ResetTimer()
-
-			for b.Loop() {
-				c.ReloadConfig()
-			}
-		})
-	}
-}
-
-func buildCABundle(b *testing.B, count int) (cert.Certificate, []byte, []byte) {
-	b.Helper()
-	require.GreaterOrEqual(b, count, 1)
-
-	before := time.Now().Add(-24 * time.Hour)
-	after := time.Now().Add(24 * time.Hour)
-
-	ca, _, caKey, pem := cert_test.NewTestCaCert(
-		cert.Version2,
-		cert.Curve_CURVE25519,
-		before,
-		after,
-		nil,
-		nil,
-		nil,
-	)
-
-	buf := bytes.NewBuffer(pem)
-	buf.Write([]byte("\n# a comment!\n"))
-
-	for i := 1; i < count; i++ {
-		_, _, _, extraPEM := cert_test.NewTestCaCert(
-			cert.Version2,
-			cert.Curve_CURVE25519,
-			time.Now(),
-			time.Now().Add(time.Hour),
-			nil,
-			nil,
-			nil,
-		)
-		buf.Write([]byte("\n# a comment!\n"))
-		buf.Write(extraPEM)
-	}
-
-	return ca, caKey, buf.Bytes()
-}
-
-func writePKIFiles(b *testing.B, dir string, ca cert.Certificate, caKey []byte, caBundle []byte) (string, string, string) {
-	b.Helper()
-
-	networks := []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")}
-
-	_, _, keyPEM, certPEM := cert_test.NewTestCert(
-		cert.Version2,
-		cert.Curve_CURVE25519,
-		ca,
-		caKey,
-		"reload-benchmark",
-		time.Now(),
-		time.Now().Add(time.Hour),
-		networks,
-		nil,
-		nil,
-	)
-
-	caPath := filepath.Join(dir, "ca.pem")
-	certPath := filepath.Join(dir, "cert.pem")
-	keyPath := filepath.Join(dir, "key.pem")
-
-	require.NoError(b, os.WriteFile(caPath, caBundle, 0o600))
-	require.NoError(b, os.WriteFile(certPath, certPEM, 0o600))
-	require.NoError(b, os.WriteFile(keyPath, keyPEM, 0o600))
-
-	return caPath, certPath, keyPath
-}

ssh.go

@@ -10,7 +10,6 @@ import (
 	"net"
 	"net/netip"
 	"os"
-	"path/filepath"
 	"reflect"
 	"runtime"
 	"runtime/pprof"
@@ -189,12 +188,6 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 }
 
 func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Interface) {
-	// sandboxDir defaults to a dir in temp. The intention is that end user will
-	// create this dir as needed. Overriding this config value to "" allows
-	// writing to anywhere in the system.
-	defaultDir := filepath.Join(os.TempDir(), "nebula-debug")
-	sandboxDir := c.GetString("sshd.sandbox_dir", defaultDir)
-
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "list-hostmap",
 		ShortDescription: "List all known previously connected hosts",
@@ -253,9 +246,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "start-cpu-profile",
 		ShortDescription: "Starts a cpu profile and write output to the provided file, ex: `cpu-profile.pb.gz`",
-		Callback: func(fs any, a []string, w sshd.StringWriter) error {
-			return sshStartCpuProfile(sandboxDir, fs, a, w)
-		},
+		Callback:         sshStartCpuProfile,
 	})
 
 	ssh.RegisterCommand(&sshd.Command{
@@ -270,9 +261,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "save-heap-profile",
 		ShortDescription: "Saves a heap profile to the provided path, ex: `heap-profile.pb.gz`",
-		Callback: func(fs any, a []string, w sshd.StringWriter) error {
-			return sshGetHeapProfile(sandboxDir, fs, a, w)
-		},
+		Callback:         sshGetHeapProfile,
 	})
 
 	ssh.RegisterCommand(&sshd.Command{
@@ -284,9 +273,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "save-mutex-profile",
 		ShortDescription: "Saves a mutex profile to the provided path, ex: `mutex-profile.pb.gz`",
-		Callback: func(fs any, a []string, w sshd.StringWriter) error {
-			return sshGetMutexProfile(sandboxDir, fs, a, w)
-		},
+		Callback:         sshGetMutexProfile,
 	})
 
 	ssh.RegisterCommand(&sshd.Command{
@@ -519,43 +506,13 @@ func sshListLighthouseMap(lightHouse *LightHouse, a any, w sshd.StringWriter) er
 	return nil
 }
 
-// sshSanitizeFilePath validates that the given file path is within the sandbox directory.
-// If sandboxDir is empty, the path is returned as-is for backwards compatibility.
-func sshSanitizeFilePath(sandboxDir, filePath string) (string, error) {
-	if sandboxDir == "" {
-		return filePath, nil
-	}
-
-	// Clean and resolve the path relative to the sandbox directory
-	if !filepath.IsAbs(filePath) {
-		filePath = filepath.Join(sandboxDir, filePath)
-	}
-	cleaned := filepath.Clean(filePath)
-
-	// Ensure the resolved path is within the sandbox directory
-	cleanedSandbox := filepath.Clean(sandboxDir)
-	if cleaned == cleanedSandbox {
-		return "", fmt.Errorf("path %q resolves to the sandbox directory itself %q", filePath, sandboxDir)
-	}
-	if !strings.HasPrefix(cleaned, cleanedSandbox+string(filepath.Separator)) {
-		return "", fmt.Errorf("path %q is outside the sandbox directory %q", filePath, sandboxDir)
-	}
-
-	return cleaned, nil
-}
-
-func sshStartCpuProfile(sandboxDir string, fs any, a []string, w sshd.StringWriter) error {
+func sshStartCpuProfile(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		err := w.WriteLine("No path to write profile provided")
 		return err
 	}
 
-	filePath, err := sshSanitizeFilePath(sandboxDir, a[0])
-	if err != nil {
-		return w.WriteLine(err.Error())
-	}
-
-	file, err := os.Create(filePath)
+	file, err := os.Create(a[0])
 	if err != nil {
 		err = w.WriteLine(fmt.Sprintf("Unable to create profile file: %s", err))
 		return err
@@ -675,6 +632,9 @@ func sshCreateTunnel(ifce *Interface, fs any, a []string, w sshd.StringWriter) e
 	}
 
 	hostInfo = ifce.handshakeManager.StartHandshake(vpnAddr, nil)
+	if hostInfo == nil {
+		return w.WriteLine("Handshake rate limit reached")
+	}
 	if addr.IsValid() {
 		hostInfo.SetRemote(addr)
 	}
@@ -719,17 +679,12 @@ func sshChangeRemote(ifce *Interface, fs any, a []string, w sshd.StringWriter) e
 	return w.WriteLine("Changed")
 }
 
-func sshGetHeapProfile(sandboxDir string, fs any, a []string, w sshd.StringWriter) error {
+func sshGetHeapProfile(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine("No path to write profile provided")
 	}
 
-	filePath, err := sshSanitizeFilePath(sandboxDir, a[0])
-	if err != nil {
-		return w.WriteLine(err.Error())
-	}
-
-	file, err := os.Create(filePath)
+	file, err := os.Create(a[0])
 	if err != nil {
 		err = w.WriteLine(fmt.Sprintf("Unable to create profile file: %s", err))
 		return err
@@ -760,17 +715,12 @@ func sshMutexProfileFraction(fs any, a []string, w sshd.StringWriter) error {
 	return w.WriteLine(fmt.Sprintf("New value: %d. Old value: %d", newRate, oldRate))
 }
 
-func sshGetMutexProfile(sandboxDir string, fs any, a []string, w sshd.StringWriter) error {
+func sshGetMutexProfile(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine("No path to write profile provided")
 	}
 
-	filePath, err := sshSanitizeFilePath(sandboxDir, a[0])
-	if err != nil {
-		return w.WriteLine(err.Error())
-	}
-
-	file, err := os.Create(filePath)
+	file, err := os.Create(a[0])
 	if err != nil {
 		return w.WriteLine(fmt.Sprintf("Unable to create profile file: %s", err))
 	}

udp/udp_linux.go

@@ -4,7 +4,6 @@
 package udp
 
 import (
-	"context"
 	"encoding/binary"
 	"fmt"
 	"net"
@@ -19,58 +18,58 @@ import (
 )
 
 type StdConn struct {
-	udpConn *net.UDPConn
-	rawConn syscall.RawConn
-	isV4    bool
-	l       *logrus.Logger
-	batch   int
+	sysFd int
+	isV4  bool
+	l     *logrus.Logger
+	batch int
 }
 
-func setReusePort(network, address string, c syscall.RawConn) error {
-	var opErr error
-	err := c.Control(func(fd uintptr) {
-		opErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, unix.SO_REUSEPORT, 1)
-		//CloseOnExec already set by the runtime
-	})
-	if err != nil {
-		return err
+func maybeIPV4(ip net.IP) (net.IP, bool) {
+	ip4 := ip.To4()
+	if ip4 != nil {
+		return ip4, true
 	}
-	return opErr
+	return ip, false
 }
 
 func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
-	listen := netip.AddrPortFrom(ip, uint16(port))
-	lc := net.ListenConfig{}
-	if multi {
-		lc.Control = setReusePort
+	af := unix.AF_INET6
+	if ip.Is4() {
+		af = unix.AF_INET
 	}
-	//this context is only used during the bind operation, you can't cancel it to kill the socket
-	pc, err := lc.ListenPacket(context.Background(), "udp", listen.String())
+	syscall.ForkLock.RLock()
+	fd, err := unix.Socket(af, unix.SOCK_DGRAM, unix.IPPROTO_UDP)
+	if err == nil {
+		unix.CloseOnExec(fd)
+	}
+	syscall.ForkLock.RUnlock()
+
 	if err != nil {
+		unix.Close(fd)
 		return nil, fmt.Errorf("unable to open socket: %s", err)
 	}
-	udpConn := pc.(*net.UDPConn)
-	rawConn, err := udpConn.SyscallConn()
-	if err != nil {
-		_ = udpConn.Close()
-		return nil, err
-	}
-	//gotta find out if we got an AF_INET6 socket or not:
-	out := &StdConn{
-		udpConn: udpConn,
-		rawConn: rawConn,
-		l:       l,
-		batch:   batch,
+
+	if multi {
+		if err = unix.SetsockoptInt(fd, unix.SOL_SOCKET, unix.SO_REUSEPORT, 1); err != nil {
+			return nil, fmt.Errorf("unable to set SO_REUSEPORT: %s", err)
+		}
 	}
 
-	af, err := out.getSockOptInt(unix.SO_DOMAIN)
-	if err != nil {
-		_ = out.Close()
-		return nil, err
+	var sa unix.Sockaddr
+	if ip.Is4() {
+		sa4 := &unix.SockaddrInet4{Port: port}
+		sa4.Addr = ip.As4()
+		sa = sa4
+	} else {
+		sa6 := &unix.SockaddrInet6{Port: port}
+		sa6.Addr = ip.As16()
+		sa = sa6
+	}
+	if err = unix.Bind(fd, sa); err != nil {
+		return nil, fmt.Errorf("unable to bind to socket: %s", err)
 	}
-	out.isV4 = af == unix.AF_INET
 
-	return out, nil
+	return &StdConn{sysFd: fd, isV4: ip.Is4(), l: l, batch: batch}, err
 }
 
 func (u *StdConn) SupportsMultipleReaders() bool {
@@ -81,137 +80,63 @@ func (u *StdConn) Rebind() error {
 	return nil
 }
 
-func (u *StdConn) getSockOptInt(opt int) (int, error) {
-	if u.rawConn == nil {
-		return 0, fmt.Errorf("no UDP connection")
-	}
-	var out int
-	var opErr error
-	err := u.rawConn.Control(func(fd uintptr) {
-		out, opErr = unix.GetsockoptInt(int(fd), unix.SOL_SOCKET, opt)
-	})
-	if err != nil {
-		return 0, err
-	}
-	return out, opErr
-}
-
-func (u *StdConn) setSockOptInt(opt int, n int) error {
-	if u.rawConn == nil {
-		return fmt.Errorf("no UDP connection")
-	}
-	var opErr error
-	err := u.rawConn.Control(func(fd uintptr) {
-		opErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, opt, n)
-	})
-	if err != nil {
-		return err
-	}
-	return opErr
-}
-
 func (u *StdConn) SetRecvBuffer(n int) error {
-	return u.setSockOptInt(unix.SO_RCVBUFFORCE, n)
+	return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_RCVBUFFORCE, n)
 }
 
 func (u *StdConn) SetSendBuffer(n int) error {
-	return u.setSockOptInt(unix.SO_SNDBUFFORCE, n)
+	return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_SNDBUFFORCE, n)
 }
 
 func (u *StdConn) SetSoMark(mark int) error {
-	return u.setSockOptInt(unix.SO_MARK, mark)
+	return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_MARK, mark)
 }
 
 func (u *StdConn) GetRecvBuffer() (int, error) {
-	return u.getSockOptInt(unix.SO_RCVBUF)
+	return unix.GetsockoptInt(int(u.sysFd), unix.SOL_SOCKET, unix.SO_RCVBUF)
 }
 
 func (u *StdConn) GetSendBuffer() (int, error) {
-	return u.getSockOptInt(unix.SO_SNDBUF)
+	return unix.GetsockoptInt(int(u.sysFd), unix.SOL_SOCKET, unix.SO_SNDBUF)
 }
 
 func (u *StdConn) GetSoMark() (int, error) {
-	return u.getSockOptInt(unix.SO_MARK)
+	return unix.GetsockoptInt(int(u.sysFd), unix.SOL_SOCKET, unix.SO_MARK)
 }
 
 func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
-	a := u.udpConn.LocalAddr()
+	sa, err := unix.Getsockname(u.sysFd)
+	if err != nil {
+		return netip.AddrPort{}, err
+	}
 
-	switch v := a.(type) {
-	case *net.UDPAddr:
-		addr, ok := netip.AddrFromSlice(v.IP)
-		if !ok {
-			return netip.AddrPort{}, fmt.Errorf("LocalAddr returned invalid IP address: %s", v.IP)
-		}
-		return netip.AddrPortFrom(addr, uint16(v.Port)), nil
+	switch sa := sa.(type) {
+	case *unix.SockaddrInet4:
+		return netip.AddrPortFrom(netip.AddrFrom4(sa.Addr), uint16(sa.Port)), nil
+
+	case *unix.SockaddrInet6:
+		return netip.AddrPortFrom(netip.AddrFrom16(sa.Addr), uint16(sa.Port)), nil
 
 	default:
-		return netip.AddrPort{}, fmt.Errorf("LocalAddr returned: %#v", a)
+		return netip.AddrPort{}, fmt.Errorf("unsupported sock type: %T", sa)
 	}
 }
 
-func recvmmsg(fd uintptr, msgs []rawMessage) (int, bool, error) {
-	var errno syscall.Errno
-	n, _, errno := unix.Syscall6(
-		unix.SYS_RECVMMSG,
-		fd,
-		uintptr(unsafe.Pointer(&msgs[0])),
-		uintptr(len(msgs)),
-		unix.MSG_WAITFORONE,
-		0,
-		0,
-	)
-	if errno == syscall.EAGAIN || errno == syscall.EWOULDBLOCK {
-		// No data available, block for I/O and try again.
-		return int(n), false, nil
-	}
-	if errno != 0 {
-		return int(n), true, &net.OpError{Op: "recvmmsg", Err: errno}
-	}
-	return int(n), true, nil
-}
-
-func (u *StdConn) listenOutSingle(r EncReader) {
-	var err error
-	var n int
-	var from netip.AddrPort
-	buffer := make([]byte, MTU)
-
-	for {
-		n, from, err = u.udpConn.ReadFromUDPAddrPort(buffer)
-		if err != nil {
-			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
-			return
-		}
-		from = netip.AddrPortFrom(from.Addr().Unmap(), from.Port())
-		r(from, buffer[:n])
-	}
-}
-
-func (u *StdConn) listenOutBatch(r EncReader) {
+func (u *StdConn) ListenOut(r EncReader) {
 	var ip netip.Addr
-	var n int
-	var operr error
 
 	msgs, buffers, names := u.PrepareRawMessages(u.batch)
-
-	//reader needs to capture variables from this function, since it's used as a lambda with rawConn.Read
-	//defining it outside the loop so it gets re-used
-	reader := func(fd uintptr) (done bool) {
-		n, done, operr = recvmmsg(fd, msgs)
-		return done
+	read := u.ReadMulti
+	if u.batch == 1 {
+		read = u.ReadSingle
 	}
 
 	for {
-		err := u.rawConn.Read(reader)
+		n, err := read(msgs)
 		if err != nil {
 			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
 			return
 		}
-		if operr != nil {
-			u.l.WithError(operr).Debug("operr: udp socket is closed, exiting read loop")
-			return
-		}
 
 		for i := 0; i < n; i++ {
 			// Its ok to skip the ok check here, the slicing is the only error that can occur and it will panic
@@ -225,20 +150,106 @@ func (u *StdConn) listenOutBatch(r EncReader) {
 	}
 }
 
-func (u *StdConn) ListenOut(r EncReader) {
-	if u.batch == 1 {
-		//save some ram by not calling PrepareRawMessages for fields we won't use
-		//we could also make this path more common by calling recvmmsg with msgs[:1],
-		//but that's still the recvmmsg syscall, which would be a change
-		u.listenOutSingle(r)
-	} else {
-		u.listenOutBatch(r)
+func (u *StdConn) ReadSingle(msgs []rawMessage) (int, error) {
+	for {
+		n, _, err := unix.Syscall6(
+			unix.SYS_RECVMSG,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&(msgs[0].Hdr))),
+			0,
+			0,
+			0,
+			0,
+		)
+
+		if err != 0 {
+			return 0, &net.OpError{Op: "recvmsg", Err: err}
+		}
+
+		msgs[0].Len = uint32(n)
+		return 1, nil
+	}
+}
+
+func (u *StdConn) ReadMulti(msgs []rawMessage) (int, error) {
+	for {
+		n, _, err := unix.Syscall6(
+			unix.SYS_RECVMMSG,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&msgs[0])),
+			uintptr(len(msgs)),
+			unix.MSG_WAITFORONE,
+			0,
+			0,
+		)
+
+		if err != 0 {
+			return 0, &net.OpError{Op: "recvmmsg", Err: err}
+		}
+
+		return int(n), nil
 	}
 }
 
 func (u *StdConn) WriteTo(b []byte, ip netip.AddrPort) error {
-	_, err := u.udpConn.WriteToUDPAddrPort(b, ip)
-	return err
+	if u.isV4 {
+		return u.writeTo4(b, ip)
+	}
+	return u.writeTo6(b, ip)
+}
+
+func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error {
+	var rsa unix.RawSockaddrInet6
+	rsa.Family = unix.AF_INET6
+	rsa.Addr = ip.Addr().As16()
+	binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], ip.Port())
+
+	for {
+		_, _, err := unix.Syscall6(
+			unix.SYS_SENDTO,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&b[0])),
+			uintptr(len(b)),
+			uintptr(0),
+			uintptr(unsafe.Pointer(&rsa)),
+			uintptr(unix.SizeofSockaddrInet6),
+		)
+
+		if err != 0 {
+			return &net.OpError{Op: "sendto", Err: err}
+		}
+
+		return nil
+	}
+}
+
+func (u *StdConn) writeTo4(b []byte, ip netip.AddrPort) error {
+	if !ip.Addr().Is4() {
+		return ErrInvalidIPv6RemoteForSocket
+	}
+
+	var rsa unix.RawSockaddrInet4
+	rsa.Family = unix.AF_INET
+	rsa.Addr = ip.Addr().As4()
+	binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], ip.Port())
+
+	for {
+		_, _, err := unix.Syscall6(
+			unix.SYS_SENDTO,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&b[0])),
+			uintptr(len(b)),
+			uintptr(0),
+			uintptr(unsafe.Pointer(&rsa)),
+			uintptr(unix.SizeofSockaddrInet4),
+		)
+
+		if err != 0 {
+			return &net.OpError{Op: "sendto", Err: err}
+		}
+
+		return nil
+	}
 }
 
 func (u *StdConn) ReloadConfig(c *config.C) {
@@ -291,28 +302,15 @@ func (u *StdConn) ReloadConfig(c *config.C) {
 
 func (u *StdConn) getMemInfo(meminfo *[unix.SK_MEMINFO_VARS]uint32) error {
 	var vallen uint32 = 4 * unix.SK_MEMINFO_VARS
-
-	if u.rawConn == nil {
-		return fmt.Errorf("no UDP connection")
-	}
-	var opErr error
-	err := u.rawConn.Control(func(fd uintptr) {
-		_, _, syserr := unix.Syscall6(unix.SYS_GETSOCKOPT, fd, uintptr(unix.SOL_SOCKET), uintptr(unix.SO_MEMINFO), uintptr(unsafe.Pointer(meminfo)), uintptr(unsafe.Pointer(&vallen)), 0)
-		if syserr != 0 {
-			opErr = syserr
-		}
-	})
-	if err != nil {
+	_, _, 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 opErr
+	return nil
 }
 
 func (u *StdConn) Close() error {
-	if u.udpConn != nil {
-		return u.udpConn.Close()
-	}
-	return nil
+	return syscall.Close(u.sysFd)
 }
 
 func NewUDPStatsEmitter(udpConns []Conn) func() {