claude does TUN virtio header support

.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}" \
+            --title "Release ${{ github.ref_name }}" \
-            "${GITHUB_REF_NAME}" \
+            "${{ github.ref_name }}" \
             SHASUM256.txt *-latest/*.zip *-latest/*.tar.gz

.github/workflows/smoke-extra.yml

@@ -30,9 +30,6 @@ jobs:
     - name: add hashicorp source
       run: wget -O- https://apt.releases.hashicorp.com/gpg | gpg --dearmor | sudo tee /usr/share/keyrings/hashicorp-archive-keyring.gpg && echo "deb [signed-by=/usr/share/keyrings/hashicorp-archive-keyring.gpg] https://apt.releases.hashicorp.com $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/hashicorp.list
 
-    - name: workaround AMD-V issue  # https://github.com/cri-o/packaging/pull/306
-      run: sudo rmmod kvm_amd
-
     - name: install vagrant
       run: sudo apt-get update && sudo apt-get install -y vagrant virtualbox
 

.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

@@ -37,18 +37,17 @@ docker run --name host4 --device /dev/net/tun:/dev/net/tun --cap-add NET_ADMIN -
 sleep 1
 
 # grab tcpdump pcaps for debugging
-docker exec lighthouse1 tcpdump -i tun0 -q -w - -U 2>logs/lighthouse1.inside.log >logs/lighthouse1.inside.pcap &
+docker exec lighthouse1 tcpdump -i nebula1 -q -w - -U 2>logs/lighthouse1.inside.log >logs/lighthouse1.inside.pcap &
 docker exec lighthouse1 tcpdump -i eth0 -q -w - -U 2>logs/lighthouse1.outside.log >logs/lighthouse1.outside.pcap &
-docker exec host2 tcpdump -i tun0 -q -w - -U 2>logs/host2.inside.log >logs/host2.inside.pcap &
+docker exec host2 tcpdump -i nebula1 -q -w - -U 2>logs/host2.inside.log >logs/host2.inside.pcap &
 docker exec host2 tcpdump -i eth0 -q -w - -U 2>logs/host2.outside.log >logs/host2.outside.pcap &
-docker exec host3 tcpdump -i tun0 -q -w - -U 2>logs/host3.inside.log >logs/host3.inside.pcap &
+docker exec host3 tcpdump -i nebula1 -q -w - -U 2>logs/host3.inside.log >logs/host3.inside.pcap &
 docker exec host3 tcpdump -i eth0 -q -w - -U 2>logs/host3.outside.log >logs/host3.outside.pcap &
-docker exec host4 tcpdump -i tun0 -q -w - -U 2>logs/host4.inside.log >logs/host4.inside.pcap &
+docker exec host4 tcpdump -i nebula1 -q -w - -U 2>logs/host4.inside.log >logs/host4.inside.pcap &
 docker exec host4 tcpdump -i eth0 -q -w - -U 2>logs/host4.outside.log >logs/host4.outside.pcap &
 
 docker exec host2 ncat -nklv 0.0.0.0 2000 &
 docker exec host3 ncat -nklv 0.0.0.0 2000 &
-docker exec host4 ncat -nkluv 0.0.0.0 4000 &
 docker exec host2 ncat -e '/usr/bin/echo host2' -nkluv 0.0.0.0 3000 &
 docker exec host3 ncat -e '/usr/bin/echo host3' -nkluv 0.0.0.0 3000 &
 
@@ -120,12 +119,11 @@ echo
 echo " *** Testing conntrack"
 echo
 set -x
-
+# host2 can ping host3 now that host3 pinged it first
-# host2 speaking to host4 on UDP 4000 should allow it to reply, when firewall rules would normally not permit this
+docker exec host2 ping -c1 192.168.100.3
-docker exec host2 sh -c "/usr/bin/echo host2 | ncat -nuv 192.168.100.4 4000"
+# host4 can ping host2 once conntrack established
-docker exec host2 ncat -e '/usr/bin/echo helloagainfromhost2' -nkluv 0.0.0.0 4000 &
+docker exec host2 ping -c1 192.168.100.4
-sleep 1
+docker exec host4 ping -c1 192.168.100.2
-docker exec host4 sh -c "/usr/bin/echo host4 | ncat -nuv 192.168.100.2 4000"
 
 docker exec host4 sh -c 'kill 1'
 docker exec host3 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

CHANGELOG.md

@@ -7,19 +7,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-## [1.10.3] - 2026-02-06
-
-### Security
-
-- Fix an issue where blocklist bypass is possible when using curve P256 since the signature can have 2 valid representations.
-  Both fingerprint representations will be tested against the blocklist.
-  Any newly issued P256 based certificates will have their signature clamped to the low-s form.
-  Nebula will assert the low-s signature form when validating certificates in a future version. [GHSA-69x3-g4r3-p962](https://github.com/slackhq/nebula/security/advisories/GHSA-69x3-g4r3-p962)
-
-### Changed
-
-- Improve error reporting if nebula fails to start due to a tun device naming issue. (#1588)
-
 ## [1.10.2] - 2026-01-21
 
 ### Fixed
@@ -788,8 +775,7 @@ created.)
 
 - Initial public release.
 
-[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.3...HEAD
+[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.2...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
 [1.10.0]: https://github.com/slackhq/nebula/releases/tag/v1.10.0

CODEOWNERS

@@ -1 +0,0 @@
-#ECCN:Open Source

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.

boring.go

@@ -1,4 +1,5 @@
 //go:build boringcrypto
+// +build boringcrypto
 
 package nebula
 

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
-
+	for {
-	scanner := bufio.NewScanner(r)
+		caPEMs, err = pool.AddCAFromPEM(caPEMs)
-	scanner.Split(SplitPEM)
+		if errors.Is(err, ErrExpired) {
-
+			expired = true
-	for scanner.Scan() {
+			err = nil
-		pemBytes := scanner.Bytes()
-
-		block, rest := pem.Decode(pemBytes)
-		if len(bytes.TrimSpace(rest)) > 0 {
-			return nil, ErrInvalidPEMBlock
 		}
-		if block == nil {
-			return nil, ErrInvalidPEMBlock
-		}
-
-		c, err := unmarshalCertificateBlock(block)
 		if err != nil {
 			return nil, err
 		}
-
+		if len(caPEMs) == 0 || strings.TrimSpace(string(caPEMs)) == "" {
-		err = pool.AddCA(c)
+			break
-		if errors.Is(err, ErrExpired) {
-			expired = true
-			continue
-		} else if err != nil {
-			return nil, err
 		}
 	}
-	if err := scanner.Err(); err != nil {
-		return nil, ErrInvalidPEMBlock
-	}
 
 	if expired {
 		return pool, ErrExpired
@@ -168,23 +141,10 @@ func (ncp *CAPool) VerifyCertificate(now time.Time, c Certificate) (*CachedCerti
 		return nil, err
 	}
 
-	// Pre nebula v1.10.3 could generate signatures in either high or low s form and validation
-	// of signatures allowed for either. Nebula v1.10.3 and beyond clamps signature generation to low-s form
-	// but validation still allows for either. Since a change in the signature bytes affects the fingerprint, we
-	// need to test both forms until such a time comes that we enforce low-s form on signature validation.
-	fp2, err := CalculateAlternateFingerprint(c)
-	if err != nil {
-		return nil, fmt.Errorf("could not calculate alternate fingerprint to verify: %w", err)
-	}
-	if fp2 != "" && ncp.IsBlocklisted(fp2) {
-		return nil, ErrBlockListed
-	}
-
 	cc := CachedCertificate{
 		Certificate:       c,
 		InvertedGroups:    make(map[string]struct{}),
 		Fingerprint:       fp,
-		fingerprint2:      fp2,
 		signerFingerprint: signer.Fingerprint,
 	}
 
@@ -198,11 +158,6 @@ func (ncp *CAPool) VerifyCertificate(now time.Time, c Certificate) (*CachedCerti
 // VerifyCachedCertificate is the same as VerifyCertificate other than it operates on a pre-verified structure and
 // is a cheaper operation to perform as a result.
 func (ncp *CAPool) VerifyCachedCertificate(now time.Time, c *CachedCertificate) error {
-	// Check any available alternate fingerprint forms for this certificate, re P256 high-s/low-s
-	if c.fingerprint2 != "" && ncp.IsBlocklisted(c.fingerprint2) {
-		return ErrBlockListed
-	}
-
 	_, err := ncp.verify(c.Certificate, now, c.Fingerprint, c.signerFingerprint)
 	return err
 }

cert/ca_pool_test.go

@@ -1,14 +1,10 @@
 package cert
 
 import (
-	"bytes"
-	"io"
 	"net/netip"
-	"strings"
 	"testing"
 	"time"
 
-	"github.com/slackhq/nebula/cert/p256"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -115,60 +111,6 @@ 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)
@@ -228,15 +170,6 @@ func TestCertificateV1_VerifyP256(t *testing.T) {
 	_, err = caPool.VerifyCertificate(time.Now(), c)
 	require.EqualError(t, err, "certificate is in the block list")
 
-	// Create a copy of the cert and swap to the alternate form for the signature
-	nc := c.Copy()
-	b, err := p256.Swap(c.Signature())
-	require.NoError(t, err)
-	require.NoError(t, nc.(*certificateV1).setSignature(b))
-
-	_, err = caPool.VerifyCertificate(time.Now(), nc)
-	require.EqualError(t, err, "certificate is in the block list")
-
 	caPool.ResetCertBlocklist()
 	_, err = caPool.VerifyCertificate(time.Now(), c)
 	require.NoError(t, err)
@@ -254,7 +187,7 @@ func TestCertificateV1_VerifyP256(t *testing.T) {
 	require.NoError(t, err)
 
 	caPool = NewCAPool()
-	b, err = caPool.AddCAFromPEM(caPem)
+	b, err := caPool.AddCAFromPEM(caPem)
 	require.NoError(t, err)
 	assert.Empty(t, b)
 
@@ -263,17 +196,7 @@ func TestCertificateV1_VerifyP256(t *testing.T) {
 	})
 
 	c, _, _, _ = NewTestCert(Version1, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), nil, nil, []string{"test1"})
-	cc, err := caPool.VerifyCertificate(time.Now(), c)
+	_, err = caPool.VerifyCertificate(time.Now(), c)
-	require.NoError(t, err)
-
-	// Reset the blocklist and block the alternate form fingerprint
-	caPool.ResetCertBlocklist()
-	caPool.BlocklistFingerprint(cc.fingerprint2)
-	err = caPool.VerifyCachedCertificate(time.Now(), cc)
-	require.EqualError(t, err, "certificate is in the block list")
-
-	caPool.ResetCertBlocklist()
-	err = caPool.VerifyCachedCertificate(time.Now(), cc)
 	require.NoError(t, err)
 }
 
@@ -471,15 +394,6 @@ func TestCertificateV2_VerifyP256(t *testing.T) {
 	_, err = caPool.VerifyCertificate(time.Now(), c)
 	require.EqualError(t, err, "certificate is in the block list")
 
-	// Create a copy of the cert and swap to the alternate form for the signature
-	nc := c.Copy()
-	b, err := p256.Swap(c.Signature())
-	require.NoError(t, err)
-	require.NoError(t, nc.(*certificateV2).setSignature(b))
-
-	_, err = caPool.VerifyCertificate(time.Now(), nc)
-	require.EqualError(t, err, "certificate is in the block list")
-
 	caPool.ResetCertBlocklist()
 	_, err = caPool.VerifyCertificate(time.Now(), c)
 	require.NoError(t, err)
@@ -497,7 +411,7 @@ func TestCertificateV2_VerifyP256(t *testing.T) {
 	require.NoError(t, err)
 
 	caPool = NewCAPool()
-	b, err = caPool.AddCAFromPEM(caPem)
+	b, err := caPool.AddCAFromPEM(caPem)
 	require.NoError(t, err)
 	assert.Empty(t, b)
 
@@ -506,17 +420,7 @@ func TestCertificateV2_VerifyP256(t *testing.T) {
 	})
 
 	c, _, _, _ = NewTestCert(Version2, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), nil, nil, []string{"test1"})
-	cc, err := caPool.VerifyCertificate(time.Now(), c)
+	_, err = caPool.VerifyCertificate(time.Now(), c)
-	require.NoError(t, err)
-
-	// Reset the blocklist and block the alternate form fingerprint
-	caPool.ResetCertBlocklist()
-	caPool.BlocklistFingerprint(cc.fingerprint2)
-	err = caPool.VerifyCachedCertificate(time.Now(), cc)
-	require.EqualError(t, err, "certificate is in the block list")
-
-	caPool.ResetCertBlocklist()
-	err = caPool.VerifyCachedCertificate(time.Now(), cc)
 	require.NoError(t, err)
 }
 

cert/cert.go

@@ -4,8 +4,6 @@ import (
 	"fmt"
 	"net/netip"
 	"time"
-
-	"github.com/slackhq/nebula/cert/p256"
 )
 
 type Version uint8
@@ -112,9 +110,6 @@ type CachedCertificate struct {
 	InvertedGroups    map[string]struct{}
 	Fingerprint       string
 	signerFingerprint string
-
-	// A place to store a 2nd fingerprint if the certificate could have one, such as with P256
-	fingerprint2 string
 }
 
 func (cc *CachedCertificate) String() string {
@@ -157,31 +152,3 @@ func Recombine(v Version, rawCertBytes, publicKey []byte, curve Curve) (Certific
 
 	return c, nil
 }
-
-// CalculateAlternateFingerprint calculates a 2nd fingerprint representation for P256 certificates
-// CAPool blocklist testing through `VerifyCertificate` and `VerifyCachedCertificate` automatically performs this step.
-func CalculateAlternateFingerprint(c Certificate) (string, error) {
-	if c.Curve() != Curve_P256 {
-		return "", nil
-	}
-
-	nc := c.Copy()
-	b, err := p256.Swap(nc.Signature())
-	if err != nil {
-		return "", err
-	}
-
-	switch v := nc.(type) {
-	case *certificateV1:
-		err = v.setSignature(b)
-	case *certificateV2:
-		err = v.setSignature(b)
-	default:
-		return "", ErrUnknownVersion
-	}
-
-	if err != nil {
-		return "", err
-	}
-	return nc.Fingerprint()
-}

cert/p256/p256.go

@@ -1,127 +0,0 @@
-package p256
-
-import (
-	"crypto/elliptic"
-	"errors"
-	"math/big"
-
-	"filippo.io/bigmod"
-
-	"golang.org/x/crypto/cryptobyte"
-	"golang.org/x/crypto/cryptobyte/asn1"
-)
-
-var halfN = new(big.Int).Rsh(elliptic.P256().Params().N, 1)
-var nMod *bigmod.Modulus
-
-func init() {
-	n, err := bigmod.NewModulus(elliptic.P256().Params().N.Bytes())
-	if err != nil {
-		panic(err)
-	}
-	nMod = n
-}
-
-func IsNormalized(sig []byte) (bool, error) {
-	r, s, err := parseSignature(sig)
-	if err != nil {
-		return false, err
-	}
-	return checkLowS(r, s), nil
-}
-
-func checkLowS(_, s []byte) bool {
-	bigS := new(big.Int).SetBytes(s)
-	// Check if S <= (N/2), because we want to include the midpoint in the set of low-s
-	return bigS.Cmp(halfN) <= 0
-}
-
-func swap(r, s []byte) ([]byte, []byte, error) {
-	var err error
-	bigS, err := bigmod.NewNat().SetBytes(s, nMod)
-	if err != nil {
-		return nil, nil, err
-	}
-	sNormalized := nMod.Nat().Sub(bigS, nMod)
-
-	result := sNormalized.Bytes(nMod)
-	for len(result) > 1 && result[0] == 0 {
-		result = result[1:]
-	}
-
-	return r, result, nil
-}
-
-func Normalize(sig []byte) ([]byte, error) {
-	r, s, err := parseSignature(sig)
-	if err != nil {
-		return nil, err
-	}
-
-	if checkLowS(r, s) {
-		return sig, nil
-	}
-
-	newR, newS, err := swap(r, s)
-	if err != nil {
-		return nil, err
-	}
-
-	return encodeSignature(newR, newS)
-}
-
-// Swap will change sig between its current form to the opposite high or low form.
-func Swap(sig []byte) ([]byte, error) {
-	r, s, err := parseSignature(sig)
-	if err != nil {
-		return nil, err
-	}
-
-	newR, newS, err := swap(r, s)
-	if err != nil {
-		return nil, err
-	}
-
-	return encodeSignature(newR, newS)
-}
-
-// parseSignature taken exactly from crypto/ecdsa/ecdsa.go
-func parseSignature(sig []byte) (r, s []byte, err error) {
-	var inner cryptobyte.String
-	input := cryptobyte.String(sig)
-	if !input.ReadASN1(&inner, asn1.SEQUENCE) ||
-		!input.Empty() ||
-		!inner.ReadASN1Integer(&r) ||
-		!inner.ReadASN1Integer(&s) ||
-		!inner.Empty() {
-		return nil, nil, errors.New("invalid ASN.1")
-	}
-	return r, s, nil
-}
-
-func encodeSignature(r, s []byte) ([]byte, error) {
-	var b cryptobyte.Builder
-	b.AddASN1(asn1.SEQUENCE, func(b *cryptobyte.Builder) {
-		addASN1IntBytes(b, r)
-		addASN1IntBytes(b, s)
-	})
-	return b.Bytes()
-}
-
-// addASN1IntBytes encodes in ASN.1 a positive integer represented as
-// a big-endian byte slice with zero or more leading zeroes.
-func addASN1IntBytes(b *cryptobyte.Builder, bytes []byte) {
-	for len(bytes) > 0 && bytes[0] == 0 {
-		bytes = bytes[1:]
-	}
-	if len(bytes) == 0 {
-		b.SetError(errors.New("invalid integer"))
-		return
-	}
-	b.AddASN1(asn1.INTEGER, func(c *cryptobyte.Builder) {
-		if bytes[0]&0x80 != 0 {
-			c.AddUint8(0)
-		}
-		c.AddBytes(bytes)
-	})
-}

cert/p256/p256_test.go

@@ -1,28 +0,0 @@
-package p256
-
-import (
-	"crypto/ecdsa"
-	"crypto/elliptic"
-	"crypto/rand"
-	"testing"
-
-	"github.com/stretchr/testify/require"
-)
-
-func TestFlipping(t *testing.T) {
-	priv, err1 := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
-	require.NoError(t, err1)
-
-	out, err := ecdsa.SignASN1(rand.Reader, priv, []byte("big chungus"))
-	require.NoError(t, err)
-
-	r, s, err := parseSignature(out)
-	require.NoError(t, err)
-
-	r, s1, err := swap(r, s)
-	require.NoError(t, err)
-	r, s2, err := swap(r, s1)
-	require.NoError(t, err)
-	require.Equal(t, s, s2)
-	require.NotEqual(t, s, s1)
-}

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

cert/sign.go

@@ -9,8 +9,6 @@ import (
 	"fmt"
 	"net/netip"
 	"time"
-
-	"github.com/slackhq/nebula/cert/p256"
 )
 
 // TBSCertificate represents a certificate intended to be signed.
@@ -128,13 +126,6 @@ func (t *TBSCertificate) SignWith(signer Certificate, curve Curve, sp SignerLamb
 		return nil, err
 	}
 
-	if curve == Curve_P256 {
-		sig, err = p256.Normalize(sig)
-		if err != nil {
-			return nil, err
-		}
-	}
-
 	err = c.setSignature(sig)
 	if err != nil {
 		return nil, err

cert/sign_test.go

@@ -9,7 +9,6 @@ import (
 	"testing"
 	"time"
 
-	"github.com/slackhq/nebula/cert/p256"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -90,48 +89,3 @@ func TestCertificateV1_SignP256(t *testing.T) {
 	require.NoError(t, err)
 	assert.NotNil(t, uc)
 }
-
-func TestCertificate_SignP256_AlwaysNormalized(t *testing.T) {
-	before := time.Now().Add(time.Second * -60).Round(time.Second)
-	after := time.Now().Add(time.Second * 60).Round(time.Second)
-	pubKey := []byte("01234567890abcedfghij1234567890ab1234567890abcedfghij1234567890ab")
-
-	tbs := TBSCertificate{
-		Version: Version1,
-		Name:    "testing",
-		Networks: []netip.Prefix{
-			mustParsePrefixUnmapped("10.1.1.1/24"),
-			mustParsePrefixUnmapped("10.1.1.2/16"),
-		},
-		UnsafeNetworks: []netip.Prefix{
-			mustParsePrefixUnmapped("9.1.1.2/24"),
-			mustParsePrefixUnmapped("9.1.1.3/16"),
-		},
-		Groups:    []string{"test-group1", "test-group2", "test-group3"},
-		NotBefore: before,
-		NotAfter:  after,
-		PublicKey: pubKey,
-		IsCA:      true,
-		Curve:     Curve_P256,
-	}
-
-	priv, err := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
-	require.NoError(t, err)
-	pub := elliptic.Marshal(elliptic.P256(), priv.PublicKey.X, priv.PublicKey.Y)
-	rawPriv := priv.D.FillBytes(make([]byte, 32))
-
-	for i := 0; i < 1000; i++ {
-		if i&1 == 1 {
-			tbs.Version = Version1
-		} else {
-			tbs.Version = Version2
-		}
-		c, err := tbs.Sign(nil, Curve_P256, rawPriv)
-		require.NoError(t, err)
-		assert.NotNil(t, c)
-		assert.True(t, c.CheckSignature(pub))
-		normie, err := p256.IsNormalized(c.Signature())
-		require.NoError(t, err)
-		assert.True(t, normie)
-	}
-}

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)
+	caPool := cert.NewCAPool()
-	if err != nil && !errors.Is(err, cert.ErrExpired) {
+	for {
-		return fmt.Errorf("error while adding ca cert to pool: %w", err)
+		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)

control_tester.go

@@ -1,4 +1,5 @@
 //go:build e2e_testing
+// +build e2e_testing
 
 package nebula
 

dist/wireshark/nebula.lua

@@ -84,24 +84,30 @@ end
 
 function nebula.prefs_changed()
     if default_settings.all_ports == nebula.prefs.all_ports and default_settings.port == nebula.prefs.port then
+        -- Nothing changed, bail
         return
     end
 
-    -- Remove all existing registrations
+    -- Remove our old dissector
     DissectorTable.get("udp.port"):remove_all(nebula)
 
-    if nebula.prefs.all_ports then
+    if nebula.prefs.all_ports and default_settings.all_ports ~= nebula.prefs.all_ports then
-        -- Register on every port for hole punch capture
+        default_settings.all_port = nebula.prefs.all_ports
+
         for i=0, 65535 do
             DissectorTable.get("udp.port"):add(i, nebula)
         end
-    else
+
-        -- Register on the configured port only
+        -- no need to establish again on specific ports
-        DissectorTable.get("udp.port"):add(nebula.prefs.port, nebula)
+        return
     end
 
-    default_settings.all_ports = nebula.prefs.all_ports
+
-    default_settings.port = nebula.prefs.port
+    if default_settings.all_ports ~= nebula.prefs.all_ports then
+        -- Add our new port dissector
+        default_settings.port = nebula.prefs.port
+        DissectorTable.get("udp.port"):add(default_settings.port, nebula)
+    end
 end
 
 DissectorTable.get("udp.port"):add(default_settings.port, nebula)

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.

e2e/tunnels_test.go

@@ -12,8 +12,6 @@ import (
 	"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"
 	"gopkg.in/yaml.v3"
 )
@@ -367,106 +365,3 @@ func TestCrossStackRelaysWork(t *testing.T) {
 	//theirControl.Stop()
 	//relayControl.Stop()
 }
-
-func TestCloseTunnelAuthenticated(t *testing.T) {
-	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{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "5s"}})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "10m"}})
-
-	// Share our underlay information
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	// Start the servers
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-
-	r.Log("Assert the tunnel between me and them works")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	r.Log("Close the tunnel")
-	myControl.CloseTunnel(theirVpnIpNet[0].Addr(), false)
-	r.FlushAll()
-
-	waitStart := time.Now()
-	for {
-		myIndexes := len(myControl.GetHostmap().Indexes)
-		theirIndexes := len(theirControl.GetHostmap().Indexes)
-		if myIndexes == 0 && theirIndexes == 0 {
-			break
-		}
-
-		since := time.Since(waitStart)
-		r.Logf("my tunnels: %v; their tunnels: %v; duration: %v", myIndexes, theirIndexes, since)
-		if since > time.Second*6 {
-			t.Fatal("Tunnel should have been declared inactive after 2 seconds and before 6 seconds")
-		}
-
-		time.Sleep(1 * time.Second)
-		//r.FlushAll()
-	}
-
-	r.Logf("Happy path success, tunnels were dropped within %v", time.Since(waitStart))
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-	r.Log("Assert another tunnel between me and them works")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-	hi := myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), false)
-	if hi == nil {
-		t.Fatal("There is no hostinfo for this tunnel")
-	}
-	myHi := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-	if myHi == nil {
-		t.Fatal("There is no hostinfo for my tunnel")
-	}
-	r.Log("It does")
-
-	buf := make([]byte, 1024)
-	hdr := header.H{
-		Version:        1,
-		Type:           header.CloseTunnel,
-		Subtype:        0,
-		Reserved:       0,
-		RemoteIndex:    hi.RemoteIndex,
-		MessageCounter: 5,
-	}
-	out, err := hdr.Encode(buf)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	pkt := &udp.Packet{
-		To:   hi.CurrentRemote,
-		From: myHi.CurrentRemote,
-		Data: out,
-	}
-	r.InjectUDPPacket(myControl, theirControl, pkt)
-	r.Log("Injected bogus close tunnel. Let's see!")
-	waitStart = time.Now()
-	for {
-		myIndexes := len(myControl.GetHostmap().Indexes)
-		theirIndexes := len(theirControl.GetHostmap().Indexes)
-		if myIndexes == 0 {
-			t.Fatal("myIndexes should not be 0")
-		}
-		if theirIndexes == 0 {
-			t.Fatal("theirIndexes should not be 0, they should have rejected this bogus packet")
-		}
-
-		since := time.Since(waitStart)
-		r.Logf("my tunnels: %v; their tunnels: %v; duration: %v", myIndexes, theirIndexes, since)
-		if since > time.Second*4 {
-			t.Log("The tunnel would have been gone by now")
-			break
-		}
-
-		time.Sleep(1 * time.Second)
-		r.FlushAll()
-	}
-
-	myControl.Stop()
-	theirControl.Stop()
-}

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:
@@ -388,8 +382,8 @@ firewall:
   # Rules are comprised of a protocol, port, and one or more of host, group, or CIDR
   # Logical evaluation is roughly: port AND proto AND (ca_sha OR ca_name) AND (host OR group OR groups OR cidr) AND (local cidr)
   # - port: Takes `0` or `any` as any, a single number `80`, a range `200-901`, or `fragment` to match second and further fragments of fragmented packets (since there is no port available).
+  #   code: same as port but makes more sense when talking about ICMP, TODO: this is not currently implemented in a way that works, use `any`
   #   proto: `any`, `tcp`, `udp`, or `icmp`
-  #          a port specification is ignored if proto is `icmp`
   #   host: `any` or a literal hostname, ie `test-host`
   #   group: `any` or a literal group name, ie `default-group`
   #   groups: Same as group but accepts a list of values. Multiple values are AND'd together and a certificate would have to contain all groups to pass

firewall.go

@@ -230,7 +230,7 @@ func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firew
 	case "drop":
 		fw.OutSendReject = false
 	default:
-		l.WithField("action", outboundAction).Warn("invalid firewall.outbound_action, defaulting to `drop`")
+		l.WithField("action", inboundAction).Warn("invalid firewall.outbound_action, defaulting to `drop`")
 		fw.OutSendReject = false
 	}
 
@@ -249,6 +249,20 @@ func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firew
 
 // AddRule properly creates the in memory rule structure for a firewall table.
 func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort int32, groups []string, host string, cidr, localCidr, caName string, caSha string) error {
+	// We need this rule string because we generate a hash. Removing this will break firewall reload.
+	ruleString := fmt.Sprintf(
+		"incoming: %v, proto: %v, startPort: %v, endPort: %v, groups: %v, host: %v, ip: %v, localIp: %v, caName: %v, caSha: %s",
+		incoming, proto, startPort, endPort, groups, host, cidr, localCidr, caName, caSha,
+	)
+	f.rules += ruleString + "\n"
+
+	direction := "incoming"
+	if !incoming {
+		direction = "outgoing"
+	}
+	f.l.WithField("firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha}).
+		Info("Firewall rule added")
+
 	var (
 		ft *FirewallTable
 		fp firewallPort
@@ -266,12 +280,6 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
 	case firewall.ProtoUDP:
 		fp = ft.UDP
 	case firewall.ProtoICMP, firewall.ProtoICMPv6:
-		//ICMP traffic doesn't have ports, so we always coerce to "any", even if a value is provided
-		if startPort != firewall.PortAny {
-			f.l.WithField("startPort", startPort).Warn("ignoring port specification for ICMP firewall rule")
-		}
-		startPort = firewall.PortAny
-		endPort = firewall.PortAny
 		fp = ft.ICMP
 	case firewall.ProtoAny:
 		fp = ft.AnyProto
@@ -279,20 +287,6 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
 		return fmt.Errorf("unknown protocol %v", proto)
 	}
 
-	// We need this rule string because we generate a hash. Removing this will break firewall reload.
-	ruleString := fmt.Sprintf(
-		"incoming: %v, proto: %v, startPort: %v, endPort: %v, groups: %v, host: %v, ip: %v, localIp: %v, caName: %v, caSha: %s",
-		incoming, proto, startPort, endPort, groups, host, cidr, localCidr, caName, caSha,
-	)
-	f.rules += ruleString + "\n"
-
-	direction := "incoming"
-	if !incoming {
-		direction = "outgoing"
-	}
-	f.l.WithField("firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha}).
-		Info("Firewall rule added")
-
 	return fp.addRule(f, startPort, endPort, groups, host, cidr, localCidr, caName, caSha)
 }
 
@@ -355,31 +349,24 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
 			sPort = r.Port
 		}
 
+		startPort, endPort, err := parsePort(sPort)
+		if err != nil {
+			return fmt.Errorf("%s rule #%v; %s %s", table, i, errPort, err)
+		}
+
 		var proto uint8
-		var startPort, endPort int32
 		switch r.Proto {
 		case "any":
 			proto = firewall.ProtoAny
-			startPort, endPort, err = parsePort(sPort)
 		case "tcp":
 			proto = firewall.ProtoTCP
-			startPort, endPort, err = parsePort(sPort)
 		case "udp":
 			proto = firewall.ProtoUDP
-			startPort, endPort, err = parsePort(sPort)
 		case "icmp":
 			proto = firewall.ProtoICMP
-			startPort = firewall.PortAny
-			endPort = firewall.PortAny
-			if sPort != "" {
-				l.WithField("port", sPort).Warn("ignoring port specification for ICMP firewall rule")
-			}
 		default:
 			return fmt.Errorf("%s rule #%v; proto was not understood; `%s`", table, i, r.Proto)
 		}
-		if err != nil {
-			return fmt.Errorf("%s rule #%v; %s %s", table, i, errPort, err)
-		}
 
 		if r.Cidr != "" && r.Cidr != "any" {
 			_, err = netip.ParsePrefix(r.Cidr)
@@ -480,7 +467,7 @@ func (f *Firewall) metrics(incoming bool) firewallMetrics {
 	}
 }
 
-// Destroy cleans up any known cyclical references so the object can be freed by GC. This should be called if a new
+// Destroy cleans up any known cyclical references so the object can be free'd my GC. This should be called if a new
 // firewall object is created
 func (f *Firewall) Destroy() {
 	//TODO: clean references if/when needed
@@ -673,13 +660,6 @@ func (fp firewallPort) match(p firewall.Packet, incoming bool, c *cert.CachedCer
 		return false
 	}
 
-	// this branch is here to catch traffic from FirewallTable.Any.match and FirewallTable.ICMP.match
-	if p.Protocol == firewall.ProtoICMP || p.Protocol == firewall.ProtoICMPv6 {
-		// port numbers are re-used for connection tracking of ICMP,
-		// but we don't want to actually filter on them.
-		return fp[firewall.PortAny].match(p, c, caPool)
-	}
-
 	var port int32
 
 	if p.Fragment {
@@ -824,8 +804,10 @@ func (fr *FirewallRule) isAny(groups []string, host string, cidr string) bool {
 		return true
 	}
 
-	if slices.Contains(groups, "any") {
+	for _, group := range groups {
-		return true
+		if group == "any" {
+			return true
+		}
 	}
 
 	if host == "any" {
@@ -1036,56 +1018,54 @@ func (r *rule) sanity() error {
 		}
 	}
 
-	if r.Code != "" {
-		return fmt.Errorf("code specified as [%s]. Support for 'code' will be dropped in a future release, as it has never been functional", r.Code)
-	}
-
 	//todo alert on cidr-any
 
 	return nil
 }
 
-func parsePort(s string) (int32, int32, error) {
+func parsePort(s string) (startPort, endPort int32, err error) {
-	var err error
-	const notAPort int32 = -2
 	if s == "any" {
-		return firewall.PortAny, firewall.PortAny, nil
+		startPort = firewall.PortAny
-	}
+		endPort = firewall.PortAny
-	if s == "fragment" {
+
-		return firewall.PortFragment, firewall.PortFragment, nil
+	} else if s == "fragment" {
-	}
+		startPort = firewall.PortFragment
-	if !strings.Contains(s, `-`) {
+		endPort = firewall.PortFragment
+
+	} else if strings.Contains(s, `-`) {
+		sPorts := strings.SplitN(s, `-`, 2)
+		sPorts[0] = strings.Trim(sPorts[0], " ")
+		sPorts[1] = strings.Trim(sPorts[1], " ")
+
+		if len(sPorts) != 2 || sPorts[0] == "" || sPorts[1] == "" {
+			return 0, 0, fmt.Errorf("appears to be a range but could not be parsed; `%s`", s)
+		}
+
+		rStartPort, err := strconv.Atoi(sPorts[0])
+		if err != nil {
+			return 0, 0, fmt.Errorf("beginning range was not a number; `%s`", sPorts[0])
+		}
+
+		rEndPort, err := strconv.Atoi(sPorts[1])
+		if err != nil {
+			return 0, 0, fmt.Errorf("ending range was not a number; `%s`", sPorts[1])
+		}
+
+		startPort = int32(rStartPort)
+		endPort = int32(rEndPort)
+
+		if startPort == firewall.PortAny {
+			endPort = firewall.PortAny
+		}
+
+	} else {
 		rPort, err := strconv.Atoi(s)
 		if err != nil {
-			return notAPort, notAPort, fmt.Errorf("was not a number; `%s`", s)
+			return 0, 0, fmt.Errorf("was not a number; `%s`", s)
 		}
-		return int32(rPort), int32(rPort), nil
+		startPort = int32(rPort)
+		endPort = startPort
 	}
 
-	sPorts := strings.SplitN(s, `-`, 2)
+	return
-	for i := range sPorts {
-		sPorts[i] = strings.Trim(sPorts[i], " ")
-	}
-	if len(sPorts) != 2 || sPorts[0] == "" || sPorts[1] == "" {
-		return notAPort, notAPort, fmt.Errorf("appears to be a range but could not be parsed; `%s`", s)
-	}
-
-	rStartPort, err := strconv.Atoi(sPorts[0])
-	if err != nil {
-		return notAPort, notAPort, fmt.Errorf("beginning range was not a number; `%s`", sPorts[0])
-	}
-
-	rEndPort, err := strconv.Atoi(sPorts[1])
-	if err != nil {
-		return notAPort, notAPort, fmt.Errorf("ending range was not a number; `%s`", sPorts[1])
-	}
-
-	startPort := int32(rStartPort)
-	endPort := int32(rEndPort)
-
-	if startPort == firewall.PortAny {
-		endPort = firewall.PortAny
-	}
-
-	return startPort, endPort, nil
 }

firewall/packet.go

@@ -22,10 +22,7 @@ const (
 type Packet struct {
 	LocalAddr  netip.Addr
 	RemoteAddr netip.Addr
-	// LocalPort is the destination port for incoming traffic, or the source port for outgoing. Zero for ICMP.
+	LocalPort  uint16
-	LocalPort uint16
-	// RemotePort is the source port for incoming traffic, or the destination port for outgoing.
-	// For ICMP, it's the "identifier". This is only used for connection tracking, actual firewall rules will not filter on ICMP identifier
 	RemotePort uint16
 	Protocol   uint8
 	Fragment   bool
@@ -49,8 +46,6 @@ func (fp Packet) MarshalJSON() ([]byte, error) {
 		proto = "tcp"
 	case ProtoICMP:
 		proto = "icmp"
-	case ProtoICMPv6:
-		proto = "icmpv6"
 	case ProtoUDP:
 		proto = "udp"
 	default:

firewall_test.go

@@ -87,10 +87,9 @@ func TestFirewall_AddRule(t *testing.T) {
 
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoICMP, 1, 1, []string{}, "h1", "", "", "", ""))
-	//no matter what port is given for icmp, it should end up as "any"
+	assert.Nil(t, fw.InRules.ICMP[1].Any.Any)
-	assert.Nil(t, fw.InRules.ICMP[firewall.PortAny].Any.Any)
+	assert.Empty(t, fw.InRules.ICMP[1].Any.Groups)
-	assert.Empty(t, fw.InRules.ICMP[firewall.PortAny].Any.Groups)
+	assert.Contains(t, fw.InRules.ICMP[1].Any.Hosts, "h1")
-	assert.Contains(t, fw.InRules.ICMP[firewall.PortAny].Any.Hosts, "h1")
 
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", ti.String(), "", "", ""))
@@ -735,150 +734,6 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
 }
 
-func TestFirewall_ICMPPortBehavior(t *testing.T) {
-	l := test.NewLogger()
-	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
-
-	network := netip.MustParsePrefix("1.2.3.4/24")
-
-	c := cert.CachedCertificate{
-		Certificate: &dummyCert{
-			name:     "host1",
-			networks: []netip.Prefix{network},
-			groups:   []string{"default-group"},
-			issuer:   "signer-shasum",
-		},
-		InvertedGroups: map[string]struct{}{"default-group": {}},
-	}
-	h := HostInfo{
-		ConnectionState: &ConnectionState{
-			peerCert: &c,
-		},
-		vpnAddrs: []netip.Addr{network.Addr()},
-	}
-	h.buildNetworks(myVpnNetworksTable, c.Certificate)
-
-	cp := cert.NewCAPool()
-
-	templ := firewall.Packet{
-		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
-		RemoteAddr: netip.MustParseAddr("1.2.3.4"),
-		Protocol:   firewall.ProtoICMP,
-		Fragment:   false,
-	}
-
-	t.Run("ICMP allowed", func(t *testing.T) {
-		fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
-		require.NoError(t, fw.AddRule(true, firewall.ProtoICMP, 0, 0, []string{"any"}, "", "", "", "", ""))
-		t.Run("zero ports", func(t *testing.T) {
-			p := templ.Copy()
-			p.LocalPort = 0
-			p.RemotePort = 0
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
-			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
-		})
-
-		t.Run("nonzero ports", func(t *testing.T) {
-			p := templ.Copy()
-			p.LocalPort = 0xabcd
-			p.RemotePort = 0x1234
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
-			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
-		})
-	})
-
-	t.Run("Any proto, some ports allowed", func(t *testing.T) {
-		fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
-		require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 80, 444, []string{"any"}, "", "", "", "", ""))
-		t.Run("zero ports, still blocked", func(t *testing.T) {
-			p := templ.Copy()
-			p.LocalPort = 0
-			p.RemotePort = 0
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
-			//now also allow outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-		})
-
-		t.Run("nonzero ports, still blocked", func(t *testing.T) {
-			p := templ.Copy()
-			p.LocalPort = 0xabcd
-			p.RemotePort = 0x1234
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
-			//now also allow outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-		})
-
-		t.Run("nonzero, matching ports, still blocked", func(t *testing.T) {
-			p := templ.Copy()
-			p.LocalPort = 80
-			p.RemotePort = 80
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
-			//now also allow outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-		})
-	})
-	t.Run("Any proto, any port", func(t *testing.T) {
-		fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
-		require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""))
-		t.Run("zero ports, allowed", func(t *testing.T) {
-			resetConntrack(fw)
-			p := templ.Copy()
-			p.LocalPort = 0
-			p.RemotePort = 0
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
-			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
-		})
-
-		t.Run("nonzero ports, allowed", func(t *testing.T) {
-			resetConntrack(fw)
-			p := templ.Copy()
-			p.LocalPort = 0xabcd
-			p.RemotePort = 0x1234
-			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-			// Allow inbound
-			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
-			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
-			//different ID is blocked
-			p.RemotePort++
-			require.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
-		})
-	})
-
-}
-
 func TestFirewall_DropIPSpoofing(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
@@ -1064,11 +919,11 @@ func TestNewFirewallFromConfig(t *testing.T) {
 
 	// Test code/port error
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh", "proto": "any"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; code was not a number; `a`")
 
-	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "a", "host": "testh", "proto": "any"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "a", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`")
 
@@ -1118,14 +973,7 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "icmp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
-	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: firewall.PortAny, endPort: firewall.PortAny, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
+	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
-
-	// Test adding icmp rule no port
-	conf = config.NewC(l)
-	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"proto": "icmp", "host": "a"}}}
-	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
-	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: firewall.PortAny, endPort: firewall.PortAny, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding any rule
 	conf = config.NewC(l)

go.mod

@@ -4,7 +4,6 @@ go 1.25
 
 require (
 	dario.cat/mergo v1.0.2
-	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
@@ -13,8 +12,8 @@ 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/dns v1.1.70
-	github.com/miekg/pkcs11 v1.1.2
+	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

@@ -1,8 +1,6 @@
 cloud.google.com/go v0.34.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
 dario.cat/mergo v1.0.2 h1:85+piFYR1tMbRrLcDwR18y4UKJ3aH1Tbzi24VRW1TK8=
 dario.cat/mergo v1.0.2/go.mod h1:E/hbnu0NxMFBjpMIE34DRGLWqDy0g5FuKDhCb31ngxA=
-filippo.io/bigmod v0.1.0 h1:UNzDk7y9ADKST+axd9skUpBQeW7fG2KrTZyOE4uGQy8=
-filippo.io/bigmod v0.1.0/go.mod h1:OjOXDNlClLblvXdwgFFOQFJEocLhhtai8vGLy0JCZlI=
 github.com/alecthomas/template v0.0.0-20160405071501-a0175ee3bccc/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
 github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
 github.com/alecthomas/units v0.0.0-20151022065526-2efee857e7cf/go.mod h1:ybxpYRFXyAe+OPACYpWeL0wqObRcbAqCMya13uyzqw0=
@@ -85,10 +83,10 @@ 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 v1.1.70 h1:DZ4u2AV35VJxdD9Fo9fIWm119BsQL5cZU1cQ9s0LkqA=
-github.com/miekg/dns v1.1.72/go.mod h1:+EuEPhdHOsfk6Wk5TT2CzssZdqkmFhf8r+aVyDEToIs=
+github.com/miekg/dns v1.1.70/go.mod h1:+EuEPhdHOsfk6Wk5TT2CzssZdqkmFhf8r+aVyDEToIs=
-github.com/miekg/pkcs11 v1.1.2 h1:/VxmeAX5qU6Q3EwafypogwWbYryHFmF2RpkJmw3m4MQ=
+github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b h1:J/AzCvg5z0Hn1rqZUJjpbzALUmkKX0Zwbc/i4fw7Sfk=
-github.com/miekg/pkcs11 v1.1.2/go.mod h1:XsNlhZGX73bx86s2hdc/FuaLm2CPZJemRLMA+WTFxgs=
+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

@@ -590,7 +590,7 @@ func (hm *HandshakeManager) allocateIndex(hh *HandshakeHostInfo) error {
 	hm.Lock()
 	defer hm.Unlock()
 
-	for range 32 {
+	for i := 0; i < 32; i++ {
 		index, err := generateIndex(hm.l)
 		if err != nil {
 			return err

hostmap_tester.go

@@ -1,4 +1,5 @@
 //go:build e2e_testing
+// +build e2e_testing
 
 package nebula
 

inside.go

@@ -9,8 +9,75 @@ import (
 	"github.com/slackhq/nebula/iputil"
 	"github.com/slackhq/nebula/noiseutil"
 	"github.com/slackhq/nebula/routing"
+	"github.com/slackhq/nebula/udp"
 )
 
+// consumeInsidePacketBatched is a variant of consumeInsidePacket that queues
+// outgoing packets into pendingPackets instead of sending them immediately.
+// The caller is responsible for flushing pendingPackets with WriteBatch.
+func (f *Interface) consumeInsidePacketBatched(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache, pendingPackets *[]udp.BatchPacket) {
+	err := newPacket(packet, false, fwPacket)
+	if err != nil {
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("packet", packet).Debugf("Error while validating outbound packet: %s", err)
+		}
+		return
+	}
+
+	// Ignore local broadcast packets
+	if f.dropLocalBroadcast {
+		if f.myBroadcastAddrsTable.Contains(fwPacket.RemoteAddr) {
+			return
+		}
+	}
+
+	if f.myVpnAddrsTable.Contains(fwPacket.RemoteAddr) {
+		if immediatelyForwardToSelf {
+			_, err := f.readers[q].Write(packet)
+			if err != nil {
+				f.l.WithError(err).Error("Failed to forward to tun")
+			}
+		}
+		return
+	}
+
+	// Ignore multicast packets
+	if f.dropMulticast && fwPacket.RemoteAddr.IsMulticast() {
+		return
+	}
+
+	hostinfo, ready := f.getOrHandshakeConsiderRouting(fwPacket, func(hh *HandshakeHostInfo) {
+		hh.cachePacket(f.l, header.Message, 0, packet, f.sendMessageNow, f.cachedPacketMetrics)
+	})
+
+	if hostinfo == nil {
+		f.rejectInside(packet, out, q)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("vpnAddr", fwPacket.RemoteAddr).
+				WithField("fwPacket", fwPacket).
+				Debugln("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks")
+		}
+		return
+	}
+
+	if !ready {
+		return
+	}
+
+	dropReason := f.firewall.Drop(*fwPacket, false, hostinfo, f.pki.GetCAPool(), localCache)
+	if dropReason == nil {
+		f.sendNoMetricsBatched(header.Message, 0, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, packet, nb, out, q, pendingPackets)
+	} else {
+		f.rejectInside(packet, out, q)
+		if f.l.Level >= logrus.DebugLevel {
+			hostinfo.logger(f.l).
+				WithField("fwPacket", fwPacket).
+				WithField("reason", dropReason).
+				Debugln("dropping outbound packet")
+		}
+	}
+}
+
 func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache) {
 	err := newPacket(packet, false, fwPacket)
 	if err != nil {
@@ -69,7 +136,6 @@ 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)
-
 	} else {
 		f.rejectInside(packet, out, q)
 		if f.l.Level >= logrus.DebugLevel {
@@ -410,3 +476,75 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 		}
 	}
 }
+
+// sendNoMetricsBatched is like sendNoMetrics but queues the packet for batched sending
+// instead of sending immediately. The caller must flush pendingPackets with WriteBatch.
+func (f *Interface) sendNoMetricsBatched(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte, q int, pendingPackets *[]udp.BatchPacket) {
+	if ci.eKey == nil {
+		return
+	}
+	useRelay := !remote.IsValid() && !hostinfo.remote.IsValid()
+	fullOut := out
+
+	if useRelay {
+		if len(out) < header.Len {
+			out = out[:header.Len]
+		}
+		out = out[header.Len:]
+	}
+
+	if noiseutil.EncryptLockNeeded {
+		ci.writeLock.Lock()
+	}
+	c := ci.messageCounter.Add(1)
+
+	out = header.Encode(out, header.Version, t, st, hostinfo.remoteIndexId, c)
+	f.connectionManager.Out(hostinfo)
+
+	if t != header.CloseTunnel && hostinfo.lastRebindCount != f.rebindCount {
+		f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
+		hostinfo.lastRebindCount = f.rebindCount
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).Debug("Lighthouse update triggered for punch due to rebind counter")
+		}
+	}
+
+	var err error
+	out, err = ci.eKey.EncryptDanger(out, out, p, c, nb)
+	if noiseutil.EncryptLockNeeded {
+		ci.writeLock.Unlock()
+	}
+	if err != nil {
+		hostinfo.logger(f.l).WithError(err).
+			WithField("udpAddr", remote).WithField("counter", c).
+			WithField("attemptedCounter", c).
+			Error("Failed to encrypt outgoing packet")
+		return
+	}
+
+	// Queue the packet for batched sending
+	var addr netip.AddrPort
+	if remote.IsValid() {
+		addr = remote
+	} else if hostinfo.remote.IsValid() {
+		addr = hostinfo.remote
+	} else {
+		// Relay path - send immediately, not batched
+		for _, relayIP := range hostinfo.relayState.CopyRelayIps() {
+			relayHostInfo, relay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relayIP)
+			if err != nil {
+				hostinfo.relayState.DeleteRelay(relayIP)
+				hostinfo.logger(f.l).WithField("relay", relayIP).WithError(err).Info("sendNoMetricsBatched failed to find HostInfo")
+				continue
+			}
+			f.SendVia(relayHostInfo, relay, out, nb, fullOut[:header.Len+len(out)], true)
+			break
+		}
+		return
+	}
+
+	// Copy the payload since the buffer will be reused
+	payload := make([]byte, len(out))
+	copy(payload, out)
+	*pendingPackets = append(*pendingPackets, udp.BatchPacket{Payload: payload, Addr: addr})
+}

inside_bsd.go

@@ -1,4 +1,5 @@
 //go:build darwin || dragonfly || freebsd || netbsd || openbsd
+// +build darwin dragonfly freebsd netbsd openbsd
 
 package nebula
 

inside_generic.go

@@ -1,4 +1,5 @@
 //go:build !darwin && !dragonfly && !freebsd && !netbsd && !openbsd
+// +build !darwin,!dragonfly,!freebsd,!netbsd,!openbsd
 
 package nebula
 

interface.go

@@ -7,6 +7,7 @@ import (
 	"io"
 	"net/netip"
 	"os"
+	"runtime"
 	"sync/atomic"
 	"time"
 
@@ -47,6 +48,8 @@ type InterfaceConfig struct {
 
 	ConntrackCacheTimeout time.Duration
 	l                     *logrus.Logger
+
+	tunBatchSize int // batch size for TUN read/write batching, 0 to disable
 }
 
 type Interface struct {
@@ -85,8 +88,9 @@ type Interface struct {
 
 	conntrackCacheTimeout time.Duration
 
-	writers []udp.Conn
+	writers      []udp.Conn
-	readers []io.ReadWriteCloser
+	readers      []io.ReadWriteCloser
+	tunBatchSize int // batch size for TUN read/write batching
 
 	metricHandshakes    metrics.Histogram
 	messageMetrics      *MessageMetrics
@@ -186,6 +190,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 		relayManager:          c.relayManager,
 		connectionManager:     c.connectionManager,
 		conntrackCacheTimeout: c.ConntrackCacheTimeout,
+		tunBatchSize:          c.tunBatchSize,
 
 		metricHandshakes: metrics.GetOrRegisterHistogram("handshakes", nil, metrics.NewExpDecaySample(1028, 0.015)),
 		messageMetrics:   c.MessageMetrics,
@@ -243,6 +248,15 @@ func (f *Interface) activate() {
 		f.readers[i] = reader
 	}
 
+	// Enable batch reading on all readers if batch size > 1
+	if f.tunBatchSize > 1 {
+		for i := 0; i < f.routines; i++ {
+			if err := overlay.EnableBatchReading(f.readers[i]); err != nil {
+				f.l.WithError(err).WithField("routine", i).Warn("Failed to enable batch reading, falling back to single reads")
+			}
+		}
+	}
+
 	if err := f.inside.Activate(); err != nil {
 		f.inside.Close()
 		f.l.Fatal(err)
@@ -262,6 +276,8 @@ func (f *Interface) run() {
 }
 
 func (f *Interface) listenOut(i int) {
+	runtime.LockOSThread()
+
 	var li udp.Conn
 	if i > 0 {
 		li = f.writers[i]
@@ -282,13 +298,23 @@ func (f *Interface) listenOut(i int) {
 }
 
 func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
+	runtime.LockOSThread()
+
+	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+
+	// Check if batch reading is available and enabled
+	batchReader := overlay.AsBatchReader(reader)
+	if batchReader != nil && f.tunBatchSize > 1 {
+		f.listenInBatched(reader, batchReader, i, conntrackCache)
+		return
+	}
+
+	// Fallback to single-packet reading
 	packet := make([]byte, mtu)
 	out := make([]byte, mtu)
 	fwPacket := &firewall.Packet{}
 	nb := make([]byte, 12, 12)
 
-	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
-
 	for {
 		n, err := reader.Read(packet)
 		if err != nil {
@@ -305,6 +331,54 @@ func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
 	}
 }
 
+func (f *Interface) listenInBatched(reader io.ReadWriteCloser, batchReader overlay.BatchReader, i int, conntrackCache *firewall.ConntrackCacheTicker) {
+	batchSize := f.tunBatchSize
+
+	// Pre-allocate buffers for batch reading
+	packets := make([][]byte, batchSize)
+	for j := range packets {
+		packets[j] = make([]byte, mtu)
+	}
+	sizes := make([]int, batchSize)
+
+	// Pre-allocate buffers for packet processing
+	out := make([]byte, mtu)
+	fwPacket := &firewall.Packet{}
+	nb := make([]byte, 12, 12)
+
+	// Pre-allocate buffer for batched UDP writes
+	pendingPackets := make([]udp.BatchPacket, 0, batchSize)
+
+	for {
+		// Read a batch of packets from TUN
+		n, err := batchReader.ReadBatch(packets, sizes)
+		if err != nil {
+			if errors.Is(err, os.ErrClosed) && f.closed.Load() {
+				return
+			}
+
+			f.l.WithError(err).Error("Error while reading outbound packets")
+			os.Exit(2)
+		}
+
+		if n == 0 {
+			continue
+		}
+
+		// Process all packets in the batch
+		cache := conntrackCache.Get(f.l)
+		for j := 0; j < n; j++ {
+			f.consumeInsidePacketBatched(packets[j][:sizes[j]], fwPacket, nb, out, i, cache, &pendingPackets)
+		}
+
+		// Flush all pending UDP writes
+		if len(pendingPackets) > 0 {
+			f.writers[i].WriteBatch(pendingPackets)
+			pendingPackets = pendingPackets[:0]
+		}
+	}
+}
+
 func (f *Interface) RegisterConfigChangeCallbacks(c *config.C) {
 	c.RegisterReloadCallback(f.reloadFirewall)
 	c.RegisterReloadCallback(f.reloadSendRecvError)

lighthouse.go

@@ -713,14 +713,21 @@ func (lh *LightHouse) unlockedShouldAddV6(vpnAddr netip.Addr, to *V6AddrPort) bo
 
 func (lh *LightHouse) IsLighthouseAddr(vpnAddr netip.Addr) bool {
 	l := lh.GetLighthouses()
-	return slices.Contains(l, vpnAddr)
+	for i := range l {
+		if l[i] == vpnAddr {
+			return true
+		}
+	}
+	return false
 }
 
 func (lh *LightHouse) IsAnyLighthouseAddr(vpnAddrs []netip.Addr) bool {
 	l := lh.GetLighthouses()
 	for i := range vpnAddrs {
-		if slices.Contains(l, vpnAddrs[i]) {
+		for j := range l {
-			return true
+			if l[j] == vpnAddrs[i] {
+				return true
+			}
 		}
 	}
 	return false

main.go

@@ -105,7 +105,11 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		// deprecated and undocumented
 		tunQueues := c.GetInt("tun.routines", 1)
 		udpQueues := c.GetInt("listen.routines", 1)
-		routines = max(tunQueues, udpQueues)
+		if tunQueues > udpQueues {
+			routines = tunQueues
+		} else {
+			routines = udpQueues
+		}
 		if routines != 1 {
 			l.WithField("routines", routines).Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead")
 		}
@@ -246,6 +250,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		punchy:                punchy,
 		ConntrackCacheTimeout: conntrackCacheTimeout,
 		l:                     l,
+		tunBatchSize:          c.GetInt("listen.batch", 64),
 	}
 
 	var ifce *Interface

notboring.go

@@ -1,4 +1,5 @@
 //go:build !boringcrypto
+// +build !boringcrypto
 
 package nebula
 

outside.go

@@ -190,13 +190,6 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 		if !f.handleEncrypted(ci, via, h) {
 			return
 		}
-		_, err = f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
-		if err != nil {
-			hostinfo.logger(f.l).WithError(err).WithField("from", via).
-				WithField("packet", packet).
-				Error("Failed to decrypt CloseTunnel packet")
-			return
-		}
 
 		hostinfo.logger(f.l).WithField("from", via).
 			Info("Close tunnel received, tearing down.")
@@ -334,29 +327,13 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 		proto := layers.IPProtocol(data[protoAt])
 
 		switch proto {
-		case layers.IPProtocolESP, layers.IPProtocolNoNextHeader:
+		case layers.IPProtocolICMPv6, 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.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
@@ -446,38 +423,34 @@ func parseV4(data []byte, incoming bool, fp *firewall.Packet) error {
 
 	// Accounting for a variable header length, do we have enough data for our src/dst tuples?
 	minLen := ihl
-	if !fp.Fragment {
+	if !fp.Fragment && fp.Protocol != firewall.ProtoICMP {
-		if fp.Protocol == firewall.ProtoICMP {
+		minLen += minFwPacketLen
-			minLen += minFwPacketLen + 2
-		} else {
-			minLen += minFwPacketLen
-		}
 	}
-
 	if len(data) < minLen {
 		return ErrIPv4InvalidHeaderLength
 	}
 
-	if incoming { // Firewall packets are locally oriented
+	// Firewall packets are locally oriented
+	if incoming {
 		fp.RemoteAddr, _ = netip.AddrFromSlice(data[12:16])
 		fp.LocalAddr, _ = netip.AddrFromSlice(data[16:20])
+		if fp.Fragment || fp.Protocol == firewall.ProtoICMP {
+			fp.RemotePort = 0
+			fp.LocalPort = 0
+		} else {
+			fp.RemotePort = binary.BigEndian.Uint16(data[ihl : ihl+2])
+			fp.LocalPort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4])
+		}
 	} else {
 		fp.LocalAddr, _ = netip.AddrFromSlice(data[12:16])
 		fp.RemoteAddr, _ = netip.AddrFromSlice(data[16:20])
-	}
+		if fp.Fragment || fp.Protocol == firewall.ProtoICMP {
-
+			fp.RemotePort = 0
-	if fp.Fragment {
+			fp.LocalPort = 0
-		fp.RemotePort = 0
+		} else {
-		fp.LocalPort = 0
+			fp.LocalPort = binary.BigEndian.Uint16(data[ihl : ihl+2])
-	} else if fp.Protocol == firewall.ProtoICMP { //note that orientation doesn't matter on ICMP
+			fp.RemotePort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4])
-		fp.RemotePort = binary.BigEndian.Uint16(data[ihl+4 : ihl+6]) //identifier
+		}
-		fp.LocalPort = 0                                             //code would be uint16(data[ihl+1])
-	} else if incoming {
-		fp.RemotePort = binary.BigEndian.Uint16(data[ihl : ihl+2])  //src port
-		fp.LocalPort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4]) //dst port
-	} else {
-		fp.LocalPort = binary.BigEndian.Uint16(data[ihl : ihl+2])    //src port
-		fp.RemotePort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4]) //dst port
 	}
 
 	return nil

outside_test.go

@@ -155,7 +155,6 @@ func Test_newPacket_v6(t *testing.T) {
 	// next layer, missing length byte
 	err = newPacket(buffer.Bytes()[:49], true, p)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
-	err = nil
 
 	// A good ICMP packet
 	ip = layers.IPv6{
@@ -166,26 +165,20 @@ func Test_newPacket_v6(t *testing.T) {
 		DstIP:      net.IPv6linklocalallnodes,
 	}
 
-	icmp := layers.ICMPv6{
+	icmp := layers.ICMPv6{}
-		TypeCode: layers.ICMPv6TypeEchoRequest,
-		Checksum: 0x1234,
-	}
 
 	buffer.Clear()
-	require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp))
+	err = gopacket.SerializeLayers(buffer, opt, &ip, &icmp)
-	require.Error(t, newPacket(buffer.Bytes(), true, p))
+	if err != nil {
-
+		panic(err)
-	buffer.Clear()
-	echo := layers.ICMPv6Echo{
-		Identifier: 0xabcd,
-		SeqNumber:  1234,
 	}
-	require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp, &echo))
+
-	require.NoError(t, newPacket(buffer.Bytes(), true, p))
+	err = newPacket(buffer.Bytes(), true, p)
+	require.NoError(t, err)
 	assert.Equal(t, uint8(layers.IPProtocolICMPv6), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
 	assert.Equal(t, netip.MustParseAddr("ff02::1"), p.LocalAddr)
-	assert.Equal(t, uint16(0xabcd), p.RemotePort)
+	assert.Equal(t, uint16(0), p.RemotePort)
 	assert.Equal(t, uint16(0), p.LocalPort)
 	assert.False(t, p.Fragment)
 
@@ -581,7 +574,7 @@ func BenchmarkParseV6(b *testing.B) {
 	}
 
 	evilBytes := buffer.Bytes()
-	for range 200 {
+	for i := 0; i < 200; i++ {
 		evilBytes = append(evilBytes, hopHeader...)
 	}
 	evilBytes = append(evilBytes, lastHopHeader...)

overlay/device.go

@@ -16,3 +16,38 @@ type Device interface {
 	SupportsMultiqueue() bool
 	NewMultiQueueReader() (io.ReadWriteCloser, error)
 }
+
+// BatchReader is an optional interface that devices can implement
+// to support reading multiple packets in a single batch operation.
+// This can significantly reduce syscall overhead under high load.
+type BatchReader interface {
+	// ReadBatch reads up to len(packets) packets into the provided buffers.
+	// Each packet is read into packets[i] and its length is stored in sizes[i].
+	// Returns the number of packets read, or an error.
+	// A return of (0, nil) indicates no packets were available (non-blocking).
+	ReadBatch(packets [][]byte, sizes []int) (int, error)
+}
+
+// AsBatchReader returns a BatchReader if the reader supports batch operations,
+// otherwise returns nil.
+func AsBatchReader(r io.ReadWriteCloser) BatchReader {
+	if br, ok := r.(BatchReader); ok {
+		return br
+	}
+	return nil
+}
+
+// BatchEnabler is an optional interface for devices that need explicit
+// enabling of batch read support (e.g., setting non-blocking mode).
+type BatchEnabler interface {
+	EnableBatchReading() error
+}
+
+// EnableBatchReading enables batch reading on the device if supported.
+// Returns nil if the device doesn't support or need explicit enabling.
+func EnableBatchReading(d interface{}) error {
+	if be, ok := d.(BatchEnabler); ok {
+		return be.EnableBatchReading()
+	}
+	return nil
+}

overlay/tun_linux.go

@@ -24,8 +24,14 @@ import (
 	"golang.org/x/sys/unix"
 )
 
+const (
+	// virtioNetHdrLen is the length of virtio_net_hdr (without mergeable buffers)
+	virtioNetHdrLen = 10
+)
+
 type tun struct {
 	io.ReadWriteCloser
+	fd          int
 	Device      string
 	vpnNetworks []netip.Prefix
 	MaxMTU      int
@@ -33,6 +39,13 @@ type tun struct {
 	TXQueueLen  int
 	deviceIndex int
 	ioctlFd     uintptr
+	nonBlocking bool // true if fd is in non-blocking mode
+	vnetHdr     bool // true if IFF_VNET_HDR is enabled on the TUN device
+
+	// readBuf is used when vnetHdr is enabled to read the full packet+header
+	// before stripping the header. This is needed because caller-provided
+	// buffers are sized for MTU but kernel writes MTU+10 with virtio header.
+	readBuf []byte
 
 	Routes                    atomic.Pointer[[]Route]
 	routeTree                 atomic.Pointer[bart.Table[routing.Gateways]]
@@ -52,6 +65,23 @@ func (t *tun) Networks() []netip.Prefix {
 	return t.vpnNetworks
 }
 
+// tunVnetHdrSupported checks if the kernel supports IFF_VNET_HDR on TUN devices
+func tunVnetHdrSupported() bool {
+	fd, err := unix.Open("/dev/net/tun", unix.O_RDONLY, 0)
+	if err != nil {
+		return false
+	}
+	defer unix.Close(fd)
+
+	var features uint32
+	err = ioctl(uintptr(fd), uintptr(unix.TUNGETFEATURES), uintptr(unsafe.Pointer(&features)))
+	if err != nil {
+		return false
+	}
+
+	return features&unix.IFF_VNET_HDR != 0
+}
+
 type ifReq struct {
 	Name  [16]byte
 	Flags uint16
@@ -72,11 +102,6 @@ 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 +114,7 @@ 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 +127,7 @@ 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)
 			}
@@ -111,11 +136,18 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 		}
 	}
 
+	// Check if VNET_HDR is supported before trying to use it
+	useVnetHdr := tunVnetHdrSupported()
+
 	var req ifReq
 	req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI)
 	if multiqueue {
 		req.Flags |= unix.IFF_MULTI_QUEUE
 	}
+	if useVnetHdr {
+		req.Flags |= unix.IFF_VNET_HDR
+	}
+
 	nameStr := c.GetString("tun.dev", "")
 	copy(req.Name[:], nameStr)
 	if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
@@ -126,6 +158,13 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	}
 	name := strings.Trim(string(req.Name[:]), "\x00")
 
+	// Track if VNET_HDR is in use
+	// Note: We don't call TUNSETOFFLOAD - just handle the headers manually
+	vnetHdrEnabled := useVnetHdr
+	if vnetHdrEnabled {
+		l.Info("TUN VNET_HDR enabled")
+	}
+
 	file := os.NewFile(uintptr(fd), "/dev/net/tun")
 	t, err := newTunGeneric(c, l, file, vpnNetworks)
 	if err != nil {
@@ -133,6 +172,13 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	}
 
 	t.Device = name
+	t.vnetHdr = vnetHdrEnabled
+
+	// Allocate read buffer for virtio header handling
+	// Buffer needs to be large enough for virtio header + max packet
+	if t.vnetHdr {
+		t.readBuf = make([]byte, t.MaxMTU+virtioNetHdrLen)
+	}
 
 	return t, nil
 }
@@ -140,6 +186,7 @@ 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),
@@ -249,14 +296,165 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 
 	var req ifReq
 	req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI | unix.IFF_MULTI_QUEUE)
+	if t.vnetHdr {
+		req.Flags |= unix.IFF_VNET_HDR
+	}
 	copy(req.Name[:], t.Device)
 	if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
 		return nil, err
 	}
 
-	file := os.NewFile(uintptr(fd), "/dev/net/tun")
+	reader := &tunBatchReader{fd: fd, device: t.Device, vnetHdr: t.vnetHdr}
+	if t.vnetHdr {
+		reader.readBuf = make([]byte, t.MaxMTU+virtioNetHdrLen)
+	}
+	return reader, nil
+}
 
-	return file, nil
+// tunBatchReader implements BatchReader for efficient batch packet reading
+type tunBatchReader struct {
+	fd      int
+	device  string
+	vnetHdr bool
+	readBuf []byte // internal buffer for virtio header handling
+}
+
+func (r *tunBatchReader) Read(b []byte) (int, error) {
+	// Choose buffer: use internal buffer for vnetHdr, caller's buffer otherwise
+	readBuf := b
+	if r.vnetHdr {
+		readBuf = r.readBuf
+	}
+
+	// Use poll to wait for data, then read
+	for {
+		n, err := unix.Read(r.fd, readBuf)
+		if err == nil {
+			if r.vnetHdr && n > virtioNetHdrLen {
+				packetLen := n - virtioNetHdrLen
+				copy(b, readBuf[virtioNetHdrLen:n])
+				return packetLen, nil
+			}
+			if r.vnetHdr {
+				return 0, nil // No packet data
+			}
+			return n, nil
+		}
+		if err == unix.EAGAIN || err == unix.EWOULDBLOCK {
+			// Wait for data
+			pfds := []unix.PollFd{{Fd: int32(r.fd), Events: unix.POLLIN}}
+			_, err = unix.Poll(pfds, -1)
+			if err != nil {
+				if err == unix.EINTR {
+					continue
+				}
+				return 0, err
+			}
+			continue
+		}
+		return n, err
+	}
+}
+
+func (r *tunBatchReader) Write(b []byte) (int, error) {
+	if !r.vnetHdr {
+		return unix.Write(r.fd, b)
+	}
+
+	// Use writev to prepend virtio header without copying the packet data
+	// Header is all zeros = no GSO, no checksum offload
+	var hdr [virtioNetHdrLen]byte
+	bufs := [][]byte{hdr[:], b}
+
+	n, err := unix.Writev(r.fd, bufs)
+	if err != nil {
+		return 0, err
+	}
+	// Return only the packet bytes written (exclude header)
+	if n > virtioNetHdrLen {
+		return n - virtioNetHdrLen, nil
+	}
+	return 0, nil
+}
+
+func (r *tunBatchReader) Close() error {
+	return unix.Close(r.fd)
+}
+
+// ReadBatch reads up to len(packets) packets from the TUN device.
+// It drains all available packets without blocking, using poll() only
+// when no packets have been read yet.
+func (r *tunBatchReader) ReadBatch(packets [][]byte, sizes []int) (int, error) {
+	count := 0
+	maxPackets := len(packets)
+	if len(sizes) < maxPackets {
+		maxPackets = len(sizes)
+	}
+
+	// Choose read buffer based on vnetHdr
+	readBuf := packets[0] // Will be updated in loop for non-vnetHdr
+	if r.vnetHdr {
+		readBuf = r.readBuf
+	}
+
+	for count < maxPackets {
+		if !r.vnetHdr {
+			readBuf = packets[count]
+		}
+
+		n, err := unix.Read(r.fd, readBuf)
+		if err == nil && n > 0 {
+			if r.vnetHdr {
+				if n > virtioNetHdrLen {
+					packetLen := n - virtioNetHdrLen
+					copy(packets[count], readBuf[virtioNetHdrLen:n])
+					sizes[count] = packetLen
+				} else {
+					// Malformed packet (no data after header), skip
+					continue
+				}
+			} else {
+				sizes[count] = n
+			}
+			count++
+			continue
+		}
+
+		if err == unix.EAGAIN || err == unix.EWOULDBLOCK {
+			// No more packets available
+			if count > 0 {
+				// We have some packets, return them
+				return count, nil
+			}
+			// No packets yet, wait for at least one
+			pfds := []unix.PollFd{{Fd: int32(r.fd), Events: unix.POLLIN}}
+			_, err = unix.Poll(pfds, -1)
+			if err != nil {
+				if err == unix.EINTR {
+					continue
+				}
+				return 0, err
+			}
+			continue
+		}
+
+		if err != nil {
+			if count > 0 {
+				// Return what we have
+				return count, nil
+			}
+			return 0, err
+		}
+
+		if n == 0 {
+			if count > 0 {
+				return count, nil
+			}
+			return 0, io.EOF
+		}
+	}
+
+	return count, nil
 }
 
 func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
@@ -264,6 +462,221 @@ func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	return r
 }
 
+func (t *tun) Write(b []byte) (int, error) {
+	if !t.vnetHdr {
+		return t.writeSimple(b)
+	}
+
+	// Use writev to prepend virtio header without copying the packet data
+	// Header is all zeros = no GSO, no checksum offload
+	var hdr [virtioNetHdrLen]byte
+	bufs := [][]byte{hdr[:], b}
+
+	n, err := unix.Writev(t.fd, bufs)
+	if err != nil {
+		return 0, err
+	}
+	// Return only the packet bytes written (exclude header)
+	if n > virtioNetHdrLen {
+		return n - virtioNetHdrLen, nil
+	}
+	return 0, nil
+}
+
+func (t *tun) writeSimple(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
+		}
+	}
+}
+
+// EnableBatchReading sets the TUN fd to non-blocking mode to enable batch reading.
+// This should be called before using ReadBatch.
+func (t *tun) EnableBatchReading() error {
+	if t.nonBlocking {
+		return nil
+	}
+	err := unix.SetNonblock(t.fd, true)
+	if err != nil {
+		return err
+	}
+	t.nonBlocking = true
+	return nil
+}
+
+// Read overrides the default Read to handle non-blocking mode and virtio headers
+func (t *tun) Read(b []byte) (int, error) {
+	if !t.vnetHdr {
+		return t.readSimple(b)
+	}
+
+	// With VNET_HDR, read into internal buffer (which has space for header)
+	// then copy packet data to caller's buffer
+	if !t.nonBlocking {
+		n, err := t.ReadWriteCloser.Read(t.readBuf)
+		if err != nil {
+			return 0, err
+		}
+		if n <= virtioNetHdrLen {
+			return 0, nil // No packet data
+		}
+		packetLen := n - virtioNetHdrLen
+		copy(b, t.readBuf[virtioNetHdrLen:n])
+		return packetLen, nil
+	}
+
+	// Non-blocking read with poll
+	for {
+		n, err := unix.Read(t.fd, t.readBuf)
+		if err == nil {
+			if n <= virtioNetHdrLen {
+				return 0, nil // No packet data
+			}
+			packetLen := n - virtioNetHdrLen
+			copy(b, t.readBuf[virtioNetHdrLen:n])
+			return packetLen, nil
+		}
+		if err == unix.EAGAIN || err == unix.EWOULDBLOCK {
+			pfds := []unix.PollFd{{Fd: int32(t.fd), Events: unix.POLLIN}}
+			_, err = unix.Poll(pfds, -1)
+			if err != nil {
+				if err == unix.EINTR {
+					continue
+				}
+				return 0, err
+			}
+			continue
+		}
+		return n, err
+	}
+}
+
+func (t *tun) readSimple(b []byte) (int, error) {
+	if !t.nonBlocking {
+		return t.ReadWriteCloser.Read(b)
+	}
+
+	for {
+		n, err := unix.Read(t.fd, b)
+		if err == nil {
+			return n, nil
+		}
+		if err == unix.EAGAIN || err == unix.EWOULDBLOCK {
+			pfds := []unix.PollFd{{Fd: int32(t.fd), Events: unix.POLLIN}}
+			_, err = unix.Poll(pfds, -1)
+			if err != nil {
+				if err == unix.EINTR {
+					continue
+				}
+				return 0, err
+			}
+			continue
+		}
+		return n, err
+	}
+}
+
+// ReadBatch reads up to len(packets) packets from the TUN device.
+// EnableBatchReading must be called first.
+func (t *tun) ReadBatch(packets [][]byte, sizes []int) (int, error) {
+	if !t.nonBlocking {
+		// Fallback to single read if non-blocking not enabled
+		n, err := t.Read(packets[0])
+		if err != nil {
+			return 0, err
+		}
+		sizes[0] = n
+		return 1, nil
+	}
+
+	count := 0
+	maxPackets := len(packets)
+	if len(sizes) < maxPackets {
+		maxPackets = len(sizes)
+	}
+
+	// Choose read buffer based on vnetHdr
+	// With vnetHdr, we need to read into internal buffer (has space for header)
+	// then copy packet data to caller's buffer
+	readBuf := packets[0] // Will be updated in the loop
+	if t.vnetHdr {
+		readBuf = t.readBuf
+	}
+
+	for count < maxPackets {
+		if !t.vnetHdr {
+			readBuf = packets[count]
+		}
+
+		n, err := unix.Read(t.fd, readBuf)
+		if err == nil && n > 0 {
+			if t.vnetHdr {
+				if n > virtioNetHdrLen {
+					packetLen := n - virtioNetHdrLen
+					copy(packets[count], readBuf[virtioNetHdrLen:n])
+					sizes[count] = packetLen
+				} else {
+					// Malformed packet (no data after header), skip
+					continue
+				}
+			} else {
+				sizes[count] = n
+			}
+			count++
+			continue
+		}
+
+		if err == unix.EAGAIN || err == unix.EWOULDBLOCK {
+			// No more packets available
+			if count > 0 {
+				return count, nil
+			}
+			// No packets yet, wait for at least one
+			pfds := []unix.PollFd{{Fd: int32(t.fd), Events: unix.POLLIN}}
+			_, err = unix.Poll(pfds, -1)
+			if err != nil {
+				if err == unix.EINTR {
+					continue
+				}
+				return 0, err
+			}
+			continue
+		}
+
+		if err != nil {
+			if count > 0 {
+				return count, nil
+			}
+			return 0, err
+		}
+
+		if n == 0 {
+			if count > 0 {
+				return count, nil
+			}
+			return 0, io.EOF
+		}
+	}
+
+	return count, nil
+}
+
 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
-}

relay_manager.go

@@ -55,7 +55,7 @@ func (rm *relayManager) setAmRelay(v bool) {
 func AddRelay(l *logrus.Logger, relayHostInfo *HostInfo, hm *HostMap, vpnIp netip.Addr, remoteIdx *uint32, relayType int, state int) (uint32, error) {
 	hm.Lock()
 	defer hm.Unlock()
-	for range 32 {
+	for i := 0; i < 32; i++ {
 		index, err := generateIndex(l)
 		if err != nil {
 			return 0, err

remote_list.go

@@ -404,7 +404,12 @@ func (r *RemoteList) Rebuild(preferredRanges []netip.Prefix) {
 
 // unlockedIsBad assumes you have the write lock and checks if the remote matches any entry in the blocked address list
 func (r *RemoteList) unlockedIsBad(remote netip.AddrPort) bool {
-	return slices.Contains(r.badRemotes, remote)
+	for _, v := range r.badRemotes {
+		if v == remote {
+			return true
+		}
+	}
+	return false
 }
 
 // unlockedSetLearnedV4 assumes you have the write lock and sets the current learned address for this owner and marks the

ssh.go

@@ -6,11 +6,9 @@ import (
 	"errors"
 	"flag"
 	"fmt"
-	"maps"
 	"net"
 	"net/netip"
 	"os"
-	"path/filepath"
 	"reflect"
 	"runtime"
 	"runtime/pprof"
@@ -189,12 +187,6 @@ 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 +245,7 @@ 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 {
+		Callback:         sshStartCpuProfile,
-			return sshStartCpuProfile(sandboxDir, fs, a, w)
-		},
 	})
 
 	ssh.RegisterCommand(&sshd.Command{
@@ -270,9 +260,7 @@ 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 {
+		Callback:         sshGetHeapProfile,
-			return sshGetHeapProfile(sandboxDir, fs, a, w)
-		},
 	})
 
 	ssh.RegisterCommand(&sshd.Command{
@@ -284,9 +272,7 @@ 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 {
+		Callback:         sshGetMutexProfile,
-			return sshGetMutexProfile(sandboxDir, fs, a, w)
-		},
 	})
 
 	ssh.RegisterCommand(&sshd.Command{
@@ -519,43 +505,13 @@ func sshListLighthouseMap(lightHouse *LightHouse, a any, w sshd.StringWriter) er
 	return nil
 }
 
-// sshSanitizeFilePath validates that the given file path is within the sandbox directory.
+func sshStartCpuProfile(fs any, a []string, w sshd.StringWriter) error {
-// 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 {
 	if len(a) == 0 {
 		err := w.WriteLine("No path to write profile provided")
 		return err
 	}
 
-	filePath, err := sshSanitizeFilePath(sandboxDir, a[0])
+	file, err := os.Create(a[0])
-	if err != nil {
-		return w.WriteLine(err.Error())
-	}
-
-	file, err := os.Create(filePath)
 	if err != nil {
 		err = w.WriteLine(fmt.Sprintf("Unable to create profile file: %s", err))
 		return err
@@ -719,17 +675,12 @@ 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])
+	file, err := os.Create(a[0])
-	if err != nil {
-		return w.WriteLine(err.Error())
-	}
-
-	file, err := os.Create(filePath)
 	if err != nil {
 		err = w.WriteLine(fmt.Sprintf("Unable to create profile file: %s", err))
 		return err
@@ -760,17 +711,12 @@ 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])
+	file, err := os.Create(a[0])
-	if err != nil {
-		return w.WriteLine(err.Error())
-	}
-
-	file, err := os.Create(filePath)
 	if err != nil {
 		return w.WriteLine(fmt.Sprintf("Unable to create profile file: %s", err))
 	}
@@ -885,7 +831,9 @@ func sshPrintRelays(ifce *Interface, fs any, a []string, w sshd.StringWriter) er
 
 	relays := map[uint32]*HostInfo{}
 	ifce.hostMap.Lock()
-	maps.Copy(relays, ifce.hostMap.Relays)
+	for k, v := range ifce.hostMap.Relays {
+		relays[k] = v
+	}
 	ifce.hostMap.Unlock()
 
 	type RelayFor struct {

timeout_test.go

@@ -134,7 +134,7 @@ func TestTimerWheel_Purge(t *testing.T) {
 	assert.True(t, tw.lastTick.After(lastTick))
 
 	// Make sure we get all 4 packets back
-	for i := range 4 {
+	for i := 0; i < 4; i++ {
 		p, has := tw.Purge()
 		assert.True(t, has)
 		assert.Equal(t, fps[i], p)
@@ -149,7 +149,7 @@ func TestTimerWheel_Purge(t *testing.T) {
 	// Make sure we cached the free'd items
 	assert.Equal(t, 4, tw.itemsCached)
 	ci := tw.itemCache
-	for range 4 {
+	for i := 0; i < 4; i++ {
 		assert.NotNil(t, ci)
 		ci = ci.Next
 	}

udp/conn.go

@@ -13,13 +13,22 @@ type EncReader func(
 	payload []byte,
 )
 
+// BatchPacket represents a single packet in a batch write operation
+type BatchPacket struct {
+	Payload []byte
+	Addr    netip.AddrPort
+}
+
 type Conn interface {
 	Rebind() error
 	LocalAddr() (netip.AddrPort, error)
 	ListenOut(r EncReader)
 	WriteTo(b []byte, addr netip.AddrPort) error
+	WriteBatch(pkts []BatchPacket) (int, error)
 	ReloadConfig(c *config.C)
 	SupportsMultipleReaders() bool
+	SupportsGSO() bool
+	SupportsGRO() bool
 	Close() error
 }
 
@@ -37,9 +46,18 @@ func (NoopConn) ListenOut(_ EncReader) {
 func (NoopConn) SupportsMultipleReaders() bool {
 	return false
 }
+func (NoopConn) SupportsGSO() bool {
+	return false
+}
+func (NoopConn) SupportsGRO() bool {
+	return false
+}
 func (NoopConn) WriteTo(_ []byte, _ netip.AddrPort) error {
 	return nil
 }
+func (NoopConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	return len(pkts), nil
+}
 func (NoopConn) ReloadConfig(_ *config.C) {
 	return
 }

udp/udp_darwin.go

@@ -188,6 +188,14 @@ func (u *StdConn) SupportsMultipleReaders() bool {
 	return false
 }
 
+func (u *StdConn) SupportsGSO() bool {
+	return false
+}
+
+func (u *StdConn) SupportsGRO() bool {
+	return false
+}
+
 func (u *StdConn) Rebind() error {
 	var err error
 	if u.isV4 {
@@ -202,3 +210,12 @@ func (u *StdConn) Rebind() error {
 
 	return nil
 }
+
+func (u *StdConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	for i := range pkts {
+		if err := u.WriteTo(pkts[i].Payload, pkts[i].Addr); err != nil {
+			return i, err
+		}
+	}
+	return len(pkts), nil
+}

udp/udp_generic.go

@@ -101,3 +101,20 @@ func (u *GenericConn) ListenOut(r EncReader) {
 func (u *GenericConn) SupportsMultipleReaders() bool {
 	return false
 }
+
+func (u *GenericConn) SupportsGSO() bool {
+	return false
+}
+
+func (u *GenericConn) SupportsGRO() bool {
+	return false
+}
+
+func (u *GenericConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	for i := range pkts {
+		if err := u.WriteTo(pkts[i].Payload, pkts[i].Addr); err != nil {
+			return i, err
+		}
+	}
+	return len(pkts), nil
+}

udp/udp_linux.go

@@ -4,8 +4,8 @@
 package udp
 
 import (
-	"context"
 	"encoding/binary"
+	"errors"
 	"fmt"
 	"net"
 	"net/netip"
@@ -19,226 +19,367 @@ import (
 )
 
 type StdConn struct {
-	udpConn *net.UDPConn
+	sysFd        int
-	rawConn syscall.RawConn
+	isV4         bool
-	isV4    bool
+	l            *logrus.Logger
-	l       *logrus.Logger
+	batch        int
-	batch   int
+	gsoSupported bool
+	groSupported bool
 }
 
-func setReusePort(network, address string, c syscall.RawConn) error {
+func maybeIPV4(ip net.IP) (net.IP, bool) {
-	var opErr error
+	ip4 := ip.To4()
-	err := c.Control(func(fd uintptr) {
+	if ip4 != nil {
-		opErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, unix.SO_REUSEPORT, 1)
+		return ip4, true
-		//CloseOnExec already set by the runtime
-	})
-	if err != nil {
-		return err
 	}
-	return opErr
+	return ip, false
+}
+
+// supportsUDPOffload checks if the kernel supports UDP GSO (Generic Segmentation Offload)
+// by attempting to get the UDP_SEGMENT socket option.
+func supportsUDPOffload(fd int) bool {
+	_, err := unix.GetsockoptInt(fd, unix.IPPROTO_UDP, unix.UDP_SEGMENT)
+	return err == nil
+}
+
+// supportsUDPGRO checks if the kernel supports UDP GRO (Generic Receive Offload)
+// and attempts to enable it on the socket.
+func supportsUDPGRO(fd int) bool {
+	// Try to enable UDP_GRO
+	err := unix.SetsockoptInt(fd, unix.IPPROTO_UDP, unix.UDP_GRO, 1)
+	return err == nil
+}
+
+const (
+	// Maximum number of datagrams that can be coalesced with GSO/GRO
+	udpSegmentMaxDatagrams = 64
+
+	// Maximum size of a GRO coalesced packet (64KB is the practical limit)
+	// This is udpSegmentMaxDatagrams * MTU but capped at 65535
+	groMaxPacketSize = 65535
+)
+
+// setGSOSize writes a UDP_SEGMENT control message to the provided buffer
+// with the given segment size. Returns the actual control message length.
+func setGSOSize(control []byte, gsoSize uint16) int {
+	// Build the cmsghdr structure
+	cmsgLen := unix.CmsgLen(2) // 2 bytes for uint16 segment size
+	cmsg := (*unix.Cmsghdr)(unsafe.Pointer(&control[0]))
+	cmsg.Level = unix.IPPROTO_UDP
+	cmsg.Type = unix.UDP_SEGMENT
+	cmsg.SetLen(cmsgLen)
+
+	// Write the segment size after the header (after cmsghdr)
+	binary.NativeEndian.PutUint16(control[unix.SizeofCmsghdr:], gsoSize)
+
+	return unix.CmsgSpace(2) // aligned size
+}
+
+// getGROSize parses a control message buffer to extract the UDP_GRO segment size.
+// Returns 0 if no GRO control message is present (meaning the packet is not coalesced).
+func getGROSize(control []byte, controlLen int) uint16 {
+	if controlLen < unix.SizeofCmsghdr {
+		return 0
+	}
+
+	// Parse control messages
+	for offset := 0; offset < controlLen; {
+		if offset+unix.SizeofCmsghdr > controlLen {
+			break
+		}
+
+		cmsg := (*unix.Cmsghdr)(unsafe.Pointer(&control[offset]))
+		cmsgDataLen := int(cmsg.Len) - unix.SizeofCmsghdr
+		if cmsgDataLen < 0 {
+			break
+		}
+
+		if cmsg.Level == unix.IPPROTO_UDP && cmsg.Type == unix.UDP_GRO {
+			if cmsgDataLen >= 2 {
+				return binary.NativeEndian.Uint16(control[offset+unix.SizeofCmsghdr:])
+			}
+		}
+
+		// Move to next control message (aligned)
+		offset += unix.CmsgSpace(cmsgDataLen)
+	}
+
+	return 0
 }
 
 func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
-	listen := netip.AddrPortFrom(ip, uint16(port))
+	af := unix.AF_INET6
-	lc := net.ListenConfig{}
+	if ip.Is4() {
-	if multi {
+		af = unix.AF_INET
-		lc.Control = setReusePort
 	}
-	//this context is only used during the bind operation, you can't cancel it to kill the socket
+	syscall.ForkLock.RLock()
-	pc, err := lc.ListenPacket(context.Background(), "udp", listen.String())
+	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 multi {
-	if err != nil {
+		if err = unix.SetsockoptInt(fd, unix.SOL_SOCKET, unix.SO_REUSEPORT, 1); err != nil {
-		_ = udpConn.Close()
+			return nil, fmt.Errorf("unable to set SO_REUSEPORT: %s", err)
-		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,
 	}
 
-	af, err := out.getSockOptInt(unix.SO_DOMAIN)
+	var sa unix.Sockaddr
-	if err != nil {
+	if ip.Is4() {
-		_ = out.Close()
+		sa4 := &unix.SockaddrInet4{Port: port}
-		return nil, err
+		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
+	gsoSupported := supportsUDPOffload(fd)
+	if gsoSupported {
+		l.Info("UDP GSO offload is supported")
+	}
+
+	groSupported := supportsUDPGRO(fd)
+	if groSupported {
+		l.Info("UDP GRO offload is supported and enabled")
+	}
+
+	return &StdConn{sysFd: fd, isV4: ip.Is4(), l: l, batch: batch, gsoSupported: gsoSupported, groSupported: groSupported}, err
 }
 
 func (u *StdConn) SupportsMultipleReaders() bool {
 	return true
 }
 
+func (u *StdConn) SupportsGSO() bool {
+	return u.gsoSupported
+}
+
+func (u *StdConn) SupportsGRO() bool {
+	return u.groSupported
+}
+
 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) {
+	switch sa := sa.(type) {
-	case *net.UDPAddr:
+	case *unix.SockaddrInet4:
-		addr, ok := netip.AddrFromSlice(v.IP)
+		return netip.AddrPortFrom(netip.AddrFrom4(sa.Addr), uint16(sa.Port)), nil
-		if !ok {
+
-			return netip.AddrPort{}, fmt.Errorf("LocalAddr returned invalid IP address: %s", v.IP)
+	case *unix.SockaddrInet6:
-		}
+		return netip.AddrPortFrom(netip.AddrFrom16(sa.Addr), uint16(sa.Port)), nil
-		return netip.AddrPortFrom(addr, uint16(v.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) {
+func (u *StdConn) ListenOut(r EncReader) {
-	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) {
 	var ip netip.Addr
-	var n int
-	var operr error
 
-	msgs, buffers, names := u.PrepareRawMessages(u.batch)
+	msgs, buffers, names, controls := u.PrepareRawMessages(u.batch)
+	read := u.ReadMulti
+	if u.batch == 1 {
+		read = u.ReadSingle
+	}
 
-	//reader needs to capture variables from this function, since it's used as a lambda with rawConn.Read
+	// Store the original control buffer size for resetting after each read
-	//defining it outside the loop so it gets re-used
+	controlLen := 0
-	reader := func(fd uintptr) (done bool) {
+	if u.groSupported && len(controls) > 0 && len(controls[0]) > 0 {
-		n, done, operr = recvmmsg(fd, msgs)
+		controlLen = len(controls[0])
-		return done
 	}
 
 	for {
-		err := u.rawConn.Read(reader)
+		// Reset Controllen before each read - the kernel updates this field
+		// after recvmsg to indicate actual received control data length
+		if controlLen > 0 {
+			for i := range msgs {
+				setMsghdrControllen(&msgs[i].Hdr, controlLen)
+			}
+		}
+
+		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
+			// Extract source address
 			if u.isV4 {
 				ip, _ = netip.AddrFromSlice(names[i][4:8])
 			} else {
 				ip, _ = netip.AddrFromSlice(names[i][8:24])
 			}
-			r(netip.AddrPortFrom(ip.Unmap(), binary.BigEndian.Uint16(names[i][2:4])), buffers[i][:msgs[i].Len])
+			srcAddr := netip.AddrPortFrom(ip.Unmap(), binary.BigEndian.Uint16(names[i][2:4]))
+
+			// Check for GRO coalesced packet
+			totalLen := int(msgs[i].Len)
+			segmentSize := uint16(0)
+			if controlLen > 0 {
+				segmentSize = getGROSize(controls[i], getMsghdrControllen(&msgs[i].Hdr))
+			}
+
+			if segmentSize > 0 && totalLen > int(segmentSize) {
+				// This is a GRO coalesced packet - split it into individual datagrams
+				for offset := 0; offset < totalLen; {
+					packetLen := int(segmentSize)
+					if offset+packetLen > totalLen {
+						// Last packet may be smaller
+						packetLen = totalLen - offset
+					}
+					r(srcAddr, buffers[i][offset:offset+packetLen])
+					offset += packetLen
+				}
+			} else {
+				// Single packet, no coalescing
+				r(srcAddr, buffers[i][:totalLen])
+			}
 		}
 	}
 }
 
-func (u *StdConn) ListenOut(r EncReader) {
+func (u *StdConn) ReadSingle(msgs []rawMessage) (int, error) {
-	if u.batch == 1 {
+	for {
-		//save some ram by not calling PrepareRawMessages for fields we won't use
+		n, _, err := unix.Syscall6(
-		//we could also make this path more common by calling recvmmsg with msgs[:1],
+			unix.SYS_RECVMSG,
-		//but that's still the recvmmsg syscall, which would be a change
+			uintptr(u.sysFd),
-		u.listenOutSingle(r)
+			uintptr(unsafe.Pointer(&(msgs[0].Hdr))),
-	} else {
+			0,
-		u.listenOutBatch(r)
+			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)
+	if u.isV4 {
-	return err
+		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 +432,235 @@ 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
-
+	_, _, 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 u.rawConn == nil {
+	if err != 0 {
-		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 {
 		return err
 	}
-	return opErr
+	return nil
 }
 
 func (u *StdConn) Close() error {
-	if u.udpConn != nil {
+	return syscall.Close(u.sysFd)
-		return u.udpConn.Close()
+}
+
+func (u *StdConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	if len(pkts) == 0 {
+		return 0, nil
 	}
-	return nil
+
+	// If GSO is supported, try to coalesce packets to the same destination
+	if u.gsoSupported {
+		return u.writeBatchGSO(pkts)
+	}
+
+	return u.writeBatchSendmmsg(pkts)
+}
+
+// writeBatchSendmmsg sends packets using sendmmsg without GSO coalescing
+func (u *StdConn) writeBatchSendmmsg(pkts []BatchPacket) (int, error) {
+	msgs := make([]rawMessage, len(pkts))
+	iovecs := make([]iovec, len(pkts))
+	var names4 []unix.RawSockaddrInet4
+	var names6 []unix.RawSockaddrInet6
+
+	if u.isV4 {
+		names4 = make([]unix.RawSockaddrInet4, len(pkts))
+	} else {
+		names6 = make([]unix.RawSockaddrInet6, len(pkts))
+	}
+
+	for i := range pkts {
+		setIovecBase(&iovecs[i], &pkts[i].Payload[0])
+		setIovecLen(&iovecs[i], len(pkts[i].Payload))
+		msgs[i].Hdr.Iov = &iovecs[i]
+		setMsghdrIovlen(&msgs[i].Hdr, 1)
+
+		if u.isV4 {
+			names4[i].Family = unix.AF_INET
+			names4[i].Addr = pkts[i].Addr.Addr().As4()
+			binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&names4[i].Port))[:], pkts[i].Addr.Port())
+			msgs[i].Hdr.Name = (*byte)(unsafe.Pointer(&names4[i]))
+			msgs[i].Hdr.Namelen = unix.SizeofSockaddrInet4
+		} else {
+			names6[i].Family = unix.AF_INET6
+			names6[i].Addr = pkts[i].Addr.Addr().As16()
+			binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&names6[i].Port))[:], pkts[i].Addr.Port())
+			msgs[i].Hdr.Name = (*byte)(unsafe.Pointer(&names6[i]))
+			msgs[i].Hdr.Namelen = unix.SizeofSockaddrInet6
+		}
+	}
+
+	var sent int
+	for sent < len(msgs) {
+		n, _, errno := unix.Syscall6(
+			unix.SYS_SENDMMSG,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&msgs[sent])),
+			uintptr(len(msgs)-sent),
+			0,
+			0,
+			0,
+		)
+
+		if errno == unix.EINTR {
+			continue
+		}
+
+		if errno != 0 {
+			return sent, &net.OpError{Op: "sendmmsg", Err: errno}
+		}
+
+		sent += int(n)
+	}
+
+	return sent, nil
+}
+
+// writeBatchGSO sends packets using GSO coalescing when possible.
+// Packets to the same destination with the same size are coalesced into a single
+// GSO message. Mixed destinations or sizes fall back to individual sendmmsg calls.
+func (u *StdConn) writeBatchGSO(pkts []BatchPacket) (int, error) {
+	// Group packets by destination and try to coalesce
+	totalSent := 0
+	i := 0
+
+	for i < len(pkts) {
+		// Find a run of packets to the same destination with compatible sizes
+		startIdx := i
+		dst := pkts[i].Addr
+		segmentSize := len(pkts[i].Payload)
+
+		// Count how many packets we can coalesce (same destination, same size except possibly last)
+		coalescedCount := 1
+		totalSize := segmentSize
+		for i+coalescedCount < len(pkts) && coalescedCount < udpSegmentMaxDatagrams {
+			next := pkts[i+coalescedCount]
+			if next.Addr != dst {
+				break
+			}
+			nextSize := len(next.Payload)
+			// For GSO, all packets except the last must have the same size
+			// The last packet can be smaller (but not larger)
+			if nextSize != segmentSize {
+				// Check if this could be the last packet (smaller is ok)
+				if nextSize < segmentSize && i+coalescedCount == len(pkts)-1 {
+					coalescedCount++
+					totalSize += nextSize
+				}
+				break
+			}
+			coalescedCount++
+			totalSize += nextSize
+		}
+
+		// If we have multiple packets to coalesce, use GSO
+		if coalescedCount > 1 {
+			err := u.sendGSO(pkts[startIdx:startIdx+coalescedCount], dst, segmentSize, totalSize)
+			if err != nil {
+				// If GSO fails (e.g., EIO due to NIC not supporting checksum offload),
+				// disable GSO and fall back to sendmmsg for the rest
+				if isGSOError(err) {
+					u.l.WithError(err).Warn("GSO send failed, disabling GSO for this connection")
+					u.gsoSupported = false
+					// Send remaining packets with sendmmsg
+					remaining, rerr := u.writeBatchSendmmsg(pkts[startIdx:])
+					return totalSent + remaining, rerr
+				}
+				return totalSent, err
+			}
+			totalSent += coalescedCount
+			i += coalescedCount
+		} else {
+			// Single packet, send without GSO overhead
+			err := u.WriteTo(pkts[i].Payload, pkts[i].Addr)
+			if err != nil {
+				return totalSent, err
+			}
+			totalSent++
+			i++
+		}
+	}
+
+	return totalSent, nil
+}
+
+// sendGSO sends coalesced packets using UDP GSO
+func (u *StdConn) sendGSO(pkts []BatchPacket, dst netip.AddrPort, segmentSize, totalSize int) error {
+	// Allocate a buffer large enough for all packet payloads
+	coalescedBuf := make([]byte, totalSize)
+	offset := 0
+	for _, pkt := range pkts {
+		copy(coalescedBuf[offset:], pkt.Payload)
+		offset += len(pkt.Payload)
+	}
+
+	// Prepare control message with GSO segment size
+	control := make([]byte, unix.CmsgSpace(2))
+	controlLen := setGSOSize(control, uint16(segmentSize))
+
+	// Prepare the iovec
+	iov := iovec{}
+	setIovecBase(&iov, &coalescedBuf[0])
+	setIovecLen(&iov, totalSize)
+
+	// Prepare the msghdr
+	var hdr msghdr
+	hdr.Iov = &iov
+	setMsghdrIovlen(&hdr, 1)
+	hdr.Control = &control[0]
+	setMsghdrControllen(&hdr, controlLen)
+
+	// Declare sockaddr at function scope so it remains valid for the syscall
+	// (must not go out of scope before the syscall is made)
+	var rsa4 unix.RawSockaddrInet4
+	var rsa6 unix.RawSockaddrInet6
+
+	// Set destination address
+	if u.isV4 {
+		rsa4.Family = unix.AF_INET
+		rsa4.Addr = dst.Addr().As4()
+		binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa4.Port))[:], dst.Port())
+		hdr.Name = (*byte)(unsafe.Pointer(&rsa4))
+		hdr.Namelen = unix.SizeofSockaddrInet4
+	} else {
+		rsa6.Family = unix.AF_INET6
+		rsa6.Addr = dst.Addr().As16()
+		binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa6.Port))[:], dst.Port())
+		hdr.Name = (*byte)(unsafe.Pointer(&rsa6))
+		hdr.Namelen = unix.SizeofSockaddrInet6
+	}
+
+	for {
+		_, _, errno := unix.Syscall6(
+			unix.SYS_SENDMSG,
+			uintptr(u.sysFd),
+			uintptr(unsafe.Pointer(&hdr)),
+			0,
+			0,
+			0,
+			0,
+		)
+
+		if errno == unix.EINTR {
+			continue
+		}
+
+		if errno != 0 {
+			return &net.OpError{Op: "sendmsg", Err: errno}
+		}
+
+		return nil
+	}
+}
+
+// isGSOError returns true if the error indicates GSO is not supported by the NIC
+func isGSOError(err error) bool {
+	var opErr *net.OpError
+	if !errors.As(err, &opErr) {
+		return false
+	}
+	// EIO typically means the NIC doesn't support checksum offload required for GSO
+	return errors.Is(opErr.Err, unix.EIO)
 }
 
 func NewUDPStatsEmitter(udpConns []Conn) func() {

udp/udp_linux_32.go

@@ -30,13 +30,26 @@ type rawMessage struct {
 	Len uint32
 }
 
-func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
+func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte, [][]byte) {
 	msgs := make([]rawMessage, n)
 	buffers := make([][]byte, n)
 	names := make([][]byte, n)
+	controls := make([][]byte, n)
+
+	// Use larger buffers if GRO is enabled to hold coalesced packets
+	bufSize := MTU
+	if u.groSupported {
+		bufSize = groMaxPacketSize
+	}
+
+	// Control buffer size for receiving UDP_GRO segment size
+	controlSize := 0
+	if u.groSupported {
+		controlSize = unix.CmsgSpace(2) // space for uint16 segment size
+	}
 
 	for i := range msgs {
-		buffers[i] = make([]byte, MTU)
+		buffers[i] = make([]byte, bufSize)
 		names[i] = make([]byte, unix.SizeofSockaddrInet6)
 
 		vs := []iovec{
@@ -48,7 +61,34 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
 
 		msgs[i].Hdr.Name = &names[i][0]
 		msgs[i].Hdr.Namelen = uint32(len(names[i]))
+
+		// Set up control message buffer for GRO
+		if controlSize > 0 {
+			controls[i] = make([]byte, controlSize)
+			msgs[i].Hdr.Control = &controls[i][0]
+			msgs[i].Hdr.Controllen = uint32(controlSize)
+		}
 	}
 
-	return msgs, buffers, names
+	return msgs, buffers, names, controls
+}
+
+func setIovecBase(iov *iovec, base *byte) {
+	iov.Base = base
+}
+
+func setIovecLen(iov *iovec, l int) {
+	iov.Len = uint32(l)
+}
+
+func setMsghdrIovlen(hdr *msghdr, l int) {
+	hdr.Iovlen = uint32(l)
+}
+
+func setMsghdrControllen(hdr *msghdr, l int) {
+	hdr.Controllen = uint32(l)
+}
+
+func getMsghdrControllen(hdr *msghdr) int {
+	return int(hdr.Controllen)
 }

udp/udp_linux_64.go

@@ -33,13 +33,26 @@ type rawMessage struct {
 	Pad0 [4]byte
 }
 
-func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
+func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte, [][]byte) {
 	msgs := make([]rawMessage, n)
 	buffers := make([][]byte, n)
 	names := make([][]byte, n)
+	controls := make([][]byte, n)
+
+	// Use larger buffers if GRO is enabled to hold coalesced packets
+	bufSize := MTU
+	if u.groSupported {
+		bufSize = groMaxPacketSize
+	}
+
+	// Control buffer size for receiving UDP_GRO segment size
+	controlSize := 0
+	if u.groSupported {
+		controlSize = unix.CmsgSpace(2) // space for uint16 segment size
+	}
 
 	for i := range msgs {
-		buffers[i] = make([]byte, MTU)
+		buffers[i] = make([]byte, bufSize)
 		names[i] = make([]byte, unix.SizeofSockaddrInet6)
 
 		vs := []iovec{
@@ -51,7 +64,34 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
 
 		msgs[i].Hdr.Name = &names[i][0]
 		msgs[i].Hdr.Namelen = uint32(len(names[i]))
+
+		// Set up control message buffer for GRO
+		if controlSize > 0 {
+			controls[i] = make([]byte, controlSize)
+			msgs[i].Hdr.Control = &controls[i][0]
+			msgs[i].Hdr.Controllen = uint64(controlSize)
+		}
 	}
 
-	return msgs, buffers, names
+	return msgs, buffers, names, controls
+}
+
+func setIovecBase(iov *iovec, base *byte) {
+	iov.Base = base
+}
+
+func setIovecLen(iov *iovec, l int) {
+	iov.Len = uint64(l)
+}
+
+func setMsghdrIovlen(hdr *msghdr, l int) {
+	hdr.Iovlen = uint64(l)
+}
+
+func setMsghdrControllen(hdr *msghdr, l int) {
+	hdr.Controllen = uint64(l)
+}
+
+func getMsghdrControllen(hdr *msghdr) int {
+	return int(hdr.Controllen)
 }

udp/udp_rio_windows.go

@@ -332,12 +332,29 @@ func (u *RIOConn) SupportsMultipleReaders() bool {
 	return false
 }
 
+func (u *RIOConn) SupportsGSO() bool {
+	return false
+}
+
+func (u *RIOConn) SupportsGRO() bool {
+	return false
+}
+
 func (u *RIOConn) Rebind() error {
 	return nil
 }
 
 func (u *RIOConn) ReloadConfig(*config.C) {}
 
+func (u *RIOConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	for i := range pkts {
+		if err := u.WriteTo(pkts[i].Payload, pkts[i].Addr); err != nil {
+			return i, err
+		}
+	}
+	return len(pkts), nil
+}
+
 func (u *RIOConn) Close() error {
 	if !u.isOpen.CompareAndSwap(true, false) {
 		return nil

udp/udp_tester.go

@@ -131,6 +131,14 @@ func (u *TesterConn) SupportsMultipleReaders() bool {
 	return false
 }
 
+func (u *TesterConn) SupportsGSO() bool {
+	return false
+}
+
+func (u *TesterConn) SupportsGRO() bool {
+	return false
+}
+
 func (u *TesterConn) Rebind() error {
 	return nil
 }
@@ -142,3 +150,12 @@ func (u *TesterConn) Close() error {
 	}
 	return nil
 }
+
+func (u *TesterConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	for i := range pkts {
+		if err := u.WriteTo(pkts[i].Payload, pkts[i].Addr); err != nil {
+			return i, err
+		}
+	}
+	return len(pkts), nil
+}