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

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

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

.github/workflows/smoke-extra.yml

@@ -30,6 +30,9 @@ 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.sh

@@ -37,17 +37,18 @@ 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 nebula1 -q -w - -U 2>logs/lighthouse1.inside.log >logs/lighthouse1.inside.pcap &
+docker exec lighthouse1 tcpdump -i tun0 -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 nebula1 -q -w - -U 2>logs/host2.inside.log >logs/host2.inside.pcap &
+docker exec host2 tcpdump -i tun0 -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 nebula1 -q -w - -U 2>logs/host3.inside.log >logs/host3.inside.pcap &
+docker exec host3 tcpdump -i tun0 -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 nebula1 -q -w - -U 2>logs/host4.inside.log >logs/host4.inside.pcap &
+docker exec host4 tcpdump -i tun0 -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 &
 
@@ -119,11 +120,11 @@ echo
 echo " *** Testing conntrack"
 echo
 set -x
-# host2 can ping host3 now that host3 pinged it first
-docker exec host2 ping -c1 192.168.100.3
-# host4 can ping host2 once conntrack established
-docker exec host2 ping -c1 192.168.100.4
-docker exec host4 ping -c1 192.168.100.2
+
+# host2 speaking to host4 on UDP 4000 should allow it to reply, when firewall rules would normally not permit this
+docker exec host2 sh -c "/usr/bin/echo host2 | ncat -nuv 192.168.100.4 4000"
+docker exec host2 ncat -e '/usr/bin/echo helloagainfromhost2' -nkluv 0.0.0.0 4000 &
+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'

CHANGELOG.md

@@ -7,6 +7,19 @@ 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
@@ -775,7 +788,8 @@ created.)
 
 - Initial public release.
 
-[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.2...HEAD
+[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.3...HEAD
+[1.10.3]: https://github.com/slackhq/nebula/releases/tag/v1.10.3
 [1.10.2]: https://github.com/slackhq/nebula/releases/tag/v1.10.2
 [1.10.1]: https://github.com/slackhq/nebula/releases/tag/v1.10.1
 [1.10.0]: https://github.com/slackhq/nebula/releases/tag/v1.10.0

CODEOWNERS

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

boring.go

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

cert/ca_pool.go

@@ -141,10 +141,23 @@ 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,
 	}
 
@@ -158,6 +171,11 @@ 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

@@ -5,6 +5,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/slackhq/nebula/cert/p256"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -170,6 +171,15 @@ 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)
@@ -187,7 +197,7 @@ 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)
 
@@ -196,7 +206,17 @@ 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"})
-	_, err = caPool.VerifyCertificate(time.Now(), c)
+	cc, 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)
 }
 
@@ -394,6 +414,15 @@ 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)
@@ -411,7 +440,7 @@ 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)
 
@@ -420,7 +449,17 @@ 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"})
-	_, err = caPool.VerifyCertificate(time.Now(), c)
+	cc, 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

@@ -4,6 +4,8 @@ import (
 	"fmt"
 	"net/netip"
 	"time"
+
+	"github.com/slackhq/nebula/cert/p256"
 )
 
 type Version uint8
@@ -110,6 +112,9 @@ 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 {
@@ -152,3 +157,31 @@ 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

@@ -0,0 +1,122 @@
+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)
+
+	return r, sNormalized.Bytes(nMod), 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

@@ -0,0 +1,28 @@
+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/sign.go

@@ -9,6 +9,8 @@ import (
 	"fmt"
 	"net/netip"
 	"time"
+
+	"github.com/slackhq/nebula/cert/p256"
 )
 
 // TBSCertificate represents a certificate intended to be signed.
@@ -126,6 +128,13 @@ 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

@@ -9,6 +9,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/slackhq/nebula/cert/p256"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -89,3 +90,48 @@ 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-service/main.go

@@ -78,16 +78,8 @@ func main() {
 	}
 
 	if !*configTest {
-		wait, err := ctrl.Start()
-		if err != nil {
-			util.LogWithContextIfNeeded("Error while running", err, l)
-			os.Exit(1)
-		}
-
-		go ctrl.ShutdownBlock()
-		wait()
-
-		l.Info("Goodbye")
+		ctrl.Start()
+		ctrl.ShutdownBlock()
 	}
 
 	os.Exit(0)

cmd/nebula/main.go

@@ -72,17 +72,9 @@ func main() {
 	}
 
 	if !*configTest {
-		wait, err := ctrl.Start()
-		if err != nil {
-			util.LogWithContextIfNeeded("Error while running", err, l)
-			os.Exit(1)
-		}
-
-		go ctrl.ShutdownBlock()
+		ctrl.Start()
 		notifyReady(l)
-		wait()
-
-		l.Info("Goodbye")
+		ctrl.ShutdownBlock()
 	}
 
 	os.Exit(0)

control.go

@@ -2,11 +2,9 @@ package nebula
 
 import (
 	"context"
-	"errors"
 	"net/netip"
 	"os"
 	"os/signal"
-	"sync"
 	"syscall"
 
 	"github.com/sirupsen/logrus"
@@ -15,16 +13,6 @@ import (
 	"github.com/slackhq/nebula/overlay"
 )
 
-type RunState int
-
-const (
-	Stopped  RunState = 0 // The control has yet to be started
-	Started  RunState = 1 // The control has been started
-	Stopping RunState = 2 // The control is stopping
-)
-
-var ErrAlreadyStarted = errors.New("nebula is already started")
-
 // Every interaction here needs to take extra care to copy memory and not return or use arguments "as is" when touching
 // core. This means copying IP objects, slices, de-referencing pointers and taking the actual value, etc
 
@@ -38,9 +26,6 @@ type controlHostLister interface {
 }
 
 type Control struct {
-	stateLock sync.Mutex
-	state     RunState
-
 	f                      *Interface
 	l                      *logrus.Logger
 	ctx                    context.Context
@@ -64,21 +49,10 @@ type ControlHostInfo struct {
 	CurrentRelaysThroughMe []netip.Addr     `json:"currentRelaysThroughMe"`
 }
 
-// Start actually runs nebula, this is a nonblocking call.
-// The returned function can be used to wait for nebula to fully stop.
-func (c *Control) Start() (func(), error) {
-	c.stateLock.Lock()
-	if c.state != Stopped {
-		c.stateLock.Unlock()
-		return nil, ErrAlreadyStarted
-	}
-
+// Start actually runs nebula, this is a nonblocking call. To block use Control.ShutdownBlock()
+func (c *Control) Start() {
 	// Activate the interface
-	err := c.f.activate()
-	if err != nil {
-		c.stateLock.Unlock()
-		return nil, err
-	}
+	c.f.activate()
 
 	// Call all the delayed funcs that waited patiently for the interface to be created.
 	if c.sshStart != nil {
@@ -98,33 +72,15 @@ func (c *Control) Start() (func(), error) {
 	}
 
 	// Start reading packets.
-	c.state = Started
-	c.stateLock.Unlock()
-	return c.f.run()
-}
-
-func (c *Control) State() RunState {
-	c.stateLock.Lock()
-	defer c.stateLock.Unlock()
-	return c.state
+	c.f.run()
 }
 
 func (c *Control) Context() context.Context {
 	return c.ctx
 }
 
-// Stop is a non-blocking call that signals nebula to close all tunnels and shut down
+// Stop signals nebula to shutdown and close all tunnels, returns after the shutdown is complete
 func (c *Control) Stop() {
-	c.stateLock.Lock()
-	if c.state != Started {
-		c.stateLock.Unlock()
-		// We are stopping or stopped already
-		return
-	}
-
-	c.state = Stopping
-	c.stateLock.Unlock()
-
 	// Stop the handshakeManager (and other services), to prevent new tunnels from
 	// being created while we're shutting them all down.
 	c.cancel()
@@ -133,7 +89,7 @@ func (c *Control) Stop() {
 	if err := c.f.Close(); err != nil {
 		c.l.WithError(err).Error("Close interface failed")
 	}
-	c.state = Stopped
+	c.l.Info("Goodbye")
 }
 
 // ShutdownBlock will listen for and block on term and interrupt signals, calling Control.Stop() once signalled

control_tester.go

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

dist/wireshark/nebula.lua

@@ -84,30 +84,24 @@ 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 our old dissector
+    -- Remove all existing registrations
     DissectorTable.get("udp.port"):remove_all(nebula)
 
-    if nebula.prefs.all_ports and default_settings.all_ports ~= nebula.prefs.all_ports then
-        default_settings.all_port = nebula.prefs.all_ports
-
+    if nebula.prefs.all_ports then
+        -- Register on every port for hole punch capture
         for i=0, 65535 do
             DissectorTable.get("udp.port"):add(i, nebula)
         end
-
-        -- no need to establish again on specific ports
-        return
+    else
+        -- Register on the configured port only
+        DissectorTable.get("udp.port"):add(nebula.prefs.port, nebula)
     end
 
-
-    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
+    default_settings.all_ports = nebula.prefs.all_ports
+    default_settings.port = nebula.prefs.port
 end
 
 DissectorTable.get("udp.port"):add(default_settings.port, nebula)

e2e/tunnels_test.go

@@ -12,6 +12,8 @@ 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"
 )
@@ -365,3 +367,106 @@ 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

@@ -342,6 +342,14 @@ logging:
   # after receiving the response for lighthouse queries
   #trigger_buffer: 64
 
+  # max_rate limits the number of new inbound handshakes per second. Once the limit is reached,
+  # new handshakes are dropped until the next second. A value of 0 means unlimited (default).
+  # This is useful for preventing DoS attacks that attempt to exhaust CPU with handshake crypto.
+  # Running `openssl speed ecdhp256` on your hardware can be a good rule of thumb for choosing
+  # a max, as each handshake performs similar DH operations. Note that this benchmarks a single
+  # core, so you may wish to scale the value by the number of `routines` configured.
+  #max_rate: 0
+
 # Tunnel manager settings
 #tunnels:
   # drop_inactive controls whether inactive tunnels are maintained or dropped after the inactive_timeout period has
@@ -382,8 +390,8 @@ 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", inboundAction).Warn("invalid firewall.outbound_action, defaulting to `drop`")
+		l.WithField("action", outboundAction).Warn("invalid firewall.outbound_action, defaulting to `drop`")
 		fw.OutSendReject = false
 	}
 
@@ -249,20 +249,6 @@ 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
@@ -280,6 +266,12 @@ 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
@@ -287,6 +279,20 @@ 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)
 }
 
@@ -349,24 +355,31 @@ 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)
@@ -467,7 +480,7 @@ func (f *Firewall) metrics(incoming bool) firewallMetrics {
 	}
 }
 
-// Destroy cleans up any known cyclical references so the object can be free'd my GC. This should be called if a new
+// Destroy cleans up any known cyclical references so the object can be freed by GC. This should be called if a new
 // firewall object is created
 func (f *Firewall) Destroy() {
 	//TODO: clean references if/when needed
@@ -660,6 +673,13 @@ 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 {
@@ -804,10 +824,8 @@ func (fr *FirewallRule) isAny(groups []string, host string, cidr string) bool {
 		return true
 	}
 
-	for _, group := range groups {
-		if group == "any" {
-			return true
-		}
+	if slices.Contains(groups, "any") {
+		return true
 	}
 
 	if host == "any" {
@@ -1018,54 +1036,56 @@ 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) (startPort, endPort int32, err error) {
+func parsePort(s string) (int32, int32, error) {
+	var err error
+	const notAPort int32 = -2
 	if s == "any" {
-		startPort = firewall.PortAny
-		endPort = firewall.PortAny
-
-	} else if s == "fragment" {
-		startPort = firewall.PortFragment
-		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 {
+		return firewall.PortAny, firewall.PortAny, nil
+	}
+	if s == "fragment" {
+		return firewall.PortFragment, firewall.PortFragment, nil
+	}
+	if !strings.Contains(s, `-`) {
 		rPort, err := strconv.Atoi(s)
 		if err != nil {
-			return 0, 0, fmt.Errorf("was not a number; `%s`", s)
+			return notAPort, notAPort, fmt.Errorf("was not a number; `%s`", s)
 		}
-		startPort = int32(rPort)
-		endPort = startPort
+		return int32(rPort), int32(rPort), nil
 	}
 
-	return
+	sPorts := strings.SplitN(s, `-`, 2)
+	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

@@ -22,7 +22,10 @@ const (
 type Packet struct {
 	LocalAddr  netip.Addr
 	RemoteAddr netip.Addr
-	LocalPort  uint16
+	// LocalPort is the destination port for incoming traffic, or the source port for outgoing. Zero for ICMP.
+	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
@@ -46,6 +49,8 @@ func (fp Packet) MarshalJSON() ([]byte, error) {
 		proto = "tcp"
 	case ProtoICMP:
 		proto = "icmp"
+	case ProtoICMPv6:
+		proto = "icmpv6"
 	case ProtoUDP:
 		proto = "udp"
 	default:

firewall_test.go

@@ -87,9 +87,10 @@ 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", "", "", "", ""))
-	assert.Nil(t, fw.InRules.ICMP[1].Any.Any)
-	assert.Empty(t, fw.InRules.ICMP[1].Any.Groups)
-	assert.Contains(t, fw.InRules.ICMP[1].Any.Hosts, "h1")
+	//no matter what port is given for icmp, it should end up as "any"
+	assert.Nil(t, fw.InRules.ICMP[firewall.PortAny].Any.Any)
+	assert.Empty(t, fw.InRules.ICMP[firewall.PortAny].Any.Groups)
+	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(), "", "", ""))
@@ -734,6 +735,150 @@ 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{}
@@ -919,11 +1064,11 @@ 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"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh", "proto": "any"}}}
 	_, 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"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "a", "host": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`")
 
@@ -973,7 +1118,14 @@ 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: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
+	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 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

@@ -4,6 +4,7 @@ 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

go.sum

@@ -1,6 +1,8 @@
 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=

handshake_manager.go

@@ -23,22 +23,25 @@ const (
 	DefaultHandshakeRetries       = 10
 	DefaultHandshakeTriggerBuffer = 64
 	DefaultUseRelays              = true
+	DefaultMaxHandshakeRate       = 0 // 0 means unlimited
 )
 
 var (
 	defaultHandshakeConfig = HandshakeConfig{
-		tryInterval:   DefaultHandshakeTryInterval,
-		retries:       DefaultHandshakeRetries,
-		triggerBuffer: DefaultHandshakeTriggerBuffer,
-		useRelays:     DefaultUseRelays,
+		tryInterval:      DefaultHandshakeTryInterval,
+		retries:          DefaultHandshakeRetries,
+		triggerBuffer:    DefaultHandshakeTriggerBuffer,
+		useRelays:        DefaultUseRelays,
+		maxHandshakeRate: DefaultMaxHandshakeRate,
 	}
 )
 
 type HandshakeConfig struct {
-	tryInterval   time.Duration
-	retries       int64
-	triggerBuffer int
-	useRelays     bool
+	tryInterval      time.Duration
+	retries          int64
+	triggerBuffer    int
+	useRelays        bool
+	maxHandshakeRate int
 
 	messageMetrics *MessageMetrics
 }
@@ -58,9 +61,15 @@ type HandshakeManager struct {
 	messageMetrics         *MessageMetrics
 	metricInitiated        metrics.Counter
 	metricTimedOut         metrics.Counter
+	metricRateLimited      metrics.Counter
 	f                      *Interface
 	l                      *logrus.Logger
 
+	// Rate limiting for new handshakes (token bucket)
+	rateBucket   int // tokens currently available
+	rateMax      int // max tokens (== max handshakes per second), 0 means unlimited
+	rateLastTick time.Time
+
 	// can be used to trigger outbound handshake for the given vpnIp
 	trigger chan netip.Addr
 }
@@ -116,10 +125,41 @@ func NewHandshakeManager(l *logrus.Logger, mainHostMap *HostMap, lightHouse *Lig
 		messageMetrics:         config.messageMetrics,
 		metricInitiated:        metrics.GetOrRegisterCounter("handshake_manager.initiated", nil),
 		metricTimedOut:         metrics.GetOrRegisterCounter("handshake_manager.timed_out", nil),
+		metricRateLimited:      metrics.GetOrRegisterCounter("handshake_manager.rate_limited", nil),
+		rateBucket:             config.maxHandshakeRate,
+		rateMax:                config.maxHandshakeRate,
+		rateLastTick:           time.Now(),
 		l:                      l,
 	}
 }
 
+// handshakeRateAllow checks the token bucket rate limiter and returns true if a
+// new handshake is allowed. Must be called with hm.Lock held.
+func (hm *HandshakeManager) handshakeRateAllow(now time.Time) bool {
+	if hm.rateMax == 0 {
+		return true
+	}
+
+	// Refill tokens based on elapsed time
+	elapsed := now.Sub(hm.rateLastTick)
+	if elapsed >= time.Second {
+		// Add tokens for full seconds elapsed
+		tokens := int(elapsed/time.Second) * hm.rateMax
+		hm.rateBucket += tokens
+		if hm.rateBucket > hm.rateMax {
+			hm.rateBucket = hm.rateMax
+		}
+		hm.rateLastTick = now
+	}
+
+	if hm.rateBucket > 0 {
+		hm.rateBucket--
+		return true
+	}
+
+	return false
+}
+
 func (hm *HandshakeManager) Run(ctx context.Context) {
 	clockSource := time.NewTicker(hm.config.tryInterval)
 	defer clockSource.Stop()
@@ -149,6 +189,15 @@ func (hm *HandshakeManager) HandleIncoming(via ViaSender, packet []byte, h *head
 	case header.HandshakeIXPSK0:
 		switch h.MessageCounter {
 		case 1:
+			// Check rate limit for new incoming handshakes
+			hm.Lock()
+			allowed := hm.handshakeRateAllow(time.Now())
+			hm.Unlock()
+			if !allowed {
+				hm.metricRateLimited.Inc(1)
+				hm.l.WithField("from", via).Debug("Handshake rate limit reached, dropping incoming handshake")
+				return
+			}
 			ixHandshakeStage1(hm.f, via, packet, h)
 
 		case 2:
@@ -590,7 +639,7 @@ func (hm *HandshakeManager) allocateIndex(hh *HandshakeHostInfo) error {
 	hm.Lock()
 	defer hm.Unlock()
 
-	for i := 0; i < 32; i++ {
+	for range 32 {
 		index, err := generateIndex(hm.l)
 		if err != nil {
 			return err

handshake_manager_test.go

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

hostmap_tester.go

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

inside_bsd.go

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

inside_generic.go

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

interface.go

@@ -6,8 +6,8 @@ import (
 	"fmt"
 	"io"
 	"net/netip"
+	"os"
 	"runtime"
-	"sync"
 	"sync/atomic"
 	"time"
 
@@ -88,7 +88,6 @@ type Interface struct {
 
 	writers []udp.Conn
 	readers []io.ReadWriteCloser
-	wg      sync.WaitGroup
 
 	metricHandshakes    metrics.Histogram
 	messageMetrics      *MessageMetrics
@@ -211,7 +210,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 // activate creates the interface on the host. After the interface is created, any
 // other services that want to bind listeners to its IP may do so successfully. However,
 // the interface isn't going to process anything until run() is called.
-func (f *Interface) activate() error {
+func (f *Interface) activate() {
 	// actually turn on tun dev
 
 	addr, err := f.outside.LocalAddr()
@@ -239,38 +238,33 @@ func (f *Interface) activate() error {
 		if i > 0 {
 			reader, err = f.inside.NewMultiQueueReader()
 			if err != nil {
-				return err
+				f.l.Fatal(err)
 			}
 		}
 		f.readers[i] = reader
 	}
 
-	if err = f.inside.Activate(); err != nil {
+	if err := f.inside.Activate(); err != nil {
 		f.inside.Close()
-		return err
+		f.l.Fatal(err)
 	}
-
-	return nil
 }
 
-func (f *Interface) run() (func(), error) {
+func (f *Interface) run() {
 	// Launch n queues to read packets from udp
 	for i := 0; i < f.routines; i++ {
-		f.wg.Add(1)
 		go f.listenOut(i)
 	}
 
 	// Launch n queues to read packets from tun dev
 	for i := 0; i < f.routines; i++ {
-		f.wg.Add(1)
 		go f.listenIn(f.readers[i], i)
 	}
-
-	return f.wg.Wait, nil
 }
 
 func (f *Interface) listenOut(i int) {
 	runtime.LockOSThread()
+
 	var li udp.Conn
 	if i > 0 {
 		li = f.writers[i]
@@ -285,21 +279,14 @@ func (f *Interface) listenOut(i int) {
 	fwPacket := &firewall.Packet{}
 	nb := make([]byte, 12, 12)
 
-	err := li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
+	li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
 		f.readOutsidePackets(ViaSender{UdpAddr: fromUdpAddr}, plaintext[:0], payload, h, fwPacket, lhh, nb, i, ctCache.Get(f.l))
 	})
-
-	if err != nil && !f.closed.Load() {
-		f.l.WithError(err).Error("Error while reading packet inbound packet, closing")
-		//TODO: Trigger Control to close
-	}
-
-	f.l.Debugf("underlay reader %v is done", i)
-	f.wg.Done()
 }
 
 func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
 	runtime.LockOSThread()
+
 	packet := make([]byte, mtu)
 	out := make([]byte, mtu)
 	fwPacket := &firewall.Packet{}
@@ -310,18 +297,17 @@ func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
 	for {
 		n, err := reader.Read(packet)
 		if err != nil {
-			if !f.closed.Load() {
-				f.l.WithError(err).Error("Error while reading outbound packet, closing")
-				//TODO: Trigger Control to close
+			if errors.Is(err, os.ErrClosed) && f.closed.Load() {
+				return
 			}
-			break
+
+			f.l.WithError(err).Error("Error while reading outbound packet")
+			// This only seems to happen when something fatal happens to the fd, so exit.
+			os.Exit(2)
 		}
 
 		f.consumeInsidePacket(packet[:n], fwPacket, nb, out, i, conntrackCache.Get(f.l))
 	}
-
-	f.l.Debugf("overlay reader %v is done", i)
-	f.wg.Done()
 }
 
 func (f *Interface) RegisterConfigChangeCallbacks(c *config.C) {
@@ -498,23 +484,21 @@ func (f *Interface) GetCertState() *CertState {
 func (f *Interface) Close() error {
 	f.closed.Store(true)
 
-	// Release the udp readers
 	for _, u := range f.writers {
 		err := u.Close()
 		if err != nil {
 			f.l.WithError(err).Error("Error while closing udp socket")
 		}
 	}
-
-	// Release the tun readers
 	for i, r := range f.readers {
 		if i == 0 {
-			continue // f.readers[0] is f.inside, which we want to save for last, since it closes other stuff too
+			continue // f.readers[0] is f.inside, which we want to save for last
 		}
 		if err := r.Close(); err != nil {
 			f.l.WithError(err).Error("Error while closing tun reader")
 		}
 	}
 
+	// Release the tun device
 	return f.inside.Close()
 }

lighthouse.go

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

main.go

@@ -105,11 +105,7 @@ 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)
-		if tunQueues > udpQueues {
-			routines = tunQueues
-		} else {
-			routines = udpQueues
-		}
+		routines = max(tunQueues, udpQueues)
 		if routines != 1 {
 			l.WithField("routines", routines).Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead")
 		}
@@ -208,10 +204,11 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	useRelays := c.GetBool("relay.use_relays", DefaultUseRelays) && !c.GetBool("relay.am_relay", false)
 
 	handshakeConfig := HandshakeConfig{
-		tryInterval:   c.GetDuration("handshakes.try_interval", DefaultHandshakeTryInterval),
-		retries:       int64(c.GetInt("handshakes.retries", DefaultHandshakeRetries)),
-		triggerBuffer: c.GetInt("handshakes.trigger_buffer", DefaultHandshakeTriggerBuffer),
-		useRelays:     useRelays,
+		tryInterval:      c.GetDuration("handshakes.try_interval", DefaultHandshakeTryInterval),
+		retries:          int64(c.GetInt("handshakes.retries", DefaultHandshakeRetries)),
+		triggerBuffer:    c.GetInt("handshakes.trigger_buffer", DefaultHandshakeTriggerBuffer),
+		useRelays:        useRelays,
+		maxHandshakeRate: c.GetInt("handshakes.max_rate", DefaultMaxHandshakeRate),
 
 		messageMetrics: messageMetrics,
 	}
@@ -292,15 +289,15 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	}
 
 	return &Control{
-		f:                      ifce,
-		l:                      l,
-		ctx:                    ctx,
-		cancel:                 cancel,
-		sshStart:               sshStart,
-		statsStart:             statsStart,
-		dnsStart:               dnsStart,
-		lighthouseStart:        lightHouse.StartUpdateWorker,
-		connectionManagerStart: connManager.Start,
+		ifce,
+		l,
+		ctx,
+		cancel,
+		sshStart,
+		statsStart,
+		dnsStart,
+		lightHouse.StartUpdateWorker,
+		connManager.Start,
 	}, nil
 }
 

notboring.go

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

outside.go

@@ -190,6 +190,13 @@ 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.")
@@ -327,13 +334,29 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 		proto := layers.IPProtocol(data[protoAt])
 
 		switch proto {
-		case layers.IPProtocolICMPv6, layers.IPProtocolESP, layers.IPProtocolNoNextHeader:
+		case layers.IPProtocolESP, layers.IPProtocolNoNextHeader:
 			fp.Protocol = uint8(proto)
 			fp.RemotePort = 0
 			fp.LocalPort = 0
 			fp.Fragment = false
 			return nil
 
+		case layers.IPProtocolICMPv6:
+			if dataLen < offset+6 {
+				return ErrIPv6PacketTooShort
+			}
+			fp.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
@@ -423,34 +446,38 @@ 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 && fp.Protocol != firewall.ProtoICMP {
-		minLen += minFwPacketLen
+	if !fp.Fragment {
+		if fp.Protocol == firewall.ProtoICMP {
+			minLen += minFwPacketLen + 2
+		} else {
+			minLen += minFwPacketLen
+		}
 	}
+
 	if len(data) < minLen {
 		return ErrIPv4InvalidHeaderLength
 	}
 
-	// Firewall packets are locally oriented
-	if incoming {
+	if incoming { // Firewall packets are locally oriented
 		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
-			fp.LocalPort = 0
-		} else {
-			fp.LocalPort = binary.BigEndian.Uint16(data[ihl : ihl+2])
-			fp.RemotePort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4])
-		}
+	}
+
+	if fp.Fragment {
+		fp.RemotePort = 0
+		fp.LocalPort = 0
+	} else if fp.Protocol == firewall.ProtoICMP { //note that orientation doesn't matter on ICMP
+		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

@@ -155,6 +155,7 @@ 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{
@@ -165,20 +166,26 @@ func Test_newPacket_v6(t *testing.T) {
 		DstIP:      net.IPv6linklocalallnodes,
 	}
 
-	icmp := layers.ICMPv6{}
-
-	buffer.Clear()
-	err = gopacket.SerializeLayers(buffer, opt, &ip, &icmp)
-	if err != nil {
-		panic(err)
+	icmp := layers.ICMPv6{
+		TypeCode: layers.ICMPv6TypeEchoRequest,
+		Checksum: 0x1234,
 	}
 
-	err = newPacket(buffer.Bytes(), true, p)
-	require.NoError(t, err)
+	buffer.Clear()
+	require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp))
+	require.Error(t, newPacket(buffer.Bytes(), true, p))
+
+	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))
 	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(0), p.RemotePort)
+	assert.Equal(t, uint16(0xabcd), p.RemotePort)
 	assert.Equal(t, uint16(0), p.LocalPort)
 	assert.False(t, p.Fragment)
 
@@ -574,7 +581,7 @@ func BenchmarkParseV6(b *testing.B) {
 	}
 
 	evilBytes := buffer.Bytes()
-	for i := 0; i < 200; i++ {
+	for range 200 {
 		evilBytes = append(evilBytes, hopHeader...)
 	}
 	evilBytes = append(evilBytes, lastHopHeader...)

overlay/tun_linux.go

@@ -72,11 +72,6 @@ type ifreqQLEN struct {
 }
 
 func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
-	err := unix.SetNonblock(deviceFd, true)
-	if err != nil {
-		return nil, err
-	}
-
 	file := os.NewFile(uintptr(deviceFd), "/dev/net/tun")
 
 	t, err := newTunGeneric(c, l, file, vpnNetworks)
@@ -127,11 +122,6 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	}
 	name := strings.Trim(string(req.Name[:]), "\x00")
 
-	err = unix.SetNonblock(fd, true)
-	if err != nil {
-		return nil, err
-	}
-
 	file := os.NewFile(uintptr(fd), "/dev/net/tun")
 	t, err := newTunGeneric(c, l, file, vpnNetworks)
 	if err != nil {
@@ -155,12 +145,7 @@ func newTunGeneric(c *config.C, l *logrus.Logger, file *os.File, vpnNetworks []n
 		l:                         l,
 	}
 
-	err := unix.SetNonblock(t.fd, true)
-	if err != nil {
-		return nil, err
-	}
-
-	err = t.reload(c, true)
+	err := t.reload(c, true)
 	if err != nil {
 		return nil, err
 	}
@@ -266,11 +251,6 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 		return nil, err
 	}
 
-	err = unix.SetNonblock(fd, true)
-	if err != nil {
-		return nil, err
-	}
-
 	file := os.NewFile(uintptr(fd), "/dev/net/tun")
 
 	return file, nil
@@ -281,6 +261,29 @@ func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	return r
 }
 
+func (t *tun) Write(b []byte) (int, error) {
+	var nn int
+	maximum := len(b)
+
+	for {
+		n, err := unix.Write(t.fd, b[nn:maximum])
+		if n > 0 {
+			nn += n
+		}
+		if nn == len(b) {
+			return nn, err
+		}
+
+		if err != nil {
+			return nn, err
+		}
+
+		if n == 0 {
+			return nn, io.ErrUnexpectedEOF
+		}
+	}
+}
+
 func (t *tun) deviceBytes() (o [16]byte) {
 	for i, c := range t.Device {
 		o[i] = byte(c)
@@ -709,17 +712,11 @@ func (t *tun) Close() error {
 	}
 
 	if t.ReadWriteCloser != nil {
-		err := t.ReadWriteCloser.Close()
-		if err != nil {
-			t.l.WithField("error", err).Error("Failed to close read/write connection")
-		}
+		_ = t.ReadWriteCloser.Close()
 	}
 
 	if t.ioctlFd > 0 {
-		err := os.NewFile(t.ioctlFd, "ioctlFd").Close()
-		if err != nil {
-			t.l.WithField("error", err).Error("Failed to close ioctl fd")
-		}
+		_ = os.NewFile(t.ioctlFd, "ioctlFd").Close()
 		t.ioctlFd = 0
 	}
 

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 i := 0; i < 32; i++ {
+	for range 32 {
 		index, err := generateIndex(l)
 		if err != nil {
 			return 0, err

remote_list.go

@@ -404,12 +404,7 @@ 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 {
-	for _, v := range r.badRemotes {
-		if v == remote {
-			return true
-		}
-	}
-	return false
+	return slices.Contains(r.badRemotes, remote)
 }
 
 // unlockedSetLearnedV4 assumes you have the write lock and sets the current learned address for this owner and marks the

service/service.go

@@ -44,10 +44,7 @@ type Service struct {
 }
 
 func New(control *nebula.Control) (*Service, error) {
-	wait, err := control.Start()
-	if err != nil {
-		return nil, err
-	}
+	control.Start()
 
 	ctx := control.Context()
 	eg, ctx := errgroup.WithContext(ctx)
@@ -144,12 +141,6 @@ func New(control *nebula.Control) (*Service, error) {
 		}
 	})
 
-	// Add the nebula wait function to the group
-	eg.Go(func() error {
-		wait()
-		return nil
-	})
-
 	return &s, nil
 }
 

ssh.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"flag"
 	"fmt"
+	"maps"
 	"net"
 	"net/netip"
 	"os"
@@ -631,6 +632,9 @@ func sshCreateTunnel(ifce *Interface, fs any, a []string, w sshd.StringWriter) e
 	}
 
 	hostInfo = ifce.handshakeManager.StartHandshake(vpnAddr, nil)
+	if hostInfo == nil {
+		return w.WriteLine("Handshake rate limit reached")
+	}
 	if addr.IsValid() {
 		hostInfo.SetRemote(addr)
 	}
@@ -831,9 +835,7 @@ func sshPrintRelays(ifce *Interface, fs any, a []string, w sshd.StringWriter) er
 
 	relays := map[uint32]*HostInfo{}
 	ifce.hostMap.Lock()
-	for k, v := range ifce.hostMap.Relays {
-		relays[k] = v
-	}
+	maps.Copy(relays, ifce.hostMap.Relays)
 	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 := 0; i < 4; i++ {
+	for i := range 4 {
 		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 i := 0; i < 4; i++ {
+	for range 4 {
 		assert.NotNil(t, ci)
 		ci = ci.Next
 	}

udp/conn.go

@@ -16,7 +16,7 @@ type EncReader func(
 type Conn interface {
 	Rebind() error
 	LocalAddr() (netip.AddrPort, error)
-	ListenOut(r EncReader) error
+	ListenOut(r EncReader)
 	WriteTo(b []byte, addr netip.AddrPort) error
 	ReloadConfig(c *config.C)
 	SupportsMultipleReaders() bool
@@ -31,8 +31,8 @@ func (NoopConn) Rebind() error {
 func (NoopConn) LocalAddr() (netip.AddrPort, error) {
 	return netip.AddrPort{}, nil
 }
-func (NoopConn) ListenOut(_ EncReader) error {
-	return nil
+func (NoopConn) ListenOut(_ EncReader) {
+	return
 }
 func (NoopConn) SupportsMultipleReaders() bool {
 	return false

udp/udp_darwin.go

@@ -165,7 +165,7 @@ func NewUDPStatsEmitter(udpConns []Conn) func() {
 	return func() {}
 }
 
-func (u *StdConn) ListenOut(r EncReader) error {
+func (u *StdConn) ListenOut(r EncReader) {
 	buffer := make([]byte, MTU)
 
 	for {
@@ -173,7 +173,8 @@ func (u *StdConn) ListenOut(r EncReader) error {
 		n, rua, err := u.ReadFromUDPAddrPort(buffer)
 		if err != nil {
 			if errors.Is(err, net.ErrClosed) {
-				return err
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
 			}
 
 			u.l.WithError(err).Error("unexpected udp socket receive error")

udp/udp_generic.go

@@ -10,9 +10,11 @@ package udp
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"net"
 	"net/netip"
+	"time"
 
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
@@ -71,14 +73,25 @@ type rawMessage struct {
 	Len uint32
 }
 
-func (u *GenericConn) ListenOut(r EncReader) error {
+func (u *GenericConn) ListenOut(r EncReader) {
 	buffer := make([]byte, MTU)
 
+	var lastRecvErr time.Time
+
 	for {
 		// Just read one packet at a time
 		n, rua, err := u.ReadFromUDPAddrPort(buffer)
 		if err != nil {
-			return err
+			if errors.Is(err, net.ErrClosed) {
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
+			}
+			// Dampen unexpected message warns to once per minute
+			if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {
+				lastRecvErr = time.Now()
+				u.l.WithError(err).Warn("unexpected udp socket receive error")
+			}
+			continue
 		}
 
 		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])

udp/udp_linux.go

@@ -9,7 +9,6 @@ import (
 	"net"
 	"net/netip"
 	"syscall"
-	"time"
 	"unsafe"
 
 	"github.com/rcrowley/go-metrics"
@@ -18,8 +17,6 @@ import (
 	"golang.org/x/sys/unix"
 )
 
-var readTimeout = unix.NsecToTimeval(int64(3 * time.Second))
-
 type StdConn struct {
 	sysFd int
 	isV4  bool
@@ -27,6 +24,14 @@ type StdConn struct {
 	batch int
 }
 
+func maybeIPV4(ip net.IP) (net.IP, bool) {
+	ip4 := ip.To4()
+	if ip4 != nil {
+		return ip4, true
+	}
+	return ip, false
+}
+
 func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	af := unix.AF_INET6
 	if ip.Is4() {
@@ -50,11 +55,6 @@ func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch in
 		}
 	}
 
-	// Set a read timeout
-	if err = unix.SetsockoptTimeval(fd, unix.SOL_SOCKET, unix.SO_RCVTIMEO, &readTimeout); err != nil {
-		return nil, fmt.Errorf("unable to set SO_RCVTIMEO: %s", err)
-	}
-
 	var sa unix.Sockaddr
 	if ip.Is4() {
 		sa4 := &unix.SockaddrInet4{Port: port}
@@ -122,7 +122,7 @@ func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
 	}
 }
 
-func (u *StdConn) ListenOut(r EncReader) error {
+func (u *StdConn) ListenOut(r EncReader) {
 	var ip netip.Addr
 
 	msgs, buffers, names := u.PrepareRawMessages(u.batch)
@@ -134,7 +134,8 @@ func (u *StdConn) ListenOut(r EncReader) error {
 	for {
 		n, err := read(msgs)
 		if err != nil {
-			return err
+			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+			return
 		}
 
 		for i := 0; i < n; i++ {
@@ -162,9 +163,6 @@ func (u *StdConn) ReadSingle(msgs []rawMessage) (int, error) {
 		)
 
 		if err != 0 {
-			if err == unix.EAGAIN || err == unix.EINTR || err == unix.EWOULDBLOCK {
-				continue
-			}
 			return 0, &net.OpError{Op: "recvmsg", Err: err}
 		}
 
@@ -184,14 +182,8 @@ func (u *StdConn) ReadMulti(msgs []rawMessage) (int, error) {
 			0,
 			0,
 		)
-		if err == unix.EAGAIN || err == unix.EINTR || err == unix.EWOULDBLOCK {
-			if int64(n) > 0 {
-				//ran out of time, but have some messages to return
-				return int(n), nil
-			} else {
-				continue
-			}
-		} else if err != 0 {
+
+		if err != 0 {
 			return 0, &net.OpError{Op: "recvmmsg", Err: err}
 		}
 

udp/udp_rio_windows.go

@@ -140,7 +140,7 @@ func (u *RIOConn) bind(l *logrus.Logger, sa windows.Sockaddr) error {
 	return nil
 }
 
-func (u *RIOConn) ListenOut(r EncReader) error {
+func (u *RIOConn) ListenOut(r EncReader) {
 	buffer := make([]byte, MTU)
 
 	var lastRecvErr time.Time
@@ -151,7 +151,8 @@ func (u *RIOConn) ListenOut(r EncReader) error {
 
 		if err != nil {
 			if errors.Is(err, net.ErrClosed) {
-				return err
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
 			}
 			// Dampen unexpected message warns to once per minute
 			if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {

udp/udp_tester.go

@@ -6,7 +6,6 @@ package udp
 import (
 	"io"
 	"net/netip"
-	"os"
 	"sync/atomic"
 
 	"github.com/sirupsen/logrus"
@@ -107,11 +106,11 @@ func (u *TesterConn) WriteTo(b []byte, addr netip.AddrPort) error {
 	return nil
 }
 
-func (u *TesterConn) ListenOut(r EncReader) error {
+func (u *TesterConn) ListenOut(r EncReader) {
 	for {
 		p, ok := <-u.RxPackets
 		if !ok {
-			return os.ErrClosed
+			return
 		}
 		r(p.From, p.Data)
 	}