feedback

Bumps [github.com/miekg/dns](https://github.com/miekg/dns) from 1.1.65 to 1.1.68.
- [Changelog](https://github.com/miekg/dns/blob/master/Makefile.release)
- [Commits](https://github.com/miekg/dns/compare/v1.1.65...v1.1.68)

---
updated-dependencies:
- dependency-name: github.com/miekg/dns
  dependency-version: 1.1.68
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

.github/ISSUE_TEMPLATE/config.yml

@@ -17,5 +17,5 @@ contact_links:
     about: 'The documentation is the best place to start if you are new to Nebula.'
 
   - name: 💁 Support/Chat
-    url: https://join.slack.com/t/nebulaoss/shared_invite/zt-2xqe6e7vn-k_KGi8s13nsr7cvHVvHvuQ
+    url: https://join.slack.com/t/nebulaoss/shared_invite/zt-39pk4xopc-CUKlGcb5Z39dQ0cK1v7ehA
     about: 'For faster support, join us on Slack for assistance!'

.github/workflows/gofmt.yml

@@ -16,9 +16,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version: '1.24'
+        go-version: '1.25'
         check-latest: true
 
     - name: Install goimports

.github/workflows/release.yml

@@ -12,9 +12,9 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@v6
         with:
-          go-version: '1.24'
+          go-version: '1.25'
           check-latest: true
 
       - name: Build
@@ -35,9 +35,9 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@v6
         with:
-          go-version: '1.24'
+          go-version: '1.25'
           check-latest: true
 
       - name: Build
@@ -68,9 +68,9 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@v6
         with:
-          go-version: '1.24'
+          go-version: '1.25'
           check-latest: true
 
       - name: Import certificates

.github/workflows/smoke-extra.yml

@@ -22,9 +22,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version-file: 'go.mod'
+        go-version: '1.25'
         check-latest: true
 
     - name: add hashicorp source

.github/workflows/smoke.yml

@@ -20,9 +20,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version: '1.24'
+        go-version: '1.25'
         check-latest: true
 
     - name: build

.github/workflows/test.yml

@@ -20,9 +20,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version: '1.24'
+        go-version: '1.25'
         check-latest: true
 
     - name: Build
@@ -32,9 +32,9 @@ jobs:
       run: make vet
 
     - name: golangci-lint
-      uses: golangci/golangci-lint-action@v7
+      uses: golangci/golangci-lint-action@v8
       with:
-        version: v2.0
+        version: v2.5
 
     - name: Test
       run: make test
@@ -58,9 +58,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version: '1.24'
+        go-version: '1.25'
         check-latest: true
 
     - name: Build
@@ -79,9 +79,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version: '1.22'
+        go-version: '1.25'
         check-latest: true
 
     - name: Build
@@ -100,9 +100,9 @@ jobs:
 
     - uses: actions/checkout@v4
 
-    - uses: actions/setup-go@v5
+    - uses: actions/setup-go@v6
       with:
-        go-version: '1.24'
+        go-version: '1.25'
         check-latest: true
 
     - name: Build nebula
@@ -115,9 +115,9 @@ jobs:
       run: make vet
 
     - name: golangci-lint
-      uses: golangci/golangci-lint-action@v7
+      uses: golangci/golangci-lint-action@v8
       with:
-        version: v2.0
+        version: v2.5
 
     - name: Test
       run: make test

README.md

@@ -12,7 +12,7 @@ Further documentation can be found [here](https://nebula.defined.net/docs/).
 
 You can read more about Nebula [here](https://medium.com/p/884110a5579).
 
-You can also join the NebulaOSS Slack group [here](https://join.slack.com/t/nebulaoss/shared_invite/zt-2xqe6e7vn-k_KGi8s13nsr7cvHVvHvuQ).
+You can also join the NebulaOSS Slack group [here](https://join.slack.com/t/nebulaoss/shared_invite/zt-39pk4xopc-CUKlGcb5Z39dQ0cK1v7ehA).
 
 ## Supported Platforms
 

calculated_remote.go

@@ -84,16 +84,11 @@ func NewCalculatedRemotesFromConfig(c *config.C, k string) (*bart.Table[[]*calcu
 
 	calculatedRemotes := new(bart.Table[[]*calculatedRemote])
 
-	rawMap, ok := value.(map[any]any)
+	rawMap, ok := value.(map[string]any)
 	if !ok {
 		return nil, fmt.Errorf("config `%s` has invalid type: %T", k, value)
 	}
-	for rawKey, rawValue := range rawMap {
+	for rawCIDR, rawValue := range rawMap {
-		rawCIDR, ok := rawKey.(string)
-		if !ok {
-			return nil, fmt.Errorf("config `%s` has invalid key (type %T): %v", k, rawKey, rawKey)
-		}
-
 		cidr, err := netip.ParsePrefix(rawCIDR)
 		if err != nil {
 			return nil, fmt.Errorf("config `%s` has invalid CIDR: %s", k, rawCIDR)
@@ -129,7 +124,7 @@ func newCalculatedRemotesListFromConfig(cidr netip.Prefix, raw any) ([]*calculat
 }
 
 func newCalculatedRemotesEntryFromConfig(cidr netip.Prefix, raw any) (*calculatedRemote, error) {
-	rawMap, ok := raw.(map[any]any)
+	rawMap, ok := raw.(map[string]any)
 	if !ok {
 		return nil, fmt.Errorf("invalid type: %T", raw)
 	}

cert/cert.go

@@ -58,6 +58,9 @@ type Certificate interface {
 	// PublicKey is the raw bytes to be used in asymmetric cryptographic operations.
 	PublicKey() []byte
 
+	// MarshalPublicKeyPEM is the value of PublicKey marshalled to PEM
+	MarshalPublicKeyPEM() []byte
+
 	// Curve identifies which curve was used for the PublicKey and Signature.
 	Curve() Curve
 
@@ -135,8 +138,7 @@ func Recombine(v Version, rawCertBytes, publicKey []byte, curve Curve) (Certific
 	case Version2:
 		c, err = unmarshalCertificateV2(rawCertBytes, publicKey, curve)
 	default:
-		//TODO: CERT-V2 make a static var
+		return nil, ErrUnknownVersion
-		return nil, fmt.Errorf("unknown certificate version %d", v)
 	}
 
 	if err != nil {

cert/cert_v1.go

@@ -83,6 +83,10 @@ func (c *certificateV1) PublicKey() []byte {
 	return c.details.publicKey
 }
 
+func (c *certificateV1) MarshalPublicKeyPEM() []byte {
+	return marshalCertPublicKeyToPEM(c)
+}
+
 func (c *certificateV1) Signature() []byte {
 	return c.signature
 }
@@ -110,8 +114,10 @@ func (c *certificateV1) CheckSignature(key []byte) bool {
 	case Curve_CURVE25519:
 		return ed25519.Verify(key, b, c.signature)
 	case Curve_P256:
-		x, y := elliptic.Unmarshal(elliptic.P256(), key)
+		pubKey, err := ecdsa.ParseUncompressedPublicKey(elliptic.P256(), key)
-		pubKey := &ecdsa.PublicKey{Curve: elliptic.P256(), X: x, Y: y}
+		if err != nil {
+			return false
+		}
 		hashed := sha256.Sum256(b)
 		return ecdsa.VerifyASN1(pubKey, hashed[:], c.signature)
 	default:

cert/cert_v1_test.go

@@ -1,6 +1,7 @@
 package cert
 
 import (
+	"crypto/ed25519"
 	"fmt"
 	"net/netip"
 	"testing"
@@ -13,6 +14,7 @@ import (
 )
 
 func TestCertificateV1_Marshal(t *testing.T) {
+	t.Parallel()
 	before := time.Now().Add(time.Second * -60).Round(time.Second)
 	after := time.Now().Add(time.Second * 60).Round(time.Second)
 	pubKey := []byte("1234567890abcedfghij1234567890ab")
@@ -60,6 +62,58 @@ func TestCertificateV1_Marshal(t *testing.T) {
 	assert.Equal(t, nc.Groups(), nc2.Groups())
 }
 
+func TestCertificateV1_PublicKeyPem(t *testing.T) {
+	t.Parallel()
+	before := time.Now().Add(time.Second * -60).Round(time.Second)
+	after := time.Now().Add(time.Second * 60).Round(time.Second)
+	pubKey := ed25519.PublicKey("1234567890abcedfghij1234567890ab")
+
+	nc := certificateV1{
+		details: detailsV1{
+			name:           "testing",
+			networks:       []netip.Prefix{},
+			unsafeNetworks: []netip.Prefix{},
+			groups:         []string{"test-group1", "test-group2", "test-group3"},
+			notBefore:      before,
+			notAfter:       after,
+			publicKey:      pubKey,
+			isCA:           false,
+			issuer:         "1234567890abcedfghij1234567890ab",
+		},
+		signature: []byte("1234567890abcedfghij1234567890ab"),
+	}
+
+	assert.Equal(t, Version1, nc.Version())
+	assert.Equal(t, Curve_CURVE25519, nc.Curve())
+	pubPem := "-----BEGIN NEBULA X25519 PUBLIC KEY-----\nMTIzNDU2Nzg5MGFiY2VkZmdoaWoxMjM0NTY3ODkwYWI=\n-----END NEBULA X25519 PUBLIC KEY-----\n"
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), pubPem)
+	assert.False(t, nc.IsCA())
+
+	nc.details.isCA = true
+	assert.Equal(t, Curve_CURVE25519, nc.Curve())
+	pubPem = "-----BEGIN NEBULA ED25519 PUBLIC KEY-----\nMTIzNDU2Nzg5MGFiY2VkZmdoaWoxMjM0NTY3ODkwYWI=\n-----END NEBULA ED25519 PUBLIC KEY-----\n"
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), pubPem)
+	assert.True(t, nc.IsCA())
+
+	pubP256KeyPem := []byte(`-----BEGIN NEBULA P256 PUBLIC KEY-----
+AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+AAAAAAAAAAAAAAAAAAAAAAA=
+-----END NEBULA P256 PUBLIC KEY-----
+`)
+	pubP256Key, _, _, err := UnmarshalPublicKeyFromPEM(pubP256KeyPem)
+	require.NoError(t, err)
+	nc.details.curve = Curve_P256
+	nc.details.publicKey = pubP256Key
+	assert.Equal(t, Curve_P256, nc.Curve())
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), string(pubP256KeyPem))
+	assert.True(t, nc.IsCA())
+
+	nc.details.isCA = false
+	assert.Equal(t, Curve_P256, nc.Curve())
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), string(pubP256KeyPem))
+	assert.False(t, nc.IsCA())
+}
+
 func TestCertificateV1_Expired(t *testing.T) {
 	nc := certificateV1{
 		details: detailsV1{

cert/cert_v2.go

@@ -114,6 +114,10 @@ func (c *certificateV2) PublicKey() []byte {
 	return c.publicKey
 }
 
+func (c *certificateV2) MarshalPublicKeyPEM() []byte {
+	return marshalCertPublicKeyToPEM(c)
+}
+
 func (c *certificateV2) Signature() []byte {
 	return c.signature
 }
@@ -149,8 +153,10 @@ func (c *certificateV2) CheckSignature(key []byte) bool {
 	case Curve_CURVE25519:
 		return ed25519.Verify(key, b, c.signature)
 	case Curve_P256:
-		x, y := elliptic.Unmarshal(elliptic.P256(), key)
+		pubKey, err := ecdsa.ParseUncompressedPublicKey(elliptic.P256(), key)
-		pubKey := &ecdsa.PublicKey{Curve: elliptic.P256(), X: x, Y: y}
+		if err != nil {
+			return false
+		}
 		hashed := sha256.Sum256(b)
 		return ecdsa.VerifyASN1(pubKey, hashed[:], c.signature)
 	default:

cert/cert_v2_test.go

@@ -15,6 +15,7 @@ import (
 )
 
 func TestCertificateV2_Marshal(t *testing.T) {
+	t.Parallel()
 	before := time.Now().Add(time.Second * -60).Round(time.Second)
 	after := time.Now().Add(time.Second * 60).Round(time.Second)
 	pubKey := []byte("1234567890abcedfghij1234567890ab")
@@ -75,6 +76,58 @@ func TestCertificateV2_Marshal(t *testing.T) {
 	assert.Equal(t, nc.Groups(), nc2.Groups())
 }
 
+func TestCertificateV2_PublicKeyPem(t *testing.T) {
+	t.Parallel()
+	before := time.Now().Add(time.Second * -60).Round(time.Second)
+	after := time.Now().Add(time.Second * 60).Round(time.Second)
+	pubKey := ed25519.PublicKey("1234567890abcedfghij1234567890ab")
+
+	nc := certificateV2{
+		details: detailsV2{
+			name:           "testing",
+			networks:       []netip.Prefix{},
+			unsafeNetworks: []netip.Prefix{},
+			groups:         []string{"test-group1", "test-group2", "test-group3"},
+			notBefore:      before,
+			notAfter:       after,
+			isCA:           false,
+			issuer:         "1234567890abcedfghij1234567890ab",
+		},
+		publicKey: pubKey,
+		signature: []byte("1234567890abcedfghij1234567890ab"),
+	}
+
+	assert.Equal(t, Version2, nc.Version())
+	assert.Equal(t, Curve_CURVE25519, nc.Curve())
+	pubPem := "-----BEGIN NEBULA X25519 PUBLIC KEY-----\nMTIzNDU2Nzg5MGFiY2VkZmdoaWoxMjM0NTY3ODkwYWI=\n-----END NEBULA X25519 PUBLIC KEY-----\n"
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), pubPem)
+	assert.False(t, nc.IsCA())
+
+	nc.details.isCA = true
+	assert.Equal(t, Curve_CURVE25519, nc.Curve())
+	pubPem = "-----BEGIN NEBULA ED25519 PUBLIC KEY-----\nMTIzNDU2Nzg5MGFiY2VkZmdoaWoxMjM0NTY3ODkwYWI=\n-----END NEBULA ED25519 PUBLIC KEY-----\n"
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), pubPem)
+	assert.True(t, nc.IsCA())
+
+	pubP256KeyPem := []byte(`-----BEGIN NEBULA P256 PUBLIC KEY-----
+AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+AAAAAAAAAAAAAAAAAAAAAAA=
+-----END NEBULA P256 PUBLIC KEY-----
+`)
+	pubP256Key, _, _, err := UnmarshalPublicKeyFromPEM(pubP256KeyPem)
+	require.NoError(t, err)
+	nc.curve = Curve_P256
+	nc.publicKey = pubP256Key
+	assert.Equal(t, Curve_P256, nc.Curve())
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), string(pubP256KeyPem))
+	assert.True(t, nc.IsCA())
+
+	nc.details.isCA = false
+	assert.Equal(t, Curve_P256, nc.Curve())
+	assert.Equal(t, string(nc.MarshalPublicKeyPEM()), string(pubP256KeyPem))
+	assert.False(t, nc.IsCA())
+}
+
 func TestCertificateV2_Expired(t *testing.T) {
 	nc := certificateV2{
 		details: detailsV2{

cert/errors.go

@@ -20,6 +20,7 @@ var (
 	ErrPublicPrivateKeyMismatch   = errors.New("public key and private key are not a pair")
 	ErrPrivateKeyEncrypted        = errors.New("private key must be decrypted")
 	ErrCaNotFound                 = errors.New("could not find ca for the certificate")
+	ErrUnknownVersion             = errors.New("certificate version unrecognized")
 
 	ErrInvalidPEMBlock                   = errors.New("input did not contain a valid PEM encoded block")
 	ErrInvalidPEMCertificateBanner       = errors.New("bytes did not contain a proper certificate banner")

cert/pem.go

@@ -7,19 +7,26 @@ import (
 	"golang.org/x/crypto/ed25519"
 )
 
-const (
+const ( //cert banners
-	CertificateBanner                = "NEBULA CERTIFICATE"
+	CertificateBanner   = "NEBULA CERTIFICATE"
-	CertificateV2Banner              = "NEBULA CERTIFICATE V2"
+	CertificateV2Banner = "NEBULA CERTIFICATE V2"
-	X25519PrivateKeyBanner           = "NEBULA X25519 PRIVATE KEY"
+)
-	X25519PublicKeyBanner            = "NEBULA X25519 PUBLIC KEY"
-	EncryptedEd25519PrivateKeyBanner = "NEBULA ED25519 ENCRYPTED PRIVATE KEY"
-	Ed25519PrivateKeyBanner          = "NEBULA ED25519 PRIVATE KEY"
-	Ed25519PublicKeyBanner           = "NEBULA ED25519 PUBLIC KEY"
 
-	P256PrivateKeyBanner               = "NEBULA P256 PRIVATE KEY"
+const ( //key-agreement-key banners
-	P256PublicKeyBanner                = "NEBULA P256 PUBLIC KEY"
+	X25519PrivateKeyBanner = "NEBULA X25519 PRIVATE KEY"
+	X25519PublicKeyBanner  = "NEBULA X25519 PUBLIC KEY"
+	P256PrivateKeyBanner   = "NEBULA P256 PRIVATE KEY"
+	P256PublicKeyBanner    = "NEBULA P256 PUBLIC KEY"
+)
+
+/* including "ECDSA" in the P256 banners is a clue that these keys should be used only for signing */
+const ( //signing key banners
 	EncryptedECDSAP256PrivateKeyBanner = "NEBULA ECDSA P256 ENCRYPTED PRIVATE KEY"
 	ECDSAP256PrivateKeyBanner          = "NEBULA ECDSA P256 PRIVATE KEY"
+	ECDSAP256PublicKeyBanner           = "NEBULA ECDSA P256 PUBLIC KEY"
+	EncryptedEd25519PrivateKeyBanner   = "NEBULA ED25519 ENCRYPTED PRIVATE KEY"
+	Ed25519PrivateKeyBanner            = "NEBULA ED25519 PRIVATE KEY"
+	Ed25519PublicKeyBanner             = "NEBULA ED25519 PUBLIC KEY"
 )
 
 // UnmarshalCertificateFromPEM will try to unmarshal the first pem block in a byte array, returning any non consumed
@@ -51,6 +58,16 @@ func UnmarshalCertificateFromPEM(b []byte) (Certificate, []byte, error) {
 
 }
 
+func marshalCertPublicKeyToPEM(c Certificate) []byte {
+	if c.IsCA() {
+		return MarshalSigningPublicKeyToPEM(c.Curve(), c.PublicKey())
+	} else {
+		return MarshalPublicKeyToPEM(c.Curve(), c.PublicKey())
+	}
+}
+
+// MarshalPublicKeyToPEM returns a PEM representation of a public key used for ECDH.
+// if your public key came from a certificate, prefer Certificate.PublicKeyPEM() if possible, to avoid mistakes!
 func MarshalPublicKeyToPEM(curve Curve, b []byte) []byte {
 	switch curve {
 	case Curve_CURVE25519:
@@ -62,6 +79,19 @@ func MarshalPublicKeyToPEM(curve Curve, b []byte) []byte {
 	}
 }
 
+// MarshalSigningPublicKeyToPEM returns a PEM representation of a public key used for signing.
+// if your public key came from a certificate, prefer Certificate.PublicKeyPEM() if possible, to avoid mistakes!
+func MarshalSigningPublicKeyToPEM(curve Curve, b []byte) []byte {
+	switch curve {
+	case Curve_CURVE25519:
+		return pem.EncodeToMemory(&pem.Block{Type: Ed25519PublicKeyBanner, Bytes: b})
+	case Curve_P256:
+		return pem.EncodeToMemory(&pem.Block{Type: P256PublicKeyBanner, Bytes: b})
+	default:
+		return nil
+	}
+}
+
 func UnmarshalPublicKeyFromPEM(b []byte) ([]byte, []byte, Curve, error) {
 	k, r := pem.Decode(b)
 	if k == nil {
@@ -73,7 +103,7 @@ func UnmarshalPublicKeyFromPEM(b []byte) ([]byte, []byte, Curve, error) {
 	case X25519PublicKeyBanner, Ed25519PublicKeyBanner:
 		expectedLen = 32
 		curve = Curve_CURVE25519
-	case P256PublicKeyBanner:
+	case P256PublicKeyBanner, ECDSAP256PublicKeyBanner:
 		// Uncompressed
 		expectedLen = 65
 		curve = Curve_P256

cert/pem_test.go

@@ -177,6 +177,7 @@ AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
 }
 
 func TestUnmarshalPublicKeyFromPEM(t *testing.T) {
+	t.Parallel()
 	pubKey := []byte(`# A good key
 -----BEGIN NEBULA ED25519 PUBLIC KEY-----
 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
@@ -230,6 +231,7 @@ AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
 }
 
 func TestUnmarshalX25519PublicKey(t *testing.T) {
+	t.Parallel()
 	pubKey := []byte(`# A good key
 -----BEGIN NEBULA X25519 PUBLIC KEY-----
 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
@@ -240,6 +242,12 @@ AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
 AAAAAAAAAAAAAAAAAAAAAAA=
 -----END NEBULA P256 PUBLIC KEY-----
+`)
+	oldPubP256Key := []byte(`# A good key
+-----BEGIN NEBULA ECDSA P256 PUBLIC KEY-----
+AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+AAAAAAAAAAAAAAAAAAAAAAA=
+-----END NEBULA ECDSA P256 PUBLIC KEY-----
 `)
 	shortKey := []byte(`# A short key
 -----BEGIN NEBULA X25519 PUBLIC KEY-----
@@ -256,15 +264,22 @@ AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
 -END NEBULA X25519 PUBLIC KEY-----`)
 
-	keyBundle := appendByteSlices(pubKey, pubP256Key, shortKey, invalidBanner, invalidPem)
+	keyBundle := appendByteSlices(pubKey, pubP256Key, oldPubP256Key, shortKey, invalidBanner, invalidPem)
 
 	// Success test case
 	k, rest, curve, err := UnmarshalPublicKeyFromPEM(keyBundle)
 	assert.Len(t, k, 32)
 	require.NoError(t, err)
-	assert.Equal(t, rest, appendByteSlices(pubP256Key, shortKey, invalidBanner, invalidPem))
+	assert.Equal(t, rest, appendByteSlices(pubP256Key, oldPubP256Key, shortKey, invalidBanner, invalidPem))
 	assert.Equal(t, Curve_CURVE25519, curve)
 
+	// Success test case
+	k, rest, curve, err = UnmarshalPublicKeyFromPEM(rest)
+	assert.Len(t, k, 65)
+	require.NoError(t, err)
+	assert.Equal(t, rest, appendByteSlices(oldPubP256Key, shortKey, invalidBanner, invalidPem))
+	assert.Equal(t, Curve_P256, curve)
+
 	// Success test case
 	k, rest, curve, err = UnmarshalPublicKeyFromPEM(rest)
 	assert.Len(t, k, 65)

cert/sign.go

@@ -7,7 +7,6 @@ import (
 	"crypto/rand"
 	"crypto/sha256"
 	"fmt"
-	"math/big"
 	"net/netip"
 	"time"
 )
@@ -55,15 +54,10 @@ func (t *TBSCertificate) Sign(signer Certificate, curve Curve, key []byte) (Cert
 		}
 		return t.SignWith(signer, curve, sp)
 	case Curve_P256:
-		pk := &ecdsa.PrivateKey{
+		pk, err := ecdsa.ParseRawPrivateKey(elliptic.P256(), key)
-			PublicKey: ecdsa.PublicKey{
+		if err != nil {
-				Curve: elliptic.P256(),
+			return nil, err
-			},
-			// ref: https://github.com/golang/go/blob/go1.19/src/crypto/x509/sec1.go#L95
-			D: new(big.Int).SetBytes(key),
 		}
-		// ref: https://github.com/golang/go/blob/go1.19/src/crypto/x509/sec1.go#L119
-		pk.X, pk.Y = pk.Curve.ScalarBaseMult(key)
 		sp := func(certBytes []byte) ([]byte, error) {
 			// We need to hash first for ECDSA
 			// - https://pkg.go.dev/crypto/ecdsa#SignASN1

connection_manager.go

@@ -4,8 +4,10 @@ import (
 	"bytes"
 	"context"
 	"encoding/binary"
+	"fmt"
 	"net/netip"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/rcrowley/go-metrics"
@@ -27,311 +29,127 @@ const (
 	sendTestPacket trafficDecision = 6
 )
 
-// LastCommunication tracks when we last communicated with a host
-type LastCommunication struct {
-	timestamp time.Time
-	vpnIp     netip.Addr // To help with logging
-}
-
 type connectionManager struct {
-	in     map[uint32]struct{}
-	inLock *sync.RWMutex
-
-	out     map[uint32]struct{}
-	outLock *sync.RWMutex
-
 	// relayUsed holds which relay localIndexs are in use
 	relayUsed     map[uint32]struct{}
 	relayUsedLock *sync.RWMutex
 
-	// Track last communication with hosts
+	hostMap      *HostMap
-	lastCommMap       map[uint32]time.Time
+	trafficTimer *LockingTimerWheel[uint32]
-	lastCommLock      *sync.RWMutex
+	intf         *Interface
-	inactivityTimer   *LockingTimerWheel[uint32]
+	punchy       *Punchy
-	inactivityTimeout time.Duration
 
-	hostMap                 *HostMap
+	// Configuration settings
-	trafficTimer            *LockingTimerWheel[uint32]
-	intf                    *Interface
-	pendingDeletion         map[uint32]struct{}
-	punchy                  *Punchy
 	checkInterval           time.Duration
 	pendingDeletionInterval time.Duration
-	metricsTxPunchy         metrics.Counter
+	inactivityTimeout       atomic.Int64
+	dropInactive            atomic.Bool
+
+	metricsTxPunchy metrics.Counter
 
 	l *logrus.Logger
 }
 
-func newConnectionManager(ctx context.Context, l *logrus.Logger, intf *Interface, checkInterval, pendingDeletionInterval time.Duration, punchy *Punchy) *connectionManager {
+func newConnectionManagerFromConfig(l *logrus.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
-	var max time.Duration
+	cm := &connectionManager{
-	if checkInterval < pendingDeletionInterval {
+		hostMap:         hm,
-		max = pendingDeletionInterval
+		l:               l,
-	} else {
+		punchy:          p,
-		max = checkInterval
+		relayUsed:       make(map[uint32]struct{}),
+		relayUsedLock:   &sync.RWMutex{},
+		metricsTxPunchy: metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
 	}
 
-	nc := &connectionManager{
+	cm.reload(c, true)
-		hostMap:                 intf.hostMap,
+	c.RegisterReloadCallback(func(c *config.C) {
-		in:                      make(map[uint32]struct{}),
+		cm.reload(c, false)
-		inLock:                  &sync.RWMutex{},
+	})
-		out:                     make(map[uint32]struct{}),
-		outLock:                 &sync.RWMutex{},
-		relayUsed:               make(map[uint32]struct{}),
-		relayUsedLock:           &sync.RWMutex{},
-		lastCommMap:             make(map[uint32]time.Time),
-		lastCommLock:            &sync.RWMutex{},
-		inactivityTimeout:       1 * time.Minute, // Default inactivity timeout: 10 minutes
-		trafficTimer:            NewLockingTimerWheel[uint32](time.Millisecond*500, max),
-		intf:                    intf,
-		pendingDeletion:         make(map[uint32]struct{}),
-		checkInterval:           checkInterval,
-		pendingDeletionInterval: pendingDeletionInterval,
-		punchy:                  punchy,
-		metricsTxPunchy:         metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
-		l:                       l,
-	}
 
-	// Initialize the inactivity timer wheel - make wheel duration slightly longer than the timeout
+	return cm
-	nc.inactivityTimer = NewLockingTimerWheel[uint32](time.Minute, nc.inactivityTimeout+time.Minute)
-
-	nc.Start(ctx)
-	return nc
 }
 
-func (n *connectionManager) updateLastCommunication(localIndex uint32) {
+func (cm *connectionManager) reload(c *config.C, initial bool) {
-	// Get host info to record VPN IP for better logging
+	if initial {
-	hostInfo := n.hostMap.QueryIndex(localIndex)
+		cm.checkInterval = time.Duration(c.GetInt("timers.connection_alive_interval", 5)) * time.Second
-	if hostInfo == nil {
+		cm.pendingDeletionInterval = time.Duration(c.GetInt("timers.pending_deletion_interval", 10)) * time.Second
-		return
+
+		// We want at least a minimum resolution of 500ms per tick so that we can hit these intervals
+		// pretty close to their configured duration.
+		// The inactivity duration is checked each time a hostinfo ticks through so we don't need the wheel to contain it.
+		minDuration := min(time.Millisecond*500, cm.checkInterval, cm.pendingDeletionInterval)
+		maxDuration := max(cm.checkInterval, cm.pendingDeletionInterval)
+		cm.trafficTimer = NewLockingTimerWheel[uint32](minDuration, maxDuration)
 	}
 
-	now := time.Now()
+	if initial || c.HasChanged("tunnels.inactivity_timeout") {
-	n.lastCommLock.Lock()
+		old := cm.getInactivityTimeout()
-	n.lastCommMap[localIndex] = now
+		cm.inactivityTimeout.Store((int64)(c.GetDuration("tunnels.inactivity_timeout", 10*time.Minute)))
-	n.lastCommLock.Unlock()
+		if !initial {
+			cm.l.WithField("oldDuration", old).
+				WithField("newDuration", cm.getInactivityTimeout()).
+				Info("Inactivity timeout has changed")
+		}
+	}
 
-	// Reset the inactivity timer for this host
+	if initial || c.HasChanged("tunnels.drop_inactive") {
-	n.inactivityTimer.m.Lock()
+		old := cm.dropInactive.Load()
-	n.inactivityTimer.t.Add(localIndex, n.inactivityTimeout)
+		cm.dropInactive.Store(c.GetBool("tunnels.drop_inactive", false))
-	n.inactivityTimer.m.Unlock()
+		if !initial {
+			cm.l.WithField("oldBool", old).
+				WithField("newBool", cm.dropInactive.Load()).
+				Info("Drop inactive setting has changed")
+		}
+	}
 }
 
-func (n *connectionManager) In(localIndex uint32) {
+func (cm *connectionManager) getInactivityTimeout() time.Duration {
-	n.inLock.RLock()
+	return (time.Duration)(cm.inactivityTimeout.Load())
+}
+
+func (cm *connectionManager) In(h *HostInfo) {
+	h.in.Store(true)
+}
+
+func (cm *connectionManager) Out(h *HostInfo) {
+	h.out.Store(true)
+}
+
+func (cm *connectionManager) RelayUsed(localIndex uint32) {
+	cm.relayUsedLock.RLock()
 	// If this already exists, return
-	if _, ok := n.in[localIndex]; ok {
+	if _, ok := cm.relayUsed[localIndex]; ok {
-		n.inLock.RUnlock()
+		cm.relayUsedLock.RUnlock()
 		return
 	}
-	n.inLock.RUnlock()
+	cm.relayUsedLock.RUnlock()
-	n.inLock.Lock()
+	cm.relayUsedLock.Lock()
-	n.in[localIndex] = struct{}{}
+	cm.relayUsed[localIndex] = struct{}{}
-	n.inLock.Unlock()
+	cm.relayUsedLock.Unlock()
-
-	// Update last communication time
-	n.updateLastCommunication(localIndex)
-}
-
-func (n *connectionManager) Out(localIndex uint32) {
-	n.outLock.RLock()
-	// If this already exists, return
-	if _, ok := n.out[localIndex]; ok {
-		n.outLock.RUnlock()
-		return
-	}
-	n.outLock.RUnlock()
-	n.outLock.Lock()
-	n.out[localIndex] = struct{}{}
-	n.outLock.Unlock()
-
-	// Update last communication time
-	n.updateLastCommunication(localIndex)
-}
-
-func (n *connectionManager) RelayUsed(localIndex uint32) {
-	n.relayUsedLock.RLock()
-	// If this already exists, return
-	if _, ok := n.relayUsed[localIndex]; ok {
-		n.relayUsedLock.RUnlock()
-		return
-	}
-	n.relayUsedLock.RUnlock()
-	n.relayUsedLock.Lock()
-	n.relayUsed[localIndex] = struct{}{}
-	n.relayUsedLock.Unlock()
 }
 
 // getAndResetTrafficCheck returns if there was any inbound or outbound traffic within the last tick and
 // resets the state for this local index
-func (n *connectionManager) getAndResetTrafficCheck(localIndex uint32) (bool, bool) {
+func (cm *connectionManager) getAndResetTrafficCheck(h *HostInfo, now time.Time) (bool, bool) {
-	n.inLock.Lock()
+	in := h.in.Swap(false)
-	n.outLock.Lock()
+	out := h.out.Swap(false)
-	_, in := n.in[localIndex]
+	if in || out {
-	_, out := n.out[localIndex]
+		h.lastUsed = now
-	delete(n.in, localIndex)
+	}
-	delete(n.out, localIndex)
-	n.inLock.Unlock()
-	n.outLock.Unlock()
 	return in, out
 }
 
-func (n *connectionManager) AddTrafficWatch(localIndex uint32) {
+// AddTrafficWatch must be called for every new HostInfo.
-	// Use a write lock directly because it should be incredibly rare that we are ever already tracking this index
+// We will continue to monitor the HostInfo until the tunnel is dropped.
-	n.outLock.Lock()
+func (cm *connectionManager) AddTrafficWatch(h *HostInfo) {
-	if _, ok := n.out[localIndex]; ok {
+	if h.out.Swap(true) == false {
-		n.outLock.Unlock()
+		cm.trafficTimer.Add(h.localIndexId, cm.checkInterval)
-		return
-	}
-	n.out[localIndex] = struct{}{}
-	n.trafficTimer.Add(localIndex, n.checkInterval)
-	n.outLock.Unlock()
-}
-
-// checkInactiveTunnels checks for tunnels that have been inactive for too long and drops them
-func (n *connectionManager) checkInactiveTunnels() {
-	now := time.Now()
-
-	// First, advance the timer wheel to the current time
-	n.inactivityTimer.m.Lock()
-	n.inactivityTimer.t.Advance(now)
-	n.inactivityTimer.m.Unlock()
-
-	// Check for expired timers (inactive connections)
-	for {
-		// Get the next expired tunnel
-		n.inactivityTimer.m.Lock()
-		localIndex, ok := n.inactivityTimer.t.Purge()
-		n.inactivityTimer.m.Unlock()
-
-		if !ok {
-			// No more expired timers
-			break
-		}
-
-		n.lastCommLock.RLock()
-		lastComm, exists := n.lastCommMap[localIndex]
-		n.lastCommLock.RUnlock()
-
-		if !exists {
-			// No last communication record, odd but skip
-			continue
-		}
-
-		// Calculate inactivity duration
-		inactiveDuration := now.Sub(lastComm)
-
-		// Check if we've exceeded the inactivity timeout
-		if inactiveDuration >= n.inactivityTimeout {
-			// Get the host info (if it still exists)
-			hostInfo := n.hostMap.QueryIndex(localIndex)
-			if hostInfo == nil {
-				// Host info is gone, remove from our tracking map
-				n.lastCommLock.Lock()
-				delete(n.lastCommMap, localIndex)
-				n.lastCommLock.Unlock()
-				continue
-			}
-
-			// Log the inactivity and drop the tunnel
-			n.l.WithField("vpnIp", hostInfo.vpnAddrs[0]).
-				WithField("localIndex", localIndex).
-				WithField("inactiveDuration", inactiveDuration).
-				WithField("timeout", n.inactivityTimeout).
-				Info("Dropping tunnel due to inactivity")
-
-			// Close the tunnel using the existing mechanism
-			n.intf.closeTunnel(hostInfo)
-
-			// Clean up our tracking map
-			n.lastCommLock.Lock()
-			delete(n.lastCommMap, localIndex)
-			n.lastCommLock.Unlock()
-		} else {
-			// Re-add to the timer wheel with the remaining time
-			remainingTime := n.inactivityTimeout - inactiveDuration
-			n.inactivityTimer.m.Lock()
-			n.inactivityTimer.t.Add(localIndex, remainingTime)
-			n.inactivityTimer.m.Unlock()
-		}
 	}
 }
 
-// CleanupDeletedHostInfos removes entries from our lastCommMap for hosts that no longer exist
+func (cm *connectionManager) Start(ctx context.Context) {
-func (n *connectionManager) CleanupDeletedHostInfos() {
+	clockSource := time.NewTicker(cm.trafficTimer.t.tickDuration)
-	n.lastCommLock.Lock()
-	defer n.lastCommLock.Unlock()
-
-	// Find indexes to delete
-	var toDelete []uint32
-	for localIndex := range n.lastCommMap {
-		if n.hostMap.QueryIndex(localIndex) == nil {
-			toDelete = append(toDelete, localIndex)
-		}
-	}
-
-	// Delete them
-	for _, localIndex := range toDelete {
-		delete(n.lastCommMap, localIndex)
-	}
-
-	if len(toDelete) > 0 && n.l.Level >= logrus.DebugLevel {
-		n.l.WithField("count", len(toDelete)).Debug("Cleaned up deleted host entries from lastCommMap")
-	}
-}
-
-// ReloadConfig updates the connection manager configuration
-func (n *connectionManager) ReloadConfig(c *config.C) {
-	// Get the inactivity timeout from config
-	inactivityTimeout := c.GetDuration("timers.inactivity_timeout", 10*time.Minute)
-
-	// Only update if different
-	if inactivityTimeout != n.inactivityTimeout {
-		n.l.WithField("old", n.inactivityTimeout).
-			WithField("new", inactivityTimeout).
-			Info("Updating inactivity timeout")
-
-		n.inactivityTimeout = inactivityTimeout
-
-		// Recreate the inactivity timer wheel with the new timeout
-		n.inactivityTimer = NewLockingTimerWheel[uint32](time.Minute, n.inactivityTimeout+time.Minute)
-
-		// Re-add all existing hosts to the new timer wheel
-		n.lastCommLock.RLock()
-		for localIndex, lastComm := range n.lastCommMap {
-			// Calculate remaining time based on last communication
-			now := time.Now()
-			elapsed := now.Sub(lastComm)
-
-			// If the elapsed time exceeds the new timeout, this will be caught
-			// in the next inactivity check. Otherwise, add with remaining time.
-			if elapsed < n.inactivityTimeout {
-				remainingTime := n.inactivityTimeout - elapsed
-				n.inactivityTimer.m.Lock()
-				n.inactivityTimer.t.Add(localIndex, remainingTime)
-				n.inactivityTimer.m.Unlock()
-			}
-		}
-		n.lastCommLock.RUnlock()
-	}
-}
-
-func (n *connectionManager) Start(ctx context.Context) {
-	go n.Run(ctx)
-}
-
-func (n *connectionManager) Run(ctx context.Context) {
-	//TODO: this tick should be based on the min wheel tick? Check firewall
-	clockSource := time.NewTicker(500 * time.Millisecond)
 	defer clockSource.Stop()
 
-	// Create ticker for inactivity checks (every minute)
-	inactivityTicker := time.NewTicker(time.Minute)
-	defer inactivityTicker.Stop()
-
-	// Create ticker for cleanup (every 5 minutes)
-	cleanupTicker := time.NewTicker(5 * time.Minute)
-	defer cleanupTicker.Stop()
-
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -342,69 +160,61 @@ func (n *connectionManager) Run(ctx context.Context) {
 			return
 
 		case now := <-clockSource.C:
-			n.trafficTimer.Advance(now)
+			cm.trafficTimer.Advance(now)
 			for {
-				localIndex, has := n.trafficTimer.Purge()
+				localIndex, has := cm.trafficTimer.Purge()
 				if !has {
 					break
 				}
 
-				n.doTrafficCheck(localIndex, p, nb, out, now)
+				cm.doTrafficCheck(localIndex, p, nb, out, now)
 			}
-
-		case <-inactivityTicker.C:
-			// Check for inactive tunnels
-			n.checkInactiveTunnels()
-
-		case <-cleanupTicker.C:
-			// Periodically clean up deleted hosts
-			n.CleanupDeletedHostInfos()
 		}
 	}
 }
 
-func (n *connectionManager) doTrafficCheck(localIndex uint32, p, nb, out []byte, now time.Time) {
+func (cm *connectionManager) doTrafficCheck(localIndex uint32, p, nb, out []byte, now time.Time) {
-	decision, hostinfo, primary := n.makeTrafficDecision(localIndex, now)
+	decision, hostinfo, primary := cm.makeTrafficDecision(localIndex, now)
 
 	switch decision {
 	case deleteTunnel:
-		if n.hostMap.DeleteHostInfo(hostinfo) {
+		if cm.hostMap.DeleteHostInfo(hostinfo) {
 			// Only clearing the lighthouse cache if this is the last hostinfo for this vpn ip in the hostmap
-			n.intf.lightHouse.DeleteVpnAddrs(hostinfo.vpnAddrs)
+			cm.intf.lightHouse.DeleteVpnAddrs(hostinfo.vpnAddrs)
 		}
 
 	case closeTunnel:
-		n.intf.sendCloseTunnel(hostinfo)
+		cm.intf.sendCloseTunnel(hostinfo)
-		n.intf.closeTunnel(hostinfo)
+		cm.intf.closeTunnel(hostinfo)
 
 	case swapPrimary:
-		n.swapPrimary(hostinfo, primary)
+		cm.swapPrimary(hostinfo, primary)
 
 	case migrateRelays:
-		n.migrateRelayUsed(hostinfo, primary)
+		cm.migrateRelayUsed(hostinfo, primary)
 
 	case tryRehandshake:
-		n.tryRehandshake(hostinfo)
+		cm.tryRehandshake(hostinfo)
 
 	case sendTestPacket:
-		n.intf.SendMessageToHostInfo(header.Test, header.TestRequest, hostinfo, p, nb, out)
+		cm.intf.SendMessageToHostInfo(header.Test, header.TestRequest, hostinfo, p, nb, out)
 	}
 
-	n.resetRelayTrafficCheck(hostinfo)
+	cm.resetRelayTrafficCheck(hostinfo)
 }
 
-func (n *connectionManager) resetRelayTrafficCheck(hostinfo *HostInfo) {
+func (cm *connectionManager) resetRelayTrafficCheck(hostinfo *HostInfo) {
 	if hostinfo != nil {
-		n.relayUsedLock.Lock()
+		cm.relayUsedLock.Lock()
-		defer n.relayUsedLock.Unlock()
+		defer cm.relayUsedLock.Unlock()
 		// No need to migrate any relays, delete usage info now.
 		for _, idx := range hostinfo.relayState.CopyRelayForIdxs() {
-			delete(n.relayUsed, idx)
+			delete(cm.relayUsed, idx)
 		}
 	}
 }
 
-func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo) {
+func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo) {
 	relayFor := oldhostinfo.relayState.CopyAllRelayFor()
 
 	for _, r := range relayFor {
@@ -414,46 +224,51 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 		var relayFrom netip.Addr
 		var relayTo netip.Addr
 		switch {
-		case ok && existing.State == Established:
+		case ok:
-			// This relay already exists in newhostinfo, then do nothing.
+			switch existing.State {
-			continue
+			case Established, PeerRequested, Disestablished:
-		case ok && existing.State == Requested:
+				// This relay already exists in newhostinfo, then do nothing.
-			// The relay exists in a Requested state; re-send the request
+				continue
-			index = existing.LocalIndex
+			case Requested:
-			switch r.Type {
+				// The relay exists in a Requested state; re-send the request
-			case TerminalType:
+				index = existing.LocalIndex
-				relayFrom = n.intf.myVpnAddrs[0]
+				switch r.Type {
-				relayTo = existing.PeerAddr
+				case TerminalType:
-			case ForwardingType:
+					relayFrom = cm.intf.myVpnAddrs[0]
-				relayFrom = existing.PeerAddr
+					relayTo = existing.PeerAddr
-				relayTo = newhostinfo.vpnAddrs[0]
+				case ForwardingType:
-			default:
+					relayFrom = existing.PeerAddr
-				// should never happen
+					relayTo = newhostinfo.vpnAddrs[0]
+				default:
+					// should never happen
+					panic(fmt.Sprintf("Migrating unknown relay type: %v", r.Type))
+				}
 			}
 		case !ok:
-			n.relayUsedLock.RLock()
+			cm.relayUsedLock.RLock()
-			if _, relayUsed := n.relayUsed[r.LocalIndex]; !relayUsed {
+			if _, relayUsed := cm.relayUsed[r.LocalIndex]; !relayUsed {
 				// The relay hasn't been used; don't migrate it.
-				n.relayUsedLock.RUnlock()
+				cm.relayUsedLock.RUnlock()
 				continue
 			}
-			n.relayUsedLock.RUnlock()
+			cm.relayUsedLock.RUnlock()
 			// The relay doesn't exist at all; create some relay state and send the request.
 			var err error
-			index, err = AddRelay(n.l, newhostinfo, n.hostMap, r.PeerAddr, nil, r.Type, Requested)
+			index, err = AddRelay(cm.l, newhostinfo, cm.hostMap, r.PeerAddr, nil, r.Type, Requested)
 			if err != nil {
-				n.l.WithError(err).Error("failed to migrate relay to new hostinfo")
+				cm.l.WithError(err).Error("failed to migrate relay to new hostinfo")
 				continue
 			}
 			switch r.Type {
 			case TerminalType:
-				relayFrom = n.intf.myVpnAddrs[0]
+				relayFrom = cm.intf.myVpnAddrs[0]
 				relayTo = r.PeerAddr
 			case ForwardingType:
 				relayFrom = r.PeerAddr
 				relayTo = newhostinfo.vpnAddrs[0]
 			default:
 				// should never happen
+				panic(fmt.Sprintf("Migrating unknown relay type: %v", r.Type))
 			}
 		}
 
@@ -466,12 +281,12 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 		switch newhostinfo.GetCert().Certificate.Version() {
 		case cert.Version1:
 			if !relayFrom.Is4() {
-				n.l.Error("can not migrate v1 relay with a v6 network because the relay is not running a current nebula version")
+				cm.l.Error("can not migrate v1 relay with a v6 network because the relay is not running a current nebula version")
 				continue
 			}
 
 			if !relayTo.Is4() {
-				n.l.Error("can not migrate v1 relay with a v6 remote network because the relay is not running a current nebula version")
+				cm.l.Error("can not migrate v1 relay with a v6 remote network because the relay is not running a current nebula version")
 				continue
 			}
 
@@ -483,16 +298,16 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 			req.RelayFromAddr = netAddrToProtoAddr(relayFrom)
 			req.RelayToAddr = netAddrToProtoAddr(relayTo)
 		default:
-			newhostinfo.logger(n.l).Error("Unknown certificate version found while attempting to migrate relay")
+			newhostinfo.logger(cm.l).Error("Unknown certificate version found while attempting to migrate relay")
 			continue
 		}
 
 		msg, err := req.Marshal()
 		if err != nil {
-			n.l.WithError(err).Error("failed to marshal Control message to migrate relay")
+			cm.l.WithError(err).Error("failed to marshal Control message to migrate relay")
 		} else {
-			n.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
+			cm.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
-			n.l.WithFields(logrus.Fields{
+			cm.l.WithFields(logrus.Fields{
 				"relayFrom":           req.RelayFromAddr,
 				"relayTo":             req.RelayToAddr,
 				"initiatorRelayIndex": req.InitiatorRelayIndex,
@@ -503,46 +318,45 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 	}
 }
 
-func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time) (trafficDecision, *HostInfo, *HostInfo) {
+func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time) (trafficDecision, *HostInfo, *HostInfo) {
-	n.hostMap.RLock()
+	// Read lock the main hostmap to order decisions based on tunnels being the primary tunnel
-	defer n.hostMap.RUnlock()
+	cm.hostMap.RLock()
+	defer cm.hostMap.RUnlock()
 
-	hostinfo := n.hostMap.Indexes[localIndex]
+	hostinfo := cm.hostMap.Indexes[localIndex]
 	if hostinfo == nil {
-		n.l.WithField("localIndex", localIndex).Debugf("Not found in hostmap")
+		cm.l.WithField("localIndex", localIndex).Debugln("Not found in hostmap")
-		delete(n.pendingDeletion, localIndex)
 		return doNothing, nil, nil
 	}
 
-	if n.isInvalidCertificate(now, hostinfo) {
+	if cm.isInvalidCertificate(now, hostinfo) {
-		delete(n.pendingDeletion, hostinfo.localIndexId)
 		return closeTunnel, hostinfo, nil
 	}
 
-	primary := n.hostMap.Hosts[hostinfo.vpnAddrs[0]]
+	primary := cm.hostMap.Hosts[hostinfo.vpnAddrs[0]]
 	mainHostInfo := true
 	if primary != nil && primary != hostinfo {
 		mainHostInfo = false
 	}
 
 	// Check for traffic on this hostinfo
-	inTraffic, outTraffic := n.getAndResetTrafficCheck(localIndex)
+	inTraffic, outTraffic := cm.getAndResetTrafficCheck(hostinfo, now)
 
 	// A hostinfo is determined alive if there is incoming traffic
 	if inTraffic {
 		decision := doNothing
-		if n.l.Level >= logrus.DebugLevel {
+		if cm.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(n.l).
+			hostinfo.logger(cm.l).
 				WithField("tunnelCheck", m{"state": "alive", "method": "passive"}).
 				Debug("Tunnel status")
 		}
-		delete(n.pendingDeletion, hostinfo.localIndexId)
+		hostinfo.pendingDeletion.Store(false)
 
 		if mainHostInfo {
 			decision = tryRehandshake
 
 		} else {
-			if n.shouldSwapPrimary(hostinfo, primary) {
+			if cm.shouldSwapPrimary(hostinfo) {
 				decision = swapPrimary
 			} else {
 				// migrate the relays to the primary, if in use.
@@ -550,46 +364,55 @@ func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time
 			}
 		}
 
-		n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
+		cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
 
 		if !outTraffic {
 			// Send a punch packet to keep the NAT state alive
-			n.sendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
 		}
 
 		return decision, hostinfo, primary
 	}
 
-	if _, ok := n.pendingDeletion[hostinfo.localIndexId]; ok {
+	if hostinfo.pendingDeletion.Load() {
 		// We have already sent a test packet and nothing was returned, this hostinfo is dead
-		hostinfo.logger(n.l).
+		hostinfo.logger(cm.l).
 			WithField("tunnelCheck", m{"state": "dead", "method": "active"}).
 			Info("Tunnel status")
 
-		delete(n.pendingDeletion, hostinfo.localIndexId)
 		return deleteTunnel, hostinfo, nil
 	}
 
 	decision := doNothing
 	if hostinfo != nil && hostinfo.ConnectionState != nil && mainHostInfo {
 		if !outTraffic {
+			inactiveFor, isInactive := cm.isInactive(hostinfo, now)
+			if isInactive {
+				// Tunnel is inactive, tear it down
+				hostinfo.logger(cm.l).
+					WithField("inactiveDuration", inactiveFor).
+					WithField("primary", mainHostInfo).
+					Info("Dropping tunnel due to inactivity")
+
+				return closeTunnel, hostinfo, primary
+			}
+
 			// If we aren't sending or receiving traffic then its an unused tunnel and we don't to test the tunnel.
 			// Just maintain NAT state if configured to do so.
-			n.sendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
-			n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
+			cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
 			return doNothing, nil, nil
-
 		}
 
-		if n.punchy.GetTargetEverything() {
+		if cm.punchy.GetTargetEverything() {
 			// This is similar to the old punchy behavior with a slight optimization.
 			// We aren't receiving traffic but we are sending it, punch on all known
 			// ips in case we need to re-prime NAT state
-			n.sendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
 		}
 
-		if n.l.Level >= logrus.DebugLevel {
+		if cm.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(n.l).
+			hostinfo.logger(cm.l).
 				WithField("tunnelCheck", m{"state": "testing", "method": "active"}).
 				Debug("Tunnel status")
 		}
@@ -598,17 +421,33 @@ func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time
 		decision = sendTestPacket
 
 	} else {
-		if n.l.Level >= logrus.DebugLevel {
+		if cm.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(n.l).Debugf("Hostinfo sadness")
+			hostinfo.logger(cm.l).Debugf("Hostinfo sadness")
 		}
 	}
 
-	n.pendingDeletion[hostinfo.localIndexId] = struct{}{}
+	hostinfo.pendingDeletion.Store(true)
-	n.trafficTimer.Add(hostinfo.localIndexId, n.pendingDeletionInterval)
+	cm.trafficTimer.Add(hostinfo.localIndexId, cm.pendingDeletionInterval)
 	return decision, hostinfo, nil
 }
 
-func (n *connectionManager) shouldSwapPrimary(current, primary *HostInfo) bool {
+func (cm *connectionManager) isInactive(hostinfo *HostInfo, now time.Time) (time.Duration, bool) {
+	if cm.dropInactive.Load() == false {
+		// We aren't configured to drop inactive tunnels
+		return 0, false
+	}
+
+	inactiveDuration := now.Sub(hostinfo.lastUsed)
+	if inactiveDuration < cm.getInactivityTimeout() {
+		// It's not considered inactive
+		return inactiveDuration, false
+	}
+
+	// The tunnel is inactive
+	return inactiveDuration, true
+}
+
+func (cm *connectionManager) shouldSwapPrimary(current *HostInfo) bool {
 	// The primary tunnel is the most recent handshake to complete locally and should work entirely fine.
 	// If we are here then we have multiple tunnels for a host pair and neither side believes the same tunnel is primary.
 	// Let's sort this out.
@@ -616,73 +455,80 @@ func (n *connectionManager) shouldSwapPrimary(current, primary *HostInfo) bool {
 	// Only one side should swap because if both swap then we may never resolve to a single tunnel.
 	// vpn addr is static across all tunnels for this host pair so lets
 	// use that to determine if we should consider swapping.
-	if current.vpnAddrs[0].Compare(n.intf.myVpnAddrs[0]) < 0 {
+	if current.vpnAddrs[0].Compare(cm.intf.myVpnAddrs[0]) < 0 {
 		// Their primary vpn addr is less than mine. Do not swap.
 		return false
 	}
 
-	crt := n.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
+	crt := cm.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
 	// If this tunnel is using the latest certificate then we should swap it to primary for a bit and see if things
 	// settle down.
 	return bytes.Equal(current.ConnectionState.myCert.Signature(), crt.Signature())
 }
 
-func (n *connectionManager) swapPrimary(current, primary *HostInfo) {
+func (cm *connectionManager) swapPrimary(current, primary *HostInfo) {
-	n.hostMap.Lock()
+	cm.hostMap.Lock()
 	// Make sure the primary is still the same after the write lock. This avoids a race with a rehandshake.
-	if n.hostMap.Hosts[current.vpnAddrs[0]] == primary {
+	if cm.hostMap.Hosts[current.vpnAddrs[0]] == primary {
-		n.hostMap.unlockedMakePrimary(current)
+		cm.hostMap.unlockedMakePrimary(current)
 	}
-	n.hostMap.Unlock()
+	cm.hostMap.Unlock()
 }
 
 // isInvalidCertificate will check if we should destroy a tunnel if pki.disconnect_invalid is true and
 // the certificate is no longer valid. Block listed certificates will skip the pki.disconnect_invalid
 // check and return true.
-func (n *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
+func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
 	remoteCert := hostinfo.GetCert()
 	if remoteCert == nil {
 		return false
 	}
 
-	caPool := n.intf.pki.GetCAPool()
+	caPool := cm.intf.pki.GetCAPool()
 	err := caPool.VerifyCachedCertificate(now, remoteCert)
 	if err == nil {
 		return false
 	}
 
-	if !n.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
+	if !cm.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
 		// Block listed certificates should always be disconnected
 		return false
 	}
 
-	hostinfo.logger(n.l).WithError(err).
+	hostinfo.logger(cm.l).WithError(err).
 		WithField("fingerprint", remoteCert.Fingerprint).
 		Info("Remote certificate is no longer valid, tearing down the tunnel")
 
 	return true
 }
 
-func (n *connectionManager) sendPunch(hostinfo *HostInfo) {
+func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
-	if !n.punchy.GetPunch() {
+	if !cm.punchy.GetPunch() {
 		// Punching is disabled
 		return
 	}
 
-	if n.punchy.GetTargetEverything() {
+	if cm.intf.lightHouse.IsAnyLighthouseAddr(hostinfo.vpnAddrs) {
-		hostinfo.remotes.ForEach(n.hostMap.GetPreferredRanges(), func(addr netip.AddrPort, preferred bool) {
+		// Do not punch to lighthouses, we assume our lighthouse update interval is good enough.
-			n.metricsTxPunchy.Inc(1)
+		// In the event the update interval is not sufficient to maintain NAT state then a publicly available lighthouse
-			_ = n.intf.outside.WriteTo([]byte{1}, addr)
+		// would lose the ability to notify us and punchy.respond would become unreliable.
+		return
+	}
+
+	if cm.punchy.GetTargetEverything() {
+		hostinfo.remotes.ForEach(cm.hostMap.GetPreferredRanges(), func(addr netip.AddrPort, preferred bool) {
+			cm.metricsTxPunchy.Inc(1)
+			cm.intf.outside.WriteTo([]byte{1}, addr)
 		})
 
 	} else if hostinfo.remote.IsValid() {
-		n.metricsTxPunchy.Inc(1)
+		cm.metricsTxPunchy.Inc(1)
-		_ = n.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
+		cm.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
 	}
 }
 
-func (n *connectionManager) tryRehandshake(hostinfo *HostInfo) {
+func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
-	cs := n.intf.pki.getCertState()
+	cs := cm.intf.pki.getCertState()
 	curCrt := hostinfo.ConnectionState.myCert
 	myCrt := cs.getCertificate(curCrt.Version())
 	if curCrt.Version() >= cs.initiatingVersion && bytes.Equal(curCrt.Signature(), myCrt.Signature()) == true {
@@ -690,9 +536,9 @@ func (n *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 		return
 	}
 
-	n.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+	cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
 		WithField("reason", "local certificate is not current").
 		Info("Re-handshaking with remote")
 
-	n.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
+	cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 }

connection_manager_test.go

@@ -1,7 +1,6 @@
 package nebula
 
 import (
-	"context"
 	"crypto/ed25519"
 	"crypto/rand"
 	"net/netip"
@@ -23,7 +22,7 @@ func newTestLighthouse() *LightHouse {
 		addrMap:   map[netip.Addr]*RemoteList{},
 		queryChan: make(chan netip.Addr, 10),
 	}
-	lighthouses := map[netip.Addr]struct{}{}
+	lighthouses := []netip.Addr{}
 	staticList := map[netip.Addr]struct{}{}
 
 	lh.lighthouses.Store(&lighthouses)
@@ -64,10 +63,10 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	ctx, cancel := context.WithCancel(context.Background())
+	conf := config.NewC(l)
-	defer cancel()
+	punchy := NewPunchyFromConfig(l, conf)
-	punchy := NewPunchyFromConfig(l, config.NewC(l))
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
-	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
+	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -85,32 +84,33 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
 	// We saw traffic out to vpnIp
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
-	nc.In(hostinfo.localIndexId)
+	nc.In(hostinfo)
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
-	assert.Contains(t, nc.out, hostinfo.localIndexId)
+	assert.True(t, hostinfo.out.Load())
+	assert.True(t, hostinfo.in.Load())
 
 	// Do a traffic check tick, should not be pending deletion but should not have any in/out packets recorded
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 
 	// Do another traffic check tick, this host should be pending deletion now
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
+	assert.True(t, hostinfo.out.Load())
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.Contains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.True(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 
 	// Do a final traffic check tick, the host should now be removed
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.NotContains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs)
-	assert.NotContains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 	assert.NotContains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 }
 
@@ -146,10 +146,10 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	ctx, cancel := context.WithCancel(context.Background())
+	conf := config.NewC(l)
-	defer cancel()
+	punchy := NewPunchyFromConfig(l, conf)
-	punchy := NewPunchyFromConfig(l, config.NewC(l))
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
-	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
+	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -167,33 +167,129 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
 	// We saw traffic out to vpnIp
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
-	nc.In(hostinfo.localIndexId)
+	nc.In(hostinfo)
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.vpnAddrs[0])
+	assert.True(t, hostinfo.in.Load())
+	assert.True(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.pendingDeletion.Load())
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 
 	// Do a traffic check tick, should not be pending deletion but should not have any in/out packets recorded
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 
 	// Do another traffic check tick, this host should be pending deletion now
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.Contains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.True(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 
 	// We saw traffic, should no longer be pending deletion
-	nc.In(hostinfo.localIndexId)
+	nc.In(hostinfo)
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
+	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
+	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
+}
+
+func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
+	l := test.NewLogger()
+	localrange := netip.MustParsePrefix("10.1.1.1/24")
+	vpnAddrs := []netip.Addr{netip.MustParseAddr("172.1.1.2")}
+	preferredRanges := []netip.Prefix{localrange}
+
+	// Very incomplete mock objects
+	hostMap := newHostMap(l)
+	hostMap.preferredRanges.Store(&preferredRanges)
+
+	cs := &CertState{
+		initiatingVersion: cert.Version1,
+		privateKey:        []byte{},
+		v1Cert:            &dummyCert{version: cert.Version1},
+		v1HandshakeBytes:  []byte{},
+	}
+
+	lh := newTestLighthouse()
+	ifce := &Interface{
+		hostMap:          hostMap,
+		inside:           &test.NoopTun{},
+		outside:          &udp.NoopConn{},
+		firewall:         &Firewall{},
+		lightHouse:       lh,
+		pki:              &PKI{},
+		handshakeManager: NewHandshakeManager(l, hostMap, lh, &udp.NoopConn{}, defaultHandshakeConfig),
+		l:                l,
+	}
+	ifce.pki.cs.Store(cs)
+
+	// Create manager
+	conf := config.NewC(l)
+	conf.Settings["tunnels"] = map[string]any{
+		"drop_inactive": true,
+	}
+	punchy := NewPunchyFromConfig(l, conf)
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	assert.True(t, nc.dropInactive.Load())
+	nc.intf = ifce
+
+	// Add an ip we have established a connection w/ to hostmap
+	hostinfo := &HostInfo{
+		vpnAddrs:      vpnAddrs,
+		localIndexId:  1099,
+		remoteIndexId: 9901,
+	}
+	hostinfo.ConnectionState = &ConnectionState{
+		myCert: &dummyCert{version: cert.Version1},
+		H:      &noise.HandshakeState{},
+	}
+	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
+
+	// Do a traffic check tick, in and out should be cleared but should not be pending deletion
+	nc.Out(hostinfo)
+	nc.In(hostinfo)
+	assert.True(t, hostinfo.out.Load())
+	assert.True(t, hostinfo.in.Load())
+
+	now := time.Now()
+	decision, _, _ := nc.makeTrafficDecision(hostinfo.localIndexId, now)
+	assert.Equal(t, tryRehandshake, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
+
+	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, now.Add(time.Second*5))
+	assert.Equal(t, doNothing, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
+
+	// Do another traffic check tick, should still not be pending deletion
+	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, now.Add(time.Second*10))
+	assert.Equal(t, doNothing, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
+	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
+	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
+
+	// Finally advance beyond the inactivity timeout
+	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, now.Add(time.Minute*10))
+	assert.Equal(t, closeTunnel, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 }
@@ -264,10 +360,10 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	ifce.disconnectInvalid.Store(true)
 
 	// Create manager
-	ctx, cancel := context.WithCancel(context.Background())
+	conf := config.NewC(l)
-	defer cancel()
+	punchy := NewPunchyFromConfig(l, conf)
-	punchy := NewPunchyFromConfig(l, config.NewC(l))
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
-	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
+	nc.intf = ifce
 	ifce.connectionManager = nc
 
 	hostinfo := &HostInfo{
@@ -350,6 +446,10 @@ func (d *dummyCert) PublicKey() []byte {
 	return d.publicKey
 }
 
+func (d *dummyCert) MarshalPublicKeyPEM() []byte {
+	return cert.MarshalPublicKeyToPEM(d.curve, d.publicKey)
+}
+
 func (d *dummyCert) Signature() []byte {
 	return d.signature
 }

control.go

@@ -26,14 +26,15 @@ type controlHostLister interface {
 }
 
 type Control struct {
-	f               *Interface
+	f                      *Interface
-	l               *logrus.Logger
+	l                      *logrus.Logger
-	ctx             context.Context
+	ctx                    context.Context
-	cancel          context.CancelFunc
+	cancel                 context.CancelFunc
-	sshStart        func()
+	sshStart               func()
-	statsStart      func()
+	statsStart             func()
-	dnsStart        func()
+	dnsStart               func()
-	lighthouseStart func()
+	lighthouseStart        func()
+	connectionManagerStart func(context.Context)
 }
 
 type ControlHostInfo struct {
@@ -63,6 +64,9 @@ func (c *Control) Start() {
 	if c.dnsStart != nil {
 		go c.dnsStart()
 	}
+	if c.connectionManagerStart != nil {
+		go c.connectionManagerStart(c.ctx)
+	}
 	if c.lighthouseStart != nil {
 		c.lighthouseStart()
 	}

control_test.go

@@ -53,7 +53,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		localIndexId:  201,
 		vpnAddrs:      []netip.Addr{vpnIp},
 		relayState: RelayState{
-			relays:         map[netip.Addr]struct{}{},
+			relays:         nil,
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},
@@ -72,7 +72,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		localIndexId:  201,
 		vpnAddrs:      []netip.Addr{vpnIp2},
 		relayState: RelayState{
-			relays:         map[netip.Addr]struct{}{},
+			relays:         nil,
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},

e2e/handshakes_test.go

@@ -506,7 +506,7 @@ func TestReestablishRelays(t *testing.T) {
 	curIndexes := len(myControl.GetHostmap().Indexes)
 	for curIndexes >= start {
 		curIndexes = len(myControl.GetHostmap().Indexes)
-		r.Logf("Wait for the dead index to go away:start=%v indexes, currnet=%v indexes", start, curIndexes)
+		r.Logf("Wait for the dead index to go away:start=%v indexes, current=%v indexes", start, curIndexes)
 		myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me should fail"))
 
 		r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
@@ -1052,6 +1052,9 @@ func TestRehandshakingLoser(t *testing.T) {
 	t.Log("Stand up a tunnel between me and them")
 	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
 
+	myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), false)
+	theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
+
 	r.RenderHostmaps("Starting hostmaps", myControl, theirControl)
 
 	r.Log("Renew their certificate and spin until mine sees it")

e2e/router/router.go

@@ -700,6 +700,7 @@ func (r *R) FlushAll() {
 			r.Unlock()
 			panic("Can't FlushAll for host: " + p.To.String())
 		}
+		receiver.InjectUDPPacket(p)
 		r.Unlock()
 	}
 }

e2e/tunnels_test.go

@@ -0,0 +1,57 @@
+//go:build e2e_testing
+// +build e2e_testing
+
+package e2e
+
+import (
+	"testing"
+	"time"
+
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/cert_test"
+	"github.com/slackhq/nebula/e2e/router"
+)
+
+func TestDropInactiveTunnels(t *testing.T) {
+	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
+	// under ideal conditions
+	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("Go inactive and wait for the tunnels to get dropped")
+	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*30 {
+			t.Fatal("Tunnel should have been declared inactive after 5 seconds and before 30 seconds")
+		}
+
+		time.Sleep(1 * time.Second)
+		r.FlushAll()
+	}
+
+	r.Logf("Inactive tunnels were dropped within %v", time.Since(waitStart))
+	myControl.Stop()
+	theirControl.Stop()
+}

examples/config.yml

@@ -338,6 +338,18 @@ logging:
   # after receiving the response for lighthouse queries
   #trigger_buffer: 64
 
+# Tunnel manager settings
+#tunnels:
+  # drop_inactive controls whether inactive tunnels are maintained or dropped after the inactive_timeout period has
+  # elapsed.
+  # In general, it is a good idea to enable this setting. It will be enabled by default in a future release.
+  # This setting is reloadable
+  #drop_inactive: false
+
+  # inactivity_timeout controls how long a tunnel MUST NOT see any inbound or outbound traffic before being considered
+  # inactive and eligible to be dropped.
+  # This setting is reloadable
+  #inactivity_timeout: 10m
 
 # Nebula security group configuration
 firewall:

firewall_test.go

@@ -68,6 +68,9 @@ func TestFirewall_AddRule(t *testing.T) {
 	ti, err := netip.ParsePrefix("1.2.3.4/32")
 	require.NoError(t, err)
 
+	ti6, err := netip.ParsePrefix("fd12::34/128")
+	require.NoError(t, err)
+
 	require.NoError(t, fw.AddRule(true, firewall.ProtoTCP, 1, 1, []string{}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
 	// An empty rule is any
 	assert.True(t, fw.InRules.TCP[1].Any.Any.Any)
@@ -92,12 +95,24 @@ func TestFirewall_AddRule(t *testing.T) {
 	_, ok := fw.OutRules.AnyProto[1].Any.CIDR.Get(ti)
 	assert.True(t, ok)
 
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
+	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", ti6, netip.Prefix{}, "", ""))
+	assert.Nil(t, fw.OutRules.AnyProto[1].Any.Any)
+	_, ok = fw.OutRules.AnyProto[1].Any.CIDR.Get(ti6)
+	assert.True(t, ok)
+
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, ti, "", ""))
 	assert.NotNil(t, fw.OutRules.AnyProto[1].Any.Any)
 	_, ok = fw.OutRules.AnyProto[1].Any.Any.LocalCIDR.Get(ti)
 	assert.True(t, ok)
 
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
+	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, ti6, "", ""))
+	assert.NotNil(t, fw.OutRules.AnyProto[1].Any.Any)
+	_, ok = fw.OutRules.AnyProto[1].Any.Any.LocalCIDR.Get(ti6)
+	assert.True(t, ok)
+
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoUDP, 1, 1, []string{"g1"}, "", netip.Prefix{}, netip.Prefix{}, "ca-name", ""))
 	assert.Contains(t, fw.InRules.UDP[1].CANames, "ca-name")
@@ -117,6 +132,13 @@ func TestFirewall_AddRule(t *testing.T) {
 	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 0, 0, []string{}, "", anyIp, netip.Prefix{}, "", ""))
 	assert.True(t, fw.OutRules.AnyProto[0].Any.Any.Any)
 
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
+	anyIp6, err := netip.ParsePrefix("::/0")
+	require.NoError(t, err)
+
+	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 0, 0, []string{}, "", anyIp6, netip.Prefix{}, "", ""))
+	assert.True(t, fw.OutRules.AnyProto[0].Any.Any.Any)
+
 	// Test error conditions
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.Error(t, fw.AddRule(true, math.MaxUint8, 0, 0, []string{}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -199,6 +221,82 @@ func TestFirewall_Drop(t *testing.T) {
 	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
 }
 
+func TestFirewall_DropV6(t *testing.T) {
+	l := test.NewLogger()
+	ob := &bytes.Buffer{}
+	l.SetOutput(ob)
+
+	p := firewall.Packet{
+		LocalAddr:  netip.MustParseAddr("fd12::34"),
+		RemoteAddr: netip.MustParseAddr("fd12::34"),
+		LocalPort:  10,
+		RemotePort: 90,
+		Protocol:   firewall.ProtoUDP,
+		Fragment:   false,
+	}
+
+	c := dummyCert{
+		name:     "host1",
+		networks: []netip.Prefix{netip.MustParsePrefix("fd12::34/120")},
+		groups:   []string{"default-group"},
+		issuer:   "signer-shasum",
+	}
+	h := HostInfo{
+		ConnectionState: &ConnectionState{
+			peerCert: &cert.CachedCertificate{
+				Certificate:    &c,
+				InvertedGroups: map[string]struct{}{"default-group": {}},
+			},
+		},
+		vpnAddrs: []netip.Addr{netip.MustParseAddr("fd12::34")},
+	}
+	h.buildNetworks(c.networks, c.unsafeNetworks)
+
+	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
+	cp := cert.NewCAPool()
+
+	// Drop outbound
+	assert.Equal(t, ErrNoMatchingRule, fw.Drop(p, false, &h, cp, nil))
+	// Allow inbound
+	resetConntrack(fw)
+	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	// Allow outbound because conntrack
+	require.NoError(t, fw.Drop(p, false, &h, cp, nil))
+
+	// test remote mismatch
+	oldRemote := p.RemoteAddr
+	p.RemoteAddr = netip.MustParseAddr("fd12::56")
+	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrInvalidRemoteIP)
+	p.RemoteAddr = oldRemote
+
+	// ensure signer doesn't get in the way of group checks
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", netip.Prefix{}, netip.Prefix{}, "", "signer-shasum"))
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", netip.Prefix{}, netip.Prefix{}, "", "signer-shasum-bad"))
+	assert.Equal(t, fw.Drop(p, true, &h, cp, nil), ErrNoMatchingRule)
+
+	// test caSha doesn't drop on match
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", netip.Prefix{}, netip.Prefix{}, "", "signer-shasum-bad"))
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", netip.Prefix{}, netip.Prefix{}, "", "signer-shasum"))
+	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+
+	// ensure ca name doesn't get in the way of group checks
+	cp.CAs["signer-shasum"] = &cert.CachedCertificate{Certificate: &dummyCert{name: "ca-good"}}
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", netip.Prefix{}, netip.Prefix{}, "ca-good", ""))
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", netip.Prefix{}, netip.Prefix{}, "ca-good-bad", ""))
+	assert.Equal(t, fw.Drop(p, true, &h, cp, nil), ErrNoMatchingRule)
+
+	// test caName doesn't drop on match
+	cp.CAs["signer-shasum"] = &cert.CachedCertificate{Certificate: &dummyCert{name: "ca-good"}}
+	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", netip.Prefix{}, netip.Prefix{}, "ca-good-bad", ""))
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", netip.Prefix{}, netip.Prefix{}, "ca-good", ""))
+	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+}
+
 func BenchmarkFirewallTable_match(b *testing.B) {
 	f := &Firewall{}
 	ft := FirewallTable{
@@ -208,6 +306,10 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 	pfix := netip.MustParsePrefix("172.1.1.1/32")
 	_ = ft.TCP.addRule(f, 10, 10, []string{"good-group"}, "good-host", pfix, netip.Prefix{}, "", "")
 	_ = ft.TCP.addRule(f, 100, 100, []string{"good-group"}, "good-host", netip.Prefix{}, pfix, "", "")
+
+	pfix6 := netip.MustParsePrefix("fd11::11/128")
+	_ = ft.TCP.addRule(f, 10, 10, []string{"good-group"}, "good-host", pfix6, netip.Prefix{}, "", "")
+	_ = ft.TCP.addRule(f, 100, 100, []string{"good-group"}, "good-host", netip.Prefix{}, pfix6, "", "")
 	cp := cert.NewCAPool()
 
 	b.Run("fail on proto", func(b *testing.B) {
@@ -239,6 +341,15 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 			assert.False(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 100, LocalAddr: ip.Addr()}, true, c, cp))
 		}
 	})
+	b.Run("pass proto, port, fail on local CIDRv6", func(b *testing.B) {
+		c := &cert.CachedCertificate{
+			Certificate: &dummyCert{},
+		}
+		ip := netip.MustParsePrefix("fd99::99/128")
+		for n := 0; n < b.N; n++ {
+			assert.False(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 100, LocalAddr: ip.Addr()}, true, c, cp))
+		}
+	})
 
 	b.Run("pass proto, port, any local CIDR, fail all group, name, and cidr", func(b *testing.B) {
 		c := &cert.CachedCertificate{
@@ -252,6 +363,18 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 			assert.False(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 10}, true, c, cp))
 		}
 	})
+	b.Run("pass proto, port, any local CIDRv6, fail all group, name, and cidr", func(b *testing.B) {
+		c := &cert.CachedCertificate{
+			Certificate: &dummyCert{
+				name:     "nope",
+				networks: []netip.Prefix{netip.MustParsePrefix("fd99::99/128")},
+			},
+			InvertedGroups: map[string]struct{}{"nope": {}},
+		}
+		for n := 0; n < b.N; n++ {
+			assert.False(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 10}, true, c, cp))
+		}
+	})
 
 	b.Run("pass proto, port, specific local CIDR, fail all group, name, and cidr", func(b *testing.B) {
 		c := &cert.CachedCertificate{
@@ -265,6 +388,18 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 			assert.False(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 100, LocalAddr: pfix.Addr()}, true, c, cp))
 		}
 	})
+	b.Run("pass proto, port, specific local CIDRv6, fail all group, name, and cidr", func(b *testing.B) {
+		c := &cert.CachedCertificate{
+			Certificate: &dummyCert{
+				name:     "nope",
+				networks: []netip.Prefix{netip.MustParsePrefix("fd99:99/128")},
+			},
+			InvertedGroups: map[string]struct{}{"nope": {}},
+		}
+		for n := 0; n < b.N; n++ {
+			assert.False(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 100, LocalAddr: pfix6.Addr()}, true, c, cp))
+		}
+	})
 
 	b.Run("pass on group on any local cidr", func(b *testing.B) {
 		c := &cert.CachedCertificate{
@@ -289,6 +424,17 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 			assert.True(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 100, LocalAddr: pfix.Addr()}, true, c, cp))
 		}
 	})
+	b.Run("pass on group on specific local cidr6", func(b *testing.B) {
+		c := &cert.CachedCertificate{
+			Certificate: &dummyCert{
+				name: "nope",
+			},
+			InvertedGroups: map[string]struct{}{"good-group": {}},
+		}
+		for n := 0; n < b.N; n++ {
+			assert.True(b, ft.match(firewall.Packet{Protocol: firewall.ProtoTCP, LocalPort: 100, LocalAddr: pfix6.Addr()}, true, c, cp))
+		}
+	})
 
 	b.Run("pass on name", func(b *testing.B) {
 		c := &cert.CachedCertificate{
@@ -447,6 +593,42 @@ func TestFirewall_Drop3(t *testing.T) {
 	require.NoError(t, fw.Drop(p, true, &h1, cp, nil))
 }
 
+func TestFirewall_Drop3V6(t *testing.T) {
+	l := test.NewLogger()
+	ob := &bytes.Buffer{}
+	l.SetOutput(ob)
+
+	p := firewall.Packet{
+		LocalAddr:  netip.MustParseAddr("fd12::34"),
+		RemoteAddr: netip.MustParseAddr("fd12::34"),
+		LocalPort:  1,
+		RemotePort: 1,
+		Protocol:   firewall.ProtoUDP,
+		Fragment:   false,
+	}
+
+	network := netip.MustParsePrefix("fd12::34/120")
+	c := cert.CachedCertificate{
+		Certificate: &dummyCert{
+			name:     "host-owner",
+			networks: []netip.Prefix{network},
+		},
+	}
+	h := HostInfo{
+		ConnectionState: &ConnectionState{
+			peerCert: &c,
+		},
+		vpnAddrs: []netip.Addr{network.Addr()},
+	}
+	h.buildNetworks(c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
+
+	// Test a remote address match
+	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
+	cp := cert.NewCAPool()
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "", netip.MustParsePrefix("fd12::34/120"), netip.Prefix{}, "", ""))
+	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+}
+
 func TestFirewall_DropConntrackReload(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
@@ -510,6 +692,50 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
 }
 
+func TestFirewall_DropIPSpoofing(t *testing.T) {
+	l := test.NewLogger()
+	ob := &bytes.Buffer{}
+	l.SetOutput(ob)
+
+	c := cert.CachedCertificate{
+		Certificate: &dummyCert{
+			name:     "host-owner",
+			networks: []netip.Prefix{netip.MustParsePrefix("192.0.2.1/24")},
+		},
+	}
+
+	c1 := cert.CachedCertificate{
+		Certificate: &dummyCert{
+			name:           "host",
+			networks:       []netip.Prefix{netip.MustParsePrefix("192.0.2.2/24")},
+			unsafeNetworks: []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")},
+		},
+	}
+	h1 := HostInfo{
+		ConnectionState: &ConnectionState{
+			peerCert: &c1,
+		},
+		vpnAddrs: []netip.Addr{c1.Certificate.Networks()[0].Addr()},
+	}
+	h1.buildNetworks(c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
+
+	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
+
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
+	cp := cert.NewCAPool()
+
+	// Packet spoofed by `c1`. Note that the remote addr is not a valid one.
+	p := firewall.Packet{
+		LocalAddr:  netip.MustParseAddr("192.0.2.1"),
+		RemoteAddr: netip.MustParseAddr("192.0.2.3"),
+		LocalPort:  1,
+		RemotePort: 1,
+		Protocol:   firewall.ProtoUDP,
+		Fragment:   false,
+	}
+	assert.Equal(t, fw.Drop(p, true, &h1, cp, nil), ErrInvalidRemoteIP)
+}
+
 func BenchmarkLookup(b *testing.B) {
 	ml := func(m map[string]struct{}, a [][]string) {
 		for n := 0; n < b.N; n++ {
@@ -727,6 +953,21 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: netip.Prefix{}, localIp: cidr}, mf.lastCall)
 
+	// Test adding rule with cidr ipv6
+	cidr6 := netip.MustParsePrefix("fd00::/8")
+	conf = config.NewC(l)
+	mf = &mockFirewall{}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": cidr6.String()}}}
+	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
+	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: cidr6, localIp: netip.Prefix{}}, mf.lastCall)
+
+	// Test adding rule with local_cidr ipv6
+	conf = config.NewC(l)
+	mf = &mockFirewall{}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": cidr6.String()}}}
+	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
+	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: netip.Prefix{}, localIp: cidr6}, mf.lastCall)
+
 	// Test adding rule with ca_sha
 	conf = config.NewC(l)
 	mf = &mockFirewall{}

go.mod

@@ -1,39 +1,37 @@
 module github.com/slackhq/nebula
 
-go 1.23.0
+go 1.25
-
-toolchain go1.24.1
 
 require (
-	dario.cat/mergo v1.0.1
+	dario.cat/mergo v1.0.2
 	github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be
 	github.com/armon/go-radix v1.0.0
 	github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
 	github.com/flynn/noise v1.1.0
-	github.com/gaissmai/bart v0.20.4
+	github.com/gaissmai/bart v0.25.0
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
-	github.com/kardianos/service v1.2.2
+	github.com/kardianos/service v1.2.4
-	github.com/miekg/dns v1.1.65
+	github.com/miekg/dns v1.1.68
 	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 v1.22.0
+	github.com/prometheus/client_golang v1.23.2
 	github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475
 	github.com/sirupsen/logrus v1.9.3
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e
 	github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6
-	github.com/stretchr/testify v1.10.0
+	github.com/stretchr/testify v1.11.1
-	github.com/vishvananda/netlink v1.3.0
+	github.com/vishvananda/netlink v1.3.1
-	golang.org/x/crypto v0.37.0
+	golang.org/x/crypto v0.43.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.39.0
+	golang.org/x/net v0.45.0
-	golang.org/x/sync v0.13.0
+	golang.org/x/sync v0.17.0
-	golang.org/x/sys v0.32.0
+	golang.org/x/sys v0.37.0
-	golang.org/x/term v0.31.0
+	golang.org/x/term v0.36.0
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2
 	golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b
 	golang.zx2c4.com/wireguard/windows v0.5.3
-	google.golang.org/protobuf v1.36.6
+	google.golang.org/protobuf v1.36.8
 	gopkg.in/yaml.v3 v3.0.1
 	gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe
 )
@@ -45,11 +43,12 @@ require (
 	github.com/google/btree v1.1.2 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
-	github.com/prometheus/client_model v0.6.1 // indirect
+	github.com/prometheus/client_model v0.6.2 // indirect
-	github.com/prometheus/common v0.62.0 // indirect
+	github.com/prometheus/common v0.66.1 // indirect
-	github.com/prometheus/procfs v0.15.1 // indirect
+	github.com/prometheus/procfs v0.16.1 // indirect
-	github.com/vishvananda/netns v0.0.4 // indirect
+	github.com/vishvananda/netns v0.0.5 // indirect
-	golang.org/x/mod v0.23.0 // indirect
+	go.yaml.in/yaml/v2 v2.4.2 // indirect
+	golang.org/x/mod v0.24.0 // indirect
 	golang.org/x/time v0.5.0 // indirect
-	golang.org/x/tools v0.30.0 // indirect
+	golang.org/x/tools v0.33.0 // indirect
 )

go.sum

@@ -1,6 +1,6 @@
 cloud.google.com/go v0.34.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
-dario.cat/mergo v1.0.1 h1:Ra4+bf83h2ztPIQYNP99R6m+Y7KfnARDfID+a+vLl4s=
+dario.cat/mergo v1.0.2 h1:85+piFYR1tMbRrLcDwR18y4UKJ3aH1Tbzi24VRW1TK8=
-dario.cat/mergo v1.0.1/go.mod h1:uNxQE+84aUszobStD9th8a29P2fMDhsBdgRYvZOxGmk=
+dario.cat/mergo v1.0.2/go.mod h1:E/hbnu0NxMFBjpMIE34DRGLWqDy0g5FuKDhCb31ngxA=
 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=
@@ -24,8 +24,8 @@ github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/flynn/noise v1.1.0 h1:KjPQoQCEFdZDiP03phOvGi11+SVVhBG2wOWAorLsstg=
 github.com/flynn/noise v1.1.0/go.mod h1:xbMo+0i6+IGbYdJhF31t2eR1BIU0CYc12+BNAKwUTag=
-github.com/gaissmai/bart v0.20.4 h1:Ik47r1fy3jRVU+1eYzKSW3ho2UgBVTVnUS8O993584U=
+github.com/gaissmai/bart v0.25.0 h1:eqiokVPqM3F94vJ0bTHXHtH91S8zkKL+bKh+BsGOsJM=
-github.com/gaissmai/bart v0.20.4/go.mod h1:cEed+ge8dalcbpi8wtS9x9m2hn/fNJH5suhdGQOHnYk=
+github.com/gaissmai/bart v0.25.0/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
 github.com/go-kit/kit v0.8.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/kit v0.9.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/log v0.1.0/go.mod h1:zbhenjAZHb184qTLMA9ZjW7ThYL0H2mk7Q6pNt4vbaY=
@@ -64,8 +64,8 @@ github.com/json-iterator/go v1.1.10/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/
 github.com/json-iterator/go v1.1.11/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/uOdHXbAo4=
 github.com/julienschmidt/httprouter v1.2.0/go.mod h1:SYymIcj16QtmaHHD7aYtjjsJG7VTCxuUUipMqKk8s4w=
 github.com/julienschmidt/httprouter v1.3.0/go.mod h1:JR6WtHb+2LUe8TCKY3cZOxFyyO8IZAc4RVcycCCAKdM=
-github.com/kardianos/service v1.2.2 h1:ZvePhAHfvo0A7Mftk/tEzqEZ7Q4lgnR8sGz4xu1YX60=
+github.com/kardianos/service v1.2.4 h1:XNlGtZOYNx2u91urOdg/Kfmc+gfmuIo1Dd3rEi2OgBk=
-github.com/kardianos/service v1.2.2/go.mod h1:CIMRFEJVL+0DS1a3Nx06NaMn4Dz63Ng6O7dl0qH0zVM=
+github.com/kardianos/service v1.2.4/go.mod h1:E4V9ufUuY82F7Ztlu1eN9VXWIQxg8NoLQlmFe0MtrXc=
 github.com/kisielk/errcheck v1.5.0/go.mod h1:pFxgyoBC7bSaBwPgfKdkLd5X25qrDl4LWUI2bnpBCr8=
 github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
 github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
@@ -83,8 +83,8 @@ 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 v1.1.65 h1:0+tIPHzUW0GCge7IiK3guGP57VAw7hoPDfApjkMD1Fc=
+github.com/miekg/dns v1.1.68 h1:jsSRkNozw7G/mnmXULynzMNIsgY2dHC8LO6U6Ij2JEA=
-github.com/miekg/dns v1.1.65/go.mod h1:Dzw9769uoKVaLuODMDZz9M6ynFU6Em65csPuoi8G0ck=
+github.com/miekg/dns v1.1.68/go.mod h1:fujopn7TB3Pu3JM69XaawiU0wqjpL9/8xGop5UrTPps=
 github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b h1:J/AzCvg5z0Hn1rqZUJjpbzALUmkKX0Zwbc/i4fw7Sfk=
 github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b/go.mod h1:XsNlhZGX73bx86s2hdc/FuaLm2CPZJemRLMA+WTFxgs=
 github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
@@ -106,24 +106,24 @@ github.com/prometheus/client_golang v0.9.1/go.mod h1:7SWBe2y4D6OKWSNQJUaRYU/AaXP
 github.com/prometheus/client_golang v1.0.0/go.mod h1:db9x61etRT2tGnBNRi70OPL5FsnadC4Ky3P0J6CfImo=
 github.com/prometheus/client_golang v1.7.1/go.mod h1:PY5Wy2awLA44sXw4AOSfFBetzPP4j5+D6mVACh+pe2M=
 github.com/prometheus/client_golang v1.11.0/go.mod h1:Z6t4BnS23TR94PD6BsDNk8yVqroYurpAkEiz0P2BEV0=
-github.com/prometheus/client_golang v1.22.0 h1:rb93p9lokFEsctTys46VnV1kLCDpVZ0a/Y92Vm0Zc6Q=
+github.com/prometheus/client_golang v1.23.2 h1:Je96obch5RDVy3FDMndoUsjAhG5Edi49h0RJWRi/o0o=
-github.com/prometheus/client_golang v1.22.0/go.mod h1:R7ljNsLXhuQXYZYtw6GAE9AZg8Y7vEW5scdCXrWRXC0=
+github.com/prometheus/client_golang v1.23.2/go.mod h1:Tb1a6LWHB3/SPIzCoaDXI4I8UHKeFTEQ1YCr+0Gyqmg=
 github.com/prometheus/client_model v0.0.0-20180712105110-5c3871d89910/go.mod h1:MbSGuTsp3dbXC40dX6PRTWyKYBIrTGTE9sqQNg2J8bo=
 github.com/prometheus/client_model v0.0.0-20190129233127-fd36f4220a90/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
 github.com/prometheus/client_model v0.2.0/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
-github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=
+github.com/prometheus/client_model v0.6.2 h1:oBsgwpGs7iVziMvrGhE53c/GrLUsZdHnqNwqPLxwZyk=
-github.com/prometheus/client_model v0.6.1/go.mod h1:OrxVMOVHjw3lKMa8+x6HeMGkHMQyHDk9E3jmP2AmGiY=
+github.com/prometheus/client_model v0.6.2/go.mod h1:y3m2F6Gdpfy6Ut/GBsUqTWZqCUvMVzSfMLjcu6wAwpE=
 github.com/prometheus/common v0.4.1/go.mod h1:TNfzLD0ON7rHzMJeJkieUDPYmFC7Snx/y86RQel1bk4=
 github.com/prometheus/common v0.10.0/go.mod h1:Tlit/dnDKsSWFlCLTWaA1cyBgKHSMdTB80sz/V91rCo=
 github.com/prometheus/common v0.26.0/go.mod h1:M7rCNAaPfAosfx8veZJCuw84e35h3Cfd9VFqTh1DIvc=
-github.com/prometheus/common v0.62.0 h1:xasJaQlnWAeyHdUBeGjXmutelfJHWMRr+Fg4QszZ2Io=
+github.com/prometheus/common v0.66.1 h1:h5E0h5/Y8niHc5DlaLlWLArTQI7tMrsfQjHV+d9ZoGs=
-github.com/prometheus/common v0.62.0/go.mod h1:vyBcEuLSvWos9B1+CyL7JZ2up+uFzXhkqml0W5zIY1I=
+github.com/prometheus/common v0.66.1/go.mod h1:gcaUsgf3KfRSwHY4dIMXLPV0K/Wg1oZ8+SbZk/HH/dA=
 github.com/prometheus/procfs v0.0.0-20181005140218-185b4288413d/go.mod h1:c3At6R/oaqEKCNdg8wHV1ftS6bRYblBhIjjI8uT2IGk=
 github.com/prometheus/procfs v0.0.2/go.mod h1:TjEm7ze935MbeOT/UhFTIMYKhuLP4wbCsTZCD3I8kEA=
 github.com/prometheus/procfs v0.1.3/go.mod h1:lV6e/gmhEcM9IjHGsFOCxxuZ+z1YqCvr4OA4YeYWdaU=
 github.com/prometheus/procfs v0.6.0/go.mod h1:cz+aTbrPOrUb4q7XlbU9ygM+/jj0fzG6c1xBZuNvfVA=
-github.com/prometheus/procfs v0.15.1 h1:YagwOFzUgYfKKHX6Dr+sHT7km/hxC76UB0learggepc=
+github.com/prometheus/procfs v0.16.1 h1:hZ15bTNuirocR6u0JZ6BAHHmwS1p8B4P6MRqxtzMyRg=
-github.com/prometheus/procfs v0.15.1/go.mod h1:fB45yRUv8NstnjriLhBQLuOUt+WW4BsoGhij/e3PBqk=
+github.com/prometheus/procfs v0.16.1/go.mod h1:teAbpZRB1iIAJYREa1LsoWUXykVXA1KlTmWl8x/U+Is=
 github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475 h1:N/ElC8H3+5XpJzTSTfLsJV/mx9Q9g7kxmchpfZyxgzM=
 github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475/go.mod h1:bCqnVzQkZxMG4s8nGwiZ5l3QUCyqpo9Y+/ZMZ9VjZe4=
 github.com/rogpeppe/go-internal v1.10.0 h1:TMyTOH3F/DB16zRVcYyreMH6GnZZrwQVAoYjRBZyWFQ=
@@ -143,29 +143,33 @@ github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXf
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
 github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4=
 github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-github.com/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
+github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
-github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
+github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
-github.com/vishvananda/netlink v1.3.0 h1:X7l42GfcV4S6E4vHTsw48qbrV+9PVojNfIhZcwQdrZk=
+github.com/vishvananda/netlink v1.3.1 h1:3AEMt62VKqz90r0tmNhog0r/PpWKmrEShJU0wJW6bV0=
-github.com/vishvananda/netlink v1.3.0/go.mod h1:i6NetklAujEcC6fK0JPjT8qSwWyO0HLn4UKG+hGqeJs=
+github.com/vishvananda/netlink v1.3.1/go.mod h1:ARtKouGSTGchR8aMwmkzC0qiNPrrWO5JS/XMVl45+b4=
-github.com/vishvananda/netns v0.0.4 h1:Oeaw1EM2JMxD51g9uhtC0D7erkIjgmj8+JZc26m1YX8=
+github.com/vishvananda/netns v0.0.5 h1:DfiHV+j8bA32MFM7bfEunvT8IAqQ/NzSJHtcmW5zdEY=
-github.com/vishvananda/netns v0.0.4/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
+github.com/vishvananda/netns v0.0.5/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
 github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
 github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
+go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=
+go.uber.org/goleak v1.3.0/go.mod h1:CoHD4mav9JJNrW/WLlf7HGZPjdw8EucARQHekz1X6bE=
+go.yaml.in/yaml/v2 v2.4.2 h1:DzmwEr2rDGHl7lsFgAHxmNz/1NlQ7xLIrlN2h5d1eGI=
+go.yaml.in/yaml/v2 v2.4.2/go.mod h1:081UH+NErpNdqlCXm3TtEran0rJZGxAYx9hb/ELlsPU=
 golang.org/x/crypto v0.0.0-20180904163835-0709b304e793/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/crypto v0.0.0-20210322153248-0c34fe9e7dc2/go.mod h1:T9bdIzuCu7OtxOm1hfPfRQxPLYneinmdGuTeoZ9dtd4=
-golang.org/x/crypto v0.37.0 h1:kJNSjF/Xp7kU0iB2Z+9viTPMW4EqqsrywMXLJOOsXSE=
+golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
-golang.org/x/crypto v0.37.0/go.mod h1:vg+k43peMZ0pUMhYmVAWysMK35e6ioLh3wB8ZCAfbVc=
+golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090 h1:Di6/M8l0O2lCLc6VVRWhgCiApHV8MnQurBnFSHsQtNY=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090/go.mod h1:FXUEEKJgO7OQYeo8N01OfiKP8RXMtf6e8aTskBGqWdc=
 golang.org/x/lint v0.0.0-20200302205851-738671d3881b/go.mod h1:3xt1FjdF8hUf6vQPIChWIBhFzV8gjjsPE/fR3IyQdNY=
 golang.org/x/mod v0.1.1-0.20191105210325-c90efee705ee/go.mod h1:QqPTAvyqsEbceGzBzNggFXnrqF1CaUcvgkdR5Ot7KZg=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
-golang.org/x/mod v0.23.0 h1:Zb7khfcRGKk+kqfxFaP5tZqCnDZMjC5VtUBs87Hr6QM=
+golang.org/x/mod v0.24.0 h1:ZfthKaKaT4NrhGVZHO1/WDTwGES4De8KtWO0SIbNJMU=
-golang.org/x/mod v0.23.0/go.mod h1:6SkKJ3Xj0I0BrPOZoBy3bdMptDDU9oJrpohJ3eWZ1fY=
+golang.org/x/mod v0.24.0/go.mod h1:IXM97Txy2VM4PJ3gI61r1YEk/gAj6zAHN3AdZt6S9Ww=
 golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20181114220301-adae6a3d119a/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20190108225652-1e06a53dbb7e/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
@@ -176,8 +180,8 @@ golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLL
 golang.org/x/net v0.0.0-20200625001655-4c5254603344/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
 golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
-golang.org/x/net v0.39.0 h1:ZCu7HMWDxpXpaiKdhzIfaltL9Lp31x/3fCP11bc6/fY=
+golang.org/x/net v0.45.0 h1:RLBg5JKixCy82FtLJpeNlVM0nrSqpCRYzVU1n8kj0tM=
-golang.org/x/net v0.39.0/go.mod h1:X7NRbYVEA+ewNkCNyJ513WmMdQ3BineSwVtN2zD/d+E=
+golang.org/x/net v0.45.0/go.mod h1:ECOoLqd5U3Lhyeyo/QDCEVQ4sNgYsqvCZ722XogGieY=
 golang.org/x/oauth2 v0.0.0-20190226205417-e64efc72b421/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -185,8 +189,8 @@ golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201207232520-09787c993a3a/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
-golang.org/x/sync v0.13.0 h1:AauUjRAJ9OSnvULf/ARrrVywoJDy0YS2AwQ98I37610=
+golang.org/x/sync v0.17.0 h1:l60nONMj9l5drqw6jlhIELNv9I0A4OFgRsG9k2oT9Ug=
-golang.org/x/sync v0.13.0/go.mod h1:1dzgHSNfp02xaA81J2MS99Qcpr2w7fw1gpm99rleRqA=
+golang.org/x/sync v0.17.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
 golang.org/x/sys v0.0.0-20180905080454-ebe1bf3edb33/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20181116152217-5ac8a444bdc5/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -197,18 +201,17 @@ golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20200615200032-f1bc736245b1/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200625212154-ddb9806d33ae/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20201015000850-e3ed0017c211/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210124154548-22da62e12c0c/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210603081109-ebe580a85c40/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.2.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.10.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.32.0 h1:s77OFDvIQeibCmezSnk/q6iAfkdiQaJi4VzroCFrN20=
+golang.org/x/sys v0.37.0 h1:fdNQudmxPjkdUTPnLn5mdQv7Zwvbvpaxqs831goi9kQ=
-golang.org/x/sys v0.32.0/go.mod h1:BJP2sWEmIv4KK5OTEluFJCKSidICx8ciO85XgH3Ak8k=
+golang.org/x/sys v0.37.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
-golang.org/x/term v0.31.0 h1:erwDkOK1Msy6offm1mOgvspSkslFnIGsFnxOKoufg3o=
+golang.org/x/term v0.36.0 h1:zMPR+aF8gfksFprF/Nc/rd1wRS1EI6nDBGyWAvDzx2Q=
-golang.org/x/term v0.31.0/go.mod h1:R4BeIy7D95HzImkxGkTW1UQTtP54tio2RyHz7PwK0aw=
+golang.org/x/term v0.36.0/go.mod h1:Qu394IJq6V6dCBRgwqshf3mPF85AqzYEzofzRdZkWss=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
@@ -219,8 +222,8 @@ golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtn
 golang.org/x/tools v0.0.0-20200130002326-2f3ba24bd6e7/go.mod h1:TB2adYChydJhpapKDTa4BR/hXlZSLoq2Wpct/0txZ28=
 golang.org/x/tools v0.0.0-20200619180055-7c47624df98f/go.mod h1:EkVYQZoAsY45+roYkvgYkIh4xh/qjgUK9TdY2XT94GE=
 golang.org/x/tools v0.0.0-20210106214847-113979e3529a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
-golang.org/x/tools v0.30.0 h1:BgcpHewrV5AUp2G9MebG4XPFI1E2W41zU1SaqVA9vJY=
+golang.org/x/tools v0.33.0 h1:4qz2S3zmRxbGIhDIAgjxvFutSvH5EfnsYrRBj0UI0bc=
-golang.org/x/tools v0.30.0/go.mod h1:c347cR/OJfw5TI+GfX7RUPNMdDRRbjvYTS0jPyvsVtY=
+golang.org/x/tools v0.33.0/go.mod h1:CIJMaWEY88juyUfo7UbgPqbC8rU2OqfAV1h2Qp0oMYI=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
@@ -239,8 +242,8 @@ google.golang.org/protobuf v1.20.1-0.20200309200217-e05f789c0967/go.mod h1:A+miE
 google.golang.org/protobuf v1.21.0/go.mod h1:47Nbq4nVaFHyn7ilMalzfO3qCViNmqZ2kzikPIcrTAo=
 google.golang.org/protobuf v1.23.0/go.mod h1:EGpADcykh3NcUnDUJcl1+ZksZNG86OlYog2l/sGQquU=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
-google.golang.org/protobuf v1.36.6 h1:z1NpPI8ku2WgiWnf+t9wTPsn6eP1L7ksHUlkfLvd9xY=
+google.golang.org/protobuf v1.36.8 h1:xHScyCOEuuwZEc6UtSOvPbAT4zRh0xcNRYekJwfqyMc=
-google.golang.org/protobuf v1.36.6/go.mod h1:jduwjTPXsFjZGTmRluh+L6NjiWu7pchiJ2/5YcXBHnY=
+google.golang.org/protobuf v1.36.8/go.mod h1:fuxRtAxBytpl4zzqUh6/eyUujkJdNiuEkXntxiD/uRU=
 gopkg.in/alecthomas/kingpin.v2 v2.2.6/go.mod h1:FMv+mEhP44yOT+4EoQTLFTRgOQ1FBLkstjWtayDeSgw=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20190902080502-41f04d3bba15/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

handshake_ix.go

@@ -249,7 +249,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		HandshakePacket:   make(map[uint8][]byte, 0),
 		lastHandshakeTime: hs.Details.Time,
 		relayState: RelayState{
-			relays:         map[netip.Addr]struct{}{},
+			relays:         nil,
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},
@@ -457,9 +457,9 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			Info("Handshake message sent")
 	}
 
-	f.connectionManager.AddTrafficWatch(hostinfo.localIndexId)
+	f.connectionManager.AddTrafficWatch(hostinfo)
 
-	hostinfo.remotes.ResetBlockedRemotes()
+	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
 
 	return
 }
@@ -652,7 +652,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 
 	// Complete our handshake and update metrics, this will replace any existing tunnels for the vpnAddrs here
 	f.handshakeManager.Complete(hostinfo, f)
-	f.connectionManager.AddTrafficWatch(hostinfo.localIndexId)
+	f.connectionManager.AddTrafficWatch(hostinfo)
 
 	if f.l.Level >= logrus.DebugLevel {
 		hostinfo.logger(f.l).Debugf("Sending %d stored packets", len(hh.packetStore))
@@ -667,7 +667,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 		f.cachedPacketMetrics.sent.Inc(int64(len(hh.packetStore)))
 	}
 
-	hostinfo.remotes.ResetBlockedRemotes()
+	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
 	f.metricHandshakes.Update(duration)
 
 	return false

handshake_manager.go

@@ -450,7 +450,7 @@ func (hm *HandshakeManager) StartHandshake(vpnAddr netip.Addr, cacheCb func(*Han
 		vpnAddrs:        []netip.Addr{vpnAddr},
 		HandshakePacket: make(map[uint8][]byte, 0),
 		relayState: RelayState{
-			relays:         map[netip.Addr]struct{}{},
+			relays:         nil,
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},

hostmap.go

@@ -4,6 +4,7 @@ import (
 	"errors"
 	"net"
 	"net/netip"
+	"slices"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -16,12 +17,10 @@ import (
 	"github.com/slackhq/nebula/header"
 )
 
-// const ProbeLen = 100
 const defaultPromoteEvery = 1000       // Count of packets sent before we try moving a tunnel to a preferred underlay ip address
 const defaultReQueryEvery = 5000       // Count of packets sent before re-querying a hostinfo to the lighthouse
 const defaultReQueryWait = time.Minute // Minimum amount of seconds to wait before re-querying a hostinfo the lighthouse. Evaluated every ReQueryEvery
 const MaxRemotes = 10
-const maxRecvError = 4
 
 // MaxHostInfosPerVpnIp is the max number of hostinfos we will track for a given vpn ip
 // 5 allows for an initial handshake and each host pair re-handshaking twice
@@ -68,7 +67,7 @@ type HostMap struct {
 type RelayState struct {
 	sync.RWMutex
 
-	relays map[netip.Addr]struct{} // Set of vpnAddr's of Hosts to use as relays to access this peer
+	relays []netip.Addr // Ordered set of VpnAddrs of Hosts to use as relays to access this peer
 	// For data race avoidance, the contents of a *Relay are treated immutably. To update a *Relay, copy the existing data,
 	// modify what needs to be updated, and store the new modified copy in the relayForByIp and relayForByIdx maps (with
 	// the RelayState Lock held)
@@ -79,7 +78,12 @@ type RelayState struct {
 func (rs *RelayState) DeleteRelay(ip netip.Addr) {
 	rs.Lock()
 	defer rs.Unlock()
-	delete(rs.relays, ip)
+	for idx, val := range rs.relays {
+		if val == ip {
+			rs.relays = append(rs.relays[:idx], rs.relays[idx+1:]...)
+			return
+		}
+	}
 }
 
 func (rs *RelayState) UpdateRelayForByIpState(vpnIp netip.Addr, state int) {
@@ -124,16 +128,16 @@ func (rs *RelayState) GetRelayForByAddr(addr netip.Addr) (*Relay, bool) {
 func (rs *RelayState) InsertRelayTo(ip netip.Addr) {
 	rs.Lock()
 	defer rs.Unlock()
-	rs.relays[ip] = struct{}{}
+	if !slices.Contains(rs.relays, ip) {
+		rs.relays = append(rs.relays, ip)
+	}
 }
 
 func (rs *RelayState) CopyRelayIps() []netip.Addr {
+	ret := make([]netip.Addr, len(rs.relays))
 	rs.RLock()
 	defer rs.RUnlock()
-	ret := make([]netip.Addr, 0, len(rs.relays))
+	copy(ret, rs.relays)
-	for ip := range rs.relays {
-		ret = append(ret, ip)
-	}
 	return ret
 }
 
@@ -219,8 +223,7 @@ type HostInfo struct {
 	// vpnAddrs is a list of vpn addresses assigned to this host that are within our own vpn networks
 	// The host may have other vpn addresses that are outside our
 	// vpn networks but were removed because they are not usable
-	vpnAddrs  []netip.Addr
+	vpnAddrs []netip.Addr
-	recvError atomic.Uint32
 
 	// networks are both all vpn and unsafe networks assigned to this host
 	networks   *bart.Lite
@@ -250,6 +253,14 @@ type HostInfo struct {
 	// Used to track other hostinfos for this vpn ip since only 1 can be primary
 	// Synchronised via hostmap lock and not the hostinfo lock.
 	next, prev *HostInfo
+
+	//TODO: in, out, and others might benefit from being an atomic.Int32. We could collapse connectionManager pendingDeletion, relayUsed, and in/out into this 1 thing
+	in, out, pendingDeletion atomic.Bool
+
+	// lastUsed tracks the last time ConnectionManager checked the tunnel and it was in use.
+	// This value will be behind against actual tunnel utilization in the hot path.
+	// This should only be used by the ConnectionManagers ticker routine.
+	lastUsed time.Time
 }
 
 type ViaSender struct {
@@ -719,13 +730,6 @@ func (i *HostInfo) SetRemoteIfPreferred(hm *HostMap, newRemote netip.AddrPort) b
 	return false
 }
 
-func (i *HostInfo) RecvErrorExceeded() bool {
-	if i.recvError.Add(1) >= maxRecvError {
-		return true
-	}
-	return true
-}
-
 func (i *HostInfo) buildNetworks(networks, unsafeNetworks []netip.Prefix) {
 	if len(networks) == 1 && len(unsafeNetworks) == 0 {
 		// Simple case, no CIDRTree needed
@@ -734,7 +738,8 @@ func (i *HostInfo) buildNetworks(networks, unsafeNetworks []netip.Prefix) {
 
 	i.networks = new(bart.Lite)
 	for _, network := range networks {
-		i.networks.Insert(network)
+		nprefix := netip.PrefixFrom(network.Addr(), network.Addr().BitLen())
+		i.networks.Insert(nprefix)
 	}
 
 	for _, network := range unsafeNetworks {

hostmap_test.go

@@ -7,6 +7,7 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestHostMap_MakePrimary(t *testing.T) {
@@ -215,3 +216,31 @@ func TestHostMap_reload(t *testing.T) {
 	c.ReloadConfigString("preferred_ranges: [1.1.1.1/32]")
 	assert.Equal(t, []string{"1.1.1.1/32"}, toS(hm.GetPreferredRanges()))
 }
+
+func TestHostMap_RelayState(t *testing.T) {
+	h1 := &HostInfo{vpnAddrs: []netip.Addr{netip.MustParseAddr("0.0.0.1")}, localIndexId: 1}
+	a1 := netip.MustParseAddr("::1")
+	a2 := netip.MustParseAddr("2001::1")
+
+	h1.relayState.InsertRelayTo(a1)
+	assert.Equal(t, []netip.Addr{a1}, h1.relayState.relays)
+	h1.relayState.InsertRelayTo(a2)
+	assert.Equal(t, []netip.Addr{a1, a2}, h1.relayState.relays)
+	// Ensure that the first relay added is the first one returned in the copy
+	currentRelays := h1.relayState.CopyRelayIps()
+	require.Len(t, currentRelays, 2)
+	assert.Equal(t, a1, currentRelays[0])
+
+	// Deleting the last one in the list works ok
+	h1.relayState.DeleteRelay(a2)
+	assert.Equal(t, []netip.Addr{a1}, h1.relayState.relays)
+
+	// Deleting an element not in the list works ok
+	h1.relayState.DeleteRelay(a2)
+	assert.Equal(t, []netip.Addr{a1}, h1.relayState.relays)
+
+	// Deleting the only element in the list works ok
+	h1.relayState.DeleteRelay(a1)
+	assert.Equal(t, []netip.Addr{}, h1.relayState.relays)
+
+}

inside.go

@@ -288,7 +288,7 @@ func (f *Interface) SendVia(via *HostInfo,
 	c := via.ConnectionState.messageCounter.Add(1)
 
 	out = header.Encode(out, header.Version, header.Message, header.MessageRelay, relay.RemoteIndex, c)
-	f.connectionManager.Out(via.localIndexId)
+	f.connectionManager.Out(via)
 
 	// Authenticate the header and payload, but do not encrypt for this message type.
 	// The payload consists of the inner, unencrypted Nebula header, as well as the end-to-end encrypted payload.
@@ -356,7 +356,7 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 
 	//l.WithField("trace", string(debug.Stack())).Error("out Header ", &Header{Version, t, st, 0, hostinfo.remoteIndexId, c}, p)
 	out = header.Encode(out, header.Version, t, st, hostinfo.remoteIndexId, c)
-	f.connectionManager.Out(hostinfo.localIndexId)
+	f.connectionManager.Out(hostinfo)
 
 	// Query our LH if we haven't since the last time we've been rebound, this will cause the remote to punch against
 	// all our addrs and enable a faster roaming.

interface.go

@@ -24,23 +24,23 @@ import (
 const mtu = 9001
 
 type InterfaceConfig struct {
-	HostMap                 *HostMap
+	HostMap            *HostMap
-	Outside                 udp.Conn
+	Outside            udp.Conn
-	Inside                  overlay.Device
+	Inside             overlay.Device
-	pki                     *PKI
+	pki                *PKI
-	Firewall                *Firewall
+	Cipher             string
-	ServeDns                bool
+	Firewall           *Firewall
-	HandshakeManager        *HandshakeManager
+	ServeDns           bool
-	lightHouse              *LightHouse
+	HandshakeManager   *HandshakeManager
-	checkInterval           time.Duration
+	lightHouse         *LightHouse
-	pendingDeletionInterval time.Duration
+	connectionManager  *connectionManager
-	DropLocalBroadcast      bool
+	DropLocalBroadcast bool
-	DropMulticast           bool
+	DropMulticast      bool
-	routines                int
+	routines           int
-	MessageMetrics          *MessageMetrics
+	MessageMetrics     *MessageMetrics
-	version                 string
+	version            string
-	relayManager            *relayManager
+	relayManager       *relayManager
-	punchy                  *Punchy
+	punchy             *Punchy
 
 	tryPromoteEvery uint32
 	reQueryEvery    uint32
@@ -157,6 +157,9 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	if c.Firewall == nil {
 		return nil, errors.New("no firewall rules")
 	}
+	if c.connectionManager == nil {
+		return nil, errors.New("no connection manager")
+	}
 
 	cs := c.pki.getCertState()
 	ifce := &Interface{
@@ -181,7 +184,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 		myVpnAddrsTable:       cs.myVpnAddrsTable,
 		myBroadcastAddrsTable: cs.myVpnBroadcastAddrsTable,
 		relayManager:          c.relayManager,
-
+		connectionManager:     c.connectionManager,
 		conntrackCacheTimeout: c.ConntrackCacheTimeout,
 
 		metricHandshakes: metrics.GetOrRegisterHistogram("handshakes", nil, metrics.NewExpDecaySample(1028, 0.015)),
@@ -198,7 +201,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	ifce.reQueryEvery.Store(c.reQueryEvery)
 	ifce.reQueryWait.Store(int64(c.reQueryWait))
 
-	ifce.connectionManager = newConnectionManager(ctx, c.l, ifce, c.checkInterval, c.pendingDeletionInterval, c.punchy)
+	ifce.connectionManager.intf = ifce
 
 	return ifce, nil
 }

lighthouse.go

@@ -24,6 +24,7 @@ import (
 )
 
 var ErrHostNotKnown = errors.New("host not known")
+var ErrBadDetailsVpnAddr = errors.New("invalid packet, malformed detailsVpnAddr")
 
 type LightHouse struct {
 	//TODO: We need a timer wheel to kick out vpnAddrs that haven't reported in a long time
@@ -56,7 +57,7 @@ type LightHouse struct {
 	// staticList exists to avoid having a bool in each addrMap entry
 	// since static should be rare
 	staticList  atomic.Pointer[map[netip.Addr]struct{}]
-	lighthouses atomic.Pointer[map[netip.Addr]struct{}]
+	lighthouses atomic.Pointer[[]netip.Addr]
 
 	interval     atomic.Int64
 	updateCancel context.CancelFunc
@@ -107,7 +108,7 @@ func NewLightHouseFromConfig(ctx context.Context, l *logrus.Logger, c *config.C,
 		queryChan:          make(chan netip.Addr, c.GetUint32("handshakes.query_buffer", 64)),
 		l:                  l,
 	}
-	lighthouses := make(map[netip.Addr]struct{})
+	lighthouses := make([]netip.Addr, 0)
 	h.lighthouses.Store(&lighthouses)
 	staticList := make(map[netip.Addr]struct{})
 	h.staticList.Store(&staticList)
@@ -143,7 +144,7 @@ func (lh *LightHouse) GetStaticHostList() map[netip.Addr]struct{} {
 	return *lh.staticList.Load()
 }
 
-func (lh *LightHouse) GetLighthouses() map[netip.Addr]struct{} {
+func (lh *LightHouse) GetLighthouses() []netip.Addr {
 	return *lh.lighthouses.Load()
 }
 
@@ -306,13 +307,12 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 	}
 
 	if initial || c.HasChanged("lighthouse.hosts") {
-		lhMap := make(map[netip.Addr]struct{})
+		lhList, err := lh.parseLighthouses(c)
-		err := lh.parseLighthouses(c, lhMap)
 		if err != nil {
 			return err
 		}
 
-		lh.lighthouses.Store(&lhMap)
+		lh.lighthouses.Store(&lhList)
 		if !initial {
 			//NOTE: we are not tearing down existing lighthouse connections because they might be used for non lighthouse traffic
 			lh.l.Info("lighthouse.hosts has changed")
@@ -346,36 +346,37 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (lh *LightHouse) parseLighthouses(c *config.C, lhMap map[netip.Addr]struct{}) error {
+func (lh *LightHouse) parseLighthouses(c *config.C) ([]netip.Addr, error) {
 	lhs := c.GetStringSlice("lighthouse.hosts", []string{})
 	if lh.amLighthouse && len(lhs) != 0 {
 		lh.l.Warn("lighthouse.am_lighthouse enabled on node but upstream lighthouses exist in config")
 	}
+	out := make([]netip.Addr, len(lhs))
 
 	for i, host := range lhs {
 		addr, err := netip.ParseAddr(host)
 		if err != nil {
-			return util.NewContextualError("Unable to parse lighthouse host entry", m{"host": host, "entry": i + 1}, err)
+			return nil, util.NewContextualError("Unable to parse lighthouse host entry", m{"host": host, "entry": i + 1}, err)
 		}
 
 		if !lh.myVpnNetworksTable.Contains(addr) {
-			return util.NewContextualError("lighthouse host is not in our networks, invalid", m{"vpnAddr": addr, "networks": lh.myVpnNetworks}, nil)
+			return nil, util.NewContextualError("lighthouse host is not in our networks, invalid", m{"vpnAddr": addr, "networks": lh.myVpnNetworks}, nil)
 		}
-		lhMap[addr] = struct{}{}
+		out[i] = addr
 	}
 
-	if !lh.amLighthouse && len(lhMap) == 0 {
+	if !lh.amLighthouse && len(out) == 0 {
 		lh.l.Warn("No lighthouse.hosts configured, this host will only be able to initiate tunnels with static_host_map entries")
 	}
 
 	staticList := lh.GetStaticHostList()
-	for lhAddr, _ := range lhMap {
+	for i := range out {
-		if _, ok := staticList[lhAddr]; !ok {
+		if _, ok := staticList[out[i]]; !ok {
-			return fmt.Errorf("lighthouse %s does not have a static_host_map entry", lhAddr)
+			return nil, fmt.Errorf("lighthouse %s does not have a static_host_map entry", out[i])
 		}
 	}
 
-	return nil
+	return out, nil
 }
 
 func getStaticMapCadence(c *config.C) (time.Duration, error) {
@@ -486,7 +487,7 @@ func (lh *LightHouse) QueryCache(vpnAddrs []netip.Addr) *RemoteList {
 	lh.Lock()
 	defer lh.Unlock()
 	// Add an entry if we don't already have one
-	return lh.unlockedGetRemoteList(vpnAddrs)
+	return lh.unlockedGetRemoteList(vpnAddrs) //todo CERT-V2 this contains addrmap lookups we could potentially skip
 }
 
 // queryAndPrepMessage is a lock helper on RemoteList, assisting the caller to build a lighthouse message containing
@@ -519,11 +520,15 @@ func (lh *LightHouse) queryAndPrepMessage(vpnAddr netip.Addr, f func(*cache) (in
 }
 
 func (lh *LightHouse) DeleteVpnAddrs(allVpnAddrs []netip.Addr) {
-	// First we check the static mapping
+	// First we check the static host map. If any of the VpnAddrs to be deleted are present, do nothing.
-	// and do nothing if it is there
+	staticList := lh.GetStaticHostList()
-	if _, ok := lh.GetStaticHostList()[allVpnAddrs[0]]; ok {
+	for _, addr := range allVpnAddrs {
-		return
+		if _, ok := staticList[addr]; ok {
+			return
+		}
 	}
+
+	// None of the VpnAddrs were present. Now we can do the deletes.
 	lh.Lock()
 	rm, ok := lh.addrMap[allVpnAddrs[0]]
 	if ok {
@@ -565,7 +570,7 @@ func (lh *LightHouse) addStaticRemotes(i int, d time.Duration, network string, t
 	am.unlockedSetHostnamesResults(hr)
 
 	for _, addrPort := range hr.GetAddrs() {
-		if !lh.shouldAdd(vpnAddr, addrPort.Addr()) {
+		if !lh.shouldAdd([]netip.Addr{vpnAddr}, addrPort.Addr()) {
 			continue
 		}
 		switch {
@@ -627,23 +632,30 @@ func (lh *LightHouse) addCalculatedRemotes(vpnAddr netip.Addr) bool {
 	return len(calculatedV4) > 0 || len(calculatedV6) > 0
 }
 
-// unlockedGetRemoteList
+// unlockedGetRemoteList assumes you have the lh lock
-// assumes you have the lh lock
 func (lh *LightHouse) unlockedGetRemoteList(allAddrs []netip.Addr) *RemoteList {
-	am, ok := lh.addrMap[allAddrs[0]]
+	// before we go and make a new remotelist, we need to make sure we don't have one for any of this set of vpnaddrs yet
-	if !ok {
+	for i, addr := range allAddrs {
-		am = NewRemoteList(allAddrs, func(a netip.Addr) bool { return lh.shouldAdd(allAddrs[0], a) })
+		am, ok := lh.addrMap[addr]
-		for _, addr := range allAddrs {
+		if ok {
-			lh.addrMap[addr] = am
+			if i != 0 {
+				lh.addrMap[allAddrs[0]] = am
+			}
+			return am
 		}
 	}
+
+	am := NewRemoteList(allAddrs, lh.shouldAdd)
+	for _, addr := range allAddrs {
+		lh.addrMap[addr] = am
+	}
 	return am
 }
 
-func (lh *LightHouse) shouldAdd(vpnAddr netip.Addr, to netip.Addr) bool {
+func (lh *LightHouse) shouldAdd(vpnAddrs []netip.Addr, to netip.Addr) bool {
-	allow := lh.GetRemoteAllowList().Allow(vpnAddr, to)
+	allow := lh.GetRemoteAllowList().AllowAll(vpnAddrs, to)
 	if lh.l.Level >= logrus.TraceLevel {
-		lh.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", to).WithField("allow", allow).
+		lh.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", to).WithField("allow", allow).
 			Trace("remoteAllowList.Allow")
 	}
 	if !allow {
@@ -698,19 +710,22 @@ func (lh *LightHouse) unlockedShouldAddV6(vpnAddr netip.Addr, to *V6AddrPort) bo
 }
 
 func (lh *LightHouse) IsLighthouseAddr(vpnAddr netip.Addr) bool {
-	if _, ok := lh.GetLighthouses()[vpnAddr]; ok {
+	l := lh.GetLighthouses()
-		return true
+	for i := range l {
+		if l[i] == vpnAddr {
+			return true
+		}
 	}
 	return false
 }
 
-// TODO: CERT-V2 IsLighthouseAddr should be sufficient, we just need to update the vpnAddrs for lighthouses after a handshake
+func (lh *LightHouse) IsAnyLighthouseAddr(vpnAddrs []netip.Addr) bool {
-// so that we know all the lighthouse vpnAddrs, not just the ones we were configured to talk to initially
-func (lh *LightHouse) IsAnyLighthouseAddr(vpnAddr []netip.Addr) bool {
 	l := lh.GetLighthouses()
-	for _, a := range vpnAddr {
+	for i := range vpnAddrs {
-		if _, ok := l[a]; ok {
+		for j := range l {
-			return true
+			if l[j] == vpnAddrs[i] {
+				return true
+			}
 		}
 	}
 	return false
@@ -752,7 +767,7 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 	queried := 0
 	lighthouses := lh.GetLighthouses()
 
-	for lhVpnAddr := range lighthouses {
+	for _, lhVpnAddr := range lighthouses {
 		hi := lh.ifce.GetHostInfo(lhVpnAddr)
 		if hi != nil {
 			v = hi.ConnectionState.myCert.Version()
@@ -870,7 +885,7 @@ func (lh *LightHouse) SendUpdate() {
 	updated := 0
 	lighthouses := lh.GetLighthouses()
 
-	for lhVpnAddr := range lighthouses {
+	for _, lhVpnAddr := range lighthouses {
 		var v cert.Version
 		hi := lh.ifce.GetHostInfo(lhVpnAddr)
 		if hi != nil {
@@ -928,7 +943,6 @@ func (lh *LightHouse) SendUpdate() {
 						V4AddrPorts:   v4,
 						V6AddrPorts:   v6,
 						RelayVpnAddrs: relays,
-						VpnAddr:       netAddrToProtoAddr(lh.myVpnNetworks[0].Addr()),
 					},
 				}
 
@@ -1048,19 +1062,19 @@ func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []neti
 		return
 	}
 
-	useVersion := cert.Version1
+	queryVpnAddr, useVersion, err := n.Details.GetVpnAddrAndVersion()
-	var queryVpnAddr netip.Addr
+	if err != nil {
-	if n.Details.OldVpnAddr != 0 {
-		b := [4]byte{}
-		binary.BigEndian.PutUint32(b[:], n.Details.OldVpnAddr)
-		queryVpnAddr = netip.AddrFrom4(b)
-		useVersion = 1
-	} else if n.Details.VpnAddr != nil {
-		queryVpnAddr = protoAddrToNetAddr(n.Details.VpnAddr)
-		useVersion = 2
-	} else {
 		if lhh.l.Level >= logrus.DebugLevel {
-			lhh.l.WithField("from", fromVpnAddrs).WithField("details", n.Details).Debugln("Dropping malformed HostQuery")
+			lhh.l.WithField("from", fromVpnAddrs).WithField("details", n.Details).
+				Debugln("Dropping malformed HostQuery")
+		}
+		return
+	}
+	if useVersion == cert.Version1 && queryVpnAddr.Is6() {
+		// this case really shouldn't be possible to represent, but reject it anyway.
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("queryVpnAddr", queryVpnAddr).
+				Debugln("invalid vpn addr for v1 handleHostQuery")
 		}
 		return
 	}
@@ -1069,9 +1083,6 @@ func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []neti
 		n = lhh.resetMeta()
 		n.Type = NebulaMeta_HostQueryReply
 		if useVersion == cert.Version1 {
-			if !queryVpnAddr.Is4() {
-				return 0, fmt.Errorf("invalid vpn addr for v1 handleHostQuery")
-			}
 			b := queryVpnAddr.As4()
 			n.Details.OldVpnAddr = binary.BigEndian.Uint32(b[:])
 		} else {
@@ -1116,8 +1127,9 @@ func (lhh *LightHouseHandler) sendHostPunchNotification(n *NebulaMeta, fromVpnAd
 			if ok {
 				whereToPunch = newDest
 			} else {
-				//TODO: CERT-V2 this means the destination will have no addresses in common with the punch-ee
+				if lhh.l.Level >= logrus.DebugLevel {
-				//choosing to do nothing for now, but maybe we return an error?
+					lhh.l.WithField("to", crt.Networks()).Debugln("unable to punch to host, no addresses in common")
+				}
 			}
 		}
 
@@ -1176,19 +1188,17 @@ func (lhh *LightHouseHandler) coalesceAnswers(v cert.Version, c *cache, n *Nebul
 				if !r.Is4() {
 					continue
 				}
-
 				b = r.As4()
 				n.Details.OldRelayVpnAddrs = append(n.Details.OldRelayVpnAddrs, binary.BigEndian.Uint32(b[:]))
 			}
-
 		} else if v == cert.Version2 {
 			for _, r := range c.relay.relay {
 				n.Details.RelayVpnAddrs = append(n.Details.RelayVpnAddrs, netAddrToProtoAddr(r))
 			}
-
 		} else {
-			//TODO: CERT-V2 don't panic
+			if lhh.l.Level >= logrus.DebugLevel {
-			panic("unsupported version")
+				lhh.l.WithField("version", v).Debug("unsupported protocol version")
+			}
 		}
 	}
 }
@@ -1198,18 +1208,16 @@ func (lhh *LightHouseHandler) handleHostQueryReply(n *NebulaMeta, fromVpnAddrs [
 		return
 	}
 
-	lhh.lh.Lock()
+	certVpnAddr, _, err := n.Details.GetVpnAddrAndVersion()
-
+	if err != nil {
-	var certVpnAddr netip.Addr
+		if lhh.l.Level >= logrus.DebugLevel {
-	if n.Details.OldVpnAddr != 0 {
+			lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("dropping malformed HostQueryReply")
-		b := [4]byte{}
+		}
-		binary.BigEndian.PutUint32(b[:], n.Details.OldVpnAddr)
+		return
-		certVpnAddr = netip.AddrFrom4(b)
-	} else if n.Details.VpnAddr != nil {
-		certVpnAddr = protoAddrToNetAddr(n.Details.VpnAddr)
 	}
 	relays := n.Details.GetRelays()
 
+	lhh.lh.Lock()
 	am := lhh.lh.unlockedGetRemoteList([]netip.Addr{certVpnAddr})
 	am.Lock()
 	lhh.lh.Unlock()
@@ -1234,27 +1242,24 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 		return
 	}
 
+	// not using GetVpnAddrAndVersion because we don't want to error on a blank detailsVpnAddr
 	var detailsVpnAddr netip.Addr
-	useVersion := cert.Version1
+	var useVersion cert.Version
-	if n.Details.OldVpnAddr != 0 {
+	if n.Details.OldVpnAddr != 0 { //v1 always sets this field
 		b := [4]byte{}
 		binary.BigEndian.PutUint32(b[:], n.Details.OldVpnAddr)
 		detailsVpnAddr = netip.AddrFrom4(b)
 		useVersion = cert.Version1
-	} else if n.Details.VpnAddr != nil {
+	} else if n.Details.VpnAddr != nil { //this field is "optional" in v2, but if it's set, we should enforce it
 		detailsVpnAddr = protoAddrToNetAddr(n.Details.VpnAddr)
 		useVersion = cert.Version2
 	} else {
-		if lhh.l.Level >= logrus.DebugLevel {
+		detailsVpnAddr = netip.Addr{}
-			lhh.l.WithField("details", n.Details).Debugf("dropping invalid HostUpdateNotification")
+		useVersion = cert.Version2
-		}
-		return
 	}
 
-	//TODO: CERT-V2 hosts with only v2 certs cannot provide their ipv6 addr when contacting the lighthouse via v4?
+	//Simple check that the host sent this not someone else, if detailsVpnAddr is filled
-	//TODO: CERT-V2 why do we care about the vpnAddr in the packet? We know where it came from, right?
+	if detailsVpnAddr.IsValid() && !slices.Contains(fromVpnAddrs, detailsVpnAddr) {
-	//Simple check that the host sent this not someone else
-	if !slices.Contains(fromVpnAddrs, detailsVpnAddr) {
 		if lhh.l.Level >= logrus.DebugLevel {
 			lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("answer", detailsVpnAddr).Debugln("Host sent invalid update")
 		}
@@ -1268,24 +1273,24 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 	am.Lock()
 	lhh.lh.Unlock()
 
-	am.unlockedSetV4(fromVpnAddrs[0], detailsVpnAddr, n.Details.V4AddrPorts, lhh.lh.unlockedShouldAddV4)
+	am.unlockedSetV4(fromVpnAddrs[0], fromVpnAddrs[0], n.Details.V4AddrPorts, lhh.lh.unlockedShouldAddV4)
-	am.unlockedSetV6(fromVpnAddrs[0], detailsVpnAddr, n.Details.V6AddrPorts, lhh.lh.unlockedShouldAddV6)
+	am.unlockedSetV6(fromVpnAddrs[0], fromVpnAddrs[0], n.Details.V6AddrPorts, lhh.lh.unlockedShouldAddV6)
 	am.unlockedSetRelay(fromVpnAddrs[0], relays)
 	am.Unlock()
 
 	n = lhh.resetMeta()
 	n.Type = NebulaMeta_HostUpdateNotificationAck
-
+	switch useVersion {
-	if useVersion == cert.Version1 {
+	case cert.Version1:
 		if !fromVpnAddrs[0].Is4() {
 			lhh.l.WithField("vpnAddrs", fromVpnAddrs).Error("Can not send HostUpdateNotificationAck for a ipv6 vpn ip in a v1 message")
 			return
 		}
 		vpnAddrB := fromVpnAddrs[0].As4()
 		n.Details.OldVpnAddr = binary.BigEndian.Uint32(vpnAddrB[:])
-	} else if useVersion == cert.Version2 {
+	case cert.Version2:
-		n.Details.VpnAddr = netAddrToProtoAddr(fromVpnAddrs[0])
+		// do nothing, we want to send a blank message
-	} else {
+	default:
 		lhh.l.WithField("useVersion", useVersion).Error("invalid protocol version")
 		return
 	}
@@ -1303,13 +1308,20 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpnAddrs []netip.Addr, w EncWriter) {
 	//It's possible the lighthouse is communicating with us using a non primary vpn addr,
 	//which means we need to compare all fromVpnAddrs against all configured lighthouse vpn addrs.
-	//maybe one day we'll have a better idea, if it matters.
 	if !lhh.lh.IsAnyLighthouseAddr(fromVpnAddrs) {
 		return
 	}
 
+	detailsVpnAddr, _, err := n.Details.GetVpnAddrAndVersion()
+	if err != nil {
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithField("details", n.Details).WithError(err).Debugln("dropping invalid HostPunchNotification")
+		}
+		return
+	}
+
 	empty := []byte{0}
-	punch := func(vpnPeer netip.AddrPort) {
+	punch := func(vpnPeer netip.AddrPort, logVpnAddr netip.Addr) {
 		if !vpnPeer.IsValid() {
 			return
 		}
@@ -1321,48 +1333,31 @@ func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpn
 		}()
 
 		if lhh.l.Level >= logrus.DebugLevel {
-			var logVpnAddr netip.Addr
-			if n.Details.OldVpnAddr != 0 {
-				b := [4]byte{}
-				binary.BigEndian.PutUint32(b[:], n.Details.OldVpnAddr)
-				logVpnAddr = netip.AddrFrom4(b)
-			} else if n.Details.VpnAddr != nil {
-				logVpnAddr = protoAddrToNetAddr(n.Details.VpnAddr)
-			}
 			lhh.l.Debugf("Punching on %v for %v", vpnPeer, logVpnAddr)
 		}
 	}
 
 	for _, a := range n.Details.V4AddrPorts {
-		punch(protoV4AddrPortToNetAddrPort(a))
+		punch(protoV4AddrPortToNetAddrPort(a), detailsVpnAddr)
 	}
 
 	for _, a := range n.Details.V6AddrPorts {
-		punch(protoV6AddrPortToNetAddrPort(a))
+		punch(protoV6AddrPortToNetAddrPort(a), detailsVpnAddr)
 	}
 
 	// This sends a nebula test packet to the host trying to contact us. In the case
 	// of a double nat or other difficult scenario, this may help establish
 	// a tunnel.
 	if lhh.lh.punchy.GetRespond() {
-		var queryVpnAddr netip.Addr
-		if n.Details.OldVpnAddr != 0 {
-			b := [4]byte{}
-			binary.BigEndian.PutUint32(b[:], n.Details.OldVpnAddr)
-			queryVpnAddr = netip.AddrFrom4(b)
-		} else if n.Details.VpnAddr != nil {
-			queryVpnAddr = protoAddrToNetAddr(n.Details.VpnAddr)
-		}
-
 		go func() {
 			time.Sleep(lhh.lh.punchy.GetRespondDelay())
 			if lhh.l.Level >= logrus.DebugLevel {
-				lhh.l.Debugf("Sending a nebula test packet to vpn addr %s", queryVpnAddr)
+				lhh.l.Debugf("Sending a nebula test packet to vpn addr %s", detailsVpnAddr)
 			}
 			//NOTE: we have to allocate a new output buffer here since we are spawning a new goroutine
 			// for each punchBack packet. We should move this into a timerwheel or a single goroutine
 			// managed by a channel.
-			w.SendMessageToVpnAddr(header.Test, header.TestRequest, queryVpnAddr, []byte(""), make([]byte, 12, 12), make([]byte, mtu))
+			w.SendMessageToVpnAddr(header.Test, header.TestRequest, detailsVpnAddr, []byte(""), make([]byte, 12, 12), make([]byte, mtu))
 		}()
 	}
 }
@@ -1441,3 +1436,17 @@ func findNetworkUnion(prefixes []netip.Prefix, addrs []netip.Addr) (netip.Addr,
 	}
 	return netip.Addr{}, false
 }
+
+func (d *NebulaMetaDetails) GetVpnAddrAndVersion() (netip.Addr, cert.Version, error) {
+	if d.OldVpnAddr != 0 {
+		b := [4]byte{}
+		binary.BigEndian.PutUint32(b[:], d.OldVpnAddr)
+		detailsVpnAddr := netip.AddrFrom4(b)
+		return detailsVpnAddr, cert.Version1, nil
+	} else if d.VpnAddr != nil {
+		detailsVpnAddr := protoAddrToNetAddr(d.VpnAddr)
+		return detailsVpnAddr, cert.Version2, nil
+	} else {
+		return netip.Addr{}, cert.Version1, ErrBadDetailsVpnAddr
+	}
+}

lighthouse_test.go

@@ -493,3 +493,123 @@ func Test_findNetworkUnion(t *testing.T) {
 	out, ok = findNetworkUnion([]netip.Prefix{fc00}, []netip.Addr{a1, afe81})
 	assert.False(t, ok)
 }
+
+func TestLighthouse_Dont_Delete_Static_Hosts(t *testing.T) {
+	l := test.NewLogger()
+
+	myUdpAddr2 := netip.MustParseAddrPort("1.2.3.4:4242")
+
+	testSameHostNotStatic := netip.MustParseAddr("10.128.0.41")
+	testStaticHost := netip.MustParseAddr("10.128.0.42")
+	//myVpnIp := netip.MustParseAddr("10.128.0.2")
+
+	c := config.NewC(l)
+	lh1 := "10.128.0.2"
+	c.Settings["lighthouse"] = map[string]any{
+		"hosts":    []any{lh1},
+		"interval": "1s",
+	}
+
+	c.Settings["listen"] = map[string]any{"port": 4242}
+	c.Settings["static_host_map"] = map[string]any{
+		lh1:           []any{"1.1.1.1:4242"},
+		"10.128.0.42": []any{"1.2.3.4:4242"},
+	}
+
+	myVpnNet := netip.MustParsePrefix("10.128.0.1/24")
+	nt := new(bart.Lite)
+	nt.Insert(myVpnNet)
+	cs := &CertState{
+		myVpnNetworks:      []netip.Prefix{myVpnNet},
+		myVpnNetworksTable: nt,
+	}
+	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	require.NoError(t, err)
+	lh.ifce = &mockEncWriter{}
+
+	//test that we actually have the static entry:
+	out := lh.Query(testStaticHost)
+	assert.NotNil(t, out)
+	assert.Equal(t, out.vpnAddrs[0], testStaticHost)
+	out.Rebuild([]netip.Prefix{}) //why tho
+	assert.Equal(t, out.addrs[0], myUdpAddr2)
+
+	//bolt on a lower numbered primary IP
+	am := lh.unlockedGetRemoteList([]netip.Addr{testStaticHost})
+	am.vpnAddrs = []netip.Addr{testSameHostNotStatic, testStaticHost}
+	lh.addrMap[testSameHostNotStatic] = am
+	out.Rebuild([]netip.Prefix{}) //???
+
+	//test that we actually have the static entry:
+	out = lh.Query(testStaticHost)
+	assert.NotNil(t, out)
+	assert.Equal(t, out.vpnAddrs[0], testSameHostNotStatic)
+	assert.Equal(t, out.vpnAddrs[1], testStaticHost)
+	assert.Equal(t, out.addrs[0], myUdpAddr2)
+
+	//test that we actually have the static entry for BOTH:
+	out2 := lh.Query(testSameHostNotStatic)
+	assert.Same(t, out2, out)
+
+	//now do the delete
+	lh.DeleteVpnAddrs([]netip.Addr{testSameHostNotStatic, testStaticHost})
+	//verify
+	out = lh.Query(testSameHostNotStatic)
+	assert.NotNil(t, out)
+	if out == nil {
+		t.Fatal("expected non-nil query for the static host")
+	}
+	assert.Equal(t, out.vpnAddrs[0], testSameHostNotStatic)
+	assert.Equal(t, out.vpnAddrs[1], testStaticHost)
+	assert.Equal(t, out.addrs[0], myUdpAddr2)
+}
+
+func TestLighthouse_DeletesWork(t *testing.T) {
+	l := test.NewLogger()
+
+	myUdpAddr2 := netip.MustParseAddrPort("1.2.3.4:4242")
+	testHost := netip.MustParseAddr("10.128.0.42")
+
+	c := config.NewC(l)
+	lh1 := "10.128.0.2"
+	c.Settings["lighthouse"] = map[string]any{
+		"hosts":    []any{lh1},
+		"interval": "1s",
+	}
+
+	c.Settings["listen"] = map[string]any{"port": 4242}
+	c.Settings["static_host_map"] = map[string]any{
+		lh1: []any{"1.1.1.1:4242"},
+	}
+
+	myVpnNet := netip.MustParsePrefix("10.128.0.1/24")
+	nt := new(bart.Lite)
+	nt.Insert(myVpnNet)
+	cs := &CertState{
+		myVpnNetworks:      []netip.Prefix{myVpnNet},
+		myVpnNetworksTable: nt,
+	}
+	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	require.NoError(t, err)
+	lh.ifce = &mockEncWriter{}
+
+	//insert the host
+	am := lh.unlockedGetRemoteList([]netip.Addr{testHost})
+	am.vpnAddrs = []netip.Addr{testHost}
+	am.addrs = []netip.AddrPort{myUdpAddr2}
+	lh.addrMap[testHost] = am
+	am.Rebuild([]netip.Prefix{}) //???
+
+	//test that we actually have the entry:
+	out := lh.Query(testHost)
+	assert.NotNil(t, out)
+	assert.Equal(t, out.vpnAddrs[0], testHost)
+	out.Rebuild([]netip.Prefix{}) //why tho
+	assert.Equal(t, out.addrs[0], myUdpAddr2)
+
+	//now do the delete
+	lh.DeleteVpnAddrs([]netip.Addr{testHost})
+	//verify
+	out = lh.Query(testHost)
+	assert.Nil(t, out)
+}

main.go

@@ -185,6 +185,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 
 	hostMap := NewHostMapFromConfig(l, c)
 	punchy := NewPunchyFromConfig(l, c)
+	connManager := newConnectionManagerFromConfig(l, c, hostMap, punchy)
 	lightHouse, err := NewLightHouseFromConfig(ctx, l, c, pki.getCertState(), udpConns[0], punchy)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to initialize lighthouse handler", err)
@@ -220,31 +221,26 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 	}
 
-	checkInterval := c.GetInt("timers.connection_alive_interval", 5)
-	pendingDeletionInterval := c.GetInt("timers.pending_deletion_interval", 10)
-
 	ifConfig := &InterfaceConfig{
-		HostMap:                 hostMap,
+		HostMap:               hostMap,
-		Inside:                  tun,
+		Inside:                tun,
-		Outside:                 udpConns[0],
+		Outside:               udpConns[0],
-		pki:                     pki,
+		pki:                   pki,
-		Firewall:                fw,
+		Firewall:              fw,
-		ServeDns:                serveDns,
+		ServeDns:              serveDns,
-		HandshakeManager:        handshakeManager,
+		HandshakeManager:      handshakeManager,
-		lightHouse:              lightHouse,
+		connectionManager:     connManager,
-		checkInterval:           time.Second * time.Duration(checkInterval),
+		lightHouse:            lightHouse,
-		pendingDeletionInterval: time.Second * time.Duration(pendingDeletionInterval),
+		tryPromoteEvery:       c.GetUint32("counters.try_promote", defaultPromoteEvery),
-		tryPromoteEvery:         c.GetUint32("counters.try_promote", defaultPromoteEvery),
+		reQueryEvery:          c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
-		reQueryEvery:            c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
+		reQueryWait:           c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
-		reQueryWait:             c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
+		DropLocalBroadcast:    c.GetBool("tun.drop_local_broadcast", false),
-		DropLocalBroadcast:      c.GetBool("tun.drop_local_broadcast", false),
+		DropMulticast:         c.GetBool("tun.drop_multicast", false),
-		DropMulticast:           c.GetBool("tun.drop_multicast", false),
+		routines:              routines,
-		routines:                routines,
+		MessageMetrics:        messageMetrics,
-		MessageMetrics:          messageMetrics,
+		version:               buildVersion,
-		version:                 buildVersion,
+		relayManager:          NewRelayManager(ctx, l, hostMap, c),
-		relayManager:            NewRelayManager(ctx, l, hostMap, c),
+		punchy:                punchy,
-		punchy:                  punchy,
-
 		ConntrackCacheTimeout: conntrackCacheTimeout,
 		l:                     l,
 	}
@@ -296,5 +292,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		statsStart,
 		dnsStart,
 		lightHouse.StartUpdateWorker,
+		connManager.Start,
 	}, nil
 }

outside.go

@@ -81,7 +81,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 			// Pull the Roaming parts up here, and return in all call paths.
 			f.handleHostRoaming(hostinfo, ip)
 			// Track usage of both the HostInfo and the Relay for the received & authenticated packet
-			f.connectionManager.In(hostinfo.localIndexId)
+			f.connectionManager.In(hostinfo)
 			f.connectionManager.RelayUsed(h.RemoteIndex)
 
 			relay, ok := hostinfo.relayState.QueryRelayForByIdx(h.RemoteIndex)
@@ -213,7 +213,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 
 	f.handleHostRoaming(hostinfo, ip)
 
-	f.connectionManager.In(hostinfo.localIndexId)
+	f.connectionManager.In(hostinfo)
 }
 
 // closeTunnel closes a tunnel locally, it does not send a closeTunnel packet to the remote
@@ -254,16 +254,18 @@ func (f *Interface) handleHostRoaming(hostinfo *HostInfo, udpAddr netip.AddrPort
 
 }
 
+// handleEncrypted returns true if a packet should be processed, false otherwise
 func (f *Interface) handleEncrypted(ci *ConnectionState, addr netip.AddrPort, h *header.H) bool {
-	// If connectionstate exists and the replay protector allows, process packet
+	// If connectionstate does not exist, send a recv error, if possible, to encourage a fast reconnect
-	// Else, send recv errors for 300 seconds after a restart to allow fast reconnection.
+	if ci == nil {
-	if ci == nil || !ci.window.Check(f.l, h.MessageCounter) {
 		if addr.IsValid() {
 			f.maybeSendRecvError(addr, h.RemoteIndex)
-			return false
-		} else {
-			return false
 		}
+		return false
+	}
+	// If the window check fails, refuse to process the packet, but don't send a recv error
+	if !ci.window.Check(f.l, h.MessageCounter) {
+		return false
 	}
 
 	return true
@@ -312,12 +314,11 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 	offset := ipv6.HeaderLen // Start at the end of the ipv6 header
 	next := 0
 	for {
-		if dataLen < offset {
+		if protoAt >= dataLen {
 			break
 		}
-
 		proto := layers.IPProtocol(data[protoAt])
-		//fmt.Println(proto, protoAt)
+
 		switch proto {
 		case layers.IPProtocolICMPv6, layers.IPProtocolESP, layers.IPProtocolNoNextHeader:
 			fp.Protocol = uint8(proto)
@@ -365,7 +366,7 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 
 		case layers.IPProtocolAH:
 			// Auth headers, used by IPSec, have a different meaning for header length
-			if dataLen < offset+1 {
+			if dataLen <= offset+1 {
 				break
 			}
 
@@ -373,7 +374,7 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 
 		default:
 			// Normal ipv6 header length processing
-			if dataLen < offset+1 {
+			if dataLen <= offset+1 {
 				break
 			}
 
@@ -499,7 +500,7 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 		return false
 	}
 
-	f.connectionManager.In(hostinfo.localIndexId)
+	f.connectionManager.In(hostinfo)
 	_, err = f.readers[q].Write(out)
 	if err != nil {
 		f.l.WithError(err).Error("Failed to write to tun")
@@ -538,10 +539,6 @@ func (f *Interface) handleRecvError(addr netip.AddrPort, h *header.H) {
 		return
 	}
 
-	if !hostinfo.RecvErrorExceeded() {
-		return
-	}
-
 	if hostinfo.remote.IsValid() && hostinfo.remote != addr {
 		f.l.Infoln("Someone spoofing recv_errors? ", addr, hostinfo.remote)
 		return

outside_test.go

@@ -117,6 +117,45 @@ func Test_newPacket_v6(t *testing.T) {
 	err = newPacket(buffer.Bytes(), true, p)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
+	// A v6 packet with a hop-by-hop extension
+	// ICMPv6 Payload (Echo Request)
+	icmpLayer := layers.ICMPv6{
+		TypeCode: layers.ICMPv6TypeEchoRequest,
+	}
+	// Hop-by-Hop Extension Header
+	hopOption := layers.IPv6HopByHopOption{}
+	hopOption.OptionData = []byte{0, 0, 0, 0}
+	hopByHop := layers.IPv6HopByHop{}
+	hopByHop.Options = append(hopByHop.Options, &hopOption)
+
+	ip = layers.IPv6{
+		Version:    6,
+		HopLimit:   128,
+		NextHeader: layers.IPProtocolIPv6Destination,
+		SrcIP:      net.IPv6linklocalallrouters,
+		DstIP:      net.IPv6linklocalallnodes,
+	}
+
+	buffer.Clear()
+	err = gopacket.SerializeLayers(buffer, gopacket.SerializeOptions{
+		ComputeChecksums: false,
+		FixLengths:       true,
+	}, &ip, &hopByHop, &icmpLayer)
+	if err != nil {
+		panic(err)
+	}
+	// Ensure buffer length checks during parsing with the next 2 tests.
+
+	// A full IPv6 header and 1 byte in the first extension, but missing
+	// the length byte.
+	err = newPacket(buffer.Bytes()[:41], true, p)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
+
+	// A full IPv6 header plus 1 full extension, but only 1 byte of the
+	// next layer, missing length byte
+	err = newPacket(buffer.Bytes()[:49], true, p)
+	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
+
 	// A good ICMP packet
 	ip = layers.IPv6{
 		Version:    6,
@@ -288,6 +327,10 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.Equal(t, uint16(22), p.LocalPort)
 	assert.False(t, p.Fragment)
 
+	// Ensure buffer bounds checking during processing
+	err = newPacket(b[:41], true, p)
+	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
+
 	// Invalid AH header
 	b = buffer.Bytes()
 	err = newPacket(b, true, p)

overlay/tun.go

@@ -1,6 +1,8 @@
 package overlay
 
 import (
+	"fmt"
+	"net"
 	"net/netip"
 
 	"github.com/sirupsen/logrus"
@@ -70,3 +72,51 @@ func findRemovedRoutes(newRoutes, oldRoutes []Route) []Route {
 
 	return removed
 }
+
+func prefixToMask(prefix netip.Prefix) netip.Addr {
+	pLen := 128
+	if prefix.Addr().Is4() {
+		pLen = 32
+	}
+
+	addr, _ := netip.AddrFromSlice(net.CIDRMask(prefix.Bits(), pLen))
+	return addr
+}
+
+func flipBytes(b []byte) []byte {
+	for i := 0; i < len(b); i++ {
+		b[i] ^= 0xFF
+	}
+	return b
+}
+func orBytes(a []byte, b []byte) []byte {
+	ret := make([]byte, len(a))
+	for i := 0; i < len(a); i++ {
+		ret[i] = a[i] | b[i]
+	}
+	return ret
+}
+
+func getBroadcast(cidr netip.Prefix) netip.Addr {
+	broadcast, _ := netip.AddrFromSlice(
+		orBytes(
+			cidr.Addr().AsSlice(),
+			flipBytes(prefixToMask(cidr).AsSlice()),
+		),
+	)
+	return broadcast
+}
+
+func selectGateway(dest netip.Prefix, gateways []netip.Prefix) (netip.Prefix, error) {
+	for _, gateway := range gateways {
+		if dest.Addr().Is4() && gateway.Addr().Is4() {
+			return gateway, nil
+		}
+
+		if dest.Addr().Is6() && gateway.Addr().Is6() {
+			return gateway, nil
+		}
+	}
+
+	return netip.Prefix{}, fmt.Errorf("no gateway found for %v in the list of vpn networks", dest)
+}

overlay/tun_darwin.go

@@ -7,7 +7,6 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"net"
 	"net/netip"
 	"os"
 	"sync/atomic"
@@ -295,7 +294,6 @@ func (t *tun) activate6(network netip.Prefix) error {
 			Vltime: 0xffffffff,
 			Pltime: 0xffffffff,
 		},
-		//TODO: CERT-V2 should we disable DAD (duplicate address detection) and mark this as a secured address?
 		Flags: _IN6_IFF_NODAD,
 	}
 
@@ -554,13 +552,3 @@ func (t *tun) Name() string {
 func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 	return nil, fmt.Errorf("TODO: multiqueue not implemented for darwin")
 }
-
-func prefixToMask(prefix netip.Prefix) netip.Addr {
-	pLen := 128
-	if prefix.Addr().Is4() {
-		pLen = 32
-	}
-
-	addr, _ := netip.AddrFromSlice(net.CIDRMask(prefix.Bits(), pLen))
-	return addr
-}

overlay/tun_freebsd.go

@@ -10,11 +10,9 @@ import (
 	"io"
 	"io/fs"
 	"net/netip"
-	"os"
-	"os/exec"
-	"strconv"
 	"sync/atomic"
 	"syscall"
+	"time"
 	"unsafe"
 
 	"github.com/gaissmai/bart"
@@ -22,12 +20,18 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
+	netroute "golang.org/x/net/route"
+	"golang.org/x/sys/unix"
 )
 
 const (
 	// FIODGNAME is defined in sys/sys/filio.h on FreeBSD
 	// For 32-bit systems, use FIODGNAME_32 (not defined in this file: 0x80086678)
-	FIODGNAME = 0x80106678
+	FIODGNAME        = 0x80106678
+	TUNSIFMODE       = 0x8004745e
+	TUNSIFHEAD       = 0x80047460
+	OSIOCAIFADDR_IN6 = 0x8088691b
+	IN6_IFF_NODAD    = 0x0020
 )
 
 type fiodgnameArg struct {
@@ -37,43 +41,159 @@ type fiodgnameArg struct {
 }
 
 type ifreqRename struct {
-	Name [16]byte
+	Name [unix.IFNAMSIZ]byte
 	Data uintptr
 }
 
 type ifreqDestroy struct {
-	Name [16]byte
+	Name [unix.IFNAMSIZ]byte
 	pad  [16]byte
 }
 
+type ifReq struct {
+	Name  [unix.IFNAMSIZ]byte
+	Flags uint16
+}
+
+type ifreqMTU struct {
+	Name [unix.IFNAMSIZ]byte
+	MTU  int32
+}
+
+type addrLifetime struct {
+	Expire    uint64
+	Preferred uint64
+	Vltime    uint32
+	Pltime    uint32
+}
+
+type ifreqAlias4 struct {
+	Name     [unix.IFNAMSIZ]byte
+	Addr     unix.RawSockaddrInet4
+	DstAddr  unix.RawSockaddrInet4
+	MaskAddr unix.RawSockaddrInet4
+	VHid     uint32
+}
+
+type ifreqAlias6 struct {
+	Name       [unix.IFNAMSIZ]byte
+	Addr       unix.RawSockaddrInet6
+	DstAddr    unix.RawSockaddrInet6
+	PrefixMask unix.RawSockaddrInet6
+	Flags      uint32
+	Lifetime   addrLifetime
+	VHid       uint32
+}
+
 type tun struct {
 	Device      string
 	vpnNetworks []netip.Prefix
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
+	linkAddr    *netroute.LinkAddr
 	l           *logrus.Logger
+	devFd       int
+}
 
-	io.ReadWriteCloser
+func (t *tun) Read(to []byte) (int, error) {
+	// use readv() to read from the tunnel device, to eliminate the need for copying the buffer
+	if t.devFd < 0 {
+		return -1, syscall.EINVAL
+	}
+
+	// first 4 bytes is protocol family, in network byte order
+	head := make([]byte, 4)
+
+	iovecs := []syscall.Iovec{
+		{&head[0], 4},
+		{&to[0], uint64(len(to))},
+	}
+
+	n, _, errno := syscall.Syscall(syscall.SYS_READV, uintptr(t.devFd), uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
+
+	var err error
+	if errno != 0 {
+		err = syscall.Errno(errno)
+	} else {
+		err = nil
+	}
+	// fix bytes read number to exclude header
+	bytesRead := int(n)
+	if bytesRead < 0 {
+		return bytesRead, err
+	} else if bytesRead < 4 {
+		return 0, err
+	} else {
+		return bytesRead - 4, err
+	}
+}
+
+// Write is only valid for single threaded use
+func (t *tun) Write(from []byte) (int, error) {
+	// use writev() to write to the tunnel device, to eliminate the need for copying the buffer
+	if t.devFd < 0 {
+		return -1, syscall.EINVAL
+	}
+
+	if len(from) <= 1 {
+		return 0, syscall.EIO
+	}
+	ipVer := from[0] >> 4
+	var head []byte
+	// first 4 bytes is protocol family, in network byte order
+	if ipVer == 4 {
+		head = []byte{0, 0, 0, syscall.AF_INET}
+	} else if ipVer == 6 {
+		head = []byte{0, 0, 0, syscall.AF_INET6}
+	} else {
+		return 0, fmt.Errorf("unable to determine IP version from packet")
+	}
+	iovecs := []syscall.Iovec{
+		{&head[0], 4},
+		{&from[0], uint64(len(from))},
+	}
+
+	n, _, errno := syscall.Syscall(syscall.SYS_WRITEV, uintptr(t.devFd), uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
+
+	var err error
+	if errno != 0 {
+		err = syscall.Errno(errno)
+	} else {
+		err = nil
+	}
+
+	return int(n) - 4, err
 }
 
 func (t *tun) Close() error {
-	if t.ReadWriteCloser != nil {
+	if t.devFd >= 0 {
-		if err := t.ReadWriteCloser.Close(); err != nil {
+		err := syscall.Close(t.devFd)
-			return err
-		}
-
-		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
 		if err != nil {
-			return err
+			t.l.WithError(err).Error("Error closing device")
 		}
-		defer syscall.Close(s)
+		t.devFd = -1
 
-		ifreq := ifreqDestroy{Name: t.deviceBytes()}
+		c := make(chan struct{})
+		go func() {
+			// destroying the interface can block if a read() is still pending. Do this asynchronously.
+			defer close(c)
+			s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
+			if err == nil {
+				defer syscall.Close(s)
+				ifreq := ifreqDestroy{Name: t.deviceBytes()}
+				err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
+			}
+			if err != nil {
+				t.l.WithError(err).Error("Error destroying tunnel")
+			}
+		}()
 
-		// Destroy the interface
+		// wait up to 1 second so we start blocking at the ioctl
-		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
+		select {
-		return err
+		case <-c:
+		case <-time.After(1 * time.Second):
+		}
 	}
 
 	return nil
@@ -85,32 +205,37 @@ func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun,
 
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open existing tun device
-	var file *os.File
+	var fd int
 	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName != "" {
-		file, err = os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
+		fd, err = syscall.Open("/dev/"+deviceName, syscall.O_RDWR, 0)
 	}
 	if errors.Is(err, fs.ErrNotExist) || deviceName == "" {
 		// If the device doesn't already exist, request a new one and rename it
-		file, err = os.OpenFile("/dev/tun", os.O_RDWR, 0)
+		fd, err = syscall.Open("/dev/tun", syscall.O_RDWR, 0)
 	}
 	if err != nil {
 		return nil, err
 	}
 
-	rawConn, err := file.SyscallConn()
+	// Read the name of the interface
-	if err != nil {
+	var name [16]byte
-		return nil, fmt.Errorf("SyscallConn: %v", err)
+	arg := fiodgnameArg{length: 16, buf: unsafe.Pointer(&name)}
+	ctrlErr := ioctl(uintptr(fd), FIODGNAME, uintptr(unsafe.Pointer(&arg)))
+
+	if ctrlErr == nil {
+		// set broadcast mode and multicast
+		ifmode := uint32(unix.IFF_BROADCAST | unix.IFF_MULTICAST)
+		ctrlErr = ioctl(uintptr(fd), TUNSIFMODE, uintptr(unsafe.Pointer(&ifmode)))
+	}
+
+	if ctrlErr == nil {
+		// turn on link-layer mode, to support ipv6
+		ifhead := uint32(1)
+		ctrlErr = ioctl(uintptr(fd), TUNSIFHEAD, uintptr(unsafe.Pointer(&ifhead)))
 	}
 
-	var name [16]byte
-	var ctrlErr error
-	rawConn.Control(func(fd uintptr) {
-		// Read the name of the interface
-		arg := fiodgnameArg{length: 16, buf: unsafe.Pointer(&name)}
-		ctrlErr = ioctl(fd, FIODGNAME, uintptr(unsafe.Pointer(&arg)))
-	})
 	if ctrlErr != nil {
 		return nil, err
 	}
@@ -122,11 +247,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 
 	// If the name doesn't match the desired interface name, rename it now
 	if ifName != deviceName {
-		s, err := syscall.Socket(
+		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-			syscall.AF_INET,
-			syscall.SOCK_DGRAM,
-			syscall.IPPROTO_IP,
-		)
 		if err != nil {
 			return nil, err
 		}
@@ -149,11 +270,11 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	}
 
 	t := &tun{
-		ReadWriteCloser: file,
+		Device:      deviceName,
-		Device:          deviceName,
+		vpnNetworks: vpnNetworks,
-		vpnNetworks:     vpnNetworks,
+		MTU:         c.GetInt("tun.mtu", DefaultMTU),
-		MTU:             c.GetInt("tun.mtu", DefaultMTU),
+		l:           l,
-		l:               l,
+		devFd:       fd,
 	}
 
 	err = t.reload(c, true)
@@ -172,38 +293,111 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 }
 
 func (t *tun) addIp(cidr netip.Prefix) error {
-	var err error
+	if cidr.Addr().Is4() {
-	// TODO use syscalls instead of exec.Command
+		ifr := ifreqAlias4{
-	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
+			Name: t.deviceBytes(),
-	t.l.Debug("command: ", cmd.String())
+			Addr: unix.RawSockaddrInet4{
-	if err = cmd.Run(); err != nil {
+				Len:    unix.SizeofSockaddrInet4,
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+				Family: unix.AF_INET,
+				Addr:   cidr.Addr().As4(),
+			},
+			DstAddr: unix.RawSockaddrInet4{
+				Len:    unix.SizeofSockaddrInet4,
+				Family: unix.AF_INET,
+				Addr:   getBroadcast(cidr).As4(),
+			},
+			MaskAddr: unix.RawSockaddrInet4{
+				Len:    unix.SizeofSockaddrInet4,
+				Family: unix.AF_INET,
+				Addr:   prefixToMask(cidr).As4(),
+			},
+			VHid: 0,
+		}
+		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+		// Note: unix.SIOCAIFADDR corresponds to FreeBSD's OSIOCAIFADDR
+		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&ifr))); err != nil {
+			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
+		}
+		return nil
 	}
 
-	cmd = exec.Command("/sbin/route", "-n", "add", "-net", cidr.String(), "-interface", t.Device)
+	if cidr.Addr().Is6() {
-	t.l.Debug("command: ", cmd.String())
+		ifr := ifreqAlias6{
-	if err = cmd.Run(); err != nil {
+			Name: t.deviceBytes(),
-		return fmt.Errorf("failed to run 'route add': %s", err)
+			Addr: unix.RawSockaddrInet6{
+				Len:    unix.SizeofSockaddrInet6,
+				Family: unix.AF_INET6,
+				Addr:   cidr.Addr().As16(),
+			},
+			PrefixMask: unix.RawSockaddrInet6{
+				Len:    unix.SizeofSockaddrInet6,
+				Family: unix.AF_INET6,
+				Addr:   prefixToMask(cidr).As16(),
+			},
+			Lifetime: addrLifetime{
+				Expire:    0,
+				Preferred: 0,
+				Vltime:    0xffffffff,
+				Pltime:    0xffffffff,
+			},
+			Flags: IN6_IFF_NODAD,
+		}
+		s, err := syscall.Socket(syscall.AF_INET6, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		if err := ioctl(uintptr(s), OSIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&ifr))); err != nil {
+			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
+		}
+		return nil
 	}
 
-	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
+	return fmt.Errorf("unknown address type %v", cidr)
-	t.l.Debug("command: ", cmd.String())
-	if err = cmd.Run(); err != nil {
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
-	}
-
-	// Unsafe path routes
-	return t.addRoutes(false)
 }
 
 func (t *tun) Activate() error {
+	// Setup our default MTU
+	err := t.setMTU()
+	if err != nil {
+		return err
+	}
+
+	linkAddr, err := getLinkAddr(t.Device)
+	if err != nil {
+		return err
+	}
+	if linkAddr == nil {
+		return fmt.Errorf("unable to discover link_addr for tun interface")
+	}
+	t.linkAddr = linkAddr
+
 	for i := range t.vpnNetworks {
 		err := t.addIp(t.vpnNetworks[i])
 		if err != nil {
 			return err
 		}
 	}
-	return nil
+
+	return t.addRoutes(false)
+}
+
+func (t *tun) setMTU() error {
+	// Set the MTU on the device
+	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+	if err != nil {
+		return err
+	}
+	defer syscall.Close(s)
+
+	ifm := ifreqMTU{Name: t.deviceBytes(), MTU: int32(t.MTU)}
+	err = ioctl(uintptr(s), unix.SIOCSIFMTU, uintptr(unsafe.Pointer(&ifm)))
+	return err
 }
 
 func (t *tun) reload(c *config.C, initial bool) error {
@@ -268,15 +462,16 @@ func (t *tun) addRoutes(logErrors bool) error {
 			continue
 		}
 
-		cmd := exec.Command("/sbin/route", "-n", "add", "-net", r.Cidr.String(), "-interface", t.Device)
+		err := addRoute(r.Cidr, t.linkAddr)
-		t.l.Debug("command: ", cmd.String())
+		if err != nil {
-		if err := cmd.Run(); err != nil {
+			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
-			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
+		} else {
+			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -289,9 +484,8 @@ func (t *tun) removeRoutes(routes []Route) error {
 			continue
 		}
 
-		cmd := exec.Command("/sbin/route", "-n", "delete", "-net", r.Cidr.String(), "-interface", t.Device)
+		err := delRoute(r.Cidr, t.linkAddr)
-		t.l.Debug("command: ", cmd.String())
+		if err != nil {
-		if err := cmd.Run(); err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
 			t.l.WithField("route", r).Info("Removed route")
@@ -306,3 +500,120 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	}
 	return
 }
+
+func addRoute(prefix netip.Prefix, gateway netroute.Addr) error {
+	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	if err != nil {
+		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
+	}
+	defer unix.Close(sock)
+
+	route := &netroute.RouteMessage{
+		Version: unix.RTM_VERSION,
+		Type:    unix.RTM_ADD,
+		Flags:   unix.RTF_UP,
+		Seq:     1,
+	}
+
+	if prefix.Addr().Is4() {
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
+			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
+			unix.RTAX_GATEWAY: gateway,
+		}
+	} else {
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
+			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
+			unix.RTAX_GATEWAY: gateway,
+		}
+	}
+
+	data, err := route.Marshal()
+	if err != nil {
+		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
+	}
+
+	_, err = unix.Write(sock, data[:])
+	if err != nil {
+		if errors.Is(err, unix.EEXIST) {
+			// Try to do a change
+			route.Type = unix.RTM_CHANGE
+			data, err = route.Marshal()
+			if err != nil {
+				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
+			}
+			_, err = unix.Write(sock, data[:])
+			fmt.Println("DOING CHANGE")
+			return err
+		}
+		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
+	}
+
+	return nil
+}
+
+func delRoute(prefix netip.Prefix, gateway netroute.Addr) error {
+	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	if err != nil {
+		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
+	}
+	defer unix.Close(sock)
+
+	route := netroute.RouteMessage{
+		Version: unix.RTM_VERSION,
+		Type:    unix.RTM_DELETE,
+		Seq:     1,
+	}
+
+	if prefix.Addr().Is4() {
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
+			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
+			unix.RTAX_GATEWAY: gateway,
+		}
+	} else {
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
+			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
+			unix.RTAX_GATEWAY: gateway,
+		}
+	}
+
+	data, err := route.Marshal()
+	if err != nil {
+		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
+	}
+	_, err = unix.Write(sock, data[:])
+	if err != nil {
+		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
+	}
+
+	return nil
+}
+
+// getLinkAddr Gets the link address for the interface of the given name
+func getLinkAddr(name string) (*netroute.LinkAddr, error) {
+	rib, err := netroute.FetchRIB(unix.AF_UNSPEC, unix.NET_RT_IFLIST, 0)
+	if err != nil {
+		return nil, err
+	}
+	msgs, err := netroute.ParseRIB(unix.NET_RT_IFLIST, rib)
+	if err != nil {
+		return nil, err
+	}
+
+	for _, m := range msgs {
+		switch m := m.(type) {
+		case *netroute.InterfaceMessage:
+			if m.Name == name {
+				sa, ok := m.Addrs[unix.RTAX_IFP].(*netroute.LinkAddr)
+				if ok {
+					return sa, nil
+				}
+			}
+		}
+	}
+
+	return nil, nil
+}

overlay/tun_linux.go

@@ -4,6 +4,7 @@
 package overlay
 
 import (
+	"errors"
 	"fmt"
 	"io"
 	"net"
@@ -101,12 +102,18 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 		}
 	}
 
+	tunNameTemplate := c.GetString("tun.dev", "nebula%d")
+	tunName, err := findNextTunName(tunNameTemplate)
+	if err != nil {
+		return nil, err
+	}
+
 	var req ifReq
 	req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI)
 	if multiqueue {
 		req.Flags |= unix.IFF_MULTI_QUEUE
 	}
-	copy(req.Name[:], c.GetString("tun.dev", ""))
+	copy(req.Name[:], tunName)
 	if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
 		return nil, err
 	}
@@ -123,6 +130,44 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	return t, nil
 }
 
+func findNextTunName(tunName string) (string, error) {
+	if !strings.HasSuffix(tunName, "%d") {
+		return tunName, nil
+	}
+	if len(tunName) == 2 {
+		return "", errors.New("please don't name your tun device '%d'")
+	}
+
+	if (len(tunName) - len("%d") + len("0")) > unix.IFNAMSIZ {
+		return "", fmt.Errorf("your tun device name template %s would result in a name longer than the maximum allowed length of %d", tunName, unix.IFNAMSIZ)
+	}
+
+	tunNameTemplate := tunName[:len(tunName)-len("%d")]
+	links, err := netlink.LinkList()
+	if err != nil {
+		return "", err
+	}
+	var candidateName string
+	i := 0
+	for {
+		candidateName = fmt.Sprintf("%s%d", tunNameTemplate, i)
+		good := true
+		for _, link := range links {
+			if candidateName == link.Attrs().Name {
+				good = false
+				break
+			}
+		}
+		if len(candidateName) > unix.IFNAMSIZ {
+			return "", fmt.Errorf("first available tun device is %s, which is longer than the max allowed size of %d", candidateName, unix.IFNAMSIZ)
+		}
+		if good {
+			return candidateName, nil
+		}
+	}
+	return "", errors.New("failed to find a tun device name")
+}
+
 func newTunGeneric(c *config.C, l *logrus.Logger, file *os.File, vpnNetworks []netip.Prefix) (*tun, error) {
 	t := &tun{
 		ReadWriteCloser:           file,
@@ -293,7 +338,6 @@ func (t *tun) addIPs(link netlink.Link) error {
 
 	//add all new addresses
 	for i := range newAddrs {
-		//TODO: CERT-V2 do we want to stack errors and try as many ops as possible?
 		//AddrReplace still adds new IPs, but if their properties change it will change them as well
 		if err := netlink.AddrReplace(link, newAddrs[i]); err != nil {
 			return err
@@ -361,6 +405,11 @@ func (t *tun) Activate() error {
 		t.l.WithError(err).Error("Failed to set tun tx queue length")
 	}
 
+	const modeNone = 1
+	if err = netlink.LinkSetIP6AddrGenMode(link, modeNone); err != nil {
+		t.l.WithError(err).Warn("Failed to disable link local address generation")
+	}
+
 	if err = t.addIPs(link); err != nil {
 		return err
 	}
@@ -578,9 +627,7 @@ func (t *tun) isGatewayInVpnNetworks(gwAddr netip.Addr) bool {
 }
 
 func (t *tun) getGatewaysFromRoute(r *netlink.Route) routing.Gateways {
-
 	var gateways routing.Gateways
-
 	link, err := netlink.LinkByName(t.Device)
 	if err != nil {
 		t.l.WithField("Devicename", t.Device).Error("Ignoring route update: failed to get link by name")
@@ -629,15 +676,18 @@ func (t *tun) getGatewaysFromRoute(r *netlink.Route) routing.Gateways {
 }
 
 func (t *tun) updateRoutes(r netlink.RouteUpdate) {
-
 	gateways := t.getGatewaysFromRoute(&r.Route)
-
 	if len(gateways) == 0 {
 		// No gateways relevant to our network, no routing changes required.
 		t.l.WithField("route", r).Debug("Ignoring route update, no gateways")
 		return
 	}
 
+	if r.Dst == nil {
+		t.l.WithField("route", r).Debug("Ignoring route update, no destination address")
+		return
+	}
+
 	dstAddr, ok := netip.AddrFromSlice(r.Dst.IP)
 	if !ok {
 		t.l.WithField("route", r).Debug("Ignoring route update, invalid destination address")

overlay/tun_netbsd.go

@@ -4,13 +4,12 @@
 package overlay
 
 import (
+	"errors"
 	"fmt"
 	"io"
 	"net/netip"
 	"os"
-	"os/exec"
 	"regexp"
-	"strconv"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
@@ -20,11 +19,42 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
+	netroute "golang.org/x/net/route"
+	"golang.org/x/sys/unix"
 )
 
-type ifreqDestroy struct {
+const (
-	Name [16]byte
+	SIOCAIFADDR_IN6 = 0x8080696b
-	pad  [16]byte
+	TUNSIFHEAD      = 0x80047442
+	TUNSIFMODE      = 0x80047458
+)
+
+type ifreqAlias4 struct {
+	Name     [unix.IFNAMSIZ]byte
+	Addr     unix.RawSockaddrInet4
+	DstAddr  unix.RawSockaddrInet4
+	MaskAddr unix.RawSockaddrInet4
+}
+
+type ifreqAlias6 struct {
+	Name       [unix.IFNAMSIZ]byte
+	Addr       unix.RawSockaddrInet6
+	DstAddr    unix.RawSockaddrInet6
+	PrefixMask unix.RawSockaddrInet6
+	Flags      uint32
+	Lifetime   addrLifetime
+}
+
+type ifreq struct {
+	Name [unix.IFNAMSIZ]byte
+	data int
+}
+
+type addrLifetime struct {
+	Expire    uint64
+	Preferred uint64
+	Vltime    uint32
+	Pltime    uint32
 }
 
 type tun struct {
@@ -34,40 +64,18 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
-
+	f           *os.File
-	io.ReadWriteCloser
+	fd          int
 }
 
-func (t *tun) Close() error {
+var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
-	if t.ReadWriteCloser != nil {
-		if err := t.ReadWriteCloser.Close(); err != nil {
-			return err
-		}
-
-		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		ifreq := ifreqDestroy{Name: t.deviceBytes()}
-
-		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
-
-		return err
-	}
-	return nil
-}
 
 func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in NetBSD")
 }
 
-var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
-
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open tun device
-	var file *os.File
 	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName == "" {
@@ -77,17 +85,23 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
 	}
 
-	file, err = os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
+	fd, err := unix.Open("/dev/"+deviceName, os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
 
+	err = unix.SetNonblock(fd, true)
+	if err != nil {
+		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
+	}
+
 	t := &tun{
-		ReadWriteCloser: file,
+		f:           os.NewFile(uintptr(fd), ""),
-		Device:          deviceName,
+		fd:          fd,
-		vpnNetworks:     vpnNetworks,
+		Device:      deviceName,
-		MTU:             c.GetInt("tun.mtu", DefaultMTU),
+		vpnNetworks: vpnNetworks,
-		l:               l,
+		MTU:         c.GetInt("tun.mtu", DefaultMTU),
+		l:           l,
 	}
 
 	err = t.reload(c, true)
@@ -105,40 +119,225 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	return t, nil
 }
 
+func (t *tun) Close() error {
+	if t.f != nil {
+		if err := t.f.Close(); err != nil {
+			return fmt.Errorf("error closing tun file: %w", err)
+		}
+
+		// t.f.Close should have handled it for us but let's be extra sure
+		_ = unix.Close(t.fd)
+
+		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		ifr := ifreq{Name: t.deviceBytes()}
+		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifr)))
+		return err
+	}
+	return nil
+}
+
+func (t *tun) Read(to []byte) (int, error) {
+	rc, err := t.f.SyscallConn()
+	if err != nil {
+		return 0, fmt.Errorf("failed to get syscall conn for tun: %w", err)
+	}
+
+	var errno syscall.Errno
+	var n uintptr
+	err = rc.Read(func(fd uintptr) bool {
+		// first 4 bytes is protocol family, in network byte order
+		head := [4]byte{}
+		iovecs := []syscall.Iovec{
+			{&head[0], 4},
+			{&to[0], uint64(len(to))},
+		}
+
+		n, _, errno = syscall.Syscall(syscall.SYS_READV, fd, uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
+		if errno.Temporary() {
+			// We got an EAGAIN, EINTR, or EWOULDBLOCK, go again
+			return false
+		}
+		return true
+	})
+	if err != nil {
+		if err == syscall.EBADF || err.Error() == "use of closed file" {
+			// Go doesn't export poll.ErrFileClosing but happily reports it to us so here we are
+			// https://github.com/golang/go/blob/master/src/internal/poll/fd_poll_runtime.go#L121
+			return 0, os.ErrClosed
+		}
+		return 0, fmt.Errorf("failed to make read call for tun: %w", err)
+	}
+
+	if errno != 0 {
+		return 0, fmt.Errorf("failed to make inner read call for tun: %w", errno)
+	}
+
+	// fix bytes read number to exclude header
+	bytesRead := int(n)
+	if bytesRead < 0 {
+		return bytesRead, nil
+	} else if bytesRead < 4 {
+		return 0, nil
+	} else {
+		return bytesRead - 4, nil
+	}
+}
+
+// Write is only valid for single threaded use
+func (t *tun) Write(from []byte) (int, error) {
+	if len(from) <= 1 {
+		return 0, syscall.EIO
+	}
+
+	ipVer := from[0] >> 4
+	var head [4]byte
+	// first 4 bytes is protocol family, in network byte order
+	if ipVer == 4 {
+		head[3] = syscall.AF_INET
+	} else if ipVer == 6 {
+		head[3] = syscall.AF_INET6
+	} else {
+		return 0, fmt.Errorf("unable to determine IP version from packet")
+	}
+
+	rc, err := t.f.SyscallConn()
+	if err != nil {
+		return 0, err
+	}
+
+	var errno syscall.Errno
+	var n uintptr
+	err = rc.Write(func(fd uintptr) bool {
+		iovecs := []syscall.Iovec{
+			{&head[0], 4},
+			{&from[0], uint64(len(from))},
+		}
+
+		n, _, errno = syscall.Syscall(syscall.SYS_WRITEV, fd, uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
+		// According to NetBSD documentation for TUN, writes will only return errors in which
+		// this packet will never be delivered so just go on living life.
+		return true
+	})
+	if err != nil {
+		return 0, err
+	}
+
+	if errno != 0 {
+		return 0, errno
+	}
+
+	return int(n) - 4, err
+}
+
 func (t *tun) addIp(cidr netip.Prefix) error {
-	var err error
+	if cidr.Addr().Is4() {
+		var req ifreqAlias4
+		req.Name = t.deviceBytes()
+		req.Addr = unix.RawSockaddrInet4{
+			Len:    unix.SizeofSockaddrInet4,
+			Family: unix.AF_INET,
+			Addr:   cidr.Addr().As4(),
+		}
+		req.DstAddr = unix.RawSockaddrInet4{
+			Len:    unix.SizeofSockaddrInet4,
+			Family: unix.AF_INET,
+			Addr:   cidr.Addr().As4(),
+		}
+		req.MaskAddr = unix.RawSockaddrInet4{
+			Len:    unix.SizeofSockaddrInet4,
+			Family: unix.AF_INET,
+			Addr:   prefixToMask(cidr).As4(),
+		}
 
-	// TODO use syscalls instead of exec.Command
+		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
+		if err != nil {
-	t.l.Debug("command: ", cmd.String())
+			return err
-	if err = cmd.Run(); err != nil {
+		}
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+		defer syscall.Close(s)
+
+		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&req))); err != nil {
+			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr(), err)
+		}
+
+		return nil
 	}
 
-	cmd = exec.Command("/sbin/route", "-n", "add", "-net", cidr.String(), cidr.Addr().String())
+	if cidr.Addr().Is6() {
-	t.l.Debug("command: ", cmd.String())
+		var req ifreqAlias6
-	if err = cmd.Run(); err != nil {
+		req.Name = t.deviceBytes()
-		return fmt.Errorf("failed to run 'route add': %s", err)
+		req.Addr = unix.RawSockaddrInet6{
+			Len:    unix.SizeofSockaddrInet6,
+			Family: unix.AF_INET6,
+			Addr:   cidr.Addr().As16(),
+		}
+		req.PrefixMask = unix.RawSockaddrInet6{
+			Len:    unix.SizeofSockaddrInet6,
+			Family: unix.AF_INET6,
+			Addr:   prefixToMask(cidr).As16(),
+		}
+		req.Lifetime = addrLifetime{
+			Vltime: 0xffffffff,
+			Pltime: 0xffffffff,
+		}
+
+		s, err := unix.Socket(unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		if err := ioctl(uintptr(s), SIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&req))); err != nil {
+			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
+		}
+		return nil
 	}
 
-	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
+	return fmt.Errorf("unknown address type %v", cidr)
-	t.l.Debug("command: ", cmd.String())
-	if err = cmd.Run(); err != nil {
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
-	}
-
-	// Unsafe path routes
-	return t.addRoutes(false)
 }
 
 func (t *tun) Activate() error {
+	mode := int32(unix.IFF_BROADCAST)
+	err := ioctl(uintptr(t.fd), TUNSIFMODE, uintptr(unsafe.Pointer(&mode)))
+	if err != nil {
+		return fmt.Errorf("failed to set tun device mode: %w", err)
+	}
+
+	v := 1
+	err = ioctl(uintptr(t.fd), TUNSIFHEAD, uintptr(unsafe.Pointer(&v)))
+	if err != nil {
+		return fmt.Errorf("failed to set tun device head: %w", err)
+	}
+
+	err = t.doIoctlByName(unix.SIOCSIFMTU, uint32(t.MTU))
+	if err != nil {
+		return fmt.Errorf("failed to set tun mtu: %w", err)
+	}
+
 	for i := range t.vpnNetworks {
-		err := t.addIp(t.vpnNetworks[i])
+		err = t.addIp(t.vpnNetworks[i])
 		if err != nil {
 			return err
 		}
 	}
-	return nil
+
+	return t.addRoutes(false)
+}
+
+func (t *tun) doIoctlByName(ctl uintptr, value uint32) error {
+	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+	if err != nil {
+		return err
+	}
+	defer syscall.Close(s)
+
+	ir := ifreq{Name: t.deviceBytes(), data: int(value)}
+	err = ioctl(uintptr(s), ctl, uintptr(unsafe.Pointer(&ir)))
+	return err
 }
 
 func (t *tun) reload(c *config.C, initial bool) error {
@@ -197,21 +396,23 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 
 func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
+
 	for _, r := range routes {
 		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
 
-		cmd := exec.Command("/sbin/route", "-n", "add", "-net", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		err := addRoute(r.Cidr, t.vpnNetworks)
-		t.l.Debug("command: ", cmd.String())
+		if err != nil {
-		if err := cmd.Run(); err != nil {
+			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
-			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
+		} else {
+			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -224,10 +425,8 @@ func (t *tun) removeRoutes(routes []Route) error {
 			continue
 		}
 
-		//TODO: CERT-V2 is this right?
+		err := delRoute(r.Cidr, t.vpnNetworks)
-		cmd := exec.Command("/sbin/route", "-n", "delete", "-net", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		if err != nil {
-		t.l.Debug("command: ", cmd.String())
-		if err := cmd.Run(); err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
 			t.l.WithField("route", r).Info("Removed route")
@@ -242,3 +441,109 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	}
 	return
 }
+
+func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
+	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	if err != nil {
+		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
+	}
+	defer unix.Close(sock)
+
+	route := &netroute.RouteMessage{
+		Version: unix.RTM_VERSION,
+		Type:    unix.RTM_ADD,
+		Flags:   unix.RTF_UP | unix.RTF_GATEWAY,
+		Seq:     1,
+	}
+
+	if prefix.Addr().Is4() {
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
+			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
+			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
+		}
+	} else {
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
+			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
+			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
+		}
+	}
+
+	data, err := route.Marshal()
+	if err != nil {
+		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
+	}
+
+	_, err = unix.Write(sock, data[:])
+	if err != nil {
+		if errors.Is(err, unix.EEXIST) {
+			// Try to do a change
+			route.Type = unix.RTM_CHANGE
+			data, err = route.Marshal()
+			if err != nil {
+				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
+			}
+			_, err = unix.Write(sock, data[:])
+			return err
+		}
+		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
+	}
+
+	return nil
+}
+
+func delRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
+	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	if err != nil {
+		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
+	}
+	defer unix.Close(sock)
+
+	route := netroute.RouteMessage{
+		Version: unix.RTM_VERSION,
+		Type:    unix.RTM_DELETE,
+		Seq:     1,
+	}
+
+	if prefix.Addr().Is4() {
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
+			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
+			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
+		}
+	} else {
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
+			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
+			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
+		}
+	}
+
+	data, err := route.Marshal()
+	if err != nil {
+		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
+	}
+	_, err = unix.Write(sock, data[:])
+	if err != nil {
+		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
+	}
+
+	return nil
+}

overlay/tun_openbsd.go

@@ -4,23 +4,50 @@
 package overlay
 
 import (
+	"errors"
 	"fmt"
 	"io"
 	"net/netip"
 	"os"
-	"os/exec"
 	"regexp"
-	"strconv"
 	"sync/atomic"
 	"syscall"
+	"unsafe"
 
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
+	netroute "golang.org/x/net/route"
+	"golang.org/x/sys/unix"
 )
 
+const (
+	SIOCAIFADDR_IN6 = 0x8080691a
+)
+
+type ifreqAlias4 struct {
+	Name     [unix.IFNAMSIZ]byte
+	Addr     unix.RawSockaddrInet4
+	DstAddr  unix.RawSockaddrInet4
+	MaskAddr unix.RawSockaddrInet4
+}
+
+type ifreqAlias6 struct {
+	Name       [unix.IFNAMSIZ]byte
+	Addr       unix.RawSockaddrInet6
+	DstAddr    unix.RawSockaddrInet6
+	PrefixMask unix.RawSockaddrInet6
+	Flags      uint32
+	Lifetime   [2]uint32
+}
+
+type ifreq struct {
+	Name [unix.IFNAMSIZ]byte
+	data int
+}
+
 type tun struct {
 	Device      string
 	vpnNetworks []netip.Prefix
@@ -28,48 +55,46 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
-
+	f           *os.File
-	io.ReadWriteCloser
+	fd          int
-
 	// cache out buffer since we need to prepend 4 bytes for tun metadata
 	out []byte
 }
 
-func (t *tun) Close() error {
-	if t.ReadWriteCloser != nil {
-		return t.ReadWriteCloser.Close()
-	}
-
-	return nil
-}
-
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
-	return nil, fmt.Errorf("newTunFromFd not supported in OpenBSD")
-}
-
 var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
 
+func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+	return nil, fmt.Errorf("newTunFromFd not supported in openbsd")
+}
+
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
+	// Try to open tun device
+	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName == "" {
-		return nil, fmt.Errorf("a device name in the format of tunN must be specified")
+		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
 	}
-
 	if !deviceNameRE.MatchString(deviceName) {
-		return nil, fmt.Errorf("a device name in the format of tunN must be specified")
+		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
 	}
 
-	file, err := os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
+	fd, err := unix.Open("/dev/"+deviceName, os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
 
+	err = unix.SetNonblock(fd, true)
+	if err != nil {
+		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
+	}
+
 	t := &tun{
-		ReadWriteCloser: file,
+		f:           os.NewFile(uintptr(fd), ""),
-		Device:          deviceName,
+		fd:          fd,
-		vpnNetworks:     vpnNetworks,
+		Device:      deviceName,
-		MTU:             c.GetInt("tun.mtu", DefaultMTU),
+		vpnNetworks: vpnNetworks,
-		l:               l,
+		MTU:         c.GetInt("tun.mtu", DefaultMTU),
+		l:           l,
 	}
 
 	err = t.reload(c, true)
@@ -87,6 +112,154 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	return t, nil
 }
 
+func (t *tun) Close() error {
+	if t.f != nil {
+		if err := t.f.Close(); err != nil {
+			return fmt.Errorf("error closing tun file: %w", err)
+		}
+
+		// t.f.Close should have handled it for us but let's be extra sure
+		_ = unix.Close(t.fd)
+	}
+	return nil
+}
+
+func (t *tun) Read(to []byte) (int, error) {
+	buf := make([]byte, len(to)+4)
+
+	n, err := t.f.Read(buf)
+
+	copy(to, buf[4:])
+	return n - 4, err
+}
+
+// Write is only valid for single threaded use
+func (t *tun) Write(from []byte) (int, error) {
+	buf := t.out
+	if cap(buf) < len(from)+4 {
+		buf = make([]byte, len(from)+4)
+		t.out = buf
+	}
+	buf = buf[:len(from)+4]
+
+	if len(from) == 0 {
+		return 0, syscall.EIO
+	}
+
+	// Determine the IP Family for the NULL L2 Header
+	ipVer := from[0] >> 4
+	if ipVer == 4 {
+		buf[3] = syscall.AF_INET
+	} else if ipVer == 6 {
+		buf[3] = syscall.AF_INET6
+	} else {
+		return 0, fmt.Errorf("unable to determine IP version from packet")
+	}
+
+	copy(buf[4:], from)
+
+	n, err := t.f.Write(buf)
+	return n - 4, err
+}
+
+func (t *tun) addIp(cidr netip.Prefix) error {
+	if cidr.Addr().Is4() {
+		var req ifreqAlias4
+		req.Name = t.deviceBytes()
+		req.Addr = unix.RawSockaddrInet4{
+			Len:    unix.SizeofSockaddrInet4,
+			Family: unix.AF_INET,
+			Addr:   cidr.Addr().As4(),
+		}
+		req.DstAddr = unix.RawSockaddrInet4{
+			Len:    unix.SizeofSockaddrInet4,
+			Family: unix.AF_INET,
+			Addr:   cidr.Addr().As4(),
+		}
+		req.MaskAddr = unix.RawSockaddrInet4{
+			Len:    unix.SizeofSockaddrInet4,
+			Family: unix.AF_INET,
+			Addr:   prefixToMask(cidr).As4(),
+		}
+
+		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&req))); err != nil {
+			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr(), err)
+		}
+
+		err = addRoute(cidr, t.vpnNetworks)
+		if err != nil {
+			return fmt.Errorf("failed to set route for vpn network %v: %w", cidr, err)
+		}
+
+		return nil
+	}
+
+	if cidr.Addr().Is6() {
+		var req ifreqAlias6
+		req.Name = t.deviceBytes()
+		req.Addr = unix.RawSockaddrInet6{
+			Len:    unix.SizeofSockaddrInet6,
+			Family: unix.AF_INET6,
+			Addr:   cidr.Addr().As16(),
+		}
+		req.PrefixMask = unix.RawSockaddrInet6{
+			Len:    unix.SizeofSockaddrInet6,
+			Family: unix.AF_INET6,
+			Addr:   prefixToMask(cidr).As16(),
+		}
+		req.Lifetime[0] = 0xffffffff
+		req.Lifetime[1] = 0xffffffff
+
+		s, err := unix.Socket(unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		if err := ioctl(uintptr(s), SIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&req))); err != nil {
+			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
+		}
+
+		return nil
+	}
+
+	return fmt.Errorf("unknown address type %v", cidr)
+}
+
+func (t *tun) Activate() error {
+	err := t.doIoctlByName(unix.SIOCSIFMTU, uint32(t.MTU))
+	if err != nil {
+		return fmt.Errorf("failed to set tun mtu: %w", err)
+	}
+
+	for i := range t.vpnNetworks {
+		err = t.addIp(t.vpnNetworks[i])
+		if err != nil {
+			return err
+		}
+	}
+
+	return t.addRoutes(false)
+}
+
+func (t *tun) doIoctlByName(ctl uintptr, value uint32) error {
+	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+	if err != nil {
+		return err
+	}
+	defer syscall.Close(s)
+
+	ir := ifreq{Name: t.deviceBytes(), data: int(value)}
+	err = ioctl(uintptr(s), ctl, uintptr(unsafe.Pointer(&ir)))
+	return err
+}
+
 func (t *tun) reload(c *config.C, initial bool) error {
 	change, routes, err := getAllRoutesFromConfig(c, t.vpnNetworks, initial)
 	if err != nil {
@@ -124,86 +297,11 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (t *tun) addIp(cidr netip.Prefix) error {
-	var err error
-	// TODO use syscalls instead of exec.Command
-	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
-	t.l.Debug("command: ", cmd.String())
-	if err = cmd.Run(); err != nil {
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
-	}
-
-	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
-	t.l.Debug("command: ", cmd.String())
-	if err = cmd.Run(); err != nil {
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
-	}
-
-	cmd = exec.Command("/sbin/route", "-n", "add", "-inet", cidr.String(), cidr.Addr().String())
-	t.l.Debug("command: ", cmd.String())
-	if err = cmd.Run(); err != nil {
-		return fmt.Errorf("failed to run 'route add': %s", err)
-	}
-
-	// Unsafe path routes
-	return t.addRoutes(false)
-}
-
-func (t *tun) Activate() error {
-	for i := range t.vpnNetworks {
-		err := t.addIp(t.vpnNetworks[i])
-		if err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
 func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
 
-func (t *tun) addRoutes(logErrors bool) error {
-	routes := *t.Routes.Load()
-	for _, r := range routes {
-		if len(r.Via) == 0 || !r.Install {
-			// We don't allow route MTUs so only install routes with a via
-			continue
-		}
-		//TODO: CERT-V2 is this right?
-		cmd := exec.Command("/sbin/route", "-n", "add", "-inet", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
-		t.l.Debug("command: ", cmd.String())
-		if err := cmd.Run(); err != nil {
-			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
-			if logErrors {
-				retErr.Log(t.l)
-			} else {
-				return retErr
-			}
-		}
-	}
-
-	return nil
-}
-
-func (t *tun) removeRoutes(routes []Route) error {
-	for _, r := range routes {
-		if !r.Install {
-			continue
-		}
-		//TODO: CERT-V2 is this right?
-		cmd := exec.Command("/sbin/route", "-n", "delete", "-inet", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
-		t.l.Debug("command: ", cmd.String())
-		if err := cmd.Run(); err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
-		} else {
-			t.l.WithField("route", r).Info("Removed route")
-		}
-	}
-	return nil
-}
-
 func (t *tun) Networks() []netip.Prefix {
 	return t.vpnNetworks
 }
@@ -213,43 +311,159 @@ func (t *tun) Name() string {
 }
 
 func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for freebsd")
+	return nil, fmt.Errorf("TODO: multiqueue not implemented for openbsd")
 }
 
-func (t *tun) Read(to []byte) (int, error) {
+func (t *tun) addRoutes(logErrors bool) error {
-	buf := make([]byte, len(to)+4)
+	routes := *t.Routes.Load()
 
-	n, err := t.ReadWriteCloser.Read(buf)
+	for _, r := range routes {
+		if len(r.Via) == 0 || !r.Install {
+			// We don't allow route MTUs so only install routes with a via
+			continue
+		}
 
-	copy(to, buf[4:])
+		err := addRoute(r.Cidr, t.vpnNetworks)
-	return n - 4, err
+		if err != nil {
+			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
+			if logErrors {
+				retErr.Log(t.l)
+			} else {
+				return retErr
+			}
+		} else {
+			t.l.WithField("route", r).Info("Added route")
+		}
+	}
+
+	return nil
 }
 
-// Write is only valid for single threaded use
+func (t *tun) removeRoutes(routes []Route) error {
-func (t *tun) Write(from []byte) (int, error) {
+	for _, r := range routes {
-	buf := t.out
+		if !r.Install {
-	if cap(buf) < len(from)+4 {
+			continue
-		buf = make([]byte, len(from)+4)
+		}
-		t.out = buf
-	}
-	buf = buf[:len(from)+4]
 
-	if len(from) == 0 {
+		err := delRoute(r.Cidr, t.vpnNetworks)
-		return 0, syscall.EIO
+		if err != nil {
+			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+		} else {
+			t.l.WithField("route", r).Info("Removed route")
+		}
+	}
+	return nil
+}
+
+func (t *tun) deviceBytes() (o [16]byte) {
+	for i, c := range t.Device {
+		o[i] = byte(c)
+	}
+	return
+}
+
+func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
+	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	if err != nil {
+		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
+	}
+	defer unix.Close(sock)
+
+	route := &netroute.RouteMessage{
+		Version: unix.RTM_VERSION,
+		Type:    unix.RTM_ADD,
+		Flags:   unix.RTF_UP | unix.RTF_GATEWAY,
+		Seq:     1,
 	}
 
-	// Determine the IP Family for the NULL L2 Header
+	if prefix.Addr().Is4() {
-	ipVer := from[0] >> 4
+		gw, err := selectGateway(prefix, gateways)
-	if ipVer == 4 {
+		if err != nil {
-		buf[3] = syscall.AF_INET
+			return err
-	} else if ipVer == 6 {
+		}
-		buf[3] = syscall.AF_INET6
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
+			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
+			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
+		}
 	} else {
-		return 0, fmt.Errorf("unable to determine IP version from packet")
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
+			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
+			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
+		}
 	}
 
-	copy(buf[4:], from)
+	data, err := route.Marshal()
+	if err != nil {
+		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
+	}
 
-	n, err := t.ReadWriteCloser.Write(buf)
+	_, err = unix.Write(sock, data[:])
-	return n - 4, err
+	if err != nil {
+		if errors.Is(err, unix.EEXIST) {
+			// Try to do a change
+			route.Type = unix.RTM_CHANGE
+			data, err = route.Marshal()
+			if err != nil {
+				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
+			}
+			_, err = unix.Write(sock, data[:])
+			return err
+		}
+		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
+	}
+
+	return nil
+}
+
+func delRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
+	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	if err != nil {
+		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
+	}
+	defer unix.Close(sock)
+
+	route := netroute.RouteMessage{
+		Version: unix.RTM_VERSION,
+		Type:    unix.RTM_DELETE,
+		Seq:     1,
+	}
+
+	if prefix.Addr().Is4() {
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
+			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
+			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
+		}
+	} else {
+		gw, err := selectGateway(prefix, gateways)
+		if err != nil {
+			return err
+		}
+		route.Addrs = []netroute.Addr{
+			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
+			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
+			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
+		}
+	}
+
+	data, err := route.Marshal()
+	if err != nil {
+		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
+	}
+	_, err = unix.Write(sock, data[:])
+	if err != nil {
+		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
+	}
+
+	return nil
 }

pkclient/pkclient_cgo.go

@@ -180,6 +180,7 @@ func (c *PKClient) DeriveNoise(peerPubKey []byte) ([]byte, error) {
 		pkcs11.NewAttribute(pkcs11.CKA_DECRYPT, true),
 		pkcs11.NewAttribute(pkcs11.CKA_WRAP, true),
 		pkcs11.NewAttribute(pkcs11.CKA_UNWRAP, true),
+		pkcs11.NewAttribute(pkcs11.CKA_VALUE_LEN, NoiseKeySize),
 	}
 
 	// Set up the parameters which include the peer's public key

pki.go

@@ -173,7 +173,6 @@ func (p *PKI) reloadCerts(c *config.C, initial bool) *util.ContextualError {
 
 	p.cs.Store(newState)
 
-	//TODO: CERT-V2 newState needs a stringer that does json
 	if initial {
 		p.l.WithField("cert", newState).Debug("Client nebula certificate(s)")
 	} else {
@@ -359,7 +358,9 @@ func newCertState(dv cert.Version, v1, v2 cert.Certificate, pkcs11backed bool, p
 			return nil, util.NewContextualError("v1 and v2 curve are not the same, ignoring", nil, nil)
 		}
 
-		//TODO: CERT-V2 make sure v2 has v1s address
+		if v1.Networks()[0] != v2.Networks()[0] {
+			return nil, util.NewContextualError("v1 and v2 networks are not the same", nil, nil)
+		}
 
 		cs.initiatingVersion = dv
 	}

remote_list.go

@@ -190,7 +190,7 @@ type RemoteList struct {
 	// The full list of vpn addresses assigned to this host
 	vpnAddrs []netip.Addr
 
-	// A deduplicated set of addresses. Any accessor should lock beforehand.
+	// A deduplicated set of underlay addresses. Any accessor should lock beforehand.
 	addrs []netip.AddrPort
 
 	// A set of relay addresses. VpnIp addresses that the remote identified as relays.
@@ -201,8 +201,10 @@ type RemoteList struct {
 	// For learned addresses, this is the vpnIp that sent the packet
 	cache map[netip.Addr]*cache
 
-	hr        *hostnamesResults
+	hr *hostnamesResults
-	shouldAdd func(netip.Addr) bool
+
+	// shouldAdd is a nillable function that decides if x should be added to addrs.
+	shouldAdd func(vpnAddrs []netip.Addr, x netip.Addr) bool
 
 	// This is a list of remotes that we have tried to handshake with and have returned from the wrong vpn ip.
 	// They should not be tried again during a handshake
@@ -213,7 +215,7 @@ type RemoteList struct {
 }
 
 // NewRemoteList creates a new empty RemoteList
-func NewRemoteList(vpnAddrs []netip.Addr, shouldAdd func(netip.Addr) bool) *RemoteList {
+func NewRemoteList(vpnAddrs []netip.Addr, shouldAdd func([]netip.Addr, netip.Addr) bool) *RemoteList {
 	r := &RemoteList{
 		vpnAddrs:  make([]netip.Addr, len(vpnAddrs)),
 		addrs:     make([]netip.AddrPort, 0),
@@ -368,6 +370,15 @@ func (r *RemoteList) CopyBlockedRemotes() []netip.AddrPort {
 	return c
 }
 
+// RefreshFromHandshake locks and updates the RemoteList to account for data learned upon a completed handshake
+func (r *RemoteList) RefreshFromHandshake(vpnAddrs []netip.Addr) {
+	r.Lock()
+	r.badRemotes = nil
+	r.vpnAddrs = make([]netip.Addr, len(vpnAddrs))
+	copy(r.vpnAddrs, vpnAddrs)
+	r.Unlock()
+}
+
 // ResetBlockedRemotes locks and clears the blocked remotes list
 func (r *RemoteList) ResetBlockedRemotes() {
 	r.Lock()
@@ -577,7 +588,7 @@ func (r *RemoteList) unlockedCollect() {
 
 	dnsAddrs := r.hr.GetAddrs()
 	for _, addr := range dnsAddrs {
-		if r.shouldAdd == nil || r.shouldAdd(addr.Addr()) {
+		if r.shouldAdd == nil || r.shouldAdd(r.vpnAddrs, addr.Addr()) {
 			if !r.unlockedIsBad(addr) {
 				addrs = append(addrs, addr)
 			}

udp/errors.go

@@ -0,0 +1,5 @@
+package udp
+
+import "errors"
+
+var ErrInvalidIPv6RemoteForSocket = errors.New("listener is IPv4, but writing to IPv6 remote")

udp/udp_darwin.go

@@ -3,20 +3,62 @@
 
 package udp
 
-// Darwin support is primarily implemented in udp_generic, besides NewListenConfig
-
 import (
+	"context"
+	"encoding/binary"
+	"errors"
 	"fmt"
 	"net"
 	"net/netip"
 	"syscall"
+	"unsafe"
 
 	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula/config"
 	"golang.org/x/sys/unix"
 )
 
+type StdConn struct {
+	*net.UDPConn
+	isV4  bool
+	sysFd uintptr
+	l     *logrus.Logger
+}
+
+var _ Conn = &StdConn{}
+
 func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
-	return NewGenericListener(l, ip, port, multi, batch)
+	lc := NewListenConfig(multi)
+	pc, err := lc.ListenPacket(context.TODO(), "udp", net.JoinHostPort(ip.String(), fmt.Sprintf("%v", port)))
+	if err != nil {
+		return nil, err
+	}
+
+	if uc, ok := pc.(*net.UDPConn); ok {
+		c := &StdConn{UDPConn: uc, l: l}
+
+		rc, err := uc.SyscallConn()
+		if err != nil {
+			return nil, fmt.Errorf("failed to open udp socket: %w", err)
+		}
+
+		err = rc.Control(func(fd uintptr) {
+			c.sysFd = fd
+		})
+		if err != nil {
+			return nil, fmt.Errorf("failed to get udp fd: %w", err)
+		}
+
+		la, err := c.LocalAddr()
+		if err != nil {
+			return nil, err
+		}
+		c.isV4 = la.Addr().Is4()
+
+		return c, nil
+	}
+
+	return nil, fmt.Errorf("unexpected PacketConn: %T %#v", pc, pc)
 }
 
 func NewListenConfig(multi bool) net.ListenConfig {
@@ -43,16 +85,116 @@ func NewListenConfig(multi bool) net.ListenConfig {
 	}
 }
 
-func (u *GenericConn) Rebind() error {
+//go:linkname sendto golang.org/x/sys/unix.sendto
-	rc, err := u.UDPConn.SyscallConn()
+//go:noescape
-	if err != nil {
+func sendto(s int, buf []byte, flags int, to unsafe.Pointer, addrlen int32) (err error)
-		return err
+
+func (u *StdConn) WriteTo(b []byte, ap netip.AddrPort) error {
+	var sa unsafe.Pointer
+	var addrLen int32
+
+	if u.isV4 {
+		if ap.Addr().Is6() {
+			return ErrInvalidIPv6RemoteForSocket
+		}
+
+		var rsa unix.RawSockaddrInet6
+		rsa.Family = unix.AF_INET6
+		rsa.Addr = ap.Addr().As16()
+		binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], ap.Port())
+		sa = unsafe.Pointer(&rsa)
+		addrLen = syscall.SizeofSockaddrInet4
+	} else {
+		var rsa unix.RawSockaddrInet6
+		rsa.Family = unix.AF_INET6
+		rsa.Addr = ap.Addr().As16()
+		binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], ap.Port())
+		sa = unsafe.Pointer(&rsa)
+		addrLen = syscall.SizeofSockaddrInet6
 	}
 
-	return rc.Control(func(fd uintptr) {
+	// Golang stdlib doesn't handle EAGAIN correctly in some situations so we do writes ourselves
-		err := syscall.SetsockoptInt(int(fd), unix.IPPROTO_IPV6, unix.IPV6_BOUND_IF, 0)
+	// See https://github.com/golang/go/issues/73919
-		if err != nil {
+	for {
-			u.l.WithError(err).Error("Failed to rebind udp socket")
+		//_, _, err := unix.Syscall6(unix.SYS_SENDTO, u.sysFd, uintptr(unsafe.Pointer(&b[0])), uintptr(len(b)), 0, sa, addrLen)
+		err := sendto(int(u.sysFd), b, 0, sa, addrLen)
+		if err == nil {
+			// Written, get out before the error handling
+			return nil
 		}
-	})
+
+		if errors.Is(err, syscall.EINTR) {
+			// Write was interrupted, retry
+			continue
+		}
+
+		if errors.Is(err, syscall.EAGAIN) {
+			return &net.OpError{Op: "sendto", Err: unix.EWOULDBLOCK}
+		}
+
+		if errors.Is(err, syscall.EBADF) {
+			return net.ErrClosed
+		}
+
+		return &net.OpError{Op: "sendto", Err: err}
+	}
+}
+
+func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
+	a := u.UDPConn.LocalAddr()
+
+	switch v := a.(type) {
+	case *net.UDPAddr:
+		addr, ok := netip.AddrFromSlice(v.IP)
+		if !ok {
+			return netip.AddrPort{}, fmt.Errorf("LocalAddr returned invalid IP address: %s", v.IP)
+		}
+		return netip.AddrPortFrom(addr, uint16(v.Port)), nil
+
+	default:
+		return netip.AddrPort{}, fmt.Errorf("LocalAddr returned: %#v", a)
+	}
+}
+
+func (u *StdConn) ReloadConfig(c *config.C) {
+	// TODO
+}
+
+func NewUDPStatsEmitter(udpConns []Conn) func() {
+	// No UDP stats for non-linux
+	return func() {}
+}
+
+func (u *StdConn) ListenOut(r EncReader) {
+	buffer := make([]byte, MTU)
+
+	for {
+		// Just read one packet at a time
+		n, rua, err := u.ReadFromUDPAddrPort(buffer)
+		if err != nil {
+			if errors.Is(err, net.ErrClosed) {
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
+			}
+
+			u.l.WithError(err).Error("unexpected udp socket receive error")
+		}
+
+		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])
+	}
+}
+
+func (u *StdConn) Rebind() error {
+	var err error
+	if u.isV4 {
+		err = syscall.SetsockoptInt(int(u.sysFd), syscall.IPPROTO_IP, syscall.IP_BOUND_IF, 0)
+	} else {
+		err = syscall.SetsockoptInt(int(u.sysFd), syscall.IPPROTO_IPV6, syscall.IPV6_BOUND_IF, 0)
+	}
+
+	if err != nil {
+		u.l.WithError(err).Error("Failed to rebind udp socket")
+	}
+
+	return nil
 }

udp/udp_generic.go

@@ -1,6 +1,7 @@
-//go:build (!linux || android) && !e2e_testing
+//go:build (!linux || android) && !e2e_testing && !darwin
 // +build !linux android
 // +build !e2e_testing
+// +build !darwin
 
 // udp_generic implements the nebula UDP interface in pure Go stdlib. This
 // means it can be used on platforms like Darwin and Windows.

udp/udp_linux.go

@@ -221,7 +221,7 @@ func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error {
 
 func (u *StdConn) writeTo4(b []byte, ip netip.AddrPort) error {
 	if !ip.Addr().Is4() {
-		return fmt.Errorf("Listener is IPv4, but writing to IPv6 remote")
+		return ErrInvalidIPv6RemoteForSocket
 	}
 
 	var rsa unix.RawSockaddrInet4

udp/udp_rio_windows.go

@@ -92,6 +92,25 @@ func (u *RIOConn) bind(sa windows.Sockaddr) error {
 	// Enable v4 for this socket
 	syscall.SetsockoptInt(syscall.Handle(u.sock), syscall.IPPROTO_IPV6, syscall.IPV6_V6ONLY, 0)
 
+	// Disable reporting of PORT_UNREACHABLE and NET_UNREACHABLE errors from the UDP socket receive call.
+	// These errors are returned on Windows during UDP receives based on the receipt of ICMP packets. Disable
+	// the UDP receive error returns with these ioctl calls.
+	ret := uint32(0)
+	flag := uint32(0)
+	size := uint32(unsafe.Sizeof(flag))
+	err = syscall.WSAIoctl(syscall.Handle(u.sock), syscall.SIO_UDP_CONNRESET, (*byte)(unsafe.Pointer(&flag)), size, nil, 0, &ret, nil, 0)
+	if err != nil {
+		return err
+	}
+	ret = 0
+	flag = 0
+	size = uint32(unsafe.Sizeof(flag))
+	SIO_UDP_NETRESET := uint32(syscall.IOC_IN | syscall.IOC_VENDOR | 15)
+	err = syscall.WSAIoctl(syscall.Handle(u.sock), SIO_UDP_NETRESET, (*byte)(unsafe.Pointer(&flag)), size, nil, 0, &ret, nil, 0)
+	if err != nil {
+		return err
+	}
+
 	err = u.rx.Open()
 	if err != nil {
 		return err
@@ -122,8 +141,12 @@ func (u *RIOConn) ListenOut(r EncReader) {
 		// Just read one packet at a time
 		n, rua, err := u.receive(buffer)
 		if err != nil {
-			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+			if errors.Is(err, net.ErrClosed) {
-			return
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
+			}
+			u.l.WithError(err).Error("unexpected udp socket receive error")
+			continue
 		}
 
 		r(netip.AddrPortFrom(netip.AddrFrom16(rua.Addr).Unmap(), (rua.Port>>8)|((rua.Port&0xff)<<8)), buffer[:n])