firewall can distinguish if the host connecting has an overlapping network, is a VPN peer without an overlapping network, or is a unsafe network

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

@@ -1,13 +1,21 @@
 blank_issues_enabled: true
 contact_links:
+  - name: 💨 Performance Issues
+    url: https://github.com/slackhq/nebula/discussions/new/choose
+    about: 'We ask that you create a discussion instead of an issue for performance-related questions. This allows us to have a more open conversation about the issue and helps us to better understand the problem.'
+
+  - name: 📄 Documentation Issues
+    url: https://github.com/definednet/nebula-docs
+    about: "If you've found an issue with the website documentation, please file it in the nebula-docs repository."
+
+  - name: 📱 Mobile Nebula Issues
+    url: https://github.com/definednet/mobile_nebula
+    about: "If you're using the mobile Nebula app and have found an issue, please file it in the mobile_nebula repository."
+
   - name: 📘 Documentation
     url: https://nebula.defined.net/docs/
-    about: Review documentation.
+    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/enQtOTA5MDI4NDg3MTg4LTkwY2EwNTI4NzQyMzc0M2ZlODBjNWI3NTY1MzhiOThiMmZlZjVkMTI0NGY4YTMyNjUwMWEyNzNkZTJmYzQxOGU
+    url: https://join.slack.com/t/nebulaoss/shared_invite/zt-39pk4xopc-CUKlGcb5Z39dQ0cK1v7ehA
-    about: 'This issue tracker is not for support questions. Join us on Slack for assistance!'
+    about: 'For faster support, join us on Slack for assistance!'
-
-  - name: 📱 Mobile Nebula
-    url: https://github.com/definednet/mobile_nebula
-    about: 'This issue tracker is not for mobile support. Try the Mobile Nebula repo instead!'

.github/pull_request_template.md

@@ -0,0 +1,11 @@
+<!--
+Thank you for taking the time to submit a pull request!
+
+Please be sure to provide a clear description of what you're trying to achieve with the change.
+
+- If you're submitting a new feature, please explain how to use it and document any new config options in the example config.
+- If you're submitting a bugfix, please link the related issue or describe the circumstances surrounding the issue.
+- If you're changing a default, explain why you believe the new default is appropriate for most users.
+
+P.S. If you're only updating the README or other docs, please file a pull request here instead: https://github.com/DefinedNet/nebula-docs
+-->

.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: '1.23'
+        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: '1.23'
+          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: '1.23'
+          go-version: '1.25'
           check-latest: true
 
       - name: Build
@@ -68,14 +68,14 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@v6
         with:
-          go-version: '1.23'
+          go-version: '1.25'
           check-latest: true
 
       - name: Import certificates
         if: env.HAS_SIGNING_CREDS == 'true'
-        uses: Apple-Actions/import-codesign-certs@v3
+        uses: Apple-Actions/import-codesign-certs@v5
         with:
           p12-file-base64: ${{ secrets.APPLE_DEVELOPER_CERTIFICATE_P12_BASE64 }}
           p12-password: ${{ secrets.APPLE_DEVELOPER_CERTIFICATE_PASSWORD }}

.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: '1.23'
+        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: '1.23'
+        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@v6
+      uses: golangci/golangci-lint-action@v8
       with:
-        version: v1.64
+        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: '1.23'
+        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: '1.23'
+        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@v6
+      uses: golangci/golangci-lint-action@v8
       with:
-        version: v1.64
+        version: v2.5
 
     - name: Test
       run: make test

.golangci.yaml

@@ -1,9 +1,23 @@
-# yaml-language-server: $schema=https://golangci-lint.run/jsonschema/golangci.jsonschema.json
+version: "2"
 linters:
-  # Disable all linters.
+  default: none
-  # Default: false
-  disable-all: true
-  # Enable specific linter
-  # https://golangci-lint.run/usage/linters/#enabled-by-default
   enable:
     - testifylint
+  exclusions:
+    generated: lax
+    presets:
+      - comments
+      - common-false-positives
+      - legacy
+      - std-error-handling
+    paths:
+      - third_party$
+      - builtin$
+      - examples$
+formatters:
+  exclusions:
+    generated: lax
+    paths:
+      - third_party$
+      - builtin$
+      - examples$

CHANGELOG.md

@@ -7,6 +7,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Changed
+
+- `default_local_cidr_any` now defaults to false, meaning that any firewall rule
+  intended to target an `unsafe_routes` entry must explicitly declare it via the
+  `local_cidr` field. This is almost always the intended behavior. This flag is
+  deprecated and will be removed in a future release.
+
 ## [1.9.4] - 2024-09-09
 
 ### Added

README.md

@@ -4,7 +4,7 @@ It lets you seamlessly connect computers anywhere in the world. Nebula is portab
 It can be used to connect a small number of computers, but is also able to connect tens of thousands of computers.
 
 Nebula incorporates a number of existing concepts like encryption, security groups, certificates,
-and tunneling, and each of those individual pieces existed before Nebula in various forms.
+and tunneling.
 What makes Nebula different to existing offerings is that it brings all of these ideas together,
 resulting in a sum that is greater than its individual parts.
 
@@ -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
 
@@ -28,33 +28,33 @@ Check the [releases](https://github.com/slackhq/nebula/releases/latest) page for
 #### Distribution Packages
 
 - [Arch Linux](https://archlinux.org/packages/extra/x86_64/nebula/)
-    ```
+    ```sh
-    $ sudo pacman -S nebula
+    sudo pacman -S nebula
     ```
 
 - [Fedora Linux](https://src.fedoraproject.org/rpms/nebula)
-    ```
+    ```sh
-    $ sudo dnf install nebula
+    sudo dnf install nebula
     ```
 
 - [Debian Linux](https://packages.debian.org/source/stable/nebula)
-    ```
+    ```sh
-    $ sudo apt install nebula
+    sudo apt install nebula
     ```
 
 - [Alpine Linux](https://pkgs.alpinelinux.org/packages?name=nebula)
-    ```
+    ```sh
-    $ sudo apk add nebula
+    sudo apk add nebula
     ```
 
 - [macOS Homebrew](https://github.com/Homebrew/homebrew-core/blob/HEAD/Formula/n/nebula.rb)
-    ```
+    ```sh
-    $ brew install nebula
+    brew install nebula
     ```
 
 - [Docker](https://hub.docker.com/r/nebulaoss/nebula)
-    ```
+    ```sh
-    $ docker pull nebulaoss/nebula
+    docker pull nebulaoss/nebula
     ```
 
 #### Mobile
@@ -64,10 +64,10 @@ Check the [releases](https://github.com/slackhq/nebula/releases/latest) page for
 
 ## Technical Overview
 
-Nebula is a mutually authenticated peer-to-peer software defined network based on the [Noise Protocol Framework](https://noiseprotocol.org/).
+Nebula is a mutually authenticated peer-to-peer software-defined network based on the [Noise Protocol Framework](https://noiseprotocol.org/).
 Nebula uses certificates to assert a node's IP address, name, and membership within user-defined groups.
 Nebula's user-defined groups allow for provider agnostic traffic filtering between nodes.
-Discovery nodes allow individual peers to find each other and optionally use UDP hole punching to establish connections from behind most firewalls or NATs.
+Discovery nodes (aka lighthouses) allow individual peers to find each other and optionally use UDP hole punching to establish connections from behind most firewalls or NATs.
 Users can move data between nodes in any number of cloud service providers, datacenters, and endpoints, without needing to maintain a particular addressing scheme.
 
 Nebula uses Elliptic-curve Diffie-Hellman (`ECDH`) key exchange and `AES-256-GCM` in its default configuration.
@@ -82,28 +82,34 @@ To set up a Nebula network, you'll need:
 
 #### 2. (Optional, but you really should..) At least one discovery node with a routable IP address, which we call a lighthouse.
 
-Nebula lighthouses allow nodes to find each other, anywhere in the world. A lighthouse is the only node in a Nebula network whose IP should not change. Running a lighthouse requires very few compute resources, and you can easily use the least expensive option from a cloud hosting provider. If you're not sure which provider to use, a number of us have used $5/mo [DigitalOcean](https://digitalocean.com) droplets as lighthouses.
+Nebula lighthouses allow nodes to find each other, anywhere in the world. A lighthouse is the only node in a Nebula network whose IP should not change. Running a lighthouse requires very few compute resources, and you can easily use the least expensive option from a cloud hosting provider. If you're not sure which provider to use, a number of us have used $6/mo [DigitalOcean](https://digitalocean.com) droplets as lighthouses.
-
-  Once you have launched an instance, ensure that Nebula udp traffic (default port udp/4242) can reach it over the internet.
 
+Once you have launched an instance, ensure that Nebula udp traffic (default port udp/4242) can reach it over the internet.
 
 #### 3. A Nebula certificate authority, which will be the root of trust for a particular Nebula network.
 
-  ```
+```sh
-  ./nebula-cert ca -name "Myorganization, Inc"
+./nebula-cert ca -name "Myorganization, Inc"
-  ```
+```
-  This will create files named `ca.key` and `ca.cert` in the current directory. The `ca.key` file is the most sensitive file you'll create, because it is the key used to sign the certificates for individual nebula nodes/hosts. Please store this file somewhere safe, preferably with strong encryption.
+
+This will create files named `ca.key` and `ca.cert` in the current directory. The `ca.key` file is the most sensitive file you'll create, because it is the key used to sign the certificates for individual nebula nodes/hosts. Please store this file somewhere safe, preferably with strong encryption.
+
+**Be aware!** By default, certificate authorities have a 1-year lifetime before expiration. See [this guide](https://nebula.defined.net/docs/guides/rotating-certificate-authority/) for details on rotating a CA.
 
 #### 4. Nebula host keys and certificates generated from that certificate authority
+
 This assumes you have four nodes, named lighthouse1, laptop, server1, host3. You can name the nodes any way you'd like, including FQDN. You'll also need to choose IP addresses and the associated subnet. In this example, we are creating a nebula network that will use 192.168.100.x/24 as its network range. This example also demonstrates nebula groups, which can later be used to define traffic rules in a nebula network.
-```
+```sh
 ./nebula-cert sign -name "lighthouse1" -ip "192.168.100.1/24"
 ./nebula-cert sign -name "laptop" -ip "192.168.100.2/24" -groups "laptop,home,ssh"
 ./nebula-cert sign -name "server1" -ip "192.168.100.9/24" -groups "servers"
 ./nebula-cert sign -name "host3" -ip "192.168.100.10/24"
 ```
 
+By default, host certificates will expire 1 second before the CA expires. Use the `-duration` flag to specify a shorter lifetime.
+
 #### 5. Configuration files for each host
+
 Download a copy of the nebula [example configuration](https://github.com/slackhq/nebula/blob/master/examples/config.yml).
 
 * On the lighthouse node, you'll need to ensure `am_lighthouse: true` is set.
@@ -118,10 +124,13 @@ For each host, copy the nebula binary to the host, along with `config.yml` from
 **DO NOT COPY `ca.key` TO INDIVIDUAL NODES.**
 
 #### 7. Run nebula on each host
-```
+
+```sh
 ./nebula -config /path/to/config.yml
 ```
 
+For more detailed instructions, [find the full documentation here](https://nebula.defined.net/docs/).
+
 ## Building Nebula from source
 
 Make sure you have [go](https://go.dev/doc/install) installed and clone this repo. Change to the nebula directory.
@@ -140,8 +149,10 @@ The default curve used for cryptographic handshakes and signatures is Curve25519
 
 In addition, Nebula can be built using the [BoringCrypto GOEXPERIMENT](https://github.com/golang/go/blob/go1.20/src/crypto/internal/boring/README.md) by running either of the following make targets:
 
-    make bin-boringcrypto
+```sh
-    make release-boringcrypto
+make bin-boringcrypto
+make release-boringcrypto
+```
 
 This is not the recommended default deployment, but may be useful based on your compliance requirements.
 
@@ -149,5 +160,3 @@ This is not the recommended default deployment, but may be useful based on your
 
 Nebula was created at Slack Technologies, Inc by Nate Brown and Ryan Huber, with contributions from Oliver Fross, Alan Lam, Wade Simmons, and Lining Wang.
 
-
-

allow_list.go

@@ -36,7 +36,7 @@ type AllowListNameRule struct {
 
 func NewLocalAllowListFromConfig(c *config.C, k string) (*LocalAllowList, error) {
 	var nameRules []AllowListNameRule
-	handleKey := func(key string, value interface{}) (bool, error) {
+	handleKey := func(key string, value any) (bool, error) {
 		if key == "interfaces" {
 			var err error
 			nameRules, err = getAllowListInterfaces(k, value)
@@ -70,7 +70,7 @@ func NewRemoteAllowListFromConfig(c *config.C, k, rangesKey string) (*RemoteAllo
 
 // If the handleKey func returns true, the rest of the parsing is skipped
 // for this key. This allows parsing of special values like `interfaces`.
-func newAllowListFromConfig(c *config.C, k string, handleKey func(key string, value interface{}) (bool, error)) (*AllowList, error) {
+func newAllowListFromConfig(c *config.C, k string, handleKey func(key string, value any) (bool, error)) (*AllowList, error) {
 	r := c.Get(k)
 	if r == nil {
 		return nil, nil
@@ -81,8 +81,8 @@ func newAllowListFromConfig(c *config.C, k string, handleKey func(key string, va
 
 // If the handleKey func returns true, the rest of the parsing is skipped
 // for this key. This allows parsing of special values like `interfaces`.
-func newAllowList(k string, raw interface{}, handleKey func(key string, value interface{}) (bool, error)) (*AllowList, error) {
+func newAllowList(k string, raw any, handleKey func(key string, value any) (bool, error)) (*AllowList, error) {
-	rawMap, ok := raw.(map[interface{}]interface{})
+	rawMap, ok := raw.(map[string]any)
 	if !ok {
 		return nil, fmt.Errorf("config `%s` has invalid type: %T", k, raw)
 	}
@@ -100,12 +100,7 @@ func newAllowList(k string, raw interface{}, handleKey func(key string, value in
 	rules4 := allowListRules{firstValue: true, allValuesMatch: true, defaultSet: false}
 	rules6 := allowListRules{firstValue: true, allValuesMatch: true, defaultSet: false}
 
-	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)
-		}
-
 		if handleKey != nil {
 			handled, err := handleKey(rawCIDR, rawValue)
 			if err != nil {
@@ -116,7 +111,7 @@ func newAllowList(k string, raw interface{}, handleKey func(key string, value in
 			}
 		}
 
-		value, ok := rawValue.(bool)
+		value, ok := config.AsBool(rawValue)
 		if !ok {
 			return nil, fmt.Errorf("config `%s` has invalid value (type %T): %v", k, rawValue, rawValue)
 		}
@@ -173,22 +168,18 @@ func newAllowList(k string, raw interface{}, handleKey func(key string, value in
 	return &AllowList{cidrTree: tree}, nil
 }
 
-func getAllowListInterfaces(k string, v interface{}) ([]AllowListNameRule, error) {
+func getAllowListInterfaces(k string, v any) ([]AllowListNameRule, error) {
 	var nameRules []AllowListNameRule
 
-	rawRules, ok := v.(map[interface{}]interface{})
+	rawRules, ok := v.(map[string]any)
 	if !ok {
 		return nil, fmt.Errorf("config `%s.interfaces` is invalid (type %T): %v", k, v, v)
 	}
 
 	firstEntry := true
 	var allValues bool
-	for rawName, rawAllow := range rawRules {
+	for name, rawAllow := range rawRules {
-		name, ok := rawName.(string)
+		allow, ok := config.AsBool(rawAllow)
-		if !ok {
-			return nil, fmt.Errorf("config `%s.interfaces` has invalid key (type %T): %v", k, rawName, rawName)
-		}
-		allow, ok := rawAllow.(bool)
 		if !ok {
 			return nil, fmt.Errorf("config `%s.interfaces` has invalid value (type %T): %v", k, rawAllow, rawAllow)
 		}
@@ -224,16 +215,11 @@ func getRemoteAllowRanges(c *config.C, k string) (*bart.Table[*AllowList], error
 
 	remoteAllowRanges := new(bart.Table[*AllowList])
 
-	rawMap, ok := value.(map[interface{}]interface{})
+	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)
-		}
-
 		allowList, err := newAllowList(fmt.Sprintf("%s.%s", k, rawCIDR), rawValue, nil)
 		if err != nil {
 			return nil, err

allow_list_test.go

@@ -15,27 +15,27 @@ import (
 func TestNewAllowListFromConfig(t *testing.T) {
 	l := test.NewLogger()
 	c := config.NewC(l)
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
 		"192.168.0.0": true,
 	}
 	r, err := newAllowListFromConfig(c, "allowlist", nil)
 	require.EqualError(t, err, "config `allowlist` has invalid CIDR: 192.168.0.0. netip.ParsePrefix(\"192.168.0.0\"): no '/'")
 	assert.Nil(t, r)
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
 		"192.168.0.0/16": "abc",
 	}
 	r, err = newAllowListFromConfig(c, "allowlist", nil)
 	require.EqualError(t, err, "config `allowlist` has invalid value (type string): abc")
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
 		"192.168.0.0/16": true,
 		"10.0.0.0/8":     false,
 	}
 	r, err = newAllowListFromConfig(c, "allowlist", nil)
 	require.EqualError(t, err, "config `allowlist` contains both true and false rules, but no default set for 0.0.0.0/0")
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
 		"0.0.0.0/0":      true,
 		"10.0.0.0/8":     false,
 		"10.42.42.0/24":  true,
@@ -45,7 +45,7 @@ func TestNewAllowListFromConfig(t *testing.T) {
 	r, err = newAllowListFromConfig(c, "allowlist", nil)
 	require.EqualError(t, err, "config `allowlist` contains both true and false rules, but no default set for ::/0")
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
 		"0.0.0.0/0":     true,
 		"10.0.0.0/8":    false,
 		"10.42.42.0/24": true,
@@ -55,7 +55,7 @@ func TestNewAllowListFromConfig(t *testing.T) {
 		assert.NotNil(t, r)
 	}
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
 		"0.0.0.0/0":      true,
 		"10.0.0.0/8":     false,
 		"10.42.42.0/24":  true,
@@ -70,16 +70,16 @@ func TestNewAllowListFromConfig(t *testing.T) {
 
 	// Test interface names
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
-		"interfaces": map[interface{}]interface{}{
+		"interfaces": map[string]any{
 			`docker.*`: "foo",
 		},
 	}
 	lr, err := NewLocalAllowListFromConfig(c, "allowlist")
 	require.EqualError(t, err, "config `allowlist.interfaces` has invalid value (type string): foo")
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
-		"interfaces": map[interface{}]interface{}{
+		"interfaces": map[string]any{
 			`docker.*`: false,
 			`eth.*`:    true,
 		},
@@ -87,8 +87,8 @@ func TestNewAllowListFromConfig(t *testing.T) {
 	lr, err = NewLocalAllowListFromConfig(c, "allowlist")
 	require.EqualError(t, err, "config `allowlist.interfaces` values must all be the same true/false value")
 
-	c.Settings["allowlist"] = map[interface{}]interface{}{
+	c.Settings["allowlist"] = map[string]any{
-		"interfaces": map[interface{}]interface{}{
+		"interfaces": map[string]any{
 			`docker.*`: false,
 		},
 	}

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

@@ -41,7 +41,7 @@ type detailsV1 struct {
 	curve Curve
 }
 
-type m map[string]interface{}
+type m = map[string]any
 
 func (c *certificateV1) Version() Version {
 	return Version1
@@ -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/crypto_test.go

@@ -10,14 +10,14 @@ import (
 
 func TestNewArgon2Parameters(t *testing.T) {
 	p := NewArgon2Parameters(64*1024, 4, 3)
-	assert.EqualValues(t, &Argon2Parameters{
+	assert.Equal(t, &Argon2Parameters{
 		version:     argon2.Version,
 		Memory:      64 * 1024,
 		Parallelism: 4,
 		Iterations:  3,
 	}, p)
 	p = NewArgon2Parameters(2*1024*1024, 2, 1)
-	assert.EqualValues(t, &Argon2Parameters{
+	assert.Equal(t, &Argon2Parameters{
 		version:     argon2.Version,
 		Memory:      2 * 1024 * 1024,
 		Parallelism: 2,
@@ -26,21 +26,21 @@ func TestNewArgon2Parameters(t *testing.T) {
 }
 
 func TestDecryptAndUnmarshalSigningPrivateKey(t *testing.T) {
-	passphrase := []byte("DO NOT USE THIS KEY")
+	passphrase := []byte("DO NOT USE")
 	privKey := []byte(`# A good key
 -----BEGIN NEBULA ED25519 ENCRYPTED PRIVATE KEY-----
-CjwKC0FFUy0yNTYtR0NNEi0IExCAgIABGAEgBCognnjujd67Vsv99p22wfAjQaDT
+CjsKC0FFUy0yNTYtR0NNEiwIExCAgAQYAyAEKiCPoDfGQiosxNPTbPn5EsMlc2MI
-oCMW1mdjkU3gACKNW4MSXOWR9Sts4C81yk1RUku2gvGKs3TB9LYoklLsIizSYOLl
+c0Bt4oz6gTrFQhX3aBJcimhHKeAuhyTGvllD0Z19fe+DFPcLH3h5VrdjVfIAajg0
-+Vs//O1T0I1Xbml2XBAROsb/VSoDln/6LMqR4B6fn6B3GOsLBBqRI8daDl9lRMPB
+KrbV3n9UHif/Au5skWmquNJzoW1E4MTdRbvpti6o+WdQ49DxjBFhx0YH8LBqrbPU
-qrlJ69wer3ZUHFXA
+0BGkUHmIO7daP24=
 -----END NEBULA ED25519 ENCRYPTED PRIVATE KEY-----
 `)
 	shortKey := []byte(`# A key which, once decrypted, is too short
 -----BEGIN NEBULA ED25519 ENCRYPTED PRIVATE KEY-----
-CjwKC0FFUy0yNTYtR0NNEi0IExCAgIABGAEgBCoga5h8owMEBWRSMMJKzuUvWce7
+CjsKC0FFUy0yNTYtR0NNEiwIExCAgAQYAyAEKiAVJwdfl3r+eqi/vF6S7OMdpjfo
-k0qlBkQmCxiuLh80MuASW70YcKt8jeEIS2axo2V6zAKA9TSMcCsJW1kDDXEtL/xe
+hAzmTCRnr58Su4AqmBJbCv3zleYCEKYJP6UI3S8ekLMGISsgO4hm5leukCCyqT0Z
-GLF5T7sDl5COp4LU3pGxpV+KoeQ/S3gQCAAcnaOtnJQX+aSDnbO3jCHyP7U9CHbs
+cQ76yrberpzkJKoPLGisX8f+xdy4aXSZl7oEYWQte1+vqbtl/eY9PGZhxUQdcyq7
-rQr3bdH3Oy/WiYU=
+hqzIyrRqfUgVuA==
 -----END NEBULA ED25519 ENCRYPTED PRIVATE KEY-----
 `)
 	invalidBanner := []byte(`# Invalid banner (not encrypted)

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"
 	CertificateV2Banner = "NEBULA CERTIFICATE V2"
+)
+
+const ( //key-agreement-key banners
 	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"
-
-	P256PrivateKeyBanner               = "NEBULA P256 PRIVATE KEY"
-	P256PublicKeyBanner                = "NEBULA P256 PUBLIC KEY"
-	EncryptedECDSAP256PrivateKeyBanner = "NEBULA ECDSA P256 ENCRYPTED PRIVATE KEY"
-	ECDSAP256PrivateKeyBanner          = "NEBULA ECDSA P256 PRIVATE 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

cmd/nebula-cert/ca_test.go

@@ -90,26 +90,26 @@ func Test_ca(t *testing.T) {
 	assertHelpError(t, ca(
 		[]string{"-version", "1", "-out-key", "nope", "-out-crt", "nope", "duration", "100m"}, ob, eb, nopw,
 	), "-name is required")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// ipv4 only ips
 	assertHelpError(t, ca([]string{"-version", "1", "-name", "ipv6", "-ips", "100::100/100"}, ob, eb, nopw), "invalid -networks definition: v1 certificates can only be ipv4, have 100::100/100")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// ipv4 only subnets
 	assertHelpError(t, ca([]string{"-version", "1", "-name", "ipv6", "-subnets", "100::100/100"}, ob, eb, nopw), "invalid -unsafe-networks definition: v1 certificates can only be ipv4, have 100::100/100")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// failed key write
 	ob.Reset()
 	eb.Reset()
 	args := []string{"-version", "1", "-name", "test", "-duration", "100m", "-out-crt", "/do/not/write/pleasecrt", "-out-key", "/do/not/write/pleasekey"}
 	require.EqualError(t, ca(args, ob, eb, nopw), "error while writing out-key: open /do/not/write/pleasekey: "+NoSuchDirError)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// create temp key file
 	keyF, err := os.CreateTemp("", "test.key")
@@ -121,8 +121,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-name", "test", "-duration", "100m", "-out-crt", "/do/not/write/pleasecrt", "-out-key", keyF.Name()}
 	require.EqualError(t, ca(args, ob, eb, nopw), "error while writing out-crt: open /do/not/write/pleasecrt: "+NoSuchDirError)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// create temp cert file
 	crtF, err := os.CreateTemp("", "test.crt")
@@ -135,8 +135,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-name", "test", "-duration", "100m", "-groups", "1,,   2    ,        ,,,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.NoError(t, ca(args, ob, eb, nopw))
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// read cert and key files
 	rb, _ := os.ReadFile(keyF.Name())
@@ -158,7 +158,7 @@ func Test_ca(t *testing.T) {
 	assert.Empty(t, lCrt.UnsafeNetworks())
 	assert.Len(t, lCrt.PublicKey(), 32)
 	assert.Equal(t, time.Duration(time.Minute*100), lCrt.NotAfter().Sub(lCrt.NotBefore()))
-	assert.Equal(t, "", lCrt.Issuer())
+	assert.Empty(t, lCrt.Issuer())
 	assert.True(t, lCrt.CheckSignature(lCrt.PublicKey()))
 
 	// test encrypted key
@@ -169,7 +169,7 @@ func Test_ca(t *testing.T) {
 	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.NoError(t, ca(args, ob, eb, testpw))
 	assert.Equal(t, pwPromptOb, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// read encrypted key file and verify default params
 	rb, _ = os.ReadFile(keyF.Name())
@@ -197,7 +197,7 @@ func Test_ca(t *testing.T) {
 	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.Error(t, ca(args, ob, eb, errpw))
 	assert.Equal(t, pwPromptOb, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// test when user fails to enter a password
 	os.Remove(keyF.Name())
@@ -207,7 +207,7 @@ func Test_ca(t *testing.T) {
 	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.EqualError(t, ca(args, ob, eb, nopw), "no passphrase specified, remove -encrypt flag to write out-key in plaintext")
 	assert.Equal(t, strings.Repeat(pwPromptOb, 5), ob.String()) // prompts 5 times before giving up
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// create valid cert/key for overwrite tests
 	os.Remove(keyF.Name())
@@ -222,8 +222,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-name", "test", "-duration", "100m", "-groups", "1,,   2    ,        ,,,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.EqualError(t, ca(args, ob, eb, nopw), "refusing to overwrite existing CA key: "+keyF.Name())
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// test that we won't overwrite existing key file
 	os.Remove(keyF.Name())
@@ -231,8 +231,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-name", "test", "-duration", "100m", "-groups", "1,,   2    ,        ,,,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.EqualError(t, ca(args, ob, eb, nopw), "refusing to overwrite existing CA cert: "+crtF.Name())
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	os.Remove(keyF.Name())
 
 }

cmd/nebula-cert/keygen_test.go

@@ -37,20 +37,20 @@ func Test_keygen(t *testing.T) {
 
 	// required args
 	assertHelpError(t, keygen([]string{"-out-pub", "nope"}, ob, eb), "-out-key is required")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	assertHelpError(t, keygen([]string{"-out-key", "nope"}, ob, eb), "-out-pub is required")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// failed key write
 	ob.Reset()
 	eb.Reset()
 	args := []string{"-out-pub", "/do/not/write/pleasepub", "-out-key", "/do/not/write/pleasekey"}
 	require.EqualError(t, keygen(args, ob, eb), "error while writing out-key: open /do/not/write/pleasekey: "+NoSuchDirError)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// create temp key file
 	keyF, err := os.CreateTemp("", "test.key")
@@ -62,8 +62,8 @@ func Test_keygen(t *testing.T) {
 	eb.Reset()
 	args = []string{"-out-pub", "/do/not/write/pleasepub", "-out-key", keyF.Name()}
 	require.EqualError(t, keygen(args, ob, eb), "error while writing out-pub: open /do/not/write/pleasepub: "+NoSuchDirError)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// create temp pub file
 	pubF, err := os.CreateTemp("", "test.pub")
@@ -75,8 +75,8 @@ func Test_keygen(t *testing.T) {
 	eb.Reset()
 	args = []string{"-out-pub", pubF.Name(), "-out-key", keyF.Name()}
 	require.NoError(t, keygen(args, ob, eb))
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// read cert and key files
 	rb, _ := os.ReadFile(keyF.Name())

cmd/nebula-cert/main.go

@@ -17,7 +17,7 @@ func (he *helpError) Error() string {
 	return he.s
 }
 
-func newHelpErrorf(s string, v ...interface{}) error {
+func newHelpErrorf(s string, v ...any) error {
 	return &helpError{s: fmt.Sprintf(s, v...)}
 }
 

cmd/nebula-cert/print_test.go

@@ -43,16 +43,16 @@ func Test_printCert(t *testing.T) {
 
 	// no path
 	err := printCert([]string{}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	assertHelpError(t, err, "-path is required")
 
 	// no cert at path
 	ob.Reset()
 	eb.Reset()
 	err = printCert([]string{"-path", "does_not_exist"}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.EqualError(t, err, "unable to read cert; open does_not_exist: "+NoSuchFileError)
 
 	// invalid cert at path
@@ -64,8 +64,8 @@ func Test_printCert(t *testing.T) {
 
 	tf.WriteString("-----BEGIN NOPE-----")
 	err = printCert([]string{"-path", tf.Name()}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.EqualError(t, err, "error while unmarshaling cert: input did not contain a valid PEM encoded block")
 
 	// test multiple certs
@@ -155,7 +155,7 @@ func Test_printCert(t *testing.T) {
 `,
 		ob.String(),
 	)
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// test json
 	ob.Reset()
@@ -177,7 +177,7 @@ func Test_printCert(t *testing.T) {
 `,
 		ob.String(),
 	)
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 }
 
 // NewTestCaCert will generate a CA cert

cmd/nebula-cert/verify_test.go

@@ -38,19 +38,19 @@ func Test_verify(t *testing.T) {
 
 	// required args
 	assertHelpError(t, verify([]string{"-ca", "derp"}, ob, eb), "-crt is required")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	assertHelpError(t, verify([]string{"-crt", "derp"}, ob, eb), "-ca is required")
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 
 	// no ca at path
 	ob.Reset()
 	eb.Reset()
 	err := verify([]string{"-ca", "does_not_exist", "-crt", "does_not_exist"}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.EqualError(t, err, "error while reading ca: open does_not_exist: "+NoSuchFileError)
 
 	// invalid ca at path
@@ -62,8 +62,8 @@ func Test_verify(t *testing.T) {
 
 	caFile.WriteString("-----BEGIN NOPE-----")
 	err = verify([]string{"-ca", caFile.Name(), "-crt", "does_not_exist"}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.EqualError(t, err, "error while adding ca cert to pool: input did not contain a valid PEM encoded block")
 
 	// make a ca for later
@@ -76,8 +76,8 @@ func Test_verify(t *testing.T) {
 
 	// no crt at path
 	err = verify([]string{"-ca", caFile.Name(), "-crt", "does_not_exist"}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.EqualError(t, err, "unable to read crt: open does_not_exist: "+NoSuchFileError)
 
 	// invalid crt at path
@@ -89,8 +89,8 @@ func Test_verify(t *testing.T) {
 
 	certFile.WriteString("-----BEGIN NOPE-----")
 	err = verify([]string{"-ca", caFile.Name(), "-crt", certFile.Name()}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.EqualError(t, err, "error while parsing crt: input did not contain a valid PEM encoded block")
 
 	// unverifiable cert at path
@@ -106,8 +106,8 @@ func Test_verify(t *testing.T) {
 	certFile.Write(b)
 
 	err = verify([]string{"-ca", caFile.Name(), "-crt", certFile.Name()}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.ErrorIs(t, err, cert.ErrSignatureMismatch)
 
 	// verified cert at path
@@ -118,7 +118,7 @@ func Test_verify(t *testing.T) {
 	certFile.Write(b)
 
 	err = verify([]string{"-ca", caFile.Name(), "-crt", certFile.Name()}, ob, eb)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
-	assert.Equal(t, "", eb.String())
+	assert.Empty(t, eb.String())
 	require.NoError(t, err)
 }

config/config.go

@@ -17,14 +17,14 @@ import (
 
 	"dario.cat/mergo"
 	"github.com/sirupsen/logrus"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
 type C struct {
 	path        string
 	files       []string
-	Settings    map[interface{}]interface{}
+	Settings    map[string]any
-	oldSettings map[interface{}]interface{}
+	oldSettings map[string]any
 	callbacks   []func(*C)
 	l           *logrus.Logger
 	reloadLock  sync.Mutex
@@ -32,7 +32,7 @@ type C struct {
 
 func NewC(l *logrus.Logger) *C {
 	return &C{
-		Settings: make(map[interface{}]interface{}),
+		Settings: make(map[string]any),
 		l:        l,
 	}
 }
@@ -92,8 +92,8 @@ func (c *C) HasChanged(k string) bool {
 	}
 
 	var (
-		nv interface{}
+		nv any
-		ov interface{}
+		ov any
 	)
 
 	if k == "" {
@@ -147,7 +147,7 @@ func (c *C) ReloadConfig() {
 	c.reloadLock.Lock()
 	defer c.reloadLock.Unlock()
 
-	c.oldSettings = make(map[interface{}]interface{})
+	c.oldSettings = make(map[string]any)
 	for k, v := range c.Settings {
 		c.oldSettings[k] = v
 	}
@@ -167,7 +167,7 @@ func (c *C) ReloadConfigString(raw string) error {
 	c.reloadLock.Lock()
 	defer c.reloadLock.Unlock()
 
-	c.oldSettings = make(map[interface{}]interface{})
+	c.oldSettings = make(map[string]any)
 	for k, v := range c.Settings {
 		c.oldSettings[k] = v
 	}
@@ -201,7 +201,7 @@ func (c *C) GetStringSlice(k string, d []string) []string {
 		return d
 	}
 
-	rv, ok := r.([]interface{})
+	rv, ok := r.([]any)
 	if !ok {
 		return d
 	}
@@ -215,13 +215,13 @@ func (c *C) GetStringSlice(k string, d []string) []string {
 }
 
 // GetMap will get the map for k or return the default d if not found or invalid
-func (c *C) GetMap(k string, d map[interface{}]interface{}) map[interface{}]interface{} {
+func (c *C) GetMap(k string, d map[string]any) map[string]any {
 	r := c.Get(k)
 	if r == nil {
 		return d
 	}
 
-	v, ok := r.(map[interface{}]interface{})
+	v, ok := r.(map[string]any)
 	if !ok {
 		return d
 	}
@@ -243,7 +243,7 @@ func (c *C) GetInt(k string, d int) int {
 // GetUint32 will get the uint32 for k or return the default d if not found or invalid
 func (c *C) GetUint32(k string, d uint32) uint32 {
 	r := c.GetInt(k, int(d))
-	if uint64(r) > uint64(math.MaxUint32) {
+	if r < 0 || uint64(r) > uint64(math.MaxUint32) {
 		return d
 	}
 	return uint32(r)
@@ -266,6 +266,22 @@ func (c *C) GetBool(k string, d bool) bool {
 	return v
 }
 
+func AsBool(v any) (value bool, ok bool) {
+	switch x := v.(type) {
+	case bool:
+		return x, true
+	case string:
+		switch x {
+		case "y", "yes":
+			return true, true
+		case "n", "no":
+			return false, true
+		}
+	}
+
+	return false, false
+}
+
 // GetDuration will get the duration for k or return the default d if not found or invalid
 func (c *C) GetDuration(k string, d time.Duration) time.Duration {
 	r := c.GetString(k, "")
@@ -276,7 +292,7 @@ func (c *C) GetDuration(k string, d time.Duration) time.Duration {
 	return v
 }
 
-func (c *C) Get(k string) interface{} {
+func (c *C) Get(k string) any {
 	return c.get(k, c.Settings)
 }
 
@@ -284,10 +300,10 @@ func (c *C) IsSet(k string) bool {
 	return c.get(k, c.Settings) != nil
 }
 
-func (c *C) get(k string, v interface{}) interface{} {
+func (c *C) get(k string, v any) any {
 	parts := strings.Split(k, ".")
 	for _, p := range parts {
-		m, ok := v.(map[interface{}]interface{})
+		m, ok := v.(map[string]any)
 		if !ok {
 			return nil
 		}
@@ -346,7 +362,7 @@ func (c *C) addFile(path string, direct bool) error {
 }
 
 func (c *C) parseRaw(b []byte) error {
-	var m map[interface{}]interface{}
+	var m map[string]any
 
 	err := yaml.Unmarshal(b, &m)
 	if err != nil {
@@ -358,7 +374,7 @@ func (c *C) parseRaw(b []byte) error {
 }
 
 func (c *C) parse() error {
-	var m map[interface{}]interface{}
+	var m map[string]any
 
 	for _, path := range c.files {
 		b, err := os.ReadFile(path)
@@ -366,7 +382,7 @@ func (c *C) parse() error {
 			return err
 		}
 
-		var nm map[interface{}]interface{}
+		var nm map[string]any
 		err = yaml.Unmarshal(b, &nm)
 		if err != nil {
 			return err

config/config_test.go

@@ -10,7 +10,7 @@ import (
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
 func TestConfig_Load(t *testing.T) {
@@ -19,7 +19,7 @@ func TestConfig_Load(t *testing.T) {
 	// invalid yaml
 	c := NewC(l)
 	os.WriteFile(filepath.Join(dir, "01.yaml"), []byte(" invalid yaml"), 0644)
-	require.EqualError(t, c.Load(dir), "yaml: unmarshal errors:\n  line 1: cannot unmarshal !!str `invalid...` into map[interface {}]interface {}")
+	require.EqualError(t, c.Load(dir), "yaml: unmarshal errors:\n  line 1: cannot unmarshal !!str `invalid...` into map[string]interface {}")
 
 	// simple multi config merge
 	c = NewC(l)
@@ -31,8 +31,8 @@ func TestConfig_Load(t *testing.T) {
 	os.WriteFile(filepath.Join(dir, "01.yaml"), []byte("outer:\n  inner: hi"), 0644)
 	os.WriteFile(filepath.Join(dir, "02.yml"), []byte("outer:\n  inner: override\nnew: hi"), 0644)
 	require.NoError(t, c.Load(dir))
-	expected := map[interface{}]interface{}{
+	expected := map[string]any{
-		"outer": map[interface{}]interface{}{
+		"outer": map[string]any{
 			"inner": "override",
 		},
 		"new": "hi",
@@ -44,12 +44,12 @@ func TestConfig_Get(t *testing.T) {
 	l := test.NewLogger()
 	// test simple type
 	c := NewC(l)
-	c.Settings["firewall"] = map[interface{}]interface{}{"outbound": "hi"}
+	c.Settings["firewall"] = map[string]any{"outbound": "hi"}
 	assert.Equal(t, "hi", c.Get("firewall.outbound"))
 
 	// test complex type
-	inner := []map[interface{}]interface{}{{"port": "1", "code": "2"}}
+	inner := []map[string]any{{"port": "1", "code": "2"}}
-	c.Settings["firewall"] = map[interface{}]interface{}{"outbound": inner}
+	c.Settings["firewall"] = map[string]any{"outbound": inner}
 	assert.EqualValues(t, inner, c.Get("firewall.outbound"))
 
 	// test missing
@@ -59,7 +59,7 @@ func TestConfig_Get(t *testing.T) {
 func TestConfig_GetStringSlice(t *testing.T) {
 	l := test.NewLogger()
 	c := NewC(l)
-	c.Settings["slice"] = []interface{}{"one", "two"}
+	c.Settings["slice"] = []any{"one", "two"}
 	assert.Equal(t, []string{"one", "two"}, c.GetStringSlice("slice", []string{}))
 }
 
@@ -101,14 +101,14 @@ func TestConfig_HasChanged(t *testing.T) {
 	// Test key change
 	c = NewC(l)
 	c.Settings["test"] = "hi"
-	c.oldSettings = map[interface{}]interface{}{"test": "no"}
+	c.oldSettings = map[string]any{"test": "no"}
 	assert.True(t, c.HasChanged("test"))
 	assert.True(t, c.HasChanged(""))
 
 	// No key change
 	c = NewC(l)
 	c.Settings["test"] = "hi"
-	c.oldSettings = map[interface{}]interface{}{"test": "hi"}
+	c.oldSettings = map[string]any{"test": "hi"}
 	assert.False(t, c.HasChanged("test"))
 	assert.False(t, c.HasChanged(""))
 }
@@ -184,11 +184,11 @@ firewall:
 `),
 	}
 
-	var m map[any]any
+	var m map[string]any
 
 	// merge the same way config.parse() merges
 	for _, b := range configs {
-		var nm map[any]any
+		var nm map[string]any
 		err := yaml.Unmarshal(b, &nm)
 		require.NoError(t, err)
 
@@ -205,15 +205,15 @@ firewall:
 	t.Logf("Merged Config as YAML:\n%s", mYaml)
 
 	// If a bug is present, some items might be replaced instead of merged like we expect
-	expected := map[any]any{
+	expected := map[string]any{
-		"firewall": map[any]any{
+		"firewall": map[string]any{
 			"inbound": []any{
-				map[any]any{"host": "any", "port": "any", "proto": "icmp"},
+				map[string]any{"host": "any", "port": "any", "proto": "icmp"},
-				map[any]any{"groups": []any{"server"}, "port": 443, "proto": "tcp"},
+				map[string]any{"groups": []any{"server"}, "port": 443, "proto": "tcp"},
-				map[any]any{"groups": []any{"webapp"}, "port": 443, "proto": "tcp"}},
+				map[string]any{"groups": []any{"webapp"}, "port": 443, "proto": "tcp"}},
 			"outbound": []any{
-				map[any]any{"host": "any", "port": "any", "proto": "any"}}},
+				map[string]any{"host": "any", "port": "any", "proto": "any"}}},
-		"listen": map[any]any{
+		"listen": map[string]any{
 			"host": "0.0.0.0",
 			"port": 4242,
 		},

connection_manager.go

@@ -4,13 +4,16 @@ import (
 	"bytes"
 	"context"
 	"encoding/binary"
+	"fmt"
 	"net/netip"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/rcrowley/go-metrics"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 )
 
@@ -27,12 +30,6 @@ const (
 )
 
 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
@@ -40,117 +37,117 @@ type connectionManager struct {
 	hostMap      *HostMap
 	trafficTimer *LockingTimerWheel[uint32]
 	intf         *Interface
-	pendingDeletion         map[uint32]struct{}
 	punchy       *Punchy
+
+	// Configuration settings
 	checkInterval           time.Duration
 	pendingDeletionInterval time.Duration
+	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
-	}
-
-	nc := &connectionManager{
-		hostMap:                 intf.hostMap,
-		in:                      make(map[uint32]struct{}),
-		inLock:                  &sync.RWMutex{},
-		out:                     make(map[uint32]struct{}),
-		outLock:                 &sync.RWMutex{},
 		relayUsed:       make(map[uint32]struct{}),
 		relayUsedLock:   &sync.RWMutex{},
-		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,
 	}
 
-	nc.Start(ctx)
+	cm.reload(c, true)
-	return nc
+	c.RegisterReloadCallback(func(c *config.C) {
+		cm.reload(c, false)
+	})
+
+	return cm
 }
 
-func (n *connectionManager) In(localIndex uint32) {
+func (cm *connectionManager) reload(c *config.C, initial bool) {
-	n.inLock.RLock()
+	if initial {
-	// If this already exists, return
+		cm.checkInterval = time.Duration(c.GetInt("timers.connection_alive_interval", 5)) * time.Second
-	if _, ok := n.in[localIndex]; ok {
+		cm.pendingDeletionInterval = time.Duration(c.GetInt("timers.pending_deletion_interval", 10)) * time.Second
-		n.inLock.RUnlock()
+
-		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)
+	}
+
+	if initial || c.HasChanged("tunnels.inactivity_timeout") {
+		old := cm.getInactivityTimeout()
+		cm.inactivityTimeout.Store((int64)(c.GetDuration("tunnels.inactivity_timeout", 10*time.Minute)))
+		if !initial {
+			cm.l.WithField("oldDuration", old).
+				WithField("newDuration", cm.getInactivityTimeout()).
+				Info("Inactivity timeout has changed")
+		}
+	}
+
+	if initial || c.HasChanged("tunnels.drop_inactive") {
+		old := cm.dropInactive.Load()
+		cm.dropInactive.Store(c.GetBool("tunnels.drop_inactive", false))
+		if !initial {
+			cm.l.WithField("oldBool", old).
+				WithField("newBool", cm.dropInactive.Load()).
+				Info("Drop inactive setting has changed")
+		}
 	}
-	n.inLock.RUnlock()
-	n.inLock.Lock()
-	n.in[localIndex] = struct{}{}
-	n.inLock.Unlock()
 }
 
-func (n *connectionManager) Out(localIndex uint32) {
+func (cm *connectionManager) getInactivityTimeout() time.Duration {
-	n.outLock.RLock()
+	return (time.Duration)(cm.inactivityTimeout.Load())
-	// 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()
 }
 
-func (n *connectionManager) RelayUsed(localIndex uint32) {
+func (cm *connectionManager) In(h *HostInfo) {
-	n.relayUsedLock.RLock()
+	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.relayUsed[localIndex]; ok {
+	if _, ok := cm.relayUsed[localIndex]; ok {
-		n.relayUsedLock.RUnlock()
+		cm.relayUsedLock.RUnlock()
 		return
 	}
-	n.relayUsedLock.RUnlock()
+	cm.relayUsedLock.RUnlock()
-	n.relayUsedLock.Lock()
+	cm.relayUsedLock.Lock()
-	n.relayUsed[localIndex] = struct{}{}
+	cm.relayUsed[localIndex] = struct{}{}
-	n.relayUsedLock.Unlock()
+	cm.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()
 }
 
-func (n *connectionManager) Start(ctx context.Context) {
+func (cm *connectionManager) Start(ctx context.Context) {
-	go n.Run(ctx)
+	clockSource := time.NewTicker(cm.trafficTimer.t.tickDuration)
-}
-
-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()
 
 	p := []byte("")
@@ -163,61 +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)
 			}
 		}
 	}
 }
 
-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 {
@@ -227,46 +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:
+			switch existing.State {
+			case Established, PeerRequested, Disestablished:
 				// This relay already exists in newhostinfo, then do nothing.
 				continue
-		case ok && existing.State == Requested:
+			case Requested:
 				// The relay exists in a Requested state; re-send the request
 				index = existing.LocalIndex
 				switch r.Type {
 				case TerminalType:
-				relayFrom = n.intf.myVpnAddrs[0]
+					relayFrom = cm.intf.myVpnAddrs[0]
 					relayTo = existing.PeerAddr
 				case ForwardingType:
 					relayFrom = existing.PeerAddr
 					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))
 			}
 		}
 
@@ -279,12 +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
 			}
 
@@ -296,16 +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,
@@ -316,46 +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.
@@ -363,46 +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 {
-			// If we aren't sending or receiving traffic then its an unused tunnel and we don't to test the tunnel.
+			inactiveFor, isInactive := cm.isInactive(hostinfo, now)
-			// Just maintain NAT state if configured to do so.
+			if isInactive {
-			n.sendPunch(hostinfo)
+				// Tunnel is inactive, tear it down
-			n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
+				hostinfo.logger(cm.l).
-			return doNothing, nil, nil
+					WithField("inactiveDuration", inactiveFor).
+					WithField("primary", mainHostInfo).
+					Info("Dropping tunnel due to inactivity")
 
+				return closeTunnel, hostinfo, primary
 			}
 
-		if n.punchy.GetTargetEverything() {
+			// 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.
+			cm.sendPunch(hostinfo)
+			cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
+			return doNothing, nil, nil
+		}
+
+		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")
 		}
@@ -411,17 +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.
@@ -429,83 +455,90 @@ 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.defaultVersion && bytes.Equal(curCrt.Signature(), myCrt.Signature()) == true {
+	if curCrt.Version() >= cs.initiatingVersion && bytes.Equal(curCrt.Signature(), myCrt.Signature()) == true {
 		// The current tunnel is using the latest certificate and version, no need to rehandshake.
 		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)
@@ -44,7 +43,7 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	hostMap.preferredRanges.Store(&preferredRanges)
 
 	cs := &CertState{
-		defaultVersion:   cert.Version1,
+		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
 		v1HandshakeBytes:  []byte{},
@@ -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)
 }
 
@@ -126,7 +126,7 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	hostMap.preferredRanges.Store(&preferredRanges)
 
 	cs := &CertState{
-		defaultVersion:   cert.Version1,
+		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
 		v1HandshakeBytes:  []byte{},
@@ -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
 }

connection_state.go

@@ -27,7 +27,7 @@ type ConnectionState struct {
 	writeLock      sync.Mutex
 }
 
-func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, initiator bool, pattern noise.HandshakePattern, psk []byte) (*ConnectionState, error) {
+func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, initiator bool, pattern noise.HandshakePattern) (*ConnectionState, error) {
 	var dhFunc noise.DHFunc
 	switch crt.Curve() {
 	case cert.Curve_CURVE25519:
@@ -61,7 +61,8 @@ func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, i
 		Pattern:       pattern,
 		Initiator:     initiator,
 		StaticKeypair: static,
-		PresharedKey:          psk,
+		//NOTE: These should come from CertState (pki.go) when we finally implement it
+		PresharedKey:          []byte{},
 		PresharedKeyPlacement: 0,
 	})
 	if err != nil {

control.go

@@ -34,6 +34,7 @@ type Control struct {
 	statsStart             func()
 	dnsStart               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()
 	}
@@ -131,8 +135,7 @@ func (c *Control) ListHostmapIndexes(pendingMap bool) []ControlHostInfo {
 
 // GetCertByVpnIp returns the authenticated certificate of the given vpn IP, or nil if not found
 func (c *Control) GetCertByVpnIp(vpnIp netip.Addr) cert.Certificate {
-	_, found := c.f.myVpnAddrsTable.Lookup(vpnIp)
+	if c.f.myVpnAddrsTable.Contains(vpnIp) {
-	if found {
 		// Only returning the default certificate since its impossible
 		// for any other host but ourselves to have more than 1
 		return c.f.pki.getCertState().GetDefaultCertificate().Copy()

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{},
 		},
@@ -101,7 +101,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 
 	// Make sure we don't have any unexpected fields
 	assertFields(t, []string{"VpnAddrs", "LocalIndex", "RemoteIndex", "RemoteAddrs", "Cert", "MessageCounter", "CurrentRemote", "CurrentRelaysToMe", "CurrentRelaysThroughMe"}, thi)
-	assert.EqualValues(t, &expectedInfo, thi)
+	assert.Equal(t, &expectedInfo, thi)
 	test.AssertDeepCopyEqual(t, &expectedInfo, thi)
 
 	// Make sure we don't panic if the host info doesn't have a cert yet
@@ -110,7 +110,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 	})
 }
 
-func assertFields(t *testing.T, expected []string, actualStruct interface{}) {
+func assertFields(t *testing.T, expected []string, actualStruct any) {
 	val := reflect.ValueOf(actualStruct).Elem()
 	fields := make([]string, val.NumField())
 	for i := 0; i < val.NumField(); i++ {

dns_server.go

@@ -26,7 +26,7 @@ type dnsRecords struct {
 	dnsMap4         map[string]netip.Addr
 	dnsMap6         map[string]netip.Addr
 	hostMap         *HostMap
-	myVpnAddrsTable *bart.Table[struct{}]
+	myVpnAddrsTable *bart.Lite
 }
 
 func newDnsRecords(l *logrus.Logger, cs *CertState, hostMap *HostMap) *dnsRecords {
@@ -112,8 +112,8 @@ func (d *dnsRecords) isSelfNebulaOrLocalhost(addr string) bool {
 		return true
 	}
 
-	_, found := d.myVpnAddrsTable.Lookup(b)
+	//if we found it in this table, it's good
-	return found //if we found it in this table, it's good
+	return d.myVpnAddrsTable.Contains(b)
 }
 
 func (d *dnsRecords) parseQuery(m *dns.Msg, w dns.ResponseWriter) {

dns_server_test.go

@@ -38,24 +38,24 @@ func TestParsequery(t *testing.T) {
 func Test_getDnsServerAddr(t *testing.T) {
 	c := config.NewC(nil)
 
-	c.Settings["lighthouse"] = map[interface{}]interface{}{
+	c.Settings["lighthouse"] = map[string]any{
-		"dns": map[interface{}]interface{}{
+		"dns": map[string]any{
 			"host": "0.0.0.0",
 			"port": "1",
 		},
 	}
 	assert.Equal(t, "0.0.0.0:1", getDnsServerAddr(c))
 
-	c.Settings["lighthouse"] = map[interface{}]interface{}{
+	c.Settings["lighthouse"] = map[string]any{
-		"dns": map[interface{}]interface{}{
+		"dns": map[string]any{
 			"host": "::",
 			"port": "1",
 		},
 	}
 	assert.Equal(t, "[::]:1", getDnsServerAddr(c))
 
-	c.Settings["lighthouse"] = map[interface{}]interface{}{
+	c.Settings["lighthouse"] = map[string]any{
-		"dns": map[interface{}]interface{}{
+		"dns": map[string]any{
 			"host": "[::]",
 			"port": "1",
 		},
@@ -63,8 +63,8 @@ func Test_getDnsServerAddr(t *testing.T) {
 	assert.Equal(t, "[::]:1", getDnsServerAddr(c))
 
 	// Make sure whitespace doesn't mess us up
-	c.Settings["lighthouse"] = map[interface{}]interface{}{
+	c.Settings["lighthouse"] = map[string]any{
-		"dns": map[interface{}]interface{}{
+		"dns": map[string]any{
 			"host": "[::] ",
 			"port": "1",
 		},

e2e/handshakes_test.go

@@ -20,7 +20,7 @@ import (
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
 func BenchmarkHotPath(b *testing.B) {
@@ -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 {
@@ -991,7 +991,7 @@ func TestRehandshaking(t *testing.T) {
 	require.NoError(t, err)
 	var theirNewConfig m
 	require.NoError(t, yaml.Unmarshal(rc, &theirNewConfig))
-	theirFirewall := theirNewConfig["firewall"].(map[interface{}]interface{})
+	theirFirewall := theirNewConfig["firewall"].(map[string]any)
 	theirFirewall["inbound"] = []m{{
 		"proto": "any",
 		"port":  "any",
@@ -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")
@@ -1087,7 +1090,7 @@ func TestRehandshakingLoser(t *testing.T) {
 	require.NoError(t, err)
 	var myNewConfig m
 	require.NoError(t, yaml.Unmarshal(rc, &myNewConfig))
-	theirFirewall := myNewConfig["firewall"].(map[interface{}]interface{})
+	theirFirewall := myNewConfig["firewall"].(map[string]any)
 	theirFirewall["inbound"] = []m{{
 		"proto": "any",
 		"port":  "any",
@@ -1224,135 +1227,3 @@ func TestV2NonPrimaryWithLighthouse(t *testing.T) {
 	myControl.Stop()
 	theirControl.Stop()
 }
-
-func TestPSK(t *testing.T) {
-	tests := []struct {
-		name         string
-		myPskMode    nebula.PskMode
-		theirPskMode nebula.PskMode
-	}{
-		// All accepting
-		{
-			name:         "both accepting",
-			myPskMode:    nebula.PskAccepting,
-			theirPskMode: nebula.PskAccepting,
-		},
-
-		// accepting and sending both ways
-		{
-			name:         "accepting to sending",
-			myPskMode:    nebula.PskAccepting,
-			theirPskMode: nebula.PskSending,
-		},
-		{
-			name:         "sending to accepting",
-			myPskMode:    nebula.PskSending,
-			theirPskMode: nebula.PskAccepting,
-		},
-
-		// All sending
-		{
-			name:         "sending to sending",
-			myPskMode:    nebula.PskSending,
-			theirPskMode: nebula.PskSending,
-		},
-
-		// enforced and sending both ways
-		{
-			name:         "enforced to sending",
-			myPskMode:    nebula.PskEnforced,
-			theirPskMode: nebula.PskSending,
-		},
-		{
-			name:         "sending to enforced",
-			myPskMode:    nebula.PskSending,
-			theirPskMode: nebula.PskEnforced,
-		},
-
-		// All enforced
-		{
-			name:         "both enforced",
-			myPskMode:    nebula.PskEnforced,
-			theirPskMode: nebula.PskEnforced,
-		},
-
-		// Enforced can technically handshake with an accepting node, but it is bad to be in this state
-		{
-			name:         "enforced to accepting",
-			myPskMode:    nebula.PskEnforced,
-			theirPskMode: nebula.PskAccepting,
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			var myPskSettings, theirPskSettings m
-
-			switch test.myPskMode {
-			case nebula.PskAccepting:
-				myPskSettings = m{"psk": &m{"mode": "accepting", "keys": []string{"garbage0", "this is a key"}}}
-			case nebula.PskSending:
-				myPskSettings = m{"psk": &m{"mode": "sending", "keys": []string{"this is a key", "garbage1"}}}
-			case nebula.PskEnforced:
-				myPskSettings = m{"psk": &m{"mode": "enforced", "keys": []string{"this is a key", "garbage2"}}}
-			}
-
-			switch test.theirPskMode {
-			case nebula.PskAccepting:
-				theirPskSettings = m{"psk": &m{"mode": "accepting", "keys": []string{"garbage3", "this is a key"}}}
-			case nebula.PskSending:
-				theirPskSettings = m{"psk": &m{"mode": "sending", "keys": []string{"this is a key", "garbage4"}}}
-			case nebula.PskEnforced:
-				theirPskSettings = m{"psk": &m{"mode": "enforced", "keys": []string{"this is a key", "garbage5"}}}
-			}
-
-			ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, nil)
-			myControl, myVpnIp, myUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "me", "10.0.0.1/24", myPskSettings)
-			theirControl, theirVpnIp, theirUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "them", "10.0.0.2/24", theirPskSettings)
-
-			myControl.InjectLightHouseAddr(theirVpnIp[0].Addr(), theirUdpAddr)
-			r := router.NewR(t, myControl, theirControl)
-
-			// Start the servers
-			myControl.Start()
-			theirControl.Start()
-
-			t.Log("Route until we see our cached packet flow")
-			myControl.InjectTunUDPPacket(theirVpnIp[0].Addr(), 80, myVpnIp[0].Addr(), 80, []byte("Hi from me"))
-			r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
-				h := &header.H{}
-				err := h.Parse(p.Data)
-				if err != nil {
-					panic(err)
-				}
-
-				// If this is the stage 1 handshake packet and I am configured to send with a psk, my cert name should
-				// not appear. It would likely be more obvious to unmarshal the payload and check but this works fine for now
-				if test.myPskMode == nebula.PskEnforced || test.myPskMode == nebula.PskSending {
-					if h.Type == 0 && h.MessageCounter == 1 {
-						assert.NotContains(t, string(p.Data), "test me")
-					}
-				}
-
-				if p.To == theirUdpAddr && h.Type == 1 {
-					return router.RouteAndExit
-				}
-
-				return router.KeepRouting
-			})
-
-			t.Log("My cached packet should be received by them")
-			myCachedPacket := theirControl.GetFromTun(true)
-			assertUdpPacket(t, []byte("Hi from me"), myCachedPacket, myVpnIp[0].Addr(), theirVpnIp[0].Addr(), 80, 80)
-
-			t.Log("Test the tunnel with them")
-			assertHostInfoPair(t, myUdpAddr, theirUdpAddr, myVpnIp, theirVpnIp, myControl, theirControl)
-			assertTunnel(t, myVpnIp[0].Addr(), theirVpnIp[0].Addr(), myControl, theirControl, r)
-
-			myControl.Stop()
-			theirControl.Stop()
-			//TODO: assert hostmaps
-		})
-	}
-
-}

e2e/helpers_test.go

@@ -22,15 +22,13 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/e2e/router"
 	"github.com/stretchr/testify/assert"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
-type m map[string]interface{}
+type m = map[string]any
 
 // newSimpleServer creates a nebula instance with many assumptions
 func newSimpleServer(v cert.Version, caCrt cert.Certificate, caKey []byte, name string, sVpnNetworks string, overrides m) (*nebula.Control, []netip.Prefix, netip.AddrPort, *config.C) {
-	l := NewTestLogger()
-
 	var vpnNetworks []netip.Prefix
 	for _, sn := range strings.Split(sVpnNetworks, ",") {
 		vpnIpNet, err := netip.ParsePrefix(strings.TrimSpace(sn))
@@ -56,6 +54,25 @@ func newSimpleServer(v cert.Version, caCrt cert.Certificate, caKey []byte, name
 		budpIp[3] = 239
 		udpAddr = netip.AddrPortFrom(netip.AddrFrom16(budpIp), 4242)
 	}
+	return newSimpleServerWithUdp(v, caCrt, caKey, name, sVpnNetworks, udpAddr, overrides)
+}
+
+func newSimpleServerWithUdp(v cert.Version, caCrt cert.Certificate, caKey []byte, name string, sVpnNetworks string, udpAddr netip.AddrPort, overrides m) (*nebula.Control, []netip.Prefix, netip.AddrPort, *config.C) {
+	l := NewTestLogger()
+
+	var vpnNetworks []netip.Prefix
+	for _, sn := range strings.Split(sVpnNetworks, ",") {
+		vpnIpNet, err := netip.ParsePrefix(strings.TrimSpace(sn))
+		if err != nil {
+			panic(err)
+		}
+		vpnNetworks = append(vpnNetworks, vpnIpNet)
+	}
+
+	if len(vpnNetworks) == 0 {
+		panic("no vpn networks")
+	}
+
 	_, _, myPrivKey, myPEM := cert_test.NewTestCert(v, cert.Curve_CURVE25519, caCrt, caKey, name, time.Now(), time.Now().Add(5*time.Minute), vpnNetworks, nil, []string{})
 
 	caB, err := caCrt.MarshalPEM()

e2e/router/router.go

@@ -111,6 +111,10 @@ type ExitFunc func(packet *udp.Packet, receiver *nebula.Control) ExitType
 func NewR(t testing.TB, controls ...*nebula.Control) *R {
 	ctx, cancel := context.WithCancel(context.Background())
 
+	if err := os.MkdirAll("mermaid", 0755); err != nil {
+		panic(err)
+	}
+
 	r := &R{
 		controls:     make(map[netip.AddrPort]*nebula.Control),
 		vpnControls:  make(map[netip.Addr]*nebula.Control),
@@ -190,9 +194,6 @@ func (r *R) renderFlow() {
 		return
 	}
 
-	if err := os.MkdirAll(filepath.Dir(r.fn), 0755); err != nil {
-		panic(err)
-	}
 	f, err := os.OpenFile(r.fn, os.O_CREATE|os.O_TRUNC|os.O_RDWR, 0644)
 	if err != nil {
 		panic(err)
@@ -699,6 +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 +1,105 @@
+//go:build e2e_testing
+// +build e2e_testing
+
+package e2e
+
+import (
+	"net/netip"
+	"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()
+}
+
+func TestCrossStackRelaysWork(t *testing.T) {
+	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
+	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "me     ", "10.128.0.1/24,fc00::1/64", m{"relay": m{"use_relays": true}})
+	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "relay  ", "10.128.0.128/24,fc00::128/64", m{"relay": m{"am_relay": true}})
+	theirUdp := netip.MustParseAddrPort("10.0.0.2:4242")
+	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServerWithUdp(cert.Version2, ca, caKey, "them   ", "fc00::2/64", theirUdp, m{"relay": m{"use_relays": true}})
+
+	//myVpnV4 := myVpnIpNet[0]
+	myVpnV6 := myVpnIpNet[1]
+	relayVpnV4 := relayVpnIpNet[0]
+	relayVpnV6 := relayVpnIpNet[1]
+	theirVpnV6 := theirVpnIpNet[0]
+
+	// Teach my how to get to the relay and that their can be reached via the relay
+	myControl.InjectLightHouseAddr(relayVpnV4.Addr(), relayUdpAddr)
+	myControl.InjectLightHouseAddr(relayVpnV6.Addr(), relayUdpAddr)
+	myControl.InjectRelays(theirVpnV6.Addr(), []netip.Addr{relayVpnV6.Addr()})
+	relayControl.InjectLightHouseAddr(theirVpnV6.Addr(), theirUdpAddr)
+
+	// Build a router so we don't have to reason who gets which packet
+	r := router.NewR(t, myControl, relayControl, theirControl)
+	defer r.RenderFlow()
+
+	// Start the servers
+	myControl.Start()
+	relayControl.Start()
+	theirControl.Start()
+
+	t.Log("Trigger a handshake from me to them via the relay")
+	myControl.InjectTunUDPPacket(theirVpnV6.Addr(), 80, myVpnV6.Addr(), 80, []byte("Hi from me"))
+
+	p := r.RouteForAllUntilTxTun(theirControl)
+	r.Log("Assert the tunnel works")
+	assertUdpPacket(t, []byte("Hi from me"), p, myVpnV6.Addr(), theirVpnV6.Addr(), 80, 80)
+
+	t.Log("reply?")
+	theirControl.InjectTunUDPPacket(myVpnV6.Addr(), 80, theirVpnV6.Addr(), 80, []byte("Hi from them"))
+	p = r.RouteForAllUntilTxTun(myControl)
+	assertUdpPacket(t, []byte("Hi from them"), p, theirVpnV6.Addr(), myVpnV6.Addr(), 80, 80)
+
+	r.RenderHostmaps("Final hostmaps", myControl, relayControl, theirControl)
+	//t.Log("finish up")
+	//myControl.Stop()
+	//theirControl.Stop()
+	//relayControl.Stop()
+}

examples/config.yml

@@ -13,43 +13,11 @@ pki:
   # disconnect_invalid is a toggle to force a client to be disconnected if the certificate is expired or invalid.
   #disconnect_invalid: true
 
-  # default_version controls which certificate version is used in handshakes.
+  # initiating_version controls which certificate version is used when initiating handshakes.
   # This setting only applies if both a v1 and a v2 certificate are configured, in which case it will default to `1`.
   # Once all hosts in the mesh are configured with both a v1 and v2 certificate then this should be changed to `2`.
   # After all hosts in the mesh are using a v2 certificate then v1 certificates are no longer needed.
-  # default_version: 1
+  # initiating_version: 1
-
-  # psk can be used to mask the contents of handshakes.
-  psk:
-    # `mode` defines how the pre shared keys can be used in a handshake.
-    # `accepting` (the default) will initiate handshakes using an empty key and will try to use any keys provided when
-    # receiving handshakes, including an empty key.
-    # `sending` will initiate handshakes with the first key provided and will try to use any keys provided when
-    # receiving handshakes, including an empty key.
-    # `enforced` will initiate handshakes with the first psk key provided and will try to use any keys provided when
-    # responding to handshakes. An empty key will not be allowed.
-    #
-    # To change a mesh from not using a psk to enforcing psk:
-    # 1. Leave `mode` as `accepting` and configure `psk.keys` to match on all nodes in the mesh and reload.
-    # 2. Change `mode` to `sending` on all nodes in the mesh and reload.
-    # 3. Change `mode` to `enforced` on all nodes in the mesh and reload.
-    #mode: accepting
-
-    # The keys provided are sent through hkdf to ensure the shared secret used in the noise protocol is the
-    # correct byte length.
-    #
-    # Only the first key is used for outbound handshakes but all keys provided will be tried in the order specified, on
-    # incoming handshakes. This is to allow for psk rotation.
-    #
-    # To rotate a primary key:
-    # 1. Put the new key in the 2nd slot on every node in the mesh and reload.
-    # 2. Move the key from the 2nd slot to the 1st slot, the old primary key is now in the 2nd slot, reload.
-    # 3. Remove the old primary key once it is no longer in use on every node in the mesh and reload.
-    #keys:
-    # - shared secret string, this one is used in all outbound handshakes # This is the primary key used when sending handshakes
-    # - this is a fallback key, received handshakes can use this
-    # - another fallback, received handshakes can use this one too
-    # - "\x68\x65\x6c\x6c\x6f\x20\x66\x72\x69\x65\x6e\x64\x73" # for raw bytes if you desire
 
 # The static host map defines a set of hosts with fixed IP addresses on the internet (or any network).
 # A host can have multiple fixed IP addresses defined here, and nebula will try each when establishing a tunnel.
@@ -271,7 +239,28 @@ tun:
 
   # Unsafe routes allows you to route traffic over nebula to non-nebula nodes
   # Unsafe routes should be avoided unless you have hosts/services that cannot run nebula
-  # NOTE: The nebula certificate of the "via" node *MUST* have the "route" defined as a subnet in its certificate
+  # Supports weighted ECMP if you define a list of gateways, this can be used for load balancing or redundancy to hosts outside of nebula
+  # NOTES:
+  # * You will only see a single gateway in the routing table if you are not on linux
+  # * If a gateway is not reachable through the overlay another gateway will be selected to send the traffic through, ignoring weights
+  #
+  # unsafe_routes:
+  # # Multiple gateways without defining a weight defaults to a weight of 1, this will balance traffic equally between the three gateways
+  # - route: 192.168.87.0/24
+  #   via:
+  #     - gateway: 10.0.0.1
+  #     - gateway: 10.0.0.2
+  #     - gateway: 10.0.0.3
+  # # Multiple gateways with a weight, this will balance traffic accordingly
+  # - route: 192.168.87.0/24
+  #   via:
+  #     - gateway: 10.0.0.1
+  #       weight: 10
+  #     - gateway: 10.0.0.2
+  #       weight: 5
+  #
+  # NOTE: The nebula certificate of the "via" node(s) *MUST* have the "route" defined as a subnet in its certificate
+  # `via`: single node or list of gateways to use for this route
   # `mtu`: will default to tun mtu if this option is not specified
   # `metric`: will default to 0 if this option is not specified
   # `install`: will default to true, controls whether this route is installed in the systems routing table.
@@ -286,6 +275,10 @@ tun:
   # On linux only, set to true to manage unsafe routes directly on the system route table with gateway routes instead of
   # in nebula configuration files. Default false, not reloadable.
   #use_system_route_table: false
+  # Buffer size for reading routes updates. 0 means default system buffer size. (/proc/sys/net/core/rmem_default).
+  # If using massive routes updates, for example BGP, you may need to increase this value to avoid packet loss.
+  # SO_RCVBUFFORCE is used to avoid having to raise the system wide max
+  #use_system_route_table_buffer_size: 0
 
 # Configure logging level
 logging:
@@ -345,6 +338,19 @@ 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:
   # Action to take when a packet is not allowed by the firewall rules.
@@ -356,11 +362,11 @@ firewall:
   outbound_action: drop
   inbound_action: drop
 
-  # Controls the default value for local_cidr. Default is true, will be deprecated after v1.9 and defaulted to false.
+  # THIS FLAG IS DEPRECATED AND WILL BE REMOVED IN A FUTURE RELEASE. (Defaults to false.)
-  # This setting only affects nebula hosts with subnets encoded in their certificate. A nebula host acting as an
+  # This setting only affects nebula hosts exposing unsafe_routes. When set to false, each inbound rule must contain a
-  # unsafe router with `default_local_cidr_any: true` will expose their unsafe routes to every inbound rule regardless
+  # `local_cidr` if the intention is to allow traffic to flow to an unsafe route. When set to true, every firewall rule
-  # of the actual destination for the packet. Setting this to false requires each inbound rule to contain a `local_cidr`
+  # will apply to all configured unsafe_routes regardless of the actual destination of the packet, unless `local_cidr`
-  # if the intention is to allow traffic to flow to an unsafe route.
+  # is explicitly defined. This is usually not the desired behavior and should be avoided!
   #default_local_cidr_any: false
 
   conntrack:
@@ -378,11 +384,9 @@ firewall:
   #   group: `any` or a literal group name, ie `default-group`
   #   groups: Same as group but accepts a list of values. Multiple values are AND'd together and a certificate would have to contain all groups to pass
   #   cidr: a remote CIDR, `0.0.0.0/0` is any ipv4 and `::/0` is any ipv6.
-  #   local_cidr: a local CIDR, `0.0.0.0/0` is any ipv4 and `::/0` is any ipv6. This could be used to filter destinations when using unsafe_routes.
+  #   local_cidr: a local CIDR, `0.0.0.0/0` is any ipv4 and `::/0` is any ipv6. This can be used to filter destinations when using unsafe_routes.
-  #     If no unsafe networks are present in the certificate(s) or `default_local_cidr_any` is true then the default is any ipv4 or ipv6 network.
+  #     By default, this is set to only the VPN (overlay) networks assigned via the certificate networks field unless `default_local_cidr_any` is set to true.
-  #     Otherwise the default is any vpn network assigned to via the certificate.
+  #     If there are unsafe_routes present in this config file, `local_cidr` should be set appropriately for the intended us case.
-  #     `default_local_cidr_any` defaults to false and is deprecated, it will be removed in a future release.
-  #     If there are unsafe routes present its best to set `local_cidr` to whatever best fits the situation.
   #   ca_name: An issuing CA name
   #   ca_sha: An issuing CA shasum
 

examples/go_service/main.go

@@ -5,8 +5,12 @@ import (
 	"fmt"
 	"log"
 	"net"
+	"os"
 
+	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/service"
 )
 
@@ -59,7 +63,16 @@ pki:
 	if err := cfg.LoadString(configStr); err != nil {
 		return err
 	}
-	svc, err := service.New(&cfg)
+
+	logger := logrus.New()
+	logger.Out = os.Stdout
+
+	ctrl, err := nebula.Main(&cfg, false, "custom-app", logger, overlay.NewUserDeviceFromConfig)
+	if err != nil {
+		return err
+	}
+
+	svc, err := service.New(ctrl)
 	if err != nil {
 		return err
 	}

firewall.go

@@ -53,7 +53,7 @@ type Firewall struct {
 
 	// routableNetworks describes the vpn addresses as well as any unsafe networks issued to us in the certificate.
 	// The vpn addresses are a full bit match while the unsafe networks only match the prefix
-	routableNetworks *bart.Table[struct{}]
+	routableNetworks *bart.Lite
 
 	// assignedNetworks is a list of vpn networks assigned to us in the certificate.
 	assignedNetworks  []netip.Prefix
@@ -125,7 +125,7 @@ type firewallPort map[int32]*FirewallCA
 
 type firewallLocalCIDR struct {
 	Any       bool
-	LocalCIDR *bart.Table[struct{}]
+	LocalCIDR *bart.Lite
 }
 
 // NewFirewall creates a new Firewall object. A TimerWheel is created for you from the provided timeouts.
@@ -148,17 +148,17 @@ func NewFirewall(l *logrus.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.D
 		tmax = defaultTimeout
 	}
 
-	routableNetworks := new(bart.Table[struct{}])
+	routableNetworks := new(bart.Lite)
 	var assignedNetworks []netip.Prefix
 	for _, network := range c.Networks() {
 		nprefix := netip.PrefixFrom(network.Addr(), network.Addr().BitLen())
-		routableNetworks.Insert(nprefix, struct{}{})
+		routableNetworks.Insert(nprefix)
 		assignedNetworks = append(assignedNetworks, network)
 	}
 
 	hasUnsafeNetworks := false
 	for _, n := range c.UnsafeNetworks() {
-		routableNetworks.Insert(n, struct{}{})
+		routableNetworks.Insert(n)
 		hasUnsafeNetworks = true
 	}
 
@@ -331,7 +331,7 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
 		return nil
 	}
 
-	rs, ok := r.([]interface{})
+	rs, ok := r.([]any)
 	if !ok {
 		return fmt.Errorf("%s failed to parse, should be an array of rules", table)
 	}
@@ -417,6 +417,8 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
 	return nil
 }
 
+var ErrUnknownNetworkType = errors.New("unknown network type")
+var ErrPeerRejected = errors.New("remote IP is not within a subnet that we handle")
 var ErrInvalidRemoteIP = errors.New("remote IP is not in remote certificate subnets")
 var ErrInvalidLocalIP = errors.New("local IP is not in list of handled local IPs")
 var ErrNoMatchingRule = errors.New("no matching rule in firewall table")
@@ -429,24 +431,35 @@ func (f *Firewall) Drop(fp firewall.Packet, incoming bool, h *HostInfo, caPool *
 		return nil
 	}
 
-	// Make sure remote address matches nebula certificate
+	// Make sure remote address matches nebula certificate, and determine how to treat it
-	if h.networks != nil {
+	if h.networks == nil {
-		_, ok := h.networks.Lookup(fp.RemoteAddr)
-		if !ok {
-			f.metrics(incoming).droppedRemoteAddr.Inc(1)
-			return ErrInvalidRemoteIP
-		}
-	} else {
 		// Simple case: Certificate has one address and no unsafe networks
 		if h.vpnAddrs[0] != fp.RemoteAddr {
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
 			return ErrInvalidRemoteIP
 		}
+	} else {
+		nwType, ok := h.networks.Lookup(fp.RemoteAddr)
+		if !ok {
+			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			return ErrInvalidRemoteIP
+		}
+		switch nwType {
+		case NetworkTypeVPN:
+			break // nothing special
+		case NetworkTypeVPNPeer:
+			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			return ErrPeerRejected // reject for now, one day this may have different FW rules
+		case NetworkTypeUnsafe:
+			break // nothing special, one day this may have different FW rules
+		default:
+			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			return ErrUnknownNetworkType //should never happen
+		}
 	}
 
 	// Make sure we are supposed to be handling this local ip address
-	_, ok := f.routableNetworks.Lookup(fp.LocalAddr)
+	if !f.routableNetworks.Contains(fp.LocalAddr) {
-	if !ok {
 		f.metrics(incoming).droppedLocalAddr.Inc(1)
 		return ErrInvalidLocalIP
 	}
@@ -752,7 +765,7 @@ func (fc *FirewallCA) match(p firewall.Packet, c *cert.CachedCertificate, caPool
 func (fr *FirewallRule) addRule(f *Firewall, groups []string, host string, ip, localCIDR netip.Prefix) error {
 	flc := func() *firewallLocalCIDR {
 		return &firewallLocalCIDR{
-			LocalCIDR: new(bart.Table[struct{}]),
+			LocalCIDR: new(bart.Lite),
 		}
 	}
 
@@ -879,7 +892,7 @@ func (flc *firewallLocalCIDR) addRule(f *Firewall, localIp netip.Prefix) error {
 		}
 
 		for _, network := range f.assignedNetworks {
-			flc.LocalCIDR.Insert(network, struct{}{})
+			flc.LocalCIDR.Insert(network)
 		}
 		return nil
 
@@ -888,7 +901,7 @@ func (flc *firewallLocalCIDR) addRule(f *Firewall, localIp netip.Prefix) error {
 		return nil
 	}
 
-	flc.LocalCIDR.Insert(localIp, struct{}{})
+	flc.LocalCIDR.Insert(localIp)
 	return nil
 }
 
@@ -901,8 +914,7 @@ func (flc *firewallLocalCIDR) match(p firewall.Packet, c *cert.CachedCertificate
 		return true
 	}
 
-	_, ok := flc.LocalCIDR.Lookup(p.LocalAddr)
+	return flc.LocalCIDR.Contains(p.LocalAddr)
-	return ok
 }
 
 type rule struct {
@@ -918,15 +930,15 @@ type rule struct {
 	CASha     string
 }
 
-func convertRule(l *logrus.Logger, p interface{}, table string, i int) (rule, error) {
+func convertRule(l *logrus.Logger, p any, table string, i int) (rule, error) {
 	r := rule{}
 
-	m, ok := p.(map[interface{}]interface{})
+	m, ok := p.(map[string]any)
 	if !ok {
 		return r, errors.New("could not parse rule")
 	}
 
-	toString := func(k string, m map[interface{}]interface{}) string {
+	toString := func(k string, m map[string]any) string {
 		v, ok := m[k]
 		if !ok {
 			return ""
@@ -944,7 +956,7 @@ func convertRule(l *logrus.Logger, p interface{}, table string, i int) (rule, er
 	r.CASha = toString("ca_sha", m)
 
 	// Make sure group isn't an array
-	if v, ok := m["group"].([]interface{}); ok {
+	if v, ok := m["group"].([]any); ok {
 		if len(v) > 1 {
 			return r, errors.New("group should contain a single value, an array with more than one entry was provided")
 		}

firewall/packet.go

@@ -6,7 +6,7 @@ import (
 	"net/netip"
 )
 
-type m map[string]interface{}
+type m = map[string]any
 
 const (
 	ProtoAny    = 0 // When we want to handle HOPOPT (0) we can change this, if ever

firewall_test.go

@@ -8,6 +8,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/gaissmai/bart"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
@@ -68,6 +69,9 @@ 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 +96,24 @@ 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 +133,13 @@ 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{}, "", ""))
@@ -127,7 +150,8 @@ func TestFirewall_Drop(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
-
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
 		RemoteAddr: netip.MustParseAddr("1.2.3.4"),
@@ -152,7 +176,7 @@ func TestFirewall_Drop(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{netip.MustParseAddr("1.2.3.4")},
 	}
-	h.buildNetworks(c.networks, c.unsafeNetworks)
+	h.buildNetworks(myVpnNetworksTable, 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{}, "", ""))
@@ -199,6 +223,85 @@ 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)
+
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
+
+	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(myVpnNetworksTable, 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 +311,10 @@ 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 +346,15 @@ 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 +368,18 @@ 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 +393,18 @@ 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 +429,17 @@ 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{
@@ -307,6 +458,8 @@ func TestFirewall_Drop2(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
@@ -332,7 +485,7 @@ func TestFirewall_Drop2(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h.buildNetworks(c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
+	h.buildNetworks(myVpnNetworksTable, c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
 
 	c1 := cert.CachedCertificate{
 		Certificate: &dummyCert{
@@ -347,7 +500,7 @@ func TestFirewall_Drop2(t *testing.T) {
 			peerCert: &c1,
 		},
 	}
-	h1.buildNetworks(c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
+	h1.buildNetworks(myVpnNetworksTable, 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, 0, 0, []string{"default-group", "test-group"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -364,6 +517,8 @@ func TestFirewall_Drop3(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
@@ -395,7 +550,7 @@ func TestFirewall_Drop3(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h1.buildNetworks(c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
+	h1.buildNetworks(myVpnNetworksTable, c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
 
 	c2 := cert.CachedCertificate{
 		Certificate: &dummyCert{
@@ -410,7 +565,7 @@ func TestFirewall_Drop3(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h2.buildNetworks(c2.Certificate.Networks(), c2.Certificate.UnsafeNetworks())
+	h2.buildNetworks(myVpnNetworksTable, c2.Certificate.Networks(), c2.Certificate.UnsafeNetworks())
 
 	c3 := cert.CachedCertificate{
 		Certificate: &dummyCert{
@@ -425,7 +580,7 @@ func TestFirewall_Drop3(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h3.buildNetworks(c3.Certificate.Networks(), c3.Certificate.UnsafeNetworks())
+	h3.buildNetworks(myVpnNetworksTable, c3.Certificate.Networks(), c3.Certificate.UnsafeNetworks())
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "host1", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -447,10 +602,50 @@ 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)
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
+
+	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(myVpnNetworksTable, 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{}
 	l.SetOutput(ob)
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
@@ -477,7 +672,7 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h.buildNetworks(c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
+	h.buildNetworks(myVpnNetworksTable, c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -510,6 +705,52 @@ 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)
+	myVpnNetworksTable := new(bart.Lite)
+	myVpnNetworksTable.Insert(netip.MustParsePrefix("192.0.2.1/24"))
+
+	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(myVpnNetworksTable, 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++ {
@@ -631,53 +872,53 @@ func TestNewFirewallFromConfig(t *testing.T) {
 	require.NoError(t, err)
 
 	conf := config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": "asdf"}
+	conf.Settings["firewall"] = map[string]any{"outbound": "asdf"}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound failed to parse, should be an array of rules")
 
 	// Test both port and code
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"port": "1", "code": "2"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "code": "2"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; only one of port or code should be provided")
 
 	// Test missing host, group, cidr, ca_name and ca_sha
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; at least one of host, group, cidr, local_cidr, ca_name, or ca_sha must be provided")
 
 	// Test code/port error
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"code": "a", "host": "testh"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; code was not a number; `a`")
 
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"port": "a", "host": "testh"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "a", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`")
 
 	// Test proto error
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"code": "1", "host": "testh"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; proto was not understood; ``")
 
 	// Test cidr parse error
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"code": "1", "cidr": "testh", "proto": "any"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "cidr": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; cidr did not parse; netip.ParsePrefix(\"testh\"): no '/'")
 
 	// Test local_cidr parse error
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"code": "1", "local_cidr": "testh", "proto": "any"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "local_cidr": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; local_cidr did not parse; netip.ParsePrefix(\"testh\"): no '/'")
 
 	// Test both group and groups
 	conf = config.NewC(l)
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "group": "a", "groups": []string{"b", "c"}}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "group": "a", "groups": []string{"b", "c"}}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.inbound rule #0; only one of group or groups should be defined, both provided")
 }
@@ -687,28 +928,28 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 	// Test adding tcp rule
 	conf := config.NewC(l)
 	mf := &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "tcp", "host": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "tcp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoTCP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
 	// Test adding udp rule
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "udp", "host": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "udp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoUDP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
 	// Test adding icmp rule
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"outbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "icmp", "host": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "icmp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
 	// Test adding any rule
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "host": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, host: "a", ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
@@ -716,49 +957,64 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 	cidr := netip.MustParsePrefix("10.0.0.0/8")
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "cidr": cidr.String()}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": cidr.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: cidr, localIp: netip.Prefix{}}, mf.lastCall)
 
 	// Test adding rule with local_cidr
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "local_cidr": cidr.String()}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": cidr.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: 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{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "ca_sha": "12312313123"}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "ca_sha": "12312313123"}}}
 	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: netip.Prefix{}, caSha: "12312313123"}, mf.lastCall)
 
 	// Test adding rule with ca_name
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "ca_name": "root01"}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "ca_name": "root01"}}}
 	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: netip.Prefix{}, caName: "root01"}, mf.lastCall)
 
 	// Test single group
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "group": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "group": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a"}, ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
 	// Test single groups
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "groups": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "groups": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a"}, ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
 	// Test multiple AND groups
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "groups": []string{"a", "b"}}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "groups": []string{"a", "b"}}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a", "b"}, ip: netip.Prefix{}, localIp: netip.Prefix{}}, mf.lastCall)
 
@@ -766,7 +1022,7 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	mf.nextCallReturn = errors.New("test error")
-	conf.Settings["firewall"] = map[interface{}]interface{}{"inbound": []interface{}{map[interface{}]interface{}{"port": "1", "proto": "any", "host": "a"}}}
+	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "host": "a"}}}
 	require.EqualError(t, AddFirewallRulesFromConfig(l, true, conf, mf), "firewall.inbound rule #0; `test error`")
 }
 
@@ -776,8 +1032,8 @@ func TestFirewall_convertRule(t *testing.T) {
 	l.SetOutput(ob)
 
 	// Ensure group array of 1 is converted and a warning is printed
-	c := map[interface{}]interface{}{
+	c := map[string]any{
-		"group": []interface{}{"group1"},
+		"group": []any{"group1"},
 	}
 
 	r, err := convertRule(l, c, "test", 1)
@@ -787,17 +1043,17 @@ func TestFirewall_convertRule(t *testing.T) {
 
 	// Ensure group array of > 1 is errord
 	ob.Reset()
-	c = map[interface{}]interface{}{
+	c = map[string]any{
-		"group": []interface{}{"group1", "group2"},
+		"group": []any{"group1", "group2"},
 	}
 
 	r, err = convertRule(l, c, "test", 1)
-	assert.Equal(t, "", ob.String())
+	assert.Empty(t, ob.String())
 	require.Error(t, err, "group should contain a single value, an array with more than one entry was provided")
 
 	// Make sure a well formed group is alright
 	ob.Reset()
-	c = map[interface{}]interface{}{
+	c = map[string]any{
 		"group": "group1",
 	}
 

go.mod

@@ -1,40 +1,38 @@
 module github.com/slackhq/nebula
 
-go 1.23.6
+go 1.25
-
-toolchain go1.23.7
 
 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 v0.18.1
+	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 v1.1.63
+	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 v1.21.1
+	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 v0.36.0
+	golang.org/x/crypto v0.43.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.37.0
+	golang.org/x/net v0.45.0
-	golang.org/x/sync v0.12.0
+	golang.org/x/sync v0.17.0
-	golang.org/x/sys v0.31.0
+	golang.org/x/sys v0.37.0
-	golang.org/x/term v0.30.0
+	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 v1.36.5
+	google.golang.org/protobuf v1.36.8
-	gopkg.in/yaml.v2 v2.4.0
+	gopkg.in/yaml.v3 v3.0.1
 	gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe
 )
 
@@ -43,15 +41,14 @@ require (
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/google/btree v1.1.2 // indirect
-	github.com/klauspost/compress v1.17.11 // 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 v0.18.0 // 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 v0.22.0 // indirect
+	golang.org/x/tools v0.33.0 // indirect
-	gopkg.in/yaml.v3 v3.0.1 // 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 v0.18.1 h1:bX2j560JC1MJpoEDevBGvXL5OZ1mkls320Vl8Igb5QQ=
+github.com/gaissmai/bart v0.25.0 h1:eqiokVPqM3F94vJ0bTHXHtH91S8zkKL+bKh+BsGOsJM=
-github.com/gaissmai/bart v0.18.1/go.mod h1:JJzMAhNF5Rjo4SF4jWBrANuJfqY+FvsFhW7t1UZJ+XY=
+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=
@@ -53,8 +53,8 @@ github.com/google/go-cmp v0.3.1/go.mod h1:8QqcDgzrUqlUb/G2PQTWiueGozuR1884gddMyw
 github.com/google/go-cmp v0.4.0/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
 github.com/google/go-cmp v0.5.4/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
 github.com/google/go-cmp v0.5.5/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
-github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
+github.com/google/go-cmp v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
-github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
+github.com/google/go-cmp v0.7.0/go.mod h1:pXiqmnSA92OHEEa9HXL2W4E7lf9JzCmGVUdgjX3N/iU=
 github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
 github.com/google/gopacket v1.1.19 h1:ves8RnFZPGiFnTS0uPQStjwru6uO6h+nlr9j6fL7kF8=
 github.com/google/gopacket v1.1.19/go.mod h1:iJ8V8n6KS+z2U1A8pUwu8bW5SyEMkXJB8Yo/Vo+TKTo=
@@ -64,12 +64,12 @@ 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 v1.17.11 h1:In6xLpyWOi1+C7tXUUWv2ot1QvBjxevKAaI6IXrJmUc=
+github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
-github.com/klauspost/compress v1.17.11/go.mod h1:pMDklpSncoRMuLFrf1W9Ss9KT+0rH90U12bZKk7uwG0=
+github.com/klauspost/compress v1.18.0/go.mod h1:2Pp+KzxcywXVXMr50+X0Q/Lsb43OQHYWRCY2AiWywWQ=
 github.com/konsorten/go-windows-terminal-sequences v1.0.1/go.mod h1:T0+1ngSBFLxvqU3pZ+m/2kptfBszLMUkC4ZK/EgS/cQ=
 github.com/konsorten/go-windows-terminal-sequences v1.0.3/go.mod h1:T0+1ngSBFLxvqU3pZ+m/2kptfBszLMUkC4ZK/EgS/cQ=
 github.com/kr/logfmt v0.0.0-20140226030751-b84e30acd515/go.mod h1:+0opPa2QZZtGFBFZlji/RkVcI2GknAs/DXo4wKdlNEc=
@@ -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 v1.1.63 h1:8M5aAw6OMZfFXTT7K5V0Eu5YiiL8l7nUAkyN6C9YwaY=
+github.com/miekg/dns v1.1.68 h1:jsSRkNozw7G/mnmXULynzMNIsgY2dHC8LO6U6Ij2JEA=
-github.com/miekg/dns v1.1.63/go.mod h1:6NGHfjhpmr5lt3XPLuyfDJi5AXbNIPM9PY6H6sF1Nfs=
+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 v1.21.1 h1:DOvXXTqVzvkIewV/CDPFdejpMCGeMcbGCQ8YOmu+Ibk=
+github.com/prometheus/client_golang v1.23.2 h1:Je96obch5RDVy3FDMndoUsjAhG5Edi49h0RJWRi/o0o=
-github.com/prometheus/client_golang v1.21.1/go.mod h1:U9NM32ykUErtVBxdvD3zfi+EuFkkaBvMb09mIfe0Zgg=
+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 v0.36.0 h1:AnAEvhDddvBdpY+uR+MyHmuZzzNqXSe/GvuDeob5L34=
+golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
-golang.org/x/crypto v0.36.0/go.mod h1:Y4J0ReaxCR1IMaabaSMugxJES1EpwhBHhv2bDHklZvc=
+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 v0.18.0 h1:5+9lSbEzPSdWkH32vYPBwEpX8KwDbM52Ud9xBUvNlb0=
+golang.org/x/mod v0.24.0 h1:ZfthKaKaT4NrhGVZHO1/WDTwGES4De8KtWO0SIbNJMU=
-golang.org/x/mod v0.18.0/go.mod h1:hTbmBsO62+eylJbnUtE2MGJUyE7QWk4xUqPFrRgJ+7c=
+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 v0.37.0 h1:1zLorHbz+LYj7MQlSf1+2tPIIgibq2eL5xkrGk6f+2c=
+golang.org/x/net v0.45.0 h1:RLBg5JKixCy82FtLJpeNlVM0nrSqpCRYzVU1n8kj0tM=
-golang.org/x/net v0.37.0/go.mod h1:ivrbrMbzFq5J41QOQh0siUuly180yBYtLp+CKbEaFx8=
+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 v0.12.0 h1:MHc5BpPuC30uJk597Ri8TV3CNZcTLu6B6z4lJy+g6Jw=
+golang.org/x/sync v0.17.0 h1:l60nONMj9l5drqw6jlhIELNv9I0A4OFgRsG9k2oT9Ug=
-golang.org/x/sync v0.12.0/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 v0.31.0 h1:ioabZlmFYtWhL+TRYpcnNlLwhyxaM9kWTDEmfnprqik=
+golang.org/x/sys v0.37.0 h1:fdNQudmxPjkdUTPnLn5mdQv7Zwvbvpaxqs831goi9kQ=
-golang.org/x/sys v0.31.0/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 v0.30.0 h1:PQ39fJZ+mfadBm0y5WlL4vlM7Sx1Hgf13sMIY2+QS9Y=
+golang.org/x/term v0.36.0 h1:zMPR+aF8gfksFprF/Nc/rd1wRS1EI6nDBGyWAvDzx2Q=
-golang.org/x/term v0.30.0/go.mod h1:NYYFdzHoI5wRh/h5tDMdMqCqPJZEuNqVR5xJLd/n67g=
+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 v0.22.0 h1:gqSGLZqv+AI9lIQzniJ0nZDRG5GBPsSi+DRNHWNz6yA=
+golang.org/x/tools v0.33.0 h1:4qz2S3zmRxbGIhDIAgjxvFutSvH5EfnsYrRBj0UI0bc=
-golang.org/x/tools v0.22.0/go.mod h1:aCwcsjqvq7Yqt6TNyX7QMU2enbQ/Gt0bo6krSeEri+c=
+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 v1.36.5 h1:tPhr+woSbjfYvY6/GPufUoYizxw1cF/yFoxJ2fmpwlM=
+google.golang.org/protobuf v1.36.8 h1:xHScyCOEuuwZEc6UtSOvPbAT4zRh0xcNRYekJwfqyMc=
-google.golang.org/protobuf v1.36.5/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=
+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=
@@ -251,8 +254,6 @@ gopkg.in/yaml.v2 v2.2.2/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
 gopkg.in/yaml.v2 v2.2.4/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
 gopkg.in/yaml.v2 v2.2.5/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
 gopkg.in/yaml.v2 v2.3.0/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
-gopkg.in/yaml.v2 v2.4.0 h1:D8xgwECY7CYvx+Y2n4sBz93Jn9JRvxdiyyo8CTfuKaY=
-gopkg.in/yaml.v2 v2.4.0/go.mod h1:RDklbk79AGWmwhnvt/jBztapEOGDOx6ZbXqjP6csGnQ=
 gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
 gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

handshake_ix.go

@@ -25,7 +25,7 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 
 	// If we're connecting to a v6 address we must use a v2 cert
 	cs := f.pki.getCertState()
-	v := cs.defaultVersion
+	v := cs.initiatingVersion
 	for _, a := range hh.hostinfo.vpnAddrs {
 		if a.Is6() {
 			v = cert.Version2
@@ -50,7 +50,7 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 			Error("Unable to handshake with host because no certificate handshake bytes is available")
 	}
 
-	ci, err := NewConnectionState(f.l, cs, crt, true, noise.HandshakeIX, cs.psk.primary)
+	ci, err := NewConnectionState(f.l, cs, crt, true, noise.HandshakeIX)
 	if err != nil {
 		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
 			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
@@ -71,7 +71,8 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 
 	hsBytes, err := hs.Marshal()
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).WithField("certVersion", v).
+		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("certVersion", v).
 			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to marshal handshake message")
 		return false
 	}
@@ -100,57 +101,38 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 	if crt == nil {
 		f.l.WithField("udpAddr", addr).
 			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
-			WithField("certVersion", cs.defaultVersion).
+			WithField("certVersion", cs.initiatingVersion).
 			Error("Unable to handshake with host because no certificate is available")
 	}
 
-	var (
+	ci, err := NewConnectionState(f.l, cs, crt, false, noise.HandshakeIX)
-		err error
-		ci  *ConnectionState
-		msg []byte
-	)
-
-	hs := &NebulaHandshake{}
-
-	for _, psk := range cs.psk.keys {
-		ci, err = NewConnectionState(f.l, cs, crt, false, noise.HandshakeIX, psk)
 	if err != nil {
-			//TODO: should be bother logging this, if we have multiple psks and the error is unrelated it will be verbose.
 		f.l.WithError(err).WithField("udpAddr", addr).
 			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
 			Error("Failed to create connection state")
-			continue
-		}
-
-		msg, _, _, err = ci.H.ReadMessage(nil, packet[header.Len:])
-		if err != nil {
-			// Calls to ReadMessage with an incorrect psk should fail, try the next one if we have one
-			continue
-		}
-
-		// Sometimes ReadMessage returns fine with a nil psk even if the handshake is using a psk, ensure our protobuf
-		// comes out clean as well
-		err = hs.Unmarshal(msg)
-		if err == nil {
-			// There was no error, we can continue with this handshake
-			break
-		}
-
-		// The unmarshal failed, try the next psk if we have one
-	}
-
-	// We finished with an error, log it and get out
-	if err != nil || hs.Details == nil {
-		// We aren't logging the error here because we can't be sure of the failure when using psk
-		f.l.WithField("udpAddr", addr).
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
-			Error("Was unable to decrypt the handshake")
 		return
 	}
 
 	// Mark packet 1 as seen so it doesn't show up as missed
 	ci.window.Update(f.l, 1)
 
+	msg, _, _, err := ci.H.ReadMessage(nil, packet[header.Len:])
+	if err != nil {
+		f.l.WithError(err).WithField("udpAddr", addr).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Error("Failed to call noise.ReadMessage")
+		return
+	}
+
+	hs := &NebulaHandshake{}
+	err = hs.Unmarshal(msg)
+	if err != nil || hs.Details == nil {
+		f.l.WithError(err).WithField("udpAddr", addr).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Error("Failed unmarshal handshake message")
+		return
+	}
+
 	rc, err := cert.Recombine(cert.Version(hs.Details.CertVersion), hs.Details.Cert, ci.H.PeerStatic(), ci.Curve())
 	if err != nil {
 		f.l.WithError(err).WithField("udpAddr", addr).
@@ -201,40 +183,29 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		return
 	}
 
-	var vpnAddrs []netip.Addr
-	var filteredNetworks []netip.Prefix
 	certName := remoteCert.Certificate.Name()
+	certVersion := remoteCert.Certificate.Version()
 	fingerprint := remoteCert.Fingerprint
 	issuer := remoteCert.Certificate.Issuer()
+	vpnNetworks := remoteCert.Certificate.Networks()
 
-	for _, network := range remoteCert.Certificate.Networks() {
+	anyVpnAddrsInCommon := false
+	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
+	for i, network := range vpnNetworks {
 		vpnAddr := network.Addr()
-		_, found := f.myVpnAddrsTable.Lookup(vpnAddr)
+		if f.myVpnAddrsTable.Contains(vpnAddr) {
-		if found {
+			f.l.WithField("vpnNetworks", vpnNetworks).WithField("udpAddr", addr).
-			f.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", addr).
 				WithField("certName", certName).
+				WithField("certVersion", certVersion).
 				WithField("fingerprint", fingerprint).
 				WithField("issuer", issuer).
 				WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Refusing to handshake with myself")
 			return
 		}
-
+		vpnAddrs[i] = network.Addr()
-		// vpnAddrs outside our vpn networks are of no use to us, filter them out
+		if f.myVpnNetworksTable.Contains(vpnAddr) {
-		if _, ok := f.myVpnNetworksTable.Lookup(vpnAddr); !ok {
+			anyVpnAddrsInCommon = true
-			continue
 		}
-
-		filteredNetworks = append(filteredNetworks, network)
-		vpnAddrs = append(vpnAddrs, vpnAddr)
-	}
-
-	if len(vpnAddrs) == 0 {
-		f.l.WithError(err).WithField("udpAddr", addr).
-			WithField("certName", certName).
-			WithField("fingerprint", fingerprint).
-			WithField("issuer", issuer).
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("No usable vpn addresses from host, refusing handshake")
-		return
 	}
 
 	if addr.IsValid() {
@@ -250,6 +221,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 	if err != nil {
 		f.l.WithError(err).WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 			WithField("certName", certName).
+			WithField("certVersion", certVersion).
 			WithField("fingerprint", fingerprint).
 			WithField("issuer", issuer).
 			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to generate index")
@@ -264,30 +236,36 @@ 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{},
 		},
 	}
 
-	f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
+	msgRxL := f.l.WithFields(m{
-		WithField("certName", certName).
+		"vpnAddrs":       vpnAddrs,
-		WithField("fingerprint", fingerprint).
+		"udpAddr":        addr,
-		WithField("issuer", issuer).
+		"certName":       certName,
-		WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+		"certVersion":    certVersion,
-		WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+		"fingerprint":    fingerprint,
-		Info("Handshake message received")
+		"issuer":         issuer,
+		"initiatorIndex": hs.Details.InitiatorIndex,
+		"responderIndex": hs.Details.ResponderIndex,
+		"remoteIndex":    h.RemoteIndex,
+		"handshake":      m{"stage": 1, "style": "ix_psk0"},
+	})
+
+	if anyVpnAddrsInCommon {
+		msgRxL.Info("Handshake message received")
+	} else {
+		//todo warn if not lighthouse or relay?
+		msgRxL.Info("Handshake message received, but no vpnNetworks in common.")
+	}
 
 	hs.Details.ResponderIndex = myIndex
 	hs.Details.Cert = cs.getHandshakeBytes(ci.myCert.Version())
 	if hs.Details.Cert == nil {
-		f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
+		msgRxL.WithField("myCertVersion", ci.myCert.Version()).
-			WithField("certName", certName).
-			WithField("fingerprint", fingerprint).
-			WithField("issuer", issuer).
-			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
-			WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
-			WithField("certVersion", ci.myCert.Version()).
 			Error("Unable to handshake with host because no certificate handshake bytes is available")
 		return
 	}
@@ -300,6 +278,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 	if err != nil {
 		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("udpAddr", addr).
 			WithField("certName", certName).
+			WithField("certVersion", certVersion).
 			WithField("fingerprint", fingerprint).
 			WithField("issuer", issuer).
 			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to marshal handshake message")
@@ -311,6 +290,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 	if err != nil {
 		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("udpAddr", addr).
 			WithField("certName", certName).
+			WithField("certVersion", certVersion).
 			WithField("fingerprint", fingerprint).
 			WithField("issuer", issuer).
 			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to call noise.WriteMessage")
@@ -318,6 +298,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 	} else if dKey == nil || eKey == nil {
 		f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("udpAddr", addr).
 			WithField("certName", certName).
+			WithField("certVersion", certVersion).
 			WithField("fingerprint", fingerprint).
 			WithField("issuer", issuer).
 			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Noise did not arrive at a key")
@@ -342,7 +323,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 
 	hostinfo.remotes = f.lightHouse.QueryCache(vpnAddrs)
 	hostinfo.SetRemote(addr)
-	hostinfo.buildNetworks(filteredNetworks, remoteCert.Certificate.UnsafeNetworks())
+	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate.Networks(), remoteCert.Certificate.UnsafeNetworks())
 
 	existing, err := f.handshakeManager.CheckAndComplete(hostinfo, 0, f)
 	if err != nil {
@@ -385,6 +366,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			// This means there was an existing tunnel and this handshake was older than the one we are currently based on
 			f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 				WithField("certName", certName).
+				WithField("certVersion", certVersion).
 				WithField("oldHandshakeTime", existing.lastHandshakeTime).
 				WithField("newHandshakeTime", hostinfo.lastHandshakeTime).
 				WithField("fingerprint", fingerprint).
@@ -400,6 +382,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			// This means we failed to insert because of collision on localIndexId. Just let the next handshake packet retry
 			f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 				WithField("certName", certName).
+				WithField("certVersion", certVersion).
 				WithField("fingerprint", fingerprint).
 				WithField("issuer", issuer).
 				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
@@ -412,6 +395,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			// And we forget to update it here
 			f.l.WithError(err).WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 				WithField("certName", certName).
+				WithField("certVersion", certVersion).
 				WithField("fingerprint", fingerprint).
 				WithField("issuer", issuer).
 				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
@@ -428,6 +412,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		if err != nil {
 			f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 				WithField("certName", certName).
+				WithField("certVersion", certVersion).
 				WithField("fingerprint", fingerprint).
 				WithField("issuer", issuer).
 				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
@@ -436,6 +421,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		} else {
 			f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 				WithField("certName", certName).
+				WithField("certVersion", certVersion).
 				WithField("fingerprint", fingerprint).
 				WithField("issuer", issuer).
 				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
@@ -454,6 +440,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		f.SendVia(via.relayHI, via.relay, msg, make([]byte, 12), make([]byte, mtu), false)
 		f.l.WithField("vpnAddrs", vpnAddrs).WithField("relay", via.relayHI.vpnAddrs[0]).
 			WithField("certName", certName).
+			WithField("certVersion", certVersion).
 			WithField("fingerprint", fingerprint).
 			WithField("issuer", issuer).
 			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
@@ -461,9 +448,9 @@ 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
 }
@@ -558,6 +545,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 
 	vpnNetworks := remoteCert.Certificate.Networks()
 	certName := remoteCert.Certificate.Name()
+	certVersion := remoteCert.Certificate.Version()
 	fingerprint := remoteCert.Fingerprint
 	issuer := remoteCert.Certificate.Issuer()
 
@@ -576,32 +564,22 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 		hostinfo.relayState.InsertRelayTo(via.relayHI.vpnAddrs[0])
 	}
 
-	var vpnAddrs []netip.Addr
+	anyVpnAddrsInCommon := false
-	var filteredNetworks []netip.Prefix
+	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
-	for _, network := range vpnNetworks {
+	for i, network := range vpnNetworks {
-		// vpnAddrs outside our vpn networks are of no use to us, filter them out
+		vpnAddrs[i] = network.Addr()
-		vpnAddr := network.Addr()
+		if f.myVpnNetworksTable.Contains(network.Addr()) {
-		if _, ok := f.myVpnNetworksTable.Lookup(vpnAddr); !ok {
+			anyVpnAddrsInCommon = true
-			continue
 		}
-
-		filteredNetworks = append(filteredNetworks, network)
-		vpnAddrs = append(vpnAddrs, vpnAddr)
-	}
-
-	if len(vpnAddrs) == 0 {
-		f.l.WithError(err).WithField("udpAddr", addr).
-			WithField("certName", certName).
-			WithField("fingerprint", fingerprint).
-			WithField("issuer", issuer).
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).Error("No usable vpn addresses from host, refusing handshake")
-		return true
 	}
 
 	// Ensure the right host responded
+	// todo is it more correct to see if any of hostinfo.vpnAddrs are in the cert? it should have len==1, but one day it might not?
 	if !slices.Contains(vpnAddrs, hostinfo.vpnAddrs[0]) {
 		f.l.WithField("intendedVpnAddrs", hostinfo.vpnAddrs).WithField("haveVpnNetworks", vpnNetworks).
-			WithField("udpAddr", addr).WithField("certName", certName).
+			WithField("udpAddr", addr).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
 			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
 			Info("Incorrect host responded to handshake")
 
@@ -609,6 +587,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 		f.handshakeManager.DeleteHostInfo(hostinfo)
 
 		// Create a new hostinfo/handshake for the intended vpn ip
+		//TODO is hostinfo.vpnAddrs[0] always the address to use?
 		f.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(newHH *HandshakeHostInfo) {
 			// Block the current used address
 			newHH.hostinfo.remotes = hostinfo.remotes
@@ -635,23 +614,29 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 	ci.window.Update(f.l, 2)
 
 	duration := time.Since(hh.startTime).Nanoseconds()
-	f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
+	msgRxL := f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 		WithField("certName", certName).
+		WithField("certVersion", certVersion).
 		WithField("fingerprint", fingerprint).
 		WithField("issuer", issuer).
 		WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
 		WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
 		WithField("durationNs", duration).
-		WithField("sentCachedPackets", len(hh.packetStore)).
+		WithField("sentCachedPackets", len(hh.packetStore))
-		Info("Handshake message received")
+	if anyVpnAddrsInCommon {
+		msgRxL.Info("Handshake message received")
+	} else {
+		//todo warn if not lighthouse or relay?
+		msgRxL.Info("Handshake message received, but no vpnNetworks in common.")
+	}
 
 	// Build up the radix for the firewall if we have subnets in the cert
 	hostinfo.vpnAddrs = vpnAddrs
-	hostinfo.buildNetworks(filteredNetworks, remoteCert.Certificate.UnsafeNetworks())
+	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate.Networks(), remoteCert.Certificate.UnsafeNetworks())
 
 	// 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))
@@ -666,7 +651,7 @@ 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

@@ -274,8 +274,7 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 			}
 
 			// Don't relay through the host I'm trying to connect to
-			_, found := hm.f.myVpnAddrsTable.Lookup(relay)
+			if hm.f.myVpnAddrsTable.Contains(relay) {
-			if found {
 				continue
 			}
 
@@ -451,7 +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{},
 		},

handshake_manager_test.go

@@ -10,7 +10,6 @@ import (
 	"github.com/slackhq/nebula/test"
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
 )
 
 func Test_NewHandshakeManagerVpnIp(t *testing.T) {
@@ -24,15 +23,11 @@ func Test_NewHandshakeManagerVpnIp(t *testing.T) {
 
 	lh := newTestLighthouse()
 
-	psk, err := NewPsk(PskAccepting, nil)
-	require.NoError(t, err)
-
 	cs := &CertState{
-		defaultVersion:   cert.Version1,
+		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
 		v1HandshakeBytes:  []byte{},
-		psk:              psk,
 	}
 
 	blah := NewHandshakeManager(l, mainHM, lh, &udp.NoopConn{}, defaultHandshakeConfig)
@@ -103,5 +98,5 @@ func (mw *mockEncWriter) GetHostInfo(_ netip.Addr) *HostInfo {
 }
 
 func (mw *mockEncWriter) GetCertState() *CertState {
-	return &CertState{defaultVersion: cert.Version2}
+	return &CertState{initiatingVersion: cert.Version2}
 }

header/header.go

@@ -19,7 +19,7 @@ import (
 // |-----------------------------------------------------------------------|
 // |                               payload...                              |
 
-type m map[string]interface{}
+type m = map[string]any
 
 const (
 	Version uint8 = 1

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
 }
 
@@ -208,6 +212,18 @@ func (rs *RelayState) InsertRelay(ip netip.Addr, idx uint32, r *Relay) {
 	rs.relayForByIdx[idx] = r
 }
 
+type NetworkType uint8
+
+const (
+	NetworkTypeUnknown NetworkType = iota
+	// NetworkTypeVPN is a network that overlaps one or more of the vpnNetworks in our certificate
+	NetworkTypeVPN
+	// NetworkTypeVPNPeer is a network that does not overlap one of our networks
+	NetworkTypeVPNPeer
+	// NetworkTypeUnsafe is a network from Certificate.UnsafeNetworks()
+	NetworkTypeUnsafe
+)
+
 type HostInfo struct {
 	remote          netip.AddrPort
 	remotes         *RemoteList
@@ -216,14 +232,11 @@ type HostInfo struct {
 	remoteIndexId   uint32
 	localIndexId    uint32
 
-	// vpnAddrs is a list of vpn addresses assigned to this host that are within our own vpn networks
+	// vpnAddrs is a list of vpn addresses assigned to this host
-	// The host may have other vpn addresses that are outside our
-	// vpn networks but were removed because they are not usable
 	vpnAddrs []netip.Addr
-	recvError atomic.Uint32
 
-	// networks are both all vpn and unsafe networks assigned to this host
+	// networks is a combination of specific vpn addresses (not prefixes!) and full unsafe networks assigned to this host.
-	networks   *bart.Table[struct{}]
+	networks   *bart.Table[NetworkType]
 	relayState RelayState
 
 	// HandshakePacket records the packets used to create this hostinfo
@@ -250,6 +263,14 @@ 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 {
@@ -719,26 +740,27 @@ func (i *HostInfo) SetRemoteIfPreferred(hm *HostMap, newRemote netip.AddrPort) b
 	return false
 }
 
-func (i *HostInfo) RecvErrorExceeded() bool {
+func (i *HostInfo) buildNetworks(myVpnNetworksTable *bart.Lite, networks, unsafeNetworks []netip.Prefix) {
-	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
+		if myVpnNetworksTable.Contains(networks[0].Addr()) {
-		return
+			return // Simple case, no CIDRTree needed
+		}
 	}
 
-	i.networks = new(bart.Table[struct{}])
+	i.networks = new(bart.Table[NetworkType])
 	for _, network := range networks {
-		i.networks.Insert(network, struct{}{})
+		var nwType NetworkType
+		if myVpnNetworksTable.Contains(network.Addr()) {
+			nwType = NetworkTypeVPN
+		} else {
+			nwType = NetworkTypeVPNPeer
+		}
+		nprefix := netip.PrefixFrom(network.Addr(), network.Addr().BitLen())
+		i.networks.Insert(nprefix, nwType)
 	}
 
 	for _, network := range unsafeNetworks {
-		i.networks.Insert(network, struct{}{})
+		i.networks.Insert(network, NetworkTypeUnsafe)
 	}
 }
 

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) {
@@ -210,8 +211,36 @@ func TestHostMap_reload(t *testing.T) {
 	assert.Empty(t, hm.GetPreferredRanges())
 
 	c.ReloadConfigString("preferred_ranges: [1.1.1.0/24, 10.1.1.0/24]")
-	assert.EqualValues(t, []string{"1.1.1.0/24", "10.1.1.0/24"}, toS(hm.GetPreferredRanges()))
+	assert.Equal(t, []string{"1.1.1.0/24", "10.1.1.0/24"}, toS(hm.GetPreferredRanges()))
 
 	c.ReloadConfigString("preferred_ranges: [1.1.1.1/32]")
-	assert.EqualValues(t, []string{"1.1.1.1/32"}, toS(hm.GetPreferredRanges()))
+	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

@@ -8,6 +8,7 @@ import (
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/iputil"
 	"github.com/slackhq/nebula/noiseutil"
+	"github.com/slackhq/nebula/routing"
 )
 
 func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache) {
@@ -21,14 +22,12 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 
 	// Ignore local broadcast packets
 	if f.dropLocalBroadcast {
-		_, found := f.myBroadcastAddrsTable.Lookup(fwPacket.RemoteAddr)
+		if f.myBroadcastAddrsTable.Contains(fwPacket.RemoteAddr) {
-		if found {
 			return
 		}
 	}
 
-	_, found := f.myVpnAddrsTable.Lookup(fwPacket.RemoteAddr)
+	if f.myVpnAddrsTable.Contains(fwPacket.RemoteAddr) {
-	if found {
 		// Immediately forward packets from self to self.
 		// This should only happen on Darwin-based and FreeBSD hosts, which
 		// routes packets from the Nebula addr to the Nebula addr through the Nebula
@@ -49,7 +48,7 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 		return
 	}
 
-	hostinfo, ready := f.getOrHandshake(fwPacket.RemoteAddr, func(hh *HandshakeHostInfo) {
+	hostinfo, ready := f.getOrHandshakeConsiderRouting(fwPacket, func(hh *HandshakeHostInfo) {
 		hh.cachePacket(f.l, header.Message, 0, packet, f.sendMessageNow, f.cachedPacketMetrics)
 	})
 
@@ -121,22 +120,93 @@ func (f *Interface) rejectOutside(packet []byte, ci *ConnectionState, hostinfo *
 	f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, out, nb, packet, q)
 }
 
+// Handshake will attempt to initiate a tunnel with the provided vpn address if it is within our vpn networks. This is a no-op if the tunnel is already established or being established
 func (f *Interface) Handshake(vpnAddr netip.Addr) {
-	f.getOrHandshake(vpnAddr, nil)
+	f.getOrHandshakeNoRouting(vpnAddr, nil)
 }
 
-// getOrHandshake returns nil if the vpnAddr is not routable.
+// getOrHandshakeNoRouting returns nil if the vpnAddr is not routable.
 // If the 2nd return var is false then the hostinfo is not ready to be used in a tunnel
-func (f *Interface) getOrHandshake(vpnAddr netip.Addr, cacheCallback func(*HandshakeHostInfo)) (*HostInfo, bool) {
+func (f *Interface) getOrHandshakeNoRouting(vpnAddr netip.Addr, cacheCallback func(*HandshakeHostInfo)) (*HostInfo, bool) {
-	_, found := f.myVpnNetworksTable.Lookup(vpnAddr)
+	if f.myVpnNetworksTable.Contains(vpnAddr) {
-	if !found {
+		return f.handshakeManager.GetOrHandshake(vpnAddr, cacheCallback)
-		vpnAddr = f.inside.RouteFor(vpnAddr)
+	}
-		if !vpnAddr.IsValid() {
+
 	return nil, false
+}
+
+// getOrHandshakeConsiderRouting will try to find the HostInfo to handle this packet, starting a handshake if necessary.
+// If the 2nd return var is false then the hostinfo is not ready to be used in a tunnel.
+func (f *Interface) getOrHandshakeConsiderRouting(fwPacket *firewall.Packet, cacheCallback func(*HandshakeHostInfo)) (*HostInfo, bool) {
+
+	destinationAddr := fwPacket.RemoteAddr
+
+	hostinfo, ready := f.getOrHandshakeNoRouting(destinationAddr, cacheCallback)
+
+	// Host is inside the mesh, no routing required
+	if hostinfo != nil {
+		return hostinfo, ready
+	}
+
+	gateways := f.inside.RoutesFor(destinationAddr)
+
+	switch len(gateways) {
+	case 0:
+		return nil, false
+	case 1:
+		// Single gateway route
+		return f.handshakeManager.GetOrHandshake(gateways[0].Addr(), cacheCallback)
+	default:
+		// Multi gateway route, perform ECMP categorization
+		gatewayAddr, balancingOk := routing.BalancePacket(fwPacket, gateways)
+
+		if !balancingOk {
+			// This happens if the gateway buckets were not calculated, this _should_ never happen
+			f.l.Error("Gateway buckets not calculated, fallback from ECMP to random routing. Please report this bug.")
+		}
+
+		var handshakeInfoForChosenGateway *HandshakeHostInfo
+		var hhReceiver = func(hh *HandshakeHostInfo) {
+			handshakeInfoForChosenGateway = hh
+		}
+
+		// Store the handshakeHostInfo for later.
+		// If this node is not reachable we will attempt other nodes, if none are reachable we will
+		// cache the packet for this gateway.
+		if hostinfo, ready = f.handshakeManager.GetOrHandshake(gatewayAddr, hhReceiver); ready {
+			return hostinfo, true
+		}
+
+		// It appears the selected gateway cannot be reached, find another gateway to fallback on.
+		// The current implementation breaks ECMP but that seems better than no connectivity.
+		// If ECMP is also required when a gateway is down then connectivity status
+		// for each gateway needs to be kept and the weights recalculated when they go up or down.
+		// This would also need to interact with unsafe_route updates through reloading the config or
+		// use of the use_system_route_table option
+
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("destination", destinationAddr).
+				WithField("originalGateway", gatewayAddr).
+				Debugln("Calculated gateway for ECMP not available, attempting other gateways")
+		}
+
+		for i := range gateways {
+			// Skip the gateway that failed previously
+			if gateways[i].Addr() == gatewayAddr {
+				continue
+			}
+
+			// We do not need the HandshakeHostInfo since we cache the packet in the originally chosen gateway
+			if hostinfo, ready = f.handshakeManager.GetOrHandshake(gateways[i].Addr(), nil); ready {
+				return hostinfo, true
 			}
 		}
 
-	return f.handshakeManager.GetOrHandshake(vpnAddr, cacheCallback)
+		// No gateways reachable, cache the packet in the originally chosen gateway
+		cacheCallback(handshakeInfoForChosenGateway)
+		return hostinfo, false
+	}
+
 }
 
 func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubType, hostinfo *HostInfo, p, nb, out []byte) {
@@ -163,7 +233,7 @@ func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubTyp
 
 // SendMessageToVpnAddr handles real addr:port lookup and sends to the current best known address for vpnAddr
 func (f *Interface) SendMessageToVpnAddr(t header.MessageType, st header.MessageSubType, vpnAddr netip.Addr, p, nb, out []byte) {
-	hostInfo, ready := f.getOrHandshake(vpnAddr, func(hh *HandshakeHostInfo) {
+	hostInfo, ready := f.getOrHandshakeNoRouting(vpnAddr, func(hh *HandshakeHostInfo) {
 		hh.cachePacket(f.l, t, st, p, f.SendMessageToHostInfo, f.cachedPacketMetrics)
 	})
 
@@ -218,7 +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.
@@ -286,7 +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

@@ -28,12 +28,12 @@ type InterfaceConfig struct {
 	Outside            udp.Conn
 	Inside             overlay.Device
 	pki                *PKI
+	Cipher             string
 	Firewall           *Firewall
 	ServeDns           bool
 	HandshakeManager   *HandshakeManager
 	lightHouse         *LightHouse
-	checkInterval           time.Duration
+	connectionManager  *connectionManager
-	pendingDeletionInterval time.Duration
 	DropLocalBroadcast bool
 	DropMulticast      bool
 	routines           int
@@ -61,11 +61,11 @@ type Interface struct {
 	serveDns              bool
 	createTime            time.Time
 	lightHouse            *LightHouse
-	myBroadcastAddrsTable *bart.Table[struct{}]
+	myBroadcastAddrsTable *bart.Lite
 	myVpnAddrs            []netip.Addr // A list of addresses assigned to us via our certificate
-	myVpnAddrsTable       *bart.Table[struct{}] // A table of addresses assigned to us via our certificate
+	myVpnAddrsTable       *bart.Lite
 	myVpnNetworks         []netip.Prefix // A list of networks assigned to us via our certificate
-	myVpnNetworksTable    *bart.Table[struct{}] // A table of networks assigned to us via our certificate
+	myVpnNetworksTable    *bart.Lite
 	dropLocalBroadcast    bool
 	dropMulticast         bool
 	routines              int
@@ -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
 }
@@ -410,7 +413,7 @@ func (f *Interface) emitStats(ctx context.Context, i time.Duration) {
 	udpStats := udp.NewUDPStatsEmitter(f.writers)
 
 	certExpirationGauge := metrics.GetOrRegisterGauge("certificate.ttl_seconds", nil)
-	certDefaultVersion := metrics.GetOrRegisterGauge("certificate.default_version", nil)
+	certInitiatingVersion := metrics.GetOrRegisterGauge("certificate.initiating_version", nil)
 	certMaxVersion := metrics.GetOrRegisterGauge("certificate.max_version", nil)
 
 	for {
@@ -425,7 +428,7 @@ func (f *Interface) emitStats(ctx context.Context, i time.Duration) {
 			certState := f.pki.getCertState()
 			defaultCrt := certState.GetDefaultCertificate()
 			certExpirationGauge.Update(int64(defaultCrt.NotAfter().Sub(time.Now()) / time.Second))
-			certDefaultVersion.Update(int64(defaultCrt.Version()))
+			certInitiatingVersion.Update(int64(defaultCrt.Version()))
 
 			// Report the max certificate version we are capable of using
 			if certState.v2Cert != 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
@@ -32,7 +33,7 @@ type LightHouse struct {
 	amLighthouse bool
 
 	myVpnNetworks      []netip.Prefix
-	myVpnNetworksTable *bart.Table[struct{}]
+	myVpnNetworksTable *bart.Lite
 	punchConn          udp.Conn
 	punchy             *Punchy
 
@@ -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()
 }
 
@@ -201,8 +202,7 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 
 			//TODO: we could technically insert all returned addrs instead of just the first one if a dns lookup was used
 			addr := addrs[0].Unmap()
-			_, found := lh.myVpnNetworksTable.Lookup(addr)
+			if lh.myVpnNetworksTable.Contains(addr) {
-			if found {
 				lh.l.WithField("addr", rawAddr).WithField("entry", i+1).
 					Warn("Ignoring lighthouse.advertise_addrs report because it is within the nebula network range")
 				continue
@@ -307,13 +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")
@@ -347,37 +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)
 		}
 
-		_, found := lh.myVpnNetworksTable.Lookup(addr)
+		if !lh.myVpnNetworksTable.Contains(addr) {
-		if !found {
+			return nil, util.NewContextualError("lighthouse host is not in our networks, invalid", m{"vpnAddr": addr, "networks": lh.myVpnNetworks}, nil)
-			return 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) {
@@ -422,7 +421,7 @@ func (lh *LightHouse) loadStaticMap(c *config.C, staticList map[netip.Addr]struc
 		return err
 	}
 
-	shm := c.GetMap("static_host_map", map[interface{}]interface{}{})
+	shm := c.GetMap("static_host_map", map[string]any{})
 	i := 0
 
 	for k, v := range shm {
@@ -431,14 +430,13 @@ func (lh *LightHouse) loadStaticMap(c *config.C, staticList map[netip.Addr]struc
 			return util.NewContextualError("Unable to parse static_host_map entry", m{"host": k, "entry": i + 1}, err)
 		}
 
-		_, found := lh.myVpnNetworksTable.Lookup(vpnAddr)
+		if !lh.myVpnNetworksTable.Contains(vpnAddr) {
-		if !found {
 			return util.NewContextualError("static_host_map key is not in our network, invalid", m{"vpnAddr": vpnAddr, "networks": lh.myVpnNetworks, "entry": i + 1}, nil)
 		}
 
-		vals, ok := v.([]interface{})
+		vals, ok := v.([]any)
 		if !ok {
-			vals = []interface{}{v}
+			vals = []any{v}
 		}
 		remoteAddrs := []string{}
 		for _, v := range vals {
@@ -489,7 +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
@@ -522,11 +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 {
+		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 {
@@ -568,7 +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 {
@@ -630,31 +632,37 @@ 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]
+		if ok {
+			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 {
 		return false
 	}
 
-	_, found := lh.myVpnNetworksTable.Lookup(to)
+	if lh.myVpnNetworksTable.Contains(to) {
-	if found {
 		return false
 	}
 
@@ -674,8 +682,7 @@ func (lh *LightHouse) unlockedShouldAddV4(vpnAddr netip.Addr, to *V4AddrPort) bo
 		return false
 	}
 
-	_, found := lh.myVpnNetworksTable.Lookup(udpAddr.Addr())
+	if lh.myVpnNetworksTable.Contains(udpAddr.Addr()) {
-	if found {
 		return false
 	}
 
@@ -695,8 +702,7 @@ func (lh *LightHouse) unlockedShouldAddV6(vpnAddr netip.Addr, to *V6AddrPort) bo
 		return false
 	}
 
-	_, found := lh.myVpnNetworksTable.Lookup(udpAddr.Addr())
+	if lh.myVpnNetworksTable.Contains(udpAddr.Addr()) {
-	if found {
 		return false
 	}
 
@@ -704,21 +710,24 @@ 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()
+	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 {
+			if l[j] == vpnAddrs[i] {
 				return true
 			}
 		}
+	}
 	return false
 }
 
@@ -758,12 +767,12 @@ 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()
 		} else {
-			v = lh.ifce.GetCertState().defaultVersion
+			v = lh.ifce.GetCertState().initiatingVersion
 		}
 
 		if v == cert.Version1 {
@@ -856,8 +865,7 @@ func (lh *LightHouse) SendUpdate() {
 
 	lal := lh.GetLocalAllowList()
 	for _, e := range localAddrs(lh.l, lal) {
-		_, found := lh.myVpnNetworksTable.Lookup(e)
+		if lh.myVpnNetworksTable.Contains(e) {
-		if found {
 			continue
 		}
 
@@ -877,13 +885,13 @@ 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 {
 			v = hi.ConnectionState.myCert.Version()
 		} else {
-			v = lh.ifce.GetCertState().defaultVersion
+			v = lh.ifce.GetCertState().initiatingVersion
 		}
 		if v == cert.Version1 {
 			if v1Update == nil {
@@ -935,7 +943,6 @@ func (lh *LightHouse) SendUpdate() {
 						V4AddrPorts:   v4,
 						V6AddrPorts:   v6,
 						RelayVpnAddrs: relays,
-						VpnAddr:       netAddrToProtoAddr(lh.myVpnNetworks[0].Addr()),
 					},
 				}
 
@@ -1010,17 +1017,17 @@ func (lhh *LightHouseHandler) resetMeta() *NebulaMeta {
 	return lhh.meta
 }
 
-func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, fromVpnAddrs []netip.Addr, p []byte, w EncWriter) {
+func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, hostinfo *HostInfo, p []byte, w EncWriter) {
 	n := lhh.resetMeta()
 	err := n.Unmarshal(p)
 	if err != nil {
-		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).WithField("udpAddr", rAddr).
+		lhh.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("udpAddr", rAddr).
 			Error("Failed to unmarshal lighthouse packet")
 		return
 	}
 
 	if n.Details == nil {
-		lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("udpAddr", rAddr).
+		lhh.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("udpAddr", rAddr).
 			Error("Invalid lighthouse update")
 		return
 	}
@@ -1029,24 +1036,24 @@ func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, fromVpnAddrs [
 
 	switch n.Type {
 	case NebulaMeta_HostQuery:
-		lhh.handleHostQuery(n, fromVpnAddrs, rAddr, w)
+		lhh.handleHostQuery(n, hostinfo, rAddr, w)
 
 	case NebulaMeta_HostQueryReply:
-		lhh.handleHostQueryReply(n, fromVpnAddrs)
+		lhh.handleHostQueryReply(n, hostinfo.vpnAddrs)
 
 	case NebulaMeta_HostUpdateNotification:
-		lhh.handleHostUpdateNotification(n, fromVpnAddrs, w)
+		lhh.handleHostUpdateNotification(n, hostinfo, w)
 
 	case NebulaMeta_HostMovedNotification:
 	case NebulaMeta_HostPunchNotification:
-		lhh.handleHostPunchNotification(n, fromVpnAddrs, w)
+		lhh.handleHostPunchNotification(n, hostinfo.vpnAddrs, w)
 
 	case NebulaMeta_HostUpdateNotificationAck:
 		// noop
 	}
 }
 
-func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []netip.Addr, addr netip.AddrPort, w EncWriter) {
+func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, hostinfo *HostInfo, addr netip.AddrPort, w EncWriter) {
 	// Exit if we don't answer queries
 	if !lhh.lh.amLighthouse {
 		if lhh.l.Level >= logrus.DebugLevel {
@@ -1055,19 +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", hostinfo.vpnAddrs).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", hostinfo.vpnAddrs).WithField("queryVpnAddr", queryVpnAddr).
+				Debugln("invalid vpn addr for v1 handleHostQuery")
 		}
 		return
 	}
@@ -1076,9 +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 {
@@ -1095,14 +1099,14 @@ func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []neti
 	}
 
 	if err != nil {
-		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host query reply")
+		lhh.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).Error("Failed to marshal lighthouse host query reply")
 		return
 	}
 
 	lhh.lh.metricTx(NebulaMeta_HostQueryReply, 1)
-	w.SendMessageToVpnAddr(header.LightHouse, 0, fromVpnAddrs[0], lhh.pb[:ln], lhh.nb, lhh.out[:0])
+	w.SendMessageToHostInfo(header.LightHouse, 0, hostinfo, lhh.pb[:ln], lhh.nb, lhh.out[:0])
 
-	lhh.sendHostPunchNotification(n, fromVpnAddrs, queryVpnAddr, w)
+	lhh.sendHostPunchNotification(n, hostinfo.vpnAddrs, queryVpnAddr, w)
 }
 
 // sendHostPunchNotification signals the other side to punch some zero byte udp packets
@@ -1111,20 +1115,35 @@ func (lhh *LightHouseHandler) sendHostPunchNotification(n *NebulaMeta, fromVpnAd
 	found, ln, err := lhh.lh.queryAndPrepMessage(whereToPunch, func(c *cache) (int, error) {
 		n = lhh.resetMeta()
 		n.Type = NebulaMeta_HostPunchNotification
-		targetHI := lhh.lh.ifce.GetHostInfo(punchNotifDest)
+		punchNotifDestHI := lhh.lh.ifce.GetHostInfo(punchNotifDest)
 		var useVersion cert.Version
-		if targetHI == nil {
+		if punchNotifDestHI == nil {
-			useVersion = lhh.lh.ifce.GetCertState().defaultVersion
+			useVersion = lhh.lh.ifce.GetCertState().initiatingVersion
 		} else {
-			crt := targetHI.GetCert().Certificate
-			useVersion = crt.Version()
 			// we can only retarget if we have a hostinfo
-			newDest, ok := findNetworkUnion(crt.Networks(), fromVpnAddrs)
+			punchNotifDestCrt := punchNotifDestHI.GetCert().Certificate
+			useVersion = punchNotifDestCrt.Version()
+			punchNotifDestNetworks := punchNotifDestCrt.Networks()
+
+			//if we (the lighthouse) don't have a network in common with punchNotifDest, try to find one
+			if !lhh.lh.myVpnNetworksTable.Contains(punchNotifDest) {
+				newPunchNotifDest, ok := findNetworkUnion(lhh.lh.myVpnNetworks, punchNotifDestHI.vpnAddrs)
 				if ok {
-				whereToPunch = newDest
+					punchNotifDest = newPunchNotifDest
 				} 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", punchNotifDestNetworks).Debugln("unable to notify host to host, no addresses in common")
+					}
+				}
+			}
+
+			newWhereToPunch, ok := findNetworkUnion(punchNotifDestNetworks, fromVpnAddrs)
+			if ok {
+				whereToPunch = newWhereToPunch
+			} else {
+				if lhh.l.Level >= logrus.DebugLevel {
+					lhh.l.WithFields(m{"from": fromVpnAddrs, "to": punchNotifDestNetworks}).Debugln("unable to punch to host, no addresses in common with requestor")
+				}
 			}
 		}
 
@@ -1183,19 +1202,17 @@ 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")
+			}
 		}
 	}
 }
@@ -1205,18 +1222,16 @@ 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()
@@ -1233,7 +1248,8 @@ func (lhh *LightHouseHandler) handleHostQueryReply(n *NebulaMeta, fromVpnAddrs [
 	}
 }
 
-func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVpnAddrs []netip.Addr, w EncWriter) {
+func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, hostinfo *HostInfo, w EncWriter) {
+	fromVpnAddrs := hostinfo.vpnAddrs
 	if !lhh.lh.amLighthouse {
 		if lhh.l.Level >= logrus.DebugLevel {
 			lhh.l.Debugln("I am not a lighthouse, do not take host updates: ", fromVpnAddrs)
@@ -1241,27 +1257,24 @@ 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")
 		}
@@ -1275,24 +1288,24 @@ 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
 	}
@@ -1304,19 +1317,26 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 	}
 
 	lhh.lh.metricTx(NebulaMeta_HostUpdateNotificationAck, 1)
-	w.SendMessageToVpnAddr(header.LightHouse, 0, fromVpnAddrs[0], lhh.pb[:ln], lhh.nb, lhh.out[:0])
+	w.SendMessageToHostInfo(header.LightHouse, 0, hostinfo, lhh.pb[:ln], lhh.nb, lhh.out[:0])
 }
 
 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
 		}
@@ -1328,48 +1348,31 @@ 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))
 		}()
 	}
 }
@@ -1448,3 +1451,17 @@ 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

@@ -14,7 +14,7 @@ import (
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
 func TestOldIPv4Only(t *testing.T) {
@@ -31,8 +31,8 @@ func TestOldIPv4Only(t *testing.T) {
 func Test_lhStaticMapping(t *testing.T) {
 	l := test.NewLogger()
 	myVpnNet := netip.MustParsePrefix("10.128.0.1/16")
-	nt := new(bart.Table[struct{}])
+	nt := new(bart.Lite)
-	nt.Insert(myVpnNet, struct{}{})
+	nt.Insert(myVpnNet)
 	cs := &CertState{
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
@@ -40,15 +40,15 @@ func Test_lhStaticMapping(t *testing.T) {
 	lh1 := "10.128.0.2"
 
 	c := config.NewC(l)
-	c.Settings["lighthouse"] = map[interface{}]interface{}{"hosts": []interface{}{lh1}}
+	c.Settings["lighthouse"] = map[string]any{"hosts": []any{lh1}}
-	c.Settings["static_host_map"] = map[interface{}]interface{}{lh1: []interface{}{"1.1.1.1:4242"}}
+	c.Settings["static_host_map"] = map[string]any{lh1: []any{"1.1.1.1:4242"}}
 	_, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 
 	lh2 := "10.128.0.3"
 	c = config.NewC(l)
-	c.Settings["lighthouse"] = map[interface{}]interface{}{"hosts": []interface{}{lh1, lh2}}
+	c.Settings["lighthouse"] = map[string]any{"hosts": []any{lh1, lh2}}
-	c.Settings["static_host_map"] = map[interface{}]interface{}{lh1: []interface{}{"100.1.1.1:4242"}}
+	c.Settings["static_host_map"] = map[string]any{lh1: []any{"100.1.1.1:4242"}}
 	_, err = NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
 	require.EqualError(t, err, "lighthouse 10.128.0.3 does not have a static_host_map entry")
 }
@@ -56,8 +56,8 @@ func Test_lhStaticMapping(t *testing.T) {
 func TestReloadLighthouseInterval(t *testing.T) {
 	l := test.NewLogger()
 	myVpnNet := netip.MustParsePrefix("10.128.0.1/16")
-	nt := new(bart.Table[struct{}])
+	nt := new(bart.Lite)
-	nt.Insert(myVpnNet, struct{}{})
+	nt.Insert(myVpnNet)
 	cs := &CertState{
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
@@ -65,12 +65,12 @@ func TestReloadLighthouseInterval(t *testing.T) {
 	lh1 := "10.128.0.2"
 
 	c := config.NewC(l)
-	c.Settings["lighthouse"] = map[interface{}]interface{}{
+	c.Settings["lighthouse"] = map[string]any{
-		"hosts":    []interface{}{lh1},
+		"hosts":    []any{lh1},
 		"interval": "1s",
 	}
 
-	c.Settings["static_host_map"] = map[interface{}]interface{}{lh1: []interface{}{"1.1.1.1:4242"}}
+	c.Settings["static_host_map"] = map[string]any{lh1: []any{"1.1.1.1:4242"}}
 	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 	lh.ifce = &mockEncWriter{}
@@ -91,8 +91,8 @@ func TestReloadLighthouseInterval(t *testing.T) {
 func BenchmarkLighthouseHandleRequest(b *testing.B) {
 	l := test.NewLogger()
 	myVpnNet := netip.MustParsePrefix("10.128.0.1/0")
-	nt := new(bart.Table[struct{}])
+	nt := new(bart.Lite)
-	nt.Insert(myVpnNet, struct{}{})
+	nt.Insert(myVpnNet)
 	cs := &CertState{
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
@@ -132,8 +132,13 @@ func BenchmarkLighthouseHandleRequest(b *testing.B) {
 	)
 
 	mw := &mockEncWriter{}
-
+	hostinfo := &HostInfo{
-	hi := []netip.Addr{vpnIp2}
+		ConnectionState: &ConnectionState{
+			eKey: nil,
+			dKey: nil,
+		},
+		vpnAddrs: []netip.Addr{vpnIp2},
+	}
 	b.Run("notfound", func(b *testing.B) {
 		lhh := lh.NewRequestHandler()
 		req := &NebulaMeta{
@@ -146,7 +151,7 @@ func BenchmarkLighthouseHandleRequest(b *testing.B) {
 		p, err := req.Marshal()
 		require.NoError(b, err)
 		for n := 0; n < b.N; n++ {
-			lhh.HandleRequest(rAddr, hi, p, mw)
+			lhh.HandleRequest(rAddr, hostinfo, p, mw)
 		}
 	})
 	b.Run("found", func(b *testing.B) {
@@ -162,7 +167,7 @@ func BenchmarkLighthouseHandleRequest(b *testing.B) {
 		require.NoError(b, err)
 
 		for n := 0; n < b.N; n++ {
-			lhh.HandleRequest(rAddr, hi, p, mw)
+			lhh.HandleRequest(rAddr, hostinfo, p, mw)
 		}
 	})
 }
@@ -192,12 +197,12 @@ func TestLighthouse_Memory(t *testing.T) {
 	theirVpnIp := netip.MustParseAddr("10.128.0.3")
 
 	c := config.NewC(l)
-	c.Settings["lighthouse"] = map[interface{}]interface{}{"am_lighthouse": true}
+	c.Settings["lighthouse"] = map[string]any{"am_lighthouse": true}
-	c.Settings["listen"] = map[interface{}]interface{}{"port": 4242}
+	c.Settings["listen"] = map[string]any{"port": 4242}
 
 	myVpnNet := netip.MustParsePrefix("10.128.0.1/24")
-	nt := new(bart.Table[struct{}])
+	nt := new(bart.Lite)
-	nt.Insert(myVpnNet, struct{}{})
+	nt.Insert(myVpnNet)
 	cs := &CertState{
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
@@ -277,12 +282,12 @@ func TestLighthouse_Memory(t *testing.T) {
 func TestLighthouse_reload(t *testing.T) {
 	l := test.NewLogger()
 	c := config.NewC(l)
-	c.Settings["lighthouse"] = map[interface{}]interface{}{"am_lighthouse": true}
+	c.Settings["lighthouse"] = map[string]any{"am_lighthouse": true}
-	c.Settings["listen"] = map[interface{}]interface{}{"port": 4242}
+	c.Settings["listen"] = map[string]any{"port": 4242}
 
 	myVpnNet := netip.MustParsePrefix("10.128.0.1/24")
-	nt := new(bart.Table[struct{}])
+	nt := new(bart.Lite)
-	nt.Insert(myVpnNet, struct{}{})
+	nt.Insert(myVpnNet)
 	cs := &CertState{
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
@@ -291,9 +296,9 @@ func TestLighthouse_reload(t *testing.T) {
 	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 
-	nc := map[interface{}]interface{}{
+	nc := map[string]any{
-		"static_host_map": map[interface{}]interface{}{
+		"static_host_map": map[string]any{
-			"10.128.0.2": []interface{}{"1.1.1.1:4242"},
+			"10.128.0.2": []any{"1.1.1.1:4242"},
 		},
 	}
 	rc, err := yaml.Marshal(nc)
@@ -326,7 +331,14 @@ func newLHHostRequest(fromAddr netip.AddrPort, myVpnIp, queryVpnIp netip.Addr, l
 	w := &testEncWriter{
 		metaFilter: &filter,
 	}
-	lhh.HandleRequest(fromAddr, []netip.Addr{myVpnIp}, b, w)
+	hostinfo := &HostInfo{
+		ConnectionState: &ConnectionState{
+			eKey: nil,
+			dKey: nil,
+		},
+		vpnAddrs: []netip.Addr{myVpnIp},
+	}
+	lhh.HandleRequest(fromAddr, hostinfo, b, w)
 	return w.lastReply
 }
 
@@ -355,9 +367,15 @@ func newLHHostUpdate(fromAddr netip.AddrPort, vpnIp netip.Addr, addrs []netip.Ad
 	if err != nil {
 		panic(err)
 	}
-
+	hostinfo := &HostInfo{
+		ConnectionState: &ConnectionState{
+			eKey: nil,
+			dKey: nil,
+		},
+		vpnAddrs: []netip.Addr{vpnIp},
+	}
 	w := &testEncWriter{}
-	lhh.HandleRequest(fromAddr, []netip.Addr{vpnIp}, b, w)
+	lhh.HandleRequest(fromAddr, hostinfo, b, w)
 }
 
 type testLhReply struct {
@@ -417,7 +435,7 @@ func (tw *testEncWriter) GetHostInfo(vpnIp netip.Addr) *HostInfo {
 }
 
 func (tw *testEncWriter) GetCertState() *CertState {
-	return &CertState{defaultVersion: tw.protocolVersion}
+	return &CertState{initiatingVersion: tw.protocolVersion}
 }
 
 // assertIp4InArray asserts every address in want is at the same position in have and that the lengths match
@@ -493,3 +511,123 @@ 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

@@ -13,10 +13,10 @@ import (
 	"github.com/slackhq/nebula/sshd"
 	"github.com/slackhq/nebula/udp"
 	"github.com/slackhq/nebula/util"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
-type m map[string]interface{}
+type m = map[string]any
 
 func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logger, deviceFactory overlay.DeviceFactory) (retcon *Control, reterr error) {
 	ctx, cancel := context.WithCancel(context.Background())
@@ -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)
@@ -220,9 +221,6 @@ 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,
 		Inside:                tun,
@@ -231,9 +229,8 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		Firewall:              fw,
 		ServeDns:              serveDns,
 		HandshakeManager:      handshakeManager,
+		connectionManager:     connManager,
 		lightHouse:            lightHouse,
-		checkInterval:           time.Second * time.Duration(checkInterval),
-		pendingDeletionInterval: time.Second * time.Duration(pendingDeletionInterval),
 		tryPromoteEvery:       c.GetUint32("counters.try_promote", defaultPromoteEvery),
 		reQueryEvery:          c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
 		reQueryWait:           c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
@@ -244,7 +241,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		version:               buildVersion,
 		relayManager:          NewRelayManager(ctx, l, hostMap, c),
 		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

@@ -31,8 +31,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 
 	//l.Error("in packet ", header, packet[HeaderLen:])
 	if ip.IsValid() {
-		_, found := f.myVpnNetworksTable.Lookup(ip.Addr())
+		if f.myVpnNetworksTable.Contains(ip.Addr()) {
-		if found {
 			if f.l.Level >= logrus.DebugLevel {
 				f.l.WithField("udpAddr", ip).Debug("Refusing to process double encrypted packet")
 			}
@@ -82,7 +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)
@@ -139,7 +138,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 			return
 		}
 
-		lhf.HandleRequest(ip, hostinfo.vpnAddrs, d, f)
+		lhf.HandleRequest(ip, hostinfo, d, f)
 
 		// Fallthrough to the bottom to record incoming traffic
 
@@ -214,7 +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
@@ -255,16 +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
 	}
+	// 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
@@ -313,12 +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)
@@ -366,7 +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
 			}
 
@@ -374,7 +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
 			}
 
@@ -500,7 +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")
@@ -539,10 +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/device.go

@@ -3,6 +3,8 @@ package overlay
 import (
 	"io"
 	"net/netip"
+
+	"github.com/slackhq/nebula/routing"
 )
 
 type Device interface {
@@ -10,6 +12,6 @@ type Device interface {
 	Activate() error
 	Networks() []netip.Prefix
 	Name() string
-	RouteFor(netip.Addr) netip.Addr
+	RoutesFor(netip.Addr) routing.Gateways
 	NewMultiQueueReader() (io.ReadWriteCloser, error)
 }

overlay/route.go

@@ -11,13 +11,14 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 )
 
 type Route struct {
 	MTU     int
 	Metric  int
 	Cidr    netip.Prefix
-	Via     netip.Addr
+	Via     routing.Gateways
 	Install bool
 }
 
@@ -47,15 +48,17 @@ func (r Route) String() string {
 	return s
 }
 
-func makeRouteTree(l *logrus.Logger, routes []Route, allowMTU bool) (*bart.Table[netip.Addr], error) {
+func makeRouteTree(l *logrus.Logger, routes []Route, allowMTU bool) (*bart.Table[routing.Gateways], error) {
-	routeTree := new(bart.Table[netip.Addr])
+	routeTree := new(bart.Table[routing.Gateways])
 	for _, r := range routes {
 		if !allowMTU && r.MTU > 0 {
 			l.WithField("route", r).Warnf("route MTU is not supported in %s", runtime.GOOS)
 		}
 
-		if r.Via.IsValid() {
+		gateways := r.Via
-			routeTree.Insert(r.Cidr, r.Via)
+		if len(gateways) > 0 {
+			routing.CalculateBucketsForGateways(gateways)
+			routeTree.Insert(r.Cidr, gateways)
 		}
 	}
 	return routeTree, nil
@@ -69,7 +72,7 @@ func parseRoutes(c *config.C, networks []netip.Prefix) ([]Route, error) {
 		return []Route{}, nil
 	}
 
-	rawRoutes, ok := r.([]interface{})
+	rawRoutes, ok := r.([]any)
 	if !ok {
 		return nil, fmt.Errorf("tun.routes is not an array")
 	}
@@ -80,7 +83,7 @@ func parseRoutes(c *config.C, networks []netip.Prefix) ([]Route, error) {
 
 	routes := make([]Route, len(rawRoutes))
 	for i, r := range rawRoutes {
-		m, ok := r.(map[interface{}]interface{})
+		m, ok := r.(map[string]any)
 		if !ok {
 			return nil, fmt.Errorf("entry %v in tun.routes is invalid", i+1)
 		}
@@ -148,7 +151,7 @@ func parseUnsafeRoutes(c *config.C, networks []netip.Prefix) ([]Route, error) {
 		return []Route{}, nil
 	}
 
-	rawRoutes, ok := r.([]interface{})
+	rawRoutes, ok := r.([]any)
 	if !ok {
 		return nil, fmt.Errorf("tun.unsafe_routes is not an array")
 	}
@@ -159,7 +162,7 @@ func parseUnsafeRoutes(c *config.C, networks []netip.Prefix) ([]Route, error) {
 
 	routes := make([]Route, len(rawRoutes))
 	for i, r := range rawRoutes {
-		m, ok := r.(map[interface{}]interface{})
+		m, ok := r.(map[string]any)
 		if !ok {
 			return nil, fmt.Errorf("entry %v in tun.unsafe_routes is invalid", i+1)
 		}
@@ -201,16 +204,65 @@ func parseUnsafeRoutes(c *config.C, networks []netip.Prefix) ([]Route, error) {
 			return nil, fmt.Errorf("entry %v.via in tun.unsafe_routes is not present", i+1)
 		}
 
-		via, ok := rVia.(string)
+		var gateways routing.Gateways
-		if !ok {
-			return nil, fmt.Errorf("entry %v.via in tun.unsafe_routes is not a string: found %T", i+1, rVia)
-		}
 
-		viaVpnIp, err := netip.ParseAddr(via)
+		switch via := rVia.(type) {
+		case string:
+			viaIp, err := netip.ParseAddr(via)
 			if err != nil {
 				return nil, fmt.Errorf("entry %v.via in tun.unsafe_routes failed to parse address: %v", i+1, err)
 			}
 
+			gateways = routing.Gateways{routing.NewGateway(viaIp, 1)}
+
+		case []any:
+			gateways = make(routing.Gateways, len(via))
+			for ig, v := range via {
+				gatewayMap, ok := v.(map[string]any)
+				if !ok {
+					return nil, fmt.Errorf("entry %v in tun.unsafe_routes[%v].via is invalid", i+1, ig+1)
+				}
+
+				rGateway, ok := gatewayMap["gateway"]
+				if !ok {
+					return nil, fmt.Errorf("entry .gateway in tun.unsafe_routes[%v].via[%v] is not present", i+1, ig+1)
+				}
+
+				parsedGateway, ok := rGateway.(string)
+				if !ok {
+					return nil, fmt.Errorf("entry .gateway in tun.unsafe_routes[%v].via[%v] is not a string", i+1, ig+1)
+				}
+
+				gatewayIp, err := netip.ParseAddr(parsedGateway)
+				if err != nil {
+					return nil, fmt.Errorf("entry .gateway in tun.unsafe_routes[%v].via[%v] failed to parse address: %v", i+1, ig+1, err)
+				}
+
+				rGatewayWeight, ok := gatewayMap["weight"]
+				if !ok {
+					rGatewayWeight = 1
+				}
+
+				gatewayWeight, ok := rGatewayWeight.(int)
+				if !ok {
+					_, err = strconv.ParseInt(rGatewayWeight.(string), 10, 32)
+					if err != nil {
+						return nil, fmt.Errorf("entry .weight in tun.unsafe_routes[%v].via[%v] is not an integer", i+1, ig+1)
+					}
+				}
+
+				if gatewayWeight < 1 || gatewayWeight > math.MaxInt32 {
+					return nil, fmt.Errorf("entry .weight in tun.unsafe_routes[%v].via[%v] is not in range (1-%d) : %v", i+1, ig+1, math.MaxInt32, gatewayWeight)
+				}
+
+				gateways[ig] = routing.NewGateway(gatewayIp, gatewayWeight)
+
+			}
+
+		default:
+			return nil, fmt.Errorf("entry %v.via in tun.unsafe_routes is not a string or list of gateways: found %T", i+1, rVia)
+		}
+
 		rRoute, ok := m["route"]
 		if !ok {
 			return nil, fmt.Errorf("entry %v.route in tun.unsafe_routes is not present", i+1)
@@ -226,7 +278,7 @@ func parseUnsafeRoutes(c *config.C, networks []netip.Prefix) ([]Route, error) {
 		}
 
 		r := Route{
-			Via:     viaVpnIp,
+			Via:     gateways,
 			MTU:     mtu,
 			Metric:  metric,
 			Install: install,

overlay/route_test.go

@@ -6,6 +6,7 @@ import (
 	"testing"
 
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -23,75 +24,75 @@ func Test_parseRoutes(t *testing.T) {
 	assert.Empty(t, routes)
 
 	// not an array
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": "hi"}
+	c.Settings["tun"] = map[string]any{"routes": "hi"}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "tun.routes is not an array")
 
 	// no routes
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{}}
+	c.Settings["tun"] = map[string]any{"routes": []any{}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
 	assert.Empty(t, routes)
 
 	// weird route
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{"asdf"}}
+	c.Settings["tun"] = map[string]any{"routes": []any{"asdf"}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1 in tun.routes is invalid")
 
 	// no mtu
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.mtu in tun.routes is not present")
 
 	// bad mtu
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "nope"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "nope"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.mtu in tun.routes is not an integer: strconv.Atoi: parsing \"nope\": invalid syntax")
 
 	// low mtu
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "499"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "499"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.mtu in tun.routes is below 500: 499")
 
 	// missing route
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "500"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "500"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.routes is not present")
 
 	// unparsable route
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "500", "route": "nope"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "500", "route": "nope"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.routes failed to parse: netip.ParsePrefix(\"nope\"): no '/'")
 
 	// below network range
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "500", "route": "1.0.0.0/8"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "500", "route": "1.0.0.0/8"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.routes is not contained within the configured vpn networks; route: 1.0.0.0/8, networks: [10.0.0.0/24]")
 
 	// above network range
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "500", "route": "10.0.1.0/24"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "500", "route": "10.0.1.0/24"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.routes is not contained within the configured vpn networks; route: 10.0.1.0/24, networks: [10.0.0.0/24]")
 
 	// Not in multiple ranges
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{map[interface{}]interface{}{"mtu": "500", "route": "192.0.0.0/24"}}}
+	c.Settings["tun"] = map[string]any{"routes": []any{map[string]any{"mtu": "500", "route": "192.0.0.0/24"}}}
 	routes, err = parseRoutes(c, []netip.Prefix{n, netip.MustParsePrefix("192.1.0.0/24")})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.routes is not contained within the configured vpn networks; route: 192.0.0.0/24, networks: [10.0.0.0/24 192.1.0.0/24]")
 
 	// happy case
-	c.Settings["tun"] = map[interface{}]interface{}{"routes": []interface{}{
+	c.Settings["tun"] = map[string]any{"routes": []any{
-		map[interface{}]interface{}{"mtu": "9000", "route": "10.0.0.0/29"},
+		map[string]any{"mtu": "9000", "route": "10.0.0.0/29"},
-		map[interface{}]interface{}{"mtu": "8000", "route": "10.0.0.1/32"},
+		map[string]any{"mtu": "8000", "route": "10.0.0.1/32"},
 	}}
 	routes, err = parseRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
@@ -128,105 +129,129 @@ func Test_parseUnsafeRoutes(t *testing.T) {
 	assert.Empty(t, routes)
 
 	// not an array
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": "hi"}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": "hi"}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "tun.unsafe_routes is not an array")
 
 	// no routes
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
 	assert.Empty(t, routes)
 
 	// weird route
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{"asdf"}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{"asdf"}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1 in tun.unsafe_routes is invalid")
 
 	// no via
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.via in tun.unsafe_routes is not present")
 
 	// invalid via
-	for _, invalidValue := range []interface{}{
+	for _, invalidValue := range []any{
 		127, false, nil, 1.0, []string{"1", "2"},
 	} {
-		c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": invalidValue}}}
+		c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": invalidValue}}}
 		routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 		assert.Nil(t, routes)
-		require.EqualError(t, err, fmt.Sprintf("entry 1.via in tun.unsafe_routes is not a string: found %T", invalidValue))
+		require.EqualError(t, err, fmt.Sprintf("entry 1.via in tun.unsafe_routes is not a string or list of gateways: found %T", invalidValue))
 	}
 
+	// Unparsable list of via
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": []string{"1", "2"}}}}
+	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
+	assert.Nil(t, routes)
+	require.EqualError(t, err, "entry 1.via in tun.unsafe_routes is not a string or list of gateways: found []string")
+
 	// unparsable via
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"mtu": "500", "via": "nope"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"mtu": "500", "via": "nope"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.via in tun.unsafe_routes failed to parse address: ParseAddr(\"nope\"): unable to parse IP")
 
+	// unparsable gateway
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"mtu": "500", "via": []any{map[string]any{"gateway": "1"}}}}}
+	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
+	assert.Nil(t, routes)
+	require.EqualError(t, err, "entry .gateway in tun.unsafe_routes[1].via[1] failed to parse address: ParseAddr(\"1\"): unable to parse IP")
+
+	// missing gateway element
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"mtu": "500", "via": []any{map[string]any{"weight": "1"}}}}}
+	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
+	assert.Nil(t, routes)
+	require.EqualError(t, err, "entry .gateway in tun.unsafe_routes[1].via[1] is not present")
+
+	// unparsable weight element
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"mtu": "500", "via": []any{map[string]any{"gateway": "10.0.0.1", "weight": "a"}}}}}
+	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
+	assert.Nil(t, routes)
+	require.EqualError(t, err, "entry .weight in tun.unsafe_routes[1].via[1] is not an integer")
+
 	// missing route
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "500"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "mtu": "500"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.unsafe_routes is not present")
 
 	// unparsable route
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "500", "route": "nope"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "mtu": "500", "route": "nope"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.unsafe_routes failed to parse: netip.ParsePrefix(\"nope\"): no '/'")
 
 	// within network range
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "route": "10.0.0.0/24"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "route": "10.0.0.0/24"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.route in tun.unsafe_routes is contained within the configured vpn networks; route: 10.0.0.0/24, network: 10.0.0.0/24")
 
 	// below network range
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "route": "1.0.0.0/8"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "route": "1.0.0.0/8"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Len(t, routes, 1)
 	require.NoError(t, err)
 
 	// above network range
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "route": "10.0.1.0/24"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "route": "10.0.1.0/24"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Len(t, routes, 1)
 	require.NoError(t, err)
 
 	// no mtu
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "route": "1.0.0.0/8"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "route": "1.0.0.0/8"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Len(t, routes, 1)
 	assert.Equal(t, 0, routes[0].MTU)
 
 	// bad mtu
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "nope"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "mtu": "nope"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.mtu in tun.unsafe_routes is not an integer: strconv.Atoi: parsing \"nope\": invalid syntax")
 
 	// low mtu
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "499"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "mtu": "499"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.mtu in tun.unsafe_routes is below 500: 499")
 
 	// bad install
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "9000", "route": "1.0.0.0/29", "install": "nope"}}}
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{map[string]any{"via": "127.0.0.1", "mtu": "9000", "route": "1.0.0.0/29", "install": "nope"}}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	assert.Nil(t, routes)
 	require.EqualError(t, err, "entry 1.install in tun.unsafe_routes is not a boolean: strconv.ParseBool: parsing \"nope\": invalid syntax")
 
 	// happy case
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{
-		map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "9000", "route": "1.0.0.0/29", "install": "t"},
+		map[string]any{"via": "127.0.0.1", "mtu": "9000", "route": "1.0.0.0/29", "install": "t"},
-		map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "8000", "route": "1.0.0.1/32", "install": 0},
+		map[string]any{"via": "127.0.0.1", "mtu": "8000", "route": "1.0.0.1/32", "install": 0},
-		map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "1500", "metric": 1234, "route": "1.0.0.2/32", "install": 1},
+		map[string]any{"via": "127.0.0.1", "mtu": "1500", "metric": 1234, "route": "1.0.0.2/32", "install": 1},
-		map[interface{}]interface{}{"via": "127.0.0.1", "mtu": "1500", "metric": 1234, "route": "1.0.0.2/32"},
+		map[string]any{"via": "127.0.0.1", "mtu": "1500", "metric": 1234, "route": "1.0.0.2/32"},
 	}}
 	routes, err = parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
@@ -263,9 +288,9 @@ func Test_makeRouteTree(t *testing.T) {
 	n, err := netip.ParsePrefix("10.0.0.0/24")
 	require.NoError(t, err)
 
-	c.Settings["tun"] = map[interface{}]interface{}{"unsafe_routes": []interface{}{
+	c.Settings["tun"] = map[string]any{"unsafe_routes": []any{
-		map[interface{}]interface{}{"via": "192.168.0.1", "route": "1.0.0.0/28"},
+		map[string]any{"via": "192.168.0.1", "route": "1.0.0.0/28"},
-		map[interface{}]interface{}{"via": "192.168.0.2", "route": "1.0.0.1/32"},
+		map[string]any{"via": "192.168.0.2", "route": "1.0.0.1/32"},
 	}}
 	routes, err := parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
@@ -280,7 +305,7 @@ func Test_makeRouteTree(t *testing.T) {
 
 	nip, err := netip.ParseAddr("192.168.0.1")
 	require.NoError(t, err)
-	assert.Equal(t, nip, r)
+	assert.Equal(t, nip, r[0].Addr())
 
 	ip, err = netip.ParseAddr("1.0.0.1")
 	require.NoError(t, err)
@@ -289,10 +314,91 @@ func Test_makeRouteTree(t *testing.T) {
 
 	nip, err = netip.ParseAddr("192.168.0.2")
 	require.NoError(t, err)
-	assert.Equal(t, nip, r)
+	assert.Equal(t, nip, r[0].Addr())
 
 	ip, err = netip.ParseAddr("1.1.0.1")
 	require.NoError(t, err)
 	r, ok = routeTree.Lookup(ip)
 	assert.False(t, ok)
 }
+
+func Test_makeMultipathUnsafeRouteTree(t *testing.T) {
+	l := test.NewLogger()
+	c := config.NewC(l)
+	n, err := netip.ParsePrefix("10.0.0.0/24")
+	require.NoError(t, err)
+
+	c.Settings["tun"] = map[string]any{
+		"unsafe_routes": []any{
+			map[string]any{
+				"route": "192.168.86.0/24",
+				"via":   "192.168.100.10",
+			},
+			map[string]any{
+				"route": "192.168.87.0/24",
+				"via": []any{
+					map[string]any{
+						"gateway": "10.0.0.1",
+					},
+					map[string]any{
+						"gateway": "10.0.0.2",
+					},
+					map[string]any{
+						"gateway": "10.0.0.3",
+					},
+				},
+			},
+			map[string]any{
+				"route": "192.168.89.0/24",
+				"via": []any{
+					map[string]any{
+						"gateway": "10.0.0.1",
+						"weight":  10,
+					},
+					map[string]any{
+						"gateway": "10.0.0.2",
+						"weight":  5,
+					},
+				},
+			},
+		},
+	}
+
+	routes, err := parseUnsafeRoutes(c, []netip.Prefix{n})
+	require.NoError(t, err)
+	assert.Len(t, routes, 3)
+	routeTree, err := makeRouteTree(l, routes, true)
+	require.NoError(t, err)
+
+	ip, err := netip.ParseAddr("192.168.86.1")
+	require.NoError(t, err)
+	r, ok := routeTree.Lookup(ip)
+	assert.True(t, ok)
+
+	nip, err := netip.ParseAddr("192.168.100.10")
+	require.NoError(t, err)
+	assert.Equal(t, nip, r[0].Addr())
+
+	ip, err = netip.ParseAddr("192.168.87.1")
+	require.NoError(t, err)
+	r, ok = routeTree.Lookup(ip)
+	assert.True(t, ok)
+
+	expectedGateways := routing.Gateways{routing.NewGateway(netip.MustParseAddr("10.0.0.1"), 1),
+		routing.NewGateway(netip.MustParseAddr("10.0.0.2"), 1),
+		routing.NewGateway(netip.MustParseAddr("10.0.0.3"), 1)}
+
+	routing.CalculateBucketsForGateways(expectedGateways)
+	assert.ElementsMatch(t, expectedGateways, r)
+
+	ip, err = netip.ParseAddr("192.168.89.1")
+	require.NoError(t, err)
+	r, ok = routeTree.Lookup(ip)
+	assert.True(t, ok)
+
+	expectedGateways = routing.Gateways{routing.NewGateway(netip.MustParseAddr("10.0.0.1"), 10),
+		routing.NewGateway(netip.MustParseAddr("10.0.0.2"), 5)}
+
+	routing.CalculateBucketsForGateways(expectedGateways)
+	assert.ElementsMatch(t, expectedGateways, r)
+}

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_android.go

@@ -13,6 +13,7 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
 )
 
@@ -21,7 +22,7 @@ type tun struct {
 	fd          int
 	vpnNetworks []netip.Prefix
 	Routes      atomic.Pointer[[]Route]
-	routeTree   atomic.Pointer[bart.Table[netip.Addr]]
+	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
 }
 
@@ -56,7 +57,7 @@ func newTun(_ *config.C, _ *logrus.Logger, _ []netip.Prefix, _ bool) (*tun, erro
 	return nil, fmt.Errorf("newTun not supported in Android")
 }
 
-func (t *tun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }

overlay/tun_darwin.go

@@ -7,7 +7,6 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"net"
 	"net/netip"
 	"os"
 	"sync/atomic"
@@ -17,6 +16,7 @@ import (
 	"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"
@@ -28,7 +28,7 @@ type tun struct {
 	vpnNetworks []netip.Prefix
 	DefaultMTU  int
 	Routes      atomic.Pointer[[]Route]
-	routeTree   atomic.Pointer[bart.Table[netip.Addr]]
+	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	linkAddr    *netroute.LinkAddr
 	l           *logrus.Logger
 
@@ -294,7 +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,
 	}
 
@@ -342,12 +341,12 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (t *tun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, ok := t.routeTree.Load().Lookup(ip)
 	if ok {
 		return r
 	}
-	return netip.Addr{}
+	return routing.Gateways{}
 }
 
 // Get the LinkAddr for the interface of the given name
@@ -382,7 +381,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
 
 	for _, r := range routes {
-		if !r.Via.IsValid() || !r.Install {
+		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
@@ -393,7 +392,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 				t.l.WithField("route", r.Cidr).
 					Warnf("unable to add unsafe_route, identical route already exists")
 			} else {
-				retErr := util.NewContextualError("Failed to add route", map[string]interface{}{"route": r}, err)
+				retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
 				if logErrors {
 					retErr.Log(t.l)
 				} else {
@@ -553,13 +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_disabled.go

@@ -9,6 +9,7 @@ import (
 	"github.com/rcrowley/go-metrics"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/iputil"
+	"github.com/slackhq/nebula/routing"
 )
 
 type disabledTun struct {
@@ -43,8 +44,8 @@ func (*disabledTun) Activate() error {
 	return nil
 }
 
-func (*disabledTun) RouteFor(addr netip.Addr) netip.Addr {
+func (*disabledTun) RoutesFor(addr netip.Addr) routing.Gateways {
-	return netip.Addr{}
+	return routing.Gateways{}
 }
 
 func (t *disabledTun) Networks() []netip.Prefix {

overlay/tun_freebsd.go

@@ -10,23 +10,28 @@ import (
 	"io"
 	"io/fs"
 	"net/netip"
-	"os"
-	"os/exec"
-	"strconv"
 	"sync/atomic"
 	"syscall"
+	"time"
 	"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 (
 	// 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
+	TUNSIFMODE       = 0x8004745e
+	TUNSIFHEAD       = 0x80047460
+	OSIOCAIFADDR_IN6 = 0x8088691b
+	IN6_IFF_NODAD    = 0x0020
 )
 
 type fiodgnameArg struct {
@@ -36,43 +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[netip.Addr]]
+	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")
 		}
+		t.devFd = -1
+
+		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()}
-
-		// Destroy the interface
 				err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
-		return err
+			}
+			if err != nil {
+				t.l.WithError(err).Error("Error destroying tunnel")
+			}
+		}()
+
+		// wait up to 1 second so we start blocking at the ioctl
+		select {
+		case <-c:
+		case <-time.After(1 * time.Second):
+		}
 	}
 
 	return nil
@@ -84,32 +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
 	}
@@ -121,11 +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
 		}
@@ -148,11 +270,11 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	}
 
 	t := &tun{
-		ReadWriteCloser: file,
 		Device:      deviceName,
 		vpnNetworks: vpnNetworks,
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
 		l:           l,
+		devFd:       fd,
 	}
 
 	err = t.reload(c, true)
@@ -171,38 +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 {
@@ -242,7 +437,7 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (t *tun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
@@ -262,20 +457,21 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
 	for _, r := range routes {
-		if !r.Via.IsValid() || !r.Install {
+		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(), "-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]interface{}{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
+		} else {
+			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -288,9 +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")
@@ -305,3 +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_ios.go

@@ -16,6 +16,7 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
 )
 
@@ -23,7 +24,7 @@ type tun struct {
 	io.ReadWriteCloser
 	vpnNetworks []netip.Prefix
 	Routes      atomic.Pointer[[]Route]
-	routeTree   atomic.Pointer[bart.Table[netip.Addr]]
+	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
 }
 
@@ -79,7 +80,7 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (t *tun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }

overlay/tun_linux.go

@@ -17,6 +17,7 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
 	"github.com/vishvananda/netlink"
 	"golang.org/x/sys/unix"
@@ -34,9 +35,10 @@ type tun struct {
 	ioctlFd     uintptr
 
 	Routes                    atomic.Pointer[[]Route]
-	routeTree       atomic.Pointer[bart.Table[netip.Addr]]
+	routeTree                 atomic.Pointer[bart.Table[routing.Gateways]]
 	routeChan                 chan struct{}
 	useSystemRoutes           bool
+	useSystemRoutesBufferSize int
 
 	l *logrus.Logger
 }
@@ -128,6 +130,7 @@ func newTunGeneric(c *config.C, l *logrus.Logger, file *os.File, vpnNetworks []n
 		vpnNetworks:               vpnNetworks,
 		TXQueueLen:                c.GetInt("tun.tx_queue", 500),
 		useSystemRoutes:           c.GetBool("tun.use_system_route_table", false),
+		useSystemRoutesBufferSize: c.GetInt("tun.use_system_route_table_buffer_size", 0),
 		l:                         l,
 	}
 
@@ -231,7 +234,7 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 	return file, nil
 }
 
-func (t *tun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
@@ -290,7 +293,6 @@ 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
@@ -358,6 +360,11 @@ 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
 	}
@@ -463,7 +470,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 
 		err := netlink.RouteReplace(&nr)
 		if err != nil {
-			retErr := util.NewContextualError("Failed to add route", map[string]interface{}{"route": r}, err)
+			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
@@ -530,7 +537,13 @@ func (t *tun) watchRoutes() {
 	rch := make(chan netlink.RouteUpdate)
 	doneChan := make(chan struct{})
 
-	if err := netlink.RouteSubscribe(rch, doneChan); err != nil {
+	netlinkOptions := netlink.RouteSubscribeOptions{
+		ReceiveBufferSize:      t.useSystemRoutesBufferSize,
+		ReceiveBufferForceSize: t.useSystemRoutesBufferSize != 0,
+		ErrorCallback:          func(e error) { t.l.WithError(e).Errorf("netlink error") },
+	}
+
+	if err := netlink.RouteSubscribeWithOptions(rch, doneChan, netlinkOptions); err != nil {
 		t.l.WithError(err).Errorf("failed to subscribe to system route changes")
 		return
 	}
@@ -540,8 +553,14 @@ func (t *tun) watchRoutes() {
 	go func() {
 		for {
 			select {
-			case r := <-rch:
+			case r, ok := <-rch:
+				if ok {
 					t.updateRoutes(r)
+				} else {
+					// may be should do something here as
+					// netlink stops sending updates
+					return
+				}
 			case <-doneChan:
 				// netlink.RouteSubscriber will close the rch for us
 				return
@@ -550,20 +569,7 @@ func (t *tun) watchRoutes() {
 	}()
 }
 
-func (t *tun) updateRoutes(r netlink.RouteUpdate) {
+func (t *tun) isGatewayInVpnNetworks(gwAddr netip.Addr) bool {
-	if r.Gw == nil {
-		// Not a gateway route, ignore
-		t.l.WithField("route", r).Debug("Ignoring route update, not a gateway route")
-		return
-	}
-
-	gwAddr, ok := netip.AddrFromSlice(r.Gw)
-	if !ok {
-		t.l.WithField("route", r).Debug("Ignoring route update, invalid gateway address")
-		return
-	}
-
-	gwAddr = gwAddr.Unmap()
 	withinNetworks := false
 	for i := range t.vpnNetworks {
 		if t.vpnNetworks[i].Contains(gwAddr) {
@@ -571,9 +577,73 @@ func (t *tun) updateRoutes(r netlink.RouteUpdate) {
 			break
 		}
 	}
-	if !withinNetworks {
+
+	return withinNetworks
+}
+
+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")
+		return gateways
+	}
+
+	// If this route is relevant to our interface and there is a gateway then add it
+	if r.LinkIndex == link.Attrs().Index && len(r.Gw) > 0 {
+		gwAddr, ok := netip.AddrFromSlice(r.Gw)
+		if !ok {
+			t.l.WithField("route", r).Debug("Ignoring route update, invalid gateway address")
+		} else {
+			gwAddr = gwAddr.Unmap()
+
+			if !t.isGatewayInVpnNetworks(gwAddr) {
 				// Gateway isn't in our overlay network, ignore
-		t.l.WithField("route", r).Debug("Ignoring route update, not in our networks")
+				t.l.WithField("route", r).Debug("Ignoring route update, not in our network")
+			} else {
+				gateways = append(gateways, routing.NewGateway(gwAddr, 1))
+			}
+		}
+	}
+
+	for _, p := range r.MultiPath {
+		// If this route is relevant to our interface and there is a gateway then add it
+		if p.LinkIndex == link.Attrs().Index && len(p.Gw) > 0 {
+			gwAddr, ok := netip.AddrFromSlice(p.Gw)
+			if !ok {
+				t.l.WithField("route", r).Debug("Ignoring multipath route update, invalid gateway address")
+			} else {
+				gwAddr = gwAddr.Unmap()
+
+				if !t.isGatewayInVpnNetworks(gwAddr) {
+					// Gateway isn't in our overlay network, ignore
+					t.l.WithField("route", r).Debug("Ignoring route update, not in our network")
+				} else {
+					// p.Hops+1 = weight of the route
+					gateways = append(gateways, routing.NewGateway(gwAddr, p.Hops+1))
+				}
+			}
+		}
+	}
+
+	routing.CalculateBucketsForGateways(gateways)
+	return 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
 	}
 
@@ -589,12 +659,12 @@ func (t *tun) updateRoutes(r netlink.RouteUpdate) {
 	newTree := t.routeTree.Load().Clone()
 
 	if r.Type == unix.RTM_NEWROUTE {
-		t.l.WithField("destination", r.Dst).WithField("via", r.Gw).Info("Adding route")
+		t.l.WithField("destination", dst).WithField("via", gateways).Info("Adding route")
-		newTree.Insert(dst, gwAddr)
+		newTree.Insert(dst, gateways)
 
 	} else {
+		t.l.WithField("destination", dst).WithField("via", gateways).Info("Removing route")
 		newTree.Delete(dst)
-		t.l.WithField("destination", r.Dst).WithField("via", r.Gw).Info("Removing route")
 	}
 	t.routeTree.Store(newTree)
 }

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"
@@ -18,12 +17,44 @@ import (
 	"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"
 )
 
-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 {
@@ -31,42 +62,20 @@ type tun struct {
 	vpnNetworks []netip.Prefix
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
-	routeTree   atomic.Pointer[bart.Table[netip.Addr]]
+	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 == "" {
@@ -76,13 +85,19 @@ 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), ""),
+		fd:          fd,
 		Device:      deviceName,
 		vpnNetworks: vpnNetworks,
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
@@ -104,40 +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
-	// TODO use syscalls instead of exec.Command
+		req.Name = t.deviceBytes()
-	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
+		req.Addr = unix.RawSockaddrInet4{
-	t.l.Debug("command: ", cmd.String())
+			Len:    unix.SizeofSockaddrInet4,
-	if err = cmd.Run(); err != nil {
+			Family: unix.AF_INET,
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+			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(),
 		}
 
-	cmd = exec.Command("/sbin/route", "-n", "add", "-net", cidr.String(), cidr.Addr().String())
+		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-	t.l.Debug("command: ", cmd.String())
+		if err != nil {
-	if err = cmd.Run(); err != nil {
+			return err
-		return fmt.Errorf("failed to run 'route add': %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)
 		}
 
-	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
+		return nil
-	t.l.Debug("command: ", cmd.String())
-	if err = cmd.Run(); err != nil {
-		return fmt.Errorf("failed to run 'ifconfig': %s", err)
 	}
 
-	// Unsafe path routes
+	if cidr.Addr().Is6() {
-	return t.addRoutes(false)
+		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 = 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
+	}
+
+	return fmt.Errorf("unknown address type %v", cidr)
 }
 
 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 {
@@ -177,7 +377,7 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (t *tun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
@@ -196,21 +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 !r.Via.IsValid() || !r.Install {
+		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]interface{}{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
+		} else {
+			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -223,10 +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")
@@ -241,3 +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

@@ -4,67 +4,93 @@
 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
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
-	routeTree   atomic.Pointer[bart.Table[netip.Addr]]
+	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), ""),
+		fd:          fd,
 		Device:      deviceName,
 		vpnNetworks: vpnNetworks,
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
@@ -86,6 +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 {
@@ -123,86 +297,11 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
-func (t *tun) addIp(cidr netip.Prefix) error {
+func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
-	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) RouteFor(ip netip.Addr) netip.Addr {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
 
-func (t *tun) addRoutes(logErrors bool) error {
-	routes := *t.Routes.Load()
-	for _, r := range routes {
-		if !r.Via.IsValid() || !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]interface{}{"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
 }
@@ -212,43 +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 {
-	copy(to, buf[4:])
+			// We don't allow route MTUs so only install routes with a via
-	return n - 4, err
+			continue
-}
-
-// 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
+		err := addRoute(r.Cidr, t.vpnNetworks)
-	ipVer := from[0] >> 4
+		if err != nil {
-	if ipVer == 4 {
+			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
-		buf[3] = syscall.AF_INET
+			if logErrors {
-	} else if ipVer == 6 {
+				retErr.Log(t.l)
-		buf[3] = syscall.AF_INET6
 			} else {
-		return 0, fmt.Errorf("unable to determine IP version from packet")
+				return retErr
+			}
+		} else {
+			t.l.WithField("route", r).Info("Added route")
+		}
 	}
 
-	copy(buf[4:], from)
+	return nil
-
+}
-	n, err := t.ReadWriteCloser.Write(buf)
+
-	return n - 4, err
+func (t *tun) removeRoutes(routes []Route) error {
+	for _, r := range routes {
+		if !r.Install {
+			continue
+		}
+
+		err := delRoute(r.Cidr, t.vpnNetworks)
+		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,
+	}
+
+	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_tester.go

@@ -13,13 +13,14 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 )
 
 type TestTun struct {
 	Device      string
 	vpnNetworks []netip.Prefix
 	Routes      []Route
-	routeTree   *bart.Table[netip.Addr]
+	routeTree   *bart.Table[routing.Gateways]
 	l           *logrus.Logger
 
 	closed    atomic.Bool
@@ -86,7 +87,7 @@ func (t *TestTun) Get(block bool) []byte {
 // Below this is boilerplate implementation to make nebula actually work
 //********************************************************************************************************************//
 
-func (t *TestTun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *TestTun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Lookup(ip)
 	return r
 }

overlay/tun_windows.go

@@ -18,6 +18,7 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
 	"github.com/slackhq/nebula/wintun"
 	"golang.org/x/sys/windows"
@@ -31,7 +32,7 @@ type winTun struct {
 	vpnNetworks []netip.Prefix
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
-	routeTree   atomic.Pointer[bart.Table[netip.Addr]]
+	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
 
 	tun *wintun.NativeTun
@@ -147,15 +148,18 @@ func (t *winTun) addRoutes(logErrors bool) error {
 	foundDefault4 := false
 
 	for _, r := range routes {
-		if !r.Via.IsValid() || !r.Install {
+		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
 
 		// Add our unsafe route
-		err := luid.AddRoute(r.Cidr, r.Via, uint32(r.Metric))
+		// Windows does not support multipath routes natively, so we install only a single route.
+		// This is not a problem as traffic will always be sent to Nebula which handles the multipath routing internally.
+		// In effect this provides multipath routing support to windows supporting loadbalancing and redundancy.
+		err := luid.AddRoute(r.Cidr, r.Via[0].Addr(), uint32(r.Metric))
 		if err != nil {
-			retErr := util.NewContextualError("Failed to add route", map[string]interface{}{"route": r}, err)
+			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 				continue
@@ -198,7 +202,8 @@ func (t *winTun) removeRoutes(routes []Route) error {
 			continue
 		}
 
-		err := luid.DeleteRoute(r.Cidr, r.Via)
+		// See comment on luid.AddRoute
+		err := luid.DeleteRoute(r.Cidr, r.Via[0].Addr())
 		if err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
@@ -208,7 +213,7 @@ func (t *winTun) removeRoutes(routes []Route) error {
 	return nil
 }
 
-func (t *winTun) RouteFor(ip netip.Addr) netip.Addr {
+func (t *winTun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }

overlay/user.go

@@ -6,6 +6,7 @@ import (
 
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/routing"
 )
 
 func NewUserDeviceFromConfig(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
@@ -38,9 +39,13 @@ type UserDevice struct {
 func (d *UserDevice) Activate() error {
 	return nil
 }
+
 func (d *UserDevice) Networks() []netip.Prefix { return d.vpnNetworks }
 func (d *UserDevice) Name() string             { return "faketun0" }
-func (d *UserDevice) RouteFor(ip netip.Addr) netip.Addr { return ip }
+func (d *UserDevice) RoutesFor(ip netip.Addr) routing.Gateways {
+	return routing.Gateways{routing.NewGateway(ip, 1)}
+}
+
 func (d *UserDevice) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 	return d, 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

@@ -33,18 +33,16 @@ type CertState struct {
 	v2Cert           cert.Certificate
 	v2HandshakeBytes []byte
 
-	defaultVersion cert.Version
+	initiatingVersion cert.Version
 	privateKey        []byte
 	pkcs11Backed      bool
 	cipher            string
 
-	psk *Psk
-
 	myVpnNetworks            []netip.Prefix
-	myVpnNetworksTable       *bart.Table[struct{}]
+	myVpnNetworksTable       *bart.Lite
 	myVpnAddrs               []netip.Addr
-	myVpnAddrsTable          *bart.Table[struct{}]
+	myVpnAddrsTable          *bart.Lite
-	myVpnBroadcastAddrsTable *bart.Table[struct{}]
+	myVpnBroadcastAddrsTable *bart.Lite
 }
 
 func NewPKIFromConfig(l *logrus.Logger, c *config.C) (*PKI, error) {
@@ -173,19 +171,8 @@ func (p *PKI) reloadCerts(c *config.C, initial bool) *util.ContextualError {
 		}
 	}
 
-	psk, err := NewPskFromConfig(c)
-	if err != nil {
-		return util.NewContextualError("Failed to load psk from config", nil, err)
-	}
-	if len(psk.keys) > 0 {
-		p.l.WithField("pskMode", psk.mode).WithField("keysLen", len(psk.keys)).
-			Info("pre shared keys are in use")
-	}
-	newState.psk = psk
-
 	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 {
@@ -206,7 +193,7 @@ func (p *PKI) reloadCAPool(c *config.C) *util.ContextualError {
 }
 
 func (cs *CertState) GetDefaultCertificate() cert.Certificate {
-	c := cs.getCertificate(cs.defaultVersion)
+	c := cs.getCertificate(cs.initiatingVersion)
 	if c == nil {
 		panic("No default certificate found")
 	}
@@ -329,37 +316,37 @@ func newCertStateFromConfig(c *config.C) (*CertState, error) {
 		return nil, errors.New("no certificates found in pki.cert")
 	}
 
-	useDefaultVersion := uint32(1)
+	useInitiatingVersion := uint32(1)
 	if v1 == nil {
 		// The only condition that requires v2 as the default is if only a v2 certificate is present
 		// We do this to avoid having to configure it specifically in the config file
-		useDefaultVersion = 2
+		useInitiatingVersion = 2
 	}
 
-	rawDefaultVersion := c.GetUint32("pki.default_version", useDefaultVersion)
+	rawInitiatingVersion := c.GetUint32("pki.initiating_version", useInitiatingVersion)
-	var defaultVersion cert.Version
+	var initiatingVersion cert.Version
-	switch rawDefaultVersion {
+	switch rawInitiatingVersion {
 	case 1:
 		if v1 == nil {
-			return nil, fmt.Errorf("can not use pki.default_version 1 without a v1 certificate in pki.cert")
+			return nil, fmt.Errorf("can not use pki.initiating_version 1 without a v1 certificate in pki.cert")
 		}
-		defaultVersion = cert.Version1
+		initiatingVersion = cert.Version1
 	case 2:
-		defaultVersion = cert.Version2
+		initiatingVersion = cert.Version2
 	default:
-		return nil, fmt.Errorf("unknown pki.default_version: %v", rawDefaultVersion)
+		return nil, fmt.Errorf("unknown pki.initiating_version: %v", rawInitiatingVersion)
 	}
 
-	return newCertState(defaultVersion, v1, v2, isPkcs11, curve, rawKey)
+	return newCertState(initiatingVersion, v1, v2, isPkcs11, curve, rawKey)
 }
 
 func newCertState(dv cert.Version, v1, v2 cert.Certificate, pkcs11backed bool, privateKeyCurve cert.Curve, privateKey []byte) (*CertState, error) {
 	cs := CertState{
 		privateKey:               privateKey,
 		pkcs11Backed:             pkcs11backed,
-		myVpnNetworksTable:       new(bart.Table[struct{}]),
+		myVpnNetworksTable:       new(bart.Lite),
-		myVpnAddrsTable:          new(bart.Table[struct{}]),
+		myVpnAddrsTable:          new(bart.Lite),
-		myVpnBroadcastAddrsTable: new(bart.Table[struct{}]),
+		myVpnBroadcastAddrsTable: new(bart.Lite),
 	}
 
 	if v1 != nil && v2 != nil {
@@ -371,9 +358,11 @@ 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.defaultVersion = dv
+		cs.initiatingVersion = dv
 	}
 
 	if v1 != nil {
@@ -392,8 +381,8 @@ func newCertState(dv cert.Version, v1, v2 cert.Certificate, pkcs11backed bool, p
 		cs.v1Cert = v1
 		cs.v1HandshakeBytes = v1hs
 
-		if cs.defaultVersion == 0 {
+		if cs.initiatingVersion == 0 {
-			cs.defaultVersion = cert.Version1
+			cs.initiatingVersion = cert.Version1
 		}
 	}
 
@@ -413,8 +402,8 @@ func newCertState(dv cert.Version, v1, v2 cert.Certificate, pkcs11backed bool, p
 		cs.v2Cert = v2
 		cs.v2HandshakeBytes = v2hs
 
-		if cs.defaultVersion == 0 {
+		if cs.initiatingVersion == 0 {
-			cs.defaultVersion = cert.Version2
+			cs.initiatingVersion = cert.Version2
 		}
 	}
 
@@ -427,16 +416,16 @@ func newCertState(dv cert.Version, v1, v2 cert.Certificate, pkcs11backed bool, p
 
 	for _, network := range crt.Networks() {
 		cs.myVpnNetworks = append(cs.myVpnNetworks, network)
-		cs.myVpnNetworksTable.Insert(network, struct{}{})
+		cs.myVpnNetworksTable.Insert(network)
 
 		cs.myVpnAddrs = append(cs.myVpnAddrs, network.Addr())
-		cs.myVpnAddrsTable.Insert(netip.PrefixFrom(network.Addr(), network.Addr().BitLen()), struct{}{})
+		cs.myVpnAddrsTable.Insert(netip.PrefixFrom(network.Addr(), network.Addr().BitLen()))
 
 		if network.Addr().Is4() {
 			addr := network.Masked().Addr().As4()
 			mask := net.CIDRMask(network.Bits(), network.Addr().BitLen())
 			binary.BigEndian.PutUint32(addr[:], binary.BigEndian.Uint32(addr[:])|^binary.BigEndian.Uint32(mask))
-			cs.myVpnBroadcastAddrsTable.Insert(netip.PrefixFrom(netip.AddrFrom4(addr), network.Addr().BitLen()), struct{}{})
+			cs.myVpnBroadcastAddrsTable.Insert(netip.PrefixFrom(netip.AddrFrom4(addr), network.Addr().BitLen()))
 		}
 	}
 

psk.go

@@ -1,150 +0,0 @@
-package nebula
-
-import (
-	"crypto/sha256"
-	"errors"
-	"fmt"
-	"io"
-
-	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/util"
-	"golang.org/x/crypto/hkdf"
-)
-
-var ErrNotAPskMode = errors.New("not a psk mode")
-var ErrKeyTooShort = errors.New("key is too short")
-var ErrNotEnoughPskKeys = errors.New("at least 1 key is required")
-
-// MinPskLength is the minimum bytes that we accept for a user defined psk, the choice is arbitrary
-const MinPskLength = 8
-
-type PskMode int
-
-const (
-	PskAccepting PskMode = 0
-	PskSending   PskMode = 1
-	PskEnforced  PskMode = 2
-)
-
-func NewPskMode(m string) (PskMode, error) {
-	switch m {
-	case "accepting":
-		return PskAccepting, nil
-	case "sending":
-		return PskSending, nil
-	case "enforced":
-		return PskEnforced, nil
-	}
-	return PskAccepting, ErrNotAPskMode
-}
-
-func (p PskMode) String() string {
-	switch p {
-	case PskAccepting:
-		return "accepting"
-	case PskSending:
-		return "sending"
-	case PskEnforced:
-		return "enforced"
-	}
-
-	return "unknown"
-}
-
-func (p PskMode) IsValid() bool {
-	switch p {
-	case PskAccepting, PskSending, PskEnforced:
-		return true
-	default:
-		return false
-	}
-}
-
-type Psk struct {
-	// pskMode sets how psk works, ignored, allowed for incoming, or enforced for all
-	mode PskMode
-
-	// primary is the key to use when sending, it may be nil
-	primary []byte
-
-	// keys holds all pre-computed psk hkdfs
-	// Handshakes iterate this directly
-	keys [][]byte
-}
-
-// NewPskFromConfig is a helper for initial boot and config reloading.
-func NewPskFromConfig(c *config.C) (*Psk, error) {
-	sMode := c.GetString("psk.mode", "accepting")
-	mode, err := NewPskMode(sMode)
-	if err != nil {
-		return nil, util.NewContextualError("Could not parse psk.mode", m{"mode": mode}, err)
-	}
-
-	return NewPsk(
-		mode,
-		c.GetStringSlice("psk.keys", nil),
-	)
-}
-
-// NewPsk creates a new Psk object and handles the caching of all accepted keys
-func NewPsk(mode PskMode, keys []string) (*Psk, error) {
-	if !mode.IsValid() {
-		return nil, ErrNotAPskMode
-	}
-
-	psk := &Psk{
-		mode: mode,
-	}
-
-	err := psk.cachePsks(keys)
-	if err != nil {
-		return nil, err
-	}
-
-	return psk, nil
-}
-
-// cachePsks generates all psks we accept and caches them to speed up handshaking
-func (p *Psk) cachePsks(keys []string) error {
-	if p.mode != PskAccepting && len(keys) < 1 {
-		return ErrNotEnoughPskKeys
-	}
-
-	p.keys = [][]byte{}
-
-	for i, rk := range keys {
-		k, err := sha256KdfFromString(rk)
-		if err != nil {
-			return fmt.Errorf("failed to generate key for position %v: %w", i, err)
-		}
-
-		p.keys = append(p.keys, k)
-	}
-
-	if p.mode != PskAccepting {
-		// We are either sending or enforcing, the primary key must the first slot
-		p.primary = p.keys[0]
-	}
-
-	if p.mode != PskEnforced {
-		// If we are not enforcing psk use then a nil psk is allowed
-		p.keys = append(p.keys, nil)
-	}
-
-	return nil
-}
-
-// sha256KdfFromString generates a useful key to use from a provided secret
-func sha256KdfFromString(secret string) ([]byte, error) {
-	if len(secret) < MinPskLength {
-		return nil, ErrKeyTooShort
-	}
-
-	hmacKey := make([]byte, sha256.Size)
-	_, err := io.ReadFull(hkdf.New(sha256.New, []byte(secret), nil, nil), hmacKey)
-	if err != nil {
-		return nil, err
-	}
-
-	return hmacKey, nil
-}

psk_test.go

@@ -1,72 +0,0 @@
-package nebula
-
-import (
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestNewPsk(t *testing.T) {
-	t.Run("mode accepting", func(t *testing.T) {
-		p, err := NewPsk(PskAccepting, nil)
-		require.NoError(t, err)
-		assert.Equal(t, PskAccepting, p.mode)
-		assert.Nil(t, p.keys[0])
-		assert.Nil(t, p.primary)
-
-		p, err = NewPsk(PskAccepting, []string{"1234567"})
-		require.ErrorIs(t, err, ErrKeyTooShort)
-
-		p, err = NewPsk(PskAccepting, []string{"hi there friends"})
-		require.NoError(t, err)
-		assert.Equal(t, PskAccepting, p.mode)
-		assert.Nil(t, p.primary)
-		assert.Len(t, p.keys, 2)
-		assert.Nil(t, p.keys[1])
-
-		expectedCache := []byte{
-			0xb9, 0x8c, 0xdc, 0xac, 0x77, 0xf4, 0x8c, 0xf8, 0x1d, 0xe7, 0xe7, 0xb, 0x53, 0x25, 0xd3, 0x65,
-			0xa3, 0x9f, 0x78, 0xb2, 0xc7, 0x2d, 0xa5, 0xd8, 0x84, 0x81, 0x7b, 0xb5, 0xdb, 0xe0, 0x9a, 0xef,
-		}
-		assert.Equal(t, expectedCache, p.keys[0])
-	})
-
-	t.Run("mode sending", func(t *testing.T) {
-		p, err := NewPsk(PskSending, nil)
-		require.ErrorIs(t, err, ErrNotEnoughPskKeys)
-
-		p, err = NewPsk(PskSending, []string{"1234567"})
-		require.ErrorIs(t, err, ErrKeyTooShort)
-
-		p, err = NewPsk(PskSending, []string{"hi there friends"})
-		require.NoError(t, err)
-		assert.Equal(t, PskSending, p.mode)
-		assert.Len(t, p.keys, 2)
-		assert.Nil(t, p.keys[1])
-
-		expectedCache := []byte{
-			0xb9, 0x8c, 0xdc, 0xac, 0x77, 0xf4, 0x8c, 0xf8, 0x1d, 0xe7, 0xe7, 0xb, 0x53, 0x25, 0xd3, 0x65,
-			0xa3, 0x9f, 0x78, 0xb2, 0xc7, 0x2d, 0xa5, 0xd8, 0x84, 0x81, 0x7b, 0xb5, 0xdb, 0xe0, 0x9a, 0xef,
-		}
-		assert.Equal(t, expectedCache, p.keys[0])
-		assert.Equal(t, p.keys[0], p.primary)
-	})
-
-	t.Run("mode enforced", func(t *testing.T) {
-		p, err := NewPsk(PskEnforced, nil)
-		require.ErrorIs(t, err, ErrNotEnoughPskKeys)
-
-		p, err = NewPsk(PskEnforced, []string{"hi there friends"})
-		require.NoError(t, err)
-		assert.Equal(t, PskEnforced, p.mode)
-		assert.Len(t, p.keys, 1)
-
-		expectedCache := []byte{
-			0xb9, 0x8c, 0xdc, 0xac, 0x77, 0xf4, 0x8c, 0xf8, 0x1d, 0xe7, 0xe7, 0xb, 0x53, 0x25, 0xd3, 0x65,
-			0xa3, 0x9f, 0x78, 0xb2, 0xc7, 0x2d, 0xa5, 0xd8, 0x84, 0x81, 0x7b, 0xb5, 0xdb, 0xe0, 0x9a, 0xef,
-		}
-		assert.Equal(t, expectedCache, p.keys[0])
-		assert.Equal(t, p.keys[0], p.primary)
-	})
-}

punchy_test.go

@@ -27,7 +27,7 @@ func TestNewPunchyFromConfig(t *testing.T) {
 	assert.True(t, p.GetPunch())
 
 	// punchy.punch
-	c.Settings["punchy"] = map[interface{}]interface{}{"punch": true}
+	c.Settings["punchy"] = map[string]any{"punch": true}
 	p = NewPunchyFromConfig(l, c)
 	assert.True(t, p.GetPunch())
 
@@ -37,18 +37,18 @@ func TestNewPunchyFromConfig(t *testing.T) {
 	assert.True(t, p.GetRespond())
 
 	// punchy.respond
-	c.Settings["punchy"] = map[interface{}]interface{}{"respond": true}
+	c.Settings["punchy"] = map[string]any{"respond": true}
 	c.Settings["punch_back"] = false
 	p = NewPunchyFromConfig(l, c)
 	assert.True(t, p.GetRespond())
 
 	// punchy.delay
-	c.Settings["punchy"] = map[interface{}]interface{}{"delay": "1m"}
+	c.Settings["punchy"] = map[string]any{"delay": "1m"}
 	p = NewPunchyFromConfig(l, c)
 	assert.Equal(t, time.Minute, p.GetDelay())
 
 	// punchy.respond_delay
-	c.Settings["punchy"] = map[interface{}]interface{}{"respond_delay": "1m"}
+	c.Settings["punchy"] = map[string]any{"respond_delay": "1m"}
 	p = NewPunchyFromConfig(l, c)
 	assert.Equal(t, time.Minute, p.GetRespondDelay())
 }

relay_manager.go

@@ -241,15 +241,13 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 	logMsg.Info("handleCreateRelayRequest")
 	// Is the source of the relay me? This should never happen, but did happen due to
 	// an issue migrating relays over to newly re-handshaked host info objects.
-	_, found := f.myVpnAddrsTable.Lookup(from)
+	if f.myVpnAddrsTable.Contains(from) {
-	if found {
 		logMsg.WithField("myIP", from).Error("Discarding relay request from myself")
 		return
 	}
 
 	// Is the target of the relay me?
-	_, found = f.myVpnAddrsTable.Lookup(target)
+	if f.myVpnAddrsTable.Contains(target) {
-	if found {
 		existingRelay, ok := h.relayState.QueryRelayForByIp(from)
 		if ok {
 			switch existingRelay.State {

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.
@@ -202,7 +202,9 @@ type RemoteList struct {
 	cache map[netip.Addr]*cache
 
 	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)
 			}

routing/balance.go

@@ -0,0 +1,39 @@
+package routing
+
+import (
+	"net/netip"
+
+	"github.com/slackhq/nebula/firewall"
+)
+
+// Hashes the packet source and destination port and always returns a positive integer
+// Based on 'Prospecting for Hash Functions'
+//   - https://nullprogram.com/blog/2018/07/31/
+//   - https://github.com/skeeto/hash-prospector
+//     [16 21f0aaad 15 d35a2d97 15] = 0.10760229515479501
+func hashPacket(p *firewall.Packet) int {
+	x := (uint32(p.LocalPort) << 16) | uint32(p.RemotePort)
+	x ^= x >> 16
+	x *= 0x21f0aaad
+	x ^= x >> 15
+	x *= 0xd35a2d97
+	x ^= x >> 15
+
+	return int(x) & 0x7FFFFFFF
+}
+
+// For this function to work correctly it requires that the buckets for the gateways have been calculated
+// If the contract is violated balancing will not work properly and the second return value will return false
+func BalancePacket(fwPacket *firewall.Packet, gateways []Gateway) (netip.Addr, bool) {
+	hash := hashPacket(fwPacket)
+
+	for i := range gateways {
+		if hash <= gateways[i].BucketUpperBound() {
+			return gateways[i].Addr(), true
+		}
+	}
+
+	// If you land here then the buckets for the gateways are not properly calculated
+	// Fallback to random routing and let the caller know
+	return gateways[hash%len(gateways)].Addr(), false
+}

routing/balance_test.go

@@ -0,0 +1,144 @@
+package routing
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/slackhq/nebula/firewall"
+	"github.com/stretchr/testify/assert"
+)
+
+func TestPacketsAreBalancedEqually(t *testing.T) {
+
+	gateways := []Gateway{}
+
+	gw1Addr := netip.MustParseAddr("1.0.0.1")
+	gw2Addr := netip.MustParseAddr("1.0.0.2")
+	gw3Addr := netip.MustParseAddr("1.0.0.3")
+
+	gateways = append(gateways, NewGateway(gw1Addr, 1))
+	gateways = append(gateways, NewGateway(gw2Addr, 1))
+	gateways = append(gateways, NewGateway(gw3Addr, 1))
+
+	CalculateBucketsForGateways(gateways)
+
+	gw1count := 0
+	gw2count := 0
+	gw3count := 0
+
+	iterationCount := uint16(65535)
+	for i := uint16(0); i < iterationCount; i++ {
+		packet := firewall.Packet{
+			LocalAddr:  netip.MustParseAddr("192.168.1.1"),
+			RemoteAddr: netip.MustParseAddr("10.0.0.1"),
+			LocalPort:  i,
+			RemotePort: 65535 - i,
+			Protocol:   6, // TCP
+			Fragment:   false,
+		}
+
+		selectedGw, ok := BalancePacket(&packet, gateways)
+		assert.True(t, ok)
+
+		switch selectedGw {
+		case gw1Addr:
+			gw1count += 1
+		case gw2Addr:
+			gw2count += 1
+		case gw3Addr:
+			gw3count += 1
+		}
+
+	}
+
+	// Assert packets are balanced, allow variation of up to 100 packets per gateway
+	assert.InDeltaf(t, iterationCount/3, gw1count, 100, "Expected %d +/- 100, but got %d", iterationCount/3, gw1count)
+	assert.InDeltaf(t, iterationCount/3, gw2count, 100, "Expected %d +/- 100, but got %d", iterationCount/3, gw1count)
+	assert.InDeltaf(t, iterationCount/3, gw3count, 100, "Expected %d +/- 100, but got %d", iterationCount/3, gw1count)
+
+}
+
+func TestPacketsAreBalancedByPriority(t *testing.T) {
+
+	gateways := []Gateway{}
+
+	gw1Addr := netip.MustParseAddr("1.0.0.1")
+	gw2Addr := netip.MustParseAddr("1.0.0.2")
+
+	gateways = append(gateways, NewGateway(gw1Addr, 10))
+	gateways = append(gateways, NewGateway(gw2Addr, 5))
+
+	CalculateBucketsForGateways(gateways)
+
+	gw1count := 0
+	gw2count := 0
+
+	iterationCount := uint16(65535)
+	for i := uint16(0); i < iterationCount; i++ {
+		packet := firewall.Packet{
+			LocalAddr:  netip.MustParseAddr("192.168.1.1"),
+			RemoteAddr: netip.MustParseAddr("10.0.0.1"),
+			LocalPort:  i,
+			RemotePort: 65535 - i,
+			Protocol:   6, // TCP
+			Fragment:   false,
+		}
+
+		selectedGw, ok := BalancePacket(&packet, gateways)
+		assert.True(t, ok)
+
+		switch selectedGw {
+		case gw1Addr:
+			gw1count += 1
+		case gw2Addr:
+			gw2count += 1
+		}
+
+	}
+
+	iterationCountAsFloat := float32(iterationCount)
+
+	assert.InDeltaf(t, iterationCountAsFloat*(2.0/3.0), gw1count, 100, "Expected %d +/- 100, but got %d", iterationCountAsFloat*(2.0/3.0), gw1count)
+	assert.InDeltaf(t, iterationCountAsFloat*(1.0/3.0), gw2count, 100, "Expected %d +/- 100, but got %d", iterationCountAsFloat*(1.0/3.0), gw2count)
+}
+
+func TestBalancePacketDistributsRandomlyAndReturnsFalseIfBucketsNotCalculated(t *testing.T) {
+	gateways := []Gateway{}
+
+	gw1Addr := netip.MustParseAddr("1.0.0.1")
+	gw2Addr := netip.MustParseAddr("1.0.0.2")
+
+	gateways = append(gateways, NewGateway(gw1Addr, 10))
+	gateways = append(gateways, NewGateway(gw2Addr, 5))
+
+	iterationCount := uint16(65535)
+	gw1count := 0
+	gw2count := 0
+
+	for i := uint16(0); i < iterationCount; i++ {
+		packet := firewall.Packet{
+			LocalAddr:  netip.MustParseAddr("192.168.1.1"),
+			RemoteAddr: netip.MustParseAddr("10.0.0.1"),
+			LocalPort:  i,
+			RemotePort: 65535 - i,
+			Protocol:   6, // TCP
+			Fragment:   false,
+		}
+
+		selectedGw, ok := BalancePacket(&packet, gateways)
+		assert.False(t, ok)
+
+		switch selectedGw {
+		case gw1Addr:
+			gw1count += 1
+		case gw2Addr:
+			gw2count += 1
+		}
+
+	}
+
+	assert.Equal(t, int(iterationCount), (gw1count + gw2count))
+	assert.NotEqual(t, 0, gw1count)
+	assert.NotEqual(t, 0, gw2count)
+
+}

routing/gateway.go

@@ -0,0 +1,70 @@
+package routing
+
+import (
+	"fmt"
+	"net/netip"
+)
+
+const (
+	// Sentinal value
+	BucketNotCalculated = -1
+)
+
+type Gateways []Gateway
+
+func (g Gateways) String() string {
+	str := ""
+	for i, gw := range g {
+		str += gw.String()
+		if i < len(g)-1 {
+			str += ", "
+		}
+	}
+	return str
+}
+
+type Gateway struct {
+	addr             netip.Addr
+	weight           int
+	bucketUpperBound int
+}
+
+func NewGateway(addr netip.Addr, weight int) Gateway {
+	return Gateway{addr: addr, weight: weight, bucketUpperBound: BucketNotCalculated}
+}
+
+func (g *Gateway) BucketUpperBound() int {
+	return g.bucketUpperBound
+}
+
+func (g *Gateway) Addr() netip.Addr {
+	return g.addr
+}
+
+func (g *Gateway) String() string {
+	return fmt.Sprintf("{addr: %s, weight: %d}", g.addr, g.weight)
+}
+
+// Divide and round to nearest integer
+func divideAndRound(v uint64, d uint64) uint64 {
+	var tmp uint64 = v + d/2
+	return tmp / d
+}
+
+// Implements Hash-Threshold mapping, equivalent to the implementation in the linux kernel.
+// After this function returns each gateway will have a
+// positive bucketUpperBound with a maximum value of 2147483647 (INT_MAX)
+func CalculateBucketsForGateways(gateways []Gateway) {
+
+	var totalWeight int = 0
+	for i := range gateways {
+		totalWeight += gateways[i].weight
+	}
+
+	var loopWeight int = 0
+	for i := range gateways {
+		loopWeight += gateways[i].weight
+		gateways[i].bucketUpperBound = int(divideAndRound(uint64(loopWeight)<<31, uint64(totalWeight))) - 1
+	}
+
+}

routing/gateway_test.go

@@ -0,0 +1,34 @@
+package routing
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestRebalance3_2Split(t *testing.T) {
+	gateways := []Gateway{}
+
+	gateways = append(gateways, Gateway{addr: netip.Addr{}, weight: 10})
+	gateways = append(gateways, Gateway{addr: netip.Addr{}, weight: 5})
+
+	CalculateBucketsForGateways(gateways)
+
+	assert.Equal(t, 1431655764, gateways[0].bucketUpperBound) // INT_MAX/3*2
+	assert.Equal(t, 2147483647, gateways[1].bucketUpperBound) // INT_MAX
+}
+
+func TestRebalanceEqualSplit(t *testing.T) {
+	gateways := []Gateway{}
+
+	gateways = append(gateways, Gateway{addr: netip.Addr{}, weight: 1})
+	gateways = append(gateways, Gateway{addr: netip.Addr{}, weight: 1})
+	gateways = append(gateways, Gateway{addr: netip.Addr{}, weight: 1})
+
+	CalculateBucketsForGateways(gateways)
+
+	assert.Equal(t, 715827882, gateways[0].bucketUpperBound)  // INT_MAX/3
+	assert.Equal(t, 1431655764, gateways[1].bucketUpperBound) // INT_MAX/3*2
+	assert.Equal(t, 2147483647, gateways[2].bucketUpperBound) // INT_MAX
+}

service/service.go

@@ -9,13 +9,10 @@ import (
 	"math"
 	"net"
 	"net/netip"
-	"os"
 	"strings"
 	"sync"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
-	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay"
 	"golang.org/x/sync/errgroup"
 	"gvisor.dev/gvisor/pkg/buffer"
@@ -46,14 +43,7 @@ type Service struct {
 	}
 }
 
-func New(config *config.C) (*Service, error) {
+func New(control *nebula.Control) (*Service, error) {
-	logger := logrus.New()
-	logger.Out = os.Stdout
-
-	control, err := nebula.Main(config, false, "custom-app", logger, overlay.NewUserDeviceFromConfig)
-	if err != nil {
-		return nil, err
-	}
 	control.Start()
 
 	ctx := control.Context()

service/service_test.go

@@ -5,18 +5,22 @@ import (
 	"context"
 	"errors"
 	"net/netip"
+	"os"
 	"testing"
 	"time"
 
 	"dario.cat/mergo"
+	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/overlay"
 	"golang.org/x/sync/errgroup"
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
-type m map[string]interface{}
+type m = map[string]any
 
 func newSimpleService(caCrt cert.Certificate, caKey []byte, name string, udpIp netip.Addr, overrides m) *Service {
 	_, _, myPrivKey, myPEM := cert_test.NewTestCert(cert.Version2, cert.Curve_CURVE25519, caCrt, caKey, "a", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.PrefixFrom(udpIp, 24)}, nil, []string{})
@@ -71,7 +75,15 @@ func newSimpleService(caCrt cert.Certificate, caKey []byte, name string, udpIp n
 		panic(err)
 	}
 
-	s, err := New(&c)
+	logger := logrus.New()
+	logger.Out = os.Stdout
+
+	control, err := nebula.Main(&c, false, "custom-app", logger, overlay.NewUserDeviceFromConfig)
+	if err != nil {
+		panic(err)
+	}
+
+	s, err := New(control)
 	if err != nil {
 		panic(err)
 	}

ssh.go

@@ -124,10 +124,10 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 	}
 
 	rawKeys := c.Get("sshd.authorized_users")
-	keys, ok := rawKeys.([]interface{})
+	keys, ok := rawKeys.([]any)
 	if ok {
 		for _, rk := range keys {
-			kDef, ok := rk.(map[interface{}]interface{})
+			kDef, ok := rk.(map[string]any)
 			if !ok {
 				l.WithField("sshKeyConfig", rk).Warn("Authorized user had an error, ignoring")
 				continue
@@ -148,7 +148,7 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 					continue
 				}
 
-			case []interface{}:
+			case []any:
 				for _, subK := range v {
 					sk, ok := subK.(string)
 					if !ok {
@@ -190,7 +190,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "list-hostmap",
 		ShortDescription: "List all known previously connected hosts",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshListHostMapFlags{}
 			fl.BoolVar(&s.Json, "json", false, "outputs as json with more information")
@@ -198,7 +198,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 			fl.BoolVar(&s.ByIndex, "by-index", false, "gets all hosts in the hostmap from the index table")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshListHostMap(f.hostMap, fs, w)
 		},
 	})
@@ -206,7 +206,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "list-pending-hostmap",
 		ShortDescription: "List all handshaking hosts",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshListHostMapFlags{}
 			fl.BoolVar(&s.Json, "json", false, "outputs as json with more information")
@@ -214,7 +214,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 			fl.BoolVar(&s.ByIndex, "by-index", false, "gets all hosts in the hostmap from the index table")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshListHostMap(f.handshakeManager, fs, w)
 		},
 	})
@@ -222,14 +222,14 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "list-lighthouse-addrmap",
 		ShortDescription: "List all lighthouse map entries",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshListHostMapFlags{}
 			fl.BoolVar(&s.Json, "json", false, "outputs as json with more information")
 			fl.BoolVar(&s.Pretty, "pretty", false, "pretty prints json, assumes -json")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshListLighthouseMap(f.lightHouse, fs, w)
 		},
 	})
@@ -237,7 +237,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "reload",
 		ShortDescription: "Reloads configuration from disk, same as sending HUP to the process",
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshReload(c, w)
 		},
 	})
@@ -251,7 +251,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "stop-cpu-profile",
 		ShortDescription: "Stops a cpu profile and writes output to the previously provided file",
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			pprof.StopCPUProfile()
 			return w.WriteLine("If a CPU profile was running it is now stopped")
 		},
@@ -278,7 +278,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "log-level",
 		ShortDescription: "Gets or sets the current log level",
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshLogLevel(l, fs, a, w)
 		},
 	})
@@ -286,7 +286,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "log-format",
 		ShortDescription: "Gets or sets the current log format",
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshLogFormat(l, fs, a, w)
 		},
 	})
@@ -294,7 +294,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "version",
 		ShortDescription: "Prints the currently running version of nebula",
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshVersion(f, fs, a, w)
 		},
 	})
@@ -302,14 +302,14 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "device-info",
 		ShortDescription: "Prints information about the network device.",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshDeviceInfoFlags{}
 			fl.BoolVar(&s.Json, "json", false, "outputs as json with more information")
 			fl.BoolVar(&s.Pretty, "pretty", false, "pretty prints json, assumes -json")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshDeviceInfo(f, fs, w)
 		},
 	})
@@ -317,7 +317,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "print-cert",
 		ShortDescription: "Prints the current certificate being used or the certificate for the provided vpn addr",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshPrintCertFlags{}
 			fl.BoolVar(&s.Json, "json", false, "outputs as json")
@@ -325,7 +325,7 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 			fl.BoolVar(&s.Raw, "raw", false, "raw prints the PEM encoded certificate, not compatible with -json or -pretty")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshPrintCert(f, fs, a, w)
 		},
 	})
@@ -333,13 +333,13 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "print-tunnel",
 		ShortDescription: "Prints json details about a tunnel for the provided vpn addr",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshPrintTunnelFlags{}
 			fl.BoolVar(&s.Pretty, "pretty", false, "pretty prints json")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshPrintTunnel(f, fs, a, w)
 		},
 	})
@@ -347,13 +347,13 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "print-relays",
 		ShortDescription: "Prints json details about all relay info",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshPrintTunnelFlags{}
 			fl.BoolVar(&s.Pretty, "pretty", false, "pretty prints json")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshPrintRelays(f, fs, a, w)
 		},
 	})
@@ -361,13 +361,13 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "change-remote",
 		ShortDescription: "Changes the remote address used in the tunnel for the provided vpn addr",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshChangeRemoteFlags{}
 			fl.StringVar(&s.Address, "address", "", "The new remote address, ip:port")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshChangeRemote(f, fs, a, w)
 		},
 	})
@@ -375,13 +375,13 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 	ssh.RegisterCommand(&sshd.Command{
 		Name:             "close-tunnel",
 		ShortDescription: "Closes a tunnel for the provided vpn addr",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshCloseTunnelFlags{}
 			fl.BoolVar(&s.LocalOnly, "local-only", false, "Disables notifying the remote that the tunnel is shutting down")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshCloseTunnel(f, fs, a, w)
 		},
 	})
@@ -390,13 +390,13 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 		Name:             "create-tunnel",
 		ShortDescription: "Creates a tunnel for the provided vpn address",
 		Help:             "The lighthouses will be queried for real addresses but you can provide one as well.",
-		Flags: func() (*flag.FlagSet, interface{}) {
+		Flags: func() (*flag.FlagSet, any) {
 			fl := flag.NewFlagSet("", flag.ContinueOnError)
 			s := sshCreateTunnelFlags{}
 			fl.StringVar(&s.Address, "address", "", "Optionally provide a real remote address, ip:port ")
 			return fl, &s
 		},
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshCreateTunnel(f, fs, a, w)
 		},
 	})
@@ -405,13 +405,13 @@ func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Inter
 		Name:             "query-lighthouse",
 		ShortDescription: "Query the lighthouses for the provided vpn address",
 		Help:             "This command is asynchronous. Only currently known udp addresses will be printed.",
-		Callback: func(fs interface{}, a []string, w sshd.StringWriter) error {
+		Callback: func(fs any, a []string, w sshd.StringWriter) error {
 			return sshQueryLighthouse(f, fs, a, w)
 		},
 	})
 }
 
-func sshListHostMap(hl controlHostLister, a interface{}, w sshd.StringWriter) error {
+func sshListHostMap(hl controlHostLister, a any, w sshd.StringWriter) error {
 	fs, ok := a.(*sshListHostMapFlags)
 	if !ok {
 		return nil
@@ -451,7 +451,7 @@ func sshListHostMap(hl controlHostLister, a interface{}, w sshd.StringWriter) er
 	return nil
 }
 
-func sshListLighthouseMap(lightHouse *LightHouse, a interface{}, w sshd.StringWriter) error {
+func sshListLighthouseMap(lightHouse *LightHouse, a any, w sshd.StringWriter) error {
 	fs, ok := a.(*sshListHostMapFlags)
 	if !ok {
 		return nil
@@ -505,7 +505,7 @@ func sshListLighthouseMap(lightHouse *LightHouse, a interface{}, w sshd.StringWr
 	return nil
 }
 
-func sshStartCpuProfile(fs interface{}, a []string, w sshd.StringWriter) error {
+func sshStartCpuProfile(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		err := w.WriteLine("No path to write profile provided")
 		return err
@@ -527,11 +527,11 @@ func sshStartCpuProfile(fs interface{}, a []string, w sshd.StringWriter) error {
 	return err
 }
 
-func sshVersion(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshVersion(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	return w.WriteLine(fmt.Sprintf("%s", ifce.version))
 }
 
-func sshQueryLighthouse(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshQueryLighthouse(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine("No vpn address was provided")
 	}
@@ -553,7 +553,7 @@ func sshQueryLighthouse(ifce *Interface, fs interface{}, a []string, w sshd.Stri
 	return json.NewEncoder(w.GetWriter()).Encode(cm)
 }
 
-func sshCloseTunnel(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshCloseTunnel(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	flags, ok := fs.(*sshCloseTunnelFlags)
 	if !ok {
 		return nil
@@ -593,7 +593,7 @@ func sshCloseTunnel(ifce *Interface, fs interface{}, a []string, w sshd.StringWr
 	return w.WriteLine("Closed")
 }
 
-func sshCreateTunnel(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshCreateTunnel(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	flags, ok := fs.(*sshCreateTunnelFlags)
 	if !ok {
 		return nil
@@ -638,7 +638,7 @@ func sshCreateTunnel(ifce *Interface, fs interface{}, a []string, w sshd.StringW
 	return w.WriteLine("Created")
 }
 
-func sshChangeRemote(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshChangeRemote(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	flags, ok := fs.(*sshChangeRemoteFlags)
 	if !ok {
 		return nil
@@ -675,7 +675,7 @@ func sshChangeRemote(ifce *Interface, fs interface{}, a []string, w sshd.StringW
 	return w.WriteLine("Changed")
 }
 
-func sshGetHeapProfile(fs interface{}, a []string, w sshd.StringWriter) error {
+func sshGetHeapProfile(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine("No path to write profile provided")
 	}
@@ -696,7 +696,7 @@ func sshGetHeapProfile(fs interface{}, a []string, w sshd.StringWriter) error {
 	return err
 }
 
-func sshMutexProfileFraction(fs interface{}, a []string, w sshd.StringWriter) error {
+func sshMutexProfileFraction(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		rate := runtime.SetMutexProfileFraction(-1)
 		return w.WriteLine(fmt.Sprintf("Current value: %d", rate))
@@ -711,7 +711,7 @@ func sshMutexProfileFraction(fs interface{}, a []string, w sshd.StringWriter) er
 	return w.WriteLine(fmt.Sprintf("New value: %d. Old value: %d", newRate, oldRate))
 }
 
-func sshGetMutexProfile(fs interface{}, a []string, w sshd.StringWriter) error {
+func sshGetMutexProfile(fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine("No path to write profile provided")
 	}
@@ -735,7 +735,7 @@ func sshGetMutexProfile(fs interface{}, a []string, w sshd.StringWriter) error {
 	return w.WriteLine(fmt.Sprintf("Mutex profile created at %s", a))
 }
 
-func sshLogLevel(l *logrus.Logger, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshLogLevel(l *logrus.Logger, fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine(fmt.Sprintf("Log level is: %s", l.Level))
 	}
@@ -749,7 +749,7 @@ func sshLogLevel(l *logrus.Logger, fs interface{}, a []string, w sshd.StringWrit
 	return w.WriteLine(fmt.Sprintf("Log level is: %s", l.Level))
 }
 
-func sshLogFormat(l *logrus.Logger, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshLogFormat(l *logrus.Logger, fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
 		return w.WriteLine(fmt.Sprintf("Log format is: %s", reflect.TypeOf(l.Formatter)))
 	}
@@ -767,7 +767,7 @@ func sshLogFormat(l *logrus.Logger, fs interface{}, a []string, w sshd.StringWri
 	return w.WriteLine(fmt.Sprintf("Log format is: %s", reflect.TypeOf(l.Formatter)))
 }
 
-func sshPrintCert(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshPrintCert(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	args, ok := fs.(*sshPrintCertFlags)
 	if !ok {
 		return nil
@@ -822,7 +822,7 @@ func sshPrintCert(ifce *Interface, fs interface{}, a []string, w sshd.StringWrit
 	return w.WriteLine(cert.String())
 }
 
-func sshPrintRelays(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshPrintRelays(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	args, ok := fs.(*sshPrintTunnelFlags)
 	if !ok {
 		w.WriteLine(fmt.Sprintf("sshPrintRelays failed to convert args type"))
@@ -919,7 +919,7 @@ func sshPrintRelays(ifce *Interface, fs interface{}, a []string, w sshd.StringWr
 	return nil
 }
 
-func sshPrintTunnel(ifce *Interface, fs interface{}, a []string, w sshd.StringWriter) error {
+func sshPrintTunnel(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {
 	args, ok := fs.(*sshPrintTunnelFlags)
 	if !ok {
 		return nil
@@ -951,7 +951,7 @@ func sshPrintTunnel(ifce *Interface, fs interface{}, a []string, w sshd.StringWr
 	return enc.Encode(copyHostInfo(hostInfo, ifce.hostMap.GetPreferredRanges()))
 }
 
-func sshDeviceInfo(ifce *Interface, fs interface{}, w sshd.StringWriter) error {
+func sshDeviceInfo(ifce *Interface, fs any, w sshd.StringWriter) error {
 
 	data := struct {
 		Name string         `json:"name"`

sshd/command.go

@@ -12,7 +12,7 @@ import (
 
 // CommandFlags is a function called before help or command execution to parse command line flags
 // It should return a flag.FlagSet instance and a pointer to the struct that will contain parsed flags
-type CommandFlags func() (*flag.FlagSet, interface{})
+type CommandFlags func() (*flag.FlagSet, any)
 
 // CommandCallback is the function called when your command should execute.
 // fs will be a a pointer to the struct provided by Command.Flags callback, if there was one. -h and -help are reserved
@@ -21,7 +21,7 @@ type CommandFlags func() (*flag.FlagSet, interface{})
 // w is the writer to use when sending messages back to the client.
 // If an error is returned by the callback it is logged locally, the callback should handle messaging errors to the user
 // where appropriate
-type CommandCallback func(fs interface{}, a []string, w StringWriter) error
+type CommandCallback func(fs any, a []string, w StringWriter) error
 
 type Command struct {
 	Name             string
@@ -34,7 +34,7 @@ type Command struct {
 func execCommand(c *Command, args []string, w StringWriter) error {
 	var (
 		fl *flag.FlagSet
-		fs interface{}
+		fs any
 	)
 
 	if c.Flags != nil {
@@ -85,7 +85,7 @@ func lookupCommand(c *radix.Tree, sCmd string) (*Command, error) {
 
 func matchCommand(c *radix.Tree, cmd string) []string {
 	cmds := make([]string, 0)
-	c.WalkPrefix(cmd, func(found string, v interface{}) bool {
+	c.WalkPrefix(cmd, func(found string, v any) bool {
 		cmds = append(cmds, found)
 		return false
 	})
@@ -95,7 +95,7 @@ func matchCommand(c *radix.Tree, cmd string) []string {
 
 func allCommands(c *radix.Tree) []*Command {
 	cmds := make([]*Command, 0)
-	c.WalkPrefix("", func(found string, v interface{}) bool {
+	c.WalkPrefix("", func(found string, v any) bool {
 		cmd, ok := v.(*Command)
 		if ok {
 			cmds = append(cmds, cmd)

sshd/server.go

@@ -86,7 +86,7 @@ func NewSSHServer(l *logrus.Entry) (*SSHServer, error) {
 	s.RegisterCommand(&Command{
 		Name:             "help",
 		ShortDescription: "prints available commands or help <command> for specific usage info",
-		Callback: func(a interface{}, args []string, w StringWriter) error {
+		Callback: func(a any, args []string, w StringWriter) error {
 			return helpCallback(s.commands, args, w)
 		},
 	})

sshd/session.go

@@ -9,13 +9,13 @@ import (
 	"github.com/armon/go-radix"
 	"github.com/sirupsen/logrus"
 	"golang.org/x/crypto/ssh"
-	"golang.org/x/crypto/ssh/terminal"
+	"golang.org/x/term"
 )
 
 type session struct {
 	l        *logrus.Entry
 	c        *ssh.ServerConn
-	term     *terminal.Terminal
+	term     *term.Terminal
 	commands *radix.Tree
 	exitChan chan bool
 }
@@ -31,7 +31,7 @@ func NewSession(commands *radix.Tree, conn *ssh.ServerConn, chans <-chan ssh.New
 	s.commands.Insert("logout", &Command{
 		Name:             "logout",
 		ShortDescription: "Ends the current session",
-		Callback: func(a interface{}, args []string, w StringWriter) error {
+		Callback: func(a any, args []string, w StringWriter) error {
 			s.Close()
 			return nil
 		},
@@ -106,8 +106,8 @@ func (s *session) handleRequests(in <-chan *ssh.Request, channel ssh.Channel) {
 	}
 }
 
-func (s *session) createTerm(channel ssh.Channel) *terminal.Terminal {
+func (s *session) createTerm(channel ssh.Channel) *term.Terminal {
-	term := terminal.NewTerminal(channel, s.c.User()+"@nebula > ")
+	term := term.NewTerminal(channel, s.c.User()+"@nebula > ")
 	term.AutoCompleteCallback = func(line string, pos int, key rune) (newLine string, newPos int, ok bool) {
 		// key 9 is tab
 		if key == 9 {

test/assert.go

@@ -13,7 +13,7 @@ import (
 
 // AssertDeepCopyEqual checks to see if two variables have the same values but DO NOT share any memory
 // There is currently a special case for `time.loc` (as this code traverses into unexported fields)
-func AssertDeepCopyEqual(t *testing.T, a interface{}, b interface{}) {
+func AssertDeepCopyEqual(t *testing.T, a any, b any) {
 	v1 := reflect.ValueOf(a)
 	v2 := reflect.ValueOf(b)
 

test/tun.go

@@ -4,12 +4,14 @@ import (
 	"errors"
 	"io"
 	"net/netip"
+
+	"github.com/slackhq/nebula/routing"
 )
 
 type NoopTun struct{}
 
-func (NoopTun) RouteFor(addr netip.Addr) netip.Addr {
+func (NoopTun) RoutesFor(addr netip.Addr) routing.Gateways {
-	return netip.Addr{}
+	return routing.Gateways{}
 }
 
 func (NoopTun) Activate() error {

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
+	}
+
+	// Golang stdlib doesn't handle EAGAIN correctly in some situations so we do writes ourselves
+	// See https://github.com/golang/go/issues/73919
+	for {
+		//_, _, 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)
 	}
 
-	return rc.Control(func(fd uintptr) {
-		err := syscall.SetsockoptInt(int(fd), unix.IPPROTO_IPV6, unix.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
@@ -122,9 +141,13 @@ func (u *RIOConn) ListenOut(r EncReader) {
 		// Just read one packet at a time
 		n, rua, err := u.receive(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")
+			continue
+		}
 
 		r(netip.AddrPortFrom(netip.AddrFrom16(rua.Addr).Unmap(), (rua.Port>>8)|((rua.Port&0xff)<<8)), buffer[:n])
 	}