Merge pull request #372 from fosrl/dev

1.13.0

Former-commit-id: 0beee679ca

.env.example

@@ -1,5 +1,5 @@
 # Copy this file to .env and fill in your values
 # Required for connecting to Pangolin service
-PANGOLIN_ENDPOINT=https://example.com
+PANGOLIN_ENDPOINT=https://app.pangolin.net
 NEWT_ID=changeme-id
 NEWT_SECRET=changeme-secret

.github/CODEOWNERS

@@ -0,0 +1 @@
+* @oschwartz10612 @miloschwartz

.github/ISSUE_TEMPLATE/1.bug_report.yml

@@ -14,12 +14,13 @@ body:
           label: Environment
           description: Please fill out the relevant details below for your environment.
           value: |
-              - OS Type & Version: (e.g., Ubuntu 22.04)
+              - OS Type & Version:
               - Pangolin Version:
+              - Edition (Community or Enterprise):
               - Gerbil Version:
               - Traefik Version:
               - Newt Version:
-              - Olm Version: (if applicable)
+              - Client Version:
       validations:
           required: true
 

.github/workflows/mirror.yaml

@@ -23,7 +23,7 @@ jobs:
           skopeo --version
 
       - name: Install cosign
-        uses: sigstore/cosign-installer@faadad0cce49287aee09b3a48701e75088a2c6ad # v4.0.0
+        uses: sigstore/cosign-installer@6f9f17788090df1f26f669e9d70d6ae9567deba6 # v4.1.2
 
       - name: Input check
         run: |

.github/workflows/nix-build.yml

@@ -0,0 +1,23 @@
+name: Build Nix package
+
+on:
+  workflow_dispatch:
+  pull_request:
+    paths:
+      - go.mod
+      - go.sum
+
+jobs:
+  nix-build:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+
+      - name: Install Nix
+        uses: DeterminateSystems/nix-installer-action@main
+
+      - name: Build flake package
+        run: |
+          nix build .#pangolin-newt -L

.github/workflows/nix-dependabot-update-hash.yml

@@ -0,0 +1,48 @@
+name: Update Nix Package Hash On Dependabot PRs
+
+on:
+  pull_request:
+    types: [opened, synchronize]
+    branches:
+      - main
+
+jobs:
+  nix-update:
+    if: github.actor == 'dependabot[bot]'
+    runs-on: ubuntu-latest
+
+    permissions:
+      contents: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          ref: ${{ github.head_ref }}
+          token: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Install Nix
+        uses: DeterminateSystems/nix-installer-action@main
+
+      - name: Run nix-update
+        run: |
+          nix run nixpkgs#nix-update -- --flake pangolin-newt --no-src --version skip
+
+      - name: Check for changes
+        id: changes
+        run: |
+          if git diff --quiet; then
+            echo "changed=false" >> "$GITHUB_OUTPUT"
+          else
+            echo "changed=true" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Commit and push changes
+        if: steps.changes.outputs.changed == 'true'
+        run: |
+          git config user.name "dependabot[bot]"
+          git config user.email "dependabot[bot]@users.noreply.github.com"
+
+          git add .
+          git commit -m "chore(nix): fix hash for updated go dependencies"
+          git push

.github/workflows/stale-bot.yml

@@ -0,0 +1,37 @@
+name: Mark and Close Stale Issues
+
+on:
+  schedule:
+    - cron: '0 0 * * *'
+  workflow_dispatch: # Allow manual trigger
+
+permissions:
+  contents: write # only for delete-branch option
+  issues: write
+  pull-requests: write
+
+jobs:
+  stale:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/stale@eb5cf3af3ac0a1aa4c9c45633dd1ae542a27a899 # v10.3.0
+        with:
+          days-before-stale: 14
+          days-before-close: 14
+          stale-issue-message: 'This issue has been automatically marked as stale due to 14 days of inactivity. It will be closed in 14 days if no further activity occurs.'
+          close-issue-message: 'This issue has been automatically closed due to inactivity. If you believe this is still relevant, please open a new issue with up-to-date information.'
+          stale-issue-label: 'stale'
+          
+          exempt-issue-labels: 'needs investigating, networking, new feature, reverse proxy, bug, api, authentication, documentation, enhancement, help wanted, good first issue, question'
+          
+          exempt-all-issue-assignees: true
+          
+          only-labels: ''
+          exempt-pr-labels: ''
+          days-before-pr-stale: -1
+          days-before-pr-close: -1
+          
+          operations-per-run: 100
+          remove-stale-when-updated: true
+          delete-branch: false
+          enable-statistics: true

.github/workflows/test.yml

@@ -11,21 +11,43 @@ on:
 
 jobs:
   test:
-    runs-on: amd64-runner
-
+    runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+      - name: Checkout repository
+        uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 # v6.0.1
 
       - name: Set up Go
-        uses: actions/setup-go@4dc6199c7b1a012772edbd06daecab0f50c9053c # v6.1.0
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6.4.0
         with:
-          go-version: 1.25
+          go-version-file: go.mod
 
-      - name: Build go
-        run: go build
+      - name: Run Go tests
+        run: make test
 
-      - name: Build Docker image
-        run: make build
+  build:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        target:
+          - local
+          - docker-build
+          - go-build-release-darwin-amd64
+          - go-build-release-darwin-arm64
+          - go-build-release-freebsd-amd64
+          - go-build-release-freebsd-arm64
+          - go-build-release-linux-amd64
+          - go-build-release-linux-arm32-v6
+          - go-build-release-linux-arm32-v7
+          - go-build-release-linux-riscv64
+          - go-build-release-windows-amd64
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 # v6.0.1
 
-      - name: Build binaries
-        run: make go-build-release
+      - name: Set up Go
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6.4.0
+        with:
+          go-version-file: go.mod
+
+      - name: Build targets via `make`
+        run: make ${{ matrix.target }}

.gitignore

@@ -5,4 +5,6 @@ nohup.out
 *.iml
 certs/
 newt_arm64
-key
+key
+/.direnv/
+/result*

Dockerfile

@@ -1,4 +1,5 @@
-FROM golang:1.25-alpine AS builder
+# FROM golang:1.25-alpine AS builder
+FROM public.ecr.aws/docker/library/golang:1.25-alpine AS builder
 
 # Install git and ca-certificates
 RUN apk --no-cache add ca-certificates git tzdata
@@ -16,15 +17,20 @@ RUN go mod download
 COPY . .
 
 # Build the application
-RUN CGO_ENABLED=0 GOOS=linux go build -ldflags="-s -w" -o /newt
+ARG VERSION=dev
+RUN CGO_ENABLED=0 GOOS=linux go build -ldflags="-s -w -X main.newtVersion=${VERSION}" -o /newt
 
-FROM alpine:3.23 AS runner
+FROM public.ecr.aws/docker/library/alpine:3.23 AS runner
 
-RUN apk --no-cache add ca-certificates tzdata
+RUN apk --no-cache add ca-certificates tzdata iputils
 
 COPY --from=builder /newt /usr/local/bin/
 COPY entrypoint.sh /
 
+# Marks this as an official Fossorial container image.
+# Auto-update is disabled in official images — update by pulling a new image tag.
+ENV NEWT_SYSTEM_SUBSTRATE="CONTAINER"
+
 # Admin/metrics endpoint (Prometheus scrape)
 EXPOSE 2112
 

Makefile

@@ -1,25 +1,76 @@
+.PHONY: all local test docker-build docker-build-release
 
 all: local
 
+VERSION ?= dev
+LDFLAGS = -X main.newtVersion=$(VERSION)
+
+local:
+	CGO_ENABLED=0 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=$(shell go env GOOS)_$(shell go env GOARCH)" -o ./bin/newt
+
+test:
+	go test ./...
+
+docker-build:
+	docker build -t fosrl/newt:latest .
+
 docker-build-release:
 	@if [ -z "$(tag)" ]; then \
 		echo "Error: tag is required. Usage: make docker-build-release tag=<tag>"; \
 		exit 1; \
 	fi
-	docker buildx build --platform linux/arm/v7,linux/arm64,linux/amd64 -t fosrl/newt:latest -f Dockerfile --push .
-	docker buildx build --platform linux/arm/v7,linux/arm64,linux/amd64 -t fosrl/newt:$(tag) -f Dockerfile --push .
+	docker buildx build . \
+		--platform linux/arm/v7,linux/arm64,linux/amd64 \
+		-t fosrl/newt:latest \
+		-t fosrl/newt:$(tag) \
+		-f Dockerfile \
+		--push
 
-local: 
-	CGO_ENABLED=0 go build -o ./bin/newt
+.PHONY: go-build-release \
+        go-build-release-linux-arm64 go-build-release-linux-arm32-v7 \
+        go-build-release-linux-arm32-v6 go-build-release-linux-amd64 \
+        go-build-release-linux-riscv64 go-build-release-darwin-arm64 \
+        go-build-release-darwin-amd64 go-build-release-windows-amd64 \
+        go-build-release-freebsd-amd64 go-build-release-freebsd-arm64
 
-go-build-release:
-	CGO_ENABLED=0 GOOS=linux GOARCH=arm64 go build -o bin/newt_linux_arm64
-	CGO_ENABLED=0 GOOS=linux GOARCH=arm GOARM=7 go build -o bin/newt_linux_arm32
-	CGO_ENABLED=0 GOOS=linux GOARCH=arm GOARM=6 go build -o bin/newt_linux_arm32v6
-	CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -o bin/newt_linux_amd64
-	CGO_ENABLED=0 GOOS=linux GOARCH=riscv64 go build -o bin/newt_linux_riscv64
-	CGO_ENABLED=0 GOOS=darwin GOARCH=arm64 go build -o bin/newt_darwin_arm64
-	CGO_ENABLED=0 GOOS=darwin GOARCH=amd64 go build -o bin/newt_darwin_amd64
-	CGO_ENABLED=0 GOOS=windows GOARCH=amd64 go build -o bin/newt_windows_amd64.exe
-	CGO_ENABLED=0 GOOS=freebsd GOARCH=amd64 go build -o bin/newt_freebsd_amd64
-	CGO_ENABLED=0 GOOS=freebsd GOARCH=arm64 go build -o bin/newt_freebsd_arm64
+go-build-release: \
+    go-build-release-linux-arm64 \
+    go-build-release-linux-arm32-v7 \
+    go-build-release-linux-arm32-v6 \
+    go-build-release-linux-amd64 \
+    go-build-release-linux-riscv64 \
+    go-build-release-darwin-arm64 \
+    go-build-release-darwin-amd64 \
+    go-build-release-windows-amd64 \
+    go-build-release-freebsd-amd64 \
+    go-build-release-freebsd-arm64
+
+go-build-release-linux-arm64:
+	CGO_ENABLED=0 GOOS=linux GOARCH=arm64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=linux_arm64" -o bin/newt_linux_arm64
+
+go-build-release-linux-arm32-v7:
+	CGO_ENABLED=0 GOOS=linux GOARCH=arm GOARM=7 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=linux_arm32" -o bin/newt_linux_arm32
+
+go-build-release-linux-arm32-v6:
+	CGO_ENABLED=0 GOOS=linux GOARCH=arm GOARM=6 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=linux_arm32v6" -o bin/newt_linux_arm32v6
+
+go-build-release-linux-amd64:
+	CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=linux_amd64" -o bin/newt_linux_amd64
+
+go-build-release-linux-riscv64:
+	CGO_ENABLED=0 GOOS=linux GOARCH=riscv64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=linux_riscv64" -o bin/newt_linux_riscv64
+
+go-build-release-darwin-arm64:
+	CGO_ENABLED=0 GOOS=darwin GOARCH=arm64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=darwin_arm64" -o bin/newt_darwin_arm64
+
+go-build-release-darwin-amd64:
+	CGO_ENABLED=0 GOOS=darwin GOARCH=amd64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=darwin_amd64" -o bin/newt_darwin_amd64
+
+go-build-release-windows-amd64:
+	CGO_ENABLED=0 GOOS=windows GOARCH=amd64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=windows_amd64" -o bin/newt_windows_amd64.exe
+
+go-build-release-freebsd-amd64:
+	CGO_ENABLED=0 GOOS=freebsd GOARCH=amd64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=freebsd_amd64" -o bin/newt_freebsd_amd64
+
+go-build-release-freebsd-arm64:
+	CGO_ENABLED=0 GOOS=freebsd GOARCH=arm64 go build -ldflags "$(LDFLAGS) -X main.newtPlatform=freebsd_arm64" -o bin/newt_freebsd_arm64

authdaemon.go

@@ -0,0 +1,150 @@
+package main
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"runtime"
+
+	"github.com/fosrl/newt/authdaemon"
+	"github.com/fosrl/newt/logger"
+)
+
+const (
+	defaultPrincipalsPath = "/var/run/auth-daemon/principals"
+	defaultCACertPath     = "/etc/ssh/ca.pem"
+)
+
+var (
+	errPresharedKeyRequired = errors.New("auth-daemon-key is required when --auth-daemon is enabled")
+	errRootRequired         = errors.New("auth-daemon must be run as root (use sudo)")
+	authDaemonServer        *authdaemon.Server // Global auth daemon server instance
+)
+
+// startAuthDaemon initializes and starts the auth daemon in the background.
+// It validates requirements (Linux, root, preshared key) and starts the server
+// in a goroutine so it runs alongside normal newt operation.
+func startAuthDaemon(ctx context.Context) error {
+	// Validation
+	if runtime.GOOS != "linux" {
+		return fmt.Errorf("auth-daemon is only supported on Linux, not %s", runtime.GOOS)
+	}
+	if os.Geteuid() != 0 {
+		return errRootRequired
+	}
+
+	// Use defaults if not set
+	principalsFile := authDaemonPrincipalsFile
+	if principalsFile == "" {
+		principalsFile = defaultPrincipalsPath
+	}
+	caCertPath := authDaemonCACertPath
+	if caCertPath == "" {
+		caCertPath = defaultCACertPath
+	}
+
+	// Create auth daemon server
+	cfg := authdaemon.Config{
+		DisableHTTPS:           true,                   // We run without HTTP server in newt
+		PresharedKey:           "this-key-is-not-used", // Not used in embedded mode, but set to non-empty to satisfy validation
+		PrincipalsFilePath:     principalsFile,
+		CACertPath:             caCertPath,
+		Force:                  true,
+		GenerateRandomPassword: authDaemonGenerateRandomPassword,
+	}
+
+	srv, err := authdaemon.NewServer(cfg)
+	if err != nil {
+		return fmt.Errorf("create auth daemon server: %w", err)
+	}
+
+	authDaemonServer = srv
+
+	// Start the auth daemon in a goroutine so it runs alongside newt
+	go func() {
+		logger.Info("Auth daemon starting (native mode, no HTTP server)")
+		if err := srv.Run(ctx); err != nil {
+			logger.Error("Auth daemon error: %v", err)
+		}
+		logger.Info("Auth daemon stopped")
+	}()
+
+	return nil
+}
+
+// runPrincipalsCmd executes the principals subcommand logic
+func runPrincipalsCmd(args []string) {
+	opts := struct {
+		PrincipalsFile string
+		Username       string
+	}{
+		PrincipalsFile: defaultPrincipalsPath,
+	}
+
+	// Parse flags manually
+	for i := 0; i < len(args); i++ {
+		switch args[i] {
+		case "--principals-file":
+			if i+1 >= len(args) {
+				fmt.Fprintf(os.Stderr, "Error: --principals-file requires a value\n")
+				os.Exit(1)
+			}
+			opts.PrincipalsFile = args[i+1]
+			i++
+		case "--username":
+			if i+1 >= len(args) {
+				fmt.Fprintf(os.Stderr, "Error: --username requires a value\n")
+				os.Exit(1)
+			}
+			opts.Username = args[i+1]
+			i++
+		case "--help", "-h":
+			printPrincipalsHelp()
+			os.Exit(0)
+		default:
+			fmt.Fprintf(os.Stderr, "Error: unknown flag: %s\n", args[i])
+			printPrincipalsHelp()
+			os.Exit(1)
+		}
+	}
+
+	// Validation
+	if opts.Username == "" {
+		fmt.Fprintf(os.Stderr, "Error: username is required\n")
+		printPrincipalsHelp()
+		os.Exit(1)
+	}
+
+	// Get principals
+	list, err := authdaemon.GetPrincipals(opts.PrincipalsFile, opts.Username)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "Error: %v\n", err)
+		os.Exit(1)
+	}
+	if len(list) == 0 {
+		fmt.Println("")
+		return
+	}
+	for _, principal := range list {
+		fmt.Println(principal)
+	}
+}
+
+func printPrincipalsHelp() {
+	fmt.Fprintf(os.Stderr, `Usage: newt principals [flags]
+
+Output principals for a username (for AuthorizedPrincipalsCommand in sshd_config).
+Read the principals file and print principals that match the given username, one per line.
+Configure in sshd_config with AuthorizedPrincipalsCommand and %%u for the username.
+
+Flags:
+  --principals-file string   Path to the principals file (default "%s")
+  --username string          Username to look up (required)
+  --help, -h                 Show this help message
+
+Example:
+  newt principals --username alice
+
+`, defaultPrincipalsPath)
+}

authdaemon/connection.go

@@ -0,0 +1,27 @@
+package authdaemon
+
+import (
+	"github.com/fosrl/newt/logger"
+)
+
+// ProcessConnection runs the same logic as POST /connection: CA cert, user create/reconcile, principals.
+// Use this when DisableHTTPS is true (e.g. embedded in Newt) instead of calling the API.
+func (s *Server) ProcessConnection(req ConnectionRequest) {
+	logger.Info("connection: niceId=%q username=%q metadata.sudoMode=%q metadata.sudoCommands=%v metadata.homedir=%v metadata.groups=%v",
+		req.NiceId, req.Username, req.Metadata.SudoMode, req.Metadata.SudoCommands, req.Metadata.Homedir, req.Metadata.Groups)
+
+	cfg := &s.cfg
+	if cfg.CACertPath != "" {
+		if err := writeCACertIfNotExists(cfg.CACertPath, req.CaCert, cfg.Force); err != nil {
+			logger.Warn("auth-daemon: write CA cert: %v", err)
+		}
+	}
+	if err := ensureUser(req.Username, req.Metadata, s.cfg.GenerateRandomPassword); err != nil {
+		logger.Warn("auth-daemon: ensure user: %v", err)
+	}
+	if cfg.PrincipalsFilePath != "" && req.NiceId != "" {
+		if err := writePrincipals(cfg.PrincipalsFilePath, req.Username, req.NiceId); err != nil {
+			logger.Warn("auth-daemon: write principals: %v", err)
+		}
+	}
+}

authdaemon/host_linux.go

@@ -0,0 +1,396 @@
+//go:build linux
+
+package authdaemon
+
+import (
+	"bufio"
+	"crypto/rand"
+	"encoding/hex"
+	"encoding/json"
+	"fmt"
+	"os"
+	"os/exec"
+	"os/user"
+	"path/filepath"
+	"strconv"
+	"strings"
+
+	"github.com/fosrl/newt/logger"
+)
+
+// writeCACertIfNotExists writes contents to path. If the file already exists: when force is false, skip; when force is true, overwrite only if content differs.
+func writeCACertIfNotExists(path, contents string, force bool) error {
+	contents = strings.TrimSpace(contents)
+	if contents != "" && !strings.HasSuffix(contents, "\n") {
+		contents += "\n"
+	}
+	existing, err := os.ReadFile(path)
+	if err == nil {
+		existingStr := strings.TrimSpace(string(existing))
+		if existingStr != "" && !strings.HasSuffix(existingStr, "\n") {
+			existingStr += "\n"
+		}
+		if existingStr == contents {
+			logger.Debug("auth-daemon: CA cert unchanged at %s, skipping write", path)
+			return nil
+		}
+		if !force {
+			logger.Debug("auth-daemon: CA cert already exists at %s, skipping write (Force disabled)", path)
+			return nil
+		}
+	} else if !os.IsNotExist(err) {
+		return fmt.Errorf("read %s: %w", path, err)
+	}
+	dir := filepath.Dir(path)
+	if err := os.MkdirAll(dir, 0755); err != nil {
+		return fmt.Errorf("mkdir %s: %w", dir, err)
+	}
+	if err := os.WriteFile(path, []byte(contents), 0644); err != nil {
+		return fmt.Errorf("write CA cert: %w", err)
+	}
+	logger.Info("auth-daemon: wrote CA cert to %s", path)
+	return nil
+}
+
+// writePrincipals updates the principals file at path: JSON object keyed by username, value is array of principals. Adds username and niceId to that user's list (deduped).
+func writePrincipals(path, username, niceId string) error {
+	if path == "" {
+		return nil
+	}
+	username = strings.TrimSpace(username)
+	niceId = strings.TrimSpace(niceId)
+	if username == "" {
+		return nil
+	}
+	data := make(map[string][]string)
+	if raw, err := os.ReadFile(path); err == nil {
+		_ = json.Unmarshal(raw, &data)
+	}
+	list := data[username]
+	seen := make(map[string]struct{}, len(list)+2)
+	for _, p := range list {
+		seen[p] = struct{}{}
+	}
+	for _, p := range []string{username, niceId} {
+		if p == "" {
+			continue
+		}
+		if _, ok := seen[p]; !ok {
+			seen[p] = struct{}{}
+			list = append(list, p)
+		}
+	}
+	data[username] = list
+	body, err := json.Marshal(data)
+	if err != nil {
+		return fmt.Errorf("marshal principals: %w", err)
+	}
+	dir := filepath.Dir(path)
+	if err := os.MkdirAll(dir, 0755); err != nil {
+		return fmt.Errorf("mkdir %s: %w", dir, err)
+	}
+	if err := os.WriteFile(path, body, 0644); err != nil {
+		return fmt.Errorf("write principals: %w", err)
+	}
+	logger.Debug("auth-daemon: wrote principals to %s", path)
+	return nil
+}
+
+// sudoGroup returns the name of the sudo group (wheel or sudo) that exists on the system. Prefers wheel.
+func sudoGroup() string {
+	f, err := os.Open("/etc/group")
+	if err != nil {
+		return "sudo"
+	}
+	defer f.Close()
+	sc := bufio.NewScanner(f)
+	hasWheel := false
+	hasSudo := false
+	for sc.Scan() {
+		line := sc.Text()
+		if strings.HasPrefix(line, "wheel:") {
+			hasWheel = true
+		}
+		if strings.HasPrefix(line, "sudo:") {
+			hasSudo = true
+		}
+	}
+	if hasWheel {
+		return "wheel"
+	}
+	if hasSudo {
+		return "sudo"
+	}
+	return "sudo"
+}
+
+// setRandomPassword generates a random password and sets it for username via chpasswd.
+// Used when GenerateRandomPassword is true so SSH with PermitEmptyPasswords no can accept the user.
+func setRandomPassword(username string) error {
+	b := make([]byte, 16)
+	if _, err := rand.Read(b); err != nil {
+		return fmt.Errorf("generate password: %w", err)
+	}
+	password := hex.EncodeToString(b)
+	cmd := exec.Command("chpasswd")
+	cmd.Stdin = strings.NewReader(username + ":" + password)
+	if out, err := cmd.CombinedOutput(); err != nil {
+		return fmt.Errorf("chpasswd: %w (output: %s)", err, string(out))
+	}
+	return nil
+}
+
+const skelDir = "/etc/skel"
+
+// copySkelInto copies files from srcDir (e.g. /etc/skel) into dstDir (e.g. user's home).
+// Only creates files that don't already exist. All created paths are chowned to uid:gid.
+func copySkelInto(srcDir, dstDir string, uid, gid int) {
+	entries, err := os.ReadDir(srcDir)
+	if err != nil {
+		if !os.IsNotExist(err) {
+			logger.Warn("auth-daemon: read %s: %v", srcDir, err)
+		}
+		return
+	}
+	for _, e := range entries {
+		name := e.Name()
+		src := filepath.Join(srcDir, name)
+		dst := filepath.Join(dstDir, name)
+		if e.IsDir() {
+			if st, err := os.Stat(dst); err == nil && st.IsDir() {
+				copySkelInto(src, dst, uid, gid)
+				continue
+			}
+			if err := os.MkdirAll(dst, 0755); err != nil {
+				logger.Warn("auth-daemon: mkdir %s: %v", dst, err)
+				continue
+			}
+			if err := os.Chown(dst, uid, gid); err != nil {
+				logger.Warn("auth-daemon: chown %s: %v", dst, err)
+			}
+			copySkelInto(src, dst, uid, gid)
+			continue
+		}
+		if _, err := os.Stat(dst); err == nil {
+			continue
+		}
+		data, err := os.ReadFile(src)
+		if err != nil {
+			logger.Warn("auth-daemon: read %s: %v", src, err)
+			continue
+		}
+		if err := os.WriteFile(dst, data, 0644); err != nil {
+			logger.Warn("auth-daemon: write %s: %v", dst, err)
+			continue
+		}
+		if err := os.Chown(dst, uid, gid); err != nil {
+			logger.Warn("auth-daemon: chown %s: %v", dst, err)
+		}
+	}
+}
+
+// ensureUser creates the system user if missing, or reconciles sudo and homedir to match meta.
+func ensureUser(username string, meta ConnectionMetadata, generateRandomPassword bool) error {
+	if username == "" {
+		return nil
+	}
+	u, err := user.Lookup(username)
+	if err != nil {
+		if _, ok := err.(user.UnknownUserError); !ok {
+			return fmt.Errorf("lookup user %s: %w", username, err)
+		}
+		return createUser(username, meta, generateRandomPassword)
+	}
+	return reconcileUser(u, meta)
+}
+
+// desiredGroups returns the exact list of supplementary groups the user should have:
+// meta.Groups plus the sudo group when meta.SudoMode is "full" (deduped).
+func desiredGroups(meta ConnectionMetadata) []string {
+	seen := make(map[string]struct{})
+	var out []string
+	for _, g := range meta.Groups {
+		g = strings.TrimSpace(g)
+		if g == "" {
+			continue
+		}
+		if _, ok := seen[g]; ok {
+			continue
+		}
+		seen[g] = struct{}{}
+		out = append(out, g)
+	}
+	if meta.SudoMode == "full" {
+		sg := sudoGroup()
+		if _, ok := seen[sg]; !ok {
+			out = append(out, sg)
+		}
+	}
+	return out
+}
+
+// setUserGroups sets the user's supplementary groups to exactly groups (local mirrors metadata).
+// When groups is empty, clears all supplementary groups (usermod -G "").
+func setUserGroups(username string, groups []string) {
+	list := strings.Join(groups, ",")
+	cmd := exec.Command("usermod", "-G", list, username)
+	if out, err := cmd.CombinedOutput(); err != nil {
+		logger.Warn("auth-daemon: usermod -G %s: %v (output: %s)", list, err, string(out))
+	} else {
+		logger.Info("auth-daemon: set %s supplementary groups to %s", username, list)
+	}
+}
+
+func createUser(username string, meta ConnectionMetadata, generateRandomPassword bool) error {
+	args := []string{"-s", "/bin/bash"}
+	if meta.Homedir {
+		args = append(args, "-m")
+	} else {
+		args = append(args, "-M")
+	}
+	args = append(args, username)
+	cmd := exec.Command("useradd", args...)
+	if out, err := cmd.CombinedOutput(); err != nil {
+		return fmt.Errorf("useradd %s: %w (output: %s)", username, err, string(out))
+	}
+	logger.Info("auth-daemon: created user %s (homedir=%v)", username, meta.Homedir)
+	if generateRandomPassword {
+		if err := setRandomPassword(username); err != nil {
+			logger.Warn("auth-daemon: set random password for %s: %v", username, err)
+		} else {
+			logger.Info("auth-daemon: set random password for %s (PermitEmptyPasswords no)", username)
+		}
+	}
+	if meta.Homedir {
+		if u, err := user.Lookup(username); err == nil && u.HomeDir != "" {
+			uid, gid := mustAtoi(u.Uid), mustAtoi(u.Gid)
+			copySkelInto(skelDir, u.HomeDir, uid, gid)
+		}
+	}
+	setUserGroups(username, desiredGroups(meta))
+	switch meta.SudoMode {
+	case "full":
+		if err := configurePasswordlessSudo(username); err != nil {
+			logger.Warn("auth-daemon: configure passwordless sudo for %s: %v", username, err)
+		}
+	case "commands":
+		if len(meta.SudoCommands) > 0 {
+			if err := configureSudoCommands(username, meta.SudoCommands); err != nil {
+				logger.Warn("auth-daemon: configure sudo commands for %s: %v", username, err)
+			}
+		}
+	default:
+		removeSudoers(username)
+	}
+	return nil
+}
+
+const sudoersFilePrefix = "90-pangolin-"
+
+func sudoersPath(username string) string {
+	return filepath.Join("/etc/sudoers.d", sudoersFilePrefix+username)
+}
+
+// writeSudoersFile writes content to the user's sudoers.d file and validates with visudo.
+func writeSudoersFile(username, content string) error {
+	sudoersFile := sudoersPath(username)
+	tmpFile := sudoersFile + ".tmp"
+	if err := os.WriteFile(tmpFile, []byte(content), 0440); err != nil {
+		return fmt.Errorf("write temp sudoers file: %w", err)
+	}
+	cmd := exec.Command("visudo", "-c", "-f", tmpFile)
+	if out, err := cmd.CombinedOutput(); err != nil {
+		os.Remove(tmpFile)
+		return fmt.Errorf("visudo validation failed: %w (output: %s)", err, string(out))
+	}
+	if err := os.Rename(tmpFile, sudoersFile); err != nil {
+		os.Remove(tmpFile)
+		return fmt.Errorf("move sudoers file: %w", err)
+	}
+	return nil
+}
+
+// configurePasswordlessSudo creates a sudoers.d file to allow passwordless sudo for the user.
+func configurePasswordlessSudo(username string) error {
+	content := fmt.Sprintf("# Created by Pangolin auth-daemon\n%s ALL=(ALL) NOPASSWD:ALL\n", username)
+	if err := writeSudoersFile(username, content); err != nil {
+		return err
+	}
+	logger.Info("auth-daemon: configured passwordless sudo for %s", username)
+	return nil
+}
+
+// configureSudoCommands creates a sudoers.d file allowing only the listed commands (NOPASSWD).
+// Each command should be a full path (e.g. /usr/bin/systemctl).
+func configureSudoCommands(username string, commands []string) error {
+	var b strings.Builder
+	b.WriteString("# Created by Pangolin auth-daemon (restricted commands)\n")
+	n := 0
+	for _, c := range commands {
+		c = strings.TrimSpace(c)
+		if c == "" {
+			continue
+		}
+		fmt.Fprintf(&b, "%s ALL=(ALL) NOPASSWD: %s\n", username, c)
+		n++
+	}
+	if n == 0 {
+		return fmt.Errorf("no valid sudo commands")
+	}
+	if err := writeSudoersFile(username, b.String()); err != nil {
+		return err
+	}
+	logger.Info("auth-daemon: configured restricted sudo for %s (%d commands)", username, len(commands))
+	return nil
+}
+
+// removeSudoers removes the sudoers.d file for the user.
+func removeSudoers(username string) {
+	sudoersFile := sudoersPath(username)
+	if err := os.Remove(sudoersFile); err != nil && !os.IsNotExist(err) {
+		logger.Warn("auth-daemon: remove sudoers for %s: %v", username, err)
+	} else if err == nil {
+		logger.Info("auth-daemon: removed sudoers for %s", username)
+	}
+}
+
+func mustAtoi(s string) int {
+	n, _ := strconv.Atoi(s)
+	return n
+}
+
+func reconcileUser(u *user.User, meta ConnectionMetadata) error {
+	setUserGroups(u.Username, desiredGroups(meta))
+	switch meta.SudoMode {
+	case "full":
+		if err := configurePasswordlessSudo(u.Username); err != nil {
+			logger.Warn("auth-daemon: configure passwordless sudo for %s: %v", u.Username, err)
+		}
+	case "commands":
+		if len(meta.SudoCommands) > 0 {
+			if err := configureSudoCommands(u.Username, meta.SudoCommands); err != nil {
+				logger.Warn("auth-daemon: configure sudo commands for %s: %v", u.Username, err)
+			}
+		} else {
+			removeSudoers(u.Username)
+		}
+	default:
+		removeSudoers(u.Username)
+	}
+	if meta.Homedir && u.HomeDir != "" {
+		uid, gid := mustAtoi(u.Uid), mustAtoi(u.Gid)
+		if st, err := os.Stat(u.HomeDir); err != nil || !st.IsDir() {
+			if err := os.MkdirAll(u.HomeDir, 0755); err != nil {
+				logger.Warn("auth-daemon: mkdir %s: %v", u.HomeDir, err)
+			} else {
+				_ = os.Chown(u.HomeDir, uid, gid)
+				copySkelInto(skelDir, u.HomeDir, uid, gid)
+				logger.Info("auth-daemon: created home %s for %s", u.HomeDir, u.Username)
+			}
+		} else {
+			// Ensure .bashrc etc. exist (e.g. home existed but was empty or skel was minimal)
+			copySkelInto(skelDir, u.HomeDir, uid, gid)
+		}
+	}
+	return nil
+}

authdaemon/host_stub.go

@@ -0,0 +1,22 @@
+//go:build !linux
+
+package authdaemon
+
+import "fmt"
+
+var errLinuxOnly = fmt.Errorf("auth-daemon PAM agent is only supported on Linux")
+
+// writeCACertIfNotExists returns an error on non-Linux.
+func writeCACertIfNotExists(path, contents string, force bool) error {
+	return errLinuxOnly
+}
+
+// ensureUser returns an error on non-Linux.
+func ensureUser(username string, meta ConnectionMetadata, generateRandomPassword bool) error {
+	return errLinuxOnly
+}
+
+// writePrincipals returns an error on non-Linux.
+func writePrincipals(path, username, niceId string) error {
+	return errLinuxOnly
+}

authdaemon/principals.go

@@ -0,0 +1,28 @@
+package authdaemon
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+)
+
+// GetPrincipals reads the principals data file at path, looks up the given user, and returns that user's principals as a string slice.
+// The file format is JSON: object with username keys and array-of-principals values, e.g. {"alice":["alice","usr-123"],"bob":["bob","usr-456"]}.
+// If the user is not found or the file is missing, returns nil and nil.
+func GetPrincipals(path, user string) ([]string, error) {
+	if path == "" {
+		return nil, fmt.Errorf("principals file path is required")
+	}
+	data, err := os.ReadFile(path)
+	if err != nil {
+		if os.IsNotExist(err) {
+			return nil, nil
+		}
+		return nil, fmt.Errorf("read principals file: %w", err)
+	}
+	var m map[string][]string
+	if err := json.Unmarshal(data, &m); err != nil {
+		return nil, fmt.Errorf("parse principals file: %w", err)
+	}
+	return m[user], nil
+}

authdaemon/routes.go

@@ -0,0 +1,58 @@
+package authdaemon
+
+import (
+	"encoding/json"
+	"net/http"
+)
+
+// registerRoutes registers all API routes. Add new endpoints here.
+func (s *Server) registerRoutes() {
+	s.mux.HandleFunc("/health", s.handleHealth)
+	s.mux.HandleFunc("/connection", s.handleConnection)
+}
+
+// ConnectionMetadata is the metadata object in POST /connection.
+type ConnectionMetadata struct {
+	SudoMode     string   `json:"sudoMode"`     // "none" | "full" | "commands"
+	SudoCommands []string `json:"sudoCommands"` // used when sudoMode is "commands"
+	Homedir      bool     `json:"homedir"`
+	Groups       []string `json:"groups"` // system groups to add the user to
+}
+
+// ConnectionRequest is the JSON body for POST /connection.
+type ConnectionRequest struct {
+	CaCert   string             `json:"caCert"`
+	NiceId   string             `json:"niceId"`
+	Username string             `json:"username"`
+	Metadata ConnectionMetadata `json:"metadata"`
+}
+
+// healthResponse is the JSON body for GET /health.
+type healthResponse struct {
+	Status string `json:"status"`
+}
+
+// handleHealth responds with 200 and {"status":"ok"}.
+func (s *Server) handleHealth(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		http.Error(w, "Method Not Allowed", http.StatusMethodNotAllowed)
+		return
+	}
+	w.Header().Set("Content-Type", "application/json")
+	_ = json.NewEncoder(w).Encode(healthResponse{Status: "ok"})
+}
+
+// handleConnection accepts POST with connection payload and delegates to ProcessConnection.
+func (s *Server) handleConnection(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		http.Error(w, "Method Not Allowed", http.StatusMethodNotAllowed)
+		return
+	}
+	var req ConnectionRequest
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		http.Error(w, "Bad Request", http.StatusBadRequest)
+		return
+	}
+	s.ProcessConnection(req)
+	w.WriteHeader(http.StatusOK)
+}

authdaemon/server.go

@@ -0,0 +1,180 @@
+package authdaemon
+
+import (
+	"context"
+	"crypto/ecdsa"
+	"crypto/elliptic"
+	"crypto/rand"
+	"crypto/subtle"
+	"crypto/tls"
+	"crypto/x509"
+	"crypto/x509/pkix"
+	"encoding/pem"
+	"fmt"
+	"math/big"
+	"net/http"
+	"os"
+	"runtime"
+	"strings"
+	"time"
+
+	"github.com/fosrl/newt/logger"
+)
+
+type Config struct {
+	// DisableHTTPS: when true, Run() does not start the HTTPS server (for embedded use inside Newt). Call ProcessConnection directly for connection events.
+	DisableHTTPS       bool
+	Port               int    // Required when DisableHTTPS is false. Listen port for the HTTPS server. No default.
+	PresharedKey       string // Required when DisableHTTPS is false. HTTP auth (Authorization: Bearer <key> or X-Preshared-Key: <key>). No default.
+	CACertPath         string // Required. Where to write the CA cert (e.g. /etc/ssh/ca.pem). No default.
+	Force                   bool   // If true, overwrite existing CA cert (and other items) when content differs. Default false.
+	PrincipalsFilePath      string // Required. Path to the principals data file (JSON: username -> array of principals). No default.
+	GenerateRandomPassword  bool   // If true, set a random password on users when they are provisioned (for SSH PermitEmptyPasswords no).
+}
+
+type Server struct {
+	cfg          Config
+	addr         string
+	presharedKey string
+	mux          *http.ServeMux
+	tlsCert      tls.Certificate
+}
+
+// generateTLSCert creates a self-signed certificate and key in memory (no disk).
+func generateTLSCert() (tls.Certificate, error) {
+	key, err := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
+	if err != nil {
+		return tls.Certificate{}, fmt.Errorf("generate key: %w", err)
+	}
+	serial, err := rand.Int(rand.Reader, new(big.Int).Lsh(big.NewInt(1), 128))
+	if err != nil {
+		return tls.Certificate{}, fmt.Errorf("serial: %w", err)
+	}
+	tmpl := &x509.Certificate{
+		SerialNumber: serial,
+		Subject: pkix.Name{
+			CommonName: "localhost",
+		},
+		NotBefore:             time.Now(),
+		NotAfter:              time.Now().Add(365 * 24 * time.Hour),
+		KeyUsage:              x509.KeyUsageDigitalSignature,
+		ExtKeyUsage:           []x509.ExtKeyUsage{x509.ExtKeyUsageServerAuth},
+		BasicConstraintsValid: true,
+		DNSNames:              []string{"localhost", "127.0.0.1"},
+	}
+	certDER, err := x509.CreateCertificate(rand.Reader, tmpl, tmpl, key.Public(), key)
+	if err != nil {
+		return tls.Certificate{}, fmt.Errorf("create certificate: %w", err)
+	}
+	keyDER, err := x509.MarshalECPrivateKey(key)
+	if err != nil {
+		return tls.Certificate{}, fmt.Errorf("marshal key: %w", err)
+	}
+	certPEM := pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: certDER})
+	keyPEM := pem.EncodeToMemory(&pem.Block{Type: "EC PRIVATE KEY", Bytes: keyDER})
+	cert, err := tls.X509KeyPair(certPEM, keyPEM)
+	if err != nil {
+		return tls.Certificate{}, fmt.Errorf("x509 key pair: %w", err)
+	}
+	return cert, nil
+}
+
+// authMiddleware wraps next and requires a valid preshared key on every request.
+// Accepts Authorization: Bearer <key> or X-Preshared-Key: <key>.
+func (s *Server) authMiddleware(next http.Handler) http.Handler {
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		key := ""
+		if v := r.Header.Get("Authorization"); strings.HasPrefix(v, "Bearer ") {
+			key = strings.TrimSpace(strings.TrimPrefix(v, "Bearer "))
+		}
+		if key == "" {
+			key = strings.TrimSpace(r.Header.Get("X-Preshared-Key"))
+		}
+		if key == "" || subtle.ConstantTimeCompare([]byte(key), []byte(s.presharedKey)) != 1 {
+			http.Error(w, "Unauthorized", http.StatusUnauthorized)
+			return
+		}
+		next.ServeHTTP(w, r)
+	})
+}
+
+// NewServer builds a new auth-daemon server from cfg. Port, PresharedKey, CACertPath, and PrincipalsFilePath are required (no defaults).
+func NewServer(cfg Config) (*Server, error) {
+	if runtime.GOOS != "linux" {
+		return nil, fmt.Errorf("auth-daemon is only supported on Linux, not %s", runtime.GOOS)
+	}
+	if !cfg.DisableHTTPS {
+		if cfg.Port <= 0 {
+			return nil, fmt.Errorf("port is required and must be positive")
+		}
+		if cfg.PresharedKey == "" {
+			return nil, fmt.Errorf("preshared key is required")
+		}
+	}
+	if cfg.CACertPath == "" {
+		return nil, fmt.Errorf("CACertPath is required")
+	}
+	if cfg.PrincipalsFilePath == "" {
+		return nil, fmt.Errorf("PrincipalsFilePath is required")
+	}
+	s := &Server{cfg: cfg}
+	if !cfg.DisableHTTPS {
+		cert, err := generateTLSCert()
+		if err != nil {
+			return nil, err
+		}
+		s.addr = fmt.Sprintf(":%d", cfg.Port)
+		s.presharedKey = cfg.PresharedKey
+		s.mux = http.NewServeMux()
+		s.tlsCert = cert
+		s.registerRoutes()
+	}
+	return s, nil
+}
+
+// Run starts the HTTPS server (unless DisableHTTPS) and blocks until ctx is cancelled or the server errors.
+// When DisableHTTPS is true, Run() blocks on ctx only and does not listen; use ProcessConnection for connection events.
+func (s *Server) Run(ctx context.Context) error {
+	if s.cfg.DisableHTTPS {
+		logger.Info("auth-daemon running (HTTPS disabled)")
+		<-ctx.Done()
+		s.cleanupPrincipalsFile()
+		return nil
+	}
+	tcfg := &tls.Config{
+		Certificates: []tls.Certificate{s.tlsCert},
+		MinVersion:   tls.VersionTLS12,
+	}
+	handler := s.authMiddleware(s.mux)
+	srv := &http.Server{
+		Addr:              s.addr,
+		Handler:           handler,
+		TLSConfig:         tcfg,
+		ReadTimeout:       10 * time.Second,
+		WriteTimeout:      10 * time.Second,
+		ReadHeaderTimeout: 5 * time.Second,
+		IdleTimeout:       60 * time.Second,
+	}
+	go func() {
+		<-ctx.Done()
+		shutdownCtx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+		defer cancel()
+		if err := srv.Shutdown(shutdownCtx); err != nil {
+			logger.Warn("auth-daemon shutdown: %v", err)
+		}
+	}()
+	logger.Info("auth-daemon listening on https://127.0.0.1%s", s.addr)
+	if err := srv.ListenAndServeTLS("", ""); err != nil && err != http.ErrServerClosed {
+		return err
+	}
+	s.cleanupPrincipalsFile()
+	return nil
+}
+
+func (s *Server) cleanupPrincipalsFile() {
+	if s.cfg.PrincipalsFilePath != "" {
+		if err := os.Remove(s.cfg.PrincipalsFilePath); err != nil && !os.IsNotExist(err) {
+			logger.Warn("auth-daemon: remove principals file: %v", err)
+		}
+	}
+}

bind/shared_bind.go

@@ -144,6 +144,10 @@ type SharedBind struct {
 
 	// Callback for magic test responses (used for holepunch testing)
 	magicResponseCallback atomic.Pointer[func(addr netip.AddrPort, echoData []byte)]
+
+	// Rebinding state - used to keep receive goroutines alive during socket transition
+	rebinding     bool       // true when socket is being replaced
+	rebindingCond *sync.Cond // signaled when rebind completes
 }
 
 // MagicResponseCallback is the function signature for magic packet response callbacks
@@ -163,6 +167,9 @@ func New(udpConn *net.UDPConn) (*SharedBind, error) {
 		closeChan:       make(chan struct{}),
 	}
 
+	// Initialize the rebinding condition variable
+	bind.rebindingCond = sync.NewCond(&bind.mu)
+
 	// Initialize reference count to 1 (the creator holds the first reference)
 	bind.refCount.Store(1)
 
@@ -310,6 +317,109 @@ func (b *SharedBind) IsClosed() bool {
 	return b.closed.Load()
 }
 
+// GetPort returns the current UDP port the bind is using.
+// This is useful when rebinding to try to reuse the same port.
+func (b *SharedBind) GetPort() uint16 {
+	b.mu.RLock()
+	defer b.mu.RUnlock()
+	return b.port
+}
+
+// CloseSocket closes the underlying UDP connection to release the port,
+// but keeps the SharedBind in a state where it can accept a new connection via Rebind.
+// This allows the caller to close the old socket first, then bind a new socket
+// to the same port before calling Rebind.
+//
+// Returns the port that was being used, so the caller can attempt to rebind to it.
+// Sets the rebinding flag so receive goroutines will wait for the new socket
+// instead of exiting.
+func (b *SharedBind) CloseSocket() (uint16, error) {
+	b.mu.Lock()
+	defer b.mu.Unlock()
+
+	if b.closed.Load() {
+		return 0, fmt.Errorf("bind is closed")
+	}
+
+	port := b.port
+
+	// Set rebinding flag BEFORE closing the socket so receive goroutines
+	// know to wait instead of exit
+	b.rebinding = true
+
+	// Close the old connection to release the port
+	if b.udpConn != nil {
+		logger.Debug("Closing UDP connection to release port %d (rebinding)", port)
+		b.udpConn.Close()
+		b.udpConn = nil
+	}
+
+	return port, nil
+}
+
+// Rebind replaces the underlying UDP connection with a new one.
+// This is necessary when network connectivity changes (e.g., WiFi to cellular
+// transition on macOS/iOS) and the old socket becomes stale.
+//
+// The caller is responsible for creating the new UDP connection and passing it here.
+// After rebind, the caller should trigger a hole punch to re-establish NAT mappings.
+//
+// Note: Call CloseSocket() first if you need to rebind to the same port, as the
+// old socket must be closed before a new socket can bind to the same port.
+func (b *SharedBind) Rebind(newConn *net.UDPConn) error {
+	if newConn == nil {
+		return fmt.Errorf("newConn cannot be nil")
+	}
+
+	b.mu.Lock()
+	defer b.mu.Unlock()
+
+	if b.closed.Load() {
+		return fmt.Errorf("bind is closed")
+	}
+
+	// Close the old connection if it's still open
+	// (it may have already been closed via CloseSocket)
+	if b.udpConn != nil {
+		logger.Debug("Closing old UDP connection during rebind")
+		b.udpConn.Close()
+	}
+
+	// Set up the new connection
+	b.udpConn = newConn
+
+	// Update packet connections for the new socket
+	if runtime.GOOS == "linux" || runtime.GOOS == "android" {
+		b.ipv4PC = ipv4.NewPacketConn(newConn)
+		b.ipv6PC = ipv6.NewPacketConn(newConn)
+
+		// Re-initialize message buffers for batch operations
+		batchSize := wgConn.IdealBatchSize
+		b.ipv4Msgs = make([]ipv4.Message, batchSize)
+		for i := range b.ipv4Msgs {
+			b.ipv4Msgs[i].OOB = make([]byte, 0)
+		}
+	} else {
+		// For non-Linux platforms, still set up ipv4PC for consistency
+		b.ipv4PC = ipv4.NewPacketConn(newConn)
+		b.ipv6PC = ipv6.NewPacketConn(newConn)
+	}
+
+	// Update the port
+	if addr, ok := newConn.LocalAddr().(*net.UDPAddr); ok {
+		b.port = uint16(addr.Port)
+		logger.Info("Rebound UDP socket to port %d", b.port)
+	}
+
+	// Clear the rebinding flag and wake up any waiting receive goroutines
+	b.rebinding = false
+	b.rebindingCond.Broadcast()
+
+	logger.Debug("Rebind complete, signaled waiting receive goroutines")
+
+	return nil
+}
+
 // SetMagicResponseCallback sets a callback function that will be called when
 // a magic test response packet is received. This is used for holepunch testing.
 // Pass nil to clear the callback.
@@ -392,24 +502,77 @@ func (b *SharedBind) Open(uport uint16) ([]wgConn.ReceiveFunc, uint16, error) {
 // makeReceiveSocket creates a receive function for physical UDP socket packets
 func (b *SharedBind) makeReceiveSocket() wgConn.ReceiveFunc {
 	return func(bufs [][]byte, sizes []int, eps []wgConn.Endpoint) (n int, err error) {
-		if b.closed.Load() {
-			return 0, net.ErrClosed
-		}
+		for {
+			if b.closed.Load() {
+				return 0, net.ErrClosed
+			}
 
-		b.mu.RLock()
-		conn := b.udpConn
-		pc := b.ipv4PC
-		b.mu.RUnlock()
+			b.mu.RLock()
+			conn := b.udpConn
+			pc := b.ipv4PC
+			b.mu.RUnlock()
 
-		if conn == nil {
-			return 0, net.ErrClosed
-		}
+			if conn == nil {
+				// Socket is nil - check if we're rebinding or truly closed
+				if b.closed.Load() {
+					return 0, net.ErrClosed
+				}
 
-		// Use batch reading on Linux for performance
-		if pc != nil && (runtime.GOOS == "linux" || runtime.GOOS == "android") {
-			return b.receiveIPv4Batch(pc, bufs, sizes, eps)
+				// Wait for rebind to complete
+				b.mu.Lock()
+				for b.rebinding && !b.closed.Load() {
+					logger.Debug("Receive goroutine waiting for socket rebind to complete")
+					b.rebindingCond.Wait()
+				}
+				b.mu.Unlock()
+
+				// Check again after waking up
+				if b.closed.Load() {
+					return 0, net.ErrClosed
+				}
+
+				// Loop back to retry with new socket
+				continue
+			}
+
+			// Use batch reading on Linux for performance
+			var n int
+			var err error
+			if pc != nil && (runtime.GOOS == "linux" || runtime.GOOS == "android") {
+				n, err = b.receiveIPv4Batch(pc, bufs, sizes, eps)
+			} else {
+				n, err = b.receiveIPv4Simple(conn, bufs, sizes, eps)
+			}
+
+			if err != nil {
+				// Check if this error is due to rebinding
+				b.mu.RLock()
+				rebinding := b.rebinding
+				b.mu.RUnlock()
+
+				if rebinding {
+					logger.Debug("Receive got error during rebind, waiting for new socket: %v", err)
+					// Wait for rebind to complete and retry
+					b.mu.Lock()
+					for b.rebinding && !b.closed.Load() {
+						b.rebindingCond.Wait()
+					}
+					b.mu.Unlock()
+
+					if b.closed.Load() {
+						return 0, net.ErrClosed
+					}
+
+					// Retry with new socket
+					continue
+				}
+
+				// Not rebinding, return the error
+				return 0, err
+			}
+
+			return n, nil
 		}
-		return b.receiveIPv4Simple(conn, bufs, sizes, eps)
 	}
 }
 
@@ -492,6 +655,8 @@ func (b *SharedBind) receiveIPv4Batch(pc *ipv4.PacketConn, bufs [][]byte, sizes
 
 // receiveIPv4Simple uses simple ReadFromUDP for non-Linux platforms
 func (b *SharedBind) receiveIPv4Simple(conn *net.UDPConn, bufs [][]byte, sizes []int, eps []wgConn.Endpoint) (int, error) {
+	// No read deadline - we rely on socket close to unblock during rebind.
+	// The caller (makeReceiveSocket) handles rebind state when errors occur.
 	for {
 		n, addr, err := conn.ReadFromUDP(bufs[0])
 		if err != nil {
@@ -523,7 +688,7 @@ func (b *SharedBind) receiveIPv4Simple(conn *net.UDPConn, bufs [][]byte, sizes [
 func (b *SharedBind) handleMagicPacket(data []byte, addr *net.UDPAddr) bool {
 	// Check if this is a test request packet
 	if len(data) >= MagicTestRequestLen && bytes.HasPrefix(data, MagicTestRequest) {
-		logger.Debug("Received magic test REQUEST from %s, sending response", addr.String())
+		// logger.Debug("Received magic test REQUEST from %s, sending response", addr.String())
 		// Extract the random data portion to echo back
 		echoData := data[len(MagicTestRequest) : len(MagicTestRequest)+MagicPacketDataLen]
 
@@ -546,7 +711,7 @@ func (b *SharedBind) handleMagicPacket(data []byte, addr *net.UDPAddr) bool {
 
 	// Check if this is a test response packet
 	if len(data) >= MagicTestResponseLen && bytes.HasPrefix(data, MagicTestResponse) {
-		logger.Debug("Received magic test RESPONSE from %s", addr.String())
+		// logger.Debug("Received magic test RESPONSE from %s", addr.String())
 		// Extract the echoed data
 		echoData := data[len(MagicTestResponse) : len(MagicTestResponse)+MagicPacketDataLen]
 

blueprint.yaml

@@ -1,37 +0,0 @@
-resources:
-  resource-nice-id:
-    name: this is my resource
-    protocol: http
-    full-domain: level1.test3.example.com
-    host-header: example.com
-    tls-server-name: example.com
-    auth:
-      pincode: 123456
-      password: sadfasdfadsf
-      sso-enabled: true
-      sso-roles:
-      - Member
-      sso-users:
-      - owen@pangolin.net
-      whitelist-users:
-      - owen@pangolin.net
-    targets:
-    # - site: glossy-plains-viscacha-rat
-    - hostname: localhost
-      method: http
-      port: 8000
-      healthcheck:
-        port: 8000
-        hostname: localhost
-    # - site: glossy-plains-viscacha-rat
-    - hostname: localhost
-      method: http
-      port: 8001
-  resource-nice-id2:
-    name: this is other resource
-    protocol: tcp
-    proxy-port: 3000
-    targets:
-    # - site: glossy-plains-viscacha-rat
-    - hostname: localhost
-      port: 3000

browsergateway/browsergateway.go

@@ -0,0 +1,144 @@
+package browsergateway
+
+import (
+	"crypto/subtle"
+	"errors"
+	"net"
+	"net/http"
+	"strings"
+	"sync"
+
+	"github.com/fosrl/newt/nativessh"
+)
+
+// Forwarding buffer size. RDP graphics traffic is bursty and TLS records cap
+// at ~16 KiB, so 64 KiB lets a couple of records pile up per syscall/frame
+// without wasting memory per session.
+const forwardBufSize = 64 * 1024
+
+// ListenPort is the port the browser gateway HTTP server listens on inside the
+// WireGuard netstack. This is a fixed value shared between newt and pangolin.
+// Targets do not overlap with this port because they start at 40000.
+const ListenPort = 39999
+
+// Target represents an allowed proxy destination for the browser gateway.
+// Only connections whose (Type, Destination, DestinationPort, AuthToken) match a
+// registered Target will be forwarded; all others are rejected.
+type Target struct {
+	ID              int
+	Type            string // "rdp" | "ssh" | "vnc"
+	Destination     string
+	DestinationPort int
+	AuthToken       string // per-target secret; must match the token supplied by the client
+}
+
+// Config holds the configuration for a Gateway.
+type Config struct {
+	// AuthToken is used only for NativeSSH mode (which has no external target
+	// to match against). For all proxy targets (RDP/SSH/VNC), auth tokens are
+	// stored per-Target and validated by isAllowed.
+	AuthToken string
+	// SSHCredentials are used by native SSH browser sessions for certificate
+	// validation against the in-memory CA/principal store.
+	SSHCredentials *nativessh.CredentialStore
+}
+
+// Gateway is a browser-based RDP/SSH/VNC WebSocket proxy.
+// Create one with New and mount it via RegisterHandlers or the individual
+// HandleRDP / HandleSSH / HandleVNC http.HandlerFunc methods.
+type Gateway struct {
+	authToken string
+	sshCreds  *nativessh.CredentialStore
+
+	mu      sync.RWMutex
+	targets map[int]Target // keyed by Target.ID
+
+	server *http.Server
+}
+
+// New creates a new Gateway from the provided Config.
+func New(cfg Config) *Gateway {
+	return &Gateway{
+		authToken: cfg.AuthToken,
+		sshCreds:  cfg.SSHCredentials,
+		targets:   make(map[int]Target),
+	}
+}
+
+// SetTargets replaces the entire allowed-destination list atomically.
+func (g *Gateway) SetTargets(targets []Target) {
+	g.mu.Lock()
+	defer g.mu.Unlock()
+	g.targets = make(map[int]Target, len(targets))
+	for _, t := range targets {
+		g.targets[t.ID] = t
+	}
+}
+
+// AddTarget adds or updates a single allowed destination.
+func (g *Gateway) AddTarget(t Target) {
+	g.mu.Lock()
+	defer g.mu.Unlock()
+	g.targets[t.ID] = t
+}
+
+// RemoveTarget removes an allowed destination by its ID.
+func (g *Gateway) RemoveTarget(id int) {
+	g.mu.Lock()
+	defer g.mu.Unlock()
+	delete(g.targets, id)
+}
+
+// isAllowed reports whether a connection to (targetType, host, port) with the
+// given authToken is permitted. The token is compared against the per-target
+// AuthToken using constant-time comparison to prevent timing attacks.
+func (g *Gateway) isAllowed(targetType, host string, port int, authToken string) bool {
+	g.mu.RLock()
+	defer g.mu.RUnlock()
+	for _, t := range g.targets {
+		if t.Type == targetType && t.Destination == host && t.DestinationPort == port {
+			return subtle.ConstantTimeCompare([]byte(authToken), []byte(t.AuthToken)) == 1
+		}
+	}
+	return false
+}
+
+// isTokenValid reports whether the given authToken matches any registered
+// target of the specified type. Used for native SSH mode where there is no
+// external destination to match against.
+func (g *Gateway) isTokenValid(targetType, authToken string) bool {
+	g.mu.RLock()
+	defer g.mu.RUnlock()
+	for _, t := range g.targets {
+		if t.Type == targetType {
+			if subtle.ConstantTimeCompare([]byte(authToken), []byte(t.AuthToken)) == 1 {
+				return true
+			}
+		}
+	}
+	return false
+}
+
+// Start serves the browser gateway HTTP server on the provided listener.
+// It returns nil when the listener is closed (normal shutdown).
+func (g *Gateway) Start(ln net.Listener) error {
+	mux := http.NewServeMux()
+	g.RegisterHandlers(mux)
+	g.server = &http.Server{Handler: mux}
+	err := g.server.Serve(ln)
+	if err == nil ||
+		errors.Is(err, net.ErrClosed) ||
+		errors.Is(err, http.ErrServerClosed) ||
+		strings.Contains(err.Error(), "use of closed") ||
+		strings.Contains(err.Error(), "invalid state") {
+		return nil
+	}
+	return err
+}
+
+// RegisterHandlers registers the /rdp, /ssh, and /vnc routes on mux.
+func (g *Gateway) RegisterHandlers(mux *http.ServeMux) {
+	mux.HandleFunc("/gateway/rdp", g.HandleRDP)
+	mux.HandleFunc("/gateway/ssh", g.HandleSSH)
+	mux.HandleFunc("/gateway/vnc", g.handleVNC)
+}

browsergateway/rdcleanpath.go

@@ -0,0 +1,88 @@
+package browsergateway
+
+import (
+	"encoding/asn1"
+	"fmt"
+)
+
+// RDCleanPath PDU version (BASE_VERSION + 1 = 3389 + 1).
+const rdCleanPathVersion = int64(3390)
+
+// rdCleanPathPdu is a Go translation of the ASN.1 SEQUENCE defined in the
+// ironrdp-rdcleanpath crate. All optional fields use EXPLICIT context-specific
+// tagging, matching the Rust `der::Sequence` derivation with
+// `tag_mode = "EXPLICIT"`.
+//
+// We only need a subset of fields for the basic proxy flow, but the struct
+// declares every tag we may encounter so that decoding does not fail on an
+// unexpected element.
+type rdCleanPathPdu struct {
+	Version           int64    `asn1:"explicit,tag:0"`
+	Destination       string   `asn1:"explicit,tag:2,optional,utf8"`
+	ProxyAuth         string   `asn1:"explicit,tag:3,optional,utf8"`
+	ServerAuth        string   `asn1:"explicit,tag:4,optional,utf8"`
+	PreconnectionBlob string   `asn1:"explicit,tag:5,optional,utf8"`
+	X224              []byte   `asn1:"explicit,tag:6,optional"`
+	ServerCertChain   [][]byte `asn1:"explicit,tag:7,optional"`
+	ServerAddr        string   `asn1:"explicit,tag:9,optional,utf8"`
+}
+
+// decodeRDCleanPathRequest parses a client-to-proxy RDCleanPath PDU.
+func decodeRDCleanPathRequest(buf []byte) (*rdCleanPathPdu, error) {
+	var pdu rdCleanPathPdu
+	rest, err := asn1.Unmarshal(buf, &pdu)
+	if err != nil {
+		return nil, fmt.Errorf("asn1 unmarshal: %w", err)
+	}
+	if len(rest) != 0 {
+		return nil, fmt.Errorf("trailing data after RDCleanPath PDU: %d bytes", len(rest))
+	}
+	if pdu.Version != rdCleanPathVersion {
+		return nil, fmt.Errorf("unexpected RDCleanPath version: %d", pdu.Version)
+	}
+	return &pdu, nil
+}
+
+// encodeRDCleanPathResponse builds a proxy-to-client RDCleanPath response PDU
+// containing the server address, X.224 connection confirm and server TLS chain.
+func encodeRDCleanPathResponse(serverAddr string, x224Rsp []byte, certChain [][]byte) ([]byte, error) {
+	pdu := rdCleanPathPdu{
+		Version:         rdCleanPathVersion,
+		X224:            x224Rsp,
+		ServerCertChain: certChain,
+		ServerAddr:      serverAddr,
+	}
+	return asn1.Marshal(pdu)
+}
+
+// detectRDCleanPathLength returns the total DER length of an RDCleanPath PDU
+// if enough bytes are available, otherwise -1.
+//
+// The PDU is a DER SEQUENCE, which begins with the universal SEQUENCE tag
+// (0x30) followed by a length octet/octets. We parse just enough to know the
+// total length so we can buffer accordingly.
+func detectRDCleanPathLength(buf []byte) int {
+	if len(buf) < 2 {
+		return -1
+	}
+	if buf[0] != 0x30 {
+		// Not a SEQUENCE: cannot be RDCleanPath.
+		return -2
+	}
+	l := buf[1]
+	if l < 0x80 {
+		return 2 + int(l)
+	}
+	n := int(l & 0x7f)
+	if n == 0 || n > 4 {
+		return -2
+	}
+	if len(buf) < 2+n {
+		return -1
+	}
+	total := 0
+	for i := 0; i < n; i++ {
+		total = (total << 8) | int(buf[2+i])
+	}
+	return 2 + n + total
+}

browsergateway/rdp.go

@@ -0,0 +1,271 @@
+package browsergateway
+
+import (
+	"context"
+	"crypto/tls"
+	"encoding/binary"
+	"errors"
+	"fmt"
+	"io"
+	"log"
+	"net"
+	"net/http"
+	"strconv"
+	"time"
+
+	"github.com/coder/websocket"
+)
+
+// HandleRDP is an http.HandlerFunc for RDP-over-WebSocket connections.
+func (g *Gateway) HandleRDP(w http.ResponseWriter, r *http.Request) {
+	ctx := r.Context()
+	ws, err := websocket.Accept(w, r, &websocket.AcceptOptions{
+		InsecureSkipVerify: true, // any-origin: minimal dev proxy with no auth
+		Subprotocols:       []string{"binary"},
+	})
+	if err != nil {
+		log.Printf("websocket upgrade failed: %v", err)
+		return
+	}
+	// Disable per-message read size cap (default is 32 KiB which would break
+	// large RDP graphics messages).
+	ws.SetReadLimit(-1)
+	defer ws.CloseNow() //nolint:errcheck
+
+	if err := g.serveSession(ctx, ws); err != nil {
+		log.Printf("session error: %v", err)
+	}
+}
+
+func (g *Gateway) serveSession(ctx context.Context, ws *websocket.Conn) error {
+	// Expose the WebSocket as a streaming net.Conn. Binary messages are
+	// concatenated into a byte stream and writes become single binary frames.
+	// This is a thin wrapper with no per-message goroutine, unlike Gorilla.
+	stream := websocket.NetConn(ctx, ws, websocket.MessageBinary)
+	defer stream.Close() //nolint:errcheck
+
+	// -- Read the initial RDCleanPath request from the client --
+	pdu, err := readCleanPath(stream)
+	if err != nil {
+		return fmt.Errorf("read RDCleanPath: %w", err)
+	}
+
+	if pdu.Destination == "" {
+		return errors.New("RDCleanPath missing destination")
+	}
+	if len(pdu.X224) == 0 {
+		return errors.New("RDCleanPath missing X224 connection PDU")
+	}
+
+	target := pdu.Destination
+	// Default port for RDP if not specified.
+	if _, _, splitErr := net.SplitHostPort(target); splitErr != nil {
+		target = net.JoinHostPort(target, "3389")
+	}
+
+	// Validate destination against the registered target allowlist,
+	// including per-target auth token.
+	rdpHost, rdpPortStr, _ := net.SplitHostPort(target)
+	rdpPort, _ := strconv.Atoi(rdpPortStr)
+	if !g.isAllowed("rdp", rdpHost, rdpPort, pdu.ProxyAuth) {
+		return fmt.Errorf("RDP destination %s is not in the allowed target list or auth token mismatch", target)
+	}
+
+	log.Printf("Connecting to RDP server %s", target)
+
+	// -- Open TCP connection to the destination RDP server --
+	serverTCP, err := net.DialTimeout("tcp", target, 15*time.Second)
+	if err != nil {
+		return fmt.Errorf("dial %s: %w", target, err)
+	}
+	defer serverTCP.Close()
+	if tcp, ok := serverTCP.(*net.TCPConn); ok {
+		// NoDelay is Go's default; set explicitly. RDP wants low latency for
+		// input echo, and the bulk path is naturally chunked by TLS records.
+		_ = tcp.SetNoDelay(true)
+		_ = tcp.SetKeepAlive(true)
+		_ = tcp.SetKeepAlivePeriod(30 * time.Second)
+		_ = tcp.SetReadBuffer(forwardBufSize)
+		_ = tcp.SetWriteBuffer(forwardBufSize)
+	}
+	serverAddr := serverTCP.RemoteAddr().String()
+
+	// Forward the optional pre-connection blob, then the X.224 connection request.
+	if pdu.PreconnectionBlob != "" {
+		if _, err := serverTCP.Write([]byte(pdu.PreconnectionBlob)); err != nil {
+			return fmt.Errorf("send PCB: %w", err)
+		}
+	}
+	if _, err := serverTCP.Write(pdu.X224); err != nil {
+		return fmt.Errorf("send X224: %w", err)
+	}
+
+	// -- Read the X.224 connection confirm from the server --
+	x224Rsp, err := readX224(serverTCP)
+	if err != nil {
+		return fmt.Errorf("read X224 response: %w", err)
+	}
+	logX224Negotiation(x224Rsp)
+
+	// -- Upgrade the server connection to TLS (skip verification) --
+	//
+	// Windows RDP hosts are picky: only set SNI when the target is a hostname
+	// (Go would skip SNI for IP literals anyway, but be explicit), and accept
+	// the full range of TLS versions / ciphers since some servers only
+	// negotiate TLS 1.0 or legacy suites.
+	host, _, _ := net.SplitHostPort(target)
+	tlsCfg := &tls.Config{
+		InsecureSkipVerify: true, //nolint:gosec // proxy intentionally skips verification
+		MinVersion:         tls.VersionTLS10,
+		// Cap at TLS 1.2: Windows RDP servers commonly send a TLS "internal_error"
+		// alert when CredSSP/NLA is layered on top of a TLS 1.3 session.
+		MaxVersion: tls.VersionTLS12,
+	}
+	if net.ParseIP(host) == nil {
+		tlsCfg.ServerName = host
+	}
+	tlsConn := tls.Client(serverTCP, tlsCfg)
+	if err := tlsConn.Handshake(); err != nil {
+		return fmt.Errorf("TLS handshake with server: %w", err)
+	}
+	log.Printf("Server TLS handshake OK (version=0x%04x cipher=0x%04x)",
+		tlsConn.ConnectionState().Version, tlsConn.ConnectionState().CipherSuite)
+
+	// Collect the raw DER server certificate chain to return to the client.
+	state := tlsConn.ConnectionState()
+	if len(state.PeerCertificates) == 0 {
+		return errors.New("server did not present any certificates")
+	}
+	chain := make([][]byte, 0, len(state.PeerCertificates))
+	for _, c := range state.PeerCertificates {
+		chain = append(chain, c.Raw)
+	}
+
+	// -- Send the RDCleanPath response back to the client --
+	rsp, err := encodeRDCleanPathResponse(serverAddr, x224Rsp, chain)
+	if err != nil {
+		return fmt.Errorf("encode RDCleanPath response: %w", err)
+	}
+	if _, err := stream.Write(rsp); err != nil {
+		return fmt.Errorf("write RDCleanPath response: %w", err)
+	}
+
+	log.Printf("RDCleanPath handshake complete, forwarding traffic to %s", serverAddr)
+
+	// -- Two-way blind forwarding of the (now TLS-encrypted) RDP stream --
+	return forward(stream, tlsConn)
+}
+
+// readCleanPath buffers bytes from the stream until a full RDCleanPath PDU has
+// been received, then decodes it.
+func readCleanPath(r io.Reader) (*rdCleanPathPdu, error) {
+	buf := make([]byte, 0, 1024)
+	tmp := make([]byte, 1024)
+	for {
+		total := detectRDCleanPathLength(buf)
+		switch {
+		case total == -2:
+			return nil, errors.New("invalid RDCleanPath PDU")
+		case total > 0 && len(buf) >= total:
+			return decodeRDCleanPathRequest(buf[:total])
+		}
+		n, err := r.Read(tmp)
+		if n > 0 {
+			buf = append(buf, tmp[:n]...)
+		}
+		if err != nil {
+			return nil, err
+		}
+	}
+}
+
+// readX224 reads exactly one TPKT-framed X.224 PDU from the server.
+//
+// The TPKT header is 4 bytes: version (0x03), reserved (0x00), and a u16
+// big-endian total length that includes the header itself.
+func readX224(r io.Reader) ([]byte, error) {
+	hdr := make([]byte, 4)
+	if _, err := io.ReadFull(r, hdr); err != nil {
+		return nil, err
+	}
+	if hdr[0] != 0x03 {
+		return nil, fmt.Errorf("unexpected TPKT version 0x%02x", hdr[0])
+	}
+	total := int(binary.BigEndian.Uint16(hdr[2:4]))
+	if total < 4 || total > 4096 {
+		return nil, fmt.Errorf("unreasonable TPKT length %d", total)
+	}
+	out := make([]byte, total)
+	copy(out, hdr)
+	if _, err := io.ReadFull(r, out[4:]); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+// logX224Negotiation prints which RDP security protocol the server selected
+// (or the failure code), to help diagnose handshake issues such as the server
+// requiring NLA/CredSSP.
+//
+// X.224 Connection Confirm layout (RFC 1006 / [MS-RDPBCGR]):
+//
+//	bytes 0..3   TPKT header (03 00 LL LL)
+//	byte  4      X.224 length indicator
+//	byte  5      X.224 code (0xD0 = CC)
+//	bytes 6..10  DST-REF, SRC-REF, class
+//	byte  11     optional RDP Negotiation type (0x02 = response, 0x03 = failure)
+//	byte  12     flags
+//	bytes 13..14 length (little-endian, =8)
+//	bytes 15..18 selected protocol / failure code (u32 little-endian)
+func logX224Negotiation(pdu []byte) {
+	if len(pdu) < 19 {
+		log.Printf("X.224 response too short (%d bytes) to contain RDP negotiation", len(pdu))
+		return
+	}
+	switch pdu[11] {
+	case 0x02:
+		proto := uint32(pdu[15]) | uint32(pdu[16])<<8 | uint32(pdu[17])<<16 | uint32(pdu[18])<<24
+		name := "unknown"
+		switch proto {
+		case 0:
+			name = "RDP (standard)"
+		case 1:
+			name = "SSL/TLS"
+		case 2:
+			name = "HYBRID (CredSSP/NLA)"
+		case 8:
+			name = "HYBRID_EX"
+		}
+		log.Printf("Server selected RDP protocol 0x%x (%s)", proto, name)
+	case 0x03:
+		code := uint32(pdu[15]) | uint32(pdu[16])<<8 | uint32(pdu[17])<<16 | uint32(pdu[18])<<24
+		log.Printf("Server returned RDP negotiation failure code 0x%x", code)
+	default:
+		log.Printf("X.224 response has no RDP negotiation block (type=0x%02x)", pdu[11])
+	}
+}
+
+// forward shuttles bytes between the two streams until either side closes.
+func forward(a, b io.ReadWriteCloser) error {
+	errc := make(chan error, 2)
+	go func() {
+		buf := make([]byte, forwardBufSize)
+		_, err := io.CopyBuffer(a, b, buf)
+		_ = a.Close()
+		_ = b.Close()
+		errc <- err
+	}()
+	go func() {
+		buf := make([]byte, forwardBufSize)
+		_, err := io.CopyBuffer(b, a, buf)
+		_ = a.Close()
+		_ = b.Close()
+		errc <- err
+	}()
+	// Wait for one side to finish, then return.
+	err := <-errc
+	if errors.Is(err, io.EOF) || err == nil {
+		return nil
+	}
+	return err
+}

browsergateway/ssh.go

@@ -0,0 +1,259 @@
+package browsergateway
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log"
+	"net"
+	"net/http"
+	"strconv"
+	"time"
+
+	"github.com/coder/websocket"
+	"github.com/fosrl/newt/logger"
+	"golang.org/x/crypto/ssh"
+)
+
+// sshClientMsg is a JSON message sent from the browser to the proxy.
+type sshClientMsg struct {
+	// type: "auth" | "data" | "resize"
+	Type       string `json:"type"`
+	Password   string `json:"password,omitempty"`   // used when type="auth"
+	PrivateKey string `json:"privateKey,omitempty"` // used when type="auth"
+	Certificate string `json:"certificate,omitempty"` // used when type="auth"
+	Data       string `json:"data,omitempty"`       // used when type="data"
+	Cols       uint32 `json:"cols,omitempty"`       // used when type="resize"
+	Rows       uint32 `json:"rows,omitempty"`       // used when type="resize"
+}
+
+// sshServerMsg is a JSON message sent from the proxy back to the browser.
+type sshServerMsg struct {
+	// type: "data" | "error"
+	Type  string `json:"type"`
+	Data  string `json:"data,omitempty"`
+	Error string `json:"error,omitempty"`
+}
+
+// HandleSSH is an http.HandlerFunc for SSH-over-WebSocket connections.
+func (g *Gateway) HandleSSH(w http.ResponseWriter, r *http.Request) {
+	logger.Debug("SSH connection request from %s", r.RemoteAddr)
+	ctx := r.Context()
+
+	token := r.URL.Query().Get("authToken")
+
+	// "mode=native" (default) connects to the local SSH daemon on this host.
+	// "mode=proxy" connects to an arbitrary host+port supplied in query params.
+	nativeSSH := r.URL.Query().Get("mode") != "proxy"
+
+	// In proxy mode we also need host + username from query params.
+	var target, username string
+	if !nativeSSH {
+		host := r.URL.Query().Get("host")
+		port := r.URL.Query().Get("port")
+		username = r.URL.Query().Get("username")
+		if host == "" || username == "" {
+			http.Error(w, "missing host or username", http.StatusBadRequest)
+			return
+		}
+		if port == "" {
+			port = "22"
+		}
+		sshPort, _ := strconv.Atoi(port)
+		if !g.isAllowed("ssh", host, sshPort, token) {
+			http.Error(w, "destination not allowed or auth token mismatch", http.StatusForbidden)
+			return
+		}
+		target = net.JoinHostPort(host, port)
+	} else {
+		// Native SSH mode: validate the token against any registered ssh target.
+		if !g.isTokenValid("ssh", token) {
+			http.Error(w, "unauthorized", http.StatusUnauthorized)
+			return
+		}
+		username = r.URL.Query().Get("username")
+		if username == "" {
+			http.Error(w, "missing username", http.StatusBadRequest)
+			return
+		}
+	}
+
+	ws, err := websocket.Accept(w, r, &websocket.AcceptOptions{
+		InsecureSkipVerify: true,
+		Subprotocols:       []string{"ssh"},
+	})
+	if err != nil {
+		log.Printf("SSH websocket upgrade failed: %v", err)
+		return
+	}
+	ws.SetReadLimit(-1)
+	defer ws.CloseNow() //nolint:errcheck
+
+	if nativeSSH {
+		if err := serveNativeSSHSession(ctx, ws, username, g.sshCreds); err != nil {
+			log.Printf("SSH native session error: %v", err)
+		}
+	} else {
+		if err := serveSSHSession(ctx, ws, target, username, g.authToken); err != nil {
+			log.Printf("SSH session error: %v", err)
+		}
+	}
+}
+
+func serveSSHSession(ctx context.Context, ws *websocket.Conn, target, username, _ string) error {
+	// -- Wait for the auth message from the client to get the password --
+	_, authBytes, err := ws.Read(ctx)
+	if err != nil {
+		return fmt.Errorf("read auth message: %w", err)
+	}
+	var authMsg sshClientMsg
+	if err := json.Unmarshal(authBytes, &authMsg); err != nil || authMsg.Type != "auth" {
+		return fmt.Errorf("expected auth message, got: %s", authBytes)
+	}
+	password := authMsg.Password
+	privateKey := authMsg.PrivateKey
+
+	// Build the list of auth methods. Private key takes priority when provided.
+	var authMethods []ssh.AuthMethod
+	if privateKey != "" {
+		signer, err := ssh.ParsePrivateKey([]byte(privateKey))
+		if err != nil {
+			sendSSHError(ctx, ws, fmt.Sprintf("Failed to parse private key: %v", err))
+			return fmt.Errorf("parse private key: %w", err)
+		}
+		authMethods = append(authMethods, ssh.PublicKeys(signer))
+	}
+	if password != "" {
+		authMethods = append(authMethods, ssh.Password(password))
+	}
+	if len(authMethods) == 0 {
+		sendSSHError(ctx, ws, "No authentication credentials provided")
+		return fmt.Errorf("no auth credentials")
+	}
+	log.Printf("SSH: connecting to %s as %s", target, username)
+	sshCfg := &ssh.ClientConfig{
+		User: username,
+		Auth: authMethods,
+		// HostKeyCallback is intentionally InsecureIgnoreHostKey for this dev
+		// proxy. In production, verify against a known-hosts store.
+		HostKeyCallback: ssh.InsecureIgnoreHostKey(), //nolint:gosec
+		Timeout:         15 * time.Second,
+	}
+
+	sshClient, err := ssh.Dial("tcp", target, sshCfg)
+	if err != nil {
+		sendSSHError(ctx, ws, fmt.Sprintf("SSH dial failed: %v", err))
+		return fmt.Errorf("ssh dial %s: %w", target, err)
+	}
+	defer sshClient.Close()
+
+	// -- Open an interactive session --
+	sess, err := sshClient.NewSession()
+	if err != nil {
+		sendSSHError(ctx, ws, fmt.Sprintf("Failed to open SSH session: %v", err))
+		return fmt.Errorf("ssh new session: %w", err)
+	}
+	defer sess.Close()
+
+	// Request a PTY.
+	if err := sess.RequestPty("xterm-256color", 24, 80, ssh.TerminalModes{
+		ssh.ECHO:          1,
+		ssh.TTY_OP_ISPEED: 38400,
+		ssh.TTY_OP_OSPEED: 38400,
+	}); err != nil {
+		sendSSHError(ctx, ws, fmt.Sprintf("Failed to request PTY: %v", err))
+		return fmt.Errorf("ssh request pty: %w", err)
+	}
+
+	stdinPipe, err := sess.StdinPipe()
+	if err != nil {
+		return fmt.Errorf("ssh stdin pipe: %w", err)
+	}
+	stdoutPipe, err := sess.StdoutPipe()
+	if err != nil {
+		return fmt.Errorf("ssh stdout pipe: %w", err)
+	}
+	stderrPipe, err := sess.StderrPipe()
+	if err != nil {
+		return fmt.Errorf("ssh stderr pipe: %w", err)
+	}
+
+	if err := sess.Shell(); err != nil {
+		sendSSHError(ctx, ws, fmt.Sprintf("Failed to start shell: %v", err))
+		return fmt.Errorf("ssh shell: %w", err)
+	}
+
+	log.Printf("SSH: session established with %s", target)
+
+	// -- Pump SSH stdout/stderr → WebSocket --
+	sessCtx, cancelSess := context.WithCancel(ctx)
+	defer cancelSess()
+
+	go func() {
+		buf := make([]byte, 4096)
+		for {
+			n, readErr := stdoutPipe.Read(buf)
+			if n > 0 {
+				msg := sshServerMsg{Type: "data", Data: string(buf[:n])}
+				b, _ := json.Marshal(msg)
+				if writeErr := ws.Write(sessCtx, websocket.MessageText, b); writeErr != nil {
+					return
+				}
+			}
+			if readErr != nil {
+				cancelSess()
+				return
+			}
+		}
+	}()
+
+	go func() {
+		buf := make([]byte, 4096)
+		for {
+			n, readErr := stderrPipe.Read(buf)
+			if n > 0 {
+				msg := sshServerMsg{Type: "data", Data: string(buf[:n])}
+				b, _ := json.Marshal(msg)
+				if writeErr := ws.Write(sessCtx, websocket.MessageText, b); writeErr != nil {
+					return
+				}
+			}
+			if readErr != nil {
+				return
+			}
+		}
+	}()
+
+	// -- Pump WebSocket input → SSH stdin / resize --
+	for {
+		_, msgBytes, readErr := ws.Read(sessCtx)
+		if readErr != nil {
+			break
+		}
+
+		var msg sshClientMsg
+		if err := json.Unmarshal(msgBytes, &msg); err != nil {
+			continue
+		}
+
+		switch msg.Type {
+		case "data":
+			if _, err := stdinPipe.Write([]byte(msg.Data)); err != nil {
+				return fmt.Errorf("write ssh stdin: %w", err)
+			}
+		case "resize":
+			if msg.Cols > 0 && msg.Rows > 0 {
+				_ = sess.WindowChange(int(msg.Rows), int(msg.Cols))
+			}
+		}
+	}
+
+	return nil
+}
+
+// sendSSHError sends an error message to the browser and logs it.
+func sendSSHError(ctx context.Context, ws *websocket.Conn, msg string) {
+	log.Printf("SSH error: %s", msg)
+	b, _ := json.Marshal(sshServerMsg{Type: "error", Error: msg})
+	_ = ws.Write(ctx, websocket.MessageText, b)
+}

browsergateway/ssh_native.go

@@ -0,0 +1,94 @@
+//go:build !windows
+
+package browsergateway
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log"
+
+	"github.com/coder/websocket"
+	"github.com/fosrl/newt/nativessh"
+)
+
+// serveNativeSSHSession handles a WebSocket SSH session by authenticating the
+// user against the host OS (authorized_keys then PAM password) and then
+// spawning a PTY+shell running as that user.
+//
+// The auth frame from the browser must be a JSON sshClientMsg with type="auth"
+// carrying the same password/privateKey fields used by the proxy SSH path.
+// The target username is passed in from the HTTP layer (query param).
+func serveNativeSSHSession(ctx context.Context, ws *websocket.Conn, username string, creds *nativessh.CredentialStore) error {
+	// Read the auth frame.
+	_, authBytes, err := ws.Read(ctx)
+	if err != nil {
+		return fmt.Errorf("read auth message: %w", err)
+	}
+	var authMsg sshClientMsg
+	if err := json.Unmarshal(authBytes, &authMsg); err != nil || authMsg.Type != "auth" {
+		return fmt.Errorf("expected auth message, got: %s", authBytes)
+	}
+
+	// Authenticate using host authorized_keys or PAM password.
+	if err := nativessh.AuthenticateWithCertificate(creds, username, authMsg.Password, authMsg.PrivateKey, authMsg.Certificate); err != nil {
+		sendSSHError(ctx, ws, "Authentication failed")
+		return fmt.Errorf("auth for user %q: %w", username, err)
+	}
+
+	log.Printf("SSH native: spawning shell as user %q", username)
+
+	sess, err := nativessh.NewPTYSessionAs(username)
+	if err != nil {
+		sendSSHError(ctx, ws, fmt.Sprintf("Failed to spawn shell: %v", err))
+		return fmt.Errorf("pty session as %q: %w", username, err)
+	}
+	defer sess.Close()
+
+	// Cancel context to unblock the WebSocket read loop when the shell exits.
+	sessCtx, cancelSess := context.WithCancel(ctx)
+	defer cancelSess()
+
+	// Pump PTY output → WebSocket.
+	go func() {
+		defer cancelSess()
+		buf := make([]byte, 4096)
+		for {
+			n, readErr := sess.Read(buf)
+			if n > 0 {
+				msg := sshServerMsg{Type: "data", Data: string(buf[:n])}
+				b, _ := json.Marshal(msg)
+				if writeErr := ws.Write(sessCtx, websocket.MessageText, b); writeErr != nil {
+					return
+				}
+			}
+			if readErr != nil {
+				return
+			}
+		}
+	}()
+
+	// Pump WebSocket input → PTY stdin / resize.
+	for {
+		_, msgBytes, readErr := ws.Read(sessCtx)
+		if readErr != nil {
+			break
+		}
+		var msg sshClientMsg
+		if err := json.Unmarshal(msgBytes, &msg); err != nil {
+			continue
+		}
+		switch msg.Type {
+		case "data":
+			if _, writeErr := sess.Write([]byte(msg.Data)); writeErr != nil {
+				return fmt.Errorf("write pty: %w", writeErr)
+			}
+		case "resize":
+			if msg.Cols > 0 && msg.Rows > 0 {
+				_ = sess.Resize(uint16(msg.Cols), uint16(msg.Rows))
+			}
+		}
+	}
+
+	return nil
+}

browsergateway/ssh_native_windows.go

@@ -0,0 +1,16 @@
+//go:build windows
+
+package browsergateway
+
+import (
+	"context"
+	"errors"
+
+	"github.com/coder/websocket"
+	"github.com/fosrl/newt/nativessh"
+)
+
+// serveNativeSSHSession is not supported on Windows.
+func serveNativeSSHSession(_ context.Context, _ *websocket.Conn, _ string, _ *nativessh.CredentialStore) error {
+	return errors.New("native SSH is not supported on Windows")
+}

browsergateway/vnc.go

@@ -0,0 +1,98 @@
+package browsergateway
+
+import (
+	"context"
+	"io"
+	"log"
+	"net"
+	"net/http"
+	"strconv"
+	"time"
+
+	"github.com/coder/websocket"
+)
+
+const (
+	vncDialTimeout    = 10 * time.Second
+	vncKeepAlive      = 30 * time.Second
+	vncForwardBufSize = 32 * 1024
+)
+
+// handleVNC proxies a noVNC WebSocket connection to a raw TCP VNC backend.
+// It follows the same auth-token-in-query-param pattern as handleSSH.
+//
+// Query parameters:
+//
+//	authToken – shared secret matching the -auth-token flag
+//	host      – VNC backend hostname or IP
+//	port      – VNC backend port (default: 5900)
+func (g *Gateway) handleVNC(w http.ResponseWriter, r *http.Request) {
+	host := r.URL.Query().Get("host")
+	port := r.URL.Query().Get("port")
+	if host == "" {
+		http.Error(w, "missing host", http.StatusBadRequest)
+		return
+	}
+	if port == "" {
+		port = "5900"
+	}
+	vncPort, _ := strconv.Atoi(port)
+	authToken := r.URL.Query().Get("authToken")
+	if !g.isAllowed("vnc", host, vncPort, authToken) {
+		http.Error(w, "destination not allowed or auth token mismatch", http.StatusForbidden)
+		return
+	}
+	target := net.JoinHostPort(host, port)
+
+	// Accept the WebSocket. noVNC negotiates the "binary" subprotocol;
+	// fall back gracefully when the client sends "base64" as well.
+	ws, err := websocket.Accept(w, r, &websocket.AcceptOptions{
+		InsecureSkipVerify: true,
+		Subprotocols:       []string{"binary", "base64"},
+	})
+	if err != nil {
+		log.Printf("vnc: websocket upgrade failed: %v", err)
+		return
+	}
+	ws.SetReadLimit(-1)
+	defer ws.CloseNow() //nolint:errcheck
+
+	ctx := r.Context()
+	if err := serveVNC(ctx, ws, target); err != nil {
+		log.Printf("vnc: session error (%s): %v", target, err)
+	}
+}
+
+func serveVNC(ctx context.Context, ws *websocket.Conn, target string) error {
+	// Dial the VNC backend TCP server.
+	dialer := &net.Dialer{
+		Timeout:   vncDialTimeout,
+		KeepAlive: vncKeepAlive,
+	}
+	conn, err := dialer.DialContext(ctx, "tcp", target)
+	if err != nil {
+		return err
+	}
+	defer conn.Close() //nolint:errcheck
+
+	// Expose the WebSocket as a plain net.Conn byte stream (binary frames).
+	stream := websocket.NetConn(ctx, ws, websocket.MessageBinary)
+	defer stream.Close() //nolint:errcheck
+
+	// Proxy bidirectionally: VNC backend <-> browser.
+	errc := make(chan error, 2)
+	go func() {
+		buf := make([]byte, vncForwardBufSize)
+		_, err := io.CopyBuffer(conn, stream, buf)
+		errc <- err
+	}()
+	go func() {
+		buf := make([]byte, vncForwardBufSize)
+		_, err := io.CopyBuffer(stream, conn, buf)
+		errc <- err
+	}()
+
+	// Return when either direction closes.
+	err = <-errc
+	return err
+}

clients.go

@@ -7,6 +7,7 @@ import (
 	wgnetstack "github.com/fosrl/newt/clients"
 	"github.com/fosrl/newt/clients/permissions"
 	"github.com/fosrl/newt/logger"
+	"github.com/fosrl/newt/nativessh"
 	"github.com/fosrl/newt/websocket"
 	"golang.zx2c4.com/wireguard/tun/netstack"
 )
@@ -14,7 +15,7 @@ import (
 var wgService *clients.WireGuardService
 var ready bool
 
-func setupClients(client *websocket.Client) {
+func setupClients(client *websocket.Client, credStore *nativessh.CredentialStore) {
 	var host = endpoint
 	if strings.HasPrefix(host, "http://") {
 		host = strings.TrimPrefix(host, "http://")
@@ -24,7 +25,7 @@ func setupClients(client *websocket.Client) {
 
 	host = strings.TrimSuffix(host, "/")
 
-	logger.Info("Setting up clients with netstack2...")
+	logger.Debug("Setting up clients with netstack2...")
 
 	// if useNativeInterface is true make sure we have permission to use native interface
 	if useNativeInterface {
@@ -42,6 +43,8 @@ func setupClients(client *websocket.Client) {
 		logger.Fatal("Failed to create WireGuard service: %v", err)
 	}
 
+	wgService.SetCredentialStore(credStore)
+
 	client.OnTokenUpdate(func(token string) {
 		wgService.SetToken(token)
 	})
@@ -63,7 +66,14 @@ func closeClients() {
 	}
 }
 
-func clientsHandleNewtConnection(publicKey string, endpoint string) {
+// setClientsBlocked enables or disables connection blocking on the WireGuard service.
+func setClientsBlocked(v bool) {
+	if wgService != nil {
+		wgService.SetBlocked(v)
+	}
+}
+
+func clientsHandleNewtConnection(publicKey string, endpoint string, relayPort uint16) {
 	if !ready {
 		return
 	}
@@ -77,7 +87,7 @@ func clientsHandleNewtConnection(publicKey string, endpoint string) {
 	endpoint = strings.Join(parts[:len(parts)-1], ":")
 
 	if wgService != nil {
-		wgService.StartHolepunch(publicKey, endpoint)
+		wgService.StartHolepunch(publicKey, endpoint, relayPort)
 	}
 }
 

clients/clients.go

@@ -2,6 +2,8 @@ package clients
 
 import (
 	"context"
+	"crypto/rand"
+	"encoding/hex"
 	"encoding/json"
 	"fmt"
 	"net"
@@ -11,12 +13,14 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/fosrl/newt/bind"
 	newtDevice "github.com/fosrl/newt/device"
 	"github.com/fosrl/newt/holepunch"
 	"github.com/fosrl/newt/logger"
+	"github.com/fosrl/newt/nativessh"
 	"github.com/fosrl/newt/netstack2"
 	"github.com/fosrl/newt/network"
 	"github.com/fosrl/newt/util"
@@ -34,18 +38,27 @@ type WgConfig struct {
 	IpAddress string   `json:"ipAddress"`
 	Peers     []Peer   `json:"peers"`
 	Targets   []Target `json:"targets"`
+	ChainId   string   `json:"chainId"`
 }
 
 type Target struct {
-	SourcePrefix string      `json:"sourcePrefix"`
-	DestPrefix   string      `json:"destPrefix"`
-	RewriteTo    string      `json:"rewriteTo,omitempty"`
-	PortRange    []PortRange `json:"portRange,omitempty"`
+	SourcePrefix   string                 `json:"sourcePrefix"`
+	SourcePrefixes []string               `json:"sourcePrefixes"`
+	DestPrefix     string                 `json:"destPrefix"`
+	RewriteTo      string                 `json:"rewriteTo,omitempty"`
+	DisableIcmp    bool                   `json:"disableIcmp,omitempty"`
+	PortRange      []PortRange            `json:"portRange,omitempty"`
+	ResourceId     int                    `json:"resourceId,omitempty"`
+	Protocol       string                 `json:"protocol,omitempty"`    // for now practicably either http or https
+	HTTPTargets    []netstack2.HTTPTarget `json:"httpTargets,omitempty"` // for http protocol, list of downstream services to load balance across
+	TLSCert        string                 `json:"tlsCert,omitempty"`     // PEM-encoded certificate for incoming HTTPS termination
+	TLSKey         string                 `json:"tlsKey,omitempty"`      // PEM-encoded private key for incoming HTTPS termination
 }
 
 type PortRange struct {
-	Min uint16 `json:"min"`
-	Max uint16 `json:"max"`
+	Min      uint16 `json:"min"`
+	Max      uint16 `json:"max"`
+	Protocol string `json:"protocol"` // "tcp" or "udp"
 }
 
 type Peer struct {
@@ -67,19 +80,20 @@ type PeerReading struct {
 }
 
 type WireGuardService struct {
-	interfaceName string
-	mtu           int
-	client        *websocket.Client
-	config        WgConfig
-	key           wgtypes.Key
-	newtId        string
-	lastReadings  map[string]PeerReading
-	mu            sync.Mutex
-	Port          uint16
-	host          string
-	serverPubKey  string
-	token         string
-	stopGetConfig func()
+	interfaceName        string
+	mtu                  int
+	client               *websocket.Client
+	config               WgConfig
+	key                  wgtypes.Key
+	newtId               string
+	lastReadings         map[string]PeerReading
+	mu                   sync.Mutex
+	Port                 uint16
+	host                 string
+	serverPubKey         string
+	token                string
+	stopGetConfig        func()
+	pendingConfigChainId string
 	// Netstack fields
 	tun    tun.Device
 	tnet   *netstack2.Net
@@ -96,12 +110,25 @@ type WireGuardService struct {
 	sharedBind         *bind.SharedBind
 	holePunchManager   *holepunch.Manager
 	useNativeInterface bool
+	// SSH server running on the clients' netstack
+	sshServer *sshServerHandle
+	credStore *nativessh.CredentialStore
 	// Direct UDP relay from main tunnel to clients' WireGuard
 	directRelayStop    chan struct{}
 	directRelayWg      sync.WaitGroup
 	netstackListener   net.PacketConn
 	netstackListenerMu sync.Mutex
 	wgTesterServer     *wgtester.Server
+
+	// connection blocking: when true, all new incoming connections are dropped
+	blocked atomic.Bool
+}
+
+// generateChainId generates a random chain ID for deduplicating round-trip messages.
+func generateChainId() string {
+	b := make([]byte, 8)
+	_, _ = rand.Read(b)
+	return hex.EncodeToString(b)
 }
 
 func NewWireGuardService(interfaceName string, port uint16, mtu int, host string, newtId string, wsClient *websocket.Client, dns string, useNativeInterface bool) (*WireGuardService, error) {
@@ -139,7 +166,7 @@ func NewWireGuardService(interfaceName string, port uint16, mtu int, host string
 	// Add a reference for the hole punch manager (creator already has one reference for WireGuard)
 	sharedBind.AddRef()
 
-	logger.Info("Created shared UDP socket on port %d (refcount: %d)", port, sharedBind.GetRefCount())
+	logger.Debug("Created shared UDP socket on port %d (refcount: %d)", port, sharedBind.GetRefCount())
 
 	// Parse DNS addresses
 	dnsAddrs := []netip.Addr{netip.MustParseAddr(dns)}
@@ -158,9 +185,8 @@ func NewWireGuardService(interfaceName string, port uint16, mtu int, host string
 		useNativeInterface: useNativeInterface,
 	}
 
-	// Create the holepunch manager with ResolveDomain function
-	// We'll need to pass a domain resolver function
-	service.holePunchManager = holepunch.NewManager(sharedBind, newtId, "newt", key.PublicKey().String())
+	// Create the holepunch manager
+	service.holePunchManager = holepunch.NewManager(sharedBind, newtId, "newt", key.PublicKey().String(), nil)
 
 	// Register websocket handlers
 	wsClient.RegisterHandler("newt/wg/receive-config", service.handleConfig)
@@ -170,10 +196,18 @@ func NewWireGuardService(interfaceName string, port uint16, mtu int, host string
 	wsClient.RegisterHandler("newt/wg/targets/add", service.handleAddTarget)
 	wsClient.RegisterHandler("newt/wg/targets/remove", service.handleRemoveTarget)
 	wsClient.RegisterHandler("newt/wg/targets/update", service.handleUpdateTarget)
+	wsClient.RegisterHandler("newt/wg/sync", service.handleSyncConfig)
 
 	return service, nil
 }
 
+// SetCredentialStore sets the in-memory SSH credential store used by the
+// native SSH server. It must be called before the netstack is configured
+// (i.e. before the first newt/wg/receive-config message is processed).
+func (s *WireGuardService) SetCredentialStore(store *nativessh.CredentialStore) {
+	s.credStore = store
+}
+
 // ReportRTT allows reporting native RTTs to telemetry, rate-limited externally.
 func (s *WireGuardService) ReportRTT(seconds float64) {
 	if s.serverPubKey == "" {
@@ -192,6 +226,21 @@ func (s *WireGuardService) Close() {
 		s.stopGetConfig = nil
 	}
 
+	// Stop SSH server before tearing down the netstack
+	if s.sshServer != nil {
+		s.sshServer.stop()
+		s.sshServer = nil
+	}
+
+	// Flush access logs before tearing down the tunnel
+	if s.tnet != nil {
+		if ph := s.tnet.GetProxyHandler(); ph != nil {
+			if al := ph.GetAccessLogger(); al != nil {
+				al.Close()
+			}
+		}
+	}
+
 	// Stop the direct UDP relay first
 	s.StopDirectUDPRelay()
 
@@ -258,6 +307,21 @@ func (s *WireGuardService) GetPublicKey() wgtypes.Key {
 	return s.key.PublicKey()
 }
 
+// SetBlocked enables or disables connection blocking for this WireGuard service.
+// The state is persisted and applied immediately to any active proxy handler.
+func (s *WireGuardService) SetBlocked(v bool) {
+	s.blocked.Store(v)
+	s.mu.Lock()
+	tnet := s.tnet
+	s.mu.Unlock()
+	if tnet == nil {
+		return
+	}
+	if ph := tnet.GetProxyHandler(); ph != nil {
+		ph.SetBlocked(v)
+	}
+}
+
 // SetOnNetstackReady sets a callback function to be called when the netstack interface is ready
 func (s *WireGuardService) SetOnNetstackReady(callback func(*netstack2.Net)) {
 	s.onNetstackReady = callback
@@ -268,16 +332,21 @@ func (s *WireGuardService) SetOnNetstackClose(callback func()) {
 }
 
 // StartHolepunch starts hole punching to a specific endpoint
-func (s *WireGuardService) StartHolepunch(publicKey string, endpoint string) {
+func (s *WireGuardService) StartHolepunch(publicKey string, endpoint string, relayPort uint16) {
 	if s.holePunchManager == nil {
 		logger.Warn("Hole punch manager not initialized")
 		return
 	}
 
+	if relayPort == 0 {
+		relayPort = 21820
+	}
+
 	// Convert websocket.ExitNode to holepunch.ExitNode
 	hpExitNodes := []holepunch.ExitNode{
 		{
 			Endpoint:  endpoint,
+			RelayPort: relayPort,
 			PublicKey: publicKey,
 		},
 	}
@@ -287,7 +356,7 @@ func (s *WireGuardService) StartHolepunch(publicKey string, endpoint string) {
 		logger.Warn("Failed to start hole punch: %v", err)
 	}
 
-	logger.Info("Starting hole punch to %s with public key: %s", endpoint, publicKey)
+	logger.Debug("Starting hole punch to %s with public key: %s", endpoint, publicKey)
 }
 
 // StartDirectUDPRelay starts a direct UDP relay from the main tunnel netstack to the clients' WireGuard.
@@ -334,7 +403,7 @@ func (s *WireGuardService) StartDirectUDPRelay(tunnelIP string) error {
 	// Set the netstack connection on the SharedBind so responses go back through the tunnel
 	s.sharedBind.SetNetstackConn(listener)
 
-	logger.Info("Started direct UDP relay on %s:%d (bidirectional via SharedBind)", tunnelIP, s.Port)
+	logger.Debug("Started direct UDP relay on %s:%d (bidirectional via SharedBind)", tunnelIP, s.Port)
 
 	// Start the relay goroutine to read from netstack and inject into SharedBind
 	s.directRelayWg.Add(1)
@@ -352,7 +421,7 @@ func (s *WireGuardService) runDirectUDPRelay(listener net.PacketConn) {
 	// Note: Don't close listener here - it's also used by SharedBind for sending responses
 	// It will be closed when the relay is stopped
 
-	logger.Info("Direct UDP relay started (bidirectional through SharedBind)")
+	logger.Debug("Direct UDP relay started (bidirectional through SharedBind)")
 
 	buf := make([]byte, 65535) // Max UDP packet size
 
@@ -433,12 +502,15 @@ func (s *WireGuardService) LoadRemoteConfig() error {
 		s.stopGetConfig()
 		s.stopGetConfig = nil
 	}
+	chainId := generateChainId()
+	s.pendingConfigChainId = chainId
 	s.stopGetConfig = s.client.SendMessageInterval("newt/wg/get-config", map[string]interface{}{
 		"publicKey": s.key.PublicKey().String(),
 		"port":      s.Port,
+		"chainId":   chainId,
 	}, 2*time.Second)
 
-	logger.Info("Requesting WireGuard configuration from remote server")
+	logger.Debug("Requesting WireGuard configuration from remote server")
 	go s.periodicBandwidthCheck()
 
 	return nil
@@ -448,7 +520,7 @@ func (s *WireGuardService) handleConfig(msg websocket.WSMessage) {
 	var config WgConfig
 
 	logger.Debug("Received message: %v", msg)
-	logger.Info("Received WireGuard clients configuration from remote server")
+	logger.Debug("Received WireGuard clients configuration from remote server")
 
 	jsonData, err := json.Marshal(msg.Data)
 	if err != nil {
@@ -460,6 +532,17 @@ func (s *WireGuardService) handleConfig(msg websocket.WSMessage) {
 		logger.Info("Error unmarshaling target data: %v", err)
 		return
 	}
+
+	// Deduplicate using chainId: discard responses that don't match the
+	// pending request, or that we have already processed.
+	if config.ChainId != "" {
+		if config.ChainId != s.pendingConfigChainId {
+			logger.Debug("Discarding duplicate/stale newt/wg/get-config response (chainId=%s, expected=%s)", config.ChainId, s.pendingConfigChainId)
+			return
+		}
+		s.pendingConfigChainId = "" // consume – further duplicates are rejected
+	}
+
 	s.config = config
 
 	if s.stopGetConfig != nil {
@@ -470,6 +553,8 @@ func (s *WireGuardService) handleConfig(msg websocket.WSMessage) {
 	// Ensure the WireGuard interface and peers are configured
 	if err := s.ensureWireguardInterface(config); err != nil {
 		logger.Error("Failed to ensure WireGuard interface: %v", err)
+		logger.Error("Clients functionality will be disabled until the interface can be created")
+		return
 	}
 
 	if err := s.ensureWireguardPeers(config.Peers); err != nil {
@@ -479,6 +564,196 @@ func (s *WireGuardService) handleConfig(msg websocket.WSMessage) {
 	if err := s.ensureTargets(config.Targets); err != nil {
 		logger.Error("Failed to ensure WireGuard targets: %v", err)
 	}
+
+	logger.Info("Client connectivity setup. Ready to accept connections from clients!")
+}
+
+// SyncConfig represents the configuration sent from server for syncing
+type SyncConfig struct {
+	Targets []Target `json:"targets"`
+	Peers   []Peer   `json:"peers"`
+}
+
+func (s *WireGuardService) handleSyncConfig(msg websocket.WSMessage) {
+	var syncConfig SyncConfig
+
+	logger.Debug("Received sync message: %v", msg)
+	logger.Info("Received sync configuration from remote server")
+
+	jsonData, err := json.Marshal(msg.Data)
+	if err != nil {
+		logger.Error("Error marshaling sync data: %v", err)
+		return
+	}
+
+	if err := json.Unmarshal(jsonData, &syncConfig); err != nil {
+		logger.Error("Error unmarshaling sync data: %v", err)
+		return
+	}
+
+	// Sync peers
+	if err := s.syncPeers(syncConfig.Peers); err != nil {
+		logger.Error("Failed to sync peers: %v", err)
+	}
+
+	// Sync targets
+	if err := s.syncTargets(syncConfig.Targets); err != nil {
+		logger.Error("Failed to sync targets: %v", err)
+	}
+}
+
+// syncPeers synchronizes the current peers with the desired state
+// It removes peers not in the desired list and adds missing ones
+func (s *WireGuardService) syncPeers(desiredPeers []Peer) error {
+	if s.device == nil {
+		return fmt.Errorf("WireGuard device is not initialized")
+	}
+
+	// Get current peers from the device
+	currentConfig, err := s.device.IpcGet()
+	if err != nil {
+		return fmt.Errorf("failed to get current device config: %v", err)
+	}
+
+	// Parse current peer public keys
+	lines := strings.Split(currentConfig, "\n")
+	currentPeerKeys := make(map[string]bool)
+	for _, line := range lines {
+		if strings.HasPrefix(line, "public_key=") {
+			pubKey := strings.TrimPrefix(line, "public_key=")
+			currentPeerKeys[pubKey] = true
+		}
+	}
+
+	// Build a map of desired peers by their public key (normalized)
+	desiredPeerMap := make(map[string]Peer)
+	for _, peer := range desiredPeers {
+		// Normalize the public key for comparison
+		pubKey, err := wgtypes.ParseKey(peer.PublicKey)
+		if err != nil {
+			logger.Warn("Invalid public key in desired peers: %s", peer.PublicKey)
+			continue
+		}
+		normalizedKey := util.FixKey(pubKey.String())
+		desiredPeerMap[normalizedKey] = peer
+	}
+
+	// Remove peers that are not in the desired list
+	for currentKey := range currentPeerKeys {
+		if _, exists := desiredPeerMap[currentKey]; !exists {
+			// Parse the key back to get the original format for removal
+			removeConfig := fmt.Sprintf("public_key=%s\nremove=true", currentKey)
+			if err := s.device.IpcSet(removeConfig); err != nil {
+				logger.Warn("Failed to remove peer %s during sync: %v", currentKey, err)
+			} else {
+				logger.Info("Removed peer %s during sync", currentKey)
+			}
+		}
+	}
+
+	// Add peers that are missing
+	for normalizedKey, peer := range desiredPeerMap {
+		if _, exists := currentPeerKeys[normalizedKey]; !exists {
+			if err := s.addPeerToDevice(peer); err != nil {
+				logger.Warn("Failed to add peer %s during sync: %v", peer.PublicKey, err)
+			} else {
+				logger.Info("Added peer %s during sync", peer.PublicKey)
+			}
+		}
+	}
+
+	return nil
+}
+
+// syncTargets synchronizes the current targets with the desired state
+// It removes targets not in the desired list and adds missing ones
+func (s *WireGuardService) syncTargets(desiredTargets []Target) error {
+	if s.tnet == nil {
+		// Native interface mode - proxy features not available, skip silently
+		logger.Debug("Skipping target sync - using native interface (no proxy support)")
+		return nil
+	}
+
+	// Get current rules from the proxy handler
+	currentRules := s.tnet.GetProxySubnetRules()
+
+	// Build a map of current rules by source+dest prefix
+	type ruleKey struct {
+		sourcePrefix string
+		destPrefix   string
+	}
+	currentRuleMap := make(map[ruleKey]bool)
+	for _, rule := range currentRules {
+		key := ruleKey{
+			sourcePrefix: rule.SourcePrefix.String(),
+			destPrefix:   rule.DestPrefix.String(),
+		}
+		currentRuleMap[key] = true
+	}
+
+	// Build a map of desired targets
+	desiredTargetMap := make(map[ruleKey]Target)
+	for _, target := range desiredTargets {
+		key := ruleKey{
+			sourcePrefix: target.SourcePrefix,
+			destPrefix:   target.DestPrefix,
+		}
+		desiredTargetMap[key] = target
+	}
+
+	// Remove targets that are not in the desired list
+	for _, rule := range currentRules {
+		key := ruleKey{
+			sourcePrefix: rule.SourcePrefix.String(),
+			destPrefix:   rule.DestPrefix.String(),
+		}
+		if _, exists := desiredTargetMap[key]; !exists {
+			s.tnet.RemoveProxySubnetRule(rule.SourcePrefix, rule.DestPrefix)
+			logger.Info("Removed target %s -> %s during sync", rule.SourcePrefix.String(), rule.DestPrefix.String())
+		}
+	}
+
+	// Add targets that are missing
+	for key, target := range desiredTargetMap {
+		if _, exists := currentRuleMap[key]; !exists {
+			sourcePrefix, err := netip.ParsePrefix(target.SourcePrefix)
+			if err != nil {
+				logger.Warn("Invalid source prefix %s during sync: %v", target.SourcePrefix, err)
+				continue
+			}
+
+			destPrefix, err := netip.ParsePrefix(target.DestPrefix)
+			if err != nil {
+				logger.Warn("Invalid dest prefix %s during sync: %v", target.DestPrefix, err)
+				continue
+			}
+
+			var portRanges []netstack2.PortRange
+			for _, pr := range target.PortRange {
+				portRanges = append(portRanges, netstack2.PortRange{
+					Min:      pr.Min,
+					Max:      pr.Max,
+					Protocol: pr.Protocol,
+				})
+			}
+
+			s.tnet.AddProxySubnetRule(netstack2.SubnetRule{
+				SourcePrefix: sourcePrefix,
+				DestPrefix:   destPrefix,
+				RewriteTo:    target.RewriteTo,
+				PortRanges:   portRanges,
+				DisableIcmp:  target.DisableIcmp,
+				ResourceId:   target.ResourceId,
+				Protocol:     target.Protocol,
+				HTTPTargets:  target.HTTPTargets,
+				TLSCert:      target.TLSCert,
+				TLSKey:       target.TLSKey,
+			})
+			logger.Info("Added target %s -> %s during sync", target.SourcePrefix, target.DestPrefix)
+		}
+	}
+
+	return nil
 }
 
 func (s *WireGuardService) ensureWireguardInterface(wgconfig WgConfig) error {
@@ -592,8 +867,9 @@ func (s *WireGuardService) ensureWireguardInterface(wgconfig WgConfig) error {
 		s.dns,
 		s.mtu,
 		netstack2.NetTunOptions{
-			EnableTCPProxy: true,
-			EnableUDPProxy: true,
+			EnableTCPProxy:  true,
+			EnableUDPProxy:  true,
+			EnableICMPProxy: true,
 		},
 	)
 	if err != nil {
@@ -603,6 +879,20 @@ func (s *WireGuardService) ensureWireguardInterface(wgconfig WgConfig) error {
 
 	s.TunnelIP = tunnelIP.String()
 
+	// Configure the access log sender to ship compressed session logs via websocket
+	s.tnet.SetAccessLogSender(func(data string) error {
+		return s.client.SendMessageNoLog("newt/access-log", map[string]interface{}{
+			"compressed": data,
+		})
+	})
+
+	// Configure the HTTP request log sender to ship compressed request logs via websocket
+	s.tnet.SetHTTPRequestLogSender(func(data string) error {
+		return s.client.SendMessageNoLog("newt/request-log", map[string]interface{}{
+			"compressed": data,
+		})
+	})
+
 	// Create WireGuard device using the shared bind
 	s.device = device.NewDevice(s.tun, s.sharedBind, device.NewLogger(
 		device.LogLevelSilent, // Use silent logging by default - could be made configurable
@@ -625,7 +915,7 @@ func (s *WireGuardService) ensureWireguardInterface(wgconfig WgConfig) error {
 		return fmt.Errorf("failed to bring up WireGuard device: %v", err)
 	}
 
-	logger.Info("WireGuard netstack device created and configured")
+	logger.Debug("WireGuard netstack device created and configured")
 
 	// Release the mutex before calling the callback
 	s.mu.Unlock()
@@ -636,7 +926,25 @@ func (s *WireGuardService) ensureWireguardInterface(wgconfig WgConfig) error {
 		logger.Error("Failed to start WireGuard tester server: %v", err)
 	}
 
+	// Start the SSH server on the clients' netstack (port 22).
+	// A nil credStore means SSH is disabled (--disable-ssh), so skip starting the server.
+	if s.credStore != nil {
+		if h, sshErr := startSSHOnNetstack(s.tnet, s.credStore); sshErr != nil {
+			logger.Warn("nativessh: not starting SSH server on clients netstack: %v", sshErr)
+		} else {
+			s.sshServer = h
+		}
+	}
+
 	// Note: we already unlocked above, so don't use defer unlock
+
+	// Apply any pending blocked state to the newly created proxy handler
+	if s.blocked.Load() {
+		if ph := s.tnet.GetProxyHandler(); ph != nil {
+			ph.SetBlocked(true)
+		}
+	}
+
 	return nil
 }
 
@@ -644,6 +952,11 @@ func (s *WireGuardService) ensureWireguardPeers(peers []Peer) error {
 	// For netstack, we need to manage peers differently
 	// We'll configure peers directly on the device using IPC
 
+	// Check if device is initialized
+	if s.device == nil {
+		return fmt.Errorf("WireGuard device is not initialized")
+	}
+
 	// First, clear all existing peers by getting current config and removing them
 	currentConfig, err := s.device.IpcGet()
 	if err != nil {
@@ -678,6 +991,19 @@ func (s *WireGuardService) ensureWireguardPeers(peers []Peer) error {
 	return nil
 }
 
+// resolveSourcePrefixes returns the effective list of source prefixes for a target,
+// supporting both the legacy single SourcePrefix field and the new SourcePrefixes array.
+// If SourcePrefixes is non-empty it takes precedence; otherwise SourcePrefix is used.
+func resolveSourcePrefixes(target Target) []string {
+	if len(target.SourcePrefixes) > 0 {
+		return target.SourcePrefixes
+	}
+	if target.SourcePrefix != "" {
+		return []string{target.SourcePrefix}
+	}
+	return nil
+}
+
 func (s *WireGuardService) ensureTargets(targets []Target) error {
 	if s.tnet == nil {
 		// Native interface mode - proxy features not available, skip silently
@@ -686,11 +1012,6 @@ func (s *WireGuardService) ensureTargets(targets []Target) error {
 	}
 
 	for _, target := range targets {
-		sourcePrefix, err := netip.ParsePrefix(target.SourcePrefix)
-		if err != nil {
-			return fmt.Errorf("invalid CIDR %s: %v", target.SourcePrefix, err)
-		}
-
 		destPrefix, err := netip.ParsePrefix(target.DestPrefix)
 		if err != nil {
 			return fmt.Errorf("invalid CIDR %s: %v", target.DestPrefix, err)
@@ -699,14 +1020,31 @@ func (s *WireGuardService) ensureTargets(targets []Target) error {
 		var portRanges []netstack2.PortRange
 		for _, pr := range target.PortRange {
 			portRanges = append(portRanges, netstack2.PortRange{
-				Min: pr.Min,
-				Max: pr.Max,
+				Min:      pr.Min,
+				Max:      pr.Max,
+				Protocol: pr.Protocol,
 			})
 		}
 
-		s.tnet.AddProxySubnetRule(sourcePrefix, destPrefix, target.RewriteTo, portRanges)
-
-		logger.Info("Added target subnet from %s to %s rewrite to %s with port ranges: %v", target.SourcePrefix, target.DestPrefix, target.RewriteTo, target.PortRange)
+		for _, sp := range resolveSourcePrefixes(target) {
+			sourcePrefix, err := netip.ParsePrefix(sp)
+			if err != nil {
+				return fmt.Errorf("invalid CIDR %s: %v", sp, err)
+			}
+			s.tnet.AddProxySubnetRule(netstack2.SubnetRule{
+				SourcePrefix: sourcePrefix,
+				DestPrefix:   destPrefix,
+				RewriteTo:    target.RewriteTo,
+				PortRanges:   portRanges,
+				DisableIcmp:  target.DisableIcmp,
+				ResourceId:   target.ResourceId,
+				Protocol:     target.Protocol,
+				HTTPTargets:  target.HTTPTargets,
+				TLSCert:      target.TLSCert,
+				TLSKey:       target.TLSKey,
+			})
+			logger.Info("Added target subnet from %s to %s rewrite to %s with port ranges: %v", sp, target.DestPrefix, target.RewriteTo, target.PortRange)
+		}
 	}
 
 	return nil
@@ -1017,21 +1355,22 @@ func (s *WireGuardService) processPeerBandwidth(publicKey string, rxBytes, txByt
 			// Update the last reading
 			s.lastReadings[publicKey] = currentReading
 
-			return &PeerBandwidth{
-				PublicKey: publicKey,
-				BytesIn:   bytesInMB,
-				BytesOut:  bytesOutMB,
+			// Only return bandwidth data if there was an increase
+			if bytesInDiff > 0 || bytesOutDiff > 0 {
+				return &PeerBandwidth{
+					PublicKey: publicKey,
+					BytesIn:   bytesInMB,
+					BytesOut:  bytesOutMB,
+				}
 			}
+
+			return nil
 		}
 	}
 
-	// For first reading or if readings are too close together, report 0
+	// For first reading or if readings are too close together, don't report
 	s.lastReadings[publicKey] = currentReading
-	return &PeerBandwidth{
-		PublicKey: publicKey,
-		BytesIn:   0,
-		BytesOut:  0,
-	}
+	return nil
 }
 
 func (s *WireGuardService) reportPeerBandwidth() error {
@@ -1075,12 +1414,6 @@ func (s *WireGuardService) handleAddTarget(msg websocket.WSMessage) {
 
 	// Process all targets
 	for _, target := range targets {
-		sourcePrefix, err := netip.ParsePrefix(target.SourcePrefix)
-		if err != nil {
-			logger.Info("Invalid CIDR %s: %v", target.SourcePrefix, err)
-			continue
-		}
-
 		destPrefix, err := netip.ParsePrefix(target.DestPrefix)
 		if err != nil {
 			logger.Info("Invalid CIDR %s: %v", target.DestPrefix, err)
@@ -1090,14 +1423,32 @@ func (s *WireGuardService) handleAddTarget(msg websocket.WSMessage) {
 		var portRanges []netstack2.PortRange
 		for _, pr := range target.PortRange {
 			portRanges = append(portRanges, netstack2.PortRange{
-				Min: pr.Min,
-				Max: pr.Max,
+				Min:      pr.Min,
+				Max:      pr.Max,
+				Protocol: pr.Protocol,
 			})
 		}
 
-		s.tnet.AddProxySubnetRule(sourcePrefix, destPrefix, target.RewriteTo, portRanges)
-
-		logger.Info("Added target subnet from %s to %s rewrite to %s with port ranges: %v", target.SourcePrefix, target.DestPrefix, target.RewriteTo, target.PortRange)
+		for _, sp := range resolveSourcePrefixes(target) {
+			sourcePrefix, err := netip.ParsePrefix(sp)
+			if err != nil {
+				logger.Info("Invalid CIDR %s: %v", sp, err)
+				continue
+			}
+			s.tnet.AddProxySubnetRule(netstack2.SubnetRule{
+				SourcePrefix: sourcePrefix,
+				DestPrefix:   destPrefix,
+				RewriteTo:    target.RewriteTo,
+				PortRanges:   portRanges,
+				DisableIcmp:  target.DisableIcmp,
+				ResourceId:   target.ResourceId,
+				Protocol:     target.Protocol,
+				HTTPTargets:  target.HTTPTargets,
+				TLSCert:      target.TLSCert,
+				TLSKey:       target.TLSKey,
+			})
+			logger.Info("Added target subnet from %s to %s rewrite to %s with port ranges: %v", sp, target.DestPrefix, target.RewriteTo, target.PortRange)
+		}
 	}
 }
 
@@ -1126,21 +1477,21 @@ func (s *WireGuardService) handleRemoveTarget(msg websocket.WSMessage) {
 
 	// Process all targets
 	for _, target := range targets {
-		sourcePrefix, err := netip.ParsePrefix(target.SourcePrefix)
-		if err != nil {
-			logger.Info("Invalid CIDR %s: %v", target.SourcePrefix, err)
-			continue
-		}
-
 		destPrefix, err := netip.ParsePrefix(target.DestPrefix)
 		if err != nil {
 			logger.Info("Invalid CIDR %s: %v", target.DestPrefix, err)
 			continue
 		}
 
-		s.tnet.RemoveProxySubnetRule(sourcePrefix, destPrefix)
-
-		logger.Info("Removed target subnet %s with destination %s", target.SourcePrefix, target.DestPrefix)
+		for _, sp := range resolveSourcePrefixes(target) {
+			sourcePrefix, err := netip.ParsePrefix(sp)
+			if err != nil {
+				logger.Info("Invalid CIDR %s: %v", sp, err)
+				continue
+			}
+			s.tnet.RemoveProxySubnetRule(sourcePrefix, destPrefix)
+			logger.Info("Removed target subnet %s with destination %s", sp, target.DestPrefix)
+		}
 	}
 }
 
@@ -1174,30 +1525,24 @@ func (s *WireGuardService) handleUpdateTarget(msg websocket.WSMessage) {
 
 	// Process all update requests
 	for _, target := range requests.OldTargets {
-		sourcePrefix, err := netip.ParsePrefix(target.SourcePrefix)
-		if err != nil {
-			logger.Info("Invalid CIDR %s: %v", target.SourcePrefix, err)
-			continue
-		}
-
 		destPrefix, err := netip.ParsePrefix(target.DestPrefix)
 		if err != nil {
 			logger.Info("Invalid CIDR %s: %v", target.DestPrefix, err)
 			continue
 		}
 
-		s.tnet.RemoveProxySubnetRule(sourcePrefix, destPrefix)
-		logger.Info("Removed target subnet %s with destination %s", target.SourcePrefix, target.DestPrefix)
+		for _, sp := range resolveSourcePrefixes(target) {
+			sourcePrefix, err := netip.ParsePrefix(sp)
+			if err != nil {
+				logger.Info("Invalid CIDR %s: %v", sp, err)
+				continue
+			}
+			s.tnet.RemoveProxySubnetRule(sourcePrefix, destPrefix)
+			logger.Info("Removed target subnet %s with destination %s", sp, target.DestPrefix)
+		}
 	}
 
 	for _, target := range requests.NewTargets {
-		// Now add the new target
-		sourcePrefix, err := netip.ParsePrefix(target.SourcePrefix)
-		if err != nil {
-			logger.Info("Invalid CIDR %s: %v", target.SourcePrefix, err)
-			continue
-		}
-
 		destPrefix, err := netip.ParsePrefix(target.DestPrefix)
 		if err != nil {
 			logger.Info("Invalid CIDR %s: %v", target.DestPrefix, err)
@@ -1207,13 +1552,32 @@ func (s *WireGuardService) handleUpdateTarget(msg websocket.WSMessage) {
 		var portRanges []netstack2.PortRange
 		for _, pr := range target.PortRange {
 			portRanges = append(portRanges, netstack2.PortRange{
-				Min: pr.Min,
-				Max: pr.Max,
+				Min:      pr.Min,
+				Max:      pr.Max,
+				Protocol: pr.Protocol,
 			})
 		}
 
-		s.tnet.AddProxySubnetRule(sourcePrefix, destPrefix, target.RewriteTo, portRanges)
-		logger.Info("Added target subnet from %s to %s rewrite to %s with port ranges: %v", target.SourcePrefix, target.DestPrefix, target.RewriteTo, target.PortRange)
+		for _, sp := range resolveSourcePrefixes(target) {
+			sourcePrefix, err := netip.ParsePrefix(sp)
+			if err != nil {
+				logger.Info("Invalid CIDR %s: %v", sp, err)
+				continue
+			}
+			s.tnet.AddProxySubnetRule(netstack2.SubnetRule{
+				SourcePrefix: sourcePrefix,
+				DestPrefix:   destPrefix,
+				RewriteTo:    target.RewriteTo,
+				PortRanges:   portRanges,
+				DisableIcmp:  target.DisableIcmp,
+				ResourceId:   target.ResourceId,
+				Protocol:     target.Protocol,
+				HTTPTargets:  target.HTTPTargets,
+				TLSCert:      target.TLSCert,
+				TLSKey:       target.TLSKey,
+			})
+			logger.Info("Added target subnet from %s to %s rewrite to %s with port ranges: %v", sp, target.DestPrefix, target.RewriteTo, target.PortRange)
+		}
 	}
 }
 
@@ -1253,3 +1617,33 @@ func (s *WireGuardService) filterReadOnlyFields(config string) string {
 
 	return strings.Join(filteredLines, "\n")
 }
+
+// sshServerHandle holds the listener so the SSH server can be stopped by
+// closing it.
+type sshServerHandle struct {
+	ln net.Listener
+}
+
+func (h *sshServerHandle) stop() {
+	_ = h.ln.Close()
+}
+
+// startSSHOnNetstack creates a TCP listener on port 22 of the clients' netstack
+// and starts serving SSH connections on it in the background. The returned
+// handle can be used to stop the server by closing the listener.
+func startSSHOnNetstack(tnet *netstack2.Net, creds *nativessh.CredentialStore) (*sshServerHandle, error) {
+	srv := nativessh.NewServer(nativessh.ServerConfig{
+		Credentials: creds,
+	})
+	ln, err := tnet.ListenTCP(&net.TCPAddr{Port: 22})
+	if err != nil {
+		return nil, fmt.Errorf("listen on netstack port 22: %w", err)
+	}
+	h := &sshServerHandle{ln: ln}
+	go func() {
+		if err := srv.Serve(ln); err != nil {
+			logger.Debug("nativessh: clients netstack server stopped: %v", err)
+		}
+	}()
+	return h, nil
+}

clients/permissions/permissions_android.go

@@ -0,0 +1,8 @@
+//go:build android
+
+package permissions
+
+// CheckNativeInterfacePermissions always allows permission on Android.
+func CheckNativeInterfacePermissions() error {
+	return nil
+}

clients/permissions/permissions_darwin.go

@@ -1,4 +1,4 @@
-//go:build darwin
+//go:build darwin && !ios
 
 package permissions
 

clients/permissions/permissions_ios.go

@@ -0,0 +1,8 @@
+//go:build ios
+
+package permissions
+
+// CheckNativeInterfacePermissions always allows permission on iOS.
+func CheckNativeInterfacePermissions() error {
+	return nil
+}

clients/permissions/permissions_linux.go

@@ -1,4 +1,4 @@
-//go:build linux
+//go:build linux && !android
 
 package permissions
 

common.go

@@ -5,8 +5,10 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"net"
 	"os"
 	"os/exec"
+	"regexp"
 	"strings"
 	"time"
 
@@ -119,19 +121,11 @@ func reliablePing(tnet *netstack.Net, dst string, baseTimeout time.Duration, max
 		totalLatency += latency
 		successCount++
 
-		// If we get at least one success, we can return early for health checks
-		if successCount > 0 {
-			avgLatency := totalLatency / time.Duration(successCount)
-			// logger.Debug("Reliable ping succeeded after %d attempts, avg latency: %v", attempt, avgLatency)
-			return avgLatency, nil
-		}
+		// Return on first success
+		return totalLatency / time.Duration(successCount), nil
 	}
 
-	if successCount == 0 {
-		return 0, fmt.Errorf("all %d ping attempts failed, last error: %v", maxAttempts, lastErr)
-	}
-
-	return totalLatency / time.Duration(successCount), nil
+	return 0, fmt.Errorf("all %d ping attempts failed, last error: %v", maxAttempts, lastErr)
 }
 
 func pingWithRetry(tnet *netstack.Net, dst string, timeout time.Duration) (stopChan chan struct{}, err error) {
@@ -206,6 +200,7 @@ func pingWithRetry(tnet *netstack.Net, dst string, timeout time.Duration) (stopC
 							logger.Warn(msgHealthFileWriteFailed, err)
 						}
 					}
+					return
 				}
 			case <-pingStopChan:
 				// Stop the goroutine when signaled
@@ -218,6 +213,25 @@ func pingWithRetry(tnet *netstack.Net, dst string, timeout time.Duration) (stopC
 	return stopChan, fmt.Errorf("initial ping attempts failed, continuing in background")
 }
 
+// shouldFireRecovery decides whether the data-plane recovery flow in
+// startPingCheck should run on this tick. Recovery fires once when the
+// consecutive-failure counter first crosses the threshold; the connectionLost
+// flag prevents re-firing until a successful ping resets the state.
+//
+// This condition was previously inlined into startPingCheck and AND-ed with
+// `currentInterval < maxInterval`, which silently broke recovery once
+// pingInterval's default was bumped to 15s while maxInterval stayed at 6s
+// (commit 8161fa6, March 2026): the gate became permanently false on default
+// settings, so the recovery code never executed and ping failures climbed
+// forever — the proximate cause of fosrl/newt#284, #310 and pangolin#1004.
+//
+// Recovery and backoff are independent concerns; the backoff ramp is now
+// computed separately in the caller. Do not re-introduce currentInterval
+// here.
+func shouldFireRecovery(consecutiveFailures, failureThreshold int, connectionLost bool) bool {
+	return consecutiveFailures >= failureThreshold && !connectionLost
+}
+
 func startPingCheck(tnet *netstack.Net, serverIP string, client *websocket.Client, tunnelID string) chan struct{} {
 	maxInterval := 6 * time.Second
 	currentInterval := pingInterval
@@ -277,35 +291,44 @@ func startPingCheck(tnet *netstack.Net, serverIP string, client *websocket.Clien
 
 					// More lenient threshold for declaring connection lost under load
 					failureThreshold := 4
-					if consecutiveFailures >= failureThreshold && currentInterval < maxInterval {
-						if !connectionLost {
-							connectionLost = true
-							logger.Warn("Connection to server lost after %d failures. Continuous reconnection attempts will be made.", consecutiveFailures)
-							if tunnelID != "" {
-								telemetry.IncReconnect(context.Background(), tunnelID, "client", telemetry.ReasonTimeout)
-							}
-							stopFunc = client.SendMessageInterval("newt/ping/request", map[string]interface{}{}, 3*time.Second)
-							// Send registration message to the server for backward compatibility
-							err := client.SendMessage("newt/wg/register", map[string]interface{}{
-								"publicKey":           publicKey.String(),
-								"backwardsCompatible": true,
-							})
+					if shouldFireRecovery(consecutiveFailures, failureThreshold, connectionLost) {
+						connectionLost = true
+						logger.Warn("Connection to server lost after %d failures. Continuous reconnection attempts will be made.", consecutiveFailures)
+						if tunnelID != "" {
+							telemetry.IncReconnect(context.Background(), tunnelID, "client", telemetry.ReasonTimeout)
+						}
+						pingChainId := generateChainId()
+						pendingPingChainId = pingChainId
+						stopFunc = client.SendMessageInterval("newt/ping/request", map[string]interface{}{
+							"chainId": pingChainId,
+						}, 3*time.Second)
+						// Send registration message to the server for backward compatibility
+						bcChainId := generateChainId()
+						pendingRegisterChainId = bcChainId
+						err := client.SendMessage("newt/wg/register", map[string]interface{}{
+							"publicKey":           publicKey.String(),
+							"backwardsCompatible": true,
+							"chainId":             bcChainId,
+						})
+						if err != nil {
+							logger.Error("Failed to send registration message: %v", err)
+						}
+						if healthFile != "" {
+							err = os.Remove(healthFile)
 							if err != nil {
-								logger.Error("Failed to send registration message: %v", err)
-							}
-							if healthFile != "" {
-								err = os.Remove(healthFile)
-								if err != nil {
-									logger.Error("Failed to remove health file: %v", err)
-								}
+								logger.Error("Failed to remove health file: %v", err)
 							}
 						}
-						currentInterval = time.Duration(float64(currentInterval) * 1.3) // Slower increase
+					}
+					// Backoff: ramp the periodic-ping interval up while we are
+					// past the failure threshold, capped at maxInterval. Kept
+					// independent of the recovery trigger above so the trigger
+					// fires on every outage regardless of pingInterval.
+					if consecutiveFailures >= failureThreshold && currentInterval < maxInterval {
+						currentInterval = time.Duration(float64(currentInterval) * 1.3)
 						if currentInterval > maxInterval {
 							currentInterval = maxInterval
 						}
-						ticker.Reset(currentInterval)
-						logger.Debug("Increased ping check interval to %v due to consecutive failures", currentInterval)
 					}
 				} else {
 					// Track recent latencies
@@ -363,27 +386,62 @@ func parseTargetData(data interface{}) (TargetData, error) {
 	return targetData, nil
 }
 
+// parseTargetString parses a target string in the format "listenPort:host:targetPort"
+// It properly handles IPv6 addresses which must be in brackets: "listenPort:[ipv6]:targetPort"
+// Examples:
+//   - IPv4: "3001:192.168.1.1:80"
+//   - IPv6: "3001:[::1]:8080" or "3001:[fd70:1452:b736:4dd5:caca:7db9:c588:f5b3]:80"
+//
+// Returns listenPort, targetAddress (in host:port format suitable for net.Dial), and error
+func parseTargetString(target string) (int, string, error) {
+	// Find the first colon to extract the listen port
+	firstColon := strings.Index(target, ":")
+	if firstColon == -1 {
+		return 0, "", fmt.Errorf("invalid target format, no colon found: %s", target)
+	}
+
+	listenPortStr := target[:firstColon]
+	var listenPort int
+	_, err := fmt.Sscanf(listenPortStr, "%d", &listenPort)
+	if err != nil {
+		return 0, "", fmt.Errorf("invalid listen port: %s", listenPortStr)
+	}
+	if listenPort <= 0 || listenPort > 65535 {
+		return 0, "", fmt.Errorf("listen port out of range: %d", listenPort)
+	}
+
+	// The remainder is host:targetPort - use net.SplitHostPort which handles IPv6 brackets
+	remainder := target[firstColon+1:]
+	host, targetPort, err := net.SplitHostPort(remainder)
+	if err != nil {
+		return 0, "", fmt.Errorf("invalid host:port format '%s': %w", remainder, err)
+	}
+
+	// Reject empty host or target port
+	if host == "" {
+		return 0, "", fmt.Errorf("empty host in target: %s", target)
+	}
+	if targetPort == "" {
+		return 0, "", fmt.Errorf("empty target port in target: %s", target)
+	}
+
+	// Reconstruct the target address using JoinHostPort (handles IPv6 properly)
+	targetAddr := net.JoinHostPort(host, targetPort)
+
+	return listenPort, targetAddr, nil
+}
+
 func updateTargets(pm *proxy.ProxyManager, action string, tunnelIP string, proto string, targetData TargetData) error {
 	for _, t := range targetData.Targets {
-		// Split the first number off of the target with : separator and use as the port
-		parts := strings.Split(t, ":")
-		if len(parts) != 3 {
-			logger.Info("Invalid target format: %s", t)
-			continue
-		}
-
-		// Get the port as an int
-		port := 0
-		_, err := fmt.Sscanf(parts[0], "%d", &port)
+		// Parse the target string, handling both IPv4 and IPv6 addresses
+		port, target, err := parseTargetString(t)
 		if err != nil {
-			logger.Info("Invalid port: %s", parts[0])
+			logger.Info("Invalid target format: %s (%v)", t, err)
 			continue
 		}
 
 		switch action {
 		case "add":
-			target := parts[1] + ":" + parts[2]
-
 			// Call updown script if provided
 			processedTarget := target
 			if updownScript != "" {
@@ -410,8 +468,6 @@ func updateTargets(pm *proxy.ProxyManager, action string, tunnelIP string, proto
 		case "remove":
 			logger.Info("Removing target with port %d", port)
 
-			target := parts[1] + ":" + parts[2]
-
 			// Call updown script if provided
 			if updownScript != "" {
 				_, err := executeUpdownScript(action, proto, target)
@@ -420,7 +476,7 @@ func updateTargets(pm *proxy.ProxyManager, action string, tunnelIP string, proto
 				}
 			}
 
-			err := pm.RemoveTarget(proto, tunnelIP, port)
+			err = pm.RemoveTarget(proto, tunnelIP, port)
 			if err != nil {
 				logger.Error("Failed to remove target: %v", err)
 				return err
@@ -475,15 +531,41 @@ func executeUpdownScript(action, proto, target string) (string, error) {
 	return target, nil
 }
 
-func sendBlueprint(client *websocket.Client) error {
-	if blueprintFile == "" {
+// interpolateBlueprint finds all {{...}} tokens in the raw blueprint bytes and
+// replaces recognised schemes with their resolved values. Currently supported:
+//
+//   - env.<VAR>  – replaced with the value of the named environment variable
+//
+// Any token that does not match a supported scheme is left as-is so that
+// future schemes (e.g. tag., api.) are preserved rather than silently dropped.
+func interpolateBlueprint(data []byte) []byte {
+	re := regexp.MustCompile(`\{\{([^}]+)\}\}`)
+	return re.ReplaceAllFunc(data, func(match []byte) []byte {
+		// strip the surrounding {{ }}
+		inner := strings.TrimSpace(string(match[2 : len(match)-2]))
+
+		if strings.HasPrefix(inner, "env.") {
+			varName := strings.TrimPrefix(inner, "env.")
+			return []byte(os.Getenv(varName))
+		}
+
+		// unrecognised scheme – leave the token untouched
+		return match
+	})
+}
+
+func sendBlueprint(client *websocket.Client, file string) error {
+	if file == "" {
 		return nil
 	}
 	// try to read the blueprint file
-	blueprintData, err := os.ReadFile(blueprintFile)
+	blueprintData, err := os.ReadFile(file)
 	if err != nil {
 		logger.Error("Failed to read blueprint file: %v", err)
 	} else {
+		// interpolate {{env.VAR}} (and any future schemes) before parsing
+		blueprintData = interpolateBlueprint(blueprintData)
+
 		// first we should convert the yaml to json and error if the yaml is bad
 		var yamlObj interface{}
 		var blueprintJsonData string

common_test.go

@@ -0,0 +1,251 @@
+package main
+
+import (
+	"net"
+	"testing"
+)
+
+func TestParseTargetString(t *testing.T) {
+	tests := []struct {
+		name           string
+		input          string
+		wantListenPort int
+		wantTargetAddr string
+		wantErr        bool
+	}{
+		// IPv4 test cases
+		{
+			name:           "valid IPv4 basic",
+			input:          "3001:192.168.1.1:80",
+			wantListenPort: 3001,
+			wantTargetAddr: "192.168.1.1:80",
+			wantErr:        false,
+		},
+		{
+			name:           "valid IPv4 localhost",
+			input:          "8080:127.0.0.1:3000",
+			wantListenPort: 8080,
+			wantTargetAddr: "127.0.0.1:3000",
+			wantErr:        false,
+		},
+		{
+			name:           "valid IPv4 same ports",
+			input:          "443:10.0.0.1:443",
+			wantListenPort: 443,
+			wantTargetAddr: "10.0.0.1:443",
+			wantErr:        false,
+		},
+
+		// IPv6 test cases
+		{
+			name:           "valid IPv6 loopback",
+			input:          "3001:[::1]:8080",
+			wantListenPort: 3001,
+			wantTargetAddr: "[::1]:8080",
+			wantErr:        false,
+		},
+		{
+			name:           "valid IPv6 full address",
+			input:          "80:[fd70:1452:b736:4dd5:caca:7db9:c588:f5b3]:8080",
+			wantListenPort: 80,
+			wantTargetAddr: "[fd70:1452:b736:4dd5:caca:7db9:c588:f5b3]:8080",
+			wantErr:        false,
+		},
+		{
+			name:           "valid IPv6 link-local",
+			input:          "443:[fe80::1]:443",
+			wantListenPort: 443,
+			wantTargetAddr: "[fe80::1]:443",
+			wantErr:        false,
+		},
+		{
+			name:           "valid IPv6 all zeros compressed",
+			input:          "8000:[::]:9000",
+			wantListenPort: 8000,
+			wantTargetAddr: "[::]:9000",
+			wantErr:        false,
+		},
+		{
+			name:           "valid IPv6 mixed notation",
+			input:          "5000:[::ffff:192.168.1.1]:6000",
+			wantListenPort: 5000,
+			wantTargetAddr: "[::ffff:192.168.1.1]:6000",
+			wantErr:        false,
+		},
+
+		// Hostname test cases
+		{
+			name:           "valid hostname",
+			input:          "8080:example.com:80",
+			wantListenPort: 8080,
+			wantTargetAddr: "example.com:80",
+			wantErr:        false,
+		},
+		{
+			name:           "valid hostname with subdomain",
+			input:          "443:api.example.com:8443",
+			wantListenPort: 443,
+			wantTargetAddr: "api.example.com:8443",
+			wantErr:        false,
+		},
+		{
+			name:           "valid localhost hostname",
+			input:          "3000:localhost:3000",
+			wantListenPort: 3000,
+			wantTargetAddr: "localhost:3000",
+			wantErr:        false,
+		},
+
+		// Error cases
+		{
+			name:    "invalid - no colons",
+			input:   "invalid",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - empty string",
+			input:   "",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - non-numeric listen port",
+			input:   "abc:192.168.1.1:80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - missing target port",
+			input:   "3001:192.168.1.1",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - IPv6 without brackets",
+			input:   "3001:fd70:1452:b736:4dd5:caca:7db9:c588:f5b3:80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - only listen port",
+			input:   "3001:",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - missing host",
+			input:   "3001::80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - IPv6 unclosed bracket",
+			input:   "3001:[::1:80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - listen port zero",
+			input:   "0:192.168.1.1:80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - listen port negative",
+			input:   "-1:192.168.1.1:80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - listen port out of range",
+			input:   "70000:192.168.1.1:80",
+			wantErr: true,
+		},
+		{
+			name:    "invalid - empty target port",
+			input:   "3001:192.168.1.1:",
+			wantErr: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			listenPort, targetAddr, err := parseTargetString(tt.input)
+
+			if (err != nil) != tt.wantErr {
+				t.Errorf("parseTargetString(%q) error = %v, wantErr %v", tt.input, err, tt.wantErr)
+				return
+			}
+
+			if tt.wantErr {
+				return // Don't check other values if we expected an error
+			}
+
+			if listenPort != tt.wantListenPort {
+				t.Errorf("parseTargetString(%q) listenPort = %d, want %d", tt.input, listenPort, tt.wantListenPort)
+			}
+
+			if targetAddr != tt.wantTargetAddr {
+				t.Errorf("parseTargetString(%q) targetAddr = %q, want %q", tt.input, targetAddr, tt.wantTargetAddr)
+			}
+		})
+	}
+}
+
+// TestParseTargetStringNetDialCompatibility verifies that the output is compatible with net.Dial
+func TestParseTargetStringNetDialCompatibility(t *testing.T) {
+	tests := []struct {
+		name  string
+		input string
+	}{
+		{"IPv4", "8080:127.0.0.1:80"},
+		{"IPv6 loopback", "8080:[::1]:80"},
+		{"IPv6 full", "8080:[2001:db8::1]:80"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			_, targetAddr, err := parseTargetString(tt.input)
+			if err != nil {
+				t.Fatalf("parseTargetString(%q) unexpected error: %v", tt.input, err)
+			}
+
+			// Verify the format is valid for net.Dial by checking it can be split back
+			// This doesn't actually dial, just validates the format
+			_, _, err = net.SplitHostPort(targetAddr)
+			if err != nil {
+				t.Errorf("parseTargetString(%q) produced invalid net.Dial format %q: %v", tt.input, targetAddr, err)
+			}
+		})
+	}
+}
+
+// TestShouldFireRecovery is the regression guard for the broken trigger gate
+// that prevented data-plane recovery from ever firing under default settings
+// (fosrl/newt#284, #310, pangolin#1004). The pre-fix condition was
+//
+//	consecutiveFailures >= failureThreshold && currentInterval < maxInterval
+//
+// which became permanently false once pingInterval's default was bumped from
+// 3s to 15s in commit 8161fa6 — currentInterval starts at pingInterval=15s,
+// maxInterval stayed at 6s, so 15<6 is false and the recovery branch never
+// executed.
+//
+// The fix is to drop currentInterval from the trigger condition entirely;
+// backoff is a separate concern computed in the caller. The cases below
+// exercise the documented contract.
+func TestShouldFireRecovery(t *testing.T) {
+	const threshold = 4
+	cases := []struct {
+		name           string
+		failures       int
+		connectionLost bool
+		want           bool
+	}{
+		{"below threshold, fresh", 3, false, false},
+		{"below threshold, already lost", 3, true, false},
+		{"at threshold, fresh — recovery must fire", threshold, false, true},
+		{"at threshold, already lost — gate prevents re-fire", threshold, true, false},
+		{"far above threshold, fresh", 100, false, true},
+		{"far above threshold, already lost", 100, true, false},
+	}
+	for _, c := range cases {
+		t.Run(c.name, func(t *testing.T) {
+			if got := shouldFireRecovery(c.failures, threshold, c.connectionLost); got != c.want {
+				t.Errorf("shouldFireRecovery(failures=%d, threshold=%d, lost=%v) = %v, want %v",
+					c.failures, threshold, c.connectionLost, got, c.want)
+			}
+		})
+	}
+}

docker-compose.yml

@@ -4,7 +4,7 @@ services:
     container_name: newt
     restart: unless-stopped
     environment:
-      - PANGOLIN_ENDPOINT=https://example.com
+      - PANGOLIN_ENDPOINT=https://app.pangolin.net
       - NEWT_ID=2ix2t8xk22ubpfy 
       - NEWT_SECRET=nnisrfsdfc7prqsp9ewo1dvtvci50j5uiqotez00dgap0ii2 
       - LOG_LEVEL=DEBUG

docker/docker.go

@@ -9,10 +9,9 @@ import (
 	"strings"
 	"time"
 
-	"github.com/docker/docker/api/types/container"
-	"github.com/docker/docker/api/types/events"
-	"github.com/docker/docker/api/types/filters"
-	"github.com/docker/docker/client"
+	"github.com/moby/moby/api/types/container"
+	"github.com/moby/moby/api/types/events"
+	"github.com/moby/moby/client"
 	"github.com/fosrl/newt/logger"
 )
 
@@ -170,7 +169,7 @@ func ListContainers(socketPath string, enforceNetworkValidation bool) ([]Contain
 	}
 
 	// Used to filter down containers returned to Pangolin
-	containerFilters := filters.NewArgs()
+	containerFilters := make(client.Filters)
 
 	// Used to determine if we will send IP addresses or hostnames to Pangolin
 	useContainerIpAddresses := true
@@ -215,21 +214,28 @@ func ListContainers(socketPath string, enforceNetworkValidation bool) ([]Contain
 	}
 
 	// List containers
-	containers, err := cli.ContainerList(ctx, container.ListOptions{All: true, Filters: containerFilters})
+	containerListResult, err := cli.ContainerList(ctx, client.ContainerListOptions{All: true, Filters: containerFilters})
 	if err != nil {
 		return nil, fmt.Errorf("failed to list containers: %v", err)
 	}
 
+	addrToString := func(addr interface{ IsValid() bool; String() string }) string {
+		if addr.IsValid() {
+			return addr.String()
+		}
+		return ""
+	}
+
 	var dockerContainers []Container
-	for _, c := range containers {
+	for _, c := range containerListResult.Items {
 		// Short ID like docker ps
 		shortId := c.ID[:12]
 
 		// Inspect container to get hostname
 		hostname := ""
-		containerInfo, err := cli.ContainerInspect(ctx, c.ID)
-		if err == nil && containerInfo.Config != nil {
-			hostname = containerInfo.Config.Hostname
+		containerInfo, err := cli.ContainerInspect(ctx, c.ID, client.ContainerInspectOptions{})
+		if err == nil && containerInfo.Container.Config != nil {
+			hostname = containerInfo.Container.Config.Hostname
 		}
 
 		// Skip host container if set
@@ -253,8 +259,8 @@ func ListContainers(socketPath string, enforceNetworkValidation bool) ([]Contain
 			if port.PublicPort != 0 {
 				dockerPort.PublicPort = int(port.PublicPort)
 			}
-			if port.IP != "" {
-				dockerPort.IP = port.IP
+			if port.IP.IsValid() {
+				dockerPort.IP = port.IP.String()
 			}
 			ports = append(ports, dockerPort)
 		}
@@ -268,19 +274,19 @@ func ListContainers(socketPath string, enforceNetworkValidation bool) ([]Contain
 				dockerNetwork := Network{
 					NetworkID:           endpoint.NetworkID,
 					EndpointID:          endpoint.EndpointID,
-					Gateway:             endpoint.Gateway,
+					Gateway:             addrToString(endpoint.Gateway),
 					IPPrefixLen:         endpoint.IPPrefixLen,
-					IPv6Gateway:         endpoint.IPv6Gateway,
-					GlobalIPv6Address:   endpoint.GlobalIPv6Address,
+					IPv6Gateway:         addrToString(endpoint.IPv6Gateway),
+					GlobalIPv6Address:   addrToString(endpoint.GlobalIPv6Address),
 					GlobalIPv6PrefixLen: endpoint.GlobalIPv6PrefixLen,
-					MacAddress:          endpoint.MacAddress,
+					MacAddress:          endpoint.MacAddress.String(),
 					Aliases:             endpoint.Aliases,
 					DNSNames:            endpoint.DNSNames,
 				}
 
 				// Use IPs over hostnames/containers as we're on the bridge network
 				if useContainerIpAddresses {
-					dockerNetwork.IPAddress = endpoint.IPAddress
+					dockerNetwork.IPAddress = addrToString(endpoint.IPAddress)
 				}
 
 				networks[networkName] = dockerNetwork
@@ -291,7 +297,7 @@ func ListContainers(socketPath string, enforceNetworkValidation bool) ([]Contain
 			ID:       shortId,
 			Name:     name,
 			Image:    c.Image,
-			State:    c.State,
+			State:    string(c.State),
 			Status:   c.Status,
 			Ports:    ports,
 			Labels:   c.Labels,
@@ -315,12 +321,12 @@ func getHostContainer(dockerContext context.Context, dockerClient *client.Client
 	}
 
 	// Get host container from the docker socket
-	hostContainer, err := dockerClient.ContainerInspect(dockerContext, hostContainerName)
+	hostContainer, err := dockerClient.ContainerInspect(dockerContext, hostContainerName, client.ContainerInspectOptions{})
 	if err != nil {
 		return nil, fmt.Errorf("failed to find host container")
 	}
 
-	return &hostContainer, nil
+	return &hostContainer.Container, nil
 }
 
 // EventCallback defines the function signature for handling Docker events
@@ -371,7 +377,7 @@ func (em *EventMonitor) Start() error {
 	logger.Debug("Starting Docker event monitoring")
 
 	// Filter for container events we care about
-	eventFilters := filters.NewArgs()
+	eventFilters := make(client.Filters)
 	eventFilters.Add("type", "container")
 	// eventFilters.Add("event", "create")
 	eventFilters.Add("event", "start")
@@ -382,9 +388,10 @@ func (em *EventMonitor) Start() error {
 	// eventFilters.Add("event", "unpause")
 
 	// Start listening for events
-	eventCh, errCh := em.client.Events(em.ctx, events.ListOptions{
+	eventsResult := em.client.Events(em.ctx, client.EventsListOptions{
 		Filters: eventFilters,
 	})
+	eventCh, errCh := eventsResult.Messages, eventsResult.Err
 
 	go func() {
 		defer func() {
@@ -408,9 +415,10 @@ func (em *EventMonitor) Start() error {
 					time.Sleep(5 * time.Second)
 					if em.ctx.Err() == nil {
 						logger.Info("Attempting to reconnect to Docker event stream")
-						eventCh, errCh = em.client.Events(em.ctx, events.ListOptions{
+						eventsResult = em.client.Events(em.ctx, client.EventsListOptions{
 							Filters: eventFilters,
 						})
+						eventCh, errCh = eventsResult.Messages, eventsResult.Err
 					}
 				}
 				return

flake.nix

@@ -25,7 +25,7 @@
           inherit (pkgs) lib;
 
           # Update version when releasing
-          version = "1.6.1";
+          version = "1.12.5";
         in
         {
           default = self.packages.${system}.pangolin-newt;
@@ -35,16 +35,28 @@
             inherit version;
             src = pkgs.nix-gitignore.gitignoreSource [ ] ./.;
 
-            vendorHash = "sha256-krxkfH+4z0rgmFi8OyCzWIrxU5Rpb7gjtGcn3LnZCig=";
+            vendorHash = "sha256-X70emc3uPN2YpsbudtoeEZDHilaTvFMqfRaDmIgRVZE=";
+
+            nativeInstallCheckInputs = [ pkgs.versionCheckHook ];
 
             env = {
               CGO_ENABLED = 0;
             };
 
             ldflags = [
+              "-s"
+              "-w"
               "-X main.newtVersion=${version}"
             ];
 
+            # Tests are broken due to a lack of Internet.
+            # Disable running `go test`, and instead do
+            # a simple version check instead.
+            doCheck = false;
+            doInstallCheck = true;
+
+            versionCheckProgramArg = [ "-version" ];
+
             meta = {
               description = "A tunneling client for Pangolin";
               homepage = "https://github.com/fosrl/newt";
@@ -52,6 +64,7 @@
               maintainers = [
                 lib.maintainers.water-sucks
               ];
+              mainProgram = "newt";
             };
           };
         }

get-newt.sh

@@ -1,7 +1,7 @@
-#!/bin/bash
+#!/bin/sh
 
 # Get Newt - Cross-platform installation script
-# Usage: curl -fsSL https://raw.githubusercontent.com/fosrl/newt/refs/heads/main/get-newt.sh | bash
+# Usage: curl -fsSL https://raw.githubusercontent.com/fosrl/newt/refs/heads/main/get-newt.sh | sh
 
 set -e
 
@@ -17,54 +17,51 @@ GITHUB_API_URL="https://api.github.com/repos/${REPO}/releases/latest"
 
 # Function to print colored output
 print_status() {
-    echo -e "${GREEN}[INFO]${NC} $1"
+    printf '%b[INFO]%b %s\n' "${GREEN}" "${NC}" "$1"
 }
 
 print_warning() {
-    echo -e "${YELLOW}[WARN]${NC} $1"
+    printf '%b[WARN]%b %s\n' "${YELLOW}" "${NC}" "$1"
 }
 
 print_error() {
-    echo -e "${RED}[ERROR]${NC} $1"
+    printf '%b[ERROR]%b %s\n' "${RED}" "${NC}" "$1"
 }
 
 # Function to get latest version from GitHub API
 get_latest_version() {
-    local latest_info
-    
+    latest_info=""
+
     if command -v curl >/dev/null 2>&1; then
         latest_info=$(curl -fsSL "$GITHUB_API_URL" 2>/dev/null)
     elif command -v wget >/dev/null 2>&1; then
         latest_info=$(wget -qO- "$GITHUB_API_URL" 2>/dev/null)
     else
-        print_error "Neither curl nor wget is available. Please install one of them." >&2
+        print_error "Neither curl nor wget is available."
         exit 1
     fi
-    
+
     if [ -z "$latest_info" ]; then
-        print_error "Failed to fetch latest version information" >&2
+        print_error "Failed to fetch latest version info"
         exit 1
     fi
-    
-    # Extract version from JSON response (works without jq)
-    local version=$(echo "$latest_info" | grep '"tag_name"' | head -1 | sed 's/.*"tag_name": *"\([^"]*\)".*/\1/')
-    
+
+    version=$(printf '%s' "$latest_info" | grep '"tag_name"' | head -1 | sed 's/.*"tag_name": *"\([^"]*\)".*/\1/')
+
     if [ -z "$version" ]; then
-        print_error "Could not parse version from GitHub API response" >&2
+        print_error "Could not parse version from GitHub API response"
         exit 1
     fi
-    
-    # Remove 'v' prefix if present
-    version=$(echo "$version" | sed 's/^v//')
-    
-    echo "$version"
+
+    version=$(printf '%s' "$version" | sed 's/^v//')
+    printf '%s' "$version"
 }
 
 # Detect OS and architecture
 detect_platform() {
-    local os arch
-    
-    # Detect OS
+    os=""
+    arch=""
+
     case "$(uname -s)" in
         Linux*)     os="linux" ;;
         Darwin*)    os="darwin" ;;
@@ -75,12 +72,11 @@ detect_platform() {
             exit 1
             ;;
     esac
-    
-    # Detect architecture
+
     case "$(uname -m)" in
         x86_64|amd64)   arch="amd64" ;;
         arm64|aarch64)  arch="arm64" ;;
-        armv7l|armv6l)  
+        armv7l|armv6l)
             if [ "$os" = "linux" ]; then
                 if [ "$(uname -m)" = "armv6l" ]; then
                     arch="arm32v6"
@@ -88,10 +84,10 @@ detect_platform() {
                     arch="arm32"
                 fi
             else
-                arch="arm64"  # Default for non-Linux ARM
+                arch="arm64"
             fi
             ;;
-        riscv64)        
+        riscv64)
             if [ "$os" = "linux" ]; then
                 arch="riscv64"
             else
@@ -104,90 +100,166 @@ detect_platform() {
             exit 1
             ;;
     esac
-    
-    echo "${os}_${arch}"
+
+    printf '%s_%s' "$os" "$arch"
 }
 
-# Get installation directory
+# Determine installation directory (default fallback)
 get_install_dir() {
-    if [ "$OS" = "windows" ]; then
-        echo "$HOME/bin"
-    else
-        # Try to use a directory in PATH, fallback to ~/.local/bin
-        if echo "$PATH" | grep -q "/usr/local/bin"; then
-            if [ -w "/usr/local/bin" ] 2>/dev/null; then
-                echo "/usr/local/bin"
-            else
-                echo "$HOME/.local/bin"
-            fi
-        else
-            echo "$HOME/.local/bin"
+    case "$PLATFORM" in
+        *windows*)
+            echo "$HOME/bin"
+            ;;
+        *)
+            echo "/usr/local/bin"
+            ;;
+    esac
+}
+
+# Parse --path argument from args
+# Returns the value after --path, or empty string if not provided
+parse_path_arg() {
+    while [ $# -gt 0 ]; do
+        case "$1" in
+            --path)
+                if [ -n "$2" ]; then
+                    printf '%s' "$2"
+                    return
+                fi
+                ;;
+            --path=*)
+                printf '%s' "${1#--path=}"
+                return
+                ;;
+        esac
+        shift
+    done
+}
+
+# Detect an existing newt binary location.
+# Tries unprivileged which first, then sudo which (for binaries only visible to root).
+# Returns the full path of the binary, or empty string if not found.
+detect_existing_binary() {
+    existing=""
+
+    # Try unprivileged which first
+    existing=$(command -v newt 2>/dev/null || true)
+    if [ -n "$existing" ]; then
+        printf '%s' "$existing"
+        return
+    fi
+
+    # Try sudo which — some installations land in paths only root can see in $PATH
+    if command -v sudo >/dev/null 2>&1; then
+        existing=$(sudo which newt 2>/dev/null || true)
+        if [ -n "$existing" ]; then
+            printf '%s' "$existing"
+            return
         fi
     fi
 }
 
+# Check if we need sudo for installation
+needs_sudo() {
+    install_dir="$1"
+    if [ -w "$install_dir" ] 2>/dev/null; then
+        return 1  # No sudo needed
+    else
+        return 0  # Sudo needed
+    fi
+}
+
+# Get the appropriate command prefix (sudo or empty)
+get_sudo_cmd() {
+    install_dir="$1"
+    if needs_sudo "$install_dir"; then
+        if command -v sudo >/dev/null 2>&1; then
+            echo "sudo"
+        else
+            print_error "Cannot write to ${install_dir} and sudo is not available."
+            print_error "Please run this script as root or install sudo."
+            exit 1
+        fi
+    else
+        echo ""
+    fi
+}
+
 # Download and install newt
 install_newt() {
-    local platform="$1"
-    local install_dir="$2"
-    local binary_name="newt_${platform}"
-    local exe_suffix=""
-    
-    # Add .exe suffix for Windows
-    if [[ "$platform" == *"windows"* ]]; then
-        binary_name="${binary_name}.exe"
-        exe_suffix=".exe"
+    platform="$1"
+    install_dir="$2"
+    sudo_cmd="$3"
+    custom_path="$4"
+    binary_name="newt_${platform}"
+    final_name="newt"
+
+    case "$platform" in
+        *windows*)
+            binary_name="${binary_name}.exe"
+            final_name="newt.exe"
+            ;;
+    esac
+
+    download_url="${BASE_URL}/${binary_name}"
+    temp_file="/tmp/${final_name}"
+
+    # If a custom path is provided, use it directly; otherwise use install_dir/final_name
+    if [ -n "$custom_path" ]; then
+        final_path="$custom_path"
+        install_dir=$(dirname "$final_path")
+    else
+        final_path="${install_dir}/${final_name}"
     fi
-    
-    local download_url="${BASE_URL}/${binary_name}"
-    local temp_file="/tmp/newt${exe_suffix}"
-    local final_path="${install_dir}/newt${exe_suffix}"
-    
+
     print_status "Downloading newt from ${download_url}"
-    
-    # Download the binary
+
     if command -v curl >/dev/null 2>&1; then
         curl -fsSL "$download_url" -o "$temp_file"
     elif command -v wget >/dev/null 2>&1; then
         wget -q "$download_url" -O "$temp_file"
     else
-        print_error "Neither curl nor wget is available. Please install one of them."
+        print_error "Neither curl nor wget is available."
         exit 1
     fi
-    
-    # Create install directory if it doesn't exist
-    mkdir -p "$install_dir"
-    
-    # Move binary to install directory
-    mv "$temp_file" "$final_path"
-    
-    # Make executable (not needed on Windows, but doesn't hurt)
-    chmod +x "$final_path"
-    
+
+    # Make executable before moving
+    chmod +x "$temp_file"
+
+    # Create install directory if it doesn't exist and move binary
+    if [ -n "$sudo_cmd" ]; then
+        $sudo_cmd mkdir -p "$install_dir"
+        print_status "Using sudo to install to ${install_dir}"
+        $sudo_cmd mv "$temp_file" "$final_path"
+    else
+        mkdir -p "$install_dir"
+        mv "$temp_file" "$final_path"
+    fi
+
     print_status "newt installed to ${final_path}"
-    
+
     # Check if install directory is in PATH
     if ! echo "$PATH" | grep -q "$install_dir"; then
         print_warning "Install directory ${install_dir} is not in your PATH."
-        print_warning "Add it to your PATH by adding this line to your shell profile:"
+        print_warning "Add it with:"
         print_warning "  export PATH=\"${install_dir}:\$PATH\""
     fi
 }
 
 # Verify installation
 verify_installation() {
-    local install_dir="$1"
-    local exe_suffix=""
-    
-    if [[ "$PLATFORM" == *"windows"* ]]; then
-        exe_suffix=".exe"
-    fi
-    
-    local newt_path="${install_dir}/newt${exe_suffix}"
-    
-    if [ -f "$newt_path" ] && [ -x "$newt_path" ]; then
+    install_dir="$1"
+    exe_suffix=""
+
+    case "$PLATFORM" in
+        *windows*) exe_suffix=".exe" ;;
+    esac
+
+    newt_path="${install_dir}/newt${exe_suffix}"
+
+    if [ -x "$newt_path" ]; then
         print_status "Installation successful!"
-        print_status "newt version: $("$newt_path" --version 2>/dev/null || echo "unknown")"
+        print_status "newt version: $("$newt_path" --version 2>/dev/null || printf 'unknown')"
         return 0
     else
         print_error "Installation failed. Binary not found or not executable."
@@ -197,39 +269,66 @@ verify_installation() {
 
 # Main installation process
 main() {
-    print_status "Installing latest version of newt..."
-    
-    # Get latest version
-    print_status "Fetching latest version from GitHub..."
+    # --path explicitly overrides everything
+    CUSTOM_PATH=$(parse_path_arg "$@")
+
+    if [ -n "$CUSTOM_PATH" ]; then
+        print_status "Installing latest version of newt to ${CUSTOM_PATH}..."
+    else
+        print_status "Installing latest version of newt..."
+    fi
+
+    print_status "Fetching latest version..."
     VERSION=$(get_latest_version)
     print_status "Latest version: v${VERSION}"
-    
-    # Set base URL with the fetched version
+
     BASE_URL="https://github.com/${REPO}/releases/download/${VERSION}"
-    
-    # Detect platform
+
     PLATFORM=$(detect_platform)
     print_status "Detected platform: ${PLATFORM}"
-    
-    # Get install directory
-    INSTALL_DIR=$(get_install_dir)
-    print_status "Install directory: ${INSTALL_DIR}"
-    
-    # Install newt
-    install_newt "$PLATFORM" "$INSTALL_DIR"
-    
-    # Verify installation
-    if verify_installation "$INSTALL_DIR"; then
-        print_status "newt is ready to use!"
-        if [[ "$PLATFORM" == *"windows"* ]]; then
-            print_status "Run 'newt --help' to get started"
+
+    if [ -n "$CUSTOM_PATH" ]; then
+        # --path wins; derive INSTALL_DIR from it
+        INSTALL_DIR=$(dirname "$CUSTOM_PATH")
+    else
+        # Try to find an existing installation so we update the right place
+        EXISTING_BINARY=$(detect_existing_binary)
+        if [ -n "$EXISTING_BINARY" ]; then
+            print_status "Found existing newt binary at ${EXISTING_BINARY}"
+            CUSTOM_PATH="$EXISTING_BINARY"
+            INSTALL_DIR=$(dirname "$EXISTING_BINARY")
+            print_status "Will update existing installation at ${INSTALL_DIR}"
         else
-            print_status "Run 'newt --help' to get started"
+            INSTALL_DIR=$(get_install_dir)
         fi
+    fi
+
+    print_status "Install directory: ${INSTALL_DIR}"
+
+    # Check if we need sudo
+    SUDO_CMD=$(get_sudo_cmd "$INSTALL_DIR")
+    if [ -n "$SUDO_CMD" ]; then
+        print_status "Root privileges required for installation to ${INSTALL_DIR}"
+    fi
+
+    install_newt "$PLATFORM" "$INSTALL_DIR" "$SUDO_CMD" "$CUSTOM_PATH"
+
+    if [ -n "$CUSTOM_PATH" ]; then
+        if [ -x "$CUSTOM_PATH" ]; then
+            print_status "Installation successful!"
+            print_status "newt version: $("$CUSTOM_PATH" --version 2>/dev/null || printf 'unknown')"
+            print_status "newt is ready to use!"
+        else
+            print_error "Installation failed. Binary not found or not executable at ${CUSTOM_PATH}."
+            exit 1
+        fi
+    elif verify_installation "$INSTALL_DIR"; then
+        print_status "newt is ready to use!"
+        print_status "Run 'newt --help' to get started."
     else
         exit 1
     fi
 }
 
 # Run main function
-main "$@"
+main "$@"

go.mod

@@ -1,77 +1,74 @@
 module github.com/fosrl/newt
 
-go 1.25
+go 1.25.0
 
 require (
-	github.com/docker/docker v28.5.2+incompatible
+	github.com/coder/websocket v1.8.14
+	github.com/creack/pty v1.1.24
+	github.com/gaissmai/bart v0.26.1
+	github.com/go-crypt/crypt v0.14.15
+	github.com/go-crypt/x v0.4.16
 	github.com/gorilla/websocket v1.5.3
+	github.com/moby/moby/api v1.54.2
+	github.com/moby/moby/client v0.4.1
 	github.com/prometheus/client_golang v1.23.2
 	github.com/vishvananda/netlink v1.3.1
-	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.63.0
-	go.opentelemetry.io/contrib/instrumentation/runtime v0.63.0
-	go.opentelemetry.io/otel v1.38.0
-	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.38.0
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.38.0
-	go.opentelemetry.io/otel/exporters/prometheus v0.60.0
-	go.opentelemetry.io/otel/metric v1.38.0
-	go.opentelemetry.io/otel/sdk v1.38.0
-	go.opentelemetry.io/otel/sdk/metric v1.38.0
-	golang.org/x/crypto v0.45.0
+	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.68.0
+	go.opentelemetry.io/contrib/instrumentation/runtime v0.68.0
+	go.opentelemetry.io/otel v1.43.0
+	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.43.0
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.43.0
+	go.opentelemetry.io/otel/exporters/prometheus v0.65.0
+	go.opentelemetry.io/otel/metric v1.43.0
+	go.opentelemetry.io/otel/sdk v1.43.0
+	go.opentelemetry.io/otel/sdk/metric v1.43.0
+	golang.org/x/crypto v0.52.0
 	golang.org/x/exp v0.0.0-20251113190631-e25ba8c21ef6
-	golang.org/x/net v0.47.0
-	golang.org/x/sys v0.38.0
+	golang.org/x/net v0.55.0
+	golang.org/x/sys v0.45.0
 	golang.zx2c4.com/wireguard v0.0.0-20250521234502-f333402bd9cb
 	golang.zx2c4.com/wireguard/wgctrl v0.0.0-20241231184526-a9ab2273dd10
-	golang.zx2c4.com/wireguard/windows v0.5.3
-	google.golang.org/grpc v1.76.0
+	golang.zx2c4.com/wireguard/windows v1.0.1
+	google.golang.org/grpc v1.81.1
 	gopkg.in/yaml.v3 v3.0.1
 	gvisor.dev/gvisor v0.0.0-20250503011706-39ed1f5ac29c
-	software.sslmate.com/src/go-pkcs12 v0.6.0
+	software.sslmate.com/src/go-pkcs12 v0.7.1
 )
 
 require (
-	github.com/Microsoft/go-winio v0.6.0 // indirect
+	github.com/Microsoft/go-winio v0.6.2 // indirect
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/cenkalti/backoff/v5 v5.0.3 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
-	github.com/containerd/errdefs v0.3.0 // indirect
+	github.com/containerd/errdefs v1.0.0 // indirect
 	github.com/containerd/errdefs/pkg v0.3.0 // indirect
 	github.com/distribution/reference v0.6.0 // indirect
-	github.com/docker/go-connections v0.6.0 // indirect
-	github.com/docker/go-units v0.4.0 // indirect
+	github.com/docker/go-connections v0.7.0 // indirect
+	github.com/docker/go-units v0.5.0 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
 	github.com/go-logr/logr v1.4.3 // indirect
 	github.com/go-logr/stdr v1.2.2 // indirect
 	github.com/google/btree v1.1.3 // indirect
 	github.com/google/uuid v1.6.0 // indirect
-	github.com/grafana/regexp v0.0.0-20240518133315-a468a5bfb3bc // indirect
-	github.com/grpc-ecosystem/grpc-gateway/v2 v2.27.2 // indirect
+	github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0 // indirect
 	github.com/moby/docker-image-spec v1.3.1 // indirect
-	github.com/moby/sys/atomicwriter v0.1.0 // indirect
-	github.com/moby/term v0.5.2 // indirect
-	github.com/morikuni/aec v1.0.0 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/opencontainers/go-digest v1.0.0 // indirect
-	github.com/opencontainers/image-spec v1.1.0 // indirect
-	github.com/pkg/errors v0.9.1 // indirect
+	github.com/opencontainers/image-spec v1.1.1 // indirect
 	github.com/prometheus/client_model v0.6.2 // indirect
-	github.com/prometheus/common v0.66.1 // indirect
-	github.com/prometheus/otlptranslator v0.0.2 // indirect
-	github.com/prometheus/procfs v0.17.0 // indirect
+	github.com/prometheus/common v0.67.5 // indirect
+	github.com/prometheus/otlptranslator v1.0.0 // indirect
+	github.com/prometheus/procfs v0.20.1 // indirect
 	github.com/vishvananda/netns v0.0.5 // indirect
-	go.opentelemetry.io/auto/sdk v1.1.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.38.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.38.0 // indirect
-	go.opentelemetry.io/otel/trace v1.38.0 // indirect
-	go.opentelemetry.io/proto/otlp v1.7.1 // indirect
-	go.yaml.in/yaml/v2 v2.4.2 // indirect
-	golang.org/x/mod v0.30.0 // indirect
-	golang.org/x/sync v0.18.0 // indirect
-	golang.org/x/text v0.31.0 // indirect
+	go.opentelemetry.io/auto/sdk v1.2.1 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.43.0 // indirect
+	go.opentelemetry.io/otel/trace v1.43.0 // indirect
+	go.opentelemetry.io/proto/otlp v1.10.0 // indirect
+	go.yaml.in/yaml/v2 v2.4.4 // indirect
+	golang.org/x/text v0.37.0 // indirect
 	golang.org/x/time v0.12.0 // indirect
-	golang.org/x/tools v0.39.0 // indirect
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 // indirect
-	google.golang.org/genproto/googleapis/api v0.0.0-20250825161204-c5933d9347a5 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20250825161204-c5933d9347a5 // indirect
-	google.golang.org/protobuf v1.36.8 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9 // indirect
+	google.golang.org/protobuf v1.36.11 // indirect
 )

go.sum

@@ -1,31 +1,35 @@
-github.com/Azure/go-ansiterm v0.0.0-20250102033503-faa5f7b0171c h1:udKWzYgxTojEKWjV8V+WSxDXJ4NFATAsZjh8iIbsQIg=
-github.com/Azure/go-ansiterm v0.0.0-20250102033503-faa5f7b0171c/go.mod h1:xomTg63KZ2rFqZQzSB4Vz2SUXa1BpHTVz9L5PTmPC4E=
-github.com/Microsoft/go-winio v0.6.0 h1:slsWYD/zyx7lCXoZVlvQrj0hPTM1HI4+v1sIda2yDvg=
-github.com/Microsoft/go-winio v0.6.0/go.mod h1:cTAf44im0RAYeL23bpB+fzCyDH2MJiz2BO69KH/soAE=
+github.com/Microsoft/go-winio v0.6.2 h1:F2VQgta7ecxGYO8k3ZZz3RS8fVIXVxONVUPlNERoyfY=
+github.com/Microsoft/go-winio v0.6.2/go.mod h1:yd8OoFMLzJbo9gZq8j5qaps8bJ9aShtEA8Ipt1oGCvU=
 github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
 github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
 github.com/cenkalti/backoff/v5 v5.0.3 h1:ZN+IMa753KfX5hd8vVaMixjnqRZ3y8CuJKRKj1xcsSM=
 github.com/cenkalti/backoff/v5 v5.0.3/go.mod h1:rkhZdG3JZukswDf7f0cwqPNk4K0sa+F97BxZthm/crw=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
-github.com/containerd/errdefs v0.3.0 h1:FSZgGOeK4yuT/+DnF07/Olde/q4KBoMsaamhXxIMDp4=
-github.com/containerd/errdefs v0.3.0/go.mod h1:+YBYIdtsnF4Iw6nWZhJcqGSg/dwvV7tyJ/kCkyJ2k+M=
+github.com/coder/websocket v1.8.14 h1:9L0p0iKiNOibykf283eHkKUHHrpG7f65OE3BhhO7v9g=
+github.com/coder/websocket v1.8.14/go.mod h1:NX3SzP+inril6yawo5CQXx8+fk145lPDC6pumgx0mVg=
+github.com/containerd/errdefs v1.0.0 h1:tg5yIfIlQIrxYtu9ajqY42W3lpS19XqdxRQeEwYG8PI=
+github.com/containerd/errdefs v1.0.0/go.mod h1:+YBYIdtsnF4Iw6nWZhJcqGSg/dwvV7tyJ/kCkyJ2k+M=
 github.com/containerd/errdefs/pkg v0.3.0 h1:9IKJ06FvyNlexW690DXuQNx2KA2cUJXx151Xdx3ZPPE=
 github.com/containerd/errdefs/pkg v0.3.0/go.mod h1:NJw6s9HwNuRhnjJhM7pylWwMyAkmCQvQ4GpJHEqRLVk=
-github.com/containerd/log v0.1.0 h1:TCJt7ioM2cr/tfR8GPbGf9/VRAX8D2B4PjzCpfX540I=
-github.com/containerd/log v0.1.0/go.mod h1:VRRf09a7mHDIRezVKTRCrOq78v577GXq3bSa3EhrzVo=
+github.com/creack/pty v1.1.24 h1:bJrF4RRfyJnbTJqzRLHzcGaZK1NeM5kTC9jGgovnR1s=
+github.com/creack/pty v1.1.24/go.mod h1:08sCNb52WyoAwi2QDyzUCTgcvVFhUzewun7wtTfvcwE=
 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/distribution/reference v0.6.0 h1:0IXCQ5g4/QMHHkarYzh5l+u8T3t73zM5QvfrDyIgxBk=
 github.com/distribution/reference v0.6.0/go.mod h1:BbU0aIcezP1/5jX/8MP0YiH4SdvB5Y4f/wlDRiLyi3E=
-github.com/docker/docker v28.5.2+incompatible h1:DBX0Y0zAjZbSrm1uzOkdr1onVghKaftjlSWt4AFexzM=
-github.com/docker/docker v28.5.2+incompatible/go.mod h1:eEKB0N0r5NX/I1kEveEz05bcu8tLC/8azJZsviup8Sk=
-github.com/docker/go-connections v0.6.0 h1:LlMG9azAe1TqfR7sO+NJttz1gy6KO7VJBh+pMmjSD94=
-github.com/docker/go-connections v0.6.0/go.mod h1:AahvXYshr6JgfUJGdDCs2b5EZG/vmaMAntpSFH5BFKE=
-github.com/docker/go-units v0.4.0 h1:3uh0PgVws3nIA0Q+MwDC8yjEPf9zjRfZZWXZYDct3Tw=
-github.com/docker/go-units v0.4.0/go.mod h1:fgPhTUdO+D/Jk86RDLlptpiXQzgHJF7gydDDbaIK4Dk=
+github.com/docker/go-connections v0.7.0 h1:6SsRfJddP22WMrCkj19x9WKjEDTB+ahsdiGYf0mN39c=
+github.com/docker/go-connections v0.7.0/go.mod h1:no1qkHdjq7kLMGUXYAduOhYPSJxxvgWBh7ogVvptn3Q=
+github.com/docker/go-units v0.5.0 h1:69rxXcBk27SvSaaxTtLh/8llcHD8vYHT7WSdRZ/jvr4=
+github.com/docker/go-units v0.5.0/go.mod h1:fgPhTUdO+D/Jk86RDLlptpiXQzgHJF7gydDDbaIK4Dk=
 github.com/felixge/httpsnoop v1.0.4 h1:NFTV2Zj1bL4mc9sqWACXbQFVBBg2W3GPvqp8/ESS2Wg=
 github.com/felixge/httpsnoop v1.0.4/go.mod h1:m8KPJKqk1gH5J9DgRY2ASl2lWCfGKXixSwevea8zH2U=
+github.com/gaissmai/bart v0.26.1 h1:+w4rnLGNlA2GDVn382Tfe3jOsK5vOr5n4KmigJ9lbTo=
+github.com/gaissmai/bart v0.26.1/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
+github.com/go-crypt/crypt v0.14.15 h1:q1i5OMpL05r935IxWmXgpDAVF0nvi4SMoHhGXLBQUEQ=
+github.com/go-crypt/crypt v0.14.15/go.mod h1:0n/to1VqIZPENj2yEUa/sLLYYnmupma6cp+QMX4zfF0=
+github.com/go-crypt/x v0.4.16 h1:WXdY28H/0MsXnH+gwerxuCcvBTJPkBG90u6oS4gIPZI=
+github.com/go-crypt/x v0.4.16/go.mod h1:vmVFA/d/oLrEaCbqsLcjBMlTqF8u8pvH/c4+EJ/ped8=
 github.com/go-logr/logr v1.2.2/go.mod h1:jdQByPbusPIv2/zmleS9BjJVeZ6kBagPoEUsqbVz/1A=
 github.com/go-logr/logr v1.4.3 h1:CjnDlHq8ikf6E492q6eKboGOC0T8CDaOvkHCIg8idEI=
 github.com/go-logr/logr v1.4.3/go.mod h1:9T104GzyrTigFIr8wt5mBrctHMim0Nb2HLGrmQ40KvY=
@@ -41,10 +45,8 @@ github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/gorilla/websocket v1.5.3 h1:saDtZ6Pbx/0u+bgYQ3q96pZgCzfhKXGPqt7kZ72aNNg=
 github.com/gorilla/websocket v1.5.3/go.mod h1:YR8l580nyteQvAITg2hZ9XVh4b55+EU/adAjf1fMHhE=
-github.com/grafana/regexp v0.0.0-20240518133315-a468a5bfb3bc h1:GN2Lv3MGO7AS6PrRoT6yV5+wkrOpcszoIsO4+4ds248=
-github.com/grafana/regexp v0.0.0-20240518133315-a468a5bfb3bc/go.mod h1:+JKpmjMGhpgPL+rXZ5nsZieVzvarn86asRlBg4uNGnk=
-github.com/grpc-ecosystem/grpc-gateway/v2 v2.27.2 h1:8Tjv8EJ+pM1xP8mK6egEbD1OgnVTyacbefKhmbLhIhU=
-github.com/grpc-ecosystem/grpc-gateway/v2 v2.27.2/go.mod h1:pkJQ2tZHJ0aFOVEEot6oZmaVEZcRme73eIFmhiVuRWs=
+github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0 h1:HWRh5R2+9EifMyIHV7ZV+MIZqgz+PMpZ14Jynv3O2Zs=
+github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0/go.mod h1:JfhWUomR1baixubs02l85lZYYOm7LV6om4ceouMv45c=
 github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
 github.com/klauspost/compress v1.18.0/go.mod h1:2Pp+KzxcywXVXMr50+X0Q/Lsb43OQHYWRCY2AiWywWQ=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
@@ -55,122 +57,110 @@ github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
 github.com/moby/docker-image-spec v1.3.1 h1:jMKff3w6PgbfSa69GfNg+zN/XLhfXJGnEx3Nl2EsFP0=
 github.com/moby/docker-image-spec v1.3.1/go.mod h1:eKmb5VW8vQEh/BAr2yvVNvuiJuY6UIocYsFu/DxxRpo=
-github.com/moby/sys/atomicwriter v0.1.0 h1:kw5D/EqkBwsBFi0ss9v1VG3wIkVhzGvLklJ+w3A14Sw=
-github.com/moby/sys/atomicwriter v0.1.0/go.mod h1:Ul8oqv2ZMNHOceF643P6FKPXeCmYtlQMvpizfsSoaWs=
-github.com/moby/sys/sequential v0.6.0 h1:qrx7XFUd/5DxtqcoH1h438hF5TmOvzC/lspjy7zgvCU=
-github.com/moby/sys/sequential v0.6.0/go.mod h1:uyv8EUTrca5PnDsdMGXhZe6CCe8U/UiTWd+lL+7b/Ko=
-github.com/moby/term v0.5.2 h1:6qk3FJAFDs6i/q3W/pQ97SX192qKfZgGjCQqfCJkgzQ=
-github.com/moby/term v0.5.2/go.mod h1:d3djjFCrjnB+fl8NJux+EJzu0msscUP+f8it8hPkFLc=
-github.com/morikuni/aec v1.0.0 h1:nP9CBfwrvYnBRgY6qfDQkygYDmYwOilePFkwzv4dU8A=
-github.com/morikuni/aec v1.0.0/go.mod h1:BbKIizmSmc5MMPqRYbxO4ZU0S0+P200+tUnFx7PXmsc=
+github.com/moby/moby/api v1.54.2 h1:wiat9QAhnDQjA7wk1kh/TqHz2I1uUA7M7t9SAl/JNXg=
+github.com/moby/moby/api v1.54.2/go.mod h1:+RQ6wluLwtYaTd1WnPLykIDPekkuyD/ROWQClE83pzs=
+github.com/moby/moby/client v0.4.1 h1:DMQgisVoMkmMs7fp3ROSdiBnoAu8+vo3GggFl06M/wY=
+github.com/moby/moby/client v0.4.1/go.mod h1:z52C9O2POPOsnxZAy//WtKcQ32P+jT/NGeXu/7nfjGQ=
 github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 h1:C3w9PqII01/Oq1c1nUAm88MOHcQC9l5mIlSMApZMrHA=
 github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822/go.mod h1:+n7T8mK8HuQTcFwEeznm/DIxMOiR9yIdICNftLE1DvQ=
 github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8Oi/yOhh5U=
 github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
-github.com/opencontainers/image-spec v1.1.0 h1:8SG7/vwALn54lVB/0yZ/MMwhFrPYtpEHQb2IpWsCzug=
-github.com/opencontainers/image-spec v1.1.0/go.mod h1:W4s4sFTMaBeK1BQLXbG4AdM2szdn85PY75RI83NrTrM=
-github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4=
-github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINEl0=
+github.com/opencontainers/image-spec v1.1.1 h1:y0fUlFfIZhPF1W537XOLg0/fcx6zcHCJwooC2xJA040=
+github.com/opencontainers/image-spec v1.1.1/go.mod h1:qpqAh3Dmcf36wStyyWU+kCeDgrGnAve2nCC8+7h8Q0M=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/prometheus/client_golang v1.23.2 h1:Je96obch5RDVy3FDMndoUsjAhG5Edi49h0RJWRi/o0o=
 github.com/prometheus/client_golang v1.23.2/go.mod h1:Tb1a6LWHB3/SPIzCoaDXI4I8UHKeFTEQ1YCr+0Gyqmg=
 github.com/prometheus/client_model v0.6.2 h1:oBsgwpGs7iVziMvrGhE53c/GrLUsZdHnqNwqPLxwZyk=
 github.com/prometheus/client_model v0.6.2/go.mod h1:y3m2F6Gdpfy6Ut/GBsUqTWZqCUvMVzSfMLjcu6wAwpE=
-github.com/prometheus/common v0.66.1 h1:h5E0h5/Y8niHc5DlaLlWLArTQI7tMrsfQjHV+d9ZoGs=
-github.com/prometheus/common v0.66.1/go.mod h1:gcaUsgf3KfRSwHY4dIMXLPV0K/Wg1oZ8+SbZk/HH/dA=
-github.com/prometheus/otlptranslator v0.0.2 h1:+1CdeLVrRQ6Psmhnobldo0kTp96Rj80DRXRd5OSnMEQ=
-github.com/prometheus/otlptranslator v0.0.2/go.mod h1:P8AwMgdD7XEr6QRUJ2QWLpiAZTgTE2UYgjlu3svompI=
-github.com/prometheus/procfs v0.17.0 h1:FuLQ+05u4ZI+SS/w9+BWEM2TXiHKsUQ9TADiRH7DuK0=
-github.com/prometheus/procfs v0.17.0/go.mod h1:oPQLaDAMRbA+u8H5Pbfq+dl3VDAvHxMUOVhe0wYB2zw=
-github.com/rogpeppe/go-internal v1.13.1 h1:KvO1DLK/DRN07sQ1LQKScxyZJuNnedQ5/wKSR38lUII=
-github.com/rogpeppe/go-internal v1.13.1/go.mod h1:uMEvuHeurkdAXX61udpOXGD/AzZDWNMNyH2VO9fmH0o=
-github.com/sirupsen/logrus v1.9.3 h1:dueUQJ1C2q9oE3F7wvmSGAaVtTmUizReu6fjN8uqzbQ=
-github.com/sirupsen/logrus v1.9.3/go.mod h1:naHLuLoDiP4jHNo9R0sCBMtWGeIprob74mVsIT4qYEQ=
+github.com/prometheus/common v0.67.5 h1:pIgK94WWlQt1WLwAC5j2ynLaBRDiinoAb86HZHTUGI4=
+github.com/prometheus/common v0.67.5/go.mod h1:SjE/0MzDEEAyrdr5Gqc6G+sXI67maCxzaT3A2+HqjUw=
+github.com/prometheus/otlptranslator v1.0.0 h1:s0LJW/iN9dkIH+EnhiD3BlkkP5QVIUVEoIwkU+A6qos=
+github.com/prometheus/otlptranslator v1.0.0/go.mod h1:vRYWnXvI6aWGpsdY/mOT/cbeVRBlPWtBNDb7kGR3uKM=
+github.com/prometheus/procfs v0.20.1 h1:XwbrGOIplXW/AU3YhIhLODXMJYyC1isLFfYCsTEycfc=
+github.com/prometheus/procfs v0.20.1/go.mod h1:o9EMBZGRyvDrSPH1RqdxhojkuXstoe4UlK79eF5TGGo=
+github.com/rogpeppe/go-internal v1.14.1 h1:UQB4HGPB6osV0SQTLymcB4TgvyWu6ZyliaW0tI/otEQ=
+github.com/rogpeppe/go-internal v1.14.1/go.mod h1:MaRKkUm5W0goXpeCfT7UZI6fk/L7L7so1lCWt35ZSgc=
 github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
 github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
 github.com/vishvananda/netlink v1.3.1 h1:3AEMt62VKqz90r0tmNhog0r/PpWKmrEShJU0wJW6bV0=
 github.com/vishvananda/netlink v1.3.1/go.mod h1:ARtKouGSTGchR8aMwmkzC0qiNPrrWO5JS/XMVl45+b4=
 github.com/vishvananda/netns v0.0.5 h1:DfiHV+j8bA32MFM7bfEunvT8IAqQ/NzSJHtcmW5zdEY=
 github.com/vishvananda/netns v0.0.5/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
-go.opentelemetry.io/auto/sdk v1.1.0 h1:cH53jehLUN6UFLY71z+NDOiNJqDdPRaXzTel0sJySYA=
-go.opentelemetry.io/auto/sdk v1.1.0/go.mod h1:3wSPjt5PWp2RhlCcmmOial7AvC4DQqZb7a7wCow3W8A=
-go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.63.0 h1:RbKq8BG0FI8OiXhBfcRtqqHcZcka+gU3cskNuf05R18=
-go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.63.0/go.mod h1:h06DGIukJOevXaj/xrNjhi/2098RZzcLTbc0jDAUbsg=
-go.opentelemetry.io/contrib/instrumentation/runtime v0.63.0 h1:PeBoRj6af6xMI7qCupwFvTbbnd49V7n5YpG6pg8iDYQ=
-go.opentelemetry.io/contrib/instrumentation/runtime v0.63.0/go.mod h1:ingqBCtMCe8I4vpz/UVzCW6sxoqgZB37nao91mLQ3Bw=
-go.opentelemetry.io/otel v1.38.0 h1:RkfdswUDRimDg0m2Az18RKOsnI8UDzppJAtj01/Ymk8=
-go.opentelemetry.io/otel v1.38.0/go.mod h1:zcmtmQ1+YmQM9wrNsTGV/q/uyusom3P8RxwExxkZhjM=
-go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.38.0 h1:vl9obrcoWVKp/lwl8tRE33853I8Xru9HFbw/skNeLs8=
-go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.38.0/go.mod h1:GAXRxmLJcVM3u22IjTg74zWBrRCKq8BnOqUVLodpcpw=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.38.0 h1:GqRJVj7UmLjCVyVJ3ZFLdPRmhDUp2zFmQe3RHIOsw24=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.38.0/go.mod h1:ri3aaHSmCTVYu2AWv44YMauwAQc0aqI9gHKIcSbI1pU=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.38.0 h1:lwI4Dc5leUqENgGuQImwLo4WnuXFPetmPpkLi2IrX54=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.38.0/go.mod h1:Kz/oCE7z5wuyhPxsXDuaPteSWqjSBD5YaSdbxZYGbGk=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.38.0 h1:aTL7F04bJHUlztTsNGJ2l+6he8c+y/b//eR0jjjemT4=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.38.0/go.mod h1:kldtb7jDTeol0l3ewcmd8SDvx3EmIE7lyvqbasU3QC4=
-go.opentelemetry.io/otel/exporters/prometheus v0.60.0 h1:cGtQxGvZbnrWdC2GyjZi0PDKVSLWP/Jocix3QWfXtbo=
-go.opentelemetry.io/otel/exporters/prometheus v0.60.0/go.mod h1:hkd1EekxNo69PTV4OWFGZcKQiIqg0RfuWExcPKFvepk=
-go.opentelemetry.io/otel/metric v1.38.0 h1:Kl6lzIYGAh5M159u9NgiRkmoMKjvbsKtYRwgfrA6WpA=
-go.opentelemetry.io/otel/metric v1.38.0/go.mod h1:kB5n/QoRM8YwmUahxvI3bO34eVtQf2i4utNVLr9gEmI=
-go.opentelemetry.io/otel/sdk v1.38.0 h1:l48sr5YbNf2hpCUj/FoGhW9yDkl+Ma+LrVl8qaM5b+E=
-go.opentelemetry.io/otel/sdk v1.38.0/go.mod h1:ghmNdGlVemJI3+ZB5iDEuk4bWA3GkTpW+DOoZMYBVVg=
-go.opentelemetry.io/otel/sdk/metric v1.38.0 h1:aSH66iL0aZqo//xXzQLYozmWrXxyFkBJ6qT5wthqPoM=
-go.opentelemetry.io/otel/sdk/metric v1.38.0/go.mod h1:dg9PBnW9XdQ1Hd6ZnRz689CbtrUp0wMMs9iPcgT9EZA=
-go.opentelemetry.io/otel/trace v1.38.0 h1:Fxk5bKrDZJUH+AMyyIXGcFAPah0oRcT+LuNtJrmcNLE=
-go.opentelemetry.io/otel/trace v1.38.0/go.mod h1:j1P9ivuFsTceSWe1oY+EeW3sc+Pp42sO++GHkg4wwhs=
-go.opentelemetry.io/proto/otlp v1.7.1 h1:gTOMpGDb0WTBOP8JaO72iL3auEZhVmAQg4ipjOVAtj4=
-go.opentelemetry.io/proto/otlp v1.7.1/go.mod h1:b2rVh6rfI/s2pHWNlB7ILJcRALpcNDzKhACevjI+ZnE=
+go.opentelemetry.io/auto/sdk v1.2.1 h1:jXsnJ4Lmnqd11kwkBV2LgLoFMZKizbCi5fNZ/ipaZ64=
+go.opentelemetry.io/auto/sdk v1.2.1/go.mod h1:KRTj+aOaElaLi+wW1kO/DZRXwkF4C5xPbEe3ZiIhN7Y=
+go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.68.0 h1:CqXxU8VOmDefoh0+ztfGaymYbhdB/tT3zs79QaZTNGY=
+go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.68.0/go.mod h1:BuhAPThV8PBHBvg8ZzZ/Ok3idOdhWIodywz2xEcRbJo=
+go.opentelemetry.io/contrib/instrumentation/runtime v0.68.0 h1:jhVIQEprwUTV+KfzzliLidclhoTOoHTgdz96kAyR8mU=
+go.opentelemetry.io/contrib/instrumentation/runtime v0.68.0/go.mod h1:4HsdbLUbernaTnA8CNaNE+1g026SciXb3juRYe3l8EY=
+go.opentelemetry.io/otel v1.43.0 h1:mYIM03dnh5zfN7HautFE4ieIig9amkNANT+xcVxAj9I=
+go.opentelemetry.io/otel v1.43.0/go.mod h1:JuG+u74mvjvcm8vj8pI5XiHy1zDeoCS2LB1spIq7Ay0=
+go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.43.0 h1:8UQVDcZxOJLtX6gxtDt3vY2WTgvZqMQRzjsqiIHQdkc=
+go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.43.0/go.mod h1:2lmweYCiHYpEjQ/lSJBYhj9jP1zvCvQW4BqL9dnT7FQ=
+go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.43.0 h1:88Y4s2C8oTui1LGM6bTWkw0ICGcOLCAI5l6zsD1j20k=
+go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.43.0/go.mod h1:Vl1/iaggsuRlrHf/hfPJPvVag77kKyvrLeD10kpMl+A=
+go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.43.0 h1:RAE+JPfvEmvy+0LzyUA25/SGawPwIUbZ6u0Wug54sLc=
+go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.43.0/go.mod h1:AGmbycVGEsRx9mXMZ75CsOyhSP6MFIcj/6dnG+vhVjk=
+go.opentelemetry.io/otel/exporters/prometheus v0.65.0 h1:jOveH/b4lU9HT7y+Gfamf18BqlOuz2PWEvs8yM7Q6XE=
+go.opentelemetry.io/otel/exporters/prometheus v0.65.0/go.mod h1:i1P8pcumauPtUI4YNopea1dhzEMuEqWP1xoUZDylLHo=
+go.opentelemetry.io/otel/metric v1.43.0 h1:d7638QeInOnuwOONPp4JAOGfbCEpYb+K6DVWvdxGzgM=
+go.opentelemetry.io/otel/metric v1.43.0/go.mod h1:RDnPtIxvqlgO8GRW18W6Z/4P462ldprJtfxHxyKd2PY=
+go.opentelemetry.io/otel/sdk v1.43.0 h1:pi5mE86i5rTeLXqoF/hhiBtUNcrAGHLKQdhg4h4V9Dg=
+go.opentelemetry.io/otel/sdk v1.43.0/go.mod h1:P+IkVU3iWukmiit/Yf9AWvpyRDlUeBaRg6Y+C58QHzg=
+go.opentelemetry.io/otel/sdk/metric v1.43.0 h1:S88dyqXjJkuBNLeMcVPRFXpRw2fuwdvfCGLEo89fDkw=
+go.opentelemetry.io/otel/sdk/metric v1.43.0/go.mod h1:C/RJtwSEJ5hzTiUz5pXF1kILHStzb9zFlIEe85bhj6A=
+go.opentelemetry.io/otel/trace v1.43.0 h1:BkNrHpup+4k4w+ZZ86CZoHHEkohws8AY+WTX09nk+3A=
+go.opentelemetry.io/otel/trace v1.43.0/go.mod h1:/QJhyVBUUswCphDVxq+8mld+AvhXZLhe+8WVFxiFff0=
+go.opentelemetry.io/proto/otlp v1.10.0 h1:IQRWgT5srOCYfiWnpqUYz9CVmbO8bFmKcwYxpuCSL2g=
+go.opentelemetry.io/proto/otlp v1.10.0/go.mod h1:/CV4QoCR/S9yaPj8utp3lvQPoqMtxXdzn7ozvvozVqk=
 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 v0.45.0 h1:jMBrvKuj23MTlT0bQEOBcAE0mjg8mK9RXFhRH6nyF3Q=
-golang.org/x/crypto v0.45.0/go.mod h1:XTGrrkGJve7CYK7J8PEww4aY7gM3qMCElcJQ8n8JdX4=
+go.yaml.in/yaml/v2 v2.4.4 h1:tuyd0P+2Ont/d6e2rl3be67goVK4R6deVxCUX5vyPaQ=
+go.yaml.in/yaml/v2 v2.4.4/go.mod h1:gMZqIpDtDqOfM0uNfy0SkpRhvUryYH0Z6wdMYcacYXQ=
+golang.org/x/crypto v0.52.0 h1:RMs7fP2rXdep0CftQlK8Uf+kibLm7qkCcradZWYz988=
+golang.org/x/crypto v0.52.0/go.mod h1:1QgfPxDqh0T2M/elOJtp9RvuR95kVjir0e6/BvEmGbc=
 golang.org/x/exp v0.0.0-20251113190631-e25ba8c21ef6 h1:zfMcR1Cs4KNuomFFgGefv5N0czO2XZpUbxGUy8i8ug0=
 golang.org/x/exp v0.0.0-20251113190631-e25ba8c21ef6/go.mod h1:46edojNIoXTNOhySWIWdix628clX9ODXwPsQuG6hsK0=
-golang.org/x/mod v0.30.0 h1:fDEXFVZ/fmCKProc/yAXXUijritrDzahmwwefnjoPFk=
-golang.org/x/mod v0.30.0/go.mod h1:lAsf5O2EvJeSFMiBxXDki7sCgAxEUcZHXoXMKT4GJKc=
-golang.org/x/net v0.47.0 h1:Mx+4dIFzqraBXUugkia1OOvlD6LemFo1ALMHjrXDOhY=
-golang.org/x/net v0.47.0/go.mod h1:/jNxtkgq5yWUGYkaZGqo27cfGZ1c5Nen03aYrrKpVRU=
-golang.org/x/sync v0.18.0 h1:kr88TuHDroi+UVf+0hZnirlk8o8T+4MrK6mr60WkH/I=
-golang.org/x/sync v0.18.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+golang.org/x/net v0.55.0 h1:bcvxaJn3e1U6InsFWt1JUq1aSjnRxLzT2rtD2KfkDF8=
+golang.org/x/net v0.55.0/go.mod h1:L5U2KuzuOe1lY7Z+aWVIKK6qEeJXnXV9yzGA+WCHJww=
 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 v0.38.0 h1:3yZWxaJjBmCWXqhN1qh02AkOnCQ1poK6oF+a7xWL6Gc=
-golang.org/x/sys v0.38.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
-golang.org/x/text v0.31.0 h1:aC8ghyu4JhP8VojJ2lEHBnochRno1sgL6nEi9WGFGMM=
-golang.org/x/text v0.31.0/go.mod h1:tKRAlv61yKIjGGHX/4tP1LTbc13YSec1pxVEWXzfoeM=
+golang.org/x/sys v0.45.0 h1:dO4czNzziLiiXplLQgBCEpCvXQ3dnkn0SdaZSYdQ+FY=
+golang.org/x/sys v0.45.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/term v0.43.0 h1:S4RLU2sB31O/NCl+zFN9Aru9A/Cq2aqKpTZJ6B+DwT4=
+golang.org/x/term v0.43.0/go.mod h1:lrhlHNdQJHO+1qVYiHfFKVuVioJIheAc3fBSMFYEIsk=
+golang.org/x/text v0.37.0 h1:Cqjiwd9eSg8e0QAkyCaQTNHFIIzWtidPahFWR83rTrc=
+golang.org/x/text v0.37.0/go.mod h1:a5sjxXGs9hsn/AJVwuElvCAo9v8QYLzvavO5z2PiM38=
 golang.org/x/time v0.12.0 h1:ScB/8o8olJvc+CQPWrK3fPZNfh7qgwCrY0zJmoEQLSE=
 golang.org/x/time v0.12.0/go.mod h1:CDIdPxbZBQxdj6cxyCIdrNogrJKMJ7pr37NYpMcMDSg=
-golang.org/x/tools v0.39.0 h1:ik4ho21kwuQln40uelmciQPp9SipgNDdrafrYA4TmQQ=
-golang.org/x/tools v0.39.0/go.mod h1:JnefbkDPyD8UU2kI5fuf8ZX4/yUeh9W877ZeBONxUqQ=
 golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 h1:B82qJJgjvYKsXS9jeunTOisW56dUokqW/FOteYJJ/yg=
 golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2/go.mod h1:deeaetjYA+DHMHg+sMSMI58GrEteJUUzzw7en6TJQcI=
 golang.zx2c4.com/wireguard v0.0.0-20250521234502-f333402bd9cb h1:whnFRlWMcXI9d+ZbWg+4sHnLp52d5yiIPUxMBSt4X9A=
 golang.zx2c4.com/wireguard v0.0.0-20250521234502-f333402bd9cb/go.mod h1:rpwXGsirqLqN2L0JDJQlwOboGHmptD5ZD6T2VmcqhTw=
 golang.zx2c4.com/wireguard/wgctrl v0.0.0-20241231184526-a9ab2273dd10 h1:3GDAcqdIg1ozBNLgPy4SLT84nfcBjr6rhGtXYtrkWLU=
 golang.zx2c4.com/wireguard/wgctrl v0.0.0-20241231184526-a9ab2273dd10/go.mod h1:T97yPqesLiNrOYxkwmhMI0ZIlJDm+p0PMR8eRVeR5tQ=
-golang.zx2c4.com/wireguard/windows v0.5.3 h1:On6j2Rpn3OEMXqBq00QEDC7bWSZrPIHKIus8eIuExIE=
-golang.zx2c4.com/wireguard/windows v0.5.3/go.mod h1:9TEe8TJmtwyQebdFwAkEWOPr3prrtqm+REGFifP60hI=
-gonum.org/v1/gonum v0.16.0 h1:5+ul4Swaf3ESvrOnidPp4GZbzf0mxVQpDCYUQE7OJfk=
-gonum.org/v1/gonum v0.16.0/go.mod h1:fef3am4MQ93R2HHpKnLk4/Tbh/s0+wqD5nfa6Pnwy4E=
-google.golang.org/genproto/googleapis/api v0.0.0-20250825161204-c5933d9347a5 h1:BIRfGDEjiHRrk0QKZe3Xv2ieMhtgRGeLcZQ0mIVn4EY=
-google.golang.org/genproto/googleapis/api v0.0.0-20250825161204-c5933d9347a5/go.mod h1:j3QtIyytwqGr1JUDtYXwtMXWPKsEa5LtzIFN1Wn5WvE=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20250825161204-c5933d9347a5 h1:eaY8u2EuxbRv7c3NiGK0/NedzVsCcV6hDuU5qPX5EGE=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20250825161204-c5933d9347a5/go.mod h1:M4/wBTSeyLxupu3W3tJtOgB14jILAS/XWPSSa3TAlJc=
-google.golang.org/grpc v1.76.0 h1:UnVkv1+uMLYXoIz6o7chp59WfQUYA2ex/BXQ9rHZu7A=
-google.golang.org/grpc v1.76.0/go.mod h1:Ju12QI8M6iQJtbcsV+awF5a4hfJMLi4X0JLo94ULZ6c=
-google.golang.org/protobuf v1.36.8 h1:xHScyCOEuuwZEc6UtSOvPbAT4zRh0xcNRYekJwfqyMc=
-google.golang.org/protobuf v1.36.8/go.mod h1:fuxRtAxBytpl4zzqUh6/eyUujkJdNiuEkXntxiD/uRU=
+golang.zx2c4.com/wireguard/windows v1.0.1 h1:eOxiDVbywPC+ZQqvdCK7x+ZwWXKbYv50TtH8ysFIbw8=
+golang.zx2c4.com/wireguard/windows v1.0.1/go.mod h1:+fbT3FFdX4zzYDLwJh5+HPEcNN/3HyNdzhNSVsQM+zs=
+gonum.org/v1/gonum v0.17.0 h1:VbpOemQlsSMrYmn7T2OUvQ4dqxQXU+ouZFQsZOx50z4=
+gonum.org/v1/gonum v0.17.0/go.mod h1:El3tOrEuMpv2UdMrbNlKEh9vd86bmQ6vqIcDwxEOc1E=
+google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9 h1:VPWxll4HlMw1Vs/qXtN7BvhZqsS9cdAittCNvVENElA=
+google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9/go.mod h1:7QBABkRtR8z+TEnmXTqIqwJLlzrZKVfAUm7tY3yGv0M=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9 h1:m8qni9SQFH0tJc1X0vmnpw/0t+AImlSvp30sEupozUg=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9/go.mod h1:4Hqkh8ycfw05ld/3BWL7rJOSfebL2Q+DVDeRgYgxUU8=
+google.golang.org/grpc v1.81.1 h1:VnnIIZ88UzOOKLukQi+ImGz8O1Wdp8nAGGnvOfEIWQQ=
+google.golang.org/grpc v1.81.1/go.mod h1:xGH9GfzOyMTGIOXBJmXt+BX/V0kcdQbdcuwQ/zNw42I=
+google.golang.org/protobuf v1.36.11 h1:fV6ZwhNocDyBLK0dj+fg8ektcVegBBuEolpbTQyBNVE=
+google.golang.org/protobuf v1.36.11/go.mod h1:HTf+CrKn2C3g5S8VImy6tdcUvCska2kB7j23XfzDpco=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c h1:Hei/4ADfdWqJk1ZMxUNpqntNwaWcugrBjAiHlqqRiVk=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c/go.mod h1:JHkPIbrfpd72SG/EVd6muEfDQjcINNoR0C8j2r3qZ4Q=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
 gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
-gotest.tools/v3 v3.4.0 h1:ZazjZUfuVeZGLAmlKKuyv3IKP5orXcwtOwDQH6YVr6o=
-gotest.tools/v3 v3.4.0/go.mod h1:CtbdzLSsqVhDgMtKsx03ird5YTGB3ar27v0u/yKBW5g=
+gotest.tools/v3 v3.5.2 h1:7koQfIKdy+I8UTetycgUqXWSDwpgv193Ka+qRsmBY8Q=
+gotest.tools/v3 v3.5.2/go.mod h1:LtdLGcnqToBH83WByAAi/wiwSFCArdFIUV/xxN4pcjA=
 gvisor.dev/gvisor v0.0.0-20250503011706-39ed1f5ac29c h1:m/r7OM+Y2Ty1sgBQ7Qb27VgIMBW8ZZhT4gLnUyDIhzI=
 gvisor.dev/gvisor v0.0.0-20250503011706-39ed1f5ac29c/go.mod h1:3r5CMtNQMKIvBlrmM9xWUNamjKBYPOWyXOjmg5Kts3g=
-software.sslmate.com/src/go-pkcs12 v0.6.0 h1:f3sQittAeF+pao32Vb+mkli+ZyT+VwKaD014qFGq6oU=
-software.sslmate.com/src/go-pkcs12 v0.6.0/go.mod h1:Qiz0EyvDRJjjxGyUQa2cCNZn/wMyzrRJ/qcDXOQazLI=
+pgregory.net/rapid v1.2.0 h1:keKAYRcjm+e1F0oAuU5F5+YPAWcyxNNRK2wud503Gnk=
+pgregory.net/rapid v1.2.0/go.mod h1:PY5XlDGj0+V1FCq0o192FdRhpKHGTRIWBgqjDBTrq04=
+software.sslmate.com/src/go-pkcs12 v0.7.1 h1:bxkUPRsvTPNRBZa4M/aSX4PyMOEbq3V8I6hbkG4F4Q8=
+software.sslmate.com/src/go-pkcs12 v0.7.1/go.mod h1:Qiz0EyvDRJjjxGyUQa2cCNZn/wMyzrRJ/qcDXOQazLI=

healthcheck/healthcheck.go

@@ -5,7 +5,9 @@ import (
 	"crypto/tls"
 	"encoding/json"
 	"fmt"
+	"net"
 	"net/http"
+	"strconv"
 	"strings"
 	"sync"
 	"time"
@@ -35,32 +37,38 @@ func (s Health) String() string {
 
 // Config holds the health check configuration for a target
 type Config struct {
-	ID                int               `json:"id"`
-	Enabled           bool              `json:"hcEnabled"`
-	Path              string            `json:"hcPath"`
-	Scheme            string            `json:"hcScheme"`
-	Mode              string            `json:"hcMode"`
-	Hostname          string            `json:"hcHostname"`
-	Port              int               `json:"hcPort"`
-	Interval          int               `json:"hcInterval"`          // in seconds
-	UnhealthyInterval int               `json:"hcUnhealthyInterval"` // in seconds
-	Timeout           int               `json:"hcTimeout"`           // in seconds
-	Headers           map[string]string `json:"hcHeaders"`
-	Method            string            `json:"hcMethod"`
-	Status            int               `json:"hcStatus"` // HTTP status code
-	TLSServerName     string            `json:"hcTlsServerName"`
+	ID                 int               `json:"id"`
+	Enabled            bool              `json:"hcEnabled"`
+	Path               string            `json:"hcPath"`
+	Scheme             string            `json:"hcScheme"`
+	Mode               string            `json:"hcMode"`
+	Hostname           string            `json:"hcHostname"`
+	Port               int               `json:"hcPort"`
+	Interval           int               `json:"hcInterval"`          // in seconds
+	UnhealthyInterval  int               `json:"hcUnhealthyInterval"` // in seconds
+	Timeout            int               `json:"hcTimeout"`           // in seconds
+	FollowRedirects    *bool             `json:"hcFollowRedirects"`
+	Headers            map[string]string `json:"hcHeaders"`
+	Method             string            `json:"hcMethod"`
+	Status             int               `json:"hcStatus"` // HTTP status code
+	TLSServerName      string            `json:"hcTlsServerName"`
+	HealthyThreshold   int               `json:"hcHealthyThreshold"`   // consecutive successes required to become healthy
+	UnhealthyThreshold int               `json:"hcUnhealthyThreshold"` // consecutive failures required to become unhealthy
 }
 
 // Target represents a health check target with its current status
 type Target struct {
-	Config     Config    `json:"config"`
-	Status     Health    `json:"status"`
-	LastCheck  time.Time `json:"lastCheck"`
-	LastError  string    `json:"lastError,omitempty"`
-	CheckCount int       `json:"checkCount"`
-	ticker     *time.Ticker
-	ctx        context.Context
-	cancel     context.CancelFunc
+	Config               Config    `json:"config"`
+	Status               Health    `json:"status"`
+	LastCheck            time.Time `json:"lastCheck"`
+	LastError            string    `json:"lastError,omitempty"`
+	CheckCount           int       `json:"checkCount"`
+	timer                *time.Timer
+	ctx                  context.Context
+	cancel               context.CancelFunc
+	client               *http.Client
+	consecutiveSuccesses int
+	consecutiveFailures  int
 }
 
 // StatusChangeCallback is called when any target's status changes
@@ -162,9 +170,16 @@ func (m *Monitor) addTargetUnsafe(config Config) error {
 	if config.Timeout == 0 {
 		config.Timeout = 5
 	}
+	if config.HealthyThreshold == 0 {
+		config.HealthyThreshold = 1
+	}
+	if config.UnhealthyThreshold == 0 {
+		config.UnhealthyThreshold = 1
+	}
 
-	logger.Debug("Target %d configuration: scheme=%s, method=%s, interval=%ds, timeout=%ds",
-		config.ID, config.Scheme, config.Method, config.Interval, config.Timeout)
+	logger.Debug("Target %d configuration: mode=%s, scheme=%s, method=%s, interval=%ds, timeout=%ds, healthyThreshold=%d, unhealthyThreshold=%d",
+		config.ID, config.Mode, config.Scheme, config.Method, config.Interval, config.Timeout,
+		config.HealthyThreshold, config.UnhealthyThreshold)
 
 	// Parse headers if provided as string
 	if len(config.Headers) == 0 && config.Path != "" {
@@ -185,6 +200,26 @@ func (m *Monitor) addTargetUnsafe(config Config) error {
 		Status: StatusUnknown,
 		ctx:    ctx,
 		cancel: cancel,
+		client: &http.Client{
+			CheckRedirect: func() func(*http.Request, []*http.Request) error {
+				// Default to following redirects if not explicitly configured
+				followRedirects := config.FollowRedirects == nil || *config.FollowRedirects
+				if !followRedirects {
+					return func(req *http.Request, via []*http.Request) error {
+						return http.ErrUseLastResponse
+					}
+				}
+				return nil
+			}(),
+			Transport: &http.Transport{
+				TLSClientConfig: &tls.Config{
+					// Configure TLS settings based on certificate enforcement
+					InsecureSkipVerify: !m.enforceCert,
+					// Use SNI TLS header if present
+					ServerName: config.TLSServerName,
+				},
+			},
+		},
 	}
 
 	m.targets[config.ID] = target
@@ -217,7 +252,7 @@ func (m *Monitor) RemoveTarget(id int) error {
 
 	// Notify callback of status change
 	if m.callback != nil {
-		go m.callback(m.GetTargets())
+		go m.callback(m.getAllTargetsUnsafe())
 	}
 
 	logger.Info("Successfully removed target %d", id)
@@ -250,7 +285,7 @@ func (m *Monitor) RemoveTargets(ids []int) error {
 
 	// Notify callback of status change if any targets were removed
 	if len(notFound) != len(ids) && m.callback != nil {
-		go m.callback(m.GetTargets())
+		go m.callback(m.getAllTargetsUnsafe())
 	}
 
 	if len(notFound) > 0 {
@@ -304,26 +339,26 @@ func (m *Monitor) monitorTarget(target *Target) {
 		go m.callback(m.GetTargets())
 	}
 
-	// Set up ticker based on current status
+	// Set up timer based on current status
 	interval := time.Duration(target.Config.Interval) * time.Second
 	if target.Status == StatusUnhealthy {
 		interval = time.Duration(target.Config.UnhealthyInterval) * time.Second
 	}
 
 	logger.Debug("Target %d: initial check interval set to %v", target.Config.ID, interval)
-	target.ticker = time.NewTicker(interval)
-	defer target.ticker.Stop()
+	target.timer = time.NewTimer(interval)
+	defer target.timer.Stop()
 
 	for {
 		select {
 		case <-target.ctx.Done():
 			logger.Info("Stopping health check monitoring for target %d", target.Config.ID)
 			return
-		case <-target.ticker.C:
+		case <-target.timer.C:
 			oldStatus := target.Status
 			m.performHealthCheck(target)
 
-			// Update ticker interval if status changed
+			// Update timer interval if status changed
 			newInterval := time.Duration(target.Config.Interval) * time.Second
 			if target.Status == StatusUnhealthy {
 				newInterval = time.Duration(target.Config.UnhealthyInterval) * time.Second
@@ -332,11 +367,12 @@ func (m *Monitor) monitorTarget(target *Target) {
 			if newInterval != interval {
 				logger.Debug("Target %d: updating check interval from %v to %v due to status change",
 					target.Config.ID, interval, newInterval)
-				target.ticker.Stop()
-				target.ticker = time.NewTicker(newInterval)
 				interval = newInterval
 			}
 
+			// Reset timer for next check with current interval
+			target.timer.Reset(interval)
+
 			// Notify callback if status changed
 			if oldStatus != target.Status && m.callback != nil {
 				logger.Info("Target %d status changed: %s -> %s",
@@ -347,17 +383,75 @@ func (m *Monitor) monitorTarget(target *Target) {
 	}
 }
 
-// performHealthCheck performs a health check on a target
+// performHealthCheck performs a health check on a target and applies threshold logic
 func (m *Monitor) performHealthCheck(target *Target) {
 	target.CheckCount++
 	target.LastCheck = time.Now()
-	target.LastError = ""
 
-	// Build URL
-	url := fmt.Sprintf("%s://%s", target.Config.Scheme, target.Config.Hostname)
-	if target.Config.Port > 0 {
-		url = fmt.Sprintf("%s:%d", url, target.Config.Port)
+	var passed bool
+	var checkErr string
+
+	switch strings.ToLower(target.Config.Mode) {
+	case "tcp":
+		passed, checkErr = m.performTCPCheck(target)
+	default:
+		// "http", "https", or anything else falls through to HTTP
+		passed, checkErr = m.performHTTPCheck(target)
 	}
+
+	if passed {
+		target.consecutiveFailures = 0
+		target.consecutiveSuccesses++
+
+		logger.Debug("Target %d: check passed (consecutive successes: %d / threshold: %d)",
+			target.Config.ID, target.consecutiveSuccesses, target.Config.HealthyThreshold)
+
+		if target.consecutiveSuccesses >= target.Config.HealthyThreshold {
+			target.Status = StatusHealthy
+			target.LastError = ""
+		}
+	} else {
+		target.consecutiveSuccesses = 0
+		target.consecutiveFailures++
+		target.LastError = checkErr
+
+		logger.Debug("Target %d: check failed (consecutive failures: %d / threshold: %d): %s",
+			target.Config.ID, target.consecutiveFailures, target.Config.UnhealthyThreshold, checkErr)
+
+		if target.consecutiveFailures >= target.Config.UnhealthyThreshold {
+			target.Status = StatusUnhealthy
+		}
+	}
+}
+
+// performTCPCheck dials the target's host:port over TCP and returns whether it succeeded
+func (m *Monitor) performTCPCheck(target *Target) (bool, string) {
+	address := net.JoinHostPort(target.Config.Hostname, strconv.Itoa(target.Config.Port))
+	timeout := time.Duration(target.Config.Timeout) * time.Second
+
+	logger.Debug("Target %d: performing TCP health check to %s (timeout: %v)",
+		target.Config.ID, address, timeout)
+
+	conn, err := net.DialTimeout("tcp", address, timeout)
+	if err != nil {
+		msg := fmt.Sprintf("TCP dial failed: %v", err)
+		logger.Warn("Target %d: %s", target.Config.ID, msg)
+		return false, msg
+	}
+	conn.Close()
+
+	logger.Debug("Target %d: TCP health check passed", target.Config.ID)
+	return true, ""
+}
+
+// performHTTPCheck performs an HTTP/HTTPS health check and returns whether it succeeded
+func (m *Monitor) performHTTPCheck(target *Target) (bool, string) {
+	// Build URL (use net.JoinHostPort to properly handle IPv6 addresses with ports)
+	host := target.Config.Hostname
+	if target.Config.Port > 0 {
+		host = net.JoinHostPort(target.Config.Hostname, strconv.Itoa(target.Config.Port))
+	}
+	url := fmt.Sprintf("%s://%s", target.Config.Scheme, host)
 	if target.Config.Path != "" {
 		if !strings.HasPrefix(target.Config.Path, "/") {
 			url += "/"
@@ -365,7 +459,7 @@ func (m *Monitor) performHealthCheck(target *Target) {
 		url += target.Config.Path
 	}
 
-	logger.Debug("Target %d: performing health check %d to %s",
+	logger.Debug("Target %d: performing HTTP health check %d to %s",
 		target.Config.ID, target.CheckCount, url)
 
 	if target.Config.Scheme == "https" {
@@ -373,27 +467,15 @@ func (m *Monitor) performHealthCheck(target *Target) {
 			target.Config.ID, m.enforceCert)
 	}
 
-	// Create request
+	// Create request with timeout context
 	ctx, cancel := context.WithTimeout(context.Background(), time.Duration(target.Config.Timeout)*time.Second)
 	defer cancel()
 
-	client := &http.Client{
-		Transport: &http.Transport{
-			TLSClientConfig: &tls.Config{
-				// Configure TLS settings based on certificate enforcement
-				InsecureSkipVerify: !m.enforceCert,
-				// Use SNI TLS header if present
-				ServerName: target.Config.TLSServerName,
-			},
-		},
-	}
-
 	req, err := http.NewRequestWithContext(ctx, target.Config.Method, url, nil)
 	if err != nil {
-		target.Status = StatusUnhealthy
-		target.LastError = fmt.Sprintf("failed to create request: %v", err)
-		logger.Warn("Target %d: failed to create request: %v", target.Config.ID, err)
-		return
+		msg := fmt.Sprintf("failed to create request: %v", err)
+		logger.Warn("Target %d: %s", target.Config.ID, msg)
+		return false, msg
 	}
 
 	// Add headers
@@ -407,45 +489,36 @@ func (m *Monitor) performHealthCheck(target *Target) {
 	}
 
 	// Perform request
-	resp, err := client.Do(req)
+	resp, err := target.client.Do(req)
 	if err != nil {
-		target.Status = StatusUnhealthy
-		target.LastError = fmt.Sprintf("request failed: %v", err)
+		msg := fmt.Sprintf("request failed: %v", err)
 		logger.Warn("Target %d: health check failed: %v", target.Config.ID, err)
-		return
+		return false, msg
 	}
 	defer resp.Body.Close()
 
 	// Check response status
-	var expectedStatus int
 	if target.Config.Status > 0 {
-		expectedStatus = target.Config.Status
-	} else {
-		expectedStatus = 0 // Use range check for 200-299
+		// Check for specific status code
+		logger.Debug("Target %d: checking status against expected code %d", target.Config.ID, target.Config.Status)
+		if resp.StatusCode == target.Config.Status {
+			logger.Debug("Target %d: health check passed (status: %d)", target.Config.ID, resp.StatusCode)
+			return true, ""
+		}
+		msg := fmt.Sprintf("unexpected status code: %d (expected: %d)", resp.StatusCode, target.Config.Status)
+		logger.Warn("Target %d: %s", target.Config.ID, msg)
+		return false, msg
 	}
 
-	if expectedStatus > 0 {
-		logger.Debug("Target %d: checking health status against expected code %d", target.Config.ID, expectedStatus)
-		// Check for specific status code
-		if resp.StatusCode == expectedStatus {
-			target.Status = StatusHealthy
-			logger.Debug("Target %d: health check passed (status: %d, expected: %d)", target.Config.ID, resp.StatusCode, expectedStatus)
-		} else {
-			target.Status = StatusUnhealthy
-			target.LastError = fmt.Sprintf("unexpected status code: %d (expected: %d)", resp.StatusCode, expectedStatus)
-			logger.Warn("Target %d: health check failed with status code %d (expected: %d)", target.Config.ID, resp.StatusCode, expectedStatus)
-		}
-	} else {
-		// Check for 2xx range
-		if resp.StatusCode >= 200 && resp.StatusCode < 300 {
-			target.Status = StatusHealthy
-			logger.Debug("Target %d: health check passed (status: %d)", target.Config.ID, resp.StatusCode)
-		} else {
-			target.Status = StatusUnhealthy
-			target.LastError = fmt.Sprintf("unhealthy status code: %d", resp.StatusCode)
-			logger.Warn("Target %d: health check failed with status code %d", target.Config.ID, resp.StatusCode)
-		}
+	// Default: check for 2xx range
+	if resp.StatusCode >= 200 && resp.StatusCode < 300 {
+		logger.Debug("Target %d: health check passed (status: %d)", target.Config.ID, resp.StatusCode)
+		return true, ""
 	}
+
+	msg := fmt.Sprintf("unhealthy status code: %d", resp.StatusCode)
+	logger.Warn("Target %d: health check failed with status code %d", target.Config.ID, resp.StatusCode)
+	return false, msg
 }
 
 // Stop stops monitoring all targets
@@ -512,7 +585,7 @@ func (m *Monitor) DisableTarget(id int) error {
 
 		// Notify callback of status change
 		if m.callback != nil {
-			go m.callback(m.GetTargets())
+			go m.callback(m.getAllTargetsUnsafe())
 		}
 	} else {
 		logger.Debug("Target %d is already disabled", id)
@@ -520,3 +593,82 @@ func (m *Monitor) DisableTarget(id int) error {
 
 	return nil
 }
+
+// GetTargetIDs returns a slice of all current target IDs
+func (m *Monitor) GetTargetIDs() []int {
+	m.mutex.RLock()
+	defer m.mutex.RUnlock()
+
+	ids := make([]int, 0, len(m.targets))
+	for id := range m.targets {
+		ids = append(ids, id)
+	}
+	return ids
+}
+
+// SyncTargets synchronizes the current targets to match the desired set.
+// It removes targets not in the desired set and adds targets that are missing.
+func (m *Monitor) SyncTargets(desiredConfigs []Config) error {
+	m.mutex.Lock()
+	defer m.mutex.Unlock()
+
+	logger.Info("Syncing health check targets: %d desired targets", len(desiredConfigs))
+
+	// Build a set of desired target IDs
+	desiredIDs := make(map[int]Config)
+	for _, config := range desiredConfigs {
+		desiredIDs[config.ID] = config
+	}
+
+	// Find targets to remove (exist but not in desired set)
+	var toRemove []int
+	for id := range m.targets {
+		if _, exists := desiredIDs[id]; !exists {
+			toRemove = append(toRemove, id)
+		}
+	}
+
+	// Remove targets that are not in the desired set
+	for _, id := range toRemove {
+		logger.Info("Sync: removing health check target %d", id)
+		if target, exists := m.targets[id]; exists {
+			target.cancel()
+			delete(m.targets, id)
+		}
+	}
+
+	// Add or update targets from the desired set
+	var addedCount, updatedCount int
+	for id, config := range desiredIDs {
+		if existing, exists := m.targets[id]; exists {
+			// Target exists - check if config changed and update if needed
+			// For now, we'll replace it to ensure config is up to date
+			logger.Debug("Sync: updating health check target %d", id)
+			existing.cancel()
+			delete(m.targets, id)
+			if err := m.addTargetUnsafe(config); err != nil {
+				logger.Error("Sync: failed to update target %d: %v", id, err)
+				return fmt.Errorf("failed to update target %d: %v", id, err)
+			}
+			updatedCount++
+		} else {
+			// Target doesn't exist - add it
+			logger.Debug("Sync: adding health check target %d", id)
+			if err := m.addTargetUnsafe(config); err != nil {
+				logger.Error("Sync: failed to add target %d: %v", id, err)
+				return fmt.Errorf("failed to add target %d: %v", id, err)
+			}
+			addedCount++
+		}
+	}
+
+	logger.Info("Sync complete: removed %d, added %d, updated %d targets",
+		len(toRemove), addedCount, updatedCount)
+
+	// Notify callback if any changes were made
+	if (len(toRemove) > 0 || addedCount > 0 || updatedCount > 0) && m.callback != nil {
+		go m.callback(m.getAllTargetsUnsafe())
+	}
+
+	return nil
+}

holepunch/holepunch.go

@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"net"
+	"strconv"
 	"sync"
 	"time"
 
@@ -19,37 +20,49 @@ import (
 // ExitNode represents a WireGuard exit node for hole punching
 type ExitNode struct {
 	Endpoint  string `json:"endpoint"`
+	RelayPort uint16 `json:"relayPort"`
 	PublicKey string `json:"publicKey"`
+	SiteIds   []int  `json:"siteIds,omitempty"`
 }
 
 // Manager handles UDP hole punching operations
 type Manager struct {
-	mu         sync.Mutex
-	running    bool
-	stopChan   chan struct{}
-	sharedBind *bind.SharedBind
-	ID         string
-	token      string
-	publicKey  string
-	clientType string
-	exitNodes  map[string]ExitNode // key is endpoint
-	updateChan chan struct{}       // signals the goroutine to refresh exit nodes
+	mu          sync.Mutex
+	running     bool
+	stopChan    chan struct{}
+	sharedBind  *bind.SharedBind
+	ID          string
+	token       string
+	publicKey   string
+	clientType  string
+	exitNodes   map[string]ExitNode // key is endpoint
+	updateChan  chan struct{}        // signals the goroutine to refresh exit nodes
+	publicDNS []string
 
-	sendHolepunchInterval time.Duration
+	sendHolepunchInterval    time.Duration
+	sendHolepunchIntervalMin time.Duration
+	sendHolepunchIntervalMax time.Duration
+	defaultIntervalMin       time.Duration
+	defaultIntervalMax       time.Duration
 }
 
-const sendHolepunchIntervalMax = 60 * time.Second
-const sendHolepunchIntervalMin = 1 * time.Second
+const defaultSendHolepunchIntervalMax = 60 * time.Second
+const defaultSendHolepunchIntervalMin = 1 * time.Second
 
 // NewManager creates a new hole punch manager
-func NewManager(sharedBind *bind.SharedBind, ID string, clientType string, publicKey string) *Manager {
+func NewManager(sharedBind *bind.SharedBind, ID string, clientType string, publicKey string, publicDNS []string) *Manager {
 	return &Manager{
-		sharedBind:            sharedBind,
-		ID:                    ID,
-		clientType:            clientType,
-		publicKey:             publicKey,
-		exitNodes:             make(map[string]ExitNode),
-		sendHolepunchInterval: sendHolepunchIntervalMin,
+		sharedBind:               sharedBind,
+		ID:                       ID,
+		clientType:               clientType,
+		publicKey:                publicKey,
+		publicDNS:              publicDNS,
+		exitNodes:                make(map[string]ExitNode),
+		sendHolepunchInterval:    defaultSendHolepunchIntervalMin,
+		sendHolepunchIntervalMin: defaultSendHolepunchIntervalMin,
+		sendHolepunchIntervalMax: defaultSendHolepunchIntervalMax,
+		defaultIntervalMin:       defaultSendHolepunchIntervalMin,
+		defaultIntervalMax:       defaultSendHolepunchIntervalMax,
 	}
 }
 
@@ -140,6 +153,51 @@ func (m *Manager) RemoveExitNode(endpoint string) bool {
 	return true
 }
 
+/*
+RemoveExitNodesByPeer removes the peer ID from the SiteIds list in each exit node.
+If the SiteIds list becomes empty after removal, the exit node is removed entirely.
+Returns the number of exit nodes removed.
+*/
+func (m *Manager) RemoveExitNodesByPeer(peerID int) int {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	removed := 0
+	for endpoint, node := range m.exitNodes {
+		// Remove peerID from SiteIds if present
+		newSiteIds := make([]int, 0, len(node.SiteIds))
+		for _, id := range node.SiteIds {
+			if id != peerID {
+				newSiteIds = append(newSiteIds, id)
+			}
+		}
+		if len(newSiteIds) != len(node.SiteIds) {
+			node.SiteIds = newSiteIds
+			if len(node.SiteIds) == 0 {
+				delete(m.exitNodes, endpoint)
+				logger.Info("Removed exit node %s as no more site IDs remain after removing peer %d", endpoint, peerID)
+				removed++
+			} else {
+				m.exitNodes[endpoint] = node
+				logger.Info("Removed peer %d from exit node %s site IDs", peerID, endpoint)
+			}
+		}
+	}
+
+	if removed > 0 {
+		// Signal the goroutine to refresh if running
+		if m.running && m.updateChan != nil {
+			select {
+			case m.updateChan <- struct{}{}:
+			default:
+				// Channel full or closed, skip
+			}
+		}
+	}
+
+	return removed
+}
+
 // GetExitNodes returns a copy of the current exit nodes
 func (m *Manager) GetExitNodes() []ExitNode {
 	m.mu.Lock()
@@ -152,17 +210,46 @@ func (m *Manager) GetExitNodes() []ExitNode {
 	return nodes
 }
 
-// ResetInterval resets the hole punch interval back to the minimum value,
-// allowing it to climb back up through exponential backoff.
-// This is useful when network conditions change or connectivity is restored.
-func (m *Manager) ResetInterval() {
+// SetServerHolepunchInterval sets custom min and max intervals for hole punching.
+// This is useful for low power mode where longer intervals are desired.
+func (m *Manager) SetServerHolepunchInterval(min, max time.Duration) {
 	m.mu.Lock()
 	defer m.mu.Unlock()
 
-	if m.sendHolepunchInterval != sendHolepunchIntervalMin {
-		m.sendHolepunchInterval = sendHolepunchIntervalMin
-		logger.Info("Reset hole punch interval to minimum (%v)", sendHolepunchIntervalMin)
+	m.sendHolepunchIntervalMin = min
+	m.sendHolepunchIntervalMax = max
+	m.sendHolepunchInterval = min
+
+	logger.Info("Set hole punch intervals: min=%v, max=%v", min, max)
+
+	// Signal the goroutine to apply the new interval if running
+	if m.running && m.updateChan != nil {
+		select {
+		case m.updateChan <- struct{}{}:
+		default:
+			// Channel full or closed, skip
+		}
 	}
+}
+
+// GetInterval returns the current min and max intervals
+func (m *Manager) GetServerHolepunchInterval() (min, max time.Duration) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	return m.sendHolepunchIntervalMin, m.sendHolepunchIntervalMax
+}
+
+// ResetServerHolepunchInterval resets the hole punch interval back to the default values.
+// This restores normal operation after low power mode or other custom settings.
+func (m *Manager) ResetServerHolepunchInterval() {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	m.sendHolepunchIntervalMin = m.defaultIntervalMin
+	m.sendHolepunchIntervalMax = m.defaultIntervalMax
+	m.sendHolepunchInterval = m.defaultIntervalMin
+
+	logger.Info("Reset hole punch intervals to defaults: min=%v, max=%v", m.defaultIntervalMin, m.defaultIntervalMax)
 
 	// Signal the goroutine to apply the new interval if running
 	if m.running && m.updateChan != nil {
@@ -196,13 +283,19 @@ func (m *Manager) TriggerHolePunch() error {
 	// Send hole punch to all exit nodes
 	successCount := 0
 	for _, exitNode := range currentExitNodes {
-		host, err := util.ResolveDomain(exitNode.Endpoint)
+		var host string
+		var err error
+		if len(m.publicDNS) > 0 {
+			host, err = util.ResolveDomainUpstream(exitNode.Endpoint, m.publicDNS)
+		} else {
+			host, err = util.ResolveDomain(exitNode.Endpoint)
+		}
 		if err != nil {
 			logger.Warn("Failed to resolve endpoint %s: %v", exitNode.Endpoint, err)
 			continue
 		}
 
-		serverAddr := net.JoinHostPort(host, "21820")
+		serverAddr := net.JoinHostPort(host, strconv.Itoa(int(exitNode.RelayPort)))
 		remoteAddr, err := net.ResolveUDPAddr("udp", serverAddr)
 		if err != nil {
 			logger.Error("Failed to resolve UDP address %s: %v", serverAddr, err)
@@ -247,7 +340,7 @@ func (m *Manager) StartMultipleExitNodes(exitNodes []ExitNode) error {
 	m.updateChan = make(chan struct{}, 1)
 	m.mu.Unlock()
 
-	logger.Info("Starting UDP hole punch to %d exit nodes with shared bind", len(exitNodes))
+	logger.Debug("Starting UDP hole punch to %d exit nodes with shared bind", len(exitNodes))
 
 	go m.runMultipleExitNodes()
 
@@ -307,13 +400,19 @@ func (m *Manager) runMultipleExitNodes() {
 
 		var resolvedNodes []resolvedExitNode
 		for _, exitNode := range currentExitNodes {
-			host, err := util.ResolveDomain(exitNode.Endpoint)
+			var host string
+			var err error
+			if len(m.publicDNS) > 0 {
+				host, err = util.ResolveDomainUpstream(exitNode.Endpoint, m.publicDNS)
+			} else {
+				host, err = util.ResolveDomain(exitNode.Endpoint)
+			}
 			if err != nil {
 				logger.Warn("Failed to resolve endpoint %s: %v", exitNode.Endpoint, err)
 				continue
 			}
 
-			serverAddr := net.JoinHostPort(host, "21820")
+			serverAddr := net.JoinHostPort(host, strconv.Itoa(int(exitNode.RelayPort)))
 			remoteAddr, err := net.ResolveUDPAddr("udp", serverAddr)
 			if err != nil {
 				logger.Error("Failed to resolve UDP address %s: %v", serverAddr, err)
@@ -325,7 +424,7 @@ func (m *Manager) runMultipleExitNodes() {
 				publicKey:    exitNode.PublicKey,
 				endpointName: exitNode.Endpoint,
 			})
-			logger.Info("Resolved exit node: %s -> %s", exitNode.Endpoint, remoteAddr.String())
+			logger.Debug("Resolved exit node: %s -> %s", exitNode.Endpoint, remoteAddr.String())
 		}
 		return resolvedNodes
 	}
@@ -345,7 +444,7 @@ func (m *Manager) runMultipleExitNodes() {
 
 	// Start with minimum interval
 	m.mu.Lock()
-	m.sendHolepunchInterval = sendHolepunchIntervalMin
+	m.sendHolepunchInterval = m.sendHolepunchIntervalMin
 	m.mu.Unlock()
 
 	ticker := time.NewTicker(m.sendHolepunchInterval)
@@ -367,7 +466,7 @@ func (m *Manager) runMultipleExitNodes() {
 			}
 			// Reset interval to minimum on update
 			m.mu.Lock()
-			m.sendHolepunchInterval = sendHolepunchIntervalMin
+			m.sendHolepunchInterval = m.sendHolepunchIntervalMin
 			m.mu.Unlock()
 			ticker.Reset(m.sendHolepunchInterval)
 			// Send immediate hole punch to newly resolved nodes
@@ -387,8 +486,8 @@ func (m *Manager) runMultipleExitNodes() {
 				// Exponential backoff: double the interval up to max
 				m.mu.Lock()
 				newInterval := m.sendHolepunchInterval * 2
-				if newInterval > sendHolepunchIntervalMax {
-					newInterval = sendHolepunchIntervalMax
+				if newInterval > m.sendHolepunchIntervalMax {
+					newInterval = m.sendHolepunchIntervalMax
 				}
 				if newInterval != m.sendHolepunchInterval {
 					m.sendHolepunchInterval = newInterval

holepunch/tester.go

@@ -41,18 +41,30 @@ func DefaultTestOptions() TestConnectionOptions {
 	}
 }
 
+// cachedAddr holds a cached resolved UDP address
+type cachedAddr struct {
+	addr      *net.UDPAddr
+	resolvedAt time.Time
+}
+
 // HolepunchTester monitors holepunch connectivity using magic packets
 type HolepunchTester struct {
-	sharedBind *bind.SharedBind
-	mu         sync.RWMutex
-	running    bool
-	stopChan   chan struct{}
+	sharedBind  *bind.SharedBind
+	publicDNS []string
+	mu          sync.RWMutex
+	running     bool
+	stopChan    chan struct{}
 
 	// Pending requests waiting for responses (key: echo data as string)
 	pendingRequests sync.Map // map[string]*pendingRequest
 
 	// Callback when connection status changes
 	callback HolepunchStatusCallback
+
+	// Address cache to avoid repeated DNS/UDP resolution
+	addrCache    map[string]*cachedAddr
+	addrCacheMu  sync.RWMutex
+	addrCacheTTL time.Duration // How long cached addresses are valid
 }
 
 // HolepunchStatus represents the status of a holepunch connection
@@ -73,9 +85,12 @@ type pendingRequest struct {
 }
 
 // NewHolepunchTester creates a new holepunch tester using the given SharedBind
-func NewHolepunchTester(sharedBind *bind.SharedBind) *HolepunchTester {
+func NewHolepunchTester(sharedBind *bind.SharedBind, publicDNS []string) *HolepunchTester {
 	return &HolepunchTester{
-		sharedBind: sharedBind,
+		sharedBind:   sharedBind,
+		publicDNS:  publicDNS,
+		addrCache:    make(map[string]*cachedAddr),
+		addrCacheTTL: 5 * time.Minute, // Cache addresses for 5 minutes
 	}
 }
 
@@ -135,12 +150,76 @@ func (t *HolepunchTester) Stop() {
 		return true
 	})
 
+	// Clear address cache
+	t.addrCacheMu.Lock()
+	t.addrCache = make(map[string]*cachedAddr)
+	t.addrCacheMu.Unlock()
+
 	logger.Debug("HolepunchTester stopped")
 }
 
+// resolveEndpoint resolves an endpoint to a UDP address, using cache when possible
+func (t *HolepunchTester) resolveEndpoint(endpoint string) (*net.UDPAddr, error) {
+	// Check cache first
+	t.addrCacheMu.RLock()
+	cached, ok := t.addrCache[endpoint]
+	ttl := t.addrCacheTTL
+	t.addrCacheMu.RUnlock()
+
+	if ok && time.Since(cached.resolvedAt) < ttl {
+		return cached.addr, nil
+	}
+
+	// Resolve the endpoint
+	var host string
+	var err error
+	if len(t.publicDNS) > 0 {
+		host, err = util.ResolveDomainUpstream(endpoint, t.publicDNS)
+	} else {
+		host, err = util.ResolveDomain(endpoint)
+	}
+	if err != nil {
+		host = endpoint
+	}
+
+	_, _, err = net.SplitHostPort(host)
+	if err != nil {
+		host = net.JoinHostPort(host, "21820")
+	}
+
+	remoteAddr, err := net.ResolveUDPAddr("udp", host)
+	if err != nil {
+		return nil, fmt.Errorf("failed to resolve UDP address %s: %w", host, err)
+	}
+
+	// Cache the result
+	t.addrCacheMu.Lock()
+	t.addrCache[endpoint] = &cachedAddr{
+		addr:       remoteAddr,
+		resolvedAt: time.Now(),
+	}
+	t.addrCacheMu.Unlock()
+
+	return remoteAddr, nil
+}
+
+// InvalidateCache removes a specific endpoint from the address cache
+func (t *HolepunchTester) InvalidateCache(endpoint string) {
+	t.addrCacheMu.Lock()
+	delete(t.addrCache, endpoint)
+	t.addrCacheMu.Unlock()
+}
+
+// ClearCache clears all cached addresses
+func (t *HolepunchTester) ClearCache() {
+	t.addrCacheMu.Lock()
+	t.addrCache = make(map[string]*cachedAddr)
+	t.addrCacheMu.Unlock()
+}
+
 // handleResponse is called by SharedBind when a magic response is received
 func (t *HolepunchTester) handleResponse(addr netip.AddrPort, echoData []byte) {
-	logger.Debug("Received magic response from %s", addr.String())
+	// logger.Debug("Received magic response from %s", addr.String())
 	key := string(echoData)
 
 	value, ok := t.pendingRequests.LoadAndDelete(key)
@@ -152,7 +231,7 @@ func (t *HolepunchTester) handleResponse(addr netip.AddrPort, echoData []byte) {
 
 	req := value.(*pendingRequest)
 	rtt := time.Since(req.sentAt)
-	logger.Debug("Magic response matched pending request for %s (RTT: %v)", req.endpoint, rtt)
+	// logger.Debug("Magic response matched pending request for %s (RTT: %v)", req.endpoint, rtt)
 
 	// Send RTT to the waiting goroutine (non-blocking)
 	select {
@@ -183,20 +262,10 @@ func (t *HolepunchTester) TestEndpoint(endpoint string, timeout time.Duration) T
 		return result
 	}
 
-	// Resolve the endpoint
-	host, err := util.ResolveDomain(endpoint)
+	// Resolve the endpoint (using cache)
+	remoteAddr, err := t.resolveEndpoint(endpoint)
 	if err != nil {
-		host = endpoint
-	}
-
-	_, _, err = net.SplitHostPort(host)
-	if err != nil {
-		host = net.JoinHostPort(host, "21820")
-	}
-
-	remoteAddr, err := net.ResolveUDPAddr("udp", host)
-	if err != nil {
-		result.Error = fmt.Errorf("failed to resolve UDP address %s: %w", host, err)
+		result.Error = err
 		return result
 	}
 

nativessh/auth.go

@@ -0,0 +1,158 @@
+package nativessh
+
+import (
+	"bufio"
+	"fmt"
+	"log"
+	"net"
+	"os"
+	"os/user"
+	"path/filepath"
+	"strings"
+
+	"golang.org/x/crypto/ssh"
+)
+
+type staticConnMeta struct {
+	user string
+}
+
+func (m staticConnMeta) User() string              { return m.user }
+func (m staticConnMeta) SessionID() []byte         { return nil }
+func (m staticConnMeta) ClientVersion() []byte     { return nil }
+func (m staticConnMeta) ServerVersion() []byte     { return nil }
+func (m staticConnMeta) RemoteAddr() net.Addr      { return nil }
+func (m staticConnMeta) LocalAddr() net.Addr       { return nil }
+
+// CheckAuthorizedKeys reports whether key matches any entry in the system
+// user's ~/.ssh/authorized_keys file.  Returns false (not an error) when the
+// user or file does not exist.
+func CheckAuthorizedKeys(username string, key ssh.PublicKey) bool {
+	u, err := user.Lookup(username)
+	if err != nil {
+		return false
+	}
+	f, err := os.Open(filepath.Join(u.HomeDir, ".ssh", "authorized_keys"))
+	if err != nil {
+		return false
+	}
+	defer f.Close()
+
+	want := ssh.FingerprintSHA256(key)
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := strings.TrimSpace(scanner.Text())
+		if line == "" || strings.HasPrefix(line, "#") {
+			continue
+		}
+		parsed, _, _, _, err := ssh.ParseAuthorizedKey([]byte(line))
+		if err != nil {
+			continue
+		}
+		if ssh.FingerprintSHA256(parsed) == want {
+			return true
+		}
+	}
+	return false
+}
+
+// SystemUserExists reports whether a user account with the given name exists
+// on the host OS.
+func SystemUserExists(username string) bool {
+	_, err := user.Lookup(username)
+	return err == nil
+}
+
+// Authenticate authenticates a user for a browser-based native SSH session.
+// It tries, in order:
+//  1. Private key — parses privateKeyPEM and checks it against the user's
+//     ~/.ssh/authorized_keys.
+//  2. Password — verifies password via the host OS PAM stack (Linux only).
+//
+// Returns nil on the first method that succeeds, or an error if all fail.
+func Authenticate(username, password, privateKeyPEM string) error {
+	return AuthenticateWithCertificate(nil, username, password, privateKeyPEM, "")
+}
+
+// AuthenticateWithCertificate authenticates a user for a browser-based native
+// SSH session using the same method ordering as the native SSH server:
+//  1. Private key in host ~/.ssh/authorized_keys.
+//  2. SSH certificate signed by the configured CA (when provided).
+//  3. Password via host PAM stack.
+func AuthenticateWithCertificate(store *CredentialStore, username, password, privateKeyPEM, certificate string) error {
+	log.Printf("nativessh: authenticating user %q (hasPassword=%v, hasPrivateKey=%v)", username, password != "", privateKeyPEM != "")
+	if !SystemUserExists(username) {
+		log.Printf("nativessh: user %q not found on system", username)
+		return fmt.Errorf("user %q does not exist", username)
+	}
+
+	var signer ssh.Signer
+	if privateKeyPEM != "" {
+		parsedSigner, err := ssh.ParsePrivateKey([]byte(privateKeyPEM))
+		if err != nil {
+			log.Printf("nativessh: failed to parse private key for %q: %v", username, err)
+		} else if CheckAuthorizedKeys(username, parsedSigner.PublicKey()) {
+			log.Printf("nativessh: private key auth succeeded for %q", username)
+			return nil
+		} else {
+			signer = parsedSigner
+			log.Printf("nativessh: private key not in authorized_keys for %q", username)
+		}
+	}
+
+	if store != nil && certificate != "" {
+		if signer == nil {
+			log.Printf("nativessh: certificate provided for %q but no matching private key was provided", username)
+		} else {
+			pub, _, _, _, err := ssh.ParseAuthorizedKey([]byte(certificate))
+			if err != nil {
+				log.Printf("nativessh: failed to parse certificate for %q: %v", username, err)
+			} else {
+				cert, ok := pub.(*ssh.Certificate)
+				if !ok {
+					log.Printf("nativessh: provided cert data for %q is not an SSH certificate", username)
+				} else if ssh.FingerprintSHA256(cert.Key) != ssh.FingerprintSHA256(signer.PublicKey()) {
+					log.Printf("nativessh: certificate key mismatch for %q", username)
+				} else {
+					caKey, userPrincipals := store.get(username)
+					if caKey == nil {
+						log.Printf("nativessh: CA key is not set for certificate auth user %q", username)
+					} else if len(userPrincipals) == 0 {
+						log.Printf("nativessh: no allowed principals found for certificate auth user %q", username)
+					} else {
+						checker := &ssh.CertChecker{
+							IsUserAuthority: func(auth ssh.PublicKey) bool {
+								return ssh.FingerprintSHA256(auth) == ssh.FingerprintSHA256(caKey)
+							},
+						}
+
+						var lastErr error
+						for principal := range userPrincipals {
+							_, authErr := checker.Authenticate(staticConnMeta{user: principal}, cert)
+							if authErr == nil {
+								log.Printf("nativessh: certificate auth succeeded for %q (principal=%q)", username, principal)
+								return nil
+							}
+							lastErr = authErr
+						}
+						if lastErr != nil {
+							log.Printf("nativessh: certificate auth failed for %q: %v", username, lastErr)
+						}
+					}
+				}
+			}
+		}
+	}
+
+	if password != "" {
+		if err := VerifySystemPassword(username, password); err != nil {
+			log.Printf("nativessh: password auth failed for %q: %v", username, err)
+		} else {
+			log.Printf("nativessh: password auth succeeded for %q", username)
+			return nil
+		}
+	} else {
+		log.Printf("nativessh: no password provided for %q", username)
+	}
+	return fmt.Errorf("authentication failed for user %q", username)
+}

nativessh/pam_linux.go

@@ -0,0 +1,99 @@
+//go:build linux
+
+package nativessh
+
+import (
+	"bufio"
+	"bytes"
+	"errors"
+	"fmt"
+	"log"
+	"os"
+	"strings"
+
+	"github.com/go-crypt/crypt"
+	"github.com/go-crypt/x/yescrypt"
+)
+
+// VerifySystemPassword authenticates username/password by reading /etc/shadow.
+// Supports yescrypt ($y$), bcrypt ($2b$/$2a$/$2y$), SHA-512 ($6$), SHA-256
+// ($5$), argon2, scrypt, and other schemes handled by go-crypt/crypt.
+func VerifySystemPassword(username, password string) error {
+	hash, err := readShadowHash(username)
+	if err != nil {
+		log.Printf("nativessh/pam: readShadowHash for %q failed: %v", username, err)
+		return fmt.Errorf("shadow: %w", err)
+	}
+
+	// Log the scheme prefix only (never the full hash).
+	scheme := "unknown"
+	for _, prefix := range []string{"$y$", "$2a$", "$2b$", "$2y$", "$6$", "$5$", "$1$"} {
+		if strings.HasPrefix(hash, prefix) {
+			scheme = prefix
+			break
+		}
+	}
+	log.Printf("nativessh/pam: verifying password for %q using scheme %s", username, scheme)
+
+	// Yescrypt ($y$) is not in go-crypt/crypt's default decoder; handle it directly.
+	if strings.HasPrefix(hash, "$y$") {
+		computed, err := yescrypt.Hash([]byte(password), []byte(hash))
+		if err != nil {
+			log.Printf("nativessh/pam: yescrypt.Hash for %q failed: %v", username, err)
+			return fmt.Errorf("yescrypt: %w", err)
+		}
+		if !bytes.Equal(computed, []byte(hash)) {
+			log.Printf("nativessh/pam: yescrypt mismatch for %q", username)
+			return errors.New("authentication failed")
+		}
+		return nil
+	}
+
+	decoder, err := crypt.NewDefaultDecoder()
+	if err != nil {
+		return fmt.Errorf("crypt decoder: %w", err)
+	}
+
+	digest, err := decoder.Decode(hash)
+	if err != nil {
+		log.Printf("nativessh/pam: failed to decode hash for %q: %v", username, err)
+		return fmt.Errorf("unsupported password hash scheme %q: %w", scheme, err)
+	}
+
+	match, err := digest.MatchAdvanced(password)
+	if err != nil {
+		log.Printf("nativessh/pam: MatchAdvanced for %q failed: %v", username, err)
+		return err
+	}
+	if !match {
+		log.Printf("nativessh/pam: password mismatch for %q", username)
+		return errors.New("authentication failed")
+	}
+	return nil
+}
+
+// readShadowHash reads /etc/shadow and returns the password hash for username.
+func readShadowHash(username string) (string, error) {
+	f, err := os.Open("/etc/shadow")
+	if err != nil {
+		return "", err
+	}
+	defer f.Close()
+
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		fields := strings.SplitN(scanner.Text(), ":", 3)
+		if len(fields) < 2 || fields[0] != username {
+			continue
+		}
+		h := fields[1]
+		if h == "" || h == "*" || strings.HasPrefix(h, "!") || h == "x" {
+			return "", errors.New("account locked or has no password")
+		}
+		return h, nil
+	}
+	if err := scanner.Err(); err != nil {
+		return "", err
+	}
+	return "", errors.New("user not found in shadow database")
+}

nativessh/pam_other.go

@@ -0,0 +1,11 @@
+//go:build !linux
+
+package nativessh
+
+import "errors"
+
+// VerifySystemPassword is not supported on non-Linux platforms; it always
+// returns an error so that password authentication is never accepted.
+func VerifySystemPassword(username, password string) error {
+	return errors.New("password authentication not supported on this platform")
+}

nativessh/pty.go

@@ -0,0 +1,130 @@
+//go:build !windows
+
+package nativessh
+
+import (
+	"bufio"
+	"errors"
+	"fmt"
+	"os"
+	"os/exec"
+	"os/user"
+	"strings"
+	"sync"
+
+	"github.com/creack/pty"
+)
+
+// PTYSession is a running shell process attached to a PTY.
+// It implements io.ReadWriteCloser so it can be bridged to any transport.
+type PTYSession struct {
+	ptmx     *os.File
+	cmd      *exec.Cmd
+	waitOnce sync.Once
+	exitCode int
+}
+
+// findShell returns the path to the best available interactive shell by
+// checking preferred shells in order, falling back to /bin/sh.
+func findShell() string {
+	preferred := []string{"zsh", "bash", "fish", "ksh", "sh"}
+	for _, name := range preferred {
+		if path, err := exec.LookPath(name); err == nil {
+			return path
+		}
+	}
+	return "/bin/sh"
+}
+
+// userShell returns the login shell configured for u in /etc/passwd.
+// If the field is empty or the binary does not exist, it falls back to
+// findShell so there is always a usable shell.
+func userShell(u *user.User) string {
+	if shell := passwdShell(u.Username); shell != "" {
+		if _, err := exec.LookPath(shell); err == nil {
+			return shell
+		}
+	}
+	return findShell()
+}
+
+// passwdShell reads /etc/passwd and returns the login shell for the named user.
+// Returns "" if the user is not found or the file cannot be read.
+func passwdShell(username string) string {
+	f, err := os.Open("/etc/passwd")
+	if err != nil {
+		return ""
+	}
+	defer f.Close()
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if line == "" || line[0] == '#' {
+			continue
+		}
+		// Fields: username:password:uid:gid:gecos:home:shell
+		fields := strings.SplitN(line, ":", 7)
+		if len(fields) == 7 && fields[0] == username {
+			return fields[6]
+		}
+	}
+	_ = scanner.Err()
+	return ""
+}
+
+// NewPTYSession spawns the best available shell in a PTY.
+func NewPTYSession() (*PTYSession, error) {
+	shell := findShell()
+	cmd := exec.Command(shell)
+	cmd.Env = append(os.Environ(), "TERM=xterm-256color")
+	ptmx, err := pty.Start(cmd)
+	if err != nil {
+		return nil, fmt.Errorf("pty start: %w", err)
+	}
+	return &PTYSession{ptmx: ptmx, cmd: cmd}, nil
+}
+
+// Read reads output from the PTY.
+func (p *PTYSession) Read(b []byte) (int, error) {
+	return p.ptmx.Read(b)
+}
+
+// Write writes input to the PTY.
+func (p *PTYSession) Write(b []byte) (int, error) {
+	return p.ptmx.Write(b)
+}
+
+// Resize changes the PTY window size.
+func (p *PTYSession) Resize(cols, rows uint16) error {
+	return pty.Setsize(p.ptmx, &pty.Winsize{Cols: cols, Rows: rows})
+}
+
+// wait waits for the child process to exit exactly once and records its exit
+// code. Safe to call concurrently or multiple times.
+func (p *PTYSession) wait() {
+	p.waitOnce.Do(func() {
+		err := p.cmd.Wait()
+		if err != nil {
+			var exitErr *exec.ExitError
+			if errors.As(err, &exitErr) {
+				p.exitCode = exitErr.ExitCode()
+				return
+			}
+			p.exitCode = 1
+		}
+	})
+}
+
+// ExitCode waits for the shell process to exit and returns its exit code.
+// It is safe to call before or after Close.
+func (p *PTYSession) ExitCode() int {
+	p.wait()
+	return p.exitCode
+}
+
+// Close closes the PTY and waits for the child process to exit.
+func (p *PTYSession) Close() error {
+	err := p.ptmx.Close()
+	p.wait()
+	return err
+}

nativessh/pty_unix.go

@@ -0,0 +1,78 @@
+//go:build !windows
+
+package nativessh
+
+import (
+	"fmt"
+	"os"
+	"os/exec"
+	"os/user"
+	"strconv"
+	"syscall"
+
+	"github.com/creack/pty"
+)
+
+// NewPTYSessionAs spawns an interactive shell in a PTY running as the given
+// system user.  The calling process must have sufficient privileges (typically
+// root / CAP_SETUID) to switch to a different UID/GID.
+func NewPTYSessionAs(username string) (*PTYSession, error) {
+	u, err := user.Lookup(username)
+	if err != nil {
+		return nil, fmt.Errorf("user lookup %q: %w", username, err)
+	}
+	uid, err := strconv.ParseUint(u.Uid, 10, 32)
+	if err != nil {
+		return nil, fmt.Errorf("parse uid: %w", err)
+	}
+	gid, err := strconv.ParseUint(u.Gid, 10, 32)
+	if err != nil {
+		return nil, fmt.Errorf("parse gid: %w", err)
+	}
+
+	// Collect supplementary group IDs.
+	groupIDs, err := u.GroupIds()
+	if err != nil {
+		groupIDs = []string{}
+	}
+	var groups []uint32
+	for _, g := range groupIDs {
+		gval, err := strconv.ParseUint(g, 10, 32)
+		if err == nil {
+			groups = append(groups, uint32(gval))
+		}
+	}
+
+	shell := userShell(u)
+
+	// Prefer the user's home directory as the working directory, but fall back
+	// to / if it does not exist (e.g. useradd was run without -m).
+	homeDir := u.HomeDir
+	if _, err := os.Stat(homeDir); err != nil {
+		homeDir = "/"
+	}
+
+	cmd := exec.Command(shell, "--login")
+	cmd.Env = []string{
+		"TERM=xterm-256color",
+		"HOME=" + u.HomeDir,
+		"USER=" + username,
+		"LOGNAME=" + username,
+		"SHELL=" + shell,
+		"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
+	}
+	cmd.Dir = homeDir
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		Credential: &syscall.Credential{
+			Uid:    uint32(uid),
+			Gid:    uint32(gid),
+			Groups: groups,
+		},
+	}
+
+	ptmx, err := pty.Start(cmd)
+	if err != nil {
+		return nil, fmt.Errorf("pty start: %w", err)
+	}
+	return &PTYSession{ptmx: ptmx, cmd: cmd}, nil
+}

nativessh/server.go

@@ -0,0 +1,363 @@
+//go:build !windows
+
+package nativessh
+
+import (
+	"crypto/ed25519"
+	"crypto/rand"
+	"fmt"
+	"io"
+	"log"
+	"net"
+	"strings"
+	"sync"
+
+	"golang.org/x/crypto/ssh"
+)
+
+// CredentialStore holds in-memory SSH credentials that can be updated at runtime.
+// It is safe for concurrent use.
+type CredentialStore struct {
+	mu         sync.RWMutex
+	caKey      ssh.PublicKey
+	principals map[string]map[string]struct{} // username -> set of allowed principals
+}
+
+// connMetaWithUser wraps ConnMetadata while overriding User() for cert checks.
+type connMetaWithUser struct {
+	ssh.ConnMetadata
+	user string
+}
+
+func (m connMetaWithUser) User() string { return m.user }
+
+// NewCredentialStore returns an empty, ready-to-use CredentialStore.
+func NewCredentialStore() *CredentialStore {
+	return &CredentialStore{
+		principals: make(map[string]map[string]struct{}),
+	}
+}
+
+// SetCAKey parses and stores the CA public key from authorized_keys-format data.
+func (s *CredentialStore) SetCAKey(authorizedKeyData string) error {
+	key, _, _, _, err := ssh.ParseAuthorizedKey([]byte(authorizedKeyData))
+	if err != nil {
+		return fmt.Errorf("parse CA key: %w", err)
+	}
+	s.mu.Lock()
+	s.caKey = key
+	s.mu.Unlock()
+	return nil
+}
+
+// AddPrincipals records username and niceId as allowed principals for username.
+// Both values are stored; either can appear in the certificate's ValidPrincipals
+// field to satisfy the standard cert-auth principal check.
+func (s *CredentialStore) AddPrincipals(username, niceId string) {
+	username = strings.TrimSpace(username)
+	niceId = strings.TrimSpace(niceId)
+	if username == "" {
+		return
+	}
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	if s.principals[username] == nil {
+		s.principals[username] = make(map[string]struct{})
+	}
+	s.principals[username][username] = struct{}{}
+	if niceId != "" {
+		s.principals[username][niceId] = struct{}{}
+	}
+}
+
+// get returns the CA key and the principal set for username under a read lock.
+func (s *CredentialStore) get(username string) (ssh.PublicKey, map[string]struct{}) {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+	return s.caKey, s.principals[username]
+}
+
+// ServerConfig holds configuration for the native SSH server.
+type ServerConfig struct {
+	// ListenAddr is the TCP address to listen on. Defaults to ":2222".
+	ListenAddr string
+	// Credentials provides in-memory CA key and per-user principals.
+	// Updates to the store are reflected immediately for new connections.
+	// If nil or the store has no CA key set, all connections are rejected.
+	Credentials *CredentialStore
+}
+
+// Server is a simple SSH server that authenticates clients via SSH certificate
+// auth only. Certificates must be signed by the configured CA and the
+// connecting username must appear in both the certificate's principal list and
+// the local principals file.
+type Server struct {
+	cfg ServerConfig
+}
+
+// NewServer creates a new Server. The ListenAddr defaults to ":2222" when empty.
+func NewServer(cfg ServerConfig) *Server {
+	if cfg.ListenAddr == "" {
+		cfg.ListenAddr = ":2222"
+	}
+	return &Server{cfg: cfg}
+}
+
+// buildSSHConfig builds the ssh.ServerConfig with multi-method authentication:
+//  1. Public key: host ~/.ssh/authorized_keys, then CA certificate.
+//  2. Password: system PAM stack (Linux only).
+func (s *Server) buildSSHConfig() (*ssh.ServerConfig, error) {
+	hostSigner, err := generateHostKey()
+	if err != nil {
+		return nil, fmt.Errorf("host key: %w", err)
+	}
+	cfg := &ssh.ServerConfig{
+		PublicKeyCallback: makePublicKeyCallback(s.cfg.Credentials),
+		PasswordCallback:  makePasswordCallback(),
+	}
+	cfg.AddHostKey(hostSigner)
+	return cfg, nil
+}
+
+// Serve accepts connections on ln and handles them. It returns when ln is
+// closed or a non-temporary Accept error occurs.
+func (s *Server) Serve(ln net.Listener) error {
+	sshCfg, err := s.buildSSHConfig()
+	if err != nil {
+		return err
+	}
+	log.Printf("nativessh: server listening on %s", ln.Addr())
+	for {
+		conn, err := ln.Accept()
+		if err != nil {
+			return err
+		}
+		go s.handleConn(conn, sshCfg)
+	}
+}
+
+// ListenAndServe starts the SSH server on the host network and blocks until
+// the listener is closed.
+func (s *Server) ListenAndServe() error {
+	ln, err := net.Listen("tcp", s.cfg.ListenAddr)
+	if err != nil {
+		return fmt.Errorf("listen %s: %w", s.cfg.ListenAddr, err)
+	}
+	defer ln.Close()
+	return s.Serve(ln)
+}
+
+func (s *Server) handleConn(conn net.Conn, cfg *ssh.ServerConfig) {
+	defer conn.Close()
+	sshConn, chans, reqs, err := ssh.NewServerConn(conn, cfg)
+	if err != nil {
+		log.Printf("nativessh: handshake failed from %s: %v", conn.RemoteAddr(), err)
+		return
+	}
+	defer sshConn.Close()
+	log.Printf("nativessh: connection from %s user=%s", conn.RemoteAddr(), sshConn.User())
+
+	go ssh.DiscardRequests(reqs)
+
+	for newChan := range chans {
+		if newChan.ChannelType() != "session" {
+			_ = newChan.Reject(ssh.UnknownChannelType, "unknown channel type")
+			continue
+		}
+		ch, requests, err := newChan.Accept()
+		if err != nil {
+			log.Printf("nativessh: channel accept error: %v", err)
+			return
+		}
+		go s.handleSession(ch, requests, sshConn.User())
+	}
+}
+
+// handleSession drives a single SSH session channel. It waits for a pty-req
+// followed by a shell request and then bridges the PTY to the channel.
+func (s *Server) handleSession(ch ssh.Channel, requests <-chan *ssh.Request, username string) {
+	defer ch.Close()
+
+	var (
+		sess    *PTYSession
+		started bool
+	)
+
+	for req := range requests {
+		switch req.Type {
+		case "pty-req":
+			var err error
+			if sess == nil {
+				sess, err = NewPTYSessionAs(username)
+				if err != nil {
+					log.Printf("nativessh: PTY start error: %v", err)
+					if req.WantReply {
+						_ = req.Reply(false, nil)
+					}
+					return
+				}
+			}
+			cols, rows := parsePTYReq(req.Payload)
+			_ = sess.Resize(cols, rows)
+			if req.WantReply {
+				_ = req.Reply(true, nil)
+			}
+
+		case "shell":
+			if req.WantReply {
+				_ = req.Reply(true, nil)
+			}
+			if started || sess == nil {
+				continue
+			}
+			started = true
+			// PTY output → SSH channel.
+			go func() {
+				_, _ = io.Copy(ch, sess)
+				// Notify the client of the shell's exit status so it can
+				// disconnect cleanly instead of requiring a manual disconnect.
+				exitCode := sess.ExitCode()
+				exitStatusPayload := ssh.Marshal(struct{ Status uint32 }{uint32(exitCode)})
+				_, _ = ch.SendRequest("exit-status", false, exitStatusPayload)
+				_ = ch.CloseWrite()
+				sess.Close() //nolint:errcheck
+				// Close the channel so the ssh library closes the requests
+				// channel, which unblocks the for-range loop in handleSession
+				// and allows the deferred ch.Close() to run. Without this,
+				// handleSession blocks forever waiting for requests to drain.
+				_ = ch.Close()
+			}()
+			// SSH channel input → PTY stdin.
+			go func() {
+				_, _ = io.Copy(sess, ch)
+			}()
+
+		case "window-change":
+			if sess != nil {
+				cols, rows := parseWindowChange(req.Payload)
+				_ = sess.Resize(cols, rows)
+			}
+			if req.WantReply {
+				_ = req.Reply(true, nil)
+			}
+
+		default:
+			if req.WantReply {
+				_ = req.Reply(false, nil)
+			}
+		}
+	}
+
+	if sess != nil && !started {
+		sess.Close() //nolint:errcheck
+	}
+}
+
+// makePublicKeyCallback returns a PublicKeyCallback that tries, in order:
+//  1. Host authorized_keys  – matches any key in the OS user's
+//     ~/.ssh/authorized_keys file.
+//  2. CA certificate        – validates an SSH certificate signed by the
+//     configured CA and checks that the user appears in the principals map.
+//
+// store may be nil or empty; those paths are simply skipped.
+func makePublicKeyCallback(store *CredentialStore) func(ssh.ConnMetadata, ssh.PublicKey) (*ssh.Permissions, error) {
+	return func(meta ssh.ConnMetadata, key ssh.PublicKey) (*ssh.Permissions, error) {
+		// 1. Host authorized_keys.
+		if CheckAuthorizedKeys(meta.User(), key) {
+			log.Printf("nativessh: authorized_keys auth for user %q", meta.User())
+			return &ssh.Permissions{}, nil
+		}
+
+		// 2. CA certificate.
+		if store != nil {
+			caKey, userPrincipals := store.get(meta.User())
+			if caKey != nil {
+				checker := &ssh.CertChecker{
+					IsUserAuthority: func(auth ssh.PublicKey) bool {
+						return ssh.FingerprintSHA256(auth) == ssh.FingerprintSHA256(caKey)
+					},
+				}
+
+				if len(userPrincipals) == 0 {
+					return nil, fmt.Errorf("user %q not in allowed principals list", meta.User())
+				}
+
+				var lastErr error
+				for principal := range userPrincipals {
+					perms, err := checker.Authenticate(connMetaWithUser{ConnMetadata: meta, user: principal}, key)
+					if err == nil {
+						log.Printf("nativessh: CA cert auth for user %q principal=%q", meta.User(), principal)
+						return perms, nil
+					}
+					lastErr = err
+				}
+
+				if lastErr != nil {
+					log.Printf("nativessh: CA cert rejected for user %q: %v", meta.User(), lastErr)
+				}
+			}
+		}
+
+		return nil, fmt.Errorf("public key not authorized for user %q", meta.User())
+	}
+}
+
+// makePasswordCallback returns a PasswordCallback that validates the supplied
+// password via the host OS PAM stack.  On non-Linux platforms this always
+// fails (see pam_other.go).
+func makePasswordCallback() func(ssh.ConnMetadata, []byte) (*ssh.Permissions, error) {
+	return func(meta ssh.ConnMetadata, password []byte) (*ssh.Permissions, error) {
+		if err := VerifySystemPassword(meta.User(), string(password)); err != nil {
+			// Return a generic message to the client; log the real reason.
+			log.Printf("nativessh: password auth failed for user %q: %v", meta.User(), err)
+			return nil, fmt.Errorf("permission denied")
+		}
+		log.Printf("nativessh: password auth for user %q", meta.User())
+		return &ssh.Permissions{}, nil
+	}
+}
+
+// generateHostKey generates a fresh ephemeral Ed25519 host key in memory.
+// A new key is created on every server start; nothing is written to disk.
+func generateHostKey() (ssh.Signer, error) {
+	_, priv, err := ed25519.GenerateKey(rand.Reader)
+	if err != nil {
+		return nil, fmt.Errorf("generate host key: %w", err)
+	}
+	log.Printf("nativessh: generated ephemeral Ed25519 host key")
+	return ssh.NewSignerFromKey(priv)
+}
+
+// ptyRequestMsg mirrors the SSH wire format for pty-req (RFC 4254 §6.2).
+type ptyRequestMsg struct {
+	Term     string
+	Columns  uint32
+	Rows     uint32
+	Width    uint32
+	Height   uint32
+	Modelist string
+}
+
+func parsePTYReq(payload []byte) (cols, rows uint16) {
+	var req ptyRequestMsg
+	if err := ssh.Unmarshal(payload, &req); err != nil {
+		return 80, 24
+	}
+	return uint16(req.Columns), uint16(req.Rows)
+}
+
+// windowChangeMsg mirrors the SSH wire format for window-change (RFC 4254 §6.7).
+type windowChangeMsg struct {
+	Columns uint32
+	Rows    uint32
+	Width   uint32
+	Height  uint32
+}
+
+func parseWindowChange(payload []byte) (cols, rows uint16) {
+	var msg windowChangeMsg
+	if err := ssh.Unmarshal(payload, &msg); err != nil {
+		return 80, 24
+	}
+	return uint16(msg.Columns), uint16(msg.Rows)
+}

nativessh/server_windows.go

@@ -0,0 +1,64 @@
+//go:build windows
+
+package nativessh
+
+import (
+	"errors"
+	"log"
+	"net"
+	"sync"
+
+	"golang.org/x/crypto/ssh"
+)
+
+// CredentialStore is a stub on Windows. Native SSH is not supported on Windows.
+type CredentialStore struct {
+	mu         sync.RWMutex
+	principals map[string]map[string]struct{}
+}
+
+// NewCredentialStore returns an empty CredentialStore stub.
+// Native SSH is not supported on Windows; a warning is logged.
+func NewCredentialStore() *CredentialStore {
+	log.Println("WARNING: native SSH is not supported on Windows and will be disabled")
+	return &CredentialStore{
+		principals: make(map[string]map[string]struct{}),
+	}
+}
+
+// SetCAKey is a no-op stub on Windows.
+func (s *CredentialStore) SetCAKey(_ string) error {
+	return errors.New("native SSH not supported on Windows")
+}
+
+// AddPrincipals is a no-op stub on Windows.
+func (s *CredentialStore) AddPrincipals(_, _ string) {}
+
+// get returns nil CA key and empty principals on Windows.
+func (s *CredentialStore) get(_ string) (ssh.PublicKey, map[string]struct{}) {
+	return nil, nil
+}
+
+// ServerConfig holds configuration for the native SSH server (stub on Windows).
+type ServerConfig struct {
+	ListenAddr  string
+	Credentials *CredentialStore
+}
+
+// Server is a stub on Windows.
+type Server struct{}
+
+// NewServer returns a stub Server and logs a warning.
+func NewServer(cfg ServerConfig) *Server {
+	return &Server{}
+}
+
+// ListenAndServe always returns an error on Windows.
+func (s *Server) ListenAndServe() error {
+	return errors.New("native SSH not supported on Windows")
+}
+
+// Serve always returns an error on Windows.
+func (s *Server) Serve(_ net.Listener) error {
+	return errors.New("native SSH not supported on Windows")
+}

netstack2/access_log.go

@@ -0,0 +1,514 @@
+package netstack2
+
+import (
+	"bytes"
+	"compress/zlib"
+	"crypto/rand"
+	"encoding/base64"
+	"encoding/hex"
+	"encoding/json"
+	"net"
+	"sort"
+	"sync"
+	"time"
+
+	"github.com/fosrl/newt/logger"
+)
+
+const (
+	// flushInterval is how often the access logger flushes completed sessions to the server
+	flushInterval = 60 * time.Second
+
+	// maxBufferedSessions is the max number of completed sessions to buffer before forcing a flush
+	maxBufferedSessions = 100
+
+	// sessionGapThreshold is the maximum gap between the end of one connection
+	// and the start of the next for them to be considered part of the same session.
+	// If the gap exceeds this, a new consolidated session is created.
+	sessionGapThreshold = 5 * time.Second
+
+	// minConnectionsToConsolidate is the minimum number of connections in a group
+	// before we bother consolidating. Groups smaller than this are sent as-is.
+	minConnectionsToConsolidate = 2
+)
+
+// SendFunc is a callback that sends compressed access log data to the server.
+// The data is a base64-encoded zlib-compressed JSON array of AccessSession objects.
+type SendFunc func(data string) error
+
+// AccessSession represents a tracked access session through the proxy
+type AccessSession struct {
+	SessionID       string    `json:"sessionId"`
+	ResourceID      int       `json:"resourceId"`
+	SourceAddr      string    `json:"sourceAddr"`
+	DestAddr        string    `json:"destAddr"`
+	Protocol        string    `json:"protocol"`
+	StartedAt       time.Time `json:"startedAt"`
+	EndedAt         time.Time `json:"endedAt,omitempty"`
+	BytesTx         int64     `json:"bytesTx"`
+	BytesRx         int64     `json:"bytesRx"`
+	ConnectionCount int       `json:"connectionCount,omitempty"` // number of raw connections merged into this session (0 or 1 = single)
+}
+
+// udpSessionKey identifies a unique UDP "session" by src -> dst
+type udpSessionKey struct {
+	srcAddr  string
+	dstAddr  string
+	protocol string
+}
+
+// consolidationKey groups connections that may be part of the same logical session.
+// Source port is intentionally excluded so that many ephemeral-port connections
+// from the same source IP to the same destination are grouped together.
+type consolidationKey struct {
+	sourceIP   string // IP only, no port
+	destAddr   string // full host:port of the destination
+	protocol   string
+	resourceID int
+}
+
+// AccessLogger tracks access sessions for resources and periodically
+// flushes completed sessions to the server via a configurable SendFunc.
+type AccessLogger struct {
+	mu                sync.Mutex
+	sessions          map[string]*AccessSession       // active sessions: sessionID -> session
+	udpSessions       map[udpSessionKey]*AccessSession // active UDP sessions for dedup
+	completedSessions []*AccessSession                 // completed sessions waiting to be flushed
+	udpTimeout        time.Duration
+	sendFn            SendFunc
+	stopCh            chan struct{}
+	flushDone         chan struct{} // closed after the flush goroutine exits
+}
+
+// NewAccessLogger creates a new access logger.
+// udpTimeout controls how long a UDP session is kept alive without traffic before being ended.
+func NewAccessLogger(udpTimeout time.Duration) *AccessLogger {
+	al := &AccessLogger{
+		sessions:          make(map[string]*AccessSession),
+		udpSessions:       make(map[udpSessionKey]*AccessSession),
+		completedSessions: make([]*AccessSession, 0),
+		udpTimeout:        udpTimeout,
+		stopCh:            make(chan struct{}),
+		flushDone:         make(chan struct{}),
+	}
+	go al.backgroundLoop()
+	return al
+}
+
+// SetSendFunc sets the callback used to send compressed access log batches
+// to the server. This can be called after construction once the websocket
+// client is available.
+func (al *AccessLogger) SetSendFunc(fn SendFunc) {
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	al.sendFn = fn
+}
+
+// generateSessionID creates a random session identifier
+func generateSessionID() string {
+	b := make([]byte, 8)
+	rand.Read(b)
+	return hex.EncodeToString(b)
+}
+
+// StartTCPSession logs the start of a TCP session and returns a session ID.
+func (al *AccessLogger) StartTCPSession(resourceID int, srcAddr, dstAddr string) string {
+	sessionID := generateSessionID()
+	now := time.Now()
+
+	session := &AccessSession{
+		SessionID:  sessionID,
+		ResourceID: resourceID,
+		SourceAddr: srcAddr,
+		DestAddr:   dstAddr,
+		Protocol:   "tcp",
+		StartedAt:  now,
+	}
+
+	al.mu.Lock()
+	al.sessions[sessionID] = session
+	al.mu.Unlock()
+
+	logger.Info("ACCESS START session=%s resource=%d proto=tcp src=%s dst=%s time=%s",
+		sessionID, resourceID, srcAddr, dstAddr, now.Format(time.RFC3339))
+
+	return sessionID
+}
+
+// EndTCPSession logs the end of a TCP session and queues it for sending.
+func (al *AccessLogger) EndTCPSession(sessionID string) {
+	now := time.Now()
+
+	al.mu.Lock()
+	session, ok := al.sessions[sessionID]
+	if ok {
+		session.EndedAt = now
+		delete(al.sessions, sessionID)
+		al.completedSessions = append(al.completedSessions, session)
+	}
+	shouldFlush := len(al.completedSessions) >= maxBufferedSessions
+	al.mu.Unlock()
+
+	if ok {
+		duration := now.Sub(session.StartedAt)
+		logger.Info("ACCESS END session=%s resource=%d proto=tcp src=%s dst=%s started=%s ended=%s duration=%s",
+			sessionID, session.ResourceID, session.SourceAddr, session.DestAddr,
+			session.StartedAt.Format(time.RFC3339), now.Format(time.RFC3339), duration)
+	}
+
+	if shouldFlush {
+		al.flush()
+	}
+}
+
+// TrackUDPSession starts or returns an existing UDP session. Returns the session ID.
+func (al *AccessLogger) TrackUDPSession(resourceID int, srcAddr, dstAddr string) string {
+	key := udpSessionKey{
+		srcAddr:  srcAddr,
+		dstAddr:  dstAddr,
+		protocol: "udp",
+	}
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+
+	if existing, ok := al.udpSessions[key]; ok {
+		return existing.SessionID
+	}
+
+	sessionID := generateSessionID()
+	now := time.Now()
+
+	session := &AccessSession{
+		SessionID:  sessionID,
+		ResourceID: resourceID,
+		SourceAddr: srcAddr,
+		DestAddr:   dstAddr,
+		Protocol:   "udp",
+		StartedAt:  now,
+	}
+
+	al.sessions[sessionID] = session
+	al.udpSessions[key] = session
+
+	logger.Info("ACCESS START session=%s resource=%d proto=udp src=%s dst=%s time=%s",
+		sessionID, resourceID, srcAddr, dstAddr, now.Format(time.RFC3339))
+
+	return sessionID
+}
+
+// EndUDPSession ends a UDP session and queues it for sending.
+func (al *AccessLogger) EndUDPSession(sessionID string) {
+	now := time.Now()
+
+	al.mu.Lock()
+	session, ok := al.sessions[sessionID]
+	if ok {
+		session.EndedAt = now
+		delete(al.sessions, sessionID)
+		key := udpSessionKey{
+			srcAddr:  session.SourceAddr,
+			dstAddr:  session.DestAddr,
+			protocol: "udp",
+		}
+		delete(al.udpSessions, key)
+		al.completedSessions = append(al.completedSessions, session)
+	}
+	shouldFlush := len(al.completedSessions) >= maxBufferedSessions
+	al.mu.Unlock()
+
+	if ok {
+		duration := now.Sub(session.StartedAt)
+		logger.Info("ACCESS END session=%s resource=%d proto=udp src=%s dst=%s started=%s ended=%s duration=%s",
+			sessionID, session.ResourceID, session.SourceAddr, session.DestAddr,
+			session.StartedAt.Format(time.RFC3339), now.Format(time.RFC3339), duration)
+	}
+
+	if shouldFlush {
+		al.flush()
+	}
+}
+
+// backgroundLoop handles periodic flushing and stale session reaping.
+func (al *AccessLogger) backgroundLoop() {
+	defer close(al.flushDone)
+
+	flushTicker := time.NewTicker(flushInterval)
+	defer flushTicker.Stop()
+
+	reapTicker := time.NewTicker(30 * time.Second)
+	defer reapTicker.Stop()
+
+	for {
+		select {
+		case <-al.stopCh:
+			return
+		case <-flushTicker.C:
+			al.flush()
+		case <-reapTicker.C:
+			al.reapStaleSessions()
+		}
+	}
+}
+
+// reapStaleSessions cleans up UDP sessions that were not properly ended.
+func (al *AccessLogger) reapStaleSessions() {
+	al.mu.Lock()
+	defer al.mu.Unlock()
+
+	staleThreshold := time.Now().Add(-5 * time.Minute)
+
+	for key, session := range al.udpSessions {
+		if session.StartedAt.Before(staleThreshold) && session.EndedAt.IsZero() {
+			now := time.Now()
+			session.EndedAt = now
+			duration := now.Sub(session.StartedAt)
+			logger.Info("ACCESS END (reaped) session=%s resource=%d proto=udp src=%s dst=%s started=%s ended=%s duration=%s",
+				session.SessionID, session.ResourceID, session.SourceAddr, session.DestAddr,
+				session.StartedAt.Format(time.RFC3339), now.Format(time.RFC3339), duration)
+			al.completedSessions = append(al.completedSessions, session)
+			delete(al.sessions, session.SessionID)
+			delete(al.udpSessions, key)
+		}
+	}
+}
+
+// extractIP strips the port from an address string and returns just the IP.
+// If the address has no port component it is returned as-is.
+func extractIP(addr string) string {
+	host, _, err := net.SplitHostPort(addr)
+	if err != nil {
+		// Might already be a bare IP
+		return addr
+	}
+	return host
+}
+
+// consolidateSessions takes a slice of completed sessions and merges bursts of
+// short-lived connections from the same source IP to the same destination into
+// single higher-level session entries.
+//
+// The algorithm:
+//  1. Group sessions by (sourceIP, destAddr, protocol, resourceID).
+//  2. Within each group, sort by StartedAt.
+//  3. Walk through the sorted list and merge consecutive sessions whose gap
+//     (previous EndedAt → next StartedAt) is ≤ sessionGapThreshold.
+//  4. For merged sessions the earliest StartedAt and latest EndedAt are kept,
+//     bytes are summed, and ConnectionCount records how many raw connections
+//     were folded in. If the merged connections used more than one source port,
+//     SourceAddr is set to just the IP (port omitted).
+//  5. Groups with fewer than minConnectionsToConsolidate members are passed
+//     through unmodified.
+func consolidateSessions(sessions []*AccessSession) []*AccessSession {
+	if len(sessions) <= 1 {
+		return sessions
+	}
+
+	// Group sessions by consolidation key
+	groups := make(map[consolidationKey][]*AccessSession)
+	for _, s := range sessions {
+		key := consolidationKey{
+			sourceIP:   extractIP(s.SourceAddr),
+			destAddr:   s.DestAddr,
+			protocol:   s.Protocol,
+			resourceID: s.ResourceID,
+		}
+		groups[key] = append(groups[key], s)
+	}
+
+	result := make([]*AccessSession, 0, len(sessions))
+
+	for key, group := range groups {
+		// Small groups don't need consolidation
+		if len(group) < minConnectionsToConsolidate {
+			result = append(result, group...)
+			continue
+		}
+
+		// Sort the group by start time so we can detect gaps
+		sort.Slice(group, func(i, j int) bool {
+			return group[i].StartedAt.Before(group[j].StartedAt)
+		})
+
+		// Walk through and merge runs that are within the gap threshold
+		var merged []*AccessSession
+		cur := cloneSession(group[0])
+		cur.ConnectionCount = 1
+		sourcePorts := make(map[string]struct{})
+		sourcePorts[cur.SourceAddr] = struct{}{}
+
+		for i := 1; i < len(group); i++ {
+			s := group[i]
+
+			// Determine the gap: from the latest end time we've seen so far to the
+			// start of the next connection.
+			gapRef := cur.EndedAt
+			if gapRef.IsZero() {
+				gapRef = cur.StartedAt
+			}
+			gap := s.StartedAt.Sub(gapRef)
+
+			if gap <= sessionGapThreshold {
+				// Merge into the current consolidated session
+				cur.ConnectionCount++
+				cur.BytesTx += s.BytesTx
+				cur.BytesRx += s.BytesRx
+				sourcePorts[s.SourceAddr] = struct{}{}
+
+				// Extend EndedAt to the latest time
+				endTime := s.EndedAt
+				if endTime.IsZero() {
+					endTime = s.StartedAt
+				}
+				if endTime.After(cur.EndedAt) {
+					cur.EndedAt = endTime
+				}
+			} else {
+				// Gap exceeded — finalize the current session and start a new one
+				finalizeMergedSourceAddr(cur, key.sourceIP, sourcePorts)
+				merged = append(merged, cur)
+
+				cur = cloneSession(s)
+				cur.ConnectionCount = 1
+				sourcePorts = make(map[string]struct{})
+				sourcePorts[s.SourceAddr] = struct{}{}
+			}
+		}
+
+		// Finalize the last accumulated session
+		finalizeMergedSourceAddr(cur, key.sourceIP, sourcePorts)
+		merged = append(merged, cur)
+
+		result = append(result, merged...)
+	}
+
+	return result
+}
+
+// cloneSession creates a shallow copy of an AccessSession.
+func cloneSession(s *AccessSession) *AccessSession {
+	cp := *s
+	return &cp
+}
+
+// finalizeMergedSourceAddr sets the SourceAddr on a consolidated session.
+// If multiple distinct source addresses (ports) were seen, the port is
+// stripped and only the IP is kept so the log isn't misleading.
+func finalizeMergedSourceAddr(s *AccessSession, sourceIP string, ports map[string]struct{}) {
+	if len(ports) > 1 {
+		// Multiple source ports — just report the IP
+		s.SourceAddr = sourceIP
+	}
+	// Otherwise keep the original SourceAddr which already has ip:port
+}
+
+// flush drains the completed sessions buffer, consolidates bursts of
+// short-lived connections, compresses with zlib, and sends via the SendFunc.
+func (al *AccessLogger) flush() {
+	al.mu.Lock()
+	if len(al.completedSessions) == 0 {
+		al.mu.Unlock()
+		return
+	}
+	batch := al.completedSessions
+	al.completedSessions = make([]*AccessSession, 0)
+	sendFn := al.sendFn
+	al.mu.Unlock()
+
+	if sendFn == nil {
+		logger.Debug("Access logger: no send function configured, discarding %d sessions", len(batch))
+		return
+	}
+
+	// Consolidate bursts of short-lived connections into higher-level sessions
+	originalCount := len(batch)
+	batch = consolidateSessions(batch)
+	if len(batch) != originalCount {
+		logger.Info("Access logger: consolidated %d raw connections into %d sessions", originalCount, len(batch))
+	}
+
+	compressed, err := compressSessions(batch)
+	if err != nil {
+		logger.Error("Access logger: failed to compress %d sessions: %v", len(batch), err)
+		return
+	}
+
+	if err := sendFn(compressed); err != nil {
+		logger.Error("Access logger: failed to send %d sessions: %v", len(batch), err)
+		// Re-queue the batch so we don't lose data
+		al.mu.Lock()
+		al.completedSessions = append(batch, al.completedSessions...)
+		// Cap re-queued data to prevent unbounded growth if server is unreachable
+		if len(al.completedSessions) > maxBufferedSessions*5 {
+			dropped := len(al.completedSessions) - maxBufferedSessions*5
+			al.completedSessions = al.completedSessions[:maxBufferedSessions*5]
+			logger.Warn("Access logger: buffer overflow, dropped %d oldest sessions", dropped)
+		}
+		al.mu.Unlock()
+		return
+	}
+
+	logger.Info("Access logger: sent %d sessions to server", len(batch))
+}
+
+// compressSessions JSON-encodes the sessions, compresses with zlib, and returns
+// a base64-encoded string suitable for embedding in a JSON message.
+func compressSessions(sessions []*AccessSession) (string, error) {
+	jsonData, err := json.Marshal(sessions)
+	if err != nil {
+		return "", err
+	}
+
+	var buf bytes.Buffer
+	w, err := zlib.NewWriterLevel(&buf, zlib.BestCompression)
+	if err != nil {
+		return "", err
+	}
+	if _, err := w.Write(jsonData); err != nil {
+		w.Close()
+		return "", err
+	}
+	if err := w.Close(); err != nil {
+		return "", err
+	}
+
+	return base64.StdEncoding.EncodeToString(buf.Bytes()), nil
+}
+
+// Close shuts down the background loop, ends all active sessions,
+// and performs one final flush to send everything to the server.
+func (al *AccessLogger) Close() {
+	// Signal the background loop to stop
+	select {
+	case <-al.stopCh:
+		// Already closed
+		return
+	default:
+		close(al.stopCh)
+	}
+
+	// Wait for the background loop to exit so we don't race on flush
+	<-al.flushDone
+
+	al.mu.Lock()
+	now := time.Now()
+
+	// End all active sessions and move them to the completed buffer
+	for _, session := range al.sessions {
+		if session.EndedAt.IsZero() {
+			session.EndedAt = now
+			duration := now.Sub(session.StartedAt)
+			logger.Info("ACCESS END (shutdown) session=%s resource=%d proto=%s src=%s dst=%s started=%s ended=%s duration=%s",
+				session.SessionID, session.ResourceID, session.Protocol, session.SourceAddr, session.DestAddr,
+				session.StartedAt.Format(time.RFC3339), now.Format(time.RFC3339), duration)
+			al.completedSessions = append(al.completedSessions, session)
+		}
+	}
+
+	al.sessions = make(map[string]*AccessSession)
+	al.udpSessions = make(map[udpSessionKey]*AccessSession)
+	al.mu.Unlock()
+
+	// Final flush to send all remaining sessions to the server
+	al.flush()
+}

netstack2/access_log_test.go

@@ -0,0 +1,811 @@
+package netstack2
+
+import (
+	"testing"
+	"time"
+)
+
+func TestExtractIP(t *testing.T) {
+	tests := []struct {
+		name     string
+		addr     string
+		expected string
+	}{
+		{"ipv4 with port", "192.168.1.1:12345", "192.168.1.1"},
+		{"ipv4 without port", "192.168.1.1", "192.168.1.1"},
+		{"ipv6 with port", "[::1]:12345", "::1"},
+		{"ipv6 without port", "::1", "::1"},
+		{"empty string", "", ""},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := extractIP(tt.addr)
+			if result != tt.expected {
+				t.Errorf("extractIP(%q) = %q, want %q", tt.addr, result, tt.expected)
+			}
+		})
+	}
+}
+
+func TestConsolidateSessions_Empty(t *testing.T) {
+	result := consolidateSessions(nil)
+	if result != nil {
+		t.Errorf("expected nil, got %v", result)
+	}
+
+	result = consolidateSessions([]*AccessSession{})
+	if len(result) != 0 {
+		t.Errorf("expected empty slice, got %d items", len(result))
+	}
+}
+
+func TestConsolidateSessions_SingleSession(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "abc123",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(1 * time.Second),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 session, got %d", len(result))
+	}
+	if result[0].SourceAddr != "10.0.0.1:5000" {
+		t.Errorf("expected source addr preserved, got %q", result[0].SourceAddr)
+	}
+}
+
+func TestConsolidateSessions_MergesBurstFromSameSourceIP(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+			BytesTx:    100,
+			BytesRx:    200,
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+			BytesTx:    150,
+			BytesRx:    250,
+		},
+		{
+			SessionID:  "s3",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5002",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(400 * time.Millisecond),
+			EndedAt:    now.Add(500 * time.Millisecond),
+			BytesTx:    50,
+			BytesRx:    75,
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 consolidated session, got %d", len(result))
+	}
+
+	s := result[0]
+	if s.ConnectionCount != 3 {
+		t.Errorf("expected ConnectionCount=3, got %d", s.ConnectionCount)
+	}
+	if s.SourceAddr != "10.0.0.1" {
+		t.Errorf("expected source addr to be IP only (multiple ports), got %q", s.SourceAddr)
+	}
+	if s.DestAddr != "192.168.1.100:443" {
+		t.Errorf("expected dest addr preserved, got %q", s.DestAddr)
+	}
+	if s.StartedAt != now {
+		t.Errorf("expected StartedAt to be earliest time")
+	}
+	if s.EndedAt != now.Add(500*time.Millisecond) {
+		t.Errorf("expected EndedAt to be latest time")
+	}
+	expectedTx := int64(300)
+	expectedRx := int64(525)
+	if s.BytesTx != expectedTx {
+		t.Errorf("expected BytesTx=%d, got %d", expectedTx, s.BytesTx)
+	}
+	if s.BytesRx != expectedRx {
+		t.Errorf("expected BytesRx=%d, got %d", expectedRx, s.BytesRx)
+	}
+}
+
+func TestConsolidateSessions_SameSourcePortPreserved(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 session, got %d", len(result))
+	}
+	if result[0].SourceAddr != "10.0.0.1:5000" {
+		t.Errorf("expected source addr with port preserved when all ports are the same, got %q", result[0].SourceAddr)
+	}
+	if result[0].ConnectionCount != 2 {
+		t.Errorf("expected ConnectionCount=2, got %d", result[0].ConnectionCount)
+	}
+}
+
+func TestConsolidateSessions_GapSplitsSessions(t *testing.T) {
+	now := time.Now()
+
+	// First burst
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+		// Big gap here (10 seconds)
+		{
+			SessionID:  "s3",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5002",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(10 * time.Second),
+			EndedAt:    now.Add(10*time.Second + 100*time.Millisecond),
+		},
+		{
+			SessionID:  "s4",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5003",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(10*time.Second + 200*time.Millisecond),
+			EndedAt:    now.Add(10*time.Second + 300*time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 2 {
+		t.Fatalf("expected 2 consolidated sessions (gap split), got %d", len(result))
+	}
+
+	// Find the sessions by their start time
+	var first, second *AccessSession
+	for _, s := range result {
+		if s.StartedAt.Equal(now) {
+			first = s
+		} else {
+			second = s
+		}
+	}
+
+	if first == nil || second == nil {
+		t.Fatal("could not find both consolidated sessions")
+	}
+
+	if first.ConnectionCount != 2 {
+		t.Errorf("first burst: expected ConnectionCount=2, got %d", first.ConnectionCount)
+	}
+	if second.ConnectionCount != 2 {
+		t.Errorf("second burst: expected ConnectionCount=2, got %d", second.ConnectionCount)
+	}
+}
+
+func TestConsolidateSessions_DifferentDestinationsNotMerged(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:8080",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	// Each goes to a different dest port so they should not be merged
+	if len(result) != 2 {
+		t.Fatalf("expected 2 sessions (different destinations), got %d", len(result))
+	}
+}
+
+func TestConsolidateSessions_DifferentProtocolsNotMerged(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "udp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 2 {
+		t.Fatalf("expected 2 sessions (different protocols), got %d", len(result))
+	}
+}
+
+func TestConsolidateSessions_DifferentResourceIDsNotMerged(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 2,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 2 {
+		t.Fatalf("expected 2 sessions (different resource IDs), got %d", len(result))
+	}
+}
+
+func TestConsolidateSessions_DifferentSourceIPsNotMerged(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.2:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 2 {
+		t.Fatalf("expected 2 sessions (different source IPs), got %d", len(result))
+	}
+}
+
+func TestConsolidateSessions_OutOfOrderInput(t *testing.T) {
+	now := time.Now()
+	// Provide sessions out of chronological order to verify sorting
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s3",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5002",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(400 * time.Millisecond),
+			EndedAt:    now.Add(500 * time.Millisecond),
+			BytesTx:    30,
+		},
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+			BytesTx:    10,
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+			BytesTx:    20,
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 consolidated session, got %d", len(result))
+	}
+
+	s := result[0]
+	if s.ConnectionCount != 3 {
+		t.Errorf("expected ConnectionCount=3, got %d", s.ConnectionCount)
+	}
+	if s.StartedAt != now {
+		t.Errorf("expected StartedAt to be earliest time")
+	}
+	if s.EndedAt != now.Add(500*time.Millisecond) {
+		t.Errorf("expected EndedAt to be latest time")
+	}
+	if s.BytesTx != 60 {
+		t.Errorf("expected BytesTx=60, got %d", s.BytesTx)
+	}
+}
+
+func TestConsolidateSessions_ExactlyAtGapThreshold(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			// Starts exactly sessionGapThreshold after s1 ends — should still merge
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(100*time.Millisecond + sessionGapThreshold),
+			EndedAt:    now.Add(100*time.Millisecond + sessionGapThreshold + 50*time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 session (gap exactly at threshold merges), got %d", len(result))
+	}
+	if result[0].ConnectionCount != 2 {
+		t.Errorf("expected ConnectionCount=2, got %d", result[0].ConnectionCount)
+	}
+}
+
+func TestConsolidateSessions_JustOverGapThreshold(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			// Starts 1ms over the gap threshold after s1 ends — should split
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(100*time.Millisecond + sessionGapThreshold + 1*time.Millisecond),
+			EndedAt:    now.Add(100*time.Millisecond + sessionGapThreshold + 50*time.Millisecond),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 2 {
+		t.Fatalf("expected 2 sessions (gap just over threshold splits), got %d", len(result))
+	}
+}
+
+func TestConsolidateSessions_UDPSessions(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		{
+			SessionID:  "u1",
+			ResourceID: 5,
+			SourceAddr: "10.0.0.1:6000",
+			DestAddr:   "192.168.1.100:53",
+			Protocol:   "udp",
+			StartedAt:  now,
+			EndedAt:    now.Add(50 * time.Millisecond),
+			BytesTx:    64,
+			BytesRx:    512,
+		},
+		{
+			SessionID:  "u2",
+			ResourceID: 5,
+			SourceAddr: "10.0.0.1:6001",
+			DestAddr:   "192.168.1.100:53",
+			Protocol:   "udp",
+			StartedAt:  now.Add(100 * time.Millisecond),
+			EndedAt:    now.Add(150 * time.Millisecond),
+			BytesTx:    64,
+			BytesRx:    256,
+		},
+		{
+			SessionID:  "u3",
+			ResourceID: 5,
+			SourceAddr: "10.0.0.1:6002",
+			DestAddr:   "192.168.1.100:53",
+			Protocol:   "udp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(250 * time.Millisecond),
+			BytesTx:    64,
+			BytesRx:    128,
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 consolidated UDP session, got %d", len(result))
+	}
+
+	s := result[0]
+	if s.Protocol != "udp" {
+		t.Errorf("expected protocol=udp, got %q", s.Protocol)
+	}
+	if s.ConnectionCount != 3 {
+		t.Errorf("expected ConnectionCount=3, got %d", s.ConnectionCount)
+	}
+	if s.SourceAddr != "10.0.0.1" {
+		t.Errorf("expected source addr to be IP only, got %q", s.SourceAddr)
+	}
+	if s.BytesTx != 192 {
+		t.Errorf("expected BytesTx=192, got %d", s.BytesTx)
+	}
+	if s.BytesRx != 896 {
+		t.Errorf("expected BytesRx=896, got %d", s.BytesRx)
+	}
+}
+
+func TestConsolidateSessions_MixedGroupsSomeConsolidatedSomeNot(t *testing.T) {
+	now := time.Now()
+	sessions := []*AccessSession{
+		// Group 1: 3 connections to :443 from same IP — should consolidate
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(100 * time.Millisecond),
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(200 * time.Millisecond),
+			EndedAt:    now.Add(300 * time.Millisecond),
+		},
+		{
+			SessionID:  "s3",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5002",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(400 * time.Millisecond),
+			EndedAt:    now.Add(500 * time.Millisecond),
+		},
+		// Group 2: 1 connection to :8080 from different IP — should pass through
+		{
+			SessionID:  "s4",
+			ResourceID: 2,
+			SourceAddr: "10.0.0.2:6000",
+			DestAddr:   "192.168.1.200:8080",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(1 * time.Second),
+			EndedAt:    now.Add(2 * time.Second),
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 2 {
+		t.Fatalf("expected 2 sessions total, got %d", len(result))
+	}
+
+	var consolidated, passthrough *AccessSession
+	for _, s := range result {
+		if s.ConnectionCount > 1 {
+			consolidated = s
+		} else {
+			passthrough = s
+		}
+	}
+
+	if consolidated == nil {
+		t.Fatal("expected a consolidated session")
+	}
+	if consolidated.ConnectionCount != 3 {
+		t.Errorf("consolidated: expected ConnectionCount=3, got %d", consolidated.ConnectionCount)
+	}
+
+	if passthrough == nil {
+		t.Fatal("expected a passthrough session")
+	}
+	if passthrough.SessionID != "s4" {
+		t.Errorf("passthrough: expected session s4, got %s", passthrough.SessionID)
+	}
+}
+
+func TestConsolidateSessions_OverlappingConnections(t *testing.T) {
+	now := time.Now()
+	// Connections that overlap in time (not sequential)
+	sessions := []*AccessSession{
+		{
+			SessionID:  "s1",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5000",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now,
+			EndedAt:    now.Add(5 * time.Second),
+			BytesTx:    100,
+		},
+		{
+			SessionID:  "s2",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5001",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(1 * time.Second),
+			EndedAt:    now.Add(3 * time.Second),
+			BytesTx:    200,
+		},
+		{
+			SessionID:  "s3",
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:5002",
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(2 * time.Second),
+			EndedAt:    now.Add(6 * time.Second),
+			BytesTx:    300,
+		},
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 1 {
+		t.Fatalf("expected 1 consolidated session, got %d", len(result))
+	}
+
+	s := result[0]
+	if s.ConnectionCount != 3 {
+		t.Errorf("expected ConnectionCount=3, got %d", s.ConnectionCount)
+	}
+	if s.StartedAt != now {
+		t.Error("expected StartedAt to be earliest")
+	}
+	if s.EndedAt != now.Add(6*time.Second) {
+		t.Error("expected EndedAt to be the latest end time")
+	}
+	if s.BytesTx != 600 {
+		t.Errorf("expected BytesTx=600, got %d", s.BytesTx)
+	}
+}
+
+func TestConsolidateSessions_DoesNotMutateOriginals(t *testing.T) {
+	now := time.Now()
+	s1 := &AccessSession{
+		SessionID:  "s1",
+		ResourceID: 1,
+		SourceAddr: "10.0.0.1:5000",
+		DestAddr:   "192.168.1.100:443",
+		Protocol:   "tcp",
+		StartedAt:  now,
+		EndedAt:    now.Add(100 * time.Millisecond),
+		BytesTx:    100,
+	}
+	s2 := &AccessSession{
+		SessionID:  "s2",
+		ResourceID: 1,
+		SourceAddr: "10.0.0.1:5001",
+		DestAddr:   "192.168.1.100:443",
+		Protocol:   "tcp",
+		StartedAt:  now.Add(200 * time.Millisecond),
+		EndedAt:    now.Add(300 * time.Millisecond),
+		BytesTx:    200,
+	}
+
+	// Save original values
+	origS1Addr := s1.SourceAddr
+	origS1Bytes := s1.BytesTx
+	origS2Addr := s2.SourceAddr
+	origS2Bytes := s2.BytesTx
+
+	_ = consolidateSessions([]*AccessSession{s1, s2})
+
+	if s1.SourceAddr != origS1Addr {
+		t.Errorf("s1.SourceAddr was mutated: %q -> %q", origS1Addr, s1.SourceAddr)
+	}
+	if s1.BytesTx != origS1Bytes {
+		t.Errorf("s1.BytesTx was mutated: %d -> %d", origS1Bytes, s1.BytesTx)
+	}
+	if s2.SourceAddr != origS2Addr {
+		t.Errorf("s2.SourceAddr was mutated: %q -> %q", origS2Addr, s2.SourceAddr)
+	}
+	if s2.BytesTx != origS2Bytes {
+		t.Errorf("s2.BytesTx was mutated: %d -> %d", origS2Bytes, s2.BytesTx)
+	}
+}
+
+func TestConsolidateSessions_ThreeBurstsWithGaps(t *testing.T) {
+	now := time.Now()
+
+	sessions := make([]*AccessSession, 0, 9)
+
+	// Burst 1: 3 connections at t=0
+	for i := 0; i < 3; i++ {
+		sessions = append(sessions, &AccessSession{
+			SessionID:  generateSessionID(),
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:" + string(rune('A'+i)),
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(time.Duration(i*100) * time.Millisecond),
+			EndedAt:    now.Add(time.Duration(i*100+50) * time.Millisecond),
+		})
+	}
+
+	// Burst 2: 3 connections at t=20s (well past the 5s gap)
+	for i := 0; i < 3; i++ {
+		sessions = append(sessions, &AccessSession{
+			SessionID:  generateSessionID(),
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:" + string(rune('D'+i)),
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(20*time.Second + time.Duration(i*100)*time.Millisecond),
+			EndedAt:    now.Add(20*time.Second + time.Duration(i*100+50)*time.Millisecond),
+		})
+	}
+
+	// Burst 3: 3 connections at t=40s
+	for i := 0; i < 3; i++ {
+		sessions = append(sessions, &AccessSession{
+			SessionID:  generateSessionID(),
+			ResourceID: 1,
+			SourceAddr: "10.0.0.1:" + string(rune('G'+i)),
+			DestAddr:   "192.168.1.100:443",
+			Protocol:   "tcp",
+			StartedAt:  now.Add(40*time.Second + time.Duration(i*100)*time.Millisecond),
+			EndedAt:    now.Add(40*time.Second + time.Duration(i*100+50)*time.Millisecond),
+		})
+	}
+
+	result := consolidateSessions(sessions)
+	if len(result) != 3 {
+		t.Fatalf("expected 3 consolidated sessions (3 bursts), got %d", len(result))
+	}
+
+	for _, s := range result {
+		if s.ConnectionCount != 3 {
+			t.Errorf("expected each burst to have ConnectionCount=3, got %d (started=%v)", s.ConnectionCount, s.StartedAt)
+		}
+	}
+}
+
+func TestFinalizeMergedSourceAddr(t *testing.T) {
+	s := &AccessSession{SourceAddr: "10.0.0.1:5000"}
+	ports := map[string]struct{}{"10.0.0.1:5000": {}}
+	finalizeMergedSourceAddr(s, "10.0.0.1", ports)
+	if s.SourceAddr != "10.0.0.1:5000" {
+		t.Errorf("single port: expected addr preserved, got %q", s.SourceAddr)
+	}
+
+	s2 := &AccessSession{SourceAddr: "10.0.0.1:5000"}
+	ports2 := map[string]struct{}{"10.0.0.1:5000": {}, "10.0.0.1:5001": {}}
+	finalizeMergedSourceAddr(s2, "10.0.0.1", ports2)
+	if s2.SourceAddr != "10.0.0.1" {
+		t.Errorf("multiple ports: expected IP only, got %q", s2.SourceAddr)
+	}
+}
+
+func TestCloneSession(t *testing.T) {
+	original := &AccessSession{
+		SessionID:  "test",
+		ResourceID: 42,
+		SourceAddr: "1.2.3.4:100",
+		DestAddr:   "5.6.7.8:443",
+		Protocol:   "tcp",
+		BytesTx:    999,
+	}
+
+	clone := cloneSession(original)
+
+	if clone == original {
+		t.Error("clone should be a different pointer")
+	}
+	if clone.SessionID != original.SessionID {
+		t.Error("clone should have same SessionID")
+	}
+
+	// Mutating clone should not affect original
+	clone.BytesTx = 0
+	clone.SourceAddr = "changed"
+	if original.BytesTx != 999 {
+		t.Error("mutating clone affected original BytesTx")
+	}
+	if original.SourceAddr != "1.2.3.4:100" {
+		t.Error("mutating clone affected original SourceAddr")
+	}
+}

netstack2/handlers.go

@@ -1,8 +1,3 @@
-/* SPDX-License-Identifier: MIT
- *
- * Copyright (C) 2017-2025 WireGuard LLC. All Rights Reserved.
- */
-
 package netstack2
 
 import (
@@ -10,12 +5,18 @@ import (
 	"fmt"
 	"io"
 	"net"
+	"net/netip"
+	"os/exec"
 	"sync"
 	"time"
 
 	"github.com/fosrl/newt/logger"
+	"golang.org/x/net/icmp"
+	"golang.org/x/net/ipv4"
 	"gvisor.dev/gvisor/pkg/tcpip"
 	"gvisor.dev/gvisor/pkg/tcpip/adapters/gonet"
+	"gvisor.dev/gvisor/pkg/tcpip/checksum"
+	"gvisor.dev/gvisor/pkg/tcpip/header"
 	"gvisor.dev/gvisor/pkg/tcpip/stack"
 	"gvisor.dev/gvisor/pkg/tcpip/transport/tcp"
 	"gvisor.dev/gvisor/pkg/tcpip/transport/udp"
@@ -58,6 +59,9 @@ const (
 
 	// Buffer size for copying data
 	bufferSize = 32 * 1024
+
+	// icmpTimeout is the default timeout for ICMP ping requests.
+	icmpTimeout = 5 * time.Second
 )
 
 // TCPHandler handles TCP connections from netstack
@@ -72,6 +76,12 @@ type UDPHandler struct {
 	proxyHandler *ProxyHandler
 }
 
+// ICMPHandler handles ICMP packets from netstack
+type ICMPHandler struct {
+	stack        *stack.Stack
+	proxyHandler *ProxyHandler
+}
+
 // NewTCPHandler creates a new TCP handler
 func NewTCPHandler(s *stack.Stack, ph *ProxyHandler) *TCPHandler {
 	return &TCPHandler{stack: s, proxyHandler: ph}
@@ -82,6 +92,11 @@ func NewUDPHandler(s *stack.Stack, ph *ProxyHandler) *UDPHandler {
 	return &UDPHandler{stack: s, proxyHandler: ph}
 }
 
+// NewICMPHandler creates a new ICMP handler
+func NewICMPHandler(s *stack.Stack, ph *ProxyHandler) *ICMPHandler {
+	return &ICMPHandler{stack: s, proxyHandler: ph}
+}
+
 // InstallTCPHandler installs the TCP forwarder on the stack
 func (h *TCPHandler) InstallTCPHandler() error {
 	tcpForwarder := tcp.NewForwarder(h.stack, defaultWndSize, maxConnAttempts, func(r *tcp.ForwarderRequest) {
@@ -117,14 +132,40 @@ func (h *TCPHandler) InstallTCPHandler() error {
 
 // handleTCPConn handles a TCP connection by proxying it to the actual target
 func (h *TCPHandler) handleTCPConn(netstackConn *gonet.TCPConn, id stack.TransportEndpointID) {
-	defer netstackConn.Close()
-
-	// Extract source and target address from the connection ID
+	// Extract source and target address from the connection ID first so they
+	// are available for HTTP routing before any defer is set up.
 	srcIP := id.RemoteAddress.String()
 	srcPort := id.RemotePort
 	dstIP := id.LocalAddress.String()
 	dstPort := id.LocalPort
 
+	// Drop connection if blocking is enabled
+	if h.proxyHandler != nil && h.proxyHandler.IsBlocked() {
+		logger.Debug("TCP Forwarder: connection blocked: %s:%d -> %s:%d", srcIP, srcPort, dstIP, dstPort)
+		netstackConn.Close()
+		return
+	}
+
+	// For HTTP/HTTPS ports, look up the matching subnet rule. If the rule has
+	// Protocol configured, hand the connection off to the HTTP handler which
+	// takes full ownership of the lifecycle (the defer close must not be
+	// installed before this point).
+	if (dstPort == 80 || dstPort == 443) && h.proxyHandler != nil && h.proxyHandler.httpHandler != nil {
+		srcAddr, _ := netip.ParseAddr(srcIP)
+		dstAddr, _ := netip.ParseAddr(dstIP)
+		rule := h.proxyHandler.subnetLookup.Match(srcAddr, dstAddr, dstPort, tcp.ProtocolNumber)
+		if rule != nil && rule.Protocol != "" && len(rule.HTTPTargets) > 0 {
+			logger.Info("TCP Forwarder: Routing %s:%d -> %s:%d to HTTP handler (%s)",
+				srcIP, srcPort, dstIP, dstPort, rule.Protocol)
+			h.proxyHandler.httpHandler.HandleConn(netstackConn, rule)
+			return
+		}
+		// Otherwise fall through to raw TCP forwarding (e.g. CIDR resources
+		// that happen to use port 80/443 without HTTP configuration).
+	}
+
+	defer netstackConn.Close()
+
 	logger.Info("TCP Forwarder: Handling connection %s:%d -> %s:%d", srcIP, srcPort, dstIP, dstPort)
 
 	// Check if there's a destination rewrite for this connection (e.g., localhost targets)
@@ -138,6 +179,18 @@ func (h *TCPHandler) handleTCPConn(netstackConn *gonet.TCPConn, id stack.Transpo
 
 	targetAddr := fmt.Sprintf("%s:%d", actualDstIP, dstPort)
 
+	// Look up resource ID and start access session if applicable
+	var accessSessionID string
+	if h.proxyHandler != nil {
+		resourceId := h.proxyHandler.LookupResourceId(srcIP, dstIP, dstPort, uint8(tcp.ProtocolNumber))
+		if resourceId != 0 {
+			if al := h.proxyHandler.GetAccessLogger(); al != nil {
+				srcAddr := fmt.Sprintf("%s:%d", srcIP, srcPort)
+				accessSessionID = al.StartTCPSession(resourceId, srcAddr, targetAddr)
+			}
+		}
+	}
+
 	// Create context with timeout for connection establishment
 	ctx, cancel := context.WithTimeout(context.Background(), tcpConnectTimeout)
 	defer cancel()
@@ -147,11 +200,26 @@ func (h *TCPHandler) handleTCPConn(netstackConn *gonet.TCPConn, id stack.Transpo
 	targetConn, err := d.DialContext(ctx, "tcp", targetAddr)
 	if err != nil {
 		logger.Info("TCP Forwarder: Failed to connect to %s: %v", targetAddr, err)
+		// End access session on connection failure
+		if accessSessionID != "" {
+			if al := h.proxyHandler.GetAccessLogger(); al != nil {
+				al.EndTCPSession(accessSessionID)
+			}
+		}
 		// Connection failed, netstack will handle RST
 		return
 	}
 	defer targetConn.Close()
 
+	// End access session when connection closes
+	if accessSessionID != "" {
+		defer func() {
+			if al := h.proxyHandler.GetAccessLogger(); al != nil {
+				al.EndTCPSession(accessSessionID)
+			}
+		}()
+	}
+
 	logger.Info("TCP Forwarder: Successfully connected to %s, starting bidirectional copy", targetAddr)
 
 	// Bidirectional copy between netstack and target
@@ -249,6 +317,12 @@ func (h *UDPHandler) handleUDPConn(netstackConn *gonet.UDPConn, id stack.Transpo
 
 	logger.Info("UDP Forwarder: Handling connection %s:%d -> %s:%d", srcIP, srcPort, dstIP, dstPort)
 
+	// Drop connection if blocking is enabled
+	if h.proxyHandler != nil && h.proxyHandler.IsBlocked() {
+		logger.Debug("UDP Forwarder: connection blocked: %s:%d -> %s:%d", srcIP, srcPort, dstIP, dstPort)
+		return
+	}
+
 	// Check if there's a destination rewrite for this connection (e.g., localhost targets)
 	actualDstIP := dstIP
 	if h.proxyHandler != nil {
@@ -260,6 +334,27 @@ func (h *UDPHandler) handleUDPConn(netstackConn *gonet.UDPConn, id stack.Transpo
 
 	targetAddr := fmt.Sprintf("%s:%d", actualDstIP, dstPort)
 
+	// Look up resource ID and start access session if applicable
+	var accessSessionID string
+	if h.proxyHandler != nil {
+		resourceId := h.proxyHandler.LookupResourceId(srcIP, dstIP, dstPort, uint8(udp.ProtocolNumber))
+		if resourceId != 0 {
+			if al := h.proxyHandler.GetAccessLogger(); al != nil {
+				srcAddr := fmt.Sprintf("%s:%d", srcIP, srcPort)
+				accessSessionID = al.TrackUDPSession(resourceId, srcAddr, targetAddr)
+			}
+		}
+	}
+
+	// End access session when UDP handler returns (timeout or error)
+	if accessSessionID != "" {
+		defer func() {
+			if al := h.proxyHandler.GetAccessLogger(); al != nil {
+				al.EndUDPSession(accessSessionID)
+			}
+		}()
+	}
+
 	// Resolve target address
 	remoteUDPAddr, err := net.ResolveUDPAddr("udp", targetAddr)
 	if err != nil {
@@ -348,3 +443,334 @@ func copyPacketData(dst, src net.PacketConn, to net.Addr, timeout time.Duration)
 		dst.SetReadDeadline(time.Now().Add(timeout))
 	}
 }
+
+// InstallICMPHandler installs the ICMP handler on the stack
+func (h *ICMPHandler) InstallICMPHandler() error {
+	h.stack.SetTransportProtocolHandler(header.ICMPv4ProtocolNumber, h.handleICMPPacket)
+	logger.Debug("ICMP Handler: Installed ICMP protocol handler")
+	return nil
+}
+
+// handleICMPPacket handles incoming ICMP packets
+func (h *ICMPHandler) handleICMPPacket(id stack.TransportEndpointID, pkt *stack.PacketBuffer) bool {
+	logger.Debug("ICMP Handler: Received ICMP packet from %s to %s", id.RemoteAddress, id.LocalAddress)
+
+	// Get the ICMP header from the packet
+	icmpData := pkt.TransportHeader().Slice()
+	if len(icmpData) < header.ICMPv4MinimumSize {
+		logger.Debug("ICMP Handler: Packet too small for ICMP header: %d bytes", len(icmpData))
+		return false
+	}
+
+	icmpHdr := header.ICMPv4(icmpData)
+	icmpType := icmpHdr.Type()
+	icmpCode := icmpHdr.Code()
+
+	logger.Debug("ICMP Handler: Type=%d, Code=%d, Ident=%d, Seq=%d",
+		icmpType, icmpCode, icmpHdr.Ident(), icmpHdr.Sequence())
+
+	// Only handle Echo Request (ping)
+	if icmpType != header.ICMPv4Echo {
+		logger.Debug("ICMP Handler: Ignoring non-echo ICMP type: %d", icmpType)
+		return false
+	}
+
+	// Extract source and destination addresses
+	srcIP := id.RemoteAddress.String()
+	dstIP := id.LocalAddress.String()
+
+	logger.Info("ICMP Handler: Echo Request from %s to %s (ident=%d, seq=%d)",
+		srcIP, dstIP, icmpHdr.Ident(), icmpHdr.Sequence())
+
+	// Convert to netip.Addr for subnet matching
+	srcAddr, err := netip.ParseAddr(srcIP)
+	if err != nil {
+		logger.Debug("ICMP Handler: Failed to parse source IP %s: %v", srcIP, err)
+		return false
+	}
+	dstAddr, err := netip.ParseAddr(dstIP)
+	if err != nil {
+		logger.Debug("ICMP Handler: Failed to parse dest IP %s: %v", dstIP, err)
+		return false
+	}
+
+	// Check subnet rules (use port 0 for ICMP since it doesn't have ports)
+	if h.proxyHandler == nil {
+		logger.Debug("ICMP Handler: No proxy handler configured")
+		return false
+	}
+
+	matchedRule := h.proxyHandler.subnetLookup.Match(srcAddr, dstAddr, 0, header.ICMPv4ProtocolNumber)
+	if matchedRule == nil {
+		logger.Debug("ICMP Handler: No matching subnet rule for %s -> %s", srcIP, dstIP)
+		return false
+	}
+
+	logger.Info("ICMP Handler: Matched subnet rule for %s -> %s", srcIP, dstIP)
+
+	// Determine actual destination (with possible rewrite)
+	actualDstIP := dstIP
+	if matchedRule.RewriteTo != "" {
+		resolvedAddr, err := h.proxyHandler.resolveRewriteAddress(matchedRule.RewriteTo)
+		if err != nil {
+			logger.Info("ICMP Handler: Failed to resolve rewrite address %s: %v", matchedRule.RewriteTo, err)
+		} else {
+			actualDstIP = resolvedAddr.String()
+			logger.Info("ICMP Handler: Using rewritten destination %s (original: %s)", actualDstIP, dstIP)
+		}
+	}
+
+	// Get the full ICMP payload (including the data after the header)
+	icmpPayload := pkt.Data().AsRange().ToSlice()
+
+	// Handle the ping in a goroutine to avoid blocking
+	go h.proxyPing(srcIP, dstIP, actualDstIP, icmpHdr.Ident(), icmpHdr.Sequence(), icmpPayload)
+
+	return true
+}
+
+// proxyPing sends a ping to the actual destination and injects the reply back
+func (h *ICMPHandler) proxyPing(srcIP, originalDstIP, actualDstIP string, ident, seq uint16, payload []byte) {
+	logger.Debug("ICMP Handler: Proxying ping from %s to %s (actual: %s), ident=%d, seq=%d",
+		srcIP, originalDstIP, actualDstIP, ident, seq)
+
+	// Try three methods in order: ip4:icmp -> udp4 -> ping command
+	// Track which method succeeded so we can handle identifier matching correctly
+	method, success := h.tryICMPMethods(actualDstIP, ident, seq, payload)
+
+	if !success {
+		logger.Info("ICMP Handler: All ping methods failed for %s", actualDstIP)
+		return
+	}
+
+	logger.Info("ICMP Handler: Ping successful to %s using %s, injecting reply (ident=%d, seq=%d)",
+		actualDstIP, method, ident, seq)
+
+	// Build the reply packet to inject back into the netstack
+	// The reply should appear to come from the original destination (before rewrite)
+	h.injectICMPReply(srcIP, originalDstIP, ident, seq, payload)
+}
+
+// tryICMPMethods tries all available ICMP methods in order
+func (h *ICMPHandler) tryICMPMethods(actualDstIP string, ident, seq uint16, payload []byte) (string, bool) {
+	if h.tryRawICMP(actualDstIP, ident, seq, payload, false) {
+		return "raw ICMP", true
+	}
+	if h.tryUnprivilegedICMP(actualDstIP, ident, seq, payload) {
+		return "unprivileged ICMP", true
+	}
+	if h.tryPingCommand(actualDstIP, ident, seq, payload) {
+		return "ping command", true
+	}
+	return "", false
+}
+
+// tryRawICMP attempts to ping using raw ICMP sockets (requires CAP_NET_RAW or root)
+func (h *ICMPHandler) tryRawICMP(actualDstIP string, ident, seq uint16, payload []byte, ignoreIdent bool) bool {
+	conn, err := icmp.ListenPacket("ip4:icmp", "0.0.0.0")
+	if err != nil {
+		logger.Debug("ICMP Handler: Raw ICMP socket not available: %v", err)
+		return false
+	}
+	defer conn.Close()
+
+	logger.Debug("ICMP Handler: Using raw ICMP socket")
+	return h.sendAndReceiveICMP(conn, actualDstIP, ident, seq, payload, false, ignoreIdent)
+}
+
+// tryUnprivilegedICMP attempts to ping using unprivileged ICMP (requires ping_group_range configured)
+func (h *ICMPHandler) tryUnprivilegedICMP(actualDstIP string, ident, seq uint16, payload []byte) bool {
+	conn, err := icmp.ListenPacket("udp4", "0.0.0.0")
+	if err != nil {
+		logger.Debug("ICMP Handler: Unprivileged ICMP socket not available: %v", err)
+		return false
+	}
+	defer conn.Close()
+
+	logger.Debug("ICMP Handler: Using unprivileged ICMP socket")
+	// Unprivileged ICMP doesn't let us control the identifier, so we ignore it in matching
+	return h.sendAndReceiveICMP(conn, actualDstIP, ident, seq, payload, true, true)
+}
+
+// sendAndReceiveICMP sends an ICMP echo request and waits for the reply
+func (h *ICMPHandler) sendAndReceiveICMP(conn *icmp.PacketConn, actualDstIP string, ident, seq uint16, payload []byte, isUnprivileged bool, ignoreIdent bool) bool {
+	// Build the ICMP echo request message
+	echoMsg := &icmp.Message{
+		Type: ipv4.ICMPTypeEcho,
+		Code: 0,
+		Body: &icmp.Echo{
+			ID:   int(ident),
+			Seq:  int(seq),
+			Data: payload,
+		},
+	}
+
+	msgBytes, err := echoMsg.Marshal(nil)
+	if err != nil {
+		logger.Debug("ICMP Handler: Failed to marshal ICMP message: %v", err)
+		return false
+	}
+
+	// Resolve destination address based on socket type
+	var writeErr error
+	if isUnprivileged {
+		// For unprivileged ICMP, use UDP-style addressing
+		udpAddr := &net.UDPAddr{IP: net.ParseIP(actualDstIP)}
+		logger.Debug("ICMP Handler: Sending ping to %s (unprivileged)", udpAddr.String())
+		conn.SetDeadline(time.Now().Add(icmpTimeout))
+		_, writeErr = conn.WriteTo(msgBytes, udpAddr)
+	} else {
+		// For raw ICMP, use IP addressing
+		dst, err := net.ResolveIPAddr("ip4", actualDstIP)
+		if err != nil {
+			logger.Debug("ICMP Handler: Failed to resolve destination %s: %v", actualDstIP, err)
+			return false
+		}
+		logger.Debug("ICMP Handler: Sending ping to %s (raw)", dst.String())
+		conn.SetDeadline(time.Now().Add(icmpTimeout))
+		_, writeErr = conn.WriteTo(msgBytes, dst)
+	}
+
+	if writeErr != nil {
+		logger.Debug("ICMP Handler: Failed to send ping to %s: %v", actualDstIP, writeErr)
+		return false
+	}
+
+	logger.Debug("ICMP Handler: Ping sent to %s, waiting for reply (ident=%d, seq=%d)", actualDstIP, ident, seq)
+
+	// Wait for reply - loop to filter out non-matching packets
+	replyBuf := make([]byte, 1500)
+
+	for {
+		n, peer, err := conn.ReadFrom(replyBuf)
+		if err != nil {
+			logger.Debug("ICMP Handler: Failed to receive ping reply from %s: %v", actualDstIP, err)
+			return false
+		}
+
+		logger.Debug("ICMP Handler: Received %d bytes from %s", n, peer.String())
+
+		// Parse the reply
+		replyMsg, err := icmp.ParseMessage(1, replyBuf[:n])
+		if err != nil {
+			logger.Debug("ICMP Handler: Failed to parse ICMP message: %v", err)
+			continue
+		}
+
+		// Check if it's an echo reply (type 0), not an echo request (type 8)
+		if replyMsg.Type != ipv4.ICMPTypeEchoReply {
+			logger.Debug("ICMP Handler: Received non-echo-reply type: %v, continuing to wait", replyMsg.Type)
+			continue
+		}
+
+		reply, ok := replyMsg.Body.(*icmp.Echo)
+		if !ok {
+			logger.Debug("ICMP Handler: Invalid echo reply body type, continuing to wait")
+			continue
+		}
+
+		// Verify the sequence matches what we sent
+		// For unprivileged ICMP, the kernel controls the identifier, so we only check sequence
+		if reply.Seq != int(seq) {
+			logger.Debug("ICMP Handler: Reply seq mismatch: got seq=%d, want seq=%d", reply.Seq, seq)
+			continue
+		}
+
+		if !ignoreIdent && reply.ID != int(ident) {
+			logger.Debug("ICMP Handler: Reply ident mismatch: got ident=%d, want ident=%d", reply.ID, ident)
+			continue
+		}
+
+		// Found matching reply
+		logger.Debug("ICMP Handler: Received valid echo reply")
+		return true
+	}
+}
+
+// tryPingCommand attempts to ping using the system ping command (always works, but less control)
+func (h *ICMPHandler) tryPingCommand(actualDstIP string, ident, seq uint16, payload []byte) bool {
+	logger.Debug("ICMP Handler: Attempting to use system ping command")
+
+	ctx, cancel := context.WithTimeout(context.Background(), icmpTimeout)
+	defer cancel()
+
+	// Send one ping with timeout
+	// -c 1: count = 1 packet
+	// -W 5: timeout = 5 seconds
+	// -q: quiet output (just summary)
+	cmd := exec.CommandContext(ctx, "ping", "-c", "1", "-W", "5", "-q", actualDstIP)
+	output, err := cmd.CombinedOutput()
+
+	if err != nil {
+		logger.Debug("ICMP Handler: System ping command failed: %v, output: %s", err, string(output))
+		return false
+	}
+
+	logger.Debug("ICMP Handler: System ping command succeeded")
+	return true
+}
+
+// injectICMPReply creates an ICMP echo reply packet and queues it to be sent back through the tunnel
+func (h *ICMPHandler) injectICMPReply(dstIP, srcIP string, ident, seq uint16, payload []byte) {
+	logger.Debug("ICMP Handler: Creating reply from %s to %s (ident=%d, seq=%d)",
+		srcIP, dstIP, ident, seq)
+
+	// Parse addresses
+	srcAddr, err := netip.ParseAddr(srcIP)
+	if err != nil {
+		logger.Info("ICMP Handler: Failed to parse source IP for reply: %v", err)
+		return
+	}
+	dstAddr, err := netip.ParseAddr(dstIP)
+	if err != nil {
+		logger.Info("ICMP Handler: Failed to parse dest IP for reply: %v", err)
+		return
+	}
+
+	// Calculate total packet size
+	ipHeaderLen := header.IPv4MinimumSize
+	icmpHeaderLen := header.ICMPv4MinimumSize
+	totalLen := ipHeaderLen + icmpHeaderLen + len(payload)
+
+	// Create the packet buffer
+	pkt := make([]byte, totalLen)
+
+	// Build IPv4 header
+	ipHdr := header.IPv4(pkt[:ipHeaderLen])
+	ipHdr.Encode(&header.IPv4Fields{
+		TotalLength: uint16(totalLen),
+		TTL:         64,
+		Protocol:    uint8(header.ICMPv4ProtocolNumber),
+		SrcAddr:     tcpip.AddrFrom4(srcAddr.As4()),
+		DstAddr:     tcpip.AddrFrom4(dstAddr.As4()),
+	})
+	ipHdr.SetChecksum(^ipHdr.CalculateChecksum())
+
+	// Build ICMP header
+	icmpHdr := header.ICMPv4(pkt[ipHeaderLen : ipHeaderLen+icmpHeaderLen])
+	icmpHdr.SetType(header.ICMPv4EchoReply)
+	icmpHdr.SetCode(0)
+	icmpHdr.SetIdent(ident)
+	icmpHdr.SetSequence(seq)
+
+	// Copy payload
+	copy(pkt[ipHeaderLen+icmpHeaderLen:], payload)
+
+	// Calculate ICMP checksum (covers ICMP header + payload)
+	icmpHdr.SetChecksum(0)
+	icmpData := pkt[ipHeaderLen:]
+	icmpHdr.SetChecksum(^checksum.Checksum(icmpData, 0))
+
+	logger.Debug("ICMP Handler: Built reply packet, total length=%d", totalLen)
+
+	// Queue the packet to be sent back through the tunnel
+	if h.proxyHandler != nil {
+		if h.proxyHandler.QueueICMPReply(pkt) {
+			logger.Info("ICMP Handler: Queued echo reply packet for transmission")
+		} else {
+			logger.Info("ICMP Handler: Failed to queue echo reply packet")
+		}
+	} else {
+		logger.Info("ICMP Handler: Cannot queue reply - proxy handler not available")
+	}
+}

netstack2/http_handler.go

@@ -0,0 +1,420 @@
+package netstack2
+
+import (
+	"bufio"
+	"context"
+	"crypto/tls"
+	"errors"
+	"fmt"
+	"net"
+	"net/http"
+	"net/http/httputil"
+	"net/url"
+	"sync"
+	"time"
+
+	"github.com/fosrl/newt/logger"
+	"gvisor.dev/gvisor/pkg/tcpip/stack"
+)
+
+// ---------------------------------------------------------------------------
+// HTTPTarget
+// ---------------------------------------------------------------------------
+
+// HTTPTarget describes a single downstream HTTP or HTTPS service that the
+// proxy should forward requests to.
+type HTTPTarget struct {
+	DestAddr string `json:"destAddr"` // IP address or hostname of the downstream service
+	DestPort uint16 `json:"destPort"` // TCP port of the downstream service
+	Scheme   string `json:"scheme"`   // When true the outbound leg uses HTTPS
+}
+
+// ---------------------------------------------------------------------------
+// HTTPHandler
+// ---------------------------------------------------------------------------
+
+// HTTPHandler intercepts TCP connections from the netstack forwarder on ports
+// 80 and 443 and services them as HTTP or HTTPS, reverse-proxying each request
+// to downstream targets specified by the matching SubnetRule.
+//
+// HTTP and raw TCP are fully separate: a connection is only routed here when
+// its SubnetRule has Protocol set ("http" or "https"). All other connections
+// on those ports fall through to the normal raw-TCP path.
+//
+// Incoming TLS termination (Protocol == "https") is performed per-connection
+// using the certificate and key stored in the rule, so different subnet rules
+// can present different certificates without sharing any state.
+//
+// Outbound connections to downstream targets honour HTTPTarget.UseHTTPS
+// independently of the incoming protocol.
+type HTTPHandler struct {
+	stack         *stack.Stack
+	proxyHandler  *ProxyHandler
+	requestLogger *HTTPRequestLogger
+
+	listener *chanListener
+	server   *http.Server
+
+	// proxyCache holds pre-built *httputil.ReverseProxy values keyed by the
+	// canonical target URL string ("scheme://host:port"). Building a proxy is
+	// cheap, but reusing one preserves the underlying http.Transport connection
+	// pool, which matters for throughput.
+	proxyCache sync.Map // map[string]*httputil.ReverseProxy
+
+	// tlsCache holds pre-parsed *tls.Config values keyed by the concatenation
+	// of the PEM certificate and key. Parsing a keypair is relatively expensive
+	// and the same cert is likely reused across many connections.
+	tlsCache sync.Map // map[string]*tls.Config
+}
+
+// ---------------------------------------------------------------------------
+// chanListener – net.Listener backed by a channel
+// ---------------------------------------------------------------------------
+
+// chanListener implements net.Listener by receiving net.Conn values over a
+// buffered channel. This lets the netstack TCP forwarder hand off connections
+// directly to a running http.Server without any real OS socket.
+type chanListener struct {
+	connCh chan net.Conn
+	closed chan struct{}
+	once   sync.Once
+}
+
+func newChanListener() *chanListener {
+	return &chanListener{
+		connCh: make(chan net.Conn, 128),
+		closed: make(chan struct{}),
+	}
+}
+
+// Accept blocks until a connection is available or the listener is closed.
+func (l *chanListener) Accept() (net.Conn, error) {
+	select {
+	case conn, ok := <-l.connCh:
+		if !ok {
+			return nil, net.ErrClosed
+		}
+		return conn, nil
+	case <-l.closed:
+		return nil, net.ErrClosed
+	}
+}
+
+// Close shuts down the listener; subsequent Accept calls return net.ErrClosed.
+func (l *chanListener) Close() error {
+	l.once.Do(func() { close(l.closed) })
+	return nil
+}
+
+// Addr returns a placeholder address (the listener has no real OS socket).
+func (l *chanListener) Addr() net.Addr {
+	return &net.TCPAddr{}
+}
+
+// send delivers conn to the listener. Returns false if the listener is already
+// closed, in which case the caller is responsible for closing conn.
+func (l *chanListener) send(conn net.Conn) bool {
+	select {
+	case l.connCh <- conn:
+		return true
+	case <-l.closed:
+		return false
+	}
+}
+
+// ---------------------------------------------------------------------------
+// httpConnCtx – conn wrapper that carries a SubnetRule through the listener
+// ---------------------------------------------------------------------------
+
+// httpConnCtx wraps a net.Conn so the matching SubnetRule and TLS state can
+// be passed through the chanListener into the http.Server's ConnContext
+// callback, making them available to request handlers via the request context.
+type httpConnCtx struct {
+	net.Conn
+	rule  *SubnetRule
+	isTLS bool // true when the conn was wrapped with tls.Server
+}
+
+// connCtxKey is the unexported context key used to store a *SubnetRule on the
+// per-connection context created by http.Server.ConnContext.
+type connCtxKey struct{}
+
+// connTLSKey is the unexported context key used to store the isTLS flag on
+// the per-connection context created by http.Server.ConnContext.
+type connTLSKey struct{}
+
+// ---------------------------------------------------------------------------
+// Constructor and lifecycle
+// ---------------------------------------------------------------------------
+
+// NewHTTPHandler creates an HTTPHandler attached to the given stack and
+// ProxyHandler. Call Start to begin serving connections.
+func NewHTTPHandler(s *stack.Stack, ph *ProxyHandler) *HTTPHandler {
+	return &HTTPHandler{
+		stack:        s,
+		proxyHandler: ph,
+	}
+}
+
+// SetRequestLogger attaches an HTTPRequestLogger so that every proxied request
+// is recorded and periodically shipped to the server.
+func (h *HTTPHandler) SetRequestLogger(rl *HTTPRequestLogger) {
+	h.requestLogger = rl
+}
+
+// Start launches the internal http.Server that services connections delivered
+// via HandleConn. The server runs for the lifetime of the HTTPHandler; call
+// Close to stop it.
+func (h *HTTPHandler) Start() error {
+	h.listener = newChanListener()
+
+	h.server = &http.Server{
+		Handler: http.HandlerFunc(h.handleRequest),
+		// ConnContext runs once per accepted connection and attaches the
+		// SubnetRule carried by httpConnCtx to the connection's context so
+		// that handleRequest can retrieve it without any global state.
+		ConnContext: func(ctx context.Context, c net.Conn) context.Context {
+			if cc, ok := c.(*httpConnCtx); ok {
+				ctx = context.WithValue(ctx, connCtxKey{}, cc.rule)
+				ctx = context.WithValue(ctx, connTLSKey{}, cc.isTLS)
+			}
+			return ctx
+		},
+	}
+
+	go func() {
+		if err := h.server.Serve(h.listener); err != nil && err != http.ErrServerClosed {
+			logger.Error("HTTP handler: server exited unexpectedly: %v", err)
+		}
+	}()
+
+	logger.Debug("HTTP handler: ready — routing determined per SubnetRule on ports 80/443")
+	return nil
+}
+
+// HandleConn accepts a TCP connection from the netstack forwarder together
+// with the SubnetRule that matched it. The HTTP handler takes full ownership
+// of the connection's lifecycle; the caller must NOT close conn after this call.
+//
+// When rule.Protocol is "https", TLS termination is performed on conn using
+// the certificate and key stored in rule.TLSCert and rule.TLSKey before the
+// connection is passed to the HTTP server. The HTTP server itself is always
+// plain-HTTP; TLS is fully unwrapped at this layer.
+func (h *HTTPHandler) HandleConn(conn net.Conn, rule *SubnetRule) {
+	var effectiveConn net.Conn = conn
+
+	if rule.Protocol == "https" {
+		// Only perform TLS termination for connections arriving on port 443.
+		// Connections on port 80 are passed through as plain HTTP so that
+		// handleRequest can issue the HTTP→HTTPS redirect.
+		doTLS := false
+		if tcpAddr, ok := conn.LocalAddr().(*net.TCPAddr); ok {
+			doTLS = tcpAddr.Port == 443
+		}
+		if doTLS {
+			tlsCfg, err := h.getTLSConfig(rule)
+			if err != nil {
+				logger.Error("HTTP handler: cannot build TLS config for connection from %s: %v",
+					conn.RemoteAddr(), err)
+				conn.Close()
+				return
+			}
+			// tls.Server wraps the raw conn; the TLS handshake is deferred until
+			// the first Read, which the http.Server will trigger naturally.
+			effectiveConn = tls.Server(conn, tlsCfg)
+		}
+	}
+
+	wrapped := &httpConnCtx{Conn: effectiveConn, rule: rule, isTLS: effectiveConn != conn}
+	if !h.listener.send(wrapped) {
+		// Listener is already closed — clean up the orphaned connection.
+		effectiveConn.Close()
+	}
+}
+
+// Close gracefully shuts down the HTTP server and the underlying channel
+// listener, causing the goroutine started in Start to exit.
+func (h *HTTPHandler) Close() error {
+	if h.server != nil {
+		if err := h.server.Close(); err != nil {
+			return err
+		}
+	}
+	if h.listener != nil {
+		h.listener.Close()
+	}
+	return nil
+}
+
+// ---------------------------------------------------------------------------
+// Internal helpers
+// ---------------------------------------------------------------------------
+
+// getTLSConfig returns a *tls.Config for the cert/key pair in rule, using a
+// cache to avoid re-parsing the same keypair on every connection.
+// The cache key is the concatenation of the PEM cert and key strings, so
+// different rules that happen to share the same material hit the same entry.
+func (h *HTTPHandler) getTLSConfig(rule *SubnetRule) (*tls.Config, error) {
+	cacheKey := rule.TLSCert + "|" + rule.TLSKey
+	if v, ok := h.tlsCache.Load(cacheKey); ok {
+		return v.(*tls.Config), nil
+	}
+
+	cert, err := tls.X509KeyPair([]byte(rule.TLSCert), []byte(rule.TLSKey))
+	if err != nil {
+		return nil, fmt.Errorf("failed to parse TLS keypair: %w", err)
+	}
+	cfg := &tls.Config{
+		Certificates: []tls.Certificate{cert},
+	}
+	// LoadOrStore is safe under concurrent calls: if two goroutines race here
+	// both will produce a valid config; the loser's work is discarded.
+	actual, _ := h.tlsCache.LoadOrStore(cacheKey, cfg)
+	return actual.(*tls.Config), nil
+}
+
+// getProxy returns a cached *httputil.ReverseProxy for the given target,
+// creating one on first use. Reusing the proxy preserves its http.Transport
+// connection pool, avoiding repeated TCP/TLS handshakes to the downstream.
+func (h *HTTPHandler) getProxy(target HTTPTarget) *httputil.ReverseProxy {
+	scheme := target.Scheme
+	cacheKey := fmt.Sprintf("%s://%s:%d", scheme, target.DestAddr, target.DestPort)
+
+	if v, ok := h.proxyCache.Load(cacheKey); ok {
+		return v.(*httputil.ReverseProxy)
+	}
+
+	targetURL := &url.URL{
+		Scheme: scheme,
+		Host:   fmt.Sprintf("%s:%d", target.DestAddr, target.DestPort),
+	}
+	var transport http.RoundTripper = http.DefaultTransport
+	if target.Scheme == "https" {
+		// Allow self-signed certificates on downstream HTTPS targets.
+		transport = &http.Transport{
+			TLSClientConfig: &tls.Config{
+				InsecureSkipVerify: true, //nolint:gosec // downstream self-signed certs are a supported configuration
+			},
+		}
+	}
+
+	proxy := &httputil.ReverseProxy{
+		Rewrite: func(pr *httputil.ProxyRequest) {
+			pr.SetURL(targetURL)
+			if host := pr.In.Host; host != "" {
+				pr.Out.Host = host
+			}
+			// SetXForwarded sets X-Forwarded-For from the inbound request's
+			// RemoteAddr (the WireGuard/netstack client address), along with
+			// X-Forwarded-Host and X-Forwarded-Proto. Using Rewrite instead of
+			// Director means the proxy does not append its own automatic
+			// X-Forwarded-For entry, so the header is set exactly once.
+			pr.SetXForwarded()
+
+			// SetXForwarded derives X-Forwarded-Proto from pr.In.TLS,
+			// which is nil because httpConnCtx wraps *tls.Conn behind
+			// net.Conn. Override using the context flag set by ConnContext.
+			if isTLS, _ := pr.In.Context().Value(connTLSKey{}).(bool); isTLS {
+				pr.Out.Header.Set("X-Forwarded-Proto", "https")
+			}
+		},
+		Transport: transport,
+	}
+
+	proxy.ErrorHandler = func(w http.ResponseWriter, r *http.Request, err error) {
+		logger.Error("HTTP handler: upstream error (%s %s -> %s): %v",
+			r.Method, r.URL.RequestURI(), cacheKey, err)
+		http.Error(w, "Bad Gateway", http.StatusBadGateway)
+	}
+
+	actual, _ := h.proxyCache.LoadOrStore(cacheKey, proxy)
+	return actual.(*httputil.ReverseProxy)
+}
+
+// statusCapture wraps an http.ResponseWriter and records the HTTP status code
+// written by the upstream handler. If WriteHeader is never called the status
+// defaults to 200 (http.StatusOK), matching net/http semantics.
+type statusCapture struct {
+	http.ResponseWriter
+	status int
+}
+
+func (sc *statusCapture) WriteHeader(code int) {
+	sc.status = code
+	sc.ResponseWriter.WriteHeader(code)
+}
+
+func (sc *statusCapture) Unwrap() http.ResponseWriter {
+	return sc.ResponseWriter
+}
+
+func (sc *statusCapture) Flush() {
+	if flusher, ok := sc.ResponseWriter.(http.Flusher); ok {
+		flusher.Flush()
+	}
+}
+
+func (sc *statusCapture) Hijack() (net.Conn, *bufio.ReadWriter, error) {
+	hijacker, ok := sc.ResponseWriter.(http.Hijacker)
+	if !ok {
+		return nil, nil, errors.New("underlying response writer does not support hijacking")
+	}
+	return hijacker.Hijack()
+}
+
+// handleRequest is the http.Handler entry point. It retrieves the SubnetRule
+// attached to the connection by ConnContext, selects the first configured
+// downstream target, and forwards the request via the cached ReverseProxy.
+//
+// TODO: add host/path-based routing across multiple HTTPTargets once the
+// configuration model evolves beyond a single target per rule.
+func (h *HTTPHandler) handleRequest(w http.ResponseWriter, r *http.Request) {
+	rule, _ := r.Context().Value(connCtxKey{}).(*SubnetRule)
+	if rule == nil || len(rule.HTTPTargets) == 0 {
+		logger.Error("HTTP handler: no downstream targets for request %s %s", r.Method, r.URL.RequestURI())
+		http.Error(w, "no targets configured", http.StatusBadGateway)
+		return
+	}
+
+	// If the rule is HTTPS but the incoming request arrived over plain HTTP
+	// (port 80), redirect to HTTPS. We use the isTLS flag stored on the
+	// connection context rather than r.TLS, because Go's http.Server calls
+	// ConnectionState() before the TLS handshake completes, so r.TLS.Version
+	// is 0 even for genuine TLS connections at that point.
+	isTLS, _ := r.Context().Value(connTLSKey{}).(bool)
+	if rule.Protocol == "https" && !isTLS {
+		host := r.Host
+		if host == "" {
+			host = r.URL.Host
+		}
+		httpsURL := "https://" + host + r.RequestURI
+		logger.Info("HTTP handler: redirecting %s %s -> %s (TLS cert present)", r.Method, r.URL.RequestURI(), httpsURL)
+		http.Redirect(w, r, httpsURL, http.StatusPermanentRedirect)
+		return
+	}
+
+	target := rule.HTTPTargets[0]
+	scheme := target.Scheme
+	logger.Info("HTTP handler: %s %s -> %s://%s:%d",
+		r.Method, r.URL.RequestURI(), scheme, target.DestAddr, target.DestPort)
+
+	timestamp := time.Now()
+	sc := &statusCapture{ResponseWriter: w, status: http.StatusOK}
+
+	h.getProxy(target).ServeHTTP(sc, r)
+
+	if h.requestLogger != nil && rule.ResourceId != 0 {
+		h.requestLogger.LogRequest(HTTPRequestLog{
+			ResourceID: rule.ResourceId,
+			Timestamp:  timestamp,
+			Method:     r.Method,
+			Scheme:     rule.Protocol,
+			Host:       r.Host,
+			Path:       r.URL.Path,
+			RawQuery:   r.URL.RawQuery,
+			UserAgent:  r.UserAgent(),
+			SourceAddr: r.RemoteAddr,
+			TLS:        rule.Protocol == "https",
+		})
+	}
+}

netstack2/http_handler_test.go

@@ -0,0 +1,97 @@
+package netstack2
+
+import (
+	"context"
+	"net"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"testing"
+
+	"github.com/gorilla/websocket"
+)
+
+func TestHTTPHandlerProxiesWebSocketUpgrade(t *testing.T) {
+	upgrader := websocket.Upgrader{}
+	backend := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		conn, err := upgrader.Upgrade(w, r, nil)
+		if err != nil {
+			t.Errorf("upgrade failed: %v", err)
+			return
+		}
+		defer conn.Close()
+
+		messageType, payload, err := conn.ReadMessage()
+		if err != nil {
+			t.Errorf("read failed: %v", err)
+			return
+		}
+		if err := conn.WriteMessage(messageType, append([]byte("echo:"), payload...)); err != nil {
+			t.Errorf("write failed: %v", err)
+		}
+	}))
+	defer backend.Close()
+
+	backendURL, err := url.Parse(backend.URL)
+	if err != nil {
+		t.Fatalf("parse backend URL: %v", err)
+	}
+	backendHost, backendPort, err := net.SplitHostPort(backendURL.Host)
+	if err != nil {
+		t.Fatalf("split backend host: %v", err)
+	}
+	port, err := net.LookupPort("tcp", backendPort)
+	if err != nil {
+		t.Fatalf("parse backend port: %v", err)
+	}
+
+	handler := NewHTTPHandler(nil, nil)
+	rule := &SubnetRule{
+		Protocol: "http",
+		HTTPTargets: []HTTPTarget{
+			{
+				DestAddr: backendHost,
+				DestPort: uint16(port),
+				Scheme:   backendURL.Scheme,
+			},
+		},
+	}
+
+	frontend := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		ctx := context.WithValue(r.Context(), connCtxKey{}, rule)
+		handler.handleRequest(w, r.WithContext(ctx))
+	}))
+	defer frontend.Close()
+
+	frontendURL, err := url.Parse(frontend.URL)
+	if err != nil {
+		t.Fatalf("parse frontend URL: %v", err)
+	}
+	wsURL := url.URL{
+		Scheme:   "ws",
+		Host:     frontendURL.Host,
+		Path:     "/socket",
+		RawQuery: "token=test",
+	}
+
+	conn, _, err := websocket.DefaultDialer.Dial(wsURL.String(), nil)
+	if err != nil {
+		t.Fatalf("dial websocket through proxy: %v", err)
+	}
+	defer conn.Close()
+
+	if err := conn.WriteMessage(websocket.TextMessage, []byte("hello")); err != nil {
+		t.Fatalf("write websocket message: %v", err)
+	}
+
+	messageType, payload, err := conn.ReadMessage()
+	if err != nil {
+		t.Fatalf("read websocket message: %v", err)
+	}
+	if messageType != websocket.TextMessage {
+		t.Fatalf("message type = %d, want %d", messageType, websocket.TextMessage)
+	}
+	if got, want := string(payload), "echo:hello"; got != want {
+		t.Fatalf("payload = %q, want %q", got, want)
+	}
+}

netstack2/http_request_log.go

@@ -0,0 +1,175 @@
+package netstack2
+
+import (
+	"bytes"
+	"compress/zlib"
+	"encoding/base64"
+	"encoding/json"
+	"sync"
+	"time"
+
+	"github.com/fosrl/newt/logger"
+)
+
+// HTTPRequestLog represents a single HTTP/HTTPS request proxied through the handler.
+type HTTPRequestLog struct {
+	RequestID  string    `json:"requestId"`
+	ResourceID int       `json:"resourceId"`
+	Timestamp  time.Time `json:"timestamp"`
+	Method     string    `json:"method"`
+	Scheme     string    `json:"scheme"`
+	Host       string    `json:"host"`
+	Path       string    `json:"path"`
+	RawQuery   string    `json:"rawQuery,omitempty"`
+	UserAgent  string    `json:"userAgent,omitempty"`
+	SourceAddr string    `json:"sourceAddr"`
+	TLS        bool      `json:"tls"`
+}
+
+// HTTPRequestLogger buffers HTTP request logs and periodically flushes them
+// to the server via a configurable SendFunc.
+type HTTPRequestLogger struct {
+	mu        sync.Mutex
+	pending   []HTTPRequestLog
+	sendFn    SendFunc
+	stopCh    chan struct{}
+	flushDone chan struct{}
+}
+
+// NewHTTPRequestLogger creates a new HTTPRequestLogger and starts its background flush loop.
+func NewHTTPRequestLogger() *HTTPRequestLogger {
+	rl := &HTTPRequestLogger{
+		pending:   make([]HTTPRequestLog, 0),
+		stopCh:    make(chan struct{}),
+		flushDone: make(chan struct{}),
+	}
+	go rl.backgroundLoop()
+	return rl
+}
+
+// SetSendFunc sets the callback used to send compressed HTTP request log batches
+// to the server. This can be called after construction once the websocket
+// client is available.
+func (rl *HTTPRequestLogger) SetSendFunc(fn SendFunc) {
+	rl.mu.Lock()
+	defer rl.mu.Unlock()
+	rl.sendFn = fn
+}
+
+// LogRequest adds an HTTP request log entry to the buffer. If the buffer
+// reaches maxBufferedSessions entries a flush is triggered immediately.
+func (rl *HTTPRequestLogger) LogRequest(log HTTPRequestLog) {
+	if log.RequestID == "" {
+		log.RequestID = generateSessionID()
+	}
+
+	rl.mu.Lock()
+	rl.pending = append(rl.pending, log)
+	shouldFlush := len(rl.pending) >= maxBufferedSessions
+	rl.mu.Unlock()
+
+	if shouldFlush {
+		rl.flush()
+	}
+}
+
+// backgroundLoop handles periodic flushing of buffered request logs.
+func (rl *HTTPRequestLogger) backgroundLoop() {
+	defer close(rl.flushDone)
+
+	ticker := time.NewTicker(flushInterval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-rl.stopCh:
+			return
+		case <-ticker.C:
+			rl.flush()
+		}
+	}
+}
+
+// flush drains the pending buffer, compresses with zlib, and sends via the SendFunc.
+// On send failure the batch is re-queued, capped at maxBufferedSessions*5 entries
+// to prevent unbounded memory growth when the server is unreachable.
+func (rl *HTTPRequestLogger) flush() {
+	rl.mu.Lock()
+	if len(rl.pending) == 0 {
+		rl.mu.Unlock()
+		return
+	}
+	batch := rl.pending
+	rl.pending = make([]HTTPRequestLog, 0)
+	sendFn := rl.sendFn
+	rl.mu.Unlock()
+
+	if sendFn == nil {
+		logger.Debug("HTTP request logger: no send function configured, discarding %d requests", len(batch))
+		return
+	}
+
+	compressed, err := compressRequestLogs(batch)
+	if err != nil {
+		logger.Error("HTTP request logger: failed to compress %d requests: %v", len(batch), err)
+		return
+	}
+
+	if err := sendFn(compressed); err != nil {
+		logger.Error("HTTP request logger: failed to send %d requests: %v", len(batch), err)
+		// Re-queue the batch so we don't lose data
+		rl.mu.Lock()
+		rl.pending = append(batch, rl.pending...)
+		// Cap re-queued data to prevent unbounded growth if server is unreachable
+		if len(rl.pending) > maxBufferedSessions*5 {
+			dropped := len(rl.pending) - maxBufferedSessions*5
+			rl.pending = rl.pending[:maxBufferedSessions*5]
+			logger.Warn("HTTP request logger: buffer overflow, dropped %d oldest requests", dropped)
+		}
+		rl.mu.Unlock()
+		return
+	}
+
+	logger.Info("HTTP request logger: sent %d requests to server", len(batch))
+}
+
+// compressRequestLogs JSON-encodes the request logs, compresses with zlib, and
+// returns a base64-encoded string suitable for embedding in a JSON message.
+func compressRequestLogs(logs []HTTPRequestLog) (string, error) {
+	jsonData, err := json.Marshal(logs)
+	if err != nil {
+		return "", err
+	}
+
+	var buf bytes.Buffer
+	w, err := zlib.NewWriterLevel(&buf, zlib.BestCompression)
+	if err != nil {
+		return "", err
+	}
+	if _, err := w.Write(jsonData); err != nil {
+		w.Close()
+		return "", err
+	}
+	if err := w.Close(); err != nil {
+		return "", err
+	}
+
+	return base64.StdEncoding.EncodeToString(buf.Bytes()), nil
+}
+
+// Close shuts down the background loop and performs one final flush to send
+// any remaining buffered requests to the server.
+func (rl *HTTPRequestLogger) Close() {
+	select {
+	case <-rl.stopCh:
+		// Already closed
+		return
+	default:
+		close(rl.stopCh)
+	}
+
+	// Wait for the background loop to exit so we don't race on flush
+	<-rl.flushDone
+
+	rl.flush()
+}

netstack2/proxy.go

@@ -6,6 +6,7 @@ import (
 	"net"
 	"net/netip"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/fosrl/newt/logger"
@@ -22,10 +23,18 @@ import (
 	"gvisor.dev/gvisor/pkg/tcpip/transport/udp"
 )
 
-// PortRange represents an allowed range of ports (inclusive)
+const (
+	// udpAccessSessionTimeout is how long a UDP access session stays alive without traffic
+	// before being considered ended by the access logger
+	udpAccessSessionTimeout = 120 * time.Second
+)
+
+// PortRange represents an allowed range of ports (inclusive) with optional protocol filtering
+// Protocol can be "tcp", "udp", or "" (empty string means both protocols)
 type PortRange struct {
-	Min uint16
-	Max uint16
+	Min      uint16
+	Max      uint16
+	Protocol string // "tcp", "udp", or "" for both
 }
 
 // SubnetRule represents a subnet with optional port restrictions and source address
@@ -41,99 +50,35 @@ type PortRange struct {
 type SubnetRule struct {
 	SourcePrefix netip.Prefix // Source IP prefix (who is sending)
 	DestPrefix   netip.Prefix // Destination IP prefix (where it's going)
+	DisableIcmp  bool         // If true, ICMP traffic is blocked for this subnet
 	RewriteTo    string       // Optional rewrite address for DNAT - can be IP/CIDR or domain name
 	PortRanges   []PortRange  // empty slice means all ports allowed
+	ResourceId   int          // Optional resource ID from the server for access logging
+
+	// HTTP proxy configuration (optional).
+	// When Protocol is non-empty the TCP connection is handled by HTTPHandler
+	// instead of the raw TCP forwarder.
+	Protocol    string       // "", "http", or "https" — controls the incoming (client-facing) protocol
+	HTTPTargets []HTTPTarget // downstream services to proxy requests to
+	TLSCert     string       // PEM-encoded certificate for incoming HTTPS termination
+	TLSKey      string       // PEM-encoded private key for incoming HTTPS termination
 }
 
-// ruleKey is used as a map key for fast O(1) lookups
-type ruleKey struct {
-	sourcePrefix string
-	destPrefix   string
-}
-
-// SubnetLookup provides fast IP subnet and port matching with O(1) lookup performance
-type SubnetLookup struct {
-	mu    sync.RWMutex
-	rules map[ruleKey]*SubnetRule // Map for O(1) lookups by prefix combination
-}
-
-// NewSubnetLookup creates a new subnet lookup table
-func NewSubnetLookup() *SubnetLookup {
-	return &SubnetLookup{
-		rules: make(map[ruleKey]*SubnetRule),
-	}
-}
-
-// AddSubnet adds a subnet rule with source and destination prefixes and optional port restrictions
-// If portRanges is nil or empty, all ports are allowed for this subnet
-// rewriteTo can be either an IP/CIDR (e.g., "192.168.1.1/32") or a domain name (e.g., "example.com")
-func (sl *SubnetLookup) AddSubnet(sourcePrefix, destPrefix netip.Prefix, rewriteTo string, portRanges []PortRange) {
-	sl.mu.Lock()
-	defer sl.mu.Unlock()
-
-	key := ruleKey{
-		sourcePrefix: sourcePrefix.String(),
-		destPrefix:   destPrefix.String(),
-	}
-
-	sl.rules[key] = &SubnetRule{
-		SourcePrefix: sourcePrefix,
-		DestPrefix:   destPrefix,
-		RewriteTo:    rewriteTo,
-		PortRanges:   portRanges,
-	}
-}
-
-// RemoveSubnet removes a subnet rule from the lookup table
-func (sl *SubnetLookup) RemoveSubnet(sourcePrefix, destPrefix netip.Prefix) {
-	sl.mu.Lock()
-	defer sl.mu.Unlock()
-
-	key := ruleKey{
-		sourcePrefix: sourcePrefix.String(),
-		destPrefix:   destPrefix.String(),
-	}
-
-	delete(sl.rules, key)
-}
-
-// Match checks if a source IP, destination IP, and port match any subnet rule
-// Returns the matched rule if BOTH:
-//   - The source IP is in the rule's source prefix
-//   - The destination IP is in the rule's destination prefix
-//   - The port is in an allowed range (or no port restrictions exist)
-//
-// Returns nil if no rule matches
-func (sl *SubnetLookup) Match(srcIP, dstIP netip.Addr, port uint16) *SubnetRule {
+// GetAllRules returns a copy of all subnet rules
+func (sl *SubnetLookup) GetAllRules() []SubnetRule {
 	sl.mu.RLock()
 	defer sl.mu.RUnlock()
 
-	// Iterate through all rules to find matching source and destination prefixes
-	// This is O(n) but necessary since we need to check prefix containment, not exact match
-	for _, rule := range sl.rules {
-		// Check if source and destination IPs match their respective prefixes
-		if !rule.SourcePrefix.Contains(srcIP) {
+	var rules []SubnetRule
+	for _, destTriePtr := range sl.sourceTrie.All() {
+		if destTriePtr == nil {
 			continue
 		}
-		if !rule.DestPrefix.Contains(dstIP) {
-			continue
-		}
-
-		// Both IPs match - now check port restrictions
-		// If no port ranges specified, all ports are allowed
-		if len(rule.PortRanges) == 0 {
-			return rule
-		}
-
-		// Check if port is in any of the allowed ranges
-		for _, pr := range rule.PortRanges {
-			if port >= pr.Min && port <= pr.Max {
-				return rule
-			}
+		for _, rule := range destTriePtr.rules {
+			rules = append(rules, *rule)
 		}
 	}
-
-	return nil
+	return rules
 }
 
 // connKey uniquely identifies a connection for NAT tracking
@@ -145,6 +90,17 @@ type connKey struct {
 	proto   uint8
 }
 
+// reverseConnKey uniquely identifies a connection for reverse NAT lookup (reply direction)
+// Key structure: (rewrittenTo, originalSrcIP, originalSrcPort, originalDstPort, proto)
+// This allows O(1) lookup of NAT entries for reply packets
+type reverseConnKey struct {
+	rewrittenTo     string // The address we rewrote to (becomes src in replies)
+	originalSrcIP   string // Original source IP (becomes dst in replies)
+	originalSrcPort uint16 // Original source port (becomes dst port in replies)
+	originalDstPort uint16 // Original destination port (becomes src port in replies)
+	proto           uint8
+}
+
 // destKey identifies a destination for handler lookups (without source port since it may change)
 type destKey struct {
 	srcIP   string
@@ -166,23 +122,33 @@ type ProxyHandler struct {
 	proxyNotifyHandle *channel.NotificationHandle
 	tcpHandler        *TCPHandler
 	udpHandler        *UDPHandler
+	icmpHandler       *ICMPHandler
+	httpHandler       *HTTPHandler
 	subnetLookup      *SubnetLookup
 	natTable          map[connKey]*natState
-	destRewriteTable  map[destKey]netip.Addr // Maps original dest to rewritten dest for handler lookups
+	reverseNatTable   map[reverseConnKey]*natState // Reverse lookup map for O(1) reply packet NAT
+	destRewriteTable  map[destKey]netip.Addr       // Maps original dest to rewritten dest for handler lookups
+	resourceTable     map[destKey]int              // Maps connection key to resource ID for access logging
 	natMu             sync.RWMutex
 	enabled           bool
+	icmpReplies       chan []byte          // Channel for ICMP reply packets to be sent back through the tunnel
+	notifiable        channel.Notification // Notification handler for triggering reads
+	accessLogger      *AccessLogger        // Access logger for tracking sessions
+	httpRequestLogger *HTTPRequestLogger   // HTTP request logger for proxied HTTP/HTTPS requests
+	blocked           atomic.Bool          // when true, all new connections are dropped
 }
 
 // ProxyHandlerOptions configures the proxy handler
 type ProxyHandlerOptions struct {
-	EnableTCP bool
-	EnableUDP bool
-	MTU       int
+	EnableTCP  bool
+	EnableUDP  bool
+	EnableICMP bool
+	MTU        int
 }
 
 // NewProxyHandler creates a new proxy handler for promiscuous mode
 func NewProxyHandler(options ProxyHandlerOptions) (*ProxyHandler, error) {
-	if !options.EnableTCP && !options.EnableUDP {
+	if !options.EnableTCP && !options.EnableUDP && !options.EnableICMP {
 		return nil, nil // No proxy needed
 	}
 
@@ -190,7 +156,11 @@ func NewProxyHandler(options ProxyHandlerOptions) (*ProxyHandler, error) {
 		enabled:          true,
 		subnetLookup:     NewSubnetLookup(),
 		natTable:         make(map[connKey]*natState),
+		reverseNatTable:  make(map[reverseConnKey]*natState),
 		destRewriteTable: make(map[destKey]netip.Addr),
+		resourceTable:    make(map[destKey]int),
+		icmpReplies:      make(chan []byte, 256), // Buffer for ICMP reply packets
+		accessLogger:     NewAccessLogger(udpAccessSessionTimeout),
 		proxyEp:          channel.New(1024, uint32(options.MTU), ""),
 		proxyStack: stack.New(stack.Options{
 			NetworkProtocols: []stack.NetworkProtocolFactory{
@@ -206,12 +176,24 @@ func NewProxyHandler(options ProxyHandlerOptions) (*ProxyHandler, error) {
 		}),
 	}
 
-	// Initialize TCP handler if enabled
+	// Initialize TCP handler if enabled. The HTTP handler piggybacks on the
+	// TCP forwarder — TCPHandler.handleTCPConn checks the subnet rule for
+	// ports 80/443 and routes matching connections to the HTTP handler, so
+	// the HTTP handler is always initialised alongside TCP.
 	if options.EnableTCP {
 		handler.tcpHandler = NewTCPHandler(handler.proxyStack, handler)
 		if err := handler.tcpHandler.InstallTCPHandler(); err != nil {
 			return nil, fmt.Errorf("failed to install TCP handler: %v", err)
 		}
+
+		handler.httpHandler = NewHTTPHandler(handler.proxyStack, handler)
+		if err := handler.httpHandler.Start(); err != nil {
+			return nil, fmt.Errorf("failed to start HTTP handler: %v", err)
+		}
+
+		handler.httpRequestLogger = NewHTTPRequestLogger()
+		handler.httpHandler.SetRequestLogger(handler.httpRequestLogger)
+		logger.Debug("ProxyHandler: HTTP handler enabled")
 	}
 
 	// Initialize UDP handler if enabled
@@ -222,6 +204,15 @@ func NewProxyHandler(options ProxyHandlerOptions) (*ProxyHandler, error) {
 		}
 	}
 
+	// Initialize ICMP handler if enabled
+	if options.EnableICMP {
+		handler.icmpHandler = NewICMPHandler(handler.proxyStack, handler)
+		if err := handler.icmpHandler.InstallICMPHandler(); err != nil {
+			return nil, fmt.Errorf("failed to install ICMP handler: %v", err)
+		}
+		logger.Debug("ProxyHandler: ICMP handler enabled")
+	}
+
 	// // Example 1: Add a rule with no port restrictions (all ports allowed)
 	// // This accepts all traffic FROM 10.0.0.0/24 TO 10.20.20.0/24
 	// sourceSubnet := netip.MustParsePrefix("10.0.0.0/24")
@@ -241,16 +232,36 @@ func NewProxyHandler(options ProxyHandlerOptions) (*ProxyHandler, error) {
 	return handler, nil
 }
 
-// AddSubnetRule adds a subnet with optional port restrictions to the proxy handler
-// sourcePrefix: The IP prefix of the peer sending the data
-// destPrefix: The IP prefix of the destination
-// rewriteTo: Optional address to rewrite destination to - can be IP/CIDR or domain name
-// If portRanges is nil or empty, all ports are allowed for this subnet
-func (p *ProxyHandler) AddSubnetRule(sourcePrefix, destPrefix netip.Prefix, rewriteTo string, portRanges []PortRange) {
+// AddSubnetRule adds a subnet rule to the proxy handler.
+// HTTP proxy behaviour is configured via rule.Protocol, rule.HTTPTargets,
+// rule.TLSCert, and rule.TLSKey; leave Protocol empty for raw TCP/UDP.
+func (p *ProxyHandler) AddSubnetRule(rule SubnetRule) {
 	if p == nil || !p.enabled {
 		return
 	}
-	p.subnetLookup.AddSubnet(sourcePrefix, destPrefix, rewriteTo, portRanges)
+	p.subnetLookup.AddSubnet(rule)
+}
+
+// SetBlocked enables or disables connection blocking on this proxy handler.
+// When enabled, all new TCP/UDP connections from the tunnel are dropped immediately.
+func (p *ProxyHandler) SetBlocked(v bool) {
+	if p == nil {
+		return
+	}
+	p.blocked.Store(v)
+	if v {
+		logger.Debug("ProxyHandler: connection blocking enabled")
+	} else {
+		logger.Debug("ProxyHandler: connection blocking disabled")
+	}
+}
+
+// IsBlocked returns true if connection blocking is currently enabled.
+func (p *ProxyHandler) IsBlocked() bool {
+	if p == nil {
+		return false
+	}
+	return p.blocked.Load()
 }
 
 // RemoveSubnetRule removes a subnet from the proxy handler
@@ -261,6 +272,69 @@ func (p *ProxyHandler) RemoveSubnetRule(sourcePrefix, destPrefix netip.Prefix) {
 	p.subnetLookup.RemoveSubnet(sourcePrefix, destPrefix)
 }
 
+// GetAllRules returns all subnet rules from the proxy handler
+func (p *ProxyHandler) GetAllRules() []SubnetRule {
+	if p == nil || !p.enabled {
+		return nil
+	}
+	return p.subnetLookup.GetAllRules()
+}
+
+// LookupResourceId looks up the resource ID for a connection
+// Returns 0 if no resource ID is associated with this connection
+func (p *ProxyHandler) LookupResourceId(srcIP, dstIP string, dstPort uint16, proto uint8) int {
+	if p == nil || !p.enabled {
+		return 0
+	}
+
+	key := destKey{
+		srcIP:   srcIP,
+		dstIP:   dstIP,
+		dstPort: dstPort,
+		proto:   proto,
+	}
+
+	p.natMu.RLock()
+	defer p.natMu.RUnlock()
+
+	return p.resourceTable[key]
+}
+
+// GetAccessLogger returns the access logger for session tracking
+func (p *ProxyHandler) GetAccessLogger() *AccessLogger {
+	if p == nil {
+		return nil
+	}
+	return p.accessLogger
+}
+
+// SetAccessLogSender configures the function used to send compressed access log
+// batches to the server. This should be called once the websocket client is available.
+func (p *ProxyHandler) SetAccessLogSender(fn SendFunc) {
+	if p == nil || !p.enabled || p.accessLogger == nil {
+		return
+	}
+	p.accessLogger.SetSendFunc(fn)
+}
+
+// GetHTTPRequestLogger returns the HTTP request logger.
+func (p *ProxyHandler) GetHTTPRequestLogger() *HTTPRequestLogger {
+	if p == nil {
+		return nil
+	}
+	return p.httpRequestLogger
+}
+
+// SetHTTPRequestLogSender configures the function used to send compressed HTTP
+// request log batches to the server. This should be called once the websocket
+// client is available.
+func (p *ProxyHandler) SetHTTPRequestLogSender(fn SendFunc) {
+	if p == nil || !p.enabled || p.httpRequestLogger == nil {
+		return
+	}
+	p.httpRequestLogger.SetSendFunc(fn)
+}
+
 // LookupDestinationRewrite looks up the rewritten destination for a connection
 // This is used by TCP/UDP handlers to find the actual target address
 func (p *ProxyHandler) LookupDestinationRewrite(srcIP, dstIP string, dstPort uint16, proto uint8) (netip.Addr, bool) {
@@ -329,6 +403,9 @@ func (p *ProxyHandler) Initialize(notifiable channel.Notification) error {
 		return nil
 	}
 
+	// Store notifiable for triggering notifications on ICMP replies
+	p.notifiable = notifiable
+
 	// Add notification handler
 	p.proxyNotifyHandle = p.proxyEp.AddNotify(notifiable)
 
@@ -407,14 +484,35 @@ func (p *ProxyHandler) HandleIncomingPacket(packet []byte) bool {
 		}
 		udpHeader := header.UDP(packet[headerLen:])
 		dstPort = udpHeader.DestinationPort()
-	default:
-		// For other protocols (ICMP, etc.), use port 0 (must match rules with no port restrictions)
+	case header.ICMPv4ProtocolNumber:
+		// ICMP doesn't have ports, use port 0 (must match rules with no port restrictions)
 		dstPort = 0
+		logger.Debug("HandleIncomingPacket: ICMP packet from %s to %s", srcAddr, dstAddr)
+	default:
+		// For other protocols, use port 0 (must match rules with no port restrictions)
+		dstPort = 0
+		logger.Debug("HandleIncomingPacket: Unknown protocol %d from %s to %s", protocol, srcAddr, dstAddr)
 	}
 
-	// Check if the source IP, destination IP, and port match any subnet rule
-	matchedRule := p.subnetLookup.Match(srcAddr, dstAddr, dstPort)
+	// Check if the source IP, destination IP, port, and protocol match any subnet rule
+	matchedRule := p.subnetLookup.Match(srcAddr, dstAddr, dstPort, protocol)
 	if matchedRule != nil {
+		logger.Debug("HandleIncomingPacket: Matched rule for %s -> %s (proto=%d, port=%d, resourceId=%d)",
+			srcAddr, dstAddr, protocol, dstPort, matchedRule.ResourceId)
+
+		// Store resource ID for connections without DNAT as well
+		if matchedRule.ResourceId != 0 && matchedRule.RewriteTo == "" {
+			dKey := destKey{
+				srcIP:   srcAddr.String(),
+				dstIP:   dstAddr.String(),
+				dstPort: dstPort,
+				proto:   uint8(protocol),
+			}
+			p.natMu.Lock()
+			p.resourceTable[dKey] = matchedRule.ResourceId
+			p.natMu.Unlock()
+		}
+
 		// Check if we need to perform DNAT
 		if matchedRule.RewriteTo != "" {
 			// Create connection tracking key using original destination
@@ -446,6 +544,13 @@ func (p *ProxyHandler) HandleIncomingPacket(packet []byte) bool {
 				proto:   uint8(protocol),
 			}
 
+			// Store resource ID for access logging if present
+			if matchedRule.ResourceId != 0 {
+				p.natMu.Lock()
+				p.resourceTable[dKey] = matchedRule.ResourceId
+				p.natMu.Unlock()
+			}
+
 			// Check if we already have a NAT entry for this connection
 			p.natMu.RLock()
 			existingEntry, exists := p.natTable[key]
@@ -472,12 +577,37 @@ func (p *ProxyHandler) HandleIncomingPacket(packet []byte) bool {
 
 				// Store NAT state for this connection
 				p.natMu.Lock()
-				p.natTable[key] = &natState{
+				natEntry := &natState{
 					originalDst: dstAddr,
 					rewrittenTo: newDst,
 				}
+				p.natTable[key] = natEntry
+
+				// Create reverse lookup key for O(1) reply packet lookups
+				// Key: (rewrittenTo, originalSrcIP, originalSrcPort, originalDstPort, proto)
+				reverseKey := reverseConnKey{
+					rewrittenTo:     newDst.String(),
+					originalSrcIP:   srcAddr.String(),
+					originalSrcPort: srcPort,
+					originalDstPort: dstPort,
+					proto:           uint8(protocol),
+				}
+				p.reverseNatTable[reverseKey] = natEntry
+
 				// Store destination rewrite for handler lookups
 				p.destRewriteTable[dKey] = newDst
+
+				// Also store the resource ID under the rewritten destination key so that
+				// TCP/UDP handlers can find it after DNAT (they see the post-NAT dst IP).
+				if matchedRule.ResourceId != 0 {
+					rewrittenKey := destKey{
+						srcIP:   srcAddr.String(),
+						dstIP:   newDst.String(),
+						dstPort: dstPort,
+						proto:   uint8(protocol),
+					}
+					p.resourceTable[rewrittenKey] = matchedRule.ResourceId
+				}
 				p.natMu.Unlock()
 				logger.Debug("New NAT entry for connection: %s -> %s", dstAddr, newDst)
 			}
@@ -501,9 +631,12 @@ func (p *ProxyHandler) HandleIncomingPacket(packet []byte) bool {
 			Payload: buffer.MakeWithData(packet),
 		})
 		p.proxyEp.InjectInbound(header.IPv4ProtocolNumber, pkb)
+		logger.Debug("HandleIncomingPacket: Injected packet into proxy stack (proto=%d)", protocol)
 		return true
 	}
 
+	// logger.Debug("HandleIncomingPacket: No matching rule for %s -> %s (proto=%d, port=%d)",
+	// srcAddr, dstAddr, protocol, dstPort)
 	return false
 }
 
@@ -626,6 +759,15 @@ func (p *ProxyHandler) ReadOutgoingPacket() *buffer.View {
 		return nil
 	}
 
+	// First check for ICMP reply packets (non-blocking)
+	select {
+	case icmpReply := <-p.icmpReplies:
+		logger.Debug("ReadOutgoingPacket: Returning ICMP reply packet (%d bytes)", len(icmpReply))
+		return buffer.NewViewWithData(icmpReply)
+	default:
+		// No ICMP reply available, continue to check proxy endpoint
+	}
+
 	pkt := p.proxyEp.Read()
 	if pkt != nil {
 		view := pkt.ToView()
@@ -655,22 +797,29 @@ func (p *ProxyHandler) ReadOutgoingPacket() *buffer.View {
 					srcPort = udpHeader.SourcePort()
 					dstPort = udpHeader.DestinationPort()
 				}
+			case header.ICMPv4ProtocolNumber:
+				// ICMP packets don't need NAT translation in our implementation
+				// since we construct reply packets with the correct addresses
+				logger.Debug("ReadOutgoingPacket: ICMP packet from %s to %s", srcIP, dstIP)
+				return view
 			}
 
-			// Look up NAT state for reverse translation
-			// The key uses the original dst (before rewrite), so for replies we need to
-			// find the entry where the rewritten address matches the current source
+			// Look up NAT state for reverse translation using O(1) reverse lookup map
+			// Key: (rewrittenTo, originalSrcIP, originalSrcPort, originalDstPort, proto)
+			// For reply packets:
+			// - reply's srcIP = rewrittenTo (the address we rewrote to)
+			// - reply's dstIP = originalSrcIP (original source IP)
+			// - reply's srcPort = originalDstPort (original destination port)
+			// - reply's dstPort = originalSrcPort (original source port)
 			p.natMu.RLock()
-			var natEntry *natState
-			for k, entry := range p.natTable {
-				// Match: reply's dst should be original src, reply's src should be rewritten dst
-				if k.srcIP == dstIP.String() && k.srcPort == dstPort &&
-					entry.rewrittenTo.String() == srcIP.String() && k.dstPort == srcPort &&
-					k.proto == uint8(protocol) {
-					natEntry = entry
-					break
-				}
+			reverseKey := reverseConnKey{
+				rewrittenTo:     srcIP.String(), // Reply's source is the rewritten address
+				originalSrcIP:   dstIP.String(), // Reply's destination is the original source
+				originalSrcPort: dstPort,        // Reply's destination port is the original source port
+				originalDstPort: srcPort,        // Reply's source port is the original destination port
+				proto:           uint8(protocol),
 			}
+			natEntry := p.reverseNatTable[reverseKey]
 			p.natMu.RUnlock()
 
 			if natEntry != nil {
@@ -688,12 +837,52 @@ func (p *ProxyHandler) ReadOutgoingPacket() *buffer.View {
 	return nil
 }
 
+// QueueICMPReply queues an ICMP reply packet to be sent back through the tunnel
+func (p *ProxyHandler) QueueICMPReply(packet []byte) bool {
+	if p == nil || !p.enabled {
+		return false
+	}
+
+	select {
+	case p.icmpReplies <- packet:
+		logger.Debug("QueueICMPReply: Queued ICMP reply packet (%d bytes)", len(packet))
+		// Trigger notification so WriteNotify picks up the packet
+		if p.notifiable != nil {
+			p.notifiable.WriteNotify()
+		}
+		return true
+	default:
+		logger.Info("QueueICMPReply: ICMP reply channel full, dropping packet")
+		return false
+	}
+}
+
 // Close cleans up the proxy handler resources
 func (p *ProxyHandler) Close() error {
 	if p == nil || !p.enabled {
 		return nil
 	}
 
+	// Shut down access logger
+	if p.accessLogger != nil {
+		p.accessLogger.Close()
+	}
+
+	// Shut down HTTP request logger
+	if p.httpRequestLogger != nil {
+		p.httpRequestLogger.Close()
+	}
+
+	// Shut down HTTP handler
+	if p.httpHandler != nil {
+		p.httpHandler.Close()
+	}
+
+	// Close ICMP replies channel
+	if p.icmpReplies != nil {
+		close(p.icmpReplies)
+	}
+
 	if p.proxyStack != nil {
 		p.proxyStack.RemoveNIC(1)
 		p.proxyStack.Close()

netstack2/subnet_lookup.go

@@ -0,0 +1,201 @@
+package netstack2
+
+import (
+	"net/netip"
+	"sync"
+
+	"github.com/gaissmai/bart"
+	"gvisor.dev/gvisor/pkg/tcpip"
+	"gvisor.dev/gvisor/pkg/tcpip/header"
+)
+
+// SubnetLookup provides fast IP subnet and port matching using BART (Binary Aggregated Range Tree)
+// This uses BART Table for O(log n) prefix matching with Supernets() for efficient lookups
+//
+// Architecture:
+// - Two-level BART structure for matching both source AND destination prefixes
+// - Level 1: Source prefix -> Level 2 (destination prefix -> rules)
+// - This reduces search space: only check destination prefixes for matching source prefixes
+type SubnetLookup struct {
+	mu sync.RWMutex
+	// Two-level BART structure:
+	// Level 1: Source prefix -> Level 2 (destination prefix -> rules)
+	// This allows us to first match source prefix, then only check destination prefixes
+	// for matching source prefixes, reducing the search space significantly
+	sourceTrie *bart.Table[*destTrie]
+}
+
+// destTrie is a BART for destination prefixes, containing the actual rules
+type destTrie struct {
+	trie  *bart.Table[[]*SubnetRule]
+	rules []*SubnetRule // All rules for this source prefix (for iteration if needed)
+}
+
+// NewSubnetLookup creates a new subnet lookup table using BART
+func NewSubnetLookup() *SubnetLookup {
+	return &SubnetLookup{
+		sourceTrie: &bart.Table[*destTrie]{},
+	}
+}
+
+// prefixEqual compares two prefixes after masking to handle host bits correctly.
+// For example, 10.0.0.5/24 and 10.0.0.0/24 are treated as equal.
+func prefixEqual(a, b netip.Prefix) bool {
+	return a.Masked() == b.Masked()
+}
+
+// AddSubnet adds a subnet rule to the lookup table.
+// If rule.PortRanges is nil or empty, all ports are allowed.
+// rule.RewriteTo can be either an IP/CIDR (e.g., "192.168.1.1/32") or a domain name (e.g., "example.com").
+// HTTP proxy behaviour is driven by rule.Protocol, rule.HTTPTargets, rule.TLSCert, and rule.TLSKey.
+func (sl *SubnetLookup) AddSubnet(rule SubnetRule) {
+	sl.mu.Lock()
+	defer sl.mu.Unlock()
+
+	rulePtr := &rule
+
+	// Canonicalize source prefix to handle host bits correctly
+	canonicalSourcePrefix := rule.SourcePrefix.Masked()
+
+	// Get or create destination trie for this source prefix
+	destTriePtr, exists := sl.sourceTrie.Get(canonicalSourcePrefix)
+	if !exists {
+		// Create new destination trie for this source prefix
+		destTriePtr = &destTrie{
+			trie:  &bart.Table[[]*SubnetRule]{},
+			rules: make([]*SubnetRule, 0),
+		}
+		sl.sourceTrie.Insert(canonicalSourcePrefix, destTriePtr)
+	}
+
+	// Canonicalize destination prefix to handle host bits correctly
+	// BART masks prefixes internally, so we need to match that behavior in our bookkeeping
+	canonicalDestPrefix := rule.DestPrefix.Masked()
+
+	// Add rule to destination trie
+	// Original behavior: overwrite if same (sourcePrefix, destPrefix) exists
+	// Store as single-element slice to match original overwrite behavior
+	destTriePtr.trie.Insert(canonicalDestPrefix, []*SubnetRule{rulePtr})
+
+	// Update destTriePtr.rules - remove old rule with same canonical prefix if exists, then add new one
+	// Use canonical comparison to handle cases like 10.0.0.5/24 vs 10.0.0.0/24
+	newRules := make([]*SubnetRule, 0, len(destTriePtr.rules)+1)
+	for _, r := range destTriePtr.rules {
+		if !prefixEqual(r.DestPrefix, canonicalDestPrefix) || !prefixEqual(r.SourcePrefix, canonicalSourcePrefix) {
+			newRules = append(newRules, r)
+		}
+	}
+	newRules = append(newRules, rulePtr)
+	destTriePtr.rules = newRules
+}
+
+// RemoveSubnet removes a subnet rule from the lookup table
+func (sl *SubnetLookup) RemoveSubnet(sourcePrefix, destPrefix netip.Prefix) {
+	sl.mu.Lock()
+	defer sl.mu.Unlock()
+
+	// Canonicalize prefixes to handle host bits correctly
+	canonicalSourcePrefix := sourcePrefix.Masked()
+	canonicalDestPrefix := destPrefix.Masked()
+
+	destTriePtr, exists := sl.sourceTrie.Get(canonicalSourcePrefix)
+	if !exists {
+		return
+	}
+
+	// Remove the rule - original behavior: delete exact (sourcePrefix, destPrefix) combination
+	// BART masks prefixes internally, so Delete works with canonical form
+	destTriePtr.trie.Delete(canonicalDestPrefix)
+
+	// Also remove from destTriePtr.rules using canonical comparison
+	// This ensures we remove rules even if they were added with host bits set
+	newDestRules := make([]*SubnetRule, 0, len(destTriePtr.rules))
+	for _, r := range destTriePtr.rules {
+		if !prefixEqual(r.DestPrefix, canonicalDestPrefix) || !prefixEqual(r.SourcePrefix, canonicalSourcePrefix) {
+			newDestRules = append(newDestRules, r)
+		}
+	}
+	destTriePtr.rules = newDestRules
+
+	// Check if the trie is actually empty using BART's Size() method
+	// This is more efficient than iterating and ensures we clean up empty tries
+	// even if there were stale entries in the rules slice (which shouldn't happen
+	// with proper canonicalization, but this provides a definitive check)
+	if destTriePtr.trie.Size() == 0 {
+		sl.sourceTrie.Delete(canonicalSourcePrefix)
+	}
+}
+
+// Match checks if a source IP, destination IP, port, and protocol match any subnet rule
+// Returns the matched rule if ALL of these conditions are met:
+//   - The source IP is in the rule's source prefix
+//   - The destination IP is in the rule's destination prefix
+//   - The port is in an allowed range (or no port restrictions exist)
+//   - The protocol matches (or the port range allows both protocols)
+//
+// proto should be header.TCPProtocolNumber, header.UDPProtocolNumber, or header.ICMPv4ProtocolNumber
+// Returns nil if no rule matches
+// This uses BART's Supernets() for O(log n) prefix matching instead of O(n) iteration
+func (sl *SubnetLookup) Match(srcIP, dstIP netip.Addr, port uint16, proto tcpip.TransportProtocolNumber) *SubnetRule {
+	sl.mu.RLock()
+	defer sl.mu.RUnlock()
+
+	// Convert IP addresses to /32 (IPv4) or /128 (IPv6) prefixes
+	// Supernets() finds all prefixes that contain this IP (i.e., are supernets of /32 or /128)
+	srcPrefix := netip.PrefixFrom(srcIP, srcIP.BitLen())
+	dstPrefix := netip.PrefixFrom(dstIP, dstIP.BitLen())
+
+	// Step 1: Find all source prefixes that contain srcIP using BART's Supernets
+	// This is O(log n) instead of O(n) iteration
+	// Supernets returns all prefixes that are supernets (contain) the given prefix
+	for _, destTriePtr := range sl.sourceTrie.Supernets(srcPrefix) {
+		if destTriePtr == nil {
+			continue
+		}
+
+		// Step 2: Find all destination prefixes that contain dstIP
+		// This is also O(log n) for each matching source prefix
+		for _, rules := range destTriePtr.trie.Supernets(dstPrefix) {
+			if rules == nil {
+				continue
+			}
+
+			// Step 3: Check each rule for ICMP and port restrictions
+			for _, rule := range rules {
+				// Handle ICMP before port range check — ICMP has no ports
+				if proto == header.ICMPv4ProtocolNumber || proto == header.ICMPv6ProtocolNumber {
+					if rule.DisableIcmp {
+						return nil
+					}
+					// ICMP is allowed; port ranges don't apply to ICMP
+					return rule
+				}
+
+				// Check port restrictions
+				if len(rule.PortRanges) == 0 {
+					// No port restrictions, match!
+					return rule
+				}
+
+				// Check if port and protocol are in any of the allowed ranges
+				for _, pr := range rule.PortRanges {
+					if port >= pr.Min && port <= pr.Max {
+						// Check protocol compatibility
+						if pr.Protocol == "" {
+							// Empty protocol means allow both TCP and UDP
+							return rule
+						}
+						// Check if the packet protocol matches the port range protocol
+						if (pr.Protocol == "tcp" && proto == header.TCPProtocolNumber) ||
+							(pr.Protocol == "udp" && proto == header.UDPProtocolNumber) {
+							return rule
+						}
+						// Port matches but protocol doesn't - continue checking other ranges
+					}
+				}
+			}
+		}
+	}
+
+	return nil
+}

netstack2/tun.go

@@ -1,8 +1,3 @@
-/* SPDX-License-Identifier: MIT
- *
- * Copyright (C) 2017-2025 WireGuard LLC. All Rights Reserved.
- */
-
 package netstack2
 
 import (
@@ -56,15 +51,17 @@ type Net netTun
 
 // NetTunOptions contains options for creating a NetTUN device
 type NetTunOptions struct {
-	EnableTCPProxy bool
-	EnableUDPProxy bool
+	EnableTCPProxy  bool
+	EnableUDPProxy  bool
+	EnableICMPProxy bool
 }
 
 // CreateNetTUN creates a new TUN device with netstack without proxying
 func CreateNetTUN(localAddresses, dnsServers []netip.Addr, mtu int) (tun.Device, *Net, error) {
 	return CreateNetTUNWithOptions(localAddresses, dnsServers, mtu, NetTunOptions{
-		EnableTCPProxy: true,
-		EnableUDPProxy: true,
+		EnableTCPProxy:  true,
+		EnableUDPProxy:  true,
+		EnableICMPProxy: true,
 	})
 }
 
@@ -84,13 +81,14 @@ func CreateNetTUNWithOptions(localAddresses, dnsServers []netip.Addr, mtu int, o
 		mtu:            mtu,
 	}
 
-	// Initialize proxy handler if TCP or UDP proxying is enabled
-	if options.EnableTCPProxy || options.EnableUDPProxy {
+	// Initialize proxy handler if TCP, UDP, or ICMP proxying is enabled
+	if options.EnableTCPProxy || options.EnableUDPProxy || options.EnableICMPProxy {
 		var err error
 		dev.proxyHandler, err = NewProxyHandler(ProxyHandlerOptions{
-			EnableTCP: options.EnableTCPProxy,
-			EnableUDP: options.EnableUDPProxy,
-			MTU:       mtu,
+			EnableTCP:  options.EnableTCPProxy,
+			EnableUDP:  options.EnableUDPProxy,
+			EnableICMP: options.EnableICMPProxy,
+			MTU:        mtu,
 		})
 		if err != nil {
 			return nil, nil, fmt.Errorf("failed to create proxy handler: %v", err)
@@ -348,13 +346,13 @@ func (net *Net) ListenUDP(laddr *net.UDPAddr) (*gonet.UDPConn, error) {
 	return net.DialUDP(laddr, nil)
 }
 
-// AddProxySubnetRule adds a subnet rule to the proxy handler
-// If portRanges is nil or empty, all ports are allowed for this subnet
-// rewriteTo can be either an IP/CIDR (e.g., "192.168.1.1/32") or a domain name (e.g., "example.com")
-func (net *Net) AddProxySubnetRule(sourcePrefix, destPrefix netip.Prefix, rewriteTo string, portRanges []PortRange) {
+// AddProxySubnetRule adds a subnet rule to the proxy handler.
+// HTTP proxy behaviour is configured via rule.Protocol, rule.HTTPTargets,
+// rule.TLSCert, and rule.TLSKey; leave Protocol empty for raw TCP/UDP.
+func (net *Net) AddProxySubnetRule(rule SubnetRule) {
 	tun := (*netTun)(net)
 	if tun.proxyHandler != nil {
-		tun.proxyHandler.AddSubnetRule(sourcePrefix, destPrefix, rewriteTo, portRanges)
+		tun.proxyHandler.AddSubnetRule(rule)
 	}
 }
 
@@ -366,6 +364,15 @@ func (net *Net) RemoveProxySubnetRule(sourcePrefix, destPrefix netip.Prefix) {
 	}
 }
 
+// GetProxySubnetRules returns all subnet rules from the proxy handler
+func (net *Net) GetProxySubnetRules() []SubnetRule {
+	tun := (*netTun)(net)
+	if tun.proxyHandler != nil {
+		return tun.proxyHandler.GetAllRules()
+	}
+	return nil
+}
+
 // GetProxyHandler returns the proxy handler (for advanced use cases)
 // Returns nil if proxy is not enabled
 func (net *Net) GetProxyHandler() *ProxyHandler {
@@ -373,6 +380,25 @@ func (net *Net) GetProxyHandler() *ProxyHandler {
 	return tun.proxyHandler
 }
 
+// SetAccessLogSender configures the function used to send compressed access log
+// batches to the server. This should be called once the websocket client is available.
+func (net *Net) SetAccessLogSender(fn SendFunc) {
+	tun := (*netTun)(net)
+	if tun.proxyHandler != nil {
+		tun.proxyHandler.SetAccessLogSender(fn)
+	}
+}
+
+// SetHTTPRequestLogSender configures the function used to send compressed HTTP
+// request log batches to the server. This should be called once the websocket
+// client is available.
+func (net *Net) SetHTTPRequestLogSender(fn SendFunc) {
+	tun := (*netTun)(net)
+	if tun.proxyHandler != nil {
+		tun.proxyHandler.SetHTTPRequestLogSender(fn)
+	}
+}
+
 type PingConn struct {
 	laddr    PingAddr
 	raddr    PingAddr

network/interface.go

@@ -44,9 +44,13 @@ func ConfigureInterface(interfaceName string, tunnelIp string, mtu int) error {
 		return configureDarwin(interfaceName, ip, ipNet)
 	case "windows":
 		return configureWindows(interfaceName, ip, ipNet)
-	default:
-		return fmt.Errorf("unsupported operating system: %s", runtime.GOOS)
+	case "android":
+		return nil
+	case "ios":
+		return nil
 	}
+
+	return nil
 }
 
 // waitForInterfaceUp polls the network interface until it's up or times out
@@ -116,7 +120,7 @@ func configureDarwin(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
 	prefix, _ := ipNet.Mask.Size()
 	ipStr := fmt.Sprintf("%s/%d", ip.String(), prefix)
 
-	cmd := exec.Command("ifconfig", interfaceName, "inet", ipStr, ip.String(), "alias")
+	cmd := exec.Command("/sbin/ifconfig", interfaceName, "inet", ipStr, ip.String(), "alias")
 	logger.Info("Running command: %v", cmd)
 
 	out, err := cmd.CombinedOutput()
@@ -125,7 +129,7 @@ func configureDarwin(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
 	}
 
 	// Bring up the interface
-	cmd = exec.Command("ifconfig", interfaceName, "up")
+	cmd = exec.Command("/sbin/ifconfig", interfaceName, "up")
 	logger.Info("Running command: %v", cmd)
 
 	out, err = cmd.CombinedOutput()

network/route.go

@@ -126,13 +126,14 @@ func LinuxRemoveRoute(destination string) error {
 
 // addRouteForServerIP adds an OS-specific route for the server IP
 func AddRouteForServerIP(serverIP, interfaceName string) error {
-	if err := AddRouteForNetworkConfig(serverIP); err != nil {
-		return err
-	}
 	if interfaceName == "" {
 		return nil
 	}
-	if runtime.GOOS == "darwin" {
+	// TODO: does this also need to be ios?
+	if runtime.GOOS == "darwin" { // macos requires routes for each peer to be added but this messes with other platforms
+		if err := AddRouteForNetworkConfig(serverIP); err != nil {
+			return err
+		}
 		return DarwinAddRoute(serverIP, "", interfaceName)
 	}
 	// else if runtime.GOOS == "windows" {
@@ -145,13 +146,14 @@ func AddRouteForServerIP(serverIP, interfaceName string) error {
 
 // removeRouteForServerIP removes an OS-specific route for the server IP
 func RemoveRouteForServerIP(serverIP string, interfaceName string) error {
-	if err := RemoveRouteForNetworkConfig(serverIP); err != nil {
-		return err
-	}
 	if interfaceName == "" {
 		return nil
 	}
-	if runtime.GOOS == "darwin" {
+	// TODO: does this also need to be ios?
+	if runtime.GOOS == "darwin" { // macos requires routes for each peer to be added but this messes with other platforms
+		if err := RemoveRouteForNetworkConfig(serverIP); err != nil {
+			return err
+		}
 		return DarwinRemoveRoute(serverIP)
 	}
 	// else if runtime.GOOS == "windows" {
@@ -217,21 +219,22 @@ func AddRoutes(remoteSubnets []string, interfaceName string) error {
 			continue
 		}
 
-		if runtime.GOOS == "darwin" {
+		switch runtime.GOOS {
+		case "darwin":
 			if err := DarwinAddRoute(subnet, "", interfaceName); err != nil {
 				logger.Error("Failed to add Darwin route for subnet %s: %v", subnet, err)
-				return err
 			}
-		} else if runtime.GOOS == "windows" {
+		case "windows":
 			if err := WindowsAddRoute(subnet, "", interfaceName); err != nil {
 				logger.Error("Failed to add Windows route for subnet %s: %v", subnet, err)
-				return err
 			}
-		} else if runtime.GOOS == "linux" {
+		case "linux":
 			if err := LinuxAddRoute(subnet, "", interfaceName); err != nil {
 				logger.Error("Failed to add Linux route for subnet %s: %v", subnet, err)
-				return err
 			}
+		case "android", "ios":
+			// Routes handled by the OS/VPN service
+			continue
 		}
 
 		logger.Info("Added route for remote subnet: %s", subnet)
@@ -258,21 +261,22 @@ func RemoveRoutes(remoteSubnets []string) error {
 		}
 
 		// Remove route based on operating system
-		if runtime.GOOS == "darwin" {
+		switch runtime.GOOS {
+		case "darwin":
 			if err := DarwinRemoveRoute(subnet); err != nil {
 				logger.Error("Failed to remove Darwin route for subnet %s: %v", subnet, err)
-				return err
 			}
-		} else if runtime.GOOS == "windows" {
+		case "windows":
 			if err := WindowsRemoveRoute(subnet); err != nil {
 				logger.Error("Failed to remove Windows route for subnet %s: %v", subnet, err)
-				return err
 			}
-		} else if runtime.GOOS == "linux" {
+		case "linux":
 			if err := LinuxRemoveRoute(subnet); err != nil {
 				logger.Error("Failed to remove Linux route for subnet %s: %v", subnet, err)
-				return err
 			}
+		case "android", "ios":
+			// Routes handled by the OS/VPN service
+			continue
 		}
 
 		logger.Info("Removed route for remote subnet: %s", subnet)

network/settings.go

@@ -115,7 +115,7 @@ func RemoveIPv4IncludedRoute(route IPv4Route) {
 		if r == route {
 			networkSettings.IPv4IncludedRoutes = append(routes[:i], routes[i+1:]...)
 			logger.Info("Removed IPv4 included route: %+v", route)
-			return
+			break
 		}
 	}
 	incrementor++

newt.REMOVED.git-id

@@ -0,0 +1 @@
+34062da7bd1b98e4a7953d2f8b43860d2222add0

newt.iss

@@ -32,7 +32,7 @@ DefaultGroupName={#MyAppName}
 DisableProgramGroupPage=yes
 ; Uncomment the following line to run in non administrative install mode (install for current user only).
 ;PrivilegesRequired=lowest
-OutputBaseFilename=mysetup
+OutputBaseFilename=newt_windows_installer
 SolidCompression=yes
 WizardStyle=modern
 ; Add this to ensure PATH changes are applied and the system is prompted for a restart if needed

packages/advantech/.gitignore

@@ -0,0 +1,3 @@
+.build
+*.tgz
+bin

packages/advantech/Makefile

@@ -0,0 +1,54 @@
+MODNAME     := newt
+VERSION     := 1.12.5
+RELEASE_URL := https://github.com/fosrl/newt/releases/download/$(VERSION)
+PLATFORM    ?= v4
+
+# Map Advantech platform to newt release binary name
+arch_v4  := arm64
+arch_v4i := arm64
+arch_v3  := arm32
+arch_v2  := arm32
+arch_v2i := arm32
+
+NEWT_ARCH := $(arch_$(PLATFORM))
+ifeq ($(NEWT_ARCH),)
+$(error Unknown platform '$(PLATFORM)'. Supported: v4, v4i, v3, v2, v2i)
+endif
+
+BINARY   := newt_linux_$(NEWT_ARCH)
+BINDIR   := bin
+OUTFILE  := $(MODNAME).$(PLATFORM).tgz
+STAGEDIR := .build/$(MODNAME)
+
+.PHONY: all clean
+
+all: $(OUTFILE)
+
+# Cache the downloaded binary in bin/ (re-download only if missing)
+$(BINDIR)/$(BINARY):
+	mkdir -p $(BINDIR)
+	@echo "Downloading $(BINARY) $(VERSION) for platform $(PLATFORM)..."
+	wget -q -O $@ "$(RELEASE_URL)/$(BINARY)" 2>/dev/null || \
+		curl -fsSL -o $@ "$(RELEASE_URL)/$(BINARY)"
+	chmod +x $@
+	@echo "Binary ready: $@"
+
+# Build the package
+$(OUTFILE): $(BINDIR)/$(BINARY) $(shell find merge -type f 2>/dev/null)
+	@rm -rf $(STAGEDIR)
+	@mkdir -p $(STAGEDIR)/bin
+	@cp $(BINDIR)/$(BINARY) $(STAGEDIR)/bin/newt
+	@chmod +x $(STAGEDIR)/bin/newt
+	@cp -r merge/. $(STAGEDIR)/
+	@chmod +x \
+		$(STAGEDIR)/etc/init \
+		$(STAGEDIR)/etc/install \
+		$(STAGEDIR)/etc/uninstall \
+		$(STAGEDIR)/www/index.cgi
+	tar -c --owner=0 --group=0 --mtime="2001-01-01 UTC" \
+		-C .build $(MODNAME) | gzip -n > $@
+	@echo "Created: $@"
+
+clean:
+	rm -rf .build $(MODNAME).*.tgz
+	@echo "Cleaned."

packages/advantech/merge/etc/defaults

@@ -0,0 +1,61 @@
+MOD_PANGOLIN_SITE_ENABLED=0
+MOD_PANGOLIN_SITE_ENDPOINT=
+MOD_PANGOLIN_SITE_ID=
+MOD_PANGOLIN_SITE_SECRET=
+
+# Core networking
+MOD_PANGOLIN_SITE_MTU=
+MOD_PANGOLIN_SITE_DNS=
+MOD_PANGOLIN_SITE_INTERFACE=
+MOD_PANGOLIN_SITE_PORT=
+MOD_PANGOLIN_SITE_UPDOWN_SCRIPT=
+MOD_PANGOLIN_SITE_PREFER_ENDPOINT=
+
+# Logging
+MOD_PANGOLIN_SITE_LOG_LEVEL=
+
+# Behavior toggles (true/false)
+MOD_PANGOLIN_SITE_DISABLE_CLIENTS=
+MOD_PANGOLIN_SITE_USE_NATIVE_INTERFACE=
+MOD_PANGOLIN_SITE_ENFORCE_HC_CERT=
+MOD_PANGOLIN_SITE_NO_CLOUD=
+
+# Timers
+MOD_PANGOLIN_SITE_PING_INTERVAL=
+MOD_PANGOLIN_SITE_PING_TIMEOUT=
+MOD_PANGOLIN_SITE_UDP_PROXY_IDLE_TIMEOUT=
+
+# Docker integration
+MOD_PANGOLIN_SITE_DOCKER_SOCKET=
+MOD_PANGOLIN_SITE_DOCKER_ENFORCE_NETWORK_VALIDATION=
+
+# Health / files
+MOD_PANGOLIN_SITE_HEALTH_FILE=
+MOD_PANGOLIN_SITE_BLUEPRINT_FILE=
+MOD_PANGOLIN_SITE_PROVISIONING_BLUEPRINT_FILE=
+MOD_PANGOLIN_SITE_CONFIG_FILE=
+
+# Provisioning
+MOD_PANGOLIN_SITE_PROVISIONING_KEY=
+MOD_PANGOLIN_SITE_NAME=
+
+# Auth daemon
+MOD_PANGOLIN_SITE_AUTH_DAEMON_ENABLED=
+MOD_PANGOLIN_SITE_AD_GENERATE_RANDOM_PASSWORD=
+MOD_PANGOLIN_SITE_AD_KEY=
+MOD_PANGOLIN_SITE_AD_PRINCIPALS_FILE=
+MOD_PANGOLIN_SITE_AD_CA_CERT_PATH=
+
+# Metrics / observability
+MOD_PANGOLIN_SITE_METRICS_PROMETHEUS_ENABLED=
+MOD_PANGOLIN_SITE_METRICS_OTLP_ENABLED=
+MOD_PANGOLIN_SITE_ADMIN_ADDR=
+MOD_PANGOLIN_SITE_REGION=
+MOD_PANGOLIN_SITE_METRICS_ASYNC_BYTES=
+MOD_PANGOLIN_SITE_PPROF_ENABLED=
+
+# mTLS
+MOD_PANGOLIN_SITE_TLS_CLIENT_CERT=
+MOD_PANGOLIN_SITE_TLS_CLIENT_KEY=
+MOD_PANGOLIN_SITE_TLS_CLIENT_CAS=
+MOD_PANGOLIN_SITE_TLS_CLIENT_CERT_PKCS12=

packages/advantech/merge/etc/description

@@ -0,0 +1,4 @@
+Pangolin Site (newt) is a WireGuard-based tunnel client that connects this router to a
+Pangolin server, enabling secure remote access to local resources without opening
+firewall ports. Configure the server endpoint, ID, and secret via the web
+interface to establish the tunnel automatically on startup.

packages/advantech/merge/etc/init

@@ -0,0 +1,124 @@
+#!/bin/sh
+
+MODNAME=newt
+PIDFILE=/tmp/newt.pid
+LOGFILE=/tmp/newt.log
+
+set_setting_raw() {
+    key="$1"
+    value="$2"
+    [ -f "/opt/$MODNAME/etc/settings" ] || cp "/opt/$MODNAME/etc/defaults" "/opt/$MODNAME/etc/settings" 2>/dev/null
+    awk -v k="$key" -v v="$value" '
+        BEGIN { done=0 }
+        index($0, k "=") == 1 { print k "=" v; done=1; next }
+        { print }
+        END { if (!done) print k "=" v }
+    ' "/opt/$MODNAME/etc/settings" > "/tmp/${MODNAME}.settings.tmp" && mv "/tmp/${MODNAME}.settings.tmp" "/opt/$MODNAME/etc/settings"
+}
+
+/usr/bin/logger -t $MODNAME "DEBUG: $0 $@"
+
+case "$1" in
+    start)
+        . /opt/$MODNAME/etc/settings
+        if [ "$MOD_PANGOLIN_SITE_ENABLED" != "1" ]; then
+            echo "$MODNAME is disabled in settings, skipping start"
+            exit 0
+        fi
+        if [ -f "$PIDFILE" ] && kill -0 "$(cat "$PIDFILE" 2>/dev/null)" 2>/dev/null; then
+            echo "$MODNAME is already running (PID: $(cat "$PIDFILE"))"
+            exit 0
+        fi
+
+        # Starting newt implies enable at boot.
+        set_setting_raw "MOD_PANGOLIN_SITE_ENABLED" "1"
+
+        # Map module settings to Newt environment variables.
+        [ -n "$MOD_PANGOLIN_SITE_ENDPOINT" ] && export PANGOLIN_ENDPOINT="$MOD_PANGOLIN_SITE_ENDPOINT"
+        [ -n "$MOD_PANGOLIN_SITE_ID" ] && export NEWT_ID="$MOD_PANGOLIN_SITE_ID"
+        [ -n "$MOD_PANGOLIN_SITE_SECRET" ] && export NEWT_SECRET="$MOD_PANGOLIN_SITE_SECRET"
+        [ -n "$MOD_PANGOLIN_SITE_MTU" ] && export MTU="$MOD_PANGOLIN_SITE_MTU"
+        [ -n "$MOD_PANGOLIN_SITE_DNS" ] && export DNS="$MOD_PANGOLIN_SITE_DNS"
+        [ -n "$MOD_PANGOLIN_SITE_LOG_LEVEL" ] && export LOG_LEVEL="$MOD_PANGOLIN_SITE_LOG_LEVEL"
+        [ -n "$MOD_PANGOLIN_SITE_UPDOWN_SCRIPT" ] && export UPDOWN_SCRIPT="$MOD_PANGOLIN_SITE_UPDOWN_SCRIPT"
+        [ -n "$MOD_PANGOLIN_SITE_INTERFACE" ] && export INTERFACE="$MOD_PANGOLIN_SITE_INTERFACE"
+        [ -n "$MOD_PANGOLIN_SITE_PORT" ] && export PORT="$MOD_PANGOLIN_SITE_PORT"
+        [ -n "$MOD_PANGOLIN_SITE_DISABLE_CLIENTS" ] && export DISABLE_CLIENTS="$MOD_PANGOLIN_SITE_DISABLE_CLIENTS"
+        [ -n "$MOD_PANGOLIN_SITE_USE_NATIVE_INTERFACE" ] && export USE_NATIVE_INTERFACE="$MOD_PANGOLIN_SITE_USE_NATIVE_INTERFACE"
+        [ -n "$MOD_PANGOLIN_SITE_ENFORCE_HC_CERT" ] && export ENFORCE_HC_CERT="$MOD_PANGOLIN_SITE_ENFORCE_HC_CERT"
+        [ -n "$MOD_PANGOLIN_SITE_DOCKER_SOCKET" ] && export DOCKER_SOCKET="$MOD_PANGOLIN_SITE_DOCKER_SOCKET"
+        [ -n "$MOD_PANGOLIN_SITE_PING_INTERVAL" ] && export PING_INTERVAL="$MOD_PANGOLIN_SITE_PING_INTERVAL"
+        [ -n "$MOD_PANGOLIN_SITE_PING_TIMEOUT" ] && export PING_TIMEOUT="$MOD_PANGOLIN_SITE_PING_TIMEOUT"
+        [ -n "$MOD_PANGOLIN_SITE_UDP_PROXY_IDLE_TIMEOUT" ] && export NEWT_UDP_PROXY_IDLE_TIMEOUT="$MOD_PANGOLIN_SITE_UDP_PROXY_IDLE_TIMEOUT"
+        [ -n "$MOD_PANGOLIN_SITE_DOCKER_ENFORCE_NETWORK_VALIDATION" ] && export DOCKER_ENFORCE_NETWORK_VALIDATION="$MOD_PANGOLIN_SITE_DOCKER_ENFORCE_NETWORK_VALIDATION"
+        [ -n "$MOD_PANGOLIN_SITE_HEALTH_FILE" ] && export HEALTH_FILE="$MOD_PANGOLIN_SITE_HEALTH_FILE"
+        [ -n "$MOD_PANGOLIN_SITE_AUTH_DAEMON_ENABLED" ] && export AUTH_DAEMON_ENABLED="$MOD_PANGOLIN_SITE_AUTH_DAEMON_ENABLED"
+        [ -n "$MOD_PANGOLIN_SITE_AD_GENERATE_RANDOM_PASSWORD" ] && export AD_GENERATE_RANDOM_PASSWORD="$MOD_PANGOLIN_SITE_AD_GENERATE_RANDOM_PASSWORD"
+        [ -n "$MOD_PANGOLIN_SITE_AD_KEY" ] && export AD_KEY="$MOD_PANGOLIN_SITE_AD_KEY"
+        [ -n "$MOD_PANGOLIN_SITE_AD_PRINCIPALS_FILE" ] && export AD_PRINCIPALS_FILE="$MOD_PANGOLIN_SITE_AD_PRINCIPALS_FILE"
+        [ -n "$MOD_PANGOLIN_SITE_AD_CA_CERT_PATH" ] && export AD_CA_CERT_PATH="$MOD_PANGOLIN_SITE_AD_CA_CERT_PATH"
+        [ -n "$MOD_PANGOLIN_SITE_METRICS_PROMETHEUS_ENABLED" ] && export NEWT_METRICS_PROMETHEUS_ENABLED="$MOD_PANGOLIN_SITE_METRICS_PROMETHEUS_ENABLED"
+        [ -n "$MOD_PANGOLIN_SITE_METRICS_OTLP_ENABLED" ] && export NEWT_METRICS_OTLP_ENABLED="$MOD_PANGOLIN_SITE_METRICS_OTLP_ENABLED"
+        [ -n "$MOD_PANGOLIN_SITE_ADMIN_ADDR" ] && export NEWT_ADMIN_ADDR="$MOD_PANGOLIN_SITE_ADMIN_ADDR"
+        [ -n "$MOD_PANGOLIN_SITE_REGION" ] && export NEWT_REGION="$MOD_PANGOLIN_SITE_REGION"
+        [ -n "$MOD_PANGOLIN_SITE_METRICS_ASYNC_BYTES" ] && export NEWT_METRICS_ASYNC_BYTES="$MOD_PANGOLIN_SITE_METRICS_ASYNC_BYTES"
+        [ -n "$MOD_PANGOLIN_SITE_PPROF_ENABLED" ] && export NEWT_PPROF_ENABLED="$MOD_PANGOLIN_SITE_PPROF_ENABLED"
+        [ -n "$MOD_PANGOLIN_SITE_TLS_CLIENT_CERT" ] && export TLS_CLIENT_CERT="$MOD_PANGOLIN_SITE_TLS_CLIENT_CERT"
+        [ -n "$MOD_PANGOLIN_SITE_TLS_CLIENT_KEY" ] && export TLS_CLIENT_KEY="$MOD_PANGOLIN_SITE_TLS_CLIENT_KEY"
+        [ -n "$MOD_PANGOLIN_SITE_TLS_CLIENT_CAS" ] && export TLS_CLIENT_CAS="$MOD_PANGOLIN_SITE_TLS_CLIENT_CAS"
+        [ -n "$MOD_PANGOLIN_SITE_TLS_CLIENT_CERT_PKCS12" ] && export TLS_CLIENT_CERT_PKCS12="$MOD_PANGOLIN_SITE_TLS_CLIENT_CERT_PKCS12"
+        [ -n "$MOD_PANGOLIN_SITE_BLUEPRINT_FILE" ] && export BLUEPRINT_FILE="$MOD_PANGOLIN_SITE_BLUEPRINT_FILE"
+        [ -n "$MOD_PANGOLIN_SITE_PROVISIONING_BLUEPRINT_FILE" ] && export PROVISIONING_BLUEPRINT_FILE="$MOD_PANGOLIN_SITE_PROVISIONING_BLUEPRINT_FILE"
+        [ -n "$MOD_PANGOLIN_SITE_NO_CLOUD" ] && export NO_CLOUD="$MOD_PANGOLIN_SITE_NO_CLOUD"
+        [ -n "$MOD_PANGOLIN_SITE_PROVISIONING_KEY" ] && export NEWT_PROVISIONING_KEY="$MOD_PANGOLIN_SITE_PROVISIONING_KEY"
+        [ -n "$MOD_PANGOLIN_SITE_NAME" ] && export NEWT_NAME="$MOD_PANGOLIN_SITE_NAME"
+        [ -n "$MOD_PANGOLIN_SITE_CONFIG_FILE" ] && export CONFIG_FILE="$MOD_PANGOLIN_SITE_CONFIG_FILE"
+
+        export NEWT_SYSTEM_SUBSTRATE="ADVANTECH_ROUTER_APP"
+        export NEWT_PID_FILE="$PIDFILE"
+
+        echo "Starting $MODNAME..."
+        if [ -n "$MOD_PANGOLIN_SITE_PREFER_ENDPOINT" ]; then
+            /opt/$MODNAME/bin/newt --prefer-endpoint "$MOD_PANGOLIN_SITE_PREFER_ENDPOINT" >> "$LOGFILE" 2>&1 &
+        else
+            /opt/$MODNAME/bin/newt >> "$LOGFILE" 2>&1 &
+        fi
+        echo $! > "$PIDFILE"
+        echo "Started $MODNAME (PID: $(cat "$PIDFILE"))"
+        exit 0
+        ;;
+    stop)
+        echo "Stopping $MODNAME..."
+        if [ -f "$PIDFILE" ]; then
+            kill "$(cat "$PIDFILE" 2>/dev/null)" 2>/dev/null
+            rm -f "$PIDFILE"
+        fi
+        killall newt 2>/dev/null
+        # Stopping newt implies disable at boot.
+        set_setting_raw "MOD_PANGOLIN_SITE_ENABLED" "0"
+        echo "Stopped $MODNAME"
+        exit 0
+        ;;
+    restart)
+        $0 stop
+        sleep 1
+        # Restart should remain enabled after reboot.
+        set_setting_raw "MOD_PANGOLIN_SITE_ENABLED" "1"
+        $0 start
+        ;;
+    status)
+        if [ -f "$PIDFILE" ] && kill -0 "$(cat "$PIDFILE" 2>/dev/null)" 2>/dev/null; then
+            echo "$MODNAME is running (PID: $(cat "$PIDFILE"))"
+            exit 0
+        else
+            echo "$MODNAME is not running"
+            exit 1
+        fi
+        ;;
+    defaults)
+        cp /opt/$MODNAME/etc/defaults /opt/$MODNAME/etc/settings 2>/dev/null
+        ;;
+    *)
+        echo "Usage: $0 {start|stop|restart|status|defaults}"
+        exit 1
+esac

packages/advantech/merge/etc/install

@@ -0,0 +1,15 @@
+#!/bin/sh
+
+MODNAME=newt
+
+# Ensure scripts are executable
+chmod +x /opt/$MODNAME/etc/init
+chmod +x /opt/$MODNAME/etc/install
+chmod +x /opt/$MODNAME/etc/uninstall
+chmod +x /opt/$MODNAME/bin/newt
+chmod +x /opt/$MODNAME/www/index.cgi
+
+# Secure the web interface with router authentication
+ln -sf /etc/htpasswd /opt/$MODNAME/www/.htpasswd
+
+exit 0

packages/advantech/merge/etc/name

@@ -0,0 +1 @@
+Pangolin Site

packages/advantech/merge/etc/summary

@@ -0,0 +1 @@
+Tunnel client for Pangolin secure remote access.

packages/advantech/merge/etc/uninstall

@@ -0,0 +1,9 @@
+#!/bin/sh
+
+MODNAME=newt
+
+rm -f /opt/$MODNAME/www/.htpasswd
+rm -f /tmp/newt.pid
+rm -f /tmp/newt.log
+
+exit 0

packages/advantech/merge/etc/version

@@ -0,0 +1 @@
+1.12.5

packages/advantech/merge/www/index.cgi

@@ -0,0 +1,282 @@
+#!/bin/sh
+
+MODNAME=newt
+SETTINGS=/opt/$MODNAME/etc/settings
+LOGFILE=/tmp/newt.log
+PIDFILE=/tmp/newt.pid
+
+# Escape special HTML characters
+htmlesc() {
+    printf '%s' "$1" | sed \
+        -e 's/&/\&/g' \
+        -e 's/</\&lt;/g' \
+        -e 's/>/\&gt;/g' \
+        -e 's/"/\&quot;/g'
+}
+
+# Decode URL-encoded form values (handles common chars in endpoints/ids/secrets)
+urldecode() {
+    printf '%s' "$1" | sed \
+        -e 's/+/ /g' \
+        -e 's/%3[Aa]/:/g' -e 's/%3[aa]/:/g' \
+        -e 's/%2[Ff]/\//g' -e 's/%2[ff]/\//g' \
+        -e 's/%40/@/g' \
+        -e 's/%2[Ee]/./g' \
+        -e 's/%2[Dd]/-/g' \
+        -e 's/%5[Ff]/_/g' \
+        -e 's/%3[Dd]/=/g' \
+        -e 's/%3[Ff]/?/g' \
+        -e 's/%23/#/g' \
+        -e 's/%25/%/g'
+}
+
+# Extract a named field from URL-encoded POST data (awk splits on & without tr)
+get_field() {
+    printf '%s' "$2" | awk -v f="${1}=" 'BEGIN{RS="&"} index($0,f)==1 {print substr($0,length(f)+1); exit}'
+}
+
+# Check whether newt process is alive
+is_running() {
+    [ -f "$PIDFILE" ] && kill -0 "$(cat "$PIDFILE" 2>/dev/null)" 2>/dev/null
+}
+
+ensure_settings_file() {
+  [ -f "$SETTINGS" ] && return
+  if [ -f "/opt/$MODNAME/etc/defaults" ]; then
+    cp "/opt/$MODNAME/etc/defaults" "$SETTINGS" 2>/dev/null
+  else
+    printf 'MOD_PANGOLIN_SITE_ENABLED=0\n' > "$SETTINGS"
+  fi
+}
+
+set_setting_raw() {
+  key="$1"
+  value="$2"
+  ensure_settings_file
+  awk -v k="$key" -v v="$value" '
+    BEGIN { done=0 }
+    index($0, k "=") == 1 { print k "=" v; done=1; next }
+    { print }
+    END { if (!done) print k "=" v }
+  ' "$SETTINGS" > "$SETTINGS.tmp" && mv "$SETTINGS.tmp" "$SETTINGS"
+}
+
+quote_sh_value() {
+  printf '%s' "$1" | sed -e 's/\\/\\\\/g' -e 's/"/\\"/g'
+}
+
+# ── Read POST body ──────────────────────────────────────────────────────────
+POST_DATA=""
+if [ "$REQUEST_METHOD" = "POST" ] && [ -n "$CONTENT_LENGTH" ] && [ "$CONTENT_LENGTH" -gt 0 ] 2>/dev/null; then
+    POST_DATA=$(dd bs="$CONTENT_LENGTH" count=1 2>/dev/null)
+fi
+
+# ── Load current settings ───────────────────────────────────────────────────
+MOD_PANGOLIN_SITE_ENABLED=0
+MOD_PANGOLIN_SITE_ENDPOINT=""
+MOD_PANGOLIN_SITE_ID=""
+MOD_PANGOLIN_SITE_SECRET=""
+[ -f "$SETTINGS" ] && . "$SETTINGS"
+
+# ── Handle actions ──────────────────────────────────────────────────────────
+if [ -n "$POST_DATA" ]; then
+    ACTION=$(get_field "action" "$POST_DATA")
+    case "$ACTION" in
+        save)
+            EP=$(urldecode "$(get_field "endpoint" "$POST_DATA")")
+            ID=$(urldecode "$(get_field "id"       "$POST_DATA")")
+            SEC=$(urldecode "$(get_field "secret"  "$POST_DATA")")
+        set_setting_raw "MOD_PANGOLIN_SITE_ENDPOINT" "\"$(quote_sh_value "$EP")\""
+        set_setting_raw "MOD_PANGOLIN_SITE_ID" "\"$(quote_sh_value "$ID")\""
+        set_setting_raw "MOD_PANGOLIN_SITE_SECRET" "\"$(quote_sh_value "$SEC")\""
+            ;;
+        start)
+        set_setting_raw "MOD_PANGOLIN_SITE_ENABLED" "1"
+            /opt/$MODNAME/etc/init start >/dev/null 2>&1
+            ;;
+        stop)
+            /opt/$MODNAME/etc/init stop >/dev/null 2>&1
+        set_setting_raw "MOD_PANGOLIN_SITE_ENABLED" "0"
+            ;;
+        restart)
+        set_setting_raw "MOD_PANGOLIN_SITE_ENABLED" "1"
+            /opt/$MODNAME/etc/init restart >/dev/null 2>&1
+            ;;
+        clearlog)
+            printf '' > "$LOGFILE"
+            ;;
+    esac
+fi
+
+# Reload settings after actions.
+MOD_PANGOLIN_SITE_ENABLED=0
+MOD_PANGOLIN_SITE_ENDPOINT=""
+MOD_PANGOLIN_SITE_ID=""
+MOD_PANGOLIN_SITE_SECRET=""
+[ -f "$SETTINGS" ] && . "$SETTINGS"
+
+# ── Status ──────────────────────────────────────────────────────────────────
+if is_running; then
+    STATUS_TEXT="Running"
+    STATUS_CLASS="running"
+    DISABLED="disabled"
+  SETTINGS_HINT='<div class="notice">Stop Newt first to edit settings.</div>'
+else
+    STATUS_TEXT="Stopped"
+    STATUS_CLASS="stopped"
+    DISABLED=""
+    SETTINGS_HINT=""
+fi
+
+EP_ESC=$(htmlesc  "$MOD_PANGOLIN_SITE_ENDPOINT")
+ID_ESC=$(htmlesc  "$MOD_PANGOLIN_SITE_ID")
+SEC_ESC=$(htmlesc "$MOD_PANGOLIN_SITE_SECRET")
+EP_INPUT_ESC="$EP_ESC"
+[ -z "$MOD_PANGOLIN_SITE_ENDPOINT" ] && EP_INPUT_ESC="https://app.pangolin.net"
+
+# ── Log (escape HTML and dollar signs to prevent shell expansion in heredoc) ─
+LOG_HTML=""
+if [ -f "$LOGFILE" ]; then
+    LOG_HTML=$(tail -100 "$LOGFILE" 2>/dev/null | sed \
+        -e 's/&/\&amp;/g' \
+        -e 's/</\&lt;/g'  \
+        -e 's/>/\&gt;/g'  \
+        -e 's/\$/\&#36;/g')
+fi
+
+# ── Output ──────────────────────────────────────────────────────────────────
+printf 'Content-type: text/html\r\n\r\n'
+
+# Static head — split around the conditional refresh meta tag
+cat << 'HEAD_START'
+<!DOCTYPE html>
+<html>
+<head>
+<meta charset="UTF-8">
+HEAD_START
+
+# Only auto-refresh while newt is running (avoids wiping settings form mid-edit)
+[ "$STATUS_CLASS" = "running" ] && printf '<meta http-equiv="refresh" content="10">\n'
+
+cat << 'STATIC_HEAD'
+<title>Pangolin Site</title>
+<style>
+*{box-sizing:border-box}
+body{margin:0;padding:8px;border-top:3px solid #79BD28;background:#fff;color:#000}
+body,table,tr,td,a,input,select,textarea,button{font-family:Verdana,Arial,Helvetica,sans-serif;font-size:12px}
+a{color:#004280;text-decoration:none}
+a:hover{text-decoration:underline}
+.topline{display:flex;align-items:center;justify-content:space-between;gap:8px;flex-wrap:wrap;margin-bottom:6px}
+.app-title{background:#004280;color:#fff;font-weight:bold;padding:6px 8px;border:1px solid #00315f;border-bottom:none}
+.window{background:#f4f4f4;border:2px solid #004280;padding:0}
+.section{border-top:1px solid #9fb7d5}
+.section:first-child{border-top:none}
+.section-head{background:#c0e0ff;color:#000;font-weight:bold;padding:5px 8px;border-bottom:1px solid #9fb7d5;display:flex;align-items:center;justify-content:space-between;gap:8px;flex-wrap:wrap}
+.section-body{padding:10px 8px}
+.badge{display:inline-block;padding:2px 10px;border-radius:2px;color:#fff;font-weight:bold;line-height:1.4}
+.running{background:#008000}
+.stopped{background:#800000}
+.row{display:flex;align-items:center;gap:8px;margin:6px 0}
+.row label{width:90px;font-weight:bold}
+.row input{width:360px;max-width:100%;padding:2px 4px;border:1px solid #7f9db9;background:#fff;height:23px}
+.btn{height:24px;padding:0 10px;border:1px solid #7f9db9;background:#efefef;color:#000;cursor:pointer}
+.btn:disabled{color:#777;background:#e5e5e5}
+.btn + .btn{margin-left:6px}
+.btn-start{border-color:#2d7a2d;background:#dff0df}
+.btn-stop{border-color:#8a2c2c;background:#f7dddd}
+.btn-restart{border-color:#9a6a1f;background:#f7ecd8}
+.btn-save{border-color:#2e5f92;background:#dce9f8}
+.hint{color:#808080;font-weight:normal}
+.notice{margin:0 0 10px 0;color:#ff0000;font-style:italic}
+.log{background:#fff;border:1px solid #7f9db9;color:#000;font-family:monospace;font-size:12px;line-height:1.35;padding:6px;height:300px;overflow-y:auto;white-space:pre-wrap;word-break:break-word}
+</style>
+</head>
+<body>
+<div class="topline">
+<a href="/">&laquo; Back to Router</a>
+</div>
+<div class="app-title">Router Apps - Pangolin Site</div>
+<div class="window">
+STATIC_HEAD
+
+# Status card (double-quoted heredoc — variables expand)
+cat << STATUS_CARD
+<div class="section">
+<div class="section-head">Status</div>
+<div class="section-body">
+<div><span class="badge ${STATUS_CLASS}">${STATUS_TEXT}</span></div>
+<div style="margin-top:8px">
+<form method="post" style="display:inline">
+  <input type="hidden" name="action" value="start">
+  <button class="btn btn-start" type="submit">Start</button>
+</form>
+<form method="post" style="display:inline">
+  <input type="hidden" name="action" value="stop">
+  <button class="btn btn-stop" type="submit">Stop</button>
+</form>
+<form method="post" style="display:inline">
+  <input type="hidden" name="action" value="restart">
+  <button class="btn btn-restart" type="submit">Restart</button>
+</form>
+</div>
+</div>
+</div>
+STATUS_CARD
+
+# Settings card
+cat << SETTINGS_CARD
+<div class="section">
+<div class="section-head">Settings</div>
+<div class="section-body">
+${SETTINGS_HINT}<form method="post">
+  <input type="hidden" name="action" value="save">
+  <div class="row">
+    <label>Endpoint</label>
+    <input type="text" name="endpoint" value="${EP_INPUT_ESC}" ${DISABLED}>
+  </div>
+  <div class="row">
+    <label>ID</label>
+    <input type="text" name="id" value="${ID_ESC}" ${DISABLED}>
+  </div>
+  <div class="row">
+    <label>Secret</label>
+    <input type="password" name="secret" value="${SEC_ESC}" ${DISABLED}>
+  </div>
+  <div style="margin-top:12px">
+    <button class="btn btn-save" type="submit" ${DISABLED}>Save</button>
+  </div>
+</form>
+</div>
+</div>
+SETTINGS_CARD
+
+# Log card header
+cat << 'LOG_HEADER'
+<div class="section">
+<div class="section-head">
+<span>Log <span class="hint">(last 100 lines, auto-refreshes every 10s while running)</span></span>
+<span style="display:inline-flex;gap:6px">
+<form method="post" style="margin:0">
+  <input type="hidden" name="action" value="clearlog">
+  <button class="btn" type="submit">Clear Log</button>
+</form>
+<form method="get" style="margin:0">
+  <button class="btn" type="submit">Refresh</button>
+</form>
+</span>
+</div>
+<div class="section-body">
+<div class="log">
+LOG_HEADER
+
+# Log content — printed with printf to prevent any shell expansion
+printf '%s' "$LOG_HTML"
+
+# Close tags (static)
+cat << 'STATIC_FOOTER'
+</div>
+</div>
+</div>
+</body>
+</html>
+STATIC_FOOTER

proxy/manager.go

@@ -21,7 +21,32 @@ import (
 	"gvisor.dev/gvisor/pkg/tcpip/adapters/gonet"
 )
 
-const errUnsupportedProtoFmt = "unsupported protocol: %s"
+const (
+	errUnsupportedProtoFmt = "unsupported protocol: %s"
+	maxUDPPacketSize       = 65507 // Maximum UDP packet size
+	defaultUDPIdleTimeout  = 90 * time.Second
+)
+
+// udpBufferPool provides reusable buffers for UDP packet handling.
+// This reduces GC pressure from frequent large allocations.
+var udpBufferPool = sync.Pool{
+	New: func() any {
+		buf := make([]byte, maxUDPPacketSize)
+		return &buf
+	},
+}
+
+// getUDPBuffer retrieves a buffer from the pool.
+func getUDPBuffer() *[]byte {
+	return udpBufferPool.Get().(*[]byte)
+}
+
+// putUDPBuffer clears and returns a buffer to the pool.
+func putUDPBuffer(buf *[]byte) {
+	// Clear the buffer to prevent data leakage
+	clear(*buf)
+	udpBufferPool.Put(buf)
+}
 
 // Target represents a proxy target with its address and port
 type Target struct {
@@ -44,6 +69,10 @@ type ProxyManager struct {
 	tunnels         map[string]*tunnelEntry
 	asyncBytes      bool
 	flushStop       chan struct{}
+	udpIdleTimeout  time.Duration
+
+	// connection blocking
+	blocked atomic.Bool
 }
 
 // tunnelEntry holds per-tunnel attributes and (optional) async counters.
@@ -105,13 +134,9 @@ func classifyProxyError(err error) string {
 	if errors.Is(err, net.ErrClosed) {
 		return "closed"
 	}
-	if ne, ok := err.(net.Error); ok {
-		if ne.Timeout() {
-			return "timeout"
-		}
-		if ne.Temporary() {
-			return "temporary"
-		}
+	var ne net.Error
+	if errors.As(err, &ne) && ne.Timeout() {
+		return "timeout"
 	}
 	msg := strings.ToLower(err.Error())
 	switch {
@@ -127,12 +152,13 @@ func classifyProxyError(err error) string {
 // NewProxyManager creates a new proxy manager instance
 func NewProxyManager(tnet *netstack.Net) *ProxyManager {
 	return &ProxyManager{
-		tnet:       tnet,
-		tcpTargets: make(map[string]map[int]string),
-		udpTargets: make(map[string]map[int]string),
-		listeners:  make([]*gonet.TCPListener, 0),
-		udpConns:   make([]*gonet.UDPConn, 0),
-		tunnels:    make(map[string]*tunnelEntry),
+		tnet:           tnet,
+		tcpTargets:     make(map[string]map[int]string),
+		udpTargets:     make(map[string]map[int]string),
+		listeners:      make([]*gonet.TCPListener, 0),
+		udpConns:       make([]*gonet.UDPConn, 0),
+		tunnels:        make(map[string]*tunnelEntry),
+		udpIdleTimeout: defaultUDPIdleTimeout,
 	}
 }
 
@@ -206,10 +232,11 @@ func (pm *ProxyManager) ClearTunnelID() {
 // init function without tnet
 func NewProxyManagerWithoutTNet() *ProxyManager {
 	return &ProxyManager{
-		tcpTargets: make(map[string]map[int]string),
-		udpTargets: make(map[string]map[int]string),
-		listeners:  make([]*gonet.TCPListener, 0),
-		udpConns:   make([]*gonet.UDPConn, 0),
+		tcpTargets:     make(map[string]map[int]string),
+		udpTargets:     make(map[string]map[int]string),
+		listeners:      make([]*gonet.TCPListener, 0),
+		udpConns:       make([]*gonet.UDPConn, 0),
+		udpIdleTimeout: defaultUDPIdleTimeout,
 	}
 }
 
@@ -220,6 +247,18 @@ func (pm *ProxyManager) SetTNet(tnet *netstack.Net) {
 	pm.tnet = tnet
 }
 
+// SetBlocked enables or disables connection blocking.
+// When enabled, all new incoming TCP connections are immediately closed
+// and all incoming UDP packets are silently dropped.
+func (pm *ProxyManager) SetBlocked(v bool) {
+	pm.blocked.Store(v)
+	if v {
+		logger.Debug("ProxyManager: connection blocking enabled, new connections will be dropped")
+	} else {
+		logger.Debug("ProxyManager: connection blocking disabled, accepting connections")
+	}
+}
+
 // AddTarget adds as new target for proxying
 func (pm *ProxyManager) AddTarget(proto, listenIP string, port int, targetAddr string) error {
 	pm.mutex.Lock()
@@ -346,6 +385,17 @@ func (pm *ProxyManager) SetAsyncBytes(b bool) {
 		go pm.flushLoop()
 	}
 }
+
+// SetUDPIdleTimeout configures when idle UDP client flows are reclaimed.
+func (pm *ProxyManager) SetUDPIdleTimeout(d time.Duration) {
+	pm.mutex.Lock()
+	defer pm.mutex.Unlock()
+	if d <= 0 {
+		pm.udpIdleTimeout = defaultUDPIdleTimeout
+		return
+	}
+	pm.udpIdleTimeout = d
+}
 func (pm *ProxyManager) flushLoop() {
 	flushInterval := 2 * time.Second
 	if v := os.Getenv("OTEL_METRIC_EXPORT_INTERVAL"); v != "" {
@@ -437,14 +487,6 @@ func (pm *ProxyManager) Stop() error {
 		pm.udpConns = append(pm.udpConns[:i], pm.udpConns[i+1:]...)
 	}
 
-	// // Clear the target maps
-	// for k := range pm.tcpTargets {
-	// 	delete(pm.tcpTargets, k)
-	// }
-	// for k := range pm.udpTargets {
-	// 	delete(pm.udpTargets, k)
-	// }
-
 	// Give active connections a chance to close gracefully
 	time.Sleep(100 * time.Millisecond)
 
@@ -498,7 +540,7 @@ func (pm *ProxyManager) handleTCPProxy(listener net.Listener, targetAddr string)
 			if !pm.running {
 				return
 			}
-			if ne, ok := err.(net.Error); ok && !ne.Temporary() {
+			if errors.Is(err, net.ErrClosed) {
 				logger.Info("TCP listener closed, stopping proxy handler for %v", listener.Addr())
 				return
 			}
@@ -508,6 +550,12 @@ func (pm *ProxyManager) handleTCPProxy(listener net.Listener, targetAddr string)
 		}
 
 		tunnelID := pm.currentTunnelID
+		// Drop connection if blocking is enabled
+		if pm.blocked.Load() {
+			conn.Close()
+			logger.Debug("TCP proxy: connection dropped (blocking enabled)")
+			continue
+		}
 		telemetry.IncProxyAccept(context.Background(), tunnelID, "tcp", "success", "")
 		telemetry.IncProxyConnectionEvent(context.Background(), tunnelID, "tcp", telemetry.ProxyConnectionOpened)
 		if tunnelID != "" {
@@ -564,7 +612,9 @@ func (pm *ProxyManager) handleTCPProxy(listener net.Listener, targetAddr string)
 }
 
 func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
-	buffer := make([]byte, 65507) // Max UDP packet size
+	bufPtr := getUDPBuffer()
+	defer putUDPBuffer(bufPtr)
+	buffer := *bufPtr
 	clientConns := make(map[string]*net.UDPConn)
 	var clientsMutex sync.RWMutex
 
@@ -583,7 +633,7 @@ func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
 			}
 
 			// Check for connection closed conditions
-			if err == io.EOF || strings.Contains(err.Error(), "use of closed network connection") {
+			if errors.Is(err, io.EOF) || errors.Is(err, net.ErrClosed) {
 				logger.Info("UDP connection closed, stopping proxy handler")
 
 				// Clean up existing client connections
@@ -602,6 +652,11 @@ func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
 		}
 
 		clientKey := remoteAddr.String()
+		// Drop packet if blocking is enabled
+		if pm.blocked.Load() {
+			logger.Debug("UDP proxy: packet dropped (blocking enabled)")
+			continue
+		}
 		// bytes from client -> target (direction=in)
 		if pm.currentTunnelID != "" && n > 0 {
 			if pm.asyncBytes {
@@ -632,6 +687,9 @@ func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
 				telemetry.IncProxyAccept(context.Background(), pm.currentTunnelID, "udp", "failure", classifyProxyError(err))
 				continue
 			}
+			// Prevent idle UDP client goroutines from living forever and
+			// retaining large per-connection buffers.
+			_ = targetConn.SetReadDeadline(time.Now().Add(pm.udpIdleTimeout))
 			tunnelID := pm.currentTunnelID
 			telemetry.IncProxyAccept(context.Background(), tunnelID, "udp", "success", "")
 			telemetry.IncProxyConnectionEvent(context.Background(), tunnelID, "udp", telemetry.ProxyConnectionOpened)
@@ -647,7 +705,10 @@ func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
 			go func(clientKey string, targetConn *net.UDPConn, remoteAddr net.Addr, tunnelID string) {
 				start := time.Now()
 				result := "success"
+				bufPtr := getUDPBuffer()
 				defer func() {
+					// Return buffer to pool first
+					putUDPBuffer(bufPtr)
 					// Always clean up when this goroutine exits
 					clientsMutex.Lock()
 					if storedConn, exists := clientConns[clientKey]; exists && storedConn == targetConn {
@@ -662,10 +723,18 @@ func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
 					telemetry.IncProxyConnectionEvent(context.Background(), tunnelID, "udp", telemetry.ProxyConnectionClosed)
 				}()
 
-				buffer := make([]byte, 65507)
+				buffer := *bufPtr
 				for {
 					n, _, err := targetConn.ReadFromUDP(buffer)
 					if err != nil {
+						var netErr net.Error
+						if errors.As(err, &netErr) && netErr.Timeout() {
+							return
+						}
+						// Connection closed is normal during cleanup
+						if errors.Is(err, net.ErrClosed) || errors.Is(err, io.EOF) {
+							return // defer will handle cleanup, result stays "success"
+						}
 						logger.Error("Error reading from target: %v", err)
 						result = "failure"
 						return // defer will handle cleanup
@@ -704,6 +773,8 @@ func (pm *ProxyManager) handleUDPProxy(conn *gonet.UDPConn, targetAddr string) {
 			delete(clientConns, clientKey)
 			clientsMutex.Unlock()
 		} else if pm.currentTunnelID != "" && written > 0 {
+			// Extend idle timeout whenever client traffic is observed.
+			_ = targetConn.SetReadDeadline(time.Now().Add(pm.udpIdleTimeout))
 			if pm.asyncBytes {
 				if e := pm.getEntry(pm.currentTunnelID); e != nil {
 					e.bytesInUDP.Add(uint64(written))
@@ -736,3 +807,28 @@ func (pm *ProxyManager) PrintTargets() {
 		}
 	}
 }
+
+// GetTargets returns a copy of the current TCP and UDP targets
+// Returns map[listenIP]map[port]targetAddress for both TCP and UDP
+func (pm *ProxyManager) GetTargets() (tcpTargets map[string]map[int]string, udpTargets map[string]map[int]string) {
+	pm.mutex.RLock()
+	defer pm.mutex.RUnlock()
+
+	tcpTargets = make(map[string]map[int]string)
+	for listenIP, targets := range pm.tcpTargets {
+		tcpTargets[listenIP] = make(map[int]string)
+		for port, targetAddr := range targets {
+			tcpTargets[listenIP][port] = targetAddr
+		}
+	}
+
+	udpTargets = make(map[string]map[int]string)
+	for listenIP, targets := range pm.udpTargets {
+		udpTargets[listenIP] = make(map[int]string)
+		for port, targetAddr := range targets {
+			udpTargets[listenIP][port] = targetAddr
+		}
+	}
+
+	return tcpTargets, udpTargets
+}

reexec_unix.go

@@ -0,0 +1,21 @@
+//go:build !windows
+
+package main
+
+import (
+	"fmt"
+	"os"
+	"syscall"
+)
+
+// reexec replaces the current process image with a fresh copy of itself,
+// preserving all arguments and environment variables. On success it never
+// returns (execve replaces the process in-place). On failure it returns an
+// error describing why the exec could not be performed.
+func reexec() error {
+	exe, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("failed to get executable path: %w", err)
+	}
+	return syscall.Exec(exe, os.Args, os.Environ())
+}

reexec_windows.go

@@ -0,0 +1,30 @@
+//go:build windows
+
+package main
+
+import (
+	"fmt"
+	"os"
+	"os/exec"
+)
+
+// reexec spawns a new copy of the current process with the same arguments and
+// environment, then exits the current process. On Windows, execve is not
+// available, so we start a child process and exit. On success it never returns
+// (os.Exit terminates the current process). On failure it returns an error.
+func reexec() error {
+	exe, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("failed to get executable path: %w", err)
+	}
+	cmd := exec.Command(exe, os.Args[1:]...)
+	cmd.Stdout = os.Stdout
+	cmd.Stderr = os.Stderr
+	cmd.Stdin = os.Stdin
+	cmd.Env = os.Environ()
+	if err := cmd.Start(); err != nil {
+		return fmt.Errorf("failed to start new process: %w", err)
+	}
+	os.Exit(0)
+	return nil // unreachable
+}

testing/udp_client.py

@@ -0,0 +1,49 @@
+import socket
+import sys
+
+# Argument parsing: Check if IP and Port are provided
+if len(sys.argv) != 3:
+    print("Usage: python udp_client.py <HOST_IP> <HOST_PORT>")
+    # Example: python udp_client.py 127.0.0.1 12000
+    sys.exit(1)
+
+HOST = sys.argv[1]
+try:
+    PORT = int(sys.argv[2])
+except ValueError:
+    print("Error: HOST_PORT must be an integer.")
+    sys.exit(1)
+
+# The message to send to the server
+MESSAGE = "Hello UDP Server! How are you?"
+
+# Create a UDP socket
+try:
+    client_socket = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+except socket.error as err:
+    print(f"Failed to create socket: {err}")
+    sys.exit()
+
+try:
+    print(f"Sending message to {HOST}:{PORT}...")
+
+    # Send the message (data must be encoded to bytes)
+    client_socket.sendto(MESSAGE.encode('utf-8'), (HOST, PORT))
+
+    # Wait for the server's response (buffer size 1024 bytes)
+    data, server_address = client_socket.recvfrom(1024)
+
+    # Decode and print the server's response
+    response = data.decode('utf-8')
+    print("-" * 30)
+    print(f"Received response from server {server_address[0]}:{server_address[1]}:")
+    print(f"-> Data: '{response}'")
+
+except socket.error as err:
+    print(f"Error during communication: {err}")
+
+finally:
+    # Close the socket
+    client_socket.close()
+    print("-" * 30)
+    print("Client finished and socket closed.")

testing/udp_server.py

@@ -0,0 +1,58 @@
+import socket
+import sys
+
+# optionally take in some positional args for the port
+if len(sys.argv) > 1:
+    try:
+        PORT = int(sys.argv[1])
+    except ValueError:
+        print("Invalid port number. Using default port 12000.")
+        PORT = 12000
+else:
+    PORT = 12000
+
+# Define the server host and port
+HOST = '0.0.0.0'  # Standard loopback interface address (localhost)
+
+# Create a UDP socket
+try:
+    server_socket = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+except socket.error as err:
+    print(f"Failed to create socket: {err}")
+    sys.exit()
+
+# Bind the socket to the address
+try:
+    server_socket.bind((HOST, PORT))
+    print(f"UDP Server listening on {HOST}:{PORT}")
+except socket.error as err:
+    print(f"Bind failed: {err}")
+    server_socket.close()
+    sys.exit()
+
+# Wait for and process incoming data
+while True:
+    try:
+        # Receive data and the client's address (buffer size 1024 bytes)
+        data, client_address = server_socket.recvfrom(1024)
+
+        # Decode the data and print the message
+        message = data.decode('utf-8')
+        print("-" * 30)
+        print(f"Received message from {client_address[0]}:{client_address[1]}:")
+        print(f"-> Data: '{message}'")
+
+        # Prepare the response message
+        response_message = f"Hello client! Server received: '{message.upper()}'"
+
+        # Send the response back to the client
+        server_socket.sendto(response_message.encode('utf-8'), client_address)
+        print(f"Sent response back to client.")
+
+    except Exception as e:
+        print(f"An error occurred: {e}")
+        break
+
+# Clean up (though usually unreachable in an infinite server loop)
+server_socket.close()
+print("Server stopped.")

testing/ws_client.py

@@ -0,0 +1,60 @@
+import asyncio
+import sys
+import websockets
+
+# Argument parsing: Check if HOST and PORT are provided
+if len(sys.argv) < 3 or len(sys.argv) > 4:
+    print("Usage: python ws_client.py <HOST_IP> <HOST_PORT> [ws|wss]")
+    # Example: python ws_client.py 127.0.0.1 8765
+    # Example: python ws_client.py 127.0.0.1 8765 wss
+    sys.exit(1)
+
+HOST = sys.argv[1]
+try:
+    PORT = int(sys.argv[2])
+except ValueError:
+    print("Error: HOST_PORT must be an integer.")
+    sys.exit(1)
+
+if len(sys.argv) == 4:
+    SCHEME = sys.argv[3].lower()
+    if SCHEME not in ("ws", "wss"):
+        print("Error: scheme must be 'ws' or 'wss'.")
+        sys.exit(1)
+else:
+    SCHEME = "ws"
+
+URI = f"{SCHEME}://{HOST}:{PORT}"
+
+# The message to send to the server
+MESSAGE = "Hello WebSocket Server! How are you?"
+
+
+async def main():
+    print(f"Connecting to {URI}...")
+
+    try:
+        async with websockets.connect(URI) as websocket:
+            print(f"Connected to server.")
+            print(f"Sending message: '{MESSAGE}'")
+
+            await websocket.send(MESSAGE)
+
+            response = await websocket.recv()
+
+            print("-" * 30)
+            print(f"Received response from server:")
+            print(f"-> Data: '{response}'")
+
+    except ConnectionRefusedError:
+        print(f"Error: Connection to {URI} was refused. Is the server running?")
+    except websockets.exceptions.InvalidMessage as e:
+        print(f"Error: Server did not respond with a valid WebSocket handshake: {e}")
+    except Exception as e:
+        print(f"Error during communication: {e}")
+
+    print("-" * 30)
+    print("Client finished.")
+
+
+asyncio.run(main())

testing/ws_server.py

@@ -0,0 +1,49 @@
+import asyncio
+import sys
+import websockets
+
+# Optionally take in a positional arg for the port
+if len(sys.argv) > 1:
+    try:
+        PORT = int(sys.argv[1])
+    except ValueError:
+        print("Invalid port number. Using default port 8765.")
+        PORT = 8765
+else:
+    PORT = 8765
+
+# Define the server host
+HOST = "0.0.0.0"
+
+
+async def handle_client(websocket):
+    client_address = websocket.remote_address
+    print(f"Client connected: {client_address[0]}:{client_address[1]}")
+
+    try:
+        async for message in websocket:
+            print("-" * 30)
+            print(f"Received message from {client_address[0]}:{client_address[1]}:")
+            print(f"-> Data: '{message}'")
+
+            response = f"Hello client! Server received: '{message.upper()}'"
+
+            await websocket.send(response)
+            print(f"Sent response back to client.")
+
+    except websockets.exceptions.ConnectionClosedOK:
+        print(f"Client {client_address[0]}:{client_address[1]} disconnected cleanly.")
+    except websockets.exceptions.ConnectionClosedError as e:
+        print(f"Client {client_address[0]}:{client_address[1]} disconnected with error: {e}")
+
+
+async def main():
+    print(f"WebSocket Server listening on {HOST}:{PORT}")
+    async with websockets.serve(handle_client, HOST, PORT):
+        await asyncio.Future()  # Run forever
+
+
+try:
+    asyncio.run(main())
+except KeyboardInterrupt:
+    print("\nServer stopped.")

updates/advantech.go

@@ -0,0 +1,48 @@
+//go:build !windows
+
+package updates
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+
+	"github.com/fosrl/newt/logger"
+)
+
+const advantechVersionFile = "/opt/newt/etc/version"
+
+// postUpdateAdvantech performs Advantech Router App-specific post-update steps:
+//   - Writes the new version string to /opt/newt/etc/version so that the
+//     router firmware's package management reflects the installed version.
+//   - Updates the PID file at pidFile (if non-empty) with the current process
+//     PID, in case the re-exec lands with a new PID on platforms where
+//     syscall.Exec behaviour differs.
+func postUpdateAdvantech(newVersion, pidFile string) error {
+	// --- Write version file ---
+	versionDir := filepath.Dir(advantechVersionFile)
+	if err := os.MkdirAll(versionDir, 0755); err != nil {
+		return fmt.Errorf("advantech: failed to create version directory %s: %w", versionDir, err)
+	}
+
+	if err := os.WriteFile(advantechVersionFile, []byte(newVersion+"\n"), 0644); err != nil {
+		return fmt.Errorf("advantech: failed to write version file %s: %w", advantechVersionFile, err)
+	}
+	logger.Debug("postUpdateAdvantech: wrote version %s to %s", newVersion, advantechVersionFile)
+
+	// --- Update PID file ---
+	// syscall.Exec replaces the process image in-place so the PID is preserved.
+	// We update the PID file here anyway so that any race between the old and
+	// new binary is covered (e.g. on platforms that fork before exec).
+	if pidFile != "" {
+		pid := fmt.Sprintf("%d\n", os.Getpid())
+		if err := os.WriteFile(pidFile, []byte(pid), 0644); err != nil {
+			// Non-fatal: log and continue so the update still proceeds.
+			logger.Debug("postUpdateAdvantech: warning: failed to update PID file %s: %v", pidFile, err)
+		} else {
+			logger.Debug("postUpdateAdvantech: updated PID file %s with PID %d", pidFile, os.Getpid())
+		}
+	}
+
+	return nil
+}

updates/advantech_windows.go

@@ -0,0 +1,8 @@
+//go:build windows
+
+package updates
+
+// postUpdateAdvantech is not supported on Windows.
+func postUpdateAdvantech(newVersion, pidFile string) error {
+	return nil
+}

updates/reexec_unix.go

@@ -0,0 +1,14 @@
+//go:build !windows
+
+package updates
+
+import (
+	"os"
+	"syscall"
+)
+
+// reexec replaces the current process image with the binary at exePath,
+// forwarding all original arguments and environment variables.
+func reexec(exePath string) error {
+	return syscall.Exec(exePath, os.Args, os.Environ())
+}

updates/reexec_windows.go

@@ -0,0 +1,28 @@
+//go:build windows
+
+package updates
+
+import (
+	"fmt"
+	"os"
+	"os/exec"
+)
+
+// reexec on Windows cannot use syscall.Exec (there is no exec syscall that
+// replaces the process image).  Instead we start a new process and exit the
+// current one.
+func reexec(exePath string) error {
+	cmd := exec.Command(exePath, os.Args[1:]...)
+	cmd.Stdin = os.Stdin
+	cmd.Stdout = os.Stdout
+	cmd.Stderr = os.Stderr
+	cmd.Env = os.Environ()
+
+	if err := cmd.Start(); err != nil {
+		return fmt.Errorf("failed to start updated binary: %w", err)
+	}
+
+	// Exit the current process so the new binary takes over.
+	os.Exit(0)
+	return nil // unreachable
+}

updates/selfupdate.go

@@ -0,0 +1,295 @@
+package updates
+
+import (
+	"bytes"
+	"context"
+	"crypto/sha256"
+	"crypto/tls"
+	"encoding/hex"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strings"
+	"time"
+
+	"github.com/fosrl/newt/logger"
+)
+
+// SelfUpdateConfig holds the configuration required to perform a self-update.
+type SelfUpdateConfig struct {
+	// Endpoint is the base URL of the pangolin server (e.g. "https://app.pangolin.net")
+	Endpoint string
+	// NewtID is the newt client identifier used for authentication.
+	NewtID string
+	// Secret is the newt client secret used for authentication.
+	Secret string
+	// CurrentVersion is the version of the currently running binary.
+	CurrentVersion string
+	// Platform is the OS+arch string embedded at build time via ldflags
+	// (e.g. "linux_amd64", "darwin_arm64").  When non-empty it is used
+	// directly; when empty the value is derived from runtime.GOOS/GOARCH.
+	Platform string
+	// TLSConfig is an optional TLS configuration for the HTTP client (may be nil).
+	TLSConfig *tls.Config
+}
+
+// versionResponse mirrors the JSON returned by POST /api/v1/auth/newt/version
+type versionResponse struct {
+	Data struct {
+		LatestVersion   string `json:"latestVersion"`
+		CurrentIsLatest bool   `json:"currentIsLatest"`
+		DownloadUrl     string `json:"downloadUrl"`
+		Sha256          string `json:"sha256"`
+	} `json:"data"`
+	Success bool   `json:"success"`
+	Message string `json:"message"`
+}
+
+// isOfficialContainer returns true when the process is running inside an
+// official Fossorial-built container image.  The image sets
+// NEWT_SYSTEM_SUBSTRATE="CONTAINER" at build time; users running newt in their own
+// containers (or bare-metal) will not have this variable set.
+func isOfficialContainer() bool {
+	return os.Getenv("NEWT_SYSTEM_SUBSTRATE") == "CONTAINER"
+}
+
+// platform returns the OS+arch string used in the newt release binary names,
+// e.g. "linux_amd64", "darwin_arm64", "windows_amd64".
+// It is used as a fallback when no platform was embedded at build time.
+func platform() string {
+	goarch := runtime.GOARCH
+	goos := runtime.GOOS
+
+	// Map Go arch names to the names used by newt releases
+	archMap := map[string]string{
+		"amd64":   "amd64",
+		"arm64":   "arm64",
+		"arm":     "arm32",
+		"riscv64": "riscv64",
+	}
+
+	arch, ok := archMap[goarch]
+	if !ok {
+		arch = goarch
+	}
+
+	return fmt.Sprintf("%s_%s", goos, arch)
+}
+
+// verifySHA256 checks that the file at path has the expected SHA-256 hex digest.
+func verifySHA256(path, expected string) error {
+	f, err := os.Open(path)
+	if err != nil {
+		return fmt.Errorf("failed to open file for hashing: %w", err)
+	}
+	defer f.Close()
+
+	h := sha256.New()
+	if _, err := io.Copy(h, f); err != nil {
+		return fmt.Errorf("failed to hash file: %w", err)
+	}
+
+	got := hex.EncodeToString(h.Sum(nil))
+	if !strings.EqualFold(got, expected) {
+		return fmt.Errorf("sha256 mismatch: expected %s, got %s", expected, got)
+	}
+	return nil
+}
+
+// CheckAndSelfUpdate contacts the pangolin server, checks whether a newer
+// version of newt is available, downloads it if so, replaces the running
+// binary on disk, and re-executes from the new binary.
+//
+// It returns an error when the check or update fails.  On a successful update
+// the function does not return – the process is replaced by the new binary via
+// syscall.Exec.
+func CheckAndSelfUpdate(cfg SelfUpdateConfig) error {
+	logger.Debug("checkAndSelfUpdate: starting update check (currentVersion=%s)", cfg.CurrentVersion)
+
+	if isOfficialContainer() {
+		logger.Debug("checkAndSelfUpdate: running inside official container, skipping auto-update")
+		return fmt.Errorf("auto-update is not supported in official Fossorial container images; pull a new image tag instead")
+	}
+
+	if cfg.CurrentVersion == "version_replaceme" {
+		logger.Debug("checkAndSelfUpdate: development build detected, skipping auto-update")
+		return fmt.Errorf("cannot auto-update a development build (version_replaceme)")
+	}
+
+	baseEndpoint := strings.TrimRight(cfg.Endpoint, "/")
+
+	// Build the HTTP client.
+	httpClient := &http.Client{Timeout: 30 * time.Second}
+	if cfg.TLSConfig != nil {
+		httpClient.Transport = &http.Transport{TLSClientConfig: cfg.TLSConfig}
+	}
+
+	// Check the current binary path before we do anything else so we can fail
+	// fast if the executable path cannot be determined.
+	exePath, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("failed to determine current executable path: %w", err)
+	}
+	exePath, err = filepath.EvalSymlinks(exePath)
+	if err != nil {
+		return fmt.Errorf("failed to resolve symlinks for executable path: %w", err)
+	}
+	logger.Debug("checkAndSelfUpdate: current executable path: %s", exePath)
+
+	// --- Step 1: Ask the server for the latest version ---
+	plat := cfg.Platform
+	if plat == "" {
+		plat = platform()
+	}
+	logger.Debug("checkAndSelfUpdate: querying server for latest version (platform=%s, endpoint=%s)", plat, baseEndpoint)
+	reqBody, err := json.Marshal(map[string]string{
+		"newtId":   cfg.NewtID,
+		"secret":   cfg.Secret,
+		"platform": plat,
+	})
+	if err != nil {
+		return fmt.Errorf("failed to marshal version request: %w", err)
+	}
+
+	versionURL, err := url.JoinPath(baseEndpoint, "/api/v1/auth/newt/version")
+	if err != nil {
+		return fmt.Errorf("failed to build version URL: %w", err)
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	req, err := http.NewRequestWithContext(ctx, "POST", versionURL, bytes.NewBuffer(reqBody))
+	if err != nil {
+		return fmt.Errorf("failed to create version request: %w", err)
+	}
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("X-CSRF-Token", "x-csrf-protection")
+
+	resp, err := httpClient.Do(req)
+	if err != nil {
+		return fmt.Errorf("failed to request version info: %w", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode == http.StatusNoContent { // updates are disabled
+		logger.Debug("checkAndSelfUpdate: server indicated updates are disabled (204 No Content)")
+		return nil
+	}
+
+	if resp.StatusCode == http.StatusNotFound { // older server without version endpoint
+		logger.Debug("checkAndSelfUpdate: server does not support version endpoint (404 Not Found), skipping")
+		return nil
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		body, _ := io.ReadAll(resp.Body)
+		return fmt.Errorf("server returned status %d: %s", resp.StatusCode, string(body))
+	}
+
+	var verResp versionResponse
+	if err := json.NewDecoder(resp.Body).Decode(&verResp); err != nil {
+		return fmt.Errorf("failed to parse version response: %w", err)
+	}
+
+	if !verResp.Success {
+		return fmt.Errorf("server error: %s", verResp.Message)
+	}
+
+	if verResp.Data.CurrentIsLatest {
+		logger.Debug("checkAndSelfUpdate: already up to date (%s)", cfg.CurrentVersion)
+		return nil
+	}
+
+	logger.Debug("checkAndSelfUpdate: update available %s → %s", cfg.CurrentVersion, verResp.Data.LatestVersion)
+
+	// --- Pre-download: verify we can write to the binary's directory ---
+	// Do this before downloading so a permission failure doesn't waste bandwidth.
+	exeDir := filepath.Dir(exePath)
+	logger.Debug("checkAndSelfUpdate: verifying write access to %s", exeDir)
+	writeTestFile, err := os.CreateTemp(exeDir, ".newt-write-test-*")
+	if err != nil {
+		return fmt.Errorf("cannot write to %s (you may need to run as root or with elevated permissions): %w", exeDir, err)
+	}
+	writeTestFile.Close()
+	_ = os.Remove(writeTestFile.Name())
+
+	// --- Step 2: Download the new binary ---
+	logger.Debug("checkAndSelfUpdate: beginning download of new binary")
+	dlCtx, dlCancel := context.WithTimeout(context.Background(), 5*time.Minute)
+	defer dlCancel()
+
+	dlReq, err := http.NewRequestWithContext(dlCtx, "GET", verResp.Data.DownloadUrl, nil)
+	if err != nil {
+		return fmt.Errorf("failed to create download request: %w", err)
+	}
+
+	dlResp, err := httpClient.Do(dlReq)
+	if err != nil {
+		return fmt.Errorf("failed to download new binary: %w", err)
+	}
+	defer dlResp.Body.Close()
+
+	if dlResp.StatusCode != http.StatusOK {
+		return fmt.Errorf("download failed with status %d", dlResp.StatusCode)
+	}
+
+	// Write to a temp file in the same directory as the current binary so that
+	// an atomic rename works even across filesystem boundaries.
+	tmpFile, err := os.CreateTemp(exeDir, ".newt-update-*")
+	tmpPath := tmpFile.Name()
+
+	// Ensure the temp file is cleaned up on any error path.
+	defer func() {
+		_ = os.Remove(tmpPath)
+	}()
+
+	if _, err := io.Copy(tmpFile, dlResp.Body); err != nil {
+		_ = tmpFile.Close()
+		return fmt.Errorf("failed to write downloaded binary: %w", err)
+	}
+	if err := tmpFile.Close(); err != nil {
+		return fmt.Errorf("failed to close temp file: %w", err)
+	}
+
+	// --- Verify SHA256 checksum if the server provided one ---
+	if verResp.Data.Sha256 != "" {
+		logger.Debug("checkAndSelfUpdate: verifying SHA256 checksum")
+		if err := verifySHA256(tmpPath, verResp.Data.Sha256); err != nil {
+			return fmt.Errorf("binary integrity check failed: %w", err)
+		}
+		logger.Debug("checkAndSelfUpdate: SHA256 checksum verified")
+	} else {
+		logger.Debug("checkAndSelfUpdate: no SHA256 checksum provided by server, skipping verification")
+	}
+
+	// Make the new binary executable.
+	if err := os.Chmod(tmpPath, 0755); err != nil {
+		return fmt.Errorf("failed to set executable permission: %w", err)
+	}
+
+	// --- Step 3: Replace the running binary ---
+	// On Unix an atomic rename works even while the file is running.
+	logger.Debug("checkAndSelfUpdate: replacing binary at %s", exePath)
+	if err := os.Rename(tmpPath, exePath); err != nil {
+		return fmt.Errorf("failed to replace binary (you may need to run as root): %w", err)
+	}
+
+	systemSubstrate := os.Getenv("NEWT_SYSTEM_SUBSTRATE")
+	logger.Debug("checkAndSelfUpdate: system substrate: %s", systemSubstrate)
+	if systemSubstrate == "ADVANTECH_ROUTER_APP" {
+		pidFile := os.Getenv("NEWT_PID_FILE")
+		if err := postUpdateAdvantech(verResp.Data.LatestVersion, pidFile); err != nil {
+			logger.Debug("checkAndSelfUpdate: advantech post-update steps failed: %v", err)
+		}
+	}
+
+	// --- Step 4: Re-exec ---
+	logger.Debug("checkAndSelfUpdate: re-executing new binary")
+	return reexec(exePath)
+}

updates/updates.go

@@ -119,7 +119,7 @@ func CheckForUpdate(owner, repo, currentVersion string) error {
 
 	// Check if update is available
 	if currentVer.isNewer(latestVer) {
-		printUpdateBanner(currentVer.String(), latestVer.String(), release.HTMLURL)
+		printUpdateBanner(currentVer.String(), latestVer.String(), "curl -fsSL https://static.pangolin.net/get-newt.sh | bash")
 	}
 
 	return nil
@@ -145,7 +145,7 @@ func printUpdateBanner(currentVersion, latestVersion, releaseURL string) {
 		"║  A newer version is available! Please update to get the" + padRight("", contentWidth-56) + "║",
 		"║  latest features, bug fixes, and security improvements." + padRight("", contentWidth-56) + "║",
 		emptyLine,
-		"║  Release URL: " + padRight(releaseURL, contentWidth-15) + "║",
+		"║  Update: " + padRight(releaseURL, contentWidth-10) + "║",
 		emptyLine,
 		borderBot,
 	}

util/util.go

@@ -1,6 +1,7 @@
 package util
 
 import (
+	"context"
 	"encoding/base64"
 	"encoding/binary"
 	"encoding/hex"
@@ -14,6 +15,99 @@ import (
 	"golang.zx2c4.com/wireguard/device"
 )
 
+func ResolveDomainUpstream(domain string, publicDNS []string) (string, error) {
+	// trim whitespace
+	domain = strings.TrimSpace(domain)
+
+	// Remove any protocol prefix if present (do this first, before splitting host/port)
+	domain = strings.TrimPrefix(domain, "http://")
+	domain = strings.TrimPrefix(domain, "https://")
+
+	// if there are any trailing slashes, remove them
+	domain = strings.TrimSuffix(domain, "/")
+
+	// Check if there's a port in the domain
+	host, port, err := net.SplitHostPort(domain)
+	if err != nil {
+		// No port found, use the domain as is
+		host = domain
+		port = ""
+	}
+
+	// Check if host is already an IP address (IPv4 or IPv6)
+	// For IPv6, the host from SplitHostPort will already have brackets stripped
+	// but if there was no port, we need to handle bracketed IPv6 addresses
+	cleanHost := strings.TrimPrefix(strings.TrimSuffix(host, "]"), "[")
+	if ip := net.ParseIP(cleanHost); ip != nil {
+		// It's already an IP address, no need to resolve
+		ipAddr := ip.String()
+		if port != "" {
+			return net.JoinHostPort(ipAddr, port), nil
+		}
+		return ipAddr, nil
+	}
+
+	// Lookup IP addresses using the upstream DNS servers if provided
+	var ips []net.IP
+	if len(publicDNS) > 0 {
+		var lastErr error
+		for _, server := range publicDNS {
+			// Ensure the upstream DNS address has a port
+			dnsAddr := server
+			if _, _, err := net.SplitHostPort(dnsAddr); err != nil {
+				// No port specified, default to 53
+				dnsAddr = net.JoinHostPort(server, "53")
+			}
+
+			resolver := &net.Resolver{
+				PreferGo: true,
+				Dial: func(ctx context.Context, network, address string) (net.Conn, error) {
+					d := net.Dialer{}
+					return d.DialContext(ctx, "udp", dnsAddr)
+				},
+			}
+			ips, lastErr = resolver.LookupIP(context.Background(), "ip", host)
+			if lastErr == nil {
+				break
+			}
+		}
+		if lastErr != nil {
+			return "", fmt.Errorf("DNS lookup failed using all upstream servers: %v", lastErr)
+		}
+	} else {
+		ips, err = net.LookupIP(host)
+		if err != nil {
+			return "", fmt.Errorf("DNS lookup failed: %v", err)
+		}
+	}
+
+	if len(ips) == 0 {
+		return "", fmt.Errorf("no IP addresses found for domain %s", host)
+	}
+
+	// Get the first IPv4 address if available
+	var ipAddr string
+	for _, ip := range ips {
+		if ipv4 := ip.To4(); ipv4 != nil {
+			ipAddr = ipv4.String()
+			break
+		}
+	}
+
+	// If no IPv4 found, use the first IP (might be IPv6)
+	if ipAddr == "" {
+		ipAddr = ips[0].String()
+	}
+
+	// Add port back if it existed
+	if port != "" {
+		ipAddr = net.JoinHostPort(ipAddr, port)
+	}
+
+	return ipAddr, nil
+}
+
+
 func ResolveDomain(domain string) (string, error) {
 	// trim whitespace
 	domain = strings.TrimSpace(domain)

websocket/client.go

@@ -2,6 +2,7 @@ package websocket
 
 import (
 	"bytes"
+	"compress/gzip"
 	"crypto/tls"
 	"crypto/x509"
 	"encoding/json"
@@ -37,16 +38,22 @@ type Client struct {
 	isConnected       bool
 	reconnectMux      sync.RWMutex
 	pingInterval      time.Duration
-	pingTimeout       time.Duration
 	onConnect         func() error
 	onTokenUpdate     func(token string)
 	writeMux          sync.Mutex
 	clientType        string // Type of client (e.g., "newt", "olm")
+	configFilePath    string // Optional override for the config file path
 	tlsConfig         TLSConfig
 	metricsCtxMu      sync.RWMutex
 	metricsCtx        context.Context
 	configNeedsSave   bool // Flag to track if config needs to be saved
 	serverVersion     string
+	configVersion     int64 // Latest config version received from server
+	configVersionMux  sync.RWMutex
+	processingMessage bool           // Flag to track if a message is currently being processed
+	processingMux     sync.RWMutex   // Protects processingMessage
+	processingWg      sync.WaitGroup // WaitGroup to wait for message processing to complete
+	justProvisioned   bool           // Set to true when provisionIfNeeded exchanges a key for permanent credentials
 }
 
 type ClientOption func(*Client)
@@ -72,6 +79,12 @@ func WithBaseURL(url string) ClientOption {
 }
 
 // WithTLSConfig sets the TLS configuration for the client
+func WithConfigFile(path string) ClientOption {
+	return func(c *Client) {
+		c.configFilePath = path
+	}
+}
+
 func WithTLSConfig(config TLSConfig) ClientOption {
 	return func(c *Client) {
 		c.tlsConfig = config
@@ -90,6 +103,16 @@ func (c *Client) OnTokenUpdate(callback func(token string)) {
 	c.onTokenUpdate = callback
 }
 
+// WasJustProvisioned reports whether the client exchanged a provisioning key
+// for permanent credentials during the most recent connection attempt. It
+// consumes the flag – subsequent calls return false until provisioning occurs
+// again (which, in practice, never happens once credentials are persisted).
+func (c *Client) WasJustProvisioned() bool {
+	v := c.justProvisioned
+	c.justProvisioned = false
+	return v
+}
+
 func (c *Client) metricsContext() context.Context {
 	c.metricsCtxMu.RLock()
 	defer c.metricsCtxMu.RUnlock()
@@ -111,7 +134,7 @@ func (c *Client) MetricsContext() context.Context {
 }
 
 // NewClient creates a new websocket client
-func NewClient(clientType string, ID, secret string, endpoint string, pingInterval time.Duration, pingTimeout time.Duration, opts ...ClientOption) (*Client, error) {
+func NewClient(clientType string, ID, secret string, endpoint string, pingInterval time.Duration, opts ...ClientOption) (*Client, error) {
 	config := &Config{
 		ID:       ID,
 		Secret:   secret,
@@ -126,7 +149,6 @@ func NewClient(clientType string, ID, secret string, endpoint string, pingInterv
 		reconnectInterval: 3 * time.Second,
 		isConnected:       false,
 		pingInterval:      pingInterval,
-		pingTimeout:       pingTimeout,
 		clientType:        clientType,
 	}
 
@@ -150,10 +172,29 @@ func (c *Client) GetConfig() *Config {
 	return c.config
 }
 
+// GetConfigFilePath returns the resolved path to the config file used by this client.
+func (c *Client) GetConfigFilePath() string {
+	return getConfigPath(c.clientType, c.configFilePath)
+}
+
 func (c *Client) GetServerVersion() string {
 	return c.serverVersion
 }
 
+// GetConfigVersion returns the latest config version received from server
+func (c *Client) GetConfigVersion() int64 {
+	c.configVersionMux.RLock()
+	defer c.configVersionMux.RUnlock()
+	return c.configVersion
+}
+
+// setConfigVersion updates the config version
+func (c *Client) setConfigVersion(version int64) {
+	c.configVersionMux.Lock()
+	defer c.configVersionMux.Unlock()
+	c.configVersion = version
+}
+
 // Connect establishes the WebSocket connection
 func (c *Client) Connect() error {
 	go c.connectWithRetry()
@@ -235,13 +276,17 @@ func (c *Client) SendMessageInterval(messageType string, data interface{}, inter
 	stopChan := make(chan struct{})
 	go func() {
 		count := 0
-		maxAttempts := 10
+		maxAttempts := 16
 
+		c.reconnectMux.RLock()
+		connected := c.isConnected
+		c.reconnectMux.RUnlock()
 		err := c.SendMessage(messageType, data) // Send immediately
 		if err != nil {
 			logger.Error("Failed to send initial message: %v", err)
+		} else if connected {
+			count++
 		}
-		count++
 
 		ticker := time.NewTicker(interval)
 		defer ticker.Stop()
@@ -252,11 +297,15 @@ func (c *Client) SendMessageInterval(messageType string, data interface{}, inter
 					logger.Info("SendMessageInterval timed out after %d attempts for message type: %s", maxAttempts, messageType)
 					return
 				}
+				c.reconnectMux.RLock()
+				connected = c.isConnected
+				c.reconnectMux.RUnlock()
 				err = c.SendMessage(messageType, data)
 				if err != nil {
 					logger.Error("Failed to send message: %v", err)
+				} else if connected {
+					count++
 				}
-				count++
 			case <-stopChan:
 				return
 			}
@@ -463,6 +512,11 @@ func (c *Client) connectWithRetry() {
 func (c *Client) establishConnection() error {
 	ctx := context.Background()
 
+	// Exchange provisioning key for permanent credentials if needed.
+	if err := c.provisionIfNeeded(); err != nil {
+		return fmt.Errorf("failed to provision newt credentials: %w", err)
+	}
+
 	// Get token for authentication
 	token, err := c.getToken()
 	if err != nil {
@@ -641,7 +695,61 @@ func (c *Client) setupPKCS12TLS() (*tls.Config, error) {
 }
 
 // pingMonitor sends pings at a short interval and triggers reconnect on failure
+func (c *Client) sendPing() {
+	if c.conn == nil {
+		return
+	}
+
+	// Skip ping if a message is currently being processed
+	c.processingMux.RLock()
+	isProcessing := c.processingMessage
+	c.processingMux.RUnlock()
+	if isProcessing {
+		logger.Debug("Skipping ping, message is being processed")
+		return
+	}
+
+	c.configVersionMux.RLock()
+	configVersion := c.configVersion
+	c.configVersionMux.RUnlock()
+
+	pingMsg := WSMessage{
+		Type:          "newt/ping",
+		Data:          map[string]interface{}{},
+		ConfigVersion: configVersion,
+	}
+
+	c.writeMux.Lock()
+	if c.conn == nil {
+		c.writeMux.Unlock()
+		return
+	}
+	err := c.conn.WriteJSON(pingMsg)
+	if err == nil {
+		telemetry.IncWSMessage(c.metricsContext(), "out", "ping")
+	}
+	c.writeMux.Unlock()
+
+	if err != nil {
+		// Check if we're shutting down before logging error and reconnecting
+		select {
+		case <-c.done:
+			// Expected during shutdown
+			return
+		default:
+			logger.Error("Ping failed: %v", err)
+			telemetry.IncWSKeepaliveFailure(c.metricsContext(), "ping_write")
+			telemetry.IncWSReconnect(c.metricsContext(), "ping_write")
+			c.reconnect()
+			return
+		}
+	}
+}
+
 func (c *Client) pingMonitor() {
+	// Send an immediate ping as soon as we connect
+	c.sendPing()
+
 	ticker := time.NewTicker(c.pingInterval)
 	defer ticker.Stop()
 
@@ -650,29 +758,7 @@ func (c *Client) pingMonitor() {
 		case <-c.done:
 			return
 		case <-ticker.C:
-			if c.conn == nil {
-				return
-			}
-			c.writeMux.Lock()
-			err := c.conn.WriteControl(websocket.PingMessage, []byte{}, time.Now().Add(c.pingTimeout))
-			if err == nil {
-				telemetry.IncWSMessage(c.metricsContext(), "out", "ping")
-			}
-			c.writeMux.Unlock()
-			if err != nil {
-				// Check if we're shutting down before logging error and reconnecting
-				select {
-				case <-c.done:
-					// Expected during shutdown
-					return
-				default:
-					logger.Error("Ping failed: %v", err)
-					telemetry.IncWSKeepaliveFailure(c.metricsContext(), "ping_write")
-					telemetry.IncWSReconnect(c.metricsContext(), "ping_write")
-					c.reconnect()
-					return
-				}
-			}
+			c.sendPing()
 		}
 	}
 }
@@ -709,10 +795,13 @@ func (c *Client) readPumpWithDisconnectDetection(started time.Time) {
 			disconnectResult = "success"
 			return
 		default:
-			var msg WSMessage
-			err := c.conn.ReadJSON(&msg)
+			msgType, p, err := c.conn.ReadMessage()
 			if err == nil {
-				telemetry.IncWSMessage(c.metricsContext(), "in", "text")
+				if msgType == websocket.BinaryMessage {
+					telemetry.IncWSMessage(c.metricsContext(), "in", "binary")
+				} else {
+					telemetry.IncWSMessage(c.metricsContext(), "in", "text")
+				}
 			}
 			if err != nil {
 				// Check if we're shutting down before logging error
@@ -737,9 +826,47 @@ func (c *Client) readPumpWithDisconnectDetection(started time.Time) {
 				}
 			}
 
+			// Update config version from incoming message
+			var data []byte
+			if msgType == websocket.BinaryMessage {
+				gr, err := gzip.NewReader(bytes.NewReader(p))
+				if err != nil {
+					logger.Error("WebSocket failed to create gzip reader: %v", err)
+					continue
+				}
+				data, err = io.ReadAll(gr)
+				gr.Close()
+				if err != nil {
+					logger.Error("WebSocket failed to decompress message: %v", err)
+					continue
+				}
+			} else {
+				data = p
+			}
+
+			var msg WSMessage
+			if err = json.Unmarshal(data, &msg); err != nil {
+				logger.Error("WebSocket failed to parse message: %v", err)
+				continue
+			}
+
+			c.setConfigVersion(msg.ConfigVersion)
+
 			c.handlersMux.RLock()
 			if handler, ok := c.handlers[msg.Type]; ok {
+				// Mark that we're processing a message
+				c.processingMux.Lock()
+				c.processingMessage = true
+				c.processingMux.Unlock()
+				c.processingWg.Add(1)
+
 				handler(msg)
+
+				// Mark that we're done processing
+				c.processingWg.Done()
+				c.processingMux.Lock()
+				c.processingMessage = false
+				c.processingMux.Unlock()
 			}
 			c.handlersMux.RUnlock()
 		}
@@ -749,10 +876,12 @@ func (c *Client) readPumpWithDisconnectDetection(started time.Time) {
 func (c *Client) reconnect() {
 	c.setConnected(false)
 	telemetry.SetWSConnectionState(false)
+	c.writeMux.Lock()
 	if c.conn != nil {
 		c.conn.Close()
 		c.conn = nil
 	}
+	c.writeMux.Unlock()
 
 	// Only reconnect if we're not shutting down
 	select {
@@ -807,3 +936,19 @@ func loadClientCertificate(p12Path string) (*tls.Config, error) {
 		RootCAs:      rootCAs,
 	}, nil
 }
+
+// BuildTLSConfig creates a *tls.Config from the provided certificate files.
+// Pass empty strings / nil slice for fields that are not used.
+// Returns nil, nil when no TLS credentials are provided.
+func BuildTLSConfig(certFile, keyFile string, caFiles []string, pkcs12File string) (*tls.Config, error) {
+	c := &Client{
+		tlsConfig: TLSConfig{
+			ClientCertFile: certFile,
+			ClientKeyFile:  keyFile,
+			CAFiles:        caFiles,
+			PKCS12File:     pkcs12File,
+		},
+		config: &Config{},
+	}
+	return c.setupTLS()
+}

websocket/config.go

@@ -1,16 +1,29 @@
 package websocket
 
 import (
+	"bytes"
+	"context"
+	"crypto/tls"
 	"encoding/json"
+	"fmt"
+	"io"
 	"log"
+	"net/http"
+	"net/url"
 	"os"
 	"path/filepath"
+	"regexp"
 	"runtime"
+	"strings"
+	"time"
 
 	"github.com/fosrl/newt/logger"
 )
 
-func getConfigPath(clientType string) string {
+func getConfigPath(clientType string, overridePath string) string {
+	if overridePath != "" {
+		return overridePath
+	}
 	configFile := os.Getenv("CONFIG_FILE")
 	if configFile == "" {
 		var configDir string
@@ -36,7 +49,7 @@ func getConfigPath(clientType string) string {
 
 func (c *Client) loadConfig() error {
 	originalConfig := *c.config // Store original config to detect changes
-	configPath := getConfigPath(c.clientType)
+	configPath := getConfigPath(c.clientType, c.configFilePath)
 
 	if c.config.ID != "" && c.config.Secret != "" && c.config.Endpoint != "" {
 		logger.Debug("Config already provided, skipping loading from file")
@@ -58,6 +71,11 @@ func (c *Client) loadConfig() error {
 		}
 		return err
 	}
+	if len(bytes.TrimSpace(data)) == 0 {
+		logger.Info("Config file at %s is empty, will initialize it with provided values", configPath)
+		c.configNeedsSave = true
+		return nil
+	}
 
 	var config Config
 	if err := json.Unmarshal(data, &config); err != nil {
@@ -83,6 +101,14 @@ func (c *Client) loadConfig() error {
 		c.config.Endpoint = config.Endpoint
 		c.baseURL = config.Endpoint
 	}
+	// Always load the provisioning key from the file if not already set
+	if c.config.ProvisioningKey == "" {
+		c.config.ProvisioningKey = config.ProvisioningKey
+	}
+	// Always load the name from the file if not already set
+	if c.config.Name == "" {
+		c.config.Name = config.Name
+	}
 
 	// Check if CLI args provided values that override file values
 	if (!fileHadID && originalConfig.ID != "") ||
@@ -105,7 +131,7 @@ func (c *Client) saveConfig() error {
 		return nil
 	}
 
-	configPath := getConfigPath(c.clientType)
+	configPath := getConfigPath(c.clientType, c.configFilePath)
 	data, err := json.MarshalIndent(c.config, "", "  ")
 	if err != nil {
 		return err
@@ -118,3 +144,139 @@ func (c *Client) saveConfig() error {
 	}
 	return err
 }
+
+// interpolateString replaces {{env.VAR}} tokens in s with the corresponding
+// environment variable values. Tokens that do not match a supported scheme are
+// left unchanged, mirroring the blueprint interpolation logic.
+func interpolateString(s string) string {
+	re := regexp.MustCompile(`\{\{([^}]+)\}\}`)
+	return re.ReplaceAllStringFunc(s, func(match string) string {
+		inner := strings.TrimSpace(match[2 : len(match)-2])
+		if strings.HasPrefix(inner, "env.") {
+			varName := strings.TrimPrefix(inner, "env.")
+			return os.Getenv(varName)
+		}
+		return match
+	})
+}
+
+// provisionIfNeeded checks whether a provisioning key is present and, if so,
+// exchanges it for a newt ID and secret by calling the registration endpoint.
+// On success the config is updated in-place and flagged for saving so that
+// subsequent runs use the permanent credentials directly.
+func (c *Client) provisionIfNeeded() error {
+	if c.config.ProvisioningKey == "" {
+		return nil
+	}
+
+	// If we already have both credentials there is nothing to provision.
+	if c.config.ID != "" && c.config.Secret != "" {
+		logger.Debug("Credentials already present, skipping provisioning")
+		return nil
+	}
+
+	logger.Info("Provisioning key found – exchanging for newt credentials...")
+
+	baseURL, err := url.Parse(c.baseURL)
+	if err != nil {
+		return fmt.Errorf("failed to parse base URL for provisioning: %w", err)
+	}
+	baseEndpoint := strings.TrimRight(baseURL.String(), "/")
+
+	// Interpolate any {{env.VAR}} tokens in the name before sending.
+	name := interpolateString(c.config.Name)
+
+	reqBody := map[string]interface{}{
+		"provisioningKey": c.config.ProvisioningKey,
+	}
+	if name != "" {
+		reqBody["name"] = name
+	}
+	jsonData, err := json.Marshal(reqBody)
+	if err != nil {
+		return fmt.Errorf("failed to marshal provisioning request: %w", err)
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	req, err := http.NewRequestWithContext(
+		ctx,
+		"POST",
+		baseEndpoint+"/api/v1/auth/newt/register",
+		bytes.NewBuffer(jsonData),
+	)
+	if err != nil {
+		return fmt.Errorf("failed to create provisioning request: %w", err)
+	}
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("X-CSRF-Token", "x-csrf-protection")
+
+	// Mirror the TLS setup used by getToken so mTLS / self-signed CAs work.
+	var tlsCfg *tls.Config
+	if c.tlsConfig.ClientCertFile != "" || c.tlsConfig.ClientKeyFile != "" ||
+		len(c.tlsConfig.CAFiles) > 0 || c.tlsConfig.PKCS12File != "" {
+		tlsCfg, err = c.setupTLS()
+		if err != nil {
+			return fmt.Errorf("failed to setup TLS for provisioning: %w", err)
+		}
+	}
+	if os.Getenv("SKIP_TLS_VERIFY") == "true" {
+		if tlsCfg == nil {
+			tlsCfg = &tls.Config{}
+		}
+		tlsCfg.InsecureSkipVerify = true
+		logger.Debug("TLS certificate verification disabled for provisioning via SKIP_TLS_VERIFY")
+	}
+
+	httpClient := &http.Client{}
+	if tlsCfg != nil {
+		httpClient.Transport = &http.Transport{TLSClientConfig: tlsCfg}
+	}
+
+	resp, err := httpClient.Do(req)
+	if err != nil {
+		return fmt.Errorf("provisioning request failed: %w", err)
+	}
+	defer resp.Body.Close()
+
+	body, _ := io.ReadAll(resp.Body)
+	logger.Debug("Provisioning response body: %s", string(body))
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return fmt.Errorf("provisioning endpoint returned status %d: %s", resp.StatusCode, string(body))
+	}
+
+	var provResp ProvisioningResponse
+	if err := json.Unmarshal(body, &provResp); err != nil {
+		return fmt.Errorf("failed to decode provisioning response: %w", err)
+	}
+
+	if !provResp.Success {
+		return fmt.Errorf("provisioning failed: %s", provResp.Message)
+	}
+
+	if provResp.Data.NewtID == "" || provResp.Data.Secret == "" {
+		return fmt.Errorf("provisioning response is missing newt ID or secret")
+	}
+
+	logger.Info("Successfully provisioned – newt ID: %s", provResp.Data.NewtID)
+
+	// Persist the returned credentials and clear the one-time provisioning key
+	// so subsequent runs authenticate normally.
+	c.config.ID = provResp.Data.NewtID
+	c.config.Secret = provResp.Data.Secret
+	c.config.ProvisioningKey = ""
+	c.config.Name = ""
+	c.configNeedsSave = true
+	c.justProvisioned = true
+
+	// Save immediately so that if the subsequent connection attempt fails the
+	// provisioning key is already gone from disk and the next retry uses the
+	// permanent credentials instead of trying to provision again.
+	if err := c.saveConfig(); err != nil {
+		logger.Error("Failed to save config after provisioning: %v", err)
+	}
+
+	return nil
+}

websocket/config_test.go

@@ -0,0 +1,35 @@
+package websocket
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestLoadConfig_EmptyFileMarksConfigForSave(t *testing.T) {
+	t.Setenv("CONFIG_FILE", "")
+
+	tmpDir := t.TempDir()
+	configPath := filepath.Join(tmpDir, "config.json")
+	if err := os.WriteFile(configPath, []byte(""), 0o644); err != nil {
+		t.Fatalf("failed to create empty config file: %v", err)
+	}
+
+	client := &Client{
+		config: &Config{
+			Endpoint:        "https://example.com",
+			ProvisioningKey: "spk-test",
+		},
+		clientType:     "newt",
+		configFilePath: configPath,
+	}
+
+	if err := client.loadConfig(); err != nil {
+		t.Fatalf("loadConfig returned error for empty file: %v", err)
+	}
+
+	if !client.configNeedsSave {
+		t.Fatal("expected empty config file to mark configNeedsSave")
+	}
+}
+

websocket/types.go

@@ -1,10 +1,13 @@
 package websocket
 
 type Config struct {
-	ID            string `json:"id"`
-	Secret        string `json:"secret"`
-	Endpoint      string `json:"endpoint"`
-	TlsClientCert string `json:"tlsClientCert"`
+	ID              string `json:"id"`
+	Secret          string `json:"secret"`
+	Endpoint        string `json:"endpoint"`
+	TlsClientCert   string `json:"tlsClientCert"`
+	ProvisioningKey string `json:"provisioningKey,omitempty"`
+	Name            string `json:"name,omitempty"`
+	Blocked         bool   `json:"blocked,omitempty"`
 }
 
 type TokenResponse struct {
@@ -16,7 +19,17 @@ type TokenResponse struct {
 	Message string `json:"message"`
 }
 
-type WSMessage struct {
-	Type string      `json:"type"`
-	Data interface{} `json:"data"`
+type ProvisioningResponse struct {
+	Data struct {
+		NewtID string `json:"newtId"`
+		Secret string `json:"secret"`
+	} `json:"data"`
+	Success bool   `json:"success"`
+	Message string `json:"message"`
+}
+
+type WSMessage struct {
+	Type          string      `json:"type"`
+	Data          interface{} `json:"data"`
+	ConfigVersion int64       `json:"configVersion,omitempty"`
 }

wgtester/wgtester.go

@@ -38,7 +38,6 @@ type Server struct {
 	isRunning    bool
 	runningLock  sync.Mutex
 	newtID       string
-	outputPrefix string
 	useNetstack  bool
 	tnet         interface{} // Will be *netstack2.Net when using netstack
 }
@@ -50,7 +49,6 @@ func NewServer(serverAddr string, serverPort uint16, newtID string) *Server {
 		serverPort:   serverPort + 1, // use the next port for the server
 		shutdownCh:   make(chan struct{}),
 		newtID:       newtID,
-		outputPrefix: "[WGTester] ",
 		useNetstack:  false,
 		tnet:         nil,
 	}
@@ -63,7 +61,6 @@ func NewServerWithNetstack(serverAddr string, serverPort uint16, newtID string,
 		serverPort:   serverPort + 1, // use the next port for the server
 		shutdownCh:   make(chan struct{}),
 		newtID:       newtID,
-		outputPrefix: "[WGTester] ",
 		useNetstack:  true,
 		tnet:         tnet,
 	}
@@ -109,7 +106,7 @@ func (s *Server) Start() error {
 	s.isRunning = true
 	go s.handleConnections()
 
-	logger.Info("%sServer started on %s:%d", s.outputPrefix, s.serverAddr, s.serverPort)
+	logger.Debug("WGTester Server started on %s:%d", s.serverAddr, s.serverPort)
 	return nil
 }
 
@@ -127,7 +124,7 @@ func (s *Server) Stop() {
 		s.conn.Close()
 	}
 	s.isRunning = false
-	logger.Info("%sServer stopped", s.outputPrefix)
+	logger.Info("WGTester Server stopped")
 }
 
 // RestartWithNetstack stops the current server and restarts it with netstack
@@ -162,7 +159,7 @@ func (s *Server) handleConnections() {
 			// Set read deadline to avoid blocking forever
 			err := s.conn.SetReadDeadline(time.Now().Add(1 * time.Second))
 			if err != nil {
-				logger.Error("%sError setting read deadline: %v", s.outputPrefix, err)
+				logger.Error("Error setting read deadline: %v", err)
 				continue
 			}
 
@@ -192,7 +189,7 @@ func (s *Server) handleConnections() {
 					if err == io.EOF {
 						return
 					}
-					logger.Error("%sError reading from UDP: %v", s.outputPrefix, err)
+					logger.Error("Error reading from UDP: %v", err)
 				}
 				continue
 			}
@@ -224,7 +221,7 @@ func (s *Server) handleConnections() {
 			copy(responsePacket[5:13], buffer[5:13])
 
 			// Log response being sent for debugging
-			// logger.Debug("%sSending response to %s", s.outputPrefix, addr.String())
+			// logger.Debug("Sending response to %s", addr.String())
 
 			// Send the response packet - handle both regular UDP and netstack UDP
 			if s.useNetstack {
@@ -238,9 +235,9 @@ func (s *Server) handleConnections() {
 			}
 
 			if err != nil {
-				logger.Error("%sError sending response: %v", s.outputPrefix, err)
+				logger.Error("Error sending response: %v", err)
 			} else {
-				// logger.Debug("%sResponse sent successfully", s.outputPrefix)
+				// logger.Debug("Response sent successfully")
 			}
 		}
 	}