Experimenting with fw event reporting

.github/actions/code-sign/action.yml

@@ -1,113 +0,0 @@
-name: Code-sign Windows binaries
-description: >
-  Sign every .exe under a given path in place via the DefinedNet code-signer
-  Lambda. If `role` or `bucket` is empty, logs a notice and skips signing so
-  forks and dev branches without AWS access still produce usable builds.
-
-inputs:
-  path:
-    description: "Directory whose .exe files should be signed in place"
-    required: true
-  role:
-    description: "IAM role ARN to assume via OIDC; empty disables signing"
-    required: false
-    default: ""
-  bucket:
-    description: "S3 staging bucket the code-signer Lambda reads from; empty disables signing"
-    required: false
-    default: ""
-  region:
-    description: "AWS region for the role and Lambda"
-    required: false
-    default: "us-east-2"
-  function-name:
-    description: "Code-signer Lambda function name"
-    required: false
-    default: "code-signer"
-  key-prefix:
-    description: "S3 key prefix the caller is authorized to write under"
-    required: false
-    default: "code-signing/slackhq/nebula"
-
-runs:
-  using: composite
-  steps:
-    - name: Skip notice
-      if: inputs.role == '' || inputs.bucket == ''
-      shell: sh
-      run: echo "::notice::code-signer role or bucket not set; skipping code signing."
-
-    - name: Configure AWS credentials
-      if: inputs.role != '' && inputs.bucket != ''
-      uses: aws-actions/configure-aws-credentials@v6
-      with:
-        role-to-assume: ${{ inputs.role }}
-        aws-region: ${{ inputs.region }}
-        # Default is 12 retries to ride out IAM trust-policy propagation; once
-        # the role is stable we want a real misconfiguration to fail fast.
-        retry-max-attempts: 5
-
-    - name: Sign .exe files
-      if: inputs.role != '' && inputs.bucket != ''
-      shell: sh
-      env:
-        SIGN_PATH: ${{ inputs.path }}
-        BUCKET: ${{ inputs.bucket }}
-        FUNCTION_NAME: ${{ inputs.function-name }}
-        KEY_PREFIX: ${{ inputs.key-prefix }}
-      run: |
-        set -eu
-        RUN="${GITHUB_RUN_ID}-${GITHUB_RUN_ATTEMPT}"
-
-        find "$SIGN_PATH" -name '*.exe' -print | while read -r path
-        do
-            rel=${path#"$SIGN_PATH"/}
-            file=$(basename "$path")
-            name=${file%.exe}
-            prefix="${KEY_PREFIX}/${RUN}"
-            src="${prefix}/unsigned/${rel}"
-            dst="${prefix}/signed/${rel}"
-
-            echo "::group::Sign ${rel}"
-            echo "Uploading unsigned to s3://${BUCKET}/${src}"
-            aws s3 cp --no-progress "$path" "s3://${BUCKET}/${src}" >/dev/null
-
-            echo "Invoking ${FUNCTION_NAME} Lambda"
-            payload=$(jq -nc \
-                --arg s "$src" \
-                --arg d "$dst" \
-                --arg p "$name" \
-                '{source_key: $s, dest_key: $d, program_name: $p}')
-            meta=$(aws lambda invoke \
-                --function-name "$FUNCTION_NAME" \
-                --cli-binary-format raw-in-base64-out \
-                --payload "$payload" \
-                --output json \
-                /tmp/sign-resp.json)
-            if echo "$meta" | jq -e '.FunctionError != null' >/dev/null
-            then
-                echo "::endgroup::"
-                echo "::error::code-signer Lambda failed for ${rel}"
-                cat /tmp/sign-resp.json >&2
-                exit 1
-            fi
-
-            echo "Downloading signed back to ${path}"
-            aws s3 cp --no-progress "s3://${BUCKET}/${dst}" "$path" >/dev/null
-
-            aws s3 rm "s3://${BUCKET}/${src}" >/dev/null 2>&1 || true
-            aws s3 rm "s3://${BUCKET}/${dst}" >/dev/null 2>&1 || true
-
-            # Sanity-check the bytes we got back actually carry an Authenticode
-            # signature that this machine can validate end to end.
-            status=$(powershell -NoProfile -Command "(Get-AuthenticodeSignature -FilePath '$path').Status" | tr -d '\r')
-            if [ "$status" != "Valid" ]
-            then
-                echo "::endgroup::"
-                echo "::error::${rel} signature status: ${status} (expected Valid)"
-                exit 1
-            fi
-
-            echo "Signed ${rel} (sha256=$(jq -r '.sha256' /tmp/sign-resp.json), status=${status})"
-            echo "::endgroup::"
-        done

.github/workflows/release.yml

@@ -24,7 +24,7 @@ jobs:
           mv build/*.tar.gz release
 
       - name: Upload artifacts
-        uses: actions/upload-artifact@v7
+        uses: actions/upload-artifact@v6
         with:
           name: linux-latest
           path: release
@@ -32,9 +32,6 @@ jobs:
   build-windows:
     name: Build Windows
     runs-on: windows-latest
-    permissions:
-      id-token: write
-      contents: read
     steps:
       - uses: actions/checkout@v6
 
@@ -57,15 +54,8 @@ jobs:
           mkdir build\dist\windows
           mv dist\windows\wintun build\dist\windows\
 
-      - name: Code-sign
-        uses: ./.github/actions/code-sign
-        with:
-          path: build
-          role: ${{ secrets.DEFINED_CODE_SIGNER_ROLE }}
-          bucket: ${{ secrets.DEFINED_CODE_SIGNER_BUCKET }}
-
       - name: Upload artifacts
-        uses: actions/upload-artifact@v7
+        uses: actions/upload-artifact@v6
         with:
           name: windows-latest
           path: build
@@ -85,7 +75,7 @@ jobs:
 
       - name: Import certificates
         if: env.HAS_SIGNING_CREDS == 'true'
-        uses: Apple-Actions/import-codesign-certs@v7
+        uses: Apple-Actions/import-codesign-certs@v6
         with:
           p12-file-base64: ${{ secrets.APPLE_DEVELOPER_CERTIFICATE_P12_BASE64 }}
           p12-password: ${{ secrets.APPLE_DEVELOPER_CERTIFICATE_PASSWORD }}
@@ -114,7 +104,7 @@ jobs:
           fi
 
       - name: Upload artifacts
-        uses: actions/upload-artifact@v7
+        uses: actions/upload-artifact@v6
         with:
           name: darwin-latest
           path: ./release/*
@@ -138,21 +128,21 @@ jobs:
 
       - name: Download artifacts
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
-        uses: actions/download-artifact@v8
+        uses: actions/download-artifact@v7
         with:
           name: linux-latest
           path: artifacts
 
       - name: Login to Docker Hub
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
-        uses: docker/login-action@v4
+        uses: docker/login-action@v3
         with:
           username: ${{ vars.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
 
       - name: Set up Docker Buildx
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
-        uses: docker/setup-buildx-action@v4
+        uses: docker/setup-buildx-action@v3
 
       - name: Build and push images
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
@@ -173,7 +163,7 @@ jobs:
       - uses: actions/checkout@v6
 
       - name: Download artifacts
-        uses: actions/download-artifact@v8
+        uses: actions/download-artifact@v7
         with:
           path: artifacts
 

.github/workflows/smoke-extra.yml

@@ -14,18 +14,10 @@ on:
       - 'go.sum'
 jobs:
 
-  smoke-extra-libvirt:
+  smoke-extra:
     if: github.ref == 'refs/heads/master' || contains(github.event.pull_request.labels.*.name, 'smoke-test-extra')
-    name: ${{ matrix.target }}
+    name: Run extra smoke tests
     runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        target:
-          - freebsd-amd64
-          - openbsd-amd64
-          - netbsd-amd64
-          - linux-amd64-ipv6disable
     env:
       VAGRANT_DEFAULT_PROVIDER: libvirt
     steps:
@@ -48,85 +40,28 @@ jobs:
         sudo chmod 666 /var/run/libvirt/libvirt-sock
         vagrant plugin install vagrant-libvirt
 
-    - name: ${{ matrix.target }}
-      run: make smoke-vagrant/${{ matrix.target }}
+    - name: freebsd-amd64
+      run: make smoke-vagrant/freebsd-amd64
 
-    timeout-minutes: 30
+    - name: openbsd-amd64
+      run: make smoke-vagrant/openbsd-amd64
 
-  # linux-386 needs VirtualBox, which conflicts with KVM/libvirt -- isolated job.
-  smoke-extra-virtualbox:
-    if: github.ref == 'refs/heads/master' || contains(github.event.pull_request.labels.*.name, 'smoke-test-extra')
-    name: linux-386
-    runs-on: ubuntu-latest
-    env:
-      VAGRANT_DEFAULT_PROVIDER: virtualbox
-    steps:
+    - name: netbsd-amd64
+      run: make smoke-vagrant/netbsd-amd64
 
-    - uses: actions/checkout@v6
+    - name: linux-amd64-ipv6disable
+      run: make smoke-vagrant/linux-amd64-ipv6disable
 
-    - uses: actions/setup-go@v6
-      with:
-        go-version: '1.25'
-        check-latest: true
-
-    - name: add hashicorp source
-      run: wget -O- https://apt.releases.hashicorp.com/gpg | gpg --dearmor | sudo tee /usr/share/keyrings/hashicorp-archive-keyring.gpg && echo "deb [signed-by=/usr/share/keyrings/hashicorp-archive-keyring.gpg] https://apt.releases.hashicorp.com $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/hashicorp.list
-
-    - name: install vagrant and virtualbox
+    # linux-386 runs last because it requires disabling KVM to use VirtualBox,
+    # which prevents libvirt (used by the other tests) from working after this point.
+    - name: install virtualbox for i386 test
       run: |
-        sudo apt-get update && sudo apt-get install -y vagrant virtualbox
+        sudo apt-get install -y virtualbox
         sudo rmmod kvm_amd kvm_intel kvm 2>/dev/null || true
 
     - name: linux-386
+      env:
+        VAGRANT_DEFAULT_PROVIDER: virtualbox
       run: make smoke-vagrant/linux-386
 
     timeout-minutes: 30
-
-  smoke-windows:
-    if: github.ref == 'refs/heads/master' || contains(github.event.pull_request.labels.*.name, 'smoke-test-extra')
-    name: Run windows smoke test
-    runs-on: windows-latest
-    steps:
-
-    - uses: actions/checkout@v6
-
-    - uses: actions/setup-go@v6
-      with:
-        go-version: '1.25'
-        check-latest: true
-
-    # WSL2 + Ubuntu so the smoke can run a real linux peer with its own
-    # netns. iputils-ping is needed for the in-WSL ping check. WSL1 has no
-    # real kernel and would lack /dev/net/tun, so we have to force WSL2.
-    - uses: Vampire/setup-wsl@v3
-      with:
-        distribution: Ubuntu-24.04
-        additional-packages: iputils-ping iproute2
-
-    # Vampire/setup-wsl provisions WSL1 even when the WSL2 platform is present.
-    # Convert the distro to WSL2 explicitly before we try to use /dev/net/tun.
-    - name: convert distro to WSL2
-      shell: pwsh
-      run: |
-        wsl --set-version Ubuntu-24.04 2
-        wsl --shutdown
-        wsl --list --verbose
-
-    - name: build windows nebula
-      run: make bin-windows
-
-    - name: build linux nebula for WSL
-      shell: bash
-      env:
-        GOOS: linux
-        GOARCH: amd64
-      run: |
-        mkdir -p build/linux-amd64
-        go build -o build/linux-amd64/nebula ./cmd/nebula
-
-    - name: run smoke-windows
-      shell: pwsh
-      working-directory: ./.github/workflows/smoke
-      run: ./smoke-windows.ps1
-
-    timeout-minutes: 15

.github/workflows/smoke/smoke-windows.ps1

@@ -1,272 +0,0 @@
-#!/usr/bin/env pwsh
-# Windows smoke test for the nebula tun + UDP + NLM code paths.
-#
-# Topology:
-#   - lighthouse runs natively on the Windows host (wintun + windows UDP)
-#   - peer runs inside WSL2 (Linux build of nebula, /dev/net/tun)
-#
-# WSL2 gives us a real netns boundary so the loopback fast-path on Windows
-# does not short-circuit the overlay -- when WSL pings the lighthouse VPN IP,
-# Linux has no idea that IP is local to the Windows host, so the packet is
-# forced through nebula. Same in reverse.
-
-$ErrorActionPreference = 'Stop'
-
-# wsl.exe emits UTF-16 LE by default which PowerShell reads as bytes, mangling
-# every captured string. WSL_UTF8 makes wsl.exe emit UTF-8 instead.
-$env:WSL_UTF8 = '1'
-
-$RepoRoot = Resolve-Path "$PSScriptRoot\..\..\.."
-$Nebula = Join-Path $RepoRoot 'nebula.exe'
-$NebulaCert = Join-Path $RepoRoot 'nebula-cert.exe'
-$NebulaLinux = Join-Path $RepoRoot 'build\linux-amd64\nebula'
-
-if (-not (Test-Path $Nebula)) { throw "missing $Nebula; run 'make bin-windows' first" }
-if (-not (Test-Path $NebulaCert)) { throw "missing $NebulaCert; run 'make bin-windows' first" }
-if (-not (Test-Path $NebulaLinux)) { throw "missing $NebulaLinux; build the linux nebula first" }
-
-# Matches the distro installed by Vampire/setup-wsl in smoke-extra.yml.
-$Distro = 'Ubuntu-24.04'
-$listed = (wsl --list --quiet 2>$null) -join "`n"
-if ($listed -notmatch [regex]::Escape($Distro)) {
-    throw "WSL distro $Distro not registered. Got: $listed"
-}
-Write-Host "Using WSL distro: $Distro"
-
-# Windows host as seen from inside WSL: WSL's default-route gateway. We extract
-# it with a regex rather than awk fields so PowerShell does not eat any '$N'
-# tokens, and tabs/double-spaces in `ip route` output do not confuse a cut.
-$ipCmd = 'ip route show default | grep -oE "([0-9]+\.){3}[0-9]+" | head -1'
-$WindowsIp = (wsl -d $Distro -- bash -c $ipCmd).Trim()
-if (-not $WindowsIp) { throw "could not determine Windows host IP from WSL" }
-Write-Host "Windows host IP from WSL: $WindowsIp"
-
-$WorkDir = Join-Path $env:TEMP 'nebula-smoke-windows'
-if (Test-Path $WorkDir) { Remove-Item -Recurse -Force $WorkDir }
-New-Item -ItemType Directory -Path $WorkDir | Out-Null
-
-$WslDir = '/tmp/nebula-smoke'
-wsl -d $Distro -- bash -c "rm -rf $WslDir && mkdir -p $WslDir" | Out-Null
-
-$DevName = 'nebula-smoke'
-$Ip1 = '192.168.241.1'
-$Ip2 = '192.168.241.2'
-$Port = 4242
-
-& $NebulaCert ca -name 'smoke-ca' -out-crt "$WorkDir\ca.crt" -out-key "$WorkDir\ca.key"
-if ($LASTEXITCODE -ne 0) { throw "nebula-cert ca failed (exit $LASTEXITCODE)" }
-
-& $NebulaCert sign -name 'lighthouse' -networks "$Ip1/24" -ca-crt "$WorkDir\ca.crt" -ca-key "$WorkDir\ca.key" -out-crt "$WorkDir\lighthouse.crt" -out-key "$WorkDir\lighthouse.key"
-if ($LASTEXITCODE -ne 0) { throw "nebula-cert sign lighthouse failed (exit $LASTEXITCODE)" }
-
-& $NebulaCert sign -name 'peer' -networks "$Ip2/24" -ca-crt "$WorkDir\ca.crt" -ca-key "$WorkDir\ca.key" -out-crt "$WorkDir\peer.crt" -out-key "$WorkDir\peer.key"
-if ($LASTEXITCODE -ne 0) { throw "nebula-cert sign peer failed (exit $LASTEXITCODE)" }
-
-# Windows lighthouse config.
-@"
-pki:
-  ca: $WorkDir\ca.crt
-  cert: $WorkDir\lighthouse.crt
-  key: $WorkDir\lighthouse.key
-static_host_map: {}
-lighthouse:
-  am_lighthouse: true
-  interval: 60
-  hosts: []
-listen:
-  host: 0.0.0.0
-  port: $Port
-tun:
-  disabled: false
-  dev: $DevName
-  drop_local_broadcast: false
-  drop_multicast: false
-  tx_queue: 500
-  mtu: 1300
-  network_category: private
-logging:
-  level: info
-  format: text
-firewall:
-  outbound_action: drop
-  inbound_action: drop
-  conntrack:
-    tcp_timeout: 12m
-    udp_timeout: 3m
-    default_timeout: 10m
-  outbound:
-    - port: any
-      proto: any
-      host: any
-  inbound:
-    - port: any
-      proto: any
-      host: any
-"@ | Out-File -FilePath "$WorkDir\lighthouse.yml" -Encoding utf8
-
-# WSL peer config (paths are POSIX, deliberately).
-@"
-pki:
-  ca: $WslDir/ca.crt
-  cert: $WslDir/peer.crt
-  key: $WslDir/peer.key
-static_host_map:
-  "${Ip1}": ["${WindowsIp}:$Port"]
-lighthouse:
-  am_lighthouse: false
-  interval: 60
-  hosts:
-    - "${Ip1}"
-listen:
-  host: 0.0.0.0
-  port: 0
-tun:
-  disabled: false
-  dev: nebula1
-  drop_local_broadcast: false
-  drop_multicast: false
-  tx_queue: 500
-  mtu: 1300
-logging:
-  level: info
-  format: text
-firewall:
-  outbound_action: drop
-  inbound_action: drop
-  conntrack:
-    tcp_timeout: 12m
-    udp_timeout: 3m
-    default_timeout: 10m
-  outbound:
-    - port: any
-      proto: any
-      host: any
-  inbound:
-    - port: any
-      proto: any
-      host: any
-"@ | Out-File -FilePath "$WorkDir\peer.yml" -Encoding utf8
-
-# Stage WSL artifacts. Convert Windows paths to WSL paths ourselves rather than
-# calling `wslpath`, because PowerShell's argument-passing to external EXEs
-# strips backslashes from path arguments in ways that are hard to escape around.
-function ConvertTo-WslPath {
-    param([string]$WindowsPath)
-    if ($WindowsPath -notmatch '^([A-Za-z]):\\(.*)$') {
-        throw "cannot convert path to WSL: $WindowsPath"
-    }
-    return "/mnt/$($matches[1].ToLower())/$($matches[2].Replace('\','/'))"
-}
-
-$WslWorkDir = ConvertTo-WslPath $WorkDir
-$WslNebulaPath = ConvertTo-WslPath $NebulaLinux
-wsl -d $Distro -- bash -c "cp '$WslWorkDir/ca.crt' '$WslWorkDir/peer.crt' '$WslWorkDir/peer.key' '$WslWorkDir/peer.yml' $WslDir/ && cp '$WslNebulaPath' $WslDir/nebula && chmod +x $WslDir/nebula"
-
-# Make sure WSL has tun support and /dev/net/tun is usable before starting
-# nebula. Diagnostics first so a fail here points at the real problem (e.g.
-# WSL1 distros do not have a real kernel and will not have tun).
-Write-Host '=== WSL diagnostic ==='
-wsl --version 2>&1 | Out-Host
-wsl --list --verbose 2>&1 | Out-Host
-wsl -d $Distro -u root -- uname -a | Out-Host
-wsl -d $Distro -u root -- bash -c "modprobe tun 2>&1 || true; mkdir -p /dev/net; [ -c /dev/net/tun ] || mknod /dev/net/tun c 10 200; chmod 600 /dev/net/tun; ls -l /dev/net/tun"
-if ($LASTEXITCODE -ne 0) { throw "failed to prepare /dev/net/tun in WSL (TUN support missing?)" }
-
-# Deliberately no New-NetFirewallRule calls here -- nebula's windows_bypass_wdf
-# feature is supposed to install WFP permit filters that let inbound traffic
-# through Windows Defender Firewall on its own. If this smoke regresses, that
-# feature regressed.
-
-$lhOut = Join-Path $WorkDir 'lighthouse.out.log'
-$lhErr = Join-Path $WorkDir 'lighthouse.err.log'
-$lhProc = Start-Process -FilePath $Nebula -ArgumentList @('-config', "$WorkDir\lighthouse.yml") `
-    -PassThru -NoNewWindow `
-    -RedirectStandardOutput $lhOut `
-    -RedirectStandardError $lhErr
-
-# Run nebula in WSL as root with no sudo + no shell wrapper. PowerShell's
-# Start-Process arg quoting mangles `bash -c "..."` strings that contain
-# spaces/redirections, so we skip bash entirely and let Start-Process do the
-# stdout/stderr capture itself.
-$peerOut = Join-Path $WorkDir 'peer.out.log'
-$peerErr = Join-Path $WorkDir 'peer.err.log'
-$peerProc = Start-Process -FilePath 'wsl' `
-    -ArgumentList @('-d', $Distro, '-u', 'root', '--', "$WslDir/nebula", '-config', "$WslDir/peer.yml") `
-    -PassThru -NoNewWindow `
-    -RedirectStandardOutput $peerOut `
-    -RedirectStandardError $peerErr
-
-function Wait-Until {
-    param([scriptblock]$Predicate, [int]$TimeoutSec, [string]$What)
-    $deadline = (Get-Date).AddSeconds($TimeoutSec)
-    while ((Get-Date) -lt $deadline) {
-        if (& $Predicate) { return }
-        Start-Sleep -Milliseconds 500
-    }
-    throw "timed out waiting for: $What"
-}
-
-try {
-    Wait-Until -TimeoutSec 30 -What "windows wintun adapter $DevName with NetworkCategory=Private" -Predicate {
-        if ($lhProc.HasExited) { throw "lighthouse exited (code $($lhProc.ExitCode)) before tun was ready" }
-        $p = Get-NetConnectionProfile -InterfaceAlias $DevName -ErrorAction SilentlyContinue
-        $p -and ("$($p.NetworkCategory)" -ieq 'Private')
-    }
-    Write-Host "OK: $DevName NetworkCategory=Private"
-
-    Wait-Until -TimeoutSec 30 -What "WSL nebula1 with $Ip2" -Predicate {
-        if ($peerProc.HasExited) { throw "peer exited (code $($peerProc.ExitCode)) before tun was ready" }
-        $r = wsl -d $Distro -u root -- bash -c "ip -o addr show nebula1 2>/dev/null | grep -q 'inet $Ip2' && echo yes"
-        ("$r").Trim() -eq 'yes'
-    }
-    Write-Host "OK: WSL nebula1 has $Ip2"
-
-    Wait-Until -TimeoutSec 30 -What "ping from WSL peer to windows lighthouse ($Ip1)" -Predicate {
-        if ($peerProc.HasExited) { throw "peer exited (code $($peerProc.ExitCode)) before ping succeeded" }
-        $r = wsl -d $Distro -u root -- bash -c "ping -c1 -W1 $Ip1 >/dev/null 2>&1 && echo OK"
-        ("$r").Trim() -eq 'OK'
-    }
-    Write-Host "OK: WSL peer -> windows lighthouse"
-
-    Wait-Until -TimeoutSec 30 -What "ping from windows lighthouse to WSL peer ($Ip2)" -Predicate {
-        $null = & ping.exe -n 1 -w 1000 $Ip2
-        $LASTEXITCODE -eq 0
-    }
-    Write-Host "OK: windows lighthouse -> WSL peer"
-
-    Write-Host ''
-    Write-Host 'All smoke checks passed.'
-}
-catch {
-    Write-Host ''
-    Write-Host '=== lighthouse stdout ==='
-    Get-Content $lhOut -ErrorAction SilentlyContinue | Out-Host
-    Write-Host '=== lighthouse stderr ==='
-    Get-Content $lhErr -ErrorAction SilentlyContinue | Out-Host
-    Write-Host '=== peer stdout ==='
-    Get-Content $peerOut -ErrorAction SilentlyContinue | Out-Host
-    Write-Host '=== peer stderr ==='
-    Get-Content $peerErr -ErrorAction SilentlyContinue | Out-Host
-    Write-Host '=== nebula WFP filters ==='
-    # Dump nebula-installed filters so we can verify they got registered with
-    # the conditions we expect.
-    $wfpDump = Join-Path $WorkDir 'wfp.xml'
-    netsh wfp show filters file=$wfpDump 2>&1 | Out-Null
-    if (Test-Path $wfpDump) {
-        Select-String -Path $wfpDump -Pattern 'Nebula' -Context 0,80 -ErrorAction SilentlyContinue | Out-Host
-    }
-    throw
-}
-finally {
-    if (-not $lhProc.HasExited) {
-        Stop-Process -Id $lhProc.Id -Force -ErrorAction SilentlyContinue
-        $lhProc.WaitForExit(5000) | Out-Null
-    }
-    wsl -d $Distro -u root -- bash -c "pkill -f $WslDir/nebula 2>/dev/null; true" | Out-Null
-    # pkill returns 1 when no match and wsl propagates that; the smoke is done
-    # so we don't want it to leak into the script's exit code.
-    $global:LASTEXITCODE = 0
-    if ($peerProc -and -not $peerProc.HasExited) {
-        Stop-Process -Id $peerProc.Id -Force -ErrorAction SilentlyContinue
-    }
-}

.github/workflows/smoke/smoke.sh

@@ -82,7 +82,7 @@ docker exec host4 tcpdump -i eth0 -q -w - -U 2>logs/host4.outside.log >logs/host
 
 docker exec host2 ncat -nklv 0.0.0.0 2000 &
 docker exec host3 ncat -nklv 0.0.0.0 2000 &
-docker exec host4 ncat -e '/usr/bin/echo helloagainfromhost4' -nkluv 0.0.0.0 4000 &
+docker exec host4 ncat -nkluv 0.0.0.0 4000 &
 docker exec host2 ncat -e '/usr/bin/echo host2' -nkluv 0.0.0.0 3000 &
 docker exec host3 ncat -e '/usr/bin/echo host3' -nkluv 0.0.0.0 3000 &
 
@@ -155,11 +155,11 @@ echo " *** Testing conntrack"
 echo
 set -x
 
-# host4's outbound firewall only allows ICMP to the lighthouse, so host4
-# cannot initiate UDP to host2. Once host2 initiates a flow to host4:4000,
-# conntrack must let host4's listener reply on that flow. If it doesn't,
-# the echo back from host4 never reaches host2.
-docker exec host2 sh -c "(/usr/bin/echo host2; sleep 2) | ncat -nuv 192.168.100.4 4000" | grep -q helloagainfromhost4
+# host2 speaking to host4 on UDP 4000 should allow it to reply, when firewall rules would normally not permit this
+docker exec host2 sh -c "/usr/bin/echo host2 | ncat -nuv 192.168.100.4 4000"
+docker exec host2 ncat -e '/usr/bin/echo helloagainfromhost2' -nkluv 0.0.0.0 4000 &
+sleep 1
+docker exec host4 sh -c "/usr/bin/echo host4 | ncat -nuv 192.168.100.2 4000"
 
 docker exec host4 sh -c 'kill 1'
 docker exec host3 sh -c 'kill 1'

.github/workflows/smoke/vagrant-netbsd-amd64/Vagrantfile

@@ -1,7 +1,7 @@
 # -*- mode: ruby -*-
 # vi: set ft=ruby :
 Vagrant.configure("2") do |config|
-  config.vm.box = "DefinedNet/netbsd10"
+  config.vm.box = "generic/netbsd9"
 
   config.vm.synced_folder "../build", "/nebula", type: "rsync"
 end

.github/workflows/test.yml

@@ -45,7 +45,7 @@ jobs:
     - name: Build test mobile
       run: make build-test-mobile
 
-    - uses: actions/upload-artifact@v7
+    - uses: actions/upload-artifact@v6
       with:
         name: e2e packet flow linux-latest
         path: e2e/mermaid/linux-latest
@@ -125,7 +125,7 @@ jobs:
     - name: End 2 end
       run: make e2evv
 
-    - uses: actions/upload-artifact@v7
+    - uses: actions/upload-artifact@v6
       with:
         name: e2e packet flow ${{ matrix.os }}
         path: e2e/mermaid/${{ matrix.os }}

.golangci.yaml

@@ -2,21 +2,7 @@ version: "2"
 linters:
   default: none
   enable:
-    - sloglint
     - testifylint
-  settings:
-    sloglint:
-      # Enforce key-value pair form for Info/Debug/Warn/Error/Log/With and
-      # the package-level slog equivalents. Use l.Log(ctx, level, ...) for
-      # custom levels instead of LogAttrs when you can.
-      #
-      # LogAttrs is also flagged by this rule because it takes ...slog.Attr;
-      # the few legitimate sites (where attrs is built up as a []slog.Attr)
-      # carry a //nolint:sloglint with rationale.
-      kv-only: true
-      # no-mixed-args is on by default: forbids mixing kv and attrs in one call.
-      # discard-handler is on by default (since Go 1.24): suggests
-      # slog.DiscardHandler over slog.NewTextHandler(io.Discard, nil).
   exclusions:
     generated: lax
     presets:

bits.go

@@ -1,43 +1,23 @@
 package nebula
 
 import (
-	"context"
-	"fmt"
-	"log/slog"
-	"math"
-	mathbits "math/bits"
-
 	"github.com/rcrowley/go-metrics"
+	"github.com/sirupsen/logrus"
 )
 
-const bitsPerWord = 64
-
-// Bits is a sliding-window anti-replay tracker. The window is stored as a
-// circular bitmap packed into uint64 words (8x denser than a []bool), so a
-// length-N window costs N/8 bytes. length must be a power of two.
 type Bits struct {
 	length             uint64
-	lengthMask         uint64
 	current            uint64
-	bits               []uint64
+	bits               []bool
 	lostCounter        metrics.Counter
 	dupeCounter        metrics.Counter
 	outOfWindowCounter metrics.Counter
 }
 
-func NewBits(length uint64) *Bits {
-	if length == 0 || length&(length-1) != 0 {
-		panic(fmt.Sprintf("Bits length must be a power of two, got %d", length))
-	}
-
-	nWords := length / bitsPerWord
-	if nWords == 0 {
-		nWords = 1
-	}
+func NewBits(bits uint64) *Bits {
 	b := &Bits{
-		length:             length,
-		lengthMask:         length - 1,
-		bits:               make([]uint64, nWords),
+		length:             bits,
+		bits:               make([]bool, bits, bits),
 		current:            0,
 		lostCounter:        metrics.GetOrRegisterCounter("network.packets.lost", nil),
 		dupeCounter:        metrics.GetOrRegisterCounter("network.packets.duplicate", nil),
@@ -45,219 +25,88 @@ func NewBits(length uint64) *Bits {
 	}
 
 	// There is no counter value 0, mark it to avoid counting a lost packet later.
-	b.bits[0] = 1
+	b.bits[0] = true
+	b.current = 0
 	return b
 }
 
-func (b *Bits) get(i uint64) bool {
-	pos := i & b.lengthMask
-	//bit-shifting by 6 because i is a bit index, not a u64 index, and we need to find the u64 without bit in it
-	return b.bits[pos>>6]&(uint64(1)<<(pos&63)) != 0
-}
-
-func (b *Bits) set(i uint64) {
-	pos := i & b.lengthMask
-	b.bits[pos>>6] |= uint64(1) << (pos & 63)
-}
-
-// clearRange clears `count` bits starting at circular position `startPos`
-// (already masked to [0, length)) and returns how many of them were set
-// before the clear. count must be in [1, length].
-func (b *Bits) clearRange(startPos, count uint64) uint64 {
-	wasSet := uint64(0)
-	if count >= b.length {
-		for _, w := range b.bits {
-			wasSet += uint64(mathbits.OnesCount64(w))
-		}
-		clear(b.bits)
-		return wasSet
-	}
-
-	pos := startPos
-	remaining := count
-
-	// handle the potential partial word before pos becomes u64 aligned
-	word := pos >> 6
-	bit := pos & 63
-	take := uint64(64) - bit
-	if take > remaining {
-		take = remaining
-	}
-	if take > b.length-pos {
-		take = b.length - pos
-	}
-	var mask uint64
-	if take == 64 {
-		mask = math.MaxUint64
-	} else {
-		mask = ((uint64(1) << take) - 1) << bit
-	}
-	wasSet += uint64(mathbits.OnesCount64(b.bits[word] & mask))
-	b.bits[word] &^= mask
-	remaining -= take
-	pos = (pos + take) & b.lengthMask
-
-	// Clear whole words, keeping track of the number of set bits
-	for remaining >= 64 {
-		word = pos >> 6
-		wasSet += uint64(mathbits.OnesCount64(b.bits[word]))
-		b.bits[word] = 0
-		remaining -= 64
-		pos = (pos + 64) & b.lengthMask
-	}
-
-	// Clear the remaining partial word
-	if remaining > 0 {
-		word = pos >> 6
-		mask = (uint64(1) << remaining) - 1
-		wasSet += uint64(mathbits.OnesCount64(b.bits[word] & mask))
-		b.bits[word] &^= mask
-	}
-
-	return wasSet
-}
-
-func (b *Bits) strictlyWithinWindow(i uint64) bool {
-	// Handle the case where the window hasn't slid yet. This avoids u64 underflow.
-	inWarmup := b.current < b.length
-	if i < b.length && inWarmup {
-		return true
-	}
-
-	// Next, if the packet is in-window, see if we've seen it before
-	if i > b.current-b.length {
-		return true
-	}
-	return false //not within window!
-}
-
-// Check returns true if i is within (or way out in front of) the window, and not a replay
-func (b *Bits) Check(l *slog.Logger, i uint64) bool {
+func (b *Bits) Check(l *logrus.Logger, i uint64) bool {
 	// If i is the next number, return true.
 	if i > b.current {
 		return true
 	}
 
-	if b.strictlyWithinWindow(i) {
-		return !b.get(i)
+	// If i is within the window, check if it's been set already.
+	if i > b.current-b.length || i < b.length && b.current < b.length {
+		return !b.bits[i%b.length]
 	}
 
 	// Not within the window
-	if l.Enabled(context.Background(), slog.LevelDebug) {
-		l.Debug("rejected a packet (top)", "current", b.current, "incoming", i)
+	if l.Level >= logrus.DebugLevel {
+		l.Debugf("rejected a packet (top) %d %d\n", b.current, i)
 	}
 	return false
 }
 
-// Update has three branches:
-//   - i == b.current+1: fast path; advance the cursor by one and lose-count
-//     the slot we just stomped (only past warmup; see the i > b.length guard
-//     below).
-//   - i  >  b.current+1: jump path; clear all slots between current and i
-//     (or up to a full window's worth, whichever is smaller) via clearRange,
-//     then mark i. Two arms here: a warmup arm that handles the very first
-//     window before the cursor has slid, and a steady-state arm that treats
-//     every cleared empty slot as a lost packet.
-//   - i  <= b.current: in-window check for duplicates; out-of-window otherwise.
-//
-// NewBits seeds bits[0]=1 so counter 0 looks "received" — Update never
-// clears that marker during warmup (clearRange skips position 0 when
-// startPos=1), and once b.current >= b.length the marker is no longer
-// consulted. The marker prevents a fictitious "lost" hit on the first real
-// counter.
-func (b *Bits) Update(l *slog.Logger, i uint64) bool {
-	// Fast path: i is the next expected counter. Split out so the function
-	// stays small and avoids paying for the slow paths' slog argument-build
-	// stack frame on every call. The bit read/test/write is inlined to
-	// touch the backing word once.
+func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
+	// If i is the next number, return true and update current.
 	if i == b.current+1 {
-		pos := i & b.lengthMask
-		word := pos >> 6
-		mask := uint64(1) << (pos & 63)
-		w := b.bits[word]
-		if i > b.length && w&mask == 0 {
+		// Check if the oldest bit was lost since we are shifting the window by 1 and occupying it with this counter
+		// The very first window can only be tracked as lost once we are on the 2nd window or greater
+		if b.bits[i%b.length] == false && i > b.length {
 			b.lostCounter.Inc(1)
 		}
-		b.bits[word] = w | mask
+		b.bits[i%b.length] = true
 		b.current = i
 		return true
 	}
-	return b.updateSlow(l, i)
-}
 
-// updateSlow handles jumps, in-window backfill, dupes, and out-of-window.
-func (b *Bits) updateSlow(l *slog.Logger, i uint64) bool {
 	// If i is a jump, adjust the window, record lost, update current, and return true
 	if i > b.current {
-		end := i
-		if end > b.current+b.length {
-			end = b.current + b.length
-		}
-		count := end - b.current
-		startPos := (b.current + 1) & b.lengthMask
-
-		var lost int64
-		if b.current >= b.length {
-			// Steady state: every cleared slot is past warmup, so any unset
-			// bit we evict is a lost packet from the previous cycle.
-			wasSet := b.clearRange(startPos, count)
-			lost = int64(count) - int64(wasSet)
-		} else {
-			// Warmup (the very first window). Some cleared slots represent
-			// packets <= length where eviction is not "lost" in the usual
-			// sense. This branch is taken at most once per connection so we
-			// don't bother optimizing it.
-			for n := b.current + 1; n <= end; n++ {
-				if !b.get(n) && n > b.length {
-					lost++
-				}
+		lost := int64(0)
+		// Zero out the bits between the current and the new counter value, limited by the window size,
+		// since the window is shifting
+		for n := b.current + 1; n <= min(i, b.current+b.length); n++ {
+			if b.bits[n%b.length] == false && n > b.length {
+				lost++
 			}
-			b.clearRange(startPos, count)
+			b.bits[n%b.length] = false
 		}
 
-		// Anything past the new window can never be backfilled, so it's lost.
-		if i > b.current+b.length {
-			lost += int64(i - b.current - b.length)
-		}
+		// Only record any skipped packets as a result of the window moving further than the window length
+		// Any loss within the new window will be accounted for in future calls
+		lost += max(0, int64(i-b.current-b.length))
 		b.lostCounter.Inc(lost)
 
-		b.set(i)
+		b.bits[i%b.length] = true
 		b.current = i
 		return true
 	}
 
-	// If i is within the current window but below the current counter, check to see if it's a duplicate
-	if b.strictlyWithinWindow(i) {
-		pos := i & b.lengthMask
-		word := pos >> 6
-		mask := uint64(1) << (pos & 63)
-		w := b.bits[word]
-		if b.current == i || w&mask != 0 {
-			if l.Enabled(context.Background(), slog.LevelDebug) {
-				l.Debug("Receive window",
-					"accepted", false,
-					"currentCounter", b.current,
-					"incomingCounter", i,
-					"reason", "duplicate",
-				)
+	// If i is within the current window but below the current counter,
+	// Check to see if it's a duplicate
+	if i > b.current-b.length || i < b.length && b.current < b.length {
+		if b.current == i || b.bits[i%b.length] == true {
+			if l.Level >= logrus.DebugLevel {
+				l.WithField("receiveWindow", m{"accepted": false, "currentCounter": b.current, "incomingCounter": i, "reason": "duplicate"}).
+					Debug("Receive window")
 			}
 			b.dupeCounter.Inc(1)
 			return false
 		}
 
-		b.bits[word] = w | mask
+		b.bits[i%b.length] = true
 		return true
 	}
 
 	// In all other cases, fail and don't change current.
 	b.outOfWindowCounter.Inc(1)
-	if l.Enabled(context.Background(), slog.LevelDebug) {
-		l.Debug("Receive window",
-			"accepted", false,
-			"currentCounter", b.current,
-			"incomingCounter", i,
-			"reason", "nonsense",
-		)
+	if l.Level >= logrus.DebugLevel {
+		l.WithField("accepted", false).
+			WithField("currentCounter", b.current).
+			WithField("incomingCounter", i).
+			WithField("reason", "nonsense").
+			Debug("Receive window")
 	}
 	return false
 }

bits_test.go

@@ -7,79 +7,61 @@ import (
 	"github.com/stretchr/testify/assert"
 )
 
-// snapshot returns the bitmap as a []bool of length b.length, for readable
-// test assertions against the now-packed []uint64 storage.
-func (b *Bits) snapshot() []bool {
-	out := make([]bool, b.length)
-	for i := uint64(0); i < b.length; i++ {
-		out[i] = b.get(i)
-	}
-	return out
-}
-
-func TestBitsRequiresPowerOfTwo(t *testing.T) {
-	assert.Panics(t, func() { NewBits(10) })
-	assert.Panics(t, func() { NewBits(0) })
-	assert.NotPanics(t, func() { NewBits(1) })
-	assert.NotPanics(t, func() { NewBits(16) })
-	assert.NotPanics(t, func() { NewBits(1024) })
-	assert.NotPanics(t, func() { NewBits(16384) })
-}
-
 func TestBits(t *testing.T) {
 	l := test.NewLogger()
-	b := NewBits(16)
-	assert.EqualValues(t, 16, b.length)
+	b := NewBits(10)
+
+	// make sure it is the right size
+	assert.Len(t, b.bits, 10)
 
 	// This is initialized to zero - receive one. This should work.
 	assert.True(t, b.Check(l, 1))
 	assert.True(t, b.Update(l, 1))
 	assert.EqualValues(t, 1, b.current)
-	g := []bool{true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false}
-	assert.Equal(t, g, b.snapshot())
+	g := []bool{true, true, false, false, false, false, false, false, false, false}
+	assert.Equal(t, g, b.bits)
 
 	// Receive two
 	assert.True(t, b.Check(l, 2))
 	assert.True(t, b.Update(l, 2))
 	assert.EqualValues(t, 2, b.current)
-	g = []bool{true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false}
-	assert.Equal(t, g, b.snapshot())
+	g = []bool{true, true, true, false, false, false, false, false, false, false}
+	assert.Equal(t, g, b.bits)
 
 	// Receive two again - it will fail
 	assert.False(t, b.Check(l, 2))
 	assert.False(t, b.Update(l, 2))
 	assert.EqualValues(t, 2, b.current)
 
-	// Jump ahead to 25, which clears the window and sets slot 25%16 = 9.
-	assert.True(t, b.Check(l, 25))
-	assert.True(t, b.Update(l, 25))
-	assert.EqualValues(t, 25, b.current)
-	g = []bool{false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false}
-	assert.Equal(t, g, b.snapshot())
+	// Jump ahead to 15, which should clear everything and set the 6th element
+	assert.True(t, b.Check(l, 15))
+	assert.True(t, b.Update(l, 15))
+	assert.EqualValues(t, 15, b.current)
+	g = []bool{false, false, false, false, false, true, false, false, false, false}
+	assert.Equal(t, g, b.bits)
 
-	// Mark 24, which is in window (current 25, length 16, window covers [10,25]).
-	assert.True(t, b.Check(l, 24))
-	assert.True(t, b.Update(l, 24))
-	assert.EqualValues(t, 25, b.current)
-	g = []bool{false, false, false, false, false, false, false, false, true, true, false, false, false, false, false, false}
-	assert.Equal(t, g, b.snapshot())
+	// Mark 14, which is allowed because it is in the window
+	assert.True(t, b.Check(l, 14))
+	assert.True(t, b.Update(l, 14))
+	assert.EqualValues(t, 15, b.current)
+	g = []bool{false, false, false, false, true, true, false, false, false, false}
+	assert.Equal(t, g, b.bits)
 
-	// Mark 5, not allowed because 5 <= current-length (25-16=9).
+	// Mark 5, which is not allowed because it is not in the window
 	assert.False(t, b.Check(l, 5))
 	assert.False(t, b.Update(l, 5))
-	assert.EqualValues(t, 25, b.current)
-	g = []bool{false, false, false, false, false, false, false, false, true, true, false, false, false, false, false, false}
-	assert.Equal(t, g, b.snapshot())
+	assert.EqualValues(t, 15, b.current)
+	g = []bool{false, false, false, false, true, true, false, false, false, false}
+	assert.Equal(t, g, b.bits)
 
-	// Make sure we handle wrapping around once to the same slot. With
-	// length=16, packets 1 and 17 share slot 1.
-	b = NewBits(16)
+	// make sure we handle wrapping around once to the current position
+	b = NewBits(10)
 	assert.True(t, b.Update(l, 1))
-	assert.True(t, b.Update(l, 17))
-	assert.Equal(t, []bool{false, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false}, b.snapshot())
+	assert.True(t, b.Update(l, 11))
+	assert.Equal(t, []bool{false, true, false, false, false, false, false, false, false, false}, b.bits)
 
 	// Walk through a few windows in order
-	b = NewBits(16)
+	b = NewBits(10)
 	for i := uint64(1); i <= 100; i++ {
 		assert.True(t, b.Check(l, i), "Error while checking %v", i)
 		assert.True(t, b.Update(l, i), "Error while updating %v", i)
@@ -90,31 +72,24 @@ func TestBits(t *testing.T) {
 
 func TestBitsLargeJumps(t *testing.T) {
 	l := test.NewLogger()
-
-	// length=16. Update(55) from current=0:
-	//   warmup, per-bit loop sees no n>16 with unset bits (slot 0 was set by
-	//   NewBits and gets re-evaluated when n=16; n=16 is not strictly > 16),
-	//   so the loop contributes 0. The jump exceeds the window so we record
-	//   55 - 0 - 16 = 39 packets fell out the back.
-	b := NewBits(16)
+	b := NewBits(10)
 	b.lostCounter.Clear()
-	assert.True(t, b.Update(l, 55))
-	assert.Equal(t, int64(39), b.lostCounter.Count())
 
-	// Update(100): clears 16 slots starting at slot 56%16=8. Only slot 7 (for
-	// packet 55) was set, so 16 - 1 = 15 evicted slots had unset bits.
-	// Plus 100 - 55 - 16 = 29 packets fell past the window. Total 44.
-	assert.True(t, b.Update(l, 100))
-	assert.Equal(t, int64(39+44), b.lostCounter.Count())
+	b = NewBits(10)
+	b.lostCounter.Clear()
+	assert.True(t, b.Update(l, 55)) // We saw packet 55 and can still track 45,46,47,48,49,50,51,52,53,54
+	assert.Equal(t, int64(45), b.lostCounter.Count())
 
-	// Update(200): same shape: 16 - 1 = 15 evicted unset, plus 200 - 100 - 16 = 84 past window. Total 99.
-	assert.True(t, b.Update(l, 200))
-	assert.Equal(t, int64(39+44+99), b.lostCounter.Count())
+	assert.True(t, b.Update(l, 100)) // We saw packet 55 and 100 and can still track 90,91,92,93,94,95,96,97,98,99
+	assert.Equal(t, int64(89), b.lostCounter.Count())
+
+	assert.True(t, b.Update(l, 200)) // We saw packet 55, 100, and 200 and can still track 190,191,192,193,194,195,196,197,198,199
+	assert.Equal(t, int64(188), b.lostCounter.Count())
 }
 
 func TestBitsDupeCounter(t *testing.T) {
 	l := test.NewLogger()
-	b := NewBits(16)
+	b := NewBits(10)
 	b.lostCounter.Clear()
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
@@ -139,117 +114,120 @@ func TestBitsDupeCounter(t *testing.T) {
 
 func TestBitsOutOfWindowCounter(t *testing.T) {
 	l := test.NewLogger()
-	b := NewBits(16)
+	b := NewBits(10)
 	b.lostCounter.Clear()
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
 
-	// Jump to 20 (warmup branch + 4 past-window packets).
 	assert.True(t, b.Update(l, 20))
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 
-	// 9 single-step advances, each evicts a slot whose bit was cleared during
-	// the jump above and whose value was never seen, so each contributes 1
-	// to lostCounter.
-	for n := uint64(21); n <= 29; n++ {
-		assert.True(t, b.Update(l, n))
-	}
+	assert.True(t, b.Update(l, 21))
+	assert.True(t, b.Update(l, 22))
+	assert.True(t, b.Update(l, 23))
+	assert.True(t, b.Update(l, 24))
+	assert.True(t, b.Update(l, 25))
+	assert.True(t, b.Update(l, 26))
+	assert.True(t, b.Update(l, 27))
+	assert.True(t, b.Update(l, 28))
+	assert.True(t, b.Update(l, 29))
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 
-	// 0 is below current-length (29-16=13) so it falls outside the window.
 	assert.False(t, b.Update(l, 0))
 	assert.Equal(t, int64(1), b.outOfWindowCounter.Count())
 
-	// 4 from the Update(20) jump + 9 from 21..29.
-	assert.Equal(t, int64(13), b.lostCounter.Count())
+	assert.Equal(t, int64(19), b.lostCounter.Count()) // packet 0 wasn't lost
 	assert.Equal(t, int64(0), b.dupeCounter.Count())
 	assert.Equal(t, int64(1), b.outOfWindowCounter.Count())
 }
 
 func TestBitsLostCounter(t *testing.T) {
 	l := test.NewLogger()
-	b := NewBits(16)
+	b := NewBits(10)
 	b.lostCounter.Clear()
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
 
-	// Walk 20..29 like the original, just with a bigger window. Same
-	// reasoning as TestBitsOutOfWindowCounter: 4 past-window from Update(20),
-	// then 9 more from the unit advances.
-	for n := uint64(20); n <= 29; n++ {
-		assert.True(t, b.Update(l, n))
-	}
-	assert.Equal(t, int64(13), b.lostCounter.Count())
+	assert.True(t, b.Update(l, 20))
+	assert.True(t, b.Update(l, 21))
+	assert.True(t, b.Update(l, 22))
+	assert.True(t, b.Update(l, 23))
+	assert.True(t, b.Update(l, 24))
+	assert.True(t, b.Update(l, 25))
+	assert.True(t, b.Update(l, 26))
+	assert.True(t, b.Update(l, 27))
+	assert.True(t, b.Update(l, 28))
+	assert.True(t, b.Update(l, 29))
+	assert.Equal(t, int64(19), b.lostCounter.Count()) // packet 0 wasn't lost
 	assert.Equal(t, int64(0), b.dupeCounter.Count())
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 
-	b = NewBits(16)
+	b = NewBits(10)
 	b.lostCounter.Clear()
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
 
-	// Update(15) clears the warmup window (no lost), sets slot 15.
-	assert.True(t, b.Update(l, 15))
+	assert.True(t, b.Update(l, 9))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
-
-	// Update(16): slot 0 was already set (NewBits seeded it), and 16 is not
-	// strictly > length, so nothing is recorded as lost.
-	assert.True(t, b.Update(l, 16))
+	// 10 will set 0 index, 0 was already set, no lost packets
+	assert.True(t, b.Update(l, 10))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
-
-	// Update(17): we jumped straight from 0 to 15, so slot 1 was cleared
-	// (and never re-set). 17 > 16 is past warmup, so packet 1 is recorded lost.
-	assert.True(t, b.Update(l, 17))
+	// 11 will set 1 index, 1 was missed, we should see 1 packet lost
+	assert.True(t, b.Update(l, 11))
 	assert.Equal(t, int64(1), b.lostCounter.Count())
+	// Now let's fill in the window, should end up with 8 lost packets
+	assert.True(t, b.Update(l, 12))
+	assert.True(t, b.Update(l, 13))
+	assert.True(t, b.Update(l, 14))
+	assert.True(t, b.Update(l, 15))
+	assert.True(t, b.Update(l, 16))
+	assert.True(t, b.Update(l, 17))
+	assert.True(t, b.Update(l, 18))
+	assert.True(t, b.Update(l, 19))
+	assert.Equal(t, int64(8), b.lostCounter.Count())
 
-	// Fill in 18..30 in single steps. Each i evicts slot i%16. Slots 2..14
-	// were all cleared during Update(15), and we never re-set any of them,
-	// so each i in 18..30 is a fresh lost packet — 13 more.
-	for n := uint64(18); n <= 30; n++ {
-		assert.True(t, b.Update(l, n))
-	}
-	assert.Equal(t, int64(14), b.lostCounter.Count())
+	// Jump ahead by a window size
+	assert.True(t, b.Update(l, 29))
+	assert.Equal(t, int64(8), b.lostCounter.Count())
+	// Now lets walk ahead normally through the window, the missed packets should fill in
+	assert.True(t, b.Update(l, 30))
+	assert.True(t, b.Update(l, 31))
+	assert.True(t, b.Update(l, 32))
+	assert.True(t, b.Update(l, 33))
+	assert.True(t, b.Update(l, 34))
+	assert.True(t, b.Update(l, 35))
+	assert.True(t, b.Update(l, 36))
+	assert.True(t, b.Update(l, 37))
+	assert.True(t, b.Update(l, 38))
+	// 39 packets tracked, 22 seen, 17 lost
+	assert.Equal(t, int64(17), b.lostCounter.Count())
 
-	// Jump ahead by exactly one window size.
-	assert.True(t, b.Update(l, 46))
-	// end = min(46, 30+16) = 46, count = 16, all slots cleared. Before the
-	// jump every slot 0..15 had been set (Update(15), (16), (17), 18..30),
-	// so wasSet=16 and 46 == current+length means no past-window slack:
-	// lost contribution = 0.
-	assert.Equal(t, int64(14), b.lostCounter.Count())
-
-	// Walk 47..55. The Update(46) jump cleared every slot, so only slot 14
-	// (for packet 46) is set when we start. Each subsequent unit step lands
-	// on a slot that was cleared and is past warmup, so it counts as lost.
-	// 9 more = 23.
-	for n := uint64(47); n <= 55; n++ {
-		assert.True(t, b.Update(l, n))
-	}
-	assert.Equal(t, int64(23), b.lostCounter.Count())
-
-	// Jump ahead by two windows: clears the window plus past-window loss.
-	assert.True(t, b.Update(l, 87))
-	// current=55, length=16. end = min(87, 71) = 71. count=16, all slots
-	// cleared. Slots set before the clear are slots 14,15,0..7 (10 total).
-	// Lost from clear = 16 - 10 = 6. Past window: 87 - 55 - 16 = 16. +22.
-	assert.Equal(t, int64(45), b.lostCounter.Count())
+	// Jump ahead by 2 windows, should have recording 1 full window missing
+	assert.True(t, b.Update(l, 58))
+	assert.Equal(t, int64(27), b.lostCounter.Count())
+	// Now lets walk ahead normally through the window, the missed packets should fill in from this window
+	assert.True(t, b.Update(l, 59))
+	assert.True(t, b.Update(l, 60))
+	assert.True(t, b.Update(l, 61))
+	assert.True(t, b.Update(l, 62))
+	assert.True(t, b.Update(l, 63))
+	assert.True(t, b.Update(l, 64))
+	assert.True(t, b.Update(l, 65))
+	assert.True(t, b.Update(l, 66))
+	assert.True(t, b.Update(l, 67))
+	// 68 packets tracked, 32 seen, 36 missed
+	assert.Equal(t, int64(36), b.lostCounter.Count())
 	assert.Equal(t, int64(0), b.dupeCounter.Count())
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 }
 
 func TestBitsLostCounterIssue1(t *testing.T) {
 	l := test.NewLogger()
-	b := NewBits(16)
+	b := NewBits(10)
 	b.lostCounter.Clear()
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
 
-	// Receive 4, backfill 1, then 9, 2, 3, 5, 6, 7 (skip 8), 10, 11, 14.
-	// Then jump to 25 — slot 25%16=9 is being evicted, but it had been set
-	// (we received packet 9), so no spurious lost increment. The original
-	// regression was about double-counting a missing packet when its slot
-	// got cleared on a jump. With the jump path now using clearRange's
-	// word-level wasSet count, the same semantics hold.
 	assert.True(t, b.Update(l, 4))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
 	assert.True(t, b.Update(l, 1))
@@ -266,7 +244,7 @@ func TestBitsLostCounterIssue1(t *testing.T) {
 	assert.Equal(t, int64(0), b.lostCounter.Count())
 	assert.True(t, b.Update(l, 7))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
-	// Skip packet 8.
+	// assert.True(t, b.Update(l, 8))
 	assert.True(t, b.Update(l, 10))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
 	assert.True(t, b.Update(l, 11))
@@ -274,23 +252,9 @@ func TestBitsLostCounterIssue1(t *testing.T) {
 
 	assert.True(t, b.Update(l, 14))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
-
-	// Jump to 25. With length=16, slot 25%16=9 corresponds to packet 9
-	// (which we DID receive), so its bit is set and no lost++ from that
-	// eviction. The trace below shows the only loss is packet 8.
-	assert.True(t, b.Update(l, 25))
-	// current was 14, i=25. end=min(25,30)=25. count=11. startPos=15.
-	// steady? current=14<16, so warmup branch: per-bit n=15..25, count those
-	// with !get(n) AND n>16. n=17..25 are >16. Among slots 17%16=1..25%16=9
-	// did we set slots 1..9 (packets 1..9)? Yes for all but slot 8 (packet 8
-	// was skipped). n=24 maps to slot 8 which is FALSE → lost++. All other
-	// n in 17..25 map to slots that are set. n=16 is not strictly > 16. So
-	// lost = 1.
+	// Issue seems to be here, we reset missing packet 8 to false here and don't increment the lost counter
+	assert.True(t, b.Update(l, 19))
 	assert.Equal(t, int64(1), b.lostCounter.Count())
-
-	// Fill in 12, 13, 15, 16. Each is below current=25 (in-window). 16 must
-	// recheck slot 0 — it was set by NewBits and then cleared by the
-	// Update(25) jump, so 16 backfills cleanly.
 	assert.True(t, b.Update(l, 12))
 	assert.Equal(t, int64(1), b.lostCounter.Count())
 	assert.True(t, b.Update(l, 13))
@@ -299,140 +263,29 @@ func TestBitsLostCounterIssue1(t *testing.T) {
 	assert.Equal(t, int64(1), b.lostCounter.Count())
 	assert.True(t, b.Update(l, 16))
 	assert.Equal(t, int64(1), b.lostCounter.Count())
+	assert.True(t, b.Update(l, 17))
+	assert.Equal(t, int64(1), b.lostCounter.Count())
+	assert.True(t, b.Update(l, 18))
+	assert.Equal(t, int64(1), b.lostCounter.Count())
+	assert.True(t, b.Update(l, 20))
+	assert.Equal(t, int64(1), b.lostCounter.Count())
+	assert.True(t, b.Update(l, 21))
 
-	// We missed packet 8 above and that loss is still recorded once, never
-	// double-counted, never zeroed.
+	// We missed packet 8 above
 	assert.Equal(t, int64(1), b.lostCounter.Count())
 	assert.Equal(t, int64(0), b.dupeCounter.Count())
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 }
 
-// TestBitsWarmupOvershoot exercises the jump path's warmup arm with an
-// overshoot past one full window. NewBits leaves current=0 with only slot 0
-// "set" by the marker. Jumping straight to length+k must (a) clear every
-// slot the jump straddles, (b) count only past-window slack (not the
-// in-window slots, which never had a "lost" tenant during warmup), and
-// (c) leave the cursor at the new counter so subsequent unit advances
-// count from steady state. The marker bit at slot 0 is irrelevant once
-// current >= length.
-func TestBitsWarmupOvershoot(t *testing.T) {
-	l := test.NewLogger()
-	b := NewBits(16)
-	b.lostCounter.Clear()
-
-	// Jump from current=0 to i=20 (length=16, overshoot=4).
-	// Warmup arm: counts slots in [1..16] where bit unset and n>length.
-	// Only n=16 was unset and >length: but slot 16%16=0 is the marker,
-	// so b.get(16) reads bits[0]=1 and skips. Result: 0 lost from the loop.
-	// Past-window: i - current - length = 20 - 0 - 16 = 4 lost.
-	assert.True(t, b.Update(l, 20))
-	assert.Equal(t, int64(4), b.lostCounter.Count())
-	assert.Equal(t, uint64(20), b.current)
-
-	// Steady state now (current=20 >= length=16). Unit advance to 21
-	// stomps slot 21%16=5, which was cleared by the jump and not reset,
-	// so this is +1 lost.
-	assert.True(t, b.Update(l, 21))
-	assert.Equal(t, int64(5), b.lostCounter.Count())
-}
-
-// TestBitsCheckAcrossWarmupBoundary pins the underflow trick in Check's
-// in-window clause. While in warmup, b.current-b.length underflows uint64
-// to a huge value so the first OR-clause is always false; the second
-// clause (i < length && current < length) carries the in-window check.
-// Once current >= length the regimes flip cleanly.
-func TestBitsCheckAcrossWarmupBoundary(t *testing.T) {
-	l := test.NewLogger()
-	b := NewBits(16)
-
-	// Warmup: current=0. Check(0) must read the marker (set) and return false.
-	assert.False(t, b.Check(l, 0), "marker slot should look already-received")
-	// Warmup: any 0 < i < length is in-window and unset → accepted.
-	for i := uint64(1); i < 16; i++ {
-		assert.True(t, b.Check(l, i), "warmup in-window i=%d should be accepted", i)
-	}
-	// Warmup: i >= length but > current is "next number" so accepted.
-	assert.True(t, b.Check(l, 16))
-	assert.True(t, b.Check(l, 1_000_000))
-
-	// Cross into steady state.
-	assert.True(t, b.Update(l, 100))
-	// Now current=100, length=16. In-window range is [85..100].
-	// 84 is just outside: the underflow clause activates; 84 > 100-16=84 is false.
-	// And the warmup clause is false (current >= length). So out of window.
-	assert.False(t, b.Check(l, 84))
-	// 85 sits at the boundary. 85 > 84 is true → in window, unset → accept.
-	assert.True(t, b.Check(l, 85))
-	// 100 is current itself; not strictly greater, in-window, but already set.
-	assert.False(t, b.Check(l, 100))
-	// Way out: clearly out of window.
-	assert.False(t, b.Check(l, 50))
-}
-
-// TestBitsMarkerInvariant verifies the seeded bits[0]=1 marker behaves
-// correctly across warmup and beyond. Update should never clear the marker
-// during warmup (clearRange skips position 0 when startPos=1), and once
-// current >= length the marker is no longer consulted by Check/Update on
-// the live path — but it must still report counter 0 as a duplicate while
-// we are in warmup.
-func TestBitsMarkerInvariant(t *testing.T) {
-	l := test.NewLogger()
-	b := NewBits(8)
-
-	// Counter 0 is the seeded marker; Check sees it as already received.
-	assert.False(t, b.Check(l, 0))
-	// Update(0) at current=0 hits the duplicate branch.
-	b.dupeCounter.Clear()
-	assert.False(t, b.Update(l, 0))
-	assert.Equal(t, int64(1), b.dupeCounter.Count())
-
-	// Walk forward through warmup; the marker must remain set.
-	for n := uint64(1); n <= 7; n++ {
-		assert.True(t, b.Update(l, n))
-	}
-	// Position 0 (the marker) should still read as set because we never
-	// cleared it; Update(0) still looks like a duplicate.
-	assert.False(t, b.Check(l, 0))
-
-	// Cross into steady state with a unit advance to 8: pos=0, evicts the
-	// marker bit. The lost-counter guard (i > b.length) is false (8 == 8),
-	// so this advance does NOT charge a lost packet — exactly what the
-	// marker is there to prevent.
-	b.lostCounter.Clear()
-	assert.True(t, b.Update(l, 8))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	// The slot at pos 0 is now occupied by counter 8.
-	assert.False(t, b.Check(l, 8))
-}
-
-// BenchmarkBitsUpdateInOrder is the steady-state hot path: each call is
-// i == current+1.
-func BenchmarkBitsUpdateInOrder(b *testing.B) {
-	l := test.NewLogger()
-	z := NewBits(16384)
+func BenchmarkBits(b *testing.B) {
+	z := NewBits(10)
 	for n := 0; n < b.N; n++ {
-		z.Update(l, uint64(n)+1)
-	}
-}
+		for i := range z.bits {
+			z.bits[i] = true
+		}
+		for i := range z.bits {
+			z.bits[i] = false
+		}
 
-// BenchmarkBitsUpdateReorder simulates light reorder within the window:
-// every other packet arrives one slot behind its predecessor (forces the
-// in-window backfill branch).
-func BenchmarkBitsUpdateReorder(b *testing.B) {
-	l := test.NewLogger()
-	z := NewBits(16384)
-	for n := 0; n < b.N; n++ {
-		base := uint64(n) * 2
-		z.Update(l, base+2)
-		z.Update(l, base+1)
-	}
-}
-
-// BenchmarkBitsUpdateLargeJumps stresses the clearRange word-level path.
-func BenchmarkBitsUpdateLargeJumps(b *testing.B) {
-	l := test.NewLogger()
-	z := NewBits(16384)
-	for n := 0; n < b.N; n++ {
-		z.Update(l, uint64(n+1)*1000)
 	}
 }

cert/ca_pool.go

@@ -217,10 +217,6 @@ func (ncp *CAPool) verify(c Certificate, now time.Time, certFp string, signerFp
 		return nil, err
 	}
 
-	if signer.Certificate.Curve() != c.Curve() {
-		return nil, ErrCurveMismatch
-	}
-
 	if signer.Certificate.Expired(now) {
 		return nil, ErrRootExpired
 	}

cert/ca_pool_test.go

@@ -654,31 +654,3 @@ func TestCertificateV2_Verify_Subnets(t *testing.T) {
 	_, err = caPool.VerifyCertificate(time.Now(), c)
 	require.NoError(t, err)
 }
-
-func TestCertificateV2_CurveMismatch(t *testing.T) {
-	caIp1 := mustParsePrefixUnmapped("10.0.0.0/16")
-	caIp2 := mustParsePrefixUnmapped("192.168.0.0/24")
-	ca, _, caKey, _ := NewTestCaCert(Version2, Curve_P256, time.Now(), time.Now().Add(10*time.Minute), []netip.Prefix{caIp1, caIp2}, nil, []string{"test"})
-
-	caPem, err := ca.MarshalPEM()
-	require.NoError(t, err)
-
-	caPool := NewCAPool()
-	b, err := caPool.AddCAFromPEM(caPem)
-	require.NoError(t, err)
-	assert.Empty(t, b)
-
-	// ip is outside the network
-	cIp1 := mustParsePrefixUnmapped("10.0.0.1/24")
-	c, _, _, _ := NewTestCert(Version2, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{cIp1}, nil, []string{"test"})
-
-	fp, _ := c.Fingerprint()
-	_, err = caPool.verify(c, time.Now(), fp, c.Issuer())
-	require.NoError(t, err)
-	//
-	c2 := c.(*certificateV2)
-	c2.curve = Curve_CURVE25519
-	fp, _ = c.Fingerprint()
-	_, err = caPool.verify(c, time.Now(), fp, c.Issuer())
-	require.Error(t, err)
-}

cert/cert_v1.go

@@ -112,9 +112,6 @@ func (c *certificateV1) CheckSignature(key []byte) bool {
 	}
 	switch c.details.curve {
 	case Curve_CURVE25519:
-		if len(key) != ed25519.PublicKeySize {
-			return false //avoids a panic internal to ed25519
-		}
 		return ed25519.Verify(key, b, c.signature)
 	case Curve_P256:
 		pubKey, err := ecdsa.ParseUncompressedPublicKey(elliptic.P256(), key)

cert/cert_v2.go

@@ -151,9 +151,6 @@ func (c *certificateV2) CheckSignature(key []byte) bool {
 
 	switch c.curve {
 	case Curve_CURVE25519:
-		if len(key) != ed25519.PublicKeySize {
-			return false //avoids a panic internal to ed25519
-		}
 		return ed25519.Verify(key, b, c.signature)
 	case Curve_P256:
 		pubKey, err := ecdsa.ParseUncompressedPublicKey(elliptic.P256(), key)

cert/errors.go

@@ -22,7 +22,6 @@ var (
 	ErrCaNotFound                 = errors.New("could not find ca for the certificate")
 	ErrUnknownVersion             = errors.New("certificate version unrecognized")
 	ErrCertPubkeyPresent          = errors.New("certificate has unexpected pubkey present")
-	ErrCurveMismatch              = errors.New("certificate curve does not match CA")
 
 	ErrInvalidPEMBlock                   = errors.New("input did not contain a valid PEM encoded block")
 	ErrInvalidPEMCertificateBanner       = errors.New("bytes did not contain a proper certificate banner")

cert_test/cert.go

@@ -163,55 +163,3 @@ func P256Keypair() ([]byte, []byte) {
 	pubkey := privkey.PublicKey()
 	return pubkey.Bytes(), privkey.Bytes()
 }
-
-// DummyCert is a minimal cert.Certificate implementation for testing error paths.
-type DummyCert struct {
-	Version_        cert.Version
-	Curve_          cert.Curve
-	Groups_         []string
-	IsCA_           bool
-	Issuer_         string
-	Name_           string
-	Networks_       []netip.Prefix
-	NotAfter_       time.Time
-	NotBefore_      time.Time
-	PublicKey_      []byte
-	Signature_      []byte
-	UnsafeNetworks_ []netip.Prefix
-}
-
-func (d *DummyCert) Version() cert.Version                         { return d.Version_ }
-func (d *DummyCert) Curve() cert.Curve                             { return d.Curve_ }
-func (d *DummyCert) Groups() []string                              { return d.Groups_ }
-func (d *DummyCert) IsCA() bool                                    { return d.IsCA_ }
-func (d *DummyCert) Issuer() string                                { return d.Issuer_ }
-func (d *DummyCert) Name() string                                  { return d.Name_ }
-func (d *DummyCert) Networks() []netip.Prefix                      { return d.Networks_ }
-func (d *DummyCert) NotAfter() time.Time                           { return d.NotAfter_ }
-func (d *DummyCert) NotBefore() time.Time                          { return d.NotBefore_ }
-func (d *DummyCert) PublicKey() []byte                             { return d.PublicKey_ }
-func (d *DummyCert) Signature() []byte                             { return d.Signature_ }
-func (d *DummyCert) UnsafeNetworks() []netip.Prefix                { return d.UnsafeNetworks_ }
-func (d *DummyCert) Fingerprint() (string, error)                  { return "", nil }
-func (d *DummyCert) CheckSignature(key []byte) bool                { return false }
-func (d *DummyCert) MarshalForHandshakes() ([]byte, error)         { return nil, nil }
-func (d *DummyCert) MarshalPEM() ([]byte, error)                   { return nil, nil }
-func (d *DummyCert) MarshalJSON() ([]byte, error)                  { return nil, nil }
-func (d *DummyCert) Marshal() ([]byte, error)                      { return nil, nil }
-func (d *DummyCert) String() string                                { return "dummy" }
-func (d *DummyCert) Copy() cert.Certificate                        { return d }
-func (d *DummyCert) VerifyPrivateKey(c cert.Curve, k []byte) error { return nil }
-func (d *DummyCert) Expired(time.Time) bool                        { return false }
-func (d *DummyCert) MarshalPublicKeyPEM() []byte                   { return nil }
-func (d *DummyCert) PublicKeyPEM() []byte                          { return nil }
-
-// NewTestCAPool creates a CAPool from the given CA certificates, panicking on error.
-func NewTestCAPool(cas ...cert.Certificate) *cert.CAPool {
-	pool := cert.NewCAPool()
-	for _, ca := range cas {
-		if err := pool.AddCA(ca); err != nil {
-			panic(err)
-		}
-	}
-	return pool
-}

cmd/nebula-service/logs_generic.go

@@ -3,15 +3,8 @@
 
 package main
 
-import (
-	"log/slog"
-	"os"
+import "github.com/sirupsen/logrus"
 
-	"github.com/slackhq/nebula/logging"
-)
-
-// newPlatformLogger returns a *slog.Logger that writes to stdout. Non-Windows
-// platforms have no special sink to integrate with.
-func newPlatformLogger() *slog.Logger {
-	return logging.NewLogger(os.Stdout)
+func HookLogger(l *logrus.Logger) {
+	// Do nothing, let the logs flow to stdout/stderr
 }

cmd/nebula-service/logs_windows.go

@@ -1,86 +1,54 @@
 package main
 
 import (
-	"context"
-	"log/slog"
-	"strings"
-	"sync"
+	"fmt"
+	"io/ioutil"
+	"os"
 
-	"github.com/slackhq/nebula/logging"
+	"github.com/kardianos/service"
+	"github.com/sirupsen/logrus"
 )
 
-// newPlatformLogger returns a *slog.Logger that routes every log record
-// through the Windows service logger so records end up in the Windows
-// Event Log. All the heavy lifting (level management, format swap,
-// timestamp toggle, WithAttrs/WithGroup) comes from logging.NewHandler;
-// this file only contributes:
-//
-//   - an io.Writer that forwards each formatted line to the service
-//     logger at the current record's Event Log severity, and
-//   - a thin severityTag that embeds *logging.Handler and overrides
-//     only Handle / WithAttrs / WithGroup, so Event Viewer's severity
-//     column and severity-based filters keep working the way they did
-//     before the slog migration.
-//
-// Format (text vs json) is carried by the embedded *logging.Handler, so
-// logging.format: json in config still produces JSON lines in Event
-// Viewer, same as the pre-slog logrus setup.
-func newPlatformLogger() *slog.Logger {
-	w := &eventLogWriter{}
-	return slog.New(&severityTag{Handler: logging.NewHandler(w), w: w})
+// HookLogger routes the logrus logs through the service logger so that they end up in the Windows Event Viewer
+// logrus output will be discarded
+func HookLogger(l *logrus.Logger) {
+	l.AddHook(newLogHook(logger))
+	l.SetOutput(ioutil.Discard)
 }
 
-// eventLogWriter forwards slog-formatted lines to the Windows service
-// logger at the severity most recently stashed by severityTag.Handle.
-// The mutex serializes the stash + inner.Handle + Write cycle per record
-// across all concurrent goroutines; slog's builtin text/json handlers
-// each hold their own mutex around Write, but that only protects the
-// Write call itself, not our stash-then-handle sequence.
-type eventLogWriter struct {
-	mu    sync.Mutex
-	level slog.Level
+type logHook struct {
+	sl service.Logger
 }
 
-func (w *eventLogWriter) Write(p []byte) (int, error) {
-	line := strings.TrimRight(string(p), "\n")
-	switch {
-	case w.level >= slog.LevelError:
-		return len(p), logger.Error(line)
-	case w.level >= slog.LevelWarn:
-		return len(p), logger.Warning(line)
+func newLogHook(sl service.Logger) *logHook {
+	return &logHook{sl: sl}
+}
+
+func (h *logHook) Fire(entry *logrus.Entry) error {
+	line, err := entry.String()
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "Unable to read entry, %v", err)
+		return err
+	}
+
+	switch entry.Level {
+	case logrus.PanicLevel:
+		return h.sl.Error(line)
+	case logrus.FatalLevel:
+		return h.sl.Error(line)
+	case logrus.ErrorLevel:
+		return h.sl.Error(line)
+	case logrus.WarnLevel:
+		return h.sl.Warning(line)
+	case logrus.InfoLevel:
+		return h.sl.Info(line)
+	case logrus.DebugLevel:
+		return h.sl.Info(line)
 	default:
-		return len(p), logger.Info(line)
+		return nil
 	}
 }
 
-// severityTag embeds *logging.Handler to pick up everything it does for
-// free (Enabled, SetLevel, GetLevel, SetFormat, GetFormat,
-// SetDisableTimestamp) and overrides only Handle / WithAttrs / WithGroup
-// so each record's slog.Level is stashed on the writer before formatting
-// and so derived handlers stay wrapped as severityTag rather than
-// downgrading to bare *logging.Handler.
-type severityTag struct {
-	*logging.Handler
-	w *eventLogWriter
-}
-
-func (s *severityTag) Handle(ctx context.Context, r slog.Record) error {
-	s.w.mu.Lock()
-	defer s.w.mu.Unlock()
-	s.w.level = r.Level
-	return s.Handler.Handle(ctx, r)
-}
-
-func (s *severityTag) WithAttrs(attrs []slog.Attr) slog.Handler {
-	if len(attrs) == 0 {
-		return s
-	}
-	return &severityTag{Handler: s.Handler.WithAttrs(attrs).(*logging.Handler), w: s.w}
-}
-
-func (s *severityTag) WithGroup(name string) slog.Handler {
-	if name == "" {
-		return s
-	}
-	return &severityTag{Handler: s.Handler.WithGroup(name).(*logging.Handler), w: s.w}
+func (h *logHook) Levels() []logrus.Level {
+	return logrus.AllLevels
 }

cmd/nebula-service/main.go

@@ -7,9 +7,9 @@ import (
 	"runtime/debug"
 	"strings"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/util"
 )
 
@@ -50,11 +50,12 @@ func main() {
 		os.Exit(0)
 	}
 
-	l := logging.NewLogger(os.Stdout)
+	l := logrus.New()
+	l.Out = os.Stdout
 
 	if *serviceFlag != "" {
 		if err := doService(configPath, configTest, Build, serviceFlag); err != nil {
-			l.Error("Service command failed", "error", err)
+			l.WithError(err).Error("Service command failed")
 			os.Exit(1)
 		}
 		return
@@ -73,16 +74,6 @@ func main() {
 		os.Exit(1)
 	}
 
-	if err := logging.ApplyConfig(l, c); err != nil {
-		fmt.Printf("failed to apply logging config: %s", err)
-		os.Exit(1)
-	}
-	c.RegisterReloadCallback(func(c *config.C) {
-		if err := logging.ApplyConfig(l, c); err != nil {
-			l.Error("Failed to reconfigure logger on reload", "error", err)
-		}
-	})
-
 	ctrl, err := nebula.Main(c, *configTest, Build, l, nil)
 	if err != nil {
 		util.LogWithContextIfNeeded("Failed to start", err, l)
@@ -99,7 +90,7 @@ func main() {
 		go ctrl.ShutdownBlock()
 
 		if err := wait(); err != nil {
-			l.Error("Nebula stopped due to fatal error", "error", err)
+			l.WithError(err).Error("Nebula stopped due to fatal error")
 			os.Exit(2)
 		}
 

cmd/nebula-service/service.go

@@ -7,9 +7,9 @@ import (
 	"path/filepath"
 
 	"github.com/kardianos/service"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/logging"
 )
 
 var logger service.Logger
@@ -25,7 +25,8 @@ func (p *program) Start(s service.Service) error {
 	// Start should not block.
 	logger.Info("Nebula service starting.")
 
-	l := newPlatformLogger()
+	l := logrus.New()
+	HookLogger(l)
 
 	c := config.NewC(l)
 	err := c.Load(*p.configPath)
@@ -33,15 +34,6 @@ func (p *program) Start(s service.Service) error {
 		return fmt.Errorf("failed to load config: %s", err)
 	}
 
-	if err := logging.ApplyConfig(l, c); err != nil {
-		return fmt.Errorf("failed to apply logging config: %s", err)
-	}
-	c.RegisterReloadCallback(func(c *config.C) {
-		if err := logging.ApplyConfig(l, c); err != nil {
-			l.Error("Failed to reconfigure logger on reload", "error", err)
-		}
-	})
-
 	p.control, err = nebula.Main(c, *p.configTest, Build, l, nil)
 	if err != nil {
 		return err
@@ -93,7 +85,7 @@ func doService(configPath *string, configTest *bool, build string, serviceFlag *
 	// Here are what the different loggers are doing:
 	// - `log` is the standard go log utility, meant to be used while the process is still attached to stdout/stderr
 	// - `logger` is the service log utility that may be attached to a special place depending on OS (Windows will have it attached to the event log)
-	// - in program.Start we build a *slog.Logger via newPlatformLogger; on non-Windows that is a stdout-backed slog logger, on Windows it routes records through the service logger
+	// - above, in `Run` we create a `logrus.Logger` which is what nebula expects to use
 	s, err := service.New(prg, svcConfig)
 	if err != nil {
 		return err

cmd/nebula/main.go

@@ -7,9 +7,9 @@ import (
 	"runtime/debug"
 	"strings"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/util"
 )
 
@@ -55,7 +55,8 @@ func main() {
 		os.Exit(1)
 	}
 
-	l := logging.NewLogger(os.Stdout)
+	l := logrus.New()
+	l.Out = os.Stdout
 
 	c := config.NewC(l)
 	err := c.Load(*configPath)
@@ -64,16 +65,6 @@ func main() {
 		os.Exit(1)
 	}
 
-	if err := logging.ApplyConfig(l, c); err != nil {
-		fmt.Printf("failed to apply logging config: %s", err)
-		os.Exit(1)
-	}
-	c.RegisterReloadCallback(func(c *config.C) {
-		if err := logging.ApplyConfig(l, c); err != nil {
-			l.Error("Failed to reconfigure logger on reload", "error", err)
-		}
-	})
-
 	ctrl, err := nebula.Main(c, *configTest, Build, l, nil)
 	if err != nil {
 		util.LogWithContextIfNeeded("Failed to start", err, l)
@@ -91,7 +82,7 @@ func main() {
 		notifyReady(l)
 
 		if err := wait(); err != nil {
-			l.Error("Nebula stopped due to fatal error", "error", err)
+			l.WithError(err).Error("Nebula stopped due to fatal error")
 			os.Exit(2)
 		}
 

cmd/nebula/notify_linux.go

@@ -1,10 +1,11 @@
 package main
 
 import (
-	"log/slog"
 	"net"
 	"os"
 	"time"
+
+	"github.com/sirupsen/logrus"
 )
 
 // SdNotifyReady tells systemd the service is ready and dependent services can now be started
@@ -12,30 +13,30 @@ import (
 // https://www.freedesktop.org/software/systemd/man/systemd.service.html
 const SdNotifyReady = "READY=1"
 
-func notifyReady(l *slog.Logger) {
+func notifyReady(l *logrus.Logger) {
 	sockName := os.Getenv("NOTIFY_SOCKET")
 	if sockName == "" {
-		l.Debug("NOTIFY_SOCKET systemd env var not set, not sending ready signal")
+		l.Debugln("NOTIFY_SOCKET systemd env var not set, not sending ready signal")
 		return
 	}
 
 	conn, err := net.DialTimeout("unixgram", sockName, time.Second)
 	if err != nil {
-		l.Error("failed to connect to systemd notification socket", "error", err)
+		l.WithError(err).Error("failed to connect to systemd notification socket")
 		return
 	}
 	defer conn.Close()
 
 	err = conn.SetWriteDeadline(time.Now().Add(time.Second))
 	if err != nil {
-		l.Error("failed to set the write deadline for the systemd notification socket", "error", err)
+		l.WithError(err).Error("failed to set the write deadline for the systemd notification socket")
 		return
 	}
 
 	if _, err = conn.Write([]byte(SdNotifyReady)); err != nil {
-		l.Error("failed to signal the systemd notification socket", "error", err)
+		l.WithError(err).Error("failed to signal the systemd notification socket")
 		return
 	}
 
-	l.Debug("notified systemd the service is ready")
+	l.Debugln("notified systemd the service is ready")
 }

cmd/nebula/notify_notlinux.go

@@ -3,8 +3,8 @@
 
 package main
 
-import "log/slog"
+import "github.com/sirupsen/logrus"
 
-func notifyReady(_ *slog.Logger) {
+func notifyReady(_ *logrus.Logger) {
 	// No init service to notify
 }

config/config.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"log/slog"
 	"math"
 	"os"
 	"os/signal"
@@ -17,6 +16,7 @@ import (
 	"time"
 
 	"dario.cat/mergo"
+	"github.com/sirupsen/logrus"
 	"go.yaml.in/yaml/v3"
 )
 
@@ -26,11 +26,11 @@ type C struct {
 	Settings    map[string]any
 	oldSettings map[string]any
 	callbacks   []func(*C)
-	l           *slog.Logger
+	l           *logrus.Logger
 	reloadLock  sync.Mutex
 }
 
-func NewC(l *slog.Logger) *C {
+func NewC(l *logrus.Logger) *C {
 	return &C{
 		Settings: make(map[string]any),
 		l:        l,
@@ -107,18 +107,12 @@ func (c *C) HasChanged(k string) bool {
 
 	newVals, err := yaml.Marshal(nv)
 	if err != nil {
-		c.l.Error("Error while marshaling new config",
-			"config_path", k,
-			"error", err,
-		)
+		c.l.WithField("config_path", k).WithError(err).Error("Error while marshaling new config")
 	}
 
 	oldVals, err := yaml.Marshal(ov)
 	if err != nil {
-		c.l.Error("Error while marshaling old config",
-			"config_path", k,
-			"error", err,
-		)
+		c.l.WithField("config_path", k).WithError(err).Error("Error while marshaling old config")
 	}
 
 	return string(newVals) != string(oldVals)
@@ -160,10 +154,7 @@ func (c *C) ReloadConfig() {
 
 	err := c.Load(c.path)
 	if err != nil {
-		c.l.Error("Error occurred while reloading config",
-			"config_path", c.path,
-			"error", err,
-		)
+		c.l.WithField("config_path", c.path).WithError(err).Error("Error occurred while reloading config")
 		return
 	}
 

connection_manager.go

@@ -5,12 +5,13 @@ import (
 	"context"
 	"encoding/binary"
 	"fmt"
-	"log/slog"
 	"net/netip"
 	"sync"
 	"sync/atomic"
 	"time"
 
+	"github.com/rcrowley/go-metrics"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
@@ -44,16 +45,19 @@ type connectionManager struct {
 	inactivityTimeout       atomic.Int64
 	dropInactive            atomic.Bool
 
-	l *slog.Logger
+	metricsTxPunchy metrics.Counter
+
+	l *logrus.Logger
 }
 
-func newConnectionManagerFromConfig(l *slog.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
+func newConnectionManagerFromConfig(l *logrus.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
 	cm := &connectionManager{
-		hostMap:       hm,
-		l:             l,
-		punchy:        p,
-		relayUsed:     make(map[uint32]struct{}),
-		relayUsedLock: &sync.RWMutex{},
+		hostMap:         hm,
+		l:               l,
+		punchy:          p,
+		relayUsed:       make(map[uint32]struct{}),
+		relayUsedLock:   &sync.RWMutex{},
+		metricsTxPunchy: metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
 	}
 
 	cm.reload(c, true)
@@ -81,10 +85,9 @@ func (cm *connectionManager) reload(c *config.C, initial bool) {
 		old := cm.getInactivityTimeout()
 		cm.inactivityTimeout.Store((int64)(c.GetDuration("tunnels.inactivity_timeout", 10*time.Minute)))
 		if !initial {
-			cm.l.Info("Inactivity timeout has changed",
-				"oldDuration", old,
-				"newDuration", cm.getInactivityTimeout(),
-			)
+			cm.l.WithField("oldDuration", old).
+				WithField("newDuration", cm.getInactivityTimeout()).
+				Info("Inactivity timeout has changed")
 		}
 	}
 
@@ -92,10 +95,9 @@ func (cm *connectionManager) reload(c *config.C, initial bool) {
 		old := cm.dropInactive.Load()
 		cm.dropInactive.Store(c.GetBool("tunnels.drop_inactive", false))
 		if !initial {
-			cm.l.Info("Drop inactive setting has changed",
-				"oldBool", old,
-				"newBool", cm.dropInactive.Load(),
-			)
+			cm.l.WithField("oldBool", old).
+				WithField("newBool", cm.dropInactive.Load()).
+				Info("Drop inactive setting has changed")
 		}
 	}
 }
@@ -254,7 +256,7 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 			var err error
 			index, err = AddRelay(cm.l, newhostinfo, cm.hostMap, r.PeerAddr, nil, r.Type, Requested)
 			if err != nil {
-				cm.l.Error("failed to migrate relay to new hostinfo", "error", err)
+				cm.l.WithError(err).Error("failed to migrate relay to new hostinfo")
 				continue
 			}
 			switch r.Type {
@@ -302,16 +304,16 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 
 		msg, err := req.Marshal()
 		if err != nil {
-			cm.l.Error("failed to marshal Control message to migrate relay", "error", err)
+			cm.l.WithError(err).Error("failed to marshal Control message to migrate relay")
 		} else {
 			cm.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
-			cm.l.Info("send CreateRelayRequest",
-				"relayFrom", req.RelayFromAddr,
-				"relayTo", req.RelayToAddr,
-				"initiatorRelayIndex", req.InitiatorRelayIndex,
-				"responderRelayIndex", req.ResponderRelayIndex,
-				"vpnAddrs", newhostinfo.vpnAddrs,
-			)
+			cm.l.WithFields(logrus.Fields{
+				"relayFrom":           req.RelayFromAddr,
+				"relayTo":             req.RelayToAddr,
+				"initiatorRelayIndex": req.InitiatorRelayIndex,
+				"responderRelayIndex": req.ResponderRelayIndex,
+				"vpnAddrs":            newhostinfo.vpnAddrs}).
+				Info("send CreateRelayRequest")
 		}
 	}
 }
@@ -323,7 +325,7 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 
 	hostinfo := cm.hostMap.Indexes[localIndex]
 	if hostinfo == nil {
-		cm.l.Debug("Not found in hostmap", "localIndex", localIndex)
+		cm.l.WithField("localIndex", localIndex).Debugln("Not found in hostmap")
 		return doNothing, nil, nil
 	}
 
@@ -343,10 +345,10 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 	// A hostinfo is determined alive if there is incoming traffic
 	if inTraffic {
 		decision := doNothing
-		if cm.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(cm.l).Debug("Tunnel status",
-				"tunnelCheck", m{"state": "alive", "method": "passive"},
-			)
+		if cm.l.Level >= logrus.DebugLevel {
+			hostinfo.logger(cm.l).
+				WithField("tunnelCheck", m{"state": "alive", "method": "passive"}).
+				Debug("Tunnel status")
 		}
 		hostinfo.pendingDeletion.Store(false)
 
@@ -365,7 +367,7 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 
 		if !outTraffic {
 			// Send a punch packet to keep the NAT state alive
-			cm.punchy.SendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
 		}
 
 		return decision, hostinfo, primary
@@ -373,9 +375,9 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 
 	if hostinfo.pendingDeletion.Load() {
 		// We have already sent a test packet and nothing was returned, this hostinfo is dead
-		hostinfo.logger(cm.l).Info("Tunnel status",
-			"tunnelCheck", m{"state": "dead", "method": "active"},
-		)
+		hostinfo.logger(cm.l).
+			WithField("tunnelCheck", m{"state": "dead", "method": "active"}).
+			Info("Tunnel status")
 
 		return deleteTunnel, hostinfo, nil
 	}
@@ -386,39 +388,40 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 			inactiveFor, isInactive := cm.isInactive(hostinfo, now)
 			if isInactive {
 				// Tunnel is inactive, tear it down
-				hostinfo.logger(cm.l).Info("Dropping tunnel due to inactivity",
-					"inactiveDuration", inactiveFor,
-					"primary", mainHostInfo,
-				)
+				hostinfo.logger(cm.l).
+					WithField("inactiveDuration", inactiveFor).
+					WithField("primary", mainHostInfo).
+					Info("Dropping tunnel due to inactivity")
 
 				return closeTunnel, hostinfo, primary
 			}
 
 			// If we aren't sending or receiving traffic then its an unused tunnel and we don't to test the tunnel.
 			// Just maintain NAT state if configured to do so.
-			cm.punchy.SendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
 			cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
 			return doNothing, nil, nil
 		}
 
-		// We aren't receiving traffic but we are sending it. The outbound
-		// traffic itself refreshes the primary remote's NAT state; this
-		// fans out to non-primary remotes, but only if target_all_remotes
-		// is configured.
-		cm.punchy.SendPunchToAll(hostinfo)
+		if cm.punchy.GetTargetEverything() {
+			// This is similar to the old punchy behavior with a slight optimization.
+			// We aren't receiving traffic but we are sending it, punch on all known
+			// ips in case we need to re-prime NAT state
+			cm.sendPunch(hostinfo)
+		}
 
-		if cm.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(cm.l).Debug("Tunnel status",
-				"tunnelCheck", m{"state": "testing", "method": "active"},
-			)
+		if cm.l.Level >= logrus.DebugLevel {
+			hostinfo.logger(cm.l).
+				WithField("tunnelCheck", m{"state": "testing", "method": "active"}).
+				Debug("Tunnel status")
 		}
 
 		// Send a test packet to trigger an authenticated tunnel test, this should suss out any lingering tunnel issues
 		decision = sendTestPacket
 
 	} else {
-		if cm.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(cm.l).Debug("Hostinfo sadness")
+		if cm.l.Level >= logrus.DebugLevel {
+			hostinfo.logger(cm.l).Debugf("Hostinfo sadness")
 		}
 	}
 
@@ -490,16 +493,14 @@ func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostI
 		return false //cert is still valid! yay!
 	} else if err == cert.ErrBlockListed { //avoiding errors.Is for speed
 		// Block listed certificates should always be disconnected
-		hostinfo.logger(cm.l).Info("Remote certificate is blocked, tearing down the tunnel",
-			"error", err,
-			"fingerprint", remoteCert.Fingerprint,
-		)
+		hostinfo.logger(cm.l).WithError(err).
+			WithField("fingerprint", remoteCert.Fingerprint).
+			Info("Remote certificate is blocked, tearing down the tunnel")
 		return true
 	} else if cm.intf.disconnectInvalid.Load() {
-		hostinfo.logger(cm.l).Info("Remote certificate is no longer valid, tearing down the tunnel",
-			"error", err,
-			"fingerprint", remoteCert.Fingerprint,
-		)
+		hostinfo.logger(cm.l).WithError(err).
+			WithField("fingerprint", remoteCert.Fingerprint).
+			Info("Remote certificate is no longer valid, tearing down the tunnel")
 		return true
 	} else {
 		//if we reach here, the cert is no longer valid, but we're configured to keep tunnels from now-invalid certs open
@@ -507,17 +508,41 @@ func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostI
 	}
 }
 
+func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
+	if !cm.punchy.GetPunch() {
+		// Punching is disabled
+		return
+	}
+
+	if cm.intf.lightHouse.IsAnyLighthouseAddr(hostinfo.vpnAddrs) {
+		// Do not punch to lighthouses, we assume our lighthouse update interval is good enough.
+		// In the event the update interval is not sufficient to maintain NAT state then a publicly available lighthouse
+		// would lose the ability to notify us and punchy.respond would become unreliable.
+		return
+	}
+
+	if cm.punchy.GetTargetEverything() {
+		hostinfo.remotes.ForEach(cm.hostMap.GetPreferredRanges(), func(addr netip.AddrPort, preferred bool) {
+			cm.metricsTxPunchy.Inc(1)
+			cm.intf.outside.WriteTo([]byte{1}, addr)
+		})
+
+	} else if hostinfo.remote.IsValid() {
+		cm.metricsTxPunchy.Inc(1)
+		cm.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
+	}
+}
+
 func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 	cs := cm.intf.pki.getCertState()
 	curCrt := hostinfo.ConnectionState.myCert
 	curCrtVersion := curCrt.Version()
 	myCrt := cs.getCertificate(curCrtVersion)
 	if myCrt == nil {
-		cm.l.Info("Re-handshaking with remote",
-			"vpnAddrs", hostinfo.vpnAddrs,
-			"version", curCrtVersion,
-			"reason", "local certificate removed",
-		)
+		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+			WithField("version", curCrtVersion).
+			WithField("reason", "local certificate removed").
+			Info("Re-handshaking with remote")
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return
 	}
@@ -525,12 +550,11 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 	if peerCrt != nil && curCrtVersion < peerCrt.Certificate.Version() {
 		// if our certificate version is less than theirs, and we have a matching version available, rehandshake?
 		if cs.getCertificate(peerCrt.Certificate.Version()) != nil {
-			cm.l.Info("Re-handshaking with remote",
-				"vpnAddrs", hostinfo.vpnAddrs,
-				"version", curCrtVersion,
-				"peerVersion", peerCrt.Certificate.Version(),
-				"reason", "local certificate version lower than peer, attempting to correct",
-			)
+			cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+				WithField("version", curCrtVersion).
+				WithField("peerVersion", peerCrt.Certificate.Version()).
+				WithField("reason", "local certificate version lower than peer, attempting to correct").
+				Info("Re-handshaking with remote")
 			cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(hh *HandshakeHostInfo) {
 				hh.initiatingVersionOverride = peerCrt.Certificate.Version()
 			})
@@ -538,19 +562,17 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 		}
 	}
 	if !bytes.Equal(curCrt.Signature(), myCrt.Signature()) {
-		cm.l.Info("Re-handshaking with remote",
-			"vpnAddrs", hostinfo.vpnAddrs,
-			"reason", "local certificate is not current",
-		)
+		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+			WithField("reason", "local certificate is not current").
+			Info("Re-handshaking with remote")
 
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return
 	}
 	if curCrtVersion < cs.initiatingVersion {
-		cm.l.Info("Re-handshaking with remote",
-			"vpnAddrs", hostinfo.vpnAddrs,
-			"reason", "current cert version < pki.initiatingVersion",
-		)
+		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+			WithField("reason", "current cert version < pki.initiatingVersion").
+			Info("Re-handshaking with remote")
 
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return

connection_manager_test.go

@@ -7,9 +7,9 @@ import (
 	"testing"
 	"time"
 
+	"github.com/flynn/noise"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/overlay/overlaytest"
 	"github.com/slackhq/nebula/test"
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
@@ -46,13 +46,13 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
-		v1Credential:      nil,
+		v1HandshakeBytes:  []byte{},
 	}
 
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlaytest.NoopTun{},
+		inside:           &test.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -63,9 +63,9 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(test.NewLogger(), conf, nil)
-	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
+	conf := config.NewC(l)
+	punchy := NewPunchyFromConfig(l, conf)
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
 	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
@@ -79,6 +79,7 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	}
 	hostinfo.ConnectionState = &ConnectionState{
 		myCert: &dummyCert{version: cert.Version1},
+		H:      &noise.HandshakeState{},
 	}
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
@@ -128,13 +129,13 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
-		v1Credential:      nil,
+		v1HandshakeBytes:  []byte{},
 	}
 
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlaytest.NoopTun{},
+		inside:           &test.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -145,9 +146,9 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(test.NewLogger(), conf, nil)
-	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
+	conf := config.NewC(l)
+	punchy := NewPunchyFromConfig(l, conf)
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
 	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
@@ -161,6 +162,7 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	}
 	hostinfo.ConnectionState = &ConnectionState{
 		myCert: &dummyCert{version: cert.Version1},
+		H:      &noise.HandshakeState{},
 	}
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
@@ -212,13 +214,13 @@ func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
 		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
-		v1Credential:      nil,
+		v1HandshakeBytes:  []byte{},
 	}
 
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlaytest.NoopTun{},
+		inside:           &test.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -229,12 +231,12 @@ func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(test.NewLogger())
+	conf := config.NewC(l)
 	conf.Settings["tunnels"] = map[string]any{
 		"drop_inactive": true,
 	}
-	punchy := NewPunchyFromConfig(test.NewLogger(), conf, nil)
-	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
+	punchy := NewPunchyFromConfig(l, conf)
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
 	assert.True(t, nc.dropInactive.Load())
 	nc.intf = ifce
 
@@ -246,6 +248,7 @@ func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
 	}
 	hostinfo.ConnectionState = &ConnectionState{
 		myCert: &dummyCert{version: cert.Version1},
+		H:      &noise.HandshakeState{},
 	}
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
@@ -336,15 +339,15 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	cachedPeerCert, err := ncp.VerifyCertificate(now.Add(time.Second), peerCert)
 
 	cs := &CertState{
-		privateKey:   []byte{},
-		v1Cert:       &dummyCert{},
-		v1Credential: nil,
+		privateKey:       []byte{},
+		v1Cert:           &dummyCert{},
+		v1HandshakeBytes: []byte{},
 	}
 
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlaytest.NoopTun{},
+		inside:           &test.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -357,9 +360,9 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	ifce.disconnectInvalid.Store(true)
 
 	// Create manager
-	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(test.NewLogger(), conf, nil)
-	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
+	conf := config.NewC(l)
+	punchy := NewPunchyFromConfig(l, conf)
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
 	nc.intf = ifce
 	ifce.connectionManager = nc
 
@@ -368,6 +371,7 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 		ConnectionState: &ConnectionState{
 			myCert:   &dummyCert{},
 			peerCert: cachedPeerCert,
+			H:        &noise.HandshakeState{},
 		},
 	}
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)

connection_state.go

@@ -1,20 +1,24 @@
 package nebula
 
 import (
+	"crypto/rand"
 	"encoding/json"
+	"fmt"
 	"sync"
 	"sync/atomic"
 
+	"github.com/flynn/noise"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
-	"github.com/slackhq/nebula/handshake"
 	"github.com/slackhq/nebula/noiseutil"
 )
 
 const ReplayWindow = 1024
 
 type ConnectionState struct {
-	eKey           noiseutil.CipherState
-	dKey           noiseutil.CipherState
+	eKey           *NebulaCipherState
+	dKey           *NebulaCipherState
+	H              *noise.HandshakeState
 	myCert         cert.Certificate
 	peerCert       *cert.CachedCertificate
 	initiator      bool
@@ -23,24 +27,55 @@ type ConnectionState struct {
 	writeLock      sync.Mutex
 }
 
-// newConnectionStateFromResult builds a fully-populated ConnectionState from a
-// completed handshake.Result. It seeds messageCounter and the replay window so
-// that the post-handshake message indices already used on the wire don't count
-// as missed traffic in the data plane.
-func newConnectionStateFromResult(r *handshake.Result) *ConnectionState {
+func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, initiator bool, pattern noise.HandshakePattern) (*ConnectionState, error) {
+	var dhFunc noise.DHFunc
+	switch crt.Curve() {
+	case cert.Curve_CURVE25519:
+		dhFunc = noise.DH25519
+	case cert.Curve_P256:
+		if cs.pkcs11Backed {
+			dhFunc = noiseutil.DHP256PKCS11
+		} else {
+			dhFunc = noiseutil.DHP256
+		}
+	default:
+		return nil, fmt.Errorf("invalid curve: %s", crt.Curve())
+	}
+
+	var ncs noise.CipherSuite
+	if cs.cipher == "chachapoly" {
+		ncs = noise.NewCipherSuite(dhFunc, noise.CipherChaChaPoly, noise.HashSHA256)
+	} else {
+		ncs = noise.NewCipherSuite(dhFunc, noiseutil.CipherAESGCM, noise.HashSHA256)
+	}
+
+	static := noise.DHKey{Private: cs.privateKey, Public: crt.PublicKey()}
+	hs, err := noise.NewHandshakeState(noise.Config{
+		CipherSuite:   ncs,
+		Random:        rand.Reader,
+		Pattern:       pattern,
+		Initiator:     initiator,
+		StaticKeypair: static,
+		//NOTE: These should come from CertState (pki.go) when we finally implement it
+		PresharedKey:          []byte{},
+		PresharedKeyPlacement: 0,
+	})
+	if err != nil {
+		return nil, fmt.Errorf("NewConnectionState: %s", err)
+	}
+
+	// The queue and ready params prevent a counter race that would happen when
+	// sending stored packets and simultaneously accepting new traffic.
 	ci := &ConnectionState{
-		myCert:    r.MyCert,
-		initiator: r.Initiator,
-		peerCert:  r.RemoteCert,
-		eKey:      noiseutil.NewCipherState(r.EKey, r.Cipher),
-		dKey:      noiseutil.NewCipherState(r.DKey, r.Cipher),
+		H:         hs,
+		initiator: initiator,
 		window:    NewBits(ReplayWindow),
+		myCert:    crt,
 	}
-	ci.messageCounter.Add(r.MessageIndex)
-	for i := uint64(1); i <= r.MessageIndex; i++ {
-		ci.window.Update(nil, i)
-	}
-	return ci
+	// always start the counter from 2, as packet 1 and packet 2 are handshake packets.
+	ci.messageCounter.Add(2)
+
+	return ci, nil
 }
 
 func (cs *ConnectionState) MarshalJSON() ([]byte, error) {

connection_state_test.go

@@ -1,114 +0,0 @@
-package nebula
-
-import (
-	"net/netip"
-	"testing"
-	"time"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/cert"
-	ct "github.com/slackhq/nebula/cert_test"
-	"github.com/slackhq/nebula/handshake"
-	"github.com/slackhq/nebula/header"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-// runTestHandshake runs a complete IX handshake between two freshly-built
-// peers and returns the initiator and responder Results. Used to produce
-// real cipher states for tests that need to exercise post-handshake glue.
-func runTestHandshake(t *testing.T) (initR, respR *handshake.Result) {
-	t.Helper()
-
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-
-	makeCreds := func(name string, networks []netip.Prefix) handshake.GetCredentialFunc {
-		c, _, rawKey, _ := ct.NewTestCert(
-			cert.Version2, cert.Curve_CURVE25519, ca, caKey,
-			name, ca.NotBefore(), ca.NotAfter(), networks, nil, nil,
-		)
-		priv, _, _, err := cert.UnmarshalPrivateKeyFromPEM(rawKey)
-		require.NoError(t, err)
-		hsBytes, err := c.MarshalForHandshakes()
-		require.NoError(t, err)
-		ncs := noise.NewCipherSuite(noise.DH25519, noise.CipherChaChaPoly, noise.HashSHA256)
-		cred := handshake.NewCredential(c, hsBytes, priv, ncs)
-		return func(v cert.Version) *handshake.Credential {
-			if v == cert.Version2 {
-				return cred
-			}
-			return nil
-		}
-	}
-
-	verifier := func(c cert.Certificate) (*cert.CachedCertificate, error) {
-		return caPool.VerifyCertificate(time.Now(), c)
-	}
-
-	initCreds := makeCreds("initiator", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	respCreds := makeCreds("responder", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-
-	initM, err := handshake.NewMachine(
-		cert.Version2, initCreds, verifier,
-		func() (uint32, error) { return 1000, nil },
-		true, header.HandshakeIXPSK0,
-	)
-	require.NoError(t, err)
-
-	respM, err := handshake.NewMachine(
-		cert.Version2, respCreds, verifier,
-		func() (uint32, error) { return 2000, nil },
-		false, header.HandshakeIXPSK0,
-	)
-	require.NoError(t, err)
-
-	msg1, err := initM.Initiate(nil)
-	require.NoError(t, err)
-
-	resp, respR, err := respM.ProcessPacket(nil, msg1)
-	require.NoError(t, err)
-	require.NotNil(t, respR)
-
-	_, initR, err = initM.ProcessPacket(nil, resp)
-	require.NoError(t, err)
-	require.NotNil(t, initR)
-
-	return initR, respR
-}
-
-func TestNewConnectionStateFromResult(t *testing.T) {
-	initR, respR := runTestHandshake(t)
-
-	t.Run("initiator", func(t *testing.T) {
-		ci := newConnectionStateFromResult(initR)
-		assert.True(t, ci.initiator)
-		assert.Equal(t, initR.MyCert, ci.myCert)
-		assert.Equal(t, initR.RemoteCert, ci.peerCert)
-		assert.NotNil(t, ci.eKey)
-		assert.NotNil(t, ci.dKey)
-
-		// IX has 2 handshake messages; the next data-plane send is counter=3.
-		assert.Equal(t, uint64(2), ci.messageCounter.Load(),
-			"messageCounter must equal Result.MessageIndex so the next send is N+1")
-
-		// Both handshake counters must be marked seen so they don't appear lost.
-		// Check returns false if an index has already been recorded.
-		assert.False(t, ci.window.Check(nil, 1), "counter 1 must already be seen")
-		assert.False(t, ci.window.Check(nil, 2), "counter 2 must already be seen")
-		// Counter 3 is the next data-plane message and must NOT be pre-marked.
-		assert.True(t, ci.window.Check(nil, 3), "counter 3 must not be pre-seeded")
-	})
-
-	t.Run("responder", func(t *testing.T) {
-		ci := newConnectionStateFromResult(respR)
-		assert.False(t, ci.initiator)
-		assert.Equal(t, respR.MyCert, ci.myCert)
-		assert.Equal(t, respR.RemoteCert, ci.peerCert)
-		assert.NotNil(t, ci.eKey)
-		assert.NotNil(t, ci.dKey)
-		assert.Equal(t, uint64(2), ci.messageCounter.Load())
-	})
-}

control.go

@@ -3,14 +3,15 @@ package nebula
 import (
 	"context"
 	"errors"
-	"log/slog"
 	"net/netip"
 	"os"
 	"os/signal"
 	"sync"
 	"syscall"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/firewall/events"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
 )
@@ -46,7 +47,7 @@ type Control struct {
 	state     RunState
 
 	f                      *Interface
-	l                      *slog.Logger
+	l                      *logrus.Logger
 	ctx                    context.Context
 	cancel                 context.CancelFunc
 	sshStart               func()
@@ -151,7 +152,7 @@ func (c *Control) Stop() {
 
 	c.CloseAllTunnels(false)
 	if err := c.f.Close(); err != nil {
-		c.l.Error("Close interface failed", "error", err)
+		c.l.WithError(err).Error("Close interface failed")
 	}
 	c.stateLock.Lock()
 	c.state = StateStopped
@@ -166,7 +167,7 @@ func (c *Control) ShutdownBlock() {
 
 	rawSig := <-sigChan
 	sig := rawSig.String()
-	c.l.Info("Caught signal, shutting down", "signal", sig)
+	c.l.WithField("signal", sig).Info("Caught signal, shutting down")
 	c.Stop()
 }
 
@@ -303,10 +304,8 @@ func (c *Control) CloseAllTunnels(excludeLighthouses bool) (closed int) {
 		c.f.send(header.CloseTunnel, 0, h.ConnectionState, h, []byte{}, make([]byte, 12, 12), make([]byte, mtu))
 		c.f.closeTunnel(h)
 
-		c.l.Debug("Sending close tunnel message",
-			"vpnAddrs", h.vpnAddrs,
-			"udpAddr", h.remote,
-		)
+		c.l.WithField("vpnAddrs", h.vpnAddrs).WithField("udpAddr", h.remote).
+			Debug("Sending close tunnel message")
 		closed++
 	}
 
@@ -342,6 +341,36 @@ func (c *Control) Device() overlay.Device {
 	return c.f.inside
 }
 
+// SetFirewallEventReporter installs an event reporter on the current firewall.
+// Passing nil clears any installed reporter. The reporter is carried across
+// firewall rule reloads. Report* methods are invoked while nebula holds
+// internal locks and must be non-blocking; in particular they must not call
+// back into *Control methods that touch the firewall, or deadlock will
+// result.
+//
+// Installation is performed by shallow-copying the current *Firewall,
+// setting the reporter field on the copy, and swapping the pointer under
+// the conntrack lock. Every Firewall the data path sees therefore has an
+// immutable reporter slot, and emit sites can read it without any
+// synchronization of their own.
+func (c *Control) SetFirewallEventReporter(r events.Reporter) {
+	old := c.f.firewall
+	if old == nil {
+		return
+	}
+	old.Conntrack.Lock()
+	defer old.Conntrack.Unlock()
+
+	// Re-read under the lock in case a concurrent reload swapped in a new
+	// Firewall between the unlocked load above and here. Both Firewalls share
+	// the same Conntrack pointer in the normal (non-overflow) reload path,
+	// so the lock we hold is the right one for whichever we see now.
+	current := c.f.firewall
+	fw := *current
+	fw.reporter = r
+	c.f.firewall = &fw
+}
+
 func copyHostInfo(h *HostInfo, preferredRanges []netip.Prefix) ControlHostInfo {
 	chi := ControlHostInfo{
 		VpnAddrs:               make([]netip.Addr, len(h.vpnAddrs)),

control_test.go

@@ -6,6 +6,7 @@ import (
 	"reflect"
 	"testing"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
@@ -82,7 +83,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		f: &Interface{
 			hostMap: hm,
 		},
-		l: test.NewLogger(),
+		l: logrus.New(),
 	}
 
 	thi := c.GetHostInfoByVpnAddr(vpnIp, false)

control_tester.go

@@ -5,6 +5,8 @@ package nebula
 import (
 	"net/netip"
 
+	"github.com/google/gopacket"
+	"github.com/google/gopacket/layers"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/udp"
@@ -20,9 +22,7 @@ func (c *Control) WaitForType(msgType header.MessageType, subType header.Message
 			panic(err)
 		}
 		pipeTo.InjectUDPPacket(p)
-		match := h.Type == msgType && h.Subtype == subType
-		p.Release()
-		if match {
+		if h.Type == msgType && h.Subtype == subType {
 			return
 		}
 	}
@@ -38,9 +38,7 @@ func (c *Control) WaitForTypeByIndex(toIndex uint32, msgType header.MessageType,
 			panic(err)
 		}
 		pipeTo.InjectUDPPacket(p)
-		match := h.RemoteIndex == toIndex && h.Type == msgType && h.Subtype == subType
-		p.Release()
-		if match {
+		if h.RemoteIndex == toIndex && h.Type == msgType && h.Subtype == subType {
 			return
 		}
 	}
@@ -92,15 +90,65 @@ func (c *Control) GetTunTxChan() <-chan []byte {
 	return c.f.inside.(*overlay.TestTun).TxPackets
 }
 
-// InjectUDPPacket injects a packet into the udp side. We copy internally so the caller keeps ownership of p.
-// The copy comes from the freelist so steady-state alloc is zero.
+// InjectUDPPacket will inject a packet into the udp side of nebula
 func (c *Control) InjectUDPPacket(p *udp.Packet) {
-	c.f.outside.(*udp.TesterConn).Send(p.Copy())
+	c.f.outside.(*udp.TesterConn).Send(p)
 }
 
-// InjectTunPacket pushes an IP packet onto the tun interface.
-func (c *Control) InjectTunPacket(packet []byte) {
-	c.f.inside.(*overlay.TestTun).Send(packet)
+// InjectTunUDPPacket puts a udp packet on the tun interface. Using UDP here because it's a simpler protocol
+func (c *Control) InjectTunUDPPacket(toAddr netip.Addr, toPort uint16, fromAddr netip.Addr, fromPort uint16, data []byte) {
+	serialize := make([]gopacket.SerializableLayer, 0)
+	var netLayer gopacket.NetworkLayer
+	if toAddr.Is6() {
+		if !fromAddr.Is6() {
+			panic("Cant send ipv6 to ipv4")
+		}
+		ip := &layers.IPv6{
+			Version:    6,
+			NextHeader: layers.IPProtocolUDP,
+			SrcIP:      fromAddr.Unmap().AsSlice(),
+			DstIP:      toAddr.Unmap().AsSlice(),
+		}
+		serialize = append(serialize, ip)
+		netLayer = ip
+	} else {
+		if !fromAddr.Is4() {
+			panic("Cant send ipv4 to ipv6")
+		}
+
+		ip := &layers.IPv4{
+			Version:  4,
+			TTL:      64,
+			Protocol: layers.IPProtocolUDP,
+			SrcIP:    fromAddr.Unmap().AsSlice(),
+			DstIP:    toAddr.Unmap().AsSlice(),
+		}
+		serialize = append(serialize, ip)
+		netLayer = ip
+	}
+
+	udp := layers.UDP{
+		SrcPort: layers.UDPPort(fromPort),
+		DstPort: layers.UDPPort(toPort),
+	}
+	err := udp.SetNetworkLayerForChecksum(netLayer)
+	if err != nil {
+		panic(err)
+	}
+
+	buffer := gopacket.NewSerializeBuffer()
+	opt := gopacket.SerializeOptions{
+		ComputeChecksums: true,
+		FixLengths:       true,
+	}
+
+	serialize = append(serialize, &udp, gopacket.Payload(data))
+	err = gopacket.SerializeLayers(buffer, opt, serialize...)
+	if err != nil {
+		panic(err)
+	}
+
+	c.f.inside.(*overlay.TestTun).Send(buffer.Bytes())
 }
 
 func (c *Control) GetVpnAddrs() []netip.Addr {

dns_server.go

@@ -3,7 +3,6 @@ package nebula
 import (
 	"context"
 	"fmt"
-	"log/slog"
 	"net"
 	"net/netip"
 	"strconv"
@@ -13,12 +12,13 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/miekg/dns"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 )
 
 type dnsServer struct {
 	sync.RWMutex
-	l               *slog.Logger
+	l               *logrus.Logger
 	ctx             context.Context
 	dnsMap4         map[string]netip.Addr
 	dnsMap6         map[string]netip.Addr
@@ -55,7 +55,7 @@ type dnsServer struct {
 // they no-op when DNS isn't enabled. Each Start invocation owns a ctx-cancel
 // watcher that tears the listener down on nebula shutdown. The returned
 // pointer is always non-nil, even on error.
-func newDnsServerFromConfig(ctx context.Context, l *slog.Logger, cs *CertState, hostMap *HostMap, c *config.C) (*dnsServer, error) {
+func newDnsServerFromConfig(ctx context.Context, l *logrus.Logger, cs *CertState, hostMap *HostMap, c *config.C) (*dnsServer, error) {
 	ds := &dnsServer{
 		l:               l,
 		ctx:             ctx,
@@ -69,7 +69,7 @@ func newDnsServerFromConfig(ctx context.Context, l *slog.Logger, cs *CertState,
 
 	c.RegisterReloadCallback(func(c *config.C) {
 		if err := ds.reload(c, false); err != nil {
-			ds.l.Error("Failed to reload DNS responder from config", "error", err)
+			l.WithError(err).Error("Failed to reload DNS responder from config")
 		}
 	})
 
@@ -145,7 +145,7 @@ func (d *dnsServer) shutdownServer(srv *dns.Server, started chan struct{}, reaso
 		<-started
 	}
 	if err := srv.Shutdown(); err != nil {
-		d.l.Warn("Failed to shut down the DNS responder", "reason", reason, "error", err)
+		d.l.WithError(err).WithField("reason", reason).Warn("Failed to shut down the DNS responder")
 	}
 }
 
@@ -188,7 +188,7 @@ func (d *dnsServer) Start() {
 		}
 	}()
 
-	d.l.Info("Starting DNS responder", "dnsListener", addr)
+	d.l.WithField("dnsListener", addr).Info("Starting DNS responder")
 	err := server.ListenAndServe()
 	close(done)
 
@@ -201,7 +201,7 @@ func (d *dnsServer) Start() {
 	}
 
 	if err != nil {
-		d.l.Warn("Failed to run the DNS responder", "error", err)
+		d.l.WithError(err).Warn("Failed to run the DNS responder")
 	}
 }
 
@@ -216,34 +216,25 @@ func (d *dnsServer) Stop() {
 	d.shutdownServer(srv, started, "stop")
 }
 
-// Query returns the address for the given name and query type. The second
-// return value reports whether the name is known at all (in either A or AAAA),
-// which lets callers distinguish NODATA from NXDOMAIN.
-func (d *dnsServer) Query(q uint16, data string) (netip.Addr, bool) {
+func (d *dnsServer) Query(q uint16, data string) netip.Addr {
 	data = strings.ToLower(data)
 	d.RLock()
 	defer d.RUnlock()
-	addr4, haveV4 := d.dnsMap4[data]
-	addr6, haveV6 := d.dnsMap6[data]
-	nameExists := haveV4 || haveV6
 	switch q {
 	case dns.TypeA:
-		if haveV4 {
-			return addr4, nameExists
+		if r, ok := d.dnsMap4[data]; ok {
+			return r
 		}
 	case dns.TypeAAAA:
-		if haveV6 {
-			return addr6, nameExists
+		if r, ok := d.dnsMap6[data]; ok {
+			return r
 		}
 	}
 
-	return netip.Addr{}, nameExists
+	return netip.Addr{}
 }
 
 func (d *dnsServer) QueryCert(data string) string {
-	if len(data) < 2 {
-		return ""
-	}
 	ip, err := netip.ParseAddr(data[:len(data)-1])
 	if err != nil {
 		return ""
@@ -314,23 +305,12 @@ func (d *dnsServer) isSelfNebulaOrLocalhost(addr string) bool {
 }
 
 func (d *dnsServer) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
-	debugEnabled := d.l.Enabled(context.Background(), slog.LevelDebug)
-	// Per RFC 2308 §2.2, a name that exists but has no record of the requested
-	// type must be answered with NOERROR and an empty answer section (NODATA),
-	// not NXDOMAIN (RFC 2308 §2.1), which is reserved for names that do not
-	// exist at all.
-	anyNameExists := false
 	for _, q := range m.Question {
 		switch q.Qtype {
 		case dns.TypeA, dns.TypeAAAA:
 			qType := dns.TypeToString[q.Qtype]
-			if debugEnabled {
-				d.l.Debug("DNS query", "type", qType, "name", q.Name)
-			}
-			ip, nameExists := d.Query(q.Qtype, q.Name)
-			if nameExists {
-				anyNameExists = true
-			}
+			d.l.Debugf("Query for %s %s", qType, q.Name)
+			ip := d.Query(q.Qtype, q.Name)
 			if ip.IsValid() {
 				rr, err := dns.NewRR(fmt.Sprintf("%s %s %s", q.Name, qType, ip))
 				if err == nil {
@@ -342,9 +322,7 @@ func (d *dnsServer) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
 			if !d.isSelfNebulaOrLocalhost(w.RemoteAddr().String()) {
 				return
 			}
-			if debugEnabled {
-				d.l.Debug("DNS query", "type", "TXT", "name", q.Name)
-			}
+			d.l.Debugf("Query for TXT %s", q.Name)
 			ip := d.QueryCert(q.Name)
 			if ip != "" {
 				rr, err := dns.NewRR(fmt.Sprintf("%s TXT %s", q.Name, ip))
@@ -355,7 +333,7 @@ func (d *dnsServer) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
 		}
 	}
 
-	if len(m.Answer) == 0 && !anyNameExists {
+	if len(m.Answer) == 0 {
 		m.Rcode = dns.RcodeNameError
 	}
 }

dns_server_test.go

@@ -2,7 +2,7 @@ package nebula
 
 import (
 	"context"
-	"log/slog"
+	"io"
 	"net"
 	"net/netip"
 	"strconv"
@@ -10,26 +10,14 @@ import (
 	"time"
 
 	"github.com/miekg/dns"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
 
-type stubDNSWriter struct{}
-
-func (stubDNSWriter) LocalAddr() net.Addr { return &net.UDPAddr{} }
-func (stubDNSWriter) RemoteAddr() net.Addr {
-	return &net.UDPAddr{IP: net.ParseIP("127.0.0.1"), Port: 5353}
-}
-func (stubDNSWriter) Write([]byte) (int, error) { return 0, nil }
-func (stubDNSWriter) WriteMsg(*dns.Msg) error   { return nil }
-func (stubDNSWriter) Close() error              { return nil }
-func (stubDNSWriter) TsigStatus() error         { return nil }
-func (stubDNSWriter) TsigTimersOnly(bool)       {}
-func (stubDNSWriter) Hijack()                   {}
-
 func TestParsequery(t *testing.T) {
-	l := slog.New(slog.DiscardHandler)
+	l := logrus.New()
 	hostMap := &HostMap{}
 	ds := &dnsServer{
 		l:       l,
@@ -45,56 +33,18 @@ func TestParsequery(t *testing.T) {
 		netip.MustParseAddr("fd01::25"),
 	}
 	ds.Add("test.com.com", addrs)
-	ds.Add("v4only.com.com", []netip.Addr{netip.MustParseAddr("1.2.3.6")})
-	ds.Add("v6only.com.com", []netip.Addr{netip.MustParseAddr("fd01::26")})
 
 	m := &dns.Msg{}
 	m.SetQuestion("test.com.com", dns.TypeA)
 	ds.parseQuery(m, nil)
 	assert.NotNil(t, m.Answer)
 	assert.Equal(t, "1.2.3.4", m.Answer[0].(*dns.A).A.String())
-	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
 
 	m = &dns.Msg{}
 	m.SetQuestion("test.com.com", dns.TypeAAAA)
 	ds.parseQuery(m, nil)
 	assert.NotNil(t, m.Answer)
 	assert.Equal(t, "fd01::24", m.Answer[0].(*dns.AAAA).AAAA.String())
-	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
-
-	// A known name with no record of the requested type should return NODATA
-	// (NOERROR with empty answer), not NXDOMAIN.
-	m = &dns.Msg{}
-	m.SetQuestion("v4only.com.com", dns.TypeAAAA)
-	ds.parseQuery(m, nil)
-	assert.Empty(t, m.Answer)
-	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
-
-	m = &dns.Msg{}
-	m.SetQuestion("v6only.com.com", dns.TypeA)
-	ds.parseQuery(m, nil)
-	assert.Empty(t, m.Answer)
-	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
-
-	// An unknown name should still return NXDOMAIN.
-	m = &dns.Msg{}
-	m.SetQuestion("unknown.com.com", dns.TypeA)
-	ds.parseQuery(m, nil)
-	assert.Empty(t, m.Answer)
-	assert.Equal(t, dns.RcodeNameError, m.Rcode)
-
-	// short lookups should not fail
-	m = &dns.Msg{}
-	m.Question = []dns.Question{{Name: "", Qtype: dns.TypeTXT, Qclass: dns.ClassINET}}
-	ds.parseQuery(m, stubDNSWriter{})
-	assert.Empty(t, m.Answer)
-	assert.Equal(t, dns.RcodeNameError, m.Rcode)
-
-	m = &dns.Msg{}
-	m.Question = []dns.Question{{Name: ".", Qtype: dns.TypeTXT, Qclass: dns.ClassINET}}
-	ds.parseQuery(m, stubDNSWriter{})
-	assert.Empty(t, m.Answer)
-	assert.Equal(t, dns.RcodeNameError, m.Rcode)
 }
 
 func Test_getDnsServerAddr(t *testing.T) {
@@ -136,9 +86,10 @@ func Test_getDnsServerAddr(t *testing.T) {
 
 func newTestDnsServer(t *testing.T) (*dnsServer, *config.C) {
 	t.Helper()
-	sl := slog.New(slog.DiscardHandler)
+	l := logrus.New()
+	l.Out = io.Discard
 	ds := &dnsServer{
-		l:       sl,
+		l:       l,
 		ctx:     context.Background(),
 		dnsMap4: make(map[string]netip.Addr),
 		dnsMap6: make(map[string]netip.Addr),
@@ -146,7 +97,7 @@ func newTestDnsServer(t *testing.T) (*dnsServer, *config.C) {
 	}
 	ds.mux = dns.NewServeMux()
 	ds.mux.HandleFunc(".", ds.handleDnsRequest)
-	return ds, config.NewC(nil)
+	return ds, config.NewC(l)
 }
 
 func setDnsConfig(c *config.C, host string, port string, amLighthouse, serveDns bool) {

e2e/handshake_manager_test.go

@@ -28,7 +28,6 @@ func makeHandshakePacket(from, to netip.AddrPort, subtype header.MessageSubType,
 }
 
 func TestHandshakeRetransmitDuplicate(t *testing.T) {
-	t.Parallel()
 	// Verify the responder correctly handles receiving the same msg1 multiple times
 	// (retransmission). The duplicate goes through CheckAndComplete -> ErrAlreadySeen
 	// and the cached response is resent.
@@ -47,7 +46,7 @@ func TestHandshakeRetransmitDuplicate(t *testing.T) {
 	defer r.RenderFlow()
 
 	t.Log("Trigger handshake from me to them")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
 
 	t.Log("Grab my msg1")
 	msg1 := myControl.GetFromUDP(true)
@@ -79,7 +78,6 @@ func TestHandshakeRetransmitDuplicate(t *testing.T) {
 }
 
 func TestHandshakeTruncatedPacketRecovery(t *testing.T) {
-	t.Parallel()
 	// Verify that a truncated handshake packet is ignored and the real
 	// packet can still complete the handshake.
 
@@ -97,7 +95,7 @@ func TestHandshakeTruncatedPacketRecovery(t *testing.T) {
 	defer r.RenderFlow()
 
 	t.Log("Trigger handshake")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
 
 	t.Log("Get msg1 and deliver to responder")
 	msg1 := myControl.GetFromUDP(true)
@@ -128,7 +126,6 @@ func TestHandshakeTruncatedPacketRecovery(t *testing.T) {
 }
 
 func TestHandshakeOrphanedMsg2Dropped(t *testing.T) {
-	t.Parallel()
 	// A msg2 arriving with no matching pending index should be silently dropped
 	// with no response sent and no state changes.
 
@@ -146,7 +143,7 @@ func TestHandshakeOrphanedMsg2Dropped(t *testing.T) {
 	defer r.RenderFlow()
 
 	t.Log("Complete a normal handshake")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
 	r.RouteForAllUntilTxTun(theirControl)
 	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
 
@@ -171,7 +168,6 @@ func TestHandshakeOrphanedMsg2Dropped(t *testing.T) {
 }
 
 func TestHandshakeUnknownMessageCounter(t *testing.T) {
-	t.Parallel()
 	// A handshake packet with an unexpected message counter should be silently
 	// dropped with no side effects and no UDP response.
 
@@ -203,7 +199,6 @@ func TestHandshakeUnknownMessageCounter(t *testing.T) {
 }
 
 func TestHandshakeUnknownSubtype(t *testing.T) {
-	t.Parallel()
 	// A handshake packet with an unknown subtype should be silently dropped.
 
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -229,7 +224,6 @@ func TestHandshakeUnknownSubtype(t *testing.T) {
 }
 
 func TestHandshakeLateResponse(t *testing.T) {
-	t.Parallel()
 	// After a handshake times out, a late response should be silently ignored
 	// with no new tunnels created.
 
@@ -248,7 +242,7 @@ func TestHandshakeLateResponse(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger handshake from me")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
 
 	t.Log("Grab msg1 but don't deliver")
 	msg1 := myControl.GetFromUDP(true)
@@ -279,7 +273,6 @@ func TestHandshakeLateResponse(t *testing.T) {
 }
 
 func TestHandshakeSelfConnectionRejected(t *testing.T) {
-	t.Parallel()
 	// Verify that a node rejects a handshake containing its own VPN IP in the
 	// peer cert. We do this by sending the initiator's own msg1 back to itself.
 
@@ -292,7 +285,7 @@ func TestHandshakeSelfConnectionRejected(t *testing.T) {
 	myControl.Start()
 
 	t.Log("Trigger handshake from me")
-	myControl.InjectTunPacket(BuildTunUDPPacket(netip.MustParseAddr("10.128.0.2"), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	myControl.InjectTunUDPPacket(netip.MustParseAddr("10.128.0.2"), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
 	msg1 := myControl.GetFromUDP(true)
 
 	t.Log("Drain any handshake retransmits before injecting")
@@ -328,7 +321,6 @@ func TestHandshakeSelfConnectionRejected(t *testing.T) {
 }
 
 func TestHandshakeMessageCounter0Dropped(t *testing.T) {
-	t.Parallel()
 	// MessageCounter=0 is not a valid handshake message and should be dropped.
 
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -349,7 +341,6 @@ func TestHandshakeMessageCounter0Dropped(t *testing.T) {
 }
 
 func TestHandshakeRemoteAllowList(t *testing.T) {
-	t.Parallel()
 	// Verify that a handshake from a blocked underlay IP is dropped with no
 	// response and no state changes. Then verify the same packet from an
 	// allowed IP succeeds.
@@ -375,7 +366,7 @@ func TestHandshakeRemoteAllowList(t *testing.T) {
 	defer r.RenderFlow()
 
 	t.Log("Trigger handshake from them")
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi"))
 	msg1 := theirControl.GetFromUDP(true)
 
 	t.Log("Rewrite the source to a blocked IP and inject")
@@ -408,7 +399,6 @@ func TestHandshakeRemoteAllowList(t *testing.T) {
 }
 
 func TestHandshakeAlreadySeenPreferredRemote(t *testing.T) {
-	t.Parallel()
 	// When a duplicate msg1 arrives via ErrAlreadySeen, verify the tunnel
 	// remains functional and hostmap index count is stable.
 
@@ -426,7 +416,7 @@ func TestHandshakeAlreadySeenPreferredRemote(t *testing.T) {
 	defer r.RenderFlow()
 
 	t.Log("Complete a normal handshake via the router")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi"))
 	r.RouteForAllUntilTxTun(theirControl)
 	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
 
@@ -437,7 +427,7 @@ func TestHandshakeAlreadySeenPreferredRemote(t *testing.T) {
 	originalRemote := hi.CurrentRemote
 
 	t.Log("Re-trigger traffic to cause a new handshake attempt (ErrAlreadySeen)")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("roam")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("roam"))
 	r.RouteForAllUntilTxTun(theirControl)
 
 	t.Log("Verify tunnel still works")
@@ -455,7 +445,6 @@ func TestHandshakeAlreadySeenPreferredRemote(t *testing.T) {
 }
 
 func TestHandshakeWrongResponderPacketStore(t *testing.T) {
-	t.Parallel()
 	// Verify that when the wrong host responds, the cached packets are
 	// transferred to the new handshake, the evil tunnel is closed, evil's
 	// address is blocked, and the correct tunnel is eventually established.
@@ -475,8 +464,8 @@ func TestHandshakeWrongResponderPacketStore(t *testing.T) {
 	evilControl.Start()
 
 	t.Log("Send multiple packets to them (cached during handshake)")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("packet1")))
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("packet2")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("packet1"))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("packet2"))
 
 	t.Log("Route until evil tunnel is closed")
 	h := &header.H{}
@@ -519,7 +508,6 @@ func TestHandshakeWrongResponderPacketStore(t *testing.T) {
 }
 
 func TestHandshakeRelayComplete(t *testing.T) {
-	t.Parallel()
 	// Verify that a relay handshake completes correctly and relay state is
 	// properly maintained on all three nodes.
 
@@ -540,7 +528,7 @@ func TestHandshakeRelayComplete(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger handshake via relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi via relay")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi via relay"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	assertUdpPacket(t, []byte("Hi via relay"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
@@ -568,7 +556,7 @@ func TestHandshakeRelayComplete(t *testing.T) {
 }
 
 // NOTE: Relay V1 cert + IPv6 rejection is not tested here because
-// BuildTunUDPPacket from a V4 node to a V6 address panics in the test
+// InjectTunUDPPacket from a V4 node to a V6 address panics in the test
 // framework. The check is in handshake_manager.go handleOutbound relay
 // logic (lines ~304-313): if the relay host has a V1 cert and either
 // address is IPv6, the relay is skipped.

e2e/handshakes_test.go

@@ -11,12 +11,12 @@ import (
 
 	"github.com/google/gopacket"
 	"github.com/google/gopacket/layers"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/e2e/router"
 	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -40,22 +40,11 @@ func BenchmarkHotPath(b *testing.B) {
 	r.CancelFlowLogs()
 
 	assertTunnel(b, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	// Pre-build the IP packet bytes once so the bench measures the data plane,
-	// not gopacket SerializeLayers overhead.
-	prebuilt := BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
-
-	// EnableFanIn switches the router to a 0-alloc routing path. Required
-	// for hot-path benchmarks; would conflict with GetFromUDP-using tests.
-	r.EnableFanIn()
-
 	b.ResetTimer()
 
 	for n := 0; n < b.N; n++ {
-		myControl.InjectTunPacket(prebuilt)
-		// Release the TUN-side bytes back to the harness freelist; the bench
-		// just confirms a packet arrived, the contents aren't inspected.
-		overlay.ReleaseTunBuf(r.RouteForAllUntilTxTun(theirControl))
+		myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
+		_ = r.RouteForAllUntilTxTun(theirControl)
 	}
 
 	myControl.Stop()
@@ -83,15 +72,11 @@ func BenchmarkHotPathRelay(b *testing.B) {
 	theirControl.Start()
 
 	assertTunnel(b, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
-
-	prebuilt := BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
-	r.EnableFanIn()
-
 	b.ResetTimer()
 
 	for n := 0; n < b.N; n++ {
-		myControl.InjectTunPacket(prebuilt)
-		overlay.ReleaseTunBuf(r.RouteForAllUntilTxTun(theirControl))
+		myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
+		_ = r.RouteForAllUntilTxTun(theirControl)
 	}
 
 	myControl.Stop()
@@ -100,7 +85,6 @@ func BenchmarkHotPathRelay(b *testing.B) {
 }
 
 func TestGoodHandshake(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
@@ -113,7 +97,7 @@ func TestGoodHandshake(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Send a udp packet through to begin standing up the tunnel, this should come out the other side")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	t.Log("Have them consume my stage 0 packet. They have a tunnel now")
 	theirControl.InjectUDPPacket(myControl.GetFromUDP(true))
@@ -151,7 +135,6 @@ func TestGoodHandshake(t *testing.T) {
 }
 
 func TestGoodHandshakeNoOverlap(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "me", "10.128.0.1/24", nil)
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "them", "2001::69/24", nil) //look ma, cross-stack!
@@ -187,7 +170,6 @@ func TestGoodHandshakeNoOverlap(t *testing.T) {
 }
 
 func TestWrongResponderHandshake(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.100/24", nil)
@@ -207,7 +189,7 @@ func TestWrongResponderHandshake(t *testing.T) {
 	evilControl.Start()
 
 	t.Log("Start the handshake process, we will route until we see the evil tunnel closed")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	h := &header.H{}
 	r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
@@ -264,7 +246,6 @@ func TestWrongResponderHandshake(t *testing.T) {
 }
 
 func TestWrongResponderHandshakeStaticHostMap(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.99/24", nil)
@@ -289,7 +270,7 @@ func TestWrongResponderHandshakeStaticHostMap(t *testing.T) {
 	evilControl.Start()
 
 	t.Log("Start the handshake process, we will route until we see the evil tunnel closed")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	h := &header.H{}
 	r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
@@ -347,7 +328,6 @@ func TestWrongResponderHandshakeStaticHostMap(t *testing.T) {
 }
 
 func TestStage1Race(t *testing.T) {
-	t.Parallel()
 	// This tests ensures that two hosts handshaking with each other at the same time will allow traffic to flow
 	// But will eventually collapse down to a single tunnel
 
@@ -368,8 +348,8 @@ func TestStage1Race(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake to start on both me and them")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 	t.Log("Get both stage 1 handshake packets")
 	myHsForThem := myControl.GetFromUDP(true)
@@ -428,7 +408,6 @@ func TestStage1Race(t *testing.T) {
 }
 
 func TestUncleanShutdownRaceLoser(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me  ", "10.128.0.1/24", nil)
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
@@ -446,7 +425,7 @@ func TestUncleanShutdownRaceLoser(t *testing.T) {
 	theirControl.Start()
 
 	r.Log("Trigger a handshake from me to them")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	assertUdpPacket(t, []byte("Hi from me"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
@@ -457,7 +436,7 @@ func TestUncleanShutdownRaceLoser(t *testing.T) {
 	myHostmap.Indexes = map[uint32]*nebula.HostInfo{}
 	myHostmap.RemoteIndexes = map[uint32]*nebula.HostInfo{}
 
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me again")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me again"))
 	p = r.RouteForAllUntilTxTun(theirControl)
 	assertUdpPacket(t, []byte("Hi from me again"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
 
@@ -478,7 +457,6 @@ func TestUncleanShutdownRaceLoser(t *testing.T) {
 }
 
 func TestUncleanShutdownRaceWinner(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me  ", "10.128.0.1/24", nil)
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
@@ -496,7 +474,7 @@ func TestUncleanShutdownRaceWinner(t *testing.T) {
 	theirControl.Start()
 
 	r.Log("Trigger a handshake from me to them")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	assertUdpPacket(t, []byte("Hi from me"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
@@ -508,7 +486,7 @@ func TestUncleanShutdownRaceWinner(t *testing.T) {
 	theirHostmap.Indexes = map[uint32]*nebula.HostInfo{}
 	theirHostmap.RemoteIndexes = map[uint32]*nebula.HostInfo{}
 
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them again")))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them again"))
 	p = r.RouteForAllUntilTxTun(myControl)
 	assertUdpPacket(t, []byte("Hi from them again"), p, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), 80, 80)
 	r.RenderHostmaps("Derp hostmaps", myControl, theirControl)
@@ -530,7 +508,6 @@ func TestUncleanShutdownRaceWinner(t *testing.T) {
 }
 
 func TestRelays(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "relay  ", "10.128.0.128/24", m{"relay": m{"am_relay": true}})
@@ -551,7 +528,7 @@ func TestRelays(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	r.Log("Assert the tunnel works")
@@ -560,7 +537,6 @@ func TestRelays(t *testing.T) {
 }
 
 func TestRelaysDontCareAboutIps(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "relay  ", "2001::9999/24", m{"relay": m{"am_relay": true}})
@@ -581,7 +557,7 @@ func TestRelaysDontCareAboutIps(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	r.Log("Assert the tunnel works")
@@ -590,7 +566,6 @@ func TestRelaysDontCareAboutIps(t *testing.T) {
 }
 
 func TestReestablishRelays(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "relay  ", "10.128.0.128/24", m{"relay": m{"am_relay": true}})
@@ -611,14 +586,14 @@ func TestReestablishRelays(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	r.Log("Assert the tunnel works")
 	assertUdpPacket(t, []byte("Hi from me"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
 
 	t.Log("Ensure packet traversal from them to me via the relay")
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them")))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 	p = r.RouteForAllUntilTxTun(myControl)
 	r.Log("Assert the tunnel works")
@@ -633,7 +608,7 @@ func TestReestablishRelays(t *testing.T) {
 	for curIndexes >= start {
 		curIndexes = len(myControl.GetHostmap().Indexes)
 		r.Logf("Wait for the dead index to go away:start=%v indexes, current=%v indexes", start, curIndexes)
-		myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me should fail")))
+		myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me should fail"))
 
 		r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
 			return router.RouteAndExit
@@ -650,7 +625,7 @@ func TestReestablishRelays(t *testing.T) {
 		myControl.InjectLightHouseAddr(relayVpnIpNet[0].Addr(), relayUdpAddr)
 		myControl.InjectRelays(theirVpnIpNet[0].Addr(), []netip.Addr{relayVpnIpNet[0].Addr()})
 		relayControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-		myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+		myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 		p = r.RouteForAllUntilTxTun(theirControl)
 		r.Log("Assert the tunnel works")
@@ -685,7 +660,7 @@ func TestReestablishRelays(t *testing.T) {
 	t.Log("Assert the tunnel works the other way, too")
 	for {
 		t.Log("RouteForAllUntilTxTun")
-		theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them")))
+		theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 		p = r.RouteForAllUntilTxTun(myControl)
 		r.Log("Assert the tunnel works")
@@ -722,7 +697,6 @@ func TestReestablishRelays(t *testing.T) {
 }
 
 func TestStage1RaceRelays(t *testing.T) {
-	t.Parallel()
 	//NOTE: this is a race between me and relay resulting in a full tunnel from me to them via relay
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
@@ -755,8 +729,8 @@ func TestStage1RaceRelays(t *testing.T) {
 	assertTunnel(t, theirVpnIpNet[0].Addr(), relayVpnIpNet[0].Addr(), theirControl, relayControl, r)
 
 	r.Log("Trigger a handshake from both them and me via relay to them and me")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 	r.Log("Wait for a packet from them to me")
 	p := r.RouteForAllUntilTxTun(myControl)
@@ -770,12 +744,12 @@ func TestStage1RaceRelays(t *testing.T) {
 }
 
 func TestStage1RaceRelays2(t *testing.T) {
-	t.Parallel()
 	//NOTE: this is a race between me and relay resulting in a full tunnel from me to them via relay
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "relay  ", "10.128.0.128/24", m{"relay": m{"am_relay": true}})
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them   ", "10.128.0.2/24", m{"relay": m{"use_relays": true}})
+	l := NewTestLogger()
 
 	// Teach my how to get to the relay and that their can be reached via the relay
 	myControl.InjectLightHouseAddr(relayVpnIpNet[0].Addr(), relayUdpAddr)
@@ -797,41 +771,49 @@ func TestStage1RaceRelays2(t *testing.T) {
 	theirControl.Start()
 
 	r.Log("Get a tunnel between me and relay")
+	l.Info("Get a tunnel between me and relay")
 	assertTunnel(t, myVpnIpNet[0].Addr(), relayVpnIpNet[0].Addr(), myControl, relayControl, r)
 
 	r.Log("Get a tunnel between them and relay")
+	l.Info("Get a tunnel between them and relay")
 	assertTunnel(t, theirVpnIpNet[0].Addr(), relayVpnIpNet[0].Addr(), theirControl, relayControl, r)
 
 	r.Log("Trigger a handshake from both them and me via relay to them and me")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them")))
+	l.Info("Trigger a handshake from both them and me via relay to them and me")
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 	//r.RouteUntilAfterMsgType(myControl, header.Control, header.MessageNone)
 	//r.RouteUntilAfterMsgType(theirControl, header.Control, header.MessageNone)
 
-	r.Log("Wait for a packet from them to me; myControl")
+	r.Log("Wait for a packet from them to me")
+	l.Info("Wait for a packet from them to me; myControl")
 	r.RouteForAllUntilTxTun(myControl)
-	r.Log("Wait for a packet from them to me; theirControl")
+	l.Info("Wait for a packet from them to me; theirControl")
 	r.RouteForAllUntilTxTun(theirControl)
 
 	r.Log("Assert the tunnel works")
+	l.Info("Assert the tunnel works")
 	assertTunnel(t, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
 
 	t.Log("Wait until we remove extra tunnels")
-	t.Logf("Waiting for hostinfos to be removed... myControl=%d theirControl=%d relayControl=%d",
-		len(myControl.GetHostmap().Indexes),
-		len(theirControl.GetHostmap().Indexes),
-		len(relayControl.GetHostmap().Indexes),
-	)
+	l.Info("Wait until we remove extra tunnels")
+	l.WithFields(
+		logrus.Fields{
+			"myControl":    len(myControl.GetHostmap().Indexes),
+			"theirControl": len(theirControl.GetHostmap().Indexes),
+			"relayControl": len(relayControl.GetHostmap().Indexes),
+		}).Info("Waiting for hostinfos to be removed...")
 	hostInfos := len(myControl.GetHostmap().Indexes) + len(theirControl.GetHostmap().Indexes) + len(relayControl.GetHostmap().Indexes)
 	retries := 60
 	for hostInfos > 6 && retries > 0 {
 		hostInfos = len(myControl.GetHostmap().Indexes) + len(theirControl.GetHostmap().Indexes) + len(relayControl.GetHostmap().Indexes)
-		t.Logf("Waiting for hostinfos to be removed... myControl=%d theirControl=%d relayControl=%d",
-			len(myControl.GetHostmap().Indexes),
-			len(theirControl.GetHostmap().Indexes),
-			len(relayControl.GetHostmap().Indexes),
-		)
+		l.WithFields(
+			logrus.Fields{
+				"myControl":    len(myControl.GetHostmap().Indexes),
+				"theirControl": len(theirControl.GetHostmap().Indexes),
+				"relayControl": len(relayControl.GetHostmap().Indexes),
+			}).Info("Waiting for hostinfos to be removed...")
 		assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
 		t.Log("Connection manager hasn't ticked yet")
 		time.Sleep(time.Second)
@@ -839,6 +821,7 @@ func TestStage1RaceRelays2(t *testing.T) {
 	}
 
 	r.Log("Assert the tunnel works")
+	l.Info("Assert the tunnel works")
 	assertTunnel(t, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
 
 	myControl.Stop()
@@ -847,7 +830,6 @@ func TestStage1RaceRelays2(t *testing.T) {
 }
 
 func TestRehandshakingRelays(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, relayConfig := newSimpleServer(cert.Version1, ca, caKey, "relay  ", "10.128.0.128/24", m{"relay": m{"am_relay": true}})
@@ -868,7 +850,7 @@ func TestRehandshakingRelays(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	r.Log("Assert the tunnel works")
@@ -951,7 +933,6 @@ func TestRehandshakingRelays(t *testing.T) {
 }
 
 func TestRehandshakingRelaysPrimary(t *testing.T) {
-	t.Parallel()
 	// This test is the same as TestRehandshakingRelays but one of the terminal types is a primary swap winner
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.128/24", m{"relay": m{"use_relays": true}})
@@ -973,7 +954,7 @@ func TestRehandshakingRelaysPrimary(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	r.Log("Assert the tunnel works")
@@ -1056,7 +1037,6 @@ func TestRehandshakingRelaysPrimary(t *testing.T) {
 }
 
 func TestRehandshaking(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, myConfig := newSimpleServer(cert.Version1, ca, caKey, "me  ", "10.128.0.2/24", nil)
 	theirControl, theirVpnIpNet, theirUdpAddr, theirConfig := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.1/24", nil)
@@ -1152,7 +1132,6 @@ func TestRehandshaking(t *testing.T) {
 }
 
 func TestRehandshakingLoser(t *testing.T) {
-	t.Parallel()
 	// The purpose of this test is that the race loser renews their certificate and rehandshakes. The final tunnel
 	// Should be the one with the new certificate
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -1251,7 +1230,6 @@ func TestRehandshakingLoser(t *testing.T) {
 }
 
 func TestRaceRegression(t *testing.T) {
-	t.Parallel()
 	// This test forces stage 1, stage 2, stage 1 to be received by me from them
 	// We had a bug where we were not finding the duplicate handshake and responding to the final stage 1 which
 	// caused a cross-linked hostinfo
@@ -1275,8 +1253,8 @@ func TestRaceRegression(t *testing.T) {
 	//them rx stage:2 initiatorIndex=120607833 responderIndex=4209862089
 
 	t.Log("Start both handshakes")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them")))
+	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 	t.Log("Get both stage 1")
 	myStage1ForThem := myControl.GetFromUDP(true)
@@ -1312,7 +1290,6 @@ func TestRaceRegression(t *testing.T) {
 }
 
 func TestV2NonPrimaryWithLighthouse(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	lhControl, lhVpnIpNet, lhUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "lh  ", "10.128.0.1/24, ff::1/64", m{"lighthouse": m{"am_lighthouse": true}})
 
@@ -1353,7 +1330,6 @@ func TestV2NonPrimaryWithLighthouse(t *testing.T) {
 }
 
 func TestV2NonPrimaryWithOffNetLighthouse(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	lhControl, lhVpnIpNet, lhUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "lh  ", "2001::1/64", m{"lighthouse": m{"am_lighthouse": true}})
 
@@ -1393,84 +1369,7 @@ func TestV2NonPrimaryWithOffNetLighthouse(t *testing.T) {
 	theirControl.Stop()
 }
 
-func TestLighthouseUpdateOnReload(t *testing.T) {
-	t.Parallel()
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-
-	// Create the lighthouse
-	lhControl, lhVpnIpNet, lhUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "lh", "10.128.0.1/24", m{"lighthouse": m{"am_lighthouse": true}})
-
-	// Create a client with NO lighthouse configured and a long update interval.
-	// The initial SendUpdate at startup will be a no-op since no lighthouses are known.
-	myControl, myVpnIpNet, _, myConfig := newSimpleServer(cert.Version2, ca, caKey, "me", "10.128.0.2/24", m{
-		"lighthouse": m{
-			"interval": 600,
-			"local_allow_list": m{
-				"10.0.0.0/24": true,
-				"::/0":        false,
-			},
-		},
-	})
-
-	r := router.NewR(t, lhControl, myControl)
-	defer r.RenderFlow()
-
-	lhControl.Start()
-	myControl.Start()
-
-	// Drain any startup packets (there should be none meaningful)
-	r.FlushAll()
-
-	// Verify lighthouse has no knowledge of the client
-	assert.Nil(t, lhControl.QueryLighthouse(myVpnIpNet[0].Addr()))
-
-	// Build a new config that adds the lighthouse
-	newSettings := make(m)
-	for k, v := range myConfig.Settings {
-		newSettings[k] = v
-	}
-	newSettings["static_host_map"] = m{
-		lhVpnIpNet[0].Addr().String(): []any{lhUdpAddr.String()},
-	}
-	newSettings["lighthouse"] = m{
-		"hosts":    []any{lhVpnIpNet[0].Addr().String()},
-		"interval": 600,
-		"local_allow_list": m{
-			"10.0.0.0/24": true,
-			"::/0":        false,
-		},
-	}
-	newCfg, err := yaml.Marshal(newSettings)
-	require.NoError(t, err)
-
-	// Reload the config. The lighthouse.hosts change triggers TriggerUpdate,
-	// which wakes the update worker. It calls SendUpdate, initiating a
-	// handshake to the new lighthouse and caching the HostUpdateNotification.
-	require.NoError(t, myConfig.ReloadConfigString(string(newCfg)))
-
-	// Route until the lighthouse receives the HostUpdateNotification.
-	// This covers: handshake stage 1, stage 2, then the cached update.
-	done := make(chan struct{})
-	go func() {
-		r.RouteForAllUntilAfterMsgTypeTo(lhControl, header.LightHouse, 0)
-		close(done)
-	}()
-	select {
-	case <-done:
-	case <-time.After(5 * time.Second):
-		t.Fatal("timed out waiting for lighthouse update after config reload")
-	}
-
-	// Verify lighthouse now has the client's addresses
-	assert.NotNil(t, lhControl.QueryLighthouse(myVpnIpNet[0].Addr()))
-
-	r.RenderHostmaps("Final hostmaps", lhControl, myControl)
-	lhControl.Stop()
-	myControl.Stop()
-}
-
 func TestGoodHandshakeUnsafeDest(t *testing.T) {
-	t.Parallel()
 	unsafePrefix := "192.168.6.0/24"
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServerWithUdpAndUnsafeNetworks(cert.Version2, ca, caKey, "spooky", "10.128.0.2/24", netip.MustParseAddrPort("10.64.0.2:4242"), unsafePrefix, nil)
@@ -1492,7 +1391,7 @@ func TestGoodHandshakeUnsafeDest(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Send a udp packet through to begin standing up the tunnel, this should come out the other side")
-	myControl.InjectTunPacket(BuildTunUDPPacket(spookyDest, 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(spookyDest, 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 
 	t.Log("Have them consume my stage 0 packet. They have a tunnel now")
 	theirControl.InjectUDPPacket(myControl.GetFromUDP(true))
@@ -1520,7 +1419,7 @@ func TestGoodHandshakeUnsafeDest(t *testing.T) {
 	assertUdpPacket(t, []byte("Hi from me"), myCachedPacket, myVpnIpNet[0].Addr(), spookyDest, 80, 80)
 
 	//reply
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnIpNet[0].Addr(), 80, spookyDest, 80, []byte("Hi from the spookyman")))
+	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, spookyDest, 80, []byte("Hi from the spookyman"))
 	//wait for reply
 	theirControl.WaitForType(1, 0, myControl)
 	theirCachedPacket := myControl.GetFromTun(true)

e2e/helpers_test.go

@@ -4,7 +4,8 @@
 package e2e
 
 import (
-	"log/slog"
+	"fmt"
+	"io"
 	"net/netip"
 	"os"
 	"strings"
@@ -14,13 +15,12 @@ import (
 	"dario.cat/mergo"
 	"github.com/google/gopacket"
 	"github.com/google/gopacket/layers"
-
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/e2e/router"
-	"github.com/slackhq/nebula/logging"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"go.yaml.in/yaml/v3"
@@ -132,7 +132,8 @@ func newSimpleServerWithUdpAndUnsafeNetworks(v cert.Version, caCrt cert.Certific
 			"port": udpAddr.Port(),
 		},
 		"logging": m{
-			"level": testLogLevelName(),
+			"timestamp_format": fmt.Sprintf("%v 15:04:05.000000", name),
+			"level":            l.Level.String(),
 		},
 		"timers": m{
 			"pending_deletion_interval": 2,
@@ -233,7 +234,8 @@ func newServer(caCrt []cert.Certificate, certs []cert.Certificate, key []byte, o
 			"port": udpAddr.Port(),
 		},
 		"logging": m{
-			"level": testLogLevelName(),
+			"timestamp_format": fmt.Sprintf("%v 15:04:05.000000", certs[0].Name()),
+			"level":            l.Level.String(),
 		},
 		"timers": m{
 			"pending_deletion_interval": 2,
@@ -292,12 +294,12 @@ func deadline(t *testing.T, seconds time.Duration) doneCb {
 
 func assertTunnel(t testing.TB, vpnIpA, vpnIpB netip.Addr, controlA, controlB *nebula.Control, r *router.R) {
 	// Send a packet from them to me
-	controlB.InjectTunPacket(BuildTunUDPPacket(vpnIpA, 80, vpnIpB, 90, []byte("Hi from B")))
+	controlB.InjectTunUDPPacket(vpnIpA, 80, vpnIpB, 90, []byte("Hi from B"))
 	bPacket := r.RouteForAllUntilTxTun(controlA)
 	assertUdpPacket(t, []byte("Hi from B"), bPacket, vpnIpB, vpnIpA, 90, 80)
 
 	// And once more from me to them
-	controlA.InjectTunPacket(BuildTunUDPPacket(vpnIpB, 80, vpnIpA, 90, []byte("Hello from A")))
+	controlA.InjectTunUDPPacket(vpnIpB, 80, vpnIpA, 90, []byte("Hello from A"))
 	aPacket := r.RouteForAllUntilTxTun(controlB)
 	assertUdpPacket(t, []byte("Hello from A"), aPacket, vpnIpA, vpnIpB, 90, 80)
 }
@@ -377,87 +379,24 @@ func getAddrs(ns []netip.Prefix) []netip.Addr {
 	return a
 }
 
-func NewTestLogger() *slog.Logger {
+func NewTestLogger() *logrus.Logger {
+	l := logrus.New()
+
 	v := os.Getenv("TEST_LOGS")
 	if v == "" {
-		return slog.New(slog.DiscardHandler)
+		l.SetOutput(io.Discard)
+		l.SetLevel(logrus.PanicLevel)
+		return l
 	}
 
-	level := slog.LevelInfo
 	switch v {
 	case "2":
-		level = slog.LevelDebug
+		l.SetLevel(logrus.DebugLevel)
 	case "3":
-		level = logging.LevelTrace
+		l.SetLevel(logrus.TraceLevel)
+	default:
+		l.SetLevel(logrus.InfoLevel)
 	}
-	return slog.New(slog.NewTextHandler(os.Stderr, &slog.HandlerOptions{Level: level}))
-}
-
-// testLogLevelName returns the level name string accepted by logging.ApplyConfig
-// for the current TEST_LOGS setting. Kept in sync with NewTestLogger.
-func testLogLevelName() string {
-	switch os.Getenv("TEST_LOGS") {
-	case "2":
-		return "debug"
-	case "3":
-		return "trace"
-	case "":
-		return "info"
-	}
-	return "info"
-}
-
-// BuildTunUDPPacket assembles an IP+UDP packet suitable for Control.InjectTunPacket.
-// Using UDP here because it's a simpler protocol.
-func BuildTunUDPPacket(toAddr netip.Addr, toPort uint16, fromAddr netip.Addr, fromPort uint16, data []byte) []byte {
-	serialize := make([]gopacket.SerializableLayer, 0)
-	var netLayer gopacket.NetworkLayer
-	if toAddr.Is6() {
-		if !fromAddr.Is6() {
-			panic("Cant send ipv6 to ipv4")
-		}
-		ip := &layers.IPv6{
-			Version:    6,
-			NextHeader: layers.IPProtocolUDP,
-			SrcIP:      fromAddr.Unmap().AsSlice(),
-			DstIP:      toAddr.Unmap().AsSlice(),
-		}
-		serialize = append(serialize, ip)
-		netLayer = ip
-	} else {
-		if !fromAddr.Is4() {
-			panic("Cant send ipv4 to ipv6")
-		}
-
-		ip := &layers.IPv4{
-			Version:  4,
-			TTL:      64,
-			Protocol: layers.IPProtocolUDP,
-			SrcIP:    fromAddr.Unmap().AsSlice(),
-			DstIP:    toAddr.Unmap().AsSlice(),
-		}
-		serialize = append(serialize, ip)
-		netLayer = ip
-	}
-
-	udp := layers.UDP{
-		SrcPort: layers.UDPPort(fromPort),
-		DstPort: layers.UDPPort(toPort),
-	}
-	if err := udp.SetNetworkLayerForChecksum(netLayer); err != nil {
-		panic(err)
-	}
-
-	buffer := gopacket.NewSerializeBuffer()
-	opt := gopacket.SerializeOptions{
-		ComputeChecksums: true,
-		FixLengths:       true,
-	}
-
-	serialize = append(serialize, &udp, gopacket.Payload(data))
-	if err := gopacket.SerializeLayers(buffer, opt, serialize...); err != nil {
-		panic(err)
-	}
-
-	return buffer.Bytes()
+
+	return l
 }

e2e/leak_test.go

@@ -1,47 +0,0 @@
-//go:build e2e_testing
-// +build e2e_testing
-
-package e2e
-
-import (
-	"testing"
-	"time"
-
-	"github.com/slackhq/nebula/cert"
-	"github.com/slackhq/nebula/cert_test"
-	"github.com/slackhq/nebula/e2e/router"
-	"go.uber.org/goleak"
-)
-
-// TestNoGoroutineLeaks brings up two nebula instances, completes a tunnel,
-// stops both, and asserts no goroutines leak past the shutdown. goleak's
-// retry mechanism gives the wg.Wait()-driven goroutines a moment to drain
-// before failing the assertion.
-//
-// Intentionally NOT t.Parallel()'d: concurrent tests would have their own
-// goroutines running and trip the assertion.
-func TestNoGoroutineLeaks(t *testing.T) {
-	defer goleak.VerifyNone(t)
-
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", nil)
-
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	myControl.Stop()
-	theirControl.Stop()
-	r.RenderFlow()
-
-	// Settle period: Stop() is non-blocking; the wg-driven goroutines need
-	// a moment to drain. goleak retries internally too, but a short explicit
-	// settle reduces flakes when the suite is busy.
-	time.Sleep(50 * time.Millisecond)
-}

e2e/router/router.go

@@ -13,7 +13,6 @@ import (
 	"regexp"
 	"sort"
 	"sync"
-	"sync/atomic"
 	"testing"
 	"time"
 
@@ -25,19 +24,6 @@ import (
 	"golang.org/x/exp/maps"
 )
 
-// outNatKey is the (from, to) pair used by outNat. Comparable struct, so it works as a map key without the
-// allocation cost of a string-concat key.
-type outNatKey struct {
-	from, to netip.AddrPort
-}
-
-// fannedPacket pairs a UDP TX packet with its source control so the router can route it after popping from
-// the fan-in channel.
-type fannedPacket struct {
-	from *nebula.Control
-	pkt  *udp.Packet
-}
-
 type R struct {
 	// Simple map of the ip:port registered on a control to the control
 	// Basically a router, right?
@@ -48,28 +34,12 @@ type R struct {
 
 	// A last used map, if an inbound packet hit the inNat map then
 	// all return packets should use the same last used inbound address for the outbound sender
-	outNat map[outNatKey]netip.AddrPort
+	// map[from address + ":" + to address] => ip:port to rewrite in the udp packet to receiver
+	outNat map[string]netip.AddrPort
 
 	// A map of vpn ip to the nebula control it belongs to
 	vpnControls map[netip.Addr]*nebula.Control
 
-	// Cached select infrastructure for RouteForAllUntilTxTun.
-	// The controls map is immutable after NewR so the cases are good for the test lifetime.
-	// We only rebuild if a different receiver is asked.
-	selRecvCtl *nebula.Control
-	selCases   []reflect.SelectCase
-	selCtls    []*nebula.Control
-
-	// Optional fan-in mode for hot-path benchmarks: one forwarder goroutine per control drains UDP TX into udpFanIn,
-	// so RouteForAllUntilTxTun can do a fixed 2-way native select instead of paying reflect.Select per call.
-	// Off by default (would otherwise interleave with tests that use GetFromUDP directly on the same control).
-	// Enabled by EnableFanIn.
-	udpFanIn  chan fannedPacket
-	stopFanIn chan struct{}
-	fanInWG   sync.WaitGroup
-	fanInMu   sync.Mutex
-	fanInOn   atomic.Bool
-
 	ignoreFlows []ignoreFlow
 	flow        []flowEntry
 
@@ -149,7 +119,7 @@ func NewR(t testing.TB, controls ...*nebula.Control) *R {
 		controls:     make(map[netip.AddrPort]*nebula.Control),
 		vpnControls:  make(map[netip.Addr]*nebula.Control),
 		inNat:        make(map[netip.AddrPort]*nebula.Control),
-		outNat:       make(map[outNatKey]netip.AddrPort),
+		outNat:       make(map[string]netip.AddrPort),
 		flow:         []flowEntry{},
 		ignoreFlows:  []ignoreFlow{},
 		fn:           filepath.Join("mermaid", fmt.Sprintf("%s.md", t.Name())),
@@ -183,10 +153,8 @@ func NewR(t testing.TB, controls ...*nebula.Control) *R {
 			case <-ctx.Done():
 				return
 			case <-clockSource.C:
-				r.Lock()
 				r.renderHostmaps("clock tick")
 				r.renderFlow()
-				r.Unlock()
 			}
 		}
 	}()
@@ -212,21 +180,15 @@ func (r *R) AddRoute(ip netip.Addr, port uint16, c *nebula.Control) {
 // RenderFlow renders the packet flow seen up until now and stops further automatic renders from happening.
 func (r *R) RenderFlow() {
 	r.cancelRender()
-	r.Lock()
-	defer r.Unlock()
 	r.renderFlow()
 }
 
 // CancelFlowLogs stops flow logs from being tracked and destroys any logs already collected
 func (r *R) CancelFlowLogs() {
 	r.cancelRender()
-	r.Lock()
 	r.flow = nil
-	r.Unlock()
 }
 
-// renderFlow writes the flow log to disk. Caller must hold r.Lock. renderFlow reads r.flow / r.additionalGraphs and
-// the *packet pointers stashed inside, all of which are mutated under the same lock by routing paths.
 func (r *R) renderFlow() {
 	if r.flow == nil {
 		return
@@ -472,157 +434,68 @@ func (r *R) RouteUntilTxTun(sender *nebula.Control, receiver *nebula.Control) []
 				panic("No control for udp tx " + a.String())
 			}
 			fp := r.unlockedInjectFlow(sender, c, p, false)
-			c.InjectUDPPacket(p) // copies internally; original is ours to release
+			c.InjectUDPPacket(p)
 			fp.WasReceived()
 			r.Unlock()
-			p.Release()
 		}
 	}
 }
 
-// RouteForAllUntilTxTun will route for everyone and return when a packet is seen on the receiver's tun.
-// If a control's UDP TX address can't be matched to a registered control, we panic.
-//
-// For allocation-sensitive callers (hot-path benchmarks, in particular relay
-// benches with 3+ controls), call EnableFanIn() first.
+// RouteForAllUntilTxTun will route for everyone and return when a packet is seen on receivers tun
+// If the router doesn't have the nebula controller for that address, we panic
 func (r *R) RouteForAllUntilTxTun(receiver *nebula.Control) []byte {
-	if r.fanInOn.Load() {
-		return r.routeFanIn(receiver)
-	}
-	return r.routeReflect(receiver)
-}
-
-// routeFanIn is the alloc-free path used when EnableFanIn is in effect.
-func (r *R) routeFanIn(receiver *nebula.Control) []byte {
-	tunTx := receiver.GetTunTxChan()
-	for {
-		select {
-		case p := <-tunTx:
-			r.Lock()
-			if r.flow != nil {
-				np := udp.Packet{Data: make([]byte, len(p))}
-				copy(np.Data, p)
-				r.unlockedInjectFlow(receiver, receiver, &np, true)
-			}
-			r.Unlock()
-			return p
-		case fp := <-r.udpFanIn:
-			r.routeUDP(fp.from, fp.pkt)
-		}
-	}
-}
-
-// routeReflect is the default reflect.Select-based path. Pays the boxing allocation per call but doesn't interfere
-// with tests that pull packets directly from controls' UDP TX channels via GetFromUDP.
-func (r *R) routeReflect(receiver *nebula.Control) []byte {
-	sc, cm := r.selectCasesFor(receiver)
-	for {
-		x, rx, _ := reflect.Select(sc)
-		if x == 0 {
-			p := rx.Interface().([]byte)
-			r.Lock()
-			if r.flow != nil {
-				np := udp.Packet{Data: make([]byte, len(p))}
-				copy(np.Data, p)
-				r.unlockedInjectFlow(cm[x], cm[x], &np, true)
-			}
-			r.Unlock()
-			return p
-		}
-		r.routeUDP(cm[x], rx.Interface().(*udp.Packet))
-	}
-}
-
-// EnableFanIn switches RouteForAllUntilTxTun to the alloc-free fan-in path.
-// One forwarder goroutine per registered control drains UDP TX into a shared channel that RouteForAllUntilTxTun selects
-// on alongside the receiver's TUN TX channel.
-func (r *R) EnableFanIn() {
-	r.fanInMu.Lock()
-	defer r.fanInMu.Unlock()
-	if r.fanInOn.Load() {
-		return
-	}
-	r.udpFanIn = make(chan fannedPacket, 32)
-	r.stopFanIn = make(chan struct{})
-	for _, c := range r.controls {
-		r.startFanInWorker(c)
-	}
-	r.fanInOn.Store(true)
-	r.t.Cleanup(r.stopFanInWorkers)
-}
-
-// startFanInWorker spawns a goroutine that drains c's UDP TX into r.udpFanIn.
-func (r *R) startFanInWorker(c *nebula.Control) {
-	r.fanInWG.Add(1)
-	udpTx := c.GetUDPTxChan()
-	go func() {
-		defer r.fanInWG.Done()
-		for {
-			select {
-			case <-r.stopFanIn:
-				return
-			case p := <-udpTx:
-				select {
-				case <-r.stopFanIn:
-					p.Release()
-					return
-				case r.udpFanIn <- fannedPacket{from: c, pkt: p}:
-				}
-			}
-		}
-	}()
-}
-
-// stopFanInWorkers signals the fan-in goroutines to exit and waits for them.
-func (r *R) stopFanInWorkers() {
-	r.fanInMu.Lock()
-	wasOn := r.fanInOn.Swap(false)
-	r.fanInMu.Unlock()
-	if !wasOn {
-		return
-	}
-	close(r.stopFanIn)
-	r.fanInWG.Wait()
-}
-
-// routeUDP forwards a UDP TX packet from the named source control to the destination control derived from p.To,
-// releasing the source packet after InjectUDPPacket has copied its bytes into a fresh pool slot.
-func (r *R) routeUDP(from *nebula.Control, p *udp.Packet) {
-	r.Lock()
-	defer r.Unlock()
-	a := from.GetUDPAddr()
-	c := r.getControl(a, p.To, p)
-	if c == nil {
-		panic(fmt.Sprintf("No control for udp tx %s", p.To))
-	}
-	fp := r.unlockedInjectFlow(from, c, p, false)
-	c.InjectUDPPacket(p) // copies internally; original is ours to release
-	fp.WasReceived()
-	p.Release()
-}
-
-// selectCasesFor returns the SelectCase array used by routeReflect: one slot for the receiver's TUN TX channel followed
-// by one per control's UDP TX channel. Cached for the test lifetime, only rebuilt if the receiver changes.
-func (r *R) selectCasesFor(receiver *nebula.Control) ([]reflect.SelectCase, []*nebula.Control) {
-	r.Lock()
-	defer r.Unlock()
-	if r.selRecvCtl == receiver && r.selCases != nil {
-		return r.selCases, r.selCtls
-	}
 	sc := make([]reflect.SelectCase, len(r.controls)+1)
 	cm := make([]*nebula.Control, len(r.controls)+1)
-	sc[0] = reflect.SelectCase{Dir: reflect.SelectRecv, Chan: reflect.ValueOf(receiver.GetTunTxChan())}
-	cm[0] = receiver
-	i := 1
+
+	i := 0
+	sc[i] = reflect.SelectCase{
+		Dir:  reflect.SelectRecv,
+		Chan: reflect.ValueOf(receiver.GetTunTxChan()),
+		Send: reflect.Value{},
+	}
+	cm[i] = receiver
+
+	i++
 	for _, c := range r.controls {
-		sc[i] = reflect.SelectCase{Dir: reflect.SelectRecv, Chan: reflect.ValueOf(c.GetUDPTxChan())}
+		sc[i] = reflect.SelectCase{
+			Dir:  reflect.SelectRecv,
+			Chan: reflect.ValueOf(c.GetUDPTxChan()),
+			Send: reflect.Value{},
+		}
+
 		cm[i] = c
 		i++
 	}
-	r.selRecvCtl = receiver
-	r.selCases = sc
-	r.selCtls = cm
-	return sc, cm
+
+	for {
+		x, rx, _ := reflect.Select(sc)
+		r.Lock()
+
+		if x == 0 {
+			// we are the tun tx, we can exit
+			p := rx.Interface().([]byte)
+			np := udp.Packet{Data: make([]byte, len(p))}
+			copy(np.Data, p)
+
+			r.unlockedInjectFlow(cm[x], cm[x], &np, true)
+			r.Unlock()
+			return p
+
+		} else {
+			// we are a udp tx, route and continue
+			p := rx.Interface().(*udp.Packet)
+			a := cm[x].GetUDPAddr()
+			c := r.getControl(a, p.To, p)
+			if c == nil {
+				r.Unlock()
+				panic(fmt.Sprintf("No control for udp tx %s", p.To))
+			}
+			fp := r.unlockedInjectFlow(cm[x], c, p, false)
+			c.InjectUDPPacket(p)
+			fp.WasReceived()
+		}
+		r.Unlock()
+	}
 }
 
 // RouteExitFunc will call the whatDo func with each udp packet from sender.
@@ -649,7 +522,6 @@ func (r *R) RouteExitFunc(sender *nebula.Control, whatDo ExitFunc) {
 		switch e {
 		case ExitNow:
 			r.Unlock()
-			p.Release()
 			return
 
 		case RouteAndExit:
@@ -657,7 +529,6 @@ func (r *R) RouteExitFunc(sender *nebula.Control, whatDo ExitFunc) {
 			receiver.InjectUDPPacket(p)
 			fp.WasReceived()
 			r.Unlock()
-			p.Release()
 			return
 
 		case KeepRouting:
@@ -670,7 +541,6 @@ func (r *R) RouteExitFunc(sender *nebula.Control, whatDo ExitFunc) {
 		}
 
 		r.Unlock()
-		p.Release()
 	}
 }
 
@@ -771,7 +641,6 @@ func (r *R) RouteForAllExitFunc(whatDo ExitFunc) {
 		switch e {
 		case ExitNow:
 			r.Unlock()
-			p.Release()
 			return
 
 		case RouteAndExit:
@@ -779,7 +648,6 @@ func (r *R) RouteForAllExitFunc(whatDo ExitFunc) {
 			receiver.InjectUDPPacket(p)
 			fp.WasReceived()
 			r.Unlock()
-			p.Release()
 			return
 
 		case KeepRouting:
@@ -791,7 +659,6 @@ func (r *R) RouteForAllExitFunc(whatDo ExitFunc) {
 			panic(fmt.Sprintf("Unknown exitFunc return: %v", e))
 		}
 		r.Unlock()
-		p.Release()
 	}
 }
 
@@ -835,20 +702,19 @@ func (r *R) FlushAll() {
 		}
 		receiver.InjectUDPPacket(p)
 		r.Unlock()
-		p.Release()
 	}
 }
 
 // getControl performs or seeds NAT translation and returns the control for toAddr, p from fields may change
 // This is an internal router function, the caller must hold the lock
 func (r *R) getControl(fromAddr, toAddr netip.AddrPort, p *udp.Packet) *nebula.Control {
-	if newAddr, ok := r.outNat[outNatKey{from: fromAddr, to: toAddr}]; ok {
+	if newAddr, ok := r.outNat[fromAddr.String()+":"+toAddr.String()]; ok {
 		p.From = newAddr
 	}
 
 	c, ok := r.inNat[toAddr]
 	if ok {
-		r.outNat[outNatKey{from: c.GetUDPAddr(), to: fromAddr}] = toAddr
+		r.outNat[c.GetUDPAddr().String()+":"+fromAddr.String()] = toAddr
 		return c
 	}
 

e2e/sshd_test.go

@@ -1,125 +0,0 @@
-//go:build e2e_testing
-// +build e2e_testing
-
-package e2e
-
-import (
-	"crypto/ed25519"
-	"crypto/rand"
-	"encoding/pem"
-	"net"
-	"strings"
-	"testing"
-	"time"
-
-	"github.com/slackhq/nebula/cert"
-	"github.com/slackhq/nebula/cert_test"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"golang.org/x/crypto/ssh"
-)
-
-func TestSSHDLifecycle(t *testing.T) {
-	// TestSSHDLifecycle exercises the in-process sshd through several config reloads and a Control.Stop.
-	ca, _, caKey, _ := cert_test.NewTestCaCert(
-		cert.Version1, cert.Curve_CURVE25519,
-		time.Now(), time.Now().Add(10*time.Minute),
-		nil, nil, []string{},
-	)
-
-	hostKeyPEM := generateSSHHostKey(t)
-	clientSigner, clientAuthKey := generateSSHClientKey(t)
-	sshdAddr := allocLoopbackPort(t)
-
-	overrides := m{
-		"sshd": m{
-			"enabled":  true,
-			"listen":   sshdAddr,
-			"host_key": hostKeyPEM,
-			"authorized_users": []m{{
-				"user": "tester",
-				"keys": []string{clientAuthKey},
-			}},
-		},
-	}
-	control, _, _, _ := newSimpleServer(cert.Version1, ca, caKey, "sshd-test", "10.222.0.1/24", overrides)
-	control.Start()
-	t.Cleanup(func() { control.Stop() })
-
-	// sshd binds in a goroutine after Start returns; wait for it.
-	require.Eventually(t, func() bool { return canDial(sshdAddr) }, 2*time.Second, 25*time.Millisecond,
-		"sshd never started listening")
-
-	for i := 1; i <= 3; i++ {
-		out := sshExecReload(t, sshdAddr, clientSigner)
-		assert.Contains(t, out, "Reloading config", "reload cycle %d", i)
-		require.Eventually(t, func() bool { return canDial(sshdAddr) }, 2*time.Second, 25*time.Millisecond,
-			"sshd not listening after reload cycle %d", i)
-	}
-
-	control.Stop()
-	require.Eventually(t, func() bool { return !canDial(sshdAddr) }, 2*time.Second, 25*time.Millisecond,
-		"sshd still listening after Control.Stop")
-}
-
-func canDial(addr string) bool {
-	c, err := net.DialTimeout("tcp", addr, 100*time.Millisecond)
-	if err != nil {
-		return false
-	}
-	_ = c.Close()
-	return true
-}
-
-// allocLoopbackPort grabs an unused TCP port on 127.0.0.1, closes it, and returns the address. There
-// is a small race between releasing the port and the sshd reclaiming it; in practice the OS keeps the
-// port available long enough for the test to bind it.
-func allocLoopbackPort(t *testing.T) string {
-	t.Helper()
-	l, err := net.Listen("tcp", "127.0.0.1:0")
-	require.NoError(t, err)
-	addr := l.Addr().String()
-	require.NoError(t, l.Close())
-	return addr
-}
-
-func generateSSHHostKey(t *testing.T) string {
-	t.Helper()
-	_, priv, err := ed25519.GenerateKey(rand.Reader)
-	require.NoError(t, err)
-	block, err := ssh.MarshalPrivateKey(priv, "nebula-e2e-host")
-	require.NoError(t, err)
-	return string(pem.EncodeToMemory(block))
-}
-
-func generateSSHClientKey(t *testing.T) (ssh.Signer, string) {
-	t.Helper()
-	_, priv, err := ed25519.GenerateKey(rand.Reader)
-	require.NoError(t, err)
-	signer, err := ssh.NewSignerFromKey(priv)
-	require.NoError(t, err)
-	auth := strings.TrimSpace(string(ssh.MarshalAuthorizedKey(signer.PublicKey())))
-	return signer, auth
-}
-
-func sshExecReload(t *testing.T, addr string, signer ssh.Signer) string {
-	t.Helper()
-	cfg := &ssh.ClientConfig{
-		User:            "tester",
-		Auth:            []ssh.AuthMethod{ssh.PublicKeys(signer)},
-		HostKeyCallback: ssh.InsecureIgnoreHostKey(),
-		Timeout:         2 * time.Second,
-	}
-	client, err := ssh.Dial("tcp", addr, cfg)
-	require.NoError(t, err)
-	defer client.Close()
-
-	sess, err := client.NewSession()
-	require.NoError(t, err)
-	defer sess.Close()
-
-	// reload tears the channel down before sending exit-status, so Output returns an error on the
-	// channel close. The output buffer still has whatever the reload callback wrote before that.
-	out, _ := sess.Output("reload")
-	return string(out)
-}

e2e/tunnels_test.go

@@ -19,7 +19,6 @@ import (
 )
 
 func TestDropInactiveTunnels(t *testing.T) {
-	t.Parallel()
 	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
 	// under ideal conditions
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -64,7 +63,6 @@ func TestDropInactiveTunnels(t *testing.T) {
 }
 
 func TestCertUpgrade(t *testing.T) {
-	t.Parallel()
 	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
 	// under ideal conditions
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -159,7 +157,6 @@ func TestCertUpgrade(t *testing.T) {
 }
 
 func TestCertDowngrade(t *testing.T) {
-	t.Parallel()
 	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
 	// under ideal conditions
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -258,7 +255,6 @@ func TestCertDowngrade(t *testing.T) {
 }
 
 func TestCertMismatchCorrection(t *testing.T) {
-	t.Parallel()
 	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
 	// under ideal conditions
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
@@ -326,7 +322,6 @@ func TestCertMismatchCorrection(t *testing.T) {
 }
 
 func TestCrossStackRelaysWork(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "me     ", "10.128.0.1/24,fc00::1/64", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "relay  ", "10.128.0.128/24,fc00::128/64", m{"relay": m{"am_relay": true}})
@@ -355,14 +350,14 @@ func TestCrossStackRelaysWork(t *testing.T) {
 	theirControl.Start()
 
 	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnV6.Addr(), 80, myVpnV6.Addr(), 80, []byte("Hi from me")))
+	myControl.InjectTunUDPPacket(theirVpnV6.Addr(), 80, myVpnV6.Addr(), 80, []byte("Hi from me"))
 
 	p := r.RouteForAllUntilTxTun(theirControl)
 	r.Log("Assert the tunnel works")
 	assertUdpPacket(t, []byte("Hi from me"), p, myVpnV6.Addr(), theirVpnV6.Addr(), 80, 80)
 
 	t.Log("reply?")
-	theirControl.InjectTunPacket(BuildTunUDPPacket(myVpnV6.Addr(), 80, theirVpnV6.Addr(), 80, []byte("Hi from them")))
+	theirControl.InjectTunUDPPacket(myVpnV6.Addr(), 80, theirVpnV6.Addr(), 80, []byte("Hi from them"))
 	p = r.RouteForAllUntilTxTun(myControl)
 	assertUdpPacket(t, []byte("Hi from them"), p, theirVpnV6.Addr(), myVpnV6.Addr(), 80, 80)
 
@@ -374,7 +369,6 @@ func TestCrossStackRelaysWork(t *testing.T) {
 }
 
 func TestCloseTunnelAuthenticated(t *testing.T) {
-	t.Parallel()
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "5s"}})
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "10m"}})

examples/config.yml

@@ -138,14 +138,6 @@ listen:
   # max, net.core.rmem_max and net.core.wmem_max
   #read_buffer: 10485760
   #write_buffer: 10485760
-
-  # On Windows only
-  # When true, Nebula installs a WFP (Windows Filtering Platform) PERMIT filter scoped to UDP at the listener port.
-  # WFP sits below Windows Defender Firewall, so this lets peer handshakes reach Nebula's outside socket regardless
-  # of WDF's inbound rules.
-  # Default true; set to false to leave WDF in charge of inbound decisions on the listener port. Not reloadable.
-  #windows_bypass_wdf: true
-
   # By default, Nebula replies to packets it has no tunnel for with a "recv_error" packet. This packet helps speed up reconnection
   # in the case that Nebula on either side did not shut down cleanly. This response can be abused as a way to discover if Nebula is running
   # on a host though. This option lets you configure if you want to send "recv_error" packets always, never, or only to private network remotes.
@@ -171,21 +163,17 @@ listen:
 
 punchy:
   # Continues to punch inbound/outbound at a regular interval to avoid expiration of firewall nat mappings
-  # This setting is reloadable.
   punch: true
 
   # respond means that a node you are trying to reach will connect back out to you if your hole punching fails
   # this is extremely useful if one node is behind a difficult nat, such as a symmetric NAT
   # Default is false
-  # This setting is reloadable.
   #respond: true
 
   # delays a punch response for misbehaving NATs, default is 1 second.
-  # This setting is reloadable.
   #delay: 1s
 
   # set the delay before attempting punchy.respond. Default is 5 seconds. respond must be true to take effect.
-  # This setting is reloadable.
   #respond_delay: 5s
 
 # Cipher allows you to choose between the available ciphers for your network. Options are chachapoly or aes
@@ -294,24 +282,6 @@ tun:
     #  metric: 100
     #  install: true
 
-  # On Windows only, sets the network category of the nebula interface. Without this, Windows often
-  # leaves the network as "Unidentified" and treats it as Public, which makes the host firewall more
-  # restrictive than you usually want for an overlay between trusted peers. Valid values:
-  #   private             - treat the nebula network as a private/trusted network (default)
-  #   public              - treat it as a public/untrusted network
-  #   domain              - treat it as a domain-authenticated network
-  #   unset               - leave whatever Windows decided alone
-  # Not reloadable.
-  #network_category: private
-
-  # On Windows only
-  # When true, Nebula installs a WFP (Windows Filtering Platform) PERMIT filter scoped to the nebula adapter LUID.
-  # WFP sits below Windows Defender Firewall, so this lets inbound traffic through regardless of WDF rules.
-  # Filters are auto-removed when the adapter goes away.
-  # See listen.windows_bypass_wdf for the matching control over inbound to nebula's outside UDP listener.
-  # Default true; set to false to leave WDF in charge of inbound decisions on the nebula interface. Not reloadable.
-  #windows_bypass_wdf: true
-
   # On linux only, set to true to manage unsafe routes directly on the system route table with gateway routes instead of
   # in nebula configuration files. Default false, not reloadable.
   #use_system_route_table: false
@@ -322,21 +292,24 @@ tun:
 
 # Configure logging level
 logging:
-  # trace, debug, info, warn, or error. Default is info and is reloadable.
-  # fatal and panic are accepted for backwards compatibility and map to error.
-  #NOTE: Debug and trace modes can log remotely controlled/untrusted data which can quickly fill a disk in some
-  # scenarios. Debug and trace logging are also CPU intensive and will decrease performance overall.
-  # Only enable debug or trace logging while actively investigating an issue.
+  # panic, fatal, error, warning, info, or debug. Default is info and is reloadable.
+  #NOTE: Debug mode can log remotely controlled/untrusted data which can quickly fill a disk in some
+  # scenarios. Debug logging is also CPU intensive and will decrease performance overall.
+  # Only enable debug logging while actively investigating an issue.
   level: info
-  # json or text formats currently available. Default is text.
+  # json or text formats currently available. Default is text
   format: text
-  # Disable timestamp logging. Useful when output is redirected to a logging system that already adds timestamps. Default is false.
+  # Disable timestamp logging. useful when output is redirected to logging system that already adds timestamps. Default is false
   #disable_timestamp: true
-  # Timestamps use RFC3339Nano ("2006-01-02T15:04:05.999999999Z07:00") and are not configurable.
+  # timestamp format is specified in Go time format, see:
+  #     https://golang.org/pkg/time/#pkg-constants
+  # default when `format: json`: "2006-01-02T15:04:05Z07:00" (RFC3339)
+  # default when `format: text`:
+  #     when TTY attached: seconds since beginning of execution
+  #     otherwise: "2006-01-02T15:04:05Z07:00" (RFC3339)
+  # As an example, to log as RFC3339 with millisecond precision, set to:
+  #timestamp_format: "2006-01-02T15:04:05.000Z07:00"
 
-# The stats section is reloadable. A HUP may change the backend, toggle stats
-# on or off, switch the listen/host address, or pick up new DNS for the
-# configured graphite host.
 #stats:
   #type: graphite
   #prefix: nebula
@@ -354,12 +327,10 @@ logging:
   # enables counter metrics for meta packets
   #   e.g.: `messages.tx.handshake`
   # NOTE: `message.{tx,rx}.recv_error` is always emitted
-  # Not reloadable.
   #message_metrics: false
 
   # enables detailed counter metrics for lighthouse packets
   #   e.g.: `lighthouse.rx.HostQuery`
-  # Not reloadable.
   #lighthouse_metrics: false
 
 # Handshake Manager Settings

examples/go_service/main.go

@@ -7,9 +7,9 @@ import (
 	"net"
 	"os"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/service"
 )
@@ -64,7 +64,8 @@ pki:
 		return err
 	}
 
-	logger := logging.NewLogger(os.Stdout)
+	logger := logrus.New()
+	logger.Out = os.Stdout
 
 	ctrl, err := nebula.Main(&cfg, false, "custom-app", logger, overlay.NewUserDeviceFromConfig)
 	if err != nil {

firewall.go

@@ -1,13 +1,11 @@
 package nebula
 
 import (
-	"context"
 	"crypto/sha256"
 	"encoding/hex"
 	"errors"
 	"fmt"
 	"hash/fnv"
-	"log/slog"
 	"net/netip"
 	"reflect"
 	"slices"
@@ -18,9 +16,11 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
+	"github.com/slackhq/nebula/firewall/events"
 )
 
 type FirewallInterface interface {
@@ -68,7 +68,15 @@ type Firewall struct {
 	incomingMetrics     firewallMetrics
 	outgoingMetrics     firewallMetrics
 
-	l *slog.Logger
+	// reporter is the optional embedder-supplied event sink. Immutable for
+	// the lifetime of this Firewall; Control.SetFirewallEventReporter
+	// installs it by shallow-copying the Firewall under the conntrack lock
+	// and swapping the pointer, and reloadFirewall carries it forward.
+	// Read unsynchronized on the data path: the preceding Firewall-pointer
+	// read pins the field's value for the duration of that call.
+	reporter events.Reporter
+
+	l *logrus.Logger
 }
 
 type firewallMetrics struct {
@@ -132,7 +140,7 @@ type firewallLocalCIDR struct {
 
 // NewFirewall creates a new Firewall object. A TimerWheel is created for you from the provided timeouts.
 // The certificate provided should be the highest version loaded in memory.
-func NewFirewall(l *slog.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Duration, c cert.Certificate) *Firewall {
+func NewFirewall(l *logrus.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Duration, c cert.Certificate) *Firewall {
 	//TODO: error on 0 duration
 	var tmin, tmax time.Duration
 
@@ -192,7 +200,7 @@ func NewFirewall(l *slog.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Dur
 	}
 }
 
-func NewFirewallFromConfig(l *slog.Logger, cs *CertState, c *config.C) (*Firewall, error) {
+func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firewall, error) {
 	certificate := cs.getCertificate(cert.Version2)
 	if certificate == nil {
 		certificate = cs.getCertificate(cert.Version1)
@@ -220,7 +228,7 @@ func NewFirewallFromConfig(l *slog.Logger, cs *CertState, c *config.C) (*Firewal
 	case "drop":
 		fw.InSendReject = false
 	default:
-		l.Warn("invalid firewall.inbound_action, defaulting to `drop`", "action", inboundAction)
+		l.WithField("action", inboundAction).Warn("invalid firewall.inbound_action, defaulting to `drop`")
 		fw.InSendReject = false
 	}
 
@@ -231,7 +239,7 @@ func NewFirewallFromConfig(l *slog.Logger, cs *CertState, c *config.C) (*Firewal
 	case "drop":
 		fw.OutSendReject = false
 	default:
-		l.Warn("invalid firewall.outbound_action, defaulting to `drop`", "action", outboundAction)
+		l.WithField("action", outboundAction).Warn("invalid firewall.outbound_action, defaulting to `drop`")
 		fw.OutSendReject = false
 	}
 
@@ -269,7 +277,7 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
 	case firewall.ProtoICMP, firewall.ProtoICMPv6:
 		//ICMP traffic doesn't have ports, so we always coerce to "any", even if a value is provided
 		if startPort != firewall.PortAny {
-			f.l.Warn("ignoring port specification for ICMP firewall rule", "startPort", startPort)
+			f.l.WithField("startPort", startPort).Warn("ignoring port specification for ICMP firewall rule")
 		}
 		startPort = firewall.PortAny
 		endPort = firewall.PortAny
@@ -291,9 +299,8 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
 	if !incoming {
 		direction = "outgoing"
 	}
-	f.l.Info("Firewall rule added",
-		"firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha},
-	)
+	f.l.WithField("firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha}).
+		Info("Firewall rule added")
 
 	return fp.addRule(f, startPort, endPort, groups, host, cidr, localCidr, caName, caSha)
 }
@@ -316,7 +323,7 @@ func (f *Firewall) GetRuleHashes() string {
 	return "SHA:" + f.GetRuleHash() + ",FNV:" + strconv.FormatUint(uint64(f.GetRuleHashFNV()), 10)
 }
 
-func AddFirewallRulesFromConfig(l *slog.Logger, inbound bool, c *config.C, fw FirewallInterface) error {
+func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw FirewallInterface) error {
 	var table string
 	if inbound {
 		table = "firewall.inbound"
@@ -374,7 +381,7 @@ func AddFirewallRulesFromConfig(l *slog.Logger, inbound bool, c *config.C, fw Fi
 			startPort = firewall.PortAny
 			endPort = firewall.PortAny
 			if sPort != "" {
-				l.Warn("ignoring port specification for ICMP firewall rule", "port", sPort)
+				l.WithField("port", sPort).Warn("ignoring port specification for ICMP firewall rule")
 			}
 		default:
 			return fmt.Errorf("%s rule #%v; proto was not understood; `%s`", table, i, r.Proto)
@@ -398,11 +405,7 @@ func AddFirewallRulesFromConfig(l *slog.Logger, inbound bool, c *config.C, fw Fi
 		}
 
 		if warning := r.sanity(); warning != nil {
-			l.Warn("firewall rule sanity check",
-				"table", table,
-				"rule", i,
-				"warning", warning,
-			)
+			l.Warnf("%s rule #%v; %s", table, i, warning)
 		}
 
 		err = fw.AddRule(inbound, proto, startPort, endPort, r.Groups, r.Host, r.Cidr, r.LocalCidr, r.CAName, r.CASha)
@@ -422,23 +425,27 @@ var ErrNoMatchingRule = errors.New("no matching rule in firewall table")
 
 // Drop returns an error if the packet should be dropped, explaining why. It
 // returns nil if the packet should not be dropped.
-func (f *Firewall) Drop(fp firewall.Packet, incoming bool, h *HostInfo, caPool *cert.CAPool, localCache firewall.ConntrackCache) error {
+func (f *Firewall) Drop(fp firewall.Packet, ctx firewall.PacketContext, incoming bool, h *HostInfo, caPool *cert.CAPool, localCache firewall.ConntrackCache) error {
 	// Check if we spoke to this tuple, if we did then allow this packet
 	if f.inConns(fp, h, caPool, localCache) {
 		return nil
 	}
 
+	peerCert := h.ConnectionState.peerCert
+
 	// Make sure remote address matches nebula certificate, and determine how to treat it
 	if h.networks == nil {
 		// Simple case: Certificate has one address and no unsafe networks
 		if h.vpnAddrs[0] != fp.RemoteAddr {
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			f.reportDrop(incoming, events.DropInvalidRemoteIP, fp, ctx, peerCert)
 			return ErrInvalidRemoteIP
 		}
 	} else {
 		nwType, ok := h.networks.Lookup(fp.RemoteAddr)
 		if !ok {
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			f.reportDrop(incoming, events.DropInvalidRemoteIP, fp, ctx, peerCert)
 			return ErrInvalidRemoteIP
 		}
 		switch nwType {
@@ -446,11 +453,13 @@ func (f *Firewall) Drop(fp firewall.Packet, incoming bool, h *HostInfo, caPool *
 			break // nothing special
 		case NetworkTypeVPNPeer:
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			f.reportDrop(incoming, events.DropPeerRejected, fp, ctx, peerCert)
 			return ErrPeerRejected // reject for now, one day this may have different FW rules
 		case NetworkTypeUnsafe:
 			break // nothing special, one day this may have different FW rules
 		default:
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
+			f.reportDrop(incoming, events.DropUnknownNetwork, fp, ctx, peerCert)
 			return ErrUnknownNetworkType //should never happen
 		}
 	}
@@ -458,6 +467,7 @@ func (f *Firewall) Drop(fp firewall.Packet, incoming bool, h *HostInfo, caPool *
 	// Make sure we are supposed to be handling this local ip address
 	if !f.routableNetworks.Contains(fp.LocalAddr) {
 		f.metrics(incoming).droppedLocalAddr.Inc(1)
+		f.reportDrop(incoming, events.DropInvalidLocalIP, fp, ctx, peerCert)
 		return ErrInvalidLocalIP
 	}
 
@@ -467,13 +477,14 @@ func (f *Firewall) Drop(fp firewall.Packet, incoming bool, h *HostInfo, caPool *
 	}
 
 	// We now know which firewall table to check against
-	if !table.match(fp, incoming, h.ConnectionState.peerCert, caPool) {
+	if !table.match(fp, incoming, peerCert, caPool) {
 		f.metrics(incoming).droppedNoRule.Inc(1)
+		f.reportDrop(incoming, events.DropNoMatchingRule, fp, ctx, peerCert)
 		return ErrNoMatchingRule
 	}
 
 	// We always want to conntrack since it is a faster operation
-	f.addConn(fp, incoming)
+	f.addConn(fp, ctx, incoming, peerCert)
 
 	return nil
 }
@@ -492,6 +503,59 @@ func (f *Firewall) Destroy() {
 	//TODO: clean references if/when needed
 }
 
+func (f *Firewall) reportDrop(incoming bool, reason events.DropReason, fp firewall.Packet, ctx firewall.PacketContext, peerCert *cert.CachedCertificate) {
+	r := f.reporter
+	if r == nil {
+		return
+	}
+	r.ReportDrop(events.DropEvent{
+		Incoming:     incoming,
+		Reason:       reason,
+		Packet:       fp,
+		Context:      ctx,
+		PeerCert:     peerCert,
+		RulesVersion: f.rulesVersion,
+	})
+}
+
+func (f *Firewall) reportFlowCreate(incoming bool, fp firewall.Packet, ctx firewall.PacketContext, peerCert *cert.CachedCertificate) {
+	r := f.reporter
+	if r == nil {
+		return
+	}
+	r.ReportFlowCreate(events.FlowCreateEvent{
+		Incoming:     incoming,
+		Packet:       fp,
+		Context:      ctx,
+		PeerCert:     peerCert,
+		RulesVersion: f.rulesVersion,
+	})
+}
+
+func (f *Firewall) reportFlowEvict(incoming bool, fp firewall.Packet, rulesVersion uint16, expired bool) {
+	r := f.reporter
+	if r == nil {
+		return
+	}
+	r.ReportFlowEvict(events.FlowEvictEvent{
+		Incoming:     incoming,
+		Packet:       fp,
+		RulesVersion: rulesVersion,
+		Expired:      expired,
+	})
+}
+
+func (f *Firewall) reportRulesReload(oldVersion, newVersion uint16) {
+	r := f.reporter
+	if r == nil {
+		return
+	}
+	r.ReportRulesReload(events.RulesReloadEvent{
+		OldVersion: oldVersion,
+		NewVersion: newVersion,
+	})
+}
+
 func (f *Firewall) EmitStats() {
 	conntrack := f.Conntrack
 	conntrack.Lock()
@@ -534,26 +598,28 @@ func (f *Firewall) inConns(fp firewall.Packet, h *HostInfo, caPool *cert.CAPool,
 
 		// We now know which firewall table to check against
 		if !table.match(fp, c.incoming, h.ConnectionState.peerCert, caPool) {
-			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				h.logger(f.l).Debug("dropping old conntrack entry, does not match new ruleset",
-					"fwPacket", fp,
-					"incoming", c.incoming,
-					"rulesVersion", f.rulesVersion,
-					"oldRulesVersion", c.rulesVersion,
-				)
+			if f.l.Level >= logrus.DebugLevel {
+				h.logger(f.l).
+					WithField("fwPacket", fp).
+					WithField("incoming", c.incoming).
+					WithField("rulesVersion", f.rulesVersion).
+					WithField("oldRulesVersion", c.rulesVersion).
+					Debugln("dropping old conntrack entry, does not match new ruleset")
 			}
+			oldRulesVersion := c.rulesVersion
 			delete(conntrack.Conns, fp)
+			f.reportFlowEvict(c.incoming, fp, oldRulesVersion, false)
 			conntrack.Unlock()
 			return false
 		}
 
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			h.logger(f.l).Debug("keeping old conntrack entry, does match new ruleset",
-				"fwPacket", fp,
-				"incoming", c.incoming,
-				"rulesVersion", f.rulesVersion,
-				"oldRulesVersion", c.rulesVersion,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			h.logger(f.l).
+				WithField("fwPacket", fp).
+				WithField("incoming", c.incoming).
+				WithField("rulesVersion", f.rulesVersion).
+				WithField("oldRulesVersion", c.rulesVersion).
+				Debugln("keeping old conntrack entry, does match new ruleset")
 		}
 
 		c.rulesVersion = f.rulesVersion
@@ -577,7 +643,7 @@ func (f *Firewall) inConns(fp firewall.Packet, h *HostInfo, caPool *cert.CAPool,
 	return true
 }
 
-func (f *Firewall) addConn(fp firewall.Packet, incoming bool) {
+func (f *Firewall) addConn(fp firewall.Packet, ctx firewall.PacketContext, incoming bool, peerCert *cert.CachedCertificate) {
 	var timeout time.Duration
 	c := &conn{}
 
@@ -592,7 +658,8 @@ func (f *Firewall) addConn(fp firewall.Packet, incoming bool) {
 
 	conntrack := f.Conntrack
 	conntrack.Lock()
-	if _, ok := conntrack.Conns[fp]; !ok {
+	_, existing := conntrack.Conns[fp]
+	if !existing {
 		conntrack.TimerWheel.Advance(time.Now())
 		conntrack.TimerWheel.Add(fp, timeout)
 	}
@@ -603,6 +670,13 @@ func (f *Firewall) addConn(fp firewall.Packet, incoming bool) {
 	c.rulesVersion = f.rulesVersion
 	c.Expires = time.Now().Add(timeout)
 	conntrack.Conns[fp] = c
+
+	// Report only when this represents a genuinely new flow. Fires under the
+	// conntrack lock so FlowCreate/FlowEvict events stay ordered relative to
+	// RulesReloadEvent, which also fires under this lock.
+	if !existing {
+		f.reportFlowCreate(incoming, fp, ctx, peerCert)
+	}
 	conntrack.Unlock()
 }
 
@@ -626,7 +700,10 @@ func (f *Firewall) evict(p firewall.Packet) {
 	}
 
 	// This conn is done
+	rulesVersion := t.rulesVersion
+	incoming := t.incoming
 	delete(conntrack.Conns, p)
+	f.reportFlowEvict(incoming, p, rulesVersion, true)
 }
 
 func (ft *FirewallTable) match(p firewall.Packet, incoming bool, c *cert.CachedCertificate, caPool *cert.CAPool) bool {
@@ -941,7 +1018,7 @@ type rule struct {
 	CASha     string
 }
 
-func convertRule(l *slog.Logger, p any, table string, i int) (rule, error) {
+func convertRule(l *logrus.Logger, p any, table string, i int) (rule, error) {
 	r := rule{}
 
 	m, ok := p.(map[string]any)
@@ -972,10 +1049,7 @@ func convertRule(l *slog.Logger, p any, table string, i int) (rule, error) {
 			return r, errors.New("group should contain a single value, an array with more than one entry was provided")
 		}
 
-		l.Warn("group was an array with a single value, converting to simple value",
-			"table", table,
-			"rule", i,
-		)
+		l.Warnf("%s rule #%v; group was an array with a single value, converting to simple value", table, i)
 		m["group"] = v[0]
 	}
 

firewall/cache.go

@@ -2,9 +2,10 @@ package firewall
 
 import (
 	"context"
-	"log/slog"
 	"sync/atomic"
 	"time"
+
+	"github.com/sirupsen/logrus"
 )
 
 // ConntrackCache is used as a local routine cache to know if a given flow
@@ -15,17 +16,15 @@ type ConntrackCacheTicker struct {
 	cacheV    uint64
 	cacheTick atomic.Uint64
 
-	l     *slog.Logger
 	cache ConntrackCache
 }
 
-func NewConntrackCacheTicker(ctx context.Context, l *slog.Logger, d time.Duration) *ConntrackCacheTicker {
+func NewConntrackCacheTicker(ctx context.Context, d time.Duration) *ConntrackCacheTicker {
 	if d == 0 {
 		return nil
 	}
 
 	c := &ConntrackCacheTicker{
-		l:     l,
 		cache: ConntrackCache{},
 	}
 
@@ -49,15 +48,15 @@ func (c *ConntrackCacheTicker) tick(ctx context.Context, d time.Duration) {
 
 // Get checks if the cache ticker has moved to the next version before returning
 // the map. If it has moved, we reset the map.
-func (c *ConntrackCacheTicker) Get() ConntrackCache {
+func (c *ConntrackCacheTicker) Get(l *logrus.Logger) ConntrackCache {
 	if c == nil {
 		return nil
 	}
 	if tick := c.cacheTick.Load(); tick != c.cacheV {
 		c.cacheV = tick
 		if ll := len(c.cache); ll > 0 {
-			if c.l.Enabled(context.Background(), slog.LevelDebug) {
-				c.l.Debug("resetting conntrack cache", "len", ll)
+			if l.Level == logrus.DebugLevel {
+				l.WithField("len", ll).Debug("resetting conntrack cache")
 			}
 			c.cache = make(ConntrackCache, ll)
 		}

firewall/cache_test.go

@@ -1,69 +0,0 @@
-package firewall
-
-import (
-	"bytes"
-	"log/slog"
-	"strings"
-	"testing"
-
-	"github.com/slackhq/nebula/test"
-	"github.com/stretchr/testify/assert"
-)
-
-// The tests below pin the log format produced by ConntrackCacheTicker.Get
-// so changes cannot silently break what operators are grepping for. The
-// ticker's internal state (cache + cacheTick) is poked directly to avoid
-// racing a goroutine-driven tick in tests.
-
-func newFixedTicker(t *testing.T, l *slog.Logger, cacheLen int) *ConntrackCacheTicker {
-	t.Helper()
-	c := &ConntrackCacheTicker{
-		l:     l,
-		cache: make(ConntrackCache, cacheLen),
-	}
-	for i := 0; i < cacheLen; i++ {
-		c.cache[Packet{LocalPort: uint16(i) + 1}] = struct{}{}
-	}
-	c.cacheTick.Store(1) // cacheV starts at 0, so Get() takes the reset path
-	return c
-}
-
-func TestConntrackCacheTicker_Get_TextFormat(t *testing.T) {
-	buf := &bytes.Buffer{}
-	l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelDebug)
-
-	c := newFixedTicker(t, l, 3)
-	c.Get()
-
-	assert.Equal(t, "level=DEBUG msg=\"resetting conntrack cache\" len=3\n", buf.String())
-}
-
-func TestConntrackCacheTicker_Get_JSONFormat(t *testing.T) {
-	buf := &bytes.Buffer{}
-	l := test.NewJSONLoggerWithOutput(buf, slog.LevelDebug)
-
-	c := newFixedTicker(t, l, 2)
-	c.Get()
-
-	assert.JSONEq(t, `{"level":"DEBUG","msg":"resetting conntrack cache","len":2}`, strings.TrimSpace(buf.String()))
-}
-
-func TestConntrackCacheTicker_Get_QuietBelowDebug(t *testing.T) {
-	buf := &bytes.Buffer{}
-	l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelInfo)
-
-	c := newFixedTicker(t, l, 5)
-	c.Get()
-
-	assert.Empty(t, buf.String())
-}
-
-func TestConntrackCacheTicker_Get_QuietWhenCacheEmpty(t *testing.T) {
-	buf := &bytes.Buffer{}
-	l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelDebug)
-
-	c := newFixedTicker(t, l, 0)
-	c.Get()
-
-	assert.Empty(t, buf.String())
-}

firewall/events/events.go

@@ -0,0 +1,102 @@
+// Package events defines the opt-in firewall event reporting interface.
+//
+// Nebula emits raw packet-level events (drops, flow creations, flow evictions,
+// rule reloads) and does no aggregation, counting, batching, rule-description,
+// transport, or timestamping. Embedders correlate events back to yaml rules
+// out of band and capture whatever clock they need themselves. All Report*
+// methods are invoked while nebula holds internal locks and must be
+// non-blocking.
+//
+// Events are passed to Report* methods by value. Implementations must not
+// take the address of a received event: doing so forces Go's escape
+// analysis to move the event to the heap and costs one allocation per call.
+// To forward an event, either copy its fields into the reporter's own
+// pooled record or send it through a value-typed channel (chan DropEvent,
+// not chan *DropEvent).
+package events
+
+import (
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/firewall"
+)
+
+type DropReason uint8
+
+const (
+	DropInvalidLocalIP DropReason = iota
+	DropInvalidRemoteIP
+	DropPeerRejected
+	DropUnknownNetwork
+	DropNoMatchingRule
+)
+
+func (r DropReason) String() string {
+	switch r {
+	case DropInvalidLocalIP:
+		return "invalid_local_ip"
+	case DropInvalidRemoteIP:
+		return "invalid_remote_ip"
+	case DropPeerRejected:
+		return "peer_rejected"
+	case DropUnknownNetwork:
+		return "unknown_network"
+	case DropNoMatchingRule:
+		return "no_matching_rule"
+	default:
+		return "unknown"
+	}
+}
+
+// DropEvent is emitted for every packet that fails the firewall check. Drops
+// are not aggregated; every drop produces one event.
+type DropEvent struct {
+	Incoming     bool
+	Reason       DropReason
+	Packet       firewall.Packet
+	Context      firewall.PacketContext
+	PeerCert     *cert.CachedCertificate
+	RulesVersion uint16
+}
+
+// FlowCreateEvent is emitted when a packet is allowed and a new conntrack
+// entry is created. Subsequent packets in the same flow do not re-emit.
+type FlowCreateEvent struct {
+	Incoming     bool
+	Packet       firewall.Packet
+	Context      firewall.PacketContext
+	PeerCert     *cert.CachedCertificate
+	RulesVersion uint16
+}
+
+// FlowEvictEvent is emitted when a conntrack entry is removed. Context is
+// not carried: timer-wheel eviction has no packet in hand, and reload
+// revalidation evicts the OLD flow rather than the triggering packet.
+// RulesVersion is the version under which the flow was originally allowed,
+// which may differ from the current firewall version.
+type FlowEvictEvent struct {
+	Incoming     bool
+	Packet       firewall.Packet
+	RulesVersion uint16
+	// Expired is true when eviction was due to conntrack timeout; false when
+	// the entry was removed because it failed re-validation after a reload.
+	Expired bool
+}
+
+// RulesReloadEvent is emitted once after each successful firewall reload.
+// Reporters that bucket state by RulesVersion should close the old bucket
+// and open a new one on receipt.
+type RulesReloadEvent struct {
+	OldVersion uint16
+	NewVersion uint16
+}
+
+// Reporter is the embedder-supplied sink for firewall events. Implementations
+// that want a timestamp should call time.Now() themselves at the top of the
+// method; nebula does not provide one. See the package doc for the
+// do-not-take-address rule.
+type Reporter interface {
+	ReportDrop(DropEvent)
+	ReportFlowCreate(FlowCreateEvent)
+	ReportFlowEvict(FlowEvictEvent)
+	ReportRulesReload(RulesReloadEvent)
+}

firewall/packet.go

@@ -31,6 +31,27 @@ type Packet struct {
 	Fragment   bool
 }
 
+// PacketContext carries additional parsed details about a packet that are
+// useful for event reporting but deliberately kept out of Packet so Packet
+// can keep being used as a conntrack map key. Populated alongside Packet by
+// newPacket.
+//
+// Fields are interpreted based on Packet.Protocol:
+//   - ProtoTCP: TCPFlags is meaningful; ICMPType / ICMPCode are zero
+//   - ProtoICMP, ProtoICMPv6: ICMPType / ICMPCode are meaningful; TCPFlags is zero
+//   - ProtoUDP and others: only Length is meaningful
+type PacketContext struct {
+	// Length is the total IP packet length in bytes, including headers.
+	Length uint16
+	// TCPFlags is the flag byte from the TCP header (bits for FIN, SYN, RST,
+	// PSH, ACK, URG, ECE, CWR).
+	TCPFlags uint8
+	// ICMPType is the type field of the ICMP / ICMPv6 header.
+	ICMPType uint8
+	// ICMPCode is the code field of the ICMP / ICMPv6 header.
+	ICMPCode uint8
+}
+
 func (fp *Packet) Copy() *Packet {
 	return &Packet{
 		LocalAddr:  fp.LocalAddr,

firewall_events_test.go

@@ -0,0 +1,731 @@
+package nebula
+
+import (
+	"net"
+	"net/netip"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/gaissmai/bart"
+	"github.com/google/gopacket"
+	"github.com/google/gopacket/layers"
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/firewall"
+	"github.com/slackhq/nebula/firewall/events"
+	"github.com/slackhq/nebula/test"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// recordingReporter captures every event fired against it. Its methods take
+// the conntrack lock implicitly (via the firewall code path that invokes
+// them), so we synchronize accumulator mutations with a small mutex to keep
+// the race detector happy across goroutines in case a test introduces any.
+type recordingReporter struct {
+	mu      sync.Mutex
+	drops   []recordedDrop
+	creates []recordedCreate
+	evicts  []recordedEvict
+	reloads []recordedReload
+}
+
+type recordedDrop struct {
+	incoming     bool
+	reason       events.DropReason
+	remote       netip.Addr
+	local        netip.Addr
+	peerName     string
+	rulesVersion uint16
+	ctx          firewall.PacketContext
+}
+
+type recordedCreate struct {
+	incoming     bool
+	remote       netip.Addr
+	local        netip.Addr
+	peerName     string
+	rulesVersion uint16
+	ctx          firewall.PacketContext
+}
+
+type recordedEvict struct {
+	incoming     bool
+	remote       netip.Addr
+	local        netip.Addr
+	rulesVersion uint16
+	expired      bool
+}
+
+type recordedReload struct {
+	oldVersion uint16
+	newVersion uint16
+}
+
+func (r *recordingReporter) ReportDrop(e events.DropEvent) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	name := ""
+	if e.PeerCert != nil && e.PeerCert.Certificate != nil {
+		name = e.PeerCert.Certificate.Name()
+	}
+	r.drops = append(r.drops, recordedDrop{
+		incoming:     e.Incoming,
+		reason:       e.Reason,
+		remote:       e.Packet.RemoteAddr,
+		local:        e.Packet.LocalAddr,
+		peerName:     name,
+		rulesVersion: e.RulesVersion,
+		ctx:          e.Context,
+	})
+}
+
+func (r *recordingReporter) ReportFlowCreate(e events.FlowCreateEvent) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	name := ""
+	if e.PeerCert != nil && e.PeerCert.Certificate != nil {
+		name = e.PeerCert.Certificate.Name()
+	}
+	r.creates = append(r.creates, recordedCreate{
+		incoming:     e.Incoming,
+		remote:       e.Packet.RemoteAddr,
+		local:        e.Packet.LocalAddr,
+		peerName:     name,
+		rulesVersion: e.RulesVersion,
+		ctx:          e.Context,
+	})
+}
+
+func (r *recordingReporter) ReportFlowEvict(e events.FlowEvictEvent) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.evicts = append(r.evicts, recordedEvict{
+		incoming:     e.Incoming,
+		remote:       e.Packet.RemoteAddr,
+		local:        e.Packet.LocalAddr,
+		rulesVersion: e.RulesVersion,
+		expired:      e.Expired,
+	})
+}
+
+func (r *recordingReporter) ReportRulesReload(e events.RulesReloadEvent) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.reloads = append(r.reloads, recordedReload{
+		oldVersion: e.OldVersion,
+		newVersion: e.NewVersion,
+	})
+}
+
+// eventFixture builds a Firewall wired to a Control plus a packet/hostinfo
+// pair that a test can reuse. By default the ruleset allows the packet;
+// callers mutate fw / p / h as needed before invoking Drop.
+type eventFixture struct {
+	ctl *Control
+	fw  *Firewall
+	p   firewall.Packet
+	h   *HostInfo
+	cp  *cert.CAPool
+}
+
+func newEventFixture(t *testing.T) *eventFixture {
+	t.Helper()
+	l := test.NewLogger()
+
+	// myVpnNetworksTable covers our single peer address so buildNetworks takes
+	// the "simple case" path (h.networks stays nil); tests that want a populated
+	// BART table overwrite h.networks directly.
+	vpnNetworks := new(bart.Lite)
+	vpnNetworks.Insert(netip.MustParsePrefix("1.2.3.0/24"))
+
+	// Use the same cert for "peer" and "local" endpoints, matching the
+	// TestFirewall_Drop fixture style: LocalAddr == RemoteAddr == peer vpn addr.
+	c := &dummyCert{
+		name:     "host1",
+		networks: []netip.Prefix{netip.MustParsePrefix("1.2.3.4/24")},
+		groups:   []string{"default-group"},
+		issuer:   "signer-shasum",
+	}
+	h := &HostInfo{
+		ConnectionState: &ConnectionState{
+			peerCert: &cert.CachedCertificate{
+				Certificate:    c,
+				InvertedGroups: map[string]struct{}{"default-group": {}},
+			},
+		},
+		vpnAddrs: []netip.Addr{netip.MustParseAddr("1.2.3.4")},
+	}
+	h.buildNetworks(vpnNetworks, c)
+
+	fw := NewFirewall(l, time.Minute, time.Minute, time.Minute, c)
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""))
+	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""))
+
+	ctl := &Control{
+		f: &Interface{firewall: fw},
+		l: l,
+	}
+
+	return &eventFixture{
+		ctl: ctl,
+		fw:  fw,
+		p: firewall.Packet{
+			LocalAddr:  netip.MustParseAddr("1.2.3.4"),
+			RemoteAddr: netip.MustParseAddr("1.2.3.4"),
+			LocalPort:  10,
+			RemotePort: 90,
+			Protocol:   firewall.ProtoUDP,
+		},
+		h:  h,
+		cp: cert.NewCAPool(),
+	}
+}
+
+// firewall() returns the currently-installed firewall. Needed because
+// SetFirewallEventReporter replaces it via shallow-copy swap.
+func (f *eventFixture) firewall() *Firewall {
+	return f.ctl.f.firewall
+}
+
+func TestEvents_ReportDrop_InvalidRemoteIP(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	// Packet to an address not in the cert's networks.
+	f.p.RemoteAddr = netip.MustParseAddr("9.9.9.9")
+	assert.Equal(t, ErrInvalidRemoteIP, f.firewall().Drop(f.p, firewall.PacketContext{}, false, f.h, f.cp, nil))
+
+	require.Len(t, r.drops, 1)
+	assert.Equal(t, events.DropInvalidRemoteIP, r.drops[0].reason)
+	assert.False(t, r.drops[0].incoming)
+	assert.Equal(t, "host1", r.drops[0].peerName)
+	assert.Empty(t, r.creates)
+	assert.Empty(t, r.evicts)
+}
+
+func TestEvents_ReportDrop_InvalidLocalIP(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	// LocalAddr outside our routable networks.
+	f.p.LocalAddr = netip.MustParseAddr("9.9.9.9")
+	assert.Equal(t, ErrInvalidLocalIP, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+
+	require.Len(t, r.drops, 1)
+	assert.Equal(t, events.DropInvalidLocalIP, r.drops[0].reason)
+	assert.True(t, r.drops[0].incoming)
+}
+
+func TestEvents_ReportDrop_NoMatchingRule(t *testing.T) {
+	f := newEventFixture(t)
+	// Reset to a firewall with no matching rule.
+	l := test.NewLogger()
+	fw := NewFirewall(l, time.Minute, time.Minute, time.Minute, f.h.ConnectionState.peerCert.Certificate)
+	// Rule that won't match (group not in peer's groups).
+	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "", ""))
+	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "", ""))
+	f.ctl.f.firewall = fw
+
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	assert.Equal(t, ErrNoMatchingRule, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.drops, 1)
+	assert.Equal(t, events.DropNoMatchingRule, r.drops[0].reason)
+}
+
+func TestEvents_ReportDrop_PeerRejected(t *testing.T) {
+	f := newEventFixture(t)
+	// Re-classify the remote as VPNPeer so it triggers DropPeerRejected.
+	f.h.networks = new(bart.Table[NetworkType])
+	f.h.networks.Insert(netip.MustParsePrefix("1.2.3.0/24"), NetworkTypeVPNPeer)
+
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	assert.Equal(t, ErrPeerRejected, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.drops, 1)
+	assert.Equal(t, events.DropPeerRejected, r.drops[0].reason)
+}
+
+func TestEvents_ReportDrop_UnknownNetwork(t *testing.T) {
+	f := newEventFixture(t)
+	// Insert an unrecognized NetworkType value to hit the default branch.
+	f.h.networks = new(bart.Table[NetworkType])
+	f.h.networks.Insert(netip.MustParsePrefix("1.2.3.0/24"), NetworkTypeUnknown)
+
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	assert.Equal(t, ErrUnknownNetworkType, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.drops, 1)
+	assert.Equal(t, events.DropUnknownNetwork, r.drops[0].reason)
+}
+
+func TestEvents_ReportFlowCreate_OnceOnly(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	// First allowed packet creates the conntrack entry.
+	require.NoError(t, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	// Second matching packet on the same tuple is short-circuited by conntrack
+	// and must not fire another FlowCreate.
+	require.NoError(t, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+
+	require.Len(t, r.creates, 1)
+	assert.True(t, r.creates[0].incoming)
+	assert.Equal(t, f.p.RemoteAddr, r.creates[0].remote)
+	assert.Empty(t, r.drops)
+}
+
+func TestEvents_ReportFlowEvict_OnReloadPurge(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	// Create a flow under the current rules.
+	require.NoError(t, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.creates, 1)
+
+	// Simulate a reload that produces rules the existing flow no longer
+	// matches. Bump rulesVersion and replace InRules with an empty table so
+	// revalidation fails.
+	fw := f.firewall()
+	fw.Conntrack.Lock()
+	fw.rulesVersion++
+	fw.InRules = newFirewallTable()
+	fw.Conntrack.Unlock()
+
+	// Next packet triggers re-validation, which fails and evicts the entry.
+	err := fw.Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil)
+	assert.Equal(t, ErrNoMatchingRule, err)
+
+	require.Len(t, r.evicts, 1)
+	assert.False(t, r.evicts[0].expired, "evict from reload purge is not expiration")
+	assert.True(t, r.evicts[0].incoming)
+}
+
+func TestEvents_ReportFlowEvict_OnTimeout(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	require.NoError(t, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.creates, 1)
+
+	// Force expiration by rewinding the entry's deadline.
+	fw := f.firewall()
+	fw.Conntrack.Lock()
+	c := fw.Conntrack.Conns[f.p]
+	require.NotNil(t, c)
+	c.Expires = time.Now().Add(-time.Hour)
+	fw.evict(f.p)
+	fw.Conntrack.Unlock()
+
+	require.Len(t, r.evicts, 1)
+	assert.True(t, r.evicts[0].expired)
+}
+
+func TestEvents_SetNil_Clears(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+	require.NoError(t, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.creates, 1)
+
+	f.ctl.SetFirewallEventReporter(nil)
+	resetConntrack(f.firewall())
+	require.NoError(t, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	// No second create should be recorded.
+	assert.Len(t, r.creates, 1)
+}
+
+func TestEvents_ReporterSurvivesSwap(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	// Simulate a reload by swapping in a fresh Firewall that carries the
+	// reporter forward. Mirrors what reloadFirewall does with the shared
+	// conntrack pointer.
+	l := test.NewLogger()
+	oldFw := f.firewall()
+	newFw := NewFirewall(l, time.Minute, time.Minute, time.Minute, f.h.ConnectionState.peerCert.Certificate)
+	require.NoError(t, newFw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""))
+	require.NoError(t, newFw.AddRule(false, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""))
+	newFw.Conntrack = oldFw.Conntrack
+	newFw.rulesVersion = oldFw.rulesVersion + 1
+	newFw.reporter = oldFw.reporter
+	f.ctl.f.firewall = newFw
+	newFw.reportRulesReload(oldFw.rulesVersion, newFw.rulesVersion)
+
+	require.Len(t, r.reloads, 1)
+	assert.Equal(t, oldFw.rulesVersion, r.reloads[0].oldVersion)
+	assert.Equal(t, newFw.rulesVersion, r.reloads[0].newVersion)
+
+	// Events on the new firewall should still reach the same reporter.
+	require.NoError(t, newFw.Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+	require.Len(t, r.creates, 1)
+	assert.Equal(t, newFw.rulesVersion, r.creates[0].rulesVersion)
+}
+
+func TestEvents_InstallDoesNotMutateOldFirewall(t *testing.T) {
+	f := newEventFixture(t)
+	before := f.firewall()
+
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	after := f.firewall()
+	assert.NotSame(t, before, after, "SetFirewallEventReporter must replace the Firewall pointer")
+	assert.Nil(t, before.reporter, "the pre-install Firewall must remain untouched")
+	assert.NotNil(t, after.reporter)
+}
+
+// --- PacketContext parse tests --------------------------------------------
+
+func mustSerialize(t *testing.T, lrs ...gopacket.SerializableLayer) []byte {
+	t.Helper()
+	buf := gopacket.NewSerializeBuffer()
+	opt := gopacket.SerializeOptions{ComputeChecksums: false, FixLengths: true}
+	require.NoError(t, gopacket.SerializeLayers(buf, opt, lrs...))
+	return buf.Bytes()
+}
+
+func TestPacketContext_IPv4_TCPFlags(t *testing.T) {
+	ip := &layers.IPv4{
+		Version: 4, TTL: 64, Protocol: layers.IPProtocolTCP,
+		SrcIP: net.IPv4(10, 0, 0, 1), DstIP: net.IPv4(10, 0, 0, 2),
+	}
+	tcp := &layers.TCP{SrcPort: 1234, DstPort: 80, SYN: true, ACK: true}
+	require.NoError(t, tcp.SetNetworkLayerForChecksum(ip))
+	data := mustSerialize(t, ip, tcp, gopacket.Payload([]byte("hello")))
+
+	var fp firewall.Packet
+	var ctx firewall.PacketContext
+	require.NoError(t, newPacket(data, true, &fp, &ctx))
+
+	assert.Equal(t, uint8(firewall.ProtoTCP), fp.Protocol)
+	// SYN (0x02) + ACK (0x10) = 0x12
+	assert.Equal(t, uint8(0x12), ctx.TCPFlags)
+	assert.Equal(t, uint16(len(data)), ctx.Length)
+	assert.Equal(t, uint8(0), ctx.ICMPType)
+	assert.Equal(t, uint8(0), ctx.ICMPCode)
+}
+
+func TestPacketContext_IPv4_ICMPTypeCode(t *testing.T) {
+	ip := &layers.IPv4{
+		Version: 4, TTL: 64, Protocol: layers.IPProtocolICMPv4,
+		SrcIP: net.IPv4(10, 0, 0, 1), DstIP: net.IPv4(10, 0, 0, 2),
+	}
+	// Destination Unreachable, code 3 (port unreachable)
+	icmp := &layers.ICMPv4{
+		TypeCode: layers.CreateICMPv4TypeCode(layers.ICMPv4TypeDestinationUnreachable, layers.ICMPv4CodePort),
+	}
+	data := mustSerialize(t, ip, icmp, gopacket.Payload([]byte{0, 0, 0, 0}))
+
+	var fp firewall.Packet
+	var ctx firewall.PacketContext
+	require.NoError(t, newPacket(data, true, &fp, &ctx))
+
+	assert.Equal(t, uint8(firewall.ProtoICMP), fp.Protocol)
+	assert.Equal(t, uint8(layers.ICMPv4TypeDestinationUnreachable), ctx.ICMPType)
+	assert.Equal(t, uint8(layers.ICMPv4CodePort), ctx.ICMPCode)
+	assert.Equal(t, uint16(len(data)), ctx.Length)
+	assert.Equal(t, uint8(0), ctx.TCPFlags)
+}
+
+func TestPacketContext_IPv4_UDPLengthOnly(t *testing.T) {
+	ip := &layers.IPv4{
+		Version: 4, TTL: 64, Protocol: layers.IPProtocolUDP,
+		SrcIP: net.IPv4(10, 0, 0, 1), DstIP: net.IPv4(10, 0, 0, 2),
+	}
+	udp := &layers.UDP{SrcPort: 1234, DstPort: 53}
+	require.NoError(t, udp.SetNetworkLayerForChecksum(ip))
+	data := mustSerialize(t, ip, udp, gopacket.Payload([]byte("query")))
+
+	var fp firewall.Packet
+	var ctx firewall.PacketContext
+	require.NoError(t, newPacket(data, true, &fp, &ctx))
+
+	assert.Equal(t, uint8(firewall.ProtoUDP), fp.Protocol)
+	assert.Equal(t, uint16(len(data)), ctx.Length)
+	assert.Zero(t, ctx.TCPFlags)
+	assert.Zero(t, ctx.ICMPType)
+	assert.Zero(t, ctx.ICMPCode)
+}
+
+func TestPacketContext_IPv6_TCPFlags(t *testing.T) {
+	ip := &layers.IPv6{
+		Version: 6, HopLimit: 64, NextHeader: layers.IPProtocolTCP,
+		SrcIP: net.ParseIP("fd00::1"), DstIP: net.ParseIP("fd00::2"),
+	}
+	tcp := &layers.TCP{SrcPort: 1234, DstPort: 443, FIN: true, ACK: true}
+	require.NoError(t, tcp.SetNetworkLayerForChecksum(ip))
+	data := mustSerialize(t, ip, tcp, gopacket.Payload([]byte("bye")))
+
+	var fp firewall.Packet
+	var ctx firewall.PacketContext
+	require.NoError(t, newPacket(data, true, &fp, &ctx))
+
+	assert.Equal(t, uint8(firewall.ProtoTCP), fp.Protocol)
+	// FIN (0x01) + ACK (0x10) = 0x11
+	assert.Equal(t, uint8(0x11), ctx.TCPFlags)
+	assert.Equal(t, uint16(len(data)), ctx.Length)
+}
+
+func TestPacketContext_IPv6_ICMPv6TypeCode(t *testing.T) {
+	ip := &layers.IPv6{
+		Version: 6, HopLimit: 64, NextHeader: layers.IPProtocolICMPv6,
+		SrcIP: net.ParseIP("fd00::1"), DstIP: net.ParseIP("fd00::2"),
+	}
+	icmp := &layers.ICMPv6{
+		TypeCode: layers.CreateICMPv6TypeCode(layers.ICMPv6TypeDestinationUnreachable, layers.ICMPv6CodePortUnreachable),
+	}
+	require.NoError(t, icmp.SetNetworkLayerForChecksum(ip))
+	data := mustSerialize(t, ip, icmp, gopacket.Payload([]byte{0, 0, 0, 0, 0, 0, 0, 0}))
+
+	var fp firewall.Packet
+	var ctx firewall.PacketContext
+	require.NoError(t, newPacket(data, true, &fp, &ctx))
+
+	assert.Equal(t, uint8(firewall.ProtoICMPv6), fp.Protocol)
+	assert.Equal(t, uint8(layers.ICMPv6TypeDestinationUnreachable), ctx.ICMPType)
+	assert.Equal(t, uint8(layers.ICMPv6CodePortUnreachable), ctx.ICMPCode)
+	assert.Equal(t, uint16(len(data)), ctx.Length)
+}
+
+// TestPacketContext_NilOK confirms a nil context pointer is accepted by
+// newPacket (the hot path may elect not to pass one).
+func TestPacketContext_NilOK(t *testing.T) {
+	ip := &layers.IPv4{
+		Version: 4, TTL: 64, Protocol: layers.IPProtocolUDP,
+		SrcIP: net.IPv4(10, 0, 0, 1), DstIP: net.IPv4(10, 0, 0, 2),
+	}
+	udp := &layers.UDP{SrcPort: 1, DstPort: 2}
+	require.NoError(t, udp.SetNetworkLayerForChecksum(ip))
+	data := mustSerialize(t, ip, udp)
+
+	var fp firewall.Packet
+	require.NoError(t, newPacket(data, true, &fp, nil))
+}
+
+// TestPacketContext_FlowCreateCarriesContext exercises the full Drop -> addConn
+// -> ReportFlowCreate path with a realistic TCP packet and confirms the
+// context makes it into the reporter.
+func TestPacketContext_FlowCreateCarriesContext(t *testing.T) {
+	f := newEventFixture(t)
+	r := &recordingReporter{}
+	f.ctl.SetFirewallEventReporter(r)
+
+	// Hand-construct a matching TCP packet.
+	ctx := firewall.PacketContext{Length: 1500, TCPFlags: 0x12}
+	p := f.p
+	p.Protocol = firewall.ProtoTCP
+	require.NoError(t, f.firewall().Drop(p, ctx, true, f.h, f.cp, nil))
+
+	require.Len(t, r.creates, 1)
+	assert.Equal(t, uint16(1500), r.creates[0].ctx.Length)
+	assert.Equal(t, uint8(0x12), r.creates[0].ctx.TCPFlags)
+}
+
+// --- benchmarks ------------------------------------------------------------
+
+// noopReporter is the cheapest possible reporter. Methods discard the event.
+type noopReporter struct{}
+
+func (noopReporter) ReportDrop(events.DropEvent)             {}
+func (noopReporter) ReportFlowCreate(events.FlowCreateEvent) {}
+func (noopReporter) ReportFlowEvict(events.FlowEvictEvent)   {}
+func (noopReporter) ReportRulesReload(events.RulesReloadEvent) {
+}
+
+// bufferedReporter demonstrates a realistic zero-alloc reporter: each event
+// is forwarded to a value-typed channel. The channel send is a memcpy into
+// the channel's pre-allocated ring buffer -- no heap traffic. A background
+// goroutine would drain these; the bench skips draining to keep the report
+// path pure.
+type bufferedReporter struct {
+	drops  chan events.DropEvent
+	flows  chan events.FlowCreateEvent
+	evicts chan events.FlowEvictEvent
+}
+
+func newBufferedReporter(cap int) *bufferedReporter {
+	return &bufferedReporter{
+		drops:  make(chan events.DropEvent, cap),
+		flows:  make(chan events.FlowCreateEvent, cap),
+		evicts: make(chan events.FlowEvictEvent, cap),
+	}
+}
+
+func (r *bufferedReporter) ReportDrop(e events.DropEvent) {
+	select {
+	case r.drops <- e:
+	default:
+	}
+}
+
+func (r *bufferedReporter) ReportFlowCreate(e events.FlowCreateEvent) {
+	select {
+	case r.flows <- e:
+	default:
+	}
+}
+
+func (r *bufferedReporter) ReportFlowEvict(e events.FlowEvictEvent) {
+	select {
+	case r.evicts <- e:
+	default:
+	}
+}
+
+func (r *bufferedReporter) ReportRulesReload(events.RulesReloadEvent) {}
+
+// pointerReporter is the anti-pattern: it takes the address of the incoming
+// event struct, which forces the callee-side copy onto the heap. Kept for
+// comparison so we can see the alloc cost an unwary reporter would incur.
+type pointerReporter struct {
+	last *events.DropEvent
+}
+
+func (r *pointerReporter) ReportDrop(e events.DropEvent) {
+	r.last = &e
+}
+
+func (r *pointerReporter) ReportFlowCreate(events.FlowCreateEvent) {}
+func (r *pointerReporter) ReportFlowEvict(events.FlowEvictEvent)   {}
+func (r *pointerReporter) ReportRulesReload(events.RulesReloadEvent) {
+}
+
+func newBenchFixture(b *testing.B) *eventFixture {
+	b.Helper()
+	l := test.NewLogger()
+
+	vpnNetworks := new(bart.Lite)
+	vpnNetworks.Insert(netip.MustParsePrefix("1.2.3.0/24"))
+
+	c := &dummyCert{
+		name:     "host1",
+		networks: []netip.Prefix{netip.MustParsePrefix("1.2.3.4/24")},
+		groups:   []string{"default-group"},
+		issuer:   "signer-shasum",
+	}
+	h := &HostInfo{
+		ConnectionState: &ConnectionState{
+			peerCert: &cert.CachedCertificate{
+				Certificate:    c,
+				InvertedGroups: map[string]struct{}{"default-group": {}},
+			},
+		},
+		vpnAddrs: []netip.Addr{netip.MustParseAddr("1.2.3.4")},
+	}
+	h.buildNetworks(vpnNetworks, c)
+
+	fw := NewFirewall(l, time.Minute, time.Minute, time.Minute, c)
+	// Inbound rule that matches our packet; outbound has no match so we can
+	// also benchmark the no-rule drop path.
+	if err := fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""); err != nil {
+		b.Fatal(err)
+	}
+
+	ctl := &Control{f: &Interface{firewall: fw}, l: l}
+	return &eventFixture{
+		ctl: ctl,
+		fw:  fw,
+		p: firewall.Packet{
+			LocalAddr:  netip.MustParseAddr("1.2.3.4"),
+			RemoteAddr: netip.MustParseAddr("1.2.3.4"),
+			LocalPort:  10,
+			RemotePort: 90,
+			Protocol:   firewall.ProtoUDP,
+		},
+		h:  h,
+		cp: cert.NewCAPool(),
+	}
+}
+
+// BenchmarkFirewallDropPath measures the cost of Firewall.Drop on a packet
+// that reaches the no-matching-rule branch (the longest drop path). Compare
+// reporter shapes:
+//
+//	nilReporter      -- no reporter installed (feature cost when off)
+//	noopReporter     -- reporter installed, methods discard args (minimum on-cost)
+//	bufferedReporter -- realistic zero-alloc reporter: value-typed channels
+//	pointerReporter  -- anti-pattern that takes &composite-literal (allocates)
+func BenchmarkFirewallDropPath(b *testing.B) {
+	run := func(b *testing.B, install func(*Control)) {
+		f := newBenchFixture(b)
+		install(f.ctl)
+		b.ReportAllocs()
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			_ = f.firewall().Drop(f.p, firewall.PacketContext{}, false, f.h, f.cp, nil)
+		}
+	}
+
+	b.Run("nilReporter", func(b *testing.B) { run(b, func(*Control) {}) })
+	b.Run("noopReporter", func(b *testing.B) {
+		run(b, func(c *Control) { c.SetFirewallEventReporter(noopReporter{}) })
+	})
+	b.Run("bufferedReporter", func(b *testing.B) {
+		run(b, func(c *Control) { c.SetFirewallEventReporter(newBufferedReporter(1024)) })
+	})
+	b.Run("pointerReporter", func(b *testing.B) {
+		run(b, func(c *Control) { c.SetFirewallEventReporter(&pointerReporter{}) })
+	})
+}
+
+// BenchmarkConntrackCreate measures Firewall.Drop for an allowed inbound
+// packet on a fresh conntrack (so addConn fires each iteration).
+func BenchmarkConntrackCreate(b *testing.B) {
+	run := func(b *testing.B, install func(*Control)) {
+		f := newBenchFixture(b)
+		install(f.ctl)
+		b.ReportAllocs()
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			resetConntrack(f.firewall())
+			_ = f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil)
+		}
+	}
+	b.Run("nilReporter", func(b *testing.B) { run(b, func(*Control) {}) })
+	b.Run("noopReporter", func(b *testing.B) {
+		run(b, func(c *Control) { c.SetFirewallEventReporter(noopReporter{}) })
+	})
+	b.Run("bufferedReporter", func(b *testing.B) {
+		run(b, func(c *Control) { c.SetFirewallEventReporter(newBufferedReporter(1024)) })
+	})
+}
+
+// BenchmarkConntrackHit measures the hot path where a flow is already in
+// conntrack and short-circuits rule evaluation. The reporter slot is checked
+// only on create/evict, so this bench should show the reporter having zero
+// impact regardless of install state.
+func BenchmarkConntrackHit(b *testing.B) {
+	b.Run("nilReporter", func(b *testing.B) {
+		f := newBenchFixture(b)
+		// Prime conntrack.
+		require.NoError(b, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+		b.ReportAllocs()
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			_ = f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil)
+		}
+	})
+
+	b.Run("noopReporter", func(b *testing.B) {
+		f := newBenchFixture(b)
+		f.ctl.SetFirewallEventReporter(noopReporter{})
+		require.NoError(b, f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil))
+		b.ReportAllocs()
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			_ = f.firewall().Drop(f.p, firewall.PacketContext{}, true, f.h, f.cp, nil)
+		}
+	})
+}

firewall_test.go

@@ -3,13 +3,13 @@ package nebula
 import (
 	"bytes"
 	"errors"
-	"log/slog"
 	"math"
 	"net/netip"
 	"testing"
 	"time"
 
 	"github.com/gaissmai/bart"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
@@ -58,8 +58,9 @@ func TestNewFirewall(t *testing.T) {
 }
 
 func TestFirewall_AddRule(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 
 	c := &dummyCert{}
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c)
@@ -176,8 +177,9 @@ func TestFirewall_AddRule(t *testing.T) {
 }
 
 func TestFirewall_Drop(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 	p := firewall.Packet{
@@ -211,49 +213,50 @@ func TestFirewall_Drop(t *testing.T) {
 	cp := cert.NewCAPool()
 
 	// Drop outbound
-	assert.Equal(t, ErrNoMatchingRule, fw.Drop(p, false, &h, cp, nil))
+	assert.Equal(t, ErrNoMatchingRule, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil))
 	// Allow inbound
 	resetConntrack(fw)
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 	// Allow outbound because conntrack
-	require.NoError(t, fw.Drop(p, false, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil))
 
 	// test remote mismatch
 	oldRemote := p.RemoteAddr
 	p.RemoteAddr = netip.MustParseAddr("1.2.3.10")
-	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrInvalidRemoteIP)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil), ErrInvalidRemoteIP)
 	p.RemoteAddr = oldRemote
 
 	// ensure signer doesn't get in the way of group checks
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "", "signer-shasum"))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "", "signer-shasum-bad"))
-	assert.Equal(t, fw.Drop(p, true, &h, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 
 	// test caSha doesn't drop on match
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "", "signer-shasum-bad"))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "", "signer-shasum"))
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 
 	// ensure ca name doesn't get in the way of group checks
 	cp.CAs["signer-shasum"] = &cert.CachedCertificate{Certificate: &dummyCert{name: "ca-good"}}
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "ca-good", ""))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "ca-good-bad", ""))
-	assert.Equal(t, fw.Drop(p, true, &h, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 
 	// test caName doesn't drop on match
 	cp.CAs["signer-shasum"] = &cert.CachedCertificate{Certificate: &dummyCert{name: "ca-good"}}
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "ca-good-bad", ""))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "ca-good", ""))
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 }
 
 func TestFirewall_DropV6(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
@@ -289,44 +292,44 @@ func TestFirewall_DropV6(t *testing.T) {
 	cp := cert.NewCAPool()
 
 	// Drop outbound
-	assert.Equal(t, ErrNoMatchingRule, fw.Drop(p, false, &h, cp, nil))
+	assert.Equal(t, ErrNoMatchingRule, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil))
 	// Allow inbound
 	resetConntrack(fw)
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 	// Allow outbound because conntrack
-	require.NoError(t, fw.Drop(p, false, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil))
 
 	// test remote mismatch
 	oldRemote := p.RemoteAddr
 	p.RemoteAddr = netip.MustParseAddr("fd12::56")
-	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrInvalidRemoteIP)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil), ErrInvalidRemoteIP)
 	p.RemoteAddr = oldRemote
 
 	// ensure signer doesn't get in the way of group checks
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "", "signer-shasum"))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "", "signer-shasum-bad"))
-	assert.Equal(t, fw.Drop(p, true, &h, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 
 	// test caSha doesn't drop on match
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "", "signer-shasum-bad"))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "", "signer-shasum"))
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 
 	// ensure ca name doesn't get in the way of group checks
 	cp.CAs["signer-shasum"] = &cert.CachedCertificate{Certificate: &dummyCert{name: "ca-good"}}
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "ca-good", ""))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "ca-good-bad", ""))
-	assert.Equal(t, fw.Drop(p, true, &h, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 
 	// test caName doesn't drop on match
 	cp.CAs["signer-shasum"] = &cert.CachedCertificate{Certificate: &dummyCert{name: "ca-good"}}
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"nope"}, "", "", "", "ca-good-bad", ""))
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group"}, "", "", "", "ca-good", ""))
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 }
 
 func BenchmarkFirewallTable_match(b *testing.B) {
@@ -482,8 +485,9 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 }
 
 func TestFirewall_Drop2(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -533,15 +537,16 @@ func TestFirewall_Drop2(t *testing.T) {
 	cp := cert.NewCAPool()
 
 	// h1/c1 lacks the proper groups
-	require.ErrorIs(t, fw.Drop(p, true, &h1, cp, nil), ErrNoMatchingRule)
+	require.ErrorIs(t, fw.Drop(p, firewall.PacketContext{}, true, &h1, cp, nil), ErrNoMatchingRule)
 	// c has the proper groups
 	resetConntrack(fw)
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 }
 
 func TestFirewall_Drop3(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -613,23 +618,24 @@ func TestFirewall_Drop3(t *testing.T) {
 	cp := cert.NewCAPool()
 
 	// c1 should pass because host match
-	require.NoError(t, fw.Drop(p, true, &h1, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h1, cp, nil))
 	// c2 should pass because ca sha match
 	resetConntrack(fw)
-	require.NoError(t, fw.Drop(p, true, &h2, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h2, cp, nil))
 	// c3 should fail because no match
 	resetConntrack(fw)
-	assert.Equal(t, fw.Drop(p, true, &h3, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, true, &h3, cp, nil), ErrNoMatchingRule)
 
 	// Test a remote address match
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "", "1.2.3.4/24", "", "", ""))
-	require.NoError(t, fw.Drop(p, true, &h1, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h1, cp, nil))
 }
 
 func TestFirewall_Drop3V6(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
 
@@ -661,12 +667,13 @@ func TestFirewall_Drop3V6(t *testing.T) {
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	cp := cert.NewCAPool()
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "", "fd12::34/120", "", "", ""))
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 }
 
 func TestFirewall_DropConntrackReload(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -702,12 +709,12 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	cp := cert.NewCAPool()
 
 	// Drop outbound
-	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 	// Allow inbound
 	resetConntrack(fw)
-	require.NoError(t, fw.Drop(p, true, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, true, &h, cp, nil))
 	// Allow outbound because conntrack
-	require.NoError(t, fw.Drop(p, false, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil))
 
 	oldFw := fw
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
@@ -716,7 +723,7 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	fw.rulesVersion = oldFw.rulesVersion + 1
 
 	// Allow outbound because conntrack and new rules allow port 10
-	require.NoError(t, fw.Drop(p, false, &h, cp, nil))
+	require.NoError(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil))
 
 	oldFw = fw
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
@@ -725,12 +732,13 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	fw.rulesVersion = oldFw.rulesVersion + 1
 
 	// Drop outbound because conntrack doesn't match new ruleset
-	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 }
 
 func TestFirewall_ICMPPortBehavior(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -770,12 +778,12 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 0
 			p.RemotePort = 0
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil))
 			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil))
 		})
 
 		t.Run("nonzero ports", func(t *testing.T) {
@@ -783,12 +791,12 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 0xabcd
 			p.RemotePort = 0x1234
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil))
 			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil))
 		})
 	})
 
@@ -800,12 +808,12 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 0
 			p.RemotePort = 0
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 			//now also allow outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 		})
 
 		t.Run("nonzero ports, still blocked", func(t *testing.T) {
@@ -813,12 +821,12 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 0xabcd
 			p.RemotePort = 0x1234
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 			//now also allow outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 		})
 
 		t.Run("nonzero, matching ports, still blocked", func(t *testing.T) {
@@ -826,12 +834,12 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 80
 			p.RemotePort = 80
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil), ErrNoMatchingRule)
 			//now also allow outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 		})
 	})
 	t.Run("Any proto, any port", func(t *testing.T) {
@@ -843,12 +851,12 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 0
 			p.RemotePort = 0
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil))
 			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil))
 		})
 
 		t.Run("nonzero ports, allowed", func(t *testing.T) {
@@ -857,23 +865,24 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 			p.LocalPort = 0xabcd
 			p.RemotePort = 0x1234
 			// Drop outbound
-			assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			assert.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 			// Allow inbound
 			resetConntrack(fw)
-			require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, true, &h, cp, nil))
 			//now also allow outbound
-			require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
+			require.NoError(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil))
 			//different ID is blocked
 			p.RemotePort++
-			require.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
+			require.Equal(t, fw.Drop(*p, firewall.PacketContext{}, false, &h, cp, nil), ErrNoMatchingRule)
 		})
 	})
 
 }
 
 func TestFirewall_DropIPSpoofing(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("192.0.2.1/24"))
 
@@ -913,7 +922,7 @@ func TestFirewall_DropIPSpoofing(t *testing.T) {
 		Protocol:   firewall.ProtoUDP,
 		Fragment:   false,
 	}
-	assert.Equal(t, fw.Drop(p, true, &h1, cp, nil), ErrInvalidRemoteIP)
+	assert.Equal(t, fw.Drop(p, firewall.PacketContext{}, true, &h1, cp, nil), ErrInvalidRemoteIP)
 }
 
 func BenchmarkLookup(b *testing.B) {
@@ -1033,28 +1042,28 @@ func TestNewFirewallFromConfig(t *testing.T) {
 	l := test.NewLogger()
 	// Test a bad rule definition
 	c := &dummyCert{}
-	cs, err := newCertState(cert.Version2, nil, c, false, cert.Curve_CURVE25519, nil, "aes")
+	cs, err := newCertState(cert.Version2, nil, c, false, cert.Curve_CURVE25519, nil)
 	require.NoError(t, err)
 
-	conf := config.NewC(test.NewLogger())
+	conf := config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": "asdf"}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound failed to parse, should be an array of rules")
 
 	// Test both port and code
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "code": "2"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; only one of port or code should be provided")
 
 	// Test missing host, group, cidr, ca_name and ca_sha
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; at least one of host, group, cidr, local_cidr, ca_name, or ca_sha must be provided")
 
 	// Test code/port error
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; code was not a number; `a`")
@@ -1064,25 +1073,25 @@ func TestNewFirewallFromConfig(t *testing.T) {
 	require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`")
 
 	// Test proto error
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; proto was not understood; ``")
 
 	// Test cidr parse error
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "cidr": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; cidr did not parse; netip.ParsePrefix(\"testh\"): no '/'")
 
 	// Test local_cidr parse error
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "local_cidr": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; local_cidr did not parse; netip.ParsePrefix(\"testh\"): no '/'")
 
 	// Test both group and groups
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "group": "a", "groups": []string{"b", "c"}}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.inbound rule #0; only one of group or groups should be defined, both provided")
@@ -1091,35 +1100,35 @@ func TestNewFirewallFromConfig(t *testing.T) {
 func TestAddFirewallRulesFromConfig(t *testing.T) {
 	l := test.NewLogger()
 	// Test adding tcp rule
-	conf := config.NewC(test.NewLogger())
+	conf := config.NewC(l)
 	mf := &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "tcp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoTCP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding udp rule
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "udp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoUDP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding icmp rule
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "icmp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: firewall.PortAny, endPort: firewall.PortAny, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding icmp rule no port
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"proto": "icmp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: firewall.PortAny, endPort: firewall.PortAny, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding any rule
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
@@ -1127,14 +1136,14 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 
 	// Test adding rule with cidr
 	cidr := netip.MustParsePrefix("10.0.0.0/8")
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": cidr.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: cidr.String(), localIp: ""}, mf.lastCall)
 
 	// Test adding rule with local_cidr
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": cidr.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
@@ -1142,82 +1151,82 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 
 	// Test adding rule with cidr ipv6
 	cidr6 := netip.MustParsePrefix("fd00::/8")
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": cidr6.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: cidr6.String(), localIp: ""}, mf.lastCall)
 
 	// Test adding rule with any cidr
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": "any"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "any", localIp: ""}, mf.lastCall)
 
 	// Test adding rule with junk cidr
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": "junk/junk"}}}
 	require.EqualError(t, AddFirewallRulesFromConfig(l, true, conf, mf), "firewall.inbound rule #0; cidr did not parse; netip.ParsePrefix(\"junk/junk\"): ParseAddr(\"junk\"): unable to parse IP")
 
 	// Test adding rule with local_cidr ipv6
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": cidr6.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "", localIp: cidr6.String()}, mf.lastCall)
 
 	// Test adding rule with any local_cidr
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": "any"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, localIp: "any"}, mf.lastCall)
 
 	// Test adding rule with junk local_cidr
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": "junk/junk"}}}
 	require.EqualError(t, AddFirewallRulesFromConfig(l, true, conf, mf), "firewall.inbound rule #0; local_cidr did not parse; netip.ParsePrefix(\"junk/junk\"): ParseAddr(\"junk\"): unable to parse IP")
 
 	// Test adding rule with ca_sha
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "ca_sha": "12312313123"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "", localIp: "", caSha: "12312313123"}, mf.lastCall)
 
 	// Test adding rule with ca_name
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "ca_name": "root01"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "", localIp: "", caName: "root01"}, mf.lastCall)
 
 	// Test single group
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "group": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a"}, ip: "", localIp: ""}, mf.lastCall)
 
 	// Test single groups
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "groups": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a"}, ip: "", localIp: ""}, mf.lastCall)
 
 	// Test multiple AND groups
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "groups": []string{"a", "b"}}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a", "b"}, ip: "", localIp: ""}, mf.lastCall)
 
 	// Test Add error
-	conf = config.NewC(test.NewLogger())
+	conf = config.NewC(l)
 	mf = &mockFirewall{}
 	mf.nextCallReturn = errors.New("test error")
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "host": "a"}}}
@@ -1225,8 +1234,9 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 }
 
 func TestFirewall_convertRule(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 
 	// Ensure group array of 1 is converted and a warning is printed
 	c := map[string]any{
@@ -1234,9 +1244,7 @@ func TestFirewall_convertRule(t *testing.T) {
 	}
 
 	r, err := convertRule(l, c, "test", 1)
-	assert.Contains(t, ob.String(), "group was an array with a single value, converting to simple value")
-	assert.Contains(t, ob.String(), "table=test")
-	assert.Contains(t, ob.String(), "rule=1")
+	assert.Contains(t, ob.String(), "test rule #1; group was an array with a single value, converting to simple value")
 	require.NoError(t, err)
 	assert.Equal(t, []string{"group1"}, r.Groups)
 
@@ -1262,8 +1270,9 @@ func TestFirewall_convertRule(t *testing.T) {
 }
 
 func TestFirewall_convertRuleSanity(t *testing.T) {
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 
 	noWarningPlease := []map[string]any{
 		{"group": "group1"},
@@ -1327,7 +1336,7 @@ func (c *testcase) Test(t *testing.T, fw *Firewall) {
 	t.Helper()
 	cp := cert.NewCAPool()
 	resetConntrack(fw)
-	err := fw.Drop(c.p, true, c.h, cp, nil)
+	err := fw.Drop(c.p, firewall.PacketContext{}, true, c.h, cp, nil)
 	if c.err == nil {
 		require.NoError(t, err, "failed to not drop remote address %s", c.p.RemoteAddr)
 	} else {
@@ -1377,7 +1386,7 @@ type testsetup struct {
 	fw                 *Firewall
 }
 
-func newSetup(t *testing.T, l *slog.Logger, myPrefixes ...netip.Prefix) testsetup {
+func newSetup(t *testing.T, l *logrus.Logger, myPrefixes ...netip.Prefix) testsetup {
 	c := dummyCert{
 		name:     "me",
 		networks: myPrefixes,
@@ -1388,7 +1397,7 @@ func newSetup(t *testing.T, l *slog.Logger, myPrefixes ...netip.Prefix) testsetu
 	return newSetupFromCert(t, l, c)
 }
 
-func newSetupFromCert(t *testing.T, l *slog.Logger, c dummyCert) testsetup {
+func newSetupFromCert(t *testing.T, l *logrus.Logger, c dummyCert) testsetup {
 	myVpnNetworksTable := new(bart.Lite)
 	for _, prefix := range c.Networks() {
 		myVpnNetworksTable.Insert(prefix)
@@ -1405,8 +1414,9 @@ func newSetupFromCert(t *testing.T, l *slog.Logger, c dummyCert) testsetup {
 
 func TestFirewall_Drop_EnforceIPMatch(t *testing.T) {
 	t.Parallel()
+	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l := test.NewLoggerWithOutput(ob)
+	l.SetOutput(ob)
 
 	myPrefix := netip.MustParsePrefix("1.1.1.1/8")
 	// for now, it's okay that these are all "incoming", the logic this test tries to check doesn't care about in/out

go.mod

@@ -9,7 +9,7 @@ require (
 	github.com/armon/go-radix v1.0.0
 	github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
 	github.com/flynn/noise v1.1.0
-	github.com/gaissmai/bart v0.26.1
+	github.com/gaissmai/bart v0.26.0
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
 	github.com/kardianos/service v1.2.4
@@ -18,15 +18,15 @@ require (
 	github.com/nbrownus/go-metrics-prometheus v0.0.0-20210712211119-974a6260965f
 	github.com/prometheus/client_golang v1.23.2
 	github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475
+	github.com/sirupsen/logrus v1.9.4
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e
 	github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6
 	github.com/stretchr/testify v1.11.1
 	github.com/vishvananda/netlink v1.3.1
-	go.uber.org/goleak v1.3.0
 	go.yaml.in/yaml/v3 v3.0.4
 	golang.org/x/crypto v0.50.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.53.0
+	golang.org/x/net v0.52.0
 	golang.org/x/sync v0.20.0
 	golang.org/x/sys v0.43.0
 	golang.org/x/term v0.42.0

go.sum

@@ -26,8 +26,8 @@ github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/flynn/noise v1.1.0 h1:KjPQoQCEFdZDiP03phOvGi11+SVVhBG2wOWAorLsstg=
 github.com/flynn/noise v1.1.0/go.mod h1:xbMo+0i6+IGbYdJhF31t2eR1BIU0CYc12+BNAKwUTag=
-github.com/gaissmai/bart v0.26.1 h1:+w4rnLGNlA2GDVn382Tfe3jOsK5vOr5n4KmigJ9lbTo=
-github.com/gaissmai/bart v0.26.1/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
+github.com/gaissmai/bart v0.26.0 h1:xOZ57E9hJLBiQaSyeZa9wgWhGuzfGACgqp4BE77OkO0=
+github.com/gaissmai/bart v0.26.0/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
 github.com/go-kit/kit v0.8.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/kit v0.9.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/log v0.1.0/go.mod h1:zbhenjAZHb184qTLMA9ZjW7ThYL0H2mk7Q6pNt4vbaY=
@@ -133,6 +133,8 @@ github.com/rogpeppe/go-internal v1.10.0/go.mod h1:UQnix2H7Ngw/k4C5ijL5+65zddjncj
 github.com/sirupsen/logrus v1.2.0/go.mod h1:LxeOpSwHxABJmUn/MG1IvRgCAasNZTLOkJPxbbu5VWo=
 github.com/sirupsen/logrus v1.4.2/go.mod h1:tLMulIdttU9McNUspp0xgXVQah82FyeX6MwdIuYE2rE=
 github.com/sirupsen/logrus v1.6.0/go.mod h1:7uNnSEd1DgxDLC74fIahvMZmmYsHGZGEOFrfsX/uA88=
+github.com/sirupsen/logrus v1.9.4 h1:TsZE7l11zFCLZnZ+teH4Umoq5BhEIfIzfRDZ1Uzql2w=
+github.com/sirupsen/logrus v1.9.4/go.mod h1:ftWc9WdOfJ0a92nsE2jF5u5ZwH8Bv2zdeOC42RjbV2g=
 github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e h1:MRM5ITcdelLK2j1vwZ3Je0FKVCfqOLp5zO6trqMLYs0=
 github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e/go.mod h1:XV66xRDqSt+GTGFMVlhk3ULuV0y9ZmzeVGR4mloJI3M=
 github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6 h1:pnnLyeX7o/5aX8qUQ69P/mLojDqwda8hFOCBTmP/6hw=
@@ -182,8 +184,8 @@ golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLL
 golang.org/x/net v0.0.0-20200625001655-4c5254603344/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
 golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
-golang.org/x/net v0.53.0 h1:d+qAbo5L0orcWAr0a9JweQpjXF19LMXJE8Ey7hwOdUA=
-golang.org/x/net v0.53.0/go.mod h1:JvMuJH7rrdiCfbeHoo3fCQU24Lf5JJwT9W3sJFulfgs=
+golang.org/x/net v0.52.0 h1:He/TN1l0e4mmR3QqHMT2Xab3Aj3L9qjbhRm78/6jrW0=
+golang.org/x/net v0.52.0/go.mod h1:R1MAz7uMZxVMualyPXb+VaqGSa3LIaUqk0eEt3w36Sw=
 golang.org/x/oauth2 v0.0.0-20190226205417-e64efc72b421/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=

handshake/credential.go

@@ -1,57 +0,0 @@
-package handshake
-
-import (
-	"crypto/rand"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/cert"
-)
-
-// Credential holds everything needed to participate in a handshake
-// at a given cert version. Version and Curve are read from Cert; the public
-// half of the static keypair likewise comes from Cert.PublicKey().
-type Credential struct {
-	Cert        cert.Certificate  // the certificate
-	Bytes       []byte            // pre-marshaled certificate bytes
-	privateKey  []byte            // static private key (public half lives in Cert)
-	cipherSuite noise.CipherSuite // pre-built cipher suite (DH + cipher + hash)
-}
-
-// NewCredential creates a Credential with all material needed for handshake
-// participation. The cipherSuite should be pre-built by the caller with the
-// appropriate DH function, cipher, and hash.
-func NewCredential(
-	c cert.Certificate,
-	hsBytes []byte,
-	privateKey []byte,
-	cipherSuite noise.CipherSuite,
-) *Credential {
-	return &Credential{
-		Cert:        c,
-		Bytes:       hsBytes,
-		privateKey:  privateKey,
-		cipherSuite: cipherSuite,
-	}
-}
-
-// buildHandshakeState creates a noise.HandshakeState from this credential.
-func (hc *Credential) buildHandshakeState(initiator bool, pattern noise.HandshakePattern) (*noise.HandshakeState, error) {
-	return noise.NewHandshakeState(noise.Config{
-		CipherSuite:           hc.cipherSuite,
-		Random:                rand.Reader,
-		Pattern:               pattern,
-		Initiator:             initiator,
-		StaticKeypair:         noise.DHKey{Private: hc.privateKey, Public: hc.Cert.PublicKey()},
-		PresharedKey:          []byte{},
-		PresharedKeyPlacement: 0,
-	})
-}
-
-// GetCredentialFunc returns the handshake credential for the given version,
-// or nil if that version is not available.
-//
-// Implementations must return credentials drawn from a snapshot stable for
-// the lifetime of any single Machine. The Machine may call this multiple
-// times during a handshake (e.g. when negotiating to the peer's version)
-// and assumes the underlying static keypair is consistent across calls.
-type GetCredentialFunc func(v cert.Version) *Credential

handshake/errors.go

@@ -1,21 +0,0 @@
-package handshake
-
-import "errors"
-
-var (
-	ErrInitiateOnResponder     = errors.New("initiate called on responder")
-	ErrInitiateAlreadyCalled   = errors.New("initiate already called")
-	ErrInitiateNotCalled       = errors.New("initiate must be called before ProcessPacket for initiators")
-	ErrPacketTooShort          = errors.New("packet too short")
-	ErrPublicKeyMismatch       = errors.New("public key mismatch between certificate and handshake")
-	ErrIncompleteHandshake     = errors.New("handshake completed without receiving required content")
-	ErrMachineFailed           = errors.New("handshake machine has failed")
-	ErrUnknownSubtype          = errors.New("unknown handshake subtype")
-	ErrMissingContent          = errors.New("expected handshake content but message was empty")
-	ErrUnexpectedContent       = errors.New("received unexpected handshake content")
-	ErrIndexAllocation         = errors.New("failed to allocate local index")
-	ErrNoCredential            = errors.New("no handshake credential available for cert version")
-	ErrAsymmetricCipherKeys    = errors.New("noise produced only one cipher key")
-	ErrMultiMessageUnsupported = errors.New("multi-message handshake patterns are not yet supported by the manager")
-	ErrSubtypeMismatch         = errors.New("packet subtype does not match handshake machine subtype")
-)

handshake/handshake.proto

@@ -1,29 +0,0 @@
-// This file documents the wire format the nebula handshake speaks. It is
-// not run through protoc; the encoder/decoder in payload.go is hand-written
-// against this shape directly to keep the parser narrow and panic-free.
-//
-// Any change to the wire format must be reflected here, and adding a new
-// field requires updating MarshalPayload / unmarshalPayloadDetails together
-// with the field-uniqueness and wire-type checks in those functions.
-
-syntax = "proto3";
-package nebula.handshake;
-
-message NebulaHandshake {
-  NebulaHandshakeDetails Details = 1;
-  bytes Hmac = 2;
-}
-
-message NebulaHandshakeDetails {
-  bytes Cert = 1;
-  uint32 InitiatorIndex = 2;
-  uint32 ResponderIndex = 3;
-  // Cookie was reserved for an anti-DoS mechanism that was never
-  // implemented. No released version of nebula has ever populated it; the
-  // hand-written parser silently skips it on read.
-  uint64 Cookie = 4 [deprecated = true];
-  uint64 Time = 5;
-  uint32 CertVersion = 8;
-  // reserved for WIP multiport
-  reserved 6, 7;
-}

handshake/helpers_test.go

@@ -1,116 +0,0 @@
-package handshake
-
-import (
-	"net/netip"
-	"testing"
-	"time"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/cert"
-	ct "github.com/slackhq/nebula/cert_test"
-	"github.com/slackhq/nebula/header"
-	"github.com/stretchr/testify/require"
-)
-
-// testCertState holds cert material for a test peer.
-type testCertState struct {
-	version cert.Version
-	creds   map[cert.Version]*Credential
-}
-
-func (s *testCertState) getCredential(v cert.Version) *Credential {
-	return s.creds[v]
-}
-
-func newTestCertState(
-	t *testing.T, ca cert.Certificate, caKey []byte, name string, networks []netip.Prefix,
-) *testCertState {
-	return newTestCertStateWithCipher(t, ca, caKey, name, networks, noise.CipherChaChaPoly)
-}
-
-func newTestCertStateWithCipher(
-	t *testing.T, ca cert.Certificate, caKey []byte, name string, networks []netip.Prefix,
-	cipher noise.CipherFunc,
-) *testCertState {
-	t.Helper()
-	c, _, rawPrivKey, _ := ct.NewTestCert(
-		cert.Version2, cert.Curve_CURVE25519, ca, caKey,
-		name, ca.NotBefore(), ca.NotAfter(), networks, nil, nil,
-	)
-
-	priv, _, _, err := cert.UnmarshalPrivateKeyFromPEM(rawPrivKey)
-	require.NoError(t, err)
-
-	hsBytes, err := c.MarshalForHandshakes()
-	require.NoError(t, err)
-
-	ncs := noise.NewCipherSuite(noise.DH25519, cipher, noise.HashSHA256)
-	return &testCertState{
-		version: cert.Version2,
-		creds: map[cert.Version]*Credential{
-			cert.Version2: NewCredential(c, hsBytes, priv, ncs),
-		},
-	}
-}
-
-func testVerifier(pool *cert.CAPool) CertVerifier {
-	return func(c cert.Certificate) (*cert.CachedCertificate, error) {
-		return pool.VerifyCertificate(time.Now(), c)
-	}
-}
-
-func newTestMachine(
-	t *testing.T,
-	cs *testCertState,
-	verifier CertVerifier,
-	initiator bool,
-	localIndex uint32,
-) *Machine {
-	t.Helper()
-	m, err := NewMachine(
-		cs.version, cs.getCredential,
-		verifier, func() (uint32, error) { return localIndex, nil },
-		initiator, header.HandshakeIXPSK0,
-	)
-	require.NoError(t, err)
-	return m
-}
-
-func initiateHandshake(
-	t *testing.T,
-	initCS *testCertState, initVerifier CertVerifier,
-	respCS *testCertState, respVerifier CertVerifier,
-) (initM, respM *Machine, respResult *Result, resp []byte, err error) {
-	t.Helper()
-	initM = newTestMachine(t, initCS, initVerifier, true, 100)
-	msg1, merr := initM.Initiate(nil)
-	require.NoError(t, merr)
-
-	respM = newTestMachine(t, respCS, respVerifier, false, 200)
-	resp, respResult, err = respM.ProcessPacket(nil, msg1)
-	return
-}
-
-func doFullHandshake(
-	t *testing.T, initCS, respCS *testCertState, caPool *cert.CAPool,
-) (initResult, respResult *Result) {
-	t.Helper()
-	v := testVerifier(caPool)
-
-	initM := newTestMachine(t, initCS, v, true, 1000)
-	respM := newTestMachine(t, respCS, v, false, 2000)
-
-	msg1, err := initM.Initiate(nil)
-	require.NoError(t, err)
-
-	resp, respResult, err := respM.ProcessPacket(nil, msg1)
-	require.NoError(t, err)
-	require.NotNil(t, respResult)
-	require.NotEmpty(t, resp)
-
-	_, initResult, err = initM.ProcessPacket(nil, resp)
-	require.NoError(t, err)
-	require.NotNil(t, initResult)
-
-	return initResult, respResult
-}

handshake/machine.go

@@ -1,446 +0,0 @@
-package handshake
-
-import (
-	"bytes"
-	"fmt"
-	"slices"
-	"time"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/cert"
-	"github.com/slackhq/nebula/header"
-)
-
-// IndexAllocator is called by the Machine to allocate a local index for the
-// handshake. It is called at most once, when the first outgoing message that
-// carries a payload is built.
-//
-// Implementations MUST NOT return 0. Zero is reserved as a sentinel meaning
-// "no index assigned" on the wire and in the payload-presence checks. If an
-// allocator ever returned 0, a legitimate handshake's payload could be
-// indistinguishable from an empty one and would be rejected.
-type IndexAllocator func() (uint32, error)
-
-// CertVerifier is called by the Machine after reconstructing the peer's
-// certificate from the handshake. The verifier performs all validation
-// (CA trust, expiry, policy checks, allow lists).
-type CertVerifier func(cert.Certificate) (*cert.CachedCertificate, error)
-
-// Result contains the results of a successful handshake.
-// Returned by ProcessPacket when the handshake is complete.
-type Result struct {
-	EKey          *noise.CipherState
-	DKey          *noise.CipherState
-	Cipher        noise.CipherFunc // identifies which post-handshake CipherState the data plane should wrap EKey/DKey in
-	MyCert        cert.Certificate
-	RemoteCert    *cert.CachedCertificate
-	RemoteIndex   uint32
-	LocalIndex    uint32
-	HandshakeTime uint64
-	MessageIndex  uint64 // number of messages exchanged during the handshake
-	Initiator     bool
-}
-
-// Machine drives a Noise handshake through N messages. It handles Noise
-// protocol operations, certificate reconstruction, and payload encoding.
-// Certificate validation is delegated to the caller via CertVerifier.
-//
-// A Machine is not safe for concurrent use. The caller must ensure that
-// Initiate and ProcessPacket are not called concurrently.
-//
-// Error contract: when ProcessPacket or Initiate returns an error, callers
-// must check Failed() to decide what to do next. If Failed() is false the
-// underlying noise state was not advanced (the packet was rejected before
-// ReadMessage took effect, or the rejection is non-fatal like a stale
-// retransmit) and the Machine can accept another packet. If Failed() is
-// true the Machine is unrecoverable and the caller must abandon it.
-type Machine struct {
-	hs             *noise.HandshakeState
-	getCred        GetCredentialFunc
-	allocIndex     IndexAllocator
-	verifier       CertVerifier
-	result         *Result
-	msgs           []msgFlags
-	myVersion      cert.Version
-	subtype        header.MessageSubType
-	indexAllocated bool
-	remoteCertSet  bool
-	payloadSet     bool
-	failed         bool
-}
-
-// NewMachine creates a handshake state machine. The subtype determines both
-// the noise pattern and the per-message content layout. The credential for
-// `version` is fetched via getCred and used to seed the noise.HandshakeState.
-// IndexAllocator is called lazily when the first outgoing payload is built.
-func NewMachine(
-	version cert.Version,
-	getCred GetCredentialFunc,
-	verifier CertVerifier,
-	allocIndex IndexAllocator,
-	initiator bool,
-	subtype header.MessageSubType,
-) (*Machine, error) {
-	info, err := subtypeInfoFor(subtype)
-	if err != nil {
-		return nil, err
-	}
-
-	cred := getCred(version)
-	if cred == nil {
-		return nil, fmt.Errorf("%w: %v", ErrNoCredential, version)
-	}
-
-	hs, err := cred.buildHandshakeState(initiator, info.pattern)
-	if err != nil {
-		return nil, fmt.Errorf("build noise state: %w", err)
-	}
-
-	return &Machine{
-		hs:         hs,
-		subtype:    subtype,
-		msgs:       info.msgs,
-		getCred:    getCred,
-		allocIndex: allocIndex,
-		verifier:   verifier,
-		myVersion:  version,
-		result: &Result{
-			Initiator: initiator,
-			Cipher:    cred.cipherSuite,
-		},
-	}, nil
-}
-
-// Failed returns true if the Machine is in an unrecoverable state.
-func (m *Machine) Failed() bool {
-	return m.failed
-}
-
-// Subtype returns the handshake subtype this Machine was built for.
-func (m *Machine) Subtype() header.MessageSubType {
-	return m.subtype
-}
-
-// MessageIndex returns the noise handshake message index, which equals the
-// wire counter of the most recently sent or received message.
-func (m *Machine) MessageIndex() int {
-	return m.hs.MessageIndex()
-}
-
-// requireComplete checks that both a peer cert and payload have been received.
-// Marks the machine as failed if not.
-func (m *Machine) requireComplete() error {
-	if !m.payloadSet || !m.remoteCertSet {
-		m.failed = true
-		return ErrIncompleteHandshake
-	}
-	return nil
-}
-
-// myMsgFlags returns the flags for the current outgoing message.
-func (m *Machine) myMsgFlags() msgFlags {
-	idx := m.hs.MessageIndex()
-	if idx < len(m.msgs) {
-		return m.msgs[idx]
-	}
-	return msgFlags{}
-}
-
-// peerMsgFlags returns the flags for the message we just read.
-func (m *Machine) peerMsgFlags() msgFlags {
-	idx := m.hs.MessageIndex() - 1
-	if idx >= 0 && idx < len(m.msgs) {
-		return m.msgs[idx]
-	}
-	return msgFlags{}
-}
-
-// Initiate produces the first handshake message. Only valid for initiators,
-// and must be called exactly once before ProcessPacket.
-//
-// out is a destination buffer the message is appended to and returned. Pass
-// nil to allocate fresh, or pass a re-used buffer sliced to length 0 (e.g.
-// buf[:0]) with sufficient capacity to avoid allocation.
-//
-// An error return may not indicate a fatal condition, check Failed() to
-// determine if the Machine can still be used.
-func (m *Machine) Initiate(out []byte) ([]byte, error) {
-	if m.failed {
-		return nil, ErrMachineFailed
-	}
-	if !m.result.Initiator {
-		m.failed = true
-		return nil, ErrInitiateOnResponder
-	}
-	if m.hs.MessageIndex() != 0 {
-		m.failed = true
-		return nil, ErrInitiateAlreadyCalled
-	}
-
-	// At MessageIndex=0 with RemoteIndex still zero, buildResponse produces
-	// header counter 1 and remote index 0, which is what the initial message needs.
-	out, _, _, err := m.buildResponse(out)
-	if err != nil {
-		m.failed = true
-		return nil, err
-	}
-	return out, nil
-}
-
-// ProcessPacket handles an incoming handshake message. It advances the Noise
-// state, validates the peer certificate via the verifier, and optionally
-// produces a response.
-//
-// out is a destination buffer the response is appended to and returned. Pass
-// nil to allocate fresh, or pass a re-used buffer sliced to length 0 (e.g.
-// buf[:0]) with sufficient capacity to avoid allocation. The returned slice
-// is nil when no outgoing message is produced (handshake complete on this
-// side, or final message of a multi-message pattern).
-//
-// Returns a non-nil Result when the handshake is complete.
-// An error return may not indicate a fatal condition, check Failed() to
-// determine if the Machine can still be used.
-func (m *Machine) ProcessPacket(out, packet []byte) ([]byte, *Result, error) {
-	if m.failed {
-		return nil, nil, ErrMachineFailed
-	}
-	if len(packet) < header.Len {
-		return nil, nil, ErrPacketTooShort
-	}
-	// Reject packets whose subtype doesn't match the one this Machine was
-	// built for. A pending handshake that suddenly receives a different
-	// subtype on its index is either a stray packet that matched by chance
-	// or a peer protocol violation; drop it without failing the Machine so
-	// the legitimate retransmit can still complete.
-	if header.MessageSubType(packet[1]) != m.subtype {
-		return nil, nil, ErrSubtypeMismatch
-	}
-	if m.result.Initiator && m.hs.MessageIndex() == 0 {
-		m.failed = true
-		return nil, nil, ErrInitiateNotCalled
-	}
-
-	// The (eKey, dKey) ordering here is correct for IX, where the initiator
-	// completes the handshake by reading the responder's stage-2 message.
-	// noise returns (cs1, cs2) where cs1 is the initiator->responder cipher.
-	// For 3-message patterns where a responder finishes by reading the final
-	// message, this ordering would be wrong; revisit when XX/pqIX lands.
-	msg, eKey, dKey, err := m.hs.ReadMessage(nil, packet[header.Len:])
-	if err != nil {
-		// Noise ReadMessage failed. The noise library checkpoints and rolls back
-		// on failure, so the Machine is still alive. The caller can retry with
-		// a different packet.
-		return nil, nil, fmt.Errorf("noise ReadMessage: %w", err)
-	}
-
-	// From here on, noise state has advanced. Any error is fatal.
-	flags := m.peerMsgFlags()
-
-	if err := m.processPayload(msg, flags); err != nil {
-		return nil, nil, err
-	}
-
-	// If ReadMessage derived keys, the handshake is complete. Noise should
-	// always produce both keys together; asymmetry is a protocol invariant
-	// violation.
-	if eKey != nil || dKey != nil {
-		if eKey == nil || dKey == nil {
-			m.failed = true
-			return nil, nil, ErrAsymmetricCipherKeys
-		}
-		if err := m.requireComplete(); err != nil {
-			return nil, nil, err
-		}
-		return nil, m.completed(eKey, dKey), nil
-	}
-
-	// ReadMessage didn't complete, produce the next outgoing message
-	out, dk, ek, err := m.buildResponse(out)
-	if err != nil {
-		m.failed = true
-		return nil, nil, err
-	}
-
-	if ek != nil || dk != nil {
-		if ek == nil || dk == nil {
-			m.failed = true
-			return nil, nil, ErrAsymmetricCipherKeys
-		}
-		if err := m.requireComplete(); err != nil {
-			return nil, nil, err
-		}
-		return out, m.completed(ek, dk), nil
-	}
-
-	return out, nil, nil
-}
-
-func (m *Machine) completed(eKey, dKey *noise.CipherState) *Result {
-	m.result.EKey = eKey
-	m.result.DKey = dKey
-	m.result.MessageIndex = uint64(m.hs.MessageIndex())
-	return m.result
-}
-
-func (m *Machine) processPayload(msg []byte, flags msgFlags) error {
-	if len(msg) == 0 {
-		if flags.expectsPayload || flags.expectsCert {
-			m.failed = true
-			return ErrMissingContent
-		}
-		return nil
-	}
-
-	payload, err := UnmarshalPayload(msg)
-	if err != nil {
-		m.failed = true
-		return fmt.Errorf("unmarshal handshake: %w", err)
-	}
-
-	// Assert the payload contains exactly what we expect
-	hasPayloadData := payload.InitiatorIndex != 0 || payload.ResponderIndex != 0 || payload.Time != 0
-	if hasPayloadData != flags.expectsPayload {
-		m.failed = true
-		return ErrUnexpectedContent
-	}
-
-	hasCertData := len(payload.Cert) > 0
-	if hasCertData != flags.expectsCert {
-		m.failed = true
-		return ErrUnexpectedContent
-	}
-
-	// Process payload
-	if flags.expectsPayload {
-		if m.result.Initiator {
-			m.result.RemoteIndex = payload.ResponderIndex
-		} else {
-			m.result.RemoteIndex = payload.InitiatorIndex
-		}
-		m.result.HandshakeTime = payload.Time
-		m.payloadSet = true
-	}
-
-	// Process certificate
-	if flags.expectsCert {
-		if err := m.validateCert(payload); err != nil {
-			return err
-		}
-	}
-
-	return nil
-}
-
-func (m *Machine) validateCert(payload Payload) error {
-	cred := m.getCred(m.myVersion)
-	if cred == nil {
-		m.failed = true
-		return fmt.Errorf("%w: %v", ErrNoCredential, m.myVersion)
-	}
-	rc, err := cert.Recombine(
-		cert.Version(payload.CertVersion),
-		payload.Cert,
-		m.hs.PeerStatic(),
-		cred.Cert.Curve(),
-	)
-	if err != nil {
-		m.failed = true
-		return fmt.Errorf("recombine cert: %w", err)
-	}
-
-	if !bytes.Equal(rc.PublicKey(), m.hs.PeerStatic()) {
-		m.failed = true
-		return ErrPublicKeyMismatch
-	}
-
-	// Version negotiation, if the peer sent a different version and we have it, switch
-	if rc.Version() != m.myVersion {
-		if m.getCred(rc.Version()) != nil {
-			m.myVersion = rc.Version()
-		}
-	}
-
-	verified, err := m.verifier(rc)
-	if err != nil {
-		m.failed = true
-		return fmt.Errorf("verify cert: %w", err)
-	}
-
-	m.result.RemoteCert = verified
-	m.remoteCertSet = true
-	return nil
-}
-
-func (m *Machine) marshalOutgoing(flags msgFlags) ([]byte, error) {
-	if !flags.expectsPayload && !flags.expectsCert {
-		return nil, nil
-	}
-
-	var p Payload
-	if flags.expectsPayload {
-		if !m.indexAllocated {
-			index, err := m.allocIndex()
-			if err != nil {
-				return nil, fmt.Errorf("%w: %w", ErrIndexAllocation, err)
-			}
-			m.result.LocalIndex = index
-			m.indexAllocated = true
-		}
-
-		if m.result.Initiator {
-			p.InitiatorIndex = m.result.LocalIndex
-		} else {
-			p.ResponderIndex = m.result.LocalIndex
-			p.InitiatorIndex = m.result.RemoteIndex
-		}
-		p.Time = uint64(time.Now().UnixNano())
-	}
-	if flags.expectsCert {
-		cred := m.getCred(m.myVersion)
-		if cred == nil {
-			return nil, fmt.Errorf("%w: %v", ErrNoCredential, m.myVersion)
-		}
-		p.Cert = cred.Bytes
-		p.CertVersion = uint32(cred.Cert.Version())
-		m.result.MyCert = cred.Cert
-	}
-
-	return MarshalPayload(nil, p), nil
-}
-
-func (m *Machine) buildResponse(out []byte) ([]byte, *noise.CipherState, *noise.CipherState, error) {
-	flags := m.myMsgFlags()
-	hsBytes, err := m.marshalOutgoing(flags)
-	if err != nil {
-		return nil, nil, nil, err
-	}
-
-	// Extend out by header.Len to make room for the header. slices.Grow is a
-	// no-op when the cap is already sufficient (the zero-copy case where the
-	// caller passed a pre-sized buffer). header.Encode overwrites the new
-	// bytes, so they don't need to be zeroed.
-	start := len(out)
-	out = slices.Grow(out, header.Len)[:start+header.Len]
-	header.Encode(
-		out[start:],
-		header.Version, header.Handshake, m.subtype,
-		m.result.RemoteIndex,
-		uint64(m.hs.MessageIndex()+1),
-	)
-
-	// noise.WriteMessage appends the encrypted handshake message to out,
-	// reusing capacity when present.
-	//
-	// The (dKey, eKey) ordering here is correct for IX, where the responder
-	// completes the handshake by writing the stage-2 message. noise returns
-	// (cs1, cs2) where cs1 is the initiator->responder cipher (which is the
-	// responder's decrypt key). For 3-message patterns where an initiator
-	// finishes by writing the final message, this ordering would be wrong;
-	// revisit when XX/pqIX lands.
-	out, dKey, eKey, err := m.hs.WriteMessage(out, hsBytes)
-	if err != nil {
-		return nil, nil, nil, fmt.Errorf("noise WriteMessage: %w", err)
-	}
-
-	return out, dKey, eKey, nil
-}

handshake/machine_test.go

@@ -1,662 +0,0 @@
-package handshake
-
-import (
-	"net/netip"
-	"testing"
-	"time"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/cert"
-	ct "github.com/slackhq/nebula/cert_test"
-	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/noiseutil"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestMachineIXHappyPath(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-
-	initCS := newTestCertState(t, ca, caKey, "initiator", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	respCS := newTestCertState(t, ca, caKey, "responder", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-
-	initR, respR := doFullHandshake(t, initCS, respCS, caPool)
-
-	assert.Equal(t, "responder", initR.RemoteCert.Certificate.Name())
-	assert.Equal(t, "initiator", respR.RemoteCert.Certificate.Name())
-
-	assert.Equal(t, uint32(1000), initR.LocalIndex)
-	assert.Equal(t, uint32(2000), initR.RemoteIndex)
-	assert.Equal(t, uint32(2000), respR.LocalIndex)
-	assert.Equal(t, uint32(1000), respR.RemoteIndex)
-
-	assert.Equal(t, uint64(2), initR.MessageIndex, "IX has 2 messages")
-	assert.Equal(t, uint64(2), respR.MessageIndex, "IX has 2 messages")
-
-	ct1, err := initR.EKey.Encrypt(nil, nil, []byte("hello"))
-	require.NoError(t, err)
-	pt1, err := respR.DKey.Decrypt(nil, nil, ct1)
-	require.NoError(t, err)
-	assert.Equal(t, []byte("hello"), pt1)
-
-	ct2, err := respR.EKey.Encrypt(nil, nil, []byte("world"))
-	require.NoError(t, err)
-	pt2, err := initR.DKey.Decrypt(nil, nil, ct2)
-	require.NoError(t, err)
-	assert.Equal(t, []byte("world"), pt2)
-}
-
-func TestMachineInitiateErrors(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	cs := newTestCertState(t, ca, caKey, "test", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	v := testVerifier(caPool)
-
-	t.Run("initiate on responder", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		_, err := m.Initiate(nil)
-		require.ErrorIs(t, err, ErrInitiateOnResponder)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("initiate called twice", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, true, 100)
-		_, err := m.Initiate(nil)
-		require.NoError(t, err)
-		_, err = m.Initiate(nil)
-		require.ErrorIs(t, err, ErrInitiateAlreadyCalled)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("process packet before initiate on initiator", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, true, 100)
-		_, _, err := m.ProcessPacket(nil, make([]byte, 100))
-		require.ErrorIs(t, err, ErrInitiateNotCalled)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("calling failed machine", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		_, err := m.Initiate(nil) // fails: responder
-		require.Error(t, err)
-		_, err = m.Initiate(nil) // fails: already failed
-		require.ErrorIs(t, err, ErrMachineFailed)
-	})
-}
-
-func TestMachineProcessPacketErrors(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	cs := newTestCertState(t, ca, caKey, "test", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	v := testVerifier(caPool)
-
-	t.Run("packet too short", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		_, _, err := m.ProcessPacket(nil, []byte{1, 2, 3})
-		require.ErrorIs(t, err, ErrPacketTooShort)
-		assert.False(t, m.Failed(), "short packet should not kill machine")
-	})
-
-	t.Run("noise decryption failure is recoverable", func(t *testing.T) {
-		initCS := newTestCertState(t, ca, caKey, "init", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-		initM := newTestMachine(t, initCS, v, true, 100)
-		msg1, err := initM.Initiate(nil)
-		require.NoError(t, err)
-
-		respM := newTestMachine(t, cs, v, false, 200)
-		resp, _, err := respM.ProcessPacket(nil, msg1)
-		require.NoError(t, err)
-
-		corrupted := make([]byte, len(resp))
-		copy(corrupted, resp)
-		for i := header.Len; i < len(corrupted); i++ {
-			corrupted[i] ^= 0xff
-		}
-		_, _, err = initM.ProcessPacket(nil, corrupted)
-		require.Error(t, err)
-		assert.False(t, initM.Failed(), "noise failure should be recoverable")
-
-		// And the machine should still complete a real handshake afterward.
-		_, result, err := initM.ProcessPacket(nil, resp)
-		require.NoError(t, err)
-		require.NotNil(t, result, "initiator should complete on the legitimate response")
-	})
-
-	t.Run("invalid cert is fatal", func(t *testing.T) {
-		otherCA, _, otherCAKey, _ := ct.NewTestCaCert(
-			cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-		)
-		otherCS := newTestCertState(t, otherCA, otherCAKey, "other", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-
-		initM := newTestMachine(t, otherCS, testVerifier(ct.NewTestCAPool(otherCA)), true, 100)
-		msg1, err := initM.Initiate(nil)
-		require.NoError(t, err)
-
-		respM := newTestMachine(t, cs, v, false, 200)
-		_, _, err = respM.ProcessPacket(nil, msg1)
-		require.Error(t, err)
-		assert.True(t, respM.Failed(), "cert validation failure should kill machine")
-	})
-
-	t.Run("subtype mismatch is recoverable", func(t *testing.T) {
-		initCS := newTestCertState(t, ca, caKey, "init", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-		initM := newTestMachine(t, initCS, v, true, 100)
-		msg1, err := initM.Initiate(nil)
-		require.NoError(t, err)
-
-		// Mutate the subtype byte (offset 1 in the header) to a value the
-		// responder Machine wasn't built for.
-		bad := make([]byte, len(msg1))
-		copy(bad, msg1)
-		bad[1] = 0xff
-
-		respM := newTestMachine(t, cs, v, false, 200)
-		_, _, err = respM.ProcessPacket(nil, bad)
-		require.ErrorIs(t, err, ErrSubtypeMismatch)
-		assert.False(t, respM.Failed(), "subtype mismatch should not kill the machine")
-
-		// And the machine should still complete a real handshake afterward.
-		resp, result, err := respM.ProcessPacket(nil, msg1)
-		require.NoError(t, err)
-		require.NotNil(t, result, "responder should complete on the legitimate stage-1 packet")
-		assert.NotEmpty(t, resp, "responder should produce a stage-2 reply")
-	})
-}
-
-// TestMachineProcessPayload exercises processPayload's internal validation
-// directly. Most of these failure modes can't be reached black-box once the
-// subtype check at the top of ProcessPacket gates external callers, so we
-// drive them by hand here for coverage.
-func TestMachineProcessPayload(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	cs := newTestCertState(t, ca, caKey, "test", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	v := testVerifier(caPool)
-
-	t.Run("empty message with expects fails", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		err := m.processPayload(nil, msgFlags{expectsPayload: true, expectsCert: true})
-		require.ErrorIs(t, err, ErrMissingContent)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("empty message with no expects passes", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		err := m.processPayload(nil, msgFlags{})
-		require.NoError(t, err)
-		assert.False(t, m.Failed())
-	})
-
-	t.Run("malformed protobuf is fatal", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		err := m.processPayload([]byte{0xff, 0xff, 0xff}, msgFlags{expectsPayload: true, expectsCert: true})
-		require.Error(t, err)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("unexpected payload data is fatal", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		// A payload with index data when none was expected.
-		bytes := MarshalPayload(nil, Payload{InitiatorIndex: 42, Time: 1})
-		err := m.processPayload(bytes, msgFlags{expectsPayload: false, expectsCert: false})
-		require.ErrorIs(t, err, ErrUnexpectedContent)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("unexpected cert data is fatal", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		// A payload with cert when none was expected.
-		bytes := MarshalPayload(nil, Payload{Cert: []byte{1, 2, 3}, CertVersion: 2})
-		err := m.processPayload(bytes, msgFlags{expectsPayload: false, expectsCert: false})
-		require.ErrorIs(t, err, ErrUnexpectedContent)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("missing payload data when expected is fatal", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		// Cert present, but no index/time fields.
-		bytes := MarshalPayload(nil, Payload{Cert: []byte{1, 2, 3}, CertVersion: 2})
-		err := m.processPayload(bytes, msgFlags{expectsPayload: true, expectsCert: true})
-		require.ErrorIs(t, err, ErrUnexpectedContent)
-		assert.True(t, m.Failed())
-	})
-}
-
-// TestMachineRequireComplete checks the fail-on-incomplete-handshake path
-// directly. Like processPayload above this isn't reachable from a normal IX
-// flow, so we drive it by hand.
-func TestMachineRequireComplete(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	cs := newTestCertState(t, ca, caKey, "test", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	v := testVerifier(caPool)
-
-	t.Run("missing both fails", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		err := m.requireComplete()
-		require.ErrorIs(t, err, ErrIncompleteHandshake)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("payload only fails", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		m.payloadSet = true
-		err := m.requireComplete()
-		require.ErrorIs(t, err, ErrIncompleteHandshake)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("cert only fails", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		m.remoteCertSet = true
-		err := m.requireComplete()
-		require.ErrorIs(t, err, ErrIncompleteHandshake)
-		assert.True(t, m.Failed())
-	})
-
-	t.Run("both set passes", func(t *testing.T) {
-		m := newTestMachine(t, cs, v, false, 100)
-		m.payloadSet = true
-		m.remoteCertSet = true
-		err := m.requireComplete()
-		require.NoError(t, err)
-		assert.False(t, m.Failed())
-	})
-}
-
-func TestMachineAESCipher(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-
-	initCS := newTestCertStateWithCipher(
-		t, ca, caKey, "init",
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")},
-		noiseutil.CipherAESGCM,
-	)
-	respCS := newTestCertStateWithCipher(
-		t, ca, caKey, "resp",
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")},
-		noiseutil.CipherAESGCM,
-	)
-
-	initR, respR := doFullHandshake(t, initCS, respCS, caPool)
-
-	ct1, err := initR.EKey.Encrypt(nil, nil, []byte("works"))
-	require.NoError(t, err)
-	pt1, err := respR.DKey.Decrypt(nil, nil, ct1)
-	require.NoError(t, err)
-	assert.Equal(t, []byte("works"), pt1)
-
-	ct2, err := respR.EKey.Encrypt(nil, nil, []byte("back"))
-	require.NoError(t, err)
-	pt2, err := initR.DKey.Decrypt(nil, nil, ct2)
-	require.NoError(t, err)
-	assert.Equal(t, []byte("back"), pt2)
-}
-
-func TestResultFields(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	initCS := newTestCertState(t, ca, caKey, "init", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	respCS := newTestCertState(t, ca, caKey, "resp", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-
-	initR, respR := doFullHandshake(t, initCS, respCS, caPool)
-
-	assert.True(t, initR.Initiator)
-	assert.False(t, respR.Initiator)
-	assert.NotZero(t, initR.HandshakeTime)
-	assert.NotZero(t, respR.HandshakeTime)
-	assert.NotNil(t, initR.RemoteCert)
-	assert.NotNil(t, respR.RemoteCert)
-}
-
-func TestMachineBufferReuse(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	initCS := newTestCertState(t, ca, caKey, "init", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	respCS := newTestCertState(t, ca, caKey, "resp", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-	v := testVerifier(caPool)
-
-	initM := newTestMachine(t, initCS, v, true, 1000)
-	respM := newTestMachine(t, respCS, v, false, 2000)
-
-	msg1, err := initM.Initiate(nil)
-	require.NoError(t, err)
-
-	t.Run("response writes into provided buffer", func(t *testing.T) {
-		buf := make([]byte, 0, 4096)
-		resp, result, err := respM.ProcessPacket(buf, msg1)
-		require.NoError(t, err)
-		require.NotNil(t, result)
-
-		assert.NotEmpty(t, resp, "response should have content")
-		assert.Equal(t, &buf[:1][0], &resp[:1][0],
-			"response should reuse the provided buffer's backing array")
-	})
-
-	t.Run("initiate writes into provided buffer", func(t *testing.T) {
-		initM2 := newTestMachine(t, initCS, v, true, 3000)
-		buf := make([]byte, 0, 4096)
-		msg, err := initM2.Initiate(buf)
-		require.NoError(t, err)
-
-		assert.NotEmpty(t, msg, "initiate should have content")
-		assert.Equal(t, &buf[:1][0], &msg[:1][0],
-			"initiate should reuse the provided buffer's backing array")
-	})
-
-	t.Run("nil out still works", func(t *testing.T) {
-		initM2 := newTestMachine(t, initCS, v, true, 4000)
-		respM2 := newTestMachine(t, respCS, v, false, 5000)
-
-		msg1, err := initM2.Initiate(nil)
-		require.NoError(t, err)
-
-		resp, _, err := respM2.ProcessPacket(nil, msg1)
-		require.NoError(t, err)
-
-		out, result, err := initM2.ProcessPacket(nil, resp)
-		require.NoError(t, err)
-		assert.NotNil(t, result)
-		assert.Nil(t, out, "initiator should have no response for IX msg2")
-	})
-}
-
-func TestMachineMsgIndexTracking(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	initCS := newTestCertState(t, ca, caKey, "init", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	respCS := newTestCertState(t, ca, caKey, "resp", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-	v := testVerifier(caPool)
-
-	initM := newTestMachine(t, initCS, v, true, 100)
-	respM := newTestMachine(t, respCS, v, false, 200)
-
-	msg1, err := initM.Initiate(nil)
-	require.NoError(t, err)
-
-	resp1, result1, err := respM.ProcessPacket(nil, msg1)
-	require.NoError(t, err)
-	assert.NotNil(t, result1)
-
-	_, result2, err := initM.ProcessPacket(nil, resp1)
-	require.NoError(t, err)
-	assert.NotNil(t, result2)
-}
-
-func TestMachineThreeMessagePattern(t *testing.T) {
-	registerTestXXInfo(t)
-
-	// Use HandshakeXX (3 messages) to verify the Machine handles multi-message
-	// patterns correctly. XX flow:
-	//   msg1 (I->R): [E]           - payload only, no cert
-	//   msg2 (R->I): [E, ee, S, es] - payload + cert
-	//   msg3 (I->R): [S, se]       - cert only (no payload, not first two)
-
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-	v := testVerifier(caPool)
-
-	initCS := newTestCertState(t, ca, caKey, "init", []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")})
-	respCS := newTestCertState(t, ca, caKey, "resp", []netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")})
-
-	initM, err := NewMachine(
-		cert.Version2,
-		initCS.getCredential, v,
-		func() (uint32, error) { return 1000, nil },
-		true, header.HandshakeXXPSK0,
-	)
-	require.NoError(t, err)
-
-	respM, err := NewMachine(
-		cert.Version2,
-		respCS.getCredential, v,
-		func() (uint32, error) { return 2000, nil },
-		false, header.HandshakeXXPSK0,
-	)
-	require.NoError(t, err)
-
-	// msg1: initiator -> responder (E only, no cert)
-	msg1, err := initM.Initiate(nil)
-	require.NoError(t, err)
-	assert.NotEmpty(t, msg1)
-
-	// Responder processes msg1, should not complete yet, should produce msg2
-	msg2, result, err := respM.ProcessPacket(nil, msg1)
-	require.NoError(t, err)
-	assert.Nil(t, result, "XX should not complete on msg1")
-	assert.NotEmpty(t, msg2, "responder should produce msg2")
-
-	// Initiator processes msg2: gets responder's cert, produces msg3, and
-	// completes (WriteMessage for msg3 derives keys)
-	msg3, initResult, err := initM.ProcessPacket(nil, msg2)
-	require.NoError(t, err)
-	require.NotNil(t, initResult, "XX initiator should complete after reading msg2 and writing msg3")
-	assert.NotEmpty(t, msg3, "initiator should produce msg3")
-	assert.Equal(t, "resp", initResult.RemoteCert.Certificate.Name())
-
-	// Responder processes msg3: gets initiator's cert and completes
-	_, respResult, err := respM.ProcessPacket(nil, msg3)
-	require.NoError(t, err)
-	require.NotNil(t, respResult, "XX responder should complete on msg3")
-	assert.Equal(t, "init", respResult.RemoteCert.Certificate.Name())
-
-	assert.Equal(t, uint64(3), initResult.MessageIndex, "XX has 3 messages")
-	assert.Equal(t, uint64(3), respResult.MessageIndex, "XX has 3 messages")
-
-	// Verify keys work
-	ct1, err := initResult.EKey.Encrypt(nil, nil, []byte("three messages"))
-	require.NoError(t, err)
-	pt1, err := respResult.DKey.Decrypt(nil, nil, ct1)
-	require.NoError(t, err)
-	assert.Equal(t, []byte("three messages"), pt1)
-}
-
-// NOTE: ErrIncompleteHandshake is tested implicitly. It can't be triggered with
-// IX since the cert is always in the payload. A 3-message pattern test (HybridIX)
-// should exercise the case where cert arrives in msg3 and verify that completing
-// without it fails.
-
-func TestMachineExpiredCert(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519,
-		time.Now().Add(-24*time.Hour), time.Now().Add(24*time.Hour),
-		nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-
-	expCert, _, expKeyPEM, _ := ct.NewTestCert(
-		cert.Version2, cert.Curve_CURVE25519, ca, caKey,
-		"expired", time.Now().Add(-2*time.Hour), time.Now().Add(-1*time.Hour),
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")}, nil, nil,
-	)
-	expKey, _, _, err := cert.UnmarshalPrivateKeyFromPEM(expKeyPEM)
-	require.NoError(t, err)
-	expHsBytes, err := expCert.MarshalForHandshakes()
-	require.NoError(t, err)
-	ncs := noise.NewCipherSuite(noise.DH25519, noise.CipherChaChaPoly, noise.HashSHA256)
-
-	expiredCS := &testCertState{
-		version: cert.Version2,
-		creds: map[cert.Version]*Credential{
-			cert.Version2: NewCredential(expCert, expHsBytes, expKey, ncs),
-		},
-	}
-
-	respCS := newTestCertState(
-		t, ca, caKey, "responder",
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")},
-	)
-
-	_, respM, _, _, err := initiateHandshake(
-		t, expiredCS, testVerifier(caPool),
-		respCS, testVerifier(caPool),
-	)
-	require.ErrorContains(t, err, "verify cert")
-	assert.True(t, respM.Failed())
-}
-
-func TestMachineNoCertNetworks(t *testing.T) {
-	ca, _, caKey, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca)
-
-	caHsBytes, err := ca.MarshalForHandshakes()
-	require.NoError(t, err)
-	ncs := noise.NewCipherSuite(noise.DH25519, noise.CipherChaChaPoly, noise.HashSHA256)
-
-	noNetCS := &testCertState{
-		version: cert.Version2,
-		creds: map[cert.Version]*Credential{
-			cert.Version2: NewCredential(ca, caHsBytes, caKey, ncs),
-		},
-	}
-
-	respCS := newTestCertState(
-		t, ca, caKey, "responder",
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")},
-	)
-
-	_, respM, _, _, err := initiateHandshake(
-		t, noNetCS, testVerifier(caPool),
-		respCS, testVerifier(caPool),
-	)
-	require.Error(t, err)
-	assert.True(t, respM.Failed())
-}
-
-func TestMachineDifferentCAs(t *testing.T) {
-	ca1, _, caKey1, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	ca2, _, caKey2, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-
-	initCS := newTestCertState(
-		t, ca1, caKey1, "init",
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")},
-	)
-	respCS := newTestCertState(
-		t, ca2, caKey2, "resp",
-		[]netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")},
-	)
-
-	_, respM, _, _, err := initiateHandshake(
-		t, initCS, testVerifier(ct.NewTestCAPool(ca1)),
-		respCS, testVerifier(ct.NewTestCAPool(ca2)),
-	)
-	require.ErrorContains(t, err, "verify cert")
-	assert.True(t, respM.Failed())
-}
-
-func TestMachineVersionNegotiation(t *testing.T) {
-	ca1, _, caKey1, _ := ct.NewTestCaCert(
-		cert.Version1, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	ca2, _, caKey2, _ := ct.NewTestCaCert(
-		cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil,
-	)
-	caPool := ct.NewTestCAPool(ca1, ca2)
-
-	makeMultiVersionResp := func(t *testing.T) *testCertState {
-		t.Helper()
-		respCertV1, _, respKeyPEM, _ := ct.NewTestCert(
-			cert.Version1, cert.Curve_CURVE25519, ca1, caKey1, "resp",
-			ca1.NotBefore(), ca1.NotAfter(),
-			[]netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")}, nil, nil,
-		)
-		respKey, _, _, _ := cert.UnmarshalPrivateKeyFromPEM(respKeyPEM)
-		respCertV2, _ := ct.NewTestCertDifferentVersion(respCertV1, cert.Version2, ca2, caKey2)
-		respHsV1, _ := respCertV1.MarshalForHandshakes()
-		respHsV2, _ := respCertV2.MarshalForHandshakes()
-		ncs := noise.NewCipherSuite(noise.DH25519, noise.CipherChaChaPoly, noise.HashSHA256)
-		return &testCertState{
-			version: cert.Version1,
-			creds: map[cert.Version]*Credential{
-				cert.Version1: NewCredential(respCertV1, respHsV1, respKey, ncs),
-				cert.Version2: NewCredential(respCertV2, respHsV2, respKey, ncs),
-			},
-		}
-	}
-
-	t.Run("responder matches initiator version", func(t *testing.T) {
-		initCS := newTestCertState(
-			t, ca2, caKey2, "init",
-			[]netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")},
-		)
-		respCS := makeMultiVersionResp(t)
-		v := testVerifier(caPool)
-
-		initM, _, respResult, resp, err := initiateHandshake(
-			t, initCS, v,
-			respCS, v,
-		)
-		require.NoError(t, err)
-		require.NotNil(t, respResult)
-
-		assert.Equal(t, cert.Version2, respResult.MyCert.Version(),
-			"responder should negotiate to initiator's version")
-
-		_, initResult, err := initM.ProcessPacket(nil, resp)
-		require.NoError(t, err)
-		require.NotNil(t, initResult)
-		assert.Equal(t, cert.Version2, initResult.RemoteCert.Certificate.Version(),
-			"initiator should see V2 cert from responder")
-	})
-
-	t.Run("responder keeps version when no match available", func(t *testing.T) {
-		initCS := newTestCertState(
-			t, ca2, caKey2, "init",
-			[]netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")},
-		)
-
-		respCert, _, respKeyPEM, _ := ct.NewTestCert(
-			cert.Version1, cert.Curve_CURVE25519, ca1, caKey1, "resp",
-			ca1.NotBefore(), ca1.NotAfter(),
-			[]netip.Prefix{netip.MustParsePrefix("10.0.0.2/24")}, nil, nil,
-		)
-		respKey, _, _, _ := cert.UnmarshalPrivateKeyFromPEM(respKeyPEM)
-		respHs, _ := respCert.MarshalForHandshakes()
-		ncs := noise.NewCipherSuite(noise.DH25519, noise.CipherChaChaPoly, noise.HashSHA256)
-		respCS := &testCertState{
-			version: cert.Version1,
-			creds: map[cert.Version]*Credential{
-				cert.Version1: NewCredential(respCert, respHs, respKey, ncs),
-			},
-		}
-
-		v := testVerifier(caPool)
-		_, _, respResult, _, err := initiateHandshake(
-			t, initCS, v,
-			respCS, v,
-		)
-		require.NoError(t, err)
-		require.NotNil(t, respResult)
-
-		assert.Equal(t, cert.Version1, respResult.MyCert.Version(),
-			"responder should keep V1 when V2 not available")
-	})
-}

handshake/patterns.go

@@ -1,54 +0,0 @@
-package handshake
-
-import (
-	"fmt"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/header"
-)
-
-// msgFlags tracks what application data a handshake message carries.
-type msgFlags struct {
-	expectsPayload bool // message carries indexes and time
-	expectsCert    bool // message carries the certificate
-}
-
-// subtypeInfo bundles the noise pattern with the per-message flags for a
-// given handshake subtype.
-type subtypeInfo struct {
-	pattern noise.HandshakePattern
-	msgs    []msgFlags
-}
-
-// subtypeInfos defines the noise pattern and message content layout for each
-// handshake subtype.
-var subtypeInfos = map[header.MessageSubType]subtypeInfo{
-	// IX: 2 messages, both carry payload and cert
-	header.HandshakeIXPSK0: {
-		pattern: noise.HandshakeIX,
-		msgs: []msgFlags{
-			{expectsPayload: true, expectsCert: true},
-			{expectsPayload: true, expectsCert: true},
-		},
-	},
-
-	// XX: 3 messages
-	//   msg1 (I->R): payload only
-	//   msg2 (R->I): payload + cert
-	//   msg3 (I->R): cert only
-	//header.HandshakeXXPSK0: {
-	//	pattern: noise.HandshakeXX,
-	//	msgs: []msgFlags{
-	//		{expectsPayload: true, expectsCert: false},
-	//		{expectsPayload: true, expectsCert: true},
-	//		{expectsPayload: false, expectsCert: true},
-	//	},
-	//},
-}
-
-func subtypeInfoFor(subtype header.MessageSubType) (subtypeInfo, error) {
-	if info, ok := subtypeInfos[subtype]; ok {
-		return info, nil
-	}
-	return subtypeInfo{}, fmt.Errorf("%w: %d", ErrUnknownSubtype, subtype)
-}

handshake/patterns_test.go

@@ -1,63 +0,0 @@
-package handshake
-
-import (
-	"testing"
-
-	"github.com/flynn/noise"
-	"github.com/slackhq/nebula/header"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestSubtypeInfo(t *testing.T) {
-	t.Run("IX", func(t *testing.T) {
-		info, err := subtypeInfoFor(header.HandshakeIXPSK0)
-		require.NoError(t, err)
-		assert.Equal(t, noise.HandshakeIX.Name, info.pattern.Name)
-		require.Len(t, info.msgs, 2)
-		// msg1: payload + cert
-		assert.True(t, info.msgs[0].expectsPayload)
-		assert.True(t, info.msgs[0].expectsCert)
-		// msg2: payload + cert
-		assert.True(t, info.msgs[1].expectsPayload)
-		assert.True(t, info.msgs[1].expectsCert)
-	})
-
-	t.Run("XX", func(t *testing.T) {
-		registerTestXXInfo(t)
-		info, err := subtypeInfoFor(header.HandshakeXXPSK0)
-		require.NoError(t, err)
-		assert.Equal(t, noise.HandshakeXX.Name, info.pattern.Name)
-		require.Len(t, info.msgs, 3)
-		// msg1: payload only
-		assert.True(t, info.msgs[0].expectsPayload)
-		assert.False(t, info.msgs[0].expectsCert)
-		// msg2: payload + cert
-		assert.True(t, info.msgs[1].expectsPayload)
-		assert.True(t, info.msgs[1].expectsCert)
-		// msg3: cert only
-		assert.False(t, info.msgs[2].expectsPayload)
-		assert.True(t, info.msgs[2].expectsCert)
-	})
-
-	t.Run("unknown subtype returns error", func(t *testing.T) {
-		_, err := subtypeInfoFor(99)
-		require.ErrorIs(t, err, ErrUnknownSubtype)
-	})
-}
-
-// registerTestXXInfo temporarily registers XX subtype info for testing.
-func registerTestXXInfo(t *testing.T) {
-	t.Helper()
-	subtypeInfos[header.HandshakeXXPSK0] = subtypeInfo{
-		pattern: noise.HandshakeXX,
-		msgs: []msgFlags{
-			{expectsPayload: true, expectsCert: false},
-			{expectsPayload: true, expectsCert: true},
-			{expectsPayload: false, expectsCert: true},
-		},
-	}
-	t.Cleanup(func() {
-		delete(subtypeInfos, header.HandshakeXXPSK0)
-	})
-}

handshake/payload.go

@@ -1,173 +0,0 @@
-package handshake
-
-import (
-	"errors"
-	"math"
-
-	"google.golang.org/protobuf/encoding/protowire"
-)
-
-var (
-	errInvalidHandshakeMessage = errors.New("invalid handshake message")
-	errInvalidHandshakeDetails = errors.New("invalid handshake details")
-)
-
-// Payload represents the decoded fields of a handshake message.
-// Wire format is protobuf-compatible with NebulaHandshake{Details: NebulaHandshakeDetails{...}}.
-type Payload struct {
-	Cert           []byte
-	InitiatorIndex uint32
-	ResponderIndex uint32
-	Time           uint64
-	CertVersion    uint32
-}
-
-// Proto field numbers for NebulaHandshakeDetails
-const (
-	fieldCert           = 1 // bytes
-	fieldInitiatorIndex = 2 // uint32
-	fieldResponderIndex = 3 // uint32
-	fieldTime           = 5 // uint64
-	fieldCertVersion    = 8 // uint32
-)
-
-// MarshalPayload encodes a handshake payload in protobuf wire format compatible
-// with NebulaHandshake{Details: NebulaHandshakeDetails{...}}.
-// Returns out (which may be nil), with the marshalled Payload appended to it.
-func MarshalPayload(out []byte, p Payload) []byte {
-	var details []byte
-
-	if len(p.Cert) > 0 {
-		details = protowire.AppendTag(details, fieldCert, protowire.BytesType)
-		details = protowire.AppendBytes(details, p.Cert)
-	}
-	if p.InitiatorIndex != 0 {
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, uint64(p.InitiatorIndex))
-	}
-	if p.ResponderIndex != 0 {
-		details = protowire.AppendTag(details, fieldResponderIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, uint64(p.ResponderIndex))
-	}
-	if p.Time != 0 {
-		details = protowire.AppendTag(details, fieldTime, protowire.VarintType)
-		details = protowire.AppendVarint(details, p.Time)
-	}
-	if p.CertVersion != 0 {
-		details = protowire.AppendTag(details, fieldCertVersion, protowire.VarintType)
-		details = protowire.AppendVarint(details, uint64(p.CertVersion))
-	}
-
-	out = protowire.AppendTag(out, 1, protowire.BytesType)
-	out = protowire.AppendBytes(out, details)
-
-	return out
-}
-
-// UnmarshalPayload decodes a protobuf-encoded NebulaHandshake message.
-func UnmarshalPayload(b []byte) (Payload, error) {
-	var p Payload
-
-	for len(b) > 0 {
-		num, typ, n := protowire.ConsumeTag(b)
-		if n < 0 {
-			return p, errInvalidHandshakeMessage
-		}
-		b = b[n:]
-
-		switch {
-		case num == 1 && typ == protowire.BytesType:
-			details, n := protowire.ConsumeBytes(b)
-			if n < 0 {
-				return p, errInvalidHandshakeMessage
-			}
-			b = b[n:]
-			if err := unmarshalPayloadDetails(&p, details); err != nil {
-				return p, err
-			}
-		default:
-			n := protowire.ConsumeFieldValue(num, typ, b)
-			if n < 0 {
-				return p, errInvalidHandshakeMessage
-			}
-			b = b[n:]
-		}
-	}
-
-	return p, nil
-}
-
-func unmarshalPayloadDetails(p *Payload, b []byte) error {
-	for len(b) > 0 {
-		num, typ, n := protowire.ConsumeTag(b)
-		if n < 0 {
-			return errInvalidHandshakeDetails
-		}
-		b = b[n:]
-
-		// For known field numbers, reject any non-matching wire type as a
-		// hard error rather than silently skipping. The caller will catch
-		// missing-field cases downstream, but a wire-type mismatch on a tag
-		// we know is a peer protocol violation worth flagging here.
-		// Repeated occurrences of a singular field follow proto3 last-wins.
-		switch num {
-		case fieldCert:
-			if typ != protowire.BytesType {
-				return errInvalidHandshakeDetails
-			}
-			v, n := protowire.ConsumeBytes(b)
-			if n < 0 {
-				return errInvalidHandshakeDetails
-			}
-			p.Cert = append([]byte(nil), v...)
-			b = b[n:]
-		case fieldInitiatorIndex:
-			if typ != protowire.VarintType {
-				return errInvalidHandshakeDetails
-			}
-			v, n := protowire.ConsumeVarint(b)
-			if n < 0 || v > math.MaxUint32 {
-				return errInvalidHandshakeDetails
-			}
-			p.InitiatorIndex = uint32(v)
-			b = b[n:]
-		case fieldResponderIndex:
-			if typ != protowire.VarintType {
-				return errInvalidHandshakeDetails
-			}
-			v, n := protowire.ConsumeVarint(b)
-			if n < 0 || v > math.MaxUint32 {
-				return errInvalidHandshakeDetails
-			}
-			p.ResponderIndex = uint32(v)
-			b = b[n:]
-		case fieldTime:
-			if typ != protowire.VarintType {
-				return errInvalidHandshakeDetails
-			}
-			v, n := protowire.ConsumeVarint(b)
-			if n < 0 {
-				return errInvalidHandshakeDetails
-			}
-			p.Time = v
-			b = b[n:]
-		case fieldCertVersion:
-			if typ != protowire.VarintType {
-				return errInvalidHandshakeDetails
-			}
-			v, n := protowire.ConsumeVarint(b)
-			if n < 0 || v > math.MaxUint32 {
-				return errInvalidHandshakeDetails
-			}
-			p.CertVersion = uint32(v)
-			b = b[n:]
-		default:
-			n := protowire.ConsumeFieldValue(num, typ, b)
-			if n < 0 {
-				return errInvalidHandshakeDetails
-			}
-			b = b[n:]
-		}
-	}
-	return nil
-}

handshake/payload_test.go

@@ -1,361 +0,0 @@
-package handshake
-
-import (
-	"bytes"
-	"math"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"google.golang.org/protobuf/encoding/protowire"
-)
-
-func TestPayloadRoundTrip(t *testing.T) {
-	t.Run("all fields set", func(t *testing.T) {
-		data := MarshalPayload(nil, Payload{
-			Cert:           []byte("test-cert-bytes"),
-			CertVersion:    2,
-			InitiatorIndex: 12345,
-			ResponderIndex: 67890,
-			Time:           1234567890,
-		})
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-
-		assert.Equal(t, []byte("test-cert-bytes"), got.Cert)
-		assert.Equal(t, uint32(12345), got.InitiatorIndex)
-		assert.Equal(t, uint32(67890), got.ResponderIndex)
-		assert.Equal(t, uint64(1234567890), got.Time)
-		assert.Equal(t, uint32(2), got.CertVersion)
-	})
-
-	t.Run("minimal fields", func(t *testing.T) {
-		data := MarshalPayload(nil, Payload{InitiatorIndex: 1})
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-
-		assert.Equal(t, uint32(1), got.InitiatorIndex)
-		assert.Equal(t, uint32(0), got.ResponderIndex)
-		assert.Equal(t, uint64(0), got.Time)
-		assert.Nil(t, got.Cert)
-	})
-
-	t.Run("empty payload", func(t *testing.T) {
-		data := MarshalPayload(nil, Payload{})
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-
-		assert.Equal(t, uint32(0), got.InitiatorIndex)
-	})
-
-	t.Run("large cert bytes", func(t *testing.T) {
-		bigCert := make([]byte, 4096)
-		for i := range bigCert {
-			bigCert[i] = byte(i % 256)
-		}
-
-		data := MarshalPayload(nil, Payload{
-			Cert:           bigCert,
-			CertVersion:    2,
-			InitiatorIndex: 999,
-		})
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-
-		assert.Equal(t, bigCert, got.Cert)
-		assert.Equal(t, uint32(999), got.InitiatorIndex)
-	})
-
-	t.Run("append to existing buffer", func(t *testing.T) {
-		prefix := []byte("prefix")
-		data := MarshalPayload(prefix, Payload{InitiatorIndex: 42})
-
-		assert.Equal(t, []byte("prefix"), data[:6])
-
-		got, err := UnmarshalPayload(data[6:])
-		require.NoError(t, err)
-		assert.Equal(t, uint32(42), got.InitiatorIndex)
-	})
-}
-
-func TestPayloadUnknownFields(t *testing.T) {
-	t.Run("unknown field in outer message is skipped", func(t *testing.T) {
-		// Marshal a normal payload then append an unknown field (field 99, varint)
-		data := MarshalPayload(nil, Payload{InitiatorIndex: 42})
-		data = protowire.AppendTag(data, 99, protowire.VarintType)
-		data = protowire.AppendVarint(data, 12345)
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-		assert.Equal(t, uint32(42), got.InitiatorIndex)
-	})
-
-	t.Run("unknown field in details is skipped", func(t *testing.T) {
-		// Build details with a known field + unknown field
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, 77)
-		// Unknown field 50, varint
-		details = protowire.AppendTag(details, 50, protowire.VarintType)
-		details = protowire.AppendVarint(details, 9999)
-		// Another known field after the unknown one
-		details = protowire.AppendTag(details, fieldResponderIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, 88)
-
-		// Wrap in outer message
-		var data []byte
-		data = protowire.AppendTag(data, 1, protowire.BytesType)
-		data = protowire.AppendBytes(data, details)
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-		assert.Equal(t, uint32(77), got.InitiatorIndex)
-		assert.Equal(t, uint32(88), got.ResponderIndex)
-	})
-
-	t.Run("reserved fields 6 and 7 are skipped", func(t *testing.T) {
-		// Fields 6 and 7 are reserved in the proto definition
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, 100)
-		details = protowire.AppendTag(details, 6, protowire.VarintType)
-		details = protowire.AppendVarint(details, 1)
-		details = protowire.AppendTag(details, 7, protowire.VarintType)
-		details = protowire.AppendVarint(details, 2)
-
-		var data []byte
-		data = protowire.AppendTag(data, 1, protowire.BytesType)
-		data = protowire.AppendBytes(data, details)
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-		assert.Equal(t, uint32(100), got.InitiatorIndex)
-	})
-}
-
-func TestPayloadBytesConsumed(t *testing.T) {
-	t.Run("all bytes consumed on valid input", func(t *testing.T) {
-		original := Payload{
-			Cert:           []byte("cert"),
-			CertVersion:    2,
-			InitiatorIndex: 100,
-			ResponderIndex: 200,
-			Time:           999,
-		}
-		data := MarshalPayload(nil, original)
-
-		got, err := UnmarshalPayload(data)
-		require.NoError(t, err)
-
-		// Re-marshal and compare — proves we consumed and reproduced all fields
-		remarshaled := MarshalPayload(nil, got)
-		assert.Equal(t, data, remarshaled)
-	})
-}
-
-// wrapDetails wraps raw detail bytes in the outer NebulaHandshake envelope
-// so UnmarshalPayload can reach unmarshalPayloadDetails.
-func wrapDetails(details []byte) []byte {
-	var out []byte
-	out = protowire.AppendTag(out, 1, protowire.BytesType)
-	out = protowire.AppendBytes(out, details)
-	return out
-}
-
-func TestPayloadUnmarshalErrors(t *testing.T) {
-	t.Run("nil input", func(t *testing.T) {
-		got, err := UnmarshalPayload(nil)
-		require.NoError(t, err)
-		assert.Equal(t, uint32(0), got.InitiatorIndex)
-	})
-
-	t.Run("truncated outer tag", func(t *testing.T) {
-		_, err := UnmarshalPayload([]byte{0x80})
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated outer details field", func(t *testing.T) {
-		_, err := UnmarshalPayload([]byte{0x0a, 0x64, 0x01, 0x02, 0x03, 0x04, 0x05})
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated outer unknown field", func(t *testing.T) {
-		// Valid tag for unknown field 99 varint, but no value follows
-		var data []byte
-		data = protowire.AppendTag(data, 99, protowire.VarintType)
-		_, err := UnmarshalPayload(data)
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated details tag", func(t *testing.T) {
-		_, err := UnmarshalPayload(wrapDetails([]byte{0x80}))
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated cert bytes", func(t *testing.T) {
-		// Field 1 (cert), bytes type, length 10 but only 2 bytes
-		var details []byte
-		details = protowire.AppendTag(details, fieldCert, protowire.BytesType)
-		details = append(details, 0x0a, 0x01, 0x02) // length 10, only 2 bytes
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated initiator index varint", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = append(details, 0x80) // incomplete varint
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated responder index varint", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldResponderIndex, protowire.VarintType)
-		details = append(details, 0x80)
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated time varint", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldTime, protowire.VarintType)
-		details = append(details, 0x80)
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated cert version varint", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldCertVersion, protowire.VarintType)
-		details = append(details, 0x80)
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("truncated unknown field in details", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, 50, protowire.VarintType)
-		details = append(details, 0x80) // incomplete varint
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("cert with wrong wire type rejected", func(t *testing.T) {
-		// fieldCert as Varint instead of Bytes.
-		var details []byte
-		details = protowire.AppendTag(details, fieldCert, protowire.VarintType)
-		details = protowire.AppendVarint(details, 42)
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("initiator index with wrong wire type rejected", func(t *testing.T) {
-		// fieldInitiatorIndex as Bytes instead of Varint.
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.BytesType)
-		details = protowire.AppendBytes(details, []byte{1, 2, 3})
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("time with wrong wire type rejected", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldTime, protowire.BytesType)
-		details = protowire.AppendBytes(details, []byte{1, 2, 3})
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("cert version with wrong wire type rejected", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldCertVersion, protowire.BytesType)
-		details = protowire.AppendBytes(details, []byte{1, 2, 3})
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("repeated singular field follows proto3 last-wins", func(t *testing.T) {
-		// Per proto3, multiple instances of a singular field are accepted and
-		// the last value wins. We keep this behavior so that peers using
-		// alternative encoders aren't rejected.
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, 1)
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, 42)
-		got, err := UnmarshalPayload(wrapDetails(details))
-		require.NoError(t, err)
-		assert.Equal(t, uint32(42), got.InitiatorIndex)
-	})
-
-	t.Run("initiator index varint overflow rejected", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldInitiatorIndex, protowire.VarintType)
-		details = protowire.AppendVarint(details, math.MaxUint32+1)
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-	t.Run("cert version varint overflow rejected", func(t *testing.T) {
-		var details []byte
-		details = protowire.AppendTag(details, fieldCertVersion, protowire.VarintType)
-		details = protowire.AppendVarint(details, math.MaxUint32+1)
-		_, err := UnmarshalPayload(wrapDetails(details))
-		assert.Error(t, err)
-	})
-
-}
-
-// FuzzPayload feeds arbitrary bytes through UnmarshalPayload to confirm it
-// never panics, and for any input that parses cleanly, that re-marshal +
-// re-parse is a fix-point. Inputs come from an authenticated peer (post-
-// noise-decrypt), so the threat model is "valid peer behaving arbitrarily,"
-// not "unauthenticated injection."
-func FuzzPayload(f *testing.F) {
-	// Seed corpus with a handful of known-good shapes.
-	f.Add(MarshalPayload(nil, Payload{}))
-	f.Add(MarshalPayload(nil, Payload{Cert: []byte{1, 2, 3}, CertVersion: 2}))
-	f.Add(MarshalPayload(nil, Payload{InitiatorIndex: 42, Time: 1}))
-	f.Add(MarshalPayload(nil, Payload{
-		Cert:           []byte("seed-cert"),
-		InitiatorIndex: 1,
-		ResponderIndex: 2,
-		Time:           3,
-		CertVersion:    2,
-	}))
-	f.Add([]byte{})
-	f.Add([]byte{0xff})
-
-	f.Fuzz(func(t *testing.T, data []byte) {
-		p1, err := UnmarshalPayload(data)
-		if err != nil {
-			return
-		}
-
-		// For any input that parses, re-marshaling and re-parsing must
-		// yield an equivalent Payload. This catches dispatch bugs (e.g.
-		// emitting a field on marshal that we don't accept on parse) and
-		// any non-idempotent parsing behavior.
-		b2 := MarshalPayload(nil, p1)
-		p2, err := UnmarshalPayload(b2)
-		if err != nil {
-			t.Fatalf("re-parse of self-marshaled payload failed: %v\nintermediate: %x\n", err, b2)
-		}
-		if !payloadsEqual(p1, p2) {
-			t.Fatalf("re-marshal not idempotent\nfirst:  %+v\nsecond: %+v", p1, p2)
-		}
-	})
-}
-
-func payloadsEqual(a, b Payload) bool {
-	return bytes.Equal(a.Cert, b.Cert) &&
-		a.InitiatorIndex == b.InitiatorIndex &&
-		a.ResponderIndex == b.ResponderIndex &&
-		a.Time == b.Time &&
-		a.CertVersion == b.CertVersion
-}

handshake_ix.go

@@ -0,0 +1,678 @@
+package nebula
+
+import (
+	"bytes"
+	"net/netip"
+	"time"
+
+	"github.com/flynn/noise"
+	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/header"
+)
+
+// NOISE IX Handshakes
+
+// This function constructs a handshake packet, but does not actually send it
+// Sending is done by the handshake manager
+func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
+	err := f.handshakeManager.allocateIndex(hh)
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to generate index")
+		return false
+	}
+
+	cs := f.pki.getCertState()
+	v := cs.initiatingVersion
+	if hh.initiatingVersionOverride != cert.VersionPre1 {
+		v = hh.initiatingVersionOverride
+	} else if v < cert.Version2 {
+		// If we're connecting to a v6 address we should encourage use of a V2 cert
+		for _, a := range hh.hostinfo.vpnAddrs {
+			if a.Is6() {
+				v = cert.Version2
+				break
+			}
+		}
+	}
+
+	crt := cs.getCertificate(v)
+	if crt == nil {
+		f.l.WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			WithField("certVersion", v).
+			Error("Unable to handshake with host because no certificate is available")
+		return false
+	}
+
+	crtHs := cs.getHandshakeBytes(v)
+	if crtHs == nil {
+		f.l.WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			WithField("certVersion", v).
+			Error("Unable to handshake with host because no certificate handshake bytes is available")
+		return false
+	}
+
+	ci, err := NewConnectionState(f.l, cs, crt, true, noise.HandshakeIX)
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			WithField("certVersion", v).
+			Error("Failed to create connection state")
+		return false
+	}
+	hh.hostinfo.ConnectionState = ci
+
+	hs := &NebulaHandshake{
+		Details: &NebulaHandshakeDetails{
+			InitiatorIndex: hh.hostinfo.localIndexId,
+			Time:           uint64(time.Now().UnixNano()),
+			Cert:           crtHs,
+			CertVersion:    uint32(v),
+		},
+	}
+
+	hsBytes, err := hs.Marshal()
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("certVersion", v).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to marshal handshake message")
+		return false
+	}
+
+	h := header.Encode(make([]byte, header.Len), header.Version, header.Handshake, header.HandshakeIXPSK0, 0, 1)
+
+	msg, _, _, err := ci.H.WriteMessage(h, hsBytes)
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to call noise.WriteMessage")
+		return false
+	}
+
+	// We are sending handshake packet 1, so we don't expect to receive
+	// handshake packet 1 from the responder
+	ci.window.Update(f.l, 1)
+
+	hh.hostinfo.HandshakePacket[0] = msg
+	hh.ready = true
+	return true
+}
+
+func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H) {
+	cs := f.pki.getCertState()
+	crt := cs.GetDefaultCertificate()
+	if crt == nil {
+		f.l.WithField("from", via).
+			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			WithField("certVersion", cs.initiatingVersion).
+			Error("Unable to handshake with host because no certificate is available")
+		return
+	}
+
+	ci, err := NewConnectionState(f.l, cs, crt, false, noise.HandshakeIX)
+	if err != nil {
+		f.l.WithError(err).WithField("from", via).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Error("Failed to create connection state")
+		return
+	}
+
+	// Mark packet 1 as seen so it doesn't show up as missed
+	ci.window.Update(f.l, 1)
+
+	msg, _, _, err := ci.H.ReadMessage(nil, packet[header.Len:])
+	if err != nil {
+		f.l.WithError(err).WithField("from", via).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Error("Failed to call noise.ReadMessage")
+		return
+	}
+
+	hs := &NebulaHandshake{}
+	err = hs.Unmarshal(msg)
+	if err != nil || hs.Details == nil {
+		f.l.WithError(err).WithField("from", via).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Error("Failed unmarshal handshake message")
+		return
+	}
+
+	rc, err := cert.Recombine(cert.Version(hs.Details.CertVersion), hs.Details.Cert, ci.H.PeerStatic(), ci.Curve())
+	if err != nil {
+		f.l.WithError(err).WithField("from", via).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Info("Handshake did not contain a certificate")
+		return
+	}
+
+	remoteCert, err := f.pki.GetCAPool().VerifyCertificate(time.Now(), rc)
+	if err != nil {
+		fp, fperr := rc.Fingerprint()
+		if fperr != nil {
+			fp = "<error generating certificate fingerprint>"
+		}
+
+		e := f.l.WithError(err).WithField("from", via).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			WithField("certVpnNetworks", rc.Networks()).
+			WithField("certFingerprint", fp)
+
+		if f.l.Level >= logrus.DebugLevel {
+			e = e.WithField("cert", rc)
+		}
+
+		e.Info("Invalid certificate from host")
+		return
+	}
+
+	if !bytes.Equal(remoteCert.Certificate.PublicKey(), ci.H.PeerStatic()) {
+		f.l.WithField("from", via).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			WithField("cert", remoteCert).Info("public key mismatch between certificate and handshake")
+		return
+	}
+
+	if remoteCert.Certificate.Version() != ci.myCert.Version() {
+		// We started off using the wrong certificate version, lets see if we can match the version that was sent to us
+		myCertOtherVersion := cs.getCertificate(remoteCert.Certificate.Version())
+		if myCertOtherVersion == nil {
+			if f.l.Level >= logrus.DebugLevel {
+				f.l.WithError(err).WithFields(m{
+					"from":      via,
+					"handshake": m{"stage": 1, "style": "ix_psk0"},
+					"cert":      remoteCert,
+				}).Debug("Might be unable to handshake with host due to missing certificate version")
+			}
+		} else {
+			// Record the certificate we are actually using
+			ci.myCert = myCertOtherVersion
+		}
+	}
+
+	if len(remoteCert.Certificate.Networks()) == 0 {
+		f.l.WithError(err).WithField("from", via).
+			WithField("cert", remoteCert).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			Info("No networks in certificate")
+		return
+	}
+
+	certName := remoteCert.Certificate.Name()
+	certVersion := remoteCert.Certificate.Version()
+	fingerprint := remoteCert.Fingerprint
+	issuer := remoteCert.Certificate.Issuer()
+	vpnNetworks := remoteCert.Certificate.Networks()
+
+	anyVpnAddrsInCommon := false
+	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
+	for i, network := range vpnNetworks {
+		if f.myVpnAddrsTable.Contains(network.Addr()) {
+			f.l.WithField("vpnNetworks", vpnNetworks).WithField("from", via).
+				WithField("certName", certName).
+				WithField("certVersion", certVersion).
+				WithField("fingerprint", fingerprint).
+				WithField("issuer", issuer).
+				WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Refusing to handshake with myself")
+			return
+		}
+		vpnAddrs[i] = network.Addr()
+		if f.myVpnNetworksTable.Contains(network.Addr()) {
+			anyVpnAddrsInCommon = true
+		}
+	}
+
+	if !via.IsRelayed {
+		// We only want to apply the remote allow list for direct tunnels here
+		if !f.lightHouse.GetRemoteAllowList().AllowAll(vpnAddrs, via.UdpAddr.Addr()) {
+			f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+				Debug("lighthouse.remote_allow_list denied incoming handshake")
+			return
+		}
+	}
+
+	myIndex, err := generateIndex(f.l)
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to generate index")
+		return
+	}
+
+	hostinfo := &HostInfo{
+		ConnectionState:   ci,
+		localIndexId:      myIndex,
+		remoteIndexId:     hs.Details.InitiatorIndex,
+		vpnAddrs:          vpnAddrs,
+		HandshakePacket:   make(map[uint8][]byte, 0),
+		lastHandshakeTime: hs.Details.Time,
+		relayState: RelayState{
+			relays:         nil,
+			relayForByAddr: map[netip.Addr]*Relay{},
+			relayForByIdx:  map[uint32]*Relay{},
+		},
+	}
+
+	msgRxL := f.l.WithFields(m{
+		"vpnAddrs":       vpnAddrs,
+		"from":           via,
+		"certName":       certName,
+		"certVersion":    certVersion,
+		"fingerprint":    fingerprint,
+		"issuer":         issuer,
+		"initiatorIndex": hs.Details.InitiatorIndex,
+		"responderIndex": hs.Details.ResponderIndex,
+		"remoteIndex":    h.RemoteIndex,
+		"handshake":      m{"stage": 1, "style": "ix_psk0"},
+	})
+
+	if anyVpnAddrsInCommon {
+		msgRxL.Info("Handshake message received")
+	} else {
+		//todo warn if not lighthouse or relay?
+		msgRxL.Info("Handshake message received, but no vpnNetworks in common.")
+	}
+
+	hs.Details.ResponderIndex = myIndex
+	hs.Details.Cert = cs.getHandshakeBytes(ci.myCert.Version())
+	if hs.Details.Cert == nil {
+		msgRxL.WithField("myCertVersion", ci.myCert.Version()).
+			Error("Unable to handshake with host because no certificate handshake bytes is available")
+		return
+	}
+
+	hs.Details.CertVersion = uint32(ci.myCert.Version())
+	// Update the time in case their clock is way off from ours
+	hs.Details.Time = uint64(time.Now().UnixNano())
+
+	hsBytes, err := hs.Marshal()
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to marshal handshake message")
+		return
+	}
+
+	nh := header.Encode(make([]byte, header.Len), header.Version, header.Handshake, header.HandshakeIXPSK0, hs.Details.InitiatorIndex, 2)
+	msg, dKey, eKey, err := ci.H.WriteMessage(nh, hsBytes)
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to call noise.WriteMessage")
+		return
+	} else if dKey == nil || eKey == nil {
+		f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Noise did not arrive at a key")
+		return
+	}
+
+	hostinfo.HandshakePacket[0] = make([]byte, len(packet[header.Len:]))
+	copy(hostinfo.HandshakePacket[0], packet[header.Len:])
+
+	// Regardless of whether you are the sender or receiver, you should arrive here
+	// and complete standing up the connection.
+	hostinfo.HandshakePacket[2] = make([]byte, len(msg))
+	copy(hostinfo.HandshakePacket[2], msg)
+
+	// We are sending handshake packet 2, so we don't expect to receive
+	// handshake packet 2 from the initiator.
+	ci.window.Update(f.l, 2)
+
+	ci.peerCert = remoteCert
+	ci.dKey = NewNebulaCipherState(dKey)
+	ci.eKey = NewNebulaCipherState(eKey)
+
+	hostinfo.remotes = f.lightHouse.QueryCache(vpnAddrs)
+	if !via.IsRelayed {
+		hostinfo.SetRemote(via.UdpAddr)
+	}
+	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate)
+
+	existing, err := f.handshakeManager.CheckAndComplete(hostinfo, 0, f)
+	if err != nil {
+		switch err {
+		case ErrAlreadySeen:
+			// Update remote if preferred
+			if existing.SetRemoteIfPreferred(f.hostMap, via) {
+				// Send a test packet to ensure the other side has also switched to
+				// the preferred remote
+				f.SendMessageToVpnAddr(header.Test, header.TestRequest, vpnAddrs[0], []byte(""), make([]byte, 12, 12), make([]byte, mtu))
+			}
+
+			msg = existing.HandshakePacket[2]
+			f.messageMetrics.Tx(header.Handshake, header.MessageSubType(msg[1]), 1)
+			if !via.IsRelayed {
+				err := f.outside.WriteTo(msg, via.UdpAddr)
+				if err != nil {
+					f.l.WithField("vpnAddrs", existing.vpnAddrs).WithField("from", via).
+						WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("cached", true).
+						WithError(err).Error("Failed to send handshake message")
+				} else {
+					f.l.WithField("vpnAddrs", existing.vpnAddrs).WithField("from", via).
+						WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("cached", true).
+						Info("Handshake message sent")
+				}
+				return
+			} else {
+				if via.relay == nil {
+					f.l.Error("Handshake send failed: both addr and via.relay are nil.")
+					return
+				}
+				hostinfo.relayState.InsertRelayTo(via.relayHI.vpnAddrs[0])
+				f.SendVia(via.relayHI, via.relay, msg, make([]byte, 12), make([]byte, mtu), false)
+				f.l.WithField("vpnAddrs", existing.vpnAddrs).WithField("relay", via.relayHI.vpnAddrs[0]).
+					WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("cached", true).
+					Info("Handshake message sent")
+				return
+			}
+		case ErrExistingHostInfo:
+			// This means there was an existing tunnel and this handshake was older than the one we are currently based on
+			f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+				WithField("certName", certName).
+				WithField("certVersion", certVersion).
+				WithField("oldHandshakeTime", existing.lastHandshakeTime).
+				WithField("newHandshakeTime", hostinfo.lastHandshakeTime).
+				WithField("fingerprint", fingerprint).
+				WithField("issuer", issuer).
+				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+				WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				Info("Handshake too old")
+
+			// Send a test packet to trigger an authenticated tunnel test, this should suss out any lingering tunnel issues
+			f.SendMessageToVpnAddr(header.Test, header.TestRequest, vpnAddrs[0], []byte(""), make([]byte, 12, 12), make([]byte, mtu))
+			return
+		case ErrLocalIndexCollision:
+			// This means we failed to insert because of collision on localIndexId. Just let the next handshake packet retry
+			f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+				WithField("certName", certName).
+				WithField("certVersion", certVersion).
+				WithField("fingerprint", fingerprint).
+				WithField("issuer", issuer).
+				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+				WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				WithField("localIndex", hostinfo.localIndexId).WithField("collision", existing.vpnAddrs).
+				Error("Failed to add HostInfo due to localIndex collision")
+			return
+		default:
+			// Shouldn't happen, but just in case someone adds a new error type to CheckAndComplete
+			// And we forget to update it here
+			f.l.WithError(err).WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+				WithField("certName", certName).
+				WithField("certVersion", certVersion).
+				WithField("fingerprint", fingerprint).
+				WithField("issuer", issuer).
+				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+				WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				Error("Failed to add HostInfo to HostMap")
+			return
+		}
+	}
+
+	// Do the send
+	f.messageMetrics.Tx(header.Handshake, header.MessageSubType(msg[1]), 1)
+	if !via.IsRelayed {
+		err = f.outside.WriteTo(msg, via.UdpAddr)
+		log := f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+			WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"})
+		if err != nil {
+			log.WithError(err).Error("Failed to send handshake")
+		} else {
+			log.Info("Handshake message sent")
+		}
+	} else {
+		if via.relay == nil {
+			f.l.Error("Handshake send failed: both addr and via.relay are nil.")
+			return
+		}
+		hostinfo.relayState.InsertRelayTo(via.relayHI.vpnAddrs[0])
+		// I successfully received a handshake. Just in case I marked this tunnel as 'Disestablished', ensure
+		// it's correctly marked as working.
+		via.relayHI.relayState.UpdateRelayForByIdxState(via.remoteIdx, Established)
+		f.SendVia(via.relayHI, via.relay, msg, make([]byte, 12), make([]byte, mtu), false)
+		f.l.WithField("vpnAddrs", vpnAddrs).WithField("relay", via.relayHI.vpnAddrs[0]).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+			WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			Info("Handshake message sent")
+	}
+
+	f.connectionManager.AddTrafficWatch(hostinfo)
+
+	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
+
+	return
+}
+
+func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packet []byte, h *header.H) bool {
+	if hh == nil {
+		// Nothing here to tear down, got a bogus stage 2 packet
+		return true
+	}
+
+	hh.Lock()
+	defer hh.Unlock()
+
+	hostinfo := hh.hostinfo
+	if !via.IsRelayed {
+		// The vpnAddr we know about is the one we tried to handshake with, use it to apply the remote allow list.
+		if !f.lightHouse.GetRemoteAllowList().AllowAll(hostinfo.vpnAddrs, via.UdpAddr.Addr()) {
+			f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).Debug("lighthouse.remote_allow_list denied incoming handshake")
+			return false
+		}
+	}
+
+	ci := hostinfo.ConnectionState
+	msg, eKey, dKey, err := ci.H.ReadMessage(nil, packet[header.Len:])
+	if err != nil {
+		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("header", h).
+			Error("Failed to call noise.ReadMessage")
+
+		// We don't want to tear down the connection on a bad ReadMessage because it could be an attacker trying
+		// to DOS us. Every other error condition after should to allow a possible good handshake to complete in the
+		// near future
+		return false
+	} else if dKey == nil || eKey == nil {
+		f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			Error("Noise did not arrive at a key")
+
+		// This should be impossible in IX but just in case, if we get here then there is no chance to recover
+		// the handshake state machine. Tear it down
+		return true
+	}
+
+	hs := &NebulaHandshake{}
+	err = hs.Unmarshal(msg)
+	if err != nil || hs.Details == nil {
+		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).Error("Failed unmarshal handshake message")
+
+		// The handshake state machine is complete, if things break now there is no chance to recover. Tear down and start again
+		return true
+	}
+
+	rc, err := cert.Recombine(cert.Version(hs.Details.CertVersion), hs.Details.Cert, ci.H.PeerStatic(), ci.Curve())
+	if err != nil {
+		f.l.WithError(err).WithField("from", via).
+			WithField("vpnAddrs", hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			Info("Handshake did not contain a certificate")
+		return true
+	}
+
+	remoteCert, err := f.pki.GetCAPool().VerifyCertificate(time.Now(), rc)
+	if err != nil {
+		fp, err := rc.Fingerprint()
+		if err != nil {
+			fp = "<error generating certificate fingerprint>"
+		}
+
+		e := f.l.WithError(err).WithField("from", via).
+			WithField("vpnAddrs", hostinfo.vpnAddrs).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			WithField("certFingerprint", fp).
+			WithField("certVpnNetworks", rc.Networks())
+
+		if f.l.Level >= logrus.DebugLevel {
+			e = e.WithField("cert", rc)
+		}
+
+		e.Info("Invalid certificate from host")
+		return true
+	}
+	if !bytes.Equal(remoteCert.Certificate.PublicKey(), ci.H.PeerStatic()) {
+		f.l.WithField("from", via).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			WithField("cert", remoteCert).Info("public key mismatch between certificate and handshake")
+		return true
+	}
+
+	if len(remoteCert.Certificate.Networks()) == 0 {
+		f.l.WithError(err).WithField("from", via).
+			WithField("vpnAddrs", hostinfo.vpnAddrs).
+			WithField("cert", remoteCert).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			Info("No networks in certificate")
+		return true
+	}
+
+	vpnNetworks := remoteCert.Certificate.Networks()
+	certName := remoteCert.Certificate.Name()
+	certVersion := remoteCert.Certificate.Version()
+	fingerprint := remoteCert.Fingerprint
+	issuer := remoteCert.Certificate.Issuer()
+
+	hostinfo.remoteIndexId = hs.Details.ResponderIndex
+	hostinfo.lastHandshakeTime = hs.Details.Time
+
+	// Store their cert and our symmetric keys
+	ci.peerCert = remoteCert
+	ci.dKey = NewNebulaCipherState(dKey)
+	ci.eKey = NewNebulaCipherState(eKey)
+
+	// Make sure the current udpAddr being used is set for responding
+	if !via.IsRelayed {
+		hostinfo.SetRemote(via.UdpAddr)
+	} else {
+		hostinfo.relayState.InsertRelayTo(via.relayHI.vpnAddrs[0])
+	}
+
+	correctHostResponded := false
+	anyVpnAddrsInCommon := false
+	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
+	for i, network := range vpnNetworks {
+		vpnAddrs[i] = network.Addr()
+		if f.myVpnNetworksTable.Contains(network.Addr()) {
+			anyVpnAddrsInCommon = true
+		}
+		if hostinfo.vpnAddrs[0] == network.Addr() {
+			// todo is it more correct to see if any of hostinfo.vpnAddrs are in the cert? it should have len==1, but one day it might not?
+			correctHostResponded = true
+		}
+	}
+
+	// Ensure the right host responded
+	if !correctHostResponded {
+		f.l.WithField("intendedVpnAddrs", hostinfo.vpnAddrs).WithField("haveVpnNetworks", vpnNetworks).
+			WithField("from", via).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			Info("Incorrect host responded to handshake")
+
+		// Release our old handshake from pending, it should not continue
+		f.handshakeManager.DeleteHostInfo(hostinfo)
+
+		// Create a new hostinfo/handshake for the intended vpn ip
+		//TODO is hostinfo.vpnAddrs[0] always the address to use?
+		f.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(newHH *HandshakeHostInfo) {
+			// Block the current used address
+			newHH.hostinfo.remotes = hostinfo.remotes
+			newHH.hostinfo.remotes.BlockRemote(via)
+
+			f.l.WithField("blockedUdpAddrs", newHH.hostinfo.remotes.CopyBlockedRemotes()).
+				WithField("vpnNetworks", vpnNetworks).
+				WithField("remotes", newHH.hostinfo.remotes.CopyAddrs(f.hostMap.GetPreferredRanges())).
+				Info("Blocked addresses for handshakes")
+
+			// Swap the packet store to benefit the original intended recipient
+			newHH.packetStore = hh.packetStore
+			hh.packetStore = []*cachedPacket{}
+
+			// Finally, put the correct vpn addrs in the host info, tell them to close the tunnel, and return true to tear down
+			hostinfo.vpnAddrs = vpnAddrs
+			f.sendCloseTunnel(hostinfo)
+		})
+
+		return true
+	}
+
+	// Mark packet 2 as seen so it doesn't show up as missed
+	ci.window.Update(f.l, 2)
+
+	duration := time.Since(hh.startTime).Nanoseconds()
+	msgRxL := f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+		WithField("certName", certName).
+		WithField("certVersion", certVersion).
+		WithField("fingerprint", fingerprint).
+		WithField("issuer", issuer).
+		WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+		WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+		WithField("durationNs", duration).
+		WithField("sentCachedPackets", len(hh.packetStore))
+	if anyVpnAddrsInCommon {
+		msgRxL.Info("Handshake message received")
+	} else {
+		//todo warn if not lighthouse or relay?
+		msgRxL.Info("Handshake message received, but no vpnNetworks in common.")
+	}
+
+	// Build up the radix for the firewall if we have subnets in the cert
+	hostinfo.vpnAddrs = vpnAddrs
+	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate)
+
+	// Complete our handshake and update metrics, this will replace any existing tunnels for the vpnAddrs here
+	f.handshakeManager.Complete(hostinfo, f)
+	f.connectionManager.AddTrafficWatch(hostinfo)
+
+	if f.l.Level >= logrus.DebugLevel {
+		hostinfo.logger(f.l).Debugf("Sending %d stored packets", len(hh.packetStore))
+	}
+
+	if len(hh.packetStore) > 0 {
+		nb := make([]byte, 12, 12)
+		out := make([]byte, mtu)
+		for _, cp := range hh.packetStore {
+			cp.callback(cp.messageType, cp.messageSubType, hostinfo, cp.packet, nb, out)
+		}
+		f.cachedPacketMetrics.sent.Inc(int64(len(hh.packetStore)))
+	}
+
+	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
+	f.metricHandshakes.Update(duration)
+
+	return false
+}

handshake_manager_test.go

@@ -5,7 +5,6 @@ import (
 	"testing"
 	"time"
 
-	"github.com/gaissmai/bart"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/test"
@@ -28,7 +27,7 @@ func Test_NewHandshakeManagerVpnIp(t *testing.T) {
 		initiatingVersion: cert.Version1,
 		privateKey:        []byte{},
 		v1Cert:            &dummyCert{version: cert.Version1},
-		v1Credential:      nil,
+		v1HandshakeBytes:  []byte{},
 	}
 
 	blah := NewHandshakeManager(l, mainHM, lh, &udp.NoopConn{}, defaultHandshakeConfig)
@@ -101,137 +100,3 @@ func (mw *mockEncWriter) GetHostInfo(_ netip.Addr) *HostInfo {
 func (mw *mockEncWriter) GetCertState() *CertState {
 	return &CertState{initiatingVersion: cert.Version2}
 }
-
-func TestValidatePeerCert(t *testing.T) {
-	l := test.NewLogger()
-
-	myNetwork := netip.MustParsePrefix("10.0.0.1/24")
-	myAddrTable := new(bart.Lite)
-	myAddrTable.Insert(netip.PrefixFrom(myNetwork.Addr(), myNetwork.Addr().BitLen()))
-	myNetTable := new(bart.Lite)
-	myNetTable.Insert(myNetwork.Masked())
-
-	newHM := func() *HandshakeManager {
-		hm := NewHandshakeManager(l, newHostMap(l), newTestLighthouse(), &udp.NoopConn{}, defaultHandshakeConfig)
-		hm.f = &Interface{
-			handshakeManager:   hm,
-			pki:                &PKI{},
-			l:                  l,
-			myVpnAddrsTable:    myAddrTable,
-			myVpnNetworksTable: myNetTable,
-			lightHouse:         hm.lightHouse,
-		}
-		return hm
-	}
-
-	cached := func(networks ...netip.Prefix) *cert.CachedCertificate {
-		return &cert.CachedCertificate{
-			Certificate: &dummyCert{name: "peer", networks: networks},
-		}
-	}
-
-	via := ViaSender{
-		UdpAddr:   netip.MustParseAddrPort("198.51.100.7:4242"),
-		IsRelayed: true, // skip the remote allow list (covered separately)
-	}
-
-	t.Run("addr inside our networks sets anyVpnAddrsInCommon", func(t *testing.T) {
-		hm := newHM()
-		// 10.0.0.2 falls inside our 10.0.0.0/24
-		addrs, common, ok := hm.validatePeerCert(via, cached(netip.MustParsePrefix("10.0.0.2/24")))
-		assert.True(t, ok)
-		assert.True(t, common)
-		assert.Equal(t, []netip.Addr{netip.MustParseAddr("10.0.0.2")}, addrs)
-	})
-
-	t.Run("addr outside our networks leaves anyVpnAddrsInCommon false", func(t *testing.T) {
-		hm := newHM()
-		addrs, common, ok := hm.validatePeerCert(via, cached(netip.MustParsePrefix("192.168.1.5/24")))
-		assert.True(t, ok)
-		assert.False(t, common)
-		assert.Equal(t, []netip.Addr{netip.MustParseAddr("192.168.1.5")}, addrs)
-	})
-
-	t.Run("any matching network is enough", func(t *testing.T) {
-		hm := newHM()
-		addrs, common, ok := hm.validatePeerCert(via, cached(
-			netip.MustParsePrefix("192.168.1.5/24"),
-			netip.MustParsePrefix("10.0.0.42/24"),
-		))
-		assert.True(t, ok)
-		assert.True(t, common)
-		assert.Len(t, addrs, 2)
-	})
-
-	t.Run("self-handshake is rejected", func(t *testing.T) {
-		hm := newHM()
-		// 10.0.0.1 is in myVpnAddrsTable
-		addrs, common, ok := hm.validatePeerCert(via, cached(netip.MustParsePrefix("10.0.0.1/24")))
-		assert.False(t, ok)
-		assert.False(t, common)
-		assert.Nil(t, addrs)
-	})
-
-	t.Run("cert with no networks is rejected", func(t *testing.T) {
-		hm := newHM()
-		addrs, common, ok := hm.validatePeerCert(via, cached())
-		assert.False(t, ok)
-		assert.False(t, common)
-		assert.Nil(t, addrs)
-	})
-}
-
-func TestHandleIncomingDispatch(t *testing.T) {
-	l := test.NewLogger()
-
-	newHM := func() *HandshakeManager {
-		hm := NewHandshakeManager(l, newHostMap(l), newTestLighthouse(), &udp.NoopConn{}, defaultHandshakeConfig)
-		hm.f = &Interface{
-			handshakeManager: hm,
-			pki:              &PKI{},
-			l:                l,
-		}
-		return hm
-	}
-
-	via := ViaSender{
-		UdpAddr:   netip.MustParseAddrPort("198.51.100.7:4242"),
-		IsRelayed: true, // bypass remote allow list
-	}
-
-	// A packet body of zero length is fine for these tests: dispatch is
-	// gated on header fields, and we assert that we never reach noise/cert
-	// processing for any of the malformed shapes here.
-	pkt := make([]byte, header.Len)
-
-	t.Run("unsupported subtype dropped", func(t *testing.T) {
-		hm := newHM()
-		h := &header.H{Type: header.Handshake, Subtype: header.MessageSubType(99), MessageCounter: 1}
-		hm.HandleIncoming(via, pkt, h)
-		assert.Empty(t, hm.indexes, "no pending handshake should be created")
-	})
-
-	t.Run("stage-1 with non-zero RemoteIndex dropped", func(t *testing.T) {
-		hm := newHM()
-		h := &header.H{
-			Type:           header.Handshake,
-			Subtype:        header.HandshakeIXPSK0,
-			RemoteIndex:    0xdeadbeef,
-			MessageCounter: 1,
-		}
-		hm.HandleIncoming(via, pkt, h)
-		assert.Empty(t, hm.indexes, "spoofed stage-1 must not create a pending machine")
-	})
-
-	t.Run("continuation with no matching pending index dropped", func(t *testing.T) {
-		hm := newHM()
-		h := &header.H{
-			Type:           header.Handshake,
-			Subtype:        header.HandshakeIXPSK0,
-			RemoteIndex:    0xcafef00d,
-			MessageCounter: 2,
-		}
-		hm.HandleIncoming(via, pkt, h)
-		assert.Empty(t, hm.indexes, "orphan stage-2 must not create state")
-	})
-}

header/header.go

@@ -174,10 +174,6 @@ func (h *H) SubTypeName() string {
 	return SubTypeName(h.Type, h.Subtype)
 }
 
-func (h *H) IsValidSubType() bool {
-	return IsValidSubType(h.Type, h.Subtype)
-}
-
 // SubTypeName will transform a nebula message sub type into a human string
 func SubTypeName(t MessageType, s MessageSubType) string {
 	if n, ok := subTypeMap[t]; ok {
@@ -189,16 +185,6 @@ func SubTypeName(t MessageType, s MessageSubType) string {
 	return "unknown"
 }
 
-func IsValidSubType(t MessageType, s MessageSubType) bool {
-	if n, ok := subTypeMap[t]; ok {
-		if _, ok := (*n)[s]; ok {
-			return true
-		}
-	}
-
-	return false
-}
-
 // NewHeader turns bytes into a header
 func NewHeader(b []byte) (*H, error) {
 	h := new(H)

hostmap.go

@@ -1,11 +1,9 @@
 package nebula
 
 import (
-	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
-	"log/slog"
 	"net"
 	"net/netip"
 	"slices"
@@ -15,10 +13,10 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/logging"
 )
 
 const defaultPromoteEvery = 1000       // Count of packets sent before we try moving a tunnel to a preferred underlay ip address
@@ -62,7 +60,7 @@ type HostMap struct {
 	RemoteIndexes   map[uint32]*HostInfo
 	Hosts           map[netip.Addr]*HostInfo
 	preferredRanges atomic.Pointer[[]netip.Prefix]
-	l               *slog.Logger
+	l               *logrus.Logger
 }
 
 // For synchronization, treat the pointed-to Relay struct as immutable. To edit the Relay
@@ -315,7 +313,7 @@ type cachedPacketMetrics struct {
 	dropped metrics.Counter
 }
 
-func NewHostMapFromConfig(l *slog.Logger, c *config.C) *HostMap {
+func NewHostMapFromConfig(l *logrus.Logger, c *config.C) *HostMap {
 	hm := newHostMap(l)
 
 	hm.reload(c, true)
@@ -323,12 +321,13 @@ func NewHostMapFromConfig(l *slog.Logger, c *config.C) *HostMap {
 		hm.reload(c, false)
 	})
 
-	l.Info("Main HostMap created", "preferredRanges", hm.GetPreferredRanges())
+	l.WithField("preferredRanges", hm.GetPreferredRanges()).
+		Info("Main HostMap created")
 
 	return hm
 }
 
-func newHostMap(l *slog.Logger) *HostMap {
+func newHostMap(l *logrus.Logger) *HostMap {
 	return &HostMap{
 		Indexes:       map[uint32]*HostInfo{},
 		Relays:        map[uint32]*HostInfo{},
@@ -347,10 +346,7 @@ func (hm *HostMap) reload(c *config.C, initial bool) {
 			preferredRange, err := netip.ParsePrefix(rawPreferredRange)
 
 			if err != nil {
-				hm.l.Warn("Failed to parse preferred ranges, ignoring",
-					"error", err,
-					"range", rawPreferredRanges,
-				)
+				hm.l.WithError(err).WithField("range", rawPreferredRanges).Warn("Failed to parse preferred ranges, ignoring")
 				continue
 			}
 
@@ -359,10 +355,7 @@ func (hm *HostMap) reload(c *config.C, initial bool) {
 
 		oldRanges := hm.preferredRanges.Swap(&preferredRanges)
 		if !initial {
-			hm.l.Info("preferred_ranges changed",
-				"oldPreferredRanges", *oldRanges,
-				"newPreferredRanges", preferredRanges,
-			)
+			hm.l.WithField("oldPreferredRanges", *oldRanges).WithField("newPreferredRanges", preferredRanges).Info("preferred_ranges changed")
 		}
 	}
 }
@@ -495,11 +488,10 @@ func (hm *HostMap) unlockedInnerDeleteHostInfo(hostinfo *HostInfo, addr netip.Ad
 		hm.Indexes = map[uint32]*HostInfo{}
 	}
 
-	if hm.l.Enabled(context.Background(), slog.LevelDebug) {
-		hm.l.Debug("Hostmap hostInfo deleted",
-			"hostMap", m{"mapTotalSize": len(hm.Hosts),
-				"vpnAddrs": hostinfo.vpnAddrs, "indexNumber": hostinfo.localIndexId, "remoteIndexNumber": hostinfo.remoteIndexId},
-		)
+	if hm.l.Level >= logrus.DebugLevel {
+		hm.l.WithField("hostMap", m{"mapTotalSize": len(hm.Hosts),
+			"vpnAddrs": hostinfo.vpnAddrs, "indexNumber": hostinfo.localIndexId, "remoteIndexNumber": hostinfo.remoteIndexId}).
+			Debug("Hostmap hostInfo deleted")
 	}
 
 	if isLastHostinfo {
@@ -623,11 +615,10 @@ func (hm *HostMap) unlockedAddHostInfo(hostinfo *HostInfo, f *Interface) {
 	hm.Indexes[hostinfo.localIndexId] = hostinfo
 	hm.RemoteIndexes[hostinfo.remoteIndexId] = hostinfo
 
-	if hm.l.Enabled(context.Background(), slog.LevelDebug) {
-		hm.l.Debug("Hostmap vpnIp added",
-			"hostMap", m{"vpnAddrs": hostinfo.vpnAddrs, "mapTotalSize": len(hm.Hosts),
-				"hostinfo": m{"existing": true, "localIndexId": hostinfo.localIndexId, "vpnAddrs": hostinfo.vpnAddrs}},
-		)
+	if hm.l.Level >= logrus.DebugLevel {
+		hm.l.WithField("hostMap", m{"vpnAddrs": hostinfo.vpnAddrs, "mapTotalSize": len(hm.Hosts),
+			"hostinfo": m{"existing": true, "localIndexId": hostinfo.localIndexId, "vpnAddrs": hostinfo.vpnAddrs}}).
+			Debug("Hostmap vpnIp added")
 	}
 }
 
@@ -793,21 +784,18 @@ func (i *HostInfo) buildNetworks(myVpnNetworksTable *bart.Lite, c cert.Certifica
 	}
 }
 
-// logger returns a derived slog.Logger with per-hostinfo fields pre-bound.
-func (i *HostInfo) logger(l *slog.Logger) *slog.Logger {
+func (i *HostInfo) logger(l *logrus.Logger) *logrus.Entry {
 	if i == nil {
-		return l
+		return logrus.NewEntry(l)
 	}
 
-	li := l.With(
-		"vpnAddrs", i.vpnAddrs,
-		"localIndex", i.localIndexId,
-		"remoteIndex", i.remoteIndexId,
-	)
+	li := l.WithField("vpnAddrs", i.vpnAddrs).
+		WithField("localIndex", i.localIndexId).
+		WithField("remoteIndex", i.remoteIndexId)
 
 	if connState := i.ConnectionState; connState != nil {
 		if peerCert := connState.peerCert; peerCert != nil {
-			li = li.With("certName", peerCert.Certificate.Name())
+			li = li.WithField("certName", peerCert.Certificate.Name())
 		}
 	}
 
@@ -816,17 +804,14 @@ func (i *HostInfo) logger(l *slog.Logger) *slog.Logger {
 
 // Utility functions
 
-func localAddrs(l *slog.Logger, allowList *LocalAllowList) []netip.Addr {
+func localAddrs(l *logrus.Logger, allowList *LocalAllowList) []netip.Addr {
 	//FIXME: This function is pretty garbage
 	var finalAddrs []netip.Addr
 	ifaces, _ := net.Interfaces()
 	for _, i := range ifaces {
 		allow := allowList.AllowName(i.Name)
-		if l.Enabled(context.Background(), logging.LevelTrace) {
-			l.Log(context.Background(), logging.LevelTrace, "localAllowList.AllowName",
-				"interfaceName", i.Name,
-				"allow", allow,
-			)
+		if l.Level >= logrus.TraceLevel {
+			l.WithField("interfaceName", i.Name).WithField("allow", allow).Trace("localAllowList.AllowName")
 		}
 
 		if !allow {
@@ -844,8 +829,8 @@ func localAddrs(l *slog.Logger, allowList *LocalAllowList) []netip.Addr {
 			}
 
 			if !addr.IsValid() {
-				if l.Enabled(context.Background(), slog.LevelDebug) {
-					l.Debug("addr was invalid", "localAddr", rawAddr)
+				if l.Level >= logrus.DebugLevel {
+					l.WithField("localAddr", rawAddr).Debug("addr was invalid")
 				}
 				continue
 			}
@@ -853,11 +838,8 @@ func localAddrs(l *slog.Logger, allowList *LocalAllowList) []netip.Addr {
 
 			if addr.IsLoopback() == false && addr.IsLinkLocalUnicast() == false {
 				isAllowed := allowList.Allow(addr)
-				if l.Enabled(context.Background(), logging.LevelTrace) {
-					l.Log(context.Background(), logging.LevelTrace, "localAllowList.Allow",
-						"localAddr", addr,
-						"allowed", isAllowed,
-					)
+				if l.Level >= logrus.TraceLevel {
+					l.WithField("localAddr", addr).WithField("allowed", isAllowed).Trace("localAllowList.Allow")
 				}
 				if !isAllowed {
 					continue

hostmap_test.go

@@ -196,7 +196,7 @@ func TestHostMap_DeleteHostInfo(t *testing.T) {
 
 func TestHostMap_reload(t *testing.T) {
 	l := test.NewLogger()
-	c := config.NewC(test.NewLogger())
+	c := config.NewC(l)
 
 	hm := NewHostMapFromConfig(l, c)
 

inside.go

@@ -1,32 +1,21 @@
 package nebula
 
 import (
-	"context"
-	"io"
-	"log/slog"
 	"net/netip"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/iputil"
 	"github.com/slackhq/nebula/noiseutil"
-	"github.com/slackhq/nebula/overlay/batch"
 	"github.com/slackhq/nebula/routing"
-	"github.com/slackhq/nebula/wire"
 )
 
-func (f *Interface) consumeInsidePacket(pkt wire.TunPacket, fwPacket *firewall.Packet, nb []byte, sendBatch *batch.SendBatch, rejectBuf []byte, q int, localCache firewall.ConntrackCache) {
-	// borrowed: pkt.Bytes is owned by the originating tio.Queue and is
-	// only valid until the next Read on that queue. If you must keep
-	// the packet, use pkt.Clone() to detach it
-	packet := pkt.Bytes
-	err := newPacket(packet, false, fwPacket)
+func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, fwCtx *firewall.PacketContext, nb, out []byte, q int, localCache firewall.ConntrackCache) {
+	err := newPacket(packet, false, fwPacket, fwCtx)
 	if err != nil {
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("Error while validating outbound packet",
-				"packet", packet,
-				"error", err,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("packet", packet).Debugf("Error while validating outbound packet: %s", err)
 		}
 		return
 	}
@@ -44,12 +33,9 @@ func (f *Interface) consumeInsidePacket(pkt wire.TunPacket, fwPacket *firewall.P
 		// routes packets from the Nebula addr to the Nebula addr through the Nebula
 		// TUN device.
 		if immediatelyForwardToSelf {
-			err := pkt.PerSegment(func(seg []byte) error {
-				_, werr := f.readers[q].Write(seg)
-				return werr
-			})
+			_, err := f.readers[q].Write(packet)
 			if err != nil {
-				f.l.Error("Failed to forward to tun", "error", err)
+				f.l.WithError(err).Error("Failed to forward to tun")
 			}
 		}
 		// Otherwise, drop. On linux, we should never see these packets - Linux
@@ -63,28 +49,15 @@ func (f *Interface) consumeInsidePacket(pkt wire.TunPacket, fwPacket *firewall.P
 	}
 
 	hostinfo, ready := f.getOrHandshakeConsiderRouting(fwPacket, func(hh *HandshakeHostInfo) {
-		// borrowed: SegmentSuperpacket builds each segment in the kernel-supplied pkt
-		// bytes underneath. cachePacket explicitly copies its argument (handshake_manager.go cachePacket),
-		// so retaining segments past the loop is safe.
-		err := pkt.PerSegment(func(seg []byte) error {
-			hh.cachePacket(f.l, header.Message, 0, seg, f.sendMessageNow, f.cachedPacketMetrics)
-			return nil
-		})
-		if err != nil && f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("Failed to segment superpacket for handshake cache",
-				"error", err,
-				"vpnAddr", fwPacket.RemoteAddr,
-			)
-		}
+		hh.cachePacket(f.l, header.Message, 0, packet, f.sendMessageNow, f.cachedPacketMetrics)
 	})
 
 	if hostinfo == nil {
-		f.rejectInside(packet, rejectBuf, q)
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks",
-				"vpnAddr", fwPacket.RemoteAddr,
-				"fwPacket", fwPacket,
-			)
+		f.rejectInside(packet, out, q)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("vpnAddr", fwPacket.RemoteAddr).
+				WithField("fwPacket", fwPacket).
+				Debugln("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks")
 		}
 		return
 	}
@@ -93,138 +66,21 @@ func (f *Interface) consumeInsidePacket(pkt wire.TunPacket, fwPacket *firewall.P
 		return
 	}
 
-	dropReason := f.firewall.Drop(*fwPacket, false, hostinfo, f.pki.GetCAPool(), localCache)
+	dropReason := f.firewall.Drop(*fwPacket, *fwCtx, false, hostinfo, f.pki.GetCAPool(), localCache)
 	if dropReason == nil {
-		f.sendInsideMessage(hostinfo, pkt, nb, sendBatch)
+		f.sendNoMetrics(header.Message, 0, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, packet, nb, out, q)
+
 	} else {
-		f.rejectInside(packet, rejectBuf, q)
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).Debug("dropping outbound packet",
-				"fwPacket", fwPacket,
-				"reason", dropReason,
-			)
+		f.rejectInside(packet, out, q)
+		if f.l.Level >= logrus.DebugLevel {
+			hostinfo.logger(f.l).
+				WithField("fwPacket", fwPacket).
+				WithField("reason", dropReason).
+				Debugln("dropping outbound packet")
 		}
 	}
 }
 
-func (f *Interface) sendInsideEncrypt(hostinfo *HostInfo, ci *ConnectionState, seg, scratch, nb []byte) []byte {
-	if noiseutil.EncryptLockNeeded {
-		ci.writeLock.Lock()
-	}
-	c := ci.messageCounter.Add(1)
-
-	out := header.Encode(scratch, header.Version, header.Message, 0, hostinfo.remoteIndexId, c)
-	f.connectionManager.Out(hostinfo)
-
-	out, encErr := ci.eKey.EncryptDanger(out, out, seg, c, nb)
-	if noiseutil.EncryptLockNeeded {
-		ci.writeLock.Unlock()
-	}
-	if encErr != nil {
-		hostinfo.logger(f.l).Error("Failed to encrypt outgoing packet",
-			"error", encErr,
-			"udpAddr", hostinfo.remote,
-			"counter", c,
-		)
-		// Skip this segment; the rest of the superpacket can still
-		// go out — TCP will retransmit anything we drop here.
-		return nil
-	}
-
-	return out
-}
-
-// sendInsideMessage encrypts a firewall-approved inside packet (or every
-// segment of a TSO/USO superpacket) into the caller's batch slot for
-// later sendmmsg flush. Segmentation is fused with encryption here so the
-// kernel-supplied superpacket bytes never get written into a separate
-// scratch arena: PerSegment builds each segment's plaintext in
-// segScratch[:segLen] in turn, and we encrypt directly into a fresh
-// SendBatch slot.
-func (f *Interface) sendInsideMessage(hostinfo *HostInfo, pkt wire.TunPacket, nb []byte, sendBatch *batch.SendBatch) {
-	ci := hostinfo.ConnectionState
-	if ci.eKey == nil {
-		return
-	}
-
-	if hostinfo.lastRebindCount != f.rebindCount {
-		//NOTE: there is an update hole if a tunnel isn't used and exactly 256 rebinds occur before the tunnel is
-		// finally used again. This tunnel would eventually be torn down and recreated if this action didn't help.
-		f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
-		hostinfo.lastRebindCount = f.rebindCount
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).Debug("Lighthouse update triggered for punch due to rebind counter",
-				"vpnAddrs", hostinfo.vpnAddrs,
-			)
-		}
-	}
-
-	if !hostinfo.remote.IsValid() { //the relay path
-		//first, find our relay hostinfo:
-		var relayHostInfo *HostInfo
-		var relay *Relay
-		var err error
-		for _, relayIP := range hostinfo.relayState.CopyRelayIps() {
-			relayHostInfo, relay, err = f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relayIP)
-			if err != nil {
-				hostinfo.relayState.DeleteRelay(relayIP)
-				hostinfo.logger(f.l).Info("sendNoMetrics failed to find HostInfo",
-					"relay", relayIP,
-					"error", err,
-				)
-				continue
-			}
-			break
-		}
-		if relayHostInfo == nil || relay == nil {
-			//failure already logged
-			return
-		}
-
-		err = pkt.PerSegment(func(seg []byte) error {
-			//relay header + header + plaintext + AEAD tag (16 bytes for both AES-GCM and ChaCha20-Poly1305) + relay tag
-			scratch := sendBatch.Reserve(header.Len + header.Len + len(seg) + 16 + 16)
-
-			innerPacket := f.sendInsideEncrypt(hostinfo, ci, seg, scratch[header.Len:], nb)
-			if innerPacket == nil {
-				return nil
-			}
-
-			//now we need to do a relay-encrypt:
-			toSend, err := f.prepareSendVia(relayHostInfo, relay, innerPacket, nb, scratch, true)
-			if err != nil {
-				//already logged
-				return nil
-			}
-
-			sendBatch.Commit(toSend, relayHostInfo.remote, 0)
-			return nil
-		})
-		if err != nil {
-			hostinfo.logger(f.l).Error("Failed to segment superpacket for relay send", "error", err)
-		}
-		return
-	}
-
-	err := pkt.PerSegment(func(seg []byte) error {
-		// header + plaintext + AEAD tag (16 bytes for both AES-GCM and ChaCha20-Poly1305)
-		scratch := sendBatch.Reserve(header.Len + len(seg) + 16)
-
-		out := f.sendInsideEncrypt(hostinfo, ci, seg, scratch, nb)
-		if out == nil {
-			return nil
-		}
-
-		sendBatch.Commit(out, hostinfo.remote, 0)
-		return nil
-	})
-	if err != nil {
-		hostinfo.logger(f.l).Error("Failed to segment superpacket for send",
-			"error", err,
-		)
-	}
-}
-
 func (f *Interface) rejectInside(packet []byte, out []byte, q int) {
 	if !f.firewall.InSendReject {
 		return
@@ -237,7 +93,7 @@ func (f *Interface) rejectInside(packet []byte, out []byte, q int) {
 
 	_, err := f.readers[q].Write(out)
 	if err != nil {
-		f.l.Error("Failed to write to tun", "error", err)
+		f.l.WithError(err).Error("Failed to write to tun")
 	}
 }
 
@@ -252,11 +108,11 @@ func (f *Interface) rejectOutside(packet []byte, ci *ConnectionState, hostinfo *
 	}
 
 	if len(out) > iputil.MaxRejectPacketSize {
-		if f.l.Enabled(context.Background(), slog.LevelInfo) {
-			f.l.Info("rejectOutside: packet too big, not sending",
-				"packet", packet,
-				"outPacket", out,
-			)
+		if f.l.GetLevel() >= logrus.InfoLevel {
+			f.l.
+				WithField("packet", packet).
+				WithField("outPacket", out).
+				Info("rejectOutside: packet too big, not sending")
 		}
 		return
 	}
@@ -328,11 +184,10 @@ func (f *Interface) getOrHandshakeConsiderRouting(fwPacket *firewall.Packet, cac
 		// This would also need to interact with unsafe_route updates through reloading the config or
 		// use of the use_system_route_table option
 
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("Calculated gateway for ECMP not available, attempting other gateways",
-				"destination", destinationAddr,
-				"originalGateway", gatewayAddr,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("destination", destinationAddr).
+				WithField("originalGateway", gatewayAddr).
+				Debugln("Calculated gateway for ECMP not available, attempting other gateways")
 		}
 
 		for i := range gateways {
@@ -356,20 +211,20 @@ func (f *Interface) getOrHandshakeConsiderRouting(fwPacket *firewall.Packet, cac
 
 func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubType, hostinfo *HostInfo, p, nb, out []byte) {
 	fp := &firewall.Packet{}
-	err := newPacket(p, false, fp)
+	ctx := &firewall.PacketContext{}
+	err := newPacket(p, false, fp, ctx)
 	if err != nil {
-		f.l.Warn("error while parsing outgoing packet for firewall check", "error", err)
+		f.l.Warnf("error while parsing outgoing packet for firewall check; %v", err)
 		return
 	}
 
 	// check if packet is in outbound fw rules
-	dropReason := f.firewall.Drop(*fp, false, hostinfo, f.pki.GetCAPool(), nil)
+	dropReason := f.firewall.Drop(*fp, *ctx, false, hostinfo, f.pki.GetCAPool(), nil)
 	if dropReason != nil {
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("dropping cached packet",
-				"fwPacket", fp,
-				"reason", dropReason,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("fwPacket", fp).
+				WithField("reason", dropReason).
+				Debugln("dropping cached packet")
 		}
 		return
 	}
@@ -385,10 +240,9 @@ func (f *Interface) SendMessageToVpnAddr(t header.MessageType, st header.Message
 	})
 
 	if hostInfo == nil {
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("dropping SendMessageToVpnAddr, vpnAddr not in our vpn networks or in unsafe routes",
-				"vpnAddr", vpnAddr,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("vpnAddr", vpnAddr).
+				Debugln("dropping SendMessageToVpnAddr, vpnAddr not in our vpn networks or in unsafe routes")
 		}
 		return
 	}
@@ -414,13 +268,21 @@ func (f *Interface) sendTo(t header.MessageType, st header.MessageSubType, ci *C
 	f.sendNoMetrics(t, st, ci, hostinfo, remote, p, nb, out, 0)
 }
 
-func (f *Interface) prepareSendVia(via *HostInfo,
+// SendVia sends a payload through a Relay tunnel. No authentication or encryption is done
+// to the payload for the ultimate target host, making this a useful method for sending
+// handshake messages to peers through relay tunnels.
+// via is the HostInfo through which the message is relayed.
+// ad is the plaintext data to authenticate, but not encrypt
+// nb is a buffer used to store the nonce value, re-used for performance reasons.
+// out is a buffer used to store the result of the Encrypt operation
+// q indicates which writer to use to send the packet.
+func (f *Interface) SendVia(via *HostInfo,
 	relay *Relay,
 	ad,
 	nb,
 	out []byte,
 	nocopy bool,
-) ([]byte, error) {
+) {
 	if noiseutil.EncryptLockNeeded {
 		// NOTE: for goboring AESGCMTLS we need to lock because of the nonce check
 		via.ConnectionState.writeLock.Lock()
@@ -436,13 +298,13 @@ func (f *Interface) prepareSendVia(via *HostInfo,
 		if noiseutil.EncryptLockNeeded {
 			via.ConnectionState.writeLock.Unlock()
 		}
-		via.logger(f.l).Error("SendVia out buffer not large enough for relay",
-			"outCap", cap(out),
-			"payloadLen", len(ad),
-			"headerLen", len(out),
-			"cipherOverhead", via.ConnectionState.eKey.Overhead(),
-		)
-		return nil, io.ErrShortBuffer
+		via.logger(f.l).
+			WithField("outCap", cap(out)).
+			WithField("payloadLen", len(ad)).
+			WithField("headerLen", len(out)).
+			WithField("cipherOverhead", via.ConnectionState.eKey.Overhead()).
+			Error("SendVia out buffer not large enough for relay")
+		return
 	}
 
 	// The header bytes are written to the 'out' slice; Grow the slice to hold the header and associated data payload.
@@ -461,37 +323,14 @@ func (f *Interface) prepareSendVia(via *HostInfo,
 		via.ConnectionState.writeLock.Unlock()
 	}
 	if err != nil {
-		via.logger(f.l).Info("Failed to EncryptDanger in sendVia", "error", err)
-		return nil, err
-	}
-	f.connectionManager.RelayUsed(relay.LocalIndex)
-	return out, nil
-}
-
-// SendVia sends a payload through a Relay tunnel. No authentication or encryption is done
-// to the payload for the ultimate target host, making this a useful method for sending
-// handshake messages to peers through relay tunnels.
-// via is the HostInfo through which the message is relayed.
-// ad is the plaintext data to authenticate, but not encrypt
-// nb is a buffer used to store the nonce value, re-used for performance reasons.
-// out is a buffer used to store the result of the Encrypt operation
-// q indicates which writer to use to send the packet.
-func (f *Interface) SendVia(via *HostInfo,
-	relay *Relay,
-	ad,
-	nb,
-	out []byte,
-	nocopy bool,
-) {
-	toSend, err := f.prepareSendVia(via, relay, ad, nb, out, nocopy)
-	if err != nil {
-		via.logger(f.l).Info("Failed to prepareSendVia", "error", err)
+		via.logger(f.l).WithError(err).Info("Failed to EncryptDanger in sendVia")
 		return
 	}
-	err = f.writers[0].WriteTo(toSend, via.remote)
+	err = f.writers[0].WriteTo(out, via.remote)
 	if err != nil {
-		via.logger(f.l).Info("Failed to WriteTo in sendVia", "error", err)
+		via.logger(f.l).WithError(err).Info("Failed to WriteTo in sendVia")
 	}
+	f.connectionManager.RelayUsed(relay.LocalIndex)
 }
 
 func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte, q int) {
@@ -528,10 +367,8 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 		// finally used again. This tunnel would eventually be torn down and recreated if this action didn't help.
 		f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
 		hostinfo.lastRebindCount = f.rebindCount
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("Lighthouse update triggered for punch due to rebind counter",
-				"vpnAddrs", hostinfo.vpnAddrs,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).Debug("Lighthouse update triggered for punch due to rebind counter")
 		}
 	}
 
@@ -541,30 +378,24 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 		ci.writeLock.Unlock()
 	}
 	if err != nil {
-		hostinfo.logger(f.l).Error("Failed to encrypt outgoing packet",
-			"error", err,
-			"udpAddr", remote,
-			"counter", c,
-			"attemptedCounter", c,
-		)
+		hostinfo.logger(f.l).WithError(err).
+			WithField("udpAddr", remote).WithField("counter", c).
+			WithField("attemptedCounter", c).
+			Error("Failed to encrypt outgoing packet")
 		return
 	}
 
 	if remote.IsValid() {
 		err = f.writers[q].WriteTo(out, remote)
 		if err != nil {
-			hostinfo.logger(f.l).Error("Failed to write outgoing packet",
-				"error", err,
-				"udpAddr", remote,
-			)
+			hostinfo.logger(f.l).WithError(err).
+				WithField("udpAddr", remote).Error("Failed to write outgoing packet")
 		}
 	} else if hostinfo.remote.IsValid() {
 		err = f.writers[q].WriteTo(out, hostinfo.remote)
 		if err != nil {
-			hostinfo.logger(f.l).Error("Failed to write outgoing packet",
-				"error", err,
-				"udpAddr", remote,
-			)
+			hostinfo.logger(f.l).WithError(err).
+				WithField("udpAddr", remote).Error("Failed to write outgoing packet")
 		}
 	} else {
 		// Try to send via a relay
@@ -572,10 +403,7 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 			relayHostInfo, relay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relayIP)
 			if err != nil {
 				hostinfo.relayState.DeleteRelay(relayIP)
-				hostinfo.logger(f.l).Info("sendNoMetrics failed to find HostInfo",
-					"relay", relayIP,
-					"error", err,
-				)
+				hostinfo.logger(f.l).WithField("relay", relayIP).WithError(err).Info("sendNoMetrics failed to find HostInfo")
 				continue
 			}
 			f.SendVia(relayHostInfo, relay, out, nb, fullOut[:header.Len+len(out)], true)

interface.go

@@ -4,7 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"log/slog"
+	"io"
 	"net/netip"
 	"sync"
 	"sync/atomic"
@@ -12,15 +12,11 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
-	"github.com/slackhq/nebula/util"
-	"github.com/slackhq/nebula/wire"
-
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
-	"github.com/slackhq/nebula/overlay/batch"
-	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/udp"
 )
 
@@ -40,7 +36,6 @@ type InterfaceConfig struct {
 	DropLocalBroadcast bool
 	DropMulticast      bool
 	routines           int
-	batchSize          int
 	MessageMetrics     *MessageMetrics
 	version            string
 	relayManager       *relayManager
@@ -51,7 +46,7 @@ type InterfaceConfig struct {
 	reQueryWait     time.Duration
 
 	ConntrackCacheTimeout time.Duration
-	l                     *slog.Logger
+	l                     *logrus.Logger
 }
 
 type Interface struct {
@@ -73,7 +68,6 @@ type Interface struct {
 	dropLocalBroadcast    bool
 	dropMulticast         bool
 	routines              int
-	batchSize             int
 	disconnectInvalid     atomic.Bool
 	closed                atomic.Bool
 	relayManager          *relayManager
@@ -93,12 +87,8 @@ type Interface struct {
 
 	ctx     context.Context
 	writers []udp.Conn
-	readers []tio.Queue
-	// batchers is one per tun queue, wrapping readers[i].
-	// decryptToTun sends plaintext into the batch.RxBatcher;
-	// listenOut calls its Flush at the end of each UDP recvmmsg batch.
-	batchers []batch.RxBatcher
-	wg       sync.WaitGroup
+	readers []io.ReadWriteCloser
+	wg      sync.WaitGroup
 
 	// fatalErr holds the first unexpected reader error that caused shutdown.
 	// nil means "no fatal error" (yet)
@@ -110,7 +100,7 @@ type Interface struct {
 	messageMetrics      *MessageMetrics
 	cachedPacketMetrics *cachedPacketMetrics
 
-	l *slog.Logger
+	l *logrus.Logger
 }
 
 type EncWriter interface {
@@ -194,11 +184,9 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 		dropLocalBroadcast:    c.DropLocalBroadcast,
 		dropMulticast:         c.DropMulticast,
 		routines:              c.routines,
-		batchSize:             c.batchSize,
 		version:               c.version,
 		writers:               make([]udp.Conn, c.routines),
-		readers:               make([]tio.Queue, c.routines),
-		batchers:              make([]batch.RxBatcher, c.routines),
+		readers:               make([]io.ReadWriteCloser, c.routines),
 		myVpnNetworks:         cs.myVpnNetworks,
 		myVpnNetworksTable:    cs.myVpnNetworksTable,
 		myVpnAddrs:            cs.myVpnAddrs,
@@ -235,16 +223,13 @@ func (f *Interface) activate() error {
 
 	addr, err := f.outside.LocalAddr()
 	if err != nil {
-		f.l.Error("Failed to get udp listen address", "error", err)
+		f.l.WithError(err).Error("Failed to get udp listen address")
 	}
 
-	f.l.Info("Nebula interface is active",
-		"interface", f.inside.Name(),
-		"networks", f.myVpnNetworks,
-		"build", f.version,
-		"udpAddr", addr,
-		"boringcrypto", boringEnabled(),
-	)
+	f.l.WithField("interface", f.inside.Name()).WithField("networks", f.myVpnNetworks).
+		WithField("build", f.version).WithField("udpAddr", addr).
+		WithField("boringcrypto", boringEnabled()).
+		Info("Nebula interface is active")
 
 	if f.routines > 1 {
 		if !f.inside.SupportsMultiqueue() || !f.outside.SupportsMultipleReaders() {
@@ -256,17 +241,15 @@ func (f *Interface) activate() error {
 	metrics.GetOrRegisterGauge("routines", nil).Update(int64(f.routines))
 
 	// Prepare n tun queues
+	var reader io.ReadWriteCloser = f.inside
 	for i := 0; i < f.routines; i++ {
 		if i > 0 {
-			if err = f.inside.NewMultiQueueReader(); err != nil {
+			reader, err = f.inside.NewMultiQueueReader()
+			if err != nil {
 				return err
 			}
 		}
-	}
-	f.readers = f.inside.Readers()
-	for i := range f.readers {
-		arena := util.NewArena(max(f.batchSize, 1) * udp.MTU)
-		f.batchers[i] = batch.NewPassthrough(f.readers[i], f.batchSize, arena)
+		f.readers[i] = reader
 	}
 
 	f.wg.Add(1) // for us to wait on Close() to return
@@ -322,64 +305,49 @@ func (f *Interface) listenOut(i int) {
 		li = f.outside
 	}
 
-	ctCache := firewall.NewConntrackCacheTicker(f.ctx, f.l, f.conntrackCacheTimeout)
+	ctCache := firewall.NewConntrackCacheTicker(f.ctx, f.conntrackCacheTimeout)
 	lhh := f.lightHouse.NewRequestHandler()
+	plaintext := make([]byte, udp.MTU)
 	h := &header.H{}
 	fwPacket := &firewall.Packet{}
+	fwCtx := &firewall.PacketContext{}
 	nb := make([]byte, 12, 12)
 
-	listener := func(fromUdpAddr netip.AddrPort, payload []byte, meta udp.RxMeta) {
-		f.readOutsidePackets(ViaSender{UdpAddr: fromUdpAddr}, payload, h, fwPacket, lhh, nb, i, ctCache.Get())
-	}
-
-	flusher := func() {
-		if err := f.batchers[i].Flush(); err != nil {
-			f.l.Error("Failed to flush tun coalescer", "error", err)
-		}
-	}
-
-	err := li.ListenOut(listener, flusher)
+	err := li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
+		f.readOutsidePackets(ViaSender{UdpAddr: fromUdpAddr}, plaintext[:0], payload, h, fwPacket, fwCtx, lhh, nb, i, ctCache.Get(f.l))
+	})
 
 	if err != nil && !f.closed.Load() {
-		f.l.Error("Error while reading inbound packet, closing", "error", err)
+		f.l.WithError(err).Error("Error while reading inbound packet, closing")
 		f.onFatal(err)
 	}
 
-	f.l.Debug("underlay reader is done", "reader", i)
+	f.l.Debugf("underlay reader %v is done", i)
 }
 
-func (f *Interface) listenIn(reader tio.Queue, q int) {
-	packetMem := make([]byte, mtu+16) //MTU + some leading slack space for platforms that return "bonus info"
-	// TODO get the amount of bonus info from the reader
-	packets := make([]wire.TunPacket, 1)
-	rejectBuf := make([]byte, mtu)
-	arenaSize := batch.SendBatchCap * (udp.MTU + 32)
-	sb := batch.NewSendBatch(f.writers[q], batch.SendBatchCap, util.NewArena(arenaSize))
+func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
+	packet := make([]byte, mtu)
+	out := make([]byte, mtu)
 	fwPacket := &firewall.Packet{}
+	fwCtx := &firewall.PacketContext{}
 	nb := make([]byte, 12, 12)
 
-	conntrackCache := firewall.NewConntrackCacheTicker(f.ctx, f.l, f.conntrackCacheTimeout)
+	conntrackCache := firewall.NewConntrackCacheTicker(f.ctx, f.conntrackCacheTimeout)
 
 	for {
-		n, err := reader.Read(packets, packetMem)
+		n, err := reader.Read(packet)
 		if err != nil {
 			if !f.closed.Load() {
-				f.l.Error("Error while reading outbound packet, closing", "error", err, "reader", q)
+				f.l.WithError(err).WithField("reader", i).Error("Error while reading outbound packet, closing")
 				f.onFatal(err)
 			}
 			break
 		}
 
-		ctCache := conntrackCache.Get()
-		for i := range n {
-			f.consumeInsidePacket(packets[i], fwPacket, nb, sb, rejectBuf, q, ctCache)
-		}
-		if err := sb.Flush(); err != nil {
-			f.l.Error("Failed to write outgoing batch", "error", err, "writer", q)
-		}
+		f.consumeInsidePacket(packet[:n], fwPacket, fwCtx, nb, out, i, conntrackCache.Get(f.l))
 	}
 
-	f.l.Debug("overlay reader is done", "reader", q)
+	f.l.Debugf("overlay reader %v is done", i)
 }
 
 func (f *Interface) RegisterConfigChangeCallbacks(c *config.C) {
@@ -399,7 +367,7 @@ func (f *Interface) reloadDisconnectInvalid(c *config.C) {
 	if initial || c.HasChanged("pki.disconnect_invalid") {
 		f.disconnectInvalid.Store(c.GetBool("pki.disconnect_invalid", true))
 		if !initial {
-			f.l.Info("pki.disconnect_invalid changed", "value", f.disconnectInvalid.Load())
+			f.l.Infof("pki.disconnect_invalid changed to %v", f.disconnectInvalid.Load())
 		}
 	}
 }
@@ -413,7 +381,7 @@ func (f *Interface) reloadFirewall(c *config.C) {
 
 	fw, err := NewFirewallFromConfig(f.l, f.pki.getCertState(), c)
 	if err != nil {
-		f.l.Error("Error while creating firewall during reload", "error", err)
+		f.l.WithError(err).Error("Error while creating firewall during reload")
 		return
 	}
 
@@ -426,23 +394,28 @@ func (f *Interface) reloadFirewall(c *config.C) {
 	// If rulesVersion is back to zero, we have wrapped all the way around. Be
 	// safe and just reset conntrack in this case.
 	if fw.rulesVersion == 0 {
-		f.l.Warn("firewall rulesVersion has overflowed, resetting conntrack",
-			"firewallHashes", fw.GetRuleHashes(),
-			"oldFirewallHashes", oldFw.GetRuleHashes(),
-			"rulesVersion", fw.rulesVersion,
-		)
+		f.l.WithField("firewallHashes", fw.GetRuleHashes()).
+			WithField("oldFirewallHashes", oldFw.GetRuleHashes()).
+			WithField("rulesVersion", fw.rulesVersion).
+			Warn("firewall rulesVersion has overflowed, resetting conntrack")
 	} else {
 		fw.Conntrack = conntrack
 	}
 
+	fw.reporter = oldFw.reporter
+
 	f.firewall = fw
 
+	// Fire ReportRulesReload under the conntrack lock so the reporter cannot
+	// observe a FlowCreate/FlowEvict for the new rulesVersion before it
+	// observes the reload marker. Report* must be non-blocking.
+	fw.reportRulesReload(oldFw.rulesVersion, fw.rulesVersion)
+
 	oldFw.Destroy()
-	f.l.Info("New firewall has been installed",
-		"firewallHashes", fw.GetRuleHashes(),
-		"oldFirewallHashes", oldFw.GetRuleHashes(),
-		"rulesVersion", fw.rulesVersion,
-	)
+	f.l.WithField("firewallHashes", fw.GetRuleHashes()).
+		WithField("oldFirewallHashes", oldFw.GetRuleHashes()).
+		WithField("rulesVersion", fw.rulesVersion).
+		Info("New firewall has been installed")
 }
 
 func (f *Interface) reloadSendRecvError(c *config.C) {
@@ -464,7 +437,8 @@ func (f *Interface) reloadSendRecvError(c *config.C) {
 			}
 		}
 
-		f.l.Info("Loaded send_recv_error config", "sendRecvError", f.sendRecvErrorConfig.String())
+		f.l.WithField("sendRecvError", f.sendRecvErrorConfig.String()).
+			Info("Loaded send_recv_error config")
 	}
 }
 
@@ -487,7 +461,8 @@ func (f *Interface) reloadAcceptRecvError(c *config.C) {
 			}
 		}
 
-		f.l.Info("Loaded accept_recv_error config", "acceptRecvError", f.acceptRecvErrorConfig.String())
+		f.l.WithField("acceptRecvError", f.acceptRecvErrorConfig.String()).
+			Info("Loaded accept_recv_error config")
 	}
 }
 
@@ -561,7 +536,7 @@ func (f *Interface) Close() error {
 	for i, u := range f.writers {
 		err := u.Close()
 		if err != nil {
-			f.l.Error("Error while closing udp socket", "error", err, "writer", i)
+			f.l.WithError(err).WithField("writer", i).Error("Error while closing udp socket")
 			errs = append(errs, err)
 		}
 	}

lighthouse.go

@@ -5,7 +5,6 @@ import (
 	"encoding/binary"
 	"errors"
 	"fmt"
-	"log/slog"
 	"net"
 	"net/netip"
 	"slices"
@@ -15,10 +14,11 @@ import (
 	"time"
 
 	"github.com/gaissmai/bart"
+	"github.com/rcrowley/go-metrics"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
-	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/udp"
 	"github.com/slackhq/nebula/util"
 )
@@ -34,6 +34,7 @@ type LightHouse struct {
 
 	myVpnNetworks      []netip.Prefix
 	myVpnNetworksTable *bart.Lite
+	punchConn          udp.Conn
 	punchy             *Punchy
 
 	// Local cache of answers from light houses
@@ -68,18 +69,18 @@ type LightHouse struct {
 	// Addr's of relays that can be used by peers to access me
 	relaysForMe atomic.Pointer[[]netip.Addr]
 
-	updateTrigger chan struct{}
-	queryChan     chan netip.Addr
+	queryChan chan netip.Addr
 
 	calculatedRemotes atomic.Pointer[bart.Table[[]*calculatedRemote]] // Maps VpnAddr to []*calculatedRemote
 
-	metrics *MessageMetrics
-	l       *slog.Logger
+	metrics           *MessageMetrics
+	metricHolepunchTx metrics.Counter
+	l                 *logrus.Logger
 }
 
 // NewLightHouseFromConfig will build a Lighthouse struct from the values provided in the config object
 // addrMap should be nil unless this is during a config reload
-func NewLightHouseFromConfig(ctx context.Context, l *slog.Logger, c *config.C, cs *CertState, pc udp.Conn, p *Punchy) (*LightHouse, error) {
+func NewLightHouseFromConfig(ctx context.Context, l *logrus.Logger, c *config.C, cs *CertState, pc udp.Conn, p *Punchy) (*LightHouse, error) {
 	amLighthouse := c.GetBool("lighthouse.am_lighthouse", false)
 	nebulaPort := uint32(c.GetInt("listen.port", 0))
 	if amLighthouse && nebulaPort == 0 {
@@ -102,8 +103,8 @@ func NewLightHouseFromConfig(ctx context.Context, l *slog.Logger, c *config.C, c
 		myVpnNetworksTable: cs.myVpnNetworksTable,
 		addrMap:            make(map[netip.Addr]*RemoteList),
 		nebulaPort:         nebulaPort,
+		punchConn:          pc,
 		punchy:             p,
-		updateTrigger:      make(chan struct{}, 1),
 		queryChan:          make(chan netip.Addr, c.GetUint32("handshakes.query_buffer", 64)),
 		l:                  l,
 	}
@@ -114,6 +115,9 @@ func NewLightHouseFromConfig(ctx context.Context, l *slog.Logger, c *config.C, c
 
 	if c.GetBool("stats.lighthouse_metrics", false) {
 		h.metrics = newLighthouseMetrics()
+		h.metricHolepunchTx = metrics.GetOrRegisterCounter("messages.tx.holepunch", nil)
+	} else {
+		h.metricHolepunchTx = metrics.NilCounter{}
 	}
 
 	err := h.reload(c, true)
@@ -127,7 +131,7 @@ func NewLightHouseFromConfig(ctx context.Context, l *slog.Logger, c *config.C, c
 		case *util.ContextualError:
 			v.Log(l)
 		case error:
-			l.Error("failed to reload lighthouse", "error", err)
+			l.WithError(err).Error("failed to reload lighthouse")
 		}
 	})
 
@@ -199,10 +203,8 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 			//TODO: we could technically insert all returned addrs instead of just the first one if a dns lookup was used
 			addr := addrs[0].Unmap()
 			if lh.myVpnNetworksTable.Contains(addr) {
-				lh.l.Warn("Ignoring lighthouse.advertise_addrs report because it is within the nebula network range",
-					"addr", rawAddr,
-					"entry", i+1,
-				)
+				lh.l.WithField("addr", rawAddr).WithField("entry", i+1).
+					Warn("Ignoring lighthouse.advertise_addrs report because it is within the nebula network range")
 				continue
 			}
 
@@ -220,9 +222,7 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 		lh.interval.Store(int64(c.GetInt("lighthouse.interval", 10)))
 
 		if !initial {
-			lh.l.Info("lighthouse.interval changed",
-				"interval", lh.interval.Load(),
-			)
+			lh.l.Infof("lighthouse.interval changed to %v", lh.interval.Load())
 
 			if lh.updateCancel != nil {
 				// May not always have a running routine
@@ -316,7 +316,6 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 		if !initial {
 			//NOTE: we are not tearing down existing lighthouse connections because they might be used for non lighthouse traffic
 			lh.l.Info("lighthouse.hosts has changed")
-			lh.TriggerUpdate()
 		}
 	}
 
@@ -334,12 +333,9 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 			for _, v := range c.GetStringSlice("relay.relays", nil) {
 				configRIP, err := netip.ParseAddr(v)
 				if err != nil {
-					lh.l.Warn("Parse relay from config failed",
-						"relay", v,
-						"error", err,
-					)
+					lh.l.WithField("relay", v).WithError(err).Warn("Parse relay from config failed")
 				} else {
-					lh.l.Info("Read relay from config", "relay", v)
+					lh.l.WithField("relay", v).Info("Read relay from config")
 					relaysForMe = append(relaysForMe, configRIP)
 				}
 			}
@@ -364,10 +360,8 @@ func (lh *LightHouse) parseLighthouses(c *config.C) ([]netip.Addr, error) {
 		}
 
 		if !lh.myVpnNetworksTable.Contains(addr) {
-			lh.l.Warn("lighthouse host is not within our networks, lighthouse functionality will work but layer 3 network traffic to the lighthouse will not",
-				"vpnAddr", addr,
-				"networks", lh.myVpnNetworks,
-			)
+			lh.l.WithFields(m{"vpnAddr": addr, "networks": lh.myVpnNetworks}).
+				Warn("lighthouse host is not within our networks, lighthouse functionality will work but layer 3 network traffic to the lighthouse will not")
 		}
 		out[i] = addr
 	}
@@ -438,11 +432,8 @@ func (lh *LightHouse) loadStaticMap(c *config.C, staticList map[netip.Addr]struc
 		}
 
 		if !lh.myVpnNetworksTable.Contains(vpnAddr) {
-			lh.l.Warn("static_host_map key is not within our networks, layer 3 network traffic to this host will not work",
-				"vpnAddr", vpnAddr,
-				"networks", lh.myVpnNetworks,
-				"entry", i+1,
-			)
+			lh.l.WithFields(m{"vpnAddr": vpnAddr, "networks": lh.myVpnNetworks, "entry": i + 1}).
+				Warn("static_host_map key is not within our networks, layer 3 network traffic to this host will not work")
 		}
 
 		vals, ok := v.([]any)
@@ -543,13 +534,12 @@ func (lh *LightHouse) DeleteVpnAddrs(allVpnAddrs []netip.Addr) {
 	lh.Lock()
 	rm, ok := lh.addrMap[allVpnAddrs[0]]
 	if ok {
-		debugEnabled := lh.l.Enabled(context.Background(), slog.LevelDebug)
 		for _, addr := range allVpnAddrs {
 			srm := lh.addrMap[addr]
 			if srm == rm {
 				delete(lh.addrMap, addr)
-				if debugEnabled {
-					lh.l.Debug("deleting from lighthouse", "vpnAddr", addr)
+				if lh.l.Level >= logrus.DebugLevel {
+					lh.l.Debugf("deleting %s from lighthouse.", addr)
 				}
 			}
 		}
@@ -666,12 +656,9 @@ func (lh *LightHouse) unlockedGetRemoteList(allAddrs []netip.Addr) *RemoteList {
 
 func (lh *LightHouse) shouldAdd(vpnAddrs []netip.Addr, to netip.Addr) bool {
 	allow := lh.GetRemoteAllowList().AllowAll(vpnAddrs, to)
-	if lh.l.Enabled(context.Background(), logging.LevelTrace) {
-		lh.l.Log(context.Background(), logging.LevelTrace, "remoteAllowList.Allow",
-			"vpnAddrs", vpnAddrs,
-			"udpAddr", to,
-			"allow", allow,
-		)
+	if lh.l.Level >= logrus.TraceLevel {
+		lh.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", to).WithField("allow", allow).
+			Trace("remoteAllowList.Allow")
 	}
 	if !allow {
 		return false
@@ -688,12 +675,9 @@ func (lh *LightHouse) shouldAdd(vpnAddrs []netip.Addr, to netip.Addr) bool {
 func (lh *LightHouse) unlockedShouldAddV4(vpnAddr netip.Addr, to *V4AddrPort) bool {
 	udpAddr := protoV4AddrPortToNetAddrPort(to)
 	allow := lh.GetRemoteAllowList().Allow(vpnAddr, udpAddr.Addr())
-	if lh.l.Enabled(context.Background(), logging.LevelTrace) {
-		lh.l.Log(context.Background(), logging.LevelTrace, "remoteAllowList.Allow",
-			"vpnAddr", vpnAddr,
-			"udpAddr", udpAddr,
-			"allow", allow,
-		)
+	if lh.l.Level >= logrus.TraceLevel {
+		lh.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", udpAddr).WithField("allow", allow).
+			Trace("remoteAllowList.Allow")
 	}
 
 	if !allow {
@@ -711,12 +695,9 @@ func (lh *LightHouse) unlockedShouldAddV4(vpnAddr netip.Addr, to *V4AddrPort) bo
 func (lh *LightHouse) unlockedShouldAddV6(vpnAddr netip.Addr, to *V6AddrPort) bool {
 	udpAddr := protoV6AddrPortToNetAddrPort(to)
 	allow := lh.GetRemoteAllowList().Allow(vpnAddr, udpAddr.Addr())
-	if lh.l.Enabled(context.Background(), logging.LevelTrace) {
-		lh.l.Log(context.Background(), logging.LevelTrace, "remoteAllowList.Allow",
-			"vpnAddr", vpnAddr,
-			"udpAddr", udpAddr,
-			"allow", allow,
-		)
+	if lh.l.Level >= logrus.TraceLevel {
+		lh.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", udpAddr).WithField("allow", allow).
+			Trace("remoteAllowList.Allow")
 	}
 
 	if !allow {
@@ -791,10 +772,8 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 
 		if v == cert.Version1 {
 			if !addr.Is4() {
-				lh.l.Error("Can't query lighthouse for v6 address using a v1 protocol",
-					"queryVpnAddr", addr,
-					"lighthouseAddr", lhVpnAddr,
-				)
+				lh.l.WithField("queryVpnAddr", addr).WithField("lighthouseAddr", lhVpnAddr).
+					Error("Can't query lighthouse for v6 address using a v1 protocol")
 				continue
 			}
 
@@ -805,11 +784,9 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 
 				v1Query, err = msg.Marshal()
 				if err != nil {
-					lh.l.Error("Failed to marshal lighthouse v1 query payload",
-						"error", err,
-						"queryVpnAddr", addr,
-						"lighthouseAddr", lhVpnAddr,
-					)
+					lh.l.WithError(err).WithField("queryVpnAddr", addr).
+						WithField("lighthouseAddr", lhVpnAddr).
+						Error("Failed to marshal lighthouse v1 query payload")
 					continue
 				}
 			}
@@ -824,11 +801,9 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 
 				v2Query, err = msg.Marshal()
 				if err != nil {
-					lh.l.Error("Failed to marshal lighthouse v2 query payload",
-						"error", err,
-						"queryVpnAddr", addr,
-						"lighthouseAddr", lhVpnAddr,
-					)
+					lh.l.WithError(err).WithField("queryVpnAddr", addr).
+						WithField("lighthouseAddr", lhVpnAddr).
+						Error("Failed to marshal lighthouse v2 query payload")
 					continue
 				}
 			}
@@ -837,11 +812,7 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 			queried++
 
 		} else {
-			lh.l.Debug("unsupported protocol version",
-				"op", "query",
-				"queryVpnAddr", addr,
-				"version", v,
-			)
+			lh.l.Debugf("Can not query lighthouse for %v using unknown protocol version: %v", addr, v)
 			continue
 		}
 	}
@@ -870,24 +841,11 @@ func (lh *LightHouse) StartUpdateWorker() {
 				return
 			case <-clockSource.C:
 				continue
-			case <-lh.updateTrigger:
-				continue
 			}
 		}
 	}()
 }
 
-// TriggerUpdate requests an immediate lighthouse update. This is a non-blocking
-// operation intended to be called after a handshake completes with a lighthouse,
-// so the lighthouse has our current addresses without waiting for the next
-// periodic update.
-func (lh *LightHouse) TriggerUpdate() {
-	select {
-	case lh.updateTrigger <- struct{}{}:
-	default:
-	}
-}
-
 func (lh *LightHouse) SendUpdate() {
 	var v4 []*V4AddrPort
 	var v6 []*V6AddrPort
@@ -933,9 +891,8 @@ func (lh *LightHouse) SendUpdate() {
 		if v == cert.Version1 {
 			if v1Update == nil {
 				if !lh.myVpnNetworks[0].Addr().Is4() {
-					lh.l.Warn("cannot update lighthouse using v1 protocol without an IPv4 address",
-						"lighthouseAddr", lhVpnAddr,
-					)
+					lh.l.WithField("lighthouseAddr", lhVpnAddr).
+						Warn("cannot update lighthouse using v1 protocol without an IPv4 address")
 					continue
 				}
 				var relays []uint32
@@ -959,10 +916,8 @@ func (lh *LightHouse) SendUpdate() {
 
 				v1Update, err = msg.Marshal()
 				if err != nil {
-					lh.l.Error("Error while marshaling for lighthouse v1 update",
-						"error", err,
-						"lighthouseAddr", lhVpnAddr,
-					)
+					lh.l.WithError(err).WithField("lighthouseAddr", lhVpnAddr).
+						Error("Error while marshaling for lighthouse v1 update")
 					continue
 				}
 			}
@@ -988,10 +943,8 @@ func (lh *LightHouse) SendUpdate() {
 
 				v2Update, err = msg.Marshal()
 				if err != nil {
-					lh.l.Error("Error while marshaling for lighthouse v2 update",
-						"error", err,
-						"lighthouseAddr", lhVpnAddr,
-					)
+					lh.l.WithError(err).WithField("lighthouseAddr", lhVpnAddr).
+						Error("Error while marshaling for lighthouse v2 update")
 					continue
 				}
 			}
@@ -1000,10 +953,7 @@ func (lh *LightHouse) SendUpdate() {
 			updated++
 
 		} else {
-			lh.l.Debug("unsupported protocol version",
-				"op", "update",
-				"version", v,
-			)
+			lh.l.Debugf("Can not update lighthouse using unknown protocol version: %v", v)
 			continue
 		}
 	}
@@ -1017,7 +967,7 @@ type LightHouseHandler struct {
 	out  []byte
 	pb   []byte
 	meta *NebulaMeta
-	l    *slog.Logger
+	l    *logrus.Logger
 }
 
 func (lh *LightHouse) NewRequestHandler() *LightHouseHandler {
@@ -1066,19 +1016,14 @@ func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, fromVpnAddrs [
 	n := lhh.resetMeta()
 	err := n.Unmarshal(p)
 	if err != nil {
-		lhh.l.Error("Failed to unmarshal lighthouse packet",
-			"error", err,
-			"vpnAddrs", fromVpnAddrs,
-			"udpAddr", rAddr,
-		)
+		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).WithField("udpAddr", rAddr).
+			Error("Failed to unmarshal lighthouse packet")
 		return
 	}
 
 	if n.Details == nil {
-		lhh.l.Error("Invalid lighthouse update",
-			"vpnAddrs", fromVpnAddrs,
-			"udpAddr", rAddr,
-		)
+		lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("udpAddr", rAddr).
+			Error("Invalid lighthouse update")
 		return
 	}
 
@@ -1106,29 +1051,25 @@ func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, fromVpnAddrs [
 func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []netip.Addr, addr netip.AddrPort, w EncWriter) {
 	// Exit if we don't answer queries
 	if !lhh.lh.amLighthouse {
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debug("I don't answer queries, but received one", "from", addr)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.Debugln("I don't answer queries, but received from: ", addr)
 		}
 		return
 	}
 
 	queryVpnAddr, useVersion, err := n.Details.GetVpnAddrAndVersion()
 	if err != nil {
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debug("Dropping malformed HostQuery",
-				"from", fromVpnAddrs,
-				"details", n.Details,
-			)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithField("from", fromVpnAddrs).WithField("details", n.Details).
+				Debugln("Dropping malformed HostQuery")
 		}
 		return
 	}
 	if useVersion == cert.Version1 && queryVpnAddr.Is6() {
 		// this case really shouldn't be possible to represent, but reject it anyway.
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debug("invalid vpn addr for v1 handleHostQuery",
-				"vpnAddrs", fromVpnAddrs,
-				"queryVpnAddr", queryVpnAddr,
-			)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("queryVpnAddr", queryVpnAddr).
+				Debugln("invalid vpn addr for v1 handleHostQuery")
 		}
 		return
 	}
@@ -1153,10 +1094,7 @@ func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []neti
 	}
 
 	if err != nil {
-		lhh.l.Error("Failed to marshal lighthouse host query reply",
-			"error", err,
-			"vpnAddrs", fromVpnAddrs,
-		)
+		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host query reply")
 		return
 	}
 
@@ -1184,10 +1122,8 @@ func (lhh *LightHouseHandler) sendHostPunchNotification(n *NebulaMeta, fromVpnAd
 			if ok {
 				whereToPunch = newDest
 			} else {
-				if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-					lhh.l.Debug("unable to punch to host, no addresses in common",
-						"to", crt.Networks(),
-					)
+				if lhh.l.Level >= logrus.DebugLevel {
+					lhh.l.WithField("to", crt.Networks()).Debugln("unable to punch to host, no addresses in common")
 				}
 			}
 		}
@@ -1213,10 +1149,7 @@ func (lhh *LightHouseHandler) sendHostPunchNotification(n *NebulaMeta, fromVpnAd
 	}
 
 	if err != nil {
-		lhh.l.Error("Failed to marshal lighthouse host was queried for",
-			"error", err,
-			"vpnAddrs", fromVpnAddrs,
-		)
+		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host was queried for")
 		return
 	}
 
@@ -1258,11 +1191,8 @@ func (lhh *LightHouseHandler) coalesceAnswers(v cert.Version, c *cache, n *Nebul
 				n.Details.RelayVpnAddrs = append(n.Details.RelayVpnAddrs, netAddrToProtoAddr(r))
 			}
 		} else {
-			if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-				lhh.l.Debug("unsupported protocol version",
-					"op", "coalesceAnswers",
-					"version", v,
-				)
+			if lhh.l.Level >= logrus.DebugLevel {
+				lhh.l.WithField("version", v).Debug("unsupported protocol version")
 			}
 		}
 	}
@@ -1275,11 +1205,8 @@ func (lhh *LightHouseHandler) handleHostQueryReply(n *NebulaMeta, fromVpnAddrs [
 
 	certVpnAddr, _, err := n.Details.GetVpnAddrAndVersion()
 	if err != nil {
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Error("dropping malformed HostQueryReply",
-				"error", err,
-				"vpnAddrs", fromVpnAddrs,
-			)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("dropping malformed HostQueryReply")
 		}
 		return
 	}
@@ -1304,8 +1231,8 @@ func (lhh *LightHouseHandler) handleHostQueryReply(n *NebulaMeta, fromVpnAddrs [
 
 func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVpnAddrs []netip.Addr, w EncWriter) {
 	if !lhh.lh.amLighthouse {
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debug("I am not a lighthouse, do not take host updates", "from", fromVpnAddrs)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.Debugln("I am not a lighthouse, do not take host updates: ", fromVpnAddrs)
 		}
 		return
 	}
@@ -1328,11 +1255,8 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 
 	//Simple check that the host sent this not someone else, if detailsVpnAddr is filled
 	if detailsVpnAddr.IsValid() && !slices.Contains(fromVpnAddrs, detailsVpnAddr) {
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debug("Host sent invalid update",
-				"vpnAddrs", fromVpnAddrs,
-				"answer", detailsVpnAddr,
-			)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("answer", detailsVpnAddr).Debugln("Host sent invalid update")
 		}
 		return
 	}
@@ -1354,9 +1278,7 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 	switch useVersion {
 	case cert.Version1:
 		if !fromVpnAddrs[0].Is4() {
-			lhh.l.Error("Can not send HostUpdateNotificationAck for a ipv6 vpn ip in a v1 message",
-				"vpnAddrs", fromVpnAddrs,
-			)
+			lhh.l.WithField("vpnAddrs", fromVpnAddrs).Error("Can not send HostUpdateNotificationAck for a ipv6 vpn ip in a v1 message")
 			return
 		}
 		vpnAddrB := fromVpnAddrs[0].As4()
@@ -1364,16 +1286,13 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 	case cert.Version2:
 		// do nothing, we want to send a blank message
 	default:
-		lhh.l.Error("invalid protocol version", "useVersion", useVersion)
+		lhh.l.WithField("useVersion", useVersion).Error("invalid protocol version")
 		return
 	}
 
 	ln, err := n.MarshalTo(lhh.pb)
 	if err != nil {
-		lhh.l.Error("Failed to marshal lighthouse host update ack",
-			"error", err,
-			"vpnAddrs", fromVpnAddrs,
-		)
+		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host update ack")
 		return
 	}
 
@@ -1390,34 +1309,59 @@ func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpn
 
 	detailsVpnAddr, _, err := n.Details.GetVpnAddrAndVersion()
 	if err != nil {
-		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debug("dropping invalid HostPunchNotification",
-				"details", n.Details,
-				"error", err,
-			)
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.WithField("details", n.Details).WithError(err).Debugln("dropping invalid HostPunchNotification")
 		}
 		return
 	}
 
+	empty := []byte{0}
+	punch := func(vpnPeer netip.AddrPort, logVpnAddr netip.Addr) {
+		if !vpnPeer.IsValid() {
+			return
+		}
+
+		go func() {
+			time.Sleep(lhh.lh.punchy.GetDelay())
+			lhh.lh.metricHolepunchTx.Inc(1)
+			lhh.lh.punchConn.WriteTo(empty, vpnPeer)
+		}()
+
+		if lhh.l.Level >= logrus.DebugLevel {
+			lhh.l.Debugf("Punching on %v for %v", vpnPeer, logVpnAddr)
+		}
+	}
+
 	remoteAllowList := lhh.lh.GetRemoteAllowList()
 	for _, a := range n.Details.V4AddrPorts {
 		b := protoV4AddrPortToNetAddrPort(a)
 		if remoteAllowList.Allow(detailsVpnAddr, b.Addr()) {
-			lhh.lh.punchy.Schedule(b, detailsVpnAddr)
+			punch(b, detailsVpnAddr)
 		}
 	}
 
 	for _, a := range n.Details.V6AddrPorts {
 		b := protoV6AddrPortToNetAddrPort(a)
 		if remoteAllowList.Allow(detailsVpnAddr, b.Addr()) {
-			lhh.lh.punchy.Schedule(b, detailsVpnAddr)
+			punch(b, detailsVpnAddr)
 		}
 	}
 
 	// This sends a nebula test packet to the host trying to contact us. In the case
 	// of a double nat or other difficult scenario, this may help establish
-	// a tunnel. ScheduleRespond is a no-op when punchy.respond is disabled.
-	lhh.lh.punchy.ScheduleRespond(detailsVpnAddr)
+	// a tunnel.
+	if lhh.lh.punchy.GetRespond() {
+		go func() {
+			time.Sleep(lhh.lh.punchy.GetRespondDelay())
+			if lhh.l.Level >= logrus.DebugLevel {
+				lhh.l.Debugf("Sending a nebula test packet to vpn addr %s", detailsVpnAddr)
+			}
+			//NOTE: we have to allocate a new output buffer here since we are spawning a new goroutine
+			// for each punchBack packet. We should move this into a timerwheel or a single goroutine
+			// managed by a channel.
+			w.SendMessageToVpnAddr(header.Test, header.TestRequest, detailsVpnAddr, []byte(""), make([]byte, 12, 12), make([]byte, mtu))
+		}()
+	}
 }
 
 func protoAddrToNetAddr(addr *Addr) netip.Addr {

logger.go

@@ -0,0 +1,45 @@
+package nebula
+
+import (
+	"fmt"
+	"strings"
+	"time"
+
+	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula/config"
+)
+
+func configLogger(l *logrus.Logger, c *config.C) error {
+	// set up our logging level
+	logLevel, err := logrus.ParseLevel(strings.ToLower(c.GetString("logging.level", "info")))
+	if err != nil {
+		return fmt.Errorf("%s; possible levels: %s", err, logrus.AllLevels)
+	}
+	l.SetLevel(logLevel)
+
+	disableTimestamp := c.GetBool("logging.disable_timestamp", false)
+	timestampFormat := c.GetString("logging.timestamp_format", "")
+	fullTimestamp := (timestampFormat != "")
+	if timestampFormat == "" {
+		timestampFormat = time.RFC3339
+	}
+
+	logFormat := strings.ToLower(c.GetString("logging.format", "text"))
+	switch logFormat {
+	case "text":
+		l.Formatter = &logrus.TextFormatter{
+			TimestampFormat:  timestampFormat,
+			FullTimestamp:    fullTimestamp,
+			DisableTimestamp: disableTimestamp,
+		}
+	case "json":
+		l.Formatter = &logrus.JSONFormatter{
+			TimestampFormat:  timestampFormat,
+			DisableTimestamp: disableTimestamp,
+		}
+	default:
+		return fmt.Errorf("unknown log format `%s`. possible formats: %s", logFormat, []string{"text", "json"})
+	}
+
+	return nil
+}

logging/logger.go

@@ -1,233 +0,0 @@
-// Package logging wires the nebula runtime-reconfigurable slog handler used
-// by nebula.Main and the nebula CLI binaries. Callers build a logger with
-// NewLogger, then call ApplyConfig at startup and from a config reload
-// callback to push logging.level, logging.format, and
-// logging.disable_timestamp changes onto the logger without rebuilding it.
-package logging
-
-import (
-	"context"
-	"fmt"
-	"io"
-	"log/slog"
-	"strings"
-	"sync/atomic"
-	"time"
-)
-
-// Config is the subset of *config.C that ApplyConfig reads. Declaring it
-// here keeps the logging package from depending on config directly, which
-// would cycle through the shared test helpers (test.NewLogger imports
-// logging, and config's tests import test). *config.C satisfies this
-// interface structurally with no adapter.
-type Config interface {
-	GetString(key, def string) string
-	GetBool(key string, def bool) bool
-}
-
-// LevelTrace is a custom slog level below Debug, used when logging.level is
-// "trace". slog has no builtin trace level; the value is one step below
-// slog.LevelDebug in slog's 4-point spacing.
-const LevelTrace = slog.Level(-8)
-
-// NewLogger returns a *slog.Logger whose level, format, and timestamp
-// emission can be reconfigured at runtime via ApplyConfig and the SSH debug
-// commands. The default configuration is info-level text output so log
-// calls made before ApplyConfig runs still produce output. Timestamps
-// follow slog's default RFC3339Nano format; set logging.disable_timestamp
-// in config to suppress them.
-//
-// ApplyConfig and the SSH commands discover the reconfig surface via
-// structural type-assertion on l.Handler(), so replacement implementations
-// (tests, platform-specific sinks) need only implement the subset of
-// {SetLevel(slog.Level), SetFormat(string) error, SetDisableTimestamp(bool)}
-// they care about. Callers that pass a plain *slog.Logger without these
-// methods get a silent no-op; reconfiguration is always opt-in.
-func NewLogger(w io.Writer) *slog.Logger {
-	return slog.New(NewHandler(w))
-}
-
-// NewHandler builds the *Handler that NewLogger wraps. Exported for
-// platform-specific sinks (notably cmd/nebula-service/logs_windows.go)
-// that want to wrap the handler with extra behavior, such as tagging each
-// record with its Event Log severity, while still benefiting from all the
-// level / format / timestamp / WithAttrs machinery implemented here.
-func NewHandler(w io.Writer) *Handler {
-	root := &handlerRoot{}
-	root.level.Set(slog.LevelInfo)
-	opts := &slog.HandlerOptions{Level: &root.level}
-	return &Handler{
-		root: root,
-		text: slog.NewTextHandler(w, opts),
-		json: slog.NewJSONHandler(w, opts),
-	}
-}
-
-// handlerRoot carries the reconfiguration state shared by every logger
-// derived from a NewHandler call. All fields are consulted on the log
-// path and updated lock-free.
-type handlerRoot struct {
-	level            slog.LevelVar
-	disableTimestamp atomic.Bool
-	// jsonMode picks which of the pre-derived inner handlers Handler.Handle
-	// dispatches to. Flipping it propagates instantly to every derived logger
-	// without rebuilding or chain-replaying anything.
-	jsonMode atomic.Bool
-}
-
-// Handler is the slog.Handler returned by NewHandler. It holds two
-// pre-derived slog handlers -- one text, one json -- both built from the
-// same accumulated WithAttrs/WithGroup state. Handle picks which one to
-// dispatch to based on handlerRoot.jsonMode, so a SetFormat call takes
-// effect immediately across the whole process without having to rebuild
-// any derived loggers.
-type Handler struct {
-	root *handlerRoot
-	text slog.Handler
-	json slog.Handler
-}
-
-func (h *Handler) Enabled(_ context.Context, l slog.Level) bool {
-	return h.root.level.Level() <= l
-}
-
-func (h *Handler) Handle(ctx context.Context, r slog.Record) error {
-	if h.root.disableTimestamp.Load() {
-		r.Time = time.Time{}
-	}
-	if h.root.jsonMode.Load() {
-		return h.json.Handle(ctx, r)
-	}
-	return h.text.Handle(ctx, r)
-}
-
-func (h *Handler) WithAttrs(attrs []slog.Attr) slog.Handler {
-	if len(attrs) == 0 {
-		return h
-	}
-	return &Handler{
-		root: h.root,
-		text: h.text.WithAttrs(attrs),
-		json: h.json.WithAttrs(attrs),
-	}
-}
-
-func (h *Handler) WithGroup(name string) slog.Handler {
-	if name == "" {
-		return h
-	}
-	return &Handler{
-		root: h.root,
-		text: h.text.WithGroup(name),
-		json: h.json.WithGroup(name),
-	}
-}
-
-// SetLevel updates the effective log level. Propagates to every derived
-// logger via the shared LevelVar.
-func (h *Handler) SetLevel(level slog.Level) { h.root.level.Set(level) }
-
-// GetLevel reports the current log level.
-func (h *Handler) GetLevel() slog.Level { return h.root.level.Level() }
-
-// SetFormat flips the output format atomically. Valid formats are "text"
-// and "json". Every derived logger sees the new format on its next Handle
-// call; no rebuild or registration is required.
-func (h *Handler) SetFormat(format string) error {
-	switch format {
-	case "text":
-		h.root.jsonMode.Store(false)
-	case "json":
-		h.root.jsonMode.Store(true)
-	default:
-		return fmt.Errorf("unknown log format `%s`. possible formats: %s", format, []string{"text", "json"})
-	}
-	return nil
-}
-
-// GetFormat reports the currently selected format name.
-func (h *Handler) GetFormat() string {
-	if h.root.jsonMode.Load() {
-		return "json"
-	}
-	return "text"
-}
-
-// SetDisableTimestamp toggles whether Handle zeroes r.Time before
-// dispatching (slog's builtin text/json handlers skip emitting the time
-// attribute on a zero time).
-func (h *Handler) SetDisableTimestamp(v bool) { h.root.disableTimestamp.Store(v) }
-
-// ApplyConfig reads logging.level, logging.format, and (optionally)
-// logging.disable_timestamp from c and applies them to l. The reconfig
-// surface is discovered via structural type-assertion on l.Handler(), so
-// foreign handlers silently opt out of whichever capabilities they do not
-// implement.
-//
-// nebula.Main does NOT call this function on your behalf; callers that want
-// config-driven log level / format / timestamp updates invoke it at
-// startup and register it as a reload callback themselves. This keeps the
-// library from mutating an embedder's logger without their say-so.
-func ApplyConfig(l *slog.Logger, c Config) error {
-	h := l.Handler()
-
-	lvl, err := ParseLevel(strings.ToLower(c.GetString("logging.level", "info")))
-	if err != nil {
-		return err
-	}
-	if ls, ok := h.(interface{ SetLevel(slog.Level) }); ok {
-		ls.SetLevel(lvl)
-	}
-
-	format := strings.ToLower(c.GetString("logging.format", "text"))
-	if fs, ok := h.(interface{ SetFormat(string) error }); ok {
-		if err := fs.SetFormat(format); err != nil {
-			return err
-		}
-	}
-
-	if ts, ok := h.(interface{ SetDisableTimestamp(bool) }); ok {
-		ts.SetDisableTimestamp(c.GetBool("logging.disable_timestamp", false))
-	}
-	return nil
-}
-
-// ParseLevel converts a config-string level name ("trace", "debug", "info",
-// "warn"/"warning", "error", "fatal"/"panic") to a slog.Level. "fatal" and
-// "panic" are accepted for backwards compatibility with pre-slog configs
-// and both map to slog.LevelError.
-func ParseLevel(s string) (slog.Level, error) {
-	switch s {
-	case "trace":
-		return LevelTrace, nil
-	case "debug":
-		return slog.LevelDebug, nil
-	case "info":
-		return slog.LevelInfo, nil
-	case "warn", "warning":
-		return slog.LevelWarn, nil
-	case "error":
-		return slog.LevelError, nil
-	case "fatal", "panic":
-		return slog.LevelError, nil
-	default:
-		return 0, fmt.Errorf("not a valid logging level: %q", s)
-	}
-}
-
-// LevelName returns a human-readable name for a slog.Level matching the
-// strings accepted by ParseLevel.
-func LevelName(l slog.Level) string {
-	switch {
-	case l <= LevelTrace:
-		return "trace"
-	case l <= slog.LevelDebug:
-		return "debug"
-	case l <= slog.LevelInfo:
-		return "info"
-	case l <= slog.LevelWarn:
-		return "warn"
-	default:
-		return "error"
-	}
-}

logging/logger_bench_test.go

@@ -1,90 +0,0 @@
-package logging
-
-import (
-	"context"
-	"io"
-	"log/slog"
-	"testing"
-)
-
-// BenchmarkLogger_* compare the handler returned by NewLogger against a
-// stock slog text handler. The key thing we care about is the per-log
-// cost on a logger that has been derived via .With(), because that is the
-// shape subsystems store on their structs (HostInfo.logger(),
-// lh.l.With("subsystem", ...), etc.) and call from hot paths.
-
-func BenchmarkLogger_Stock_RootInfo(b *testing.B) {
-	l := slog.New(slog.DiscardHandler)
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		l.Info("hello", "i", i)
-	}
-}
-
-func BenchmarkLogger_Nebula_RootInfo(b *testing.B) {
-	l := NewLogger(io.Discard)
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		l.Info("hello", "i", i)
-	}
-}
-
-func BenchmarkLogger_Stock_DerivedInfo(b *testing.B) {
-	l := slog.New(slog.DiscardHandler).With(
-		"subsystem", "bench",
-		"localIndex", 1234,
-	)
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		l.Info("hello", "i", i)
-	}
-}
-
-func BenchmarkLogger_Nebula_DerivedInfo(b *testing.B) {
-	l := NewLogger(io.Discard).With(
-		"subsystem", "bench",
-		"localIndex", 1234,
-	)
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		l.Info("hello", "i", i)
-	}
-}
-
-// Gated-off-path benchmarks: mimic the typical hot-path shape
-// `if l.Enabled(ctx, slog.LevelDebug) { ... }` where the log is gated below
-// the active level. This is the dominant pattern in inside.go/outside.go and
-// what we pay on every packet.
-func BenchmarkLogger_Stock_DerivedEnabledGateMiss(b *testing.B) {
-	l := slog.New(slog.DiscardHandler).With(
-		"subsystem", "bench",
-		"localIndex", 1234,
-	)
-	ctx := context.Background()
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		if l.Enabled(ctx, slog.LevelDebug) {
-			l.Debug("hello", "i", i)
-		}
-	}
-}
-
-func BenchmarkLogger_Nebula_DerivedEnabledGateMiss(b *testing.B) {
-	l := NewLogger(io.Discard).With(
-		"subsystem", "bench",
-		"localIndex", 1234,
-	)
-	ctx := context.Background()
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		if l.Enabled(ctx, slog.LevelDebug) {
-			l.Debug("hello", "i", i)
-		}
-	}
-}

main.go

@@ -3,13 +3,13 @@ package nebula
 import (
 	"context"
 	"fmt"
-	"log/slog"
 	"net"
 	"net/netip"
 	"runtime/debug"
 	"strings"
 	"time"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/sshd"
@@ -20,7 +20,7 @@ import (
 
 type m = map[string]any
 
-func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, deviceFactory overlay.DeviceFactory) (retcon *Control, reterr error) {
+func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logger, deviceFactory overlay.DeviceFactory) (retcon *Control, reterr error) {
 	ctx, cancel := context.WithCancel(context.Background())
 	// Automatically cancel the context if Main returns an error, to signal all created goroutines to quit.
 	defer func() {
@@ -33,6 +33,11 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		buildVersion = moduleVersion()
 	}
 
+	l := logger
+	l.Formatter = &logrus.TextFormatter{
+		FullTimestamp: true,
+	}
+
 	// Print the config if in test, the exit comes later
 	if configTest {
 		b, err := yaml.Marshal(c.Settings)
@@ -41,9 +46,21 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		}
 
 		// Print the final config
-		l.Info(string(b))
+		l.Println(string(b))
 	}
 
+	err := configLogger(l, c)
+	if err != nil {
+		return nil, util.ContextualizeIfNeeded("Failed to configure the logger", err)
+	}
+
+	c.RegisterReloadCallback(func(c *config.C) {
+		err := configLogger(l, c)
+		if err != nil {
+			l.WithError(err).Error("Failed to configure the logger")
+		}
+	})
+
 	pki, err := NewPKIFromConfig(l, c)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to load PKI from config", err)
@@ -53,9 +70,9 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Error while loading firewall rules", err)
 	}
-	l.Info("Firewall started", "firewallHashes", fw.GetRuleHashes())
+	l.WithField("firewallHashes", fw.GetRuleHashes()).Info("Firewall started")
 
-	ssh, err := sshd.NewSSHServer(ctx, l.With("subsystem", "sshd"))
+	ssh, err := sshd.NewSSHServer(l.WithField("subsystem", "sshd"))
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Error while creating SSH server", err)
 	}
@@ -64,7 +81,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 	if c.GetBool("sshd.enabled", false) {
 		sshStart, err = configSSH(l, ssh, c)
 		if err != nil {
-			l.Warn("Failed to configure sshd, ssh debugging will not be available", "error", err)
+			l.WithError(err).Warn("Failed to configure sshd, ssh debugging will not be available")
 			sshStart = nil
 		}
 	}
@@ -82,7 +99,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 			routines = 1
 		}
 		if routines > 1 {
-			l.Info("Using multiple routines", "routines", routines)
+			l.WithField("routines", routines).Info("Using multiple routines")
 		}
 	} else {
 		// deprecated and undocumented
@@ -90,7 +107,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		udpQueues := c.GetInt("listen.routines", 1)
 		routines = max(tunQueues, udpQueues)
 		if routines != 1 {
-			l.Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead", "routines", routines)
+			l.WithField("routines", routines).Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead")
 		}
 	}
 
@@ -103,7 +120,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		conntrackCacheTimeout = 1 * time.Second
 	}
 	if conntrackCacheTimeout > 0 {
-		l.Info("Using routine-local conntrack cache", "duration", conntrackCacheTimeout)
+		l.WithField("duration", conntrackCacheTimeout).Info("Using routine-local conntrack cache")
 	}
 
 	var tun overlay.Device
@@ -149,7 +166,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		}
 
 		for i := 0; i < routines; i++ {
-			l.Info("listening", "addr", netip.AddrPortFrom(listenHost, uint16(port)))
+			l.Infof("listening on %v", netip.AddrPortFrom(listenHost, uint16(port)))
 			udpServer, err := udp.NewListener(l, listenHost, port, routines > 1, c.GetInt("listen.batch", 64))
 			if err != nil {
 				return nil, util.NewContextualError("Failed to open udp listener", m{"queue": i}, err)
@@ -170,7 +187,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 	}
 
 	hostMap := NewHostMapFromConfig(l, c)
-	punchy := NewPunchyFromConfig(l, c, udpConns[0])
+	punchy := NewPunchyFromConfig(l, c)
 	connManager := newConnectionManagerFromConfig(l, c, hostMap, punchy)
 	lightHouse, err := NewLightHouseFromConfig(ctx, l, c, pki.getCertState(), udpConns[0], punchy)
 	if err != nil {
@@ -184,10 +201,14 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		messageMetrics = newMessageMetricsOnlyRecvError()
 	}
 
+	useRelays := c.GetBool("relay.use_relays", DefaultUseRelays) && !c.GetBool("relay.am_relay", false)
+
 	handshakeConfig := HandshakeConfig{
-		tryInterval:    c.GetDuration("handshakes.try_interval", DefaultHandshakeTryInterval),
-		retries:        int64(c.GetInt("handshakes.retries", DefaultHandshakeRetries)),
-		triggerBuffer:  c.GetInt("handshakes.trigger_buffer", DefaultHandshakeTriggerBuffer),
+		tryInterval:   c.GetDuration("handshakes.try_interval", DefaultHandshakeTryInterval),
+		retries:       int64(c.GetInt("handshakes.retries", DefaultHandshakeRetries)),
+		triggerBuffer: c.GetInt("handshakes.trigger_buffer", DefaultHandshakeTriggerBuffer),
+		useRelays:     useRelays,
+
 		messageMetrics: messageMetrics,
 	}
 
@@ -196,7 +217,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 
 	ds, err := newDnsServerFromConfig(ctx, l, pki.getCertState(), hostMap, c)
 	if err != nil {
-		l.Warn("Failed to start DNS responder", "error", err)
+		l.WithError(err).Warn("Failed to start DNS responder")
 	}
 
 	ifConfig := &InterfaceConfig{
@@ -215,7 +236,6 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		DropLocalBroadcast:    c.GetBool("tun.drop_local_broadcast", false),
 		DropMulticast:         c.GetBool("tun.drop_multicast", false),
 		routines:              routines,
-		batchSize:             c.GetInt("listen.batch", 64),
 		MessageMetrics:        messageMetrics,
 		version:               buildVersion,
 		relayManager:          NewRelayManager(ctx, l, hostMap, c),
@@ -241,11 +261,9 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 
 		handshakeManager.f = ifce
 		go handshakeManager.Run(ctx)
-
-		punchy.Start(ctx, ifce, hostMap, lightHouse)
 	}
 
-	stats, err := newStatsServerFromConfig(ctx, l, c, buildVersion, configTest)
+	statsStart, err := startStats(l, c, buildVersion, configTest)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to start stats emitter", err)
 	}
@@ -265,7 +283,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 		ctx:                    ctx,
 		cancel:                 cancel,
 		sshStart:               sshStart,
-		statsStart:             stats.Start,
+		statsStart:             statsStart,
 		dnsStart:               ds.Start,
 		lighthouseStart:        lightHouse.StartUpdateWorker,
 		connectionManagerStart: connManager.Start,

message_metrics.go

@@ -13,8 +13,6 @@ type MessageMetrics struct {
 
 	rxUnknown metrics.Counter
 	txUnknown metrics.Counter
-
-	rxInvalid metrics.Counter
 }
 
 func (m *MessageMetrics) Rx(t header.MessageType, s header.MessageSubType, i int64) {
@@ -35,11 +33,6 @@ func (m *MessageMetrics) Tx(t header.MessageType, s header.MessageSubType, i int
 		}
 	}
 }
-func (m *MessageMetrics) RxInvalid(i int64) {
-	if m != nil && m.rxInvalid != nil {
-		m.rxInvalid.Inc(i)
-	}
-}
 
 func newMessageMetrics() *MessageMetrics {
 	gen := func(t string) [][]metrics.Counter {
@@ -63,7 +56,6 @@ func newMessageMetrics() *MessageMetrics {
 
 		rxUnknown: metrics.GetOrRegisterCounter("messages.rx.other", nil),
 		txUnknown: metrics.GetOrRegisterCounter("messages.tx.other", nil),
-		rxInvalid: metrics.GetOrRegisterCounter("messages.rx.invalid", nil),
 	}
 }
 

nebula.pb.go

@@ -124,7 +124,7 @@ func (x NebulaControl_MessageType) String() string {
 }
 
 func (NebulaControl_MessageType) EnumDescriptor() ([]byte, []int) {
-	return fileDescriptor_2d65afa7693df5ef, []int{6, 0}
+	return fileDescriptor_2d65afa7693df5ef, []int{8, 0}
 }
 
 type NebulaMeta struct {
@@ -489,6 +489,142 @@ func (m *NebulaPing) GetTime() uint64 {
 	return 0
 }
 
+type NebulaHandshake struct {
+	Details *NebulaHandshakeDetails `protobuf:"bytes,1,opt,name=Details,proto3" json:"Details,omitempty"`
+	Hmac    []byte                  `protobuf:"bytes,2,opt,name=Hmac,proto3" json:"Hmac,omitempty"`
+}
+
+func (m *NebulaHandshake) Reset()         { *m = NebulaHandshake{} }
+func (m *NebulaHandshake) String() string { return proto.CompactTextString(m) }
+func (*NebulaHandshake) ProtoMessage()    {}
+func (*NebulaHandshake) Descriptor() ([]byte, []int) {
+	return fileDescriptor_2d65afa7693df5ef, []int{6}
+}
+func (m *NebulaHandshake) XXX_Unmarshal(b []byte) error {
+	return m.Unmarshal(b)
+}
+func (m *NebulaHandshake) XXX_Marshal(b []byte, deterministic bool) ([]byte, error) {
+	if deterministic {
+		return xxx_messageInfo_NebulaHandshake.Marshal(b, m, deterministic)
+	} else {
+		b = b[:cap(b)]
+		n, err := m.MarshalToSizedBuffer(b)
+		if err != nil {
+			return nil, err
+		}
+		return b[:n], nil
+	}
+}
+func (m *NebulaHandshake) XXX_Merge(src proto.Message) {
+	xxx_messageInfo_NebulaHandshake.Merge(m, src)
+}
+func (m *NebulaHandshake) XXX_Size() int {
+	return m.Size()
+}
+func (m *NebulaHandshake) XXX_DiscardUnknown() {
+	xxx_messageInfo_NebulaHandshake.DiscardUnknown(m)
+}
+
+var xxx_messageInfo_NebulaHandshake proto.InternalMessageInfo
+
+func (m *NebulaHandshake) GetDetails() *NebulaHandshakeDetails {
+	if m != nil {
+		return m.Details
+	}
+	return nil
+}
+
+func (m *NebulaHandshake) GetHmac() []byte {
+	if m != nil {
+		return m.Hmac
+	}
+	return nil
+}
+
+type NebulaHandshakeDetails struct {
+	Cert           []byte `protobuf:"bytes,1,opt,name=Cert,proto3" json:"Cert,omitempty"`
+	InitiatorIndex uint32 `protobuf:"varint,2,opt,name=InitiatorIndex,proto3" json:"InitiatorIndex,omitempty"`
+	ResponderIndex uint32 `protobuf:"varint,3,opt,name=ResponderIndex,proto3" json:"ResponderIndex,omitempty"`
+	Cookie         uint64 `protobuf:"varint,4,opt,name=Cookie,proto3" json:"Cookie,omitempty"`
+	Time           uint64 `protobuf:"varint,5,opt,name=Time,proto3" json:"Time,omitempty"`
+	CertVersion    uint32 `protobuf:"varint,8,opt,name=CertVersion,proto3" json:"CertVersion,omitempty"`
+}
+
+func (m *NebulaHandshakeDetails) Reset()         { *m = NebulaHandshakeDetails{} }
+func (m *NebulaHandshakeDetails) String() string { return proto.CompactTextString(m) }
+func (*NebulaHandshakeDetails) ProtoMessage()    {}
+func (*NebulaHandshakeDetails) Descriptor() ([]byte, []int) {
+	return fileDescriptor_2d65afa7693df5ef, []int{7}
+}
+func (m *NebulaHandshakeDetails) XXX_Unmarshal(b []byte) error {
+	return m.Unmarshal(b)
+}
+func (m *NebulaHandshakeDetails) XXX_Marshal(b []byte, deterministic bool) ([]byte, error) {
+	if deterministic {
+		return xxx_messageInfo_NebulaHandshakeDetails.Marshal(b, m, deterministic)
+	} else {
+		b = b[:cap(b)]
+		n, err := m.MarshalToSizedBuffer(b)
+		if err != nil {
+			return nil, err
+		}
+		return b[:n], nil
+	}
+}
+func (m *NebulaHandshakeDetails) XXX_Merge(src proto.Message) {
+	xxx_messageInfo_NebulaHandshakeDetails.Merge(m, src)
+}
+func (m *NebulaHandshakeDetails) XXX_Size() int {
+	return m.Size()
+}
+func (m *NebulaHandshakeDetails) XXX_DiscardUnknown() {
+	xxx_messageInfo_NebulaHandshakeDetails.DiscardUnknown(m)
+}
+
+var xxx_messageInfo_NebulaHandshakeDetails proto.InternalMessageInfo
+
+func (m *NebulaHandshakeDetails) GetCert() []byte {
+	if m != nil {
+		return m.Cert
+	}
+	return nil
+}
+
+func (m *NebulaHandshakeDetails) GetInitiatorIndex() uint32 {
+	if m != nil {
+		return m.InitiatorIndex
+	}
+	return 0
+}
+
+func (m *NebulaHandshakeDetails) GetResponderIndex() uint32 {
+	if m != nil {
+		return m.ResponderIndex
+	}
+	return 0
+}
+
+func (m *NebulaHandshakeDetails) GetCookie() uint64 {
+	if m != nil {
+		return m.Cookie
+	}
+	return 0
+}
+
+func (m *NebulaHandshakeDetails) GetTime() uint64 {
+	if m != nil {
+		return m.Time
+	}
+	return 0
+}
+
+func (m *NebulaHandshakeDetails) GetCertVersion() uint32 {
+	if m != nil {
+		return m.CertVersion
+	}
+	return 0
+}
+
 type NebulaControl struct {
 	Type                NebulaControl_MessageType `protobuf:"varint,1,opt,name=Type,proto3,enum=nebula.NebulaControl_MessageType" json:"Type,omitempty"`
 	InitiatorRelayIndex uint32                    `protobuf:"varint,2,opt,name=InitiatorRelayIndex,proto3" json:"InitiatorRelayIndex,omitempty"`
@@ -503,7 +639,7 @@ func (m *NebulaControl) Reset()         { *m = NebulaControl{} }
 func (m *NebulaControl) String() string { return proto.CompactTextString(m) }
 func (*NebulaControl) ProtoMessage()    {}
 func (*NebulaControl) Descriptor() ([]byte, []int) {
-	return fileDescriptor_2d65afa7693df5ef, []int{6}
+	return fileDescriptor_2d65afa7693df5ef, []int{8}
 }
 func (m *NebulaControl) XXX_Unmarshal(b []byte) error {
 	return m.Unmarshal(b)
@@ -593,55 +729,65 @@ func init() {
 	proto.RegisterType((*V4AddrPort)(nil), "nebula.V4AddrPort")
 	proto.RegisterType((*V6AddrPort)(nil), "nebula.V6AddrPort")
 	proto.RegisterType((*NebulaPing)(nil), "nebula.NebulaPing")
+	proto.RegisterType((*NebulaHandshake)(nil), "nebula.NebulaHandshake")
+	proto.RegisterType((*NebulaHandshakeDetails)(nil), "nebula.NebulaHandshakeDetails")
 	proto.RegisterType((*NebulaControl)(nil), "nebula.NebulaControl")
 }
 
 func init() { proto.RegisterFile("nebula.proto", fileDescriptor_2d65afa7693df5ef) }
 
 var fileDescriptor_2d65afa7693df5ef = []byte{
-	// 665 bytes of a gzipped FileDescriptorProto
-	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x84, 0x54, 0xcd, 0x6e, 0xd3, 0x5c,
-	0x10, 0x8d, 0x1d, 0x27, 0x69, 0x27, 0x4d, 0x3e, 0x7f, 0x53, 0x51, 0x12, 0x24, 0xac, 0xe0, 0x45,
-	0x55, 0xb1, 0x48, 0x51, 0x5a, 0xba, 0xa6, 0x2d, 0x42, 0xa9, 0xd4, 0x9f, 0x70, 0x55, 0x8a, 0xc4,
-	0xce, 0xb5, 0x2f, 0x8d, 0x55, 0xc7, 0x37, 0xb5, 0x6f, 0x50, 0xf3, 0x16, 0x3c, 0x0c, 0x0f, 0x01,
-	0xbb, 0x2e, 0x59, 0xa2, 0x66, 0xc9, 0x92, 0x17, 0x40, 0xf7, 0xfa, 0xbf, 0x31, 0xb0, 0xbb, 0x33,
-	0xe7, 0x9c, 0x99, 0xc9, 0xc9, 0x8c, 0x61, 0xcd, 0xa7, 0x97, 0x33, 0xcf, 0xea, 0x4f, 0x03, 0xc6,
-	0x19, 0xd6, 0xa3, 0xc8, 0xfc, 0xa9, 0x02, 0x9c, 0xca, 0xe7, 0x09, 0xe5, 0x16, 0x0e, 0x40, 0x3b,
-	0x9f, 0x4f, 0x69, 0x47, 0xe9, 0x29, 0x5b, 0xed, 0x81, 0xd1, 0x8f, 0x35, 0x19, 0xa3, 0x7f, 0x42,
-	0xc3, 0xd0, 0xba, 0xa2, 0x82, 0x45, 0x24, 0x17, 0x77, 0xa0, 0xf1, 0x9a, 0x72, 0xcb, 0xf5, 0xc2,
-	0x8e, 0xda, 0x53, 0xb6, 0x9a, 0x83, 0xee, 0xb2, 0x2c, 0x26, 0x90, 0x84, 0x69, 0xfe, 0x52, 0xa0,
-	0x99, 0x2b, 0x85, 0x2b, 0xa0, 0x9d, 0x32, 0x9f, 0xea, 0x15, 0x6c, 0xc1, 0xea, 0x90, 0x85, 0xfc,
-	0xed, 0x8c, 0x06, 0x73, 0x5d, 0x41, 0x84, 0x76, 0x1a, 0x12, 0x3a, 0xf5, 0xe6, 0xba, 0x8a, 0x4f,
-	0x60, 0x43, 0xe4, 0xde, 0x4d, 0x1d, 0x8b, 0xd3, 0x53, 0xc6, 0xdd, 0x8f, 0xae, 0x6d, 0x71, 0x97,
-	0xf9, 0x7a, 0x15, 0xbb, 0xf0, 0x48, 0x60, 0x27, 0xec, 0x13, 0x75, 0x0a, 0x90, 0x96, 0x40, 0xa3,
-	0x99, 0x6f, 0x8f, 0x0b, 0x50, 0x0d, 0xdb, 0x00, 0x02, 0x7a, 0x3f, 0x66, 0xd6, 0xc4, 0xd5, 0xeb,
-	0xb8, 0x0e, 0xff, 0x65, 0x71, 0xd4, 0xb6, 0x21, 0x26, 0x1b, 0x59, 0x7c, 0x7c, 0x38, 0xa6, 0xf6,
-	0xb5, 0xbe, 0x22, 0x26, 0x4b, 0xc3, 0x88, 0xb2, 0x8a, 0x4f, 0xa1, 0x5b, 0x3e, 0xd9, 0xbe, 0x7d,
-	0xad, 0x83, 0xf9, 0x4d, 0x85, 0xff, 0x97, 0x4c, 0x41, 0x13, 0xe0, 0xcc, 0x73, 0x2e, 0xa6, 0xfe,
-	0xbe, 0xe3, 0x04, 0xd2, 0xfa, 0xd6, 0x81, 0xda, 0x51, 0x48, 0x2e, 0x8b, 0x9b, 0xd0, 0x48, 0x08,
-	0x75, 0x69, 0xf2, 0x5a, 0x62, 0xb2, 0xc8, 0x91, 0x04, 0xc4, 0x3e, 0xe8, 0x67, 0x9e, 0x43, 0xa8,
-	0x67, 0xcd, 0xe3, 0x54, 0xd8, 0xa9, 0xf5, 0xaa, 0x71, 0xc5, 0x25, 0x0c, 0x07, 0xd0, 0x2a, 0x92,
-	0x1b, 0xbd, 0xea, 0x52, 0xf5, 0x22, 0x05, 0x77, 0xa1, 0x79, 0xb1, 0x2b, 0x9e, 0x23, 0x16, 0x70,
-	0xf1, 0xa7, 0x0b, 0x05, 0x26, 0x8a, 0x0c, 0x22, 0x79, 0x9a, 0x54, 0xed, 0x65, 0x2a, 0xed, 0x81,
-	0x6a, 0x2f, 0xa7, 0xca, 0x68, 0xd8, 0x81, 0x86, 0xcd, 0x66, 0x3e, 0xa7, 0x41, 0xa7, 0x2a, 0x8c,
-	0x21, 0x49, 0x68, 0x6e, 0x82, 0x26, 0x7f, 0x71, 0x1b, 0xd4, 0xa1, 0x2b, 0x5d, 0xd3, 0x88, 0x3a,
-	0x74, 0x45, 0x7c, 0xcc, 0xe4, 0x26, 0x6a, 0x44, 0x3d, 0x66, 0xe6, 0x2e, 0x40, 0x36, 0x06, 0x62,
-	0xa4, 0x8a, 0x5c, 0x26, 0x51, 0x05, 0x04, 0x4d, 0x60, 0x52, 0xd3, 0x22, 0xf2, 0x6d, 0xbe, 0x02,
-	0xc8, 0xc6, 0xf8, 0x57, 0x8f, 0xb4, 0x42, 0x35, 0x57, 0xe1, 0x36, 0x39, 0xac, 0x91, 0xeb, 0x5f,
-	0xfd, 0xfd, 0xb0, 0x04, 0xa3, 0xe4, 0xb0, 0x10, 0xb4, 0x73, 0x77, 0x42, 0xe3, 0x3e, 0xf2, 0x6d,
-	0x9a, 0x4b, 0x67, 0x23, 0xc4, 0x7a, 0x05, 0x57, 0xa1, 0x16, 0x2d, 0xa1, 0x62, 0x7e, 0xa9, 0x42,
-	0x2b, 0x2a, 0x7c, 0xc8, 0x7c, 0x1e, 0x30, 0x0f, 0x5f, 0x16, 0xba, 0x3f, 0x2b, 0x76, 0x8f, 0x49,
-	0x25, 0x03, 0xbc, 0x80, 0xf5, 0x23, 0xdf, 0xe5, 0xae, 0xc5, 0x59, 0x20, 0x57, 0xe0, 0xc8, 0x77,
-	0xe8, 0x6d, 0xec, 0x53, 0x19, 0x24, 0x14, 0x84, 0x86, 0x53, 0xe6, 0x3b, 0x34, 0xaf, 0x88, 0x7c,
-	0x29, 0x83, 0xf0, 0x39, 0xb4, 0x93, 0xa5, 0x3c, 0x67, 0xf2, 0xaf, 0xd1, 0xd2, 0x03, 0x78, 0x80,
-	0xe4, 0x97, 0xfb, 0x4d, 0xc0, 0x26, 0x92, 0x5d, 0x4b, 0xd9, 0x4b, 0x18, 0xf6, 0xa1, 0x99, 0x2f,
-	0x5c, 0x76, 0x38, 0x79, 0x42, 0x7a, 0x0c, 0x69, 0xf1, 0x46, 0x89, 0xa2, 0x48, 0x31, 0x87, 0x7f,
-	0xfa, 0x8e, 0x6d, 0x00, 0x1e, 0x06, 0xd4, 0xe2, 0x54, 0xf2, 0x09, 0xbd, 0x99, 0xd1, 0x90, 0xeb,
-	0x0a, 0x3e, 0x86, 0xf5, 0x42, 0x5e, 0x58, 0x12, 0x52, 0x5d, 0x3d, 0xd8, 0xf9, 0x7a, 0x6f, 0x28,
-	0x77, 0xf7, 0x86, 0xf2, 0xe3, 0xde, 0x50, 0x3e, 0x2f, 0x8c, 0xca, 0xdd, 0xc2, 0xa8, 0x7c, 0x5f,
-	0x18, 0x95, 0x0f, 0xdd, 0x2b, 0x97, 0x8f, 0x67, 0x97, 0x7d, 0x9b, 0x4d, 0xb6, 0x43, 0xcf, 0xb2,
-	0xaf, 0xc7, 0x37, 0xdb, 0xd1, 0x48, 0x97, 0x75, 0xf9, 0x39, 0xdf, 0xf9, 0x1d, 0x00, 0x00, 0xff,
-	0xff, 0x51, 0x0a, 0xe3, 0xd7, 0xde, 0x05, 0x00, 0x00,
+	// 785 bytes of a gzipped FileDescriptorProto
+	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x84, 0x55, 0xcd, 0x6e, 0xeb, 0x44,
+	0x14, 0x8e, 0x1d, 0x27, 0x4e, 0x4f, 0x7e, 0xae, 0x39, 0x15, 0xc1, 0x41, 0x22, 0x0a, 0x5e, 0x54,
+	0x57, 0x2c, 0x72, 0x51, 0x5a, 0xae, 0x58, 0x72, 0x1b, 0x84, 0xd2, 0xaa, 0x3f, 0x61, 0x54, 0x8a,
+	0xc4, 0x06, 0xb9, 0xf6, 0xd0, 0x58, 0x71, 0x3c, 0xa9, 0x3d, 0x41, 0xcd, 0x5b, 0xf0, 0x30, 0x3c,
+	0x04, 0xec, 0xba, 0x42, 0x2c, 0x51, 0xbb, 0x64, 0xc9, 0x0b, 0xa0, 0x19, 0xff, 0x27, 0x86, 0xbb,
+	0x9b, 0x73, 0xbe, 0xef, 0x3b, 0x73, 0xe6, 0xf3, 0x9c, 0x31, 0x74, 0x02, 0x7a, 0xb7, 0xf1, 0xed,
+	0xf1, 0x3a, 0x64, 0x9c, 0x61, 0x33, 0x8e, 0xac, 0xbf, 0x55, 0x80, 0x2b, 0xb9, 0xbc, 0xa4, 0xdc,
+	0xc6, 0x09, 0x68, 0x37, 0xdb, 0x35, 0x35, 0x95, 0x91, 0xf2, 0xba, 0x37, 0x19, 0x8e, 0x13, 0x4d,
+	0xce, 0x18, 0x5f, 0xd2, 0x28, 0xb2, 0xef, 0xa9, 0x60, 0x11, 0xc9, 0xc5, 0x63, 0xd0, 0xbf, 0xa6,
+	0xdc, 0xf6, 0xfc, 0xc8, 0x54, 0x47, 0xca, 0xeb, 0xf6, 0x64, 0xb0, 0x2f, 0x4b, 0x08, 0x24, 0x65,
+	0x5a, 0xff, 0x28, 0xd0, 0x2e, 0x94, 0xc2, 0x16, 0x68, 0x57, 0x2c, 0xa0, 0x46, 0x0d, 0xbb, 0x70,
+	0x30, 0x63, 0x11, 0xff, 0x76, 0x43, 0xc3, 0xad, 0xa1, 0x20, 0x42, 0x2f, 0x0b, 0x09, 0x5d, 0xfb,
+	0x5b, 0x43, 0xc5, 0x8f, 0xa1, 0x2f, 0x72, 0xdf, 0xad, 0x5d, 0x9b, 0xd3, 0x2b, 0xc6, 0xbd, 0x9f,
+	0x3c, 0xc7, 0xe6, 0x1e, 0x0b, 0x8c, 0x3a, 0x0e, 0xe0, 0x43, 0x81, 0x5d, 0xb2, 0x9f, 0xa9, 0x5b,
+	0x82, 0xb4, 0x14, 0x9a, 0x6f, 0x02, 0x67, 0x51, 0x82, 0x1a, 0xd8, 0x03, 0x10, 0xd0, 0xf7, 0x0b,
+	0x66, 0xaf, 0x3c, 0xa3, 0x89, 0x87, 0xf0, 0x2a, 0x8f, 0xe3, 0x6d, 0x75, 0xd1, 0xd9, 0xdc, 0xe6,
+	0x8b, 0xe9, 0x82, 0x3a, 0x4b, 0xa3, 0x25, 0x3a, 0xcb, 0xc2, 0x98, 0x72, 0x80, 0x9f, 0xc0, 0xa0,
+	0xba, 0xb3, 0x77, 0xce, 0xd2, 0x00, 0xeb, 0x77, 0x15, 0x3e, 0xd8, 0x33, 0x05, 0x2d, 0x80, 0x6b,
+	0xdf, 0xbd, 0x5d, 0x07, 0xef, 0x5c, 0x37, 0x94, 0xd6, 0x77, 0x4f, 0x55, 0x53, 0x21, 0x85, 0x2c,
+	0x1e, 0x81, 0x9e, 0x12, 0x9a, 0xd2, 0xe4, 0x4e, 0x6a, 0xb2, 0xc8, 0x91, 0x14, 0xc4, 0x31, 0x18,
+	0xd7, 0xbe, 0x4b, 0xa8, 0x6f, 0x6f, 0x93, 0x54, 0x64, 0x36, 0x46, 0xf5, 0xa4, 0xe2, 0x1e, 0x86,
+	0x13, 0xe8, 0x96, 0xc9, 0xfa, 0xa8, 0xbe, 0x57, 0xbd, 0x4c, 0xc1, 0x13, 0x68, 0xdf, 0x9e, 0x88,
+	0xe5, 0x9c, 0x85, 0x5c, 0x7c, 0x74, 0xa1, 0xc0, 0x54, 0x91, 0x43, 0xa4, 0x48, 0x93, 0xaa, 0xb7,
+	0xb9, 0x4a, 0xdb, 0x51, 0xbd, 0x2d, 0xa8, 0x72, 0x1a, 0x9a, 0xa0, 0x3b, 0x6c, 0x13, 0x70, 0x1a,
+	0x9a, 0x75, 0x61, 0x0c, 0x49, 0x43, 0xeb, 0x08, 0x34, 0x79, 0xe2, 0x1e, 0xa8, 0x33, 0x4f, 0xba,
+	0xa6, 0x11, 0x75, 0xe6, 0x89, 0xf8, 0x82, 0xc9, 0x9b, 0xa8, 0x11, 0xf5, 0x82, 0x59, 0x27, 0x00,
+	0x79, 0x1b, 0x88, 0xb1, 0x2a, 0x76, 0x99, 0xc4, 0x15, 0x10, 0x34, 0x81, 0x49, 0x4d, 0x97, 0xc8,
+	0xb5, 0xf5, 0x15, 0x40, 0xde, 0xc6, 0xfb, 0xf6, 0xc8, 0x2a, 0xd4, 0x0b, 0x15, 0x1e, 0xd3, 0xc1,
+	0x9a, 0x7b, 0xc1, 0xfd, 0xff, 0x0f, 0x96, 0x60, 0x54, 0x0c, 0x16, 0x82, 0x76, 0xe3, 0xad, 0x68,
+	0xb2, 0x8f, 0x5c, 0x5b, 0xd6, 0xde, 0xd8, 0x08, 0xb1, 0x51, 0xc3, 0x03, 0x68, 0xc4, 0x97, 0x50,
+	0xb1, 0x7e, 0x84, 0x57, 0x71, 0xdd, 0x99, 0x1d, 0xb8, 0xd1, 0xc2, 0x5e, 0x52, 0xfc, 0x32, 0x9f,
+	0x51, 0x45, 0x5e, 0x9f, 0x9d, 0x0e, 0x32, 0xe6, 0xee, 0xa0, 0x8a, 0x26, 0x66, 0x2b, 0xdb, 0x91,
+	0x4d, 0x74, 0x88, 0x5c, 0x5b, 0x7f, 0x28, 0xd0, 0xaf, 0xd6, 0x09, 0xfa, 0x94, 0x86, 0x5c, 0xee,
+	0xd2, 0x21, 0x72, 0x8d, 0x47, 0xd0, 0x3b, 0x0b, 0x3c, 0xee, 0xd9, 0x9c, 0x85, 0x67, 0x81, 0x4b,
+	0x1f, 0x13, 0xa7, 0x77, 0xb2, 0x82, 0x47, 0x68, 0xb4, 0x66, 0x81, 0x4b, 0x13, 0x5e, 0xec, 0xe7,
+	0x4e, 0x16, 0xfb, 0xd0, 0x9c, 0x32, 0xb6, 0xf4, 0xa8, 0xa9, 0x49, 0x67, 0x92, 0x28, 0xf3, 0xab,
+	0x91, 0xfb, 0x85, 0x23, 0x68, 0x8b, 0x1e, 0x6e, 0x69, 0x18, 0x79, 0x2c, 0x30, 0x5b, 0xb2, 0x60,
+	0x31, 0x75, 0xae, 0xb5, 0x9a, 0x86, 0x7e, 0xae, 0xb5, 0x74, 0xa3, 0x65, 0xfd, 0x5a, 0x87, 0x6e,
+	0x7c, 0xb0, 0x29, 0x0b, 0x78, 0xc8, 0x7c, 0xfc, 0xa2, 0xf4, 0xdd, 0x3e, 0x2d, 0xbb, 0x96, 0x90,
+	0x2a, 0x3e, 0xdd, 0xe7, 0x70, 0x98, 0x1d, 0x4e, 0x0e, 0x4f, 0xf1, 0xdc, 0x55, 0x90, 0x50, 0x64,
+	0xc7, 0x2c, 0x28, 0x62, 0x07, 0xaa, 0x20, 0xfc, 0x0c, 0x7a, 0xe9, 0x38, 0xdf, 0x30, 0x79, 0xa9,
+	0xb5, 0xec, 0xe9, 0xd8, 0x41, 0x8a, 0xcf, 0xc2, 0x37, 0x21, 0x5b, 0x49, 0x76, 0x23, 0x63, 0xef,
+	0x61, 0x38, 0x86, 0x76, 0xb1, 0x70, 0xd5, 0x93, 0x53, 0x24, 0x64, 0xcf, 0x48, 0x56, 0x5c, 0xaf,
+	0x50, 0x94, 0x29, 0xd6, 0xec, 0xbf, 0xfe, 0x00, 0x7d, 0xc0, 0x69, 0x48, 0x6d, 0x4e, 0x25, 0x9f,
+	0xd0, 0x87, 0x0d, 0x8d, 0xb8, 0xa1, 0xe0, 0x47, 0x70, 0x58, 0xca, 0x0b, 0x4b, 0x22, 0x6a, 0xa8,
+	0xa7, 0xc7, 0xbf, 0x3d, 0x0f, 0x95, 0xa7, 0xe7, 0xa1, 0xf2, 0xd7, 0xf3, 0x50, 0xf9, 0xe5, 0x65,
+	0x58, 0x7b, 0x7a, 0x19, 0xd6, 0xfe, 0x7c, 0x19, 0xd6, 0x7e, 0x18, 0xdc, 0x7b, 0x7c, 0xb1, 0xb9,
+	0x1b, 0x3b, 0x6c, 0xf5, 0x26, 0xf2, 0x6d, 0x67, 0xb9, 0x78, 0x78, 0x13, 0xb7, 0x74, 0xd7, 0x94,
+	0x3f, 0xc2, 0xe3, 0x7f, 0x03, 0x00, 0x00, 0xff, 0xff, 0xea, 0x6f, 0xbc, 0x50, 0x18, 0x07, 0x00,
+	0x00,
 }
 
 func (m *NebulaMeta) Marshal() (dAtA []byte, err error) {
@@ -926,6 +1072,103 @@ func (m *NebulaPing) MarshalToSizedBuffer(dAtA []byte) (int, error) {
 	return len(dAtA) - i, nil
 }
 
+func (m *NebulaHandshake) Marshal() (dAtA []byte, err error) {
+	size := m.Size()
+	dAtA = make([]byte, size)
+	n, err := m.MarshalToSizedBuffer(dAtA[:size])
+	if err != nil {
+		return nil, err
+	}
+	return dAtA[:n], nil
+}
+
+func (m *NebulaHandshake) MarshalTo(dAtA []byte) (int, error) {
+	size := m.Size()
+	return m.MarshalToSizedBuffer(dAtA[:size])
+}
+
+func (m *NebulaHandshake) MarshalToSizedBuffer(dAtA []byte) (int, error) {
+	i := len(dAtA)
+	_ = i
+	var l int
+	_ = l
+	if len(m.Hmac) > 0 {
+		i -= len(m.Hmac)
+		copy(dAtA[i:], m.Hmac)
+		i = encodeVarintNebula(dAtA, i, uint64(len(m.Hmac)))
+		i--
+		dAtA[i] = 0x12
+	}
+	if m.Details != nil {
+		{
+			size, err := m.Details.MarshalToSizedBuffer(dAtA[:i])
+			if err != nil {
+				return 0, err
+			}
+			i -= size
+			i = encodeVarintNebula(dAtA, i, uint64(size))
+		}
+		i--
+		dAtA[i] = 0xa
+	}
+	return len(dAtA) - i, nil
+}
+
+func (m *NebulaHandshakeDetails) Marshal() (dAtA []byte, err error) {
+	size := m.Size()
+	dAtA = make([]byte, size)
+	n, err := m.MarshalToSizedBuffer(dAtA[:size])
+	if err != nil {
+		return nil, err
+	}
+	return dAtA[:n], nil
+}
+
+func (m *NebulaHandshakeDetails) MarshalTo(dAtA []byte) (int, error) {
+	size := m.Size()
+	return m.MarshalToSizedBuffer(dAtA[:size])
+}
+
+func (m *NebulaHandshakeDetails) MarshalToSizedBuffer(dAtA []byte) (int, error) {
+	i := len(dAtA)
+	_ = i
+	var l int
+	_ = l
+	if m.CertVersion != 0 {
+		i = encodeVarintNebula(dAtA, i, uint64(m.CertVersion))
+		i--
+		dAtA[i] = 0x40
+	}
+	if m.Time != 0 {
+		i = encodeVarintNebula(dAtA, i, uint64(m.Time))
+		i--
+		dAtA[i] = 0x28
+	}
+	if m.Cookie != 0 {
+		i = encodeVarintNebula(dAtA, i, uint64(m.Cookie))
+		i--
+		dAtA[i] = 0x20
+	}
+	if m.ResponderIndex != 0 {
+		i = encodeVarintNebula(dAtA, i, uint64(m.ResponderIndex))
+		i--
+		dAtA[i] = 0x18
+	}
+	if m.InitiatorIndex != 0 {
+		i = encodeVarintNebula(dAtA, i, uint64(m.InitiatorIndex))
+		i--
+		dAtA[i] = 0x10
+	}
+	if len(m.Cert) > 0 {
+		i -= len(m.Cert)
+		copy(dAtA[i:], m.Cert)
+		i = encodeVarintNebula(dAtA, i, uint64(len(m.Cert)))
+		i--
+		dAtA[i] = 0xa
+	}
+	return len(dAtA) - i, nil
+}
+
 func (m *NebulaControl) Marshal() (dAtA []byte, err error) {
 	size := m.Size()
 	dAtA = make([]byte, size)
@@ -1132,6 +1375,51 @@ func (m *NebulaPing) Size() (n int) {
 	return n
 }
 
+func (m *NebulaHandshake) Size() (n int) {
+	if m == nil {
+		return 0
+	}
+	var l int
+	_ = l
+	if m.Details != nil {
+		l = m.Details.Size()
+		n += 1 + l + sovNebula(uint64(l))
+	}
+	l = len(m.Hmac)
+	if l > 0 {
+		n += 1 + l + sovNebula(uint64(l))
+	}
+	return n
+}
+
+func (m *NebulaHandshakeDetails) Size() (n int) {
+	if m == nil {
+		return 0
+	}
+	var l int
+	_ = l
+	l = len(m.Cert)
+	if l > 0 {
+		n += 1 + l + sovNebula(uint64(l))
+	}
+	if m.InitiatorIndex != 0 {
+		n += 1 + sovNebula(uint64(m.InitiatorIndex))
+	}
+	if m.ResponderIndex != 0 {
+		n += 1 + sovNebula(uint64(m.ResponderIndex))
+	}
+	if m.Cookie != 0 {
+		n += 1 + sovNebula(uint64(m.Cookie))
+	}
+	if m.Time != 0 {
+		n += 1 + sovNebula(uint64(m.Time))
+	}
+	if m.CertVersion != 0 {
+		n += 1 + sovNebula(uint64(m.CertVersion))
+	}
+	return n
+}
+
 func (m *NebulaControl) Size() (n int) {
 	if m == nil {
 		return 0
@@ -1948,6 +2236,305 @@ func (m *NebulaPing) Unmarshal(dAtA []byte) error {
 	}
 	return nil
 }
+func (m *NebulaHandshake) Unmarshal(dAtA []byte) error {
+	l := len(dAtA)
+	iNdEx := 0
+	for iNdEx < l {
+		preIndex := iNdEx
+		var wire uint64
+		for shift := uint(0); ; shift += 7 {
+			if shift >= 64 {
+				return ErrIntOverflowNebula
+			}
+			if iNdEx >= l {
+				return io.ErrUnexpectedEOF
+			}
+			b := dAtA[iNdEx]
+			iNdEx++
+			wire |= uint64(b&0x7F) << shift
+			if b < 0x80 {
+				break
+			}
+		}
+		fieldNum := int32(wire >> 3)
+		wireType := int(wire & 0x7)
+		if wireType == 4 {
+			return fmt.Errorf("proto: NebulaHandshake: wiretype end group for non-group")
+		}
+		if fieldNum <= 0 {
+			return fmt.Errorf("proto: NebulaHandshake: illegal tag %d (wire type %d)", fieldNum, wire)
+		}
+		switch fieldNum {
+		case 1:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Details", wireType)
+			}
+			var msglen int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				msglen |= int(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			if msglen < 0 {
+				return ErrInvalidLengthNebula
+			}
+			postIndex := iNdEx + msglen
+			if postIndex < 0 {
+				return ErrInvalidLengthNebula
+			}
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			if m.Details == nil {
+				m.Details = &NebulaHandshakeDetails{}
+			}
+			if err := m.Details.Unmarshal(dAtA[iNdEx:postIndex]); err != nil {
+				return err
+			}
+			iNdEx = postIndex
+		case 2:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Hmac", wireType)
+			}
+			var byteLen int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				byteLen |= int(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			if byteLen < 0 {
+				return ErrInvalidLengthNebula
+			}
+			postIndex := iNdEx + byteLen
+			if postIndex < 0 {
+				return ErrInvalidLengthNebula
+			}
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.Hmac = append(m.Hmac[:0], dAtA[iNdEx:postIndex]...)
+			if m.Hmac == nil {
+				m.Hmac = []byte{}
+			}
+			iNdEx = postIndex
+		default:
+			iNdEx = preIndex
+			skippy, err := skipNebula(dAtA[iNdEx:])
+			if err != nil {
+				return err
+			}
+			if (skippy < 0) || (iNdEx+skippy) < 0 {
+				return ErrInvalidLengthNebula
+			}
+			if (iNdEx + skippy) > l {
+				return io.ErrUnexpectedEOF
+			}
+			iNdEx += skippy
+		}
+	}
+
+	if iNdEx > l {
+		return io.ErrUnexpectedEOF
+	}
+	return nil
+}
+func (m *NebulaHandshakeDetails) Unmarshal(dAtA []byte) error {
+	l := len(dAtA)
+	iNdEx := 0
+	for iNdEx < l {
+		preIndex := iNdEx
+		var wire uint64
+		for shift := uint(0); ; shift += 7 {
+			if shift >= 64 {
+				return ErrIntOverflowNebula
+			}
+			if iNdEx >= l {
+				return io.ErrUnexpectedEOF
+			}
+			b := dAtA[iNdEx]
+			iNdEx++
+			wire |= uint64(b&0x7F) << shift
+			if b < 0x80 {
+				break
+			}
+		}
+		fieldNum := int32(wire >> 3)
+		wireType := int(wire & 0x7)
+		if wireType == 4 {
+			return fmt.Errorf("proto: NebulaHandshakeDetails: wiretype end group for non-group")
+		}
+		if fieldNum <= 0 {
+			return fmt.Errorf("proto: NebulaHandshakeDetails: illegal tag %d (wire type %d)", fieldNum, wire)
+		}
+		switch fieldNum {
+		case 1:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Cert", wireType)
+			}
+			var byteLen int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				byteLen |= int(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			if byteLen < 0 {
+				return ErrInvalidLengthNebula
+			}
+			postIndex := iNdEx + byteLen
+			if postIndex < 0 {
+				return ErrInvalidLengthNebula
+			}
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.Cert = append(m.Cert[:0], dAtA[iNdEx:postIndex]...)
+			if m.Cert == nil {
+				m.Cert = []byte{}
+			}
+			iNdEx = postIndex
+		case 2:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field InitiatorIndex", wireType)
+			}
+			m.InitiatorIndex = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				m.InitiatorIndex |= uint32(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 3:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field ResponderIndex", wireType)
+			}
+			m.ResponderIndex = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				m.ResponderIndex |= uint32(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 4:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Cookie", wireType)
+			}
+			m.Cookie = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				m.Cookie |= uint64(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 5:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Time", wireType)
+			}
+			m.Time = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				m.Time |= uint64(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 8:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field CertVersion", wireType)
+			}
+			m.CertVersion = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowNebula
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				m.CertVersion |= uint32(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		default:
+			iNdEx = preIndex
+			skippy, err := skipNebula(dAtA[iNdEx:])
+			if err != nil {
+				return err
+			}
+			if (skippy < 0) || (iNdEx+skippy) < 0 {
+				return ErrInvalidLengthNebula
+			}
+			if (iNdEx + skippy) > l {
+				return io.ErrUnexpectedEOF
+			}
+			iNdEx += skippy
+		}
+	}
+
+	if iNdEx > l {
+		return io.ErrUnexpectedEOF
+	}
+	return nil
+}
 func (m *NebulaControl) Unmarshal(dAtA []byte) error {
 	l := len(dAtA)
 	iNdEx := 0

nebula.proto

@@ -60,9 +60,21 @@ message NebulaPing {
 	uint64 Time = 2;
 }
 
-// NebulaHandshake / NebulaHandshakeDetails moved to
-// handshake/handshake.proto. The handshake package speaks that wire format
-// directly via a hand-written encoder/decoder.
+message NebulaHandshake {
+  NebulaHandshakeDetails Details = 1;
+  bytes Hmac = 2;
+}
+
+message NebulaHandshakeDetails {
+  bytes Cert = 1;
+  uint32 InitiatorIndex = 2;
+  uint32 ResponderIndex = 3;
+  uint64 Cookie = 4;
+  uint64 Time = 5;
+  uint32 CertVersion = 8;
+  // reserved for WIP multiport
+  reserved 6, 7;
+}
 
 message NebulaControl {
   enum MessageType {

noise.go

@@ -0,0 +1,75 @@
+package nebula
+
+import (
+	"crypto/cipher"
+	"encoding/binary"
+	"errors"
+
+	"github.com/flynn/noise"
+)
+
+type endianness interface {
+	PutUint64(b []byte, v uint64)
+}
+
+var noiseEndianness endianness = binary.BigEndian
+
+type NebulaCipherState struct {
+	c noise.Cipher
+	//k [32]byte
+	//n uint64
+}
+
+func NewNebulaCipherState(s *noise.CipherState) *NebulaCipherState {
+	return &NebulaCipherState{c: s.Cipher()}
+
+}
+
+// EncryptDanger encrypts and authenticates a given payload.
+//
+// out is a destination slice to hold the output of the EncryptDanger operation.
+//   - ad is additional data, which will be authenticated and appended to out, but not encrypted.
+//   - plaintext is encrypted, authenticated and appended to out.
+//   - n is a nonce value which must never be re-used with this key.
+//   - nb is a buffer used for temporary storage in the implementation of this call, which should
+//     be re-used by callers to minimize garbage collection.
+func (s *NebulaCipherState) EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error) {
+	if s != nil {
+		// TODO: Is this okay now that we have made messageCounter atomic?
+		// Alternative may be to split the counter space into ranges
+		//if n <= s.n {
+		//	return nil, errors.New("CRITICAL: a duplicate counter value was used")
+		//}
+		//s.n = n
+		nb[0] = 0
+		nb[1] = 0
+		nb[2] = 0
+		nb[3] = 0
+		noiseEndianness.PutUint64(nb[4:], n)
+		out = s.c.(cipher.AEAD).Seal(out, nb, plaintext, ad)
+		//l.Debugf("Encryption: outlen: %d, nonce: %d, ad: %s, plainlen %d", len(out), n, ad, len(plaintext))
+		return out, nil
+	} else {
+		return nil, errors.New("no cipher state available to encrypt")
+	}
+}
+
+func (s *NebulaCipherState) DecryptDanger(out, ad, ciphertext []byte, n uint64, nb []byte) ([]byte, error) {
+	if s != nil {
+		nb[0] = 0
+		nb[1] = 0
+		nb[2] = 0
+		nb[3] = 0
+		noiseEndianness.PutUint64(nb[4:], n)
+		return s.c.(cipher.AEAD).Open(out, nb, ciphertext, ad)
+	} else {
+		return []byte{}, nil
+	}
+}
+
+func (s *NebulaCipherState) Overhead() int {
+	if s != nil {
+		return s.c.(cipher.AEAD).Overhead()
+	}
+	return 0
+}

noiseutil/aesgcm.go

@@ -1,53 +0,0 @@
-package noiseutil
-
-import (
-	"crypto/cipher"
-	"encoding/binary"
-	"errors"
-
-	"github.com/flynn/noise"
-)
-
-// CipherStateAESGCM is the data-plane wrapper for the AES-GCM AEAD cipher.
-// AES-GCM uses big-endian nonce encoding per the Noise spec.
-type CipherStateAESGCM struct {
-	c cipher.AEAD
-}
-
-// NewCipherStateAESGCM extracts the underlying AEAD from the post-handshake noise.CipherState.
-// The caller is responsible for ensuring the noise cipher is actually AES-GCM,
-// otherwise the type assertion still succeeds but the nonce endianness will be wrong on the wire.
-func NewCipherStateAESGCM(s *noise.CipherState) *CipherStateAESGCM {
-	return &CipherStateAESGCM{c: s.Cipher().(cipher.AEAD)}
-}
-
-func (s *CipherStateAESGCM) EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error) {
-	if s == nil {
-		return nil, errors.New("no cipher state available to encrypt")
-	}
-	nb[0] = 0
-	nb[1] = 0
-	nb[2] = 0
-	nb[3] = 0
-	binary.BigEndian.PutUint64(nb[4:], n)
-	return s.c.Seal(out, nb, plaintext, ad), nil
-}
-
-func (s *CipherStateAESGCM) DecryptDanger(out, ad, ciphertext []byte, n uint64, nb []byte) ([]byte, error) {
-	if s == nil {
-		return []byte{}, nil
-	}
-	nb[0] = 0
-	nb[1] = 0
-	nb[2] = 0
-	nb[3] = 0
-	binary.BigEndian.PutUint64(nb[4:], n)
-	return s.c.Open(out, nb, ciphertext, ad)
-}
-
-func (s *CipherStateAESGCM) Overhead() int {
-	if s == nil {
-		return 0
-	}
-	return s.c.Overhead()
-}

noiseutil/chachapoly.go

@@ -1,52 +0,0 @@
-package noiseutil
-
-import (
-	"crypto/cipher"
-	"encoding/binary"
-	"errors"
-
-	"github.com/flynn/noise"
-)
-
-// CipherStateChaChaPoly is the data-plane wrapper for the ChaCha20-Poly1305 AEAD cipher.
-// ChaCha20-Poly1305 uses little-endian nonce encoding per the Noise spec.
-type CipherStateChaChaPoly struct {
-	c cipher.AEAD
-}
-
-// NewCipherStateChaChaPoly extracts the underlying AEAD from the post-handshake noise.CipherState.
-// The caller is responsible for ensuring the noise cipher is actually ChaCha20-Poly1305.
-func NewCipherStateChaChaPoly(s *noise.CipherState) *CipherStateChaChaPoly {
-	return &CipherStateChaChaPoly{c: s.Cipher().(cipher.AEAD)}
-}
-
-func (s *CipherStateChaChaPoly) EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error) {
-	if s == nil {
-		return nil, errors.New("no cipher state available to encrypt")
-	}
-	nb[0] = 0
-	nb[1] = 0
-	nb[2] = 0
-	nb[3] = 0
-	binary.LittleEndian.PutUint64(nb[4:], n)
-	return s.c.Seal(out, nb, plaintext, ad), nil
-}
-
-func (s *CipherStateChaChaPoly) DecryptDanger(out, ad, ciphertext []byte, n uint64, nb []byte) ([]byte, error) {
-	if s == nil {
-		return []byte{}, nil
-	}
-	nb[0] = 0
-	nb[1] = 0
-	nb[2] = 0
-	nb[3] = 0
-	binary.LittleEndian.PutUint64(nb[4:], n)
-	return s.c.Open(out, nb, ciphertext, ad)
-}
-
-func (s *CipherStateChaChaPoly) Overhead() int {
-	if s == nil {
-		return 0
-	}
-	return s.c.Overhead()
-}

noiseutil/cipher_state.go

@@ -1,40 +0,0 @@
-package noiseutil
-
-import (
-	"fmt"
-
-	"github.com/flynn/noise"
-)
-
-// CipherState is the post-handshake AEAD cipher used for the data plane.
-// Each supported cipher has its own concrete implementation in this package with the nonce endianness hardcoded,
-// so the encrypt/decrypt fast path avoids interface dispatch on the byte order.
-type CipherState interface {
-	// EncryptDanger encrypts and authenticates a given payload.
-	//
-	// out is a destination slice to hold the output of the EncryptDanger operation.
-	//   - ad is additional data, which will be authenticated and appended to out, but not encrypted.
-	//   - plaintext is encrypted, authenticated and appended to out.
-	//   - n is a nonce value which must never be re-used with this key.
-	//   - nb is a scratch buffer used to assemble the nonce.
-	EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error)
-
-	// DecryptDanger authenticates and decrypts a given payload, with the same argument shape as EncryptDanger.
-	DecryptDanger(out, ad, ciphertext []byte, n uint64, nb []byte) ([]byte, error)
-
-	// Overhead returns the AEAD tag size, or 0 if the receiver is nil.
-	Overhead() int
-}
-
-// NewCipherState wraps the post-handshake noise.CipherState in the per-cipher type that matches cipherFunc.
-// cipherFunc must be the same cipher used to build the noise CipherSuite that produced s.
-func NewCipherState(s *noise.CipherState, cipherFunc noise.CipherFunc) CipherState {
-	switch cipherFunc.CipherName() {
-	case CipherAESGCM.CipherName():
-		return NewCipherStateAESGCM(s)
-	case noise.CipherChaChaPoly.CipherName():
-		return NewCipherStateChaChaPoly(s)
-	default:
-		panic(fmt.Sprintf("noiseutil: unsupported cipher %q", cipherFunc.CipherName()))
-	}
-}

noiseutil/cipher_state_test.go

@@ -1,222 +0,0 @@
-package noiseutil
-
-import (
-	"testing"
-
-	"github.com/flynn/noise"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestCipherStateAESGCMRoundtrip(t *testing.T) {
-	enc, dec := buildCipherStates(t, CipherAESGCM)
-	roundtrip(t, NewCipherStateAESGCM(enc), NewCipherStateAESGCM(dec))
-}
-
-func TestCipherStateChaChaPolyRoundtrip(t *testing.T) {
-	enc, dec := buildCipherStates(t, noise.CipherChaChaPoly)
-	roundtrip(t, NewCipherStateChaChaPoly(enc), NewCipherStateChaChaPoly(dec))
-}
-
-func TestNewCipherStateDispatch(t *testing.T) {
-	encA, _ := buildCipherStates(t, CipherAESGCM)
-	encC, _ := buildCipherStates(t, noise.CipherChaChaPoly)
-
-	assert.IsType(t, &CipherStateAESGCM{}, NewCipherState(encA, CipherAESGCM))
-	assert.IsType(t, &CipherStateChaChaPoly{}, NewCipherState(encC, noise.CipherChaChaPoly))
-}
-
-func TestNewCipherStateUnsupportedPanics(t *testing.T) {
-	enc, _ := buildCipherStates(t, CipherAESGCM)
-	assert.Panics(t, func() {
-		NewCipherState(enc, fakeCipher{})
-	})
-}
-
-type fakeCipher struct{}
-
-func (fakeCipher) Cipher(k [32]byte) noise.Cipher { return nil }
-func (fakeCipher) CipherName() string             { return "Fake" }
-
-// buildCipherStates runs an in-memory NN handshake with the requested cipher
-// to produce a pair of post-handshake CipherStates that share keys.
-func buildCipherStates(t *testing.T, c noise.CipherFunc) (*noise.CipherState, *noise.CipherState) {
-	t.Helper()
-	suite := noise.NewCipherSuite(noise.DH25519, c, noise.HashSHA256)
-	cfg := noise.Config{CipherSuite: suite, Pattern: noise.HandshakeNN}
-	cfg.Initiator = true
-	hsI, err := noise.NewHandshakeState(cfg)
-	require.NoError(t, err)
-	cfg.Initiator = false
-	hsR, err := noise.NewHandshakeState(cfg)
-	require.NoError(t, err)
-
-	msg, _, _, err := hsI.WriteMessage(nil, nil)
-	require.NoError(t, err)
-	_, _, _, err = hsR.ReadMessage(nil, msg)
-	require.NoError(t, err)
-
-	msg, dR, _, err := hsR.WriteMessage(nil, nil)
-	require.NoError(t, err)
-	_, eI, _, err := hsI.ReadMessage(nil, msg)
-	require.NoError(t, err)
-	require.NotNil(t, eI)
-	require.NotNil(t, dR)
-
-	// noise returns (cs1, cs2) where cs1 is the initiator->responder cipher.
-	return eI, dR
-}
-
-func roundtrip(t *testing.T, enc, dec CipherState) {
-	t.Helper()
-	plaintext := []byte("nebula cipher state roundtrip")
-	ad := []byte("aad")
-	nb := make([]byte, 12)
-
-	ct, err := enc.EncryptDanger(nil, ad, plaintext, 1, nb)
-	require.NoError(t, err)
-	assert.NotEqual(t, plaintext, ct)
-
-	pt, err := dec.DecryptDanger(nil, ad, ct, 1, nb)
-	require.NoError(t, err)
-	assert.Equal(t, plaintext, pt)
-
-	// Wrong nonce must fail authentication.
-	_, err = dec.DecryptDanger(nil, ad, ct, 2, nb)
-	require.Error(t, err)
-
-	assert.Equal(t, enc.Overhead(), dec.Overhead())
-	assert.Equal(t, 16, enc.Overhead())
-}
-
-func BenchmarkCipherStateEncryptAESGCM(b *testing.B) {
-	enc, _ := buildCipherStatesB(b, CipherAESGCM)
-	benchEncryptCipherState(b, NewCipherState(enc, CipherAESGCM))
-}
-
-func BenchmarkCipherStateEncryptChaChaPoly(b *testing.B) {
-	enc, _ := buildCipherStatesB(b, noise.CipherChaChaPoly)
-	benchEncryptCipherState(b, NewCipherState(enc, noise.CipherChaChaPoly))
-}
-
-func benchEncryptCipherState(b *testing.B, cs CipherState) {
-	plaintext := make([]byte, 1280)
-	ad := make([]byte, 16)
-	nb := make([]byte, 12)
-	out := make([]byte, 0, len(plaintext)+cs.Overhead())
-	b.ResetTimer()
-	b.ReportAllocs()
-	for i := 0; i < b.N; i++ {
-		var err error
-		out, err = cs.EncryptDanger(out[:0], ad, plaintext, uint64(i+1), nb)
-		if err != nil {
-			b.Fatal(err)
-		}
-	}
-}
-
-func buildCipherStatesB(b *testing.B, c noise.CipherFunc) (*noise.CipherState, *noise.CipherState) {
-	b.Helper()
-	suite := noise.NewCipherSuite(noise.DH25519, c, noise.HashSHA256)
-	cfg := noise.Config{CipherSuite: suite, Pattern: noise.HandshakeNN}
-	cfg.Initiator = true
-	hsI, err := noise.NewHandshakeState(cfg)
-	if err != nil {
-		b.Fatal(err)
-	}
-	cfg.Initiator = false
-	hsR, err := noise.NewHandshakeState(cfg)
-	if err != nil {
-		b.Fatal(err)
-	}
-	msg, _, _, err := hsI.WriteMessage(nil, nil)
-	if err != nil {
-		b.Fatal(err)
-	}
-	if _, _, _, err := hsR.ReadMessage(nil, msg); err != nil {
-		b.Fatal(err)
-	}
-	msg, dR, _, err := hsR.WriteMessage(nil, nil)
-	if err != nil {
-		b.Fatal(err)
-	}
-	_, eI, _, err := hsI.ReadMessage(nil, msg)
-	if err != nil {
-		b.Fatal(err)
-	}
-	return eI, dR
-}
-
-// TestDecryptDangerRelayShapeNoAlloc covers the AD-only relay path used in
-// outside.go's handleOutsideRelayPacket: the body is AD, the trailing 16 bytes
-// are the AEAD tag, the plaintext is empty, and the caller passes nil as the
-// destination because it only needs the auth side-effect. The call must
-// succeed, return an empty plaintext, and not allocate on the hot path.
-func TestDecryptDangerRelayShapeNoAlloc(t *testing.T) {
-	cases := []struct {
-		name string
-		c    noise.CipherFunc
-		wrap func(*noise.CipherState) CipherState
-	}{
-		{"AESGCM", CipherAESGCM, func(cs *noise.CipherState) CipherState { return NewCipherStateAESGCM(cs) }},
-		{"ChaChaPoly", noise.CipherChaChaPoly, func(cs *noise.CipherState) CipherState { return NewCipherStateChaChaPoly(cs) }},
-	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			encCS, decCS := buildCipherStates(t, tc.c)
-			enc, dec := tc.wrap(encCS), tc.wrap(decCS)
-
-			ad := make([]byte, 1200) // typical relay packet body size
-			for i := range ad {
-				ad[i] = byte(i)
-			}
-			nb := make([]byte, 12)
-
-			// Build the "signature value" the way handleOutsideRelayPacket sees it:
-			// empty plaintext encrypted with the body as AD yields just the 16-byte tag.
-			tag, err := enc.EncryptDanger(nil, ad, nil, 1, nb)
-			require.NoError(t, err)
-			require.Len(t, tag, dec.Overhead())
-
-			// Sanity: the relay-shaped call returns empty plaintext, no error.
-			out, err := dec.DecryptDanger(nil, ad, tag, 1, nb)
-			require.NoError(t, err)
-			assert.Empty(t, out)
-
-			// Tampering with the AD must fail authentication.
-			ad[0] ^= 0xff
-			_, err = dec.DecryptDanger(nil, ad, tag, 1, nb)
-			require.Error(t, err)
-			ad[0] ^= 0xff
-
-			// The hot path must not allocate. AllocsPerRun does a warm-up run, so any
-			// one-time setup is excluded. Counter has to advance so the AEAD nonce is
-			// unique per call, but we don't care whether the auth succeeds — we only
-			// care about whether the call path allocates.
-			var counter uint64 = 2
-			allocs := testing.AllocsPerRun(100, func() {
-				_, _ = dec.DecryptDanger(nil, ad, tag, counter, nb)
-				counter++
-			})
-			assert.Equal(t, 0.0, allocs, "DecryptDanger(nil, ...) must not allocate")
-		})
-	}
-}
-
-func TestCipherStateNilSafety(t *testing.T) {
-	var aes *CipherStateAESGCM
-	_, err := aes.EncryptDanger(nil, nil, nil, 0, make([]byte, 12))
-	require.Error(t, err)
-	out, err := aes.DecryptDanger(nil, nil, nil, 0, make([]byte, 12))
-	require.NoError(t, err)
-	assert.Empty(t, out)
-	assert.Equal(t, 0, aes.Overhead())
-
-	var cc *CipherStateChaChaPoly
-	_, err = cc.EncryptDanger(nil, nil, nil, 0, make([]byte, 12))
-	require.Error(t, err)
-	out, err = cc.DecryptDanger(nil, nil, nil, 0, make([]byte, 12))
-	require.NoError(t, err)
-	assert.Empty(t, out)
-	assert.Equal(t, 0, cc.Overhead())
-}

outside.go

@@ -1,16 +1,15 @@
 package nebula
 
 import (
-	"context"
 	"encoding/binary"
 	"errors"
-	"log/slog"
 	"net/netip"
 	"time"
 
 	"github.com/google/gopacket/layers"
 	"golang.org/x/net/ipv6"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"golang.org/x/net/ipv4"
@@ -20,241 +19,215 @@ const (
 	minFwPacketLen = 4
 )
 
-var ErrOutOfWindow = errors.New("out of window packet")
-
-func (f *Interface) readOutsidePackets(via ViaSender, packet []byte, h *header.H, fwPacket *firewall.Packet, lhf *LightHouseHandler, nb []byte, q int, localCache firewall.ConntrackCache) {
+func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte, h *header.H, fwPacket *firewall.Packet, fwCtx *firewall.PacketContext, lhf *LightHouseHandler, nb []byte, q int, localCache firewall.ConntrackCache) {
 	err := h.Parse(packet)
 	if err != nil {
 		// Hole punch packets are 0 or 1 byte big, so lets ignore printing those errors
-		// TODO: record metrics for rx holepunch/punchy packets?
 		if len(packet) > 1 {
-			f.messageMetrics.RxInvalid(1)
-			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				f.l.Debug("Error while parsing inbound packet",
-					"from", via,
-					"error", err,
-					"packet", packet,
-				)
-			}
-		}
-		return
-	}
-
-	if h.Version != header.Version {
-		f.messageMetrics.RxInvalid(1)
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("Unexpected header version received", "from", via)
-		}
-		return
-	}
-
-	// Check before processing to see if this is a expected type/subtype
-	if !h.IsValidSubType() {
-		f.messageMetrics.RxInvalid(1)
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.Debug("Unexpected packet received", "from", via)
+			f.l.WithField("packet", packet).Infof("Error while parsing inbound packet from %s: %s", via, err)
 		}
 		return
 	}
 
+	//l.Error("in packet ", header, packet[HeaderLen:])
 	if !via.IsRelayed {
 		if f.myVpnNetworksTable.Contains(via.UdpAddr.Addr()) {
-			f.messageMetrics.RxInvalid(1)
-			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				f.l.Debug("Refusing to process double encrypted packet", "from", via)
+			if f.l.Level >= logrus.DebugLevel {
+				f.l.WithField("from", via).Debug("Refusing to process double encrypted packet")
 			}
 			return
 		}
 	}
 
-	// don't keep Rx metrics for message type, since you can see those in the tun metrics
-	if h.Type != header.Message {
-		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
-	}
-
-	// Unencrypted packets
-	switch h.Type {
-	case header.Handshake:
-		f.handshakeManager.HandleIncoming(via, packet, h)
-		return
-
-	case header.RecvError:
-		f.handleRecvError(via.UdpAddr, h)
-		return
-	}
-
-	// Relay packets are special
-	isMessageRelay := (h.Type == header.Message && h.Subtype == header.MessageRelay)
-
 	var hostinfo *HostInfo
-	if isMessageRelay {
+	// verify if we've seen this index before, otherwise respond to the handshake initiation
+	if h.Type == header.Message && h.Subtype == header.MessageRelay {
 		hostinfo = f.hostMap.QueryRelayIndex(h.RemoteIndex)
 	} else {
 		hostinfo = f.hostMap.QueryIndex(h.RemoteIndex)
 	}
 
-	// At this point we should have a valid existing tunnel, verify and send
-	// recvError if necessary
-	if hostinfo == nil || hostinfo.ConnectionState == nil {
-		if !via.IsRelayed {
-			f.maybeSendRecvError(via.UdpAddr, h.RemoteIndex)
-		}
-		return
+	var ci *ConnectionState
+	if hostinfo != nil {
+		ci = hostinfo.ConnectionState
 	}
 
-	// All remaining packets are encrypted
-	ci := hostinfo.ConnectionState
-	if !ci.window.Check(f.l, h.MessageCounter) {
-		return
-	}
-
-	// Relay packets are special
-	if isMessageRelay {
-		f.handleOutsideRelayPacket(hostinfo, via, packet, h, fwPacket, lhf, nb, q, localCache)
-		return
-	}
-
-	out := f.batchers[q].Reserve(len(packet))[:0]
-	out, err = f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
-	if err != nil {
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).Debug("Failed to decrypt packet",
-				"error", err,
-				"from", via,
-				"header", h,
-			)
-		}
-		return
-	}
-
-	// Roam before we respond
-	f.handleHostRoaming(hostinfo, via)
-	f.connectionManager.In(hostinfo)
-
 	switch h.Type {
 	case header.Message:
+		if !f.handleEncrypted(ci, via, h) {
+			return
+		}
+
 		switch h.Subtype {
 		case header.MessageNone:
-			f.handleOutsideMessagePacket(hostinfo, out, packet, fwPacket, nb, q, localCache)
-		default:
-			hostinfo.logger(f.l).Error("IsValidSubType was true, but unexpected message subtype seen", "from", via, "header", h)
-			return
+			if !f.decryptToTun(hostinfo, h.MessageCounter, out, packet, fwPacket, fwCtx, nb, q, localCache) {
+				return
+			}
+		case header.MessageRelay:
+			// The entire body is sent as AD, not encrypted.
+			// The packet consists of a 16-byte parsed Nebula header, Associated Data-protected payload, and a trailing 16-byte AEAD signature value.
+			// The packet is guaranteed to be at least 16 bytes at this point, b/c it got past the h.Parse() call above. If it's
+			// otherwise malformed (meaning, there is no trailing 16 byte AEAD value), then this will result in at worst a 0-length slice
+			// which will gracefully fail in the DecryptDanger call.
+			signedPayload := packet[:len(packet)-hostinfo.ConnectionState.dKey.Overhead()]
+			signatureValue := packet[len(packet)-hostinfo.ConnectionState.dKey.Overhead():]
+			out, err = hostinfo.ConnectionState.dKey.DecryptDanger(out, signedPayload, signatureValue, h.MessageCounter, nb)
+			if err != nil {
+				return
+			}
+			// Successfully validated the thing. Get rid of the Relay header.
+			signedPayload = signedPayload[header.Len:]
+			// Pull the Roaming parts up here, and return in all call paths.
+			f.handleHostRoaming(hostinfo, via)
+			// Track usage of both the HostInfo and the Relay for the received & authenticated packet
+			f.connectionManager.In(hostinfo)
+			f.connectionManager.RelayUsed(h.RemoteIndex)
+
+			relay, ok := hostinfo.relayState.QueryRelayForByIdx(h.RemoteIndex)
+			if !ok {
+				// The only way this happens is if hostmap has an index to the correct HostInfo, but the HostInfo is missing
+				// its internal mapping. This should never happen.
+				hostinfo.logger(f.l).WithFields(logrus.Fields{"vpnAddrs": hostinfo.vpnAddrs, "remoteIndex": h.RemoteIndex}).Error("HostInfo missing remote relay index")
+				return
+			}
+
+			switch relay.Type {
+			case TerminalType:
+				// If I am the target of this relay, process the unwrapped packet
+				// From this recursive point, all these variables are 'burned'. We shouldn't rely on them again.
+				via = ViaSender{
+					UdpAddr:   via.UdpAddr,
+					relayHI:   hostinfo,
+					remoteIdx: relay.RemoteIndex,
+					relay:     relay,
+					IsRelayed: true,
+				}
+				f.readOutsidePackets(via, out[:0], signedPayload, h, fwPacket, fwCtx, lhf, nb, q, localCache)
+				return
+			case ForwardingType:
+				// Find the target HostInfo relay object
+				targetHI, targetRelay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relay.PeerAddr)
+				if err != nil {
+					hostinfo.logger(f.l).WithField("relayTo", relay.PeerAddr).WithError(err).WithField("hostinfo.vpnAddrs", hostinfo.vpnAddrs).Info("Failed to find target host info by ip")
+					return
+				}
+
+				// If that relay is Established, forward the payload through it
+				if targetRelay.State == Established {
+					switch targetRelay.Type {
+					case ForwardingType:
+						// Forward this packet through the relay tunnel
+						// Find the target HostInfo
+						f.SendVia(targetHI, targetRelay, signedPayload, nb, out, false)
+						return
+					case TerminalType:
+						hostinfo.logger(f.l).Error("Unexpected Relay Type of Terminal")
+					}
+				} else {
+					hostinfo.logger(f.l).WithFields(logrus.Fields{"relayTo": relay.PeerAddr, "relayFrom": hostinfo.vpnAddrs[0], "targetRelayState": targetRelay.State}).Info("Unexpected target relay state")
+					return
+				}
+			}
 		}
 
 	case header.LightHouse:
+		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
+		if !f.handleEncrypted(ci, via, h) {
+			return
+		}
+
+		d, err := f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
+		if err != nil {
+			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+				WithField("packet", packet).
+				Error("Failed to decrypt lighthouse packet")
+			return
+		}
+
 		//TODO: assert via is not relayed
-		lhf.HandleRequest(via.UdpAddr, hostinfo.vpnAddrs, out, f)
+		lhf.HandleRequest(via.UdpAddr, hostinfo.vpnAddrs, d, f)
+
+		// Fallthrough to the bottom to record incoming traffic
 
 	case header.Test:
-		switch h.Subtype {
-		case header.TestReply:
-			// No-op, useful for the Roaming and connectionManager side-effects above
-		case header.TestRequest:
-			f.send(header.Test, header.TestReply, ci, hostinfo, out, nb, out)
-		default:
-			hostinfo.logger(f.l).Error("IsValidSubType was true, but unexpected test subtype seen", "from", via, "header", h)
+		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
+		if !f.handleEncrypted(ci, via, h) {
 			return
 		}
 
+		d, err := f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
+		if err != nil {
+			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+				WithField("packet", packet).
+				Error("Failed to decrypt test packet")
+			return
+		}
+
+		if h.Subtype == header.TestRequest {
+			// This testRequest might be from TryPromoteBest, so we should roam
+			// to the new IP address before responding
+			f.handleHostRoaming(hostinfo, via)
+			f.send(header.Test, header.TestReply, ci, hostinfo, d, nb, out)
+		}
+
+		// Fallthrough to the bottom to record incoming traffic
+
+		// Non encrypted messages below here, they should not fall through to avoid tracking incoming traffic since they
+		// are unauthenticated
+
+	case header.Handshake:
+		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
+		f.handshakeManager.HandleIncoming(via, packet, h)
+		return
+
+	case header.RecvError:
+		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
+		f.handleRecvError(via.UdpAddr, h)
+		return
+
 	case header.CloseTunnel:
-		hostinfo.logger(f.l).Info("Close tunnel received, tearing down.", "from", via)
+		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
+		if !f.handleEncrypted(ci, via, h) {
+			return
+		}
+		_, err = f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
+		if err != nil {
+			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+				WithField("packet", packet).
+				Error("Failed to decrypt CloseTunnel packet")
+			return
+		}
+
+		hostinfo.logger(f.l).WithField("from", via).
+			Info("Close tunnel received, tearing down.")
+
 		f.closeTunnel(hostinfo)
+		return
 
 	case header.Control:
-		f.relayManager.HandleControlMsg(hostinfo, out, f)
-
-	default:
-		hostinfo.logger(f.l).Error("IsValidSubType was true, but unexpected message type seen", "from", via, "header", h)
-	}
-}
-
-func (f *Interface) handleOutsideRelayPacket(hostinfo *HostInfo, via ViaSender, packet []byte, h *header.H, fwPacket *firewall.Packet, lhf *LightHouseHandler, nb []byte, q int, localCache firewall.ConntrackCache) {
-	// The entire body is sent as AD, not encrypted.
-	// The packet consists of a 16-byte parsed Nebula header, Associated Data-protected payload, and a trailing 16-byte AEAD signature value.
-	// The packet is guaranteed to be at least 16 bytes at this point, b/c it got past the h.Parse() call above. If it's
-	// otherwise malformed (meaning, there is no trailing 16 byte AEAD value), then this will result in at worst a 0-length slice
-	// which will gracefully fail in the DecryptDanger call.
-	signedPayload := packet[:len(packet)-hostinfo.ConnectionState.dKey.Overhead()]
-	signatureValue := packet[len(packet)-hostinfo.ConnectionState.dKey.Overhead():]
-	// The decrypted output is empty (relay packets carry their payload as AD) and unused.
-	// The recursive readOutsidePackets call below operates on signedPayload. Passing
-	// nil avoids reserving an arena slot.
-	if _, err := hostinfo.ConnectionState.dKey.DecryptDanger(nil, signedPayload, signatureValue, h.MessageCounter, nb); err != nil {
-		return
-	}
-	// Successfully validated the thing. Get rid of the Relay header.
-	signedPayload = signedPayload[header.Len:]
-	// Pull the Roaming parts up here, and return in all call paths.
-	f.handleHostRoaming(hostinfo, via)
-	// Track usage of both the HostInfo and the Relay for the received & authenticated packet
-	f.connectionManager.In(hostinfo)
-	f.connectionManager.RelayUsed(h.RemoteIndex)
-
-	relay, ok := hostinfo.relayState.QueryRelayForByIdx(h.RemoteIndex)
-	if !ok {
-		// The only way this happens is if hostmap has an index to the correct HostInfo, but the HostInfo is missing
-		// its internal mapping. This should never happen.
-		hostinfo.logger(f.l).Error("HostInfo missing remote relay index",
-			"vpnAddrs", hostinfo.vpnAddrs,
-			"remoteIndex", h.RemoteIndex,
-		)
-		return
-	}
-
-	switch relay.Type {
-	case TerminalType:
-		// If I am the target of this relay, process the unwrapped packet
-		// From this recursive point, all these variables are 'burned'. We shouldn't rely on them again.
-		via = ViaSender{
-			UdpAddr:   via.UdpAddr,
-			relayHI:   hostinfo,
-			remoteIdx: relay.RemoteIndex,
-			relay:     relay,
-			IsRelayed: true,
+		if !f.handleEncrypted(ci, via, h) {
+			return
 		}
-		f.readOutsidePackets(via, signedPayload, h, fwPacket, lhf, nb, q, localCache)
-	case ForwardingType:
-		// Find the target HostInfo relay object
-		targetHI, targetRelay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relay.PeerAddr)
+
+		d, err := f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
 		if err != nil {
-			hostinfo.logger(f.l).Info("Failed to find target host info by ip",
-				"relayTo", relay.PeerAddr,
-				"error", err,
-				"hostinfo.vpnAddrs", hostinfo.vpnAddrs,
-			)
+			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+				WithField("packet", packet).
+				Error("Failed to decrypt Control packet")
 			return
 		}
 
-		// If that relay is Established, forward the payload through it
-		if targetRelay.State == Established {
-			switch targetRelay.Type {
-			case ForwardingType:
-				// Forward this packet through the relay tunnel
-				// Find the target HostInfo //todo it would potentially be nice to batch these
-				out := f.batchers[q].Reserve(len(packet) + header.Len + hostinfo.ConnectionState.dKey.Overhead())[:0]
-				f.SendVia(targetHI, targetRelay, signedPayload, nb, out, false)
-			case TerminalType:
-				hostinfo.logger(f.l).Error("Unexpected Relay Type of Terminal")
-				return
-			default:
-				if f.l.Enabled(context.Background(), slog.LevelDebug) {
-					hostinfo.logger(f.l).Debug("Unexpected targetRelay Type", "from", via, "relayType", targetRelay.Type)
-				}
-				return
-			}
-		} else {
-			hostinfo.logger(f.l).Info("Unexpected target relay state",
-				"relayTo", relay.PeerAddr,
-				"relayFrom", hostinfo.vpnAddrs[0],
-				"targetRelayState", targetRelay.State,
-			)
-			return
-		}
+		f.relayManager.HandleControlMsg(hostinfo, d, f)
+
 	default:
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).Debug("Unexpected relay type", "from", via, "relayType", relay.Type)
-		}
+		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
+		hostinfo.logger(f.l).Debugf("Unexpected packet received from %s", via)
+		return
 	}
+
+	f.handleHostRoaming(hostinfo, via)
+
+	f.connectionManager.In(hostinfo)
 }
 
 // closeTunnel closes a tunnel locally, it does not send a closeTunnel packet to the remote
@@ -274,27 +247,20 @@ func (f *Interface) sendCloseTunnel(h *HostInfo) {
 func (f *Interface) handleHostRoaming(hostinfo *HostInfo, via ViaSender) {
 	if !via.IsRelayed && hostinfo.remote != via.UdpAddr {
 		if !f.lightHouse.GetRemoteAllowList().AllowAll(hostinfo.vpnAddrs, via.UdpAddr.Addr()) {
-			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				hostinfo.logger(f.l).Debug("lighthouse.remote_allow_list denied roaming", "newAddr", via.UdpAddr)
-			}
+			hostinfo.logger(f.l).WithField("newAddr", via.UdpAddr).Debug("lighthouse.remote_allow_list denied roaming")
 			return
 		}
 
 		if !hostinfo.lastRoam.IsZero() && via.UdpAddr == hostinfo.lastRoamRemote && time.Since(hostinfo.lastRoam) < RoamingSuppressSeconds*time.Second {
-			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				hostinfo.logger(f.l).Debug("Suppressing roam back to previous remote",
-					"suppressSeconds", RoamingSuppressSeconds,
-					"udpAddr", hostinfo.remote,
-					"newAddr", via.UdpAddr,
-				)
+			if f.l.Level >= logrus.DebugLevel {
+				hostinfo.logger(f.l).WithField("udpAddr", hostinfo.remote).WithField("newAddr", via.UdpAddr).
+					Debugf("Suppressing roam back to previous remote for %d seconds", RoamingSuppressSeconds)
 			}
 			return
 		}
 
-		hostinfo.logger(f.l).Info("Host roamed to new udp ip/port.",
-			"udpAddr", hostinfo.remote,
-			"newAddr", via.UdpAddr,
-		)
+		hostinfo.logger(f.l).WithField("udpAddr", hostinfo.remote).WithField("newAddr", via.UdpAddr).
+			Info("Host roamed to new udp ip/port.")
 		hostinfo.lastRoam = time.Now()
 		hostinfo.lastRoamRemote = hostinfo.remote
 		hostinfo.SetRemote(via.UdpAddr)
@@ -302,6 +268,23 @@ func (f *Interface) handleHostRoaming(hostinfo *HostInfo, via ViaSender) {
 
 }
 
+// handleEncrypted returns true if a packet should be processed, false otherwise
+func (f *Interface) handleEncrypted(ci *ConnectionState, via ViaSender, h *header.H) bool {
+	// If connectionstate does not exist, send a recv error, if possible, to encourage a fast reconnect
+	if ci == nil {
+		if !via.IsRelayed {
+			f.maybeSendRecvError(via.UdpAddr, h.RemoteIndex)
+		}
+		return false
+	}
+	// If the window check fails, refuse to process the packet, but don't send a recv error
+	if !ci.window.Check(f.l, h.MessageCounter) {
+		return false
+	}
+
+	return true
+}
+
 var (
 	ErrPacketTooShort          = errors.New("packet is too short")
 	ErrUnknownIPVersion        = errors.New("packet is an unknown ip version")
@@ -312,7 +295,10 @@ var (
 )
 
 // newPacket validates and parses the interesting bits for the firewall out of the ip and sub protocol headers
-func newPacket(data []byte, incoming bool, fp *firewall.Packet) error {
+func newPacket(data []byte, incoming bool, fp *firewall.Packet, ctx *firewall.PacketContext) error {
+	if ctx != nil {
+		*ctx = firewall.PacketContext{}
+	}
 	if len(data) < 1 {
 		return ErrPacketTooShort
 	}
@@ -320,14 +306,14 @@ func newPacket(data []byte, incoming bool, fp *firewall.Packet) error {
 	version := int((data[0] >> 4) & 0x0f)
 	switch version {
 	case ipv4.Version:
-		return parseV4(data, incoming, fp)
+		return parseV4(data, incoming, fp, ctx)
 	case ipv6.Version:
-		return parseV6(data, incoming, fp)
+		return parseV6(data, incoming, fp, ctx)
 	}
 	return ErrUnknownIPVersion
 }
 
-func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
+func parseV6(data []byte, incoming bool, fp *firewall.Packet, ctx *firewall.PacketContext) error {
 	dataLen := len(data)
 	if dataLen < ipv6.HeaderLen {
 		return ErrIPv6PacketTooShort
@@ -372,6 +358,11 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 				fp.RemotePort = 0
 			}
 			fp.Fragment = false
+			if ctx != nil {
+				ctx.Length = binary.BigEndian.Uint16(data[4:6]) + uint16(ipv6.HeaderLen)
+				ctx.ICMPType = data[offset]
+				ctx.ICMPCode = data[offset+1]
+			}
 			return nil
 
 		case layers.IPProtocolTCP, layers.IPProtocolUDP:
@@ -389,6 +380,12 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 			}
 
 			fp.Fragment = false
+			if ctx != nil {
+				ctx.Length = binary.BigEndian.Uint16(data[4:6]) + uint16(ipv6.HeaderLen)
+				if proto == layers.IPProtocolTCP && dataLen >= offset+14 {
+					ctx.TCPFlags = data[offset+13]
+				}
+			}
 			return nil
 
 		case layers.IPProtocolIPv6Fragment:
@@ -440,7 +437,7 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
 	return ErrIPv6CouldNotFindPayload
 }
 
-func parseV4(data []byte, incoming bool, fp *firewall.Packet) error {
+func parseV4(data []byte, incoming bool, fp *firewall.Packet, ctx *firewall.PacketContext) error {
 	// Do we at least have an ipv4 header worth of data?
 	if len(data) < ipv4.HeaderLen {
 		return ErrIPv4PacketTooShort
@@ -497,6 +494,21 @@ func parseV4(data []byte, incoming bool, fp *firewall.Packet) error {
 		fp.RemotePort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4]) //dst port
 	}
 
+	if ctx != nil {
+		ctx.Length = binary.BigEndian.Uint16(data[2:4])
+		if !fp.Fragment {
+			switch fp.Protocol {
+			case firewall.ProtoICMP:
+				ctx.ICMPType = data[ihl]
+				ctx.ICMPCode = data[ihl+1]
+			case firewall.ProtoTCP:
+				if len(data) >= ihl+14 {
+					ctx.TCPFlags = data[ihl+13]
+				}
+			}
+		}
+	}
+
 	return nil
 }
 
@@ -508,40 +520,55 @@ func (f *Interface) decrypt(hostinfo *HostInfo, mc uint64, out []byte, packet []
 	}
 
 	if !hostinfo.ConnectionState.window.Update(f.l, mc) {
-		return nil, ErrOutOfWindow
+		hostinfo.logger(f.l).WithField("header", h).
+			Debugln("dropping out of window packet")
+		return nil, errors.New("out of window packet")
 	}
 
 	return out, nil
 }
 
-func (f *Interface) handleOutsideMessagePacket(hostinfo *HostInfo, out []byte, packet []byte, fwPacket *firewall.Packet, nb []byte, q int, localCache firewall.ConntrackCache) {
-	err := newPacket(out, true, fwPacket)
+func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out []byte, packet []byte, fwPacket *firewall.Packet, fwCtx *firewall.PacketContext, nb []byte, q int, localCache firewall.ConntrackCache) bool {
+	var err error
+
+	out, err = hostinfo.ConnectionState.dKey.DecryptDanger(out, packet[:header.Len], packet[header.Len:], messageCounter, nb)
 	if err != nil {
-		hostinfo.logger(f.l).Warn("Error while validating inbound packet",
-			"error", err,
-			"packet", out,
-		)
-		return
+		hostinfo.logger(f.l).WithError(err).Error("Failed to decrypt packet")
+		return false
 	}
 
-	dropReason := f.firewall.Drop(*fwPacket, true, hostinfo, f.pki.GetCAPool(), localCache)
+	err = newPacket(out, true, fwPacket, fwCtx)
+	if err != nil {
+		hostinfo.logger(f.l).WithError(err).WithField("packet", out).
+			Warnf("Error while validating inbound packet")
+		return false
+	}
+
+	if !hostinfo.ConnectionState.window.Update(f.l, messageCounter) {
+		hostinfo.logger(f.l).WithField("fwPacket", fwPacket).
+			Debugln("dropping out of window packet")
+		return false
+	}
+
+	dropReason := f.firewall.Drop(*fwPacket, *fwCtx, true, hostinfo, f.pki.GetCAPool(), localCache)
 	if dropReason != nil {
 		// NOTE: We give `packet` as the `out` here since we already decrypted from it and we don't need it anymore
 		// This gives us a buffer to build the reject packet in
 		f.rejectOutside(out, hostinfo.ConnectionState, hostinfo, nb, packet, q)
-		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).Debug("dropping inbound packet",
-				"fwPacket", fwPacket,
-				"reason", dropReason,
-			)
+		if f.l.Level >= logrus.DebugLevel {
+			hostinfo.logger(f.l).WithField("fwPacket", fwPacket).
+				WithField("reason", dropReason).
+				Debugln("dropping inbound packet")
 		}
-		return
+		return false
 	}
 
-	err = f.batchers[q].Commit(out)
+	f.connectionManager.In(hostinfo)
+	_, err = f.readers[q].Write(out)
 	if err != nil {
-		f.l.Error("Failed to write to tun", "error", err)
+		f.l.WithError(err).Error("Failed to write to tun")
 	}
+	return true
 }
 
 func (f *Interface) maybeSendRecvError(endpoint netip.AddrPort, index uint32) {
@@ -555,41 +582,35 @@ func (f *Interface) sendRecvError(endpoint netip.AddrPort, index uint32) {
 
 	b := header.Encode(make([]byte, header.Len), header.Version, header.RecvError, 0, index, 0)
 	_ = f.outside.WriteTo(b, endpoint)
-	if f.l.Enabled(context.Background(), slog.LevelDebug) {
-		f.l.Debug("Recv error sent",
-			"index", index,
-			"udpAddr", endpoint,
-		)
+	if f.l.Level >= logrus.DebugLevel {
+		f.l.WithField("index", index).
+			WithField("udpAddr", endpoint).
+			Debug("Recv error sent")
 	}
 }
 
 func (f *Interface) handleRecvError(addr netip.AddrPort, h *header.H) {
 	if !f.acceptRecvErrorConfig.ShouldRecvError(addr) {
-		f.l.Debug("Recv error received, ignoring",
-			"index", h.RemoteIndex,
-			"udpAddr", addr,
-		)
+		f.l.WithField("index", h.RemoteIndex).
+			WithField("udpAddr", addr).
+			Debug("Recv error received, ignoring")
 		return
 	}
 
-	if f.l.Enabled(context.Background(), slog.LevelDebug) {
-		f.l.Debug("Recv error received",
-			"index", h.RemoteIndex,
-			"udpAddr", addr,
-		)
+	if f.l.Level >= logrus.DebugLevel {
+		f.l.WithField("index", h.RemoteIndex).
+			WithField("udpAddr", addr).
+			Debug("Recv error received")
 	}
 
 	hostinfo := f.hostMap.QueryReverseIndex(h.RemoteIndex)
 	if hostinfo == nil {
-		f.l.Debug("Did not find remote index in main hostmap", "remoteIndex", h.RemoteIndex)
+		f.l.WithField("remoteIndex", h.RemoteIndex).Debugln("Did not find remote index in main hostmap")
 		return
 	}
 
 	if hostinfo.remote.IsValid() && hostinfo.remote != addr {
-		f.l.Info("Someone spoofing recv_errors?",
-			"addr", addr,
-			"hostinfoRemote", hostinfo.remote,
-		)
+		f.l.Infoln("Someone spoofing recv_errors? ", addr, hostinfo.remote)
 		return
 	}
 

outside_test.go

@@ -20,13 +20,13 @@ func Test_newPacket(t *testing.T) {
 	p := &firewall.Packet{}
 
 	// length fails
-	err := newPacket([]byte{}, true, p)
+	err := newPacket([]byte{}, true, p, nil)
 	require.ErrorIs(t, err, ErrPacketTooShort)
 
-	err = newPacket([]byte{0x40}, true, p)
+	err = newPacket([]byte{0x40}, true, p, nil)
 	require.ErrorIs(t, err, ErrIPv4PacketTooShort)
 
-	err = newPacket([]byte{0x60}, true, p)
+	err = newPacket([]byte{0x60}, true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
 
 	// length fail with ip options
@@ -39,15 +39,15 @@ func Test_newPacket(t *testing.T) {
 	}
 
 	b, _ := h.Marshal()
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.ErrorIs(t, err, ErrIPv4InvalidHeaderLength)
 
 	// not an ipv4 packet
-	err = newPacket([]byte{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, true, p)
+	err = newPacket([]byte{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, true, p, nil)
 	require.ErrorIs(t, err, ErrUnknownIPVersion)
 
 	// invalid ihl
-	err = newPacket([]byte{4<<4 | (8 >> 2 & 0x0f), 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, true, p)
+	err = newPacket([]byte{4<<4 | (8 >> 2 & 0x0f), 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, true, p, nil)
 	require.ErrorIs(t, err, ErrIPv4InvalidHeaderLength)
 
 	// account for variable ip header length - incoming
@@ -62,7 +62,7 @@ func Test_newPacket(t *testing.T) {
 
 	b, _ = h.Marshal()
 	b = append(b, []byte{0, 3, 0, 4}...)
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 
 	require.NoError(t, err)
 	assert.Equal(t, uint8(firewall.ProtoTCP), p.Protocol)
@@ -84,7 +84,7 @@ func Test_newPacket(t *testing.T) {
 
 	b, _ = h.Marshal()
 	b = append(b, []byte{0, 5, 0, 6}...)
-	err = newPacket(b, false, p)
+	err = newPacket(b, false, p, nil)
 
 	require.NoError(t, err)
 	assert.Equal(t, uint8(2), p.Protocol)
@@ -114,7 +114,7 @@ func Test_newPacket_v6(t *testing.T) {
 	err := gopacket.SerializeLayers(buffer, opt, &ip)
 	require.NoError(t, err)
 
-	err = newPacket(buffer.Bytes(), true, p)
+	err = newPacket(buffer.Bytes(), true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
 	// A v6 packet with a hop-by-hop extension
@@ -148,12 +148,12 @@ func Test_newPacket_v6(t *testing.T) {
 
 	// A full IPv6 header and 1 byte in the first extension, but missing
 	// the length byte.
-	err = newPacket(buffer.Bytes()[:41], true, p)
+	err = newPacket(buffer.Bytes()[:41], true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
 	// A full IPv6 header plus 1 full extension, but only 1 byte of the
 	// next layer, missing length byte
-	err = newPacket(buffer.Bytes()[:49], true, p)
+	err = newPacket(buffer.Bytes()[:49], true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 	err = nil
 
@@ -173,7 +173,7 @@ func Test_newPacket_v6(t *testing.T) {
 
 	buffer.Clear()
 	require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp))
-	require.Error(t, newPacket(buffer.Bytes(), true, p))
+	require.Error(t, newPacket(buffer.Bytes(), true, p, nil))
 
 	buffer.Clear()
 	echo := layers.ICMPv6Echo{
@@ -181,7 +181,7 @@ func Test_newPacket_v6(t *testing.T) {
 		SeqNumber:  1234,
 	}
 	require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp, &echo))
-	require.NoError(t, newPacket(buffer.Bytes(), true, p))
+	require.NoError(t, newPacket(buffer.Bytes(), true, p, nil))
 	assert.Equal(t, uint8(layers.IPProtocolICMPv6), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
 	assert.Equal(t, netip.MustParseAddr("ff02::1"), p.LocalAddr)
@@ -192,7 +192,7 @@ func Test_newPacket_v6(t *testing.T) {
 	// A good ESP packet
 	b := buffer.Bytes()
 	b[6] = byte(layers.IPProtocolESP)
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(layers.IPProtocolESP), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
@@ -204,7 +204,7 @@ func Test_newPacket_v6(t *testing.T) {
 	// A good None packet
 	b = buffer.Bytes()
 	b[6] = byte(layers.IPProtocolNoNextHeader)
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(layers.IPProtocolNoNextHeader), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
@@ -216,7 +216,7 @@ func Test_newPacket_v6(t *testing.T) {
 	// An unknown protocol packet
 	b = buffer.Bytes()
 	b[6] = 255 // 255 is a reserved protocol number
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 
 	// A good UDP packet
@@ -243,7 +243,7 @@ func Test_newPacket_v6(t *testing.T) {
 	b = buffer.Bytes()
 
 	// incoming
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(firewall.ProtoUDP), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
@@ -253,7 +253,7 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.False(t, p.Fragment)
 
 	// outgoing
-	err = newPacket(b, false, p)
+	err = newPacket(b, false, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(firewall.ProtoUDP), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.LocalAddr)
@@ -263,14 +263,14 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.False(t, p.Fragment)
 
 	// Too short UDP packet
-	err = newPacket(b[:len(b)-10], false, p) // pull off the last 10 bytes
+	err = newPacket(b[:len(b)-10], false, p, nil) // pull off the last 10 bytes
 	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
 
 	// A good TCP packet
 	b[6] = byte(layers.IPProtocolTCP)
 
 	// incoming
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(firewall.ProtoTCP), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
@@ -280,7 +280,7 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.False(t, p.Fragment)
 
 	// outgoing
-	err = newPacket(b, false, p)
+	err = newPacket(b, false, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(firewall.ProtoTCP), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.LocalAddr)
@@ -290,7 +290,7 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.False(t, p.Fragment)
 
 	// Too short TCP packet
-	err = newPacket(b[:len(b)-10], false, p) // pull off the last 10 bytes
+	err = newPacket(b[:len(b)-10], false, p, nil) // pull off the last 10 bytes
 	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
 
 	// A good UDP packet with an AH header
@@ -325,7 +325,7 @@ func Test_newPacket_v6(t *testing.T) {
 	b = append(b, ahb...)
 	b = append(b, udpHeader...)
 
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, uint8(firewall.ProtoUDP), p.Protocol)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
@@ -335,12 +335,12 @@ func Test_newPacket_v6(t *testing.T) {
 	assert.False(t, p.Fragment)
 
 	// Ensure buffer bounds checking during processing
-	err = newPacket(b[:41], true, p)
+	err = newPacket(b[:41], true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
 
 	// Invalid AH header
 	b = buffer.Bytes()
-	err = newPacket(b, true, p)
+	err = newPacket(b, true, p, nil)
 	require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
 }
 
@@ -388,7 +388,7 @@ func Test_newPacket_ipv6Fragment(t *testing.T) {
 	firstFrag = append(firstFrag, []byte{0xde, 0xad, 0xbe, 0xef}...)
 
 	// Test first fragment incoming
-	err = newPacket(firstFrag, true, p)
+	err = newPacket(firstFrag, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
 	assert.Equal(t, netip.MustParseAddr("ff02::1"), p.LocalAddr)
@@ -398,7 +398,7 @@ func Test_newPacket_ipv6Fragment(t *testing.T) {
 	assert.False(t, p.Fragment)
 
 	// Test first fragment outgoing
-	err = newPacket(firstFrag, false, p)
+	err = newPacket(firstFrag, false, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.LocalAddr)
 	assert.Equal(t, netip.MustParseAddr("ff02::1"), p.RemoteAddr)
@@ -427,7 +427,7 @@ func Test_newPacket_ipv6Fragment(t *testing.T) {
 	secondFrag = append(secondFrag, []byte{0xde, 0xad, 0xbe, 0xef}...)
 
 	// Test second fragment incoming
-	err = newPacket(secondFrag, true, p)
+	err = newPacket(secondFrag, true, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
 	assert.Equal(t, netip.MustParseAddr("ff02::1"), p.LocalAddr)
@@ -437,7 +437,7 @@ func Test_newPacket_ipv6Fragment(t *testing.T) {
 	assert.True(t, p.Fragment)
 
 	// Test second fragment outgoing
-	err = newPacket(secondFrag, false, p)
+	err = newPacket(secondFrag, false, p, nil)
 	require.NoError(t, err)
 	assert.Equal(t, netip.MustParseAddr("ff02::2"), p.LocalAddr)
 	assert.Equal(t, netip.MustParseAddr("ff02::1"), p.RemoteAddr)
@@ -447,7 +447,7 @@ func Test_newPacket_ipv6Fragment(t *testing.T) {
 	assert.True(t, p.Fragment)
 
 	// Too short of a fragment packet
-	err = newPacket(secondFrag[:len(secondFrag)-10], false, p)
+	err = newPacket(secondFrag[:len(secondFrag)-10], false, p, nil)
 	require.ErrorIs(t, err, ErrIPv6PacketTooShort)
 }
 
@@ -529,7 +529,7 @@ func BenchmarkParseV6(b *testing.B) {
 
 	b.Run("Normal", func(b *testing.B) {
 		for i := 0; i < b.N; i++ {
-			if err = parseV6(normalPacket, true, fp); err != nil {
+			if err = parseV6(normalPacket, true, fp, nil); err != nil {
 				b.Fatal(err)
 			}
 		}
@@ -537,7 +537,7 @@ func BenchmarkParseV6(b *testing.B) {
 
 	b.Run("FirstFragment", func(b *testing.B) {
 		for i := 0; i < b.N; i++ {
-			if err = parseV6(firstFrag, true, fp); err != nil {
+			if err = parseV6(firstFrag, true, fp, nil); err != nil {
 				b.Fatal(err)
 			}
 		}
@@ -545,7 +545,7 @@ func BenchmarkParseV6(b *testing.B) {
 
 	b.Run("SecondFragment", func(b *testing.B) {
 		for i := 0; i < b.N; i++ {
-			if err = parseV6(secondFrag, true, fp); err != nil {
+			if err = parseV6(secondFrag, true, fp, nil); err != nil {
 				b.Fatal(err)
 			}
 		}
@@ -590,7 +590,7 @@ func BenchmarkParseV6(b *testing.B) {
 
 	b.Run("200 HopByHop headers", func(b *testing.B) {
 		for i := 0; i < b.N; i++ {
-			if err = parseV6(evilBytes, false, fp); err != nil {
+			if err = parseV6(evilBytes, false, fp, nil); err != nil {
 				b.Fatal(err)
 			}
 		}

overlay/batch/coalesce_core.go

@@ -1,42 +0,0 @@
-package batch
-
-// Arena is an injectable byte-slab that hands out non-overlapping borrowed
-// slices via Reserve and releases them in bulk via Reset. Coalescers take
-// an *Arena at construction so the caller controls the slab lifetime and
-// can share one slab across multiple coalescers (MultiCoalescer hands the
-// same *Arena to every lane so the lanes don't carry their own backings).
-//
-// Reserve borrows; the slice is valid until the next Reset. The slab grows
-// (by allocating a fresh, larger backing array) if a Reserve doesn't fit;
-// pre-size the arena via NewArena to avoid that path on the hot path.
-type Arena struct {
-	buf []byte
-}
-
-// NewArena returns an Arena with a pre-allocated backing of the given
-// capacity. Pass 0 if you don't intend to call Reserve (e.g. a test that
-// only feeds the coalescer pre-made []byte packets via Commit).
-func NewArena(capacity int) *Arena {
-	return &Arena{buf: make([]byte, 0, capacity)}
-}
-
-// Reserve hands out a non-overlapping sz-byte slice from the arena. If the
-// request doesn't fit the current backing, a fresh, larger backing is
-// allocated; already-borrowed slices reference the old backing and remain
-// valid until Reset.
-func (a *Arena) Reserve(sz int) []byte {
-	if len(a.buf)+sz > cap(a.buf) {
-		newCap := max(cap(a.buf)*2, sz)
-		a.buf = make([]byte, 0, newCap)
-	}
-	start := len(a.buf)
-	a.buf = a.buf[:start+sz]
-	return a.buf[start : start+sz : start+sz]
-}
-
-// Reset releases every slice handed out since the last Reset. Callers must
-// not use any previously-borrowed slice after this returns. The underlying
-// backing array is retained so subsequent Reserves don't re-allocate.
-func (a *Arena) Reset() {
-	a.buf = a.buf[:0]
-}

overlay/batch/passthrough.go

@@ -1,46 +0,0 @@
-package batch
-
-import (
-	"io"
-
-	"github.com/slackhq/nebula/util"
-)
-
-// Passthrough is a RxBatcher that doesn't batch anything, it just accumulates and then sends packets.
-type Passthrough struct {
-	out    io.Writer
-	slots  [][]byte
-	arena  *util.Arena
-	cursor int
-}
-
-func NewPassthrough(w io.Writer, slots int, arena *util.Arena) *Passthrough {
-	return &Passthrough{
-		out:   w,
-		slots: make([][]byte, 0, slots),
-		arena: arena,
-	}
-}
-
-func (p *Passthrough) Reserve(sz int) []byte {
-	return p.arena.Reserve(sz)
-}
-
-func (p *Passthrough) Commit(pkt []byte) error {
-	p.slots = append(p.slots, pkt)
-	return nil
-}
-
-func (p *Passthrough) Flush() error {
-	var firstErr error
-	for _, s := range p.slots {
-		_, err := p.out.Write(s)
-		if err != nil && firstErr == nil {
-			firstErr = err
-		}
-	}
-	clear(p.slots)
-	p.slots = p.slots[:0]
-	p.arena.Reset()
-	return firstErr
-}

overlay/batch/rx_batch.go

@@ -1,12 +0,0 @@
-package batch
-
-type RxBatcher interface {
-	// Reserve creates a pkt to borrow
-	Reserve(sz int) []byte
-	// Commit borrows pkt. The caller must keep pkt valid until the next Flush
-	Commit(pkt []byte) error
-	// Flush emits every queued packet in arrival order. Returns the
-	// first error observed; keeps draining so one bad packet doesn't hold up
-	// the rest. After Flush returns, borrowed payload slices may be recycled.
-	Flush() error
-}

overlay/batch/tx_batch.go

@@ -1,60 +0,0 @@
-package batch
-
-import (
-	"net/netip"
-
-	"github.com/slackhq/nebula/util"
-)
-
-const SendBatchCap = 128
-
-// batchWriter is the minimal subset of udp.Conn needed by SendBatch to flush.
-type batchWriter interface {
-	WriteBatch(bufs [][]byte, addrs []netip.AddrPort, outerECNs []byte) error
-}
-
-// SendBatch accumulates encrypted UDP packets and flushes them via WriteBatch.
-// One SendBatch is owned by each listenIn goroutine; no locking is needed.
-// Slot bytes are borrowed from the injected Arena and remain valid until
-// Flush, which Resets the arena.
-type SendBatch struct {
-	out   batchWriter
-	bufs  [][]byte
-	dsts  []netip.AddrPort
-	ecns  []byte
-	arena *util.Arena
-}
-
-// NewSendBatch makes a SendBatch with batchCap slots backed by arena.
-func NewSendBatch(out batchWriter, batchCap int, arena *util.Arena) *SendBatch {
-	return &SendBatch{
-		out:   out,
-		bufs:  make([][]byte, 0, batchCap),
-		dsts:  make([]netip.AddrPort, 0, batchCap),
-		ecns:  make([]byte, 0, batchCap),
-		arena: arena,
-	}
-}
-
-func (b *SendBatch) Reserve(sz int) []byte {
-	return b.arena.Reserve(sz)
-}
-
-func (b *SendBatch) Commit(pkt []byte, dst netip.AddrPort, outerECN byte) {
-	b.bufs = append(b.bufs, pkt)
-	b.dsts = append(b.dsts, dst)
-	b.ecns = append(b.ecns, outerECN)
-}
-
-func (b *SendBatch) Flush() error {
-	var err error
-	if len(b.bufs) > 0 {
-		err = b.out.WriteBatch(b.bufs, b.dsts, b.ecns)
-	}
-	clear(b.bufs)
-	b.bufs = b.bufs[:0]
-	b.dsts = b.dsts[:0]
-	b.ecns = b.ecns[:0]
-	b.arena.Reset()
-	return err
-}

overlay/batch/tx_batch_test.go

@@ -1,126 +0,0 @@
-package batch
-
-import (
-	"net/netip"
-	"testing"
-
-	"github.com/slackhq/nebula/util"
-)
-
-type fakeBatchWriter struct {
-	bufs  [][]byte
-	addrs []netip.AddrPort
-	ecns  []byte
-}
-
-func (w *fakeBatchWriter) WriteBatch(bufs [][]byte, addrs []netip.AddrPort, ecns []byte) error {
-	// Snapshot — SendBatch.Flush nils its slot pointers right after WriteBatch
-	// returns, so tests must capture data before that happens.
-	w.bufs = make([][]byte, len(bufs))
-	for i, b := range bufs {
-		cp := make([]byte, len(b))
-		copy(cp, b)
-		w.bufs[i] = cp
-	}
-	w.addrs = append(w.addrs[:0], addrs...)
-	w.ecns = append(w.ecns[:0], ecns...)
-	return nil
-}
-
-func TestSendBatchReserveCommitFlush(t *testing.T) {
-	fw := &fakeBatchWriter{}
-	b := NewSendBatch(fw, 4, util.NewArena(32))
-
-	ap := netip.MustParseAddrPort("10.0.0.1:4242")
-	for i := 0; i < 4; i++ {
-		slot := b.Reserve(32)
-		if cap(slot) != 32 {
-			t.Fatalf("slot %d: cap=%d want 32", i, cap(slot))
-		}
-		pkt := append(slot[:0], byte(i), byte(i+1), byte(i+2))
-		b.Commit(pkt, ap, 0)
-	}
-	if err := b.Flush(); err != nil {
-		t.Fatalf("Flush: %v", err)
-	}
-	if len(fw.bufs) != 4 {
-		t.Fatalf("WriteBatch got %d bufs want 4", len(fw.bufs))
-	}
-	for i, buf := range fw.bufs {
-		if len(buf) != 3 || buf[0] != byte(i) {
-			t.Errorf("buf %d: %x", i, buf)
-		}
-		if fw.addrs[i] != ap {
-			t.Errorf("addr %d: got %v want %v", i, fw.addrs[i], ap)
-		}
-	}
-
-	// Flush again with nothing committed — should be a no-op.
-	fw.bufs = nil
-	if err := b.Flush(); err != nil {
-		t.Fatalf("empty Flush: %v", err)
-	}
-	if fw.bufs != nil {
-		t.Fatalf("empty Flush triggered WriteBatch")
-	}
-
-	// Reuse after Flush.
-	slot := b.Reserve(32)
-	if cap(slot) != 32 {
-		t.Fatalf("after Flush Reserve wrong cap: %d", cap(slot))
-	}
-}
-
-func TestSendBatchSlotsDoNotOverlap(t *testing.T) {
-	fw := &fakeBatchWriter{}
-	b := NewSendBatch(fw, 3, util.NewArena(8))
-	ap := netip.MustParseAddrPort("10.0.0.1:80")
-
-	for i := 0; i < 3; i++ {
-		s := b.Reserve(8)
-		pkt := append(s[:0], byte(0xA0+i), byte(0xB0+i))
-		b.Commit(pkt, ap, 0)
-	}
-	if err := b.Flush(); err != nil {
-		t.Fatalf("Flush: %v", err)
-	}
-
-	for i, buf := range fw.bufs {
-		if buf[0] != byte(0xA0+i) || buf[1] != byte(0xB0+i) {
-			t.Errorf("slot %d corrupted: %x", i, buf)
-		}
-	}
-}
-
-func TestSendBatchGrowPreservesCommitted(t *testing.T) {
-	fw := &fakeBatchWriter{}
-	// Tiny initial backing forces a grow on the second Reserve.
-	b := NewSendBatch(fw, 1, util.NewArena(4))
-	ap := netip.MustParseAddrPort("10.0.0.1:80")
-
-	s1 := b.Reserve(4)
-	pkt1 := append(s1[:0], 0x11, 0x22, 0x33, 0x44)
-	b.Commit(pkt1, ap, 0)
-
-	s2 := b.Reserve(8) // exceeds remaining cap, triggers grow
-	pkt2 := append(s2[:0], 0xA, 0xB, 0xC, 0xD, 0xE)
-	b.Commit(pkt2, ap, 0)
-
-	// pkt1 must still be intact even though backing reallocated.
-	if pkt1[0] != 0x11 || pkt1[3] != 0x44 {
-		t.Fatalf("first packet corrupted by grow: %x", pkt1)
-	}
-
-	if err := b.Flush(); err != nil {
-		t.Fatalf("Flush: %v", err)
-	}
-	if len(fw.bufs) != 2 {
-		t.Fatalf("got %d bufs want 2", len(fw.bufs))
-	}
-	if fw.bufs[0][0] != 0x11 || fw.bufs[0][3] != 0x44 {
-		t.Errorf("first packet on the wire: %x", fw.bufs[0])
-	}
-	if fw.bufs[1][0] != 0xA || fw.bufs[1][4] != 0xE {
-		t.Errorf("second packet on the wire: %x", fw.bufs[1])
-	}
-}

overlay/device.go

@@ -4,21 +4,15 @@ import (
 	"io"
 	"net/netip"
 
-	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
 )
 
-// defaultBatchBufSize is the per-Queue scratch size for Read on backends
-// that don't do TSO segmentation. 65535 covers any single IP packet.
-const defaultBatchBufSize = 65535
-
 type Device interface {
-	io.Closer
+	io.ReadWriteCloser
 	Activate() error
 	Networks() []netip.Prefix
 	Name() string
 	RoutesFor(netip.Addr) routing.Gateways
 	SupportsMultiqueue() bool
-	NewMultiQueueReader() error
-	Readers() []tio.Queue
+	NewMultiQueueReader() (io.ReadWriteCloser, error)
 }

overlay/network_category_windows.go

@@ -1,358 +0,0 @@
-//go:build !e2e_testing
-// +build !e2e_testing
-
-package overlay
-
-import (
-	"errors"
-	"fmt"
-	"log/slog"
-	"runtime"
-	"strings"
-	"syscall"
-	"time"
-	"unsafe"
-
-	"golang.org/x/sys/windows"
-)
-
-// networkCategory mirrors NLM_NETWORK_CATEGORY from netlistmgr.h.
-type networkCategory int32
-
-const (
-	networkCategoryPublic              networkCategory = 0
-	networkCategoryPrivate             networkCategory = 1
-	networkCategoryDomainAuthenticated networkCategory = 2
-)
-
-func (c networkCategory) String() string {
-	switch c {
-	case networkCategoryPublic:
-		return "public"
-	case networkCategoryPrivate:
-		return "private"
-	case networkCategoryDomainAuthenticated:
-		return "domain"
-	}
-	return fmt.Sprintf("unknown(%d)", c)
-}
-
-// parseNetworkCategory accepts the user-supplied tun.network_category. A
-// second return of false means "leave the category alone".
-func parseNetworkCategory(s string) (networkCategory, bool, error) {
-	switch strings.ToLower(strings.TrimSpace(s)) {
-	case "", "unset":
-		return 0, false, nil
-	case "public":
-		return networkCategoryPublic, true, nil
-	case "private":
-		return networkCategoryPrivate, true, nil
-	case "domain", "domainauthenticated":
-		return networkCategoryDomainAuthenticated, true, nil
-	}
-	return 0, false, fmt.Errorf("unknown tun.network_category %q (expected public, private, domain, or unset)", s)
-}
-
-// CLSID_NetworkListManager {DCB00C01-570F-4A9B-8D69-199FDBA5723B}
-var clsidNetworkListManager = windows.GUID{
-	Data1: 0xDCB00C01, Data2: 0x570F, Data3: 0x4A9B,
-	Data4: [8]byte{0x8D, 0x69, 0x19, 0x9F, 0xDB, 0xA5, 0x72, 0x3B},
-}
-
-// IID_INetworkListManager {DCB00000-570F-4A9B-8D69-199FDBA5723B}
-var iidINetworkListManager = windows.GUID{
-	Data1: 0xDCB00000, Data2: 0x570F, Data3: 0x4A9B,
-	Data4: [8]byte{0x8D, 0x69, 0x19, 0x9F, 0xDB, 0xA5, 0x72, 0x3B},
-}
-
-// x/sys/windows doesn't expose CoCreateInstance, so we bind it ourselves.
-var procCoCreateInstance = windows.NewLazySystemDLL("ole32.dll").NewProc("CoCreateInstance")
-
-const clsCtxAll = windows.CLSCTX_INPROC_SERVER | windows.CLSCTX_INPROC_HANDLER |
-	windows.CLSCTX_LOCAL_SERVER | windows.CLSCTX_REMOTE_SERVER
-
-const (
-	hrSFALSE          = 0x00000001
-	hrRPCEChangedMode = 0x80010106
-)
-
-type hresult uint32
-
-func (h hresult) failed() bool { return int32(h) < 0 }
-func (h hresult) String() string {
-	return fmt.Sprintf("HRESULT 0x%08x", uint32(h))
-}
-
-var errAdapterNotFound = errors.New("adapter not present in network connections enumeration")
-
-// Vtable layouts. Slot order must match the declaration order in netlistmgr.h.
-// All NLM interfaces here derive from IDispatch, which derives from IUnknown.
-
-type iUnknownVtbl struct {
-	QueryInterface uintptr
-	AddRef         uintptr
-	Release        uintptr
-}
-
-type iDispatchVtbl struct {
-	iUnknownVtbl
-	GetTypeInfoCount uintptr
-	GetTypeInfo      uintptr
-	GetIDsOfNames    uintptr
-	Invoke           uintptr
-}
-
-type iNetworkListManagerVtbl struct {
-	iDispatchVtbl
-	GetNetworks           uintptr
-	GetNetwork            uintptr
-	GetNetworkConnections uintptr
-	GetNetworkConnection  uintptr
-	IsConnectedToInternet uintptr
-	IsConnected           uintptr
-	GetConnectivity       uintptr
-}
-
-type iNetworkListManager struct{ Vtbl *iNetworkListManagerVtbl }
-
-func (n *iNetworkListManager) Release() {
-	syscall.SyscallN(n.Vtbl.Release, uintptr(unsafe.Pointer(n)))
-}
-
-func (n *iNetworkListManager) GetNetworkConnections() (*iEnumNetworkConnections, error) {
-	var enum *iEnumNetworkConnections
-	r1, _, _ := syscall.SyscallN(n.Vtbl.GetNetworkConnections,
-		uintptr(unsafe.Pointer(n)), uintptr(unsafe.Pointer(&enum)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return nil, fmt.Errorf("INetworkListManager.GetNetworkConnections: %s", hr)
-	}
-	return enum, nil
-}
-
-type iEnumNetworkConnectionsVtbl struct {
-	iDispatchVtbl
-	NewEnum uintptr
-	Next    uintptr
-	Skip    uintptr
-	Reset   uintptr
-	Clone   uintptr
-}
-
-type iEnumNetworkConnections struct{ Vtbl *iEnumNetworkConnectionsVtbl }
-
-func (e *iEnumNetworkConnections) Release() {
-	syscall.SyscallN(e.Vtbl.Release, uintptr(unsafe.Pointer(e)))
-}
-
-// Next returns the next connection, or (nil, nil) at the end of the enumeration.
-func (e *iEnumNetworkConnections) Next() (*iNetworkConnection, error) {
-	var conn *iNetworkConnection
-	var fetched uint32
-	r1, _, _ := syscall.SyscallN(e.Vtbl.Next,
-		uintptr(unsafe.Pointer(e)), 1,
-		uintptr(unsafe.Pointer(&conn)), uintptr(unsafe.Pointer(&fetched)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return nil, fmt.Errorf("IEnumNetworkConnections.Next: %s", hr)
-	}
-	if fetched == 0 {
-		return nil, nil
-	}
-	return conn, nil
-}
-
-type iNetworkConnectionVtbl struct {
-	iDispatchVtbl
-	GetNetwork            uintptr
-	IsConnectedToInternet uintptr
-	IsConnected           uintptr
-	GetConnectivity       uintptr
-	GetConnectionId       uintptr
-	GetAdapterId          uintptr
-	GetDomainType         uintptr
-}
-
-type iNetworkConnection struct{ Vtbl *iNetworkConnectionVtbl }
-
-func (c *iNetworkConnection) Release() {
-	syscall.SyscallN(c.Vtbl.Release, uintptr(unsafe.Pointer(c)))
-}
-
-func (c *iNetworkConnection) GetAdapterId() (windows.GUID, error) {
-	var g windows.GUID
-	r1, _, _ := syscall.SyscallN(c.Vtbl.GetAdapterId,
-		uintptr(unsafe.Pointer(c)), uintptr(unsafe.Pointer(&g)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return windows.GUID{}, fmt.Errorf("INetworkConnection.GetAdapterId: %s", hr)
-	}
-	return g, nil
-}
-
-func (c *iNetworkConnection) GetNetwork() (*iNetwork, error) {
-	var net *iNetwork
-	r1, _, _ := syscall.SyscallN(c.Vtbl.GetNetwork,
-		uintptr(unsafe.Pointer(c)), uintptr(unsafe.Pointer(&net)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return nil, fmt.Errorf("INetworkConnection.GetNetwork: %s", hr)
-	}
-	return net, nil
-}
-
-type iNetworkVtbl struct {
-	iDispatchVtbl
-	GetName                    uintptr
-	SetName                    uintptr
-	GetDescription             uintptr
-	SetDescription             uintptr
-	GetNetworkId               uintptr
-	GetDomainType              uintptr
-	GetNetworkConnections      uintptr
-	GetTimeCreatedAndConnected uintptr
-	IsConnectedToInternet      uintptr
-	IsConnected                uintptr
-	GetConnectivity            uintptr
-	GetCategory                uintptr
-	SetCategory                uintptr
-}
-
-type iNetwork struct{ Vtbl *iNetworkVtbl }
-
-func (n *iNetwork) Release() {
-	syscall.SyscallN(n.Vtbl.Release, uintptr(unsafe.Pointer(n)))
-}
-
-func (n *iNetwork) GetCategory() (networkCategory, error) {
-	var c networkCategory
-	r1, _, _ := syscall.SyscallN(n.Vtbl.GetCategory,
-		uintptr(unsafe.Pointer(n)), uintptr(unsafe.Pointer(&c)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return 0, fmt.Errorf("INetwork.GetCategory: %s", hr)
-	}
-	return c, nil
-}
-
-func (n *iNetwork) SetCategory(c networkCategory) error {
-	r1, _, _ := syscall.SyscallN(n.Vtbl.SetCategory,
-		uintptr(unsafe.Pointer(n)), uintptr(int32(c)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return fmt.Errorf("INetwork.SetCategory: %s", hr)
-	}
-	return nil
-}
-
-// coInit initializes COM for the current OS thread. The returned function must
-// be deferred to balance a successful init. RPC_E_CHANGED_MODE means COM is
-// already initialized in a different mode on this thread, which is still fine
-// for our calls but we must not Uninitialize in that case.
-func coInit() (func(), error) {
-	err := windows.CoInitializeEx(0, windows.COINIT_MULTITHREADED)
-	if err == nil {
-		return windows.CoUninitialize, nil
-	}
-	if e, ok := err.(syscall.Errno); ok {
-		switch uint32(e) {
-		case hrSFALSE:
-			return windows.CoUninitialize, nil
-		case hrRPCEChangedMode:
-			return func() {}, nil
-		}
-	}
-	return nil, fmt.Errorf("CoInitializeEx: %w", err)
-}
-
-func createNetworkListManager() (*iNetworkListManager, error) {
-	var nlm *iNetworkListManager
-	r1, _, _ := procCoCreateInstance.Call(
-		uintptr(unsafe.Pointer(&clsidNetworkListManager)),
-		0,
-		uintptr(clsCtxAll),
-		uintptr(unsafe.Pointer(&iidINetworkListManager)),
-		uintptr(unsafe.Pointer(&nlm)),
-	)
-	if hr := hresult(r1); hr.failed() {
-		return nil, fmt.Errorf("CoCreateInstance(NetworkListManager): %s", hr)
-	}
-	return nlm, nil
-}
-
-// setNetworkCategory locates the network connection bound to adapterGUID and
-// sets the category of its parent network. Returns errAdapterNotFound if the
-// adapter is not yet visible in the NLM enumeration.
-func setNetworkCategory(adapterGUID windows.GUID, cat networkCategory) error {
-	deinit, err := coInit()
-	if err != nil {
-		return err
-	}
-	defer deinit()
-
-	nlm, err := createNetworkListManager()
-	if err != nil {
-		return err
-	}
-	defer nlm.Release()
-
-	enum, err := nlm.GetNetworkConnections()
-	if err != nil {
-		return err
-	}
-	defer enum.Release()
-
-	for {
-		conn, err := enum.Next()
-		if err != nil {
-			return err
-		}
-		if conn == nil {
-			return errAdapterNotFound
-		}
-
-		guid, err := conn.GetAdapterId()
-		if err != nil || guid != adapterGUID {
-			conn.Release()
-			continue
-		}
-
-		net, err := conn.GetNetwork()
-		conn.Release()
-		if err != nil {
-			return err
-		}
-		err = net.SetCategory(cat)
-		net.Release()
-		return err
-	}
-}
-
-// applyNetworkCategory polls until the wintun adapter shows up in the NLM
-// enumeration, then sets the category. Intended to run in its own goroutine.
-func applyNetworkCategory(l *slog.Logger, adapterGUID windows.GUID, cat networkCategory) {
-	// COM Init/Uninit must be paired on the same OS thread.
-	runtime.LockOSThread()
-	defer runtime.UnlockOSThread()
-
-	const (
-		attempts = 30
-		interval = 500 * time.Millisecond
-	)
-	for i := 0; i < attempts; i++ {
-		err := setNetworkCategory(adapterGUID, cat)
-		if err == nil {
-			l.Info("Set Windows network category", "category", cat.String())
-			return
-		}
-		if !errors.Is(err, errAdapterNotFound) {
-			l.Warn("Failed to set Windows network category", "error", err, "category", cat.String())
-			return
-		}
-		time.Sleep(interval)
-	}
-	l.Warn("Gave up waiting for adapter to appear in NLM enumeration; network category not set",
-		"category", cat.String(),
-		"waited", time.Duration(attempts)*interval,
-	)
-}

overlay/network_category_windows_test.go

@@ -1,109 +0,0 @@
-//go:build !e2e_testing
-// +build !e2e_testing
-
-package overlay
-
-import (
-	"testing"
-)
-
-func Test_parseNetworkCategory(t *testing.T) {
-	cases := []struct {
-		in        string
-		wantCat   networkCategory
-		wantApply bool
-		wantErr   bool
-	}{
-		{"", 0, false, false},
-		{"unset", 0, false, false},
-		{"  UNSET  ", 0, false, false},
-		{"private", networkCategoryPrivate, true, false},
-		{"Private", networkCategoryPrivate, true, false},
-		{"  PRIVATE  ", networkCategoryPrivate, true, false},
-		{"public", networkCategoryPublic, true, false},
-		{"PUBLIC", networkCategoryPublic, true, false},
-		{"domain", networkCategoryDomainAuthenticated, true, false},
-		{"DomainAuthenticated", networkCategoryDomainAuthenticated, true, false},
-		{"garbage", 0, false, true},
-		{"privates", 0, false, true},
-	}
-	for _, tc := range cases {
-		cat, apply, err := parseNetworkCategory(tc.in)
-		if (err != nil) != tc.wantErr {
-			t.Errorf("parseNetworkCategory(%q) err=%v, wantErr=%v", tc.in, err, tc.wantErr)
-			continue
-		}
-		if cat != tc.wantCat || apply != tc.wantApply {
-			t.Errorf("parseNetworkCategory(%q) = (%v, %v), want (%v, %v)", tc.in, cat, apply, tc.wantCat, tc.wantApply)
-		}
-	}
-}
-
-// Test_NLM_round_trip exercises every COM call path used by setNetworkCategory
-// without mutating the host's network state. It validates the CLSID/IID
-// constants and every vtable index by enumerating connections, fetching the
-// adapter id and parent network, reading the current category, and writing it
-// back unchanged.
-//
-// Requires Windows but does not require admin or the wintun driver. Skips if
-// no network connections are available (unlikely outside of an isolated
-// container).
-func Test_NLM_round_trip(t *testing.T) {
-	deinit, err := coInit()
-	if err != nil {
-		t.Fatalf("coInit: %v", err)
-	}
-	defer deinit()
-
-	nlm, err := createNetworkListManager()
-	if err != nil {
-		t.Fatalf("createNetworkListManager: %v", err)
-	}
-	defer nlm.Release()
-
-	enum, err := nlm.GetNetworkConnections()
-	if err != nil {
-		t.Fatalf("GetNetworkConnections: %v", err)
-	}
-	defer enum.Release()
-
-	saw := 0
-	for {
-		conn, err := enum.Next()
-		if err != nil {
-			t.Fatalf("EnumNetworkConnections.Next: %v", err)
-		}
-		if conn == nil {
-			break
-		}
-		saw++
-
-		if _, err := conn.GetAdapterId(); err != nil {
-			conn.Release()
-			t.Fatalf("INetworkConnection.GetAdapterId: %v", err)
-		}
-
-		net, err := conn.GetNetwork()
-		conn.Release()
-		if err != nil {
-			t.Fatalf("INetworkConnection.GetNetwork: %v", err)
-		}
-
-		cat, err := net.GetCategory()
-		if err != nil {
-			net.Release()
-			t.Fatalf("INetwork.GetCategory: %v", err)
-		}
-		// Set to the current value so the host's NLM state is unchanged but
-		// SetCategory's vtable slot is still validated end-to-end.
-		if err := net.SetCategory(cat); err != nil {
-			net.Release()
-			t.Fatalf("INetwork.SetCategory(%v): %v", cat, err)
-		}
-		net.Release()
-	}
-
-	if saw == 0 {
-		t.Skip("no NLM network connections available; skipping round-trip")
-	}
-}

overlay/overlaytest/noop.go

@@ -1,61 +0,0 @@
-// Package overlaytest provides fakes of overlay.Device for tests that do
-// not want to touch a real tun device or route table.
-package overlaytest
-
-import (
-	"errors"
-	"net/netip"
-
-	"github.com/slackhq/nebula/overlay/tio"
-	"github.com/slackhq/nebula/routing"
-	"github.com/slackhq/nebula/wire"
-)
-
-// NoopTun is an overlay.Device that silently discards every read and write.
-// Useful in tests that need to construct a nebula Interface but do not
-// exercise the datapath.
-type NoopTun struct{}
-
-func (NoopTun) Capabilities() tio.Capabilities {
-	return tio.Capabilities{}
-}
-
-func (NoopTun) RoutesFor(addr netip.Addr) routing.Gateways {
-	return routing.Gateways{}
-}
-
-func (NoopTun) Activate() error {
-	return nil
-}
-
-func (NoopTun) Networks() []netip.Prefix {
-	return []netip.Prefix{}
-}
-
-func (NoopTun) Name() string {
-	return "noop"
-}
-
-func (NoopTun) Read(p []wire.TunPacket, mem []byte) (int, error) {
-	return 0, nil
-}
-
-func (NoopTun) Write([]byte) (int, error) {
-	return 0, nil
-}
-
-func (NoopTun) SupportsMultiqueue() bool {
-	return false
-}
-
-func (NoopTun) NewMultiQueueReader() error {
-	return errors.New("unsupported")
-}
-
-func (NoopTun) Readers() []tio.Queue {
-	return []tio.Queue{NoopTun{}}
-}
-
-func (NoopTun) Close() error {
-	return nil
-}

overlay/route.go

@@ -2,7 +2,6 @@ package overlay
 
 import (
 	"fmt"
-	"log/slog"
 	"math"
 	"net"
 	"net/netip"
@@ -10,6 +9,7 @@ import (
 	"strconv"
 
 	"github.com/gaissmai/bart"
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 )
@@ -48,14 +48,11 @@ func (r Route) String() string {
 	return s
 }
 
-func makeRouteTree(l *slog.Logger, routes []Route, allowMTU bool) (*bart.Table[routing.Gateways], error) {
+func makeRouteTree(l *logrus.Logger, routes []Route, allowMTU bool) (*bart.Table[routing.Gateways], error) {
 	routeTree := new(bart.Table[routing.Gateways])
 	for _, r := range routes {
 		if !allowMTU && r.MTU > 0 {
-			l.Warn("route MTU is not supported on this platform",
-				"goos", runtime.GOOS,
-				"route", r,
-			)
+			l.WithField("route", r).Warnf("route MTU is not supported in %s", runtime.GOOS)
 		}
 
 		gateways := r.Via

overlay/route_test.go

@@ -295,7 +295,7 @@ func Test_makeRouteTree(t *testing.T) {
 	routes, err := parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
 	assert.Len(t, routes, 2)
-	routeTree, err := makeRouteTree(test.NewLogger(), routes, true)
+	routeTree, err := makeRouteTree(l, routes, true)
 	require.NoError(t, err)
 
 	ip, err := netip.ParseAddr("1.0.0.2")
@@ -367,7 +367,7 @@ func Test_makeMultipathUnsafeRouteTree(t *testing.T) {
 	routes, err := parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
 	assert.Len(t, routes, 3)
-	routeTree, err := makeRouteTree(test.NewLogger(), routes, true)
+	routeTree, err := makeRouteTree(l, routes, true)
 	require.NoError(t, err)
 
 	ip, err := netip.ParseAddr("192.168.86.1")

overlay/tio/queueset_poll_linux.go

@@ -1,80 +0,0 @@
-package tio
-
-import (
-	"encoding/binary"
-	"errors"
-	"fmt"
-
-	"golang.org/x/sys/unix"
-)
-
-type pollQueueSet struct {
-	pq []*Poll
-	// pqi is exactly the same as pq, but stored as the interface type
-	pqi        []Queue
-	shutdownFd int
-}
-
-func NewPollQueueSet() (QueueSet, error) {
-	shutdownFd, err := unix.Eventfd(0, unix.EFD_NONBLOCK|unix.EFD_CLOEXEC)
-	if err != nil {
-		return nil, fmt.Errorf("failed to create eventfd: %w", err)
-	}
-
-	out := &pollQueueSet{
-		pq:         []*Poll{},
-		pqi:        []Queue{},
-		shutdownFd: shutdownFd,
-	}
-
-	return out, nil
-}
-
-func (c *pollQueueSet) Queues() []Queue {
-	return c.pqi
-}
-
-func (c *pollQueueSet) Add(fd int) error {
-	x, err := newPoll(fd, c.shutdownFd)
-	if err != nil {
-		return err
-	}
-	c.pq = append(c.pq, x)
-	c.pqi = append(c.pqi, x)
-
-	return nil
-}
-
-func (c *pollQueueSet) wakeForShutdown() error {
-	var buf [8]byte
-	binary.NativeEndian.PutUint64(buf[:], 1)
-	_, err := unix.Write(c.shutdownFd, buf[:])
-	return err
-}
-
-func (c *pollQueueSet) Close() error {
-	if c.shutdownFd < 0 {
-		return nil
-	}
-
-	errs := []error{}
-
-	if err := c.wakeForShutdown(); err != nil {
-		errs = append(errs, err)
-	}
-
-	for _, x := range c.pq {
-		if err := x.Close(); err != nil {
-			errs = append(errs, err)
-		}
-	}
-
-	// All Polls reference shutdownFd in their pollfd arrays, so close it
-	// only after every Poll.Close has returned.
-	if err := unix.Close(c.shutdownFd); err != nil {
-		errs = append(errs, err)
-	}
-	c.shutdownFd = -1
-
-	return errors.Join(errs...)
-}

overlay/tio/tio.go

@@ -1,122 +0,0 @@
-package tio
-
-import (
-	"io"
-
-	"github.com/slackhq/nebula/wire"
-)
-
-// QueueSet holds one or many Queue objects and helps close them in an orderly way.
-type QueueSet interface {
-	io.Closer
-	Queues() []Queue
-
-	// Add takes a tun fd, adds it to the set, and prepares it for use as a Queue.
-	Add(fd int) error
-}
-
-// Capabilities advertises which kernel offload features a Queue successfully negotiated.
-// Callers consult this to decide which coalescers to wire onto the write path.
-type Capabilities struct {
-	// TSO means the FD was opened with IFF_VNET_HDR and the kernel agreed
-	// to TUN_F_TSO4|TSO6 — i.e. WriteGSO with GSOProtoTCP is safe.
-	TSO bool
-	// USO means the kernel additionally agreed to TUN_F_USO4|USO6, so
-	// WriteGSO with GSOProtoUDP is safe. Linux ≥ 6.2.
-	USO bool
-}
-
-// Queue is a readable/writable Poll queue. One Queue is driven by a single
-// read goroutine plus a single writer (see Write below).
-type Queue interface {
-	io.Closer
-
-	// Read will read at least 1 packet from the tun (up to len(p)).
-	// mem will be used to provide the backing for each of p[n].Bytes.
-	// Callers should size mem and p to avoid exhausting mem before p.
-	// Returns the number of packets actually read, or error.
-	Read(p []wire.TunPacket, mem []byte) (int, error)
-
-	// Write emits a single packet on the plaintext (outside→inside)
-	// delivery path.
-	Write(p []byte) (int, error)
-
-	// Capabilities returns the Queue's negotiated offload capabilities,
-	// or the zero value when q does not advertise any.
-	Capabilities() Capabilities
-}
-
-// GSOInfo describes a kernel-supplied superpacket sitting in Packet.Bytes.
-// The zero value means "not a superpacket" — Bytes is one regular IP
-// datagram and no segmentation is required.
-type GSOInfo struct {
-	// Size is the GSO segment size: max payload bytes per segment
-	// (== TCP MSS for TSO, == UDP payload chunk for USO). Zero means
-	// not a superpacket.
-	Size uint16
-	// HdrLen is the total L3+L4 header length within Bytes (already
-	// corrected via correctHdrLen, so safe to slice on).
-	HdrLen uint16
-	// CsumStart is the L4 header offset inside Bytes (== L3 header
-	// length).
-	CsumStart uint16
-	// Proto picks the L4 protocol (TCP or UDP) so the segmenter knows
-	// which checksum/header layout to apply.
-	Proto GSOProto
-}
-
-// GSOProto selects the L4 protocol for a GSO superpacket. Determines which
-// VIRTIO_NET_HDR_GSO_* type the writer stamps and which checksum offset
-// inside the transport header virtio NEEDS_CSUM expects.
-type GSOProto uint8
-
-const (
-	GSOProtoNone GSOProto = iota
-	GSOProtoTCP
-	GSOProtoUDP
-)
-
-// GSOWriter is implemented by Queues that can emit a TCP or UDP superpacket
-// assembled from a header prefix plus one or more borrowed payload
-// fragments, in a single vectored write (writev with a leading
-// virtio_net_hdr). This lets the coalescer avoid copying payload bytes
-// between the caller's decrypt buffer and the TUN. Backends without GSO
-// support do not implement this interface and coalescing is skipped.
-//
-// hdr contains the IPv4/IPv6 header prefix (mutable - callers will have
-// filled in total length and IP csum). transportHdr is the TCP or UDP
-// header (mutable - the L4 checksum field must hold the pseudo-header
-// partial, single-fold not inverted, per virtio NEEDS_CSUM semantics).
-// pays are non-overlapping payload fragments whose concatenation is the
-// full superpacket payload; they are read-only from the writer's
-// perspective and must remain valid until the call returns. Every segment
-// in pays except possibly the last is exactly the same size. proto picks
-// the L4 protocol so the writer knows which GSOType / CsumOffset to set.
-//
-// Callers should also consult CapsProvider (via SupportsGSO or
-// QueueCapabilities) for the per-protocol negotiated capability; an
-// implementation of GSOWriter is necessary but not sufficient since USO
-// may not have been negotiated even when TSO was.
-type GSOWriter interface {
-	WriteGSO(hdr []byte, transportHdr []byte, pays [][]byte, proto GSOProto) error
-}
-
-// SupportsGSO reports whether w implements GSOWriter and the underlying
-// queue advertises the negotiated capability for `want`. A writer that
-// implements GSOWriter but not CapsProvider is treated as permissive
-// (used by tests and fakes that don't negotiate).
-func SupportsGSO(w Queue, want GSOProto) (GSOWriter, bool) {
-	gw, ok := w.(GSOWriter)
-	if !ok {
-		return nil, false
-	}
-	caps := w.Capabilities()
-	switch want {
-	case GSOProtoTCP:
-		return gw, caps.TSO
-	case GSOProtoUDP:
-		return gw, caps.USO
-	default:
-		return gw, false
-	}
-}

overlay/tio/tio_poll_linux.go

@@ -1,168 +0,0 @@
-package tio
-
-import (
-	"fmt"
-	"os"
-	"sync"
-	"sync/atomic"
-
-	"github.com/slackhq/nebula/wire"
-	"golang.org/x/sys/unix"
-)
-
-type Poll struct {
-	fd int
-
-	readPoll  [2]unix.PollFd
-	writePoll [2]unix.PollFd
-	writeLock sync.Mutex
-	closed    atomic.Bool
-}
-
-func newPoll(fd int, shutdownFd int) (*Poll, error) {
-	if err := unix.SetNonblock(fd, true); err != nil {
-		_ = unix.Close(fd)
-		return nil, fmt.Errorf("failed to set Poll device as nonblocking: %w", err)
-	}
-
-	out := &Poll{
-		fd: fd,
-		readPoll: [2]unix.PollFd{
-			{Fd: int32(fd), Events: unix.POLLIN},
-			{Fd: int32(shutdownFd), Events: unix.POLLIN},
-		},
-		writePoll: [2]unix.PollFd{
-			{Fd: int32(fd), Events: unix.POLLOUT},
-			{Fd: int32(shutdownFd), Events: unix.POLLIN},
-		},
-		writeLock: sync.Mutex{},
-	}
-	return out, nil
-}
-
-// blockOnRead waits until the Poll fd is readable or shutdown has been signaled.
-// Returns os.ErrClosed if Close was called.
-func (t *Poll) blockOnRead() error {
-	const problemFlags = unix.POLLHUP | unix.POLLNVAL | unix.POLLERR
-	var err error
-	for {
-		_, err = unix.Poll(t.readPoll[:], -1)
-		if err != unix.EINTR {
-			break
-		}
-	}
-	tunEvents := t.readPoll[0].Revents
-	shutdownEvents := t.readPoll[1].Revents
-	t.readPoll[0].Revents = 0
-	t.readPoll[1].Revents = 0
-	if err != nil {
-		return err
-	}
-	if shutdownEvents&(unix.POLLIN|problemFlags) != 0 {
-		return os.ErrClosed
-	}
-	if tunEvents&problemFlags != 0 {
-		return os.ErrClosed
-	}
-	return nil
-}
-
-func (t *Poll) blockOnWrite() error {
-	const problemFlags = unix.POLLHUP | unix.POLLNVAL | unix.POLLERR
-	var err error
-	for {
-		_, err = unix.Poll(t.writePoll[:], -1)
-		if err != unix.EINTR {
-			break
-		}
-	}
-	t.writeLock.Lock()
-	tunEvents := t.writePoll[0].Revents
-	shutdownEvents := t.writePoll[1].Revents
-	t.writePoll[0].Revents = 0
-	t.writePoll[1].Revents = 0
-	t.writeLock.Unlock()
-	if err != nil {
-		return err
-	}
-	if shutdownEvents&(unix.POLLIN|problemFlags) != 0 {
-		return os.ErrClosed
-	}
-	if tunEvents&problemFlags != 0 {
-		return os.ErrClosed
-	}
-	return nil
-}
-
-func (t *Poll) Read(p []wire.TunPacket, mem []byte) (int, error) {
-	if len(p) == 0 || len(mem) == 0 {
-		return 0, nil //todo should this be an err?
-	}
-	p[0].Meta = struct{}{}
-	n, err := t.readOne(mem)
-	if err != nil {
-		return 0, err
-	}
-	p[0].Bytes = mem[:n]
-	return 1, nil
-}
-
-func (t *Poll) readOne(to []byte) (int, error) {
-	for {
-		n, errno := unix.Read(t.fd, to)
-		if errno == nil {
-			return n, nil
-		}
-		switch errno {
-		case unix.EAGAIN:
-			if err := t.blockOnRead(); err != nil {
-				return 0, err
-			}
-		case unix.EINTR:
-			// retry
-		case unix.EBADF:
-			return 0, os.ErrClosed
-		default:
-			return 0, errno
-		}
-	}
-}
-
-func (t *Poll) Write(from []byte) (int, error) {
-	for {
-		n, errno := unix.Write(t.fd, from)
-		if errno == nil {
-			return n, nil
-		}
-		switch errno {
-		case unix.EAGAIN:
-			if err := t.blockOnWrite(); err != nil {
-				return 0, err
-			}
-		case unix.EINTR:
-			// retry
-		case unix.EBADF:
-			return 0, os.ErrClosed
-		default:
-			return 0, errno
-		}
-	}
-}
-
-func (t *Poll) Close() error {
-	if t.closed.Swap(true) {
-		return nil
-	}
-	//shutdownFd is owned by the container, so we should not close it
-	var err error
-	if t.fd >= 0 {
-		err = unix.Close(t.fd)
-		t.fd = -1
-	}
-
-	return err
-}
-
-func (t *Poll) Capabilities() Capabilities {
-	return Capabilities{}
-}

overlay/tio/tun_file_linux_test.go

@@ -1,106 +0,0 @@
-//go:build linux && !android && !e2e_testing
-// +build linux,!android,!e2e_testing
-
-package tio
-
-import (
-	"errors"
-	"os"
-	"sync"
-	"testing"
-	"time"
-
-	"github.com/slackhq/nebula/wire"
-	"github.com/stretchr/testify/require"
-	"golang.org/x/sys/unix"
-)
-
-// newReadPipe returns a read fd. The matching write fd is registered for cleanup.
-// The caller takes ownership of the read fd (pass it into a QueueSet).
-func newReadPipe(t *testing.T) int {
-	t.Helper()
-	var fds [2]int
-	if err := unix.Pipe2(fds[:], unix.O_CLOEXEC); err != nil {
-		t.Fatalf("pipe2: %v", err)
-	}
-	t.Cleanup(func() { _ = unix.Close(fds[1]) })
-	return fds[0]
-}
-
-func TestPoll_WakeForShutdown_WakesFriends(t *testing.T) {
-	parent, err := NewPollQueueSet()
-	require.NoError(t, err)
-	require.NoError(t, parent.Add(newReadPipe(t)))
-	require.NoError(t, parent.Add(newReadPipe(t)))
-	// QueueSet.Close owns the read fds we Added — don't register a separate
-	// Cleanup to close them or we'll double-close whatever fd the kernel
-	// has since reused.
-
-	readers := parent.Queues()
-	errs := make([]error, len(readers))
-	var wg sync.WaitGroup
-	for i, r := range readers {
-		wg.Add(1)
-		go func(i int, r Queue) {
-			defer wg.Done()
-			pkts := make([]wire.TunPacket, 1)
-			_, errs[i] = r.Read(pkts, make([]byte, 64))
-		}(i, r)
-	}
-
-	time.Sleep(50 * time.Millisecond)
-
-	if err := parent.Close(); err != nil {
-		t.Fatalf("Close: %v", err)
-	}
-
-	done := make(chan struct{})
-	go func() { wg.Wait(); close(done) }()
-	select {
-	case <-done:
-	case <-time.After(2 * time.Second):
-		t.Fatal("readers did not wake")
-	}
-
-	for i, err := range errs {
-		if !errors.Is(err, os.ErrClosed) {
-			t.Errorf("reader %d: expected os.ErrClosed, got %v", i, err)
-		}
-	}
-}
-
-func TestPoll_Close_Idempotent(t *testing.T) {
-	shutdownFd, err := unix.Eventfd(0, unix.EFD_NONBLOCK|unix.EFD_CLOEXEC)
-	require.NoError(t, err)
-	t.Cleanup(func() { _ = unix.Close(shutdownFd) })
-
-	tf, err := newPoll(newReadPipe(t), shutdownFd)
-	require.NoError(t, err)
-	if err := tf.Close(); err != nil {
-		t.Fatalf("first Close: %v", err)
-	}
-	if err := tf.Close(); err != nil {
-		t.Fatalf("second Close should be a no-op, got %v", err)
-	}
-}
-
-func TestPollQueueSet_Close_ClosesEventfd(t *testing.T) {
-	qs, err := NewPollQueueSet()
-	require.NoError(t, err)
-	require.NoError(t, qs.Add(newReadPipe(t)))
-
-	fd := qs.(*pollQueueSet).shutdownFd
-	require.NoError(t, qs.Close())
-
-	// Closing the eventfd again should fail with EBADF, proving Close
-	// actually released it.
-	if err := unix.Close(fd); err == nil {
-		t.Fatalf("eventfd %d still open after QueueSet.Close", fd)
-	}
-
-	// Second Close must be a no-op (and must not double-close the eventfd
-	// in case the kernel handed it out to another caller in the meantime).
-	if err := qs.Close(); err != nil {
-		t.Fatalf("second Close: %v", err)
-	}
-}

overlay/tun.go

@@ -2,10 +2,10 @@ package overlay
 
 import (
 	"fmt"
-	"log/slog"
 	"net"
 	"net/netip"
 
+	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/util"
 )
@@ -22,9 +22,9 @@ func (e *NameError) Error() string {
 }
 
 // TODO: We may be able to remove routines
-type DeviceFactory func(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error)
+type DeviceFactory func(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error)
 
-func NewDeviceFromConfig(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
+func NewDeviceFromConfig(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
 	switch {
 	case c.GetBool("tun.disabled", false):
 		tun := newDisabledTun(vpnNetworks, c.GetInt("tun.tx_queue", 500), c.GetBool("stats.message_metrics", false), l)
@@ -36,7 +36,7 @@ func NewDeviceFromConfig(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix
 }
 
 func NewFdDeviceFromConfig(fd *int) DeviceFactory {
-	return func(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
+	return func(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
 		return newTunFromFd(c, l, *fd, vpnNetworks)
 	}
 }