Bump Vampire/setup-wsl from 3 to 7

Bumps [Vampire/setup-wsl](https://github.com/vampire/setup-wsl) from 3 to 7.
- [Release notes](https://github.com/vampire/setup-wsl/releases)
- [Commits](https://github.com/vampire/setup-wsl/compare/v3...v7)

---
updated-dependencies:
- dependency-name: Vampire/setup-wsl
  dependency-version: '7'
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>

.github/workflows/gofmt.yml

@@ -1,34 +0,0 @@
-name: gofmt
-on:
-  push:
-    branches:
-      - master
-  pull_request:
-    paths:
-      - '.github/workflows/gofmt.yml'
-      - '**.go'
-jobs:
-
-  gofmt:
-    name: Run gofmt
-    runs-on: ubuntu-latest
-    steps:
-
-    - uses: actions/checkout@v6
-
-    - uses: actions/setup-go@v6
-      with:
-        go-version: '1.25'
-        check-latest: true
-
-    - name: Install goimports
-      run: |
-        go install golang.org/x/tools/cmd/goimports@latest
-
-    - name: gofmt
-      run: |
-        if [ "$(find . -iname '*.go' | grep -v '\.pb\.go$' | xargs goimports -l)" ]
-        then
-          find . -iname '*.go' | grep -v '\.pb\.go$' | xargs goimports -d
-          exit 1
-        fi

.github/workflows/release.yml

@@ -10,7 +10,7 @@ jobs:
     name: Build Linux/BSD All
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
 
       - uses: actions/setup-go@v6
         with:
@@ -36,7 +36,7 @@ jobs:
       id-token: write
       contents: read
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
 
       - uses: actions/setup-go@v6
         with:
@@ -76,7 +76,7 @@ jobs:
       HAS_SIGNING_CREDS: ${{ secrets.AC_USERNAME != '' }}
     runs-on: macos-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
 
       - uses: actions/setup-go@v6
         with:
@@ -134,7 +134,7 @@ jobs:
       # be overwritten
       - name: Checkout code
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
-        uses: actions/checkout@v6
+        uses: actions/checkout@v7
 
       - name: Download artifacts
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
@@ -163,14 +163,17 @@ jobs:
           mkdir -p build/linux-{amd64,arm64}
           tar -zxvf artifacts/nebula-linux-amd64.tar.gz -C build/linux-amd64/
           tar -zxvf artifacts/nebula-linux-arm64.tar.gz -C build/linux-arm64/
-          docker buildx build . --push -f docker/Dockerfile --platform linux/amd64,linux/arm64 --tag "${DOCKER_IMAGE_REPO}:${DOCKER_IMAGE_TAG}" --tag "${DOCKER_IMAGE_REPO}:${GITHUB_REF#refs/tags/v}"
+          docker buildx build . --push -f docker/Dockerfile --platform linux/amd64,linux/arm64 \
+            --build-arg VERSION="${GITHUB_REF#refs/tags/v}" \
+            --build-arg REVISION="${GITHUB_SHA}" \
+            --tag "${DOCKER_IMAGE_REPO}:${DOCKER_IMAGE_TAG}" --tag "${DOCKER_IMAGE_REPO}:${GITHUB_REF#refs/tags/v}"
 
   release:
     name: Create and Upload Release
     needs: [build-linux, build-darwin, build-windows]
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
 
       - name: Download artifacts
         uses: actions/download-artifact@v8

.github/workflows/smoke-extra.yml

@@ -30,7 +30,7 @@ jobs:
       VAGRANT_DEFAULT_PROVIDER: libvirt
     steps:
 
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
 
     - uses: actions/setup-go@v6
       with:
@@ -62,7 +62,7 @@ jobs:
       VAGRANT_DEFAULT_PROVIDER: virtualbox
     steps:
 
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
 
     - uses: actions/setup-go@v6
       with:
@@ -88,7 +88,7 @@ jobs:
     runs-on: windows-latest
     steps:
 
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
 
     - uses: actions/setup-go@v6
       with:
@@ -98,7 +98,7 @@ jobs:
     # 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
+    - uses: Vampire/setup-wsl@v7
       with:
         distribution: Ubuntu-24.04
         additional-packages: iputils-ping iproute2

.github/workflows/smoke.yml

@@ -18,7 +18,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
 
     - uses: actions/setup-go@v6
       with:
@@ -36,6 +36,14 @@ jobs:
       working-directory: ./.github/workflows/smoke
       run: ./smoke.sh
 
+    - name: setup docker image ipv6
+      working-directory: ./.github/workflows/smoke
+      run: SMOKE_OVERLAY_IPV6=1 ./build.sh
+
+    - name: run smoke ipv6
+      working-directory: ./.github/workflows/smoke
+      run: SMOKE_OVERLAY_IPV6=1 ./smoke.sh
+
     - name: setup relay docker image
       working-directory: ./.github/workflows/smoke
       run: ./build-relay.sh

.github/workflows/smoke/build.sh

@@ -5,6 +5,19 @@ set -e -x
 rm -rf ./build
 mkdir ./build
 
+if [ "$SMOKE_OVERLAY_IPV6" ]
+then
+    LIGHTHOUSE_NIP="fd00:4242:0:0:0:ffff:c0a8:6401"
+    HOST2_NIP="fd00:4242:0:0:0:ffff:c0a8:6402"
+    HOST3_NIP="fd00:4242:0:0:0:ffff:c0a8:6403"
+    HOST4_NIP="fd00:4242:0:0:0:ffff:c0a8:6404"
+else
+    LIGHTHOUSE_NIP="192.168.100.1"
+    HOST2_NIP="192.168.100.2"
+    HOST3_NIP="192.168.100.3"
+    HOST4_NIP="192.168.100.4"
+fi
+
 # Smoke containers run on a dedicated docker network whose subnet is allocated
 # at smoke time, not known at build time. Configs are written with TEST-NET-3
 # placeholder IPs (RFC 5737) and smoke.sh / smoke-vagrant.sh / smoke-relay.sh
@@ -31,24 +44,24 @@ LIGHTHOUSE_IP="203.0.113.2"
         ../genconfig.sh >lighthouse1.yml
 
     HOST="host2" \
-        LIGHTHOUSES="192.168.100.1 $LIGHTHOUSE_IP:4242" \
+        LIGHTHOUSES="$LIGHTHOUSE_NIP $LIGHTHOUSE_IP:4242" \
         ../genconfig.sh >host2.yml
 
     HOST="host3" \
-        LIGHTHOUSES="192.168.100.1 $LIGHTHOUSE_IP:4242" \
+        LIGHTHOUSES="$LIGHTHOUSE_NIP $LIGHTHOUSE_IP:4242" \
         INBOUND='[{"port": "any", "proto": "icmp", "group": "lighthouse"}]' \
         ../genconfig.sh >host3.yml
 
     HOST="host4" \
-        LIGHTHOUSES="192.168.100.1 $LIGHTHOUSE_IP:4242" \
+        LIGHTHOUSES="$LIGHTHOUSE_NIP $LIGHTHOUSE_IP:4242" \
         OUTBOUND='[{"port": "any", "proto": "icmp", "group": "lighthouse"}]' \
         ../genconfig.sh >host4.yml
 
     ../../../../nebula-cert ca -curve "${CURVE:-25519}" -name "Smoke Test"
-    ../../../../nebula-cert sign -name "lighthouse1" -groups "lighthouse,lighthouse1" -ip "192.168.100.1/24"
-    ../../../../nebula-cert sign -name "host2" -groups "host,host2" -ip "192.168.100.2/24"
-    ../../../../nebula-cert sign -name "host3" -groups "host,host3" -ip "192.168.100.3/24"
-    ../../../../nebula-cert sign -name "host4" -groups "host,host4" -ip "192.168.100.4/24"
+    ../../../../nebula-cert sign -name "lighthouse1" -groups "lighthouse,lighthouse1" -ip "$LIGHTHOUSE_NIP/24"
+    ../../../../nebula-cert sign -name "host2" -groups "host,host2" -ip "$HOST2_NIP/24"
+    ../../../../nebula-cert sign -name "host3" -groups "host,host3" -ip "$HOST3_NIP/24"
+    ../../../../nebula-cert sign -name "host4" -groups "host,host4" -ip "$HOST4_NIP/24"
 )
 
 docker build -t "nebula:${NAME:-smoke}" .

.github/workflows/smoke/smoke.sh

@@ -47,6 +47,19 @@ HOST2_IP="$PREFIX.3"
 HOST3_IP="$PREFIX.4"
 HOST4_IP="$PREFIX.5"
 
+if [ "$SMOKE_OVERLAY_IPV6" ]
+then
+    LIGHTHOUSE_NIP="fd00:4242:0:0:0:ffff:c0a8:6401"
+    HOST2_NIP="fd00:4242:0:0:0:ffff:c0a8:6402"
+    HOST3_NIP="fd00:4242:0:0:0:ffff:c0a8:6403"
+    HOST4_NIP="fd00:4242:0:0:0:ffff:c0a8:6404"
+else
+    LIGHTHOUSE_NIP="192.168.100.1"
+    HOST2_NIP="192.168.100.2"
+    HOST3_NIP="192.168.100.3"
+    HOST4_NIP="192.168.100.4"
+fi
+
 # Sed the placeholder TEST-NET-3 IPs in the host configs to the real ones.
 # build/lighthouse1.yml has no IPs to rewrite so it's skipped.
 for f in build/host2.yml build/host3.yml build/host4.yml; do
@@ -80,28 +93,28 @@ docker exec host3 tcpdump -i eth0 -q -w - -U 2>logs/host3.outside.log >logs/host
 docker exec host4 tcpdump -i tun0 -q -w - -U 2>logs/host4.inside.log >logs/host4.inside.pcap &
 docker exec host4 tcpdump -i eth0 -q -w - -U 2>logs/host4.outside.log >logs/host4.outside.pcap &
 
-docker exec host2 ncat -nklv 0.0.0.0 2000 &
-docker exec host3 ncat -nklv 0.0.0.0 2000 &
-docker exec host4 ncat -e '/usr/bin/echo helloagainfromhost4' -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 &
+docker exec host2 ncat -nklv 2000 &
+docker exec host3 ncat -nklv 2000 &
+docker exec host4 ncat -e '/usr/bin/echo helloagainfromhost4' -nkluv 4000 &
+docker exec host2 ncat -e '/usr/bin/echo host2' -nkluv 3000 &
+docker exec host3 ncat -e '/usr/bin/echo host3' -nkluv 3000 &
 
 set +x
 echo
 echo " *** Testing ping from lighthouse1"
 echo
 set -x
-docker exec lighthouse1 ping -c1 192.168.100.2
-docker exec lighthouse1 ping -c1 192.168.100.3
+docker exec lighthouse1 ping -c1 $HOST2_NIP
+docker exec lighthouse1 ping -c1 $HOST3_NIP
 
 set +x
 echo
 echo " *** Testing ping from host2"
 echo
 set -x
-docker exec host2 ping -c1 192.168.100.1
+docker exec host2 ping -c1 $LIGHTHOUSE_NIP
 # Should fail because not allowed by host3 inbound firewall
-! docker exec host2 ping -c1 192.168.100.3 -w5 || exit 1
+! docker exec host2 ping -c1 $HOST3_NIP -w5 || exit 1
 
 set +x
 echo
@@ -109,34 +122,34 @@ echo " *** Testing ncat from host2"
 echo
 set -x
 # Should fail because not allowed by host3 inbound firewall
-! docker exec host2 ncat -nzv -w5 192.168.100.3 2000 || exit 1
-! docker exec host2 ncat -nzuv -w5 192.168.100.3 3000 | grep -q host3 || exit 1
+! docker exec host2 ncat -nzv -w5 $HOST3_NIP 2000 || exit 1
+! docker exec host2 ncat -nzuv -w5 $HOST3_NIP 3000 | grep -q host3 || exit 1
 
 set +x
 echo
 echo " *** Testing ping from host3"
 echo
 set -x
-docker exec host3 ping -c1 192.168.100.1
-docker exec host3 ping -c1 192.168.100.2
+docker exec host3 ping -c1 $LIGHTHOUSE_NIP
+docker exec host3 ping -c1 $HOST2_NIP
 
 set +x
 echo
 echo " *** Testing ncat from host3"
 echo
 set -x
-docker exec host3 ncat -nzv -w5 192.168.100.2 2000
-docker exec host3 ncat -nzuv -w5 192.168.100.2 3000 | grep -q host2
+docker exec host3 ncat -nzv -w5 $HOST2_NIP 2000
+docker exec host3 ncat -nzuv -w5 $HOST2_NIP 3000 | grep -q host2
 
 set +x
 echo
 echo " *** Testing ping from host4"
 echo
 set -x
-docker exec host4 ping -c1 192.168.100.1
+docker exec host4 ping -c1 $LIGHTHOUSE_NIP
 # Should fail because not allowed by host4 outbound firewall
-! docker exec host4 ping -c1 192.168.100.2 -w5 || exit 1
-! docker exec host4 ping -c1 192.168.100.3 -w5 || exit 1
+! docker exec host4 ping -c1 $HOST2_NIP -w5 || exit 1
+! docker exec host4 ping -c1 $HOST3_NIP -w5 || exit 1
 
 set +x
 echo
@@ -144,10 +157,10 @@ echo " *** Testing ncat from host4"
 echo
 set -x
 # Should fail because not allowed by host4 outbound firewall
-! docker exec host4 ncat -nzv -w5 192.168.100.2 2000 || exit 1
-! docker exec host4 ncat -nzv -w5 192.168.100.3 2000 || exit 1
-! docker exec host4 ncat -nzuv -w5 192.168.100.2 3000 | grep -q host2 || exit 1
-! docker exec host4 ncat -nzuv -w5 192.168.100.3 3000 | grep -q host3 || exit 1
+! docker exec host4 ncat -nzv -w5 $HOST2_NIP 2000 || exit 1
+! docker exec host4 ncat -nzv -w5 $HOST3_NIP 2000 || exit 1
+! docker exec host4 ncat -nzuv -w5 $HOST2_NIP 3000 | grep -q host2 || exit 1
+! docker exec host4 ncat -nzuv -w5 $HOST3_NIP 3000 | grep -q host3 || exit 1
 
 set +x
 echo
@@ -159,7 +172,7 @@ set -x
 # 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
+docker exec host2 sh -c "(/usr/bin/echo host2; sleep 2) | ncat -nuv $HOST4_NIP 4000" | grep -q helloagainfromhost4
 
 docker exec host4 sh -c 'kill 1'
 docker exec host3 sh -c 'kill 1'

.github/workflows/test.yml

@@ -13,20 +13,28 @@ on:
       - 'go.sum'
 jobs:
 
-  test-linux:
-    name: Build all and test on ubuntu-linux
+  static:
+    name: Static checks
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
 
     - uses: actions/setup-go@v6
       with:
         go-version: '1.25'
         check-latest: true
 
-    - name: Build
-      run: make all
+    - name: Install goimports
+      run: go install golang.org/x/tools/cmd/goimports@latest
+
+    - name: gofmt
+      run: |
+        if [ "$(find . -iname '*.go' | grep -v '\.pb\.go$' | xargs goimports -l)" ]
+        then
+          find . -iname '*.go' | grep -v '\.pb\.go$' | xargs goimports -d
+          exit 1
+        fi
 
     - name: Vet
       run: make vet
@@ -36,97 +44,107 @@ jobs:
       with:
         version: v2.5
 
-    - name: Test
-      run: make test
-
-    - name: End 2 end
-      run: make e2evv
-
-    - name: Build test mobile
-      run: make build-test-mobile
-
-    - uses: actions/upload-artifact@v7
-      with:
-        name: e2e packet flow linux-latest
-        path: e2e/mermaid/linux-latest
-        if-no-files-found: warn
-
-  test-linux-boringcrypto:
-    name: Build and test on linux with boringcrypto
-    runs-on: ubuntu-latest
-    steps:
-
-    - uses: actions/checkout@v6
-
-    - uses: actions/setup-go@v6
-      with:
-        go-version: '1.25'
-        check-latest: true
-
-    - name: Build
-      run: make bin-boringcrypto
-
-    - name: Test
-      run: make test-boringcrypto
-
-    - name: End 2 end
-      run: make e2e GOEXPERIMENT=boringcrypto CGO_ENABLED=1 TEST_ENV="TEST_LOGS=1" TEST_FLAGS="-v -ldflags -checklinkname=0"
-
-  test-linux-pkcs11:
-    name: Build and test on linux with pkcs11
-    runs-on: ubuntu-latest
-    steps:
-
-    - uses: actions/checkout@v6
-
-    - uses: actions/setup-go@v6
-      with:
-        go-version: '1.25'
-        check-latest: true
-
-    - name: Build
-      run: make bin-pkcs11
-
-    - name: Test
-      run: make test-pkcs11
-
   test:
-    name: Build and test on ${{ matrix.os }}
+    name: Test ${{ matrix.name }}
     runs-on: ${{ matrix.os }}
     strategy:
+      fail-fast: false
       matrix:
-        os: [windows-latest, macos-latest]
+        include:
+        - name: linux
+          os: ubuntu-latest
+          build-cmd: go build ./cmd/nebula ./cmd/nebula-cert
+          test-cmd: make test
+          e2e-cmd: make e2evv
+        - name: linux-boringcrypto
+          os: ubuntu-latest
+          build-cmd: make bin-boringcrypto
+          test-cmd: make test-boringcrypto
+          e2e-cmd: make e2e GOEXPERIMENT=boringcrypto CGO_ENABLED=1 TEST_ENV="TEST_LOGS=1" TEST_FLAGS="-v -ldflags -checklinkname=0"
+        - name: linux-pkcs11
+          os: ubuntu-latest
+          build-cmd: make bin-pkcs11
+          test-cmd: make test-pkcs11
+          e2e-cmd: ''
+        - name: macos
+          os: macos-latest
+          build-cmd: go build ./cmd/nebula ./cmd/nebula-cert
+          test-cmd: make test
+          e2e-cmd: make e2evv
+        - name: windows
+          os: windows-latest
+          build-cmd: go build ./cmd/nebula ./cmd/nebula-cert
+          test-cmd: make test
+          e2e-cmd: make e2evv
     steps:
 
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
 
     - uses: actions/setup-go@v6
       with:
         go-version: '1.25'
         check-latest: true
 
-    - name: Build nebula
-      run: go build ./cmd/nebula
+    - name: Build
+      run: ${{ matrix.build-cmd }}
 
-    - name: Build nebula-cert
-      run: go build ./cmd/nebula-cert
-
-    - name: Vet
-      run: make vet
-
-    - name: golangci-lint
-      uses: golangci/golangci-lint-action@v9
-      with:
-        version: v2.5
+    - name: Cross-build darwin-amd64
+      if: matrix.name == 'macos'
+      run: GOARCH=amd64 go build -o /tmp/nebula-amd64 ./cmd/nebula && GOARCH=amd64 go build -o /tmp/nebula-cert-amd64 ./cmd/nebula-cert
 
     - name: Test
-      run: make test
+      run: ${{ matrix.test-cmd }}
 
     - name: End 2 end
-      run: make e2evv
+      if: matrix.e2e-cmd != ''
+      run: ${{ matrix.e2e-cmd }}
 
     - uses: actions/upload-artifact@v7
+      if: matrix.e2e-cmd != '' && always()
       with:
-        name: e2e packet flow ${{ matrix.os }}
-        path: e2e/mermaid/${{ matrix.os }}
+        name: e2e packet flow ${{ matrix.name }}
+        path: e2e/mermaid/
         if-no-files-found: warn
+
+  cross-build:
+    name: Cross-build ${{ matrix.name }}
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+        - {name: linux-arm,   make-target: all-cross-linux-arm}
+        - {name: linux-mips,  make-target: all-cross-linux-mips}
+        - {name: linux-other, make-target: all-cross-linux-other}
+        - {name: freebsd,     make-target: all-freebsd}
+        - {name: openbsd,     make-target: all-openbsd}
+        - {name: netbsd,      make-target: all-netbsd}
+        - {name: windows,     make-target: all-cross-windows}
+        - {name: mobile,      make-target: build-test-mobile}
+    steps:
+
+    - uses: actions/checkout@v7
+
+    - uses: actions/setup-go@v6
+      with:
+        go-version: '1.25'
+        check-latest: true
+
+    - name: Build ${{ matrix.name }}
+      run: make -j"$(nproc)" ${{ matrix.make-target }}
+
+  finish:
+    name: CI status
+    if: always()
+    needs: [static, test, cross-build]
+    runs-on: ubuntu-latest
+    steps:
+
+    - name: Fail if any upstream job failed
+      if: contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled')
+      run: |
+        echo "upstream results: ${{ toJSON(needs) }}"
+        exit 1
+
+    - name: All upstream jobs passed
+      run: echo "ok"

Makefile

@@ -60,6 +60,18 @@ ALL = $(ALL_LINUX) \
 	windows-amd64 \
 	windows-arm64
 
+# Cross-build shards used by .github/workflows/test.yml — same as ALL_*
+# but with the arch that has a native CI runner removed, so the cross-build
+# job is not duplicating coverage the native test jobs already give.
+ALL_CROSS_LINUX = $(filter-out linux-amd64,$(ALL_LINUX))
+
+# ALL_CROSS_LINUX further split into family sub-shards so each can run on
+# its own CI runner in parallel. Union of the three must equal
+# ALL_CROSS_LINUX; adding a new linux arch goes into the matching family.
+ALL_CROSS_LINUX_ARM   = linux-arm-5 linux-arm-6 linux-arm-7 linux-arm64
+ALL_CROSS_LINUX_MIPS  = linux-mips linux-mipsle linux-mips64 linux-mips64le linux-mips-softfloat
+ALL_CROSS_LINUX_OTHER = linux-386 linux-ppc64le linux-riscv64 linux-loong64
+
 e2e:
 	$(TEST_ENV) go test -tags=e2e_testing -count=1 $(TEST_FLAGS) ./e2e
 
@@ -82,6 +94,35 @@ DOCKER_BIN = build/linux-amd64/nebula build/linux-amd64/nebula-cert
 
 all: $(ALL:%=build/%/nebula) $(ALL:%=build/%/nebula-cert)
 
+all-linux: $(ALL_LINUX:%=build/%/nebula) $(ALL_LINUX:%=build/%/nebula-cert)
+
+all-freebsd: $(ALL_FREEBSD:%=build/%/nebula) $(ALL_FREEBSD:%=build/%/nebula-cert)
+
+all-openbsd: $(ALL_OPENBSD:%=build/%/nebula) $(ALL_OPENBSD:%=build/%/nebula-cert)
+
+all-netbsd: $(ALL_NETBSD:%=build/%/nebula) $(ALL_NETBSD:%=build/%/nebula-cert)
+
+all-darwin: build/darwin-amd64/nebula build/darwin-amd64/nebula-cert build/darwin-arm64/nebula build/darwin-arm64/nebula-cert
+
+all-windows: build/windows-amd64/nebula.exe build/windows-amd64/nebula-cert.exe build/windows-arm64/nebula.exe build/windows-arm64/nebula-cert.exe
+
+# CI cross-build shards. darwin-arm64 is covered by the native macos-latest
+# job; windows-amd64 is covered by the native windows-latest job; both are
+# omitted here to avoid building them a second time. darwin-amd64 stays in
+# all-cross-darwin because intel mac is only a labeled/master-time native
+# job, so PRs still need cross-build coverage for it.
+all-cross-linux: $(ALL_CROSS_LINUX:%=build/%/nebula) $(ALL_CROSS_LINUX:%=build/%/nebula-cert)
+
+all-cross-linux-arm:   $(ALL_CROSS_LINUX_ARM:%=build/%/nebula)   $(ALL_CROSS_LINUX_ARM:%=build/%/nebula-cert)
+
+all-cross-linux-mips:  $(ALL_CROSS_LINUX_MIPS:%=build/%/nebula)  $(ALL_CROSS_LINUX_MIPS:%=build/%/nebula-cert)
+
+all-cross-linux-other: $(ALL_CROSS_LINUX_OTHER:%=build/%/nebula) $(ALL_CROSS_LINUX_OTHER:%=build/%/nebula-cert)
+
+all-cross-darwin: build/darwin-amd64/nebula build/darwin-amd64/nebula-cert
+
+all-cross-windows: build/windows-arm64/nebula.exe build/windows-arm64/nebula-cert.exe
+
 docker: docker/linux-$(shell go env GOARCH)
 
 release: $(ALL:%=build/nebula-%.tar.gz)
@@ -227,6 +268,9 @@ smoke-relay-docker: bin-docker
 	cd .github/workflows/smoke/ && ./build-relay.sh
 	cd .github/workflows/smoke/ && ./smoke-relay.sh
 
+smoke-docker-ipv6: export SMOKE_OVERLAY_IPV6 = 1
+smoke-docker-ipv6: smoke-docker
+
 smoke-docker-race: BUILD_ARGS = -race
 smoke-docker-race: CGO_ENABLED = 1
 smoke-docker-race: smoke-docker
@@ -236,5 +280,5 @@ smoke-vagrant/%: bin-docker build/%/nebula
 	cd .github/workflows/smoke/ && ./smoke-vagrant.sh $*
 
 .FORCE:
-.PHONY: bench bench-cpu bench-cpu-long bin build-test-mobile e2e e2ev e2evv e2evvv e2evvvv proto release service smoke-docker smoke-docker-race test test-cov-html smoke-vagrant/%
+.PHONY: all all-linux all-freebsd all-openbsd all-netbsd all-darwin all-windows all-cross-linux all-cross-linux-arm all-cross-linux-mips all-cross-linux-other all-cross-darwin all-cross-windows bench bench-cpu bench-cpu-long bin build-test-mobile e2e e2ev e2evv e2evvv e2evvvv proto release service smoke-docker smoke-docker-race test test-cov-html smoke-vagrant/%
 .DEFAULT_GOAL := bin

cert/helper_test.go

@@ -13,6 +13,12 @@ import (
 	"golang.org/x/crypto/ed25519"
 )
 
+// testCertNow is the reference "now" used to derive default before/after times
+// in NewTestCaCert and NewTestCert. Holding it fixed for the lifetime of the
+// test binary keeps CA and leaf defaults aligned at the same second, so a leaf
+// signed with default times can never expire after its CA on a rounding race.
+var testCertNow = time.Now().Round(time.Second)
+
 // NewTestCaCert will create a new ca certificate
 func NewTestCaCert(version Version, curve Curve, before, after time.Time, networks, unsafeNetworks []netip.Prefix, groups []string) (Certificate, []byte, []byte, []byte) {
 	var err error
@@ -34,10 +40,10 @@ func NewTestCaCert(version Version, curve Curve, before, after time.Time, networ
 	}
 
 	if before.IsZero() {
-		before = time.Now().Add(time.Second * -60).Round(time.Second)
+		before = testCertNow.Add(time.Second * -60)
 	}
 	if after.IsZero() {
-		after = time.Now().Add(time.Second * 60).Round(time.Second)
+		after = testCertNow.Add(time.Second * 60)
 	}
 
 	t := &TBSCertificate{
@@ -70,11 +76,11 @@ func NewTestCaCert(version Version, curve Curve, before, after time.Time, networ
 // Expiry times are defaulted if you do not pass them in
 func NewTestCert(v Version, curve Curve, ca Certificate, key []byte, name string, before, after time.Time, networks, unsafeNetworks []netip.Prefix, groups []string) (Certificate, []byte, []byte, []byte) {
 	if before.IsZero() {
-		before = time.Now().Add(time.Second * -60).Round(time.Second)
+		before = testCertNow.Add(time.Second * -60)
 	}
 
 	if after.IsZero() {
-		after = time.Now().Add(time.Second * 60).Round(time.Second)
+		after = testCertNow.Add(time.Second * 60)
 	}
 
 	if len(networks) == 0 {

cert_test/cert.go

@@ -14,6 +14,12 @@ import (
 	"golang.org/x/crypto/ed25519"
 )
 
+// testCertNow is the reference "now" used to derive default before/after times
+// in NewTestCaCert and NewTestCert. Holding it fixed for the lifetime of the
+// test binary keeps CA and leaf defaults aligned at the same second, so a leaf
+// signed with default times can never expire after its CA on a rounding race.
+var testCertNow = time.Now().Round(time.Second)
+
 // NewTestCaCert will create a new ca certificate
 func NewTestCaCert(version cert.Version, curve cert.Curve, before, after time.Time, networks, unsafeNetworks []netip.Prefix, groups []string) (cert.Certificate, []byte, []byte, []byte) {
 	var err error
@@ -35,10 +41,10 @@ func NewTestCaCert(version cert.Version, curve cert.Curve, before, after time.Ti
 	}
 
 	if before.IsZero() {
-		before = time.Now().Add(time.Second * -60).Round(time.Second)
+		before = testCertNow.Add(time.Second * -60)
 	}
 	if after.IsZero() {
-		after = time.Now().Add(time.Second * 60).Round(time.Second)
+		after = testCertNow.Add(time.Second * 60)
 	}
 
 	t := &cert.TBSCertificate{
@@ -71,11 +77,11 @@ func NewTestCaCert(version cert.Version, curve cert.Curve, before, after time.Ti
 // Expiry times are defaulted if you do not pass them in
 func NewTestCert(v cert.Version, curve cert.Curve, ca cert.Certificate, key []byte, name string, before, after time.Time, networks, unsafeNetworks []netip.Prefix, groups []string) (cert.Certificate, []byte, []byte, []byte) {
 	if before.IsZero() {
-		before = time.Now().Add(time.Second * -60).Round(time.Second)
+		before = testCertNow.Add(time.Second * -60)
 	}
 
 	if after.IsZero() {
-		after = time.Now().Add(time.Second * 60).Round(time.Second)
+		after = testCertNow.Add(time.Second * 60)
 	}
 
 	var pub, priv []byte

cmd/nebula-cert/ca.go

@@ -97,6 +97,19 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 		if err = mustFlagString("out-key", cf.outKeyPath); err != nil {
 			return err
 		}
+	} else {
+		// out-key is meaningless under PKCS#11 because the private key never
+		// leaves the HSM; reject it so we never silently accept or claim a
+		// stdout slot for it.
+		outKeySet := false
+		cf.set.Visit(func(f *flag.Flag) {
+			if f.Name == "out-key" {
+				outKeySet = true
+			}
+		})
+		if outKeySet {
+			return newHelpErrorf("cannot set -out-key with -pkcs11")
+		}
 	}
 	if err := mustFlagString("out-crt", cf.outCertPath); err != nil {
 		return err
@@ -171,12 +184,21 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 		}
 	}
 
+	var claims ioClaims
+	if err := reserveOutputs(&claims,
+		"out-key", *cf.outKeyPath,
+		"out-crt", *cf.outCertPath,
+		"out-qr", *cf.outQRPath,
+	); err != nil {
+		return err
+	}
+
 	var passphrase []byte
 	if !isP11 && *cf.encryption {
 		passphrase = []byte(os.Getenv("NEBULA_CA_PASSPHRASE"))
 		if len(passphrase) == 0 {
 			for i := 0; i < 5; i++ {
-				out.Write([]byte("Enter passphrase: "))
+				errOut.Write([]byte("Enter passphrase: "))
 				passphrase, err = pr.ReadPassword()
 
 				if err == ErrNoTerminal {
@@ -261,14 +283,16 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 		Curve:          curve,
 	}
 
-	if !isP11 {
+	if !isP11 && !isStdio(*cf.outKeyPath) {
 		if _, err := os.Stat(*cf.outKeyPath); err == nil {
 			return fmt.Errorf("refusing to overwrite existing CA key: %s", *cf.outKeyPath)
 		}
 	}
 
-	if _, err := os.Stat(*cf.outCertPath); err == nil {
-		return fmt.Errorf("refusing to overwrite existing CA cert: %s", *cf.outCertPath)
+	if !isStdio(*cf.outCertPath) {
+		if _, err := os.Stat(*cf.outCertPath); err == nil {
+			return fmt.Errorf("refusing to overwrite existing CA cert: %s", *cf.outCertPath)
+		}
 	}
 
 	var c cert.Certificate
@@ -294,7 +318,7 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 			b = cert.MarshalSigningPrivateKeyToPEM(curve, rawPriv)
 		}
 
-		err = os.WriteFile(*cf.outKeyPath, b, 0600)
+		err = writeOutput(*cf.outKeyPath, b, 0600, out)
 		if err != nil {
 			return fmt.Errorf("error while writing out-key: %s", err)
 		}
@@ -305,7 +329,7 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 		return fmt.Errorf("error while marshalling certificate: %s", err)
 	}
 
-	err = os.WriteFile(*cf.outCertPath, b, 0600)
+	err = writeOutput(*cf.outCertPath, b, 0600, out)
 	if err != nil {
 		return fmt.Errorf("error while writing out-crt: %s", err)
 	}
@@ -316,7 +340,7 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 			return fmt.Errorf("error while generating qr code: %s", err)
 		}
 
-		err = os.WriteFile(*cf.outQRPath, b, 0600)
+		err = writeOutput(*cf.outQRPath, b, 0600, out)
 		if err != nil {
 			return fmt.Errorf("error while writing out-qr: %s", err)
 		}
@@ -332,6 +356,7 @@ func caSummary() string {
 func caHelp(out io.Writer) {
 	cf := newCaFlags()
 	out.Write([]byte("Usage of " + os.Args[0] + " " + caSummary() + "\n"))
+	out.Write([]byte(stdioHelpText))
 	cf.set.SetOutput(out)
 	cf.set.PrintDefaults()
 }

cmd/nebula-cert/ca_test.go

@@ -27,6 +27,7 @@ func Test_caHelp(t *testing.T) {
 	assert.Equal(
 		t,
 		"Usage of "+os.Args[0]+" ca <flags>: create a self signed certificate authority\n"+
+			"  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"+
 			"  -argon-iterations uint\n"+
 			"    \tOptional: Argon2 iterations parameter used for encrypted private key passphrase (default 1)\n"+
 			"  -argon-memory uint\n"+
@@ -84,7 +85,7 @@ func Test_ca(t *testing.T) {
 		err:      nil,
 	}
 
-	pwPromptOb := "Enter passphrase: "
+	pwPromptEB := "Enter passphrase: "
 
 	// required args
 	assertHelpError(t, ca(
@@ -168,8 +169,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.NoError(t, ca(args, ob, eb, testpw))
-	assert.Equal(t, pwPromptOb, ob.String())
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, pwPromptEB, eb.String())
 
 	// test encrypted key with passphrase environment variable
 	os.Remove(keyF.Name())
@@ -207,8 +208,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.Error(t, ca(args, ob, eb, errpw))
-	assert.Equal(t, pwPromptOb, ob.String())
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, pwPromptEB, eb.String())
 
 	// test when user fails to enter a password
 	os.Remove(keyF.Name())
@@ -217,8 +218,8 @@ func Test_ca(t *testing.T) {
 	eb.Reset()
 	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
 	require.EqualError(t, ca(args, ob, eb, nopw), "no passphrase specified, remove -encrypt flag to write out-key in plaintext")
-	assert.Equal(t, strings.Repeat(pwPromptOb, 5), ob.String()) // prompts 5 times before giving up
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, strings.Repeat(pwPromptEB, 5), eb.String()) // prompts 5 times before giving up
 
 	// create valid cert/key for overwrite tests
 	os.Remove(keyF.Name())
@@ -247,3 +248,67 @@ func Test_ca(t *testing.T) {
 	os.Remove(keyF.Name())
 
 }
+
+func Test_ca_stdio(t *testing.T) {
+	nopw := &StubPasswordReader{}
+
+	keyF, err := os.CreateTemp("", "ca.key")
+	require.NoError(t, err)
+	os.Remove(keyF.Name())
+	defer os.Remove(keyF.Name())
+
+	crtF, err := os.CreateTemp("", "ca.crt")
+	require.NoError(t, err)
+	os.Remove(crtF.Name())
+	defer os.Remove(crtF.Name())
+
+	// out-crt on stdout, out-key on disk
+	ob := &bytes.Buffer{}
+	eb := &bytes.Buffer{}
+	require.NoError(t, ca([]string{"-name", "test-ca", "-duration", "1h", "-out-crt", "-", "-out-key", keyF.Name()}, ob, eb, nopw))
+	assert.Empty(t, eb.String())
+	c, _, err := cert.UnmarshalCertificateFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.True(t, c.IsCA())
+	assert.Equal(t, "test-ca", c.Name())
+
+	// out-key on stdout, out-crt on disk
+	os.Remove(keyF.Name())
+	ob.Reset()
+	eb.Reset()
+	require.NoError(t, ca([]string{"-name", "test-ca", "-duration", "1h", "-out-crt", crtF.Name(), "-out-key", "-"}, ob, eb, nopw))
+	assert.Empty(t, eb.String())
+	_, _, curve, err := cert.UnmarshalSigningPrivateKeyFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.Equal(t, cert.Curve_CURVE25519, curve)
+
+	// dual stdout is rejected up front
+	os.Remove(crtF.Name())
+	ob.Reset()
+	eb.Reset()
+	require.EqualError(t,
+		ca([]string{"-name", "test-ca", "-duration", "1h", "-out-crt", "-", "-out-key", "-"}, ob, eb, nopw),
+		`-out-key and -out-crt both set to "-", only one output may write to stdout`)
+	assert.Empty(t, ob.String())
+
+	// an output conflict combined with -encrypt must error BEFORE prompting
+	// for a passphrase; pr would record any read attempt
+	tracker := &trackingPasswordReader{}
+	ob.Reset()
+	eb.Reset()
+	require.EqualError(t,
+		ca([]string{"-name", "test-ca", "-duration", "1h", "-encrypt", "-out-crt", "-", "-out-key", "-"}, ob, eb, tracker),
+		`-out-key and -out-crt both set to "-", only one output may write to stdout`)
+	assert.Empty(t, ob.String())
+	assert.Empty(t, eb.String())
+	assert.Zero(t, tracker.calls, "passphrase prompt should not have been called")
+}
+
+type trackingPasswordReader struct {
+	calls int
+}
+
+func (pr *trackingPasswordReader) ReadPassword() ([]byte, error) {
+	pr.calls++
+	return []byte(""), nil
+}

cmd/nebula-cert/keygen.go

@@ -42,6 +42,8 @@ func keygen(args []string, out io.Writer, errOut io.Writer) error {
 		if err = mustFlagString("out-key", cf.outKeyPath); err != nil {
 			return err
 		}
+	} else if *cf.outKeyPath != "" {
+		return newHelpErrorf("cannot set -out-key with -pkcs11")
 	}
 	if err = mustFlagString("out-pub", cf.outPubPath); err != nil {
 		return err
@@ -69,6 +71,14 @@ func keygen(args []string, out io.Writer, errOut io.Writer) error {
 		}
 	}
 
+	var claims ioClaims
+	if err := reserveOutputs(&claims,
+		"out-key", *cf.outKeyPath,
+		"out-pub", *cf.outPubPath,
+	); err != nil {
+		return err
+	}
+
 	if isP11 {
 		p11Client, err := pkclient.FromUrl(*cf.p11url)
 		if err != nil {
@@ -82,12 +92,12 @@ func keygen(args []string, out io.Writer, errOut io.Writer) error {
 			return fmt.Errorf("error while getting public key: %w", err)
 		}
 	} else {
-		err = os.WriteFile(*cf.outKeyPath, cert.MarshalPrivateKeyToPEM(curve, rawPriv), 0600)
+		err = writeOutput(*cf.outKeyPath, cert.MarshalPrivateKeyToPEM(curve, rawPriv), 0600, out)
 		if err != nil {
 			return fmt.Errorf("error while writing out-key: %s", err)
 		}
 	}
-	err = os.WriteFile(*cf.outPubPath, cert.MarshalPublicKeyToPEM(curve, pub), 0600)
+	err = writeOutput(*cf.outPubPath, cert.MarshalPublicKeyToPEM(curve, pub), 0600, out)
 	if err != nil {
 		return fmt.Errorf("error while writing out-pub: %s", err)
 	}
@@ -102,6 +112,7 @@ func keygenSummary() string {
 func keygenHelp(out io.Writer) {
 	cf := newKeygenFlags()
 	_, _ = out.Write([]byte("Usage of " + os.Args[0] + " " + keygenSummary() + "\n"))
+	_, _ = out.Write([]byte(stdioHelpText))
 	cf.set.SetOutput(out)
 	cf.set.PrintDefaults()
 }

cmd/nebula-cert/keygen_test.go

@@ -20,6 +20,7 @@ func Test_keygenHelp(t *testing.T) {
 	assert.Equal(
 		t,
 		"Usage of "+os.Args[0]+" keygen <flags>: create a public/private key pair. the public key can be passed to `nebula-cert sign`\n"+
+			"  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"+
 			"  -curve string\n"+
 			"    \tECDH Curve (25519, P256) (default \"25519\")\n"+
 			"  -out-key string\n"+
@@ -93,3 +94,43 @@ func Test_keygen(t *testing.T) {
 	require.NoError(t, err)
 	assert.Len(t, lPub, 32)
 }
+
+func Test_keygen_stdio(t *testing.T) {
+	keyF, err := os.CreateTemp("", "test.key")
+	require.NoError(t, err)
+	os.Remove(keyF.Name())
+	defer os.Remove(keyF.Name())
+
+	pubF, err := os.CreateTemp("", "test.pub")
+	require.NoError(t, err)
+	os.Remove(pubF.Name())
+	defer os.Remove(pubF.Name())
+
+	// out-pub on stdout, out-key on disk
+	ob := &bytes.Buffer{}
+	eb := &bytes.Buffer{}
+	require.NoError(t, keygen([]string{"-out-pub", "-", "-out-key", keyF.Name()}, ob, eb))
+	assert.Empty(t, eb.String())
+	lPub, _, curve, err := cert.UnmarshalPublicKeyFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.Equal(t, cert.Curve_CURVE25519, curve)
+	assert.Len(t, lPub, 32)
+
+	// out-key on stdout, out-pub on disk
+	os.Remove(keyF.Name())
+	ob.Reset()
+	eb.Reset()
+	require.NoError(t, keygen([]string{"-out-pub", pubF.Name(), "-out-key", "-"}, ob, eb))
+	assert.Empty(t, eb.String())
+	lKey, _, curve, err := cert.UnmarshalPrivateKeyFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.Equal(t, cert.Curve_CURVE25519, curve)
+	assert.Len(t, lKey, 32)
+
+	// both on stdout is a conflict caught up front
+	ob.Reset()
+	eb.Reset()
+	require.EqualError(t, keygen([]string{"-out-pub", "-", "-out-key", "-"}, ob, eb),
+		`-out-key and -out-pub both set to "-", only one output may write to stdout`)
+	assert.Empty(t, ob.String())
+}

cmd/nebula-cert/passwords.go

@@ -22,7 +22,9 @@ func (pr StdinPasswordReader) ReadPassword() ([]byte, error) {
 	}
 
 	password, err := term.ReadPassword(int(os.Stdin.Fd()))
-	fmt.Println()
+	// Terminal echo is off while reading, so the user's Enter key does not
+	// produce a visible newline. Emit one on stderr to match the prompt.
+	fmt.Fprintln(os.Stderr)
 
 	return password, err
 }

cmd/nebula-cert/print.go

@@ -40,11 +40,23 @@ func printCert(args []string, out io.Writer, errOut io.Writer) error {
 		return err
 	}
 
-	rawCert, err := os.ReadFile(*pf.path)
+	var claims ioClaims
+	if err := reserveInputs(&claims, "path", *pf.path); err != nil {
+		return err
+	}
+	if err := reserveOutputs(&claims, "out-qr", *pf.outQRPath); err != nil {
+		return err
+	}
+
+	rawCert, err := readInput("path", *pf.path, &claims)
 	if err != nil {
 		return fmt.Errorf("unable to read cert; %s", err)
 	}
 
+	// When the QR is going to stdout, suppress the human-readable text/json
+	// output so the binary stream is not contaminated.
+	qrToStdout := isStdio(*pf.outQRPath)
+
 	var c cert.Certificate
 	var qrBytes []byte
 	part := 0
@@ -57,11 +69,13 @@ func printCert(args []string, out io.Writer, errOut io.Writer) error {
 			return fmt.Errorf("error while unmarshaling cert: %s", err)
 		}
 
-		if *pf.json {
-			jsonCerts = append(jsonCerts, c)
-		} else {
-			_, _ = out.Write([]byte(c.String()))
-			_, _ = out.Write([]byte("\n"))
+		if !qrToStdout {
+			if *pf.json {
+				jsonCerts = append(jsonCerts, c)
+			} else {
+				_, _ = out.Write([]byte(c.String()))
+				_, _ = out.Write([]byte("\n"))
+			}
 		}
 
 		if *pf.outQRPath != "" {
@@ -79,7 +93,7 @@ func printCert(args []string, out io.Writer, errOut io.Writer) error {
 		part++
 	}
 
-	if *pf.json {
+	if *pf.json && !qrToStdout {
 		b, _ := json.Marshal(jsonCerts)
 		_, _ = out.Write(b)
 		_, _ = out.Write([]byte("\n"))
@@ -91,7 +105,7 @@ func printCert(args []string, out io.Writer, errOut io.Writer) error {
 			return fmt.Errorf("error while generating qr code: %s", err)
 		}
 
-		err = os.WriteFile(*pf.outQRPath, b, 0600)
+		err = writeOutput(*pf.outQRPath, b, 0600, out)
 		if err != nil {
 			return fmt.Errorf("error while writing out-qr: %s", err)
 		}
@@ -107,6 +121,7 @@ func printSummary() string {
 func printHelp(out io.Writer) {
 	pf := newPrintFlags()
 	out.Write([]byte("Usage of " + os.Args[0] + " " + printSummary() + "\n"))
+	out.Write([]byte(stdioHelpText))
 	pf.set.SetOutput(out)
 	pf.set.PrintDefaults()
 }

cmd/nebula-cert/print_test.go

@@ -25,6 +25,7 @@ func Test_printHelp(t *testing.T) {
 	assert.Equal(
 		t,
 		"Usage of "+os.Args[0]+" print <flags>: prints details about a certificate\n"+
+			"  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"+
 			"  -json\n"+
 			"    \tOptional: outputs certificates in json format\n"+
 			"  -out-qr string\n"+
@@ -178,6 +179,44 @@ func Test_printCert(t *testing.T) {
 		ob.String(),
 	)
 	assert.Empty(t, eb.String())
+
+	// read cert from stdin
+	ob.Reset()
+	eb.Reset()
+	withStdin(t, bytes.NewReader(p))
+	err = printCert([]string{"-json", "-path", "-"}, ob, eb)
+	require.NoError(t, err)
+	assert.Equal(
+		t,
+		`[{"details":{"curve":"CURVE25519","groups":["hi"],"isCa":false,"issuer":"`+c.Issuer()+`","name":"test","networks":["10.0.0.123/8"],"notAfter":"0001-01-01T00:00:00Z","notBefore":"0001-01-01T00:00:00Z","publicKey":"`+pk+`","unsafeNetworks":[]},"fingerprint":"`+fp+`","signature":"`+sig+`","version":1}]
+`,
+		ob.String(),
+	)
+	assert.Empty(t, eb.String())
+
+	// -out-qr - sends only the PNG to stdout, suppressing the cert dump
+	ob.Reset()
+	eb.Reset()
+	withStdin(t, bytes.NewReader(p))
+	err = printCert([]string{"-path", "-", "-out-qr", "-"}, ob, eb)
+	require.NoError(t, err)
+	assert.Empty(t, eb.String())
+	stdout := ob.Bytes()
+	require.NotEmpty(t, stdout)
+	// PNG magic, no PEM/JSON noise prepended
+	assert.Equal(t, []byte{0x89, 'P', 'N', 'G', 0x0d, 0x0a, 0x1a, 0x0a}, stdout[:8])
+	assert.NotContains(t, string(stdout), "NebulaCertificate")
+	assert.NotContains(t, string(stdout), `"details"`)
+
+	// json + out-qr - still suppresses json
+	ob.Reset()
+	eb.Reset()
+	withStdin(t, bytes.NewReader(p))
+	err = printCert([]string{"-json", "-path", "-", "-out-qr", "-"}, ob, eb)
+	require.NoError(t, err)
+	assert.Empty(t, eb.String())
+	assert.Equal(t, []byte{0x89, 'P', 'N', 'G'}, ob.Bytes()[:4])
+	assert.NotContains(t, ob.String(), `"details"`)
 }
 
 // NewTestCaCert will generate a CA cert

cmd/nebula-cert/sign.go

@@ -85,6 +85,9 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 	if !isP11 && *sf.inPubPath != "" && *sf.outKeyPath != "" {
 		return newHelpErrorf("cannot set both -in-pub and -out-key")
 	}
+	if isP11 && *sf.outKeyPath != "" {
+		return newHelpErrorf("cannot set -out-key with -pkcs11")
+	}
 
 	var v4Networks []netip.Prefix
 	var v6Networks []netip.Prefix
@@ -102,13 +105,35 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 		return newHelpErrorf("-version must be either %v or %v", cert.Version1, cert.Version2)
 	}
 
+	if *sf.outKeyPath == "" {
+		*sf.outKeyPath = *sf.name + ".key"
+	}
+	if *sf.outCertPath == "" {
+		*sf.outCertPath = *sf.name + ".crt"
+	}
+
+	var claims ioClaims
+	if err := reserveInputs(&claims,
+		"ca-key", *sf.caKeyPath,
+		"ca-crt", *sf.caCertPath,
+		"in-pub", *sf.inPubPath,
+	); err != nil {
+		return err
+	}
+	if err := reserveOutputs(&claims,
+		"out-key", *sf.outKeyPath,
+		"out-crt", *sf.outCertPath,
+		"out-qr", *sf.outQRPath,
+	); err != nil {
+		return err
+	}
+
 	var curve cert.Curve
 	var caKey []byte
 
 	if !isP11 {
 		var rawCAKey []byte
-		rawCAKey, err := os.ReadFile(*sf.caKeyPath)
-
+		rawCAKey, err = readInput("ca-key", *sf.caKeyPath, &claims)
 		if err != nil {
 			return fmt.Errorf("error while reading ca-key: %s", err)
 		}
@@ -121,7 +146,7 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 			if len(passphrase) == 0 {
 				// ask for a passphrase until we get one
 				for i := 0; i < 5; i++ {
-					out.Write([]byte("Enter passphrase: "))
+					errOut.Write([]byte("Enter passphrase: "))
 					passphrase, err = pr.ReadPassword()
 
 					if errors.Is(err, ErrNoTerminal) {
@@ -147,7 +172,7 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 		}
 	}
 
-	rawCACert, err := os.ReadFile(*sf.caCertPath)
+	rawCACert, err := readInput("ca-crt", *sf.caCertPath, &claims)
 	if err != nil {
 		return fmt.Errorf("error while reading ca-crt: %s", err)
 	}
@@ -245,7 +270,7 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 
 	if *sf.inPubPath != "" {
 		var pubCurve cert.Curve
-		rawPub, err := os.ReadFile(*sf.inPubPath)
+		rawPub, err := readInput("in-pub", *sf.inPubPath, &claims)
 		if err != nil {
 			return fmt.Errorf("error while reading in-pub: %s", err)
 		}
@@ -266,16 +291,10 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 		pub, rawPriv = newKeypair(curve)
 	}
 
-	if *sf.outKeyPath == "" {
-		*sf.outKeyPath = *sf.name + ".key"
-	}
-
-	if *sf.outCertPath == "" {
-		*sf.outCertPath = *sf.name + ".crt"
-	}
-
-	if _, err := os.Stat(*sf.outCertPath); err == nil {
-		return fmt.Errorf("refusing to overwrite existing cert: %s", *sf.outCertPath)
+	if !isStdio(*sf.outCertPath) {
+		if _, err := os.Stat(*sf.outCertPath); err == nil {
+			return fmt.Errorf("refusing to overwrite existing cert: %s", *sf.outCertPath)
+		}
 	}
 
 	var crts []cert.Certificate
@@ -360,11 +379,13 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 	}
 
 	if !isP11 && *sf.inPubPath == "" {
-		if _, err := os.Stat(*sf.outKeyPath); err == nil {
-			return fmt.Errorf("refusing to overwrite existing key: %s", *sf.outKeyPath)
+		if !isStdio(*sf.outKeyPath) {
+			if _, err := os.Stat(*sf.outKeyPath); err == nil {
+				return fmt.Errorf("refusing to overwrite existing key: %s", *sf.outKeyPath)
+			}
 		}
 
-		err = os.WriteFile(*sf.outKeyPath, cert.MarshalPrivateKeyToPEM(curve, rawPriv), 0600)
+		err = writeOutput(*sf.outKeyPath, cert.MarshalPrivateKeyToPEM(curve, rawPriv), 0600, out)
 		if err != nil {
 			return fmt.Errorf("error while writing out-key: %s", err)
 		}
@@ -379,7 +400,7 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 		b = append(b, sb...)
 	}
 
-	err = os.WriteFile(*sf.outCertPath, b, 0600)
+	err = writeOutput(*sf.outCertPath, b, 0600, out)
 	if err != nil {
 		return fmt.Errorf("error while writing out-crt: %s", err)
 	}
@@ -390,7 +411,7 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 			return fmt.Errorf("error while generating qr code: %s", err)
 		}
 
-		err = os.WriteFile(*sf.outQRPath, b, 0600)
+		err = writeOutput(*sf.outQRPath, b, 0600, out)
 		if err != nil {
 			return fmt.Errorf("error while writing out-qr: %s", err)
 		}
@@ -440,6 +461,7 @@ func signSummary() string {
 func signHelp(out io.Writer) {
 	sf := newSignFlags()
 	out.Write([]byte("Usage of " + os.Args[0] + " " + signSummary() + "\n"))
+	out.Write([]byte(stdioHelpText))
 	sf.set.SetOutput(out)
 	sf.set.PrintDefaults()
 }

cmd/nebula-cert/sign_test.go

@@ -27,6 +27,7 @@ func Test_signHelp(t *testing.T) {
 	assert.Equal(
 		t,
 		"Usage of "+os.Args[0]+" sign <flags>: create and sign a certificate\n"+
+			"  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"+
 			"  -ca-crt string\n"+
 			"    \tOptional: path to the signing CA cert (default \"ca.crt\")\n"+
 			"  -ca-key string\n"+
@@ -376,15 +377,18 @@ func Test_signCert(t *testing.T) {
 	// test with the proper password
 	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
 	require.NoError(t, signCert(args, ob, eb, testpw))
-	assert.Equal(t, "Enter passphrase: ", ob.String())
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, "Enter passphrase: ", eb.String())
 
 	// test with the proper password in the environment
 	os.Remove(crtF.Name())
 	os.Remove(keyF.Name())
 	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
 	os.Setenv("NEBULA_CA_PASSPHRASE", string(passphrase))
+	ob.Reset()
+	eb.Reset()
 	require.NoError(t, signCert(args, ob, eb, testpw))
+	assert.Empty(t, ob.String())
 	assert.Empty(t, eb.String())
 	os.Setenv("NEBULA_CA_PASSPHRASE", "")
 
@@ -395,8 +399,8 @@ func Test_signCert(t *testing.T) {
 	testpw.password = []byte("invalid password")
 	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
 	require.Error(t, signCert(args, ob, eb, testpw))
-	assert.Equal(t, "Enter passphrase: ", ob.String())
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, "Enter passphrase: ", eb.String())
 
 	// test with the wrong password in environment
 	ob.Reset()
@@ -416,8 +420,8 @@ func Test_signCert(t *testing.T) {
 	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
 	require.Error(t, signCert(args, ob, eb, nopw))
 	// normally the user hitting enter on the prompt would add newlines between these
-	assert.Equal(t, "Enter passphrase: Enter passphrase: Enter passphrase: Enter passphrase: Enter passphrase: ", ob.String())
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, "Enter passphrase: Enter passphrase: Enter passphrase: Enter passphrase: Enter passphrase: ", eb.String())
 
 	// test an error condition
 	ob.Reset()
@@ -425,6 +429,106 @@ func Test_signCert(t *testing.T) {
 
 	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
 	require.Error(t, signCert(args, ob, eb, errpw))
-	assert.Equal(t, "Enter passphrase: ", ob.String())
-	assert.Empty(t, eb.String())
+	assert.Empty(t, ob.String())
+	assert.Equal(t, "Enter passphrase: ", eb.String())
+}
+
+func Test_signCert_stdio(t *testing.T) {
+	nopw := &StubPasswordReader{
+		password: []byte(""),
+		err:      nil,
+	}
+
+	caPub, caPriv, _ := ed25519.GenerateKey(rand.Reader)
+	rawCAKey := cert.MarshalSigningPrivateKeyToPEM(cert.Curve_CURVE25519, caPriv)
+
+	ca, _ := NewTestCaCert("ca", caPub, caPriv, time.Now(), time.Now().Add(time.Minute*200), nil, nil, nil)
+	rawCACrt, _ := ca.MarshalPEM()
+
+	caCrtF, err := os.CreateTemp("", "sign-cert.crt")
+	require.NoError(t, err)
+	defer os.Remove(caCrtF.Name())
+	caCrtF.Write(rawCACrt)
+
+	caKeyF, err := os.CreateTemp("", "sign-cert.key")
+	require.NoError(t, err)
+	defer os.Remove(caKeyF.Name())
+	caKeyF.Write(rawCAKey)
+
+	keyF, err := os.CreateTemp("", "sign.key")
+	require.NoError(t, err)
+	os.Remove(keyF.Name())
+	defer os.Remove(keyF.Name())
+
+	// ca-key on stdin, cert to stdout
+	withStdin(t, bytes.NewReader(rawCAKey))
+	ob := &bytes.Buffer{}
+	eb := &bytes.Buffer{}
+	args := []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", "-", "-name", "stdin-test", "-ip", "1.1.1.1/24", "-out-crt", "-", "-out-key", keyF.Name(), "-duration", "100m"}
+	require.NoError(t, signCert(args, ob, eb, nopw))
+	assert.Empty(t, eb.String())
+
+	lCrt, _, err := cert.UnmarshalCertificateFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.Equal(t, "stdin-test", lCrt.Name())
+	assert.True(t, lCrt.CheckSignature(caPub))
+
+	// two flags reading from stdin should error before any read attempt;
+	// otherwise an interactive shell would hang on io.ReadAll
+	stdinIn := bytes.NewReader(rawCAKey)
+	withStdin(t, stdinIn)
+	ob.Reset()
+	eb.Reset()
+	args = []string{"-version", "1", "-ca-crt", "-", "-ca-key", "-", "-name", "stdin-test", "-ip", "1.1.1.1/24", "-out-crt", "nope", "-out-key", "nope", "-duration", "100m"}
+	require.EqualError(t, signCert(args, ob, eb, nopw),
+		`-ca-key and -ca-crt both set to "-", only one input may read from stdin`)
+	assert.Equal(t, len(rawCAKey), stdinIn.Len(), "stdin should be untouched when conflict is caught up front")
+
+	// two flags writing to stdout should error before any output is written
+	// AND before stdin is consumed
+	stdinR := bytes.NewReader(rawCAKey)
+	withStdin(t, stdinR)
+	ob.Reset()
+	eb.Reset()
+	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", "-", "-name", "stdin-test", "-ip", "1.1.1.1/24", "-out-crt", "-", "-out-key", "-", "-duration", "100m"}
+	require.EqualError(t, signCert(args, ob, eb, nopw),
+		`-out-key and -out-crt both set to "-", only one output may write to stdout`)
+	assert.Empty(t, ob.String())
+	// stdin should be untouched because the conflict was caught up front
+	assert.Equal(t, len(rawCAKey), stdinR.Len())
+
+	// out-key on stdout, cert on disk
+	keyF2, err := os.CreateTemp("", "sign.key")
+	require.NoError(t, err)
+	os.Remove(keyF2.Name())
+	defer os.Remove(keyF2.Name())
+	crtF, err := os.CreateTemp("", "sign.crt")
+	require.NoError(t, err)
+	os.Remove(crtF.Name())
+	defer os.Remove(crtF.Name())
+
+	ob.Reset()
+	eb.Reset()
+	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "stdin-test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", "-", "-duration", "100m"}
+	require.NoError(t, signCert(args, ob, eb, nopw))
+	assert.Empty(t, eb.String())
+	_, _, curve, err := cert.UnmarshalPrivateKeyFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.Equal(t, cert.Curve_CURVE25519, curve)
+
+	// in-pub on stdin (caller already has a keypair, only the cert is generated)
+	inPub, _ := x25519Keypair()
+	rawInPub := cert.MarshalPublicKeyToPEM(cert.Curve_CURVE25519, inPub)
+
+	withStdin(t, bytes.NewReader(rawInPub))
+	os.Remove(crtF.Name())
+	ob.Reset()
+	eb.Reset()
+	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "in-pub-test", "-ip", "1.1.1.1/24", "-in-pub", "-", "-out-crt", "-", "-duration", "100m"}
+	require.NoError(t, signCert(args, ob, eb, nopw))
+	assert.Empty(t, eb.String())
+	stdinCrt, _, err := cert.UnmarshalCertificateFromPEM(ob.Bytes())
+	require.NoError(t, err)
+	assert.Equal(t, "in-pub-test", stdinCrt.Name())
+	assert.Equal(t, inPub, stdinCrt.PublicKey())
 }

cmd/nebula-cert/stdio.go

@@ -0,0 +1,117 @@
+package main
+
+import (
+	"fmt"
+	"io"
+	"os"
+)
+
+// stdioPath is the special path value that selects stdin (for inputs) or
+// stdout (for outputs) instead of a file on disk.
+const stdioPath = "-"
+
+// stdioHelpText is rendered just under the Usage line of each subcommand
+// help so the - convention is documented once instead of on every flag.
+const stdioHelpText = "  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"
+
+// stdinReader is the source used when an input flag is set to "-".
+// It is a package level var so tests can swap in a deterministic reader.
+// Tests that mutate stdinReader cannot run with t.Parallel().
+var stdinReader io.Reader = os.Stdin
+
+// ioClaims tracks which flags have claimed stdin and stdout during a single
+// command invocation so we can refuse a second flag asking for the same
+// stream.
+type ioClaims struct {
+	in  string
+	out string
+}
+
+func (c *ioClaims) claimIn(flagName string) error {
+	if c.in != "" && c.in != flagName {
+		return fmt.Errorf("-%s and -%s both set to %q, only one input may read from stdin", c.in, flagName, stdioPath)
+	}
+	c.in = flagName
+	return nil
+}
+
+func (c *ioClaims) claimOut(flagName string) error {
+	if c.out != "" && c.out != flagName {
+		return fmt.Errorf("-%s and -%s both set to %q, only one output may write to stdout", c.out, flagName, stdioPath)
+	}
+	c.out = flagName
+	return nil
+}
+
+// reserveInputs walks alternating (flagName, path) pairs and claims stdin
+// for any path equal to stdioPath. It must be called before any input is
+// read so a conflict can be reported immediately instead of blocking on
+// io.ReadAll while waiting for input that will never arrive.
+func reserveInputs(claims *ioClaims, pairs ...string) error {
+	return reserveStdio(claims, "reserveInputs", (*ioClaims).claimIn, pairs)
+}
+
+// reserveOutputs walks alternating (flagName, path) pairs and claims stdout
+// for any path equal to stdioPath. It must be called before any output is
+// written so a conflict cannot leave one stream half written before the
+// second flag fails.
+func reserveOutputs(claims *ioClaims, pairs ...string) error {
+	return reserveStdio(claims, "reserveOutputs", (*ioClaims).claimOut, pairs)
+}
+
+func reserveStdio(claims *ioClaims, who string, claim func(*ioClaims, string) error, pairs []string) error {
+	if len(pairs)%2 != 0 {
+		panic(who + " requires alternating name, path pairs")
+	}
+	for i := 0; i < len(pairs); i += 2 {
+		name, path := pairs[i], pairs[i+1]
+		if path != stdioPath {
+			continue
+		}
+		if err := claim(claims, name); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// readInput returns the bytes referenced by path, reading from stdin when
+// path is stdioPath.
+func readInput(flagName, path string, claims *ioClaims) ([]byte, error) {
+	if path == stdioPath {
+		if err := claims.claimIn(flagName); err != nil {
+			return nil, err
+		}
+		return io.ReadAll(stdinReader)
+	}
+	return os.ReadFile(path)
+}
+
+// openInput returns a reader for path. When path is stdioPath the returned
+// reader wraps stdin and Close is a no-op.
+func openInput(flagName, path string, claims *ioClaims) (io.ReadCloser, error) {
+	if path == stdioPath {
+		if err := claims.claimIn(flagName); err != nil {
+			return nil, err
+		}
+		return io.NopCloser(stdinReader), nil
+	}
+	return os.Open(path)
+}
+
+// writeOutput writes data to path, or to stdout when path is stdioPath. perm
+// is only used for file output. The caller must have already claimed stdout
+// via reserveOutputs before invoking with stdioPath.
+func writeOutput(path string, data []byte, perm os.FileMode, stdout io.Writer) error {
+	if path == stdioPath {
+		_, err := stdout.Write(data)
+		return err
+	}
+	return os.WriteFile(path, data, perm)
+}
+
+// isStdio reports whether path is the stdio sentinel and so should skip
+// existence checks like "refuse to overwrite".
+func isStdio(path string) bool {
+	return path == stdioPath
+}

cmd/nebula-cert/stdio_test.go

@@ -0,0 +1,167 @@
+package main
+
+import (
+	"bytes"
+	"io"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// withStdin temporarily replaces stdinReader for the duration of t.
+func withStdin(t *testing.T, r io.Reader) {
+	t.Helper()
+	prev := stdinReader
+	stdinReader = r
+	t.Cleanup(func() { stdinReader = prev })
+}
+
+func Test_readInput_stdin(t *testing.T) {
+	withStdin(t, bytes.NewBufferString("hello"))
+	var claims ioClaims
+
+	got, err := readInput("path", "-", &claims)
+	require.NoError(t, err)
+	assert.Equal(t, []byte("hello"), got)
+	assert.Equal(t, "path", claims.in)
+}
+
+func Test_readInput_file(t *testing.T) {
+	dir := t.TempDir()
+	p := filepath.Join(dir, "f")
+	require.NoError(t, os.WriteFile(p, []byte("file"), 0600))
+	var claims ioClaims
+
+	got, err := readInput("path", p, &claims)
+	require.NoError(t, err)
+	assert.Equal(t, []byte("file"), got)
+	assert.Empty(t, claims.in)
+}
+
+func Test_readInput_doubleStdinErrors(t *testing.T) {
+	withStdin(t, bytes.NewBufferString("hello"))
+	var claims ioClaims
+
+	_, err := readInput("ca-key", "-", &claims)
+	require.NoError(t, err)
+
+	_, err = readInput("ca-crt", "-", &claims)
+	require.EqualError(t, err, `-ca-key and -ca-crt both set to "-", only one input may read from stdin`)
+}
+
+func Test_openInput_stdin(t *testing.T) {
+	withStdin(t, bytes.NewBufferString("hi"))
+	var claims ioClaims
+
+	r, err := openInput("ca", "-", &claims)
+	require.NoError(t, err)
+	defer r.Close()
+	b, err := io.ReadAll(r)
+	require.NoError(t, err)
+	assert.Equal(t, []byte("hi"), b)
+}
+
+func Test_openInput_doubleStdinErrors(t *testing.T) {
+	withStdin(t, bytes.NewBufferString("hi"))
+	var claims ioClaims
+
+	r, err := openInput("ca", "-", &claims)
+	require.NoError(t, err)
+	r.Close()
+
+	_, err = openInput("crt", "-", &claims)
+	require.EqualError(t, err, `-ca and -crt both set to "-", only one input may read from stdin`)
+}
+
+func Test_writeOutput_stdout(t *testing.T) {
+	out := &bytes.Buffer{}
+
+	err := writeOutput("-", []byte("payload"), 0600, out)
+	require.NoError(t, err)
+	assert.Equal(t, "payload", out.String())
+}
+
+func Test_writeOutput_file(t *testing.T) {
+	dir := t.TempDir()
+	p := filepath.Join(dir, "f")
+	out := &bytes.Buffer{}
+
+	err := writeOutput(p, []byte("payload"), 0600, out)
+	require.NoError(t, err)
+	assert.Empty(t, out.String())
+	got, err := os.ReadFile(p)
+	require.NoError(t, err)
+	assert.Equal(t, []byte("payload"), got)
+}
+
+func Test_reserveOutputs_noConflict(t *testing.T) {
+	var claims ioClaims
+	require.NoError(t, reserveOutputs(&claims,
+		"out-key", "/tmp/key",
+		"out-crt", "-",
+		"out-qr", "",
+	))
+	assert.Equal(t, "out-crt", claims.out)
+}
+
+func Test_reserveOutputs_conflict(t *testing.T) {
+	var claims ioClaims
+	err := reserveOutputs(&claims,
+		"out-key", "-",
+		"out-crt", "-",
+	)
+	require.EqualError(t, err, `-out-key and -out-crt both set to "-", only one output may write to stdout`)
+}
+
+func Test_reserveOutputs_panicsOnOddPairs(t *testing.T) {
+	defer func() {
+		r := recover()
+		require.NotNil(t, r)
+	}()
+	var claims ioClaims
+	_ = reserveOutputs(&claims, "out-key")
+}
+
+func Test_reserveInputs_noConflict(t *testing.T) {
+	var claims ioClaims
+	require.NoError(t, reserveInputs(&claims,
+		"ca-key", "/tmp/ca.key",
+		"ca-crt", "-",
+		"in-pub", "",
+	))
+	assert.Equal(t, "ca-crt", claims.in)
+}
+
+func Test_reserveInputs_conflict(t *testing.T) {
+	var claims ioClaims
+	err := reserveInputs(&claims,
+		"ca-key", "-",
+		"ca-crt", "-",
+	)
+	require.EqualError(t, err, `-ca-key and -ca-crt both set to "-", only one input may read from stdin`)
+}
+
+func Test_claimIn_idempotent(t *testing.T) {
+	// pre-claim then a lazy re-claim of the same flag should be a no-op
+	var claims ioClaims
+	require.NoError(t, claims.claimIn("ca-key"))
+	require.NoError(t, claims.claimIn("ca-key"))
+	assert.Equal(t, "ca-key", claims.in)
+}
+
+func Test_claimOut_idempotent(t *testing.T) {
+	var claims ioClaims
+	require.NoError(t, claims.claimOut("out-crt"))
+	require.NoError(t, claims.claimOut("out-crt"))
+	assert.Equal(t, "out-crt", claims.out)
+}
+
+func Test_isStdio(t *testing.T) {
+	assert.True(t, isStdio("-"))
+	assert.False(t, isStdio(""))
+	assert.False(t, isStdio("./-"))
+	assert.False(t, isStdio("foo"))
+}

cmd/nebula-cert/verify.go

@@ -39,18 +39,26 @@ func verify(args []string, out io.Writer, errOut io.Writer) error {
 		return err
 	}
 
-	caFile, err := os.Open(*vf.caPath)
+	var claims ioClaims
+	if err := reserveInputs(&claims,
+		"ca", *vf.caPath,
+		"crt", *vf.certPath,
+	); err != nil {
+		return err
+	}
+
+	caReader, err := openInput("ca", *vf.caPath, &claims)
 	if err != nil {
 		return fmt.Errorf("error while reading ca: %w", err)
 	}
-	defer caFile.Close()
+	defer caReader.Close()
 
-	caPool, err := cert.NewCAPoolFromPEMReader(caFile)
+	caPool, err := cert.NewCAPoolFromPEMReader(caReader)
 	if err != nil && !errors.Is(err, cert.ErrExpired) {
 		return fmt.Errorf("error while adding ca cert to pool: %w", err)
 	}
 
-	rawCert, err := os.ReadFile(*vf.certPath)
+	rawCert, err := readInput("crt", *vf.certPath, &claims)
 	if err != nil {
 		return fmt.Errorf("unable to read crt: %w", err)
 	}
@@ -85,6 +93,7 @@ func verifySummary() string {
 func verifyHelp(out io.Writer) {
 	vf := newVerifyFlags()
 	_, _ = out.Write([]byte("Usage of " + os.Args[0] + " " + verifySummary() + "\n"))
+	_, _ = out.Write([]byte(stdioHelpText))
 	vf.set.SetOutput(out)
 	vf.set.PrintDefaults()
 }

cmd/nebula-cert/verify_test.go

@@ -23,6 +23,7 @@ func Test_verifyHelp(t *testing.T) {
 	assert.Equal(
 		t,
 		"Usage of "+os.Args[0]+" verify <flags>: verifies a certificate isn't expired and was signed by a trusted authority.\n"+
+			"  Pass \"-\" to any path flag to read from stdin or write to stdout.\n"+
 			"  -ca string\n"+
 			"    \tRequired: path to a file containing one or more ca certificates\n"+
 			"  -crt string\n"+
@@ -122,3 +123,46 @@ func Test_verify(t *testing.T) {
 	assert.Empty(t, eb.String())
 	require.NoError(t, err)
 }
+
+func Test_verify_stdio(t *testing.T) {
+	ob := &bytes.Buffer{}
+	eb := &bytes.Buffer{}
+
+	caPub, caPriv, _ := ed25519.GenerateKey(rand.Reader)
+	ca, _ := NewTestCaCert("test-ca", caPub, caPriv, time.Now().Add(time.Hour*-1), time.Now().Add(time.Hour*2), nil, nil, nil)
+	caPEM, _ := ca.MarshalPEM()
+
+	crt, _ := NewTestCert(ca, caPriv, "test-cert", time.Now().Add(time.Hour*-1), time.Now().Add(time.Hour), nil, nil, nil)
+	crtPEM, _ := crt.MarshalPEM()
+
+	caFile, err := os.CreateTemp("", "verify-ca")
+	require.NoError(t, err)
+	defer os.Remove(caFile.Name())
+	caFile.Write(caPEM)
+
+	// crt on stdin, ca on disk
+	withStdin(t, bytes.NewReader(crtPEM))
+	require.NoError(t, verify([]string{"-ca", caFile.Name(), "-crt", "-"}, ob, eb))
+	assert.Empty(t, ob.String())
+	assert.Empty(t, eb.String())
+
+	// ca on stdin, crt on disk
+	certFile, err := os.CreateTemp("", "verify-cert")
+	require.NoError(t, err)
+	defer os.Remove(certFile.Name())
+	certFile.Write(crtPEM)
+
+	withStdin(t, bytes.NewReader(caPEM))
+	ob.Reset()
+	eb.Reset()
+	require.NoError(t, verify([]string{"-ca", "-", "-crt", certFile.Name()}, ob, eb))
+	assert.Empty(t, ob.String())
+	assert.Empty(t, eb.String())
+
+	// both flags on stdin should error
+	withStdin(t, bytes.NewReader(caPEM))
+	ob.Reset()
+	eb.Reset()
+	require.EqualError(t, verify([]string{"-ca", "-", "-crt", "-"}, ob, eb),
+		`-ca and -crt both set to "-", only one input may read from stdin`)
+}

cmd/nebula-service/main.go

@@ -61,9 +61,12 @@ func main() {
 	}
 
 	if *configPath == "" {
-		fmt.Println("-config flag must be set")
-		flag.Usage()
-		os.Exit(1)
+		p, err := config.DefaultPath()
+		if err != nil {
+			fmt.Println(err)
+			os.Exit(1)
+		}
+		*configPath = p
 	}
 
 	c := config.NewC(l)

cmd/nebula-service/service.go

@@ -3,8 +3,6 @@ package main
 import (
 	"fmt"
 	"log"
-	"os"
-	"path/filepath"
 
 	"github.com/kardianos/service"
 	"github.com/slackhq/nebula"
@@ -57,24 +55,13 @@ func (p *program) Stop(s service.Service) error {
 	return nil
 }
 
-func fileExists(filename string) bool {
-	_, err := os.Stat(filename)
-	if os.IsNotExist(err) {
-		return false
-	}
-	return true
-}
-
 func doService(configPath *string, configTest *bool, build string, serviceFlag *string) error {
 	if *configPath == "" {
-		ex, err := os.Executable()
+		p, err := config.DefaultPath()
 		if err != nil {
 			return err
 		}
-		*configPath = filepath.Dir(ex) + "/config.yaml"
-		if !fileExists(*configPath) {
-			*configPath = filepath.Dir(ex) + "/config.yml"
-		}
+		*configPath = p
 	}
 
 	svcConfig := &service.Config{

cmd/nebula/main.go

@@ -50,9 +50,12 @@ func main() {
 	}
 
 	if *configPath == "" {
-		fmt.Println("-config flag must be set")
-		flag.Usage()
-		os.Exit(1)
+		p, err := config.DefaultPath()
+		if err != nil {
+			fmt.Println(err)
+			os.Exit(1)
+		}
+		*configPath = p
 	}
 
 	l := logging.NewLogger(os.Stdout)

config/default.go

@@ -0,0 +1,29 @@
+package config
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+)
+
+// DefaultPath returns a path to a config file alongside the running executable, preferring config.yaml over config.yml.
+// If neither file exists an error is returned that names both paths checked.
+func DefaultPath() (string, error) {
+	ex, err := os.Executable()
+	if err != nil {
+		return "", err
+	}
+	return defaultPathInDir(filepath.Dir(ex))
+}
+
+func defaultPathInDir(dir string) (string, error) {
+	yamlPath := filepath.Join(dir, "config.yaml")
+	if _, err := os.Stat(yamlPath); err == nil {
+		return yamlPath, nil
+	}
+	ymlPath := filepath.Join(dir, "config.yml")
+	if _, err := os.Stat(ymlPath); err == nil {
+		return ymlPath, nil
+	}
+	return "", fmt.Errorf("no default config found at %s or %s", yamlPath, ymlPath)
+}

config/default_test.go

@@ -0,0 +1,67 @@
+package config
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestDefaultPathInDir(t *testing.T) {
+	t.Run("prefers config.yaml when both exist", func(t *testing.T) {
+		dir := t.TempDir()
+		want := filepath.Join(dir, "config.yaml")
+		other := filepath.Join(dir, "config.yml")
+		require.NoError(t, os.WriteFile(want, []byte("a: 1"), 0644))
+		require.NoError(t, os.WriteFile(other, []byte("a: 2"), 0644))
+
+		got, err := defaultPathInDir(dir)
+		require.NoError(t, err)
+		assert.Equal(t, want, got)
+	})
+
+	t.Run("returns config.yaml when only it exists", func(t *testing.T) {
+		dir := t.TempDir()
+		want := filepath.Join(dir, "config.yaml")
+		require.NoError(t, os.WriteFile(want, []byte("a: 1"), 0644))
+
+		got, err := defaultPathInDir(dir)
+		require.NoError(t, err)
+		assert.Equal(t, want, got)
+	})
+
+	t.Run("falls back to config.yml when only it exists", func(t *testing.T) {
+		dir := t.TempDir()
+		want := filepath.Join(dir, "config.yml")
+		require.NoError(t, os.WriteFile(want, []byte("a: 1"), 0644))
+
+		got, err := defaultPathInDir(dir)
+		require.NoError(t, err)
+		assert.Equal(t, want, got)
+	})
+
+	t.Run("errors when neither exists and names both paths", func(t *testing.T) {
+		dir := t.TempDir()
+		got, err := defaultPathInDir(dir)
+		assert.Empty(t, got)
+		require.Error(t, err)
+		assert.Contains(t, err.Error(), filepath.Join(dir, "config.yaml"))
+		assert.Contains(t, err.Error(), filepath.Join(dir, "config.yml"))
+	})
+}
+
+func TestDefaultPath(t *testing.T) {
+	got, err := DefaultPath()
+	if err != nil {
+		ex, exErr := os.Executable()
+		require.NoError(t, exErr)
+		assert.Contains(t, err.Error(), filepath.Dir(ex))
+		return
+	}
+	ex, err := os.Executable()
+	require.NoError(t, err)
+	assert.Equal(t, filepath.Dir(ex), filepath.Dir(got))
+	assert.Contains(t, []string{"config.yaml", "config.yml"}, filepath.Base(got))
+}

connection_manager.go

@@ -136,14 +136,6 @@ func (cm *connectionManager) getAndResetTrafficCheck(h *HostInfo, now time.Time)
 	return in, out
 }
 
-// AddTrafficWatch must be called for every new HostInfo.
-// We will continue to monitor the HostInfo until the tunnel is dropped.
-func (cm *connectionManager) AddTrafficWatch(h *HostInfo) {
-	if h.out.Swap(true) == false {
-		cm.trafficTimer.Add(h.localIndexId, cm.checkInterval)
-	}
-}
-
 func (cm *connectionManager) Start(ctx context.Context) {
 	clockSource := time.NewTicker(cm.trafficTimer.t.tickDuration)
 	defer clockSource.Stop()
@@ -306,8 +298,8 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 		} 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,
+				"relayFrom", relayFrom,
+				"relayTo", relayTo,
 				"initiatorRelayIndex", req.InitiatorRelayIndex,
 				"responderRelayIndex", req.ResponderRelayIndex,
 				"vpnAddrs", newhostinfo.vpnAddrs,

dns_server.go

@@ -11,19 +11,21 @@ import (
 	"sync"
 	"sync/atomic"
 
-	"github.com/gaissmai/bart"
 	"github.com/miekg/dns"
 	"github.com/slackhq/nebula/config"
 )
 
 type dnsServer struct {
 	sync.RWMutex
-	l               *slog.Logger
-	ctx             context.Context
-	dnsMap4         map[string]netip.Addr
-	dnsMap6         map[string]netip.Addr
-	hostMap         *HostMap
-	myVpnAddrsTable *bart.Lite
+	l       *slog.Logger
+	ctx     context.Context
+	dnsMap4 map[string]netip.Addr
+	dnsMap6 map[string]netip.Addr
+	hostMap *HostMap
+	pki     *PKI
+
+	// selfHost is the cached FQDN we last seeded for ourselves
+	selfHost string
 
 	mux *dns.ServeMux
 
@@ -55,14 +57,14 @@ 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 *slog.Logger, pki *PKI, hostMap *HostMap, c *config.C) (*dnsServer, error) {
 	ds := &dnsServer{
-		l:               l,
-		ctx:             ctx,
-		dnsMap4:         make(map[string]netip.Addr),
-		dnsMap6:         make(map[string]netip.Addr),
-		hostMap:         hostMap,
-		myVpnAddrsTable: cs.myVpnAddrsTable,
+		l:       l,
+		ctx:     ctx,
+		dnsMap4: make(map[string]netip.Addr),
+		dnsMap6: make(map[string]netip.Addr),
+		hostMap: hostMap,
+		pki:     pki,
 	}
 	ds.mux = dns.NewServeMux()
 	ds.mux.HandleFunc(".", ds.handleDnsRequest)
@@ -76,6 +78,7 @@ func newDnsServerFromConfig(ctx context.Context, l *slog.Logger, cs *CertState,
 	if err := ds.reload(c, true); err != nil {
 		return ds, err
 	}
+	ds.seedSelf()
 	return ds, nil
 }
 
@@ -113,7 +116,7 @@ func (d *dnsServer) reload(c *config.C, initial bool) error {
 			d.Stop()
 		}
 		// Drop any records that accumulated while enabled; a later re-enable
-		// will repopulate from fresh handshakes.
+		// will repopulate from fresh handshakes and a fresh seedSelf.
 		d.clearRecords()
 		return nil
 	}
@@ -121,17 +124,14 @@ func (d *dnsServer) reload(c *config.C, initial bool) error {
 	if running == nil {
 		// Was disabled (or never started); bring it up now.
 		go d.Start()
-		return nil
+	} else if !sameAddr {
+		d.shutdownServer(running, runningStarted, "reload")
+		// Old Start goroutine has now exited; bring up a fresh listener on the new address.
+		go d.Start()
 	}
 
-	if sameAddr {
-		return nil
-	}
-
-	d.shutdownServer(running, runningStarted, "reload")
-	// Old Start goroutine has now exited; bring up a fresh listener on the
-	// new address.
-	go d.Start()
+	// Refresh the self entry every enabled reload so cert renewals that change our name or VPN addresses are picked up.
+	d.seedSelf()
 	return nil
 }
 
@@ -249,6 +249,20 @@ func (d *dnsServer) QueryCert(data string) string {
 		return ""
 	}
 
+	// The hostmap only ever contains peers we have handshaked with, so it never carries an entry for ourselves.
+	// Answer self lookups straight from the local cert state.
+	if cs := d.certState(); cs != nil && cs.myVpnAddrsTable != nil && cs.myVpnAddrsTable.Contains(ip) {
+		c := cs.GetDefaultCertificate()
+		if c == nil {
+			return ""
+		}
+		b, err := c.MarshalJSON()
+		if err != nil {
+			return ""
+		}
+		return string(b)
+	}
+
 	hostinfo := d.hostMap.QueryVpnAddr(ip)
 	if hostinfo == nil {
 		return ""
@@ -266,12 +280,60 @@ func (d *dnsServer) QueryCert(data string) string {
 	return string(b)
 }
 
-// clearRecords drops all DNS records.
+// clearRecords drops all DNS records, including the self entry.
 func (d *dnsServer) clearRecords() {
 	d.Lock()
 	defer d.Unlock()
 	clear(d.dnsMap4)
 	clear(d.dnsMap6)
+	d.selfHost = ""
+}
+
+// seedSelf inserts (or refreshes) a record for our own cert name pointing at our VPN addresses,
+// so a single-lighthouse network can resolve the lighthouse's own hostname without the two-process workaround.
+func (d *dnsServer) seedSelf() {
+	if !d.enabled.Load() {
+		return
+	}
+	cs := d.certState()
+	if cs == nil {
+		return
+	}
+	c := cs.GetDefaultCertificate()
+	if c == nil {
+		return
+	}
+	newHost := strings.ToLower(c.Name()) + "."
+
+	d.Lock()
+	defer d.Unlock()
+	if d.selfHost != "" && d.selfHost != newHost {
+		delete(d.dnsMap4, d.selfHost)
+		delete(d.dnsMap6, d.selfHost)
+	}
+	d.selfHost = newHost
+	delete(d.dnsMap4, newHost)
+	delete(d.dnsMap6, newHost)
+	haveV4, haveV6 := false, false
+	for _, addr := range cs.myVpnAddrs {
+		if addr.Is4() && !haveV4 {
+			d.dnsMap4[newHost] = addr
+			haveV4 = true
+		} else if addr.Is6() && !haveV6 {
+			d.dnsMap6[newHost] = addr
+			haveV6 = true
+		}
+		if haveV4 && haveV6 {
+			break
+		}
+	}
+}
+
+func (d *dnsServer) certState() *CertState {
+	if d.pki == nil {
+		return nil
+	}
+	return d.pki.getCertState()
 }
 
 // Add adds the first IPv4 and IPv6 address that appears in `addresses` as the record for `host`
@@ -309,8 +371,12 @@ func (d *dnsServer) isSelfNebulaOrLocalhost(addr string) bool {
 		return true
 	}
 
+	cs := d.certState()
+	if cs == nil || cs.myVpnAddrsTable == nil {
+		return false
+	}
 	//if we found it in this table, it's good
-	return d.myVpnAddrsTable.Contains(b)
+	return cs.myVpnAddrsTable.Contains(b)
 }
 
 func (d *dnsServer) parseQuery(m *dns.Msg, w dns.ResponseWriter) {

dns_server_test.go

@@ -9,7 +9,10 @@ import (
 	"testing"
 	"time"
 
+	"github.com/gaissmai/bart"
 	"github.com/miekg/dns"
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/config"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -276,6 +279,92 @@ func TestDnsServer_Stop_beforeBind_doesNotHang(t *testing.T) {
 	}
 }
 
+// newTestPKI builds a minimal *PKI with a single v1 cert whose name and
+// VPN addresses are caller-provided, suitable for exercising seedSelf and
+// QueryCert self handling.
+func newTestPKI(t *testing.T, name string, addrs []netip.Addr) *PKI {
+	t.Helper()
+	networks := make([]netip.Prefix, 0, len(addrs))
+	for _, a := range addrs {
+		bits := 32
+		if a.Is6() {
+			bits = 128
+		}
+		networks = append(networks, netip.PrefixFrom(a, bits))
+	}
+	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Time{}, time.Time{}, nil, nil, nil)
+	c, _, _, _ := cert_test.NewTestCert(cert.Version2, cert.Curve_CURVE25519, ca, caKey, name, time.Time{}, time.Time{}, networks, nil, nil)
+
+	addrsTable := new(bart.Lite)
+	for _, a := range addrs {
+		addrsTable.Insert(netip.PrefixFrom(a, a.BitLen()))
+	}
+
+	cs := &CertState{
+		v2Cert:            c,
+		initiatingVersion: cert.Version2,
+		myVpnAddrs:        addrs,
+		myVpnAddrsTable:   addrsTable,
+	}
+	pki := &PKI{}
+	pki.cs.Store(cs)
+	return pki
+}
+
+func TestDnsServer_seedSelf_addsOwnRecord(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	myV4 := netip.MustParseAddr("10.0.0.1")
+	myV6 := netip.MustParseAddr("fd00::1")
+	ds.pki = newTestPKI(t, "lighthouse", []netip.Addr{myV4, myV6})
+	setDnsConfig(c, "127.0.0.1", "0", true, true)
+	require.NoError(t, ds.reload(c, true))
+
+	ds.seedSelf()
+	got4, exists := ds.Query(dns.TypeA, "lighthouse.")
+	assert.True(t, exists)
+	assert.Equal(t, myV4, got4)
+	got6, exists := ds.Query(dns.TypeAAAA, "lighthouse.")
+	assert.True(t, exists)
+	assert.Equal(t, myV6, got6)
+}
+
+func TestDnsServer_seedSelf_disabled_noOp(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	ds.pki = newTestPKI(t, "lighthouse", []netip.Addr{netip.MustParseAddr("10.0.0.1")})
+	setDnsConfig(c, "127.0.0.1", "0", true, false)
+	require.NoError(t, ds.reload(c, true))
+
+	ds.seedSelf()
+	_, exists := ds.Query(dns.TypeA, "lighthouse.")
+	assert.False(t, exists)
+}
+
+func TestDnsServer_clearRecords_dropsSelfHost(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	ds.pki = newTestPKI(t, "lighthouse", []netip.Addr{netip.MustParseAddr("10.0.0.1")})
+	setDnsConfig(c, "127.0.0.1", "0", true, true)
+	require.NoError(t, ds.reload(c, true))
+	ds.seedSelf()
+	require.NotEmpty(t, ds.selfHost)
+
+	ds.clearRecords()
+	assert.Empty(t, ds.selfHost)
+	_, exists := ds.Query(dns.TypeA, "lighthouse.")
+	assert.False(t, exists)
+}
+
+func TestDnsServer_QueryCert_returnsOwnCert(t *testing.T) {
+	ds, _ := newTestDnsServer(t)
+	myV4 := netip.MustParseAddr("10.0.0.1")
+	ds.pki = newTestPKI(t, "lighthouse", []netip.Addr{myV4})
+
+	got := ds.QueryCert(myV4.String() + ".")
+	assert.NotEmpty(t, got, "TXT lookup of our own VPN address should return our cert")
+
+	other := netip.MustParseAddr("10.0.0.99")
+	assert.Empty(t, ds.QueryCert(other.String()+"."), "unknown peer IP should return nothing")
+}
+
 func TestDnsServer_reload_disable_stopsRunningServer(t *testing.T) {
 	port := freeUDPPort(t)
 	ds, c := newTestDnsServer(t)

docker/Dockerfile

@@ -1,6 +1,16 @@
 FROM gcr.io/distroless/static:latest
 
 ARG TARGETOS TARGETARCH
+
+ARG VERSION=dev
+ARG REVISION=unknown
+LABEL org.opencontainers.image.title="nebula" \
+      org.opencontainers.image.description="A scalable overlay networking tool with a focus on performance, simplicity and security" \
+      org.opencontainers.image.vendor="Nebula OSS" \
+      org.opencontainers.image.source="https://github.com/slackhq/nebula" \
+      org.opencontainers.image.version="${VERSION}" \
+      org.opencontainers.image.revision="${REVISION}"
+
 COPY build/$TARGETOS-$TARGETARCH/nebula /nebula
 COPY build/$TARGETOS-$TARGETARCH/nebula-cert /nebula-cert
 

e2e/tunnels_test.go

@@ -15,6 +15,7 @@ import (
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 	"gopkg.in/yaml.v3"
 )
 
@@ -373,6 +374,100 @@ func TestCrossStackRelaysWork(t *testing.T) {
 	//relayControl.Stop()
 }
 
+// TestRelayReplayProtection asserts that a relay (forwarding-type) node rejects
+// replayed relay frames. A captured relay frame, re-injected with the same
+// message counter, must be dropped by the replay window rather than re-forwarded
+// to the relay target. Before the fix, handleOutsideRelayPacket authenticated the
+// frame but never advanced the replay window, so every replay was re-forwarded.
+func TestRelayReplayProtection(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}})
+	theirUdp := netip.MustParseAddrPort("10.0.0.2:4242")
+	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServerWithUdp(cert.Version2, ca, caKey, "them   ", "fc00::2/64", theirUdp, m{"relay": m{"use_relays": true}})
+
+	myVpnV6 := myVpnIpNet[1]
+	relayVpnV4 := relayVpnIpNet[0]
+	relayVpnV6 := relayVpnIpNet[1]
+	theirVpnV6 := theirVpnIpNet[0]
+
+	// Teach me how to reach the relay and that them is reachable via the relay
+	myControl.InjectLightHouseAddr(relayVpnV4.Addr(), relayUdpAddr)
+	myControl.InjectLightHouseAddr(relayVpnV6.Addr(), relayUdpAddr)
+	myControl.InjectRelays(theirVpnV6.Addr(), []netip.Addr{relayVpnV6.Addr()})
+	relayControl.InjectLightHouseAddr(theirVpnV6.Addr(), theirUdpAddr)
+
+	r := router.NewR(t, myControl, relayControl, theirControl)
+	defer r.RenderFlow()
+
+	myControl.Start()
+	relayControl.Start()
+	theirControl.Start()
+
+	// Establish the relayed tunnel in both directions so all handshakes complete.
+	t.Log("Establish the relayed tunnel")
+	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnV6.Addr(), 80, myVpnV6.Addr(), 80, []byte("Hi from me")))
+	p := r.RouteForAllUntilTxTun(theirControl)
+	assertUdpPacket(t, []byte("Hi from me"), p, myVpnV6.Addr(), theirVpnV6.Addr(), 80, 80)
+	theirControl.InjectTunPacket(BuildTunUDPPacket(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)
+
+	// Drain anything still queued on me's UDP tx so the next packet we pull is the
+	// relay frame we are about to generate.
+	for myControl.GetFromUDP(false) != nil {
+	}
+
+	// Capture a single legitimate relay frame that me transmits toward the relay.
+	t.Log("Capture a relay frame from me -> relay")
+	myControl.InjectTunPacket(BuildTunUDPPacket(theirVpnV6.Addr(), 80, myVpnV6.Addr(), 80, []byte("replay me")))
+	relayFrame := myControl.GetFromUDP(true)
+	require.Equal(t, relayUdpAddr, relayFrame.To, "captured frame should be addressed to the relay")
+	var fh header.H
+	require.NoError(t, fh.Parse(relayFrame.Data))
+	require.Equal(t, header.Message, fh.Type)
+	require.Equal(t, header.MessageRelay, fh.Subtype)
+
+	// drainForwards counts relay frames the relay forwards toward them within the
+	// settle window. We match on destination + (Message, MessageRelay) so the
+	// relay's own direct traffic to them can't be miscounted.
+	drainForwards := func(settle time.Duration) int {
+		ch := relayControl.GetUDPTxChan()
+		count := 0
+		for {
+			select {
+			case pkt := <-ch:
+				var ph header.H
+				if pkt.To == theirUdpAddr && ph.Parse(pkt.Data) == nil &&
+					ph.Type == header.Message && ph.Subtype == header.MessageRelay {
+					count++
+				}
+				pkt.Release()
+			case <-time.After(settle):
+				return count
+			}
+		}
+	}
+
+	// First delivery of the captured frame: the relay should forward it once.
+	t.Log("Deliver the captured frame once; relay forwards it to them")
+	relayControl.InjectUDPPacket(relayFrame)
+	require.Equal(t, 1, drainForwards(200*time.Millisecond), "relay should forward the first, legitimate copy")
+
+	// Replay the exact same frame several times. A correct replay window rejects
+	// these duplicates so the relay forwards none of them.
+	t.Log("Replay the captured frame; relay must drop the duplicates")
+	const replays = 3
+	for i := 0; i < replays; i++ {
+		relayControl.InjectUDPPacket(relayFrame)
+	}
+	forwarded := drainForwards(200 * time.Millisecond)
+	assert.Equal(t, 0, forwarded, "relay re-forwarded %d/%d replayed relay frames; replay protection is ineffective on relay tunnels", forwarded, replays)
+
+	r.RenderHostmaps("Final hostmaps", myControl, relayControl, theirControl)
+}
+
 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{})

examples/config.yml

@@ -397,7 +397,7 @@ firewall:
   #   `drop` (default): silently drop the packet.
   #   `reject`: send a reject reply.
   #     - For TCP, this will be a RST "Connection Reset" packet.
-  #     - For other protocols, this will be an ICMP port unreachable packet.
+  #     - For other protocols, this will be an ICMP "Destination unreachable: Communication administratively prohibited" packet.
   outbound_action: drop
   inbound_action: drop
 

firewall.go

@@ -58,8 +58,9 @@ type Firewall struct {
 	routableNetworks *bart.Lite
 
 	// assignedNetworks is a list of vpn networks assigned to us in the certificate.
-	assignedNetworks  []netip.Prefix
-	hasUnsafeNetworks bool
+	assignedNetworks []netip.Prefix
+	// unsafeNetworks is the list of unsafe networks issued to us in the certificate
+	unsafeNetworks []netip.Prefix
 
 	rules        string
 	rulesVersion uint16
@@ -158,10 +159,9 @@ func NewFirewall(l *slog.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Dur
 		assignedNetworks = append(assignedNetworks, network)
 	}
 
-	hasUnsafeNetworks := false
-	for _, n := range c.UnsafeNetworks() {
+	unsafeNetworks := c.UnsafeNetworks()
+	for _, n := range unsafeNetworks {
 		routableNetworks.Insert(n)
-		hasUnsafeNetworks = true
 	}
 
 	return &Firewall{
@@ -169,15 +169,15 @@ func NewFirewall(l *slog.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Dur
 			Conns:      make(map[firewall.Packet]*conn),
 			TimerWheel: NewTimerWheel[firewall.Packet](tmin, tmax),
 		},
-		InRules:           newFirewallTable(),
-		OutRules:          newFirewallTable(),
-		TCPTimeout:        tcpTimeout,
-		UDPTimeout:        UDPTimeout,
-		DefaultTimeout:    defaultTimeout,
-		routableNetworks:  routableNetworks,
-		assignedNetworks:  assignedNetworks,
-		hasUnsafeNetworks: hasUnsafeNetworks,
-		l:                 l,
+		InRules:          newFirewallTable(),
+		OutRules:         newFirewallTable(),
+		TCPTimeout:       tcpTimeout,
+		UDPTimeout:       UDPTimeout,
+		DefaultTimeout:   defaultTimeout,
+		routableNetworks: routableNetworks,
+		assignedNetworks: assignedNetworks,
+		unsafeNetworks:   unsafeNetworks,
+		l:                l,
 
 		incomingMetrics: firewallMetrics{
 			droppedLocalAddr:  metrics.GetOrRegisterCounter("firewall.incoming.dropped.local_addr", nil),
@@ -897,7 +897,7 @@ func (flc *firewallLocalCIDR) addRule(f *Firewall, localCidr string) error {
 	}
 
 	if localCidr == "" {
-		if !f.hasUnsafeNetworks || f.defaultLocalCIDRAny {
+		if len(f.unsafeNetworks) == 0 || f.defaultLocalCIDRAny {
 			flc.Any = true
 			return nil
 		}
@@ -1055,7 +1055,6 @@ func (r *rule) sanity() error {
 }
 
 func parsePort(s string) (int32, int32, error) {
-	var err error
 	const notAPort int32 = -2
 	if s == "any" {
 		return firewall.PortAny, firewall.PortAny, nil
@@ -1064,11 +1063,11 @@ func parsePort(s string) (int32, int32, error) {
 		return firewall.PortFragment, firewall.PortFragment, nil
 	}
 	if !strings.Contains(s, `-`) {
-		rPort, err := strconv.Atoi(s)
+		rPort, err := parsePortValue("", s)
 		if err != nil {
-			return notAPort, notAPort, fmt.Errorf("was not a number; `%s`", s)
+			return notAPort, notAPort, err
 		}
-		return int32(rPort), int32(rPort), nil
+		return rPort, rPort, nil
 	}
 
 	sPorts := strings.SplitN(s, `-`, 2)
@@ -1079,22 +1078,40 @@ func parsePort(s string) (int32, int32, error) {
 		return notAPort, notAPort, fmt.Errorf("appears to be a range but could not be parsed; `%s`", s)
 	}
 
-	rStartPort, err := strconv.Atoi(sPorts[0])
+	startPort, err := parsePortValue("beginning range ", sPorts[0])
 	if err != nil {
-		return notAPort, notAPort, fmt.Errorf("beginning range was not a number; `%s`", sPorts[0])
+		return notAPort, notAPort, err
 	}
 
-	rEndPort, err := strconv.Atoi(sPorts[1])
+	endPort, err := parsePortValue("ending range ", sPorts[1])
 	if err != nil {
-		return notAPort, notAPort, fmt.Errorf("ending range was not a number; `%s`", sPorts[1])
+		return notAPort, notAPort, err
 	}
 
-	startPort := int32(rStartPort)
-	endPort := int32(rEndPort)
-
 	if startPort == firewall.PortAny {
 		endPort = firewall.PortAny
 	}
 
 	return startPort, endPort, nil
 }
+
+// parsePortValue accepts a base-10 decimal in [0, 65535] and returns it
+// widened to int32. Using strconv.ParseUint with bitSize 16 rejects
+// negative input, out-of-range input (>65535), and any non-decimal byte
+// by construction, so the int32 widening that follows is provably safe
+// and cannot collide with firewall.PortAny (0) or firewall.PortFragment
+// (-1) via integer truncation.
+//
+// prefix is prepended to both error messages so callers can disambiguate
+// the single-port path (prefix="") from the range bounds (prefix="beginning
+// range " / "ending range "), preserving the historical error strings.
+func parsePortValue(prefix, s string) (int32, error) {
+	n, err := strconv.ParseUint(s, 10, 16)
+	if err == nil {
+		return int32(n), nil
+	}
+	if errors.Is(err, strconv.ErrRange) {
+		return 0, fmt.Errorf("%sout of range [0,65535]; `%s`", prefix, s)
+	}
+	return 0, fmt.Errorf("%swas not a number; `%s`", prefix, s)
+}

firewall_test.go

@@ -1029,6 +1029,75 @@ func Test_parsePort(t *testing.T) {
 	require.NoError(t, err)
 }
 
+// Test_parsePort_invalid covers inputs that must error. The named bug is
+// that int32(strconv.Atoi("4294967296")) truncates to 0 == firewall.PortAny,
+// silently turning a typo into a match-all-ports rule; the rest are
+// representative syntax/range probes.
+func Test_parsePort_invalid(t *testing.T) {
+	tests := []struct {
+		name            string
+		input           string
+		wantErrContains string
+	}{
+		// Numeric overflow (the named bug + boundary).
+		{"named bug: 2^32 truncates to PortAny", "4294967296", "out of range"},
+		{"just above max real port", "65536", "out of range"},
+
+		// Negatives route through the range branch and hit the empty-half
+		// guard; included as defense in depth so a future refactor cannot
+		// accidentally reach the int32 cast.
+		{"negative", "-1", "could not be parsed"},
+
+		// Syntax probes.
+		{"NUL between digits", "4\x002", "was not a number"},
+		{"hex notation", "0x10", "was not a number"},
+		{"scientific notation", "1e3", "was not a number"},
+		{"leading whitespace", " 42", "was not a number"},
+		{"fullwidth digits", "４２", "was not a number"},
+
+		// Range branch.
+		{"range upper out of range", "1-65536", "ending range out of range"},
+		{"range lower out of range", "65536-65537", "beginning range out of range"},
+		{"range with negative upper", "1--1", "ending range was not a number"},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			_, _, err := parsePort(tc.input)
+			require.Error(t, err, "input %q must error", tc.input)
+			require.ErrorContains(t, err, tc.wantErrContains)
+		})
+	}
+}
+
+// Test_parsePort_valid_boundaries locks in success cases at 0, 1, and 65535
+// so a future refactor cannot regress the boundaries.
+func Test_parsePort_valid_boundaries(t *testing.T) {
+	tests := []struct {
+		name      string
+		input     string
+		wantStart int32
+		wantEnd   int32
+	}{
+		{"zero is PortAny", "0", 0, 0},
+		{"min real port", "1", 1, 1},
+		{"max real port", "65535", 65535, 65535},
+		{"range zero to max forces end to zero", "0-65535", 0, 0},
+		{"range max to max", "65535-65535", 65535, 65535},
+		{"range one to max", "1-65535", 1, 65535},
+		{"range with whitespace inside", " 1 - 2    ", 1, 2},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			s, e, err := parsePort(tc.input)
+			require.NoError(t, err)
+			assert.Equal(t, tc.wantStart, s, "start port")
+			assert.Equal(t, tc.wantEnd, e, "end port")
+		})
+	}
+}
+
 func TestNewFirewallFromConfig(t *testing.T) {
 	l := test.NewLogger()
 	// Test a bad rule definition

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 v0.28.0
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
 	github.com/kardianos/service v1.2.4
@@ -24,12 +24,12 @@ require (
 	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/crypto v0.53.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.53.0
-	golang.org/x/sync v0.20.0
-	golang.org/x/sys v0.43.0
-	golang.org/x/term v0.42.0
+	golang.org/x/net v0.56.0
+	golang.org/x/sync v0.21.0
+	golang.org/x/sys v0.46.0
+	golang.org/x/term v0.44.0
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2
 	golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b
 	golang.zx2c4.com/wireguard/windows v0.6.1

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 v0.28.0 h1:89yZLo8NmyqD0RYgJ3QO9HhqqGGw+oWhf90cZm69Lko=
+github.com/gaissmai/bart v0.28.0/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=
@@ -162,8 +162,8 @@ golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACk
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/crypto v0.0.0-20210322153248-0c34fe9e7dc2/go.mod h1:T9bdIzuCu7OtxOm1hfPfRQxPLYneinmdGuTeoZ9dtd4=
-golang.org/x/crypto v0.50.0 h1:zO47/JPrL6vsNkINmLoo/PH1gcxpls50DNogFvB5ZGI=
-golang.org/x/crypto v0.50.0/go.mod h1:3muZ7vA7PBCE6xgPX7nkzzjiUq87kRItoJQM1Yo8S+Q=
+golang.org/x/crypto v0.53.0 h1:QZ4Muo8THX6CizN2vPPd5fBGHyogrdK9fG4wLPFUsto=
+golang.org/x/crypto v0.53.0/go.mod h1:DNLU434OwVakk9PzuwV8w62mAJpRJL3vsgcfp4Qnsio=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090 h1:Di6/M8l0O2lCLc6VVRWhgCiApHV8MnQurBnFSHsQtNY=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090/go.mod h1:FXUEEKJgO7OQYeo8N01OfiKP8RXMtf6e8aTskBGqWdc=
 golang.org/x/lint v0.0.0-20200302205851-738671d3881b/go.mod h1:3xt1FjdF8hUf6vQPIChWIBhFzV8gjjsPE/fR3IyQdNY=
@@ -182,8 +182,8 @@ 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 v0.56.0 h1:Rw8j/hFzGvJUZwNBXnAtf5sVDVt+65SK2C7IxCxZt5o=
+golang.org/x/net v0.56.0/go.mod h1:D3Ku6r+V6JROoZK144D2XfMHFcMq/0zSfLelVTCFKec=
 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=
@@ -191,8 +191,8 @@ golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201207232520-09787c993a3a/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
-golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
-golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
+golang.org/x/sync v0.21.0 h1:HLII4xRRTtCRkxYp4HNFF0Js/Og6q2i++KXbg0gHCwM=
+golang.org/x/sync v0.21.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
 golang.org/x/sys v0.0.0-20180905080454-ebe1bf3edb33/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20181116152217-5ac8a444bdc5/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -208,11 +208,11 @@ golang.org/x/sys v0.0.0-20210124154548-22da62e12c0c/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20210603081109-ebe580a85c40/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.2.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.10.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.43.0 h1:Rlag2XtaFTxp19wS8MXlJwTvoh8ArU6ezoyFsMyCTNI=
-golang.org/x/sys v0.43.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/sys v0.46.0 h1:noSf2Fq6F8DBgS+LysIkx7rIExoNHJsxOAtPp4rthXw=
+golang.org/x/sys v0.46.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
-golang.org/x/term v0.42.0 h1:UiKe+zDFmJobeJ5ggPwOshJIVt6/Ft0rcfrXZDLWAWY=
-golang.org/x/term v0.42.0/go.mod h1:Dq/D+snpsbazcBG5+F9Q1n2rXV8Ma+71xEjTRufARgY=
+golang.org/x/term v0.44.0 h1:0rLvDRCtNj0gZkyIXhCyOb2OAzEhLVqc4B+hrsBhrmc=
+golang.org/x/term v0.44.0/go.mod h1:7ze4MdzUzLXpSAoFP1H0bOI9aXDqveSvatT5vKcFh2Y=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=

handshake/errors.go

@@ -13,6 +13,7 @@ var (
 	ErrUnknownSubtype          = errors.New("unknown handshake subtype")
 	ErrMissingContent          = errors.New("expected handshake content but message was empty")
 	ErrUnexpectedContent       = errors.New("received unexpected handshake content")
+	ErrInvalidRemoteIndex      = errors.New("peer sent an invalid index in handshake payload")
 	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")

handshake/machine.go

@@ -312,11 +312,19 @@ func (m *Machine) processPayload(msg []byte, flags msgFlags) error {
 
 	// Process payload
 	if flags.expectsPayload {
+		var remoteIndex uint32
 		if m.result.Initiator {
-			m.result.RemoteIndex = payload.ResponderIndex
+			remoteIndex = payload.ResponderIndex
 		} else {
-			m.result.RemoteIndex = payload.InitiatorIndex
+			remoteIndex = payload.InitiatorIndex
 		}
+		// The payload presence check above can be satisfied by Time alone, so a payload
+		// could still carry a zero index here. We need to reject it.
+		if remoteIndex == 0 {
+			m.failed = true
+			return ErrInvalidRemoteIndex
+		}
+		m.result.RemoteIndex = remoteIndex
 		m.result.HandshakeTime = payload.Time
 		m.payloadSet = true
 	}

handshake/machine_test.go

@@ -229,6 +229,24 @@ func TestMachineProcessPayload(t *testing.T) {
 		require.ErrorIs(t, err, ErrUnexpectedContent)
 		assert.True(t, m.Failed())
 	})
+
+	t.Run("zero initiator index on responder is fatal", func(t *testing.T) {
+		m := newTestMachine(t, cs, v, false, 100)
+		bytes := MarshalPayload(nil, Payload{InitiatorIndex: 0, Time: 1})
+		err := m.processPayload(bytes, msgFlags{expectsPayload: true})
+		require.ErrorIs(t, err, ErrInvalidRemoteIndex)
+		assert.True(t, m.Failed())
+		assert.Zero(t, m.result.RemoteIndex)
+	})
+
+	t.Run("zero responder index on initiator is fatal", func(t *testing.T) {
+		m := newTestMachine(t, cs, v, true, 100)
+		bytes := MarshalPayload(nil, Payload{InitiatorIndex: 100, ResponderIndex: 0, Time: 1})
+		err := m.processPayload(bytes, msgFlags{expectsPayload: true})
+		require.ErrorIs(t, err, ErrInvalidRemoteIndex)
+		assert.True(t, m.Failed())
+		assert.Zero(t, m.result.RemoteIndex)
+	})
 }
 
 // TestMachineRequireComplete checks the fail-on-incomplete-handshake path

handshake_manager.go

@@ -83,6 +83,7 @@ type HandshakeHostInfo struct {
 	initiatingVersionOverride cert.Version     // Should we use a non-default cert version for this handshake?
 	counter                   int64            // How many attempts have we made so far
 	lastRemotes               []netip.AddrPort // Remotes that we sent to during the previous attempt
+	lastRelays                []netip.Addr     // Relays we attempted to use during the previous attempt
 	packetStore               []*cachedPacket  // A set of packets to be transmitted once the handshake completes
 
 	hostinfo *HostInfo
@@ -217,7 +218,6 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 		fields := []any{
 			"udpAddrs", hh.hostinfo.remotes.CopyAddrs(hm.mainHostMap.GetPreferredRanges()),
 			"initiatorIndex", hh.hostinfo.localIndexId,
-			"remoteIndex", hh.hostinfo.remoteIndexId,
 			"durationNs", time.Since(hh.startTime).Nanoseconds(),
 		}
 		// hh.machine can be nil here if buildStage0Packet never succeeded
@@ -323,7 +323,7 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 		)
 	}
 
-	hm.f.relayManager.StartRelays(hm.f, vpnIp, hostinfo, stage0)
+	hm.f.relayManager.StartRelays(hm.f, vpnIp, hh, stage0)
 
 	// If a lighthouse triggered this attempt then we are still in the timer wheel and do not need to re-add
 	if !lighthouseTriggered {
@@ -465,7 +465,6 @@ func (hm *HandshakeManager) CheckAndComplete(hostinfo *HostInfo, handshakePacket
 		// We have a collision, but this can happen since we can't control
 		// the remote ID. Just log about the situation as a note.
 		hostinfo.logger(hm.l).Info("New host shadows existing host remoteIndex",
-			"remoteIndex", hostinfo.remoteIndexId,
 			"collision", existingRemoteIndex.vpnAddrs,
 		)
 	}
@@ -488,7 +487,6 @@ func (hm *HandshakeManager) Complete(hostinfo *HostInfo, f *Interface) {
 		// We have a collision, but this can happen since we can't control
 		// the remote ID. Just log about the situation as a note.
 		hostinfo.logger(hm.l).Info("New host shadows existing host remoteIndex",
-			"remoteIndex", hostinfo.remoteIndexId,
 			"collision", existingRemoteIndex.vpnAddrs,
 		)
 	}
@@ -798,7 +796,6 @@ func (hm *HandshakeManager) beginHandshake(via ViaSender, packet []byte, h *head
 	}
 
 	hm.sendHandshakeResponse(via, response, hostinfo, false)
-	f.connectionManager.AddTrafficWatch(hostinfo)
 	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
 
 	// Don't wait for UpdateWorker
@@ -965,7 +962,6 @@ func (hm *HandshakeManager) continueHandshake(via ViaSender, hh *HandshakeHostIn
 	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate)
 
 	hm.Complete(hostinfo, f)
-	f.connectionManager.AddTrafficWatch(hostinfo)
 
 	if len(hh.packetStore) > 0 {
 		if f.l.Enabled(context.Background(), slog.LevelDebug) {

hostmap.go

@@ -138,9 +138,9 @@ func (rs *RelayState) InsertRelayTo(ip netip.Addr) {
 }
 
 func (rs *RelayState) CopyRelayIps() []netip.Addr {
-	ret := make([]netip.Addr, len(rs.relays))
 	rs.RLock()
 	defer rs.RUnlock()
+	ret := make([]netip.Addr, len(rs.relays))
 	copy(ret, rs.relays)
 	return ret
 }
@@ -623,6 +623,11 @@ func (hm *HostMap) unlockedAddHostInfo(hostinfo *HostInfo, f *Interface) {
 	hm.Indexes[hostinfo.localIndexId] = hostinfo
 	hm.RemoteIndexes[hostinfo.remoteIndexId] = hostinfo
 
+	hostinfo.out.Store(true)
+	if f.connectionManager != nil { // f.connectionManager is only nil in some unit tests
+		f.connectionManager.trafficTimer.Add(hostinfo.localIndexId, f.connectionManager.checkInterval)
+	}
+
 	if hm.l.Enabled(context.Background(), slog.LevelDebug) {
 		hm.l.Debug("Hostmap vpnIp added",
 			"hostMap", m{"vpnAddrs": hostinfo.vpnAddrs, "mapTotalSize": len(hm.Hosts),

inside.go

@@ -391,7 +391,6 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 			"error", err,
 			"udpAddr", remote,
 			"counter", c,
-			"attemptedCounter", c,
 		)
 		return
 	}

interface.go

@@ -7,6 +7,7 @@ import (
 	"io"
 	"log/slog"
 	"net/netip"
+	"slices"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -14,6 +15,7 @@ import (
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
 
+	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
@@ -375,13 +377,22 @@ func (f *Interface) reloadDisconnectInvalid(c *config.C) {
 }
 
 func (f *Interface) reloadFirewall(c *config.C) {
-	//TODO: need to trigger/detect if the certificate changed too
-	if c.HasChanged("firewall") == false {
+	cs := f.pki.getCertState()
+	curCert := cs.getCertificate(cert.Version2)
+	if curCert == nil {
+		curCert = cs.getCertificate(cert.Version1)
+	}
+
+	// The firewall builds its routableNetworks set from the certificate's UnsafeNetworks at construction.
+	// Check to see if that set has changed, and if so, rebuild the firewall.
+	certUnsafeChanged := curCert != nil && !slices.Equal(curCert.UnsafeNetworks(), f.firewall.unsafeNetworks)
+
+	if !c.HasChanged("firewall") && !certUnsafeChanged {
 		f.l.Debug("No firewall config change detected")
 		return
 	}
 
-	fw, err := NewFirewallFromConfig(f.l, f.pki.getCertState(), c)
+	fw, err := NewFirewallFromConfig(f.l, cs, c)
 	if err != nil {
 		f.l.Error("Error while creating firewall during reload", "error", err)
 		return
@@ -491,26 +502,34 @@ func (f *Interface) emitStats(ctx context.Context, i time.Duration) {
 	certInitiatingVersion := metrics.GetOrRegisterGauge("certificate.initiating_version", nil)
 	certMaxVersion := metrics.GetOrRegisterGauge("certificate.max_version", nil)
 
+	emit := func() {
+		f.firewall.EmitStats()
+		f.handshakeManager.EmitStats()
+		udpStats()
+
+		certState := f.pki.getCertState()
+		defaultCrt := certState.GetDefaultCertificate()
+		certExpirationGauge.Update(int64(defaultCrt.NotAfter().Sub(time.Now()) / time.Second))
+		certInitiatingVersion.Update(int64(defaultCrt.Version()))
+
+		// Report the max certificate version we are capable of using
+		if certState.v2Cert != nil {
+			certMaxVersion.Update(int64(certState.v2Cert.Version()))
+		} else {
+			certMaxVersion.Update(int64(certState.v1Cert.Version()))
+		}
+	}
+
+	// Prime gauges so a Prometheus scrape that lands before the first tick
+	// sees real values instead of the zero defaults (issue #907).
+	emit()
+
 	for {
 		select {
 		case <-ctx.Done():
 			return
 		case <-ticker.C:
-			f.firewall.EmitStats()
-			f.handshakeManager.EmitStats()
-			udpStats()
-
-			certState := f.pki.getCertState()
-			defaultCrt := certState.GetDefaultCertificate()
-			certExpirationGauge.Update(int64(defaultCrt.NotAfter().Sub(time.Now()) / time.Second))
-			certInitiatingVersion.Update(int64(defaultCrt.Version()))
-
-			// Report the max certificate version we are capable of using
-			if certState.v2Cert != nil {
-				certMaxVersion.Update(int64(certState.v2Cert.Version()))
-			} else {
-				certMaxVersion.Update(int64(certState.v1Cert.Version()))
-			}
+			emit()
 		}
 	}
 }

interface_emit_test.go

@@ -0,0 +1,73 @@
+//go:build linux || darwin
+
+package nebula
+
+import (
+	"context"
+	"net/netip"
+	"testing"
+	"time"
+
+	"github.com/rcrowley/go-metrics"
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/firewall"
+	"github.com/slackhq/nebula/overlay/overlaytest"
+	"github.com/slackhq/nebula/test"
+	"github.com/slackhq/nebula/udp"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// Test_emitStats_primesGauges covers issue #907: a Prometheus scrape that
+// landed before the first ticker fire used to read 0 for the cert gauges.
+// emitStats now primes the gauges before entering the ticker loop. We assert
+// the gauge is zero before the first call and non-zero after.
+func Test_emitStats_primesGauges(t *testing.T) {
+	defer metrics.DefaultRegistry.UnregisterAll()
+
+	l := test.NewLogger()
+	hostMap := newHostMap(l)
+	preferredRanges := []netip.Prefix{netip.MustParsePrefix("10.0.0.0/8")}
+	hostMap.preferredRanges.Store(&preferredRanges)
+
+	notAfter := time.Now().Add(time.Hour)
+	cs := &CertState{
+		initiatingVersion: cert.Version1,
+		privateKey:        []byte{},
+		v1Cert:            &dummyCert{version: cert.Version1, notAfter: notAfter},
+		v1Credential:      nil,
+	}
+
+	lh := newTestLighthouse()
+	ifce := &Interface{
+		hostMap:          hostMap,
+		inside:           &overlaytest.NoopTun{},
+		outside:          &udp.NoopConn{},
+		firewall:         &Firewall{Conntrack: &FirewallConntrack{Conns: map[firewall.Packet]*conn{}}},
+		lightHouse:       lh,
+		pki:              &PKI{},
+		handshakeManager: NewHandshakeManager(l, hostMap, lh, &udp.NoopConn{}, defaultHandshakeConfig),
+		l:                l,
+		// On linux, udp.NewUDPStatsEmitter indexes writers[0] and asserts to
+		// *udp.StdConn. A zero value works: getMemInfo sees a nil rawConn,
+		// returns an error, and the emitter falls through to a no-op.
+		writers: []udp.Conn{&udp.StdConn{}},
+	}
+	ifce.pki.cs.Store(cs)
+
+	ttlGauge := metrics.GetOrRegisterGauge("certificate.ttl_seconds", nil)
+	require.Zero(t, ttlGauge.Value(), "gauge should be zero before emitStats runs")
+
+	// Pre-cancel the context so emitStats returns after priming the gauges
+	// without ever reading from ticker.C. The one hour interval is just a
+	// belt-and-suspenders, the test does not expect the ticker to fire.
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	ifce.emitStats(ctx, time.Hour)
+
+	ttl := ttlGauge.Value()
+	assert.Positive(t, ttl, "ttl gauge should be primed by emitStats before its first tick")
+	assert.LessOrEqual(t, ttl, int64(3600))
+	assert.Equal(t, int64(cert.Version1), metrics.GetOrRegisterGauge("certificate.initiating_version", nil).Value())
+	assert.Equal(t, int64(cert.Version1), metrics.GetOrRegisterGauge("certificate.max_version", nil).Value())
+}

interface_test.go

@@ -0,0 +1,120 @@
+package nebula
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/test"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// TestReloadFirewall_CertUnsafeNetworksChanged verifies that reloadFirewall
+// rebuilds the firewall when only the certificate's UnsafeNetworks have changed,
+// even if the firewall section of the YAML has not.
+func TestReloadFirewall_CertUnsafeNetworksChanged(t *testing.T) {
+	l := test.NewLogger()
+
+	vpnNet := netip.MustParsePrefix("10.0.0.1/24")
+	initialUnsafe := []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")}
+
+	// dummyCert avoids dragging the real signing pipeline into a unit test.
+	c1 := &dummyCert{
+		version:        cert.Version2,
+		networks:       []netip.Prefix{vpnNet},
+		unsafeNetworks: initialUnsafe,
+	}
+	pki := &PKI{}
+	pki.cs.Store(&CertState{v2Cert: c1, initiatingVersion: cert.Version2})
+
+	rawYAML := `firewall:
+  outbound:
+    - port: any
+      proto: any
+      host: any
+  inbound:
+    - port: any
+      proto: any
+      host: any
+`
+	cfg := config.NewC(l)
+	require.NoError(t, cfg.LoadString(rawYAML))
+
+	fw, err := NewFirewallFromConfig(l, pki.getCertState(), cfg)
+	require.NoError(t, err)
+	require.Equal(t, initialUnsafe, fw.unsafeNetworks)
+
+	f := &Interface{
+		pki:      pki,
+		firewall: fw,
+		l:        l,
+	}
+
+	// Swap the cert with a different UnsafeNetworks set.
+	newUnsafe := []netip.Prefix{
+		netip.MustParsePrefix("198.51.100.0/24"),
+		netip.MustParsePrefix("203.0.113.0/24"),
+	}
+	c2 := &dummyCert{
+		version:        cert.Version2,
+		networks:       []netip.Prefix{vpnNet},
+		unsafeNetworks: newUnsafe,
+	}
+	pki.cs.Store(&CertState{v2Cert: c2, initiatingVersion: cert.Version2})
+
+	// Reload with the same YAML so HasChanged("firewall") reports false.
+	require.NoError(t, cfg.ReloadConfigString(rawYAML))
+	require.False(t, cfg.HasChanged("firewall"))
+
+	f.reloadFirewall(cfg)
+
+	assert.NotSame(t, fw, f.firewall, "firewall pointer should have been replaced")
+	assert.Equal(t, newUnsafe, f.firewall.unsafeNetworks)
+	assert.True(t, f.firewall.routableNetworks.Contains(netip.MustParseAddr("203.0.113.5")))
+}
+
+// TestReloadFirewall_NoChange verifies that reloadFirewall is a no-op when
+// neither the firewall config nor the cert's UnsafeNetworks have changed.
+func TestReloadFirewall_NoChange(t *testing.T) {
+	l := test.NewLogger()
+
+	vpnNet := netip.MustParsePrefix("10.0.0.1/24")
+	unsafe := []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")}
+
+	c1 := &dummyCert{
+		version:        cert.Version2,
+		networks:       []netip.Prefix{vpnNet},
+		unsafeNetworks: unsafe,
+	}
+	pki := &PKI{}
+	pki.cs.Store(&CertState{v2Cert: c1, initiatingVersion: cert.Version2})
+
+	rawYAML := `firewall:
+  outbound:
+    - port: any
+      proto: any
+      host: any
+  inbound:
+    - port: any
+      proto: any
+      host: any
+`
+	cfg := config.NewC(l)
+	require.NoError(t, cfg.LoadString(rawYAML))
+
+	fw, err := NewFirewallFromConfig(l, pki.getCertState(), cfg)
+	require.NoError(t, err)
+
+	f := &Interface{
+		pki:      pki,
+		firewall: fw,
+		l:        l,
+	}
+
+	require.NoError(t, cfg.ReloadConfigString(rawYAML))
+	f.reloadFirewall(cfg)
+
+	assert.Same(t, fw, f.firewall, "firewall should not have been replaced")
+}

iputil/packet.go

@@ -4,26 +4,54 @@ import (
 	"encoding/binary"
 
 	"golang.org/x/net/ipv4"
+	"golang.org/x/net/ipv6"
 )
 
 const (
-	// Need 96 bytes for the largest reject packet:
+	// MaxIPv4RejectPacketSize is the largest IPv4 reject packet:
 	// - 20 byte ipv4 header
 	// - 8 byte icmpv4 header
 	// - 68 byte body (60 byte max orig ipv4 header + 8 byte orig icmpv4 header)
-	MaxRejectPacketSize = ipv4.HeaderLen + 8 + 60 + 8
+	maxIPv4RejectPacketSize = ipv4.HeaderLen + 8 + 60 + 8
+
+	// MaxRejectPacketSize is sized for the largest possible reject packet (IPv6):
+	// - 40 byte ipv6 header
+	// - 8 byte icmpv6 header
+	// - up to 1000 byte body (original packet, possibly truncated. We want to stay
+	//   under the MTU with Nebula overhead included)
+	maxIPv6RejectPacketSize = ipv6.HeaderLen + 8 + 1000
+
+	MaxRejectPacketSize = maxIPv6RejectPacketSize
 )
 
 func CreateRejectPacket(packet []byte, out []byte) []byte {
-	if len(packet) < ipv4.HeaderLen || int(packet[0]>>4) != ipv4.Version {
+	if len(packet) < 1 {
 		return nil
 	}
 
-	switch packet[9] {
-	case 6: // tcp
-		return ipv4CreateRejectTCPPacket(packet, out)
+	version := int(packet[0] >> 4)
+	switch version {
+	case ipv4.Version:
+		if len(packet) < ipv4.HeaderLen {
+			return nil
+		}
+		// Do not send reject packets for non-first fragments
+		if packet[6]&0x1f != 0 || packet[7] != 0 {
+			return nil
+		}
+		switch packet[9] {
+		case 6: // tcp
+			return ipv4CreateRejectTCPPacket(packet, out)
+		default:
+			return ipv4CreateRejectICMPPacket(packet, out)
+		}
+	case ipv6.Version:
+		if len(packet) < ipv6.HeaderLen {
+			return nil
+		}
+		return ipv6CreateRejectPacket(packet, out)
 	default:
-		return ipv4CreateRejectICMPPacket(packet, out)
+		return nil
 	}
 }
 
@@ -35,12 +63,17 @@ func ipv4CreateRejectICMPPacket(packet []byte, out []byte) []byte {
 		return nil
 	}
 
-	// ICMP reply includes original header and first 8 bytes of the packet
-	packetLen := len(packet)
-	if packetLen > ihl+8 {
-		packetLen = ihl + 8
+	// Do not generate ICMP errors in response to ICMP error packets
+	if packet[9] == 1 && len(packet) > ihl {
+		icmpType := packet[ihl]
+		if icmpType == 3 || icmpType == 4 || icmpType == 5 || icmpType == 11 || icmpType == 12 {
+			return nil
+		}
 	}
 
+	// ICMP reply includes original header and first 8 bytes of the packet
+	packetLen := min(len(packet), ihl+8)
+
 	outLen := ipv4.HeaderLen + 8 + packetLen
 	if outLen > cap(out) {
 		return nil
@@ -71,14 +104,14 @@ func ipv4CreateRejectICMPPacket(packet []byte, out []byte) []byte {
 
 	// ICMP Destination Unreachable
 	icmpOut := out[ipv4.HeaderLen:]
-	icmpOut[0] = 3 // type (Destination unreachable)
-	icmpOut[1] = 3 // code (Port unreachable error)
-	icmpOut[2] = 0 // checksum
-	icmpOut[3] = 0 //  .
-	icmpOut[4] = 0 // unused
-	icmpOut[5] = 0 //  .
-	icmpOut[6] = 0 //  .
-	icmpOut[7] = 0 //  .
+	icmpOut[0] = 3  // type (Destination unreachable)
+	icmpOut[1] = 13 // code (Communication administratively prohibited)
+	icmpOut[2] = 0  // checksum
+	icmpOut[3] = 0  //  .
+	icmpOut[4] = 0  // unused
+	icmpOut[5] = 0  //  .
+	icmpOut[6] = 0  //  .
+	icmpOut[7] = 0  //  .
 
 	// Copy original IP header and first 8 bytes as body
 	copy(icmpOut[8:], packet[:packetLen])
@@ -165,7 +198,193 @@ func ipv4CreateRejectTCPPacket(packet []byte, out []byte) []byte {
 	return out
 }
 
+func ipv6CreateRejectPacket(packet []byte, out []byte) []byte {
+	proto, offset, isFragment := ipv6FindUpperProtocol(packet)
+	if isFragment {
+		return nil
+	}
+	switch proto {
+	case 6: // tcp
+		return ipv6CreateRejectTCPPacket(packet, out, offset)
+	default:
+		return ipv6CreateRejectICMPPacket(packet, out, proto, offset)
+	}
+}
+
+func ipv6CreateRejectICMPPacket(packet []byte, out []byte, proto uint8, offset int) []byte {
+	// Do not generate ICMPv6 errors in response to ICMPv6 error packets
+	if proto == 58 && len(packet) > offset {
+		icmpType := packet[offset]
+		if icmpType >= 1 && icmpType <= 4 {
+			return nil
+		}
+	}
+
+	// Include as much of the original packet as possible, up to 1000 bytes,
+	// so the response fits comfortably within any tunnel MTU.
+	packetLen := min(len(packet), 1000)
+
+	outLen := ipv6.HeaderLen + 8 + packetLen
+	if outLen > cap(out) {
+		return nil
+	}
+
+	out = out[:outLen]
+
+	// IPv6 header
+	ipHdr := out[0:ipv6.HeaderLen]
+	ipHdr[0] = ipv6.Version << 4 // version, traffic class (high bits)
+	ipHdr[1] = 0                 // traffic class (low bits), flow label (high bits)
+	ipHdr[2] = 0                 // flow label
+	ipHdr[3] = 0                 // flow label
+
+	payloadLen := uint16(outLen - ipv6.HeaderLen)
+	binary.BigEndian.PutUint16(ipHdr[4:], payloadLen) // payload length
+	ipHdr[6] = 58                                     // next header (ICMPv6)
+	ipHdr[7] = 64                                     // hop limit
+
+	// Swap dest / src IPs (each 16 bytes, src at 8, dst at 24)
+	copy(ipHdr[8:24], packet[24:40])
+	copy(ipHdr[24:40], packet[8:24])
+
+	// ICMPv6 Destination Unreachable
+	icmpOut := out[ipv6.HeaderLen:]
+	icmpOut[0] = 1 // type (Destination Unreachable)
+	icmpOut[1] = 1 // code (Communication with destination administratively prohibited)
+	icmpOut[2] = 0 // checksum
+	icmpOut[3] = 0 //  .
+	icmpOut[4] = 0 // unused
+	icmpOut[5] = 0 //  .
+	icmpOut[6] = 0 //  .
+	icmpOut[7] = 0 //  .
+
+	copy(icmpOut[8:], packet[:packetLen])
+
+	// ICMPv6 checksum uses a pseudo-header
+	csum := ipv6PseudoheaderChecksum(ipHdr[8:24], ipHdr[24:40], 58, uint32(payloadLen))
+	binary.BigEndian.PutUint16(icmpOut[2:], tcpipChecksum(icmpOut, csum))
+
+	return out
+}
+
+func ipv6CreateRejectTCPPacket(packet []byte, out []byte, offset int) []byte {
+	const tcpLen = 20
+
+	if len(packet) < offset+tcpLen {
+		return nil
+	}
+
+	outLen := ipv6.HeaderLen + tcpLen
+	if outLen > cap(out) {
+		return nil
+	}
+
+	out = out[:outLen]
+
+	// IPv6 header
+	ipHdr := out[0:ipv6.HeaderLen]
+	ipHdr[0] = ipv6.Version << 4 // version, traffic class (high bits)
+	ipHdr[1] = 0                 // traffic class (low bits), flow label (high bits)
+	ipHdr[2] = 0                 // flow label
+	ipHdr[3] = 0                 // flow label
+
+	binary.BigEndian.PutUint16(ipHdr[4:], tcpLen) // payload length
+	ipHdr[6] = 6                                  // next header (TCP)
+	ipHdr[7] = 64                                 // hop limit
+
+	// Swap dest / src IPs
+	copy(ipHdr[8:24], packet[24:40])
+	copy(ipHdr[24:40], packet[8:24])
+
+	// TCP RST
+	tcpIn := packet[offset:]
+	var ackSeq, seq uint32
+	outFlags := byte(0b00000100) // RST
+
+	inAck := tcpIn[13]&0b00010000 != 0
+	if inAck {
+		seq = binary.BigEndian.Uint32(tcpIn[8:])
+	} else {
+		inSyn := uint32((tcpIn[13] & 0b00000010) >> 1)
+		inFin := uint32(tcpIn[13] & 0b00000001)
+		ackSeq = binary.BigEndian.Uint32(tcpIn[4:]) + inSyn + inFin + uint32(len(tcpIn)) - uint32(tcpIn[12]>>4)<<2
+		outFlags |= 0b00010000 // ACK
+	}
+
+	tcpOut := out[ipv6.HeaderLen:]
+	// Swap dest / src ports
+	copy(tcpOut[0:2], tcpIn[2:4])
+	copy(tcpOut[2:4], tcpIn[0:2])
+	binary.BigEndian.PutUint32(tcpOut[4:], seq)
+	binary.BigEndian.PutUint32(tcpOut[8:], ackSeq)
+	tcpOut[12] = (tcpLen >> 2) << 4 // data offset, reserved, NS
+	tcpOut[13] = outFlags           // CWR, ECE, URG, ACK, PSH, RST, SYN, FIN
+	tcpOut[14] = 0                  // window size
+	tcpOut[15] = 0                  //  .
+	tcpOut[16] = 0                  // checksum
+	tcpOut[17] = 0                  //  .
+	tcpOut[18] = 0                  // URG Pointer
+	tcpOut[19] = 0                  //  .
+
+	// Calculate checksum with IPv6 pseudo-header
+	csum := ipv6PseudoheaderChecksum(ipHdr[8:24], ipHdr[24:40], 6, tcpLen)
+	binary.BigEndian.PutUint16(tcpOut[16:], tcpipChecksum(tcpOut, csum))
+
+	return out
+}
+
+func ipv6FindUpperProtocol(packet []byte) (nextHeader uint8, offset int, isFragment bool) {
+	nextHeader = packet[6]
+	offset = ipv6.HeaderLen
+
+	for {
+		switch nextHeader {
+		case 0, 43, 60: // Hop-by-Hop, Routing, Destination
+			if len(packet) < offset+2 {
+				return nextHeader, offset, isFragment
+			}
+			nextHeader = packet[offset]
+			offset += int(packet[offset+1]+1) << 3
+
+		case 44: // Fragment
+			if len(packet) < offset+8 {
+				return nextHeader, offset, isFragment
+			}
+			if packet[offset+2] != 0 || packet[offset+3]&0xf8 != 0 {
+				isFragment = true
+			}
+			nextHeader = packet[offset]
+			offset += 8
+
+		case 51: // AH
+			if len(packet) < offset+2 {
+				return nextHeader, offset, isFragment
+			}
+			nextHeader = packet[offset]
+			offset += int(packet[offset+1]+2) << 2
+
+		default:
+			return nextHeader, offset, isFragment
+		}
+	}
+}
+
 func CreateICMPEchoResponse(packet, out []byte) []byte {
+	if len(packet) < 1 {
+		return nil
+	}
+
+	switch packet[0] >> 4 {
+	case 4:
+		return createICMPv4EchoResponse(packet, out)
+	case 6:
+		return createICMPv6EchoResponse(packet, out)
+	default:
+		return nil
+	}
+}
+
+func createICMPv4EchoResponse(packet, out []byte) []byte {
 	// Return early if this is not a simple ICMP Echo Request
 	//TODO: make constants out of these
 	if !(len(packet) >= 28 && len(packet) <= 9001 && packet[0] == 0x45 && packet[9] == 0x01 && packet[20] == 0x08) {
@@ -199,6 +418,43 @@ func CreateICMPEchoResponse(packet, out []byte) []byte {
 	return out
 }
 
+func createICMPv6EchoResponse(packet, out []byte) []byte {
+	// IPv6 header (40 bytes) + ICMPv6 header (8 bytes minimum)
+	if len(packet) < ipv6.HeaderLen+8 || len(packet) > 9001 {
+		return nil
+	}
+
+	// Next Header must be ICMPv6 (58)
+	if packet[6] != 58 {
+		return nil
+	}
+
+	// ICMPv6 type must be Echo Request (128)
+	if packet[ipv6.HeaderLen] != 128 {
+		return nil
+	}
+
+	out = out[:len(packet)]
+	copy(out, packet)
+
+	// Swap src/dst addresses (bytes 8-23 and 24-39)
+	copy(out[8:24], packet[24:40])
+	copy(out[24:40], packet[8:24])
+
+	// Change ICMPv6 type to Echo Reply (129)
+	icmp := out[ipv6.HeaderLen:]
+	icmp[0] = 129
+	icmp[2] = 0
+	icmp[3] = 0
+
+	// ICMPv6 checksum uses a pseudo-header with src, dst, length, and next header
+	payloadLen := uint32(len(icmp))
+	csum := ipv6PseudoheaderChecksum(out[8:24], out[24:40], 58, payloadLen)
+	binary.BigEndian.PutUint16(icmp[2:], tcpipChecksum(icmp, csum))
+
+	return out
+}
+
 // calculates the TCP/IP checksum defined in rfc1071. The passed-in
 // csum is any initial checksum data that's already been computed.
 //
@@ -236,3 +492,18 @@ func ipv4PseudoheaderChecksum(src, dst []byte, proto, length uint32) (csum uint3
 	csum += length >> 16
 	return csum
 }
+
+// based on:
+// - https://github.com/google/gopacket/blob/v1.1.19/layers/tcpip.go#L37-L48
+func ipv6PseudoheaderChecksum(src, dst []byte, proto, length uint32) (csum uint32) {
+	for i := 0; i < 16; i += 2 {
+		csum += uint32(src[i]) << 8
+		csum += uint32(src[i+1])
+		csum += uint32(dst[i]) << 8
+		csum += uint32(dst[i+1])
+	}
+	csum += proto
+	csum += length & 0xffff
+	csum += length >> 16
+	return csum
+}

iputil/packet_test.go

@@ -1,11 +1,13 @@
 package iputil
 
 import (
+	"encoding/binary"
 	"net"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
 	"golang.org/x/net/ipv4"
+	"golang.org/x/net/ipv6"
 )
 
 func Test_CreateRejectPacket(t *testing.T) {
@@ -43,7 +45,7 @@ func Test_CreateRejectPacket(t *testing.T) {
 	}
 	b = append(b, []byte{0, 3, 0, 4, 0, 0, 0, 0}...)
 
-	expectedLen = MaxRejectPacketSize
+	expectedLen = maxIPv4RejectPacketSize
 	out = make([]byte, MaxRejectPacketSize)
 	rejectPacket = CreateRejectPacket(b, out)
 	assert.NotNil(t, rejectPacket)
@@ -71,3 +73,404 @@ func Test_CreateRejectPacket(t *testing.T) {
 	assert.NotNil(t, rejectPacket)
 	assert.Len(t, rejectPacket, expectedLen)
 }
+
+func Test_CreateRejectPacket_NoFragment(t *testing.T) {
+	out := make([]byte, MaxRejectPacketSize)
+
+	// IPv4: non-zero fragment offset should not generate reject packet
+	h := ipv4.Header{
+		Len:      20,
+		Src:      net.IPv4(10, 0, 0, 1),
+		Dst:      net.IPv4(10, 0, 0, 2),
+		Protocol: 17, // UDP
+	}
+	b, err := h.Marshal()
+	if err != nil {
+		t.Fatalf("h.Marshal: %v", err)
+	}
+	b = append(b, make([]byte, 8)...)
+	// Set fragment offset to non-zero (byte 6-7, offset in 8-byte units)
+	b[6] = 0x00
+	b[7] = 0x01
+	assert.Nil(t, CreateRejectPacket(b, out))
+
+	// MF flag with zero offset (first fragment) should still generate reject
+	b[6] = 0x20 // MF flag set
+	b[7] = 0x00
+	assert.NotNil(t, CreateRejectPacket(b, out))
+
+	// Non-fragment should still generate reject packet
+	b[6] = 0x00
+	b[7] = 0x00
+	assert.NotNil(t, CreateRejectPacket(b, out))
+
+	// DF flag only (not a fragment) should still generate reject packet
+	b[6] = 0x40
+	b[7] = 0x00
+	assert.NotNil(t, CreateRejectPacket(b, out))
+}
+
+func Test_CreateRejectPacketIPv6_NoFragment(t *testing.T) {
+	src := net.ParseIP("fd00::1")
+	dst := net.ParseIP("fd00::2")
+	out := make([]byte, MaxRejectPacketSize)
+
+	// IPv6 with Fragment header and non-zero offset should not generate reject
+	fragHeader := []byte{
+		17,   // next header: UDP
+		0,    // reserved
+		0, 9, // fragment offset=1 (shifted left 3), M=1
+		0, 0, 0, 1, // identification
+	}
+	udpPayload := make([]byte, 8)
+	payload := append(fragHeader, udpPayload...)
+	packet := makeIPv6Packet(src, dst, 44, payload) // next header 44 = Fragment
+	assert.Nil(t, CreateRejectPacket(packet, out))
+
+	// Fragment header with zero offset (first fragment) should still generate reject
+	fragHeader[2] = 0
+	fragHeader[3] = 1 // offset=0, M=1
+	payload = append(fragHeader, udpPayload...)
+	packet = makeIPv6Packet(src, dst, 44, payload)
+	assert.NotNil(t, CreateRejectPacket(packet, out))
+}
+
+func Test_CreateRejectPacket_NoICMPError(t *testing.T) {
+	out := make([]byte, MaxRejectPacketSize)
+
+	// ICMP error types should not generate reject packets
+	icmpErrorTypes := []byte{3, 4, 5, 11, 12}
+	for _, icmpType := range icmpErrorTypes {
+		h := ipv4.Header{
+			Len:      20,
+			Src:      net.IPv4(10, 0, 0, 1),
+			Dst:      net.IPv4(10, 0, 0, 2),
+			Protocol: 1, // ICMP
+		}
+
+		b, err := h.Marshal()
+		if err != nil {
+			t.Fatalf("h.Marshal: %v", err)
+		}
+		b = append(b, icmpType, 0, 0, 0, 0, 0, 0, 0)
+
+		rejectPacket := CreateRejectPacket(b, out)
+		assert.Nil(t, rejectPacket, "ICMP type %d should not generate a reject packet", icmpType)
+	}
+
+	// ICMP non-error types should still generate reject packets
+	icmpNonErrorTypes := []byte{0, 8, 13, 14}
+	for _, icmpType := range icmpNonErrorTypes {
+		h := ipv4.Header{
+			Len:      20,
+			Src:      net.IPv4(10, 0, 0, 1),
+			Dst:      net.IPv4(10, 0, 0, 2),
+			Protocol: 1, // ICMP
+		}
+
+		b, err := h.Marshal()
+		if err != nil {
+			t.Fatalf("h.Marshal: %v", err)
+		}
+		b = append(b, icmpType, 0, 0, 0, 0, 0, 0, 0)
+
+		rejectPacket := CreateRejectPacket(b, out)
+		assert.NotNil(t, rejectPacket, "ICMP type %d should generate a reject packet", icmpType)
+	}
+}
+
+func makeIPv6Packet(src, dst net.IP, nextHeader uint8, payload []byte) []byte {
+	b := make([]byte, ipv6.HeaderLen+len(payload))
+	b[0] = ipv6.Version << 4
+	binary.BigEndian.PutUint16(b[4:], uint16(len(payload)))
+	b[6] = nextHeader
+	b[7] = 64
+	copy(b[8:24], src.To16())
+	copy(b[24:40], dst.To16())
+	copy(b[ipv6.HeaderLen:], payload)
+	return b
+}
+
+func Test_CreateRejectPacketIPv6_ICMP(t *testing.T) {
+	src := net.ParseIP("fd00::1")
+	dst := net.ParseIP("fd00::2")
+
+	// Small UDP packet: entire original included in body
+	udpPayload := make([]byte, 20)
+	udpPayload[0] = 0x00 // src port high
+	udpPayload[1] = 0x50 // src port low (80)
+	udpPayload[2] = 0x01 // dst port high
+	udpPayload[3] = 0xBB // dst port low (443)
+	packet := makeIPv6Packet(src, dst, 17, udpPayload)
+
+	out := make([]byte, MaxRejectPacketSize)
+	rejectPacket := CreateRejectPacket(packet, out)
+	assert.NotNil(t, rejectPacket)
+
+	// Small packet fits entirely: 40 (ipv6 hdr) + 8 (icmpv6 hdr) + 60 (original)
+	expectedLen := ipv6.HeaderLen + 8 + len(packet)
+	assert.Len(t, rejectPacket, expectedLen)
+
+	// Verify version
+	assert.Equal(t, byte(ipv6.Version<<4), rejectPacket[0]&0xf0)
+	// Verify next header is ICMPv6 (58)
+	assert.Equal(t, byte(58), rejectPacket[6])
+	// Verify src/dst are swapped
+	assert.Equal(t, dst.To16(), net.IP(rejectPacket[8:24]))
+	assert.Equal(t, src.To16(), net.IP(rejectPacket[24:40]))
+	// Verify ICMPv6 type=1 (Dest Unreachable), code=1 (Administratively prohibited)
+	assert.Equal(t, byte(1), rejectPacket[ipv6.HeaderLen])
+	assert.Equal(t, byte(1), rejectPacket[ipv6.HeaderLen+1])
+	// Verify entire original packet is included in body
+	assert.Equal(t, packet, rejectPacket[ipv6.HeaderLen+8:])
+
+	// Large packet: body is truncated to 1000 bytes
+	largePkt := makeIPv6Packet(src, dst, 17, make([]byte, 1200))
+	rejectPacket = CreateRejectPacket(largePkt, out)
+	assert.NotNil(t, rejectPacket)
+	assert.Len(t, rejectPacket, ipv6.HeaderLen+8+1000)
+	assert.Equal(t, largePkt[:1000], rejectPacket[ipv6.HeaderLen+8:])
+}
+
+func Test_CreateRejectPacketIPv6_TCP(t *testing.T) {
+	src := net.ParseIP("fd00::1")
+	dst := net.ParseIP("fd00::2")
+
+	// TCP SYN packet (next header 6)
+	tcpPayload := make([]byte, 20)
+	tcpPayload[0] = 0x00                             // src port high
+	tcpPayload[1] = 0x50                             // src port low (80)
+	tcpPayload[2] = 0x01                             // dst port high
+	tcpPayload[3] = 0xBB                             // dst port low (443)
+	binary.BigEndian.PutUint32(tcpPayload[4:], 1000) // seq
+	binary.BigEndian.PutUint32(tcpPayload[8:], 0)    // ack seq
+	tcpPayload[12] = (20 >> 2) << 4                  // data offset
+	tcpPayload[13] = 0b00000010                      // SYN flag
+
+	packet := makeIPv6Packet(src, dst, 6, tcpPayload)
+
+	out := make([]byte, MaxRejectPacketSize)
+	rejectPacket := CreateRejectPacket(packet, out)
+	assert.NotNil(t, rejectPacket)
+
+	// Expected: 40 (ipv6 hdr) + 20 (tcp RST)
+	expectedLen := ipv6.HeaderLen + 20
+	assert.Len(t, rejectPacket, expectedLen)
+
+	// Verify version
+	assert.Equal(t, byte(ipv6.Version<<4), rejectPacket[0]&0xf0)
+	// Verify next header is TCP (6)
+	assert.Equal(t, byte(6), rejectPacket[6])
+	// Verify src/dst are swapped
+	assert.Equal(t, dst.To16(), net.IP(rejectPacket[8:24]))
+	assert.Equal(t, src.To16(), net.IP(rejectPacket[24:40]))
+	// Verify ports are swapped
+	tcpOut := rejectPacket[ipv6.HeaderLen:]
+	assert.Equal(t, uint16(443), binary.BigEndian.Uint16(tcpOut[0:2]))
+	assert.Equal(t, uint16(80), binary.BigEndian.Uint16(tcpOut[2:4]))
+	// RST+ACK flags (since input was SYN without ACK)
+	assert.Equal(t, byte(0b00010100), tcpOut[13])
+	// ack_seq = original seq (1000) + SYN (1) + FIN (0) + segment data (0)
+	assert.Equal(t, uint32(1001), binary.BigEndian.Uint32(tcpOut[8:]))
+}
+
+func Test_CreateRejectPacketIPv6_TCPWithACK(t *testing.T) {
+	src := net.ParseIP("fd00::1")
+	dst := net.ParseIP("fd00::2")
+
+	// TCP packet with ACK set
+	tcpPayload := make([]byte, 20)
+	tcpPayload[0] = 0x00
+	tcpPayload[1] = 0x50
+	tcpPayload[2] = 0x01
+	tcpPayload[3] = 0xBB
+	binary.BigEndian.PutUint32(tcpPayload[4:], 1000) // seq
+	binary.BigEndian.PutUint32(tcpPayload[8:], 2000) // ack seq
+	tcpPayload[12] = (20 >> 2) << 4                  // data offset
+	tcpPayload[13] = 0b00010000                      // ACK flag
+
+	packet := makeIPv6Packet(src, dst, 6, tcpPayload)
+
+	out := make([]byte, MaxRejectPacketSize)
+	rejectPacket := CreateRejectPacket(packet, out)
+	assert.NotNil(t, rejectPacket)
+
+	tcpOut := rejectPacket[ipv6.HeaderLen:]
+	// RST only (no ACK) since input had ACK
+	assert.Equal(t, byte(0b00000100), tcpOut[13])
+	// seq = original ack_seq
+	assert.Equal(t, uint32(2000), binary.BigEndian.Uint32(tcpOut[4:]))
+}
+
+func Test_CreateRejectPacketIPv6_NoICMPError(t *testing.T) {
+	src := net.ParseIP("fd00::1")
+	dst := net.ParseIP("fd00::2")
+	out := make([]byte, MaxRejectPacketSize)
+
+	// ICMPv6 error types (1-4) should not generate reject packets
+	for icmpType := byte(1); icmpType <= 4; icmpType++ {
+		payload := make([]byte, 8)
+		payload[0] = icmpType
+		packet := makeIPv6Packet(src, dst, 58, payload)
+
+		rejectPacket := CreateRejectPacket(packet, out)
+		assert.Nil(t, rejectPacket, "ICMPv6 type %d should not generate a reject packet", icmpType)
+	}
+
+	// ICMPv6 non-error types should still generate reject packets
+	nonErrorTypes := []byte{128, 129, 133, 134}
+	for _, icmpType := range nonErrorTypes {
+		payload := make([]byte, 8)
+		payload[0] = icmpType
+		packet := makeIPv6Packet(src, dst, 58, payload)
+
+		rejectPacket := CreateRejectPacket(packet, out)
+		assert.NotNil(t, rejectPacket, "ICMPv6 type %d should generate a reject packet", icmpType)
+	}
+}
+
+func Test_CreateRejectPacketIPv6_TooShort(t *testing.T) {
+	// Packet too short to be valid IPv6
+	out := make([]byte, MaxRejectPacketSize)
+	assert.Nil(t, CreateRejectPacket([]byte{0x60}, out))
+	assert.Nil(t, CreateRejectPacket(make([]byte, 39), out))
+}
+
+func Test_CreateRejectPacketIPv6_ExtensionHeaders(t *testing.T) {
+	src := net.ParseIP("fd00::1")
+	dst := net.ParseIP("fd00::2")
+
+	// IPv6 + Hop-by-Hop extension header + TCP
+	hopByHop := []byte{
+		6,    // next header: TCP
+		0,    // length (8 bytes total)
+		0, 0, // padding
+		0, 0, 0, 0,
+	}
+	tcpPayload := make([]byte, 20)
+	tcpPayload[0] = 0x00
+	tcpPayload[1] = 0x50
+	tcpPayload[2] = 0x01
+	tcpPayload[3] = 0xBB
+	binary.BigEndian.PutUint32(tcpPayload[4:], 1000)
+	binary.BigEndian.PutUint32(tcpPayload[8:], 2000)
+	tcpPayload[12] = (20 >> 2) << 4
+	tcpPayload[13] = 0b00010000 // ACK
+
+	payload := append(hopByHop, tcpPayload...)
+	packet := makeIPv6Packet(src, dst, 0, payload) // next header 0 = Hop-by-Hop
+
+	out := make([]byte, MaxRejectPacketSize)
+	rejectPacket := CreateRejectPacket(packet, out)
+	assert.NotNil(t, rejectPacket)
+
+	// Should produce TCP RST
+	expectedLen := ipv6.HeaderLen + 20
+	assert.Len(t, rejectPacket, expectedLen)
+	assert.Equal(t, byte(6), rejectPacket[6]) // next header is TCP
+	tcpOut := rejectPacket[ipv6.HeaderLen:]
+	assert.Equal(t, byte(0b00000100), tcpOut[13]) // RST only
+}
+
+func TestCreateICMPEchoResponse_IPv4(t *testing.T) {
+	// Build a simple IPv4 ICMP Echo Request
+	packet := make([]byte, 28)
+	packet[0] = 0x45                                   // version 4, IHL 5
+	binary.BigEndian.PutUint16(packet[2:], uint16(28)) // total length
+	packet[8] = 64                                     // TTL
+	packet[9] = 1                                      // protocol ICMP
+	copy(packet[12:16], net.IPv4(10, 0, 0, 1).To4())   // src
+	copy(packet[16:20], net.IPv4(10, 0, 0, 2).To4())   // dst
+	packet[20] = 8                                     // ICMP Echo Request
+
+	out := make([]byte, len(packet))
+	result := CreateICMPEchoResponse(packet, out)
+	assert.NotNil(t, result)
+	assert.Equal(t, byte(0x45), result[0])
+	// src/dst swapped
+	assert.Equal(t, net.IPv4(10, 0, 0, 2).To4(), net.IP(result[12:16]))
+	assert.Equal(t, net.IPv4(10, 0, 0, 1).To4(), net.IP(result[16:20]))
+	// ICMP Echo Reply
+	assert.Equal(t, byte(0), result[20])
+}
+
+func TestCreateICMPEchoResponse_IPv6(t *testing.T) {
+	src := net.ParseIP("fd00::1").To16()
+	dst := net.ParseIP("fd00::2").To16()
+
+	// Build an IPv6 ICMPv6 Echo Request packet
+	// IPv6 header (40 bytes) + ICMPv6 (8 bytes)
+	packet := make([]byte, 48)
+	packet[0] = 0x60        // version 6
+	payloadLen := uint16(8) // ICMPv6 header only
+	binary.BigEndian.PutUint16(packet[4:], payloadLen)
+	packet[6] = 58           // Next Header: ICMPv6
+	packet[7] = 64           // Hop Limit
+	copy(packet[8:24], src)  // src address
+	copy(packet[24:40], dst) // dst address
+
+	// ICMPv6 Echo Request
+	icmp := packet[40:]
+	icmp[0] = 128                           // type: Echo Request
+	icmp[1] = 0                             // code
+	binary.BigEndian.PutUint16(icmp[4:], 1) // identifier
+	binary.BigEndian.PutUint16(icmp[6:], 1) // sequence number
+
+	// Compute correct checksum for the request
+	csum := ipv6PseudoheaderChecksum(src, dst, 58, uint32(payloadLen))
+	binary.BigEndian.PutUint16(icmp[2:], tcpipChecksum(icmp, csum))
+
+	out := make([]byte, len(packet))
+	result := CreateICMPEchoResponse(packet, out)
+	assert.NotNil(t, result)
+
+	// Version should still be 6
+	assert.Equal(t, byte(6), result[0]>>4)
+	// src/dst swapped
+	assert.Equal(t, dst, net.IP(result[8:24]))
+	assert.Equal(t, src, net.IP(result[24:40]))
+	// ICMPv6 Echo Reply type
+	assert.Equal(t, byte(129), result[40])
+
+	// Verify checksum is valid (tcpipChecksum returns 0 when data+checksum is correct)
+	respIcmp := result[40:]
+	verifyCsum := ipv6PseudoheaderChecksum(result[8:24], result[24:40], 58, uint32(payloadLen))
+	assert.Equal(t, uint16(0), tcpipChecksum(respIcmp, verifyCsum))
+}
+
+func TestCreateICMPEchoResponse_IPv6_NotEchoRequest(t *testing.T) {
+	src := net.ParseIP("fd00::1").To16()
+	dst := net.ParseIP("fd00::2").To16()
+
+	packet := make([]byte, 48)
+	packet[0] = 0x60
+	binary.BigEndian.PutUint16(packet[4:], 8)
+	packet[6] = 58
+	packet[7] = 64
+	copy(packet[8:24], src)
+	copy(packet[24:40], dst)
+
+	// ICMPv6 type 1 (Destination Unreachable) - not Echo Request
+	packet[40] = 1
+
+	out := make([]byte, len(packet))
+	result := CreateICMPEchoResponse(packet, out)
+	assert.Nil(t, result)
+}
+
+func TestCreateICMPEchoResponse_IPv6_NotICMPv6(t *testing.T) {
+	src := net.ParseIP("fd00::1").To16()
+	dst := net.ParseIP("fd00::2").To16()
+
+	packet := make([]byte, 48)
+	packet[0] = 0x60
+	binary.BigEndian.PutUint16(packet[4:], 8)
+	packet[6] = 6 // TCP, not ICMPv6
+	packet[7] = 64
+	copy(packet[8:24], src)
+	copy(packet[24:40], dst)
+
+	out := make([]byte, len(packet))
+	result := CreateICMPEchoResponse(packet, out)
+	assert.Nil(t, result)
+}

lighthouse.go

@@ -272,16 +272,18 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 	//NOTE: many things will get much simpler when we combine static_host_map and lighthouse.hosts in config
 	if initial || c.HasChanged("static_host_map") || c.HasChanged("static_map.cadence") || c.HasChanged("static_map.network") || c.HasChanged("static_map.lookup_timeout") {
 		// Clean up. Entries still in the static_host_map will be re-built.
-		// Entries no longer present must have their (possible) background DNS goroutines stopped.
-		if existingStaticList := lh.staticList.Load(); existingStaticList != nil {
+		ourselves := lh.myVpnNetworks[0].Addr()
+		oldStaticList := lh.staticList.Load()
+		if oldStaticList != nil {
 			lh.RLock()
-			for staticVpnAddr := range *existingStaticList {
+			for staticVpnAddr := range *oldStaticList {
 				if am, ok := lh.addrMap[staticVpnAddr]; ok && am != nil {
-					am.hr.Cancel()
+					am.ResetForOwner(ourselves)
 				}
 			}
 			lh.RUnlock()
 		}
+
 		// Build a new list based on current config.
 		staticList := make(map[netip.Addr]struct{})
 		err := lh.loadStaticMap(c, staticList)
@@ -289,6 +291,21 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 			return err
 		}
 
+		// For entries removed from static_host_map, stop the DNS goroutine and drop the cached addrs.
+		// All addrs must come from the lighthouses now that it's no longer a static host.
+		if oldStaticList != nil {
+			lh.RLock()
+			for staticVpnAddr := range *oldStaticList {
+				if _, stillStatic := staticList[staticVpnAddr]; stillStatic {
+					continue
+				}
+				if am, ok := lh.addrMap[staticVpnAddr]; ok && am != nil {
+					am.ClearHostnameResults()
+				}
+			}
+			lh.RUnlock()
+		}
+
 		lh.staticList.Store(&staticList)
 		if !initial {
 			if c.HasChanged("static_host_map") {

lighthouse_test.go

@@ -303,6 +303,132 @@ func TestLighthouse_reload(t *testing.T) {
 	require.NoError(t, err)
 }
 
+// TestLighthouse_reloadStaticHostMap verifies that reloading static_host_map applies the new
+// config rather than appending to it. See issue #718.
+func TestLighthouse_reloadStaticHostMap(t *testing.T) {
+	l := test.NewLogger()
+	c := config.NewC(l)
+	c.Settings["lighthouse"] = map[string]any{"am_lighthouse": true}
+	c.Settings["listen"] = map[string]any{"port": 4242}
+	c.Settings["static_host_map"] = map[string]any{
+		"10.128.0.2": []any{"1.1.1.1:4242"},
+	}
+
+	myVpnNet := netip.MustParsePrefix("10.128.0.1/24")
+	nt := new(bart.Lite)
+	nt.Insert(myVpnNet)
+	cs := &CertState{
+		myVpnNetworks:      []netip.Prefix{myVpnNet},
+		myVpnNetworksTable: nt,
+	}
+
+	lh, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
+	require.NoError(t, err)
+
+	staticHost := netip.MustParseAddr("10.128.0.2")
+	otherHost := netip.MustParseAddr("10.128.0.3")
+
+	// Capture the RemoteList pointer up front; an in-flight handshake would hold the same one
+	// on hostinfo.remotes, so it must reflect every reload below.
+	pinned := lh.Query(staticHost)
+	require.NotNil(t, pinned)
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("1.1.1.1:4242")}, pinned.CopyAddrs([]netip.Prefix{}))
+
+	// Replace the remote address. The new address should be the only entry.
+	nc := map[string]any{
+		"static_host_map": map[string]any{
+			"10.128.0.2": []any{"2.2.2.2:4242"},
+		},
+	}
+	rc, err := yaml.Marshal(nc)
+	require.NoError(t, err)
+	require.NoError(t, c.ReloadConfigString(string(rc)))
+
+	rl := lh.Query(staticHost)
+	require.NotNil(t, rl)
+	assert.Same(t, pinned, rl, "RemoteList pointer must stay stable so in-flight handshakes pick up the change")
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("2.2.2.2:4242")}, rl.CopyAddrs([]netip.Prefix{}))
+
+	// Reload back to the original IP. Mirrors the round-trip in issue #718 step 6-8 where
+	// the buggy reload produced [1.1.1.1, 2.2.2.2, 1.1.1.1] instead of [1.1.1.1].
+	nc = map[string]any{
+		"static_host_map": map[string]any{
+			"10.128.0.2": []any{"1.1.1.1:4242"},
+		},
+	}
+	rc, err = yaml.Marshal(nc)
+	require.NoError(t, err)
+	require.NoError(t, c.ReloadConfigString(string(rc)))
+
+	rl = lh.Query(staticHost)
+	require.NotNil(t, rl)
+	assert.Same(t, pinned, rl)
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("1.1.1.1:4242")}, rl.CopyAddrs([]netip.Prefix{}))
+
+	// Reload with the same config. An unchanged entry must not duplicate.
+	require.NoError(t, c.ReloadConfigString(string(rc)))
+
+	rl = lh.Query(staticHost)
+	require.NotNil(t, rl)
+	assert.Same(t, pinned, rl)
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("1.1.1.1:4242")}, rl.CopyAddrs([]netip.Prefix{}))
+
+	// Switch back to 2.2.2.2 so the rest of the test continues against a known address.
+	nc = map[string]any{
+		"static_host_map": map[string]any{
+			"10.128.0.2": []any{"2.2.2.2:4242"},
+		},
+	}
+	rc, err = yaml.Marshal(nc)
+	require.NoError(t, err)
+	require.NoError(t, c.ReloadConfigString(string(rc)))
+
+	// Add a second host alongside the first. Both should be present, neither duplicated.
+	nc = map[string]any{
+		"static_host_map": map[string]any{
+			"10.128.0.2": []any{"2.2.2.2:4242"},
+			"10.128.0.3": []any{"3.3.3.3:4242"},
+		},
+	}
+	rc, err = yaml.Marshal(nc)
+	require.NoError(t, err)
+	require.NoError(t, c.ReloadConfigString(string(rc)))
+
+	rl = lh.Query(staticHost)
+	require.NotNil(t, rl)
+	assert.Same(t, pinned, rl, "adding a sibling entry must not displace the existing RemoteList")
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("2.2.2.2:4242")}, rl.CopyAddrs([]netip.Prefix{}))
+
+	rl = lh.Query(otherHost)
+	require.NotNil(t, rl)
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("3.3.3.3:4242")}, rl.CopyAddrs([]netip.Prefix{}))
+
+	// Drop the first host entirely. The vpnAddr is no longer marked static, our owner
+	// contribution is cleared, but the addrMap entry stays in place so non-static cache
+	// data (from lighthouse queries) on the same RemoteList isn't lost. In-flight handshakes
+	// that already had the pointer see an empty address list rather than retrying stale ones.
+	nc = map[string]any{
+		"static_host_map": map[string]any{
+			"10.128.0.3": []any{"3.3.3.3:4242"},
+		},
+	}
+	rc, err = yaml.Marshal(nc)
+	require.NoError(t, err)
+	require.NoError(t, c.ReloadConfigString(string(rc)))
+
+	_, isStatic := lh.GetStaticHostList()[staticHost]
+	assert.False(t, isStatic)
+
+	rl = lh.Query(staticHost)
+	require.NotNil(t, rl)
+	assert.Same(t, pinned, rl)
+	assert.Empty(t, rl.CopyAddrs([]netip.Prefix{}))
+
+	rl = lh.Query(otherHost)
+	require.NotNil(t, rl)
+	assert.Equal(t, []netip.AddrPort{netip.MustParseAddrPort("3.3.3.3:4242")}, rl.CopyAddrs([]netip.Prefix{}))
+}
+
 func newLHHostRequest(fromAddr netip.AddrPort, myVpnIp, queryVpnIp netip.Addr, lhh *LightHouseHandler) testLhReply {
 	req := &NebulaMeta{
 		Type:    NebulaMeta_HostQuery,

main.go

@@ -194,7 +194,7 @@ func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, dev
 	handshakeManager := NewHandshakeManager(l, hostMap, lightHouse, udpConns[0], handshakeConfig)
 	lightHouse.handshakeTrigger = handshakeManager.trigger
 
-	ds, err := newDnsServerFromConfig(ctx, l, pki.getCertState(), hostMap, c)
+	ds, err := newDnsServerFromConfig(ctx, l, pki, hostMap, c)
 	if err != nil {
 		l.Warn("Failed to start DNS responder", "error", err)
 	}

outside.go

@@ -181,6 +181,13 @@ func (f *Interface) handleOutsideRelayPacket(hostinfo *HostInfo, via ViaSender,
 	if err != nil {
 		return
 	}
+	// Advance the replay window now that the frame is authenticated
+	if !hostinfo.ConnectionState.window.Update(f.l, h.MessageCounter) {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
+			hostinfo.logger(f.l).Debug("dropping out of window relay packet", "header", h)
+		}
+		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.
@@ -194,8 +201,7 @@ func (f *Interface) handleOutsideRelayPacket(hostinfo *HostInfo, via ViaSender,
 		// 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,
+			"relayRemoteIndex", h.RemoteIndex,
 		)
 		return
 	}
@@ -218,8 +224,8 @@ func (f *Interface) handleOutsideRelayPacket(hostinfo *HostInfo, via ViaSender,
 		if err != nil {
 			hostinfo.logger(f.l).Info("Failed to find target host info by ip",
 				"relayTo", relay.PeerAddr,
+				"relayFrom", hostinfo.vpnAddrs[0],
 				"error", err,
-				"hostinfo.vpnAddrs", hostinfo.vpnAddrs,
 			)
 			return
 		}

relay_manager.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"log/slog"
 	"net/netip"
+	"slices"
 	"sync/atomic"
 
 	"github.com/slackhq/nebula/cert"
@@ -57,14 +58,25 @@ func (rm *relayManager) GetUseRelays() bool {
 // For each candidate relay it either kicks off a handshake to the relay, sends a CreateRelayRequest, retransmits
 // one that may have been lost, or, once the relay is Established, forwards the in-progress
 // stage 0 handshake packet for vpnIp through it.
-func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *HostInfo, stage0 []byte) {
+func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hh *HandshakeHostInfo, stage0 []byte) {
+	hostinfo := hh.hostinfo
 	if !rm.GetUseRelays() || len(hostinfo.remotes.relays) == 0 {
+		hh.lastRelays = nil
 		return
 	}
 
-	hostinfo.logger(rm.l).Info("Attempt to relay through hosts", "relays", hostinfo.remotes.relays)
+	relays := hostinfo.remotes.relays
+	listLevel := slog.LevelDebug
+	prior := hh.lastRelays
+	if !slices.Equal(relays, prior) {
+		listLevel = slog.LevelInfo
+		hh.lastRelays = slices.Clone(relays)
+	}
+	hl := hostinfo.logger(rm.l)
+	hl.Log(context.Background(), listLevel, "Attempt to relay through hosts", "relays", relays)
+
 	// Send a RelayRequest to all known Relay IP's
-	for _, relay := range hostinfo.remotes.relays {
+	for _, relay := range relays {
 		// Don't relay through the host I'm trying to connect to
 		if relay == vpnIp {
 			continue
@@ -75,12 +87,19 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 			continue
 		}
 
+		// Each relay's per-attempt log fires at Info on the first time we hit it and Debug after that.
+		level := slog.LevelInfo
+		if slices.Contains(prior, relay) {
+			level = slog.LevelDebug
+		}
+
 		relayHostInfo := rm.hostmap.QueryVpnAddr(relay)
 		if relayHostInfo == nil || !relayHostInfo.remote.IsValid() {
-			hostinfo.logger(rm.l).Info("Establish tunnel to relay target", "relay", relay.String())
+			hl.Log(context.Background(), level, "Establish tunnel to relay target", "relay", relay.String())
 			f.Handshake(relay)
 			continue
 		}
+
 		// Check the relay HostInfo to see if we already established a relay through
 		existingRelay, ok := relayHostInfo.relayState.QueryRelayForByIp(vpnIp)
 		if !ok {
@@ -88,7 +107,7 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 			if relayHostInfo.remote.IsValid() {
 				idx, err := AddRelay(rm.l, relayHostInfo, rm.hostmap, vpnIp, nil, TerminalType, Requested)
 				if err != nil {
-					hostinfo.logger(rm.l).Info("Failed to add relay to hostmap", "relay", relay.String(), "error", err)
+					hl.Info("Failed to add relay to hostmap", "relay", relay.String(), "error", err)
 				}
 
 				m := NebulaControl{
@@ -99,12 +118,12 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 				switch relayHostInfo.GetCert().Certificate.Version() {
 				case cert.Version1:
 					if !f.myVpnAddrs[0].Is4() {
-						hostinfo.logger(rm.l).Error("can not establish v1 relay with a v6 network because the relay is not running a current nebula version")
+						hl.Error("can not establish v1 relay with a v6 network because the relay is not running a current nebula version")
 						continue
 					}
 
 					if !vpnIp.Is4() {
-						hostinfo.logger(rm.l).Error("can not establish v1 relay with a v6 remote network because the relay is not running a current nebula version")
+						hl.Error("can not establish v1 relay with a v6 remote network because the relay is not running a current nebula version")
 						continue
 					}
 
@@ -116,16 +135,16 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 					m.RelayFromAddr = netAddrToProtoAddr(f.myVpnAddrs[0])
 					m.RelayToAddr = netAddrToProtoAddr(vpnIp)
 				default:
-					hostinfo.logger(rm.l).Error("Unknown certificate version found while creating relay")
+					hl.Error("Unknown certificate version found while creating relay")
 					continue
 				}
 
 				msg, err := m.Marshal()
 				if err != nil {
-					hostinfo.logger(rm.l).Error("Failed to marshal Control message to create relay", "error", err)
+					hl.Error("Failed to marshal Control message to create relay", "error", err)
 				} else {
 					f.SendMessageToHostInfo(header.Control, 0, relayHostInfo, msg, make([]byte, 12), make([]byte, mtu))
-					rm.l.Info("send CreateRelayRequest",
+					rm.l.Log(context.Background(), level, "send CreateRelayRequest",
 						"relayFrom", f.myVpnAddrs[0],
 						"relayTo", vpnIp,
 						"initiatorRelayIndex", idx,
@@ -138,14 +157,14 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 
 		switch existingRelay.State {
 		case Established:
-			hostinfo.logger(rm.l).Info("Send handshake via relay", "relay", relay.String())
+			hl.Log(context.Background(), level, "Send handshake via relay", "relay", relay.String())
 			f.SendVia(relayHostInfo, existingRelay, stage0, make([]byte, 12), make([]byte, mtu), false)
 		case Disestablished:
 			// Mark this relay as 'requested'
 			relayHostInfo.relayState.UpdateRelayForByIpState(vpnIp, Requested)
 			fallthrough
 		case Requested:
-			hostinfo.logger(rm.l).Info("Re-send CreateRelay request", "relay", relay.String())
+			hl.Log(context.Background(), level, "Re-send CreateRelay request", "relay", relay.String())
 			// Re-send the CreateRelay request, in case the previous one was lost.
 			m := NebulaControl{
 				Type:                NebulaControl_CreateRelayRequest,
@@ -155,12 +174,12 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 			switch relayHostInfo.GetCert().Certificate.Version() {
 			case cert.Version1:
 				if !f.myVpnAddrs[0].Is4() {
-					hostinfo.logger(rm.l).Error("can not establish v1 relay with a v6 network because the relay is not running a current nebula version")
+					hl.Error("can not establish v1 relay with a v6 network because the relay is not running a current nebula version")
 					continue
 				}
 
 				if !vpnIp.Is4() {
-					hostinfo.logger(rm.l).Error("can not establish v1 relay with a v6 remote network because the relay is not running a current nebula version")
+					hl.Error("can not establish v1 relay with a v6 remote network because the relay is not running a current nebula version")
 					continue
 				}
 
@@ -172,16 +191,16 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 				m.RelayFromAddr = netAddrToProtoAddr(f.myVpnAddrs[0])
 				m.RelayToAddr = netAddrToProtoAddr(vpnIp)
 			default:
-				hostinfo.logger(rm.l).Error("Unknown certificate version found while creating relay")
+				hl.Error("Unknown certificate version found while creating relay")
 				continue
 			}
 			msg, err := m.Marshal()
 			if err != nil {
-				hostinfo.logger(rm.l).Error("Failed to marshal Control message to create relay", "error", err)
+				hl.Error("Failed to marshal Control message to create relay", "error", err)
 			} else {
 				// This must send over the hostinfo, not over hm.Hosts[ip]
 				f.SendMessageToHostInfo(header.Control, 0, relayHostInfo, msg, make([]byte, 12), make([]byte, mtu))
-				rm.l.Info("send CreateRelayRequest",
+				rm.l.Log(context.Background(), level, "send CreateRelayRequest",
 					"relayFrom", f.myVpnAddrs[0],
 					"relayTo", vpnIp,
 					"initiatorRelayIndex", existingRelay.LocalIndex,
@@ -192,7 +211,7 @@ func (rm *relayManager) StartRelays(f *Interface, vpnIp netip.Addr, hostinfo *Ho
 			// PeerRequested only occurs in Forwarding relays, not Terminal relays, and this is a Terminal relay case.
 			fallthrough
 		default:
-			hostinfo.logger(rm.l).Error("Relay unexpected state",
+			hl.Error("Relay unexpected state",
 				"vpnIp", vpnIp,
 				"state", existingRelay.State,
 				"relay", relay,
@@ -299,17 +318,16 @@ func (rm *relayManager) HandleControlMsg(h *HostInfo, d []byte, f *Interface) {
 }
 
 func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f *Interface, m *NebulaControl) {
+	relayFrom := protoAddrToNetAddr(m.RelayFromAddr)
+	relayTo := protoAddrToNetAddr(m.RelayToAddr)
 	rm.l.Info("handleCreateRelayResponse",
-		"relayFrom", protoAddrToNetAddr(m.RelayFromAddr),
-		"relayTo", protoAddrToNetAddr(m.RelayToAddr),
+		"relayFrom", relayFrom,
+		"relayTo", relayTo,
 		"initiatorRelayIndex", m.InitiatorRelayIndex,
 		"responderRelayIndex", m.ResponderRelayIndex,
 		"vpnAddrs", h.vpnAddrs,
 	)
 
-	target := m.RelayToAddr
-	targetAddr := protoAddrToNetAddr(target)
-
 	relay, err := rm.EstablishRelay(h, m)
 	if err != nil {
 		rm.l.Error("Failed to update relay for relayTo", "error", err)
@@ -325,7 +343,7 @@ func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f
 		rm.l.Error("Can't find a HostInfo for peer", "relayTo", relay.PeerAddr)
 		return
 	}
-	peerRelay, ok := peerHostInfo.relayState.QueryRelayForByIp(targetAddr)
+	peerRelay, ok := peerHostInfo.relayState.QueryRelayForByIp(relayTo)
 	if !ok {
 		rm.l.Error("peerRelay does not have Relay state for relayTo", "relayTo", peerHostInfo.vpnAddrs[0])
 		return
@@ -335,19 +353,19 @@ func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f
 		// I initiated the request to this peer, but haven't heard back from the peer yet. I must wait for this peer
 		// to respond to complete the connection.
 	case PeerRequested, Disestablished, Established:
-		peerHostInfo.relayState.UpdateRelayForByIpState(targetAddr, Established)
+		peerHostInfo.relayState.UpdateRelayForByIpState(relayTo, Established)
 		resp := NebulaControl{
 			Type:                NebulaControl_CreateRelayResponse,
 			ResponderRelayIndex: peerRelay.LocalIndex,
 			InitiatorRelayIndex: peerRelay.RemoteIndex,
 		}
 
+		peer := peerHostInfo.vpnAddrs[0]
 		if v == cert.Version1 {
-			peer := peerHostInfo.vpnAddrs[0]
 			if !peer.Is4() {
 				rm.l.Error("Refusing to CreateRelayResponse for a v1 relay with an ipv6 address",
 					"relayFrom", peer,
-					"relayTo", target,
+					"relayTo", relayTo,
 					"initiatorRelayIndex", resp.InitiatorRelayIndex,
 					"responderRelayIndex", resp.ResponderRelayIndex,
 					"vpnAddrs", peerHostInfo.vpnAddrs,
@@ -357,26 +375,26 @@ func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f
 
 			b := peer.As4()
 			resp.OldRelayFromAddr = binary.BigEndian.Uint32(b[:])
-			b = targetAddr.As4()
+			b = relayTo.As4()
 			resp.OldRelayToAddr = binary.BigEndian.Uint32(b[:])
 		} else {
-			resp.RelayFromAddr = netAddrToProtoAddr(peerHostInfo.vpnAddrs[0])
-			resp.RelayToAddr = target
+			resp.RelayFromAddr = netAddrToProtoAddr(peer)
+			resp.RelayToAddr = m.RelayToAddr
 		}
 
 		msg, err := resp.Marshal()
 		if err != nil {
 			rm.l.Error("relayManager Failed to marshal Control CreateRelayResponse message to create relay", "error", err)
-		} else {
-			f.SendMessageToHostInfo(header.Control, 0, peerHostInfo, msg, make([]byte, 12), make([]byte, mtu))
-			rm.l.Info("send CreateRelayResponse",
-				"relayFrom", resp.RelayFromAddr,
-				"relayTo", resp.RelayToAddr,
-				"initiatorRelayIndex", resp.InitiatorRelayIndex,
-				"responderRelayIndex", resp.ResponderRelayIndex,
-				"vpnAddrs", peerHostInfo.vpnAddrs,
-			)
+			return
 		}
+		f.SendMessageToHostInfo(header.Control, 0, peerHostInfo, msg, make([]byte, 12), make([]byte, mtu))
+		rm.l.Info("send CreateRelayResponse",
+			"relayFrom", peer,
+			"relayTo", relayTo,
+			"initiatorRelayIndex", resp.InitiatorRelayIndex,
+			"responderRelayIndex", resp.ResponderRelayIndex,
+			"vpnAddrs", peerHostInfo.vpnAddrs,
+		)
 	}
 }
 

relay_manager_test.go

@@ -0,0 +1,97 @@
+package nebula
+
+import (
+	"bytes"
+	"log/slog"
+	"net/netip"
+	"testing"
+
+	"github.com/gaissmai/bart"
+	"github.com/slackhq/nebula/test"
+	"github.com/stretchr/testify/assert"
+)
+
+// TestStartRelaysLogDedupe verifies that repeated attempts with the same relay set drop the log
+// chatter to Debug, mirroring how the normal handshake retry loop quiets down once it's already
+// announced its targets.
+func TestStartRelaysLogDedupe(t *testing.T) {
+	vpnIp := netip.MustParseAddr("100.64.99.4")
+	otherRelay := netip.MustParseAddr("100.64.99.5")
+
+	newHH := func() *HandshakeHostInfo {
+		// Use the target's own vpnIp as the "relay" so the loop body skips it without
+		// touching any sender-side state. That isolates the test to the level-selection
+		// behavior of the top-level "Attempt to relay through hosts" log.
+		hostinfo := &HostInfo{
+			vpnAddrs:     []netip.Addr{vpnIp},
+			localIndexId: 1,
+			remotes:      NewRemoteList([]netip.Addr{vpnIp}, nil),
+		}
+		hostinfo.remotes.relays = []netip.Addr{vpnIp}
+		return &HandshakeHostInfo{hostinfo: hostinfo}
+	}
+
+	// Park any extra relay addresses we'll introduce mid-test in myVpnAddrsTable so the loop
+	// body always skips before touching f.Handshake (which would need a real handshakeManager).
+	addrTable := new(bart.Lite)
+	addrTable.Insert(netip.PrefixFrom(otherRelay, otherRelay.BitLen()))
+	f := &Interface{myVpnAddrsTable: addrTable}
+
+	newRM := func(buf *bytes.Buffer) *relayManager {
+		l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelDebug)
+		rm := &relayManager{l: l, hostmap: newHostMap(l)}
+		rm.useRelays.Store(true)
+		return rm
+	}
+
+	const msg = `msg="Attempt to relay through hosts"`
+
+	t.Run("first attempt logs at Info", func(t *testing.T) {
+		var buf bytes.Buffer
+		rm := newRM(&buf)
+		hh := newHH()
+		rm.StartRelays(f, vpnIp, hh, nil)
+		assert.Equal(t, []netip.Addr{vpnIp}, hh.lastRelays, "lastRelays should record the relay set we just attempted")
+		assert.Contains(t, buf.String(), "level=INFO "+msg, "expected Info level on first attempt")
+	})
+
+	t.Run("repeat attempt with same relays drops to Debug", func(t *testing.T) {
+		var buf bytes.Buffer
+		rm := newRM(&buf)
+		hh := newHH()
+		rm.StartRelays(f, vpnIp, hh, nil)
+		first := append([]netip.Addr(nil), hh.lastRelays...)
+		buf.Reset()
+		rm.StartRelays(f, vpnIp, hh, nil)
+		assert.Equal(t, first, hh.lastRelays)
+		assert.Contains(t, buf.String(), "level=DEBUG "+msg, "expected Debug level on identical retry")
+		assert.NotContains(t, buf.String(), "level=INFO "+msg, "Info should not fire on identical retry")
+	})
+
+	t.Run("changed relay list bumps back to Info", func(t *testing.T) {
+		var buf bytes.Buffer
+		rm := newRM(&buf)
+		hh := newHH()
+		rm.StartRelays(f, vpnIp, hh, nil)
+		buf.Reset()
+
+		// The lighthouse handed us a new set this round.
+		hh.hostinfo.remotes.relays = []netip.Addr{vpnIp, otherRelay}
+
+		rm.StartRelays(f, vpnIp, hh, nil)
+		assert.Equal(t, []netip.Addr{vpnIp, otherRelay}, hh.lastRelays)
+		assert.Contains(t, buf.String(), "level=INFO "+msg, "expected Info when the relay list changes")
+	})
+
+	t.Run("disabled relays clears lastRelays and emits no Attempt log", func(t *testing.T) {
+		var buf bytes.Buffer
+		rm := newRM(&buf)
+		rm.useRelays.Store(false)
+		hh := newHH()
+		hh.lastRelays = []netip.Addr{vpnIp}
+
+		rm.StartRelays(f, vpnIp, hh, nil)
+		assert.Nil(t, hh.lastRelays, "with relays disabled lastRelays should be cleared")
+		assert.NotContains(t, buf.String(), msg, "should not log when we shortcut out")
+	})
+}

remote_list.go

@@ -239,6 +239,31 @@ func (r *RemoteList) unlockedSetHostnamesResults(hr *hostnamesResults) {
 	r.hr = hr
 }
 
+// ResetForOwner zeros the reported address slices for the given owner and marks the addrs list dirty.
+// Any pending hostname resolution will be canceled.
+func (r *RemoteList) ResetForOwner(ownerVpnAddr netip.Addr) {
+	r.Lock()
+	defer r.Unlock()
+	r.hr.Cancel()
+	if c, ok := r.cache[ownerVpnAddr]; ok {
+		if c.v4 != nil {
+			c.v4.reported = c.v4.reported[:0]
+		}
+		if c.v6 != nil {
+			c.v6.reported = c.v6.reported[:0]
+		}
+	}
+	r.shouldRebuild = true
+}
+
+// ClearHostnameResults cancels the in-flight DNS resolver goroutine (if any) and drops the resolved IP cache.
+func (r *RemoteList) ClearHostnameResults() {
+	r.Lock()
+	defer r.Unlock()
+	r.unlockedSetHostnamesResults(nil)
+	r.shouldRebuild = true
+}
+
 // Len locks and reports the size of the deduplicated address list
 // The deduplication work may need to occur here, so you must pass preferredRanges
 func (r *RemoteList) Len(preferredRanges []netip.Prefix) int {

remote_list_test.go

@@ -6,8 +6,22 @@ import (
 	"testing"
 
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
+// trackedHostnameResults builds a *hostnamesResults with a known cancel function and a
+// pre-populated ips map so tests can assert cancellation and verify previously-resolved
+// IPs survive a cancel without spinning up a real DNS resolver.
+func trackedHostnameResults(cancelFn func(), addrs ...string) *hostnamesResults {
+	hr := &hostnamesResults{cancelFn: cancelFn}
+	ips := map[netip.AddrPort]struct{}{}
+	for _, a := range addrs {
+		ips[netip.MustParseAddrPort(a)] = struct{}{}
+	}
+	hr.ips.Store(&ips)
+	return hr
+}
+
 func TestRemoteList_Rebuild(t *testing.T) {
 	rl := NewRemoteList([]netip.Addr{netip.MustParseAddr("0.0.0.0")}, nil)
 	rl.unlockedSetV4(
@@ -112,6 +126,81 @@ func TestRemoteList_Rebuild(t *testing.T) {
 	assert.Equal(t, "172.31.0.1:10101", rl.addrs[9].String())
 }
 
+func TestRemoteList_ResetForOwner(t *testing.T) {
+	ourselves := netip.MustParseAddr("10.0.0.1")
+	otherOwner := netip.MustParseAddr("10.0.0.2")
+	vpnAddr := netip.MustParseAddr("10.0.0.99")
+
+	rl := NewRemoteList([]netip.Addr{vpnAddr}, nil)
+	rl.unlockedSetV4(ourselves, vpnAddr,
+		[]*V4AddrPort{newIp4AndPortFromString("1.1.1.1:4242")},
+		func(netip.Addr, *V4AddrPort) bool { return true },
+	)
+	rl.unlockedSetV6(ourselves, vpnAddr,
+		[]*V6AddrPort{newIp6AndPortFromString("[1::1]:4242")},
+		func(netip.Addr, *V6AddrPort) bool { return true },
+	)
+	rl.unlockedSetV4(otherOwner, vpnAddr,
+		[]*V4AddrPort{newIp4AndPortFromString("2.2.2.2:4242")},
+		func(netip.Addr, *V4AddrPort) bool { return true },
+	)
+
+	canceled := 0
+	hr := trackedHostnameResults(func() { canceled++ }, "3.3.3.3:4242")
+	rl.Lock()
+	rl.unlockedSetHostnamesResults(hr)
+	rl.Unlock()
+
+	rl.ResetForOwner(ourselves)
+
+	rl.RLock()
+	defer rl.RUnlock()
+	assert.Empty(t, rl.cache[ourselves].v4.reported, "our v4 reported should be cleared")
+	assert.Empty(t, rl.cache[ourselves].v6.reported, "our v6 reported should be cleared")
+	assert.Len(t, rl.cache[otherOwner].v4.reported, 1, "other owner's contribution must be preserved")
+	assert.Equal(t, "2.2.2.2:4242", protoV4AddrPortToNetAddrPort(rl.cache[otherOwner].v4.reported[0]).String())
+	assert.Equal(t, 1, canceled, "DNS resolution goroutine should be canceled")
+	assert.Same(t, hr, rl.hr, "hostnamesResults must be preserved so DNS-resolved IPs keep feeding addrs until replaced")
+	assert.NotEmpty(t, rl.hr.GetAddrs(), "previously-resolved IPs should still be readable after cancel")
+	assert.True(t, rl.shouldRebuild, "shouldRebuild must be set so the next Rebuild recomputes addrs")
+}
+
+func TestRemoteList_ResetForOwner_NoEntry(t *testing.T) {
+	// An owner with no cache entry must not panic; shouldRebuild is still set and any
+	// existing hostnamesResults is canceled.
+	rl := NewRemoteList([]netip.Addr{netip.MustParseAddr("10.0.0.99")}, nil)
+	canceled := 0
+	rl.Lock()
+	rl.unlockedSetHostnamesResults(trackedHostnameResults(func() { canceled++ }, "3.3.3.3:4242"))
+	rl.Unlock()
+
+	rl.ResetForOwner(netip.MustParseAddr("10.0.0.1"))
+
+	rl.RLock()
+	defer rl.RUnlock()
+	assert.Equal(t, 1, canceled)
+	assert.True(t, rl.shouldRebuild)
+}
+
+func TestRemoteList_ClearHostnameResults(t *testing.T) {
+	rl := NewRemoteList([]netip.Addr{netip.MustParseAddr("10.0.0.99")}, nil)
+
+	canceled := 0
+	hr := trackedHostnameResults(func() { canceled++ }, "3.3.3.3:4242")
+	rl.Lock()
+	rl.unlockedSetHostnamesResults(hr)
+	rl.Unlock()
+	require.NotEmpty(t, hr.GetAddrs(), "hostnamesResults should have its fastrack IPs populated")
+
+	rl.ClearHostnameResults()
+
+	rl.RLock()
+	defer rl.RUnlock()
+	assert.Equal(t, 1, canceled, "DNS resolution goroutine should be canceled")
+	assert.Nil(t, rl.hr, "hostnamesResults should be dropped")
+	assert.True(t, rl.shouldRebuild)
+}
+
 func BenchmarkFullRebuild(b *testing.B) {
 	rl := NewRemoteList([]netip.Addr{netip.MustParseAddr("0.0.0.0")}, nil)
 	rl.unlockedSetV4(

stats.go

@@ -331,7 +331,7 @@ func loadStatsConfig(c *config.C) (statsConfig, error) {
 	}
 
 	cfg.interval = c.GetDuration("stats.interval", 0)
-	if cfg.interval == 0 {
+	if cfg.interval <= 0 {
 		return cfg, fmt.Errorf("stats.interval was an invalid duration: %s", c.GetString("stats.interval", ""))
 	}
 

udp/udp_darwin.go

@@ -175,8 +175,8 @@ func (u *StdConn) ListenOut(r EncReader) error {
 			if errors.Is(err, net.ErrClosed) {
 				return err
 			}
-
 			u.l.Error("unexpected udp socket receive error", "error", err)
+			continue
 		}
 
 		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])