Merge branch 'main' into profile-ownership

.github/workflows/release.yml

@@ -23,8 +23,50 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
+  freebsd_port_trigger:
+    name: "FreeBSD Port / Trigger"
+    runs-on: ubuntu-22.04
+    outputs:
+      run: ${{ steps.decide.outputs.run }}
+      version: ${{ steps.decide.outputs.version }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+        with:
+          persist-credentials: false
+
+      - name: Decide whether the port job runs
+        id: decide
+        env:
+          BASE_SHA: ${{ github.event.pull_request.base.sha }}
+        run: |
+          set -euo pipefail
+          run=false
+          case "$GITHUB_EVENT_NAME" in
+            push)
+              if [[ "$GITHUB_REF" == "refs/heads/main" ]]; then
+                run=true
+              elif [[ "$GITHUB_REF" == refs/tags/* && "$GITHUB_REF_NAME" =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
+                # The port tracks stable releases only, so a canary or rc tag has nothing to submit.
+                run=true
+                echo "version=${GITHUB_REF_NAME#v}" >> "$GITHUB_OUTPUT"
+              fi
+              ;;
+            pull_request)
+              git fetch --depth=1 origin "$BASE_SHA"
+              if git diff --name-only "$BASE_SHA" HEAD | \
+                  grep -qE '^(release_files/freebsd-port-.*\.sh|\.github/workflows/release\.yml)$'; then
+                run=true
+              fi
+              ;;
+          esac
+          echo "run=$run" >> "$GITHUB_OUTPUT"
+          echo "FreeBSD port job runs: $run"
+
   release_freebsd_port:
     name: "FreeBSD Port / Build & Test"
+    needs: freebsd_port_trigger
+    if: needs.freebsd_port_trigger.outputs.run == 'true'
     runs-on: ubuntu-22.04
     steps:
       - name: Checkout
@@ -33,18 +75,19 @@ jobs:
           persist-credentials: false
 
       - name: Generate FreeBSD port diff
-        run: bash -x release_files/freebsd-port-diff.sh
-
-      - name: Generate FreeBSD port issue body
-        run: bash -x release_files/freebsd-port-issue-body.sh
+        env:
+          TARGET_VERSION: ${{ needs.freebsd_port_trigger.outputs.version }}
+        run: bash -x release_files/freebsd-port-diff.sh "$TARGET_VERSION"
 
       - name: Check if diff was generated
         id: check_diff
         run: |
-          if ls netbird-*.diff 1> /dev/null 2>&1; then
-            echo "diff_exists=true" >> $GITHUB_OUTPUT
+          shopt -s nullglob
+          diffs=(netbird-*.diff)
+          if [[ ${#diffs[@]} -gt 0 ]]; then
+            echo "diff_exists=true" >> "$GITHUB_OUTPUT"
           else
-            echo "diff_exists=false" >> $GITHUB_OUTPUT
+            echo "diff_exists=false" >> "$GITHUB_OUTPUT"
             echo "No diff file generated (port may already be up to date)"
           fi
 
@@ -52,10 +95,18 @@ jobs:
         if: steps.check_diff.outputs.diff_exists == 'true'
         id: version
         run: |
-          VERSION=$(ls netbird-*.diff | sed 's/netbird-\(.*\)\.diff/\1/')
-          echo "version=$VERSION" >> $GITHUB_OUTPUT
+          diffs=(netbird-*.diff)
+          VERSION="${diffs[0]#netbird-}"
+          VERSION="${VERSION%.diff}"
+          echo "version=$VERSION" >> "$GITHUB_OUTPUT"
           echo "Generated files for version: $VERSION"
-          cat netbird-*.diff
+          cat "${diffs[0]}"
+
+      - name: Generate FreeBSD port issue body
+        if: steps.check_diff.outputs.diff_exists == 'true'
+        env:
+          TARGET_VERSION: ${{ steps.version.outputs.version }}
+        run: bash -x release_files/freebsd-port-issue-body.sh "" "$TARGET_VERSION"
 
       - name: Read Go version from go.mod
         id: goversion
@@ -366,6 +417,8 @@ jobs:
           path: |
             dist/netbird_linux**
             dist/netbird-pkcs11_linux**
+            dist/netbird-deb-32bit_linux**
+            dist/netbird-rpm-*_linux**
           retention-days: 7
       - name: upload windows packages
         id: upload_windows_packages

.goreleaser.yaml

@@ -40,8 +40,9 @@ builds:
     tags:
       - load_wgnt_from_rsrc
 
-  # deb and rpm packages target glibc distributions, so they carry the PKCS#11 store, which
-  # links libc. Tarballs and the Alpine-based images keep the static build above.
+  # deb and rpm packages target glibc distributions, so on amd64 and arm64, the only
+  # architectures with a PKCS#11 driver, they carry the PKCS#11 store, which links libc.
+  # Tarballs, the Alpine-based images and the other architectures keep the static build.
   - id: netbird-pkcs11
     dir: client
     binary: netbird
@@ -49,10 +50,8 @@ builds:
     goos:
       - linux
     goarch:
-      - arm
       - amd64
       - arm64
-      - 386
     ldflags:
       - -s -w -X github.com/netbirdio/netbird/version.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}} -X main.builtBy=goreleaser
     mod_timestamp: "{{ .CommitTimestamp }}"
@@ -60,8 +59,24 @@ builds:
       - load_wgnt_from_rsrc
       - pkcs11
 
-  # Single-arch builds: nfpm provides is not templated, so the RPM splits per arch. They
-  # carry the PKCS#11 store like the deb build above.
+  # The 32-bit deb architectures, which have no PKCS#11 driver.
+  - id: netbird-deb-32bit
+    dir: client
+    binary: netbird
+    env: [CGO_ENABLED=0]
+    goos:
+      - linux
+    goarch:
+      - arm
+      - 386
+    ldflags:
+      - -s -w -X github.com/netbirdio/netbird/version.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}} -X main.builtBy=goreleaser
+    mod_timestamp: "{{ .CommitTimestamp }}"
+    tags:
+      - load_wgnt_from_rsrc
+
+  # Single-arch builds: nfpm provides is not templated, so the RPM splits per arch. The
+  # ones with a PKCS#11 driver carry the store like the deb build above.
   - &netbird_rpm_build
     id: netbird-rpm-amd64
     dir: client
@@ -83,10 +98,14 @@ builds:
   - <<: *netbird_rpm_build
     id: netbird-rpm-arm
     goarch: [arm]
+    tags:
+      - load_wgnt_from_rsrc
 
   - <<: *netbird_rpm_build
     id: netbird-rpm-386
     goarch: [386]
+    tags:
+      - load_wgnt_from_rsrc
 
   - id: netbird-static
     dir: client
@@ -323,6 +342,7 @@ nfpms:
     bindir: /usr/bin
     builds:
       - netbird-pkcs11
+      - netbird-deb-32bit
     formats:
       - deb
     scripts:

client/cmd/up.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"net"
 	"net/netip"
 	"runtime"
 	"strings"
@@ -26,16 +25,11 @@ import (
 	"github.com/netbirdio/netbird/client/proto"
 	"github.com/netbirdio/netbird/client/server"
 	"github.com/netbirdio/netbird/client/system"
+	"github.com/netbirdio/netbird/shared/externalipmap"
 	"github.com/netbirdio/netbird/shared/management/domain"
 	"github.com/netbirdio/netbird/util"
 )
 
-const (
-	invalidInputType int = iota
-	ipInputType
-	interfaceInputType
-)
-
 const (
 	dnsLabelsFlag = "extra-dns-labels"
 
@@ -106,7 +100,7 @@ func upFunc(cmd *cobra.Command, args []string) error {
 		return fmt.Errorf("failed initializing log %v", err)
 	}
 
-	err = validateNATExternalIPs(natExternalIPs)
+	err = externalipmap.Validate(natExternalIPs)
 	if err != nil {
 		return err
 	}
@@ -846,36 +840,6 @@ func setupLoginRequest(providedSetupKey string, customDNSAddressConverted []byte
 	return &loginRequest, nil
 }
 
-func validateNATExternalIPs(list []string) error {
-	for _, element := range list {
-		if element == "" {
-			return fmt.Errorf("empty string is not a valid input for %s", externalIPMapFlag)
-		}
-
-		subElements := strings.Split(element, "/")
-		if len(subElements) > 2 {
-			return fmt.Errorf("%s is not a valid input for %s. it should be formatted as \"String\" or \"String/String\"", element, externalIPMapFlag)
-		}
-
-		if len(subElements) == 1 && !isValidIP(subElements[0]) {
-			return fmt.Errorf("%s is not a valid input for %s. it should be formatted as \"IP\" or \"IP/IP\", or \"IP/Interface Name\"", element, externalIPMapFlag)
-		}
-
-		last := 0
-		for _, singleElement := range subElements {
-			inputType, err := validateElement(singleElement)
-			if err != nil {
-				return fmt.Errorf("%s is not a valid input for %s. it should be an IP string or a network name", singleElement, externalIPMapFlag)
-			}
-			if last == interfaceInputType && inputType == interfaceInputType {
-				return fmt.Errorf("%s is not a valid input for %s. it should not contain two interface names", element, externalIPMapFlag)
-			}
-			last = inputType
-		}
-	}
-	return nil
-}
-
 func parseInterfaceName(name string) error {
 	if runtime.GOOS != "darwin" {
 		return nil
@@ -888,39 +852,6 @@ func parseInterfaceName(name string) error {
 	return fmt.Errorf("invalid interface name %s. Please use the prefix utun followed by a number on MacOS. e.g., utun1 or utun199", name)
 }
 
-func validateElement(element string) (int, error) {
-	if isValidIP(element) {
-		return ipInputType, nil
-	}
-	validIface, err := isValidInterface(element)
-	if err != nil {
-		return invalidInputType, fmt.Errorf("unable to validate the network interface name, error: %s", err)
-	}
-
-	if validIface {
-		return interfaceInputType, nil
-	}
-
-	return interfaceInputType, fmt.Errorf("invalid IP or network interface name not found")
-}
-
-func isValidIP(ip string) bool {
-	return net.ParseIP(ip) != nil
-}
-
-func isValidInterface(name string) (bool, error) {
-	netInterfaces, err := net.Interfaces()
-	if err != nil {
-		return false, err
-	}
-	for _, iface := range netInterfaces {
-		if iface.Name == name {
-			return true, nil
-		}
-	}
-	return false, nil
-}
-
 func parseCustomDNSAddress(modified bool) ([]byte, error) {
 	var parsed []byte
 	if modified {

client/embed/embed.go

@@ -118,6 +118,9 @@ type Options struct {
 	MTU *uint16
 	// DNSLabels defines additional DNS labels configured in the peer.
 	DNSLabels []string
+	// NATExternalIPs maps external IPs to internal IPs or interfaces for ICE
+	// candidates, same format as the agent's --external-ip-map flag.
+	NATExternalIPs []string
 	// Performance configures the tunnel's buffer pool cap and batch size.
 	Performance Performance
 }
@@ -224,6 +227,7 @@ func New(opts Options) (*Client, error) {
 		WireguardPort:       opts.WireguardPort,
 		MTU:                 opts.MTU,
 		DNSLabels:           parsedLabels,
+		NATExternalIPs:      opts.NATExternalIPs,
 	}
 	if opts.ConfigPath != "" {
 		config, err = profilemanager.UpdateOrCreateConfig(input)

client/iface/bind/ice_bind_test.go

@@ -2,17 +2,22 @@ package bind
 
 import (
 	"fmt"
+	"io"
 	"net"
 	"net/netip"
 	"sync"
+	"sync/atomic"
 	"testing"
 	"time"
 
+	"github.com/pion/transport/v3"
 	"github.com/pion/transport/v3/stdnet"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"golang.org/x/net/ipv4"
 	"golang.org/x/net/ipv6"
+	"golang.zx2c4.com/wireguard/device"
+	"golang.zx2c4.com/wireguard/tun/netstack"
 
 	"github.com/netbirdio/netbird/client/iface/wgaddr"
 )
@@ -326,3 +331,86 @@ func listenUDP(t *testing.T, network, addr string) *net.UDPConn {
 	require.NoError(t, err)
 	return conn
 }
+
+type countingNet struct {
+	transport.Net
+	calls  atomic.Int64
+	ifaces []*transport.Interface
+	before func()
+}
+
+func (n *countingNet) Interfaces() ([]*transport.Interface, error) {
+	n.calls.Add(1)
+	if n.before != nil {
+		n.before()
+	}
+	return n.ifaces, nil
+}
+
+func TestICEBind_OpenDoesNotDiscoverInterfaces(t *testing.T) {
+	n := newCountingNet()
+	iceBind := NewICEBind(n, testAddress(), 1280)
+
+	for i := 0; i < 2; i++ {
+		_, port, err := iceBind.Open(0)
+		require.NoError(t, err)
+		require.NotZero(t, port, "the bind listens on a port")
+		assert.Equal(t, int64(0), n.calls.Load(), "opening the bind must not enumerate interfaces, attempt %d", i+1)
+		require.NoError(t, iceBind.Close())
+	}
+
+	_, _, err := iceBind.Open(0)
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = iceBind.Close() })
+
+	mux, err := iceBind.GetICEMux()
+	require.NoError(t, err)
+	assert.NotEmpty(t, mux.SingleSocketUDPMux.GetListenAddresses(), "listen addresses come from the discovered interfaces")
+	assert.Equal(t, int64(1), n.calls.Load(), "interfaces are discovered when the listen addresses are asked for")
+}
+
+func TestICEBind_DeviceUpDoesNotWaitOnInterfaceDiscovery(t *testing.T) {
+	tunDev, _, err := netstack.CreateNetTUN([]netip.Addr{netip.MustParseAddr("100.64.0.1")}, nil, 1280)
+	require.NoError(t, err)
+
+	var dev *device.Device
+	n := newCountingNet()
+	n.before = func() {
+		_ = dev.IpcGetOperation(io.Discard)
+	}
+	iceBind := NewICEBind(n, testAddress(), 1280)
+	dev = device.NewDevice(tunDev, iceBind, device.NewLogger(device.LogLevelSilent, ""))
+
+	upDone := make(chan error, 1)
+	go func() {
+		upDone <- dev.Up()
+	}()
+
+	select {
+	case err := <-upDone:
+		require.NoError(t, err)
+	case <-time.After(5 * time.Second):
+		t.Fatal("device Up did not return: interface discovery ran under the device lock and waited on the device itself")
+	}
+	t.Cleanup(dev.Close)
+
+	assert.Equal(t, int64(0), n.calls.Load(), "bringing the device up must not enumerate interfaces")
+
+	mux, err := iceBind.GetICEMux()
+	require.NoError(t, err)
+	assert.NotEmpty(t, mux.SingleSocketUDPMux.GetListenAddresses(), "discovery that talks to the device succeeds once the device is up")
+	assert.Equal(t, int64(1), n.calls.Load(), "interfaces are discovered outside the device lock")
+}
+
+func newCountingNet() *countingNet {
+	ifc := transport.NewInterface(net.Interface{Index: 1, MTU: 1500, Name: "eth0", Flags: net.FlagUp | net.FlagMulticast})
+	ifc.AddAddress(&net.IPNet{IP: net.IPv4(192, 168, 1, 10), Mask: net.CIDRMask(24, 32)})
+	return &countingNet{ifaces: []*transport.Interface{ifc}}
+}
+
+func testAddress() wgaddr.Address {
+	return wgaddr.Address{
+		IP:      netip.MustParseAddr("100.64.0.1"),
+		Network: netip.MustParsePrefix("100.64.0.0/10"),
+	}
+}

client/iface/udpmux/mux.go

@@ -47,6 +47,7 @@ type SingleSocketUDPMux struct {
 
 	// for UDP connection listen at unspecified address
 	localAddrsForUnspecified []net.Addr
+	unspecified              bool
 }
 
 const maxAddrSize = 512
@@ -157,6 +158,9 @@ func NewSingleSocketUDPMux(params Params) *SingleSocketUDPMux {
 	if params.Logger == nil {
 		params.Logger = getLogger()
 	}
+	if params.Net == nil {
+		params.Net = stdnet.NewNet(context.Background(), nil, nil)
+	}
 
 	mux := &SingleSocketUDPMux{
 		addressMap:       map[string][]*udpMuxedConn{},
@@ -165,6 +169,7 @@ func NewSingleSocketUDPMux(params Params) *SingleSocketUDPMux {
 		connsIPv6:        make(map[string]*udpMuxedConn),
 		candidateConnMap: make(map[string]*udpMuxedConn),
 		closedChan:       make(chan struct{}, 1),
+		unspecified:      isUnspecified(params.UDPConn.LocalAddr()),
 		pool: &sync.Pool{
 			New: func() interface{} {
 				// big enough buffer to fit both packet and address
@@ -173,7 +178,6 @@ func NewSingleSocketUDPMux(params Params) *SingleSocketUDPMux {
 		},
 	}
 
-	mux.updateLocalAddresses()
 	return mux
 }
 
@@ -199,10 +203,6 @@ func (m *SingleSocketUDPMux) updateLocalAddresses() {
 			m.params.Logger.Errorf("LocalAddr expected IPV4 or IPV6, got %T", m.params.UDPConn.LocalAddr())
 		}
 		if len(networks) > 0 {
-			if m.params.Net == nil {
-				m.params.Net = stdnet.NewNet(context.Background(), nil, nil)
-			}
-
 			ips, err := localInterfaces(m.params.Net, m.params.InterfaceFilter, nil, networks, true)
 			if err == nil {
 				for _, ip := range ips {
@@ -241,10 +241,7 @@ func (m *SingleSocketUDPMux) GetListenAddresses() []net.Addr {
 // creates the connection if an existing one can't be found
 func (m *SingleSocketUDPMux) GetConn(ufrag string, addr net.Addr, candidateID string) (net.PacketConn, error) {
 	// don't check addr for mux using unspecified address
-	m.mu.Lock()
-	lenLocalAddrs := len(m.localAddrsForUnspecified)
-	m.mu.Unlock()
-	if lenLocalAddrs == 0 && m.params.UDPConn.LocalAddr().String() != addr.String() {
+	if !m.unspecified && m.params.UDPConn.LocalAddr().String() != addr.String() {
 		return nil, fmt.Errorf("invalid address %s", addr.String())
 	}
 
@@ -508,3 +505,8 @@ func getLogger() logging.LeveledLogger {
 	//fac.Writer = log.StandardLogger().Writer()
 	return fac.NewLogger("ice")
 }
+
+func isUnspecified(addr net.Addr) bool {
+	udpAddr, ok := addr.(*net.UDPAddr)
+	return ok && udpAddr.IP.IsUnspecified()
+}

client/iface/udpmux/mux_test.go

@@ -0,0 +1,99 @@
+package udpmux
+
+import (
+	"fmt"
+	"net"
+	"sync"
+	"sync/atomic"
+	"testing"
+
+	"github.com/pion/transport/v3"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+type countingNet struct {
+	transport.Net
+	calls  atomic.Int64
+	ifaces []*transport.Interface
+}
+
+func (n *countingNet) Interfaces() ([]*transport.Interface, error) {
+	n.calls.Add(1)
+	return n.ifaces, nil
+}
+
+func TestNewSingleSocketUDPMux_DoesNotDiscoverInterfaces(t *testing.T) {
+	conn := listenUDP(t, net.IPv4zero)
+	n := &countingNet{ifaces: []*transport.Interface{testInterface("eth0", net.IPv4(192, 168, 1, 10))}}
+
+	mux := NewSingleSocketUDPMux(Params{UDPConn: conn, Net: n})
+	t.Cleanup(func() { _ = mux.Close() })
+
+	assert.Equal(t, int64(0), n.calls.Load(), "constructing the mux must not enumerate interfaces")
+
+	addrs := mux.GetListenAddresses()
+	assert.Equal(t, int64(1), n.calls.Load(), "listen addresses are discovered when they are asked for")
+
+	port := conn.LocalAddr().(*net.UDPAddr).Port
+	require.Len(t, addrs, 1, "one address per discovered interface IP")
+	assert.Equal(t, fmt.Sprintf("192.168.1.10:%d", port), addrs[0].String(), "the discovered IP is paired with the socket port")
+}
+
+func TestSingleSocketUDPMux_GetConnWithoutListenAddresses(t *testing.T) {
+	t.Run("unspecified socket accepts any address", func(t *testing.T) {
+		conn := listenUDP(t, net.IPv4zero)
+		n := &countingNet{}
+		mux := NewSingleSocketUDPMux(Params{UDPConn: conn, Net: n})
+		t.Cleanup(func() { _ = mux.Close() })
+
+		port := conn.LocalAddr().(*net.UDPAddr).Port
+		pc, err := mux.GetConn("ufrag", &net.UDPAddr{IP: net.IPv4(192, 168, 1, 10), Port: port}, "candidate")
+		require.NoError(t, err)
+		assert.NotNil(t, pc, "a connection is created for a local address of an unspecified socket")
+		assert.Equal(t, int64(0), n.calls.Load(), "GetConn must not depend on a prior discovery")
+	})
+
+	t.Run("specific socket rejects a foreign address", func(t *testing.T) {
+		conn := listenUDP(t, net.IPv4(127, 0, 0, 1))
+		mux := NewSingleSocketUDPMux(Params{UDPConn: conn})
+		t.Cleanup(func() { _ = mux.Close() })
+
+		_, err := mux.GetConn("ufrag", &net.UDPAddr{IP: net.IPv4(192, 168, 1, 10), Port: 1234}, "candidate")
+		assert.Error(t, err, "an address the socket is not bound to must be rejected")
+
+		pc, err := mux.GetConn("ufrag", conn.LocalAddr(), "candidate")
+		require.NoError(t, err)
+		assert.NotNil(t, pc, "the bound address is accepted")
+	})
+}
+
+func listenUDP(t *testing.T, ip net.IP) *net.UDPConn {
+	t.Helper()
+	conn, err := net.ListenUDP("udp4", &net.UDPAddr{IP: ip, Port: 0})
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = conn.Close() })
+	return conn
+}
+
+func testInterface(name string, ip net.IP) *transport.Interface {
+	ifc := transport.NewInterface(net.Interface{Index: 1, MTU: 1500, Name: name, Flags: net.FlagUp | net.FlagMulticast})
+	ifc.AddAddress(&net.IPNet{IP: ip, Mask: net.CIDRMask(24, 32)})
+	return ifc
+}
+
+func TestSingleSocketUDPMux_ConcurrentListenAddressesWithDefaultNet(t *testing.T) {
+	conn := listenUDP(t, net.IPv4zero)
+	mux := NewSingleSocketUDPMux(Params{UDPConn: conn})
+	t.Cleanup(func() { _ = mux.Close() })
+
+	var wg sync.WaitGroup
+	for i := 0; i < 8; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			assert.NotEmpty(t, mux.GetListenAddresses(), "every caller gets at least the socket address")
+		}()
+	}
+	wg.Wait()
+}

client/internal/certproof/collect.go

@@ -149,9 +149,19 @@ func prove(signer crypto.Signer, chain []*x509.Certificate, nonce, peerKey []byt
 }
 
 func chainFingerprint(chain []*x509.Certificate) [sha256.Size]byte {
-	buf := make([]byte, 0, len(chain)*sha256.Size)
+	ders := make([][]byte, 0, len(chain))
 	for _, cert := range chain {
-		certHash := sha256.Sum256(cert.Raw)
+		ders = append(ders, cert.Raw)
+	}
+	return chainDERFingerprint(ders)
+}
+
+// chainDERFingerprint identifies a chain of DER certificates by the hashes of all of them,
+// in order.
+func chainDERFingerprint(chain [][]byte) [sha256.Size]byte {
+	buf := make([]byte, 0, len(chain)*sha256.Size)
+	for _, der := range chain {
+		certHash := sha256.Sum256(der)
 		buf = append(buf, certHash[:]...)
 	}
 	return sha256.Sum256(buf)

client/internal/certproof/collect_darwin.go

@@ -60,7 +60,7 @@ func UserContext(cfg Config) string {
 	if os.Geteuid() != 0 || cfg.OwnerUnknown {
 		return ""
 	}
-	user, ok := CurrentConsoleUser()
+	user, ok := currentConsoleUser()
 	if !ok || !user.isOwner(cfg.ProfileOwner) {
 		return ""
 	}
@@ -72,7 +72,7 @@ func UserContext(cfg Config) string {
 // per-session securityd, so the helper has to enter their Mach bootstrap namespace,
 // which is what launchctl asuser does.
 func collectAsConsoleUser(ctx context.Context, owner string, challenges []*proto.CertificateChallenge, peerKey []byte) ([]certposture.Proof, error) {
-	user, ok := CurrentConsoleUser()
+	user, ok := currentConsoleUser()
 	if !ok {
 		return nil, nil
 	}
@@ -126,12 +126,12 @@ func helperStore() Store {
 	home, err := os.UserHomeDir()
 	if err != nil {
 		log.Debugf("certificate posture: no home directory, searching the default keychain list: %v", err)
-		return NewKeychainStore()
+		return newKeychainStore()
 	}
 	login := filepath.Join(home, "Library", "Keychains", "login.keychain-db")
 	if _, err := os.Stat(login); err != nil {
 		// Keychains created before macOS 10.12 keep the old file name.
 		login = filepath.Join(home, "Library", "Keychains", "login.keychain")
 	}
-	return NewKeychainStore(login)
+	return newKeychainStore(login)
 }

client/internal/certproof/collect_test.go

@@ -51,7 +51,7 @@ func TestCollect_ProvesOneMatchingCertificatePerChallenge(t *testing.T) {
 		challenge("not a pem"),
 	}
 
-	proofs := Collect(context.Background(), NewFileStore(dir), checks, peerKey)
+	proofs := Collect(context.Background(), newFileStore(dir), checks, peerKey)
 
 	require.Len(t, proofs, 2)
 	var subjects []string
@@ -74,10 +74,10 @@ func TestCollect_NothingToProve(t *testing.T) {
 		store  Store
 		checks []*proto.Checks
 	}{
-		{"no checks", NewFileStore(dir), nil},
-		{"files only", NewFileStore(dir), []*proto.Checks{{Files: []string{"/bin/x"}}}},
-		{"challenge without nonce", NewFileStore(dir), []*proto.Checks{{CertificateChallenge: &proto.CertificateChallenge{CaCertificates: []string{ca.PEM}}}}},
-		{"missing store dir", NewFileStore(filepath.Join(dir, "missing")), []*proto.Checks{{CertificateChallenge: &proto.CertificateChallenge{Nonce: make([]byte, certposture.NonceSize), CaCertificates: []string{ca.PEM}}}}},
+		{"no checks", newFileStore(dir), nil},
+		{"files only", newFileStore(dir), []*proto.Checks{{Files: []string{"/bin/x"}}}},
+		{"challenge without nonce", newFileStore(dir), []*proto.Checks{{CertificateChallenge: &proto.CertificateChallenge{CaCertificates: []string{ca.PEM}}}}},
+		{"missing store dir", newFileStore(filepath.Join(dir, "missing")), []*proto.Checks{{CertificateChallenge: &proto.CertificateChallenge{Nonce: make([]byte, certposture.NonceSize), CaCertificates: []string{ca.PEM}}}}},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
@@ -95,7 +95,7 @@ func TestFileStore_ChainWithIntermediate(t *testing.T) {
 	dir := storeDir(t)
 	writeFile(t, dir, "device.pem", certtest.CertPEM(leaf)+certtest.CertPEM(intermediate.Cert)+certtest.KeyPEM(t, key))
 
-	candidates, err := NewFileStore(dir).Candidates(context.Background())
+	candidates, err := newFileStore(dir).Candidates(context.Background())
 	require.NoError(t, err)
 	require.Len(t, candidates, 1)
 	require.Len(t, candidates[0].Chain, 2)
@@ -124,7 +124,7 @@ func TestCollectChallenges_RefusesMalformedInput(t *testing.T) {
 	dir := storeDir(t)
 	key := certtest.ECDSAKey(t)
 	writeFile(t, dir, "device.pem", certtest.CertPEM(ca.Issue(t, key, "device"))+certtest.KeyPEM(t, key))
-	store := NewFileStore(dir)
+	store := newFileStore(dir)
 	nonce := certposture.NewChallenger([]byte("secret")).Nonce(peerKey, time.Now())
 
 	tests := []struct {
@@ -159,7 +159,7 @@ func TestFileStore_SkipsKeyOfAnotherCertificate(t *testing.T) {
 	good := ca.Issue(t, goodKey, "good")
 	writeFile(t, dir, "b-good.pem", certtest.CertPEM(good)+certtest.KeyPEM(t, goodKey))
 
-	candidates, err := NewFileStore(dir).Candidates(context.Background())
+	candidates, err := newFileStore(dir).Candidates(context.Background())
 	require.NoError(t, err)
 	require.Len(t, candidates, 1, "only the certificate whose key matches is a candidate")
 	assert.True(t, good.Equal(candidates[0].Chain[0]), "the matching pair is kept")
@@ -167,7 +167,7 @@ func TestFileStore_SkipsKeyOfAnotherCertificate(t *testing.T) {
 	challenger := certposture.NewChallenger([]byte("secret"))
 	now := time.Now()
 	nonce := challenger.Nonce(peerKey, now)
-	proofs := CollectChallenges(context.Background(), NewFileStore(dir), []*proto.CertificateChallenge{{Nonce: nonce, CaCertificates: []string{ca.PEM}}}, peerKey)
+	proofs := CollectChallenges(context.Background(), newFileStore(dir), []*proto.CertificateChallenge{{Nonce: nonce, CaCertificates: []string{ca.PEM}}}, peerKey)
 	require.Len(t, proofs, 1)
 	_, err = challenger.Verify(proofs[0], peerKey, now)
 	assert.NoError(t, err, "the proof sent is one management accepts")

client/internal/certproof/collect_windows.go

@@ -51,7 +51,7 @@ func UserContext(cfg Config) string {
 	if !runningAsLocalSystem() || cfg.OwnerUnknown {
 		return ""
 	}
-	user, ok := CurrentDesktopUser(cfg.ProfileOwner)
+	user, ok := currentDesktopUser(cfg.ProfileOwner)
 	if !ok {
 		return ""
 	}
@@ -62,14 +62,14 @@ func UserContext(cfg Config) string {
 // helperStore is the store the helper reads. It runs as the signed-in user, so it wants
 // that user's store rather than the machine store the service already read.
 func helperStore() Store {
-	return NewUserStore()
+	return newUserStore()
 }
 
 // collectAsDesktopUser runs the helper inside the interactive session of the signed-in
 // user. Unlike a keychain on macOS, a Windows service can assume a user identity
 // directly, so the session token goes straight into the child process.
 func collectAsDesktopUser(ctx context.Context, owner string, challenges []*proto.CertificateChallenge, peerKey []byte) ([]certposture.Proof, error) {
-	user, ok := CurrentDesktopUser(owner)
+	user, ok := currentDesktopUser(owner)
 	if !ok {
 		return nil, nil
 	}

client/internal/certproof/consoleuser_darwin.go

@@ -35,10 +35,10 @@ type ConsoleUser struct {
 	GID  uint32
 }
 
-// CurrentConsoleUser reports the user sitting at the desktop. The second return value is
+// currentConsoleUser reports the user sitting at the desktop. The second return value is
 // false when nobody is: at the login window macOS either reports no console user at all
 // or attributes the session to root, and neither has a login keychain to offer.
-func CurrentConsoleUser() (ConsoleUser, bool) {
+func currentConsoleUser() (ConsoleUser, bool) {
 	if err := loadConsoleUser(); err != nil {
 		log.Debugf("console user lookup unavailable: %v", err)
 		return ConsoleUser{}, false

client/internal/certproof/consoleuser_darwin_test.go

@@ -27,10 +27,10 @@ func TestConsoleUser_OnlyADesktopSessionCanBeValidated(t *testing.T) {
 	}
 }
 
-// CurrentConsoleUser runs against the real SystemConfiguration framework. A machine with
+// currentConsoleUser runs against the real SystemConfiguration framework. A machine with
 // a desktop open must report a non-root user; a headless runner must report none.
 func TestCurrentConsoleUser_AgreesWithItself(t *testing.T) {
-	user, ok := CurrentConsoleUser()
+	user, ok := currentConsoleUser()
 	if !ok {
 		t.Log("no console user, running headless")
 		return

client/internal/certproof/desktopuser_windows.go

@@ -13,12 +13,8 @@ const (
 	noActiveSession = 0xFFFFFFFF
 	servicesSession = 0
 
-	// wtsCurrentServer is WTS_CURRENT_SERVER_HANDLE; wtsActive and wtsDisconnected are
-	// WTSActive and WTSDisconnected of WTS_CONNECTSTATE_CLASS. None is exported by
-	// x/sys/windows.
+	// wtsCurrentServer is WTS_CURRENT_SERVER_HANDLE, which x/sys/windows does not export.
 	wtsCurrentServer = windows.Handle(0)
-	wtsActive        = 0
-	wtsDisconnected  = 4
 )
 
 // DesktopUser is an interactive session and the account signed into it. The user's
@@ -37,7 +33,7 @@ func (u DesktopUser) Close() {
 	}
 }
 
-// CurrentDesktopUser returns a token for the session whose certificate store should be
+// currentDesktopUser returns a token for the session whose certificate store should be
 // asked: a session of owner, the account the active profile belongs to, with the
 // physical console preferred over remote sessions. With no owner only the console user
 // counts. Picking any signed-in user instead would let whoever else is logged in to a
@@ -46,7 +42,7 @@ func (u DesktopUser) Close() {
 //
 // Obtaining the token needs SE_TCB_NAME, which the LocalSystem service has and an
 // ordinary process does not.
-func CurrentDesktopUser(owner string) (DesktopUser, bool) {
+func currentDesktopUser(owner string) (DesktopUser, bool) {
 	console := windows.WTSGetActiveConsoleSessionId()
 	if owner == "" {
 		return consoleUser(console)
@@ -163,7 +159,7 @@ func userSessions(console uint32) ([]uint32, error) {
 	if console != noActiveSession && console != servicesSession {
 		sessions = append(sessions, console)
 	}
-	for _, state := range []uint32{wtsActive, wtsDisconnected} {
+	for _, state := range []uint32{windows.WTSActive, windows.WTSDisconnected} {
 		for _, session := range unsafe.Slice(info, count) {
 			if session.SessionID == servicesSession || session.SessionID == console || session.State != state {
 				continue

client/internal/certproof/desktopuser_windows_test.go

@@ -26,10 +26,10 @@ func TestSameAccountName(t *testing.T) {
 	}
 }
 
-// CurrentDesktopUser against the real session manager: an owner no session belongs to
+// currentDesktopUser against the real session manager: an owner no session belongs to
 // must never yield a session, whoever else is signed in.
 func TestCurrentDesktopUser_UnknownOwnerHasNoSession(t *testing.T) {
-	user, ok := CurrentDesktopUser(`NO-SUCH-DOMAIN\no-such-user-netbird`)
+	user, ok := currentDesktopUser(`NO-SUCH-DOMAIN\no-such-user-netbird`)
 	if ok {
 		user.Close()
 	}
@@ -44,14 +44,14 @@ func TestCurrentDesktopUser_FindsTheOwnersSession(t *testing.T) {
 	if owner == "" {
 		t.Skip("set NB_TEST_DESKTOP_OWNER to a signed-in account and run as LocalSystem")
 	}
-	user, ok := CurrentDesktopUser(owner)
+	user, ok := currentDesktopUser(owner)
 	require.True(t, ok, "the signed-in owner %s must have a session", owner)
 	defer user.Close()
 	t.Logf("owner %s resolved to session %d as %s", owner, user.Session, user.Name)
 	assert.True(t, sameAccountName(user.Name, owner), "the session found belongs to %s, not %s", owner, user.Name)
 	assert.NotZero(t, user.Session, "session 0 hosts services and never belongs to the owner")
 
-	console, consoleOK := CurrentDesktopUser("")
+	console, consoleOK := currentDesktopUser("")
 	if consoleOK {
 		defer console.Close()
 		t.Logf("without an owner the console user counts: session %d as %s", console.Session, console.Name)

client/internal/certproof/helper_spawn.go

@@ -3,12 +3,6 @@
 package certproof
 
 import (
-	"crypto/sha256"
-	"crypto/x509"
-
-	log "github.com/sirupsen/logrus"
-
-	"github.com/netbirdio/netbird/shared/management/certposture"
 	"github.com/netbirdio/netbird/shared/management/proto"
 )
 
@@ -22,42 +16,3 @@ func helperRequest(challenges []*proto.CertificateChallenge, peerKey []byte) Hel
 	}
 	return req
 }
-
-// mergeProofs appends the user session proofs to the device proofs, dropping a leaf that
-// both stores hold so the same certificate is proven once.
-func mergeProofs(device, user []certposture.Proof) []certposture.Proof {
-	if len(user) == 0 {
-		return device
-	}
-
-	seen := make(map[[sha256.Size]byte]struct{}, len(device))
-	for _, proof := range device {
-		if len(proof.Chain) > 0 {
-			seen[sha256.Sum256(proof.Chain[0])] = struct{}{}
-		}
-	}
-
-	merged := device
-	for _, proof := range user {
-		if len(proof.Chain) == 0 {
-			continue
-		}
-		fingerprint := sha256.Sum256(proof.Chain[0])
-		if _, done := seen[fingerprint]; done {
-			continue
-		}
-		seen[fingerprint] = struct{}{}
-		merged = append(merged, proof)
-		logUserProof(proof)
-	}
-	return merged
-}
-
-func logUserProof(proof certposture.Proof) {
-	leaf, err := x509.ParseCertificate(proof.Chain[0])
-	if err != nil {
-		log.Debugf("certificate posture: user proof carries an unparsable leaf: %v", err)
-		return
-	}
-	log.Debugf("certificate posture: signed-in user proved %q issued by %q", leaf.Subject, leaf.Issuer)
-}

client/internal/certproof/helper_spawn_test.go

@@ -3,42 +3,15 @@
 package certproof
 
 import (
-	"crypto/sha256"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
-	"github.com/netbirdio/netbird/shared/management/certposture"
 	"github.com/netbirdio/netbird/shared/management/certposture/certtest"
 	"github.com/netbirdio/netbird/shared/management/proto"
 )
 
-func TestMergeProofs_ProvesACertificateHeldByBothStoresOnce(t *testing.T) {
-	ca := certtest.NewCA(t, "corp-root")
-	shared := ca.Issue(t, certtest.ECDSAKey(t), "shared")
-	userOnly := ca.Issue(t, certtest.ECDSAKey(t), "user-only")
-
-	device := []certposture.Proof{{Chain: [][]byte{shared.Raw}}}
-	user := []certposture.Proof{{Chain: [][]byte{shared.Raw}}, {Chain: [][]byte{userOnly.Raw}}, {}}
-
-	merged := mergeProofs(device, user)
-
-	require.Len(t, merged, 2, "the shared leaf is proven once and the empty chain is dropped")
-	assert.Equal(t, shared.Raw, merged[0].Chain[0], "the device proof keeps its place")
-	assert.Equal(t, userOnly.Raw, merged[1].Chain[0], "the user-only certificate is appended")
-}
-
-func TestMergeProofs_KeepsDeviceProofsWhenNoUserSession(t *testing.T) {
-	ca := certtest.NewCA(t, "corp-root")
-	device := []certposture.Proof{{Chain: [][]byte{ca.Issue(t, certtest.ECDSAKey(t), "device").Raw}}}
-
-	merged := mergeProofs(device, nil)
-
-	require.Len(t, merged, 1, "a machine with nobody signed in still sends its device proof")
-	assert.Equal(t, sha256.Sum256(device[0].Chain[0]), sha256.Sum256(merged[0].Chain[0]), "the device proof is unchanged")
-}
-
 func TestHelperRequest_CarriesEveryChallenge(t *testing.T) {
 	ca := certtest.NewCA(t, "corp-root")
 	challenges := []*proto.CertificateChallenge{

client/internal/certproof/helper_test.go

@@ -14,7 +14,10 @@ import (
 	"github.com/netbirdio/netbird/shared/management/certposture/certtest"
 )
 
-func TestRunHelper_ProofSurvivesTheProcessBoundary(t *testing.T) {
+// TestRunHelper_AnswersAChallengeFromItsStore runs the helper's request and response
+// handling in-process: the proof it writes verifies as the daemon will verify it. The
+// process boundary itself is covered by the runHelperCmd tests.
+func TestRunHelper_AnswersAChallengeFromItsStore(t *testing.T) {
 	ca := certtest.NewCA(t, "corp-root")
 	dir := storeDir(t)
 	key := certtest.ECDSAKey(t)
@@ -29,7 +32,7 @@ func TestRunHelper_ProofSurvivesTheProcessBoundary(t *testing.T) {
 	require.NoError(t, err, "request must encode")
 
 	var stdout bytes.Buffer
-	require.NoError(t, runHelper(context.Background(), NewFileStore(dir), bytes.NewReader(request), &stdout))
+	require.NoError(t, runHelper(context.Background(), newFileStore(dir), bytes.NewReader(request), &stdout))
 
 	var resp HelperResponse
 	require.NoError(t, json.Unmarshal(stdout.Bytes(), &resp), "helper must emit decodable JSON")
@@ -47,7 +50,7 @@ func TestRunHelper_NoChallengesYieldsEmptyResponse(t *testing.T) {
 	require.NoError(t, err)
 
 	var stdout bytes.Buffer
-	require.NoError(t, runHelper(context.Background(), NewFileStore(storeDir(t)), bytes.NewReader(request), &stdout))
+	require.NoError(t, runHelper(context.Background(), newFileStore(storeDir(t)), bytes.NewReader(request), &stdout))
 
 	var resp HelperResponse
 	require.NoError(t, json.Unmarshal(stdout.Bytes(), &resp), "an empty request must still emit valid JSON")
@@ -56,7 +59,7 @@ func TestRunHelper_NoChallengesYieldsEmptyResponse(t *testing.T) {
 
 func TestRunHelper_RejectsMalformedRequest(t *testing.T) {
 	var stdout bytes.Buffer
-	err := runHelper(context.Background(), NewFileStore(storeDir(t)), bytes.NewReader([]byte("not json")), &stdout)
+	err := runHelper(context.Background(), newFileStore(storeDir(t)), bytes.NewReader([]byte("not json")), &stdout)
 
 	require.Error(t, err, "a malformed request must fail rather than emit an empty proof set")
 	assert.Empty(t, stdout.String(), "nothing should be written to stdout on a decode failure")

client/internal/certproof/keychain_darwin.go

@@ -71,25 +71,25 @@ const systemKeychain = "/Library/Keychains/System.keychain"
 // by an ordinary user, it is that user's keychain search list.
 func DefaultStore() Store {
 	if os.Geteuid() == 0 {
-		return NewKeychainStore(systemKeychain)
+		return newKeychainStore(systemKeychain)
 	}
-	return NewKeychainStore()
+	return newKeychainStore()
 }
 
-// KeychainStore yields the identities of a set of keychains, or of the process's keychain
+// keychainStore yields the identities of a set of keychains, or of the process's keychain
 // search list when none is named, reached through purego so the client keeps building
 // with CGO_ENABLED=0.
-type KeychainStore struct {
+type keychainStore struct {
 	keychains []string
 }
 
-// NewKeychainStore returns a store that searches the keychain files at paths, or the
+// newKeychainStore returns a store that searches the keychain files at paths, or the
 // process's keychain search list when no path is given.
-func NewKeychainStore(paths ...string) *KeychainStore {
-	return &KeychainStore{keychains: paths}
+func newKeychainStore(paths ...string) *keychainStore {
+	return &keychainStore{keychains: paths}
 }
 
-func (s *KeychainStore) Candidates(_ context.Context) ([]Candidate, error) {
+func (s *keychainStore) Candidates(_ context.Context) ([]Candidate, error) {
 	if err := loadKeychain(); err != nil {
 		return nil, err
 	}
@@ -163,17 +163,24 @@ func (s *keychainSigner) Sign(_ io.Reader, digest []byte, opts crypto.SignerOpts
 
 	algorithm := keychainAlgorithm(scheme)
 	var signature []byte
+	var attempts signAttempts
 	err = eachIdentity(searchList, func(identity uintptr, der []byte) (bool, error) {
 		if !bytes.Equal(der, s.leaf.Raw) {
 			return false, nil
 		}
-		signature, err = signWithIdentity(identity, algorithm, digest)
-		return true, err
+		// The same certificate can sit in the keychain with more than one key, of which
+		// only one may be usable, so a failure moves on to the next match.
+		var ok bool
+		signature, ok = attempts.try(func() ([]byte, error) { return signWithIdentity(identity, algorithm, digest) })
+		return ok, nil
 	})
 	if err != nil {
 		return nil, err
 	}
 	if signature == nil {
+		if err := attempts.err(); err != nil {
+			return nil, err
+		}
 		return nil, errors.New("certificate is no longer in the keychain")
 	}
 	log.Debugf("keychain signed certificate posture challenge for %q, %d bytes", s.leaf.Subject, len(signature))

client/internal/certproof/keystore_test.go

@@ -56,6 +56,8 @@ func TestSchemeFor(t *testing.T) {
 		{"ecdsa sha384", ecKey.Public(), crypto.SHA384, schemeECDSASHA384},
 		{"rsa pss sha256", rsaKey.Public(), pss, schemeRSAPSSSHA256},
 		{"rsa pkcs1v15", rsaKey.Public(), crypto.SHA256, 0},
+		{"ecdsa sha512", ecKey.Public(), crypto.SHA512, 0},
+		{"rsa pss sha384", rsaKey.Public(), &rsa.PSSOptions{SaltLength: rsa.PSSSaltLengthEqualsHash, Hash: crypto.SHA384}, 0},
 		{"ed25519", certtest.Ed25519Key(t).Public(), crypto.Hash(0), 0},
 	}
 	for _, tc := range tests {
@@ -83,5 +85,7 @@ func TestECDSASignatureASN1(t *testing.T) {
 	assert.True(t, ecdsa.VerifyASN1(&key.PublicKey, digest[:], der))
 
 	_, err = ecdsaSignatureASN1(raw[:63])
-	assert.Error(t, err)
+	assert.Error(t, err, "an odd length cannot split into r and s")
+	_, err = ecdsaSignatureASN1(nil)
+	assert.Error(t, err, "an empty signature is malformed")
 }

client/internal/certproof/merge.go

@@ -0,0 +1,50 @@
+package certproof
+
+import (
+	"crypto/sha256"
+	"crypto/x509"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/shared/management/certposture"
+)
+
+// mergeProofs appends the user session proofs to the device proofs, dropping one whose
+// chain the device already proved. Management checks each chain against each check's CAs,
+// so the same leaf proven through another chain is kept.
+func mergeProofs(device, user []certposture.Proof) []certposture.Proof {
+	if len(user) == 0 {
+		return device
+	}
+
+	seen := make(map[[sha256.Size]byte]struct{}, len(device)+len(user))
+	for _, proof := range device {
+		if len(proof.Chain) > 0 {
+			seen[chainDERFingerprint(proof.Chain)] = struct{}{}
+		}
+	}
+
+	merged := device
+	for _, proof := range user {
+		if len(proof.Chain) == 0 {
+			continue
+		}
+		fingerprint := chainDERFingerprint(proof.Chain)
+		if _, done := seen[fingerprint]; done {
+			continue
+		}
+		seen[fingerprint] = struct{}{}
+		merged = append(merged, proof)
+		logUserProof(proof)
+	}
+	return merged
+}
+
+func logUserProof(proof certposture.Proof) {
+	leaf, err := x509.ParseCertificate(proof.Chain[0])
+	if err != nil {
+		log.Debugf("certificate posture: user proof carries an unparsable leaf: %v", err)
+		return
+	}
+	log.Debugf("certificate posture: signed-in user proved %q issued by %q", leaf.Subject, leaf.Issuer)
+}

client/internal/certproof/merge_test.go

@@ -0,0 +1,54 @@
+package certproof
+
+import (
+	"crypto/sha256"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/shared/management/certposture"
+	"github.com/netbirdio/netbird/shared/management/certposture/certtest"
+)
+
+func TestMergeProofs_ProvesACertificateHeldByBothStoresOnce(t *testing.T) {
+	ca := certtest.NewCA(t, "corp-root")
+	shared := ca.Issue(t, certtest.ECDSAKey(t), "shared")
+	userOnly := ca.Issue(t, certtest.ECDSAKey(t), "user-only")
+
+	device := []certposture.Proof{{Chain: [][]byte{shared.Raw}}}
+	user := []certposture.Proof{{Chain: [][]byte{shared.Raw}}, {Chain: [][]byte{userOnly.Raw}}, {}}
+
+	merged := mergeProofs(device, user)
+
+	require.Len(t, merged, 2, "the shared leaf is proven once and the empty chain is dropped")
+	assert.Equal(t, shared.Raw, merged[0].Chain[0], "the device proof keeps its place")
+	assert.Equal(t, userOnly.Raw, merged[1].Chain[0], "the user-only certificate is appended")
+}
+
+func TestMergeProofs_KeepsDeviceProofsWhenNoUserSession(t *testing.T) {
+	ca := certtest.NewCA(t, "corp-root")
+	device := []certposture.Proof{{Chain: [][]byte{ca.Issue(t, certtest.ECDSAKey(t), "device").Raw}}}
+
+	merged := mergeProofs(device, nil)
+
+	require.Len(t, merged, 1, "a machine with nobody signed in still sends its device proof")
+	assert.Equal(t, sha256.Sum256(device[0].Chain[0]), sha256.Sum256(merged[0].Chain[0]), "the device proof is unchanged")
+}
+
+func TestMergeProofs_KeepsAnotherChainOfTheSameLeaf(t *testing.T) {
+	root := certtest.NewCA(t, "corp-root")
+	interA := root.Issue(t, certtest.ECDSAKey(t), "inter-a")
+	interB := root.Issue(t, certtest.ECDSAKey(t), "inter-b")
+	leaf := root.Issue(t, certtest.ECDSAKey(t), "leaf")
+
+	// Management checks each chain against each check's CAs, so the same leaf proven
+	// through another chain is a different proof, not a duplicate.
+	device := []certposture.Proof{{Chain: [][]byte{leaf.Raw, interA.Raw}}}
+	user := []certposture.Proof{{Chain: [][]byte{leaf.Raw, interB.Raw}}, {Chain: [][]byte{leaf.Raw, interA.Raw}}}
+
+	merged := mergeProofs(device, user)
+
+	require.Len(t, merged, 2, "the chain only the user store holds is kept, the one both hold is not repeated")
+	assert.Equal(t, interB.Raw, merged[1].Chain[1], "the user's chain through the other intermediate is appended")
+}

client/internal/certproof/pkcs11store.go

@@ -38,24 +38,24 @@ func PKCS11FromEnv() PKCS11Config {
 	return PKCS11Config{URI: os.Getenv(PKCS11URIEnv), PIN: os.Getenv(PINEnv)}
 }
 
-// PKCS11Store yields the identities of a PKCS#11 token, which is how tpm2-pkcs11 exposes
+// pkcs11Store yields the identities of a PKCS#11 token, which is how tpm2-pkcs11 exposes
 // TPM-held keys on Linux. Certificates on the token are paired with keys by CKA_ID, the
 // convention tpm2_ptool addcert and pkcs11-tool follow; certificate files in the PEM
 // directory by public key. Every signature happens on the token.
-type PKCS11Store struct {
+type pkcs11Store struct {
 	uri     *pkcs11.URI
 	pin     string
 	certDir string
 }
 
-// NewPKCS11Store parses cfg.URI, standing in the bare defaults when it is empty. Files in
+// newPKCS11Store parses cfg.URI, standing in the bare defaults when it is empty. Files in
 // certDir without a key of their own are paired with the token's keys by public key.
 //
 // A PIN is only accepted together with a token label: without one the PIN would go to
 // whichever token the module lists first, which on a machine with a smartcard plugged
 // in may be the card, and every wrong PIN counts towards that card's lockout.
-func NewPKCS11Store(cfg PKCS11Config, certDir string) (*PKCS11Store, error) {
-	store := &PKCS11Store{uri: &pkcs11.URI{}, pin: cfg.PIN, certDir: certDir}
+func newPKCS11Store(cfg PKCS11Config, certDir string) (*pkcs11Store, error) {
+	store := &pkcs11Store{uri: &pkcs11.URI{}, pin: cfg.PIN, certDir: certDir}
 	if cfg.URI != "" {
 		parsed, err := pkcs11.ParseURI(cfg.URI)
 		if err != nil {
@@ -69,7 +69,7 @@ func NewPKCS11Store(cfg PKCS11Config, certDir string) (*PKCS11Store, error) {
 	return store, nil
 }
 
-func (s *PKCS11Store) Candidates(_ context.Context) ([]Candidate, error) {
+func (s *pkcs11Store) Candidates(_ context.Context) ([]Candidate, error) {
 	session, err := s.open()
 	if err != nil {
 		return nil, err
@@ -80,7 +80,7 @@ func (s *PKCS11Store) Candidates(_ context.Context) ([]Candidate, error) {
 	if err != nil {
 		return nil, err
 	}
-	fileChains, err := s.fileChains()
+	fileChains, fileIntermediates, err := s.fileChains()
 	if err != nil {
 		return nil, err
 	}
@@ -93,6 +93,7 @@ func (s *PKCS11Store) Candidates(_ context.Context) ([]Candidate, error) {
 	for _, chain := range fileChains {
 		pool = append(pool, chain...)
 	}
+	pool = append(pool, fileIntermediates...)
 
 	var candidates []Candidate
 	for _, cert := range certs {
@@ -106,9 +107,12 @@ func (s *PKCS11Store) Candidates(_ context.Context) ([]Candidate, error) {
 		return candidates, nil
 	}
 
+	// A token that cannot list its public keys still answers for the certificates it
+	// holds, so the identities found above are kept.
 	keys, err := tokenPublicKeys(session)
 	if err != nil {
-		return nil, err
+		log.Warnf("%s: cannot read its public keys, certificate files are not paired with them: %v", s, err)
+		return candidates, nil
 	}
 	for _, chain := range fileChains {
 		leaf := chain[0]
@@ -122,41 +126,45 @@ func (s *PKCS11Store) Candidates(_ context.Context) ([]Candidate, error) {
 	return candidates, nil
 }
 
-func (s *PKCS11Store) candidate(leaf *x509.Certificate, id []byte, pool []*x509.Certificate) Candidate {
+func (s *pkcs11Store) candidate(leaf *x509.Certificate, id []byte, pool []*x509.Certificate) Candidate {
 	chain := buildChain(leaf, pool)
 	log.Debugf("%s candidate %q issued by %q built a chain of %d certificates", s, leaf.Subject, leaf.Issuer, len(chain))
 	return Candidate{Chain: chain, Signer: &pkcs11Signer{store: s, leaf: leaf, id: id}, Intermediates: pool}
 }
 
 // fileChains reads the certificate files in the PEM directory that carry no key of their
-// own; the file store answers for the ones that do.
-func (s *PKCS11Store) fileChains() ([][]*x509.Certificate, error) {
+// own; the file store answers for the ones that do. The certificates of a file that does
+// carry a key are returned too, since they can complete the chain of a token certificate.
+func (s *pkcs11Store) fileChains() (chains [][]*x509.Certificate, intermediates []*x509.Certificate, err error) {
 	if s.certDir == "" {
-		return nil, nil
+		return nil, nil, nil
 	}
 	paths, err := certFiles(s.certDir)
 	if err != nil {
-		return nil, err
+		return nil, nil, err
 	}
-	var chains [][]*x509.Certificate
 	for _, path := range paths {
 		chain, signer, err := loadPEM(path)
-		if err != nil || signer != nil {
+		if err != nil {
+			continue
+		}
+		if signer != nil {
+			intermediates = append(intermediates, chain...)
 			continue
 		}
 		chains = append(chains, chain)
 	}
-	return chains, nil
+	return chains, intermediates, nil
 }
 
-func (s *PKCS11Store) String() string {
+func (s *pkcs11Store) String() string {
 	if s.uri.Token == "" {
 		return "PKCS#11 token"
 	}
 	return fmt.Sprintf("PKCS#11 token %q", s.uri.Token)
 }
 
-func (s *PKCS11Store) open() (*pkcs11.Session, error) {
+func (s *pkcs11Store) open() (*pkcs11.Session, error) {
 	module, err := pkcs11.Load(s.uri.Module())
 	if err != nil {
 		return nil, err
@@ -185,7 +193,7 @@ func (s *PKCS11Store) open() (*pkcs11.Session, error) {
 	return session, err
 }
 
-func (s *PKCS11Store) userPIN() ([]byte, error) {
+func (s *pkcs11Store) userPIN() ([]byte, error) {
 	if s.pin != "" {
 		return []byte(s.pin), nil
 	}
@@ -208,7 +216,8 @@ func tokenPublicKeys(session *pkcs11.Session) (tokenKeys, error) {
 	for _, object := range objects {
 		id, err := session.Attribute(object, pkcs11.AttrID)
 		if err != nil {
-			return nil, err
+			log.Debugf("skipping public key on PKCS#11 token: %v", err)
+			continue
 		}
 		public, err := session.PublicKey(object)
 		if err != nil {
@@ -287,7 +296,7 @@ func privateKey(session *pkcs11.Session, id []byte) (pkcs11.Object, error) {
 // pkcs11Signer holds only the certificate and its CKA_ID; the key is looked up in a fresh
 // session at signing time so no token handle outlives a call.
 type pkcs11Signer struct {
-	store *PKCS11Store
+	store *pkcs11Store
 	leaf  *x509.Certificate
 	id    []byte
 }

client/internal/certproof/pkcs11store_test.go

@@ -44,7 +44,7 @@ func TestStores_KeepsFileCertificatesWhenTokenFails(t *testing.T) {
 	dir := storeDir(t)
 	writeFile(t, dir, "device.pem", certtest.CertPEM(ca.Issue(t, key, "device"))+certtest.KeyPEM(t, key))
 
-	candidates, err := Stores{failingStore{}, NewFileStore(dir)}.Candidates(context.Background())
+	candidates, err := Stores{failingStore{}, newFileStore(dir)}.Candidates(context.Background())
 	require.NoError(t, err)
 	assert.Len(t, candidates, 1, "the directory's certificate must survive a failing token")
 }
@@ -241,13 +241,13 @@ func attr(typ uint, value []byte) pkcs11.Attribute {
 
 // pkcs11TestStore builds the store for the token NB_TEST_PKCS11_URI names, skipping when
 // no token is configured or this build lacks PKCS#11 support.
-func pkcs11TestStore(t *testing.T, certDir string) (*PKCS11Store, string) {
+func pkcs11TestStore(t *testing.T, certDir string) (*pkcs11Store, string) {
 	t.Helper()
 	uri := os.Getenv(testPKCS11URIEnv)
 	if uri == "" {
 		t.Skipf("set %s to a PKCS#11 URI with a PIN to run", testPKCS11URIEnv)
 	}
-	store, err := NewPKCS11Store(PKCS11Config{URI: uri}, certDir)
+	store, err := newPKCS11Store(PKCS11Config{URI: uri}, certDir)
 	require.NoError(t, err)
 	if _, err := pkcs11.Load(store.uri.Module()); errors.Is(err, pkcs11.ErrUnsupported) {
 		t.Skip(err)
@@ -297,17 +297,30 @@ func TestPKCS11Store_FileChains(t *testing.T) {
 	writeFile(t, dir, "sibling.key", certtest.KeyPEM(t, keyed))
 	writeFile(t, dir, "notes.txt", "not a certificate")
 
-	chains, err := (&PKCS11Store{uri: &pkcs11.URI{}, certDir: dir}).fileChains()
+	// An intermediate that only a keyed file carries still has to reach the pool a token
+	// certificate's chain is built from.
+	root := certtest.NewCA(t, "corp-root")
+	intermediateCA := certtest.NewIntermediate(t, root, "corp-intermediate")
+	chained := certtest.ECDSAKey(t)
+	writeFile(t, dir, "chained.pem", certtest.CertPEM(intermediateCA.Issue(t, chained, "chained"))+certtest.CertPEM(intermediateCA.Cert)+certtest.KeyPEM(t, chained))
+
+	chains, extra, err := (&pkcs11Store{uri: &pkcs11.URI{}, certDir: dir}).fileChains()
 	require.NoError(t, err)
+	require.NotEmpty(t, extra, "the keyed file's certificates are offered for chain building")
+	tokenLeaf := intermediateCA.Issue(t, certtest.ECDSAKey(t), "token-leaf")
+	chain := buildChain(tokenLeaf, extra)
+	require.Len(t, chain, 2, "a token certificate is completed through the intermediate only the keyed file holds")
+	assert.True(t, intermediateCA.Cert.Equal(chain[1]), "the second certificate is that intermediate")
 	var subjects []string
 	for _, chain := range chains {
 		subjects = append(subjects, chain[0].Subject.CommonName)
 	}
 	assert.ElementsMatch(t, []string{"device", "corp"}, subjects, "only key-less certificate files are left to the token")
 
-	chains, err = (&PKCS11Store{uri: &pkcs11.URI{}}).fileChains()
+	chains, extra, err = (&pkcs11Store{uri: &pkcs11.URI{}}).fileChains()
 	require.NoError(t, err)
 	assert.Empty(t, chains, "no directory configured means no file certificates")
+	assert.Empty(t, extra, "and no extra intermediates")
 }
 
 func TestNewPKCS11Store_PIN(t *testing.T) {
@@ -324,7 +337,7 @@ func TestNewPKCS11Store_PIN(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			store, err := NewPKCS11Store(tt.cfg, "")
+			store, err := newPKCS11Store(tt.cfg, "")
 			require.NoError(t, err)
 			pin, err := store.userPIN()
 			require.NoError(t, err)
@@ -333,7 +346,7 @@ func TestNewPKCS11Store_PIN(t *testing.T) {
 		})
 	}
 
-	_, err := NewPKCS11Store(PKCS11Config{URI: "not-a-pkcs11-uri", PIN: "1234"}, "")
+	_, err := newPKCS11Store(PKCS11Config{URI: "not-a-pkcs11-uri", PIN: "1234"}, "")
 	assert.Error(t, err, "a malformed URI must not be silently replaced by the defaults")
 
 	for name, cfg := range map[string]PKCS11Config{
@@ -342,7 +355,7 @@ func TestNewPKCS11Store_PIN(t *testing.T) {
 		"inline pin-value only":    {URI: "pkcs11:?pin-value=0000"},
 		"pin-source only":          {URI: "pkcs11:?pin-source=file:/etc/netbird/pkcs11.pin"},
 	} {
-		_, err := NewPKCS11Store(cfg, "")
+		_, err := newPKCS11Store(cfg, "")
 		assert.ErrorIs(t, err, errPINNeedsToken, "%s: a PIN must not go to whichever token is listed first", name)
 	}
 }
@@ -360,7 +373,7 @@ func TestPKCS11Store_WrongPINIsTriedOnce(t *testing.T) {
 	_, uri := pkcs11TestStore(t, "")
 
 	wrongPIN := "wrong-pin-" + t.Name()
-	store, err := NewPKCS11Store(PKCS11Config{URI: uri, PIN: wrongPIN}, "")
+	store, err := newPKCS11Store(PKCS11Config{URI: uri, PIN: wrongPIN}, "")
 	require.NoError(t, err)
 
 	_, err = store.Candidates(context.Background())
@@ -370,7 +383,7 @@ func TestPKCS11Store_WrongPINIsTriedOnce(t *testing.T) {
 	_, err = store.Candidates(context.Background())
 	assert.ErrorIs(t, err, errPINRejectedBefore, "the rejected PIN is not sent to the token again")
 
-	good, err := NewPKCS11Store(PKCS11Config{URI: uri}, "")
+	good, err := newPKCS11Store(PKCS11Config{URI: uri}, "")
 	require.NoError(t, err)
 	_, err = good.Candidates(context.Background())
 	assert.NoError(t, err, "the correct PIN for the same token is unaffected")

client/internal/certproof/signattempts.go

@@ -0,0 +1,24 @@
+package certproof
+
+import "errors"
+
+// signAttempts tracks the identities tried to sign with, for a certificate that can match
+// more than one of them.
+type signAttempts struct {
+	errs []error
+}
+
+// try runs sign and reports its signature and whether it succeeded. A failure is kept.
+func (a *signAttempts) try(sign func() ([]byte, error)) ([]byte, bool) {
+	signature, err := sign()
+	if err != nil {
+		a.errs = append(a.errs, err)
+		return nil, false
+	}
+	return signature, true
+}
+
+// err joins the failures of the attempts, nil when there were none.
+func (a *signAttempts) err() error {
+	return errors.Join(a.errs...)
+}

client/internal/certproof/signattempts_test.go

@@ -0,0 +1,41 @@
+package certproof
+
+import (
+	"errors"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+var errNeedsApprovalForTest = errors.New("key needs approval")
+
+// TestSignAttempts_FallsThroughToTheNextIdentity: a certificate can sit in a keychain with
+// more than one key, of which only some can sign. A failure moves on, and an identity that
+// signs ends the search without reporting the earlier failure.
+func TestSignAttempts_FallsThroughToTheNextIdentity(t *testing.T) {
+	var attempts signAttempts
+
+	signature, ok := attempts.try(func() ([]byte, error) { return nil, errNeedsApprovalForTest })
+	assert.False(t, ok, "a failing identity does not end the search")
+	assert.Nil(t, signature)
+
+	signature, ok = attempts.try(func() ([]byte, error) { return []byte("sig"), nil })
+	require.True(t, ok, "the next identity signs")
+	assert.Equal(t, []byte("sig"), signature)
+}
+
+// TestSignAttempts_KeepsEveryFailure: when no identity signs, the approval a user has to
+// give must not be hidden behind the failure of a later identity.
+func TestSignAttempts_KeepsEveryFailure(t *testing.T) {
+	var attempts signAttempts
+	assert.NoError(t, attempts.err(), "no attempts, no error")
+
+	attempts.try(func() ([]byte, error) { return nil, errNeedsApprovalForTest })
+	attempts.try(func() ([]byte, error) { return nil, errors.New("key is gone") })
+
+	err := attempts.err()
+	require.Error(t, err)
+	assert.ErrorIs(t, err, errNeedsApprovalForTest, "the earlier approval error survives the later failure")
+	assert.ErrorContains(t, err, "key is gone")
+}

client/internal/certproof/store.go

@@ -46,7 +46,7 @@ func (c Candidate) issuers() []*x509.Certificate {
 	return append(slices.Clip(c.Chain[1:]), c.Intermediates...)
 }
 
-// Store yields the certificates a peer may prove possession of. FileStore is the PEM
+// Store yields the certificates a peer may prove possession of. fileStore is the PEM
 // directory implementation; OS keystores (CNG, Keychain, PKCS#11) slot in here.
 type Store interface {
 	Candidates(ctx context.Context) ([]Candidate, error)
@@ -68,15 +68,15 @@ type Config struct {
 	OwnerUnknown bool
 }
 
-// FileStore reads PEM files from a directory. A file holds the chain (leaf first) and
+// fileStore reads PEM files from a directory. A file holds the chain (leaf first) and
 // either its private key or a sibling "<name>.key" file holds it. The key is a plain
 // PKCS#8, EC or RSA key, or a TSS2 key the TPM signs with.
-type FileStore struct {
+type fileStore struct {
 	dir string
 }
 
-func NewFileStore(dir string) *FileStore {
-	return &FileStore{dir: dir}
+func newFileStore(dir string) *fileStore {
+	return &fileStore{dir: dir}
 }
 
 func StoreDir() string {
@@ -86,7 +86,7 @@ func StoreDir() string {
 	return defaultStoreDir
 }
 
-func (s *FileStore) Candidates(_ context.Context) ([]Candidate, error) {
+func (s *fileStore) Candidates(_ context.Context) ([]Candidate, error) {
 	paths, err := certFiles(s.dir)
 	if err != nil {
 		return nil, err

client/internal/certproof/store_other.go

@@ -14,13 +14,13 @@ var unsupportedTokenOnce sync.Once
 
 // DefaultStore is the PEM directory named by NB_CERT_STORE_DIR, or /etc/netbird/certs.
 func DefaultStore() Store {
-	return NewFileStore(StoreDir())
+	return newFileStore(StoreDir())
 }
 
 // storeWithToken reads the PEM directory cfg names, joined by the PKCS#11 token when cfg
 // names one. The token pairs the directory's key-less certificates with its own keys.
 func storeWithToken(cfg Config) Store {
-	files := NewFileStore(StoreDir())
+	files := newFileStore(StoreDir())
 	if cfg.PKCS11.URI == "" && cfg.PKCS11.PIN == "" {
 		return files
 	}
@@ -30,7 +30,7 @@ func storeWithToken(cfg Config) Store {
 		})
 		return files
 	}
-	token, err := NewPKCS11Store(cfg.PKCS11, StoreDir())
+	token, err := newPKCS11Store(cfg.PKCS11, StoreDir())
 	if err != nil {
 		log.Warnf("ignoring PKCS#11 URI: %v", err)
 		return files

client/internal/certproof/store_other_test.go

@@ -15,12 +15,12 @@ func TestStoreWithToken(t *testing.T) {
 	dir := storeDir(t)
 	t.Setenv(StoreDirEnv, dir)
 
-	files, ok := storeWithToken(Config{}).(*FileStore)
+	files, ok := storeWithToken(Config{}).(*fileStore)
 	require.True(t, ok, "no token configured reads the PEM directory alone")
 	assert.Equal(t, dir, files.dir, "the directory comes from NB_CERT_STORE_DIR")
 
-	assert.IsType(t, &FileStore{}, storeWithToken(Config{PKCS11: PKCS11Config{URI: "not-a-pkcs11-uri"}}), "an invalid URI must not hide the PEM directory")
-	assert.IsType(t, &FileStore{}, storeWithToken(Config{PKCS11: PKCS11Config{PIN: "1234"}}), "a PIN naming no token is refused and leaves the PEM directory")
+	assert.IsType(t, &fileStore{}, storeWithToken(Config{PKCS11: PKCS11Config{URI: "not-a-pkcs11-uri"}}), "an invalid URI must not hide the PEM directory")
+	assert.IsType(t, &fileStore{}, storeWithToken(Config{PKCS11: PKCS11Config{PIN: "1234"}}), "a PIN naming no token is refused and leaves the PEM directory")
 
 	configured := map[string]PKCS11Config{
 		"env pin with token uri": {URI: "pkcs11:token=netbird", PIN: "1234"},
@@ -28,7 +28,7 @@ func TestStoreWithToken(t *testing.T) {
 	}
 	if !pkcs11.Supported() {
 		for name, cfg := range configured {
-			assert.IsType(t, &FileStore{}, storeWithToken(Config{PKCS11: cfg}),
+			assert.IsType(t, &fileStore{}, storeWithToken(Config{PKCS11: cfg}),
 				"%s: a build without PKCS#11 support reads the PEM directory alone", name)
 		}
 		return
@@ -37,7 +37,7 @@ func TestStoreWithToken(t *testing.T) {
 		store, ok := storeWithToken(Config{PKCS11: cfg}).(Stores)
 		require.True(t, ok, "%s joins the token to the PEM directory", name)
 		require.Len(t, store, 2, name)
-		token, ok := store[1].(*PKCS11Store)
+		token, ok := store[1].(*pkcs11Store)
 		require.True(t, ok, name)
 		assert.Equal(t, dir, token.certDir, "%s: the token pairs certificates from the same directory", name)
 	}

client/internal/certproof/store_tpm_test.go

@@ -36,7 +36,7 @@ func TestFileStore_TPMKeyFile(t *testing.T) {
 	writeFile(t, dir, "locked.pem", certtest.CertPEM(ca.Issue(t, locked, "locked")))
 	writeFile(t, dir, "locked.key", tpmtest.KeyPEM(t, locked.Public().(*ecdsa.PublicKey), tpmtest.WithAuth()))
 
-	candidates, err := NewFileStore(dir).Candidates(context.Background())
+	candidates, err := newFileStore(dir).Candidates(context.Background())
 	require.NoError(t, err)
 	require.Len(t, candidates, 1, "only the key without an authorization value is usable")
 	assert.True(t, leaf.Equal(candidates[0].Chain[0]), "candidate must carry the device certificate")
@@ -73,7 +73,7 @@ func TestCollect_TPMKeyEndToEnd(t *testing.T) {
 	nonce := challenger.Nonce(peerKey, now)
 	checks := []*proto.Checks{{CertificateChallenge: &proto.CertificateChallenge{Nonce: nonce, CaCertificates: []string{ca.PEM}}}}
 
-	proofs := Collect(context.Background(), NewFileStore(dir), checks, peerKey)
+	proofs := Collect(context.Background(), newFileStore(dir), checks, peerKey)
 	require.Len(t, proofs, 1, "the TPM-held key must prove the certificate")
 	chain, err := challenger.Verify(proofs[0], peerKey, now)
 	require.NoError(t, err)

client/internal/certproof/storefile_unix_test.go

@@ -26,7 +26,7 @@ func devicePair(t *testing.T, dir string) {
 
 func candidates(t *testing.T, dir string) ([]Candidate, error) {
 	t.Helper()
-	return NewFileStore(dir).Candidates(context.Background())
+	return newFileStore(dir).Candidates(context.Background())
 }
 
 func TestFileStore_AcceptsAPrivatePair(t *testing.T) {

client/internal/certproof/systemstore_windows.go

@@ -49,10 +49,10 @@ type bcryptPSSPaddingInfo struct {
 // DefaultStore is the local machine's personal certificate store, where device
 // certificates enrolled through AD or Intune are kept.
 func DefaultStore() Store {
-	return NewSystemStore()
+	return newSystemStore()
 }
 
-// SystemStore yields the identities of a personal certificate store, completing their
+// systemStore yields the identities of a personal certificate store, completing their
 // chains from the matching intermediate CA store. Keys are used through CNG and never
 // exported.
 //
@@ -60,22 +60,22 @@ func DefaultStore() Store {
 // a service reads; the current user store lives in the signed-in user's registry hive
 // with keys protected against their profile, so it is only readable while running as
 // that user.
-type SystemStore struct {
+type systemStore struct {
 	location uint32
 }
 
-// NewSystemStore reads the local machine store, which is what the daemon uses.
-func NewSystemStore() *SystemStore {
-	return &SystemStore{location: windows.CERT_SYSTEM_STORE_LOCAL_MACHINE}
+// newSystemStore reads the local machine store, which is what the daemon uses.
+func newSystemStore() *systemStore {
+	return &systemStore{location: windows.CERT_SYSTEM_STORE_LOCAL_MACHINE}
 }
 
-// NewUserStore reads the calling user's personal store. It is only useful in a process
+// newUserStore reads the calling user's personal store. It is only useful in a process
 // already running as that user, which is what the posture helper is.
-func NewUserStore() *SystemStore {
-	return &SystemStore{location: windows.CERT_SYSTEM_STORE_CURRENT_USER}
+func newUserStore() *systemStore {
+	return &systemStore{location: windows.CERT_SYSTEM_STORE_CURRENT_USER}
 }
 
-func (s *SystemStore) Candidates(_ context.Context) ([]Candidate, error) {
+func (s *systemStore) Candidates(_ context.Context) ([]Candidate, error) {
 	leaves, err := storeCertificates(s.location, personalStore)
 	if err != nil {
 		return nil, err
@@ -100,7 +100,7 @@ func (s *SystemStore) Candidates(_ context.Context) ([]Candidate, error) {
 }
 
 // String names the store location the way the Windows documentation does.
-func (s *SystemStore) String() string {
+func (s *systemStore) String() string {
 	if s.location == windows.CERT_SYSTEM_STORE_CURRENT_USER {
 		return "CurrentUser"
 	}

client/internal/engine.go

@@ -387,7 +387,7 @@ func NewEngine(
 		mgmClient:          services.MgmClient,
 		relayManager:       services.RelayManager,
 		peerStore:          peerstore.NewConnStore(),
-		wgDetector:         stdnet.NewWGDetector(),
+		wgDetector:         stdnet.NewWGDetector(config.WgIfaceName),
 		syncMsgMux:         &sync.Mutex{},
 		config:             config,
 		mobileDep:          mobileDep,

client/internal/networkmonitor/check_change_linux.go

@@ -6,6 +6,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"net"
 	"syscall"
 
 	log "github.com/sirupsen/logrus"
@@ -35,7 +36,7 @@ func checkChange(ctx context.Context, nexthopv4, nexthopv6 systemops.Nexthop) er
 		// handle route changes
 		case route := <-routeChan:
 			// default route and main table
-			if route.Dst != nil || route.Table != syscall.RT_TABLE_MAIN {
+			if !isDefaultRoute(route.Dst) || route.Table != syscall.RT_TABLE_MAIN {
 				continue
 			}
 			switch route.Type {
@@ -52,3 +53,13 @@ func checkChange(ctx context.Context, nexthopv4, nexthopv6 systemops.Nexthop) er
 		}
 	}
 }
+
+// isDefaultRoute reports whether dst is a default route. netlink v1.3 reports
+// the default as 0.0.0.0/0 or ::/0, older versions as a nil Dst.
+func isDefaultRoute(dst *net.IPNet) bool {
+	if dst == nil {
+		return true
+	}
+	ones, _ := dst.Mask.Size()
+	return ones == 0
+}

client/internal/networkmonitor/check_change_linux_test.go

@@ -0,0 +1,43 @@
+//go:build !android
+
+package networkmonitor
+
+import (
+	"net"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestIsDefaultRoute(t *testing.T) {
+	mustCIDR := func(s string) *net.IPNet {
+		t.Helper()
+		_, n, err := net.ParseCIDR(s)
+		if err != nil {
+			t.Fatalf("parse %s: %v", s, err)
+		}
+		return n
+	}
+
+	tests := []struct {
+		name string
+		dst  *net.IPNet
+		want bool
+	}{
+		// netlink before v1.3 left Dst nil for the default route.
+		{name: "nil dst", dst: nil, want: true},
+		// netlink v1.3 reports the default route with an explicit zero-length prefix.
+		{name: "v4 default", dst: mustCIDR("0.0.0.0/0"), want: true},
+		{name: "v6 default", dst: mustCIDR("::/0"), want: true},
+		// The exit node split routes and ordinary prefixes are not defaults.
+		{name: "v4 split half", dst: mustCIDR("0.0.0.0/1"), want: false},
+		{name: "v4 subnet", dst: mustCIDR("10.0.0.0/8"), want: false},
+		{name: "v4 host", dst: mustCIDR("1.1.1.1/32"), want: false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assert.Equal(t, tt.want, isDefaultRoute(tt.dst), "default route detection")
+		})
+	}
+}

client/internal/pkcs11/driver_linux.go

@@ -32,6 +32,7 @@ const (
 	// signature is 2 KiB.
 	maxAttributeLen = 1 << 20
 	maxSignatureLen = 1 << 16
+	maxFoundObjects = 1 << 16
 )
 
 type version struct {
@@ -103,6 +104,15 @@ const (
 	fnSign              = 43
 )
 
+// checkFunctionListVersion refuses a module whose function list is older than the layout
+// functionList mirrors, which would otherwise be read past its end.
+func checkFunctionListVersion(v version) error {
+	if v.major < 2 {
+		return fmt.Errorf("PKCS#11 function list version %d.%d is older than 2.0", v.major, v.minor)
+	}
+	return nil
+}
+
 // module holds the entry points of one loaded library, bound straight from its
 // CK_FUNCTION_LIST.
 type module struct {
@@ -152,6 +162,10 @@ func load(path string) (_ driver, err error) {
 		return nil, Error{Op: "C_GetFunctionList", Code: uint(rv)}
 	}
 
+	if err := checkFunctionListVersion(list.version); err != nil {
+		return nil, fmt.Errorf("%s: %w", path, err)
+	}
+
 	m := &module{}
 	for _, entry := range []struct {
 		fn    any
@@ -185,21 +199,43 @@ func load(path string) (_ driver, err error) {
 	return m, nil
 }
 
+// slotListAttempts bounds how often the slot list is asked for again after it grew between
+// the size query and the read.
+const slotListAttempts = 3
+
+// slotList returns the slots with a token present. A token plugged in between the size
+// query and the read makes the module answer CKR_BUFFER_TOO_SMALL, in which case the list
+// is asked for again.
+func (m *module) slotList() ([]ulong, error) {
+	for range slotListAttempts {
+		var count ulong
+		if rv := m.cGetSlotList(1, nil, &count); rv != rvOK {
+			return nil, Error{Op: "C_GetSlotList", Code: uint(rv)}
+		}
+		if count == 0 {
+			return nil, nil
+		}
+		slots := make([]ulong, count)
+		rv := m.cGetSlotList(1, &slots[0], &count)
+		if rv == rvBufferTooSmall {
+			continue
+		}
+		if rv != rvOK {
+			return nil, Error{Op: "C_GetSlotList", Code: uint(rv)}
+		}
+		return slots[:count], nil
+	}
+	return nil, Error{Op: "C_GetSlotList", Code: rvBufferTooSmall}
+}
+
 func (m *module) tokens() ([]Token, error) {
-	var count ulong
-	if rv := m.cGetSlotList(1, nil, &count); rv != rvOK {
-		return nil, Error{Op: "C_GetSlotList", Code: uint(rv)}
-	}
-	if count == 0 {
-		return nil, nil
-	}
-	slots := make([]ulong, count)
-	if rv := m.cGetSlotList(1, &slots[0], &count); rv != rvOK {
-		return nil, Error{Op: "C_GetSlotList", Code: uint(rv)}
+	slots, err := m.slotList()
+	if err != nil {
+		return nil, err
 	}
 
-	tokens := make([]Token, 0, count)
-	for _, slot := range slots[:count] {
+	tokens := make([]Token, 0, len(slots))
+	for _, slot := range slots {
 		var info tokenInfo
 		if rv := m.cGetTokenInfo(slot, &info); rv != rvOK {
 			continue
@@ -262,11 +298,16 @@ func (m *module) findObjects(session uint, template []Attribute) ([]Object, erro
 		if count > findBatch {
 			return nil, fmt.Errorf("C_FindObjects reported %d handles for a batch of %d", count, findBatch)
 		}
+		// Only an empty batch ends the search: a module may return fewer handles than
+		// asked for before it has returned them all.
+		if count == 0 {
+			return objects, nil
+		}
 		for _, handle := range batch[:count] {
 			objects = append(objects, Object(handle))
 		}
-		if count < findBatch {
-			return objects, nil
+		if len(objects) > maxFoundObjects {
+			return nil, fmt.Errorf("C_FindObjects returned more than %d objects", maxFoundObjects)
 		}
 	}
 }

client/internal/pkcs11/driver_linux_test.go

@@ -70,3 +70,119 @@ func loadFirst(t *testing.T, paths ...string) *Module {
 	t.Skip("p11-kit trust module not installed")
 	return nil
 }
+
+// TestSlotList_RetriesWhenTheListGrows: a token plugged in between the size query and the
+// fill makes the module answer CKR_BUFFER_TOO_SMALL; the list has to be asked for again.
+func TestSlotList_RetriesWhenTheListGrows(t *testing.T) {
+	calls := 0
+	m := &module{cGetSlotList: func(_ byte, slots *ulong, count *ulong) ulong {
+		calls++
+		switch {
+		case slots == nil && calls == 1:
+			*count = 1
+		case slots != nil && calls == 2:
+			*count = 2
+			return rvBufferTooSmall
+		case slots == nil:
+			*count = 2
+		default:
+			*count = 2
+			backing := unsafe.Slice(slots, 2)
+			backing[0], backing[1] = 7, 9
+		}
+		return rvOK
+	}}
+
+	slots, err := m.slotList()
+
+	require.NoError(t, err)
+	assert.Equal(t, []ulong{7, 9}, slots, "the list is read again at its new size")
+}
+
+func TestSlotList_GivesUpOnAListThatKeepsGrowing(t *testing.T) {
+	m := &module{cGetSlotList: func(_ byte, slots *ulong, count *ulong) ulong {
+		if slots == nil {
+			*count = 1
+			return rvOK
+		}
+		return rvBufferTooSmall
+	}}
+
+	_, err := m.slotList()
+
+	assert.Error(t, err, "a module that never settles on a size is an error, not a loop")
+}
+
+func TestCheckFunctionListVersion(t *testing.T) {
+	assert.NoError(t, checkFunctionListVersion(version{major: 2, minor: 40}))
+	assert.NoError(t, checkFunctionListVersion(version{major: 3, minor: 0}))
+	assert.Error(t, checkFunctionListVersion(version{major: 1, minor: 0}), "a pre-2.0 list has no entry points to index")
+	assert.Error(t, checkFunctionListVersion(version{}), "an empty version is not a function list")
+}
+
+// fakeFindModule answers C_FindObjects from batches, a handle per entry, and reports how
+// the search was torn down.
+func fakeFindModule(batches [][]ulong, finalized *int) *module {
+	next := 0
+	return &module{
+		cFindObjectsInit:  func(ulong, *attribute, ulong) ulong { return rvOK },
+		cFindObjectsFinal: func(ulong) ulong { *finalized++; return rvOK },
+		cFindObjects: func(_ ulong, objects *ulong, max ulong, count *ulong) ulong {
+			if next >= len(batches) {
+				*count = 0
+				return rvOK
+			}
+			batch := batches[next]
+			next++
+			copy(unsafe.Slice(objects, max), batch)
+			*count = ulong(len(batch))
+			return rvOK
+		},
+	}
+}
+
+var findTemplate = []Attribute{{Type: AttrClass, Value: ULong(ClassPrivateKey)}}
+
+// TestFindObjects_ReadsPastAShortBatch: PKCS#11 ends a search with an empty batch, not with
+// one shorter than asked for, so a module that hands out its handles in small batches must
+// still have all of them returned.
+func TestFindObjects_ReadsPastAShortBatch(t *testing.T) {
+	finalized := 0
+	m := fakeFindModule([][]ulong{{1, 2}, {3}, {4, 5, 6}}, &finalized)
+
+	objects, err := m.findObjects(1, findTemplate)
+
+	require.NoError(t, err)
+	assert.Equal(t, []Object{1, 2, 3, 4, 5, 6}, objects, "every batch is read until the module returns an empty one")
+	assert.Equal(t, 1, finalized, "the search is finished once")
+}
+
+func TestFindObjects_StopsAtAnEmptyBatch(t *testing.T) {
+	finalized := 0
+	m := fakeFindModule(nil, &finalized)
+
+	objects, err := m.findObjects(1, findTemplate)
+
+	require.NoError(t, err)
+	assert.Empty(t, objects, "a search that finds nothing returns nothing")
+	assert.Equal(t, 1, finalized)
+}
+
+// TestFindObjects_BoundsAModuleThatNeverEnds: a module that keeps returning handles must
+// not make the daemon allocate without limit.
+func TestFindObjects_BoundsAModuleThatNeverEnds(t *testing.T) {
+	finalized := 0
+	endless := &module{
+		cFindObjectsInit:  func(ulong, *attribute, ulong) ulong { return rvOK },
+		cFindObjectsFinal: func(ulong) ulong { finalized++; return rvOK },
+		cFindObjects: func(_ ulong, _ *ulong, _ ulong, count *ulong) ulong {
+			*count = findBatch
+			return rvOK
+		},
+	}
+
+	_, err := endless.findObjects(1, findTemplate)
+
+	assert.Error(t, err, "an endless search is cut off")
+	assert.Equal(t, 1, finalized, "and still finished")
+}

client/internal/pkcs11/pkcs11.go

@@ -50,6 +50,7 @@ const (
 	rvPINExpired          = 0xa3
 	rvPINLocked           = 0xa4
 	rvUserAlreadyLoggedIn = 0x100
+	rvBufferTooSmall      = 0x150
 	rvAlreadyInitialized  = 0x191
 )
 

client/internal/pkcs11/publickey.go

@@ -70,8 +70,10 @@ func rsaPublicKey(modulus, exponent []byte) (*rsa.PublicKey, error) {
 // uncompressed point wrapped in a DER OCTET STRING, which some modules hand out bare.
 func ecPublicKey(params, point []byte) (*ecdsa.PublicKey, error) {
 	var oid asn1.ObjectIdentifier
-	if _, err := asn1.Unmarshal(params, &oid); err != nil {
+	if rest, err := asn1.Unmarshal(params, &oid); err != nil {
 		return nil, fmt.Errorf("CKA_EC_PARAMS: %w", err)
+	} else if len(rest) != 0 {
+		return nil, errors.New("CKA_EC_PARAMS has trailing data")
 	}
 	curve, ok := curvesByOID[oid.String()]
 	if !ok {
@@ -80,8 +82,10 @@ func ecPublicKey(params, point []byte) (*ecdsa.PublicKey, error) {
 	size := (curve.Params().BitSize + 7) / 8
 	raw := point
 	if len(raw) != 1+2*size {
-		if _, err := asn1.Unmarshal(point, &raw); err != nil {
+		if rest, err := asn1.Unmarshal(point, &raw); err != nil {
 			return nil, fmt.Errorf("CKA_EC_POINT: %w", err)
+		} else if len(rest) != 0 {
+			return nil, errors.New("CKA_EC_POINT has trailing data")
 		}
 	}
 	if len(raw) != 1+2*size || raw[0] != 4 {

client/internal/pkcs11/publickey_test.go

@@ -6,6 +6,7 @@ import (
 	"crypto/rand"
 	"crypto/rsa"
 	"encoding/asn1"
+	"math"
 	"math/big"
 	"testing"
 
@@ -21,6 +22,8 @@ func TestECPublicKey(t *testing.T) {
 	}{
 		{"P-256", elliptic.P256(), asn1.ObjectIdentifier{1, 2, 840, 10045, 3, 1, 7}},
 		{"P-384", elliptic.P384(), asn1.ObjectIdentifier{1, 3, 132, 0, 34}},
+		// The wrapped P-521 point is 133 bytes, which needs the long form of the DER length.
+		{"P-521", elliptic.P521(), asn1.ObjectIdentifier{1, 3, 132, 0, 35}},
 	}
 	for _, tt := range curves {
 		t.Run(tt.name, func(t *testing.T) {
@@ -57,6 +60,13 @@ func TestECPublicKey_Rejections(t *testing.T) {
 	assert.Error(t, err, "a truncated point must be rejected")
 	_, err = ecPublicKey([]byte("junk"), point)
 	assert.Error(t, err, "malformed parameters must be rejected")
+
+	wrapped, err := asn1.Marshal(point)
+	require.NoError(t, err)
+	_, err = ecPublicKey(append(append([]byte{}, p256...), 0), point)
+	assert.Error(t, err, "bytes after the curve parameters must be rejected")
+	_, err = ecPublicKey(p256, append(append([]byte{}, wrapped...), 0))
+	assert.Error(t, err, "bytes after the wrapped point must be rejected")
 }
 
 func TestRSAPublicKey(t *testing.T) {
@@ -69,6 +79,12 @@ func TestRSAPublicKey(t *testing.T) {
 
 	_, err = rsaPublicKey(key.N.Bytes(), nil)
 	assert.Error(t, err, "a missing exponent must be rejected")
+	_, err = rsaPublicKey(key.N.Bytes(), big.NewInt(math.MaxInt32).Bytes())
+	assert.NoError(t, err, "the largest exponent an int holds is accepted")
+	_, err = rsaPublicKey(key.N.Bytes(), big.NewInt(math.MaxInt32+1).Bytes())
+	assert.Error(t, err, "an exponent past int32 must be rejected")
+	_, err = rsaPublicKey(key.N.Bytes(), []byte{0})
+	assert.Error(t, err, "a zero exponent must be rejected")
 }
 
 func TestULongRoundTrip(t *testing.T) {

client/internal/pkcs11/uri.go

@@ -29,7 +29,7 @@ type URI struct {
 // module-path must be absolute and a module-name must be a bare name. Unknown query
 // attributes are ignored, as the RFC asks.
 func ParseURI(raw string) (*URI, error) {
-	rest, ok := strings.CutPrefix(raw, "pkcs11:")
+	rest, ok := cutPrefixFold(raw, "pkcs11:")
 	if !ok {
 		return nil, errors.New("PKCS#11 URI does not start with the pkcs11 scheme")
 	}
@@ -72,11 +72,61 @@ func (u *URI) setQueryAttribute(name, value string) error {
 	case "pin-value":
 		u.pinValue = &value
 	case "pin-source":
-		u.pinSource = value
+		path, err := pinSourcePath(value)
+		if err != nil {
+			return err
+		}
+		u.pinSource = path
 	}
 	return nil
 }
 
+// pinSourcePath turns a pin-source value into the file to read. A plain path is used as
+// it is; a file URI must name a local absolute path, with no host or localhost (RFC 8089
+// section 2), so a PIN is never read from somewhere the URI only appears to point to.
+func pinSourcePath(value string) (string, error) {
+	if value == "" {
+		return "", errors.New("PKCS#11 URI pin-source is empty")
+	}
+	rest, isFile := cutPrefixFold(value, "file:")
+	if !isFile {
+		return value, nil
+	}
+	authorityAndPath, hasAuthority := strings.CutPrefix(rest, "//")
+	if !hasAuthority {
+		if !isAbsolutePath(rest) {
+			return "", fmt.Errorf("PKCS#11 URI pin-source %q must be an absolute path", value)
+		}
+		return rest, nil
+	}
+	if isAbsolutePath(authorityAndPath) {
+		return authorityAndPath, nil
+	}
+	host, path, _ := strings.Cut(authorityAndPath, "/")
+	if host != "" && !strings.EqualFold(host, "localhost") {
+		return "", fmt.Errorf("PKCS#11 URI pin-source host %q is not the local machine", host)
+	}
+	if path == "" {
+		return "", fmt.Errorf("PKCS#11 URI pin-source %q names no file", value)
+	}
+	return "/" + path, nil
+}
+
+// isAbsolutePath reports whether p is absolute in a file URI. A path starting with a slash
+// is absolute there on every platform, including Windows, where filepath.IsAbs wants a
+// drive letter.
+func isAbsolutePath(p string) bool {
+	return strings.HasPrefix(p, "/") || filepath.IsAbs(p)
+}
+
+// cutPrefixFold is strings.CutPrefix with the prefix matched without regard to case.
+func cutPrefixFold(s, prefix string) (string, bool) {
+	if len(s) < len(prefix) || !strings.EqualFold(s[:len(prefix)], prefix) {
+		return s, false
+	}
+	return s[len(prefix):], true
+}
+
 // Module is the library to load, DefaultModule when the URI names none.
 func (u *URI) Module() string {
 	if u.ModulePath == "" {
@@ -99,8 +149,7 @@ func (u *URI) PIN() ([]byte, error) {
 	if u.pinSource == "" {
 		return nil, nil
 	}
-	path := strings.TrimPrefix(strings.TrimPrefix(u.pinSource, "file://"), "file:")
-	pin, err := os.ReadFile(path)
+	pin, err := os.ReadFile(u.pinSource)
 	if err != nil {
 		return nil, fmt.Errorf("read PIN: %w", err)
 	}

client/internal/pkcs11/uri_test.go

@@ -84,6 +84,10 @@ func TestParseURI_Rejections(t *testing.T) {
 		"module-name with dotdot":  "pkcs11:token=a?module-name=..",
 		"empty module-name":        "pkcs11:token=a?module-name=",
 		"pin-value and pin-source": "pkcs11:token=a?pin-value=1&pin-source=file:/etc/pin",
+		// Without a value the URI would silently mean "no PIN" and skip the login.
+		"empty pin-source":    "pkcs11:token=a?pin-source=",
+		"remote pin-source":   "pkcs11:token=a?pin-source=file://example.com/etc/pin",
+		"relative pin-source": "pkcs11:token=a?pin-source=file:pin",
 	}
 	for name, raw := range tests {
 		_, err := ParseURI(raw)
@@ -91,6 +95,31 @@ func TestParseURI_Rejections(t *testing.T) {
 	}
 }
 
+// TestParseURI_SchemeIsCaseInsensitive: RFC 3986 section 3.1 makes schemes case-insensitive.
+func TestParseURI_SchemeIsCaseInsensitive(t *testing.T) {
+	for _, raw := range []string{"PKCS11:token=netbird", "Pkcs11:token=netbird"} {
+		uri, err := ParseURI(raw)
+		require.NoError(t, err, raw)
+		assert.Equal(t, "netbird", uri.Token, raw)
+	}
+}
+
+func TestURI_PINFromLocalhostFileURI(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("file URIs with a host name are written differently on Windows")
+	}
+	path := filepath.Join(t.TempDir(), "pin")
+	require.NoError(t, os.WriteFile(path, []byte("secret\n"), 0o600))
+
+	for _, host := range []string{"localhost", "LocalHost"} {
+		uri, err := ParseURI("pkcs11:token=netbird?pin-source=file://" + host + path)
+		require.NoError(t, err, host)
+		pin, err := uri.PIN()
+		require.NoError(t, err, host)
+		assert.Equal(t, []byte("secret"), pin, "RFC 8089 names the local machine as localhost")
+	}
+}
+
 func TestURI_PINFromFile(t *testing.T) {
 	path := filepath.Join(t.TempDir(), "pin")
 	require.NoError(t, os.WriteFile(path, []byte("secret\n"), 0o600))
@@ -117,3 +146,30 @@ func absModule(name string) string {
 	}
 	return "/lib/" + name
 }
+
+// TestPinSourcePath covers how a pin-source value becomes a file path, without touching the
+// file system, so the same cases run on every platform. A path starting with a slash is
+// absolute in a file URI everywhere, including Windows, where filepath.IsAbs would refuse it.
+func TestPinSourcePath(t *testing.T) {
+	tests := []struct {
+		value string
+		want  string
+	}{
+		{"/etc/netbird/pkcs11.pin", "/etc/netbird/pkcs11.pin"},
+		{"file:/etc/netbird/pkcs11.pin", "/etc/netbird/pkcs11.pin"},
+		{"file:///etc/netbird/pkcs11.pin", "/etc/netbird/pkcs11.pin"},
+		{"FILE:///etc/netbird/pkcs11.pin", "/etc/netbird/pkcs11.pin"},
+		{"file://localhost/etc/netbird/pkcs11.pin", "/etc/netbird/pkcs11.pin"},
+		{"file://LocalHost/etc/netbird/pkcs11.pin", "/etc/netbird/pkcs11.pin"},
+	}
+	for _, tt := range tests {
+		got, err := pinSourcePath(tt.value)
+		require.NoError(t, err, tt.value)
+		assert.Equal(t, tt.want, got, tt.value)
+	}
+
+	for _, value := range []string{"", "file:pin", "file://example.com/etc/pin", "file://localhost", "file://localhost/"} {
+		_, err := pinSourcePath(value)
+		assert.Error(t, err, "%q must be refused", value)
+	}
+}

client/internal/routemanager/client/client.go

@@ -65,6 +65,12 @@ const (
 	// well above any usable path so that slow paths keep their order.
 	unknownLatency = 10 * time.Second
 
+	// unmeasuredLatencyHold is how long a current routing peer without a
+	// latency sample is kept against a measured one. A direct connection is
+	// sampled within seconds, so a peer still unsampled after this is ranked
+	// at unknownLatency rather than held indefinitely.
+	unmeasuredLatencyHold = 30 * time.Second
+
 	// holdReportInterval is how long a routing peer held back by a switch rule
 	// is reported at debug level only, after being reported at info level once.
 	holdReportInterval = 10 * time.Minute
@@ -99,6 +105,7 @@ type holdKind int
 const (
 	holdLatencyMargin holdKind = iota
 	holdLatencyDwell
+	holdLatencyUnmeasured
 	holdRelayedDelay
 )
 
@@ -124,6 +131,9 @@ type routeCandidate struct {
 	relayed   bool
 	latency   time.Duration
 	noise     time.Duration
+	// measured is false while the peer has no latency sample, as right after a
+	// direct connection comes up: latency then holds unknownLatency
+	measured bool
 }
 
 func newRouteCandidate(r *route.Route, status routerPeerStatus) routeCandidate {
@@ -140,6 +150,7 @@ func newRouteCandidate(r *route.Route, status routerPeerStatus) routeCandidate {
 		relayed:   status.relayed,
 		latency:   latency,
 		noise:     status.noise,
+		measured:  status.latency > 0,
 	}
 }
 
@@ -230,6 +241,9 @@ type WatcherConfig struct {
 	StatusRecorder   *peer.Status
 	Route            *route.Route
 	Handler          RouteHandler
+	// ReconcileAllowedIPs re-applies the routed allowed IPs tracked for a
+	// routing peer onto the interface. Optional.
+	ReconcileAllowedIPs func(peerKey string) error
 }
 
 // Watcher watches route and peer changes and updates allowed IPs accordingly.
@@ -271,11 +285,20 @@ type Watcher struct {
 	trackedCurrent route.ID
 	trackedPeer    string
 	trackedDirect  bool
-	// degradedSince is when the current routing peer dropped from direct to
-	// relayed, zero if it is direct or was already relayed when chosen
+	// degradedSince is when the relayedSwitch delay of the current relayed
+	// routing peer started: its drop from direct, or the first direct contender
+	// if it was never direct. Zero while it is direct or uncontended.
 	degradedSince time.Time
-	// relayedHold fires when a relayed switch being held back becomes due, so it
-	// is applied on time rather than on the next periodic re-evaluation
+	// unmeasuredRoute and unmeasuredSince track since when the current route's
+	// peer has had no latency sample, bounding the hold in shouldSwitchRoute
+	unmeasuredRoute route.ID
+	unmeasuredSince time.Time
+	// neverDirect marks a degradedSince started by a direct contender rather
+	// than by a drop from direct, which the "never" policy does not cover
+	neverDirect bool
+	// relayedHold fires when a switch held back by the relayed delay or the
+	// unmeasured latency hold becomes due, so it is applied on time rather than
+	// on the next periodic re-evaluation
 	relayedHold *time.Timer
 	// heldCandidate and heldKind identify the alternative last held back by the
 	// switch rules and the rule holding it, and heldReported is when that was
@@ -298,6 +321,9 @@ type Watcher struct {
 	stopped      bool
 	handler      RouteHandler
 	updateSerial uint64
+	// reconcileAllowedIPs re-applies a routing peer's routed allowed IPs when
+	// it connects while keeping the route, see keepChosenRoute. May be nil.
+	reconcileAllowedIPs func(peerKey string) error
 }
 
 func newPeerStateUpdate() chan struct{} {
@@ -320,6 +346,7 @@ func NewWatcher(config WatcherConfig) *Watcher {
 		handler:             config.Handler,
 		currentChosenStatus: nil,
 		relayedSwitch:       relayedSwitch(),
+		reconcileAllowedIPs: config.ReconcileAllowedIPs,
 	}
 	return client
 }
@@ -393,6 +420,8 @@ func (w *Watcher) getBestRouteFromStatuses(routePeerStatuses map[route.ID]router
 	}
 
 	w.trackRelayed(current, haveCurrent)
+	w.trackUnmeasured(current, haveCurrent)
+	w.trackContender(current, best, haveCurrent && haveBest)
 
 	if !haveBest {
 		w.reportUnassigned()
@@ -449,9 +478,8 @@ func (w *Watcher) reportUnassigned() {
 
 // trackRelayed records when the current routing peer dropped from a direct to a
 // relayed connection while carrying the route, which starts the relayedSwitch
-// delay. A peer that was never direct, because it was relayed when chosen or
-// came up over relay from idle, is not degraded: holding on to it would only
-// keep the route on the slower path.
+// delay at once. A peer that was never direct starts it later, when a direct
+// peer first contends for the route, see relayedSwitchDue.
 func (w *Watcher) trackRelayed(current routeCandidate, haveCurrent bool) {
 	if !haveCurrent {
 		w.trackedCurrent, w.trackedPeer, w.trackedDirect, w.degradedSince = "", "", false, time.Time{}
@@ -462,7 +490,7 @@ func (w *Watcher) trackRelayed(current routeCandidate, haveCurrent bool) {
 	case current.id != w.trackedCurrent, current.peer != w.trackedPeer, !current.relayed:
 		w.degradedSince = time.Time{}
 	case w.trackedDirect:
-		w.degradedSince = time.Now()
+		w.degradedSince, w.neverDirect = time.Now(), false
 		// a new degradation, reported as such even if an earlier one held
 		// back the same candidate
 		w.heldCandidate = ""
@@ -494,6 +522,20 @@ func (w *Watcher) shouldSwitchRoute(current, candidate routeCandidate) bool {
 		return true
 	}
 
+	// Without a sample on both sides there is no gain to weigh. A direct
+	// connection reports no latency until its first sample, so comparing it as
+	// unknownLatency would move the route away from a peer that just recovered.
+	// The hold is bounded so a peer that never gets a sample does not keep the
+	// route for good.
+	if w.unmeasuredHoldActive(current, candidate) {
+		if level := w.holdLevel(candidate, holdLatencyUnmeasured); log.IsLevelEnabled(level) {
+			log.StandardLogger().Logf(level,
+				"keeping routing peer %s for [%v] over %s: latency not measured on both peers yet (%s)",
+				current.peer, w.handler, candidate.peer, candidate)
+		}
+		return false
+	}
+
 	gain := current.latency - candidate.latency
 	if margin := switchMargin(current, candidate); gain < margin {
 		if level := w.holdLevel(candidate, holdLatencyMargin); log.IsLevelEnabled(level) {
@@ -518,22 +560,75 @@ func (w *Watcher) shouldSwitchRoute(current, candidate routeCandidate) bool {
 	return true
 }
 
+// trackContender restarts the relayed delay of a peer that was never direct
+// once no direct contender is left. That delay runs from the first contender,
+// so keeping it would let a later contender take over without waiting. The
+// delay of a peer that dropped from direct keeps running from the drop.
+func (w *Watcher) trackContender(current, best routeCandidate, have bool) {
+	if !w.neverDirect || w.degradedSince.IsZero() {
+		return
+	}
+	if have && best.winsOnlyOnRelay(current) {
+		return
+	}
+	w.degradedSince = time.Time{}
+}
+
+// trackUnmeasured records since when the current route's peer has had no
+// latency sample, and forgets it once the peer is measured or the route moves.
+func (w *Watcher) trackUnmeasured(current routeCandidate, haveCurrent bool) {
+	if !haveCurrent || current.measured || current.id != w.unmeasuredRoute {
+		w.unmeasuredRoute, w.unmeasuredSince = "", time.Time{}
+	}
+	if haveCurrent && !current.measured && w.unmeasuredSince.IsZero() {
+		w.unmeasuredRoute, w.unmeasuredSince = current.id, time.Now()
+	}
+}
+
+// unmeasuredHoldActive reports whether a latency comparison must wait because
+// a side has no sample: always for an unsampled candidate, and for an unsampled
+// current peer until unmeasuredLatencyHold runs out. A pending hold arms
+// relayedHold so the switch is reconsidered when it runs out.
+func (w *Watcher) unmeasuredHoldActive(current, candidate routeCandidate) bool {
+	if !candidate.measured {
+		return true
+	}
+	if current.measured {
+		return false
+	}
+
+	remaining := unmeasuredLatencyHold - time.Since(w.unmeasuredSince)
+	if remaining <= 0 {
+		return false
+	}
+	w.armRelayedHold(remaining)
+
+	return true
+}
+
 // relayedSwitchDue reports whether a relayed current routing peer should give
 // way to a candidate that is better only by being direct.
 //
-// A peer that dropped from direct to relayed while carrying the route is kept
-// for the relayedSwitch delay: the relay still carries the traffic, staying
-// keeps every flow through this routing peer alive, and a direct connection that
-// dropped often recovers within seconds. A peer that was already relayed when
-// chosen gives way at once, so the route is not stuck on the slower path when a
-// direct one becomes available. A switch that is not due yet arms relayedHold so
-// it is applied as soon as it becomes due.
+// A relayed peer is kept for the relayedSwitch delay: the relay still carries
+// the traffic, and staying keeps every flow through this routing peer alive.
+// For a peer that dropped from direct to relayed the delay runs from the drop,
+// since a direct connection that dropped often recovers within seconds. For a
+// peer that was never direct, as when the peers of a group come up over relay
+// before their direct connections, it runs from the first evaluation that
+// finds a direct contender: the current peer usually turns direct within
+// seconds too. A switch that is not due yet arms relayedHold so it is applied
+// as soon as it becomes due.
 func (w *Watcher) relayedSwitchDue(current, candidate routeCandidate) bool {
 	if w.degradedSince.IsZero() {
-		return true
+		w.degradedSince, w.neverDirect = time.Now(), true
 	}
 
 	policy := w.relayedSwitch
+	// "never" keeps a peer whose direct connection dropped; one that was never
+	// direct still moves after the default delay instead of staying relayed
+	if policy.never && w.neverDirect {
+		policy = relayedSwitchPolicy{delay: defaultRelayedSwitchDelay}
+	}
 	if !policy.never {
 		remaining := policy.delay - time.Since(w.degradedSince)
 		if remaining <= 0 {
@@ -544,7 +639,7 @@ func (w *Watcher) relayedSwitchDue(current, candidate routeCandidate) bool {
 
 	if level := w.holdLevel(candidate, holdRelayedDelay); log.IsLevelEnabled(level) {
 		log.StandardLogger().Logf(level,
-			"keeping relayed routing peer %s for [%v] over direct %s: dropped to relay %v ago, switch delay %s",
+			"keeping relayed routing peer %s for [%v] over direct %s: held for %v, switch delay %s",
 			current.peer, w.handler, candidate.peer, time.Since(w.degradedSince).Round(time.Second), policy)
 	}
 
@@ -677,21 +772,46 @@ func (w *Watcher) removeAllowedIPs(route *route.Route, rsn reason) error {
 	return nil
 }
 
-// shouldSkipRecalculation checks if we can skip route recalculation for the same route without status changes
-func (w *Watcher) shouldSkipRecalculation(newChosenID route.ID, newStatus routerPeerStatus) bool {
-	if w.currentChosen == nil {
+// keepChosenRoute reports whether the evaluation keeps the current route, with
+// the same definition, on the same routing peer. A status change of that peer
+// is recorded without removing and re-adding its allowed IPs, which would leave
+// the prefix without an owner in between. When the peer connects, its routed
+// allowed IPs are re-applied instead: a peer recreated by a peer update comes
+// back with its overlay addresses only.
+func (w *Watcher) keepChosenRoute(newChosenID route.ID, newStatus routerPeerStatus) bool {
+	if w.currentChosen == nil || w.currentChosen.ID != newChosenID {
 		return false
 	}
 
-	isSameRoute := w.currentChosen.ID == newChosenID && w.currentChosen.Equal(w.routes[newChosenID])
-	if !isSameRoute {
+	newChosenRoute := w.routes[newChosenID]
+	if !w.currentChosen.Equal(newChosenRoute) {
 		return false
 	}
 
-	if w.currentChosenStatus != nil {
-		return w.currentChosenStatus.status == newStatus.status
+	if w.currentChosenStatus == nil || w.currentChosenStatus.status == newStatus.status {
+		return true
 	}
 
+	// a failed reconcile leaves the previous status recorded, so the next
+	// evaluation sees the change again and retries
+	if newStatus.status == peer.StatusConnected && !w.reconcileRoutedAllowedIPs(newChosenRoute.Peer) {
+		return true
+	}
+
+	w.currentChosen = newChosenRoute
+	w.currentChosenStatus = &newStatus
+
+	return true
+}
+
+func (w *Watcher) reconcileRoutedAllowedIPs(peerKey string) bool {
+	if w.reconcileAllowedIPs == nil {
+		return true
+	}
+	if err := w.reconcileAllowedIPs(peerKey); err != nil {
+		log.Warnf("failed to reconcile routed allowed IPs of routing peer %s for [%v], will retry: %v", peerKey, w.handler, err)
+		return false
+	}
 	return true
 }
 
@@ -719,8 +839,7 @@ func (w *Watcher) recalculateRoutes(rsn reason, routerPeerStatuses map[route.ID]
 		return nil
 	}
 
-	// If we can skip recalculation for the same route without changes, do nothing
-	if w.shouldSkipRecalculation(newChosenID, newStatus) {
+	if w.keepChosenRoute(newChosenID, newStatus) {
 		return nil
 	}
 

client/internal/routemanager/client/client_test.go

@@ -2,6 +2,7 @@ package client
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"net/netip"
 	"sync"
@@ -1480,3 +1481,84 @@ func TestLatencyJitterDoesNotSwitchRoute(t *testing.T) {
 		require.Equal(t, route.ID("route1"), chosen, "route changed on jitter at sample %d", i)
 	}
 }
+
+// A routing peer that keeps the route while only its connection status changes,
+// as when an idle peer is woken by traffic, must keep its allowed IPs in place:
+// removing and re-adding them leaves the prefix without an owner in between.
+func TestRecalculateRoutesStatusChangeKeepsAllowedIPs(t *testing.T) {
+	w, handler := newRecalcWatcher(t)
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusIdle},
+	}))
+	require.Equal(t, []string{"peer1"}, handler.added)
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected, relayed: true},
+	}))
+	assert.Equal(t, []string{"peer1"}, handler.added, "a status change must not re-add the allowed IPs")
+	assert.Equal(t, 0, handler.removed, "a status change must not remove the allowed IPs")
+	require.NotNil(t, w.currentChosenStatus)
+	assert.Equal(t, peer.StatusConnected, w.currentChosenStatus.status, "the new status must be recorded")
+}
+
+// A routing peer recreated by a peer update comes back idle with its overlay
+// addresses only. When it connects while keeping the route, its routed allowed
+// IPs must be re-applied, since nothing else puts them back on the new peer.
+func TestRecalculateRoutesReconcilesAllowedIPsOnConnect(t *testing.T) {
+	w, handler := newRecalcWatcher(t)
+	var reconciled []string
+	w.reconcileAllowedIPs = func(peerKey string) error {
+		reconciled = append(reconciled, peerKey)
+		return nil
+	}
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected},
+	}))
+	require.Equal(t, []string{"peer1"}, handler.added)
+	require.Empty(t, reconciled, "a fresh assignment adds the allowed IPs itself")
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusIdle},
+	}))
+	assert.Empty(t, reconciled, "going idle must not reconcile")
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected},
+	}))
+	assert.Equal(t, []string{"peer1"}, reconciled, "connecting while keeping the route must reconcile its allowed IPs")
+	assert.Equal(t, []string{"peer1"}, handler.added, "the reconcile must not go through the refcounted add")
+	assert.Equal(t, 0, handler.removed, "the allowed IPs must not be removed")
+}
+
+// A failed reconcile must be retried: the new status is only recorded once the
+// routed allowed IPs are back, so the next evaluation tries again.
+func TestRecalculateRoutesRetriesFailedReconcile(t *testing.T) {
+	w, _ := newRecalcWatcher(t)
+	calls := 0
+	w.reconcileAllowedIPs = func(string) error {
+		calls++
+		if calls == 1 {
+			return errors.New("add allowed IP: transient failure")
+		}
+		return nil
+	}
+
+	connected := map[route.ID]routerPeerStatus{"route1": {status: peer.StatusConnected}}
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, connected))
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusIdle},
+	}))
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, connected))
+	require.Equal(t, 1, calls, "connecting must reconcile")
+	assert.Equal(t, peer.StatusIdle, w.currentChosenStatus.status, "a failed reconcile must not record the new status")
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, connected))
+	assert.Equal(t, 2, calls, "the next evaluation must retry the reconcile")
+	assert.Equal(t, peer.StatusConnected, w.currentChosenStatus.status, "a successful reconcile records the status")
+
+	require.NoError(t, w.recalculateRoutes(reasonPeerUpdate, connected))
+	assert.Equal(t, 2, calls, "nothing to retry once the reconcile succeeded")
+}

client/internal/routemanager/client/env.go

@@ -12,10 +12,11 @@ import (
 const (
 	// envRelayedSwitchDelay sets how long a routing peer that dropped from a
 	// direct to a relayed connection while carrying the route is kept before the
-	// route moves to an equivalent peer with a direct connection. Takes a
-	// duration ("30s", "0" to switch at once) or "never" to only leave such a
-	// peer when it disconnects. A routing peer that was already relayed when
-	// chosen always gives way to a direct one at once.
+	// route moves to an equivalent peer with a direct connection. A routing peer
+	// that was never direct is kept for the same delay from when a direct peer
+	// first contends for the route. Takes a duration ("30s", "0" to switch at
+	// once) or "never" to only leave a peer that dropped from direct when it
+	// disconnects; a peer that was never direct then uses the default delay.
 	envRelayedSwitchDelay = "NB_HA_RELAYED_SWITCH_DELAY"
 
 	relayedSwitchDelayNever = "never"

client/internal/routemanager/client/relayed_switch_test.go

@@ -103,11 +103,11 @@ func TestDegradedRoutingPeerIsReplacedAfterDelay(t *testing.T) {
 	assert.Equal(t, switchReasonDirect, w.switchReason, "the switch reason must name the relayed connection")
 }
 
-// A routing peer that was relayed when it was chosen is not degraded: when an
-// equivalent direct peer shows up the route must upgrade at once instead of
-// staying on the slower path for the delay.
-func TestRelayedRoutingPeerUpgradesAtOnce(t *testing.T) {
-	w := newRelayedTestWatcher(t, relayedSwitchPolicy{never: true})
+// A routing peer that was relayed when it was chosen is held for the delay when
+// an equivalent direct peer shows up, from that moment on: it usually turns
+// direct within seconds, and moving the route resets its masqueraded flows.
+func TestRelayedRoutingPeerIsKeptDuringDelay(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{delay: 30 * time.Second})
 
 	// route1 was chosen while relayed and route2 was still connecting
 	chosen, _ := w.getBestRouteFromStatuses(map[route.ID]routerPeerStatus{
@@ -115,36 +115,57 @@ func TestRelayedRoutingPeerUpgradesAtOnce(t *testing.T) {
 		"route2": {status: peer.StatusConnecting},
 	})
 	require.Equal(t, route.ID("route1"), chosen)
-	require.True(t, w.degradedSince.IsZero(), "a peer relayed from the start is not degraded")
+	require.True(t, w.degradedSince.IsZero(), "the delay must not run before a direct peer contends")
 
 	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
-	assert.Equal(t, route.ID("route2"), chosen, "a direct peer must take over from one that was relayed all along")
-	assert.Equal(t, switchReasonDirect, w.switchReason)
-	assert.Nil(t, w.relayedHold, "an upgrade must not schedule anything")
+	assert.Equal(t, route.ID("route1"), chosen, "a relayed peer must be kept when a direct one shows up")
+	assert.False(t, w.degradedSince.IsZero(), "the direct contender must start the delay")
+	assert.NotNil(t, w.relayedHold, "the pending switch must arm a timer so it is applied when due")
+
+	w.degradedSince = w.degradedSince.Add(-31 * time.Second)
+	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
+	assert.Equal(t, route.ID("route2"), chosen, "a peer still relayed after the delay must give way to a direct one")
+	assert.Equal(t, switchReasonDirect, w.switchReason, "the switch reason must name the relayed connection")
 }
 
-// An idle routing peer woken by traffic whose relay connection comes up before
-// the direct one was never direct: it is not degraded, and must give way to an
-// equivalent direct peer at once.
-func TestIdleRoutingPeerWokenOverRelayUpgradesAtOnce(t *testing.T) {
-	w := newRelayedTestWatcher(t, relayedSwitchPolicy{never: true})
+// The peers of an HA group woken from idle come up over relay first and turn
+// direct one after the other. The route must stay on the peer it was on when
+// the other one turns direct first and the current one follows within the
+// delay, including before the current peer's first latency sample.
+func TestIdleRoutingPeersWokenOverRelayKeepRoute(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{delay: 30 * time.Second})
 
-	chosen, _ := w.getBestRouteFromStatuses(map[route.ID]routerPeerStatus{
-		"route1": {status: peer.StatusIdle},
-		"route2": {status: peer.StatusIdle},
-	})
-	require.Equal(t, route.ID("route1"), chosen)
+	steps := []struct {
+		name     string
+		statuses map[route.ID]routerPeerStatus
+	}{
+		{"both idle", map[route.ID]routerPeerStatus{
+			"route1": {status: peer.StatusIdle},
+			"route2": {status: peer.StatusIdle},
+		}},
+		{"both relayed", map[route.ID]routerPeerStatus{
+			"route1": {status: peer.StatusConnected, relayed: true},
+			"route2": {status: peer.StatusConnected, relayed: true},
+		}},
+		{"other direct first", map[route.ID]routerPeerStatus{
+			"route1": {status: peer.StatusConnected, relayed: true},
+			"route2": {status: peer.StatusConnected, latency: 20 * time.Millisecond},
+		}},
+		{"current direct, not sampled", map[route.ID]routerPeerStatus{
+			"route1": {status: peer.StatusConnected},
+			"route2": {status: peer.StatusConnected, latency: 20 * time.Millisecond},
+		}},
+		{"current direct, sampled", map[route.ID]routerPeerStatus{
+			"route1": {status: peer.StatusConnected, latency: 38 * time.Millisecond},
+			"route2": {status: peer.StatusConnected, latency: 20 * time.Millisecond},
+		}},
+	}
 
-	chosen, _ = w.getBestRouteFromStatuses(map[route.ID]routerPeerStatus{
-		"route1": {status: peer.StatusConnected, relayed: true},
-		"route2": {status: peer.StatusConnecting},
-	})
-	require.Equal(t, route.ID("route1"), chosen)
-	require.True(t, w.degradedSince.IsZero(), "waking up over relay is not a drop from direct")
-
-	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
-	assert.Equal(t, route.ID("route2"), chosen, "a direct peer must take over from one that was never direct")
-	assert.Nil(t, w.relayedHold, "an upgrade must not schedule anything")
+	for _, step := range steps {
+		chosen, _ := w.getBestRouteFromStatuses(step.statuses)
+		require.Equal(t, route.ID("route1"), chosen, "step %q must keep the route on route1", step.name)
+	}
+	assert.True(t, w.degradedSince.IsZero(), "the delay must end once the current peer is direct")
 }
 
 // A routing peer that drops to relay again, after recovering from an earlier
@@ -356,3 +377,128 @@ func TestRoutingPeerSwitchPublishesEvent(t *testing.T) {
 	assert.Equal(t, "No routing peer available", events[1].Message)
 	assert.Equal(t, "peer2", events[1].Metadata["previous_peer"])
 }
+
+// The statuses below follow the status recorder: a drop to relay clears the
+// latency, and a recovered direct connection reports no latency until the new
+// ICE agent's first sample lands.
+var (
+	flapDirect = map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected, latency: 50 * time.Millisecond},
+		"route2": {status: peer.StatusConnected, latency: 60 * time.Millisecond},
+	}
+	flapRelayed = map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected, relayed: true},
+		"route2": {status: peer.StatusConnected, latency: 60 * time.Millisecond},
+	}
+	flapRecoveredUnsampled = map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected},
+		"route2": {status: peer.StatusConnected, latency: 60 * time.Millisecond},
+	}
+)
+
+// A routing peer whose direct connection drops to relay and comes back within
+// the delay must keep the route, including the window between the recovery and
+// the first latency sample of the new connection.
+func TestRecoveredRoutingPeerWithoutLatencySampleIsKept(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{delay: 30 * time.Second})
+
+	chosen, _ := w.getBestRouteFromStatuses(flapDirect)
+	require.Equal(t, route.ID("route1"), chosen)
+
+	chosen, _ = w.getBestRouteFromStatuses(flapRelayed)
+	require.Equal(t, route.ID("route1"), chosen, "a routing peer that just dropped to relay must be kept")
+
+	chosen, _ = w.getBestRouteFromStatuses(flapRecoveredUnsampled)
+	assert.Equal(t, route.ID("route1"), chosen, "a recovered routing peer must not lose the route before its latency is sampled")
+}
+
+// A routing peer that keeps dropping to relay for less than the delay, and
+// staying direct for less than the delay in between, must keep the route on
+// every cycle: each drop is a new degradation with its own delay.
+func TestFlappingRoutingPeerIsKept(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{delay: 30 * time.Second})
+
+	chosen, _ := w.getBestRouteFromStatuses(flapDirect)
+	require.Equal(t, route.ID("route1"), chosen)
+
+	for cycle := 0; cycle < 3; cycle++ {
+		chosen, _ = w.getBestRouteFromStatuses(flapRelayed)
+		require.Equal(t, route.ID("route1"), chosen, "cycle %d: drop to relay", cycle)
+
+		// relayed for 25s, still within the delay
+		w.degradedSince = w.degradedSince.Add(-25 * time.Second)
+		chosen, _ = w.getBestRouteFromStatuses(flapRelayed)
+		require.Equal(t, route.ID("route1"), chosen, "cycle %d: relayed within the delay", cycle)
+
+		chosen, _ = w.getBestRouteFromStatuses(flapRecoveredUnsampled)
+		require.Equal(t, route.ID("route1"), chosen, "cycle %d: recovered, latency not sampled yet", cycle)
+
+		chosen, _ = w.getBestRouteFromStatuses(flapDirect)
+		require.Equal(t, route.ID("route1"), chosen, "cycle %d: recovered and sampled", cycle)
+	}
+}
+
+// "never" only keeps a routing peer that dropped from direct. A peer that was
+// never direct must still move to a direct one after the default delay, so the
+// route is not left on a relay for good.
+func TestNeverPolicyDoesNotPinNeverDirectPeer(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{never: true})
+
+	chosen, _ := w.getBestRouteFromStatuses(map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected, relayed: true},
+		"route2": {status: peer.StatusConnecting},
+	})
+	require.Equal(t, route.ID("route1"), chosen)
+
+	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
+	require.Equal(t, route.ID("route1"), chosen, "a never-direct peer must be held for the default delay")
+	require.NotNil(t, w.relayedHold, "the default delay must arm the timer")
+
+	w.degradedSince = w.degradedSince.Add(-defaultRelayedSwitchDelay - time.Second)
+	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
+	assert.Equal(t, route.ID("route2"), chosen, "with never, a never-direct peer must still give way after the default delay")
+}
+
+// A current peer that never gets a latency sample must not keep the route
+// forever against a measured peer: once the hold runs out it is ranked as
+// unknown latency again.
+func TestUnmeasuredRoutingPeerHoldIsBounded(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{delay: 30 * time.Second})
+
+	chosen, _ := w.getBestRouteFromStatuses(flapRecoveredUnsampled)
+	require.Equal(t, route.ID("route1"), chosen, "an unsampled current peer must be held at first")
+	require.False(t, w.unmeasuredSince.IsZero(), "the hold must record when it started")
+
+	w.unmeasuredSince = w.unmeasuredSince.Add(-unmeasuredLatencyHold - time.Second)
+	chosen, _ = w.getBestRouteFromStatuses(flapRecoveredUnsampled)
+	assert.Equal(t, route.ID("route2"), chosen, "a peer still unsampled after the hold must give way to a measured one")
+	assert.Equal(t, switchReasonLatency, w.switchReason)
+}
+
+// The delay of a peer that was never direct runs from the first direct
+// contender. When that contender goes away the delay must start over, so a
+// later contender is held for the full delay instead of taking over at once.
+func TestNeverDirectDelayRestartsWhenContenderGoes(t *testing.T) {
+	w := newRelayedTestWatcher(t, relayedSwitchPolicy{delay: 30 * time.Second})
+
+	relayedAlone := map[route.ID]routerPeerStatus{
+		"route1": {status: peer.StatusConnected, relayed: true},
+		"route2": {status: peer.StatusConnecting},
+	}
+	chosen, _ := w.getBestRouteFromStatuses(relayedAlone)
+	require.Equal(t, route.ID("route1"), chosen)
+
+	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
+	require.Equal(t, route.ID("route1"), chosen, "the first direct contender must be held")
+	require.False(t, w.degradedSince.IsZero())
+
+	// the first contender was seen longer ago than the delay
+	w.degradedSince = w.degradedSince.Add(-31 * time.Second)
+
+	chosen, _ = w.getBestRouteFromStatuses(relayedAlone)
+	require.Equal(t, route.ID("route1"), chosen)
+	assert.True(t, w.degradedSince.IsZero(), "the delay must start over once no direct contender is left")
+
+	chosen, _ = w.getBestRouteFromStatuses(currentRelayed)
+	assert.Equal(t, route.ID("route1"), chosen, "a new direct contender must be held for the full delay")
+}

client/internal/routemanager/manager.go

@@ -578,12 +578,13 @@ func (m *DefaultManager) TriggerSelection(networks route.HAMap) {
 		}
 
 		config := client.WatcherConfig{
-			Context:          m.ctx,
-			DNSRouteInterval: m.dnsRouteInterval,
-			WGInterface:      m.wgInterface,
-			StatusRecorder:   m.statusRecorder,
-			Route:            routes[0],
-			Handler:          handler,
+			Context:             m.ctx,
+			DNSRouteInterval:    m.dnsRouteInterval,
+			WGInterface:         m.wgInterface,
+			StatusRecorder:      m.statusRecorder,
+			Route:               routes[0],
+			Handler:             handler,
+			ReconcileAllowedIPs: m.ReconcilePeerAllowedIPs,
 		}
 		clientNetworkWatcher := client.NewWatcher(config)
 		m.clientNetworks[id] = clientNetworkWatcher
@@ -628,12 +629,13 @@ func (m *DefaultManager) updateClientNetworks(updateSerial uint64, networks rout
 			}
 
 			config := client.WatcherConfig{
-				Context:          m.ctx,
-				DNSRouteInterval: m.dnsRouteInterval,
-				WGInterface:      m.wgInterface,
-				StatusRecorder:   m.statusRecorder,
-				Route:            routes[0],
-				Handler:          handler,
+				Context:             m.ctx,
+				DNSRouteInterval:    m.dnsRouteInterval,
+				WGInterface:         m.wgInterface,
+				StatusRecorder:      m.statusRecorder,
+				Route:               routes[0],
+				Handler:             handler,
+				ReconcileAllowedIPs: m.ReconcilePeerAllowedIPs,
 			}
 			clientNetworkWatcher = client.NewWatcher(config)
 			m.clientNetworks[id] = clientNetworkWatcher

client/internal/stdnet/wgdetector.go

@@ -34,6 +34,8 @@ type WGDetector struct {
 	// probe is replaced in tests; it is the call this type exists to avoid repeating.
 	probe func(string) bool
 
+	ownIface string
+
 	mu    sync.RWMutex
 	cache map[string]wgDetectorEntry
 
@@ -41,11 +43,12 @@ type WGDetector struct {
 }
 
 // NewWGDetector returns a detector with the default time to live.
-func NewWGDetector() *WGDetector {
+func NewWGDetector(ownIface string) *WGDetector {
 	return &WGDetector{
-		ttl:   wgDetectorTTL,
-		probe: probeWireGuard,
-		cache: make(map[string]wgDetectorEntry),
+		ttl:      wgDetectorTTL,
+		probe:    probeWireGuard,
+		ownIface: ownIface,
+		cache:    make(map[string]wgDetectorEntry),
 	}
 }
 
@@ -57,6 +60,10 @@ func (d *WGDetector) IsWireGuard(iFace string) bool {
 		return probeWireGuard(iFace)
 	}
 
+	if d.isOwnInterface(iFace) {
+		return true
+	}
+
 	if isWireGuard, ok := d.cached(iFace); ok {
 		return isWireGuard
 	}
@@ -104,6 +111,10 @@ func (d *WGDetector) store(iFace string, isWireGuard bool) {
 	d.cache[iFace] = wgDetectorEntry{isWireGuard: isWireGuard, expireAt: now.Add(d.ttl)}
 }
 
+func (d *WGDetector) isOwnInterface(iFace string) bool {
+	return d.ownIface != "" && iFace == d.ownIface
+}
+
 func probeWireGuard(iFace string) bool {
 	wg, err := wgctrl.New()
 	if err != nil {

client/internal/stdnet/wgdetector_test.go

@@ -150,3 +150,39 @@ func TestWGDetectorDropsExpiredEntries(t *testing.T) {
 	assert.Len(t, d.cache, 1, "expired entries of vanished interfaces must be dropped")
 	assert.Contains(t, d.cache, "eth0", "the fresh answer must be kept")
 }
+
+func TestWGDetectorOwnInterfaceIsWireGuardWithoutProbing(t *testing.T) {
+	d, calls := newCountingDetector(t, time.Minute, false)
+	d.ownIface = "NetBird"
+
+	assert.True(t, d.IsWireGuard("NetBird"), "the agent's own interface is a WireGuard device")
+	assert.Equal(t, int64(0), calls.Load(), "the own interface must never reach the probe")
+
+	assert.False(t, d.IsWireGuard("NetBird2"), "only the exact own name is answered without probing")
+	assert.Equal(t, int64(1), calls.Load(), "a different interface is probed")
+}
+
+func TestWGDetectorWithoutOwnInterfaceProbesEmptyName(t *testing.T) {
+	d, calls := newCountingDetector(t, time.Minute, false)
+
+	assert.False(t, d.IsWireGuard(""), "an unnamed interface is not mistaken for the own one")
+	assert.Equal(t, int64(1), calls.Load(), "with no own interface set every name is probed")
+}
+
+func TestInterfaceFilterRejectsOwnInterfaceRegardlessOfDisallowList(t *testing.T) {
+	d, calls := newCountingDetector(t, time.Minute, false)
+	d.ownIface = "NetBird"
+	filter := InterfaceFilter([]string{"wt"}, d)
+
+	assert.False(t, filter("NetBird"), "the own interface is rejected even when the disallow list does not cover it")
+	assert.Equal(t, int64(0), calls.Load(), "the own interface is not probed")
+
+	assert.True(t, filter("eth0"), "an ordinary interface stays available")
+	assert.Equal(t, int64(1), calls.Load(), "an ordinary interface is still probed")
+}
+
+func TestNewWGDetectorKeepsTheOwnInterface(t *testing.T) {
+	d := NewWGDetector("NetbirdWT")
+
+	assert.True(t, d.IsWireGuard("NetbirdWT"), "the constructor records the own interface")
+}

management/internals/network_map_db/shared_types_test.go

@@ -4,6 +4,7 @@ import (
 	"testing"
 
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestRecordTypeAndRdata(t *testing.T) {
@@ -36,3 +37,24 @@ func TestRecordTypeAndRdata(t *testing.T) {
 		})
 	}
 }
+
+// TestConvertToNmdataPeers_CarriesTheVerifiedCertificates: the certificate posture check is
+// evaluated on the chains stored in the peer meta, so a peer loaded through this path with
+// its certificates dropped would silently fail every certificate check.
+func TestConvertToNmdataPeers_CarriesTheVerifiedCertificates(t *testing.T) {
+	peers, _, err := ConvertToNmdataPeers([]Peer{
+		{MetaCertificates: []byte(`["-----BEGIN CERTIFICATE-----\nleaf\n-----END CERTIFICATE-----\n","chain-2"]`)},
+		{},
+	})
+
+	require.NoError(t, err)
+	require.Len(t, peers, 2)
+	assert.Equal(t, []string{"-----BEGIN CERTIFICATE-----\nleaf\n-----END CERTIFICATE-----\n", "chain-2"}, peers[0].Meta.Certificates)
+	assert.Empty(t, peers[1].Meta.Certificates, "a peer that proved nothing has no certificates")
+}
+
+func TestConvertToNmdataPeers_RejectsMalformedCertificates(t *testing.T) {
+	_, _, err := ConvertToNmdataPeers([]Peer{{MetaCertificates: []byte(`{"not":"a list"}`)}})
+
+	assert.Error(t, err, "a stored value that is not a list of chains must not load as no certificates")
+}

management/internals/server/boot.go

@@ -204,7 +204,7 @@ func (s *BaseServer) IDPHandler() http.Handler {
 func (s *BaseServer) Router() *mux.Router {
 	return Create(s, func() *mux.Router {
 		router := mux.NewRouter()
-		router.Use(middleware.BuildMiddleware(s.RateLimiter(), s.AuthManager(), s.AccountManager(), s.Metrics(), s.IsValidChildAccount)...)
+		router.Use(middleware.BuildMiddleware(s.RateLimiter(), s.AuthManager(), s.AccountManager(), s.Metrics())...)
 		return router
 	})
 }

management/internals/server/modules.go

@@ -248,7 +248,3 @@ func (s *BaseServer) ReverseProxyDomainManager() *manager.Manager {
 		return &m
 	})
 }
-
-func (s *BaseServer) IsValidChildAccount(_ context.Context, _, _, _ string) bool {
-	return false
-}

management/internals/shared/grpc/peer_meta_test.go

@@ -0,0 +1,45 @@
+package grpc
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+
+	"github.com/netbirdio/netbird/shared/management/proto"
+)
+
+func TestExtractPeerMeta_RemovesNulAndInvalidUTF8(t *testing.T) {
+	meta := extractPeerMeta(context.Background(), &proto.PeerSystemMeta{
+		Hostname:        "host\x00name",
+		OS:              "\x00\x00\x00\x00\x00\x00s",
+		OSVersion:       "11\xff",
+		SysSerialNumber: "SN\x00\x00",
+		Environment:     &proto.Environment{Cloud: "\x00aws"},
+		NetworkAddresses: []*proto.NetworkAddress{
+			{NetIP: "192.168.0.1/24", Mac: "e0:d5\x00"},
+		},
+		Files: []*proto.File{{Path: "C:\\x\x00.exe"}},
+	})
+
+	assert.Equal(t, "hostname", meta.Hostname)
+	assert.Equal(t, "s", meta.OS)
+	assert.Equal(t, "11", meta.OSVersion)
+	assert.Equal(t, "SN", meta.SystemSerialNumber)
+	assert.Equal(t, "aws", meta.Environment.Cloud)
+	assert.Equal(t, "e0:d5", meta.NetworkAddresses[0].Mac)
+	assert.Equal(t, "C:\\x.exe", meta.Files[0].Path)
+}
+
+func TestCleanClientString(t *testing.T) {
+	tests := map[string]string{
+		"":                    "",
+		"\x00\x00":            "",
+		"Wïndows":             "Wïndows",
+		"ssh-ed25519 AAA\x00": "ssh-ed25519 AAA",
+		"bad\xc3\x28utf8":     "bad(utf8",
+	}
+	for in, want := range tests {
+		assert.Equal(t, want, cleanClientString(in), "input %q", in)
+	}
+}

management/internals/shared/grpc/proxy.go

@@ -152,6 +152,10 @@ type ProxyServiceServer struct {
 	// initial snapshot delivery. Configurable via NB_PROXY_SNAPSHOT_BATCH_SIZE.
 	snapshotBatchSize int
 
+	// snapshotAckTimeout bounds the wait for each snapshot batch ack.
+	// Defaults to defaultSnapshotAckTimeout when zero.
+	snapshotAckTimeout time.Duration
+
 	cancel context.CancelFunc
 }
 
@@ -168,6 +172,8 @@ const defaultProxyTokenTTL = 5 * time.Minute
 
 const defaultSnapshotBatchSize = 500
 
+const defaultSnapshotAckTimeout = 5 * time.Minute
+
 func snapshotBatchSizeFromEnv() int {
 	if v := os.Getenv("NB_PROXY_SNAPSHOT_BATCH_SIZE"); v != "" {
 		if n, err := strconv.Atoi(v); err == nil && n > 0 {
@@ -185,6 +191,14 @@ func (s *ProxyServiceServer) proxyTokenTTL() time.Duration {
 	return defaultProxyTokenTTL
 }
 
+// snapshotAckWait returns the configured snapshot ack timeout or the default when unset.
+func (s *ProxyServiceServer) snapshotAckWait() time.Duration {
+	if s.snapshotAckTimeout > 0 {
+		return s.snapshotAckTimeout
+	}
+	return defaultSnapshotAckTimeout
+}
+
 // proxyConnection represents a connected proxy
 type proxyConnection struct {
 	proxyID      string
@@ -484,7 +498,7 @@ func (s *ProxyServiceServer) SyncMappings(stream proto.ProxyService_SyncMappings
 		return err
 	}
 
-	if err := s.sendSnapshotSync(stream.Context(), conn, stream); err != nil {
+	if err := s.sendSnapshotSync(conn.ctx, conn, stream); err != nil {
 		s.cleanupFailedSnapshot(stream.Context(), conn)
 		return fmt.Errorf("send snapshot to proxy %s: %w", params.proxyID, err)
 	}
@@ -728,7 +742,7 @@ func (s *ProxyServiceServer) sendSnapshotSync(ctx context.Context, conn *proxyCo
 			return fmt.Errorf("send snapshot batch: %w", err)
 		}
 
-		if err := waitForAck(stream); err != nil {
+		if err := waitForAck(ctx, stream, s.snapshotAckWait()); err != nil {
 			return err
 		}
 	}
@@ -740,7 +754,7 @@ func (s *ProxyServiceServer) sendSnapshotSync(ctx context.Context, conn *proxyCo
 			return fmt.Errorf("send snapshot completion: %w", err)
 		}
 
-		if err := waitForAck(stream); err != nil {
+		if err := waitForAck(ctx, stream, s.snapshotAckWait()); err != nil {
 			return err
 		}
 	}
@@ -748,15 +762,38 @@ func (s *ProxyServiceServer) sendSnapshotSync(ctx context.Context, conn *proxyCo
 	return nil
 }
 
-func waitForAck(stream proto.ProxyService_SyncMappingsServer) error {
-	msg, err := stream.Recv()
-	if err != nil {
-		return fmt.Errorf("receive ack: %w", err)
+// waitForAck waits for the proxy to ack a snapshot batch. It gives up when ctx
+// is canceled or no ack arrives within timeout, since a proxy that abandoned the
+// stream without closing it would otherwise block the handler indefinitely.
+func waitForAck(ctx context.Context, stream proto.ProxyService_SyncMappingsServer, timeout time.Duration) error {
+	type recvResult struct {
+		msg *proto.SyncMappingsRequest
+		err error
 	}
-	if msg.GetAck() == nil {
-		return fmt.Errorf("expected ack, got %T", msg.GetMsg())
+	// Recv returns once the handler exits and gRPC tears the stream down.
+	result := make(chan recvResult, 1)
+	go func() {
+		msg, err := stream.Recv()
+		result <- recvResult{msg: msg, err: err}
+	}()
+
+	timer := time.NewTimer(timeout)
+	defer timer.Stop()
+
+	select {
+	case r := <-result:
+		if r.err != nil {
+			return fmt.Errorf("receive ack: %w", r.err)
+		}
+		if r.msg.GetAck() == nil {
+			return fmt.Errorf("expected ack, got %T", r.msg.GetMsg())
+		}
+		return nil
+	case <-ctx.Done():
+		return fmt.Errorf("wait for ack: %w", ctx.Err())
+	case <-timer.C:
+		return fmt.Errorf("no ack within %s", timeout)
 	}
-	return nil
 }
 
 // heartbeat updates the proxy's last_seen timestamp every minute and

management/internals/shared/grpc/server.go

@@ -553,9 +553,9 @@ func extractPeerMeta(ctx context.Context, meta *proto.PeerSystemMeta) nbpeer.Pee
 		return nbpeer.PeerSystemMeta{}
 	}
 
-	osVersion := meta.GetOSVersion()
+	osVersion := cleanClientString(meta.GetOSVersion())
 	if osVersion == "" {
-		osVersion = meta.GetCore()
+		osVersion = cleanClientString(meta.GetCore())
 	}
 
 	networkAddresses := make([]nbpeer.NetworkAddress, 0, len(meta.GetNetworkAddresses()))
@@ -567,36 +567,36 @@ func extractPeerMeta(ctx context.Context, meta *proto.PeerSystemMeta) nbpeer.Pee
 		}
 		networkAddresses = append(networkAddresses, nbpeer.NetworkAddress{
 			NetIP: netAddr,
-			Mac:   addr.GetMac(),
+			Mac:   cleanClientString(addr.GetMac()),
 		})
 	}
 
 	files := make([]nbpeer.File, 0, len(meta.GetFiles()))
 	for _, file := range meta.GetFiles() {
 		files = append(files, nbpeer.File{
-			Path:             file.GetPath(),
+			Path:             cleanClientString(file.GetPath()),
 			Exist:            file.GetExist(),
 			ProcessIsRunning: file.GetProcessIsRunning(),
 		})
 	}
 
 	return nbpeer.PeerSystemMeta{
-		Hostname:           meta.GetHostname(),
-		GoOS:               meta.GetGoOS(),
-		Kernel:             meta.GetKernel(),
-		Platform:           meta.GetPlatform(),
-		OS:                 meta.GetOS(),
+		Hostname:           cleanClientString(meta.GetHostname()),
+		GoOS:               cleanClientString(meta.GetGoOS()),
+		Kernel:             cleanClientString(meta.GetKernel()),
+		Platform:           cleanClientString(meta.GetPlatform()),
+		OS:                 cleanClientString(meta.GetOS()),
 		OSVersion:          osVersion,
-		WtVersion:          meta.GetNetbirdVersion(),
-		UIVersion:          meta.GetUiVersion(),
-		KernelVersion:      meta.GetKernelVersion(),
+		WtVersion:          cleanClientString(meta.GetNetbirdVersion()),
+		UIVersion:          cleanClientString(meta.GetUiVersion()),
+		KernelVersion:      cleanClientString(meta.GetKernelVersion()),
 		NetworkAddresses:   networkAddresses,
-		SystemSerialNumber: meta.GetSysSerialNumber(),
-		SystemProductName:  meta.GetSysProductName(),
-		SystemManufacturer: meta.GetSysManufacturer(),
+		SystemSerialNumber: cleanClientString(meta.GetSysSerialNumber()),
+		SystemProductName:  cleanClientString(meta.GetSysProductName()),
+		SystemManufacturer: cleanClientString(meta.GetSysManufacturer()),
 		Environment: nbpeer.Environment{
-			Cloud:    meta.GetEnvironment().GetCloud(),
-			Platform: meta.GetEnvironment().GetPlatform(),
+			Cloud:    cleanClientString(meta.GetEnvironment().GetCloud()),
+			Platform: cleanClientString(meta.GetEnvironment().GetPlatform()),
 		},
 		Flags: nbpeer.Flags{
 			RosenpassEnabled:      meta.GetFlags().GetRosenpassEnabled(),
@@ -618,6 +618,11 @@ func extractPeerMeta(ctx context.Context, meta *proto.PeerSystemMeta) nbpeer.Pee
 	}
 }
 
+// cleanClientString removes NUL bytes and invalid UTF-8 from a client-reported string, which Postgres rejects in text columns.
+func cleanClientString(s string) string {
+	return strings.ToValidUTF8(strings.ReplaceAll(s, "\x00", ""), "")
+}
+
 func capabilitiesToInt32(caps []proto.PeerCapability) []int32 {
 	result := make([]int32, len(caps))
 	for i, c := range caps {
@@ -711,7 +716,7 @@ func (s *Server) Login(ctx context.Context, req *proto.EncryptedMessage) (*proto
 
 	peer, network, postureChecks, enableSSH, err := s.accountManager.LoginPeer(ctx, types.PeerLogin{
 		WireGuardPubKey: peerKey.String(),
-		SSHKey:          string(sshKey),
+		SSHKey:          cleanClientString(string(sshKey)),
 		Meta:            peerMeta,
 		UserID:          userID,
 		SetupKey:        loginReq.GetSetupKey(),

management/internals/shared/grpc/sync_mappings_test.go

@@ -409,3 +409,75 @@ func TestConnectionSendResponse_RoutesToLegacyStream(t *testing.T) {
 	require.Len(t, stream.messages, 1)
 	assert.Equal(t, "svc-2", stream.messages[0].Mapping[0].Id)
 }
+
+// blockingSyncStream is a ProxyService_SyncMappingsServer whose Recv blocks
+// until the test ends, like a proxy that stopped acking without closing the stream.
+type blockingSyncStream struct {
+	grpc.ServerStream
+	done chan struct{}
+}
+
+func (s *blockingSyncStream) Recv() (*proto.SyncMappingsRequest, error) {
+	<-s.done
+	return nil, context.Canceled
+}
+
+func (s *blockingSyncStream) Send(*proto.SyncMappingsResponse) error { return nil }
+
+func newBlockingSyncStream(t *testing.T) *blockingSyncStream {
+	t.Helper()
+	s := &blockingSyncStream{done: make(chan struct{})}
+	t.Cleanup(func() { close(s.done) })
+	return s
+}
+
+func TestWaitForAck_Ack(t *testing.T) {
+	stream := &syncRecordingStream{recvMsgs: []*proto.SyncMappingsRequest{ackMsg()}}
+
+	require.NoError(t, waitForAck(context.Background(), stream, time.Second))
+}
+
+func TestWaitForAck_UnexpectedMessage(t *testing.T) {
+	stream := &syncRecordingStream{recvMsgs: []*proto.SyncMappingsRequest{
+		{Msg: &proto.SyncMappingsRequest_Init{Init: &proto.SyncMappingsInit{}}},
+	}}
+
+	assert.ErrorContains(t, waitForAck(context.Background(), stream, time.Second), "expected ack")
+}
+
+func TestWaitForAck_TimesOutWhenProxyStopsAcking(t *testing.T) {
+	stream := newBlockingSyncStream(t)
+
+	start := time.Now()
+	err := waitForAck(context.Background(), stream, 50*time.Millisecond)
+
+	assert.ErrorContains(t, err, "no ack within")
+	assert.Less(t, time.Since(start), time.Second)
+}
+
+func TestWaitForAck_ReturnsWhenConnectionCanceled(t *testing.T) {
+	stream := newBlockingSyncStream(t)
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+
+	assert.ErrorIs(t, waitForAck(ctx, stream, time.Minute), context.Canceled)
+}
+
+func TestSendSnapshotSync_GivesUpWhenProxyStopsAcking(t *testing.T) {
+	const cluster = "cluster.example.com"
+
+	ctrl := gomock.NewController(t)
+	mgr := rpservice.NewMockManager(ctrl)
+	mgr.EXPECT().GetGlobalServices(gomock.Any()).Return(makeServices(1, cluster), nil)
+
+	s := newSnapshotTestServer(t, 10)
+	s.serviceManager = mgr
+	s.snapshotAckTimeout = 50 * time.Millisecond
+
+	stream := newBlockingSyncStream(t)
+	conn := &proxyConnection{proxyID: "proxy-a", address: cluster, syncStream: stream}
+
+	err := s.sendSnapshotSync(context.Background(), conn, stream)
+
+	assert.ErrorContains(t, err, "no ack within")
+}

management/server/account.go

@@ -770,7 +770,12 @@ func (am *DefaultAccountManager) peerLoginExpirationJob(ctx context.Context, acc
 		ctx = context.WithValue(ctx, hook.ExecutionContextKey, fmt.Sprintf("%s-PEER-EXPIRATION", hook.SystemSource))
 
 		expiredPeers, err := am.getExpiredPeers(ctx, accountID)
+		if isNotFound(err) {
+			log.WithContext(ctx).Debugf("account %s not found, stopping peer login expiration job", accountID)
+			return 0, false
+		}
 		if err != nil {
+			log.WithContext(ctx).Errorf("failed getting expired peers for account %s: %v", accountID, err)
 			return peerSchedulerRetryInterval, true
 		}
 
@@ -806,8 +811,12 @@ func (am *DefaultAccountManager) schedulePeerLoginExpiration(ctx context.Context
 func (am *DefaultAccountManager) peerInactivityExpirationJob(ctx context.Context, accountID string) func() (time.Duration, bool) {
 	return func() (time.Duration, bool) {
 		inactivePeers, err := am.getInactivePeers(ctx, accountID)
+		if isNotFound(err) {
+			log.WithContext(ctx).Debugf("account %s not found, stopping peer inactivity expiration job", accountID)
+			return 0, false
+		}
 		if err != nil {
-			log.WithContext(ctx).Errorf("failed getting inactive peers for account %s", accountID)
+			log.WithContext(ctx).Errorf("failed getting inactive peers for account %s: %v", accountID, err)
 			return peerSchedulerRetryInterval, true
 		}
 
@@ -827,6 +836,14 @@ func (am *DefaultAccountManager) peerInactivityExpirationJob(ctx context.Context
 	}
 }
 
+func isNotFound(err error) bool {
+	if err == nil {
+		return false
+	}
+	sErr, ok := status.FromError(err)
+	return ok && sErr.Type() == status.NotFound
+}
+
 // checkAndSchedulePeerInactivityExpiration periodically checks for inactive peers to end their sessions
 func (am *DefaultAccountManager) checkAndSchedulePeerInactivityExpiration(ctx context.Context, accountID string) {
 	am.peerInactivityExpiry.Cancel(ctx, []string{accountID})
@@ -978,8 +995,8 @@ func (am *DefaultAccountManager) DeleteAccount(ctx context.Context, accountID, u
 		log.WithContext(ctx).Errorf("failed deleting account %s. error: %s", accountID, err)
 		return err
 	}
-	// cancel peer login expiry job
 	am.peerLoginExpiry.Cancel(ctx, []string{account.Id})
+	am.peerInactivityExpiry.Cancel(ctx, []string{account.Id})
 	am.certChallenges.Forget(account.Id)
 
 	meta := map[string]any{"account_id": account.Id, "domain": account.Domain, "created_at": account.CreatedAt}
@@ -1637,10 +1654,6 @@ func (am *DefaultAccountManager) GetAccountIDFromUserAuth(ctx context.Context, u
 		return "", "", status.Errorf(status.NotFound, "user %s not found", userAuth.UserId)
 	}
 
-	if userAuth.IsChild {
-		return accountID, user.Id, nil
-	}
-
 	ctx, err = am.permissionsManager.ValidateAccountAccess(ctx, accountID, user, false)
 	if err != nil {
 		return "", "", err
@@ -1692,10 +1705,10 @@ func (am *DefaultAccountManager) updateUserAuthWithSingleMode(ctx context.Contex
 // and propagates changes to peers if group propagation is enabled.
 // requires userAuth to have been ValidateAndParseToken and EnsureUserAccessByJWTGroups by the AuthManager
 func (am *DefaultAccountManager) SyncUserJWTGroups(ctx context.Context, userAuth auth.UserAuth) error {
-	// Child accounts and PAT-authenticated requests do not sync JWT groups.
+	// PAT-authenticated requests do not sync JWT groups.
 	// Embedded-Dex local users also skip sync because local password authentication
 	// does not provide external IdP group claims.
-	if userAuth.IsChild || userAuth.IsPAT || dex.IsLocalUserID(userAuth.UserId) {
+	if userAuth.IsPAT || dex.IsLocalUserID(userAuth.UserId) {
 		return nil
 	}
 
@@ -1874,8 +1887,6 @@ func (am *DefaultAccountManager) SyncUserJWTGroups(ctx context.Context, userAuth
 // Existing user + Existing account + Existing Indexed Domain -> Nothing changes
 //
 // Existing user + Existing account + Existing domain reclassified Domain as private -> Nothing changes (index domain)
-//
-// UserAuth IsChild -> checks that account exists
 func (am *DefaultAccountManager) getAccountIDWithAuthorizationClaims(ctx context.Context, userAuth auth.UserAuth) (string, error) {
 	log.WithContext(ctx).Tracef("getting account with authorization claims. User ID: \"%s\", Account ID: \"%s\", Domain: \"%s\", Domain Category: \"%s\"",
 		userAuth.UserId, userAuth.AccountId, userAuth.Domain, userAuth.DomainCategory)
@@ -1884,14 +1895,6 @@ func (am *DefaultAccountManager) getAccountIDWithAuthorizationClaims(ctx context
 		return "", errors.New(emptyUserID)
 	}
 
-	if userAuth.IsChild {
-		exists, err := am.Store.AccountExists(ctx, store.LockingStrengthNone, userAuth.AccountId)
-		if err != nil || !exists {
-			return "", err
-		}
-		return userAuth.AccountId, nil
-	}
-
 	if userAuth.DomainCategory != types.PrivateCategory || !isDomainValid(userAuth.Domain) {
 		return am.GetAccountIDByUserID(ctx, userAuth)
 	}

management/server/account_test.go

@@ -4990,3 +4990,27 @@ func TestUpdateUserAuthWithSingleMode(t *testing.T) {
 		assert.Contains(t, err.Error(), "query failed")
 	})
 }
+
+func TestDefaultAccountManager_ExpirationJobsStopForDeletedAccount(t *testing.T) {
+	manager, _, err := createManager(t)
+	require.NoError(t, err, "unable to create account manager")
+
+	ctx := context.Background()
+	accountID, err := manager.GetAccountIDByUserID(ctx, auth.UserAuth{UserId: userID})
+	require.NoError(t, err, "unable to create an account")
+
+	account, err := manager.Store.GetAccount(ctx, accountID)
+	require.NoError(t, err, "unable to get account")
+	require.NoError(t, manager.Store.DeleteAccount(ctx, account), "unable to delete account")
+
+	jobs := map[string]func() (time.Duration, bool){
+		"login":      manager.peerLoginExpirationJob(ctx, accountID),
+		"inactivity": manager.peerInactivityExpirationJob(ctx, accountID),
+	}
+	for name, job := range jobs {
+		t.Run(name, func(t *testing.T) {
+			_, reschedule := job()
+			assert.False(t, reschedule, "job must not reschedule for a deleted account")
+		})
+	}
+}

management/server/auth/manager.go

@@ -75,10 +75,10 @@ func (m *manager) ValidateAndParseToken(ctx context.Context, value string) (auth
 }
 
 func (m *manager) EnsureUserAccessByJWTGroups(ctx context.Context, userAuth auth.UserAuth, token *jwt.Token) (auth.UserAuth, error) {
-	// Child accounts and PAT-authenticated requests do not use JWT group access checks.
+	// PAT-authenticated requests do not use JWT group access checks.
 	// Embedded-Dex local users also skip them because local password authentication
 	// does not provide external IdP group claims.
-	if userAuth.IsChild || userAuth.IsPAT || dex.IsLocalUserID(userAuth.UserId) {
+	if userAuth.IsPAT || dex.IsLocalUserID(userAuth.UserId) {
 		return userAuth, nil
 	}
 

management/server/http/middleware/auth_middleware.go

@@ -26,8 +26,6 @@ type SyncUserJWTGroupsFunc func(ctx context.Context, userAuth auth.UserAuth) err
 
 type GetUserFromUserAuthFunc func(ctx context.Context, userAuth auth.UserAuth) (*types.User, error)
 
-type IsValidChildAccountFunc func(ctx context.Context, userID, accountID, childAccountID string) bool
-
 // AuthMiddleware middleware to verify personal access tokens (PAT) and JWT tokens
 type AuthMiddleware struct {
 	authManager         serverauth.Manager
@@ -36,7 +34,6 @@ type AuthMiddleware struct {
 	syncUserJWTGroups   SyncUserJWTGroupsFunc
 	rateLimiter         *ratelimit.APIRateLimiter
 	patUsageTracker     *PATUsageTracker
-	isValidChildAccount IsValidChildAccountFunc
 }
 
 // NewAuthMiddleware instance constructor
@@ -47,7 +44,6 @@ func NewAuthMiddleware(
 	getUserFromUserAuth GetUserFromUserAuthFunc,
 	rateLimiter *ratelimit.APIRateLimiter,
 	meter metric.Meter,
-	isValidChildAccount IsValidChildAccountFunc,
 ) *AuthMiddleware {
 	var patUsageTracker *PATUsageTracker
 	if meter != nil {
@@ -65,7 +61,6 @@ func NewAuthMiddleware(
 		getUserFromUserAuth: getUserFromUserAuth,
 		rateLimiter:         rateLimiter,
 		patUsageTracker:     patUsageTracker,
-		isValidChildAccount: isValidChildAccount,
 	}
 }
 
@@ -127,13 +122,6 @@ func (m *AuthMiddleware) checkJWTFromRequest(r *http.Request, authHeaderParts []
 		return err
 	}
 
-	if impersonate, ok := r.URL.Query()["account"]; ok && len(impersonate) == 1 {
-		if m.isValidChildAccount(ctx, userAuth.UserId, userAuth.AccountId, impersonate[0]) {
-			userAuth.AccountId = impersonate[0]
-			userAuth.IsChild = true
-		}
-	}
-
 	// Email is now extracted in ToUserAuth (from claims or userinfo endpoint)
 	// Available as userAuth.Email
 
@@ -210,13 +198,6 @@ func (m *AuthMiddleware) checkPATFromRequest(r *http.Request, authHeaderParts []
 		IsPAT:          true,
 	}
 
-	if impersonate, ok := r.URL.Query()["account"]; ok && len(impersonate) == 1 {
-		if m.isValidChildAccount(r.Context(), userAuth.UserId, userAuth.AccountId, impersonate[0]) {
-			userAuth.AccountId = impersonate[0]
-			userAuth.IsChild = true
-		}
-	}
-
 	// propagates ctx change to upstream middleware
 	*r = *nbcontext.SetUserAuthInRequest(r, userAuth)
 	return nil

management/server/http/middleware/auth_middleware_test.go

@@ -104,7 +104,7 @@ func mockMarkPATUsed(_ context.Context, token string) error {
 }
 
 func mockEnsureUserAccessByJWTGroups(_ context.Context, userAuth nbauth.UserAuth, token *jwt.Token) (nbauth.UserAuth, error) {
-	if userAuth.IsChild || userAuth.IsPAT {
+	if userAuth.IsPAT {
 		return userAuth, nil
 	}
 
@@ -212,7 +212,6 @@ func TestAuthMiddleware_Handler(t *testing.T) {
 		},
 		disabledLimiter,
 		nil,
-		func(_ context.Context, _, _, _ string) bool { return false },
 	)
 
 	handlerToTest := authMiddleware.Handler(nextHandler)
@@ -279,7 +278,6 @@ func TestAuthMiddleware_SyncUserJWTGroupsDetachedFromRequestCancellation(t *test
 		},
 		disabledLimiter,
 		nil,
-		func(_ context.Context, _, _, _ string) bool { return false },
 	)
 
 	handlerToTest := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {}))
@@ -332,7 +330,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -385,7 +382,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -429,7 +425,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -474,7 +469,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -539,7 +533,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -599,7 +592,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -655,7 +647,6 @@ func TestAuthMiddleware_RateLimiting(t *testing.T) {
 			},
 			ratelimit.NewAPIRateLimiter(rateLimitConfig),
 			nil,
-			func(_ context.Context, _, _, _ string) bool { return false },
 		)
 
 		handler := authMiddleware.Handler(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -756,7 +747,6 @@ func TestAuthMiddleware_Handler_Child(t *testing.T) {
 		},
 		disabledLimiter,
 		nil,
-		func(_ context.Context, _, _, _ string) bool { return false },
 	)
 
 	for _, tc := range tt {

management/server/http/middleware/middleware_factory.go

@@ -10,7 +10,7 @@ import (
 	log "github.com/sirupsen/logrus"
 )
 
-func BuildMiddleware(rateLimiter *ratelimit.APIRateLimiter, authManager auth.Manager, accountManager account.Manager, metrics telemetry.AppMetrics, isValidChildAcctFunc IsValidChildAccountFunc) []mux.MiddlewareFunc {
+func BuildMiddleware(rateLimiter *ratelimit.APIRateLimiter, authManager auth.Manager, accountManager account.Manager, metrics telemetry.AppMetrics) []mux.MiddlewareFunc {
 	toret := make([]mux.MiddlewareFunc, 0)
 
 	toret = append(toret, metrics.HTTPMiddleware().Handler)
@@ -28,7 +28,6 @@ func BuildMiddleware(rateLimiter *ratelimit.APIRateLimiter, authManager auth.Man
 		accountManager.GetUserFromUserAuth,
 		rateLimiter,
 		metrics.GetMeter(),
-		isValidChildAcctFunc,
 	).Handler)
 
 	return toret

management/server/http/testing/testing_tools/channel/channel.go

@@ -146,7 +146,7 @@ func BuildApiBlackBoxWithDBState(t testing_tools.TB, sqlFile string, expectedPee
 	zoneRecordsManager := recordsManager.NewManager(store, am, permissionsManager)
 
 	apiRouter := mux.NewRouter().PathPrefix("/api").Subrouter()
-	apiRouter.Use(middleware.BuildMiddleware(nil, authManagerMock, am, metrics, nil)...)
+	apiRouter.Use(middleware.BuildMiddleware(nil, authManagerMock, am, metrics)...)
 	apiHandler, err := http2.NewAPIHandler(ctx, apiRouter, am, networksManager, resourcesManager, routersManager,
 		groupsManager, geoMock, permissionsManager, settingsManager, customZonesManager, zoneRecordsManager,
 		networkMapController, nil, serviceManager, nil, nil, nil, nil, nil, nil)
@@ -288,7 +288,7 @@ func BuildApiBlackBoxWithDBStateAndPeerChannel(t testing_tools.TB, sqlFile strin
 	zoneRecordsManager := recordsManager.NewManager(store, am, permissionsManager)
 
 	apiRouter := mux.NewRouter().PathPrefix("/api").Subrouter()
-	apiRouter.Use(middleware.BuildMiddleware(nil, authManagerMock, am, metrics, nil)...)
+	apiRouter.Use(middleware.BuildMiddleware(nil, authManagerMock, am, metrics)...)
 	apiHandler, err := http2.NewAPIHandler(ctx, apiRouter, am, networksManager, resourcesManager, routersManager, groupsManager,
 		geoMock, permissionsManager, settingsManager, customZonesManager, zoneRecordsManager,
 		networkMapController, nil, serviceManager, nil, nil, nil, nil, nil, nil)

management/server/peer.go

@@ -1702,6 +1702,9 @@ func (am *DefaultAccountManager) getNextPeerExpiration(ctx context.Context, acco
 	}
 
 	settings, err := am.Store.GetAccountSettings(ctx, store.LockingStrengthNone, accountID)
+	if isNotFound(err) {
+		return 0, false
+	}
 	if err != nil {
 		log.WithContext(ctx).Errorf("failed to get account settings: %v", err)
 		return peerSchedulerRetryInterval, true
@@ -1745,6 +1748,9 @@ func (am *DefaultAccountManager) getNextInactivePeerExpiration(ctx context.Conte
 	}
 
 	settings, err := am.Store.GetAccountSettings(ctx, store.LockingStrengthNone, accountID)
+	if isNotFound(err) {
+		return 0, false
+	}
 	if err != nil {
 		log.WithContext(ctx).Errorf("failed to get account settings: %v", err)
 		return peerSchedulerRetryInterval, true

proxy/cmd/proxy/cmd/root.go

@@ -6,6 +6,7 @@ import (
 	"os"
 	"os/signal"
 	"strconv"
+	"strings"
 	"syscall"
 	"time"
 
@@ -13,6 +14,7 @@ import (
 	"github.com/spf13/cobra"
 	"golang.org/x/crypto/acme"
 
+	"github.com/netbirdio/netbird/shared/externalipmap"
 	"github.com/netbirdio/netbird/shared/management/domain"
 	"github.com/netbirdio/netbird/shared/profiling"
 
@@ -74,6 +76,7 @@ var (
 	certLockMethod        string
 	wildcardCertDir       string
 	wgPort                uint16
+	externalIPMap         []string
 	proxyProtocol         bool
 	preSharedKey          string
 	supportsCustomPorts   bool
@@ -117,6 +120,7 @@ func init() {
 	rootCmd.Flags().StringVar(&certLockMethod, "cert-lock-method", envStringOrDefault("NB_PROXY_CERT_LOCK_METHOD", "auto"), "Certificate lock method for cross-replica coordination: auto, flock, or k8s-lease")
 	rootCmd.Flags().StringVar(&wildcardCertDir, "wildcard-cert-dir", envStringOrDefault("NB_PROXY_WILDCARD_CERT_DIR", ""), "Directory containing wildcard certificate pairs (<name>.crt/<name>.key). Wildcard patterns are extracted from SANs automatically")
 	rootCmd.Flags().Uint16Var(&wgPort, "wg-port", envUint16OrDefault("NB_PROXY_WG_PORT", 0), "WireGuard listen port (0 = random). Fixed port only works with single-account deployments")
+	rootCmd.Flags().StringSliceVar(&externalIPMap, "external-ip-map", envStringSliceOrDefault("NB_PROXY_EXTERNAL_IP_MAP", nil), "Comma-separated external IPs to advertise as ICE candidates, as IP or IP/internal-IP-or-interface (e.g. '203.0.113.5/10.0.0.5'). Requires a fixed --wg-port forwarded on the NAT")
 	rootCmd.Flags().BoolVar(&proxyProtocol, "proxy-protocol", envBoolOrDefault("NB_PROXY_PROXY_PROTOCOL", false), "Enable PROXY protocol on TCP listeners to preserve client IPs behind L4 proxies")
 	rootCmd.Flags().StringVar(&preSharedKey, "preshared-key", envStringOrDefault("NB_PROXY_PRESHARED_KEY", ""), "Define a pre-shared key for the tunnel between proxy and peers")
 	rootCmd.Flags().BoolVar(&supportsCustomPorts, "supports-custom-ports", envBoolOrDefault("NB_PROXY_SUPPORTS_CUSTOM_PORTS", true), "Whether the proxy can bind arbitrary ports for UDP/TCP passthrough")
@@ -221,6 +225,13 @@ func runServer(cmd *cobra.Command, args []string) error {
 		return fmt.Errorf("invalid --trusted-proxies: %w", err)
 	}
 
+	for i := range externalIPMap {
+		externalIPMap[i] = strings.TrimSpace(externalIPMap[i])
+	}
+	if err := externalipmap.Validate(externalIPMap); err != nil {
+		return fmt.Errorf("invalid --external-ip-map: %w", err)
+	}
+
 	ctx, stop := signal.NotifyContext(context.Background(), syscall.SIGTERM, syscall.SIGINT)
 	defer stop()
 
@@ -248,6 +259,7 @@ func runServer(cmd *cobra.Command, args []string) error {
 		CertLockMethod:           nbacme.CertLockMethod(certLockMethod),
 		WildcardCertDir:          wildcardCertDir,
 		WireguardPort:            wgPort,
+		NATExternalIPs:           externalIPMap,
 		Performance:              perf,
 		ProxyProtocol:            proxyProtocol,
 		PreSharedKey:             preSharedKey,
@@ -286,6 +298,14 @@ func envStringOrDefault(key string, def string) string {
 	return v
 }
 
+func envStringSliceOrDefault(key string, def []string) []string {
+	v, exists := os.LookupEnv(key)
+	if !exists || v == "" {
+		return def
+	}
+	return strings.Split(v, ",")
+}
+
 func envUint16OrDefault(key string, def uint16) uint16 {
 	v, exists := os.LookupEnv(key)
 	if !exists {

proxy/internal/roundtrip/netbird.go

@@ -144,6 +144,8 @@ type ClientConfig struct {
 	WGPort       uint16
 	PreSharedKey string
 	Performance  embed.Performance
+	// NATExternalIPs mirrors embed.Options.NATExternalIPs.
+	NATExternalIPs []string
 	// BlockInbound mirrors embed.Options.BlockInbound. Set to true on the
 	// standalone proxy where the embedded client never accepts inbound;
 	// set to false on the private/embedded proxy so the engine creates
@@ -402,6 +404,7 @@ func (n *NetBird) createClientEntry(ctx context.Context, accountID types.Account
 		WireguardPort:  &wgPort,
 		PreSharedKey:   n.clientCfg.PreSharedKey,
 		Performance:    n.clientCfg.Performance,
+		NATExternalIPs: n.clientCfg.NATExternalIPs,
 	}
 	logEmbedOptions(n.logger, accountID, serviceID, publicKey.String(), embedOpts)
 	client, err := embed.New(embedOpts)
@@ -914,6 +917,7 @@ func logEmbedOptions(logger *log.Logger, accountID types.AccountID, serviceID ty
 		"management_url":        opts.ManagementURL,
 		"log_level":             opts.LogLevel,
 		"wg_port":               wgPort,
+		"nat_external_ips":      opts.NATExternalIPs,
 		"mtu":                   mtu,
 		"block_inbound":         opts.BlockInbound,
 		"block_lan_access":      opts.BlockLANAccess,

proxy/lifecycle.go

@@ -89,6 +89,9 @@ type Config struct {
 	// WireguardPort is the UDP port for the embedded NetBird tunnel.
 	// Zero asks the OS for a random port.
 	WireguardPort uint16
+	// NATExternalIPs are external IPs advertised as ICE candidates by the
+	// embedded tunnel. Requires a fixed WireguardPort forwarded on the NAT.
+	NATExternalIPs []string
 	// ProxyProtocol enables PROXY protocol (v1/v2) on TCP listeners.
 	ProxyProtocol bool
 	// PreSharedKey is the WireGuard pre-shared key used between the
@@ -160,6 +163,7 @@ func New(ctx context.Context, cfg Config) *Server {
 		ForwardedProto:           cfg.ForwardedProto,
 		TrustedProxies:           cfg.TrustedProxies,
 		WireguardPort:            cfg.WireguardPort,
+		NATExternalIPs:           cfg.NATExternalIPs,
 		ProxyProtocol:            cfg.ProxyProtocol,
 		PreSharedKey:             cfg.PreSharedKey,
 		Performance:              cfg.Performance,

proxy/server.go

@@ -203,6 +203,9 @@ type Server struct {
 	// single-account deployments; multiple accounts will fail to bind
 	// the same port.
 	WireguardPort uint16
+	// NATExternalIPs maps external IPs advertised as ICE candidates by every
+	// embedded client. Requires a fixed WireguardPort forwarded on the NAT.
+	NATExternalIPs []string
 	// Performance configures the tunnel pool/batch sizes for every
 	// embedded client this proxy spawns.
 	Performance embed.Performance
@@ -581,10 +584,11 @@ func (s *Server) initManagementClient() error {
 // inbound listeners.
 func (s *Server) initNetBirdClient() error {
 	s.netbird = roundtrip.NewNetBird(s.ctx, s.ID, s.ProxyURL, roundtrip.ClientConfig{
-		MgmtAddr:     s.ManagementAddress,
-		WGPort:       s.WireguardPort,
-		PreSharedKey: s.PreSharedKey,
-		Performance:  s.Performance,
+		MgmtAddr:       s.ManagementAddress,
+		WGPort:         s.WireguardPort,
+		PreSharedKey:   s.PreSharedKey,
+		Performance:    s.Performance,
+		NATExternalIPs: s.NATExternalIPs,
 		// On --private the embedded client serves per-account inbound
 		// listeners and must apply management's ACL: keep BlockInbound off
 		// so the engine creates the ACL manager. On the standalone proxy
@@ -1299,6 +1303,10 @@ func (s *Server) proxyCapabilities() *proto.ProxyCapabilities {
 }
 
 func (s *Server) tryGetMappingUpdate(ctx context.Context, client proto.ProxyServiceClient, initialSyncDone *bool, onConnected func()) error {
+	// Canceling closes the stream on management; a bare return leaves it open.
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
 	connectTime := time.Now()
 	mappingClient, err := client.GetMappingUpdate(ctx, &proto.GetMappingUpdateRequest{
 		ProxyId:      s.ID,
@@ -1321,6 +1329,10 @@ func (s *Server) tryGetMappingUpdate(ctx context.Context, client proto.ProxyServ
 }
 
 func (s *Server) trySyncMappings(ctx context.Context, client proto.ProxyServiceClient, initialSyncDone *bool, onConnected func()) error {
+	// Canceling closes the stream on management; a bare return leaves it open.
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
 	connectTime := time.Now()
 	stream, err := client.SyncMappings(ctx)
 	if err != nil {

release_files/freebsd-port-diff.sh

@@ -9,62 +9,19 @@
 # Usage: ./freebsd-port-diff.sh [new_version]
 # Example: ./freebsd-port-diff.sh 0.60.7
 #
-# If no version is provided, it fetches the latest from GitHub.
+# If no version is provided, it uses the latest stable release from GitHub.
+# Prerelease versions (canary, rc, ...) are rejected: the port only tracks
+# stable releases.
 
 set -e
 
-GITHUB_REPO="netbirdio/netbird"
-PORTS_MIRROR_BASE="https://raw.githubusercontent.com/freebsd/freebsd-ports/main/security/netbird"
+# shellcheck source=release_files/freebsd-port-lib.sh
+source "$(dirname "${BASH_SOURCE[0]}")/freebsd-port-lib.sh"
+
 GO_PROXY="https://proxy.golang.org/github.com/netbirdio/netbird/@v"
 OUTPUT_DIR="${OUTPUT_DIR:-.}"
 AWK_FIRST_FIELD='{print $1}'
 
-fetch_all_tags() {
-    curl -sL "https://github.com/${GITHUB_REPO}/tags" 2>/dev/null | \
-        grep -oE '/releases/tag/v[0-9]+\.[0-9]+\.[0-9]+([^"]+)?' | \
-        grep -iv 'rc' | \
-        sed 's/.*\/v//' | \
-        sort -u -V
-    return 0
-}
-
-fetch_current_ports_version() {
-    echo "Fetching current version from FreeBSD ports..." >&2
-    local makefile version
-    makefile=$(fetch_ports_file "Makefile") || return 1
-    version=$(echo "$makefile" | \
-        grep -E "^DISTVERSION=" | \
-        sed 's/DISTVERSION=[[:space:]]*//' | \
-        tr -d '\t ')
-    if [[ -z "$version" ]]; then
-        echo "Error: Could not extract DISTVERSION from ports Makefile" >&2
-        return 1
-    fi
-    echo "$version"
-    return 0
-}
-
-fetch_latest_github_release() {
-    echo "Fetching latest release from GitHub..." >&2
-    fetch_all_tags | tail -1
-    return 0
-}
-
-fetch_ports_file() {
-    local filename="$1"
-    local content
-    if ! content=$(curl -fsL --proto '=https' --proto-redir '=https' --retry 3 "${PORTS_MIRROR_BASE}/${filename}" 2>/dev/null); then
-        echo "Error: Could not fetch ${filename} from ${PORTS_MIRROR_BASE}" >&2
-        return 1
-    fi
-    if [[ "$content" == \<* ]]; then
-        echo "Error: Received HTML instead of ${filename} from ${PORTS_MIRROR_BASE}" >&2
-        return 1
-    fi
-    printf '%s' "$content"
-    return 0
-}
-
 compute_checksums() {
     local version="$1"
     local tmpdir
@@ -77,11 +34,15 @@ compute_checksums() {
     local mod_file="${tmpdir}/v${version}.mod"
     local zip_file="${tmpdir}/v${version}.zip"
 
-    curl -sL "${GO_PROXY}/v${version}.mod" -o "$mod_file" 2>/dev/null
-    curl -sL "${GO_PROXY}/v${version}.zip" -o "$zip_file" 2>/dev/null
+    local -a curl_opts=(-fsL --proto '=https' --proto-redir '=https' --retry 3)
+    if ! curl "${curl_opts[@]}" "${GO_PROXY}/v${version}.mod" -o "$mod_file" 2>/dev/null || \
+       ! curl "${curl_opts[@]}" "${GO_PROXY}/v${version}.zip" -o "$zip_file" 2>/dev/null; then
+        echo "Error: Could not download v${version} from the Go module proxy" >&2
+        return 1
+    fi
 
     if [[ ! -s "$mod_file" ]] || [[ ! -s "$zip_file" ]]; then
-        echo "Error: Could not download files from Go module proxy" >&2
+        echo "Error: Go module proxy returned empty files for v${version}" >&2
         return 1
     fi
 
@@ -133,25 +94,12 @@ generate_new_makefile() {
 }
 
 # Parse arguments
-NEW_VERSION="${1:-}"
-
-# Auto-detect versions if not provided
-OLD_VERSION=$(fetch_current_ports_version)
-if [[ -z "$OLD_VERSION" ]]; then
-    echo "Error: Could not fetch current version from FreeBSD ports" >&2
-    exit 1
-fi
-echo "Current FreeBSD ports version: ${OLD_VERSION}" >&2
-
-if [[ -z "$NEW_VERSION" ]]; then
-    NEW_VERSION=$(fetch_latest_github_release)
-    if [[ -z "$NEW_VERSION" ]]; then
-        echo "Error: Could not fetch latest release from GitHub" >&2
-        exit 1
-    fi
-fi
+NEW_VERSION=$(resolve_target_version "${1:-}")
 echo "Target version: ${NEW_VERSION}" >&2
 
+OLD_VERSION=$(fetch_current_ports_version)
+echo "Current FreeBSD ports version: ${OLD_VERSION}" >&2
+
 if [[ "$OLD_VERSION" = "$NEW_VERSION" ]]; then
     echo "Port is already at version ${NEW_VERSION}. Nothing to do." >&2
     exit 0

release_files/freebsd-port-issue-body.sh

@@ -10,81 +10,29 @@
 #
 # If no versions are provided, the script will:
 #   - Fetch OLD version from the FreeBSD ports GitHub mirror (current version in ports tree)
-#   - Fetch NEW version from latest NetBird GitHub release tag
+#   - Fetch NEW version from the latest stable NetBird release tag on GitHub
+#
+# Prerelease versions (canary, rc, ...) are rejected: the port only tracks
+# stable releases.
 
 set -e
 
-GITHUB_REPO="netbirdio/netbird"
-PORTS_MAKEFILE_URL="https://raw.githubusercontent.com/freebsd/freebsd-ports/main/security/netbird/Makefile"
-
-fetch_current_ports_version() {
-    echo "Fetching current version from FreeBSD ports..." >&2
-    local makefile_content
-    makefile_content=$(curl -fsL --proto '=https' --proto-redir '=https' --retry 3 "$PORTS_MAKEFILE_URL" 2>/dev/null) || makefile_content=""
-    if [[ "$makefile_content" == \<* ]]; then
-        echo "Error: Received HTML instead of Makefile from ${PORTS_MAKEFILE_URL}" >&2
-        return 1
-    fi
-    if [[ -z "$makefile_content" ]]; then
-        echo "Error: Could not fetch Makefile from FreeBSD ports" >&2
-        return 1
-    fi
-    echo "$makefile_content" | grep -E "^DISTVERSION=" | sed 's/DISTVERSION=[[:space:]]*//' | tr -d '\t '
-    return 0
-}
-
-fetch_all_tags() {
-    # Fetch tags from GitHub tags page (no rate limiting, no auth needed)
-    curl -sL "https://github.com/${GITHUB_REPO}/tags" 2>/dev/null | \
-        grep -oE '/releases/tag/v[0-9]+\.[0-9]+\.[0-9]+([^"]+)?' | \
-        grep -iv 'rc' | \
-        sed 's/.*\/v//' | \
-        sort -u -V
-    return 0
-}
-
-fetch_latest_github_release() {
-    echo "Fetching latest release from GitHub..." >&2
-    local latest
-
-    # Fetch from GitHub tags page
-    latest=$(fetch_all_tags | tail -1)
-
-    if [[ -z "$latest" ]]; then
-        # Fallback to GitHub API
-        latest=$(curl -sL "https://api.github.com/repos/${GITHUB_REPO}/releases/latest" 2>/dev/null | \
-            grep '"tag_name"' | sed 's/.*"tag_name": *"v\([^"]*\)".*/\1/')
-    fi
-
-    if [[ -z "$latest" ]]; then
-        echo "Error: Could not fetch latest release from GitHub" >&2
-        return 1
-    fi
-    echo "$latest"
-    return 0
-}
+# shellcheck source=release_files/freebsd-port-lib.sh
+source "$(dirname "${BASH_SOURCE[0]}")/freebsd-port-lib.sh"
 
 OLD_VERSION="${1:-}"
 NEW_VERSION="${2:-}"
 
 if [[ -z "$OLD_VERSION" ]]; then
     OLD_VERSION=$(fetch_current_ports_version)
-    if [[ -z "$OLD_VERSION" ]]; then
-        echo "Error: Could not determine old version. Please provide it manually." >&2
-        echo "Usage: $0 <old_version> <new_version>" >&2
-        exit 1
-    fi
     echo "Detected OLD version from FreeBSD ports: $OLD_VERSION" >&2
 fi
 
 if [[ -z "$NEW_VERSION" ]]; then
-    NEW_VERSION=$(fetch_latest_github_release)
-    if [[ -z "$NEW_VERSION" ]]; then
-        echo "Error: Could not determine new version. Please provide it manually." >&2
-        echo "Usage: $0 <old_version> <new_version>" >&2
-        exit 1
-    fi
+    NEW_VERSION=$(resolve_target_version "")
     echo "Detected NEW version from GitHub: $NEW_VERSION" >&2
+else
+    NEW_VERSION=$(resolve_target_version "$NEW_VERSION")
 fi
 
 if [[ "$OLD_VERSION" = "$NEW_VERSION" ]]; then
@@ -98,8 +46,11 @@ OUTPUT_DIR="${OUTPUT_DIR:-.}"
 fetch_releases_between_versions() {
     echo "Fetching release history from GitHub..." >&2
 
-    # Fetch all tags and filter to those between OLD and NEW versions
-    fetch_all_tags | \
+    local tags
+    tags=$(fetch_stable_tags) || return 1
+
+    # List the stable releases after OLD up to and including NEW
+    echo "$tags" | \
         while read -r ver; do
             if [[ "$(printf '%s\n' "$OLD_VERSION" "$ver" | sort -V | head -n1)" = "$OLD_VERSION" ]] && \
                [[ "$(printf '%s\n' "$ver" "$NEW_VERSION" | sort -V | head -n1)" = "$ver" ]] && \
@@ -112,7 +63,7 @@ fetch_releases_between_versions() {
 
 generate_changelog_section() {
     local releases
-    releases=$(fetch_releases_between_versions)
+    releases=$(fetch_releases_between_versions) || return 1
 
     echo "Changelogs:"
     if [[ -n "$releases" ]]; then
@@ -127,7 +78,10 @@ generate_changelog_section() {
 
 OUTPUT_FILE="${OUTPUT_DIR}/netbird-${NEW_VERSION}-issue.txt"
 
-cat << EOF > "$OUTPUT_FILE"
+# Built before the heredoc: a failing substitution inside it would not stop the script.
+CHANGELOG=$(generate_changelog_section)
+
+cat << EOT > "$OUTPUT_FILE"
 BUGZILLA ISSUE DETAILS
 ======================
 
@@ -139,11 +93,11 @@ Description:
 ------------
 security/netbird: Update ${OLD_VERSION} => ${NEW_VERSION}
 
-$(generate_changelog_section)
+${CHANGELOG}
 
 Commit log:
 https://github.com/${GITHUB_REPO}/compare/v${OLD_VERSION}...v${NEW_VERSION}
-EOF
+EOT
 
 echo "========================================="
 echo "Issue body saved to: ${OUTPUT_FILE}"

release_files/freebsd-port-lib.sh

@@ -0,0 +1,98 @@
+#!/bin/bash
+#
+# Shared helpers for the FreeBSD port scripts. Source this file, do not run it.
+#
+# The FreeBSD port tracks stable NetBird releases only, so every helper here
+# works on plain MAJOR.MINOR.PATCH versions and ignores canary, release
+# candidate and any other prerelease tag.
+
+GITHUB_REPO="netbirdio/netbird"
+GITHUB_REPO_URL="https://github.com/${GITHUB_REPO}.git"
+PORTS_MIRROR_BASE="https://raw.githubusercontent.com/freebsd/freebsd-ports/main/security/netbird"
+STABLE_VERSION_RE='^[0-9]+\.[0-9]+\.[0-9]+$'
+
+# is_stable_version reports whether the version is a plain MAJOR.MINOR.PATCH release.
+is_stable_version() {
+    local version="$1"
+    if [[ "$version" =~ $STABLE_VERSION_RE ]]; then
+        return 0
+    fi
+    return 1
+}
+
+# fetch_stable_tags prints the stable release versions known to GitHub, oldest
+# first. git ls-remote returns every tag in one round trip, with no API
+# pagination or rate limit to work around.
+fetch_stable_tags() {
+    local refs
+    if ! refs=$(git ls-remote --tags --refs "$GITHUB_REPO_URL" 2>/dev/null); then
+        echo "Error: Could not list tags of ${GITHUB_REPO_URL}" >&2
+        return 1
+    fi
+
+    echo "$refs" | \
+        sed -nE 's#.*refs/tags/v([0-9]+\.[0-9]+\.[0-9]+)$#\1#p' | \
+        sort -u -V
+    return 0
+}
+
+# resolve_target_version validates an explicitly requested version, or picks the
+# latest stable release from GitHub when none was given. A leading "v" is
+# accepted so a git tag name can be passed as is.
+resolve_target_version() {
+    local raw="$1"
+    local requested="${raw#v}"
+
+    if [[ -n "$raw" ]]; then
+        if ! is_stable_version "$requested"; then
+            echo "Error: '${raw}' is not a stable release. The FreeBSD port only tracks MAJOR.MINOR.PATCH versions." >&2
+            return 1
+        fi
+        echo "$requested"
+        return 0
+    fi
+
+    echo "Fetching latest stable release from GitHub..." >&2
+    local latest
+    latest=$(fetch_stable_tags | tail -1)
+    if [[ -z "$latest" ]]; then
+        echo "Error: Could not determine the latest stable release from GitHub" >&2
+        return 1
+    fi
+    echo "$latest"
+    return 0
+}
+
+# fetch_ports_file prints a file from the security/netbird port on the FreeBSD
+# ports GitHub mirror.
+fetch_ports_file() {
+    local filename="$1"
+    local content
+    if ! content=$(curl -fsL --proto '=https' --proto-redir '=https' --retry 3 "${PORTS_MIRROR_BASE}/${filename}" 2>/dev/null); then
+        echo "Error: Could not fetch ${filename} from ${PORTS_MIRROR_BASE}" >&2
+        return 1
+    fi
+    if [[ "$content" == \<* ]]; then
+        echo "Error: Received HTML instead of ${filename} from ${PORTS_MIRROR_BASE}" >&2
+        return 1
+    fi
+    printf '%s' "$content"
+    return 0
+}
+
+# fetch_current_ports_version prints the DISTVERSION currently in the ports tree.
+fetch_current_ports_version() {
+    echo "Fetching current version from FreeBSD ports..." >&2
+    local makefile version
+    makefile=$(fetch_ports_file "Makefile") || return 1
+    version=$(echo "$makefile" | \
+        grep -E "^DISTVERSION=" | \
+        sed 's/DISTVERSION=[[:space:]]*//' | \
+        tr -d '\t ')
+    if [[ -z "$version" ]]; then
+        echo "Error: Could not extract DISTVERSION from ports Makefile" >&2
+        return 1
+    fi
+    echo "$version"
+    return 0
+}

shared/externalipmap/validate.go

@@ -0,0 +1,70 @@
+// Package externalipmap validates the --external-ip-map flag shared by the
+// agent and the proxy.
+package externalipmap
+
+import (
+	"fmt"
+	"net"
+	"strings"
+)
+
+const flagName = "external-ip-map"
+
+// Validate checks each element is formatted as "IP", "IP/IP" or
+// "IP/interface". The engine drops every mapping when one element fails to
+// parse, so callers should reject bad input up front.
+func Validate(list []string) error {
+	for _, element := range list {
+		if element == "" {
+			return fmt.Errorf("empty string is not a valid input for %s", flagName)
+		}
+
+		subElements := strings.Split(element, "/")
+		if len(subElements) > 2 {
+			return fmt.Errorf("%s is not a valid input for %s. it should be formatted as \"String\" or \"String/String\"", element, flagName)
+		}
+
+		if !isValidIP(subElements[0]) {
+			return fmt.Errorf("%s is not a valid input for %s. it should be formatted as \"IP\" or \"IP/IP\", or \"IP/Interface Name\"", element, flagName)
+		}
+
+		if len(subElements) == 2 {
+			if err := validateInternal(subElements[1]); err != nil {
+				return fmt.Errorf("%s is not a valid input for %s. it should be an IP string or a network name: %w", subElements[1], flagName, err)
+			}
+		}
+	}
+	return nil
+}
+
+func validateInternal(element string) error {
+	if isValidIP(element) {
+		return nil
+	}
+	validIface, err := isValidInterface(element)
+	if err != nil {
+		return fmt.Errorf("validate network interface name: %w", err)
+	}
+	if !validIface {
+		return fmt.Errorf("invalid IP or network interface name not found")
+	}
+	return nil
+}
+
+// isValidIP uses net.ParseIP to match the engine's parser exactly.
+func isValidIP(ip string) bool {
+	return net.ParseIP(ip) != nil
+}
+
+func isValidInterface(name string) (bool, error) {
+	netInterfaces, err := net.Interfaces()
+	if err != nil {
+		return false, err
+	}
+	for _, iface := range netInterfaces {
+		if iface.Name == name {
+			return true, nil
+		}
+	}
+	return false, nil
+}

shared/externalipmap/validate_test.go

@@ -0,0 +1,48 @@
+package externalipmap
+
+import (
+	"net"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestValidate(t *testing.T) {
+	ifaces, err := net.Interfaces()
+	require.NoError(t, err)
+	require.NotEmpty(t, ifaces, "host must have at least one interface")
+	iface := ifaces[0].Name
+
+	tests := []struct {
+		name    string
+		input   []string
+		wantErr bool
+	}{
+		{name: "nil", input: nil},
+		{name: "single IP", input: []string{"203.0.113.5"}},
+		{name: "IP to IP", input: []string{"203.0.113.5/10.0.0.5"}},
+		{name: "IP to interface", input: []string{"203.0.113.5/" + iface}},
+		{name: "several", input: []string{"203.0.113.5", "203.0.113.6/10.0.0.6"}},
+		{name: "empty element", input: []string{""}, wantErr: true},
+		// The engine rejects padded values, so validation must too.
+		{name: "space padded", input: []string{"203.0.113.5", " 203.0.113.6"}, wantErr: true},
+		{name: "not an IP", input: []string{"not-an-ip"}, wantErr: true},
+		{name: "too many slashes", input: []string{"203.0.113.5/10.0.0.5/10.0.0.6"}, wantErr: true},
+		{name: "unknown interface", input: []string{"203.0.113.5/nb-no-such-iface0"}, wantErr: true},
+		{name: "external interface", input: []string{iface}, wantErr: true},
+		{name: "interface to IP", input: []string{iface + "/10.0.0.5"}, wantErr: true},
+		{name: "two interfaces", input: []string{iface + "/" + iface}, wantErr: true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := Validate(tt.input)
+			if tt.wantErr {
+				assert.Error(t, err, "input %q should be rejected", tt.input)
+				return
+			}
+			assert.NoError(t, err, "input %q should be accepted", tt.input)
+		})
+	}
+}

shared/management/certposture/certposture_test.go

@@ -2,6 +2,8 @@ package certposture
 
 import (
 	"crypto"
+	"crypto/ecdsa"
+	"crypto/elliptic"
 	"crypto/rand"
 	"crypto/rsa"
 	"crypto/sha256"
@@ -86,6 +88,7 @@ func TestSignVerify_RoundTripPerKeyType(t *testing.T) {
 	ca := certtest.NewCA(t, "root")
 	keys := map[string]crypto.Signer{
 		SigAlgECDSASHA256:  certtest.ECDSAKey(t),
+		SigAlgECDSASHA384:  p384Key(t),
 		SigAlgRSAPSSSHA256: certtest.RSAKey(t),
 		SigAlgEd25519:      certtest.Ed25519Key(t),
 	}
@@ -103,6 +106,24 @@ func TestSignVerify_RoundTripPerKeyType(t *testing.T) {
 	}
 }
 
+func p384Key(t *testing.T) crypto.Signer {
+	t.Helper()
+	key, err := ecdsa.GenerateKey(elliptic.P384(), rand.Reader)
+	require.NoError(t, err)
+	return key
+}
+
+// TestSign_RefusesCurvesTheProofDoesNotName: a proof names its algorithm from the key, and
+// the verifier only knows P-256 and P-384, so P-521 is refused before anything is signed.
+func TestSign_RefusesCurvesTheProofDoesNotName(t *testing.T) {
+	key, err := ecdsa.GenerateKey(elliptic.P521(), rand.Reader)
+	require.NoError(t, err)
+
+	_, _, err = Sign(key, NewChallenger(secret).Nonce(peerKey, now), peerKey)
+
+	assert.ErrorIs(t, err, ErrUnsupportedKey)
+}
+
 func TestVerify_Rejections(t *testing.T) {
 	ca := certtest.NewCA(t, "root")
 	c := NewChallenger(secret)

shared/management/client/grpc_test.go

@@ -12,6 +12,8 @@ import (
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/credentials/insecure"
 
+	"github.com/netbirdio/netbird/client/system"
+	"github.com/netbirdio/netbird/shared/management/certposture"
 	mgmtProto "github.com/netbirdio/netbird/shared/management/proto"
 )
 
@@ -93,3 +95,25 @@ func TestMaxRecvMsgSizeIntegration(t *testing.T) {
 		assert.Len(t, resp.Key, payloadSize)
 	})
 }
+
+// TestInfoToMetaData_CarriesCertificateProofs: management verifies the proofs it receives
+// in the peer meta, so a field dropped by the mapping fails the certificate check without
+// any error on either side.
+func TestInfoToMetaData_CarriesCertificateProofs(t *testing.T) {
+	info := &system.Info{CertificateProofs: []certposture.Proof{
+		{Nonce: []byte("nonce-1"), Chain: [][]byte{[]byte("leaf"), []byte("intermediate")}, SigAlg: "ecdsa-sha256", Signature: []byte("sig-1")},
+		{Nonce: []byte("nonce-2"), Chain: [][]byte{[]byte("other")}, SigAlg: "rsa-pss-sha256", Signature: []byte("sig-2")},
+	}}
+
+	meta := infoToMetaData(info)
+
+	require.Len(t, meta.GetCertificateProofs(), 2, "every proof reaches the meta")
+	first := meta.GetCertificateProofs()[0]
+	assert.Equal(t, []byte("nonce-1"), first.GetNonce())
+	assert.Equal(t, [][]byte{[]byte("leaf"), []byte("intermediate")}, first.GetChain(), "the whole chain is carried, in order")
+	assert.Equal(t, "ecdsa-sha256", first.GetSigAlg())
+	assert.Equal(t, []byte("sig-1"), first.GetSignature())
+	assert.Equal(t, "rsa-pss-sha256", meta.GetCertificateProofs()[1].GetSigAlg())
+
+	assert.Empty(t, infoToMetaData(&system.Info{}).GetCertificateProofs(), "a device without proofs sends none")
+}

shared/management/http/api/openapi.yml

@@ -1821,6 +1821,7 @@ components:
         ca_certificates:
           description: PEM encoded CA certificates the peer's certificate must chain to
           type: array
+          minItems: 1
           items:
             type: string
           example: ["-----BEGIN CERTIFICATE-----\nMIIB...\n-----END CERTIFICATE-----"]