Remove partial userspace firewall mode and open foreign chains via a table-less allower

client/firewall/allower_other.go

@@ -0,0 +1,11 @@
+//go:build android || (!linux && !windows)
+
+package firewall
+
+import "github.com/netbirdio/netbird/client/firewall/uspfilter"
+
+// interfaceAllower returns no allower: these platforms have no host firewall to
+// open for the interface.
+func interfaceAllower(IFaceMapper, uint16) uspfilter.InterfaceAllower {
+	return nil
+}

client/firewall/allower_windows.go

@@ -0,0 +1,10 @@
+//go:build windows
+
+package firewall
+
+import "github.com/netbirdio/netbird/client/firewall/uspfilter"
+
+// interfaceAllower returns the Windows netsh-based interface allower.
+func interfaceAllower(iface IFaceMapper, _ uint16) uspfilter.InterfaceAllower {
+	return uspfilter.NewWindowsInterfaceAllower(iface)
+}

client/firewall/create.go

@@ -6,8 +6,6 @@ import (
 	"fmt"
 	"runtime"
 
-	log "github.com/sirupsen/logrus"
-
 	firewall "github.com/netbirdio/netbird/client/firewall/manager"
 	"github.com/netbirdio/netbird/client/firewall/uspfilter"
 	nftypes "github.com/netbirdio/netbird/client/internal/netflow/types"
@@ -21,13 +19,11 @@ func NewFirewall(iface IFaceMapper, _ *statemanager.Manager, flowLogger nftypes.
 	}
 
 	// use userspace packet filtering firewall
-	fm, err := uspfilter.Create(iface, nil, disableServerRoutes, flowLogger, mtu)
-	if err != nil {
-		return nil, err
-	}
-	err = fm.AllowNetbird()
-	if err != nil {
-		log.Warnf("failed to allow netbird interface traffic: %v", err)
-	}
-	return fm, nil
+	return uspfilter.Create(uspfilter.Config{
+		IFace:               iface,
+		DisableServerRoutes: disableServerRoutes,
+		FlowLogger:          flowLogger,
+		MTU:                 mtu,
+		InterfaceAllower:    interfaceAllower(iface, mtu),
+	})
 }

client/firewall/create_linux.go

@@ -16,6 +16,7 @@ import (
 	firewall "github.com/netbirdio/netbird/client/firewall/manager"
 	nbnftables "github.com/netbirdio/netbird/client/firewall/nftables"
 	"github.com/netbirdio/netbird/client/firewall/uspfilter"
+	"github.com/netbirdio/netbird/client/iface/netstack"
 	nftypes "github.com/netbirdio/netbird/client/internal/netflow/types"
 	"github.com/netbirdio/netbird/client/internal/statemanager"
 )
@@ -40,19 +41,29 @@ var errNoFirewallManager = errors.New("no firewall manager found")
 type FWType int
 
 func NewFirewall(iface IFaceMapper, stateManager *statemanager.Manager, flowLogger nftypes.FlowLogger, disableServerRoutes bool, mtu uint16) (firewall.Manager, error) {
-	// We run in userspace mode and force userspace firewall was requested.
-	if iface.IsUserspaceBind() && forceUserspaceFirewall() {
-		nativeFw, err := createNativeFirewall(iface, stateManager, disableServerRoutes, mtu)
-		if err != nil {
-			log.Warnf("failed to create native firewall: %v. Proceeding without it", err)
+	// Userspace firewall without a native counterpart: routing is handled
+	// entirely in userspace. The interface is opened in the kernel's foreign
+	// filter chains via a table-less allower, except in netstack mode where no
+	// kernel interface exists.
+	if netstack.IsEnabled() || (iface.IsUserspaceBind() && forceUserspaceFirewall()) {
+		if netstack.IsEnabled() {
+			log.Info("netstack mode, using userspace firewall")
+		} else {
+			log.Info("forcing userspace firewall")
+		}
+		cfg := uspfilter.Config{
+			IFace:               iface,
+			DisableServerRoutes: disableServerRoutes,
+			FlowLogger:          flowLogger,
+			MTU:                 mtu,
+			InterfaceAllower:    interfaceAllower(iface, mtu),
 		}
 
-		log.Info("forcing userspace firewall")
-		return createUserspaceFirewall(iface, nativeFw, disableServerRoutes, flowLogger, mtu)
+		return uspfilter.Create(cfg)
 	}
 
 	// Use native firewall for either kernel or userspace, the interface appears identical to netfilter
-	fm, err := createNativeFirewall(iface, stateManager, disableServerRoutes, mtu)
+	fm, err := createNativeFirewall(iface, stateManager, mtu)
 	switch {
 	case err == nil && !iface.IsUserspaceBind():
 		// Nothing to do, fall through
@@ -69,24 +80,41 @@ func NewFirewall(iface IFaceMapper, stateManager *statemanager.Manager, flowLogg
 	case err != nil && iface.IsUserspaceBind():
 		// Fall back to the userspace packet filter if native is unavailable
 		logNativeFirewallUnavailable(err)
-		return createUserspaceFirewall(iface, nil, disableServerRoutes, flowLogger, mtu)
+		return uspfilter.Create(uspfilter.Config{
+			IFace:               iface,
+			DisableServerRoutes: disableServerRoutes,
+			FlowLogger:          flowLogger,
+			MTU:                 mtu,
+			InterfaceAllower:    interfaceAllower(iface, mtu),
+		})
 	}
 
 	return fm, nil
 }
 
-// createUserspaceFirewall builds the userspace packet filter, optionally
-// backed by a native firewall, and allows netbird interface traffic.
-func createUserspaceFirewall(iface IFaceMapper, nativeFw firewall.Manager, disableServerRoutes bool, flowLogger nftypes.FlowLogger, mtu uint16) (firewall.Manager, error) {
-	fm, err := uspfilter.Create(iface, nativeFw, disableServerRoutes, flowLogger, mtu)
-	if err != nil {
-		return nil, err
+// interfaceAllower selects how the userspace firewall opens the interface in
+// foreign kernel chains: nftables when available (which also opens foreign nft
+// tables), else iptables (the legacy fallback, filter INPUT only), else nil.
+// firewalld trust is applied separately by the manager. Netstack has no kernel
+// interface to open.
+func interfaceAllower(iface IFaceMapper, mtu uint16) uspfilter.InterfaceAllower {
+	if netstack.IsEnabled() {
+		return nil
 	}
 
-	if err := fm.AllowNetbird(); err != nil {
-		log.Errorf("failed to allow netbird interface traffic: %v", err)
+	nftAllower, err := nbnftables.NewInterfaceAllower(iface, mtu)
+	if err == nil {
+		return nftAllower
 	}
-	return fm, nil
+	log.Infof("no nftables interface allower: %v", err)
+
+	iptAllower, err := nbiptables.NewInterfaceAllower(iface)
+	if err == nil {
+		return iptAllower
+	}
+	log.Infof("no iptables interface allower: %v", err)
+
+	return nil
 }
 
 // logNativeFirewallUnavailable logs the fallback to userspace at info level
@@ -99,7 +127,7 @@ func logNativeFirewallUnavailable(err error) {
 	}
 }
 
-func createNativeFirewall(iface IFaceMapper, stateManager *statemanager.Manager, routes bool, mtu uint16) (firewall.Manager, error) {
+func createNativeFirewall(iface IFaceMapper, stateManager *statemanager.Manager, mtu uint16) (firewall.Manager, error) {
 	fm, err := createFW(iface, mtu)
 	if err != nil {
 		return nil, fmt.Errorf("create firewall: %w", err)
@@ -193,15 +221,21 @@ func isIptablesClientAvailable(client *iptables.IPTables) bool {
 	return err == nil
 }
 
+// forceUserspaceFirewall reports whether the userspace firewall is forced.
+// NB_FORCE_USERSPACE_ROUTER is an alias: forcing userspace routing implies the
+// userspace firewall, since the two are no longer separable.
 func forceUserspaceFirewall() bool {
-	val := os.Getenv(EnvForceUserspaceFirewall)
+	return envForceBool(EnvForceUserspaceFirewall) || envForceBool(uspfilter.EnvForceUserspaceRouter)
+}
+
+func envForceBool(name string) bool {
+	val := os.Getenv(name)
 	if val == "" {
 		return false
 	}
-
 	force, err := strconv.ParseBool(val)
 	if err != nil {
-		log.Warnf("failed to parse %s: %v", EnvForceUserspaceFirewall, err)
+		log.Warnf("failed to parse %s: %v", name, err)
 		return false
 	}
 	return force

client/firewall/iptables/interface_allower_linux.go

@@ -0,0 +1,93 @@
+package iptables
+
+import (
+	"fmt"
+
+	"github.com/coreos/go-iptables/iptables"
+	"github.com/hashicorp/go-multierror"
+	log "github.com/sirupsen/logrus"
+
+	nberrors "github.com/netbirdio/netbird/client/errors"
+)
+
+// InterfaceAllower opens the NetBird interface on the iptables filter INPUT
+// chain so the host firewall doesn't drop traffic the userspace firewall
+// handles. It is the fallback used when nftables is unavailable (an
+// iptables-legacy host).
+//
+// It opens INPUT only: the userspace router never forwards in the kernel.
+// firewalld trust is handled by the uspfilter manager, not here.
+type InterfaceAllower struct {
+	ifaceName string
+	ipt4      *iptables.IPTables
+	// ipt6 is nil when the interface has no IPv6 overlay address.
+	ipt6 *iptables.IPTables
+}
+
+// NewInterfaceAllower builds an iptables allower for the interface. It returns
+// an error when iptables is unavailable, so the caller can fall back to
+// firewalld trust.
+func NewInterfaceAllower(wgIface iFaceMapper) (*InterfaceAllower, error) {
+	ipt4, err := iptables.NewWithProtocol(iptables.ProtocolIPv4)
+	if err != nil {
+		return nil, fmt.Errorf("iptables not available: %w", err)
+	}
+	if _, err := ipt4.ListChains(tableFilter); err != nil {
+		return nil, fmt.Errorf("iptables filter table not available: %w", err)
+	}
+
+	a := &InterfaceAllower{ifaceName: wgIface.Name(), ipt4: ipt4}
+
+	// Missing v6 must not break the v4 path: open v4 only and continue.
+	if wgIface.Address().HasIPv6() {
+		ipt6, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
+		if err != nil {
+			log.Warnf("ip6tables not available, opening interface on v4 only: %v", err)
+		} else if _, err := ipt6.ListChains(tableFilter); err != nil {
+			log.Warnf("ip6tables filter table not available, opening interface on v4 only: %v", err)
+		} else {
+			a.ipt6 = ipt6
+		}
+	}
+
+	return a, nil
+}
+
+// Apply inserts the interface accept rule on the filter INPUT chain. It removes
+// any stale rule first so an unclean exit (e.g. SIGKILL, where Close never ran)
+// is recovered deterministically rather than accumulating duplicates.
+func (a *InterfaceAllower) Apply() error {
+	var merr *multierror.Error
+	for _, ipt := range a.clients() {
+		if err := ipt.DeleteIfExists(tableFilter, chainInput, a.inputRule()...); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("clean stale interface accept rule: %w", err))
+		}
+		if err := ipt.Insert(tableFilter, chainInput, 1, a.inputRule()...); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("add interface accept rule: %w", err))
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}
+
+// Close removes the interface accept rule.
+func (a *InterfaceAllower) Close() error {
+	var merr *multierror.Error
+	for _, ipt := range a.clients() {
+		if err := ipt.DeleteIfExists(tableFilter, chainInput, a.inputRule()...); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("remove interface accept rule: %w", err))
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}
+
+func (a *InterfaceAllower) inputRule() []string {
+	return []string{"-i", a.ifaceName, "-j", "ACCEPT"}
+}
+
+func (a *InterfaceAllower) clients() []*iptables.IPTables {
+	clients := []*iptables.IPTables{a.ipt4}
+	if a.ipt6 != nil {
+		clients = append(clients, a.ipt6)
+	}
+	return clients
+}

client/firewall/iptables/manager_linux.go

@@ -312,27 +312,6 @@ func (m *Manager) Close(stateManager *statemanager.Manager) error {
 	return nberrors.FormatErrorOrNil(merr)
 }
 
-// AllowNetbird allows netbird interface traffic.
-// This is called when USPFilter wraps the native firewall, adding blanket accept
-// rules so that packet filtering is handled in userspace instead of by netfilter.
-func (m *Manager) AllowNetbird() error {
-	var merr *multierror.Error
-	if _, err := m.AddFilterRule(nil, []netip.Prefix{netip.PrefixFrom(netip.IPv4Unspecified(), 0)}, firewall.Network{}, firewall.ProtocolALL, nil, nil, firewall.ActionAccept); err != nil {
-		merr = multierror.Append(merr, fmt.Errorf("allow netbird v4 interface traffic: %w", err))
-	}
-	if m.hasIPv6() {
-		if _, err := m.AddFilterRule(nil, []netip.Prefix{netip.PrefixFrom(netip.IPv6Unspecified(), 0)}, firewall.Network{}, firewall.ProtocolALL, nil, nil, firewall.ActionAccept); err != nil {
-			merr = multierror.Append(merr, fmt.Errorf("allow netbird v6 interface traffic: %w", err))
-		}
-	}
-
-	if err := firewalld.TrustInterface(m.wgIface.Name()); err != nil {
-		log.Warnf("failed to trust interface in firewalld: %v", err)
-	}
-
-	return nberrors.FormatErrorOrNil(merr)
-}
-
 // Flush doesn't need to be implemented for this manager
 func (m *Manager) Flush() error { return nil }
 

client/firewall/iptables/routing_linux.go

@@ -22,16 +22,14 @@ func (r *family) AddNatRule(pair firewall.RouterPair) error {
 		}
 	}
 
-	if !pair.Masquerade {
-		return nil
-	}
+	if pair.Masquerade {
+		if err := r.addNatRule(pair); err != nil {
+			return fmt.Errorf("add nat rule: %w", err)
+		}
 
-	if err := r.addNatRule(pair); err != nil {
-		return fmt.Errorf("add nat rule: %w", err)
-	}
-
-	if err := r.addNatRule(firewall.GetInversePair(pair)); err != nil {
-		return fmt.Errorf("add inverse nat rule: %w", err)
+		if err := r.addNatRule(firewall.GetInversePair(pair)); err != nil {
+			return fmt.Errorf("add inverse nat rule: %w", err)
+		}
 	}
 
 	r.updateState()
@@ -244,7 +242,6 @@ func (r *family) addNatRule(pair firewall.RouterPair) error {
 
 	r.rules[ruleID] = rule
 
-	r.updateState()
 	return nil
 }
 
@@ -264,6 +261,5 @@ func (r *family) removeNatRule(pair firewall.RouterPair) error {
 		log.Debugf("marking rule %s not found", ruleID)
 	}
 
-	r.updateState()
 	return nil
 }

client/firewall/manager/firewall.go

@@ -115,9 +115,6 @@ func (d Network) IsZero() bool {
 type Manager interface {
 	Init(stateManager *statemanager.Manager) error
 
-	// AllowNetbird allows netbird interface traffic
-	AllowNetbird() error
-
 	// AddFilterRule adds a packet-filtering rule to the firewall.
 	//
 	// If destination is the zero Network, the rule applies to traffic

client/firewall/nftables/chains_linux.go

@@ -83,8 +83,9 @@ func (r *family) createContainers() error {
 		log.Errorf("failed to add MSS clamping rules: %s", err)
 	}
 
-	if err := r.acceptForwardRules(); err != nil {
-		log.Errorf("failed to add accept rules for the forward chain: %s", err)
+	// Kernel routing opens both INPUT and FORWARD.
+	if err := r.openInterface(true); err != nil {
+		log.Errorf("failed to open interface in foreign chains: %s", err)
 	}
 
 	if err := firewalld.TrustInterface(r.wgIface.Name()); err != nil {
@@ -174,21 +175,26 @@ func (r *family) setupDataPlaneMark() error {
 	return nil
 }
 
-func (r *family) acceptForwardRules() error {
+// openInterface adds passthrough accept rules for the NetBird interface to the
+// kernel's filter table and external chains so they don't drop our traffic.
+// includeForward also opens the FORWARD chains (kernel routing); when false only
+// INPUT is opened, which is all the userspace router needs since it never
+// forwards in the kernel.
+func (r *family) openInterface(includeForward bool) error {
 	var merr *multierror.Error
 
-	if err := r.acceptFilterTableRules(); err != nil {
+	if err := r.acceptFilterTableRules(includeForward); err != nil {
 		merr = multierror.Append(merr, err)
 	}
 
-	if err := r.acceptExternalChainsRules(); err != nil {
+	if err := r.acceptExternalChainsRules(includeForward); err != nil {
 		merr = multierror.Append(merr, fmt.Errorf("add accept rules to external chains: %w", err))
 	}
 
 	return nberrors.FormatErrorOrNil(merr)
 }
 
-func (r *family) acceptFilterTableRules() error {
+func (r *family) acceptFilterTableRules(includeForward bool) error {
 	if r.filterTable == nil {
 		return nil
 	}
@@ -196,7 +202,7 @@ func (r *family) acceptFilterTableRules() error {
 	fw := "iptables"
 
 	defer func() {
-		log.Debugf("Used %s to add accept forward and input rules", fw)
+		log.Debugf("Used %s to add accept input/forward rules", fw)
 	}()
 
 	// Try iptables first and fallback to nftables if iptables is not available.
@@ -206,25 +212,27 @@ func (r *family) acceptFilterTableRules() error {
 		log.Warnf("Will use nftables to manipulate the filter table because iptables is not available: %v", err)
 
 		fw = "nftables"
-		return r.acceptFilterRulesNftables(r.filterTable)
+		return r.acceptFilterRulesNftables(r.filterTable, includeForward)
 	}
 
-	if err := r.acceptFilterRulesIptables(ipt); err != nil {
+	if err := r.acceptFilterRulesIptables(ipt, includeForward); err != nil {
 		log.Warnf("iptables failed (table may be incompatible), falling back to nftables: %v", err)
 		fw = "nftables"
-		return r.acceptFilterRulesNftables(r.filterTable)
+		return r.acceptFilterRulesNftables(r.filterTable, includeForward)
 	}
 	return nil
 }
 
-func (r *family) acceptFilterRulesIptables(ipt *iptables.IPTables) error {
+func (r *family) acceptFilterRulesIptables(ipt *iptables.IPTables, includeForward bool) error {
 	var merr *multierror.Error
 
-	for _, rule := range r.getAcceptForwardRules() {
-		if err := ipt.Insert("filter", chainNameForward, 1, rule...); err != nil {
-			merr = multierror.Append(merr, fmt.Errorf("add iptables forward rule: %v", err))
-		} else {
-			log.Debugf("added iptables forward rule: %v", rule)
+	if includeForward {
+		for _, rule := range r.getAcceptForwardRules() {
+			if err := ipt.Insert("filter", chainNameForward, 1, rule...); err != nil {
+				merr = multierror.Append(merr, fmt.Errorf("add iptables forward rule: %v", err))
+			} else {
+				log.Debugf("added iptables forward rule: %v", rule)
+			}
 		}
 	}
 
@@ -252,17 +260,19 @@ func (r *family) getAcceptInputRule() []string {
 
 // acceptFilterRulesNftables adds accept rules to the ip filter table using nftables.
 // This is used when iptables is not available.
-func (r *family) acceptFilterRulesNftables(table *nftables.Table) error {
+func (r *family) acceptFilterRulesNftables(table *nftables.Table, includeForward bool) error {
 	intf := ifname(r.wgIface.Name())
 
-	forwardChain := &nftables.Chain{
-		Name:     chainNameForward,
-		Table:    table,
-		Type:     nftables.ChainTypeFilter,
-		Hooknum:  nftables.ChainHookForward,
-		Priority: nftables.ChainPriorityFilter,
+	if includeForward {
+		forwardChain := &nftables.Chain{
+			Name:     chainNameForward,
+			Table:    table,
+			Type:     nftables.ChainTypeFilter,
+			Hooknum:  nftables.ChainHookForward,
+			Priority: nftables.ChainPriorityFilter,
+		}
+		r.insertForwardAcceptRules(forwardChain, intf)
 	}
-	r.insertForwardAcceptRules(forwardChain, intf)
 
 	inputChain := &nftables.Chain{
 		Name:     chainNameInput,
@@ -278,7 +288,7 @@ func (r *family) acceptFilterRulesNftables(table *nftables.Table) error {
 
 // acceptExternalChainsRules adds accept rules to external chains (non-netbird, non-iptables tables).
 // It dynamically finds chains at call time to handle chains that may have been created after startup.
-func (r *family) acceptExternalChainsRules() error {
+func (r *family) acceptExternalChainsRules(includeForward bool) error {
 	chains := r.findExternalChains()
 	if len(chains) == 0 {
 		return nil
@@ -286,7 +296,7 @@ func (r *family) acceptExternalChainsRules() error {
 
 	intf := ifname(r.wgIface.Name())
 	for _, chain := range chains {
-		r.applyExternalChainAccept(chain, intf)
+		r.applyExternalChainAccept(chain, intf, includeForward)
 	}
 
 	if err := r.conn.Flush(); err != nil {
@@ -295,7 +305,7 @@ func (r *family) acceptExternalChainsRules() error {
 	return nil
 }
 
-func (r *family) applyExternalChainAccept(chain *nftables.Chain, intf []byte) {
+func (r *family) applyExternalChainAccept(chain *nftables.Chain, intf []byte, includeForward bool) {
 	if chain.Hooknum == nil {
 		log.Debugf("skipping external chain %s/%s: hooknum is nil", chain.Table.Name, chain.Name)
 		return
@@ -306,7 +316,9 @@ func (r *family) applyExternalChainAccept(chain *nftables.Chain, intf []byte) {
 
 	switch *chain.Hooknum {
 	case *nftables.ChainHookForward:
-		r.insertForwardAcceptRules(chain, intf)
+		if includeForward {
+			r.insertForwardAcceptRules(chain, intf)
+		}
 	case *nftables.ChainHookInput:
 		r.insertInputAcceptRule(chain, intf)
 	}
@@ -576,26 +588,6 @@ func (r *family) removeAcceptFilterRulesIptables(ipt *iptables.IPTables) error {
 	return nberrors.FormatErrorOrNil(merr)
 }
 
-func (r *family) createDefaultAllowRules() error {
-	expIn := []expr.Any{
-		&expr.Verdict{
-			Kind: expr.VerdictAccept,
-		},
-	}
-
-	_ = r.conn.InsertRule(&nftables.Rule{
-		Table:    r.workTable,
-		Chain:    r.chainInputRules,
-		Position: 0,
-		Exprs:    expIn,
-	})
-
-	if err := r.conn.Flush(); err != nil {
-		return fmt.Errorf(flushError, err)
-	}
-	return nil
-}
-
 // Flush rule/chain/set operations from the buffer
 //
 // Method also get all rules after flush and refreshes handle values in the rulesets

client/firewall/nftables/interface_allower_integration_linux_test.go

@@ -0,0 +1,90 @@
+//go:build integration && !android
+
+package nftables
+
+import (
+	"bytes"
+	"os"
+	"testing"
+
+	"github.com/google/nftables"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/iface"
+)
+
+// TestInterfaceAllowerInputOnly verifies the userspace-mode allower opens the
+// interface on the INPUT hook of foreign chains only (not FORWARD, since the
+// userspace router never forwards in the kernel), creates no netbird work
+// table, and removes its rules on Close.
+func TestInterfaceAllowerInputOnly(t *testing.T) {
+	if os.Geteuid() != 0 {
+		t.Skip("root required")
+	}
+
+	require.False(t, ipTableExists(t, getTableName()), "precondition: no stale netbird table")
+
+	conn := &nftables.Conn{}
+	extTable := conn.AddTable(&nftables.Table{Name: "nbtest_extchains", Family: nftables.TableFamilyINet})
+	inputChain := conn.AddChain(&nftables.Chain{
+		Name: "ext_input", Table: extTable,
+		Hooknum: nftables.ChainHookInput, Priority: nftables.ChainPriorityFilter, Type: nftables.ChainTypeFilter,
+	})
+	forwardChain := conn.AddChain(&nftables.Chain{
+		Name: "ext_forward", Table: extTable,
+		Hooknum: nftables.ChainHookForward, Priority: nftables.ChainPriorityFilter, Type: nftables.ChainTypeFilter,
+	})
+	require.NoError(t, conn.Flush(), "create external table and chains")
+	t.Cleanup(func() {
+		c := &nftables.Conn{}
+		c.DelTable(extTable)
+		_ = c.Flush()
+	})
+
+	allower, err := NewInterfaceAllower(ifaceMock, iface.DefaultMTU)
+	require.NoError(t, err, "create allower")
+	require.NoError(t, allower.Apply(), "apply")
+
+	require.True(t, chainHasUserData(t, extTable, inputChain, userDataAcceptInputRule),
+		"external INPUT chain should get the accept rule")
+	require.Len(t, listRules(t, extTable, forwardChain), 0,
+		"external FORWARD chain must not be opened in userspace mode")
+	require.False(t, ipTableExists(t, getTableName()),
+		"allower must not create a netbird work table")
+
+	require.NoError(t, allower.Close(), "close")
+	require.False(t, chainHasUserData(t, extTable, inputChain, userDataAcceptInputRule),
+		"accept rule should be removed on close")
+}
+
+func listRules(t *testing.T, table *nftables.Table, chain *nftables.Chain) []*nftables.Rule {
+	t.Helper()
+	c := &nftables.Conn{}
+	rules, err := c.GetRules(table, chain)
+	require.NoError(t, err)
+	return rules
+}
+
+func chainHasUserData(t *testing.T, table *nftables.Table, chain *nftables.Chain, ud string) bool {
+	for _, r := range listRules(t, table, chain) {
+		if bytes.Equal(r.UserData, []byte(ud)) {
+			return true
+		}
+	}
+	return false
+}
+
+func ipTableExists(t *testing.T, name string) bool {
+	t.Helper()
+	c := &nftables.Conn{}
+	for _, fam := range []nftables.TableFamily{nftables.TableFamilyIPv4, nftables.TableFamilyIPv6} {
+		tbls, err := c.ListTablesOfFamily(fam)
+		require.NoError(t, err)
+		for _, tb := range tbls {
+			if tb.Name == name {
+				return true
+			}
+		}
+	}
+	return false
+}

client/firewall/nftables/interface_allower_linux.go

@@ -0,0 +1,114 @@
+package nftables
+
+import (
+	"fmt"
+
+	"github.com/google/nftables"
+	"github.com/hashicorp/go-multierror"
+
+	nberrors "github.com/netbirdio/netbird/client/errors"
+)
+
+// InterfaceAllower opens the NetBird interface in the kernel's filter table and
+// external chains and keeps them reconciled via a netlink monitor, so the host
+// firewall doesn't drop traffic the NetBird firewall handles. It is used by the
+// userspace firewall, where routing happens in the forwarder, so only INPUT is
+// opened (the userspace router never forwards in the kernel).
+//
+// It owns its own families/connection and never creates a netbird work table.
+// firewalld trust is handled by the caller, not here. Its operations are serial
+// (Apply before the monitor starts; reconciles run on the single monitor
+// goroutine; Close stops the monitor before removing), so it needs no locking.
+//
+// TODO: this opens nftables and the iptables-nft filter table (detected via
+// nft), but not a legacy-iptables ruleset running in parallel with nftables.
+// Such a host would keep its legacy filter chains closed for the interface.
+type InterfaceAllower struct {
+	family4    *family
+	family6    *family
+	extMonitor *externalChainMonitor
+}
+
+// NewInterfaceAllower builds an allower for the given interface. It returns an
+// error when nftables is unavailable (e.g. an iptables-legacy host), so the
+// caller can fall back to firewalld trust.
+func NewInterfaceAllower(wgIface iFaceMapper, mtu uint16) (*InterfaceAllower, error) {
+	tableName := getTableName()
+
+	family4, err := newFamily(&nftables.Table{Name: tableName, Family: nftables.TableFamilyIPv4}, wgIface, mtu)
+	if err != nil {
+		return nil, fmt.Errorf("create family: %w", err)
+	}
+
+	// Probe nftables availability before committing to this backend.
+	if _, err := family4.conn.ListChainsOfTableFamily(nftables.TableFamilyINet); err != nil {
+		return nil, fmt.Errorf("nftables not available: %w", err)
+	}
+
+	a := &InterfaceAllower{family4: family4}
+
+	if wgIface.Address().HasIPv6() {
+		family6, err := newFamily(&nftables.Table{Name: tableName, Family: nftables.TableFamilyIPv6}, wgIface, mtu)
+		if err != nil {
+			return nil, fmt.Errorf("create v6 family: %w", err)
+		}
+		a.family6 = family6
+	}
+
+	a.extMonitor = newExternalChainMonitor(a)
+	return a, nil
+}
+
+// Apply opens the interface (INPUT only) in the foreign filter chains and starts
+// reconciling them on nftables changes.
+func (a *InterfaceAllower) Apply() error {
+	var merr *multierror.Error
+	for _, f := range a.families() {
+		// Remove any stale accepts first so a prior unclean exit (e.g. SIGKILL,
+		// where Close never ran) is recovered deterministically rather than
+		// accumulating duplicate rules on the iptables filter table.
+		if err := f.removeAcceptFilterRules(); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("clean stale accept rules: %w", err))
+		}
+		if err := f.openInterface(false); err != nil {
+			merr = multierror.Append(merr, err)
+		}
+	}
+
+	a.extMonitor.start()
+	return nberrors.FormatErrorOrNil(merr)
+}
+
+// families returns the configured address families (v4, and v6 when present).
+func (a *InterfaceAllower) families() []*family {
+	families := []*family{a.family4}
+	if a.family6 != nil {
+		families = append(families, a.family6)
+	}
+	return families
+}
+
+// reconcileExternalChains re-applies the INPUT accepts to external chains. It
+// implements externalChainReconciler for the monitor.
+func (a *InterfaceAllower) reconcileExternalChains() error {
+	var merr *multierror.Error
+	for _, f := range a.families() {
+		if err := f.acceptExternalChainsRules(false); err != nil {
+			merr = multierror.Append(merr, err)
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}
+
+// Close stops the monitor and removes the accept rules.
+func (a *InterfaceAllower) Close() error {
+	a.extMonitor.stop()
+
+	var merr *multierror.Error
+	for _, f := range a.families() {
+		if err := f.removeAcceptFilterRules(); err != nil {
+			merr = multierror.Append(merr, err)
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}

client/firewall/nftables/manager_linux.go

@@ -15,7 +15,6 @@ import (
 	"golang.org/x/sys/unix"
 
 	nberrors "github.com/netbirdio/netbird/client/errors"
-	"github.com/netbirdio/netbird/client/firewall/firewalld"
 	firewall "github.com/netbirdio/netbird/client/firewall/manager"
 	"github.com/netbirdio/netbird/client/iface/wgaddr"
 	"github.com/netbirdio/netbird/client/internal/statemanager"
@@ -140,19 +139,20 @@ func (m *Manager) Init(stateManager *statemanager.Manager) error {
 
 // reconcileExternalChains re-applies passthrough accept rules to external
 // filter chains for both IPv4 and IPv6 routers. Called by the monitor when
-// tables or chains appear (e.g. after firewalld reloads).
+// tables or chains appear (e.g. after firewalld reloads). Kernel routing opens
+// both INPUT and FORWARD.
 func (m *Manager) reconcileExternalChains() error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
 	var merr *multierror.Error
 	if m.family4 != nil {
-		if err := m.family4.acceptExternalChainsRules(); err != nil {
+		if err := m.family4.acceptExternalChainsRules(true); err != nil {
 			merr = multierror.Append(merr, fmt.Errorf("v4: %w", err))
 		}
 	}
 	if m.hasIPv6() {
-		if err := m.family6.acceptExternalChainsRules(); err != nil {
+		if err := m.family6.acceptExternalChainsRules(true); err != nil {
 			merr = multierror.Append(merr, fmt.Errorf("v6: %w", err))
 		}
 	}
@@ -370,39 +370,6 @@ func (m *Manager) RemoveNatRule(pair firewall.RouterPair) error {
 	return nberrors.FormatErrorOrNil(merr)
 }
 
-// AllowNetbird allows netbird interface traffic.
-// This is called when USPFilter wraps the native firewall, adding blanket accept
-// rules so that packet filtering is handled in userspace instead of by netfilter.
-//
-// TODO: In USP mode this only adds ACCEPT to the netbird table's own chains,
-// which doesn't override DROP rules in external tables (e.g. firewalld).
-// Should add passthrough rules to external chains (like the native mode router's
-// addExternalChainsRules does) for both the netbird table family and inet tables.
-// The netbird table itself is fine (routing chains already exist there), but
-// non-netbird tables with INPUT/FORWARD hooks can still DROP our WG traffic.
-func (m *Manager) AllowNetbird() error {
-	m.mutex.Lock()
-	defer m.mutex.Unlock()
-
-	if err := m.family4.createDefaultAllowRules(); err != nil {
-		return fmt.Errorf("create default allow rules: %w", err)
-	}
-	if m.hasIPv6() {
-		if err := m.family6.createDefaultAllowRules(); err != nil {
-			return fmt.Errorf("create v6 default allow rules: %w", err)
-		}
-	}
-	if err := m.rConn.Flush(); err != nil {
-		return fmt.Errorf("flush allow input netbird rules: %w", err)
-	}
-
-	if err := firewalld.TrustInterface(m.wgIface.Name()); err != nil {
-		log.Warnf("failed to trust interface in firewalld: %v", err)
-	}
-
-	return nil
-}
-
 // SetLegacyManagement sets the route manager to use legacy management
 func (m *Manager) SetLegacyManagement(isLegacy bool) error {
 	if err := firewall.SetLegacyManagement(m.family4, isLegacy); err != nil {

client/firewall/uspfilter/allow_netbird.go

@@ -1,37 +0,0 @@
-//go:build !windows
-
-package uspfilter
-
-import (
-	log "github.com/sirupsen/logrus"
-
-	"github.com/netbirdio/netbird/client/firewall/firewalld"
-	"github.com/netbirdio/netbird/client/internal/statemanager"
-)
-
-// Close cleans up the firewall manager by removing all rules and closing trackers
-func (m *Manager) Close(stateManager *statemanager.Manager) error {
-	m.mutex.Lock()
-	defer m.mutex.Unlock()
-
-	m.resetState()
-
-	if m.nativeFirewall != nil {
-		return m.nativeFirewall.Close(stateManager)
-	}
-	if err := firewalld.UntrustInterface(m.wgIface.Name()); err != nil {
-		log.Warnf("failed to untrust interface in firewalld: %v", err)
-	}
-	return nil
-}
-
-// AllowNetbird allows netbird interface traffic
-func (m *Manager) AllowNetbird() error {
-	if m.nativeFirewall != nil {
-		return m.nativeFirewall.AllowNetbird()
-	}
-	if err := firewalld.TrustInterface(m.wgIface.Name()); err != nil {
-		log.Warnf("failed to trust interface in firewalld: %v", err)
-	}
-	return nil
-}

client/firewall/uspfilter/common/iface.go

@@ -1,17 +0,0 @@
-package common
-
-import (
-	wgdevice "golang.zx2c4.com/wireguard/device"
-
-	"github.com/netbirdio/netbird/client/iface/device"
-	"github.com/netbirdio/netbird/client/iface/wgaddr"
-)
-
-// IFaceMapper defines subset methods of interface required for manager
-type IFaceMapper interface {
-	Name() string
-	SetFilter(device.PacketFilter) error
-	Address() wgaddr.Address
-	GetWGDevice() *wgdevice.Device
-	GetDevice() *device.FilteredDevice
-}

client/firewall/uspfilter/filter.go

@@ -19,14 +19,18 @@ import (
 	"github.com/google/uuid"
 	"github.com/hashicorp/go-multierror"
 	log "github.com/sirupsen/logrus"
+	wgdevice "golang.zx2c4.com/wireguard/device"
 
 	nberrors "github.com/netbirdio/netbird/client/errors"
+	"github.com/netbirdio/netbird/client/firewall/firewalld"
 	firewall "github.com/netbirdio/netbird/client/firewall/manager"
 	"github.com/netbirdio/netbird/client/firewall/uspfilter/common"
 	"github.com/netbirdio/netbird/client/firewall/uspfilter/conntrack"
 	"github.com/netbirdio/netbird/client/firewall/uspfilter/forwarder"
 	nblog "github.com/netbirdio/netbird/client/firewall/uspfilter/log"
+	"github.com/netbirdio/netbird/client/iface/device"
 	"github.com/netbirdio/netbird/client/iface/netstack"
+	"github.com/netbirdio/netbird/client/iface/wgaddr"
 	nbid "github.com/netbirdio/netbird/client/internal/acl/id"
 	nftypes "github.com/netbirdio/netbird/client/internal/netflow/types"
 	"github.com/netbirdio/netbird/client/internal/statemanager"
@@ -57,7 +61,10 @@ const (
 	// EnvDisableMSSClamping disables TCP MSS clamping for forwarded traffic.
 	EnvDisableMSSClamping = "NB_DISABLE_MSS_CLAMPING"
 
-	// EnvForceUserspaceRouter forces userspace routing even if native routing is available.
+	// EnvForceUserspaceRouter is a deprecated alias for
+	// NB_FORCE_USERSPACE_FIREWALL: the userspace firewall always routes in
+	// userspace, so forcing one forces the other. Kept for backward
+	// compatibility.
 	EnvForceUserspaceRouter = "NB_FORCE_USERSPACE_ROUTER"
 
 	// EnvEnableLocalForwarding enables forwarding of local traffic to the native stack for internal (non-NetBird) interfaces.
@@ -69,7 +76,10 @@ const (
 	EnvEnableNetstackLocalForwarding = "NB_ENABLE_NETSTACK_LOCAL_FORWARDING"
 )
 
-var errNatNotSupported = errors.New("nat not supported with userspace firewall")
+// errNotSupported is returned by firewall operations that only make sense with
+// a kernel firewall (kernel NAT/DNAT, eBPF) and are not implemented in
+// userspace mode, where they should not be called.
+var errNotSupported = errors.New("not supported with userspace firewall")
 
 // peerRules is the canonical list-based storage for peer ACL rules.
 // Match order is significant: drop rules come before accept rules so
@@ -105,17 +115,48 @@ type peerRuleSpec struct {
 	action   firewall.Action
 }
 
+// Iface is the network interface the userspace firewall attaches to: the
+// methods of the WireGuard device it actually uses.
+type Iface interface {
+	Name() string
+	Address() wgaddr.Address
+	SetFilter(device.PacketFilter) error
+	GetWGDevice() *wgdevice.Device
+}
+
+// InterfaceAllower opens the NetBird interface in the host firewall so it
+// doesn't drop traffic the userspace firewall handles, without taking over
+// packet filtering. Implementations (nftables, iptables, firewalld, the windows
+// netsh rule) are selected per platform and injected into Create; Apply runs at
+// creation and Close on teardown.
+type InterfaceAllower interface {
+	Apply() error
+	Close() error
+}
+
+// Config holds the dependencies and options for the userspace firewall.
+type Config struct {
+	// IFace is the overlay interface the filter attaches to.
+	IFace Iface
+	// InterfaceAllower opens the NetBird interface in foreign kernel filter
+	// chains so the kernel doesn't drop traffic the userspace firewall handles.
+	// Nil in netstack mode, on non-Linux platforms without a backend, or when
+	// neither nftables nor iptables is available. firewalld trust is applied by
+	// the manager regardless, since firewalld owns its own chains and we cannot
+	// insert into them.
+	InterfaceAllower InterfaceAllower
+	// DisableServerRoutes indicates whether server routes are disabled.
+	DisableServerRoutes bool
+	FlowLogger          nftypes.FlowLogger
+	MTU                 uint16
+}
+
 // Manager userspace firewall manager
 type Manager struct {
-	decoders sync.Pool
-	wgIface  common.IFaceMapper
-	// nativeFirewall is the kernel firewall (nftables/iptables) used for
-	// the split case where peer ACLs run in userspace here but routing
-	// stays in the kernel: when the userspace firewall is forced yet the
-	// router keeps using the kernel, route NAT/ACLs and DNAT are
-	// delegated to it. It is nil when no native backend is available.
-	nativeFirewall firewall.Manager
-	mutex          sync.RWMutex
+	decoders     sync.Pool
+	wgIface      Iface
+	ifaceAllower InterfaceAllower
+	mutex        sync.RWMutex
 
 	incomingDenyRules   peerRules
 	incomingAcceptRules peerRules
@@ -236,7 +277,7 @@ func parseCreateEnv() (bool, bool, bool) {
 	return disableConntrack, enableLocalForwarding, disableMSSClamping
 }
 
-func Create(iface common.IFaceMapper, nativeFirewall firewall.Manager, disableServerRoutes bool, flowLogger nftypes.FlowLogger, mtu uint16) (*Manager, error) {
+func Create(cfg Config) (_ *Manager, err error) {
 	disableConntrack, enableLocalForwarding, disableMSSClamping := parseCreateEnv()
 
 	m := &Manager{
@@ -259,13 +300,13 @@ func Create(iface common.IFaceMapper, nativeFirewall firewall.Manager, disableSe
 				return d
 			},
 		},
-		wgIface:             iface,
-		nativeFirewall:      nativeFirewall,
+		wgIface:             cfg.IFace,
+		ifaceAllower:        cfg.InterfaceAllower,
 		localipmanager:      newLocalIPManager(),
-		disableServerRoutes: disableServerRoutes,
+		disableServerRoutes: cfg.DisableServerRoutes,
 		stateful:            !disableConntrack,
 		logger:              nblog.NewFromLogrus(log.StandardLogger()),
-		flowLogger:          flowLogger,
+		flowLogger:          cfg.FlowLogger,
 		netstack:            netstack.IsEnabled(),
 		localForwarding:     enableLocalForwarding,
 		peerRulesMap:        make(map[nbid.RuleID]*PeerRule),
@@ -273,46 +314,117 @@ func Create(iface common.IFaceMapper, nativeFirewall firewall.Manager, disableSe
 		dnatMappings:        make(map[netip.Addr]netip.Addr),
 		portDNATRules:       []portDNATRule{},
 		netstackServices:    make(map[serviceKey]struct{}),
-		mtu:                 mtu,
+		mtu:                 cfg.MTU,
 	}
 	m.routingEnabled.Store(false)
 
+	// Release the allower (and its monitor) if setup fails after it was wired in.
+	defer func() {
+		if err != nil {
+			m.closeAllowerOnError()
+		}
+	}()
+
 	if !disableMSSClamping {
-		m.mssClampEnabled = true
-		if mtu > ipv4TCPHeaderMinSize {
-			m.mssClampValueIPv4 = mtu - ipv4TCPHeaderMinSize
-		}
-		if mtu > ipv6TCPHeaderMinSize {
-			m.mssClampValueIPv6 = mtu - ipv6TCPHeaderMinSize
-		}
+		m.enableMSSClamping(cfg.MTU)
 	}
-	if err := m.localipmanager.UpdateLocalIPs(iface); err != nil {
+	if err := m.localipmanager.UpdateLocalIPs(cfg.IFace); err != nil {
 		return nil, fmt.Errorf("update local IPs: %w", err)
 	}
-	if disableConntrack {
-		log.Info("conntrack is disabled")
-	} else {
-		m.udpTracker = conntrack.NewUDPTracker(conntrack.DefaultUDPTimeout, m.logger, flowLogger)
-		m.icmpTracker = conntrack.NewICMPTracker(conntrack.DefaultICMPTimeout, m.logger, flowLogger)
-		m.tcpTracker = conntrack.NewTCPTracker(conntrack.DefaultTCPTimeout, m.logger, flowLogger)
-	}
+	m.setupConntrack(disableConntrack)
 	if m.netstack && m.localForwarding {
 		if err := m.initForwarder(); err != nil {
 			log.Errorf("failed to initialize forwarder: %v", err)
 		}
 	}
-	if err := iface.SetFilter(m); err != nil {
+	if err := cfg.IFace.SetFilter(m); err != nil {
 		return nil, fmt.Errorf("set filter: %w", err)
 	}
+
+	m.openHostFirewall(cfg.IFace.Name())
+
 	return m, nil
 }
 
+// closeAllowerOnError releases the allower (and its monitor) when Create fails
+// after the allower was wired in.
+func (m *Manager) closeAllowerOnError() {
+	if m.ifaceAllower == nil {
+		return
+	}
+	if err := m.ifaceAllower.Close(); err != nil {
+		log.Warnf("close interface allower after failed firewall setup: %v", err)
+	}
+}
+
+// enableMSSClamping enables MSS clamping and computes the per-family clamp values.
+func (m *Manager) enableMSSClamping(mtu uint16) {
+	m.mssClampEnabled = true
+	if mtu > ipv4TCPHeaderMinSize {
+		m.mssClampValueIPv4 = mtu - ipv4TCPHeaderMinSize
+	}
+	if mtu > ipv6TCPHeaderMinSize {
+		m.mssClampValueIPv6 = mtu - ipv6TCPHeaderMinSize
+	}
+}
+
+// setupConntrack initializes the stateful trackers unless conntrack is disabled.
+func (m *Manager) setupConntrack(disabled bool) {
+	if disabled {
+		log.Info("conntrack is disabled")
+		return
+	}
+	m.udpTracker = conntrack.NewUDPTracker(conntrack.DefaultUDPTimeout, m.logger, m.flowLogger)
+	m.icmpTracker = conntrack.NewICMPTracker(conntrack.DefaultICMPTimeout, m.logger, m.flowLogger)
+	m.tcpTracker = conntrack.NewTCPTracker(conntrack.DefaultTCPTimeout, m.logger, m.flowLogger)
+}
+
+// openHostFirewall opens the NetBird interface in the kernel firewall so it
+// doesn't drop traffic the userspace firewall handles. Best-effort: failures
+// here shouldn't prevent the firewall from coming up.
+func (m *Manager) openHostFirewall(ifaceName string) {
+	if m.ifaceAllower != nil {
+		if err := m.ifaceAllower.Apply(); err != nil {
+			log.Errorf("failed to allow netbird interface traffic: %v", err)
+		}
+	}
+	// firewalld owns its own chains we can't insert into, so trust the interface
+	// there in addition to the allower. Netstack has no kernel interface.
+	if !m.netstack {
+		if err := firewalld.TrustInterface(ifaceName); err != nil {
+			log.Warnf("failed to trust interface in firewalld: %v", err)
+		}
+	}
+}
+
+// Close cleans up the firewall manager: removes rules, closes trackers, and
+// closes the interface allower.
+func (m *Manager) Close(*statemanager.Manager) error {
+	m.mutex.Lock()
+	defer m.mutex.Unlock()
+
+	m.resetState()
+
+	var merr *multierror.Error
+	if m.ifaceAllower != nil {
+		if err := m.ifaceAllower.Close(); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("close interface allower: %w", err))
+		}
+	}
+	if !m.netstack {
+		if err := firewalld.UntrustInterface(m.wgIface.Name()); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("untrust interface in firewalld: %w", err))
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}
+
 // blockInvalidRouted installs drop rules for traffic to the wg overlay that
 // arrives via the routing path. v4 and v6 are independent: a v6 install
 // failure leaves v4 protection in place (and vice versa) so the returned
 // slice always contains whatever was successfully installed, even on error.
 // Callers must persist the slice so DisableRouting can clean partial state.
-func (m *Manager) blockInvalidRouted(iface common.IFaceMapper) ([]firewall.Rule, error) {
+func (m *Manager) blockInvalidRouted(iface Iface) ([]firewall.Rule, error) {
 	wgPrefix := iface.Address().Network
 	log.Debugf("blocking invalid routed traffic for %s", wgPrefix)
 
@@ -360,20 +472,14 @@ func (m *Manager) blockInvalidRouted(iface common.IFaceMapper) ([]firewall.Rule,
 }
 
 func (m *Manager) determineRouting() error {
-	var disableUspRouting, forceUserspaceRouter bool
-	var err error
+	var disableUspRouting bool
 	if val := os.Getenv(EnvDisableUserspaceRouting); val != "" {
+		var err error
 		disableUspRouting, err = strconv.ParseBool(val)
 		if err != nil {
 			log.Warnf("failed to parse %s: %v", EnvDisableUserspaceRouting, err)
 		}
 	}
-	if val := os.Getenv(EnvForceUserspaceRouter); val != "" {
-		forceUserspaceRouter, err = strconv.ParseBool(val)
-		if err != nil {
-			log.Warnf("failed to parse %s: %v", EnvForceUserspaceRouter, err)
-		}
-	}
 
 	switch {
 	case disableUspRouting:
@@ -388,26 +494,11 @@ func (m *Manager) determineRouting() error {
 
 		log.Info("server routes are disabled")
 
-	case forceUserspaceRouter:
-		m.routingEnabled.Store(true)
-		m.nativeRouter.Store(false)
-
-		log.Info("userspace routing is forced")
-
-	case !m.netstack && m.nativeFirewall != nil:
-		// if the OS supports routing natively, then we don't need to filter/route ourselves
-		// netstack mode won't support native routing as there is no interface
-
-		m.routingEnabled.Store(true)
-		m.nativeRouter.Store(true)
-
-		log.Info("native routing is enabled")
-
 	default:
 		m.routingEnabled.Store(true)
 		m.nativeRouter.Store(false)
 
-		log.Info("userspace routing enabled by default")
+		log.Info("userspace routing enabled")
 	}
 
 	if m.routingEnabled.Load() && !m.nativeRouter.Load() {
@@ -473,21 +564,14 @@ func (m *Manager) IsStateful() bool {
 	return m.stateful
 }
 
-func (m *Manager) AddNatRule(pair firewall.RouterPair) error {
-	if m.nativeRouter.Load() && m.nativeFirewall != nil {
-		return m.nativeFirewall.AddNatRule(pair)
-	}
-
+func (m *Manager) AddNatRule(firewall.RouterPair) error {
 	// userspace routed packets are always SNATed to the inbound direction
 	// TODO: implement outbound SNAT
 	return nil
 }
 
 // RemoveNatRule removes a routing firewall rule
-func (m *Manager) RemoveNatRule(pair firewall.RouterPair) error {
-	if m.nativeRouter.Load() && m.nativeFirewall != nil {
-		return m.nativeFirewall.RemoveNatRule(pair)
-	}
+func (m *Manager) RemoveNatRule(firewall.RouterPair) error {
 	return nil
 }
 
@@ -651,10 +735,7 @@ func (m *Manager) DeleteFilterRule(rule firewall.Rule) error {
 		return m.deletePeerRuleLocked(r)
 	}
 
-	// Either our *RouteRule or, under native delegation, a native
-	// route-rule object that implements firewall.Rule but isn't one of
-	// our concrete types. The route path forwards the latter to the
-	// native firewall.
+	// Anything else is a route rule (matched on the forward path).
 	return m.deleteRouteRule(rule)
 }
 
@@ -666,10 +747,6 @@ func (m *Manager) addRouteRule(
 	sPort, dPort *firewall.Port,
 	action firewall.Action,
 ) (firewall.Rule, error) {
-	if m.nativeRouter.Load() && m.nativeFirewall != nil {
-		return m.nativeFirewall.AddFilterRule(id, sources, destination, proto, sPort, dPort, action)
-	}
-
 	ruleID := nbid.GenerateRuleID(sources, destination, proto, sPort, dPort, action)
 
 	if existingRule, ok := m.routeRulesMap[ruleID]; ok {
@@ -698,10 +775,6 @@ func (m *Manager) addRouteRule(
 }
 
 func (m *Manager) deleteRouteRule(rule firewall.Rule) error {
-	if m.nativeRouter.Load() && m.nativeFirewall != nil {
-		return m.nativeFirewall.DeleteFilterRule(rule)
-	}
-
 	ruleID := rule.ID()
 	trimmed, _, ok := removeRuleByID(m.routeRules, ruleID)
 	if !ok {
@@ -743,12 +816,11 @@ func removeRuleByID[S ~[]T, T firewall.Rule](rules S, ruleID firewall.RuleID) (S
 	return slices.Delete(rules, idx, idx+1), removed, true
 }
 
-// SetLegacyManagement doesn't need to be implemented for this manager
-func (m *Manager) SetLegacyManagement(isLegacy bool) error {
-	if m.nativeFirewall == nil {
-		return nil
-	}
-	return m.nativeFirewall.SetLegacyManagement(isLegacy)
+// SetLegacyManagement is a no-op for the userspace firewall: it only matters
+// when an old management server can't send route firewall rules, which the
+// userspace router doesn't rely on.
+func (m *Manager) SetLegacyManagement(bool) error {
+	return nil
 }
 
 // Flush doesn't need to be implemented for this manager
@@ -793,21 +865,15 @@ func (m *Manager) resetState() {
 	}
 }
 
-// SetupEBPFProxyNoTrack creates notrack rules for eBPF proxy loopback traffic.
-func (m *Manager) SetupEBPFProxyNoTrack(proxyPort, wgPort uint16) error {
-	if m.nativeFirewall == nil {
-		return nil
-	}
-	return m.nativeFirewall.SetupEBPFProxyNoTrack(proxyPort, wgPort)
+// SetupEBPFProxyNoTrack is not supported by the userspace firewall: eBPF isn't
+// used in userspace mode, so this should never be called.
+func (m *Manager) SetupEBPFProxyNoTrack(uint16, uint16) error {
+	return errNotSupported
 }
 
 // UpdateSet updates the rule destinations associated with the given set
 // by merging the existing prefixes with the new ones, then deduplicating.
 func (m *Manager) UpdateSet(set firewall.Set, prefixes []netip.Prefix) error {
-	if m.nativeRouter.Load() && m.nativeFirewall != nil {
-		return m.nativeFirewall.UpdateSet(set, prefixes)
-	}
-
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
@@ -1591,10 +1657,13 @@ func (m *Manager) EnableRouting() error {
 	}
 
 	rules, err := m.blockInvalidRouted(m.wgIface)
-	// Persist whatever was installed even on partial failure, so DisableRouting
-	// can clean it up later.
 	m.blockRules = rules
 	if err != nil {
+		// Roll back so forwarding can't stay active without the full set of
+		// block rules.
+		if derr := m.disableRouting(); derr != nil {
+			log.Warnf("roll back routing after block rule failure: %v", derr)
+		}
 		return fmt.Errorf("block invalid routed: %w", err)
 	}
 
@@ -1605,6 +1674,10 @@ func (m *Manager) DisableRouting() error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
+	return m.disableRouting()
+}
+
+func (m *Manager) disableRouting() error {
 	fwder := m.forwarder.Load()
 	if fwder == nil {
 		return nil

client/firewall/uspfilter/filter_bench_test.go

@@ -164,9 +164,12 @@ func BenchmarkCoreFiltering(b *testing.B) {
 				}
 
 				// Create manager and basic setup
-				manager, _ := Create(&IFaceMock{
-					SetFilterFunc: func(device.PacketFilter) error { return nil },
-				}, nil, false, flowLogger, iface.DefaultMTU)
+				manager, err := Create(Config{
+					IFace: &IFaceMock{
+						SetFilterFunc: func(device.PacketFilter) error { return nil },
+					},
+					FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+				require.NoError(b, err)
 				defer b.Cleanup(func() {
 					require.NoError(b, manager.Close(nil))
 				})
@@ -204,9 +207,12 @@ func BenchmarkStateScaling(b *testing.B) {
 
 	for _, count := range connCounts {
 		b.Run(fmt.Sprintf("conns_%d", count), func(b *testing.B) {
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -247,9 +253,12 @@ func BenchmarkEstablishmentOverhead(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -405,9 +414,12 @@ func BenchmarkRoutedNetworkReturn(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -532,9 +544,12 @@ func BenchmarkLongLivedConnections(b *testing.B) {
 				require.NoError(b, os.Unsetenv("NB_DISABLE_CONNTRACK"))
 			}
 
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			defer b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -615,9 +630,12 @@ func BenchmarkShortLivedConnections(b *testing.B) {
 				require.NoError(b, os.Unsetenv("NB_DISABLE_CONNTRACK"))
 			}
 
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			defer b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -726,9 +744,12 @@ func BenchmarkParallelLongLivedConnections(b *testing.B) {
 				require.NoError(b, os.Unsetenv("NB_DISABLE_CONNTRACK"))
 			}
 
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			defer b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -806,9 +827,12 @@ func BenchmarkParallelShortLivedConnections(b *testing.B) {
 				require.NoError(b, os.Unsetenv("NB_DISABLE_CONNTRACK"))
 			}
 
-			manager, _ := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+			require.NoError(b, err)
 			defer b.Cleanup(func() {
 				require.NoError(b, manager.Close(nil))
 			})
@@ -1010,9 +1034,11 @@ func BenchmarkMSSClamping(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, err := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 			require.NoError(b, err)
 			defer func() {
 				require.NoError(b, manager.Close(nil))
@@ -1075,9 +1101,11 @@ func BenchmarkMSSClampingOverhead(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, err := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 			require.NoError(b, err)
 			defer func() {
 				require.NoError(b, manager.Close(nil))
@@ -1130,9 +1158,11 @@ func BenchmarkMSSClampingMemory(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, err := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 			require.NoError(b, err)
 			defer func() {
 				require.NoError(b, manager.Close(nil))

client/firewall/uspfilter/filter_filter_test.go

@@ -32,7 +32,7 @@ func TestPeerACLFiltering(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	require.NotNil(t, manager)
 
@@ -549,7 +549,7 @@ func TestPeerACLFilteringIPv6(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() { require.NoError(t, manager.Close(nil)) })
 
@@ -788,7 +788,7 @@ func setupRoutedManager(tb testing.TB, network string) *Manager {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(tb, err)
 	require.NoError(tb, manager.EnableRouting())
 	require.NotNil(tb, manager)
@@ -1634,7 +1634,7 @@ func TestRouteACLSet(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		require.NoError(t, manager.Close(nil))

client/firewall/uspfilter/filter_routeacl_test.go

@@ -187,7 +187,7 @@ func TestBlockInvalidRoutedDualStack(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		require.NoError(t, manager.Close(nil))
@@ -235,7 +235,7 @@ func TestBlockInvalidRoutedIdempotent(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		require.NoError(t, manager.Close(nil))
@@ -298,7 +298,7 @@ func TestBlockRuleNotAccumulatedOnRepeatedEnableRouting(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		require.NoError(t, manager.Close(nil))
@@ -417,7 +417,7 @@ func setupTestManager(t *testing.T) *Manager {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	require.NoError(t, manager.EnableRouting())
 

client/firewall/uspfilter/filter_test.go

@@ -78,11 +78,12 @@ func TestManagerCreate(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	if err != nil {
 		t.Errorf("failed to create Manager: %v", err)
 		return
 	}
+	t.Cleanup(func() { require.NoError(t, m.Close(nil)) })
 
 	if m == nil {
 		t.Error("Manager is nil")
@@ -98,11 +99,12 @@ func TestManagerAddFilterRule(t *testing.T) {
 		},
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	if err != nil {
 		t.Errorf("failed to create Manager: %v", err)
 		return
 	}
+	t.Cleanup(func() { require.NoError(t, m.Close(nil)) })
 
 	ip := net.ParseIP("192.168.1.1")
 	proto := fw.ProtocolTCP
@@ -131,11 +133,12 @@ func TestManagerDeleteRule(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	if err != nil {
 		t.Errorf("failed to create Manager: %v", err)
 		return
 	}
+	t.Cleanup(func() { require.NoError(t, m.Close(nil)) })
 
 	ip := netip.MustParseAddr("192.168.1.1")
 	proto := fw.ProtocolTCP
@@ -166,9 +169,11 @@ func TestManagerDeleteRule(t *testing.T) {
 }
 
 func TestSetUDPPacketHook(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() { require.NoError(t, manager.Close(nil)) })
 
@@ -190,9 +195,11 @@ func TestSetUDPPacketHook(t *testing.T) {
 }
 
 func TestSetTCPPacketHook(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() { require.NoError(t, manager.Close(nil)) })
 
@@ -220,7 +227,7 @@ func TestPeerRuleLifecycleDenyRules(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, m.Close(nil))
@@ -267,7 +274,7 @@ func TestPeerRuleAddAndDeleteDontLeak(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, m.Close(nil))
@@ -307,7 +314,7 @@ func TestMixedAllowDenyRulesSameIP(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, m.Close(nil))
@@ -360,7 +367,7 @@ func TestManagerReset(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	if err != nil {
 		t.Errorf("failed to create Manager: %v", err)
 		return
@@ -399,7 +406,7 @@ func TestNotMatchByIP(t *testing.T) {
 		},
 	}
 
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	if err != nil {
 		t.Errorf("failed to create Manager: %v", err)
 		return
@@ -462,7 +469,7 @@ func TestRemovePacketHook(t *testing.T) {
 	}
 
 	// creating manager instance
-	manager, err := Create(iface, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{IFace: iface, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	if err != nil {
 		t.Fatalf("Failed to create Manager: %s", err)
 	}
@@ -479,9 +486,11 @@ func TestRemovePacketHook(t *testing.T) {
 }
 
 func TestProcessOutgoingHooks(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 
 	manager.udpTracker.Close()
@@ -566,7 +575,7 @@ func TestUSPFilterCreatePerformance(t *testing.T) {
 			ifaceMock := &IFaceMock{
 				SetFilterFunc: func(device.PacketFilter) error { return nil },
 			}
-			manager, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+			manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 			require.NoError(t, err)
 			time.Sleep(time.Second)
 
@@ -591,9 +600,11 @@ func TestUSPFilterCreatePerformance(t *testing.T) {
 }
 
 func TestStatefulFirewall_UDPTracking(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 
 	manager.udpTracker.Close() // Close the existing tracker
@@ -805,7 +816,7 @@ func TestUpdateSetMerge(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		require.NoError(t, manager.Close(nil))
@@ -891,7 +902,7 @@ func TestUpdateSetDeduplication(t *testing.T) {
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		require.NoError(t, manager.Close(nil))
@@ -1011,7 +1022,7 @@ func TestMSSClamping(t *testing.T) {
 		},
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, 1280)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: 1280})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))
@@ -1203,7 +1214,7 @@ func TestShouldForward(t *testing.T) {
 		return wgaddr.Address{IP: wgIP, Network: netip.PrefixFrom(wgIP, 24)}
 	}
 
-	manager, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))
@@ -1318,7 +1329,7 @@ func TestShouldForward(t *testing.T) {
 
 	// Re-create manager to pick up the new address with IPv6
 	require.NoError(t, manager.Close(nil))
-	manager, err = Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	manager, err = Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err)
 
 	v6Cases := []struct {

client/firewall/uspfilter/forwarder/forwarder.go

@@ -10,6 +10,7 @@ import (
 	"time"
 
 	log "github.com/sirupsen/logrus"
+	wgdevice "golang.zx2c4.com/wireguard/device"
 	"gvisor.dev/gvisor/pkg/buffer"
 	"gvisor.dev/gvisor/pkg/tcpip"
 	"gvisor.dev/gvisor/pkg/tcpip/header"
@@ -20,9 +21,9 @@ import (
 	"gvisor.dev/gvisor/pkg/tcpip/transport/tcp"
 	"gvisor.dev/gvisor/pkg/tcpip/transport/udp"
 
-	"github.com/netbirdio/netbird/client/firewall/uspfilter/common"
 	"github.com/netbirdio/netbird/client/firewall/uspfilter/conntrack"
 	nblog "github.com/netbirdio/netbird/client/firewall/uspfilter/log"
+	"github.com/netbirdio/netbird/client/iface/wgaddr"
 	nftypes "github.com/netbirdio/netbird/client/internal/netflow/types"
 )
 
@@ -33,6 +34,12 @@ const (
 	iosMaxInFlight       = 256
 )
 
+// IFace provides the WireGuard device and overlay addresses the forwarder needs.
+type IFace interface {
+	GetWGDevice() *wgdevice.Device
+	Address() wgaddr.Address
+}
+
 type Forwarder struct {
 	logger     *nblog.Logger
 	flowLogger nftypes.FlowLogger
@@ -51,7 +58,7 @@ type Forwarder struct {
 	pingSemaphore      chan struct{}
 }
 
-func New(iface common.IFaceMapper, logger *nblog.Logger, flowLogger nftypes.FlowLogger, netstack bool, mtu uint16) (*Forwarder, error) {
+func New(iface IFace, logger *nblog.Logger, flowLogger nftypes.FlowLogger, netstack bool, mtu uint16) (*Forwarder, error) {
 	s := stack.New(stack.Options{
 		NetworkProtocols: []stack.NetworkProtocolFactory{
 			ipv4.NewProtocol,

client/firewall/uspfilter/interface_allower_windows.go

@@ -9,7 +9,6 @@ import (
 	log "github.com/sirupsen/logrus"
 
 	nberrors "github.com/netbirdio/netbird/client/errors"
-	"github.com/netbirdio/netbird/client/internal/statemanager"
 )
 
 type action string
@@ -20,35 +19,20 @@ const (
 	firewallRuleName        = "Netbird"
 )
 
-// Close cleans up the firewall manager by removing all rules and closing trackers
-func (m *Manager) Close(*statemanager.Manager) error {
-	m.mutex.Lock()
-	defer m.mutex.Unlock()
-
-	m.resetState()
-
-	if !isWindowsFirewallReachable() {
-		return nil
-	}
-
-	var merr *multierror.Error
-	if isFirewallRuleActive(firewallRuleName) {
-		if err := manageFirewallRule(firewallRuleName, deleteRule); err != nil {
-			merr = multierror.Append(merr, fmt.Errorf("remove windows firewall rule: %w", err))
-		}
-	}
-
-	if isFirewallRuleActive(firewallRuleName + "-v6") {
-		if err := manageFirewallRule(firewallRuleName+"-v6", deleteRule); err != nil {
-			merr = multierror.Append(merr, fmt.Errorf("remove windows v6 firewall rule: %w", err))
-		}
-	}
-
-	return nberrors.FormatErrorOrNil(merr)
+// WindowsInterfaceAllower opens the NetBird interface in the Windows firewall
+// via netsh advfirewall rules. It implements InterfaceAllower for the userspace
+// firewall on Windows.
+type WindowsInterfaceAllower struct {
+	iface Iface
 }
 
-// AllowNetbird allows netbird interface traffic
-func (m *Manager) AllowNetbird() error {
+// NewWindowsInterfaceAllower builds the Windows netsh-based interface allower.
+func NewWindowsInterfaceAllower(iface Iface) *WindowsInterfaceAllower {
+	return &WindowsInterfaceAllower{iface: iface}
+}
+
+// Apply adds inbound-allow netsh rules for the interface's addresses.
+func (a *WindowsInterfaceAllower) Apply() error {
 	if !isWindowsFirewallReachable() {
 		return nil
 	}
@@ -60,13 +44,13 @@ func (m *Manager) AllowNetbird() error {
 			"enable=yes",
 			"action=allow",
 			"profile=any",
-			"localip="+m.wgIface.Address().IP.String(),
+			"localip="+a.iface.Address().IP.String(),
 		); err != nil {
 			return err
 		}
 	}
 
-	if v6 := m.wgIface.Address().IPv6; v6.IsValid() && !isFirewallRuleActive(firewallRuleName+"-v6") {
+	if v6 := a.iface.Address().IPv6; v6.IsValid() && !isFirewallRuleActive(firewallRuleName+"-v6") {
 		if err := manageFirewallRule(firewallRuleName+"-v6",
 			addRule,
 			"dir=in",
@@ -82,8 +66,27 @@ func (m *Manager) AllowNetbird() error {
 	return nil
 }
 
-func manageFirewallRule(ruleName string, action action, extraArgs ...string) error {
+// Close removes the netsh rules added by Apply.
+func (a *WindowsInterfaceAllower) Close() error {
+	if !isWindowsFirewallReachable() {
+		return nil
+	}
 
+	var merr *multierror.Error
+	if isFirewallRuleActive(firewallRuleName) {
+		if err := manageFirewallRule(firewallRuleName, deleteRule); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("remove windows firewall rule: %w", err))
+		}
+	}
+	if isFirewallRuleActive(firewallRuleName + "-v6") {
+		if err := manageFirewallRule(firewallRuleName+"-v6", deleteRule); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("remove windows v6 firewall rule: %w", err))
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}
+
+func manageFirewallRule(ruleName string, action action, extraArgs ...string) error {
 	args := []string{"advfirewall", "firewall", string(action), "rule", "name=" + ruleName}
 	if action == addRule {
 		args = append(args, extraArgs...)

client/firewall/uspfilter/localip.go

@@ -7,8 +7,6 @@ import (
 	"sync/atomic"
 
 	log "github.com/sirupsen/logrus"
-
-	"github.com/netbirdio/netbird/client/firewall/uspfilter/common"
 )
 
 // localIPSnapshot is an immutable snapshot of local IP addresses, swapped
@@ -60,7 +58,7 @@ func processInterface(iface net.Interface, ips map[netip.Addr]struct{}, addresse
 }
 
 // UpdateLocalIPs rebuilds the local IP snapshot and swaps it in atomically.
-func (m *localIPManager) UpdateLocalIPs(iface common.IFaceMapper) (err error) {
+func (m *localIPManager) UpdateLocalIPs(iface Iface) (err error) {
 	defer func() {
 		if r := recover(); r != nil {
 			err = fmt.Errorf("panic: %v", r)

client/firewall/uspfilter/nat.go

@@ -487,19 +487,13 @@ func incrementalUpdate(oldChecksum uint16, oldBytes, newBytes []byte) uint16 {
 }
 
 // AddDNATRule adds outbound DNAT rule for forwarding external traffic to NetBird network.
-func (m *Manager) AddDNATRule(rule firewall.ForwardRule) (firewall.Rule, error) {
-	if m.nativeFirewall == nil {
-		return nil, errNatNotSupported
-	}
-	return m.nativeFirewall.AddDNATRule(rule)
+func (m *Manager) AddDNATRule(firewall.ForwardRule) (firewall.Rule, error) {
+	return nil, errNotSupported
 }
 
 // DeleteDNATRule deletes outbound DNAT rule.
-func (m *Manager) DeleteDNATRule(rule firewall.Rule) error {
-	if m.nativeFirewall == nil {
-		return errNatNotSupported
-	}
-	return m.nativeFirewall.DeleteDNATRule(rule)
+func (m *Manager) DeleteDNATRule(firewall.Rule) error {
+	return errNotSupported
 }
 
 // addPortRedirection adds a port redirection rule.
@@ -521,7 +515,6 @@ func (m *Manager) addPortRedirection(targetIP netip.Addr, protocol gopacket.Laye
 }
 
 // AddInboundDNAT adds an inbound DNAT rule redirecting traffic from NetBird peers to local services.
-// TODO: also delegate to nativeFirewall when available for kernel WG mode
 func (m *Manager) AddInboundDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
 	var layerType gopacket.LayerType
 	switch protocol {
@@ -567,20 +560,16 @@ func (m *Manager) RemoveInboundDNAT(localAddr netip.Addr, protocol firewall.Prot
 	return m.removePortRedirection(localAddr, layerType, originalPort, translatedPort)
 }
 
-// AddOutputDNAT delegates to the native firewall if available.
-func (m *Manager) AddOutputDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
-	if m.nativeFirewall == nil {
-		return fmt.Errorf("output DNAT not supported without native firewall")
-	}
-	return m.nativeFirewall.AddOutputDNAT(localAddr, protocol, originalPort, translatedPort)
+// AddOutputDNAT is not supported by the userspace firewall: it backs kernel DNS
+// redirection, but userspace DNS is served in-process on the gVisor netstack, so
+// this should never be called.
+func (m *Manager) AddOutputDNAT(netip.Addr, firewall.Protocol, uint16, uint16) error {
+	return errNotSupported
 }
 
-// RemoveOutputDNAT delegates to the native firewall if available.
-func (m *Manager) RemoveOutputDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
-	if m.nativeFirewall == nil {
-		return nil
-	}
-	return m.nativeFirewall.RemoveOutputDNAT(localAddr, protocol, originalPort, translatedPort)
+// RemoveOutputDNAT is a no-op for the userspace firewall (see AddOutputDNAT).
+func (m *Manager) RemoveOutputDNAT(netip.Addr, firewall.Protocol, uint16, uint16) error {
+	return nil
 }
 
 // translateInboundPortDNAT applies port-specific DNAT translation to inbound packets.

client/firewall/uspfilter/nat_bench_test.go

@@ -64,9 +64,11 @@ func BenchmarkDNATTranslation(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, err := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 			require.NoError(b, err)
 			defer func() {
 				require.NoError(b, manager.Close(nil))
@@ -124,9 +126,11 @@ func BenchmarkDNATTranslation(b *testing.B) {
 
 // BenchmarkDNATConcurrency tests DNAT performance under concurrent load
 func BenchmarkDNATConcurrency(b *testing.B) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(b, err)
 	defer func() {
 		require.NoError(b, manager.Close(nil))
@@ -196,9 +200,11 @@ func BenchmarkDNATScaling(b *testing.B) {
 
 	for _, count := range mappingCounts {
 		b.Run(fmt.Sprintf("mappings_%d", count), func(b *testing.B) {
-			manager, err := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 			require.NoError(b, err)
 			defer func() {
 				require.NoError(b, manager.Close(nil))
@@ -308,9 +314,11 @@ func BenchmarkChecksumUpdate(b *testing.B) {
 
 // BenchmarkDNATMemoryAllocations checks for memory allocations in DNAT operations
 func BenchmarkDNATMemoryAllocations(b *testing.B) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(b, err)
 	defer func() {
 		require.NoError(b, manager.Close(nil))
@@ -481,9 +489,11 @@ func BenchmarkPortDNAT(b *testing.B) {
 
 	for _, sc := range scenarios {
 		b.Run(sc.name, func(b *testing.B) {
-			manager, err := Create(&IFaceMock{
-				SetFilterFunc: func(device.PacketFilter) error { return nil },
-			}, nil, false, flowLogger, iface.DefaultMTU)
+			manager, err := Create(Config{
+				IFace: &IFaceMock{
+					SetFilterFunc: func(device.PacketFilter) error { return nil },
+				},
+				FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 			require.NoError(b, err)
 			defer func() {
 				require.NoError(b, manager.Close(nil))

client/firewall/uspfilter/nat_stateful_test.go

@@ -13,9 +13,11 @@ import (
 
 // TestPortDNATBasic tests basic port DNAT functionality
 func TestPortDNATBasic(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))
@@ -49,9 +51,11 @@ func TestPortDNATBasic(t *testing.T) {
 
 // TestPortDNATMultipleRules tests multiple port DNAT rules
 func TestPortDNATMultipleRules(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))

client/firewall/uspfilter/nat_test.go

@@ -15,9 +15,11 @@ import (
 
 // TestDNATTranslationCorrectness verifies DNAT translation works correctly
 func TestDNATTranslationCorrectness(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))
@@ -104,9 +106,11 @@ func parsePacket(t testing.TB, packetData []byte) *decoder {
 
 // TestDNATMappingManagement tests adding/removing DNAT mappings
 func TestDNATMappingManagement(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))
@@ -152,9 +156,11 @@ func TestDNATMappingManagement(t *testing.T) {
 }
 
 func TestInboundPortDNAT(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))
@@ -202,9 +208,11 @@ func TestInboundPortDNAT(t *testing.T) {
 }
 
 func TestInboundPortDNATNegative(t *testing.T) {
-	manager, err := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	require.NoError(t, err)
 	defer func() {
 		require.NoError(t, manager.Close(nil))

client/firewall/uspfilter/peer_acl_bench_test.go

@@ -88,10 +88,13 @@ func runPeerACLBench(b *testing.B, m, k int, hit, v6 bool) {
 			IPv6Net: netip.PrefixFrom(dstIP, bits),
 		}
 	}
-	manager, _ := Create(&IFaceMock{
-		SetFilterFunc: func(device.PacketFilter) error { return nil },
-		AddressFunc:   func() wgaddr.Address { return mockAddr },
-	}, nil, false, flowLogger, iface.DefaultMTU)
+	manager, err := Create(Config{
+		IFace: &IFaceMock{
+			SetFilterFunc: func(device.PacketFilter) error { return nil },
+			AddressFunc:   func() wgaddr.Address { return mockAddr },
+		},
+		FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+	require.NoError(b, err)
 	b.Cleanup(func() { require.NoError(b, manager.Close(nil)) })
 
 	// Generate M policies × K source peers, all distinct.
@@ -169,9 +172,12 @@ func BenchmarkPeerACLIndexMemory(b *testing.B) {
 		b.Run(c.name, func(b *testing.B) {
 			b.ReportAllocs()
 			for i := 0; i < b.N; i++ {
-				mgr, _ := Create(&IFaceMock{
-					SetFilterFunc: func(device.PacketFilter) error { return nil },
-				}, nil, false, flowLogger, iface.DefaultMTU)
+				mgr, err := Create(Config{
+					IFace: &IFaceMock{
+						SetFilterFunc: func(device.PacketFilter) error { return nil },
+					},
+					FlowLogger: flowLogger, MTU: iface.DefaultMTU})
+				require.NoError(b, err)
 
 				populateIndexedRules(b, mgr, c.M, c.K, c.overlapFrac)
 

client/firewall/uspfilter/peer_acl_dedup_test.go

@@ -17,8 +17,9 @@ func newTestManager(t *testing.T) *Manager {
 	ifaceMock := &IFaceMock{
 		SetFilterFunc: func(device.PacketFilter) error { return nil },
 	}
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err, "create manager")
+	t.Cleanup(func() { require.NoError(t, m.Close(nil)) })
 	return m
 }
 

client/firewall/uspfilter/peer_acl_ipv6_test.go

@@ -29,7 +29,7 @@ func newV6TestManager(t *testing.T, localV6 string) *Manager {
 			}
 		},
 	}
-	m, err := Create(ifaceMock, nil, false, flowLogger, nbiface.DefaultMTU)
+	m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: nbiface.DefaultMTU})
 	require.NoError(t, err, "create manager")
 	t.Cleanup(func() { require.NoError(t, m.Close(nil)) })
 	return m

client/firewall/uspfilter/tracer_test.go

@@ -45,7 +45,7 @@ func TestTracePacket(t *testing.T) {
 			},
 		}
 
-		m, err := Create(ifaceMock, nil, false, flowLogger, iface.DefaultMTU)
+		m, err := Create(Config{IFace: ifaceMock, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 		require.NoError(t, err)
 
 		if !statefulMode {

client/internal/dns/server_test.go

@@ -900,7 +900,7 @@ func createWgInterfaceWithBind(t *testing.T) (*iface.WGIface, error) {
 		return nil, err
 	}
 
-	pf, err := uspfilter.Create(wgIface, nil, false, flowLogger, iface.DefaultMTU)
+	pf, err := uspfilter.Create(uspfilter.Config{IFace: wgIface, FlowLogger: flowLogger, MTU: iface.DefaultMTU})
 	if err != nil {
 		t.Fatalf("failed to create uspfilter: %v", err)
 		return nil, err