[management, client] Add IPv6 overlay support (#5631)

client/firewall/iptables/manager_linux.go

@@ -18,6 +18,10 @@ import (
 	"github.com/netbirdio/netbird/client/internal/statemanager"
 )
 
+type resetter interface {
+	Reset() error
+}
+
 // Manager of iptables firewall
 type Manager struct {
 	mutex sync.Mutex
@@ -28,6 +32,11 @@ type Manager struct {
 	aclMgr       *aclManager
 	router       *router
 	rawSupported bool
+
+	// IPv6 counterparts, nil when no v6 overlay
+	ipv6Client *iptables.IPTables
+	aclMgr6    *aclManager
+	router6    *router
 }
 
 // iFaceMapper defines subset methods of interface required for manager
@@ -58,9 +67,43 @@ func Create(wgIface iFaceMapper, mtu uint16) (*Manager, error) {
 		return nil, fmt.Errorf("create acl manager: %w", err)
 	}
 
+	if wgIface.Address().HasIPv6() {
+		if err := m.createIPv6Components(wgIface, mtu); err != nil {
+			return nil, fmt.Errorf("create IPv6 firewall: %w", err)
+		}
+	}
+
 	return m, nil
 }
 
+func (m *Manager) createIPv6Components(wgIface iFaceMapper, mtu uint16) error {
+	ip6Client, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
+	if err != nil {
+		return fmt.Errorf("init ip6tables: %w", err)
+	}
+	m.ipv6Client = ip6Client
+
+	m.router6, err = newRouter(ip6Client, wgIface, mtu)
+	if err != nil {
+		return fmt.Errorf("create v6 router: %w", err)
+	}
+
+	// Share the same IP forwarding state with the v4 router, since
+	// EnableIPForwarding controls both v4 and v6 sysctls.
+	m.router6.ipFwdState = m.router.ipFwdState
+
+	m.aclMgr6, err = newAclManager(ip6Client, wgIface)
+	if err != nil {
+		return fmt.Errorf("create v6 acl manager: %w", err)
+	}
+
+	return nil
+}
+
+func (m *Manager) hasIPv6() bool {
+	return m.ipv6Client != nil
+}
+
 func (m *Manager) Init(stateManager *statemanager.Manager) error {
 	state := &ShutdownState{
 		InterfaceState: &InterfaceState{
@@ -74,13 +117,8 @@ func (m *Manager) Init(stateManager *statemanager.Manager) error {
 		log.Errorf("failed to update state: %v", err)
 	}
 
-	if err := m.router.init(stateManager); err != nil {
-		return fmt.Errorf("router init: %w", err)
-	}
-
-	if err := m.aclMgr.init(stateManager); err != nil {
-		// TODO: cleanup router
-		return fmt.Errorf("acl manager init: %w", err)
+	if err := m.initChains(stateManager); err != nil {
+		return err
 	}
 
 	if err := m.initNoTrackChain(); err != nil {
@@ -103,6 +141,41 @@ func (m *Manager) Init(stateManager *statemanager.Manager) error {
 	return nil
 }
 
+// initChains initializes router and ACL chains for both address families,
+// rolling back on failure.
+func (m *Manager) initChains(stateManager *statemanager.Manager) error {
+	type initStep struct {
+		name string
+		init func(*statemanager.Manager) error
+		mgr  resetter
+	}
+
+	steps := []initStep{
+		{"router", m.router.init, m.router},
+		{"acl manager", m.aclMgr.init, m.aclMgr},
+	}
+	if m.hasIPv6() {
+		steps = append(steps,
+			initStep{"v6 router", m.router6.init, m.router6},
+			initStep{"v6 acl manager", m.aclMgr6.init, m.aclMgr6},
+		)
+	}
+
+	var initialized []initStep
+	for _, s := range steps {
+		if err := s.init(stateManager); err != nil {
+			for i := len(initialized) - 1; i >= 0; i-- {
+				if rerr := initialized[i].mgr.Reset(); rerr != nil {
+					log.Warnf("rollback %s: %v", initialized[i].name, rerr)
+				}
+			}
+			return fmt.Errorf("%s init: %w", s.name, err)
+		}
+		initialized = append(initialized, s)
+	}
+	return nil
+}
+
 // AddPeerFiltering adds a rule to the firewall
 //
 // Comment will be ignored because some system this feature is not supported
@@ -118,7 +191,13 @@ func (m *Manager) AddPeerFiltering(
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.aclMgr.AddPeerFiltering(id, ip, proto, sPort, dPort, action, ipsetName)
+	if ip.To4() != nil {
+		return m.aclMgr.AddPeerFiltering(id, ip, proto, sPort, dPort, action, ipsetName)
+	}
+	if !m.hasIPv6() {
+		return nil, fmt.Errorf("add peer filtering for %s: %w", ip, firewall.ErrIPv6NotInitialized)
+	}
+	return m.aclMgr6.AddPeerFiltering(id, ip, proto, sPort, dPort, action, ipsetName)
 }
 
 func (m *Manager) AddRouteFiltering(
@@ -132,25 +211,48 @@ func (m *Manager) AddRouteFiltering(
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	if destination.IsPrefix() && !destination.Prefix.Addr().Is4() {
-		return nil, fmt.Errorf("unsupported IP version: %s", destination.Prefix.Addr().String())
+	if isIPv6RouteRule(sources, destination) {
+		if !m.hasIPv6() {
+			return nil, fmt.Errorf("add route filtering: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.AddRouteFiltering(id, sources, destination, proto, sPort, dPort, action)
 	}
 
 	return m.router.AddRouteFiltering(id, sources, destination, proto, sPort, dPort, action)
 }
 
+func isIPv6RouteRule(sources []netip.Prefix, destination firewall.Network) bool {
+	if destination.IsPrefix() {
+		return destination.Prefix.Addr().Is6()
+	}
+	return len(sources) > 0 && sources[0].Addr().Is6()
+}
+
 // DeletePeerRule from the firewall by rule definition
 func (m *Manager) DeletePeerRule(rule firewall.Rule) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
+	if m.hasIPv6() && isIPv6IptRule(rule) {
+		return m.aclMgr6.DeletePeerRule(rule)
+	}
 	return m.aclMgr.DeletePeerRule(rule)
 }
 
+func isIPv6IptRule(rule firewall.Rule) bool {
+	r, ok := rule.(*Rule)
+	return ok && r.v6
+}
+
+// DeleteRouteRule deletes a routing rule.
+// Route rules are keyed by content hash. Check v4 first, try v6 if not found.
 func (m *Manager) DeleteRouteRule(rule firewall.Rule) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
+	if m.hasIPv6() && !m.router.hasRule(rule.ID()) {
+		return m.router6.DeleteRouteRule(rule)
+	}
 	return m.router.DeleteRouteRule(rule)
 }
 
@@ -166,18 +268,65 @@ func (m *Manager) AddNatRule(pair firewall.RouterPair) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.AddNatRule(pair)
+	if pair.Destination.IsPrefix() && pair.Destination.Prefix.Addr().Is6() {
+		if !m.hasIPv6() {
+			return fmt.Errorf("add NAT rule: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.AddNatRule(pair)
+	}
+
+	if err := m.router.AddNatRule(pair); err != nil {
+		return err
+	}
+
+	// Dynamic routes need NAT in both tables since resolved IPs can be
+	// either v4 or v6. This covers both DomainSet (modern) and the legacy
+	// wildcard 0.0.0.0/0 destination where the client resolves DNS.
+	if m.hasIPv6() && pair.Dynamic {
+		v6Pair := firewall.ToV6NatPair(pair)
+		if err := m.router6.AddNatRule(v6Pair); err != nil {
+			return fmt.Errorf("add v6 NAT rule: %w", err)
+		}
+	}
+
+	return nil
 }
 
 func (m *Manager) RemoveNatRule(pair firewall.RouterPair) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.RemoveNatRule(pair)
+	if pair.Destination.IsPrefix() && pair.Destination.Prefix.Addr().Is6() {
+		if !m.hasIPv6() {
+			return nil
+		}
+		return m.router6.RemoveNatRule(pair)
+	}
+
+	var merr *multierror.Error
+
+	if err := m.router.RemoveNatRule(pair); err != nil {
+		merr = multierror.Append(merr, fmt.Errorf("remove v4 NAT rule: %w", err))
+	}
+
+	if m.hasIPv6() && pair.Dynamic {
+		v6Pair := firewall.ToV6NatPair(pair)
+		if err := m.router6.RemoveNatRule(v6Pair); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("remove v6 NAT rule: %w", err))
+		}
+	}
+
+	return nberrors.FormatErrorOrNil(merr)
 }
 
 func (m *Manager) SetLegacyManagement(isLegacy bool) error {
-	return firewall.SetLegacyManagement(m.router, isLegacy)
+	if err := firewall.SetLegacyManagement(m.router, isLegacy); err != nil {
+		return err
+	}
+	if m.hasIPv6() {
+		return firewall.SetLegacyManagement(m.router6, isLegacy)
+	}
+	return nil
 }
 
 // Reset firewall to the default state
@@ -191,6 +340,15 @@ func (m *Manager) Close(stateManager *statemanager.Manager) error {
 		merr = multierror.Append(merr, fmt.Errorf("cleanup notrack chain: %w", err))
 	}
 
+	if m.hasIPv6() {
+		if err := m.aclMgr6.Reset(); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("reset v6 acl manager: %w", err))
+		}
+		if err := m.router6.Reset(); err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("reset v6 router: %w", err))
+		}
+	}
+
 	if err := m.aclMgr.Reset(); err != nil {
 		merr = multierror.Append(merr, fmt.Errorf("reset acl manager: %w", err))
 	}
@@ -218,24 +376,21 @@ func (m *Manager) Close(stateManager *statemanager.Manager) error {
 // 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 {
-	_, err := m.AddPeerFiltering(
-		nil,
-		net.IP{0, 0, 0, 0},
-		firewall.ProtocolALL,
-		nil,
-		nil,
-		firewall.ActionAccept,
-		"",
-	)
-	if err != nil {
-		return fmt.Errorf("allow netbird interface traffic: %w", err)
+	var merr *multierror.Error
+	if _, err := m.AddPeerFiltering(nil, net.IP{0, 0, 0, 0}, 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.AddPeerFiltering(nil, net.IPv6zero, 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 nil
+	return nberrors.FormatErrorOrNil(merr)
 }
 
 // Flush doesn't need to be implemented for this manager
@@ -265,6 +420,12 @@ func (m *Manager) AddDNATRule(rule firewall.ForwardRule) (firewall.Rule, error)
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
+	if rule.TranslatedAddress.Is6() {
+		if !m.hasIPv6() {
+			return nil, fmt.Errorf("add DNAT rule: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.AddDNATRule(rule)
+	}
 	return m.router.AddDNATRule(rule)
 }
 
@@ -273,6 +434,9 @@ func (m *Manager) DeleteDNATRule(rule firewall.Rule) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
+	if m.hasIPv6() && !m.router.hasRule(rule.ID()+dnatSuffix) {
+		return m.router6.DeleteDNATRule(rule)
+	}
 	return m.router.DeleteDNATRule(rule)
 }
 
@@ -281,39 +445,82 @@ func (m *Manager) UpdateSet(set firewall.Set, prefixes []netip.Prefix) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.UpdateSet(set, prefixes)
+	var v4Prefixes, v6Prefixes []netip.Prefix
+	for _, p := range prefixes {
+		if p.Addr().Is6() {
+			v6Prefixes = append(v6Prefixes, p)
+		} else {
+			v4Prefixes = append(v4Prefixes, p)
+		}
+	}
+
+	if err := m.router.UpdateSet(set, v4Prefixes); err != nil {
+		return err
+	}
+
+	if m.hasIPv6() && len(v6Prefixes) > 0 {
+		if err := m.router6.UpdateSet(set, v6Prefixes); err != nil {
+			return fmt.Errorf("update v6 set: %w", err)
+		}
+	}
+
+	return nil
 }
 
 // AddInboundDNAT adds an inbound DNAT rule redirecting traffic from NetBird peers to local services.
-func (m *Manager) AddInboundDNAT(localAddr netip.Addr, protocol firewall.Protocol, sourcePort, targetPort uint16) error {
+func (m *Manager) AddInboundDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.AddInboundDNAT(localAddr, protocol, sourcePort, targetPort)
+	if localAddr.Is6() {
+		if !m.hasIPv6() {
+			return fmt.Errorf("add inbound DNAT: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.AddInboundDNAT(localAddr, protocol, originalPort, translatedPort)
+	}
+	return m.router.AddInboundDNAT(localAddr, protocol, originalPort, translatedPort)
 }
 
 // RemoveInboundDNAT removes an inbound DNAT rule.
-func (m *Manager) RemoveInboundDNAT(localAddr netip.Addr, protocol firewall.Protocol, sourcePort, targetPort uint16) error {
+func (m *Manager) RemoveInboundDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.RemoveInboundDNAT(localAddr, protocol, sourcePort, targetPort)
+	if localAddr.Is6() {
+		if !m.hasIPv6() {
+			return fmt.Errorf("remove inbound DNAT: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.RemoveInboundDNAT(localAddr, protocol, originalPort, translatedPort)
+	}
+	return m.router.RemoveInboundDNAT(localAddr, protocol, originalPort, translatedPort)
 }
 
 // AddOutputDNAT adds an OUTPUT chain DNAT rule for locally-generated traffic.
-func (m *Manager) AddOutputDNAT(localAddr netip.Addr, protocol firewall.Protocol, sourcePort, targetPort uint16) error {
+func (m *Manager) AddOutputDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.AddOutputDNAT(localAddr, protocol, sourcePort, targetPort)
+	if localAddr.Is6() {
+		if !m.hasIPv6() {
+			return fmt.Errorf("add output DNAT: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.AddOutputDNAT(localAddr, protocol, originalPort, translatedPort)
+	}
+	return m.router.AddOutputDNAT(localAddr, protocol, originalPort, translatedPort)
 }
 
 // RemoveOutputDNAT removes an OUTPUT chain DNAT rule.
-func (m *Manager) RemoveOutputDNAT(localAddr netip.Addr, protocol firewall.Protocol, sourcePort, targetPort uint16) error {
+func (m *Manager) RemoveOutputDNAT(localAddr netip.Addr, protocol firewall.Protocol, originalPort, translatedPort uint16) error {
 	m.mutex.Lock()
 	defer m.mutex.Unlock()
 
-	return m.router.RemoveOutputDNAT(localAddr, protocol, sourcePort, targetPort)
+	if localAddr.Is6() {
+		if !m.hasIPv6() {
+			return fmt.Errorf("remove output DNAT: %w", firewall.ErrIPv6NotInitialized)
+		}
+		return m.router6.RemoveOutputDNAT(localAddr, protocol, originalPort, translatedPort)
+	}
+	return m.router.RemoveOutputDNAT(localAddr, protocol, originalPort, translatedPort)
 }
 
 const (