The L2TPv3 session ID identifies the session context on the decapsulating system. For dynamic sessions, the value of the session ID is selected to optimize the context identification efficiency of the decapsulating system. A decapsulation implementation may, therefore, elect to support a smaller session ID bit field. In this L2TPv3 implementation, an upper value for the L2TPv3 session ID was set at 023. The L2TPv3 session ID value 0 is reserved for use by the protocol. For static sessions, the session ID is manually configured.

Note	The local session ID must be unique on the decapsulating system and is restricted to the least significant ten bits.

The L2TPv3 header contains a control channel cookie field. The control channel cookie field has a variable length of 0, 4, or 8 bytes according to the cookie length supported by a given platform for packet decapsulation. The control channel cookie length can be configured manually for static sessions or determined dynamically for dynamic sessions.

The variable cookie length does not present a problem when the same platform is at both ends of an L2TPv3 control channel. However, when different platforms interoperate across an L2TPv3 control channel, both platforms need to encapsulate packets with a 4-byte cookie length.

The L2TPv3 pseudowire control encapsulation consists of 32 bits (4 bytes) and contains information used to sequence L2TP packets (see the Sequencing section). For the purposes of sequencing, only the first bit and bits 8 to 31 are relevant. Bit 1 indicates whether the Sequence Number field, bits 8 to 31, contains a valid sequence number and is to be updated.

Typically, the L2TP control plane is responsible for negotiating session parameters, such as the session ID or the cookie, to set up the session. However, some IP networks require sessions to be configured so that no signaling is required for session establishment. You can set up static L2TPv3 sessions for a PE device by configuring fixed values for the fields in the L2TP data header. A static L2TPv3 session allows the PE device to tunnel Layer 2 traffic as soon as the attachment circuit to which the session is bound comes up.

Static configuration allows sessions to be established without dynamically negotiating control connection parameters. This means that although sessions are displayed in the show l2tun session command output, no control channel information is displayed in the show l2tun tunnel command output.

Note	In an L2TPv3 static session, you can still run the L2TP control channel to perform peer authentication and dead-peer detection. If the L2TP control channel cannot be established or is torn down because of a hello failure, the static session is also torn down.

If you use a static L2TPv3 session, you cannot perform circuit interworking, such as LMI, because there is no facility to exchange control messages. To perform circuit interworking, you must use a dynamic session.

A dynamic L2TP session is established through the exchange of control messages containing attribute-value (AV) pairs. Each AV pair contains information about the nature of the Layer 2 link being forwarded, including the payload type and virtual circuit (VC) ID.

Multiple L2TP sessions, one for each forwarded Layer 2 circuit, can exist between a pair of PE devices and can be maintained by a single control channel. Session IDs and cookies are dynamically generated and exchanged as part of a dynamic session setup. Information such as sequencing configuration is also exchanged. Circuit state changes (UP/DOWN) are conveyed using the set link info (SLI) message.

The L2TP class configuration procedure creates a template of L2TP control channel parameters that can be inherited by different pseudowire classes. L2TP control channel parameters are used in control channel authentication, keepalive messages, and control channel negotiation. In an L2TPv3 session, the same L2TP class must be specified in the pseudowire configured on the PE device at each end of the control channel. Configuring L2TP control channel parameters is optional. However, the L2TP class must be configured before it is associated with a pseudowire class (see the Configuring the L2TPv3 Pseudowire task).

Two methods of control channel message authentication are available: the L2TPv3 Control Message Hashing feature and CHAP-style L2TP control channel. The L2TPv3 Control Message Hashing feature introduces a more robust authentication method than the older, CHAP-style L2TP control channel method of authentication. You may choose to enable both the methods of authentication to ensure interoperability with peers that support only one of these methods of authentication, but this configuration will yield control of the authentication method used on the peer PE device. Enabling both the methods of authentication should be considered as an interim solution to solve backward compatibility issues during software upgrades.

The principal difference between the two methods of authentication lies in the L2TPv3 Control Message Hashing feature using the entire message in the hash instead of computing the hash over selected contents of a received control message. In addition, instead of including the hash digest in only the start control channel replay (SCCRP) and start control channel connected (SCCCN) messages, it includes it in all messages.

Support for L2TP control channel authentication is maintained for backward compatibility. Either or both authentication methods can be enabled to allow interoperability with peers supporting only one of the authentication methods.

The table below shows a compatibility matrix for the different L2TPv3 authentication methods. PE1 is running the new authentication method. The possible authentication configurations for PE1 are shown in the first column. The other columns represent PE2 running software with different available authentication options. The tables cells in these columns indicate compatible configuration options for PE2. If any PE1/PE2 authentication configuration poses ambiguity about the authentication method used, the winning authentication method is indicated in bold. If both the old and new authentication methods are enabled on PE1 and PE2, both types of authentication occur.

The Ethernet over L2TPv3 feature provides support for Ethernet-based Layer 2 payload tunneling over IP core networks using L2TPv3.

The Ethernet over L2TPv3 feature supports the following like-to-like switching modes:

• Ethernet port mode

• Ethernet VLAN mode

• Ethernet VLAN mode with VLAN rewrite

• Ethernet QinQ and QinAny mode

Note	The QinQ over L2TPv3 support feature includes QinAny over L2TPv3, which has a fixed outer VLAN tag and a variable inner VLAN tag.

The Ethernet over L2TPv3 feature supports the following types of internetworking:

• Ethernet port to VLAN (routed)

• Ethernet port to VLAN (bridged)

• QinQ to Ethernet VLAN or Port Interworking (routed)

• QinQ to Ethernet VLAN or Port Interworking (bridged)

Note	QinAny Interworking is not a valid configuration because the inner VLAN tag is undetermined.

Gigabit EtherChannel (GEC) over Layer 2 Tunneling Protocol Version 3 (L2TPv3) provides support for GEC-based Layer 2 payload tunneling over IP core networks using L2TPv3. GEC also known as port channel is integrated with Ethernet and dot1q attachment circuits (ACs).

A port channel bundles physical links into a channel group to create a single logical link that provides the aggregate bandwidth of up to eight physical links. If a member port within a port channel fails, the traffic previously carried over the failed link switches to the remaining member ports within the port channel.

Interworking switching is supported in the following scenarios:

• The customer-edge-provider-edge (CE-PE) connecting interface on the local PE is a port-channel interface without dot1q encapsulation. The CE-PE connecting interface on the remote PE is a port-channel interface with dot1q encapsulation.
• The CE-PE connecting interface on the local PE is a port-channel interface with or without dot1q encapsulation. The CE-PE connecting interface on the remote PE is an Ethernet interface with or without dot1q encapsulation.

The figure below illustrates a port channel over IP core networks using L2TPv3. CE1 and CE2 are connected to L2TP Control Connection Endpoints (LCCE) and through port channels. The LCCE is connected to the IP core network using L2TPv3.

Figure 3. GEC over L2TPv3

Although the correct sequence of received Layer 2 frames is guaranteed by some Layer 2 technologies (by the nature of the link such as a serial line) or by the protocol itself, forwarded Layer 2 frames may be lost, duplicated, or reordered when they traverse a network as IP packets. If the Layer 2 protocol does not provide an explicit sequencing mechanism, you can configure L2TP to sequence its data packets according to the data channel sequencing mechanism described in the L2TPv3 IETF l2tpext working group draft.

A receiver of L2TP data packets mandates sequencing through the Sequencing Required AV pair when the session is being negotiated. A sender (or one that is manually configured to send sequenced packets) that receives this AV pair uses the Layer 2-specific pseudowire control encapsulation defined in L2TPv3.

You can configure L2TP to drop only out-of-order packets; you cannot configure L2TP to deliver the packets out-of-order. No reordering mechanism is available.

Interworking is not allowed when sequencing is enabled.

When Layer 2 traffic is tunneled across an IP network, information contained in the Type of Service (ToS) bits may be transferred to the L2TP-encapsulated IP packets in one of the following ways:

• If the tunneled Layer 2 frames themselves encapsulate IP packets, it may be desirable to simply copy the ToS bytes of the inner IP packets to the outer IP packet headers. This action is known as "ToS byte reflection."

• You can specify the ToS byte value used by all packets sent across the pseudowire. This is known as "Static ToS byte configuration".

For more details on how to configure ToS, see the Example Configuring a Negotiated L2TPv3 Session for Local HDLC Switching section.

The keepalive mechanism for L2TPv3 extends only to the endpoints of the tunneling protocol. L2TP has a reliable control message delivery mechanism that serves as the basis for the keepalive mechanism. The keepalive mechanism consists of an exchange of L2TP hello messages.

If a keepalive mechanism is required, the control plane is used, although it may not be used to bring up sessions. You can configure sessions manually.

In the case of static L2TPv3 sessions, a control channel between the two L2TP peers is negotiated through the exchange of start control channel request (SCCRQ), SCCRP, and SCCCN control messages. The control channel is responsible for maintaining only the keepalive mechanism through the exchange of hello messages.

The interval between hello messages is configurable per control channel. If one peer detects that the other peer has gone down through the keepalive mechanism, it sends a StopCCN control message and then notifies all the pseudowires to the peer about the event. This notification results in the teardown of both manually configured and dynamic sessions.

It is important that you configure a Maximum Transmission Unit (MTU) appropriate for each L2TPv3 tunneled link. The configured MTU size ensures the following:

• The lengths of the tunneled Layer 2 frames fall below the MTU of the destination attachment circuit.

• The tunneled packets are not fragmented, which forces the receiving PE to reassemble them.

L2TPv3 handles the MTU as follows:

• The default behavior is to fragment packets that are larger than the session MTU.

• If you enable the ip dfbit set command in the pseudowire class, the default MTU behavior changes so that any packets that cannot fit within the tunnel MTU are dropped.

• If you enable the ip pmtu command in the pseudowire class, the L2TPv3 control channel participates in the path MTU (PMTU) discovery.

If you enable this feature, the following processing is performed:

• Internet Control Message Protocol (ICMP) unreachable messages sent back to the L2TPv3 device are deciphered and the tunnel MTU is updated accordingly. To receive ICMP unreachable messages for fragmentation errors, the Don’t Fragment (DF) bit in the tunnel header is either set according to the DF bit value received from the CE device or set statically if the ip dfbit set option is enabled. The tunnel MTU is periodically reset to the default value based on a periodic timer.

• ICMP unreachable messages are sent back to the clients on the CE side. ICMP unreachable messages are sent to the CE whenever IP packets arrive on the CE-PE interface and have a packet size greater than the tunnel MTU. A Layer 2 header calculation is performed before the ICMP unreachable message is sent to the CE.

The L2TPv3 Control Message Hashing feature introduces a new and more secure authentication system that replaces the CHAP-like authentication system inherited from L2TPv2, which uses the Challenge and Challenge Response AV pairs in the SCCRQ, SCCRP, and SCCCN messages. The L2TPv3 Control Message Hashing feature incorporates an optional authentication or integrity check for all control messages.

The per-message authentication introduced by the L2TPv3 Control Message Hashing feature is designed to:

• Perform a mutual authentication between L2TP nodes.

• Check integrity of all control messages.

• Guard against control message spoofing and replay attacks that would otherwise be trivial to mount against the network.

The new authentication method uses the following:

• A computed, one-way hash over the header and body of the L2TP control message

• A preconfigured, shared secret that must be defined on the communicating L2TP nodes

• A local and remote random value exchanged using the Nonce AV pairs

Received control messages that lack any of the required security elements are dropped.

L2TPv3 control message integrity checking is a unidirectional mechanism that does not require the configuration of a shared secret. If integrity checking is enabled on the local PE device, control messages are sent with the message digest calculated without the shared secret or Nonce AV pairs and are verified by the remote PE device. If verification fails, the remote PE device drops the control message.

Enabling the L2TPv3 Control Message Hashing feature will impact performance during control channel and session establishment because additional digest calculation of the full message content is required for each sent and received control message. This is an expected trade-off for the additional security provided by this feature. In addition, network congestion may occur if the receive window size is too small. If the L2TPv3 Control Message Hashing feature is enabled, message digest validation must be enabled. Message digest validation deactivates the data path received sequence number update and restricts the minimum local receive window size to 35.

You may choose to configure control channel authentication or control message integrity checking. Control channel authentication requires participation by both peers and a shared secret must be configured on both devices. Control message integrity check is unidirectional and requires configuration on only one of the peers.

The L2TPv3 Control Message Rate Limiting feature was introduced to counter the possibility of a denial-of-service (DoS) attack on a device running L2TPv3. The L2TPv3 Control Message Rate Limiting feature limits the rate at which SCCRQ control packets arriving at the PE that terminates the L2TPv3 tunnel can be processed. SCCRQ control packets initiate the process of bringing up the L2TPv3 tunnel and require a large amount of control plane resources of the PE device.

No configuration is required for the L2TPv3 Control Message Rate Limiting feature. This feature automatically runs in the background in supported releases.

Authentication of L2TPv3 control channel messages occurs using a password that is configured on all participating peer PE devices. Before the introduction of this feature, changing this password required removing of the old password from the configuration before adding the new password, causing an interruption in L2TPv3 services. The authentication password must be updated on all peer PE devices, which are often at different physical locations. It is difficult for all peer PE devices to be updated with the new password simultaneously to minimize interruptions in L2TPv3 services.

The L2TPv3 Digest Secret Graceful Switchover feature allows the password used to authenticate L2TPv3 control channel messages to be changed without tearing down the established L2TPv3 tunnels. This feature works only for authentication passwords configured with the L2TPv3 Control Message Hashing feature. Authentication passwords configured with the older, CHAP-like authentication system cannot be updated without tearing down L2TPv3 tunnels.

The L2TPv3 Digest Secret Graceful Switchover feature allows two control channel passwords to be configured simultaneously, so a new control channel password can be enabled without first removing the old password. Established tunnels are rapidly updated with the new password, but continue to use the old password until it is removed from the configuration. This allows authentication to continue normally with peer PE devices that have not yet been updated to use the new password. After all peer PE devices are configured with the new password, the old password can be removed from the configuration.

During the period when both a new and an old password are configured, authentication will occur only with the new password if the attempt to authenticate using the old password fails.

The pseudowire class configuration procedure creates a configuration template for the pseudowire. Use this template or class to configure session-level parameters for L2TPv3 sessions that are used to transport attachment circuit traffic over the pseudowire.

The pseudowire configuration specifies the characteristics of the L2TPv3 signaling mechanism, including the data encapsulation type, the control protocol, sequencing, Layer 3 fragmentation, payload-specific options, and IP properties. The setting that determines whether signaling is used to set up the pseudowire is also included.

If you specify the encapsulation l2tpv3 command, you cannot remove it by using the no encapsulation l2tpv3 command. You also cannot change the command setting by using the encapsulation mpls command. These methods result in the following error message:

Encapsulation changes are not allowed on an existing pw-class.

To remove the command, you must delete the pseudowire by using the no pseudowire-class command. To change the type of encapsulation, remove the pseudowire by using the no pseudowire-class command, reestablish the pseudowire, and specify the new encapsulation type.

This feature lets you clear L2TPv3 tunnels manually. Before the introduction of this feature, there was no provision to clear a specific L2TPv3 tunnel manually. This functionality provides users more control over an L2TPv3 network.

New and enhanced commands have been introduced to facilitate the management and diagnosis of problems with xconnect configurations. No specific configuration tasks are associated with these commands.

For information about these Cisco IOS commands, go to the Command Lookup Tool at http:/​/​tools.cisco.com/​Support/​CLILookup or to the Cisco IOS Master Commands List, All Releases.

The L2TPv3 Protocol Demultiplexing feature introduces the ability to provide native IPv6 support by utilizing a specialized IPv6 network to offload IPv6 traffic from the IPv4 network. The IPv6 traffic is tunneled to the IPv6 network transparently by using L2TPv3 pseudowires without affecting the configuration of the CE devices. IPv4 traffic is routed as usual within the IPv4 network, maintaining the existing performance and reliability of the IPv4 network.

The IPv4 PE devices must be configured to demultiplex the incoming IPv6 traffic from IPv4 traffic. The PE devices facing the IPv6 network do not require the IPv6 configuration. The configuration of the IPv6 network is beyond the scope of this document. For more information on configuring an IPv6 network, see the IPv6 Configuration Guide.

The L2TPv3 Custom Ethertype for Dot1q and QinQ Encapsulations feature lets you configure an Ethertype other than 0x8100 on Gigabit Ethernet interfaces with the QinQ or Dot1Q encapsulation. You can set the custom Ethertype to 0x9100, 0x9200, or 0x88A8. This allows interoperability in a multivendor Gigabit Ethernet environment.

HDLC for Layer 2 Data Encapsulation provides encapsulation of port-to-port Layer 2 traffic. All HDLC traffic including IPv4, IPv6, and non-IP packet, such as IS-IS, is tunneled over L2TPv3. HDLC does not support interworking mode.

Note	L2TPv3 supports the IPv4 tunnel only for HDLC. The IPv4 tunnel supports IPv4 and IPv6 packets.

An Ethernet frame arriving at a PE device is simply encapsulated in its entirety with an L2TP data header. At the other end, a received L2TP data packet is stripped of its L2TP data header. The payload, an Ethernet frame, is then forwarded to the appropriate attachment circuit.

Because the L2TPv3 tunneling protocol serves essentially as a bridge, it need not examine any part of an Ethernet frame. Any Ethernet frame received on an interface is tunneled, and any L2TP-tunneled Ethernet frame is forwarded out of the interface.

Note	Because of the way in which L2TPv3 handles Ethernet frames, an Ethernet interface must be configured to promiscuous mode to capture all traffic received on the Ethernet segment attached to the device. All frames are tunneled through the L2TP pseudowire.

L2TPv3 supports VLAN memberships in the following ways:

• Port-based, in which undated Ethernet frames are received

• VLAN-based, in which tagged Ethernet frames are received

In L2TPv3, Ethernet xconnect supports port-based VLAN membership and the reception of tagged Ethernet frames. A tagged Ethernet frame contains a tag header (defined in 802.1Q), which is 4 bytes long and consists of a 2-byte tag protocol identifier (TPID) field and a 2-byte tag control information (TCI) field. The TPID indicates that a TCI follows. The TCI is further broken down into the following three fields:

• User priority field

• Canonical format indicator (CFI)

• A 12-bit VLAN ID (VID)

For L2TPv3, an Ethernet subinterface configured to support VLAN switching may be bound to an xconnect service so that all Ethernet traffic, tagged with a VID specified on the subinterface, is tunneled to another PE. The VLAN Ethernet frames are forwarded in their entirety. The receiving PE may rewrite the VID of the tunneled traffic to another value before forwarding the traffic onto an attachment circuit.

To successfully rewrite VLANs, it may be necessary to disable the Spanning Tree Protocol (STP). This can be done on a per-VLAN basis by using the no spanning-tree vlan command.

Note	Because of the way in which L2TPv3 handles VLAN packets, the Ethernet interface must be configured in promiscuous mode to capture all traffic received on the Ethernet segment attached to the device. All frames are tunneled through the L2TP pseudowire.

Upgrading a service provider network to support IPv6 is a long and expensive process. As an interim solution, the Protocol Demultiplexing for L2TPv3 feature introduces the ability to provide native IPv6 support by setting up a specialized IPv6 network and offloading IPv6 traffic from the IPv4 network. IPv6 traffic is tunneled transparently to the IPv6 network using L2TPv3 pseudowires without affecting the configuration of the CE devices. IPv4 traffic is routed as usual within the IPv4 network, maintaining the existing performance and reliability of the IPv4 network.

The figure below shows a network deployment that offloads IPv6 traffic from the IPv4 network to a specialized IPv6 network. The PE devices demultiplex the IPv6 traffic from the IPv4 traffic. IPv6 traffic is routed to the IPv6 network over an L2TPv3 pseudowire, while IPv4 traffic is routed normally. The IPv4 PE devices must be configured to demultiplex the incoming IPv6 traffic from the IPv4 traffic. The PE devices facing the IPv6 network do not require the IPv6 configuration.

Figure 4. Protocol Demultiplexing of IPv6 Traffic from IPv4 Traffic

If no IP address is configured, the protocol demultiplexing configuration is rejected. If an IP address is configured, the xconnect command configuration is rejected unless protocol demultiplexing is enabled in xconnect configuration mode before exiting that mode. If an IP address is configured with an xconnect command configuration and protocol demultiplexing is enabled, the IP address cannot be removed. To change or remove the configured IP address, the xconnect command configuration must first be disabled.

The table below shows the valid combinations of configurations.