The DHCPv4 Relay Agent Identifier Sub-Option

The Dynamic Host Configuration Protocol for IPv4 (DHCPv4) provides IP addresses and configuration information for IPv4 clients. It includes a relay agent capability, in which network elements receive broadcast messages from clients and forward them to DHCP servers as unicast messages. In many network environments, relay agents add information to the DHCP messages before forwarding them, using the Relay Agent Information option . Servers that recognize the Relay Agent Information option echo it back in their replies. This specification introduces a Relay Agent Identifier (Relay-ID) sub-option for the Relay Agent Information option. The Relay-ID sub-option carries a sequence of octets that is intended to uniquely identify the relay agent within the administrative domain. In this document, an administrative domain consists of all DHCP servers and relay agents that communicate with each other.

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in . DHCPv4 terminology is defined in , and the DHCPv4 Relay Agent Information option is defined in .

There has been quite a bit of recent interest in extending the DHCP Leasequery protocol to accommodate some additional situations. proposes a variety of enhancements to the existing Leasequery protocol. The document describes a use case where a relay agent queries DHCP servers using the relay identifier to retrieve all the leases allocated through the relay agent.

DHCP typically identifies clients based on information in their DHCP messages, such as the Client-Identifier option or the value of the chaddr field. In some networks, however, the location of a client -- its point of attachment to the network -- is a more useful identifier. In factory-floor networks (commonly called 'industrial' networks), for example, the role a device plays is often fixed and based on its location. Using manual address configuration is possible (and is common), but it would be beneficial if DHCP configuration could be applied to these networks. One way to provide connection-based identifiers for industrial networks is to have the network elements acting as DHCP relay agents supply information that a DHCP server could use as a client identifier. A straightforward way to form identifier information is to combine something that is unique within the scope of the network element, such as a port/slot value, with something that uniquely identifies that network element, such as a Relay Agent Identifier.

Format of the Relay Agent Identifier sub-option: 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |SUBOPT_RELAY_ID| length | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | . . . identifier (variable) . . . +---------------------------------------------------------------+ Where: SUBOPT_RELAY_ID 12 length the number of octets in the sub-option (excluding the sub-option ID and length fields); the minimum length is one. identifier the identifying data

If the relay identifier is to be meaningful, it has to be stable. A relay agent SHOULD use a single identifier value consistently. The identifier used by a relay device SHOULD be committed to stable storage, unless the relay device can regenerate the value upon reboot. If the Relay-ID configured in a relay agent is not unique within its administrative domain, resource allocation problems may occur as the DHCP server attempts to allocate the same resource to devices behind two different relay agents. Therefore, a Relay-ID configured in a relay agent MUST be unique within its administrative domain. To aid in ensuring uniqueness of Relay-IDs, relay agents SHOULD make their relay identifiers visible to their administrators via their user interface, through a log entry, through a MIB field, or through some other mechanism. Implementors of relay agents should note that the identifier needs to be present in all DHCP message types where its value is being used by the DHCP server. The relay agent may not be able to add the Relay Agent Information option to all messages, such as RENEW messages sent as IP unicasts. In some deployments, that might mean that the server has to be willing to continue to associate the relay identifier it has last seen with a lease that is being RENEWed. Other deployments may prefer to use the Server Identifier Override sub-option to permit the relay device to insert the Relay Agent Information option into all relayed messages. Handling situations where a relay agent device is replaced is another aspect of stability. One of the use cases for the relay identifier is to permit a server to associate clients' lease bindings with the relay device connected to the clients. If the relay device is replaced because it has failed or been upgraded, it may be desirable for the new device to continue to provide the same relay identifier as the old device. Therefore, if a relay agent supports Relay-ID, the Relay-ID should be administratively configurable.

It is strongly recommended that administrators take special care to ensure that Relay-IDs configured in their relay agents are not duplicated. There are a number of strategies that may be used to achieve this. Administrators may use a strategy to configure unique Relay-IDs. One such strategy is that a Relay-ID on a relay agent may reuse an existing identifier or set of identifiers that are already guaranteed to be unique (e.g., Universally Unique Identifier (UUID) ). For administrators who are already using a provisioning system to manage their networking infrastructure, it may work to enumerate relay agents on the basis of roles and then, as a second step, assign those roles to specific relay agents or groups of relay agents. In such a scenario, when a replacement relay agent is first seen by the DHCP server, it could trigger a configuration event on the provisioning system, and the new relay agent could be assigned to the role of the relay agent it is replacing. It may be that the DHCP server has configurable event notification and that a duplicate Relay-ID would trigger this notification. Administrators can take advantage of this feature to work out whether the duplication is real and unintended or whether the original relay agent is being replaced. A network management/provisioning system may also be able to collect a full list of all relay agents on the network. It may then notice that more than one device reports the same Relay-ID. In such a case, the provisioning system could notify the administrator of the fault, which could then be corrected. This is not an exhaustive list of strategies. We suggest an additional strategy in the Security Considerations section. If none of these strategies will work, administrators are also encouraged to consider the specifics of their own network configuration to see if there is some way to detect duplicate Relay-IDs other than the ones listed here.

Security issues with the Relay Agent Information option and its use by servers in address assignment are discussed in and . The DHCP Relay Agent Information option depends on a trusted relationship between the DHCP relay agent and the DHCP server, as described in Section 5 of . While the introduction of fraudulent DHCP Relay Agent Information options can be prevented by a perimeter defense that blocks these options unless the DHCP relay agent is trusted, a deeper defense using the authentication sub-option for the DHCP Relay Agent Information option SHOULD be deployed as well. It also helps in avoiding duplication of relay identifiers by malicious entities. However, implementation of the authentication sub-option for the DHCP Relay Agent Information option is not a must to support the Relay-ID sub-option.

One possible use case for the Relay-ID sub-option is the automated configuration of machines on a factory floor. In this situation, various sections of the factory floor might be on their own network links with a relay agent interposed between those links and the DHCP server. The Relay-ID of each relay agent might cause special configurations to be downloaded to those devices to control their behavior. If a relay agent was deployed on the factory floor in such a situation, with an incorrect Relay-ID, there is the potential that devices could be misconfigured in a way that could produce incorrect results, cause physical damage, or even create hazardous conditions for workers. In deployment scenarios like this one, administrators must use some dependable technique to ensure that such misconfigurations do not occur. It is beyond the scope of this document to provide a complete list of such techniques. However, as an example, a relay agent device intended for use in such a scenario could require the use of a hardware token that contains a Relay-ID that is physically attached to the installation location of the relay agent device and can be connected to and disconnected from the relay agent device without the use of special tools. Such a relay agent device should not be operable when this hardware token is not connected to it: either it should fail because it presents an unknown identifier to the DHCP server, or it should simply refuse to relay DHCP packets until the token is connected to it. A relay agent device that does not provide a clear mitigation strategy for a scenario where misconfiguration could have damaging or hazardous consequences should not be deployed in such a scenario.

IANA has assigned a new sub-option code from the "DHCP Relay Agent Sub-Option Codes" registry maintained at http://www.iana.org/assignments/bootp-dhcp-parameters. Relay Agent Identifier Sub-Option 12

Thanks to Bernie Volz, David W. Hankins, Pavan Kurapati, and Ted Lemon for providing valuable suggestions.