RFC 8947 | DHCPv6 Link-Layer Address Assignment | December 2020 |
Volz, et al. | Standards Track | [Page] |
In certain environments, e.g., large-scale virtualization deployments, new devices are created in an automated manner. Such devices may have their link-layer addresses assigned in an automated fashion. With sufficient scale, the likelihood of a collision using random assignment without duplication detection is not acceptable. Therefore, an allocation mechanism is required. This document proposes an extension to DHCPv6 that allows a scalable approach to link-layer address assignments where preassigned link-layer address assignments (such as by a manufacturer) are not possible or are unnecessary.¶
This is an Internet Standards Track document.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc8947.¶
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
There are several deployment types that deal with a large number of devices that need to be initialized. One of them is a scenario where virtual machines (VMs) are created on a massive scale. Typically, the new VM instances are assigned a link-layer address, but random assignment does not scale well due to the risk of a collision (see Appendix A.1 of [RFC4429]). Another use case is Internet of Things (IoT) devices (see [RFC7228]). The huge number of such devices could strain the IEEE's available Organizationally Unique Identifier (OUI) global address space. While there is typically no need to provide global link-layer address uniqueness for such devices, a link-layer assignment mechanism allows for conflicts to be avoided inside an administrative domain. For those reasons, it is desired to have some form of mechanism that would be able to assign locally unique Media Access Control (MAC) addresses.¶
This document proposes a new mechanism that extends DHCPv6 operation to handle link-layer address assignments.¶
Since DHCPv6 [RFC8415] is a protocol that can allocate various types of resources (non-temporary addresses, temporary addresses, prefixes, as well as many options) and has the necessary infrastructure to maintain such allocations (numerous server and client implementations, large deployed relay infrastructure, and supportive solutions such as leasequery and failover), it is a good candidate to address the desired functionality.¶
While this document presents a design that should be usable for any link-layer address type, some of the details are specific to IEEE 802 48-bit MAC addresses [IEEEStd802]. Future documents may provide specifics for other link-layer address types.¶
IEEE 802 originally set aside half of the 48-bit MAC address space for local use (where the Universal/Local (U/L) bit is set to 1). In 2017, IEEE published an amendment [IEEEStd802c] that divides this space into quadrants with differentiated address rules. More details are in Appendix A.¶
IEEE is also developing protocols and procedures for assignment of locally unique addresses (IEEE 802.1CQ). This work may serve as an alternative protocol for assignment. For additional background, see [IEEE-P802.1CQ-Project].¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The DHCP terminology relevant to this specification from [RFC8415] applies here. The following definitions either modify those definitions as to how they are used in this document or define new terminology used herein.¶
This mechanism is designed to be generic and usable in many deployments, but there are two scenarios it attempts to address in particular: (i) proxy client mode and (ii) direct client mode.¶
This mode is used when an entity acts as a DHCP client that requests that available DHCP servers assign one or more addresses (an address block) for the DHCP client to then assign to the final end devices to use. Large-scale virtualization is one application scenario for proxy client mode. In such environments, this entity is often called a "hypervisor" and is frequently required to spawn new VMs. The hypervisor needs to assign new addresses to those machines. The hypervisor does not use those addresses for itself, but rather it uses them to create new VMs with appropriate addresses. It is worth pointing out the cumulative nature of this scenario. Over time, the hypervisor is likely to increase its address use. Some obsolete VMs will be deleted; their addresses are potentially eligible for reuse by new VMs.¶
This mode can be used when an entity acts as a DHCP client that requests that available DHCP servers assign one or more addresses (an address block) for its own use. This usage scenario is related to IoT (see Section 1). Upon first boot, for each interface, the device uses a temporary address, as described in [IEEEStd802.11] and IEEE 802.1CQ [IEEE-P802.1CQ-Project], to send initial DHCP packets to available DHCP servers wherein the device requests a single address for that network interface. Once the server assigns an address, the device abandons its temporary address and uses the assigned (leased) address.¶
Note that a client that operates as above that does not have a globally unique link-layer address on any of its interfaces MUST NOT use a link-layer-based DHCP Unique Identifier (DUID). For more details, refer to Section 11 of [RFC8415].¶
Also, a client that operates as above may run into issues if the switch it is connected to prohibits or restricts link-layer address changes. This may limit where this capability can be used or may require the administrator to adjust the configuration of the switch(es) to allow a change in address.¶
In the scenarios described in Section 4, the protocol operates in fundamentally the same way. The device requesting an address, acting as a DHCP client, will send a Solicit message with an IA_LL option to all available DHCP servers. That IA_LL option MUST include an LLADDR option specifying the link-layer-type and link-layer-len, and it may include a specific address or address block as a hint for the server. Each available server responds with either a Reply message with committed address(es) (if Rapid Commit was requested and honored) or an Advertise message with offered address(es). The client selects a server's response, as governed by [RFC8415]. If necessary, the client sends a Request message; the target server will then assign the address(es) and send a Reply message. Once a Reply is received, the client can start using those address(es).¶
Normal DHCP mechanisms are in use. The client is expected to periodically renew the addresses as governed by T1 and T2 timers and to stop using the address once the valid lifetime expires. Renewals can be administratively disabled by the server sending "infinity" as the T1 and T2 values (see Section 7.7 of [RFC8415]). An administrator may make the address assignment permanent by specifying use of the "infinity" valid lifetime, as defined in Section 7.7 of [RFC8415].¶
The client can release addresses when they are no longer needed by sending a Release message (see Section 18.2.7 of [RFC8415]).¶
Figure 9 in [RFC8415] shows a timeline diagram of the messages exchanged between a client and two servers for the typical life cycle of one or more leases.¶
Confirm and Information-request messages are not used in link-layer address assignment. Decline should technically never be needed, but see Section 12 for one situation where this message is needed.¶
Clients implementing this mechanism SHOULD use the Rapid Commit option, as specified in Sections 5.1 and 18.2.1 of [RFC8415], to obtain addresses with a two-message exchange when possible.¶
Devices supporting this proposal MAY support the reconfigure mechanism, as defined in Section 18.2.11 of [RFC8415]. If supported by both server and client, the reconfigure mechanism allows the administrator to immediately notify clients that the configuration has changed and triggers retrieval of relevant changes immediately, rather than after the T1 timer elapses. Since this mechanism requires implementation of Reconfiguration Key Authentication Protocol (see Section 20.4 of [RFC8415]), small-footprint devices may choose not to support it.¶
One of the essential aspects of this mechanism is its cumulative nature, especially in the hypervisor scenario. The server-client relationship does not look like other DHCP transactions in the hypervisor scenario. In a typical environment, there would be one server and a rather small number of hypervisors, possibly even only one. However, over time, the number of addresses requested by the hypervisor(s) will increase as more VMs are spawned.¶
Another aspect crucial for efficient design is the observation that a single client acting as hypervisor will likely use thousands of addresses. An approach similar to what is used for IPv6 address or prefix assignment (IA container with all assigned addresses listed, one option for each address) would not work well. Therefore, the mechanism should operate on address blocks rather than single values. A single address can be treated as an address block with just one address.¶
The DHCP mechanisms are reused to a large degree, including message and option formats, transmission mechanisms, relay infrastructure, and others. However, a device wishing to support only link-layer address assignment is not required to support full DHCP. In other words, the device may support only assignment of link-layer addresses but not IPv6 addresses or prefixes.¶
A client MUST send an LLADDR option encapsulated in an IA_LL option to specify the link-layer-type and link-layer-len values. For link-layer-type 1 (Ethernet) and 6 (IEEE 802 Networks), a client sets the link-layer-address field to:¶
Encoding information for other link-layer-types may be added in the future by publishing an RFC that specifies those values.¶
A client sets the extra-addresses field to either 0 for a single address or the size of the requested address block minus 1.¶
A client MUST set the valid-lifetime field to 0 (this field MUST be ignored by the server).¶
The addresses are assigned in blocks. The smallest block is a single address. To request an assignment, the client sends a Solicit message with an IA_LL option inside. The IA_LL option MUST contain an LLADDR option, as specified in Section 7.¶
The server, upon receiving an IA_LL option, inspects its content and may offer an address or addresses for each LLADDR option according to its policy. The server MAY take into consideration the address block requested by the client in the LLADDR option. However, the server MAY choose to ignore some or all parameters of the requested address block. In particular, the server may send either a different starting address or a smaller number of addresses than requested. The server sends back an Advertise message with an IA_LL option containing an LLADDR option that specifies the addresses being offered. If the server is unable to provide any addresses, it MUST return the IA_LL option containing a Status Code option (see Section 21.13 of [RFC8415]) with status set to NoAddrsAvail.¶
Note that servers that do not support the IA_LL option will ignore the option and not return it in Advertise (and Reply) messages. Clients that send IA_LL options MUST treat this as if the server returned the NoAddrsAvail status for these IA_LL option(s).¶
The client waits for available servers to send Advertise responses and picks one server, as defined in Section 18.2.9 of [RFC8415]. The client then sends a Request message that includes the IA_LL container option with the LLADDR option copied from the Advertise message sent by the chosen server.¶
The client MUST process the address block(s) returned in the Advertise, rather than what it included in the Solicit message, and may consider the offered address block(s) in selecting the Advertise message to accept. The server may offer a smaller number of addresses or different addresses from those requested. A client MUST NOT use resources returned in an Advertise message except to select a server and in sending the Request message to that server; resources are only useable by a client when returned in a Reply message.¶
Upon reception of a Request message with the IA_LL container option, the server assigns the requested addresses. The server allocates a block of addresses according to its configured policy. The server MAY assign a different block or smaller block size than requested in the Request message. The server then generates and sends a Reply message back to the client.¶
Upon receiving a Reply message, the client parses the IA_LL container option and may start using all provided addresses. It MUST restart its T1 and T2 timers using the values specified in the IA_LL option.¶
The client MUST use the address block(s) returned in the Reply message, which may be a smaller block(s) or may have a different address(es) than requested.¶
A client that has included a Rapid Commit option in the Solicit message may receive a Reply in response to the Solicit message and skip the Advertise and Request message steps above (see Section 18.2.1 of [RFC8415]).¶
A client that changes its link-layer address on an interface SHOULD follow the recommendations in Section 7.2.6 of [RFC4861] to inform its neighbors of the new link-layer address quickly.¶
Address renewals follow the normal DHCP renewals processing described in Section 18.2.4 of [RFC8415]. Once the T1 timer elapses, the client starts sending Renew messages with the IA_LL option containing an LLADDR option for the address block being renewed. The server responds with a Reply message that contains the renewed address block. The server MUST NOT shrink or expand the address block. Once a block is assigned and has a non-zero valid lifetime, its size, starting address, and ending address MUST NOT change.¶
If the requesting client needs additional addresses (e.g., in the hypervisor scenario because addresses need to be assigned to new VMs), it MUST send an IA_LL option with a different Identity Association IDentifier (IAID) to create another "container" for more addresses.¶
If the client is unable to renew before the T2 timer elapses, it starts sending Rebind messages, as described in Section 18.2.5 of [RFC8415].¶
The client may decide to release a leased address block. A client MUST release the block in its entirety. A client releases an address block by sending a Release message that includes an IA_LL option containing the LLADDR option for the address block to release. The Release transmission mechanism is described in Section 18.2.7 of [RFC8415].¶
Note that if the client is releasing the link-layer address it is using, it MUST stop using this address before sending the Release message (as per [RFC8415]). In order to send the Release message, the client MUST use another address (such as the one originally used to initiate DHCPv6 to provide an allocated link-layer address).¶
This mechanism uses an approach similar to the existing mechanisms in DHCP. There is one container option (the IA_LL option) that contains the actual address or addresses, represented by an LLADDR option. Each LLADDR option represents an address block, which is expressed as a first address with a number that specifies how many additional addresses are included.¶
The Identity Association for Link-Layer Addresses option (the IA_LL option) is used to carry an IA_LL, the parameters associated with the IA_LL, and the address blocks associated with the IA_LL.¶
The format of the IA_LL option is:¶
An IA_LL option may only appear in the options area of a DHCP message. A DHCP message may contain multiple IA_LL options (though each must have a unique IAID).¶
The status of any operations involving this IA_LL is indicated in a Status Code option (see Section 21.13 of [RFC8415]) in the IA_LL-options field.¶
Note that an IA_LL has no explicit "lifetime" or "lease length" of its own. When the valid lifetimes of all of the addresses in an IA_LL have expired, the IA_LL can be considered to be expired. T1 and T2 are included to give servers explicit control over when a client recontacts the server about a specific IA_LL.¶
In a message sent by a client to a server, the T1 and T2 fields MUST be set to 0. The server MUST ignore any values in these fields in messages received from a client.¶
In a message sent by a server to a client, the client MUST use the values in the T1 and T2 fields for the T1 and T2 times, unless those values in those fields are 0. The values in the T1 and T2 fields are the number of seconds until T1 and T2.¶
As per Section 7.7 of [RFC8415], the value 0xffffffff is taken to mean "infinity" and should be used carefully.¶
The server selects the T1 and T2 times to allow the client to extend the lifetimes of any address block in the IA_LL before the lifetimes expire, even if the server is unavailable for some short period of time. Recommended values for T1 and T2 are .5 and .8 times the shortest valid lifetime of the address blocks in the IA that the server is willing to extend, respectively. If the "shortest" valid lifetime is 0xffffffff ("infinity"), the recommended T1 and T2 values are also 0xffffffff. If the time at which the addresses in an IA_LL are to be renewed is to be left to the discretion of the client, the server sets T1 and T2 to 0. The client MUST follow the rules defined in Section 14.2 of [RFC8415].¶
If a client receives an IA_LL with T1 greater than T2, and both T1 and T2 are greater than 0, the client discards the IA_LL option and processes the remainder of the message as though the server had not included the invalid IA_LL option.¶
The IA_LL-options field typically contains one or more LLADDR options (see Section 11.2). If a client does not include an LLADDR option in a Solicit or Request message, the server MUST treat this as a request for a single address and that the client has no hint as to the address it would like.¶
The Link-Layer Addresses option is used to specify an address block associated with an IA_LL. The option must be encapsulated in the IA_LL-options field of an IA_LL option. The LLaddr-options field encapsulates those options that are specific to this address block.¶
The format of the Link-Layer Addresses option is:¶
In a message sent by a client to a server, the valid lifetime field MUST be set to 0. The server MUST ignore any received value.¶
In a message sent by a server to a client, the client MUST use the value in the valid lifetime field for the valid lifetime for the address block. The value in the valid lifetime field is the number of seconds remaining in the lifetime.¶
As per Section 7.7 of [RFC8415], the valid lifetime of 0xffffffff is taken to mean "infinity" and should be used carefully.¶
More than one LLADDR option can appear in an IA_LL option.¶
A server selects link-layer addresses to be assigned to an IA_LL according to the assignment policies determined by the server administrator and the requirements of that address space.¶
Link-layer addresses are typically specific to a link and the server SHOULD follow the steps in Section 13.1 of [RFC8415] to determine the client's link.¶
For IEEE 802 MAC addresses (see [IEEEStd802] as amended by [IEEEStd802c]):¶
IANA has assigned the OPTION_IA_LL (138) option code from the "Option Codes" subregistry of the "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)" registry maintained at <http://www.iana.org/assignments/dhcpv6-parameters>:¶
IANA has assigned the OPTION_LLADDR (139) option code from the "Option Codes" subregistry of the "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)" registry maintained at <http://www.iana.org/assignments/dhcpv6-parameters>:¶
See Section 22 of [RFC8415] and Section 23 of [RFC7227] for the DHCP security considerations. See [RFC8200] for the IPv6 security considerations.¶
As discussed in Section 22 of [RFC8415]:¶
DHCP lacks end-to-end encryption between clients and servers; thus, hijacking, tampering, and eavesdropping attacks are all possible as a result.¶
In some network environments, it is possible to secure them, as discussed later in Section 22 of [RFC8415].¶
If not all parties on a link use this mechanism to obtain an address from the space assigned to the DHCP server, there is the possibility of the same link-layer address being used by more than one device. Note that this issue would exist on these networks even if DHCP were not used to obtain the address.¶
Server implementations SHOULD consider configuration options to limit the maximum number of addresses to allocate (both in a single request and in total) to a client. However, note that this does not prevent a bad client actor from pretending to be many different clients and consuming all available addresses.¶
See Section 23 of [RFC8415] for the DHCP privacy considerations.¶
For a client requesting a link-layer address directly from a server, as the address assigned to a client will likely be used by the client to communicate on the link, the address will be exposed to those able to listen in on this communication. For those peers on the link that are able to listen in on the DHCP exchange, they would also be able to correlate the client's identity (based on the DUID used) with the assigned address. Additional mechanisms, such as the ones described in [RFC7844], can also be used to improve anonymity by minimizing what is exposed.¶
As discussed in Section 23 of [RFC8415], DHCP servers and hypervisors may need to consider the implications of assigning addresses sequentially. Though in general, this is only of link-local concern unlike for IPv6 address assignment and prefix delegation, as these may be used for communication over the Internet.¶
This appendix provides a brief summary of IEEE 802c [IEEEStd802c].¶
The original IEEE 802 specifications assigned half of the 48-bit MAC address space to local use -- these addresses have the U/L bit set to 1 and are locally administered with no imposed structure.¶
In 2017, the IEEE issued the IEEE Std 802c specification, which defines a new optional "Structured Local Address Plan (SLAP) that specifies different assignment approaches in four specified regions of the local MAC address space". Under this plan, there are four SLAP quadrants that use different assignment policies.¶
The first octet of the MAC address Z and Y bits define the quadrant for locally assigned addresses (X-bit is 1). In IEEE representation, these bits are as follows:¶
The SLAP quadrants are:¶
Quadrant | Y-bit | Z-bit | Local Identifier Type | Local Identifier |
---|---|---|---|---|
01 | 0 | 1 | Extended Local | ELI |
11 | 1 | 1 | Standard Assigned | SAI |
00 | 0 | 0 | Administratively Assigned | AAI |
10 | 1 | 0 | Reserved | Reserved |
MAC addresses derived from an Extended Local Identifier (ELI) are based on an assigned Company ID (CID), which is 24 bits (including the M, X, Y, and Z bits) for 48-bit MAC addresses. This leaves 24 bits for the locally assigned address for each CID for unicast (M-bit = 0) and also for multicast (M-bit = 1). The CID is assigned by the IEEE Registration Authority (RA).¶
MAC addresses derived from a Standard Assigned Identifier (SAI) are assigned by a protocol specified in an IEEE 802 standard. For 48-bit MAC addresses, 44 bits are available. Multiple protocols for assigning SAIs may be specified in IEEE standards. Coexistence of multiple protocols may be supported by limiting the subspace available for assignment by each protocol.¶
MAC addresses derived from an Administratively Assigned Identifier (AAI) are assigned locally. Administrators manage the space as needed. Note that multicast IPv6 packets [RFC2464] use a destination address starting in 33-33, so AAI addresses in that range should not be assigned. For 48-bit MAC addresses, 44 bits are available.¶
The last quadrant is reserved for future use. While this quadrant may also be used similar to AAI space, administrators should be aware that future specifications may define alternate uses that could be incompatible.¶
Thanks to the DHC Working Group participants that reviewed this document and provided comments and support. With special thanks to Ian Farrer for his thorough reviews and shepherding of this document through the IETF process. Thanks also to directorate reviewers Samita Chakrabarti, Roni Even, and Tianran Zhou and IESG members Martin Duke, Benjamin Kaduk, Murray Kucherawy, Warren Kumari, Barry Leiba, Alvaro Retana, Éric Vyncke, and Robert Wilton for their suggestions. And thanks to Roger Marks, Robert Grow, and Antonio de la Oliva for comments related to IEEE work and references.¶