Stuart Cheshire Document: draft-ietf-zeroconf-ipv4-linklocal-03.txt Apple Computer Expires 22nd December 2001 Bernard Aboba Microsoft 22nd June 2001 Dynamic Configuration of IPv4 Link-Local Addresses <draft-ietf-zeroconf-ipv4-linklocal-03.txt> Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html Distribution of this memo is unlimited. Abstract As the Internet Protocol continues to increase in popularity, it becomes increasingly valuable to be able to use familiar IP tools such as ftp not only for global communication, but for local communication as well. For example, two people with laptop computers with built-in wireless Ethernet may meet and wish to exchange files. It is desirable for these people to be able to use IP application software without the inconvenience of having to manually configure static IP addresses or set up a DHCP server [RFC 2131]. This document describes a method by which a host may automatically configure an interface with an IPv4 address in the 169.254/16 prefix that is valid for link-local communication on that interface. This is especially valuable in environments where no other configuration mechanism is available. Microsoft Windows 98 (and later) and Mac OS 8.5 (and later) already implement versions of this functionality. This document standardizes this protocol and simplifies it in one important way -- in previous implementations, link-local addresses were only available after a host had tried and failed to contact a DHCP server. This standard removes that restriction, making link-local addresses available all the time, independent of DHCP. Expires 22nd December 2001 Cheshire & Aboba [Page 1]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 Table of Contents 1. Introduction..................................................2 1.1 Conventions and Terminology Used in this Document.............3 1.2 Applicability.................................................3 1.3 Issues with Autoconfiguration.................................4 1.4 169.254/16 Not To Be Used for Other Purposes..................4 1.5 Supporting Multiple Addresses per Interface...................5 1.6 Supporting Multiple Interfaces................................5 2. IPv4 Link-Local Address Selection, Defense and Delivery.......6 2.1 Selecting Link-Local Addresses................................6 2.2 Claiming Link-Local Addresses.................................6 2.3 Address Collision Detection and Address Defense...............8 2.4 Source Address Selection......................................9 2.5 Link-Local Addresses Are Not Forwarded........................9 3. Considerations for Multiple Interfaces.......................11 3.1 Link-Local Addressing on Only One Interface..................11 3.2 Single Shared Link-Local Address.............................11 3.3 Different Link-Local Address for Each Interface..............12 3.4 Collision Detection for Multi-Homed Hosts....................14 4. Considerations for Joining of Previously Separate Networks...15 5. Security Considerations......................................16 6. Acknowledgements.............................................16 7. Copyright....................................................16 8. References...................................................17 9. Authors' Addresses...........................................18 Appendix. Prior Implementations..................................18 A.1 Apple Mac OS 8.x and 9.x.....................................18 A.2 Microsoft Windows 98/98SE....................................19 A.3 Windows 2000 and Windows ME..................................20 1. Introduction As the Internet Protocol continues to increase in popularity, it becomes increasingly valuable to be able to use familiar IP tools such as ftp not only for global communication, but for local communication as well. For example, two people with laptop computers with built-in wireless Ethernet may meet and wish to exchange files. It is desirable for these people to be able to use IP application software without the inconvenience of having to manually configure static IP addresses or set up a DHCP server [RFC 2131]. This document describes a method by which a host may automatically configure an interface with an IPv4 address in the 169.254/16 prefix that is valid for link-local communication on that interface. This is especially valuable in environments where no other configuration mechanism is available. Allocation of link-local IPv6 addresses is described in [RFC 2462]. Expires 22nd December 2001 Cheshire & Aboba [Page 2]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 Microsoft Windows 98 (and later) and Mac OS 8.5 (and later) already implement versions of this functionality. This document standardizes this protocol and simplifies it in one important way -- in previous implementations, link-local addresses were only available after a host had tried and failed to contact a DHCP server. This standard removes that restriction, making link-local addresses available all the time, independent of DHCP. A host need not implement a DHCP client in order to use link-local addresses. Even for hosts that do implement a DHCP client, information received from DHCP servers has no bearing on a host's configuration of link-local addresses. This document prescribes rules for how IPv4 link-local addresses MUST be treated by hosts and routers. In particular, it describes how routers MUST behave when receiving packets with IPv4 link-local addresses in the source or destination address. With respects to hosts, it discusses claiming and defending addresses, maintaining link-local and routable addresses on the same interface, and multihoming issues. 1.1. Conventions and Terminology Used in this Document 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 [RFC 2119]. This document uses the term "routable address" to refer to all unicast addresses outside the 169.254/16 prefix, including global addresses and private addresses such as Net 10/8 [RFC 1918], all of which may be forwarded via routers. Wherever this document uses the term "host" when describing use of link-local addresses, the text applies equally to routers using link-local addresses on any or all interfaces. Wherever this document uses the term "ARP packet" or "ARP packets", [RFC 826] the text applies uniformly to both ARP request and ARP reply packets. Anywhere that the text applies only to one or other kind of ARP packet, this document explicitly states either "ARP request" or "ARP reply". 1.2. Applicability The specifications in this document apply to any IEEE 802 media [IEEE802] including Ethernet, and other similar link-layer ("layer two") network technologies that operate at data rates of at least 1Mb/s, have a round-trip latency of at most one second, and support ARP [RFC 826]. Wherever this document uses the term "Ethernet", the text applies equally to any of these network technologies. Expires 22nd December 2001 Cheshire & Aboba [Page 3]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 Link-layer network technologies that support ARP but operate at rates below 1Mb/s or latencies above one second may need to specify different values for the following parameters described in Sections 2.2 and 2.3: (a) the number of, and interval between, ARP probes, (b) the number of, and interval between, gratuitous ARPs, (c) the maximum rate at which address claiming may be attempted, and (d) the time interval between conflicting ARPs below which a host MUST reconfigure instead of attempting to defend its address. Link-layer network technologies that do not support ARP may be able to use other techniques for determining whether a particular IP address is currently in use. However, the application of claim-and- defend mechanisms to such networks is left to a future document. This document describes link-local addressing, for IP communication between two hosts on a single link. Two hosts are considered to be on the same link if, when one host sends packets to the other, the entire link-layer packet payload arrives unmodified. In particular, any device forwarding a packet whose source and/or destination address is in the 169.254/16 prefix MUST NOT modify any part of the IP header. This means that the packet may pass through devices such as Ethernet switches, bridges, hubs or repeaters, but not through devices like IP routers that modify the IP header. 1.3. Issues with Autoconfiguration Implementations of IPv4 link-local address autoconfiguration MUST expect address collisions, and MUST be prepared to handle them grace- fully by automatically selecting a new address whenever a collision is detected, as described in Section 2. This requirement to detect and handle address collisions applies during the entire period that a host is using a 169.254/16 link-local address, not just during some 'startup phase' of initial interface configuration. For example, address collisions can occur well after a host has completed booting if two previously separate networks are joined, as described in Section 4. 1.4. 169.254/16 Not To Be Used for Other Purposes Note that addresses in the 169.254/16 prefix SHOULD NOT be configured manually or by a DHCP server. Manual or DHCP configuration may cause a host to use an address in the 169.254/16 prefix without following the special rules regarding duplicate detection and automatic conf- iguration that pertain to addresses in this prefix. While RFC 2131 indicates that a DHCP client SHOULD probe a newly received address with ARP, this is not mandatory. Similarly, while RFC 2131 recommends Expires 22nd December 2001 Cheshire & Aboba [Page 4]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 that a DHCP server SHOULD probe an address using an ICMP Echo Request before allocating it, this is also not mandatory, and even if the server does this, link-local addresses are not routable, so a DHCP server not directly connected to a link cannot detect whether a host on that link is already using the desired IPv4 link-local address. Administrators wishing to configure their own local addresses (using manual configuration, a DHCP server, or any other mechanism not described in this document) should use one of the existing private address prefixes [RFC 1918], not the 169.254/16 prefix. 1.5. Supporting Multiple Addresses per Interface IPv4 link-local addresses are independent from any other IPv4 addresses that a host may have. Each interface on a host may have a link-local address in addition to zero or more other addresses configured by other means (e.g. manually or via a DHCP server). There are several reasons why it is beneficial for hosts to maintain link-local addresses in addition to any other addresses they may have. For example, a DHCP server may appear on a network where hosts are already communicating using link-local addresses, and it is beneficial for already-established link-local TCP connections to continue working even after the hosts have contacted the DHCP server and configured additional routable addresses. Another example of why it is beneficial for hosts to maintain link- local addresses in addition to other addresses is that there may be networks where some of the hosts have only link-local addresses. For example, a user with a wireless home network may have a laptop computer and an IP printer. The laptop computer may have both a self-configured link-local address and a DHCP-configured global address. The printer, in contrast, may have only a link-local address, because the user does not need (or want) the printer to be globally addressable. In this case, the laptop computer would access pages on the World Wide Web using its globally-routable address to communicate with servers world-wide, but print those web pages using its link-local address to communicate with its local printer. 1.6. Supporting Multiple Interfaces Hosts that support multi-homing have additional considerations if they wish to use IPv4 link-local addresses on more than one interface at a time. These are discussed in Section 3. Expires 22nd December 2001 Cheshire & Aboba [Page 5]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 2. IPv4 Link-Local Address Selection, Defense and Delivery The following section explains the link-local address selection algorithm, how link-local addresses are defended, and how IPv4 packets with link-local addresses are delivered. Windows and Mac OS hosts that already implement IPv4 address auto- configuration are compatible with the rules presented in this section. However, should any interoperability problem be discovered, this document, not any prior implementation, defines the standard. 2.1 Selecting Link-Local Addresses When a host wishes to configure a link-local address, it selects an address using a pseudo-random number generator with a uniform distribution in the range from 169.254.1.0 to 169.254.254.255. The IPv4 network 169.254/16 is registered with the IANA for this purpose. The first 256 and last 256 addresses in the 169.254/16 network are reserved for use by future IETF-Standard protocols and MUST NOT be selected by a host using this dynamic configuration mechanism. The pseudo-random number generation algorithm should be chosen so that different hosts do not generate the same sequence of numbers. Issues with pseudo-random number generators are discussed in "Random- ness Recommendations for Security" [RFC 1750]. If the host has access to persistent information that is different for each host, such as its burned-in Ethernet hardware address, then the pseudo-random number generator SHOULD be seeded using a value derived from this information. This means that even without using any other persistent storage, a host will usually select the same link-local address each time it is booted, which can be convenient for debugging and other operational reasons. Seeding the pseudo-random number generator using the real-time clock is NOT suitable for this purpose, since a group of hosts that are all powered on at the same time might then all generate the same sequence. 2.2 Claiming Link-Local Addresses After it has selected an address, a host MUST test to see if the address is already in use before beginning to use it. On a network such as Ethernet that supports ARP, this test is done using ARP probes. On link-layer network technologies that do not support ARP other techniques may be available for determining whether a partic- ular IP address is currently in use. However, the application of claim-and-defend mechanisms to such networks is left to a future document. A host probes to see if an address is already in use by broadcasting an ARP request for the desired address. The client MUST fill in the 'sender hardware address' field of the ARP packet with the hardware Expires 22nd December 2001 Cheshire & Aboba [Page 6]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 address of the interface through which it is sending the packet. The 'sender IP address' field MUST be set to all zeroes, to avoid polluting ARP caches in other hosts on the same link in the case where the address turns out to be already in use by another host. The 'target hardware address' field is ignored and SHOULD be set to all zeroes. The 'target IP address' field MUST be set to the address being probed. An ARP request constructed this way with an all-zero 'sender IP address' is referred to as an "ARP probe". On Ethernet four such probes should be sent, with a two-second wait after each probe (a total of eight seconds). If during this period the host receives any ARP packet where the packet's 'sender IP address' is the address being probed for, then the host MUST treat this address as being in use by some other host, and MUST select a new pseudo-random address and repeat the process. In addition, if during this period the host receives any ARP packet where the packet's 'target IP address' is the address being probed for, and the packet's 'sender hardware address' is not the hardware address of any of the host's interfaces, then the host MUST similarly treat this as an address collision and select a new address as above. This can occur if two (or more) hosts attempt to configure the same link- local address at the same time. A host should maintain a counter of the number of address collisions it has experienced in the process of trying to acquire an address, and if the number of collisions exceeds ten then the host MUST limit the rate at which it probes for new addresses to no more than one new address per second. This is to prevent catastrophic ARP storms in pathological failure cases, such as a rogue host that answers all ARP probes, causing legitimate hosts to go into an infinite loop attempting to select a usable address. After successfully configuring a link-local address, a host on Ethernet SHOULD broadcast two gratuitous ARPs, spaced two seconds apart. A gratuitous ARP is identical to the ARP probe described above, except that now the sender and target IP addresses are both set to the host's newly selected IP address. The purpose of these gratuitous ARPs is to make sure that other hosts on the link do not have stale ARP cache entries left over from some other host that may previously have been using the same address. Hosts that are equipped with persistent storage MAY, for each interface, record the IP address they have selected. On booting, hosts with a previously recorded address SHOULD use that address as their first candidate when probing. This increases the stability of addresses. For example, if a group of hosts are powered off at night, then when they are powered on the next morning they will all resume using the same addresses, instead of picking different addresses and potentially having to resolve conflicts that arise. Expires 22nd December 2001 Cheshire & Aboba [Page 7]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 2.3 Address Collision Detection and Address Defense Address collision detection is not limited to the address selection phase, when a host is sending ARP probes. Address collision detection is an ongoing process that is in effect for as long as a host is using a link-local IPv4 address. At any time, if a host receives an ARP packet where the 'sender IP address' is the host's own IP address, but the 'sender hardware address' does not match any of the host's own interface addresses, then this is a conflicting ARP packet, indicating an address collision. A host MUST respond to a conflicting ARP packet as described in either (a) or (b) below: (a) Upon receiving a conflicting ARP packet, a host MAY elect to imm- ediately configure a new link-local IP address as described above, or (b) If a host currently has active TCP connections or other reasons to prefer to keep the same IP address, and it has not seen any other conflicting ARP packets recently (for Ethernet, within the last ten seconds) then it MAY elect to attempt to defend its address, by recording the time that the conflicting ARP packet was received, and then broadcasting one single gratuitous ARP request, giving its own IP and hardware addresses as the sender addresses of the ARP. Having done this, the host can then continue to use the address normally without any further special action. However, if this is not the first conflicting ARP packet the host has seen, and the time recorded for the previous conflicting ARP packet is recent (within ten seconds for Ethernet) then the host MUST immediately cease using this address and configure a new link-local IP address as described above. This is necessary to ensure that two hosts do not get stuck in an endless loop with both hosts trying to defend the same address. A host MUST respond to conflicting ARP packets as described in either (a) or (b) above. A host MUST NOT ignore conflicting ARP packets. Forced address reconfiguration may be disruptive, causing TCP conn- ections to be broken. However, it is expected that such disruptions will be rare, and if inadvertent address duplication happens, then disruption of communication is inevitable, no matter how the addresses were assigned. It is not possible for two different hosts using the same IP address on the same network to operate reliably. Immediately configuring a new address as soon as the conflict is detected is the best way to restore useful communication as quickly as possible. The mechanism described above of broadcasting a single gratuitous ARP to defend the address mitigates the problem somewhat, by helping to improve the chance that one of the two conflicting hosts may be able to retain its address. All ARP packets (replies as well as requests) that contain a link- local sender address SHOULD be sent using link-level broadcast instead of link-level unicast. This aids timely detection of duplicate addresses. An example illustrating how this helps is given in Section 4. Expires 22nd December 2001 Cheshire & Aboba [Page 8]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 2.4 Source Address Selection Since each interface on a host may have a link-local address in addition to zero or more other addresses configured by other means (e.g. manually or via a DHCP server), a host may have to make a choice about what source address to use when it sends a packet or initiates a TCP connection. If the destination address is in the 169.254/16 prefix (including the 169.254.255.255 broadcast address), the host MUST use its link-local source address. If the destination address is: (a) The 255.255.255.255 broadcast address, or (b) a multicast address with link-local scope, or (c) a multicast address with scope larger than link-local, but the TTL in the IP header is set to 1 so the packet is limited to the local link, then the host may use either its link-local source address or a routable address. If the destination address is: (a) a unicast address outside the 169.254/16 prefix, or (b) a multicast address with scope larger than link-local and the TTL in the IP header is greater than 1, then the host MUST NOT use its link-local source address, and should instead use an appropriate routable source address. In the case where two hosts on the same link each have both a link- local address and another address configured via some other means, the host may use either its link-local source address or a routable address, depending on which is expected to be more likely to remain stable for the expected lifetime of the connection. Note that link- local addresses may be forced to change over time, as described in Section 2.3. Any host receiving an IPv4 packet whose destination addresses is in the 169.254/16 prefix MUST discard the packet if the source address is not in the 169.254/16 prefix. 2.5 Link-Local Addresses Are Not Forwarded Any host sending an IPv4 packet to a destination addresses in the 169.254/16 prefix MUST set the TTL in the IP header to 255. Any host receiving an IPv4 packet whose destination addresses is in the 169.254/16 prefix MUST discard the packet if the TTL in the IP header is not 255. This is to guard against misconfigured routers which may allow packets to leak in from outside the local link. Since even the most dysfunctional router will decrement the TTL in the IP header, Expires 22nd December 2001 Cheshire & Aboba [Page 9]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 a host receiving a packet with a destination addresses in the 169.254/16 prefix and TTL less than 255 can detect that it has leaked in from outside the local link, and correctly discard it. An IPv4 packet whose source and/or destination address is in the 169.254/16 prefix MUST NOT be sent to any router for forwarding, and any network device receiving such a packet MUST NOT forward it, regardless of the TTL in the IP header. This restriction also applies to multicast packets. IP packets with a link-local source address MUST NOT be forwarded off the local link even if they have a multicast destination address. Similar considerations apply at layers above IP. For example, DNS Resource Records containing link-local addresses SHOULD NOT be sent to hosts outside the link to which those link-local addresses apply. Automatically generated web pages SHOULD NOT contain links with embedded link-local addresses if those pages are viewable from hosts outside the local link where the addresses are valid. Since DNS treats Resource Record Sets [RFC 2181] as indivisible units (e.g. for generating DNS reply packets, signatures, etc.), RRSets SHOULD only contain A records where all the addresses have the same scope. Link- local and routable addresses SHOULD NOT be mixed in a single set. The non-forwarding rule means that hosts may assume that all 169.254/16 destination addresses are "on-link" and directly reachable. The 169.254/16 address prefix MUST NOT be subnetted. This specification utilizes ARP-based address collision detection, which functions by broadcasting on the local subnet. Since such broadcasts are not forwarded, were subnetting to be allowed then address conflicts could remain undetected. The non-forwarding rule is important because it is expected that many link-local-only devices will be extremely simple devices of the kind that currently use X10 [X10], USB [USB] or FireWire [IEEE1394]. The designers of these devices assume that they will communicate only with other local devices, and implement a degree of security appropriate for that expected environment. (For example, a typical USB mouse does not have a password, nor does it encrypt its mouse- movement data, and in most environments this is acceptable.) Any network gateway device that blindly forwards the contents of link-local IP packets off the local network (or vice-versa) exposes these link-local-only devices to a much greater degree of risk than their designers may have planned for. This does not mean that link-local devices are forbidden from communicating outside the local link. IP hosts that implement both link-local and conventional routable IP addresses may still use their routable addresses without restriction as they do today. Extremely simple IP devices that implement only link-local addresses Expires 22nd December 2001 Cheshire & Aboba [Page 10]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 may also communicate with hosts outside the local link, provided that such communication is mediated through a device capable of enforcing appropriate security controls. For example, a home heating system could be controlled via a Web server on the local network, where the Web server uses cryptographic methods to verify the authority of the remote user, and then uses link-local communication on the local network to communicate commands to the heating system's thermostat. It should be understood that this mediated communication is not mandatory; it is an option afforded to designers of very simple devices who wish to implement only link-local addresses and thereby simplify their security assumptions. Any designer of a device desiring unmediated communication outside the local link need only implement today's conventional IP host software (e.g. a DHCP client) in order to enjoy the same degree of global addressability available to other conventional IP hosts. 3. Considerations for Multiple Interfaces This section does not specify protocol requirements; it offers imple- mentation advice. Implementers of multi-homed hosts have three basic choices: (1) implement link-local addresses on only one active inter- face at a time, (2) adopt networking APIs that include interface identifiers, and implement a single shared link-local address used by all active interfaces, or (3) provide networking APIs that identify interfaces by IP address (the norm) and maintain some additional address uniqueness properties in order to support the use of IP addresses as unambiguous interface identifiers. This section discusses these three choices. 3.1 Link-Local Addressing on Only One Interface A multi-homed host may elect to configure an IPv4 link-local address on only one of its active interfaces. In some situations this will be adequate. Implementers who are not planning to support IPv4 link- local addresses simultaneously on multiple interfaces may skip the remainder of this section. 3.2 Single Shared Link-Local Address Some operating systems have networking APIs that allow applications to specify network interfaces by name, index value, or other local identifier, and to identify interfaces this way both for incoming and outgoing packets/connections. For operating systems that have this capability (and have application software that makes use of it) it is sufficient to configure all interfaces with a single common IPv4 link-local address, and defend that address on all interfaces. Expires 22nd December 2001 Cheshire & Aboba [Page 11]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 3.3 Different Link-Local Address for Each Interface Many operating systems have networking APIs that allow applications to bind endpoints only to a desired source IP address, or similarly when a packet is received, identify the interface on which it arrived only by its IP address, not by its name or other identifier. A multi- homed host in this situation should ensure that each of its inter- faces is configured with a different link-local address, to enable use of the source address as an unambiguous interface identifier. To provide this property, if the selection algorithm chooses an address that is already in use on one of the host's other interfaces, then the process should be repeated until a unique address is selected. In addition, this kind of host should probe for, and defend, all of its link-local addresses on all of its active interfaces that are using link-local addresses. When bringing up a new interface, the host SHOULD first probe for all of its existing link-local addresses on that new interface. If any of the addresses are found to be already in use by some other host on that new link, then the interfaces in question MUST be reconfigured to select new unique link-local addresses. The host should then select a link-local address for the new interface, and probe on all of its active interfaces to verify that this new address is unique on all the links that the host has connections to. This uniqueness requirement simplifies host software layers above IP (application software and transport protocols) by allowing them to identify connections using only the source and destination IP addresses, without also needing interface identifiers. Since link- local addresses are only unique per-link, hosts on different links could be using the same link-local address. By requiring uniqueness of source addresses on the multi-homed host, this ensures that TCP connections to hosts using the same link-local destination addresses on different links can be disambiguated by their different source addresses. Note that this also requires that the multi-homed host using different link-local addresses on multiple interfaces MUST implement the "strong end system" model [RFC 1122] for packets from link-local source addresses, so that packets are only sent out from the interface that matches the source address in the packet. (The "weak end system" model may still be used for other packets.) When bringing up multiple interfaces simultaneously (e.g. at system boot time) the process can be streamlined by configuring all inter- faces concurrently. First all interfaces should be brought up in an un-numbered state. Once this is done, all interfaces can then concurrently go through the process of selecting a locally unique address and probing for that address simultaneously on all active interfaces (including all other interfaces that themselves are currently in the process of selecting and probing for an address). Expires 22nd December 2001 Cheshire & Aboba [Page 12]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 3.3.1 Example Illustrating (Incorrect) Ambiguous Address Re-Use Figure 1 shows a network topology where host A has an interface on link X with link-local address P, and host B, also on link X, has address Q. If we allow host A to have another interface on a different link that also has address Q, then although there are no conflicts strictly within the scope of either link, there is potential for confusion. When host A sends a UDP packet from source address P to destination address Q without specifying an explicit outgoing interface identifier, it is ambiguous whether A intends to talk to itself, or to host B. By ensuring that all of a host's link-local addresses are distinct not only from each other, but also from all addresses currently active on all links that the host is connected to, we remove this ambiguity. | | | P ----- Q? | |-----| A |-----| | ----- | | | | | X| |Y | | | | ----- | | | B |-----| | ----- Q | | | | Figure 1. (Incorrect) Ambiguous Re-Use of Address Q 3.3.2 Example Illustrating Acceptable Address Re-Use Note that it is acceptable for different hosts on separate links to be using the same link-local address on their respective separate links. Figure 2 shows a network topology where host C on link X is using address R, while at the same time, host D on link Y is also using address R. This is entirely in keeping with the concept of link-local addresses. Link-local addresses are only unique amongst the member hosts of a single link. Hosts C and D are not on the same link, hence there is no requirement for them to have distinct addresses. Note that in this case, host A is still able to communic- ate with both hosts C and D, because a packet sent from source address P to destination address R travels on link X to host C, and a packet sent from source address Q to destination address R travels on link Y to host D. TCP connections are uniquely identified by the source and destination addresses and port numbers, not just by the Expires 22nd December 2001 Cheshire & Aboba [Page 13]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 destination address and port number alone. To support link-local addressing on multiple interfaces simultaneously, the network software API must allow applications to bind endpoints to a desired source interface as well as specifying the desired destination address for a TCP connection. Networking implementations that only allow the destination address to be specified should limit themselves to configuring only one interface for link-local addressing. | | | P ----- Q | |-----| A |-----| | ----- | | | X| |Y | | ----- | | ----- | C |-----| |-----| D | ----- R | | R ----- | | Figure 2. Acceptable Re-Use of Address R 3.4 Collision Detection for Multi-Homed Hosts If a host receives an ARP packet on an interface that's using a link- local address, where the 'sender IP address' in the ARP packet is equal to *any* of the host's current link-local addresses, and the 'sender hardware address' in the ARP packet is equal to *none* of the host's active interfaces, then this is a conflicting ARP packet and must be handled as such. If a host receives an ARP packet where the 'sender hardware address' in the ARP packet is equal to *any* of the host's active interfaces, then this is not a conflicting ARP packet, and should be silently discarded. This is because a user of a multi-homed host with two Ethernet interfaces may connect both interfaces to the same Ethernet hub, in which case the two interfaces will see each other's packets, and if it did not check and realize that the apparently conflicting ARP packets were coming from itself the host could erroneously conclude that all its addresses were in conflict. Another common example is a host with both wired and wireless Ethernet interfaces, in an environment where a wireless gateway is available, but (perhaps unknown to the user) is bridged onto the same wired Ethernet. Expires 22nd December 2001 Cheshire & Aboba [Page 14]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 4. Considerations for Joining of Previously Separate Networks Hosts on disjoint network links may configure the same link-local addresses. If these separate network links are later joined or bridged together, then there may be two hosts which are now on the same link, trying to use the same address. When either host attempts to communicate with any other host on the network, it will at some point broadcast an ARP packet which will enable the hosts in question to detect that there is an address conflict. When these address conflicts are detected, the subsequent forced reconfiguration may be disruptive, causing TCP connections to be broken. However, it is expected that such disruptions will be rare. It should be relatively uncommon for networks to be joined while hosts on those networks are active. Also, 65024 addresses are available for link-local use, so even when two small networks are joined, the chance of collision for any given host is fairly small. When joining two large networks (defined as networks with a substantial number of hosts per segment) there is a greater chance of collision. In such networks, it is likely that the joining of previously separated segments will result in one or more hosts needing to change their IPv4 link-local address, with subsequent loss of TCP connections. In cases where separation and re-joining is frequent, as in remotely bridged networks, this could prove disruptive. However, unless the number of hosts on the joined segments is very large, the traffic resulting from the join and subsequent address conflict resolution will be small. Sending ARP replies that have link-local sender addresses via broadcast instead of unicast ensures that these conflicts can be detected as soon as they become potential problems, but no sooner. For example, if two disjoint network links are joined, where hosts A and B have both configured the same link-local address, X, they can remain in this state until A, B or some other host attempts to initiate communication. If some other host C now sends an ARP request for address X, and hosts A and B were to both reply with conventional unicast ARP replies, then host C might be confused, but A and B still wouldn't know there is a problem because neither would have seen the other's packet. Sending these replies via broadcast allows A and B see each other's conflicting ARP packets and respond accordingly. Note that sending periodic gratuitous ARPs in an attempt to detect these conflicts sooner is not necessary, wastes network bandwidth, and may actually be detrimental. For example, if the network links were joined only briefly, and were separated again before any new communication involving A or B were initiated, then the temporary conflict would have been benign and no forced reconfiguration would have been required. Triggering an unnecessary forced reconfiguration in this case would not serve any useful purpose. Hosts SHOULD NOT send periodic gratuitous ARPs. Expires 22nd December 2001 Cheshire & Aboba [Page 15]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 5. Security Considerations The use of this functionality may open a network host to new attacks. In particular, a host that previously did not have an IP address, and no IP stack running, was not susceptible to IP-based attacks. By configuring a working address, the host may now be vulnerable to IP-based attacks. The ARP protocol [RFC 826] is insecure. A malicious host may send fraudulent ARP packets on the network, interfering with the correct operation of other hosts. For example, it is easy for a host to answer all ARP requests with responses giving its own hardware address, thereby claiming ownership of every address on the network. 6. Acknowledgements We would like to thank (in alphabetical order) Pavani Diwanji, Donald Eastlake 3rd, Peter Ford, Josh Graessley, Erik Guttman, Myron Hattig, Hugh Holbrook, Richard Johnson, Kim Yong-Woon, Satish Mundra, Thomas Narten, Erik Nordmark, Howard Ridenour, Daniel Senie, Dieter Siegmund, Valery Smyslov and Ryan Troll for their contributions. 7. Copyright Copyright (C) The Internet Society 8th March 2000. All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of develop- ing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Expires 22nd December 2001 Cheshire & Aboba [Page 16]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 8. References [IEEE802] IEEE Standards for Local and Metropolitan Area Networks: Overview and Architecture. Institute of Electrical and Electronic Engineers, IEEE Standard 802, 1990. [IEEE1394] Standard for a High Performance Serial Bus. Institute of Electrical and Electronic Engineers, IEEE Standard 1394-1995, 1995. [RFC 826] D. Plummer, "An Ethernet Address Resolution Protocol -or- Converting Network Addresses to 48-bit Ethernet Address for Transmission on Ethernet Hardware", STD 37, RFC 826, November 1982. [RFC 1122] R. Braden, "Requirements for Internet Hosts -- Communication Layers", RFC 1122, October 1989. [RFC 1750] D. Eastlake 3rd, S. Crocker and J. Schiller, "Randomness Recommendations for Security", RFC 1750, December 1994. [RFC 1918] Y. Rekhter et.al., "Address Allocation for Private Internets", RFC 1918, February 1996. [RFC 2119] S. Bradner, "Key words for use in RFCs to Indicate Requirement Levels", RFC 2119, March 1997. [RFC 2131] R. Droms, "Dynamic Host Configuration Protocol", RFC 2131, March 1997. [RFC 2181] R. Elz and R. Bush, "Clarifications to the DNS Specification", RFC 2181, July 1997. [RFC 2462] S. Thomson and T. Narten, "IPv6 Stateless Address Autoconfiguration", RFC 2462, December 1998. [USB] Universal Serial Bus Implementers Forum <http://www.usb.org/> [X10] X10 Ltd. <http://www.x10.com/> Expires 22nd December 2001 Cheshire & Aboba [Page 17]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 9. Authors' Addresses Stuart Cheshire Apple Computer, Inc. 1 Infinite Loop Cupertino California 95014 USA Phone: +1 408 974 3207 EMail: rfc@stuartcheshire.org Bernard Aboba Microsoft Corporation One Microsoft Way Redmond, WA 98052 USA Phone: +1 425 936 6605 Fax: +1 425 936 7329 EMail: bernarda@microsoft.com Appendix. Behavior of Legacy Implementations A.1. Apple Mac OS 8.x and 9.x. Mac OS chooses the IP address on a pseudo-random basis. The selected address is saved in persistent storage for continued use after reboot, when possible. Mac OS sends four DHCPDISCOVER packets, with timeouts of 1, 2, 4, and then 8 seconds. When no response is received from any of these requests (15 seconds), it will autoconfigure. Upon finding that a selected address is in use, Mac OS will select a new random address and try again, at a rate limited to no more than one attempt every two seconds. Autoconfigured Mac OS systems check for the presence of a DHCP server every five minutes. If a DHCP server is found but Mac OS is not successful in obtaining a new lease, it keeps the existing autoconfigured IP address. If Mac OS is successful at obtaining a new lease, it drops all existing connections without warning. This may cause users to lose sessions in progress. Once a new lease is obtained, Mac OS will not allocate further connections using the autoconfigured IP address. Expires 22nd December 2001 Cheshire & Aboba [Page 18]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 Mac OS systems do not send packets addressed to a link-local address to the default gateway if one is present; these addresses are always resolved on the local segment. Mac OS implements media sense where the hardware (and driver software) supports this. As soon as network connectivity is detected, a DHCPDISCOVER will be sent on the interface. This means that systems will immediately transition out of autoconfigured mode as soon as connectivity is restored. A.2. Microsoft Windows 98/98SE Windows 98/98SE systems choose their IP address on a pseudo- random basis. This ensures that systems rebooting will obtain the same autoconfigured address, unless a conflict is detected. The address selection algorithm is based on computing a hash on the interface's MAC address, so that a large collection of hosts should obey the uniform probability distribution in choosing addresses within the 169.254/16 address space. The Windows 98/98SE DHCP Client sends out a total of 4 DHCPDISCOVERs, with an inter-packet interval of 6 seconds. When no response is received after all 4 packets (24 seconds), it will autoconfigure an address. The autoconfigure retry count for Windows 98/98SE systems is 10. After trying 10 autoconfigured IP addresses, and finding all are taken, the host will boot without an IP address. Autoconfigured Windows 98/98SE systems check for the presence of a DHCP server every five minutes. If a DHCP server is found but Windows 98 is not successful in obtaining a new lease, it keeps the existing autoconfigured IP address. If Windows 98/98SE is successful at obtaining a new lease, it drops all existing connections without warning. This may cause users to lose sessions in progress. Once a new lease is obtained, Windows 98/98SE will not allocate further connections using the autoconfigured IP address. Windows 98/98SE systems do not send packets addressed to a link-local address to the default gateway if one is present; these addresses are always resolved on the local segment. Windows 98/98SE systems do not implement media sense. This means that network connectivity issues (such as a loose cable) may prevent a system from contacting the DHCP server, thereby causing it to auto- configure. When the connectivity problem is fixed (such as when the cable is re-connected) the situation will not immediately correct itself. Since the system will not sense the re-connection, it will remain in autoconfigured mode until an attempt is made to reach the DHCP server. Expires 22nd December 2001 Cheshire & Aboba [Page 19]
Internet Draft IPv4 Link-Local Addresses 22nd June 2001 The mini-DHCP server included with Windows 98SE Internet Connection Sharing (ICS), which functions as a NAT, allocates out of the 192.168/16 private address space by default. However, it is possible to change the allocation prefix via a registry key, and no checks are made to prevent allocation out of the link-local prefix. When configured to do so, Windows 98SE ICS will NAT packets from the link-local prefix off the local link. Windows 98SE ICS does not automatically route for the link-local prefix, so that hosts obtaining addresses via DHCP cannot communicate with autoconfigured-only devices. Other home gateways exist that allocate addresses out of the link- local prefix by default. Windows 98/98SE systems can use a 169.254/16 link-local address as the source address when communicating with non-link-local hosts. However, Windows 98/98SE does not support router solicitation/advertisement. This means that Windows 98/98SE systems will typically not be able to discover a default gateway when in autoconfigured mode. A.3. Windows 2000 and Windows ME The autoconfiguration behavior of Windows 2000 and Windows ME systems is identical to Windows 98/98SE except in the following respects: Media Sense Router Discovery Silent RIP Windows 2000 and ME implement media sense. As soon as network connectivity is detected, a DHCPDISCOVER will be sent on the interface. This means that systems will immediately transition out of autoconfigured mode as soon as connectivity is restored. Windows 2000 and ME also support router discovery, although it is turned off by default. Windows 2000 also supports a RIP listener. This means that it is possible to discover a default gateway while in autoconfigured mode. ICS on Windows 2000/ME behaves identically to Windows 98SE with respect to address allocation and NATing of link-local prefixes. Expires 22nd December 2001 Cheshire & Aboba [Page 20]
