Internet Engineering Task Force (IETF) M. Bagnulo Request for Comments: 7219 A. Garcia-Martinez Category: Standards Track UC3M ISSN: 2070-1721 May 2014 SEcure Neighbor Discovery (SEND) Source Address Validation Improvement (SAVI) Abstract This memo specifies SEcure Neighbor Discovery (SEND) Source Address Validation Improvement (SAVI), a mechanism to provide source address validation using the SEND protocol. The proposed mechanism complements ingress filtering techniques to provide a finer granularity on the control of IPv6 source addresses. Status of This Memo 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 5741. Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc7219. Copyright Notice Copyright (c) 2014 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 (http://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. Bagnulo & Garcia-Martinez Standards Track [Page 1] RFC 7219 SEND SAVI May 2014 Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1. Requirements Language . . . . . . . . . . . . . . . . . . 4 2. Background on SEND SAVI . . . . . . . . . . . . . . . . . . . 4 2.1. Address Validation Scope . . . . . . . . . . . . . . . . 4 2.2. Binding Creation for SEND SAVI . . . . . . . . . . . . . 4 2.3. SEND SAVI Protection Perimeter . . . . . . . . . . . . . 7 2.4. Special Cases . . . . . . . . . . . . . . . . . . . . . . 9 3. SEND SAVI Specification . . . . . . . . . . . . . . . . . . . 11 3.1. SEND SAVI Data Structures . . . . . . . . . . . . . . . . 11 3.2. SEND SAVI Device Configuration . . . . . . . . . . . . . 12 3.3. Traffic Processing . . . . . . . . . . . . . . . . . . . 13 3.3.1. Transit Traffic Processing . . . . . . . . . . . . . 13 3.3.2. Local Traffic Processing . . . . . . . . . . . . . . 13 3.4. SEND SAVI Port Configuration Guidelines . . . . . . . . . 27 3.5. VLAN Support . . . . . . . . . . . . . . . . . . . . . . 28 3.6. Protocol Constants . . . . . . . . . . . . . . . . . . . 28 4. Protocol Walk-Through . . . . . . . . . . . . . . . . . . . . 29 4.1. Change of the Attachment Point of a Host . . . . . . . . 29 4.1.1. Moving to a Port of the Same Switch . . . . . . . . . 29 4.1.2. Moving to a Port of a Different Switch . . . . . . . 30 4.2. Attack of a Malicious Host . . . . . . . . . . . . . . . 31 4.2.1. M Attaches to the Same Switch as the Victim's Switch 31 4.2.2. M Attaches to a Different Switch to the Victim's Switch . . . . . . . . . . . . . . . . . . . . . . . 32 5. Security Considerations . . . . . . . . . . . . . . . . . . . 33 5.1. Protection against Replay Attacks . . . . . . . . . . . . 33 5.2. Protection against Denial-of-Service Attacks . . . . . . 34 5.3. Considerations on the Deployment Model for Trust Anchors 36 5.4. Residual Threats . . . . . . . . . . . . . . . . . . . . 36 5.5. Privacy Considerations . . . . . . . . . . . . . . . . . 37 6. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 37 7. References . . . . . . . . . . . . . . . . . . . . . . . . . 37 7.1. Normative References . . . . . . . . . . . . . . . . . . 37 7.2. Informative References . . . . . . . . . . . . . . . . . 38 Bagnulo & Garcia-Martinez Standards Track [Page 2] RFC 7219 SEND SAVI May 2014 1. Introduction This memo specifies SEND SAVI, a mechanism to provide source address validation for IPv6 networks using the SEND protocol [RFC3971]. The proposed mechanism complements ingress filtering techniques to provide a finer granularity on the control of the source addresses used. SEND SAVI uses the DAD_NSOL (Duplicate Address Detection Neighbor SOLicitation) and the DAD_NADV (DAD Neighbor ADVertisement) messages defined in [RFC4862] and the NUD_NSOL (Neighbor Unreachability Detection Neighbor SOLicitation) and NUD_NADV (NUD Neighbor ADVertisement) messages defined in [RFC4861] to validate the address ownership claim of a node. Using the information contained in these messages, host IPv6 addresses are associated to switch ports, so that data packets will be validated by checking for consistency in this binding, as described in [RFC7039]. In addition, SEND SAVI prevents hosts from generating packets containing off-link IPv6 source addresses. Scalability of a distributed SAVI system comprising multiple SEND SAVI devices is preserved by means of a deployment scenario in which SEND SAVI devices form a "protection perimeter". In this deployment scenario, the distributed SAVI system only validates the packets when they ingress to the protection perimeter, not in every SEND SAVI device traversed. The SEND SAVI specification, as defined in this document, is limited to links and prefixes in which every IPv6 host and every IPv6 router uses the SEND protocol [RFC3971] to protect the exchange of Neighbor Discovery information. If the SEND protocol is not used, we can deploy other SAVI solutions relying on monitoring different address configuration mechanisms to prove address ownership. For example, FCFS (First-Come, First-Served) SAVI [RFC6620] can be used by nodes locally configuring IPv6 addresses by means of the Stateless Address Autoconfiguration mechanism [RFC4862]. SEND SAVI is designed to be deployed in SEND networks with as few changes to the deployed implementations as possible. In particular, SEND SAVI does not require any changes in the nodes whose source address is to be verified. This is because verification solely relies in the usage of already available protocols. Therefore, SEND SAVI neither defines a new protocol nor defines any new message on existing protocols, nor does it require that a host or router use an existing protocol message in a different way. An overview of the general framework about Source Address Validation Improvement is presented in [RFC7039]. Bagnulo & Garcia-Martinez Standards Track [Page 3] RFC 7219 SEND SAVI May 2014 1.1. Requirements Language 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 [RFC2119]. 2. Background on SEND SAVI 2.1. Address Validation Scope The application scenario of SEND SAVI is limited to the local link. This means that the goal of SEND SAVI is to verify that the source addresses of the packets generated by the nodes attached to the local link have not been spoofed and that only legitimate routers generate packets with off-link IPv6 source addresses. In a link, there usually are hosts and routers attached. Hosts generate packets with their own addresses as the source address. This is called "local traffic". Routers may send packets containing a source address other than their own, since they can forward packets generated by other hosts (usually located in a different link). This is the so-called transit traffic. SEND SAVI allows the validation of the source address of the local traffic, i.e., it allows verification that the source addresses of the packets generated by the nodes attached to the local link have not been spoofed. SEND SAVI also provides means to prevent hosts from generating packets with source addresses derived from off-link prefixes. However, SEND SAVI does not provide the means to verify if a given router is actually authorized to forward packets containing a particular off-link source address. Other techniques, like ingress filtering [RFC2827], are recommended to validate transit traffic. 2.2. Binding Creation for SEND SAVI SEND SAVI devices filter packets according to bindings between a layer-2 anchor (the binding anchor) and an IPv6 address. These bindings should allow legitimate nodes to use the bounded IPv6 address as source address and prevent illegitimate nodes from doing so. Any SAVI solution is not stronger than the binding anchor it uses. If the binding anchor is easily spoofable (e.g., a Media Access Control (MAC) address), then the resulting solution will be weak. The treatment of non-compliant packets needs to be tuned accordingly. In particular, if the binding anchor is easily spoofable and the SEND SAVI device is configured to drop non-compliant packets, then the usage of SEND SAVI may open a new vector of Denial-of-Service (DoS) Bagnulo & Garcia-Martinez Standards Track [Page 4] RFC 7219 SEND SAVI May 2014 attacks, based on spoofed binding anchors. For that reason, implementations of this specification use switch ports as their binding anchors. Other forms of binding anchors are out of the scope of this specification, and proper analysis of the implications of using them should be performed before their usage. SEND [RFC3971] provides tools to assure that a Neighbor Discovery (ND) message containing a Cryptographically Generated Address (CGA) [RFC3972] option and signed by an RSA option has been generated by the legitimate owner of the CGA IPv6 address. SEND SAVI uses SEND-validated messages to create bindings between the CGA and the port of the SEND SAVI device from which it is reasonable to receive packets with the CGA as the source address. The events that trigger the binding creation process in a SEND SAVI device are: o The reception of a DAD_NSOL message, indicating the attempt of a node to configure an address. This may occur when a node configures an address for the first time or after being idle for some time or when the node has changed the physical attachment point to the layer-2 infrastructure. o The reception of any other packet (including data packets) with a source address for which no binding exists. This may occur if DAD_NSOL messages were lost, a node has changed the physical attachment point to the layer-2 infrastructure without issuing a DAD_NSOL message, a SAVI device loses a binding (for example, due to a restart), or the link topology changed. When the binding creation process is triggered, the SEND SAVI device has to assure that the node for which the binding is to be created is the legitimate owner of the address. For the case in which the binding creation process is initiated by a DAD_NSOL exchange, the SEND SAVI device waits for the reception of a validated DAD_NADV message, indicating that the other node has configured the address before, or validated DAD_NSOL messages arriving from other locations, indicating that another node is trying to configure the same address at the same time. For the case in which packets other than a DAD_NSOL initiate the creation of the binding, the SEND SAVI device explicitly requires the node sending those packets to prove address ownership by issuing a secured NUD_NSOL, which has to be answered with a secured NUD_NADV by the probed node. SEND SAVI devices issue secured NUD_NSOL messages periodically in order to refresh bindings, which have to be answered with a valid NUD_NADV message by the node for which the binding exists. Bagnulo & Garcia-Martinez Standards Track [Page 5] RFC 7219 SEND SAVI May 2014 SEND SAVI devices only forward packets with off-link source addresses if they are received from a port manually configured to connect to a router. SEND SAVI needs to be protected against replay attacks, i.e., attacks in which a secured SEND message is replayed by another node. As discussed before, the SEND SAVI specification uses SEND messages to create a binding between the address contained in the message (that must be signed by a node possessing the private key associated to the address) and the port through which the message is received. If an attacker manages to obtain such a message from another node, for example, because the message was sent to the all-nodes multicast address or because the attacker has subscribed to the Solicited Node multicast address associated to a remote node, it could replay it preserving the original signature. This may create an illegitimate binding in the SEND SAVI device or could be used to abort address configuration at the other node. While SEND provides some means to limit the impact of the replay of ND messages, the emphasis for SEND anti-replay protection is to limit to a short period of time the validity of the ND information transmitted in the message, for example, the relationship between an IPv6 address and a layer-2 address. Note that the period must be long enough to assure that the information sent by the legitimate sender is considered valid despite the possible differences in clock synchronization between the sender and receiver(s). For example, with the values recommended by [RFC3971] for TIMESTAMP_FUZZ and TIMESTAMP_DRIFT, a node receiving a DAD_NSOL message would not discard replays of this message being received within a period of approximately 2 seconds (more precisely, 2/0.99 seconds). The underlying assumption for SEND security is that even if the message is replayed by another node during this period of time, the information disseminated by ND is still the same. However, allowing a node to replay a SEND message does have an impact on the SEND SAVI operation, regardless of the time elapsed since it was generated, since the node can create a new binding in a SEND SAVI device for the port to which an illegitimate node attaches. As can be concluded, the protection provided by SEND is not enough in all cases for SEND SAVI. SEND SAVI increases the protection against the replay attacks compared to SEND. First, each node is required to connect to the SEND SAVI topology through a different port to prevent eavesdropping before entering the SAVI protection perimeter. Then, SEND SAVI bindings are updated only according to messages whose dissemination can be restricted in the SEND SAVI topology without interfering with the normal SEND operation. The messages used by SEND SAVI to create bindings are DAD_NSOL messages, for which SEND SAVI limits its propagation to the ports through which a previous binding for the same IPv6 address existed (see Section 3.3.2), and NUD_NADV messages Bagnulo & Garcia-Martinez Standards Track [Page 6] RFC 7219 SEND SAVI May 2014 in response to a secured NUD_NSOL sent by the SEND SAVI device only through the tested port. Finally, SEND SAVI filtering rules prevent nodes from replaying messages generated by the SEND SAVI devices themselves. Section 5.1 discusses in more detail the protection provided by SEND SAVI against replay attacks. 2.3. SEND SAVI Protection Perimeter In order to reduce computing and state requirements in SEND SAVI devices, SEND SAVI devices can be deployed to form a "protection perimeter" [RFC7039]. With this deployment strategy, SEND SAVI devices perform source-address validation only when packets enter in the protected realm defined through the protection perimeter. The perimeter is defined by appropriate configuration of the roles of each port, which can be 'Validating' or 'Trusted': o Validating ports (VPs) are ports in which SEND SAVI filtering and binding creation are performed. o Trusted ports (TPs) are ports in which limited processing is performed. Only SEND messages related with certificates, prefix information, and DAD operation are processed in order to update the state of the SEND SAVI device or the state related with any of the Validating ports of the switch. Bagnulo & Garcia-Martinez Standards Track [Page 7] RFC 7219 SEND SAVI May 2014 Figure 1 shows a typical topology involving trusted and untrusted infrastructure. +--+ +--+ +--+ +--+ |H1| |H2| |H3| |R1| +--+ +--+ +--+ +--+ | | | | +------------SEND SAVI PROTECTION PERIMETER-----------+ | | | | | | | +-1-----2-+ +-1-----2-+ | | | SEND- | | SEND- | | | | SAVI1 | | SAVI2 | | | +-3--4----+ +--3--4---+ | | | | +--------------+ | | | | | +----------| |--------+ | | | | | SWITCH-A | | | | | +----------| | | | | | | +--------------+ | | | +-1--2----+ +-----1---+ | | | SEND- | | SEND- | | | | SAVI3 | | SAVI4 | | | +-3-----4-+ +----4----+ | | | | | | +------------SEND SAVI PROTECTION PERIMETER-----------+ | | | +--+ +--+ +--+ |R2| |H4| |H5| +--+ +--+ +--+ Figure 1: SAVI Protection Perimeter Trusted ports are used for connections with trusted infrastructures, such as routers and other SEND SAVI devices. Port 2 of SEND-SAVI2 and port 3 of SEND-SAVI3 are Validating ports because they connect to routers. Port 3 of SEND-SAVI1 and port 1 of SEND-SAVI3 as well as port 4 of SEND-SAVI2 and port 1 of SEND-SAVI4 are trusted because they connect two SAVI devices. Finally, port 4 of SEND-SAVI1, port 3 of SEND-SAVI2, and port 2 of SEND-SAVI3 are trusted because they connect to SWITCH-A to which only trusted nodes are connected. Validating ports are used for connection with non-trusted infrastructures; therefore, hosts connect normally to Validating ports. So, in Figure 1 above, ports 1 and 2 of SEND-SAVI1, port 1 of SEND-SAVI2, and port 4 of SEND-SAVI3 are Validating ports because they connect to hosts. Port 4 of SEND-SAVI4 is also a Validating port because it is connected to host H5. Bagnulo & Garcia-Martinez Standards Track [Page 8] RFC 7219 SEND SAVI May 2014 For a more detailed discussion on this, see Section 3.4. 2.4. Special Cases Multi-subnet links: In some cases, a given subnet may have several prefixes. This is supported by SEND SAVI as any port can support multiple prefixes. Multihomed hosts: A multihomed host is a host with multiple interfaces. The interaction between SEND SAVI and multihomed hosts is as follows. If the different interfaces of the host are assigned different IP addresses and packets sent from each interface and always carry the address assigned to that interface as the source address, then from the perspective of a SEND SAVI device, this is equivalent to two hosts with a single interface, each with an IP address. SEND SAVI supports this without additional considerations. If the different interfaces share the same IP address or if the interfaces have different addresses but the host sends packets using the address of one of the interfaces through any of the interfaces, then SEND SAVI does not directly support it. It would require either connecting at least one interface of the multihomed host to a Trusted port or manually configuring the SEND SAVI bindings to allow binding the address of the multihomed host to multiple anchors simultaneously. Virtual switches: A hypervisor or a host operating system may perform bridging functions between virtual hosts running on the same machine. The hypervisor or host OS may in turn connect to a SEND SAVI system. This scenario is depicted in Figure 2, with two virtual machines, VM1 and VM2, connected through a virtual switch, VS1, to SEND SAVI device SEND-SAVI1. The attachment points of VS1 to VM1 and VM2 are configured as Validating. Bagnulo & Garcia-Martinez Standards Track [Page 9] RFC 7219 SEND SAVI May 2014 Host1 +----------------+ | +---+ +---+ | | |VM1| |VM2| | | +---+ +---+ | | | | | | +-1-----2--+ | | | VS1 | | | +--3-------+ | | | | +----|-----------+ | | +--1-----2--+ | SEND- | | SAVI1 | +--3---4----+ | | Figure 2: Virtual Switches Connected to the SEND SAVI Device In order to provide proper security against replay attacks, performing SEND SAVI filtering as close to untrusted hosts as possible (see Sections 3.4 and 5.1) is recommended. In this scenario, this objective can be achieved by enabling SEND SAVI validation in VS1. Ideally, VS1 could be integrated into the SEND SAVI protection perimeter if the hypervisor or host OS at Host1 can be trusted (even though VM1 and VM2 could not be trusted). To do so, both the attachment to SEND-SAVI1 at VS1, and port 1 at SEND-SAVI1, are configured as Trusted. If the administrator of the network does not trust VS1, port 1 of SEND-SAVI1 is configured as Validating, so that every address being used at Host1 is validated at SEND-SAVI1 by SEND SAVI. The attachment point to the physical network at VS1 should be configured as Trusted if the host administrator knows that it is connected to a SEND SAVI device; in this case, VS1 relies on the infrastructure comprised by the physical SEND SAVI devices but not vice versa. Packets egressing from VM1 are validated twice: first at VS1 and then at SEND-SAVI1. Packets going in the reverse direction (from an external host to VM1) are validated once: when they first reach a SEND SAVI device. If the administrator of VS1 does not trust the physical switch to which it attaches, it can configure the attachment to SEND-SAVI1 as Validating. In Figure 2 above, this means that a packet going from another host to VM1 would be validated twice: once when entering the SEND SAVI perimeter formed by the physical devices and again when entering at VS1. Bagnulo & Garcia-Martinez Standards Track [Page 10] RFC 7219 SEND SAVI May 2014 Untrusted routers: One can envision scenarios where routers are dynamically attached to a SEND SAVI network. A typical example would be a mobile phone connecting to a SEND SAVI switch where the mobile phone is acting as a router for other personal devices that are accessing the network through it. Regarding the validation of the source address performed in a SEND SAVI device, such an untrusted router does not seem to directly fall in the category of trusted infrastructure (if this was the case, it is likely that all devices would be trusted); hence, it cannot be connected to a Trusted port, and if it is connected to a Validating port, the SEND SAVI switch would discard all the packets containing an off-link source address coming from that device. Although the SEND SAVI device to which this router attaches could be configured to permit the transit of packets with source addresses belonging to the set of prefixes reachable through the untrusted router, such a mechanism is out of the scope of this document. As a result, the default mechanism described in this specification cannot be applied in such a scenario. 3. SEND SAVI Specification 3.1. SEND SAVI Data Structures The following three data structures are defined for SEND SAVI operations. SEND SAVI Database: The SEND SAVI function relies on state information binding the source IPv6 address used in data packets to the port through which the legitimate node connects. Such information is stored in the SEND SAVI Database. The SEND SAVI Database is populated with the contents of validated SEND messages. Each entry contains the following information: o IPv6 source address o Binding anchor: the port through which the packet was received o Lifetime o Status: TENTATIVE_DAD, TENTATIVE_NUD, VALID, TESTING_VP, TESTING_VP' o Alternative binding anchor: the port from which a DAD_NSOL message or any data packet has been received while a different port was stored in the binding anchor for the address. o Creation time: the value of the local clock when the entry was first created Bagnulo & Garcia-Martinez Standards Track [Page 11] RFC 7219 SEND SAVI May 2014 SEND SAVI Prefix List: SEND SAVI devices need to know which ones are the link prefixes in order to identify local and off-link traffic. A SEND SAVI device MUST support discovering this information from the Prefix Information option [RFC4861] with the L bit set of Router Advertisement (RADV) messages coming from Trusted ports, as described in Section 3.3.2. The list of prefixes MAY also be configured manually. This information is not specific to a given port. The SEND SAVI Prefix List contains one entry per prefix in use, as follows: o Prefix: the prefix included in a Prefix Information option. o Prefix lifetime: time in seconds that the prefix is valid. Initially set to the Valid Lifetime value of the Prefix Information option of a valid RADV message or set to a value of all 1 bits (0xffffffff), which represents infinity, if configured manually. When the SEND SAVI device boots, it MUST send a Router Solicitation (RSOL) message, which does not need to be secured if the unspecified address is used (see [RFC3971], Sections 5.1.1 and 5.2.1). The SAVI device SHOULD issue a RSOL message in case the prefix entry is about to expire. 3.2. SEND SAVI Device Configuration In order to perform the SEND SAVI operation, some basic parameters of the SEND SAVI device have to be configured. Since a SEND SAVI device operates as a SEND node to generate NUD_NSOL, RSOL, or Certification Path Solicitation (CPS) messages: o The SEND SAVI device MUST be configured with a valid CGA address. When the SEND SAVI device configures this address, it MUST behave as a regular SEND node, i.e., using secured NSOL messages to perform DAD, etc., in addition to fulfilling the requirements stated for regular IPv6 nodes [RFC6434]. o The SEND SAVI device MAY be configured with at least one trust anchor if it is configured to validate RADV messages (see Section 3.3.2). In this case, the SEND SAVI device MAY be configured with certification paths. The alternative is obtaining them by means of issuing Certification Path Solicitation messages, as detailed in the SEND specification [RFC3971]. In addition, the port role for each port of the SEND SAVI device MUST be configured. The guidelines for this configuration are specified in Section 3.4. Bagnulo & Garcia-Martinez Standards Track [Page 12] RFC 7219 SEND SAVI May 2014 3.3. Traffic Processing In this section, we describe how packets are processed by a SEND SAVI device. Behavior varies depending on if the packet belongs to local or transit traffic. This is determined by checking if the prefix of the source address is included in the SEND SAVI Prefix List or in the unspecified address (local traffic) or not included in the SEND SAVI Prefix List (transit traffic). 3.3.1. Transit Traffic Processing Transit traffic processing occurs as follows: o If the SEND SAVI device receives a transit traffic packet through a Trusted port, it forwards it without any SAVI processing. o If the SEND SAVI device receives a transit traffic packet through a Validating port, it discards the packet. 3.3.2. Local Traffic Processing If the verification of the source address of a packet shows that it belongs to local traffic, this packet is processed using the state machine described in this section. For the rest of the section, the following assumptions hold: o When it is stated that a secured NUD_NSOL message is issued by a SEND SAVI device through a port P, it means that the SEND SAVI device generates a NUD_NSOL message, according to the Neighbor Unreachability Detection procedure described in [RFC4861], addressed to the IPv6 target address, which is the source address of the packet triggering the procedure. This message is secured by SEND as defined in [RFC3971]. The source address used for issuing the NUD_NSOL message is the source address of the SEND SAVI device. The message is sent only through port P. o When it is stated that a validated NUD_NADV message is received by a SEND SAVI device, it means that a SEND secured NUD_NADV message has been received by the same port P through which the corresponding NUD_NSOL message was issued, and the NUD_NADV message has been validated according to [RFC3971] to prove ownership for the IPv6 address under consideration and to prove that it is a response for the previous NUD_NSOL message issued by the SEND SAVI device (containing the same nonce value as the NUD_NSOL message to which it answers). Bagnulo & Garcia-Martinez Standards Track [Page 13] RFC 7219 SEND SAVI May 2014 We use VP to refer to a Validating port and TP to refer to a Trusted port. The state machine is defined for a binding of a given source IPv6 address in a given SEND SAVI device. In the transitions considered, packets described as inputs refer to the IPaddr IPv6 address associated to the state machine. The possible states for a given IPaddr are NO_BIND, TENTATIVE_DAD, TENTATIVE_NUD, VALID, TESTING_VP, and TESTING_VP'. The NO_BIND state represents that no binding exists for IPaddr; this is the state for all addresses unless a binding is explicitly created. The states can be classified into 'forwarding' states, i.e., states in which packets received from the port associated to the IPv6 address are forwarded, and 'non-forwarding' states, i.e., states in which packets different to the ones used for signaling are not forwarded. VALID, TENTATIVE_DAD, TESTING_VP, and TESTING_VP' are forwarding states, and NO_BIND and TENTATIVE_NUD are non-forwarding states. The SEND SAVI device MUST join the Solicited Node Multicast group for all the addresses whose state is other than NO_BIND. This is needed to make sure that the SEND SAVI device receives DAD_NSOL messages issued for those addresses. Note that it may not be enough to relay on the Multicast Listener Discovery (MLD) messages being sent by the node attached to a Validating port for which a binding for the corresponding address exists, since the node may move and packets sent to that particular Solicited Node Multicast group may no longer be forwarded to the SEND SAVI device. In order to determine which traffic is on-link and off-link, the SEND SAVI device MUST support discovery of this information from the Prefix Information option with the L bit set of RADV messages. In this case, at least one router SHOULD be configured to advertise RADV messages containing a Prefix Information option with the prefixes that the untrusted nodes can use as source addresses, and the bit L set. An alternative to this is to manually configure the SEND SAVI Prefix List or restrict the use of link-local addresses. SEND SAVI devices MUST discard RADV messages received from Validating ports. RADV messages are only accepted and processed when received through Trusted ports. SEND SAVI devices SHOULD NOT validate RADV messages to update the SEND SAVI Prefix List and forward them to other nodes. These messages can only be received from Trusted ports, and we assume that routers are trusted. Validating RADV messages would be required in Bagnulo & Garcia-Martinez Standards Track [Page 14] RFC 7219 SEND SAVI May 2014 any SEND SAVI device the node is traversing. Besides, hosts will validate this message before using the information it contains. In case SEND SAVI devices are configured to validate RADV messages, SEND SAVI devices SHOULD support the processing of validated Certification Path Advertisement (CPA) messages, sent in reply to CPS messages, to acquire certificates used to validate router messages; alternatively, it SHOULD be configured with a certification path. The state machine defined for the SEND SAVI operation adheres to the following design guidelines: o The only events that trigger state changes from forwarding to non- forwarding states, and vice versa, are the reception of DAD_NSOL, DAD_NADV, and NUD_NADV or the expiration of a timer. The other possible input to consider is 'any other packet', which could generate changes to states belonging to the same forwarding or non-forwarding class as the original state. In other words, when 'any other packet' is received, the state cannot move from forwarding to non-forwarding, and vice versa. The reduced set of messages being able to trigger a change simplifies the processing at SEND SAVI devices. o DAD_NADV and NUD_NADV are only processed when they are a response to a DAD_NSOL or a NUD_NSOL message. o SEND SAVI devices MUST only use ND messages received through Validating ports if they are valid; otherwise, they discard them. SEND SAVI devices SHOULD assume that such messages received from Trusted ports have been validated by other SEND SAVI devices, or come from a trusted device such a router, so they SHOULD NOT attempt to validate them in order to reduce the processing load at the SEND SAVI device. o The only messages the SEND SAVI device is required to generate specifically per each source IP address are MLD and NUD_NSOL messages. This also keeps the state machine simple. o Well-behaved nodes are expected to initiate communication by sending secured DAD_NSOL messages. The SEND SAVI state machine is tailored to efficiently process these events. The reception of other packet types without receiving previously validated DAD_NSOL messages is assumed to be a consequence of bad-behaving nodes or infrequent events (such as packet loss, a change in the topology connecting the switches, etc.). While a binding will ultimately be created for nodes affected by such events, simplicity of the state machine is prioritized over any possible optimization for these cases. Bagnulo & Garcia-Martinez Standards Track [Page 15] RFC 7219 SEND SAVI May 2014 o If a node has a configured address, and it can prove that it owns this address, the binding is preserved regardless of any indication that a binding for the same source address could be configured in other SEND SAVI devices. Bindings for the same source address in two or more SEND SAVI devices may occur due to several reasons, for example, when a host moves (the two bindings exist just for a short period of time) or when many nodes generate the same address and the DAD procedure has failed. In these infrequent cases, SEND SAVI preserves connectivity for the resulting bindings. Next, we describe how different inputs are processed, depending on the state of the binding of the IP address 'IPaddr'. Note that every ND message is assumed to be validated according to the SEND specification. To facilitate the reader's understanding of the most relevant transitions of the SEND SAVI state machine, a simplified version, which does not contain every possible transition, is depicted in Figure 3: Bagnulo & Garcia-Martinez Standards Track [Page 16] RFC 7219 SEND SAVI May 2014 +-------------+ | | | TESTING_VP' | | | +-------------+ Timeout/VP=VP' | ^ | | VP_NUD_NADV/- | | VP'_DAD_NSOL/ | | VP_NUD_NSOL | | v | VP_DAD_NSOL/- +--------+ +------------- | | | | VALID |< -------------------+ | +-------- >| | | | | +--------+ | | | ^ | | | | VP_NUD_ | | Timeout, | | | NADV/- | | TP_DAD_NSOL/VP_NUD_NSOL| | | | v | | | +------------+ | | | | | | | | | TESTING_VP | | | | | | | | | +------------+ | | | | | | | | Timeout/- | | | VP*, | | | | Timeout/- | VP_NUD_NADV/- | v | | | +---------------+ | +---------------+ | | | | | | TENTATIVE_DAD | | | TENTATIVE_NUD | | | | | | +---------------+ | +---------------+ ^ | | | ^ | | | Timeout/- | | | | TP_DAD_NSOL, | | | | | TP_DAD_NADV/- | | | | | v | | | | +---------+ | | | +--------- >| |< -----+ | | | NO_BIND | | +--------------| |-----------------+ VP_DAD_NSOL/- +---------+ VP*/VP_NUD_NSOL Figure 3: Simplified SEND SAVI State Machine Bagnulo & Garcia-Martinez Standards Track [Page 17] RFC 7219 SEND SAVI May 2014 Each state transition is characterized by any of the events that may trigger the change and the message(s) generated as a result of this change. The meaning of some terms are referred next: o VP_DAD_NSOL as a triggering event means that a validated DAD_NSOL message has been received from the current BINDING_ANCHOR port VP. o VP* means any packet (data packet) received from the current BINDING_ANCHOR port VP. o TP_DAD_NSOL as a triggering event means that a DAD_NSOL message was received from a Trusted port. o - means that no message is sent. VP=VP' means that the BINDING_ANCHOR is set to VP'. The notation Timeout, TP_DAD_NSOL/VP_NUD_NSOL means that the transition is triggered by either a timeout expiration or the reception of a DAD_NSOL message from a Trusted port, and in addition to the transition, a NUD_NSOL message is sent through port VP. For the rest of the description, we assume the following: o When a validated message is required (i.e., a 'validated DAD_NSOL'), messages are check for validity in the considered switch according to [RFC3971], and messages not fulfilling these conditions are discarded. o When any SEND message is received from a validated port, the SEND SAVI SHOULD assume that the message has been validated by the SEND SAVI device through which the message accessed the SEND SAVI protection perimeter (unless the SEND SAVI perimeter has been breached), or the device generating it is trusted. In this case, the SAVI device does not perform any further validation. Performing validation for SEND messages received through a Trusted port may affect performance negatively. Bagnulo & Garcia-Martinez Standards Track [Page 18] RFC 7219 SEND SAVI May 2014 NO_BIND When the node is in this state, there are no unresolved NUD_NSOL messages generated by SEND SAVI or DAD_NSOL propagated to any Validating port, so the only relevant inputs are DAD_NSOL messages coming either from a Validating port (VP) or Trusted port (TP), or any packet other than DAD_NSOL coming from a VP or TP. There are no timers configured for this state. Messages received from a Validating port: o If a validated DAD_NSOL message is received from a Validating port VP, the SEND SAVI device forwards this message to all appropriate Trusted ports (the subset of Trusted ports that belong to the forwarding layer-2 topology, with the restrictions imposed by the MLD snooping mechanism, if applied). DAD_NSOL messages are not sent through any of the ports configured as Validating ports. The SEND SAVI device sets the LIFETIME to TENT_LT, stores all the information required for future validation of the corresponding DAD_NADV message (such as the nonce of the message), creates a new entry in the SEND SAVI Database for IPaddr, sets BINDING_ANCHOR to VP, and changes the state to TENTATIVE_DAD. Creation time is set to the current value of the local clock. Note that in this case, it is not possible to check address ownership by sending a NUD_NSOL because while the node is waiting for a possible DAD_NADV, its address is in tentative state and the node cannot respond to NSOL messages [RFC4862]. o If any packet other than a DAD_NSOL is received through a Validating port VP, the SEND SAVI device issues a secured NUD_NSOL through port VP. The SEND SAVI device sets the LIFETIME to TENT_LT. The SEND SAVI device creates a new entry in the SEND SAVI Database for IPaddr, sets BINDING_ANCHOR to VP, and the state is changed to TENTATIVE_NUD. Creation time is set to the current value of the local clock. The SAVI device MAY discard the packet while the NUD procedure is being executed or MAY store it in order to send it if the next transitions are (strictly) TENTATIVE_NUD and then VALID. Messages received from a Trusted port: o If a DAD_NSOL message containing IPaddr as the target address is received through a Trusted port, it MUST NOT be forwarded through any of the Validating ports: it is sent through the proper Trusted ports. The state is not changed. Bagnulo & Garcia-Martinez Standards Track [Page 19] RFC 7219 SEND SAVI May 2014 o Any packet other than a DAD_NSOL received from a Trusted port is forwarded to its destination. This packet is assumed to come from a SEND SAVI device that has securely validated the binding, according to the SEND SAVI rules (unless the SEND SAVI perimeter has been breached). The state is not changed. TENTATIVE_DAD To arrive at this state, the SEND SAVI device has received a validated DAD_NSOL coming from the BINDING_ANCHOR port, and it has forwarded it to the appropriate TPs. The relevant events occurring in this state are the reception of a DAD_NADV message from a TP, a DAD_NSOL message from the BINDING_ANCHOR port, other Validating port or TP, a data packet from the BINDING_ANCHOR port, and the expiration of the LIFETIME timer initiated when the DAD_NSOL was received at the BINDING_ANCHOR port. Messages received from a Trusted port: o The reception of a valid DAD_NADV message from a Trusted port indicates that the binding cannot be configured for the BINDING_ANCHOR port. The state is changed to NO_BIND, and the LIFETIME is cleared. o The reception of a valid DAD_NSOL from a Trusted port indicates that a node connected to another SEND SAVI device may be trying to configure the same address at the same time. The DAD_NSOL message is forwarded to the BINDING_ANCHOR port, so that the node at this port will not configure the address, as stated in [RFC4862]. The DAD_NSOL message is also forwarded to all appropriate Trusted ports. Then, the LIFETIME is cleared, and the state is changed to NO_BIND. o Any packet other than a validated DAD_NSOL or DAD_NADV received from a Trusted port is forwarded to its destination. This packet is assumed to come from a SEND SAVI device that has securely validated the binding, according to the SEND SAVI rules (unless the SEND SAVI perimeter has been breached). The state is not changed. Bagnulo & Garcia-Martinez Standards Track [Page 20] RFC 7219 SEND SAVI May 2014 Messages received from a Validating port different from the BINDING_ANCHOR: o A validated DAD_NSOL is received from a Validating port VP' different from the BINDING_ANCHOR port. The reception of a valid DAD_NSOL from port VP' indicates that a node connected to VP' may be trying to configure the same address at the same time. The DAD_NSOL message is forwarded to the BINDING_ANCHOR port, so that the node at this port will not configure the address, as stated in [RFC4862]. The DAD_NSOL message is also forwarded to all appropriate Trusted ports. Then, the BINDING_ANCHOR is set to VP' (through which the DAD_NSOL message was received), the LIFETIME is set to TENT_LT, and the state remains in TENTATIVE_DAD. o Any packet other than a validated DAD_NSOL received from a Validating port VP' different from the BINDING_ANCHOR port is discarded. The state is not changed. Messages received from the BINDING_ANCHOR port: o If a validated DAD_NSOL is received from the BINDING_ANCHOR port, the LIFETIME is set to TENT_LT, and the state remains in TENTATIVE_DAD. o If any packet other than a DAD_NSOL is received from the BINDING_ANCHOR port, it is assumed that the node has configured its address, although it has done it in less time than expected by the SEND SAVI device (less than TENT_LT). Since the node proved address ownership by means of the validated DAD_NSOL message, the LIFETIME is set to DEFAULT_LT, and the state is changed to VALID. LIFETIME expires: o If LIFETIME expires, it is assumed that no other node has configured this address. Therefore, the Validating port VP (currently stored in the BINDING_ANCHOR) could be bound to this IPv6 address. The LIFETIME is set to DEFAULT_LT, and the state is changed to VALID. VALID To arrive at this state, the SEND SAVI device has successfully validated address ownership and has created a binding for IPaddr. Relevant transitions for this state are triggered by the reception of DAD_NSOL from the BINDING_ANCHOR port, other Validating port or a TP, and any packet other than DAD_NSOL from a Validating port other than Bagnulo & Garcia-Martinez Standards Track [Page 21] RFC 7219 SEND SAVI May 2014 the BINDING_ANCHOR or a TP. The expiration of LIFETIME is also relevant to trigger a check for address ownership for the node at the BINDING_ANCHOR port. Messages received from the BINDING_ANCHOR port: o If a validated DAD_NSOL with IPaddr as a source address is received through the BINDING_ANCHOR port, it is forwarded to the appropriate Trusted ports. The LIFETIME is set to TENT_LT, and the state is changed to TENTATIVE_DAD. o Any packet other than a DAD_NSOL containing IPaddr as a source address arriving from the BINDING_ANCHOR port is forwarded appropriately. The state is not changed. Messages received from a Trusted port: o If a DAD_NSOL with IPaddr as a source address is received through a Trusted port, the message is forwarded to VP. The LIFETIME is set to TENT_LT, a secured NUD_NSOL message is sent to IPaddr through VP, and the state is changed to TESTING_VP. o If any packet other than a DAD_NSOL with IPaddr as a source address is received through a Trusted port, the packet is forwarded to VP and to other appropriate Trusted ports. A secured NUD_NSOL is sent to the BINDING_ANCHOR port, the LIFETIME is set to TENT_LT, and the state is changed to TESTING_VP. Messages received from a Validating port different from the BINDING_ANCHOR: o If a validated DAD_NSOL packet with IPaddr as a source address is received through a Validating port VP' (a VP' different from the current BINDING ANCHOR), the message is forwarded to the BINDING_ANCHOR port. In addition, a secured NUD_NSOL is sent to the BINDING_ANCHOR port, the ALTERNATIVE BINDING ANCHOR is set to port VP' (for future use if the node at VP' is finally selected), the LIFETIME is set to TENT_LT, and the state is changed to TESTING_VP'. o If any packet other than a DAD_NSOL with IPaddr as a source address is received from a Validating port VP', different from the current BINDING_ANCHOR for this binding, VP, the packet is discarded. The SEND SAVI device MAY issue a secured NUD_NSOL through the BINDING_ANCHOR port, store VP' in the ALTERNATIVE BINDING ANCHOR for possible future use, set the LIFETIME to TENT_LT, and change the state to TESTING_VP'. An alternative to this behavior is that the SEND SAVI device MAY not do anything (in Bagnulo & Garcia-Martinez Standards Track [Page 22] RFC 7219 SEND SAVI May 2014 this case, the state would eventually change after a maximum DEFAULT_LT time; if the node at VP does not respond to a NUD_NSOL at TESTING_VP, the state is moved to NO_BIND). Then, a packet arriving from VP' would trigger a process that may end up with binding for the node connecting to VP'. LIFETIME expires: o If LIFETIME expires, a secured NUD_NSOL message is sent through the BINDING_ANCHOR port to IPaddr, the LIFETIME is set to TENT_LT, and the state is changed to TESTING_VP. In the TESTING_VP state, packets are still being forwarded until the timer expires without receiving a NUD_NADV. TESTING_VP When the SEND SAVI device enters the TESTING_VP state, the current Validating port is under check through a secured NUD_NSOL message generated by the SEND SAVI device. While testing, packets from the current Validating port are forwarded. Packets coming from Trusted ports are also forwarded. The relevant events for this state are the reception of a NUD_NADV message from VP; the reception of a DAD_NSOL message from VP, VP', or TP; the reception of any packet other than the previous cases from VP, VP', or TP; and the expiration of the timer associated to the reception of NUD_NADV. Messages received from the BINDING_ANCHOR port: o If a validated NUD_NADV is received from VP, the LIFETIME is changed to DEFAULT_LT, and the state is changed to VALID. The message is not forwarded to any other port. o If a validated DAD_NSOL message is received from VP, it is forwarded to the appropriate Trusted ports, the LIFETIME is set to DEFAULT_LT, and the state is changed to TENTATIVE_DAD. o Any packet other than DAD_NSOL or NUD_NADV containing IPaddr as a source address arriving from the BINDING_ANCHOR port is forwarded. Neither the LIFETIME nor the state are changed. Messages received from a Trusted port: o If a DAD_NSOL packet is received from a Trusted port, the message is forwarded to VP and the appropriate Trusted ports. Neither the LIFETIME nor the state are changed. The node at the BINDING_ANCHOR port is under check; if it still is at this port, it should answer with a NUD_NADV and also with a DAD_NADV. If it is not there, neither the NUD_NADV nor the DAD_NADV will be Bagnulo & Garcia-Martinez Standards Track [Page 23] RFC 7219 SEND SAVI May 2014 received, the timer will expire, and the local state will move to NO_BIND. o If a packet other than a DAD_NSOL arrives from a Trusted port, the packet is forwarded. Neither the LIFETIME nor the state are changed. Messages received from a Validating port different from the BINDING_ANCHOR: o If a valid DAD_NSOL is received from a Validating port VP' other than the current BINDING_ANCHOR port, the message is forwarded to the BINDING_ANCHOR port and to the appropriate Trusted ports. In addition, a secured NUD_NSOL is sent to the BINDING_ANCHOR port, the ALTERNATIVE BINDING ANCHOR is set to VP' (for future use if the node at VP' is finally selected), the LIFETIME is set to TENT_LT, and the state is changed to TESTING_VP'. o Any other packet received from a Validating port VP' other than the BINDING_ANCHOR port is discarded. This may occur because the node has moved but has not issued a DAD_NSOL or the DAD_NSOL message has been lost. The state will eventually move to NO_BIND, and then the packets sent from VP' will trigger the creation of the binding for VP'. LIFETIME expires: o If the LIFETIME expires, the LIFETIME is cleared and the state is changed to NO_BIND. TESTING_VP' To arrive at this state, the SEND SAVI device has received an indication that a node at VP' different from the BINDING_ANCHOR port wants to send data with IPaddr as a source address and has occurred while a binding existed for VP. The port VP' that triggered the change of the state to TESTING_VP' was stored at the ALTERNATIVE_BINDING_ANCHOR, so that it can be retrieved if the node at VP' is determined as the legitimate owner of IPaddr. The SEND SAVI device has issued a NUD_NSOL to IPaddr through the BINDING_ANCHOR port. The relevant events that may occur in this case are the reception of a NUD_NADV from port VP (the BINDING_ANCHOR port); the reception of a DAD_NSOL from VP, VP', TP, and VP" (VP" different from VP and VP'); the reception of any other packet from VP, VP', TP, or VP"; and the expiration of the timer. Bagnulo & Garcia-Martinez Standards Track [Page 24] RFC 7219 SEND SAVI May 2014 Messages received from the BINDING_ANCHOR port: o A validated NUD_NADV is received from the BINDING_ANCHOR port. The reception of a valid NUD_NADV indicates that the node at VP is defending its address. The BINDING_ANCHOR in use is kept, the LIFETIME is set to DEFAULT_LT, and the state is changed to VALID. o If a valid DAD_NSOL is received from the BINDING_ANCHOR port, it is forwarded to VP' (the port stored in the ALTERNATIVE_BINDING_ANCHOR). The BINDING_ANCHOR in use is kept, the LIFETIME is set to TENT_LT, and the state is changed to TENTATIVE_DAD. When the DAD_NSOL message is received by the node at VP', the address will not be configured. o Any packet other than a validated DAD_NSOL, or a validated NUD_NADV coming from the BINDING_ANCHOR port, is forwarded, and the state is not changed. Messages received from the ALTERNATIVE_BINDING_ANCHOR Validating port: o If a valid DAD_NSOL is received from the port stored in the ALTERNATIVE_BINDING_ANCHOR, it is forwarded to the BINDING_ANCHOR port. The BINDING_ANCHOR and the ALTERNATIVE BINDING ANCHOR are kept, the LIFETIME is set to DEFAULT_LT, and the state is not changed. o Any packet other than a validated DAD_NSOL coming from the ALTERNATIVE_BINDING_ANCHOR port is discarded, and the state is not changed. Messages received from a Validating port different from the BINDING_ANCHOR and the ALTERNATIVE_BINDING_ANCHOR ports: o If a validated DAD_NSOL is received from port VP", different from BINDING_ANCHOR and the ALTERNATIVE_BINDING_ANCHOR ports, it is forwarded to the BINDING_ANCHOR and the ALTERNATIVE_BINDING_ANCHOR ports. The node at the ALTERNATIVE BINDING ANCHOR port is expected to unconfigure its address if the message triggering the transition to this state was a DAD_NSOL message received from the ALTERNATIVE_BINDING_ANCHOR port (and not any other packet). The state remains in TESTING_VP', although VP" is stored in the ALTERNATIVE_BINDING_ANCHOR for future use if the node at VP" is finally selected. The LIFETIME is not changed. o Any packet other than a validated DAD_NSOL received from port VP" is discarded and does not affect the state. Bagnulo & Garcia-Martinez Standards Track [Page 25] RFC 7219 SEND SAVI May 2014 Messages received from a Trusted port: o If a DAD_NSOL is received from a Trusted port, the message is forwarded to the BINDING_ANCHOR, ALTERNATIVE_BINDING_ANCHOR ports, and other appropriate Trusted ports. The LIFETIME is left unchanged, and the state is changed to TESTING_VP. The node at the ALTERNATIVE_BINDING_ANCHOR port is expected to unconfigure its address if the packet triggering the transition to this state was a DAD_NSOL message received from the ALTERNATIVE_BINDING_ANCHOR port. o Any packet other than a DAD_NSOL coming from a Trusted port is forwarded appropriately, but the state is not changed. LIFETIME expires: o If LIFETIME expires, it is assumed that the node for which the binding existed is no longer connected through the BINDING_ANCHOR port. Therefore, the BINDING_ANCHOR is set to the ALTERNATIVE_BINDING_ANCHOR port value. The LIFETIME is set to DEFAULT_LT, and the state is changed to VALID. TENTATIVE_NUD To arrive at this state, a data packet has been received through the BINDING_ANCHOR port without any existing binding in the SEND SAVI device. The SEND SAVI device has sent a NUD_NSOL message to the BINDING_ANCHOR port. The relevant events for this case are the reception of a NUD_NADV from the BINDING_ANCHOR port; the reception of a DAD_NSOL from the BINDING_ANCHOR port, other VP different from the BINDING_ANCHOR port, or a TP; and the reception of any packet other than a DAD_NSOL and a NUD_NADV from the BINDING_ANCHOR port and a DAD_NSOL for other VP different from the BINDING_ANCHOR port, or TP. In addition, the LIFETIME may expire. Messages received from the BINDING_ANCHOR port: o If a validated NUD_NADV message is received through the BINDING_ANCHOR port, the LIFETIME is set to TENT_LT, and the state is changed to VALID. The message is not forwarded to any port. o If a validated DAD_NSOL message is received through the BINDING_ANCHOR port, it is forwarded to the appropriate Trusted ports, the LIFETIME is set to TENT_LT, and the state is changed to TENTATIVE_DAD. o Any packet other than NUD_NADV or DAD_NSOL received through the BINDING_ANCHOR port is discarded. Bagnulo & Garcia-Martinez Standards Track [Page 26] RFC 7219 SEND SAVI May 2014 Messages received from a Validating port different from the BINDING_ANCHOR: o If a validated DAD_NSOL message is received through port VP' different from the BINDING_ANCHOR port, it is forwarded to the appropriate Trusted ports, the LIFETIME is set to TENT_LT, the BINDING_ANCHOR is set to VP', and the state is changed to TENTATIVE_DAD. o Any packet other than validated DAD_NSOL received through port VP' MUST NOT be forwarded unless the next state for the binding is VALID. The packets received MAY be discarded or MAY be stored to be sent if the state changes later to VALID. The state is left unchanged. Messages received from a Trusted port: o If a DAD_NSOL message is received through a Trusted port, it is forwarded to the BINDING_ANCHOR port, and the state is left unchanged. o Any other packet received from a Trusted port is forwarded appropriately. This packet may come from a SEND SAVI device that has securely validated the attachment of the node to its Validating port, according to SEND SAVI rules. The state is left unchanged. LIFETIME expires: o If LIFETIME expires, the LIFETIME is cleared and the state is changed to NO_BIND. 3.4. SEND SAVI Port Configuration Guidelines The detailed guidelines for port configuration in SEND SAVI devices are: o Ports connected to another SEND SAVI device MUST be configured as Trusted ports. Not doing so will prevent off-link traffic from being forwarded, along with the following effects for on-link traffic: significantly increase the CPU time, memory consumption, and signaling traffic due to SEND SAVI validation, in both the SEND SAVI devices and the node whose address is being validated. o Ports connected to hosts SHOULD be configured as Validating ports. Not doing so will allow the host connected to that port to send packets with a spoofed source address. Bagnulo & Garcia-Martinez Standards Track [Page 27] RFC 7219 SEND SAVI May 2014 o No more than one host SHOULD be connected to each port. Connecting more than one host to a port will allow hosts to generate packets with the same source address as the other hosts connected to the same port, and will allow replaying attacks to be performed as described in Section 5.1. o Ports connected to routers MUST be configured as Trusted ports. Not doing so results in SEND SAVI devices discarding off-link traffic. Note that this means that since routers are connected through Trusted ports, they can generate traffic with any source address, even those belonging to the link. o Ports connected to a chain of one or more legacy switches that have other SEND SAVI devices but have no routers or hosts attached to them SHOULD be configured as Trusted ports. Not doing so will significantly increase the memory consumption in the SEND SAVI devices and increase the signaling traffic due to SEND SAVI validation. 3.5. VLAN Support In the case where the SEND SAVI device is a switch that supports customer VLANs [IEEE.802-1Q.2005], the SEND SAVI specification MUST behave as if there was one SEND SAVI process per customer VLAN. The SEND SAVI process of each customer VLAN will store the binding information corresponding to the nodes attached to that particular customer VLAN. 3.6. Protocol Constants TENT_LT is 500 milliseconds. DEFAULT_LT is 5 minutes. Bagnulo & Garcia-Martinez Standards Track [Page 28] RFC 7219 SEND SAVI May 2014 4. Protocol Walk-Through In this section, we include two cases that illustrate the behavior of SEND SAVI, the change of the attachment port of a host, and the attack of a malicious host. We use the topology depicted in Figure 4. +---+ | H | +---+ | | +-1-----2-+ +-1-----2-+ | | | | | SAVI1 | | SAVI2 | | | | | +-3-----4-+ +-3-----4-+ | | ------------------- Figure 4: Reference SEND SAVI Topology for Protocol Walk-Through 4.1. Change of the Attachment Point of a Host There are two cases, depending on whether the host H moves to a different port on the same switch or to a different switch. 4.1.1. Moving to a Port of the Same Switch Host H is connected to port 1 of SAVI1 and moves to port 2 of the same switch. Before moving, the SEND SAVI state associated to IPH, the IP address of H, is: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=NO_BIND In the general case, H issues a DAD_NSOL message for IPH when it is connected to a different port. When SAVI1 receives this message, it validates it and changes its state to: SAVI1=TESTING_VP', BINDING_ANCHOR=1, ALTERNATIVE_BINDING_ANCHOR=2, TIMER=TENT_LT / SAVI2=NO_BIND The DAD_NSOL message is propagated to port 1, because it is the current BINDING_ANCHOR, and the Trusted port 3; it is not propagated to Validating port 4. SAVI1 configures a timer for TENT_LT seconds. In addition, SAVI1 generates a NUD_NSOL and sends it through port 1. When SAVI2 receives this message through its Trusted port, it discards it and remains in the NO_BIND state. Bagnulo & Garcia-Martinez Standards Track [Page 29] RFC 7219 SEND SAVI May 2014 SAVI1 waits for a NUD_NADV message to be received from port 1. Since there is no node attached to 1, there is no response for either of these messages. When TENT_LT expires at SAVI1, the state changes to: SAVI1=VALID, BINDING_ANCHOR=2 / SAVI2=NO_BIND If the node moving does not issue a DAD_NSOL when it attaches to port 2, then SAVI1 will receive a data packet through this port. The data packet is discarded, SAVI1 issues a secured NUD_NSOL through port 1, and the state changes to TESTING_VP'. SAVI1=TESTING_VP', BINDING_ANCHOR=1, ALTERNATIVE_BINDING_ANCHOR=2 TIMER=TENT_LT / SAVI2=NO_BIND SAVI1 waits for a NUD_NADV message to be received from port 1. Since there is no node attached to 1, there is no response for neither of these messages. When TENT_LT expires at SAVI1, the state changes to: SAVI1=VALID, BINDING_ANCHOR=2 / SAVI2=NO_BIND An alternative behavior allowed by the specification for the case in which the host does not issue a DAD_NSOL is that SAVI1 does nothing. In this case, after some time (bounded by DEFAULT_LT), the switch will change the state for IPH to TESTING_VP, check if H is still at port 1 (which it is not), and move the state to NO_BIND. Then, a packet arriving from port 2 would trigger a process that finishes with a VALID stated with BINDING_ANCHOR=2. 4.1.2. Moving to a Port of a Different Switch Host H, connected to port 1 of SAVI1, moves to port 4 of SAVI2. Before moving, the SEND SAVI state associated to IPH, the IP address of H, is: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=NO_BIND If H issues a DAD_NSOL message for IPH when it connects to port 4 of SAVI2, the state is changed to: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=TENTATIVE_DAD, BINDING_ANCHOR=4, TIMER=TENT_LT The DAD_NSOL message is propagated only through the Trusted port of SAVI2. Then, SAVI1 changes its state as follows: SAVI1=TESTING_VP, BINDING_ANCHOR=1, TIMER=TENT_LT / SAVI2=TENTATIVE_DAD, BINDING_ANCHOR=4, TIMER=TENT_LT Bagnulo & Garcia-Martinez Standards Track [Page 30] RFC 7219 SEND SAVI May 2014 SAVI1 propagates the DAD_NSOL message to port 1. Since the only node that can answer with a secured DAD_NUD has moved, the timer at SAVI2 expires, and SAVI2 changes its state to VALID: SAVI1=TESTING_VP, BINDING_ANCHOR=1, TIMER=TENT_LT / SAVI2=VALID, BINDING_ANCHOR=4 Just a very short time after, the timer at SAVI1 expires, and the state changes to NO_BIND: SAVI1=NO_BIND / SAVI2=VALID, BINDING_ANCHOR=4 If host H does not send a DAD_NSOL when it moves to SAVI2 but instead sends a data packet, SAVI2 changes its state to TENTATIVE_NUD: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=TENTATIVE_NUD, BINDING_ANCHOR=4, TIMER=TENT_LT SAVI2 issues a secured NUD_NSOL through port 4. H is assumed to have the address configured (otherwise, it should not have generated a data packet), so it can respond with a NUD_NADV. When SAVI1 receives the NUD_NADV and validates it, the state is changed to VALID: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=VALID, BINDING_ANCHOR=4 After some time (bounded by DEFAULT_LT), the state in SAVI1 will expire, and SAVI1 will perform a check for host H: SAVI1=TESTING_VP, BINDING_ANCHOR=1, TIMER=TENT_LT / SAVI2=VALID, BINDING_ANCHOR=4 SAVI1 issues a NUD_NSOL through port 1 for IPH. No response is received in this case, so SAVI1 changes its state to NO_BIND: SAVI1=NO_BIND / SAVI2=VALID, BINDING_ANCHOR=4 4.2. Attack of a Malicious Host Host H is attached to the SEND SAVI infrastructure through port 1 of SAVI1. We consider that host M starts sending data packets using IPH (the IP address of H) as the source address, without issuing a DAD_NSOL (a similar analysis can be done for this case). 4.2.1. M Attaches to the Same Switch as the Victim's Switch The initial state before the attack of M is: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=NO_BIND Bagnulo & Garcia-Martinez Standards Track [Page 31] RFC 7219 SEND SAVI May 2014 M attaches to port 2 of SAVI1 and starts sending data packets. When SAVI1 receives the data packet, the packet is discarded. SEND SAVI may issue a secured NUD_NSOL through port 1 and changes the state to: SAVI1=TESTING_VP', BINDING_ANCHOR=1, ALTERNATIVE_BINDING_ANCHOR=2, TIMER=TENT_LT / SAVI2=NO_BIND Host H is still attached to port 1, so it receives the NUD_NSOL and responds with a secured NUD_NADV. SAVI1 receives this message, validates it, and changes its state again to: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=NO_BIND To prevent the drain of CPU resources in SAVI1, the processing of further packets received from port 2 may be rate-limited, as discussed in Section 5.2. An alternative to the previous behavior is that SAVI1 does nothing when node M starts sending packets from port 2. In this case, when the timer to renew the state triggers (this time it's bounded by DEFAULT_LT), SAVI1 moves the state to TESTING_VP, sends a NUD_NSOL through port 1, host H responds, and the state remains in VALID for BINDING_ANCHOR=1. In this way, communication of host H is also defended. 4.2.2. M Attaches to a Different Switch to the Victim's Switch The initial state before the attack of M is: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=NO_BIND M attaches to port 2 of SAVI2 and starts sending data packets. When SAVI2 receives the data packet, it changes the state to: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=TENTATIVE_DAD, BINDING_ANCHOR=2, TIMER=TENT_LT SAVI2 issues a secured NUD_NSOL through port 2. Since M does not own the IPH CGA, it cannot respond to the message. When the timer expires, the state is moved back to: SAVI1=VALID, BINDING_ANCHOR=1 / SAVI2=NO_BIND To prevent the drain of CPU resources in SAVI2, the processing of further packets received from port 2 may be rate-limited, as discussed in Section 5.2. Bagnulo & Garcia-Martinez Standards Track [Page 32] RFC 7219 SEND SAVI May 2014 5. Security Considerations SEND SAVI operates only with validated SEND messages to create bindings. Note that IPv6 packets generated by non-SEND nodes will be discarded by the first SEND SAVI device receiving it. Therefore, attackers cannot obtain any benefit by not using SEND. In order to perform address validation in a mixed scenario comprising SEND and non-SEND devices, a different solution is required, which should be addressed in another document. Nodes MUST NOT assume that all SEND messages received from a SEND SAVI device are validated, since these devices only validate the messages strictly required for SEND SAVI operation. Among the number of messages that are not validated by SEND SAVI, we can name NUD_NSOL messages generated by other nodes and its corresponding NUD_NADV responses, or RSOL messages. SEND SAVI improves protection compared to conventional SAVI as a result of the increased ability of SEND nodes to prove address ownership. A critical security consideration regarding SEND SAVI deals with the need of proper configuration of the roles of the ports in a SEND SAVI deployment scenario. Regarding security, the main requirement is that ports defining the protected perimeter SHOULD be configured as Validating ports. Not doing so will allow an attacker to send packets using any source address, regardless of the bindings established in other SEND SAVI devices. 5.1. Protection against Replay Attacks One possible concern about SEND SAVI is its behavior when an attacker tries to forge the identity of a legitimate node by replaying SEND messages used by the SEND SAVI specification. An attacker could replay any of these messages to interfere with the SEND SAVI operation. For example, it could replay a DAD_NSOL message to abort the configuration of an address for a legitimate node and to gain the right to use the address for DEFAULT_LT seconds. We can analyze two different cases when considering SEND SAVI replay attacks: o When the SEND message replayed is used to create or update binding information for SEND SAVI, since the port through which this message is received is key to the SEND SAVI operation. SEND SAVI creates and maintains bindings as a result of the reception of DAD_NSOL messages and of the exchange of NUD_NSOL/NUD_NADV messages. Bagnulo & Garcia-Martinez Standards Track [Page 33] RFC 7219 SEND SAVI May 2014 o When the SEND message replayed does not result in the update of binding information for SEND SAVI and, thus, is not related to the specific port through which it was received. Such situations are the reception of CPA messages containing certificates, and the processing of an RADV message coming from a Trusted port, which can be used in SEND SAVI to populate the SEND SAVI Prefix List. In these two cases, the security risks are equivalent to those of the SEND operation, i.e., we can consider that the information will not be changed by its legitimate sender for the time during which the SEND specification allows replaying (which depends on the values of TIMESTAMP_FUZZ and TIMESTAMP_DRIFT [RFC3971]). For replay of messages belonging to the second case, i.e., messages that do not result in changes in the SEND SAVI binding information, the security provided by SEND is sufficient. For the replay of messages belonging to the first case, DAD_NSOL and NUD_NSOL/NUD_NADV messages, protection results from the behavior of SEND SAVI, as specified in Section 3.3.2, which restricts the ports to which the messages involved in SEND SAVI binding updates are disseminated. SEND SAVI devices only forward these messages to ports for which a binding to the address being tested by the DAD_NSOL message existed. Therefore, it is not enough for an attacker to subscribe to a Solicited Node address to receive DAD_NSOL messages sent to that address, but the attacker needs to generate a valid DAD_NSOL message associated to the address for which the binding is being tested, which is deemed unfeasible [RFC3971]. 5.2. Protection against Denial-of-Service Attacks The attacks against the SEND SAVI device basically consist of making the SEND SAVI device consume its resources until it runs out of them. For instance, a possible attack would be to send packets with different source addresses, making the SEND SAVI device create state for each of the addresses and waste memory. At some point, the SEND SAVI device runs out of memory and needs to decide how to react. The result is that some form of garbage collection is needed to prune the entries. When the SEND SAVI device runs out of the memory allocated for the SEND SAVI Database, it is RECOMMENDED that it creates new entries by deleting the entries with a higher Creation time. This implies that older entries are preserved and newer entries overwrite each other. In an attack scenario where the attacker sends a batch of data packets with different source addresses, each new source address is likely to rewrite another source address created by the attack itself. It should be noted that entries are also garbage collected using the DEFAULT_LT, which is updated by NUD_NSOL/NUD_NADV exchanges. The result is that in order for an attacker to actually fill the SEND SAVI Database with false source addresses, it needs to continuously answer to NUD_NSOL for all the different source Bagnulo & Garcia-Martinez Standards Track [Page 34] RFC 7219 SEND SAVI May 2014 addresses, so that the entries grow old and compete with the legitimate entries. The result is that the cost of the attack is highly increased for the attacker. In addition, it is also RECOMMENDED that a SEND SAVI device reserves a minimum amount of memory for each available port (in the case where the port is used as part of the L2 anchor). The REQUIRED minimum is the memory needed to store four bindings associated to the port, although it SHOULD be raised if the ratio between the maximum number of bindings allowed in the device and the number of ports is high. The motivation for setting a minimum number of bindings per port is as follows. An attacker attached to a given port of a SEND SAVI device may attempt to launch a DoS attack towards the SEND SAVI device by creating many bindings for different addresses. It can do so by sending DAD_NSOL for different addresses. The result is that the attack will consume all the memory available in the SEND SAVI device. The above recommendation aims to reserve a minimum amount of memory per port, so that nodes located in different ports can make use of the reserved memory for their port even if a DoS attack is occurring in a different port. The SEND SAVI device may store data packets while the address is being verified, for example, when a DAD_NSOL is lost before arriving to the SEND SAVI device to which the host attaches; when the host sends data packets, these data packets may be stored until the SEND SAVI device verifies the binding by means of a NUD packet exchange. In this case, the memory for data packet storage may also be a target of DoS attacks. A SEND SAVI device MUST limit the amount of memory used to store data packets, allowing the other functions (such as being able to store new bindings) to have available memory even in the case of an attack, such as those described above. It is worth noting that the potential of DoS attacks against the SEND SAVI network is increased due to the use of costly cryptographic operations in order to validate the address of the nodes. An attacker could generate packets using new source addresses in order to make the closest SEND SAVI device spend CPU time to validate DAD_NSOL messages or to generate a secure NUD_NSOL. This attack can be used to drain CPU resources of SEND SAVI devices with a very low cost for the attacker. In order to solve this problem, rate-limiting the processing of packets that trigger SEND SAVI events SHOULD be enforced on a per-port basis. Bagnulo & Garcia-Martinez Standards Track [Page 35] RFC 7219 SEND SAVI May 2014 5.3. Considerations on the Deployment Model for Trust Anchors The SEND specification [RFC3971] proposes two deployment models for trust anchors: either a centralized model relaying on a globally rooted public key infrastructure or a more local, decentralized deployment model in which end hosts are configured with a collection of public keys that are trusted only on a domain. The appeal of a centralized model is the possibility for hosts to use SEND to validate routers as they move through links belonging to different organizations without additional configuration. However, without any further protection, it also enables routers authorized with a certificate path rooted on a global trust anchor to appear as legitimate routers in a link in which they were not intended to act as such. This threat already existed for SEND deployments, for which links configured to accept centralized trust anchors may send outgoing traffic and use prefix information from alien routers. In a SEND SAVI deployment, such routers may be able to deliver off-link traffic to any node of the link. In order to cope with this threat, SEND SAVI specifies that nodes are only allowed to behave as routers if they connect through Trusted ports. In particular, RADV messages and traffic with off-link source addresses are discarded when received through Validating ports, which are the ports intended for non-trusted infrastructure, as moving nodes. The protection provided by filtering RADV messages prevents SEND nodes from identifying alien routers as legitimate routers, even though the trust anchor of these routers is valid. Besides, it is worth to say that SEND SAVI supports a decentralized deployment model. 5.4. Residual Threats SEND SAVI assumes that a host will be able to defend its address when the DAD procedure is executed for its addresses, and that it will answer to a NUD_NSOL with a NUD_NADV when required. This is needed, among other things, to support mobility within a link (i.e., to allow a host to detach and reconnect to a different layer-2 anchor of the same IP subnetwork, without changing its IP address). If the SEND SAVI device does not see the DAD_NADV or the NUD_NADV, it may grant the binding to a different binding anchor. This means that if an attacker manages to prevent a host from defending its source address, it will be able to destroy the existing binding and create a new one, with a different binding anchor. An attacker may do so, for example, by launching a DoS attack to the host that will prevent it to issue proper replies. Bagnulo & Garcia-Martinez Standards Track [Page 36] RFC 7219 SEND SAVI May 2014 5.5. Privacy Considerations A SEND SAVI device MUST delete binding anchor information as soon as possible (i.e., as soon as the state for a given address is back to NO_BIND), except where there is an identified reason why that information is likely to be involved in the detection, prevention, or tracing of actual source address spoofing. Information about the majority of hosts that never spoof SHOULD NOT be logged. 6. Acknowledgments Thanks to Jean-Michel Combes, Ana Kukec, Ted Lemon, Adrian Farrel, Barry Leiba, Brian Haberman, Vicent Roca, and Benoit Claise for their reviews and comments on this document. The text has also benefited from feedback provided by Tony Cheneau and Greg Daley. Marcelo Bagnulo is partly funded by Trilogy 2, a research project supported by the European Commission under its Seventh Framework Program. Alberto Garcia-Martinez was supported, in part, by project TEC2012-38362-C03-01, granted by the Spanish Economy and Competitiveness Ministry. 7. References 7.1. Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC3971] Arkko, J., Kempf, J., Zill, B., and P. Nikander, "SEcure Neighbor Discovery (SEND)", RFC 3971, March 2005. [RFC3972] Aura, T., "Cryptographically Generated Addresses (CGA)", RFC 3972, March 2005. [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, September 2007. [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless Address Autoconfiguration", RFC 4862, September 2007. Bagnulo & Garcia-Martinez Standards Track [Page 37] RFC 7219 SEND SAVI May 2014 7.2. Informative References [IEEE.802-1Q.2005] Institute of Electrical and Electronics Engineers, "IEEE Standard for Local and Metropolitan Area Networks / Virtual Bridged Local Area Networks", IEEE Standard 802.1Q, May 2005. [RFC2827] Ferguson, P. and D. Senie, "Network Ingress Filtering: Defeating Denial of Service Attacks which employ IP Source Address Spoofing", BCP 38, RFC 2827, May 2000. [RFC6434] Jankiewicz, E., Loughney, J., and T. Narten, "IPv6 Node Requirements", RFC 6434, December 2011. [RFC6620] Nordmark, E., Bagnulo, M., and E. Levy-Abegnoli, "FCFS SAVI: First-Come, First-Served Source Address Validation Improvement for Locally Assigned IPv6 Addresses", RFC 6620, May 2012. [RFC7039] Wu, J., Bi, J., Bagnulo, M., Baker, F., and C. Vogt, "Source Address Validation Improvement (SAVI) Framework", RFC 7039, October 2013. Authors' Addresses Marcelo Bagnulo Universidad Carlos III de Madrid Av. Universidad 30 Leganes, Madrid 28911 Spain Phone: 34 91 6248814 EMail: marcelo@it.uc3m.es URI: http://www.it.uc3m.es Alberto Garcia-Martinez Universidad Carlos III de Madrid Av. Universidad 30 Leganes, Madrid 28911 Spain Phone: 34 91 6248782 EMail: alberto@it.uc3m.es URI: http://www.it.uc3m.es Bagnulo & Garcia-Martinez Standards Track [Page 38]