IPv6 and UDP Checksums for Tunneled Packets

This document constitutes an update of the IPv6 specification for cases where a tunnel protocol uses UDP with IPv6 to tunnel packets. With the rapid growth of the Internet, tunnel protocols have become increasingly important to enable the deployment of new protocols. Tunnel protocols can be deployed rapidly, while the time to upgrade and deploy a new protocol on a critical mass of routers, middleboxes, and hosts on the global Internet is now measured in decades. At the same time, the increasing use of firewalls and other security-related middleboxes means that truly new tunnel protocols, with new protocol numbers, are also unlikely to be deployable in a reasonable time frame. The result is an increasing interest in and use of UDP-based tunnel protocols. In such protocols, there is an encapsulated "inner" packet, and the "outer" packet carrying the tunneled inner packet is a UDP packet, which can pass through firewalls and other middleboxes that perform the filtering that is a fact of life on the current Internet. Tunnel endpoints may be routers or middleboxes aggregating traffic from a number of tunnel users. Therefore, the computation of an additional checksum on the outer UDP packet may be seen as an unwarranted burden on nodes that implement a tunnel protocol, especially if the inner packets are already protected by a checksum. IPv4 has a checksum over the IP packet header, and the checksum on the outer UDP packet may be set to zero. However, IPv6 has no checksum in the IP header, and RFC 2460 explicitly states that IPv6 receivers MUST discard UDP packets with a zero checksum. So, while sending a UDP datagram with a zero checksum is permitted in IPv4 packets, it is explicitly forbidden in IPv6 packets. To improve support for IPv6 UDP tunnels, this document updates RFC 2460 to allow endpoints to use a zero UDP checksum under constrained situations (primarily for IPv6 tunnel transports that carry checksum-protected packets), following the applicability statements and constraints in . When reading this document, the advice in "Unicast UDP Usage Guidelines for Application Designers" is applicable. It discusses both UDP tunnels (Section 3.1.3) and the usage of checksums (Section 3.4). While the origin of this specification is the problem raised by the draft titled "Automatic Multicast Tunnels", also known as "AMT" , we expect it to have wide applicability. Since the first draft of this RFC was written, the need for an efficient UDP tunneling mechanism has increased. Other IETF Working Groups, notably LISP and Softwires, have expressed a need to update the UDP checksum processing in RFC 2460. We therefore expect this update to be applicable in the future to other tunnel protocols specified by these and other IETF Working Groups.

This document discusses only IPv6, because the problem being addressed does not exist for IPv4. Therefore, all references to "IP" should be understood as references to IPv6. The document uses the terms "tunneling" and "tunneled" as adjectives when describing packets. When we refer to "tunneling packets", we refer to the outer packet header that provides the tunneling function. When we refer to "tunneled packets", we refer to the inner packet, i.e., the packet being carried in the tunnel.

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.

When using tunnel protocols based on UDP, there can be both a benefit and a cost to computing and checking the UDP checksum of the outer (encapsulating) UDP transport header. In certain cases, where reducing the forwarding cost is important, the cost of the computation may outweigh the benefit of the checksum. This document provides an update for usage of the UDP checksum with IPv6. The update is specified for use by a tunnel protocol that transports packets that are themselves protected by a checksum.

"Applicability Statement for the Use of IPv6 UDP Datagrams with Zero Checksums" describes issues related to allowing UDP over IPv6 to have a valid zero UDP checksum and is the starting point for this discussion. Sections 4 and 5 of , respectively, identify node implementation and usage requirements for datagrams sent and received with a zero UDP checksum. These sections introduce constraints on the usage of a zero checksum for UDP over IPv6. The remainder of this section analyzes the use of general tunnels and explains the motivations for why tunnel protocols are being permitted to use the method described in this update. It also discusses issues with middleboxes.

This section analyzes the impact of the different corruption modes in the context of a tunnel protocol. It specifies what needs to be considered by the designer and user of a tunnel protocol for the protocol to be robust. It also summarizes why use of a zero UDP checksum is thought to be safe for deployment. Context (i.e., tunneling state) should be established by exchanging application Protocol Data Units (PDUs) carried in checksummed UDP datagrams or by using other protocols that provide integrity protection against corruption. These control packets should also carry any negotiation required to enable the tunnel endpoint to accept UDP datagrams with a zero checksum and identify the set of ports that are used. It is important that the control traffic is robust against corruption, because undetected errors can lead to long-lived and significant failures that may affect much more than the single packet that was corrupted. Keepalive datagrams with a zero UDP checksum should be sent to validate the network path, because the path between tunnel endpoints can change, and therefore, the set of middleboxes along the path may change during the life of an association. Paths with middleboxes that drop datagrams with a zero UDP checksum will drop these keepalives. To enable the tunnel endpoints to discover and react to this behavior in a timely way, the keepalive traffic should include datagrams with a non-zero checksum and datagrams with a zero checksum. Receivers should attempt to detect corruption of the address information in an encapsulating packet. A robust tunnel protocol should track tunnel context based on the 5-tuple (tunneled protocol number, IPv6 source address, IPv6 destination address, UDP source port, UDP destination port). A corrupted datagram that arrives at a destination may be filtered based on this check. If the datagram header matches the 5-tuple and the node has enabled the zero checksum for this port, the payload is matched to the wrong context. The tunneled packet will then be decapsulated and forwarded by the tunnel egress. If a corrupted datagram matches a different 5-tuple and the node has enabled zero checksum for the port, the datagram payload is matched to the wrong context and may be processed by the wrong tunnel protocol, provided that it also passes the verification of that protocol. If a corrupted datagram matches a 5-tuple and node has not enabled the zero checksum for this port, the datagram will be discarded. When only the source information is corrupted, the datagram could arrive at the intended applications or protocol, which will process the datagram and try to match it against an existing tunnel context. The likelihood that a corrupted packet enters a valid context is reduced when the protocol restricts processing to only the source addresses with established contexts. When both source and destination fields are corrupted, this also decreases the likelihood of matching a context. However, the exception is when errors replace one packet header with another, so both packets could be tunneled, and therefore the corrupted packet could match a previously defined context. Receivers should attempt to detect corruption of source-fragmented encapsulating packets. A tunnel protocol may reassemble fragments associated with the wrong context at the right tunnel endpoint, it may reassemble fragments associated with a context at the wrong tunnel endpoint, or corrupted fragments may be reassembled at the right context at the right tunnel endpoint. In each of these cases, the IPv6 length of the encapsulating header may be checked (although points out the weakness in this check). In addition, if the encapsulated packet is protected by a transport (or other) checksum, these errors can be detected (with some probability). Compared to other applications, tunnel protocols using UDP have some advantages that reduce the risk for a corrupted tunnel packet reaching a destination that will receive it. These advantages result from processing by the network of the inner (tunneled) packet after it is forwarded from the tunnel egress using a wrong context: A tunneled packet may be forwarded to the wrong address domain, for example, to a private address domain where the inner packet's address is not routable, or it may fail a source address check, such as Unicast Reverse Path Forwarding, resulting in the packet being dropped. The destination address of a tunneled packet may not be reachable at all from the delivered domain. An example is an Ethernet frame where the destination MAC address is not present on the LAN segment that was reached. The type of the tunneled packet may prevent delivery. For example, an attempt to interpret an IP packet payload as an Ethernet frame would likely to result in the packet being dropped as invalid. The tunneled packet checksum or integrity mechanism may detect corruption of the inner packet caused at the same time as corruption to the outer packet header. The resulting packet would likely be dropped as invalid. Each of these checks significantly reduces the likelihood that a corrupted inner tunneled packet is finally delivered to a protocol listener that can be affected by the packet. While the methods do not guarantee correctness, they can reduce the risks of relaxing the UDP checksum requirement for a tunnel application using IPv6.

This document describes the applicability of using a zero UDP checksum to support tunnel protocols. There are good motivations behind this, and the arguments are provided here. Tunnels carry inner packets that have their own semantics, which may make any corruption less likely to reach the indicated destination and be accepted as a valid packet. This is true for IP packets with the addition of verification that can be made by the tunnel protocol, the network processing of the inner packet headers as discussed above, and verification of the inner packet checksums. Non-IP inner packets are likely to be subject to similar effects that may reduce the likelihood of a misdelivered packet being delivered to a protocol listener that can be affected by the packet. Protocols that directly consume the payload must have sufficient robustness against misdelivered packets (from any context), including ones that are corrupted in tunnels or corrupted by other usage of the zero checksum. This will require an integrity mechanism. Using a standard UDP checksum reduces the computational load in the receiver that is necessary to verify this mechanism. The design for stateful protocols or protocols where corruption causes cascade effects requires extra care. In tunnel usage, each encapsulating packet provides no functions other than a transport from tunnel ingress to tunnel egress. A corruption will commonly affect only the single tunneled packet, not the established protocol state. One common effect is that the inner packet flow will see only a corruption and a misdelivery of the outer packet as a lost packet. Some non-tunnel protocols operate with general servers that do not know the source from which they will receive a packet. In such applications, a zero UDP checksum is unsuitable, because it is necessary to provide the first level of verification that the packet was intended for the receiving server. A verification prevents the server from processing the datagram payload; without this, the server may spend significant resources processing the packet, including sending replies or error messages. Tunnel protocols that encapsulate IP will generally be safe for deployment, because all IPv4 and IPv6 packets include at least one checksum at either the network or transport layer. The network delivery of the inner packet will then further reduce the effects of corruption. Tunnel protocols carrying non-IP packets may offer equivalent protection when the non-IP networks reduce the risk of misdelivery to applications. However, further analysis is necessary to understand the implications of misdelivery of corrupted packets for each non-IP protocol. The analysis above suggests that non-tunnel protocols can be expected to have significantly more cases where a zero checksum would result in misdelivery or negative side effects. One unfortunate side effect of increased use of a zero checksum is that it also increases the likelihood of acceptance when a datagram with a zero UDP checksum is misdelivered. This requires all tunnel protocols using this method to be designed to be robust in the face of misdelivery.

"Applicability Statement for the Use of IPv6 UDP Datagrams with Zero Checksums" specifies requirements for middleboxes and tunnels that need to traverse middleboxes. Tunnel protocols intending to use a zero UDP checksum need to ensure that they have defined a method for handling cases when a middlebox prevents the path between the tunnel ingress and egress from supporting transmission of datagrams with a zero UDP checksum. This is especially important as middleboxes that conform to RFC 2460 are likely to discard datagrams with a zero UDP checksum.

This specification updates IPv6 to allow a zero UDP checksum in the outer encapsulating datagram of a tunnel protocol. UDP endpoints that implement this update MUST follow the node requirements in "Applicability Statement for the Use of IPv6 UDP Datagrams with Zero Checksums". The following text in , Section 8.1, fourth bullet should be deleted: "Unlike IPv4, when UDP packets are originated by an IPv6 node, the UDP checksum is not optional. That is, whenever originating a UDP packet, an IPv6 node must compute a UDP checksum over the packet and the pseudo-header, and, if that computation yields a result of zero, it must be changed to hex FFFF for placement in the UDP header. IPv6 receivers must discard UDP packets containing a zero checksum, and should log the error." This text should be replaced by: An IPv6 node associates a mode with each used UDP port (for sending and/or receiving packets). Whenever originating a UDP packet for a port in the default mode, an IPv6 node MUST compute a UDP checksum over the packet and the pseudo-header, and, if that computation yields a result of zero, the checksum MUST be changed to hex FFFF for placement in the UDP header, as specified in . IPv6 receivers MUST by default discard UDP packets containing a zero checksum and SHOULD log the error. As an alternative, certain protocols that use UDP as a tunnel encapsulation MAY enable zero-checksum mode for a specific port (or set of ports) for sending and/or receiving. Any node implementing zero-checksum mode MUST follow the node requirements specified in Section 4 of "Applicability Statement for the use of IPv6 UDP Datagrams with Zero Checksums". Any protocol that enables zero-checksum mode for a specific port or ports MUST follow the usage requirements specified in Section 5 of "Applicability Statement for the Use of IPv6 UDP Datagrams with Zero Checksums". Middleboxes supporting IPv6 MUST follow requirements 9, 10, and 11 of the usage requirements specified in Section 5 of "Applicability Statement for the Use of IPv6 UDP Datagrams with Zero Checksums".

This update was motivated by the existence of a number of protocols being developed in the IETF that are expected to benefit from the change. The following observations are made: An empirically based analysis of the probabilities of packet corruption (with or without checksums) has not, to our knowledge, been conducted since about 2000. At the time of publication, it is now 2013. We strongly suggest that a new empirical study be performed, along with extensive analysis of the corruption probabilities of the IPv6 header. This could potentially allow revising the recommendations in this document. A key motivation for the increase in use of UDP in tunneling is a lack of protocol support in middleboxes. Specifically, new protocols, such as LISP , may prefer to use UDP tunnels to traverse an end-to-end path successfully and avoid having their packets dropped by middleboxes. If middleboxes were updated to support UDP-Lite , UDP-Lite would provide better protection than offered by this update. UDP-Lite may be suited to a variety of applications and would be expected to be preferred over this method for many tunnel protocols. Another issue is that the UDP checksum is overloaded with the task of protecting the IPv6 header for UDP flows (as is the TCP checksum for TCP flows). Protocols that do not use a pseudo-header approach to computing a checksum or CRC have essentially no protection from misdelivered packets.

Less work is required to generate an attack using a zero UDP checksum than one using a standard full UDP checksum. However, this does not lead to significant new vulnerabilities, because checksums are not a security measure and can be easily generated by any attacker. In general, any user of zero UDP checksums should apply the checks and context verification that are possible to minimize the risk of unintended traffic to reach a particular context. This will, however, not protect against an intentional attack that creates packets with the correct information. Source address validation can help prevent injection of traffic into contexts by an attacker. Depending on the hardware design, the processing requirements may differ for tunnels that have a zero UDP checksum and those that calculate a checksum. This processing overhead may need to be considered when deciding whether to enable a tunnel and to determine an acceptable rate for transmission. This processing overhead can become a security risk for designs that can handle a significantly larger number of packets with zero UDP checksums compared to datagrams with a non-zero checksum, such as a tunnel egress. An attacker could attempt to inject non-zero checksummed UDP packets into a tunnel forwarding zero checksum UDP packets and cause overload in the processing of the non-zero checksums, e.g., if this happens in a router's slow path. Therefore, protection mechanisms should be employed when this threat exists. Protection may include source-address filtering to prevent an attacker from injecting traffic, as well as throttling the amount of non-zero checksum traffic. The latter may impact the functioning of the tunnel protocol.

We would like to thank Brian Haberman, Dan Wing, Joel Halpern, David Waltermire, J.W. Atwood, Peter Yee, Joe Touch, and the IESG of 2012 for discussions and reviews. Gorry Fairhurst has been very diligent in reviewing and helping to ensure alignment between this document and .