| RFC 0000 | IP Fragmentation Fragile | June 2020 |
| Bonica, et al. | Best Current Practice | [Page] |
This document describes IP fragmentation and explains how it introduces fragility to Internet communication.¶
This document also proposes alternatives to IP fragmentation and provides recommendations for developers and network operators.¶
This memo documents an Internet Best Current Practice.¶
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 BCPs is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc0000.¶
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
Operational experience [Kent] [Huston] [RFC7872] reveals that IP fragmentation introduces fragility to Internet communication. This document describes IP fragmentation and explains the fragility it introduces. It also proposes alternatives to IP fragmentation and provides recommendations for developers and network operators.¶
While this document identifies issues associated with IP fragmentation, it does not recommend deprecation. Legacy protocols that depend upon IP fragmentation would do well to be updated to remove that dependency. However, some applications and environments (see Section 5) require IP fragmentation. In these cases, the protocol will continue to rely on IP fragmentation, but the designer should to be aware that fragmented packets may result in black holes; a design should include appropriate safeguards.¶
Rather than deprecating IP fragmentation, this document recommends that upper-layer protocols address the problem of fragmentation at their layer, reducing their reliance on IP fragmentation to the greatest degree possible.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
An Internet path connects a source node to a destination node. A path may contain links and routers. If a path contains more than one link, the links are connected in series, and a router connects each link to the next.¶
Internet paths are dynamic. Assume that the path from one node to another contains a set of links and routers. If a link or a router fails, the path can also change so that it includes a different set of links and routers.¶
Each link is constrained by the number of bytes that it can convey in a single IP packet. This constraint is called the link Maximum Transmission Unit (MTU). IPv4 [RFC0791] requires every link to support 576 bytes or greater (see NOTE 1). IPv6 [RFC0791] similarly requires every link to support an MTU of 1280 bytes or greater. These are called the IPv4 and IPv6 minimum link MTUs.¶
Some links, and some ways of using links, result in additional variable overhead. For the simple case of tunnels, this document defers to other documents. For other cases, such as MPLS, this document considers the link MTU to include appropriate allowance for any such overhead.¶
Likewise, each Internet path is constrained by the number of bytes that it can convey in a single IP packet. This constraint is called the Path MTU (PMTU). For any given path, the PMTU is equal to the smallest of its link MTUs. Because Internet paths are dynamic, PMTU is also dynamic.¶
For reasons described below, source nodes estimate the PMTU between themselves and destination nodes. A source node can produce extremely conservative PMTU estimates in which:¶
While these conservative estimates are guaranteed to be less than or equal to the actual PMTU, they are likely to be much less than the actual PMTU. This may adversely affect upper-layer protocol performance.¶
By executing Path MTU Discovery (PMTUD) [RFC1191] [RFC8201] procedures, a source node can maintain a less conservative estimate of the PMTU between itself and a destination node. In PMTUD, the source node produces an initial PMTU estimate. This initial estimate is equal to the MTU of the first link along the path to the destination node. It can be greater than the actual PMTU.¶
Having produced an initial PMTU estimate, the source node sends non-fragmentable IP packets to the destination node (see NOTE 2). If one of these packets is larger than the actual PMTU, a downstream router will not be able to forward the packet through the next link along the path. Therefore, the downstream router drops the packet and sends an Internet Control Message Protocol (ICMP) [RFC0792] [RFC4443] Packet Too Big (PTB) message to the source node (see NOTE 3). The ICMP PTB message indicates the MTU of the link through which the packet could not be forwarded. The source node uses this information to refine its PMTU estimate.¶
PMTUD produces a running estimate of the PMTU between a source node and a destination node. Because PMTU is dynamic, the PMTU estimate can be larger than the actual PMTU. In order to detect PMTU increases, PMTUD occasionally resets the PMTU estimate to its initial value and repeats the procedure described above.¶
Ideally, PMTUD operates as described above. However, in some scenarios, PMTUD fails. For example:¶
When an upper-layer protocol submits data to the underlying IP module, and the resulting IP packet's length is greater than the PMTU, the packet is divided into fragments. Each fragment includes an IP header and a portion of the original packet.¶
[RFC0791] describes IPv4 fragmentation procedures. An IPv4 packet whose DF-bit is set to 1 may be fragmented by the source node, but may not be fragmented by a downstream router. An IPv4 packet whose DF-bit is set to 0 may be fragmented by the source node or by a downstream router. When an IPv4 packet is fragmented, all IP options (which are within the IPv4 header) appear in the first fragment, but only options whose "copy" bit is set to 1 appear in subsequent fragments.¶
[RFC8200], notably in Section 4.5, describes IPv6 fragmentation procedures. An IPv6 packet may be fragmented only at the source node. When an IPv6 packet is fragmented, all extension headers appear in the first fragment, but only per-fragment headers appear in subsequent fragments. Per-fragment headers include the following:¶
In IPv4, the upper-layer header usually appears in the first fragment, due to the sizes of the headers involved; in IPv6, it is required to.¶
Upper-layer protocols can operate in the following modes:¶
Upper-layer protocols running over IPv4 can operate in all of the above-mentioned modes. Upper-layer protocols running over IPv6 can operate in the first and second modes only.¶
Upper-layer protocols that operate in the first two modes (above) require access to the PMTU estimate. In order to fulfill this requirement, they can:¶
According to PLPMTUD procedures, the upper-layer protocol maintains a running PMTU estimate. It does so by sending probe packets of various sizes to its upper-layer peer and receiving acknowledgements. This strategy differs from PMTUD in that it relies on acknowledgement of received messages, as opposed to ICMP PTB messages concerning dropped messages. Therefore, PLPMTUD does not rely on the network's ability to deliver ICMP PTB messages to the source.¶
This section explains how IP fragmentation introduces fragility to Internet communication.¶
Virtual reassembly is a procedure in which a device conceptually reassembles a packet, forwards its fragments, and discards the reassembled copy. In A+P and CGN, virtual reassembly is required in order to correctly translate fragment addresses. It could be useful to address the problems in Section 3.2, Section 3.3, Section 3.4, and Section 3.5.¶
Virtual reassembly in the network is problematic, however, because it is computationally expensive and because it holds state for indeterminate periods of time, is prone to errors and, is prone to attacks (Section 3.7).¶
One of the benefits of fragmenting at the source, as IPv6 does, is that there is no question of temporary state or involved processes as required in virtual fragmentation. The sender has the entire message, and is fragmenting it as needed - and can apply that knowledge consistently across the fragments it produces. It is better than virtual fragmentation in that sense.¶
IP Fragmentation causes problems for routers that implement policy-based routing.¶
When a router receives a packet, it identifies the next-hop on route to the packet's destination and forwards the packet to that next-hop. In order to identify the next-hop, the router interrogates a local data structure called the Forwarding Information Base (FIB).¶
Normally, the FIB contains destination-based entries that map a destination prefix to a next-hop. Policy-based routing allows destination-based and policy-based entries to coexist in the same FIB. A policy-based FIB entry maps multiple fields, drawn from either the IP or transport-layer header, to a next-hop.¶
| Entry | Type | Dest. Prefix | Next Hdr / Dest. Port | Next-Hop |
|---|---|---|---|---|
| 1 | Destination- based | 2001:db8::1/128 | Any / Any | 2001:db8::2 |
| 2 | Policy- based | 2001:db8::1/128 | TCP / 80 | 2001:db8::3 |
Assume that a router maintains the FIB in Table 1. The first FIB entry is destination-based. It maps a destination prefix 2001:db8::1/128 to a next-hop 2001:db8::2. The second FIB entry is policy-based. It maps the same destination prefix 2001:db8::1/128 and a destination port ( TCP / 80 ) to a different next-hop (2001:db8::3). The second entry is more specific than the first.¶
When the router receives the first fragment of a packet that is destined for TCP port 80 on 2001:db8::1, it interrogates the FIB. Both FIB entries satisfy the query. The router selects the second FIB entry because it is more specific and forwards the packet to 2001:db8::3.¶
When the router receives the second fragment of the packet, it interrogates the FIB again. This time, only the first FIB entry satisfies the query, because the second fragment contains no indication that the packet is destined for TCP port 80. Therefore, the router selects the first FIB entry and forwards the packet to 2001:db8::2.¶
Policy-based routing is also known as filter-based-forwarding.¶
IP fragmentation causes problems for Network Address Translation (NAT) devices. When a NAT device detects a new, outbound flow, it maps that flow's source port and IP address to another source port and IP address. Having created that mapping, the NAT device translates:¶
A+P [RFC6346] and Carrier Grade NAT (CGN) [RFC6888] are two common NAT strategies. In both approaches the NAT device must virtually reassemble fragmented packets in order to translate and forward each fragment. (See NOTE 1.)¶
As discussed in more detail in Section 3.7, IP fragmentation causes problems for stateless firewalls whose rules include TCP and UDP ports. Because port information is only available in the first fragment and not available in the subsequent fragments the firewall is limited to the following options:¶
Neither option is attractive.¶
IP fragmentation causes problems for Equal Cost Multipath (ECMP), Link Aggregate Groups (LAG) and other stateless load-distribution technologies. In order to assign a packet or packet fragment to a link, an intermediate node executes a hash (i.e., load-distributing) algorithm. The following paragraphs describe a commonly deployed hash algorithm.¶
If the packet or packet fragment contains a transport-layer header, the algorithm accepts the following 5-tuple as input:¶
If the packet or packet fragment does not contain a transport-layer header, the algorithm accepts only the following 3-tuple as input:¶
Therefore, non-fragmented packets belonging to a flow can be assigned to one link while fragmented packets belonging to the same flow can be divided between that link and another. This can cause suboptimal load-distribution.¶
[RFC6438] offers a partial solution to this problem for IPv6 devices only. According to [RFC6438]:¶
"At intermediate routers that perform load balancing, the hash algorithm used to determine the outgoing component-link in an ECMP and/or LAG toward the next hop MUST minimally include the 3-tuple {dest addr, source addr, flow label} and MAY also include the remaining components of the 5-tuple."¶
If the algorithm includes only the 3-tuple {dest addr, source addr, flow label}, it will assign all fragments belonging to a packet to the same link. (See [RFC6437] and [RFC7098]).¶
In order to avoid the problem described above, implementations SHOULD implement the recommendations provided in Section 6.4 of this document.¶
IPv4 fragmentation is not sufficiently robust for use under some conditions in today's Internet. At high data rates, the 16-bit IP identification field is not large enough to prevent duplicate IDs resulting in frequent incorrectly assembled IP fragments, and the TCP and UDP checksums are insufficient to prevent the resulting corrupted datagrams from being delivered to higher protocol layers. [RFC4963] describes some easily reproduced experiments demonstrating the problem, and discusses some of the operational implications of these observations.¶
These reassembly issues do not occur as frequently in IPv6 because the IPv6 identification field is 32 bits long.¶
Security researchers have documented several attacks that exploit IP fragmentation. The following are examples:¶
In the overlapping fragment attack, an attacker constructs a series of packet fragments. The first fragment contains an IP header, a transport-layer header, and some transport-layer payload. This fragment complies with local security policy and is allowed to pass through a stateless firewall. A second fragment, having a non-zero offset, overlaps with the first fragment. The second fragment also passes through the stateless firewall. When the packet is reassembled, the transport layer header from the first fragment is overwritten by data from the second fragment. The reassembled packet does not comply with local security policy. Had it traversed the firewall in one piece, the firewall would have rejected it.¶
A stateless firewall cannot protect against the overlapping fragment attack. However, destination nodes can protect against the overlapping fragment attack by implementing the procedures described in RFC 1858, RFC 3128 and RFC 8200. These reassembly procedures detect the overlap and discard the packet.¶
The fragment reassembly algorithm is a stateful procedure in an otherwise stateless protocol. Therefore, it can be exploited by resource exhaustion attacks. An attacker can construct a series of fragmented packets, with one fragment missing from each packet so that the reassembly is impossible. Thus, this attack causes resource exhaustion on the destination node, possibly denying reassembly services to other flows. This type of attack can be mitigated by flushing fragment reassembly buffers when necessary, at the expense of possibly dropping legitimate fragments.¶
Each IP fragment contains an "Identification" field that destination nodes use to reassemble fragmented packets. Some implementations set the Identification field to a predictable value, thus making it easy for an attacker to forge malicious IP fragments that would cause the reassembly procedure for legitimate packets to fail.¶
NIDS aims at identifying malicious activity by analyzing network traffic. Ambiguity in the possible result of the fragment reassembly process may allow an attacker to evade these systems. Many of these systems try to mitigate some of these evasion techniques (e.g. By computing all possible outcomes of the fragment reassembly process, at the expense of increased processing requirements).¶
As mentioned in Section 2.3, upper-layer protocols can be configured to rely on PMTUD. Because PMTUD relies upon the network to deliver ICMP PTB messages, those protocols also rely on the networks to deliver ICMP PTB messages.¶
According to [RFC4890], ICMPv6 PTB messages must not be filtered. However, ICMP PTB delivery is not reliable. It is subject to both transient and persistent loss.¶
Transient loss of ICMP PTB messages can cause transient PMTU black holes. When the conditions contributing to transient loss abate, the network regains its ability to deliver ICMP PTB messages and connectivity between the source and destination nodes is restored. Section 3.8.1 of this document describes conditions that lead to transient loss of ICMP PTB messages.¶
Persistent loss of ICMP PTB messages can cause persistent black holes. Section 3.8.2, Section 3.8.3, and Section 3.8.4 of this document describe conditions that lead to persistent loss of ICMP PTB messages.¶
The problem described in this section is specific to PMTUD. It does not occur when the upper-layer protocol obtains its PMTU estimate from PLPMTUD or from any other source.¶
The following factors can contribute to transient loss of ICMP PTB messages:¶
The effect of rate limiting may be severe, as RFC 4443 recommends strict rate limiting of ICMPv6 traffic.¶
Incorrect implementation of security policy can cause persistent loss of ICMP PTB messages.¶
For example assume that a Customer Premise Equipment (CPE) router implements the following zone-based security policy:¶
When a correct implementation of the above-mentioned security policy receives an ICMP PTB message, it examines the ICMP PTB payload in order to determine whether the original packet (i.e., the packet that elicited the ICMP PTB message) belonged to an existing flow. If the original packet belonged to an existing flow, the implementation allows the ICMP PTB to flow from the outside zone to the inside zone. If not, the implementation discards the ICMP PTB message.¶
When an incorrect implementation of the above-mentioned security policy receives an ICMP PTB message, it discards the packet because its source address is not associated with an existing flow.¶
The security policy described above has been implemented incorrectly on many consumer CPE routers.¶
Anycast can cause persistent loss of ICMP PTB messages. Consider the example below:¶
A DNS client sends a request to an anycast address. The network routes that DNS request to the nearest instance of that anycast address (i.e., a DNS Server). The DNS server generates a response and sends it back to the DNS client. While the response does not exceed the DNS server's PMTU estimate, it does exceed the actual PMTU.¶
A downstream router drops the packet and sends an ICMP PTB message the packet's source (i.e., the anycast address). The network routes the ICMP PTB message to the anycast instance closest to the downstream router. That anycast instance may not be the DNS server that originated the DNS response. It may be another DNS server with the same anycast address. The DNS server that originated the response may never receive the ICMP PTB message and may never update its PMTU estimate.¶
Unidirectional routing can cause persistent loss of ICMP PTB messages. Consider the example below:¶
A source node sends a packet to a destination node. All intermediate nodes maintain a route to the destination node, but do not maintain a route to the source node. In this case, when an intermediate node encounters an MTU issue, it cannot send an ICMP PTB message to the source node.¶
In RFC 7872, researchers sampled Internet paths to determine whether they would convey packets that contain IPv6 extension headers. Sampled paths terminated at popular Internet sites (e.g., popular web, mail and DNS servers).¶
The study revealed that at least 28% of the sampled paths did not convey packets containing the IPv6 Fragment extension header. In most cases, fragments were dropped in the destination autonomous system. In other cases, the fragments were dropped in transit autonomous systems.¶
Another study [Huston] confirmed this finding. It reported that 37% of sampled endpoints used IPv6-capable DNS resolvers that were incapable of receiving a fragmented IPv6 response.¶
It is difficult to determine why network operators drop fragments. Possible causes follow:¶
The Transport Control Protocol (TCP) [RFC0793]) can be operated in a mode that does not require IP fragmentation.¶
Applications submit a stream of data to TCP. TCP divides that stream of data into segments, with no segment exceeding the TCP Maximum Segment Size (MSS). Each segment is encapsulated in a TCP header and submitted to the underlying IP module. The underlying IP module prepends an IP header and forwards the resulting packet.¶
If the TCP MSS is sufficiently small, the underlying IP module never produces a packet whose length is greater than the actual PMTU. Therefore, IP fragmentation is not required.¶
TCP offers the following mechanisms for MSS management:¶
Manual configuration is always applicable. If the MSS is configured to a sufficiently low value, the IP layer will never produce a packet whose length is greater than the protocol minimum link MTU. However, manual configuration prevents TCP from taking advantage of larger link MTUs.¶
Upper-layer protocols can implement PMTUD in order to discover and take advantage of larger path MTUs. However, as mentioned in Section 2.1, PMTUD relies upon the network to deliver ICMP PTB messages. Therefore, PMTUD can only provide an estimate of the PMTU in environments where the risk of ICMP PTB loss is acceptable (e.g., known to not be filtered).¶
By contrast, PLPMTUD does not rely upon the network's ability to deliver ICMP PTB messages. It utilises probe messages sent as TCP segments to determine whether the probed PMTU can be successfully used across the network path. In PLPMTUD, probing is separated from congestion control, so that loss of a TCP probe segment does not cause a reduction of the congestion control window. [RFC4821] defines PLPMTUD procedures for TCP.¶
While TCP will never knowingly cause the underlying IP module to emit a packet that is larger than the PMTU estimate, it can cause the underlying IP module to emit a packet that is larger than the actual PMTU. For example, if routing changes and as a result the PMTU becomes smaller, TCP will not know until the ICMP PTB message arrives. If this occurs, the packet is dropped, the PMTU estimate is updated, the segment is divided into smaller segments and each smaller segment is submitted to the underlying IP module.¶
The Datagram Congestion Control Protocol (DCCP) [RFC4340] and the Stream Control Transport Protocol (SCTP) [RFC4960] also can be operated in a mode that does not require IP fragmentation. They both accept data from an application and divide that data into segments, with no segment exceeding a maximum size.¶
DCCP offers manual configuration, PMTUD, and PLPMTUD as mechanisms for managing that maximum size. Datagram protocols can also implement PLPMTUD to estimate the PMTU via [RFCYYYY]. This proposes procedures for performing PLPMTUD with UDP, UDP-Options, SCTP, QUIC and other datagram protocols.¶
Currently, User Datagram Protocol (UDP) [RFC0768] lacks a fragmentation mechanism of its own and relies on IP fragmentation. However, [UDP-OPTIONS] proposes a fragmentation mechanism for UDP.¶
[RFC8085] recognizes that IP fragmentation reduces the reliability of Internet communication. It also recognizes that UDP lacks a fragmentation mechanism of its own and relies on IP fragmentation. Therefore, [RFC8085] offers the following advice regarding applications the run over the UDP.¶
An application SHOULD NOT send UDP datagrams that result in IP packets that exceed the Maximum Transmission Unit (MTU) along the path to the destination. Consequently, an application SHOULD either use the path MTU information provided by the IP layer or implement Path MTU Discovery (PMTUD) itself to determine whether the path to a destination will support its desired message size without fragmentation.¶
RFC 8085 continues:¶
Applications that do not follow the recommendation to do PMTU/PLPMTUD discovery SHOULD still avoid sending UDP datagrams that would result in IP packets that exceed the path MTU. Because the actual path MTU is unknown, such applications SHOULD fall back to sending messages that are shorter than the default effective MTU for sending (EMTU_S in [RFC1122]). For IPv4, EMTU_S is the smaller of 576 bytes and the first-hop MTU. For IPv6, EMTU_S is 1280 bytes [RFC8200]. The effective PMTU for a directly connected destination (with no routers on the path) is the configured interface MTU, which could be less than the maximum link payload size. Transmission of minimum-sized UDP datagrams is inefficient over paths that support a larger PMTU, which is a second reason to implement PMTU discovery.¶
RFC 8085 assumes that for IPv4, an EMTU_S of 576 is sufficiently small to be supported by most current Internet paths, even though the IPv4 minimum link MTU is 68 bytes.¶
This advice applies equally to any application that runs directly over IP.¶
The following applications rely on IPv6 fragmentation:¶
Each of these applications relies on IPv6 fragmentation to a varying degree. In some cases, that reliance is essential, and cannot be broken without fundamentally changing the protocol. In other cases, that reliance is incidental, and most implementations already take appropriate steps to avoid fragmentation.¶
This list is not comprehensive, and other protocols that rely on IP fragmentation may exist. They are not specifically considered in the context of this document.¶
DNS relies on UDP for efficiency, and the consequence is the use of IP fragmentation for large responses, as permitted by the DNS EDNS0 options in the query. It is possible to mitigate the issue of fragmentation-based packet loss by having queries use smaller EDNS0 UDP buffer sizes, or by having the DNS server limit the size of its UDP responses to some self-imposed maximum packet size that may be less than the preferred EDNS0 UDP Buffer Size. In both cases, large responses are truncated in the DNS, signaling to the client to re-query using TCP to obtain the complete response. However, the operational issue of the partial level of support for DNS over TCP, particularly in the case where IPv6 transport is being used, becomes a limiting factor of the efficacy of this approach [Damas].¶
Larger DNS responses can normally be avoided by aggressively pruning the Additional section of DNS responses. One scenario where such pruning is ineffective is in the use of DNSSEC, where large key sizes act to increase the response size to certain DNS queries. There is no effective response to this situation within the DNS other than using smaller cryptographic keys and adoption of DNSSEC administrative practices that attempt to keep DNS response as short as possible.¶
OSPF implementations can emit messages large enough to cause fragmentation. However, in order to optimize performance, most OSPF implementations restrict their maximum message size to a value that will not cause fragmentation.¶
This document acknowledges that in some cases, packets must be fragmented within IP-in-IP tunnels. Therefore, this document makes no additional recommendations regarding IP-in-IP tunnels.¶
In this document, packet-in-packet encapsulations include IP-in-IP [RFC2003], Generic Routing Encapsulation (GRE) [RFC2784], GRE-in-UDP [RFC8086] and Generic Packet Tunneling in IPv6 [RFC2473]. [RFC4459] describes fragmentation issues associated with all of the above-mentioned encapsulations.¶
The fragmentation strategy described for GRE in [RFC7588] has been deployed for all of the above-mentioned encapsulations. This strategy does not rely on IP fragmentation except in one corner case. (see Section 3.3.2.2 of RFC 7588 [RFC7588] and Section 7.1 of RFC 2473 [RFC2473]). Section 3.3 of [RFC7676] further describes this corner case.¶
Some UDP applications rely on IP fragmentation to achieve acceptable levels of performance. These applications use UDP datagram sizes that are larger than the path MTU so that more data can be conveyed between the application and the kernel in a single system call.¶
To pick one example, the Licklider Transmission Protocol (LTP), [RFC5326]which is in current use on the International Space Station (ISS), uses UDP datagram sizes larger than the path MTU to achieve acceptable levels of performance even though this invokes IP fragmentation. More generally, SNMP and video applications may transmit an application-layer quantum of data, depending on the network layer to fragment and reassemble as needed.¶
Developers SHOULD NOT develop new protocols or applications that rely on IP fragmentation. When a new protocol or application is deployed in an environment that does not fully support IP fragmentation, it SHOULD operate correctly, either in its default configuration or in a specified alternative configuration.¶
While there may be controlled environments where IP fragmentation works reliably, this is a deployment issue and can not be known to someone developing a new protocol or application. It is not recommended that new protocols or applications be developed that rely on IP fragmentation. Protocols and applications that rely on IP fragmentation will work less reliably on the Internet.¶
Legacy protocols that depend upon IP fragmentation SHOULD be updated to break that dependency. However, in some cases, there may be no viable alternative to IP fragmentation (e.g., IPSEC tunnel mode, IP-in-IP encapsulation). Applications and protocols cannot necessarily know or control whether they use lower layers or network paths that rely on such fragmentation. In these cases, the protocol will continue to rely on IP fragmentation but should only be used in environments where IP fragmentation is known to be supported.¶
Protocols may be able to avoid IP fragmentation by using a sufficiently small MTU (e.g., The protocol minimum link MTU), disabling IP fragmentation, and ensuring that the transport protocol in use adapts its segment size to the MTU. Other protocols may deploy a sufficiently reliable PMTU discovery mechanism (e.g., PLMPTUD).¶
UDP applications SHOULD abide by the recommendations stated in Section 3.2 of [RFC8085].¶
Software libraries SHOULD include provision for PLPMTUD for each supported transport protocol.¶
Middle boxes, which are systems that "transparently" perform policy functions on passing traffic but do not participate in the routing system, should process IP fragments in a manner that is consistent with [RFC0791] and [RFC8200]. In many cases, middle boxes must maintain state in order to achieve this goal.¶
Price and performance considerations frequently motivate network operators to deploy stateless middle boxes. These stateless middle boxes may perform sub-optimally, process IP fragments in a manner that is not compliant with RFC 791 or RFC 8200, or even discard IP fragments completely. Such behaviors are NOT RECOMMENDED. If a middleboxes implements non-standard behavior with respect to IP fragmentation, then that behavior MUST be clearly documented.¶
In their default configuration, when the IPv6 Flow Label is not equal to zero, IPv6 devices that implement Equal-Cost Multipath (ECMP) Routing as described in OSPF [RFC2328] and other routing protocols, Link Aggregation Grouping (LAG) [RFC7424], or other load-distribution technologies SHOULD accept only the following fields as input to their hash algorithm:¶
Operators SHOULD deploy these devices in their default configuration.¶
These recommendations are similar to those presented in [RFC6438] and [RFC7098]. They differ in that they specify a default configuration.¶
Operators MUST ensure proper PMTUD operation in their network, including making sure the network generates PTB packets when dropping packets too large compared to outgoing interface MTU. However, implementations MAY rate limit the generation of ICMP messages as per [RFC1812] and [RFC4443].¶
As per RFC 4890, network operators MUST NOT filter ICMPv6 PTB messages unless they are known to be forged or otherwise illegitimate. As stated in Section 3.8, filtering ICMPv6 PTB packets causes PMTUD to fail. Many upper-layer protocols rely on PMTUD.¶
As per RFC 8200, network operators MUST NOT deploy IPv6 links whose MTU is less than 1280 bytes.¶
Network operators SHOULD NOT filter IP fragments if they are known to have originated at a domain name server or be destined for a domain name server. This is because domain name services are critical to operation of the Internet.¶
This document has no IANA actions.¶
This document mitigates some of the security considerations associated with IP fragmentation by discouraging its use. It does not introduce any new security vulnerabilities, because it does not introduce any new alternatives to IP fragmentation. Instead, it recommends well-understood alternatives.¶
Thanks to Mikael Abrahamsson, Brian Carpenter, Silambu Chelvan, Joel Halpern, Lorenzo Colitti, Gorry Fairhurst, Mike Heard, Tom Herbert, Tatuya Jinmei, Suresh Krishnan, Jen Linkova, Paolo Lucente, Manoj Nayak, Eric Nygren, Fred Templin, and Joe Touch for their comments.¶