| RFC 9715 | Avoid IP Fragmentation | January 2025 |
| Fujiwara & Vixie | Informational | [Page] |
The widelydeployed Extension Mechanisms for DNS (EDNS(0)) feature in the DNS enables a DNS receiver to indicateits received UDP message size capacity, which supports the sending oflarge UDP responses by a DNS server.Large DNS/UDP messages are more likely to be fragmented,and IP fragmentation has exposed weaknesses in application protocols.It is possible to avoid IP fragmentation in DNS by limiting the responsesize where possible and signaling the need to upgrade from UDP to TCPtransport where necessary.This document describes techniques to avoid IP fragmentation in DNS.¶
This document is not an Internet Standards Track specification; it is published for informational purposes.¶
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). Not all documents approved by the IESG are candidates for any level of Internet Standard; see 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 athttps://www.rfc-editor.org/info/rfc9715.¶
Copyright (c) 2025 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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
This document was originally intended to be a Best Current Practice, but due tooperating system and socket option limitations, some of therecommendations have not yet gained real-world experience;therefore, this document is Informational.It is expected that, as operating systems and implementations evolve,we will gain more experience with the recommendations and will publish anupdated document as a Best Current Practice in the future.¶
DNS has an EDNS(0) mechanism[RFC6891].The widely deployed EDNS(0) feature in the DNS enables a DNS receiver to indicateits received UDP message size capacity, which supports the sending oflarge UDP responses by a DNS server.DNS over UDP invites IP fragmentation when a packet is larger than theMaximum Transmission Unit (MTU) of some network in the packet's path.¶
Fragmented DNS UDP responses have systemic weaknesses, which exposethe requestor to DNS cache poisoning from off-path attackers (seeSection 7.3 for references and details).¶
[RFC8900] states that IP fragmentationintroduces fragility to Internet communication.The transport of DNS messagesover UDP should take account of the observations stated in that document.¶
TCP avoids fragmentation by segmenting data into packets that are smaller than or equal to the Maximum Segment Size (MSS). For each transmitted segment, the size of the IP and TCP headers is known,and the IP packet size can be chosen to keep it within the estimated MTU and the MSS. This takes advantage of the elasticity of the TCP'spacketizing process, depending on how much queued data will fit into the nextsegment. In contrast, DNS over UDP has little datagram size elasticity andlacks insight into IP header and option size, so we must make moreconservative estimates about available UDP payload space.¶
[RFC7766] states that all general-purpose DNS implementationsMUST support both UDP and TCP transport.¶
DNS transaction security[RFC8945][RFC2931] does protectagainst the security risks of fragmentation, and it protectsdelegation responses. But[RFC8945] has limited applicability dueto key distribution requirements, and there is little if any deploymentof[RFC2931].¶
This document describes various techniques to avoid IP fragmentationof UDP packets in DNS.This document is primarily applicable to DNS use on the global Internet.¶
In contrast, a path MTU that deviates from therecommended value might be obtained through static configuration, serverrouting hints, or a future discovery protocol. However, addressingthis falls outside the scope of this document and may be the subjectof future specifications.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14[RFC2119][RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The definitions of "requestor" and "responder" are per[RFC6891]:¶
"Requestor" refers to the side that sends a request. "Responder"refers to an authoritative, recursive resolver or other DNS componentthat responds to questions.¶
The definition of "path MTU" is per[RFC8201]:¶
path MTU [is] the minimum link MTU of all the links in a path between a source node and a destination node.¶
In this document, the term "Path MTU Discovery" includesboth Classical Path MTU Discovery[RFC1191][RFC8201] andPacketization Layer Path MTU Discovery[RFC8899].¶
Many of the specialized terms used in this document are defined in"DNS Terminology"[RFC9499].¶
These recommendations are intendedfor nodes with global IP addresses on the Internet.Private networks or local networks are out of the scope of this document.¶
The methods to avoid IP fragmentation in DNS are described below:¶
UDP responders should configure their systems to prevent fragmentation of UDP packets when sending replies, provided it can be done safely. The mechanisms to achieve this vary across different operating systems.¶
For BSD-like operating systems, the IP Don't Fragment (DF) flag bit[RFC0791] can be used to prevent fragmentation. In contrast, Linux systems do not expose a direct API for this purpose and require the use of Path MTU socket options (IP_MTU_DISCOVER) to manage fragmentation settings. However, it is important to note that enabling IPv4 Path MTU Discovery for UDP in current Linux versions is considered harmful and dangerous. For more details, seeAppendix C.¶
Large DNS responses are typically the result of zone configuration. People who publish information in the DNS should seek configurations resulting in small responses. For example:¶
It is difficult to determine a specific upper limit for R8, R9, andR11, but it is sufficient if all responses from the DNS servers arebelow the size of R3 and R5.¶
Some authoritative servers deviate from the DNS standard as follows:¶
Some authoritative servers ignore the EDNS(0) requestor's maximum UDP payload size and return large UDP responses[Fujiwara2018].¶
Some authoritative servers do not support TCP transport.¶
Such non-compliant behavior cannot become implementation or configurationconstraints for the rest of the DNS. If failure is the result, then thatfailure must be localized to the non-compliant servers.¶
This document has no IANA actions.¶
If the Don't Fragment (DF) flag bit is not set,on-path fragmentation may happen on IPv4,and it can lead to vulnerabilities as shown inSection 7.3.To avoid this, R6 needs to be used to discard the fragmented responses and retry using TCP.¶
When avoiding fragmentation,a DNS/UDP requestor behind a small MTU network may experienceUDP timeouts, which would reduce performanceand may lead to TCP fallback.This would indicate prior reliance upon IP fragmentation,which is considered to be harmfulto both the performance and stability of applications, endpoints, and gateways.Avoiding IP fragmentation will improve operating conditions overall,and the performance of DNS/TCP has increased and will continue to increase.¶
If a UDP response packet is dropped in transit,up to and including the network stack of the initiator,it increases the attack window for poisoning the requestor's cache.¶
"Fragmentation Considered Poisonous"[Herzberg2013] notes effectiveoff-path DNS cache poisoning attack vectors using IP fragmentation."IP fragmentation attack on DNS"[Hlavacek2013] and "Domain Validation++For MitM-Resilient PKI"[Brandt2018] note that off-path attackerscan intervene in the Path MTU Discovery[RFC1191]to cause authoritative servers to produce fragmented responses.[RFC7739] states the security implications of predictable fragment identification values.¶
Section 3.2 of [RFC8085] states that"an applicationSHOULD NOT send UDP datagramsthat result in IP packets that exceed the Maximum Transmission Unit (MTU)along the path to the destination".¶
A DNS message receiver cannot trust fragmented UDP datagrams primarily due tothe small amount of entropy provided by UDP port numbers and DNS messageidentifiers, each of which is only 16 bits in size, and both are likelyto be in the first fragment of a packet if fragmentation occurs.By comparison, the TCP protocol stack controls packet size and avoids IP fragmentation under ICMP NEEDFRAG attacks.In TCP, fragmentation should be avoided for performance reasons, whereas forUDP, fragmentation should be avoided for resiliency and authenticity reasons.¶
DNSSEC is a countermeasure against cache poisoning attacks that useIP fragmentation.However, DNS delegation responses are not signed with DNSSEC,and DNSSEC does not have a mechanism to get the correct response ifan incorrect delegation is injected. This is a denial-of-servicevulnerability that can yield failed name resolutions.If cache poisoning attacks can be avoided,DNSSEC validation failures will be avoided.¶
Because this document is published as Informationalrather than a Best Current Practice,this section presents steps that resolver operators can taketo avoid vulnerabilities related to IP fragmentation.¶
To avoid vulnerabilities related to IP fragmentation, implement R5 and R6.¶
Specifically, configure the firewall functions protecting the full-service resolverto discard incoming DNS response packetswith a non-zero Fragment Offset (FO) or a More Fragments (MF) flag bit of 1 on IPv4,and discard packets with IPv6 Fragment Headers.(If the resolver's IP address is not dedicated to the DNS resolverand uses UDP communication that relies on IP Fragmentation for purposesother than DNS, discard only the first fragment that contains the UDP headerfrom port 53.)¶
The most recent resolver software is believed to implement R7.¶
Even if R7 is not implemented, it will only result in a name resolution error,preventing attacks from leading to malicious sites.¶
There are many discussions about default path MTU size and a requestor's maximum UDP payload size.¶
The minimum MTU for an IPv6 interface is 1280 octets(seeSection 5 of [RFC8200]).So, it can be used as the default path MTU value for IPv6.The corresponding minimum MTU for an IPv4 interface is 68 (60 + 8)[RFC0791].¶
[RFC4035] states that "A security-aware name serverMUST support the EDNS0 ([RFC2671]) message size extension, [and it]MUST support a message size of at least 1220 octets". Then, the smallest number ofthe maximum DNS/UDP payload size is 1220.¶
In order to avoid IP fragmentation,[DNSFlagDay2020] proposes that UDP requestors set the requestor'spayload size to 1232 and UDP responders compose UDP responses so they fitin 1232 octets.The size 1232 is based on an MTU of 1280, which is requiredby the IPv6 specification[RFC8200],minus 48 octets for the IPv6 and UDP headers.¶
Most of the Internet, especially the inner core, has an MTU of at least 1500 octets.Maximum DNS/UDP payload size for IPv6 on an MTU 1500 Ethernet is1452 (1500 minus 40 (IPv6 header size) minus 8 (UDP header size)).To allow for possible IP options and distant tunnel overhead,the recommendation of default maximum DNS/UDP payload size is 1400.¶
[Huston2021] analyzes the result of[DNSFlagDay2020] and reports thattheir measurements suggest that in the interior of the Internetbetween recursive resolvers and authoritative servers, the prevailing MTU is 1500and there is no measurable signal of use of smaller MTUsin this part of the Internet. They propose thattheir measurements suggest setting the EDNS(0) requestor's UDP payload size to1472 octets for IPv4 and 1452 octets for IPv6.¶
As a result of these discussions,this document recommends a value of 1400,with smaller values also allowed.¶
Some implementations have a "minimal responses" configuration setting/option that causesa DNS server to make response packets smaller, containing only mandatory and required data.¶
Under the minimal-responses configuration,a DNS server composes responses containing only necessary Resource Records (RRs).For delegations, see[RFC9471].In case of a non-existent domain name or non-existent type, the authority section will contain an SOA record, and the answer section is empty(seeSection 2 of [RFC2308]).¶
Some resource records (MX, SRV, SVCB, and HTTPS) requireadditional A, AAAA, and Service Binding (SVCB) recordsin the Additional sectiondefined in[RFC1035],[RFC2782], and[RFC9460].¶
In addition, if the zone is DNSSEC signed and a query has the DNSSEC OK bit,signatures are added in the answer section,or the corresponding DS RRSet and signatures are added in the authority section.Details are defined in[RFC4035] and[RFC5155].¶
This section records the status of known implementations of the proposed recommendations described inSection 3.¶
Please note that the listing of any individual implementation here does notimply endorsement by the IETF. Furthermore, no effort has been made toverify the information that was supplied by IETF contributors and presented here.¶
BIND 9 does not implement R1 and R2.¶
BIND 9 on Linux sets IP_MTU_DISCOVER to IP_PMTUDISC_OMIT with a fallback to IP_PMTUDISC_DONT.¶
When BIND 9 is on systems with IP_DONTFRAG (such as FreeBSD), IP_DONTFRAG is disabled.¶
Accepting Path MTU Discovery for UDP is considered harmful and dangerous.BIND 9's settings avoid attacks to Path MTU Discovery.¶
For R3, BIND 9 will honor the requestor's size up to theconfigured limit (max-udp-size). The UDP response packet is bound to bebetween 512 and 4096 bytes, with the default set to 1232. BIND 9 supports therequestor's size up to the configured limit (max-udp-size).¶
In the case of R4 and the send fails with EMSGSIZE, BIND 9sets the TC bit and tries to send a minimal answer again.¶
For R5, BIND 9 uses theedns-buf-sizeoption, with the default of 1232.¶
For R7, after two UDP timeouts, BIND 9 will fall back to TCP.¶
Both Knot servers set IP_PMTUDISC_OMIT to avoid path MTU spoofing. The UDP size limit is 1232 by default.¶
Fragments are ignored if they arrive over a Linux XDP interface.¶
TCP is attempted after repeated UDP timeouts.¶
Minimal responses are returned and are currently not configurable.¶
Smaller signatures are used, with ecdsap256sha256 as the default.¶
Unbound sets IP_MTU_DISCOVER to IP_PMTUDISC_OMIT with fallback toIP_PMTUDISC_DONT. It also disables IP_DONTFRAG on systems that haveit, but not on Apple systems. On systems that support it, Unbound setsIPV6_USE_MIN_MTU, with a fallback to IPV6_MTU at 1280, with a fallbackto IPV6_USER_MTU. It also sets IPV6_MTU_DISCOVER to IPV6_PMTUDISC_OMIT,with a fallback to IPV6_PMTUDISC_DONT.¶
Unbound requests a UDP size of 1232 from peers, by default. The requestor's size is limited to a max of 1232.¶
After some timeouts, Unbound retries with a smaller size, if applicable, or atsize 1232 for IPv6 and 1472 for IPv4. This does not cause any negative effects due to the "flag day"[DNSFlagDay2020] change to 1232.¶
Unbound has the "minimal responses" configuration option; set default on.¶
The authors would like to specifically thankPaul Wouters,Mukund Sivaraman,Tony Finch,Hugo Salgado,Peter van Dijk,Brian Dickson,Puneet Sood,Jim Reid,Petr Spacek,Andrew McConachie,Joe Abley,Daisuke Higashi,Joe Touch,Wouter Wijngaards,Vladimir Cunat,Benno Overeinder, andŠtěpán Němec for their extensive reviews and comments.¶