| Internet-Draft | DPLPMTUD | December 2019 |
| Fairhurst, et al. | Expires 7 June 2020 | [Page] |
This document describes a robust method for Path MTU Discovery (PMTUD) for datagram Packetization Layers (PLs). It describes an extension to RFC 1191 and RFC 8201, which specifies ICMP-based Path MTU Discovery for IPv4 and IPv6. The method allows a PL, or a datagram application that uses a PL, to discover whether a network path can support the current size of datagram. This can be used to detect and reduce the message size when a sender encounters a network black hole (where packets are discarded). The method can probe a network path with progressively larger packets to discover whether the maximum packet size can be increased. This allows a sender to determine an appropriate packet size, providing functionally for datagram transports that is equivalent to the Packetization Layer PMTUD specification for TCP, specified in RFC 4821.¶
The document also provides implementation notes for incorporating Datagram PMTUD into IETF datagram transports or applications that use datagram transports.¶
When published, this specification updates RFC 4821.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 7 June 2020.¶
Copyright (c) 2019 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.¶
The IETF has specified datagram transport using UDP, SCTP, and DCCP, as well as protocols layered on top of these transports (e.g., SCTP/UDP, DCCP/UDP, QUIC/UDP), and direct datagram transport over the IP network layer. This document describes a robust method for Path MTU Discovery (PMTUD) that may be used with these transport protocols (or the applications that use their transport service) to discover an appropriate size of packet to use across an Internet path.¶
Classical Path Maximum Transmission Unit Discovery (PMTUD) can be used with any transport that is able to process ICMP Packet Too Big (PTB) messages (e.g., [RFC1191] and [RFC8201]). In this document, the term PTB message is applied to both IPv4 ICMP Unreachable messages (type 3) that carry the error Fragmentation Needed (Type 3, Code 4) [RFC0792] and ICMPv6 Packet Too Big messages (Type 2) [RFC4443]. When a sender receives a PTB message, it reduces the effective MTU to the value reported as the Link MTU in the PTB message, and a method that from time-to-time increases the packet size in attempt to discover an increase in the supported PMTU. The packets sent with a size larger than the current effective PMTU are known as probe packets.¶
Packets not intended as probe packets are either fragmented to the current effective PMTU, or the attempt to send fails with an error code. Applications are sometimes provided with a primitive to let them read the Maximum Packet Size (MPS), derived from the current effective PMTU.¶
Classical PMTUD is subject to protocol failures. One failure arises when traffic using a packet size larger than the actual PMTU is black-holed (all datagrams sent with this size, or larger, are discarded). This could arise when the PTB messages are not delivered back to the sender for some reason (see for example [RFC2923]).¶
Examples where PTB messages are not delivered include:¶
Another failure could result if a node that is not on the network path sends a PTB message that attempts to force a sender to change the effective PMTU [RFC8201]. A sender can protect itself from reacting to such messages by utilising the quoted packet within a PTB message payload to validate that the received PTB message was generated in response to a packet that had actually originated from the sender. However, there are situations where a sender would be unable to provide this validation. Examples where validation of the PTB message is not possible include:¶
When a router issuing the ICMP message implements RFC792 [RFC0792], it is only required to include the first 64 bits of the IP payload of the packet within the quoted payload. There could be insufficient bytes remaining for the sender to interpret the quoted transport information.¶
Note: The recommendation in RFC1812 [RFC1812] is that IPv4 routers return a quoted packet with as much of the original datagram as possible without the length of the ICMP datagram exceeding 576 bytes. IPv6 routers include as much of the invoking packet as possible without the ICMPv6 packet exceeding 1280 bytes [RFC4443].¶
The term Packetization Layer (PL) has been introduced to describe the layer that is responsible for placing data blocks into the payload of IP packets and selecting an appropriate MPS. This function is often performed by a transport protocol, but can also be performed by other encapsulation methods working above the transport layer.¶
In contrast to PMTUD, Packetization Layer Path MTU Discovery (PLPMTUD) [RFC4821] does not rely upon reception and validation of PTB messages. It is therefore more robust than Classical PMTUD. This has become the recommended approach for implementing PMTU discovery.¶
It uses a general strategy where the PL sends probe packets to search for the largest size of unfragmented datagram that can be sent over a network path. Probe packets are sent with a progressively larger packet size. If a probe packet is successfully delivered (as determined by the PL), then the PLPMTU is raised to the size of the successful probe. If no response is received to a probe packet, the method reduces the probe size. The result of probing with the PLPMTU is used to set the application MPS.¶
PLPMTUD introduces flexibility in the implementation of PMTU discovery. At one extreme, it can be configured to only perform ICMP Black Hole Detection and recovery to increase the robustness of Classical PMTUD, or at the other extreme, all PTB processing can be disabled and PLPMTUD can completely replace Classical PMTUD (see Section 4.5).¶
PLPMTUD can also include additional consistency checks without increasing the risk that data is lost when probing to discover the path MTU. For example, information available at the PL, or higher layers, enables received PTB messages to be validated before being utilized.¶
Section 5 of this document presents a set of algorithms for datagram protocols to discover the largest size of unfragmented datagram that can be sent over a network path. The method described relies on features of the PL described in Section 3 and applies to transport protocols operating over IPv4 and IPv6. It does not require cooperation from the lower layers, although it can utilize PTB messages when these received messages are made available to the PL.¶
The UDP Usage Guidelines [RFC8085] state "an application SHOULD either use the Path MTU information provided by the IP layer or implement Path MTU Discovery (PMTUD)", but does not provide a mechanism for discovering the largest size of unfragmented datagram that can be used on a network path. Prior to this document, PLPMTUD had not been specified for UDP.¶
Section 10.2 of [RFC4821] recommends a PLPMTUD probing method for the Stream Control Transport Protocol (SCTP). SCTP utilizes probe packets consisting of a minimal sized HEARTBEAT chunk bundled with a PAD chunk as defined in [RFC4820], but RFC4821 does not provide a complete specification. The present document provides the details to complete that specification.¶
The Datagram Congestion Control Protocol (DCCP) [RFC4340] requires implementations to support Classical PMTUD and states that a DCCP sender "MUST maintain the MPS allowed for each active DCCP session". It also defines the current congestion control MPS (CCMPS) supported by a network path. This recommends use of PMTUD, and suggests use of control packets (DCCP-Sync) as path probe packets, because they do not risk application data loss. The method defined in this specification could be used with DCCP.¶
Section 6 specifies the method for a set of transports, and provides information to enable the implementation of PLPMTUD with other datagram transports and applications that use datagram transports.¶
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.¶
Other terminology is directly copied from [RFC4821], and the definitions in [RFC1122].¶
A Black Hole is encountered when a sender is unaware that packets are not being delivered to the destination end point. Two types of Black Hole are relevant to DPLPMTUD:¶
TCP PLPMTUD has been defined using standard TCP protocol mechanisms. All of the requirements in [RFC4821] also apply to the use of the technique with a datagram PL. Unlike TCP, some datagram PLs require additional mechanisms to implement PLPMTUD.¶
There are eight requirements for performing the datagram PLPMTUD method described in this specification:¶
In addition, the following principles are stated for design of a DPLPMTUD method:¶
This section lists the protocol mechanisms used in this specification.¶
The DPLPMTUD method relies upon the PL sender being able to generate probe packets with a specific size. TCP is able to generate these probe packets by choosing to appropriately segment data being sent [RFC4821]. In contrast, a datagram PL that needs to construct a probe packet has to either request an application to send a data block that is larger than that generated by an application, or to utilize padding functions to extend a datagram beyond the size of the application data block. Protocols that permit exchange of control messages (without an application data block) MAY prefer to generate a probe packet by extending a control message with padding data.¶
A receiver is REQUIRED to be able to distinguish an in-band data block from any added padding. This is needed to ensure that any added padding is not passed on to an application at the receiver.¶
This results in three possible ways that a sender can create a probe packet:¶
A PL that uses a probe packet carrying an application data block, could need to retransmit this application data block if the probe fails. This could need the PL to re-fragment the data block to a smaller packet size that is expected to traverse the end-to-end path (which could utilize endpoint network-layer or PL fragmentation when these are available).¶
DPLPMTUD MAY choose to use only one of these methods to simplify the implementation.¶
Probe messages sent by a PL MUST contain enough information to uniquely identify the probe within Maximum Segment Lifetime, while being robust to reordering and replay of probe response and PTB messages.¶
The PL needs a method to determine (confirm) when probe packets have been successfully received end-to-end across a network path.¶
Transport protocols can include end-to-end methods that detect and report reception of specific datagrams that they send (e.g., DCCP and SCTP provide keep-alive/heartbeat features). When supported, this mechanism SHOULD also be used by DPLPMTUD to acknowledge reception of a probe packet.¶
A PL that does not acknowledge data reception (e.g., UDP and UDP-Lite) is unable itself to detect when the packets that it sends are discarded because their size is greater than the actual PMTU. These PLs need to either rely on an application protocol to detect this loss.¶
Section 6 specifies this function for a set of IETF-specified protocols.¶
A PL sender needs to reduce the PLPMTU when it discovers the actual PMTU supported by a network path is less than the PLPMTU. This can be triggered when a validated PTB message is received, or by another event that indicates the network path no longer sustains the current packet size, such as a loss report from the PL, or repeated lack of response to probe packets sent to confirm the PLPMTU. Detection is followed by a reduction of the PLPMTU.¶
This is performed by sending packet probes of size PLPMTU to verify that a network path still supports the last acknowledged PLPMTU size. There are two alternative mechanism:¶
A PL MAY inhibit sending probe packets when no application data has been sent since the previous probe packet. A PL preferring to use an up-to-data PLPMTU once user data is sent again, MAY choose to continue PLPMTU discovery for each path. However, this may result in additional packets being sent.¶
When the method detects the current PLPMTU is not supported, DPLPMTUD sets a lower MPS. The PL then confirms that the updated PLPMTU can be successfully used across the path. The PL could need to send a probe packet with a size less than the size of the data block generated by an application. In this case, the PL could provide a way to fragment a datagram at the PL, or use a control packet as the packet probe.¶
A PL implementing this specification MUST suspend network layer processing of outgoing packets that enforces a PMTU [RFC1191][RFC8201] for each flow utilising DPLPMTUD, and instead use DPLPMTUD to control the size of packets that are sent by a flow. This removes the need for the network layer to drop or fragment sent packets that have a size greater than the PMTU.¶
This method requires the DPLPMTUD sender to validate any received PTB message before using the PTB information. The response to a PTB message depends on the PTB_SIZE indicated in the PTB message, the state of the PLPMTUD state machine, and the IP protocol being used.¶
Section 4.5.1 first describes validation for both IPv4 ICMP Unreachable messages (type 3) and ICMPv6 Packet Too Big messages, both of which are referred to as PTB messages in this document.¶
This section specifies utilization of PTB messages.¶
A PL that receives a PTB message from a router or middlebox, performs ICMP validation as specified in Section 5.2 of [RFC8085][RFC8201]. Because DPLPMTUD operates at the PL, the PL needs to check that each received PTB message is received in response to a packet transmitted by the endpoint PL performing DPLPMTUD.¶
The PL MUST check the protocol information in the quoted packet carried in an ICMP PTB message payload to validate the message originated from the sending node. This validation includes determining that the combination of the IP addresses, the protocol, the source port and destination port match those returned in the quoted packet - this is also necessary for the PTB message to be passed to the corresponding PL.¶
The validation SHOULD utilize information that it is not simple for an off-path attacker to determine [RFC8085]. For example, by checking the value of a protocol header field known only to the two PL endpoints. A datagram application that uses well-known source and destination ports ought to also rely on other information to complete this validation.¶
These checks are intended to provide protection from packets that originate from a node that is not on the network path. A PTB message that does not complete the validation MUST NOT be further utilized by the DPLPMTUD method.¶
PTB messages that have been validated MAY be utilized by the DPLPMTUD algorithm, but MUST NOT be used directly to set the PLPMTU. A method that utilizes these PTB messages can improve the speed at the which the algorithm detects an appropriate PLPMTU, compared to one that relies solely on probing. Section 4.5.2 describes this processing.¶
A set of checks are intended to provide protection from a router that reports an unexpected PTB_SIZE. The PL also needs to check that the indicated PTB_SIZE is less than the size used by probe packets and larger than minimum size accepted.¶
This section provides a summary of how PTB messages can be utilized. This processing depends on the PTB_SIZE and the current value of a set of variables:¶
This section specifies Datagram PLPMTUD (DPLPMTUD). The method can be introduced at various points (as indicated with * in the figure below) in the IP protocol stack to discover the PLPMTU so that an application can utilize an appropriate MPS for the current network path. DPLPMTUD SHOULD NOT be used by an application if it is already used in a lower layer.¶
The central idea of DPLPMTUD is probing by a sender. Probe packets are sent to find the maximum size of a user message that can be completely transferred across the network path from the sender to the destination.¶
The following sections identify the components needed for implementation, provides an overview of the phases of operation, and specifies the state machine and search algorithm.¶
This section describes the timers, constants, and variables of DPLPMTUD.¶
The method utilizes up to three timers:¶
The PROBE_TIMER is configured to expire after a period longer than the maximum time to receive an acknowledgment to a probe packet. This value MUST NOT be smaller than 1 second, and SHOULD be larger than 15 seconds. Guidance on selection of the timer value are provided in section 3.1.1 of the UDP Usage Guidelines [RFC8085].¶
If the PL has a path Round Trip Time (RTT) estimate and timely acknowledgements the PROBE_TIMER can be derived from the PL RTT estimate.¶
The PMTU_RAISE_TIMER is configured to the period a sender will continue to use the current PLPMTU, after which it re-enters the Search phase. This timer has a period of 600 seconds, as recommended by PLPMTUD [RFC4821].¶
DPLPMTUD MAY inhibit sending probe packets when no application data has been sent since the previous probe packet. A PL preferring to use an up-to-data PMTU once user data is sent again, can choose to continue PMTU discovery for each path. However, this may result in sending additional packets.¶
When an acknowledged PL is used, this timer MUST NOT be used. For other PLs, the CONFIRMATION_TIMER is configured to the period a PL sender waits before confirming the current PLPMTU is still supported. This is less than the PMTU_RAISE_TIMER and used to decrease the PLPMTU (e.g., when a black hole is encountered). Confirmation needs to be frequent enough when data is flowing that the sending PL does not black hole extensive amounts of traffic. Guidance on selection of the timer value are provided in section 3.1.1 of the UDP Usage Guidelines [RFC8085].¶
DPLPMTUD MAY inhibit sending probe packets when no application data has been sent since the previous probe packet. A PL preferring to use an up-to-data PMTU once user data is sent again, can choose to continue PMTU discovery for each path. However, this may result in sending additional packets.¶
An implementation could implement the various timers using a single timer.¶
The following constants are defined:¶
The MIN_PMTU is the smallest allowed probe packet size. For IPv6, this value is 1280 bytes, as specified in [RFC8200]. For IPv4, the minimum value is 68 bytes.¶
Note: An IPv4 router is required to be able to forward a datagram of 68 bytes without further fragmentation. This is the combined size of an IPv4 header and the minimum fragment size of 8 bytes. In addition, receivers are required to be able to reassemble fragmented datagrams at least up to 576 bytes, as stated in section 3.3.3 of [RFC1122].¶
This method utilizes a set of variables:¶
The figure below illustrates the relationship between the packet size constants and variables at a point of time when the DPLPMTUD algorithm performs path probing to increase the size of the PLPMTU. A probe packet has been sent of size PROBED_SIZE. Once this is acknowledged, the PLPMTU will raise to PROBED_SIZE allowing the DPLPMTUD algorithm to further increase PROBED_SIZE towards the actual PMTU.¶
This section provides a high-level informative view of the DPLPMTUD method, by describing the movement of the method through several phases of operation. More detail is available in the state machine Section 5.2.¶
The Base Phase confirms connectivity to the remote peer. This phase is implicit for a connection-oriented PL (where it can be performed in a PL connection handshake). A connectionless PL needs to send an acknowledged probe packet to confirm that the remote peer is reachable. The sender also confirms that BASE_PMTU is supported across the network path.¶
A PL that does not wish to support a path with a PLPMTU less than BASE_PMTU can simplify the phase into a single step by performing the connectivity checks with a probe of the BASE_PMTU size.¶
Once confirmed, DPLPMTUD enters the Search Phase. If this phase fails to confirm, DPLPMTUD enters the Error Phase.¶
The Search Phase utilizes a search algorithm to send probe packets to seek to increase the PLPMTU. The algorithm concludes when it has found a suitable PLPMTU, by entering the Search Complete Phase.¶
A PL could respond to PTB messages using the PTB to advance or terminate the search, see Section 4.5.¶
The Search Complete Phase is entered when the PLPMTU is supported across the network path. A PL can use a CONFIRMATION_TIMER to periodically repeat a probe packet for the current PLPMTU size. If the sender is unable to confirm reachability (e.g., if the CONFIRMATION_TIMER expires) or the PL signals a lack of reachability, DPLPMTUD enters the Base phase.¶
The PMTU_RAISE_TIMER is used to periodically resume the search phase to discover if the PLPMTU can be raised. Black Hole Detection or receipt of a validated PTB message (see Section 4.5.1) can cause the sender to enter the Base Phase.¶
The Error Phase is entered when there is conflicting or invalid PLPMTU information for the path (e.g. a failure to support the BASE_PMTU) that cause DPLPMTUD to be unable to progress and the PLPMTU is lowered.¶
DPLPMTUD remains in the Error Phase until a consistent view of the path can be discovered and it has also been confirmed that the path supports the BASE_PMTU (or DPLPMTUD is suspended).¶
An implementation that only reduces the PLPMTU to a suitable size would be sufficient to ensure reliable operation, but can be very inefficient when the actual PMTU changes or when the method (for whatever reason) makes a suboptimal choice for the PLPMTU.¶
A full implementation of DPLPMTUD provides an algorithm enabling the DPLPMTUD sender to increase the PLPMTU following a change in the characteristics of the path, such as when a link is reconfigured with a larger MTU, or when there is a change in the set of links traversed by an end-to-end flow (e.g., after a routing or path fail-over decision).¶
A state machine for DPLPMTUD is depicted in Figure 4. If multipath or multihoming is supported, a state machine is needed for each path.¶
Note: Not all changes are not shown to simplify the diagram.¶
The following states are defined:¶
The BASE state is used to confirm that the BASE_PMTU size is supported by the network path and is designed to allow an application to continue working when there are transient reductions in the actual PMTU. It also seeks to avoid long periods where traffic is black holed while searching for a larger PLPMTU.¶
On entry, the PROBED_SIZE is set to the BASE_PMTU size and the PROBE_COUNT is set to zero.¶
Each time a probe packet is sent, the PROBE_TIMER is started. The state is exited when the probe packet is acknowledged, and the PL sender enters the SEARCHING state.¶
The state is also left when the PROBE_COUNT reaches MAX_PROBES or a received PTB message is validated. This causes the PL sender to enter the ERROR state.¶
The SEARCHING state is the main probing state. This state is entered when probing for the BASE_PMTU was successful.¶
Each time a probe packet is acknowledged, the PROBE_COUNT is set to zero, the PLPMTU is set to the PROBED_SIZE and then the PROBED_SIZE is increased using the search algorithm.¶
When a probe packet is sent and not acknowledged within the period of the PROBE_TIMER, the PROBE_COUNT is incremented and a new probe packet is transmitted. The state is exited when the PROBE_COUNT reaches MAX_PROBES, a received PTB message is validated, a probe of size MAX_PMTU is acknowledged, or a black hole is detected.¶
The SEARCH_COMPLETE state indicates a successful end to the SEARCHING state. DPLPMTUD remains in this state until either the PMTU_RAISE_TIMER expires, a received PTB message is validated, or a black hole is detected.¶
When DPLPMTUD uses an unacknowledged PL and is in the SEARCH_COMPLETE state, a CONFIRMATION_TIMER periodically resets the PROBE_COUNT and schedules a probe packet with the size of the PLPMTU. If MAX_PROBES successive PLPMTUD sized probes fail to be acknowledged the method enters the BASE state. When used with an acknowledged PL (e.g., SCTP), DPLPMTUD SHOULD NOT continue to generate PLPMTU probes in this state.¶
The ERROR state represents the case where either the network path is not known to support a PLPMTU of at least the BASE_PMTU size or when there is contradictory information about the network path that would otherwise result in excessive variation in the MPS signalled to the higher layer. The state implements a method to mitigate oscillation in the state-event engine. It signals a conservative value of the MPS to the higher layer by the PL. The state is exited when packet probes no longer detect the error or when the PL indicates that connectivity has been lost.¶
Implementations are permitted to enable endpoint fragmentation if the DPLPMTUD is unable to validate MIN_PMTU within PROBE_COUNT probes. If DPLPMTUD is unable to validate MIN_PMTU the implementation should transition to the DISABLED state.¶
Note: MIN_PMTU may be identical to BASE_PMTU, simplifying the actions in this state.¶
This section describes the algorithms used by DPLPMTUD to search for a larger PLPMTU.¶
Implementations use a search algorithm across the search range to determine whether a larger PLPMTU can be supported across a network path.¶
The method discovers the search range by confirming the minimum PLPMTU and then using the probe method to select a PROBED_SIZE less than or equal to MAX_PMTU. MAX_PMTU is the minimum of the local MTU and EMTU_R (learned from the remote endpoint). The MAX_PMTU MAY be reduced by an application that sets a maximum to the size of datagrams it will send.¶
The PROBE_COUNT is initialized to zero when the first probe with a size greater than or equal to PLPMTUD is sent. A timer is used by the search algorithm to trigger the sending of probe packets of size PROBED_SIZE, larger than the PLPMTU. Each probe packet successfully sent to the remote peer is confirmed by acknowledgement at the PL, see Section 4.1.¶
Each time a probe packet is sent to the destination, the PROBE_TIMER is started. The timer is canceled when the PL receives acknowledgment that the probe packet has been successfully sent across the path Section 4.1. This confirms that the PROBED_SIZE is supported, and the PROBED_SIZE value is then assigned to the PLPMTU. The search algorithm can continue to send subsequent probe packets of an increasing size.¶
If the timer expires before a probe packet is acknowledged, the probe has failed to confirm the PROBED_SIZE. Each time the PROBE_TIMER expires, the PROBE_COUNT is incremented, the PROBE_TIMER is reinitialized, and a new probe of the same size or any other size (determined by the search algorithm) can be sent. The maximum number of consecutive failed probes is configured (MAX_PROBES). If the value of the PROBE_COUNT reaches MAX_PROBES, probing will stop, and the PL sender enters the SEARCH_COMPLETE state.¶
The search algorithm needs to determine a minimum useful gain in PLPMTU. It would not be constructive for a PL sender to attempt to probe for all sizes. This would incur unnecessary load on the path and has the undesirable effect of slowing the time to reach a more optimal MPS. Implementations SHOULD select the set of probe packet sizes to maximize the gain in PLPMTU from each search step.¶
Implementations could optimize the search procedure by selecting step sizes from a table of common PMTU sizes. When selecting the appropriate next size to search, an implementer ought to also consider that there can be common sizes of MPS that applications seek to use, and their could be common sizes of MTU used within the network.¶
A decision to increase the PLPMTU needs to be resilient to the possibility that information learned about the network path is inconsistent. A path is inconsistent, when, for example, probe packets are lost due to other reasons (i.e. not packet size) or due to frequent path changes. Frequent path changes could occur by unexpected "flapping" - where some packets from a flow pass along one path, but other packets follow a different path with different properties.¶
A PL sender is able to detect inconsistency from the sequence of PLPMTU probes that it sends or the sequence of PTB messages that it receives. When inconsistent path information is detected, a PL sender could use an alternate search mode that clamps the offered MPS to a smaller value for a period of time. This avoids unnecessary loss of packets due to MTU limitation.¶
Some paths could be unable to sustain packets of the BASE_PMTU size. To be robust to these paths an implementation could implement the Error State. This allows fallback to a smaller than desired PLPMTU, rather than suffer connectivity failure. This could utilize methods such as endpoint IP fragmentation to enable the PL sender to communicate using packets smaller than the BASE_PMTU.¶
DPLPMTUD requires protocol-specific details to be specified for each PL that is used.¶
The first subsection provides guidance on how to implement the DPLPMTUD method as a part of an application using UDP or UDP-Lite. The guidance also applies to other datagram services that do not include a specific transport protocol (such as a tunnel encapsulation). The following subsections describe how DPLPMTUD can be implemented as a part of the transport service, allowing applications using the service to benefit from discovery of the PLPMTU without themselves needing to implement this method.¶
The current specifications of UDP [RFC0768] and UDP-Lite [RFC3828] do not define a method in the RFC-series that supports PLPMTUD. In particular, the UDP transport does not provide the transport layer features needed to implement datagram PLPMTUD.¶
The DPLPMTUD method can be implemented as a part of an application built directly or indirectly on UDP or UDP-Lite, but relies on higher-layer protocol features to implement the method [RFC8085].¶
Some primitives used by DPLPMTUD might not be available via the Datagram API (e.g., the ability to access the PLPMTU from the IP layer cache, or interpret received PTB messages).¶
In addition, it is desirable that PMTU discovery is not performed by multiple protocol layers. An application SHOULD avoid using DPLPMTUD when the underlying transport system provides this capability. To use common method for managing the PLPMTU has benefits, both in the ability to share state between different processes and opportunities to coordinate probing.¶
An application needs an application-layer protocol mechanism (such as a message acknowledgement method) that solicits a response from a destination endpoint. The method SHOULD allow the sender to check the value returned in the response to provide additional protection from off-path insertion of data [RFC8085], suitable methods include a parameter known only to the two endpoints, such as a session ID or initialized sequence number.¶
An application needs an application-layer protocol mechanism to communicate the response from the destination endpoint. This response may indicate successful reception of the probe across the path, but could also indicate that some (or all packets) have failed to reach the destination.¶
A probe packet that may carry an application data block, but the successful transmission of this data is at risk when used for probing. Some applications may prefer to use a probe packet that does not carry an application data block to avoid disruption to data transfer.¶
An application that does not have other higher-layer information confirming connectivity with the remote peer SHOULD implement a connectivity mechanism using acknowledged probe packets before entering the BASE state.¶
An application that does not have other higher-layer information confirming correct delivery of datagrams SHOULD implement the CONFIRMATION_TIMER to periodically send probe packets while in the SEARCH_COMPLETE state.¶
An application that is able and wishes to receive PTB messages MUST perform ICMP validation as specified in Section 5.2 of [RFC8085]. This requires that the application to check each received PTB messages to validate it is received in response to transmitted traffic and that the reported PTB_SIZE is less than the current probed size (see Section 4.5.2). A validated PTB message MAY be used as input to the DPLPMTUD algorithm, but MUST NOT be used directly to set the PLPMTU.¶
Section 10.2 of [RFC4821] specifies a recommended PLPMTUD probing method for SCTP. It recommends the use of the PAD chunk, defined in [RFC4820] to be attached to a minimum length HEARTBEAT chunk to build a probe packet. This enables probing without affecting the transfer of user messages and without interfering with congestion control. This is preferred to using DATA chunks (with padding as required) as path probes.¶
The base protocol is specified in [RFC4960]. This provides an acknowledged PL. A sender can therefore enter the BASE state as soon as connectivity has been confirmed.¶
Probe packets consist of an SCTP common header followed by a HEARTBEAT chunk and a PAD chunk. The PAD chunk is used to control the length of the probe packet. The HEARTBEAT chunk is used to trigger the sending of a HEARTBEAT ACK chunk. The reception of the HEARTBEAT ACK chunk acknowledges reception of a successful probe.¶
The HEARTBEAT chunk carries a Heartbeat Information parameter which should include, besides the information suggested in [RFC4960], the probe size, which is the size of the complete datagram. The size of the PAD chunk is therefore computed by reducing the probing size by the IPv4 or IPv6 header size, the SCTP common header, the HEARTBEAT request and the PAD chunk header. The payload of the PAD chunk contains arbitrary data.¶
To avoid fragmentation of retransmitted data, probing starts right after the PL handshake, before data is sent. Assuming this behavior (i.e., the PMTU is smaller than or equal to the interface MTU), this process will take a few round trip time periods depending on the number of PMTU sizes probed. The Heartbeat timer can be used to implement the PROBE_TIMER.¶
Since SCTP provides an acknowledged PL, a sender MUST NOT implement the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state.¶
Normal ICMP validation MUST be performed as specified in Appendix C of [RFC4960]. This requires that the first 8 bytes of the SCTP common header are quoted in the payload of the PTB message, which can be the case for ICMPv4 and is normally the case for ICMPv6.¶
When a PTB message has been validated, the PTB_SIZE reported in the PTB message SHOULD be used with the DPLPMTUD algorithm, providing that the reported PTB_SIZE is less than the current probe size (see Section 4.5).¶
The UDP encapsulation of SCTP is specified in [RFC6951].¶
A sender can enter the BASE state as soon as SCTP connectivity has been confirmed.¶
Packet probing can be performed as specified in Section 6.2.1.2. The maximum payload is reduced by 8 bytes, which has to be considered when filling the PAD chunk.¶
Since SCTP provides an acknowledged PL, a sender MUST NOT implement the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state.¶
ICMP validation MUST be performed for PTB messages as specified in Appendix C of [RFC4960]. This requires that the first 8 bytes of the SCTP common header are contained in the PTB message, which can be the case for ICMPv4 (but note the UDP header also consumes a part of the quoted packet header) and is normally the case for ICMPv6. When the validation is completed, the PTB_SIZE indicated in the PTB message SHOULD be used with the DPLPMTUD providing that the reported PTB_SIZE is less than the current probe size.¶
The Datagram Transport Layer Security (DTLS) encapsulation of SCTP is specified in [RFC8261]. It is used for data channels in WebRTC implementations.¶
A sender can enter the BASE state as soon as SCTP connectivity has been confirmed.¶
Packet probing can be done as specified in Section 6.2.1.2.¶
Since SCTP provides an acknowledged PL, a sender MUST NOT implement the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state.¶
It is not possible to perform ICMP validation as specified in [RFC4960], since even if the ICMP message payload contains sufficient information, the reflected SCTP common header would be encrypted. Therefore it is not possible to process PTB messages at the PL.¶
QUIC [I-D.ietf-quic-transport] is a UDP-based transport that provides reception feedback. The UDP payload includes the QUIC packet header, protected payload, and any authentication fields. QUIC depends on a PMTU of at least 1280 bytes.¶
Section 14.1 of [I-D.ietf-quic-transport] describes the path considerations when sending QUIC packets. It recommends the use of PADDING frames to build the probe packet. Pure probe-only packets are constructed with PADDING frames and PING frames to create a padding only packet that will elicit an acknowledgement. Such padding only packets enable probing without affecting the transfer of other QUIC frames.¶
The recommendation for QUIC endpoints implementing DPLPMTUD is that a MPS is maintained for each combination of local and remote IP addresses [I-D.ietf-quic-transport]. If a QUIC endpoint determines that the PMTU between any pair of local and remote IP addresses has fallen below an acceptable MPS, it needs to immediately cease sending QUIC packets on the affected path. This could result in termination of the connection if an alternative path cannot be found [I-D.ietf-quic-transport].¶
The base protocol is specified in [I-D.ietf-quic-transport]. This provides an acknowledged PL. A sender can therefore enter the BASE state as soon as connectivity has been confirmed.¶
A probe packet consists of a QUIC Header and a payload containing PADDING Frames and a PING Frame. PADDING Frames are a single octet (0x00) and several of these can be used to create a probe packet of size PROBED_SIZE. QUIC provides an acknowledged PL, a sender can therefore enter the BASE state as soon as connectivity has been confirmed.¶
The current specification of QUIC sets the following:¶
QUIC provides an acknowledged PL. A sender therefore MUST NOT implement the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state.¶
QUIC operates over the UDP transport, and the guidelines on ICMP validation as specified in Section 5.2 of [RFC8085] therefore apply. In addition to UDP Port validation QUIC can validate an ICMP message by looking for valid Connection IDs in the quoted packet.¶
This work was partially funded by the European Union's Horizon 2020 research and innovation programme under grant agreement No. 644334 (NEAT). The views expressed are solely those of the author(s).¶
Thanks to all that have commented or contributed, the TSVWG and QUIC working groups, and Mathew Calder and Julius Flohr for providing implementations.¶
This memo includes no request to IANA.¶
If there are no requirements for IANA, the section will be removed during conversion into an RFC by the RFC Editor.¶
The security considerations for the use of UDP and SCTP are provided in the references RFCs. The interval between individual probe packets MUST be at least one RTT, and the interval between rounds of probing is determined by the PMTU_RAISE_TIMER.¶
A PL sender needs to ensure that the method used to confirm reception of probe packets offers protection from off-path attackers injecting packets into the path. This protection if provided in IETF-defined protocols (e.g., TCP, SCTP) using a randomly-initialized sequence number. A description of one way to do this when using UDP is provided in section 5.1 of [RFC8085]).¶
There are cases where ICMP Packet Too Big (PTB) messages are not delivered due to policy, configuration or equipment design (see Section 1.1), this method therefore does not rely upon PTB messages being received, but is able to utilize these when they are received by the sender. PTB messages could potentially be used to cause a node to inappropriately reduce the PLPMTU. A node supporting DPLPMTUD MUST therefore appropriately validate the payload of PTB messages to ensure these are received in response to transmitted traffic (i.e., a reported error condition that corresponds to a datagram actually sent by the path layer, see Section 4.5.1).¶
An on-path attacker, able to create a PTB message could forge PTB messages that include a valid quoted IP packet. Such an attack could be used to drive down the PLPMTU. There are two ways this method can be mitigated against such attacks: First, by ensuring that a PL sender never reduces the PLPMTU below the base size, solely in response to receiving a PTB message. This is achieved by first entering the BASE state when such a message is received. Second, the design does not require processing of PTB messages, a PL sender could therefore suspend processing of PTB messages (e.g., in a robustness mode after detecting that subsequent probes actually confirm that a size larger than the PTB_SIZE is supported by a path).¶
Parallel forwarding paths SHOULD be considered. Section 5.4 identifies the need for robustness in the method when the path information may be inconsistent.¶
A node performing DPLPMTUD could experience conflicting information about the size of supported probe packets. This could occur when there are multiple paths are concurrently in use and these exhibit a different PMTU. If not considered, this could result in data being black holed when the PLPMTU is larger than the smallest PMTU across the current paths.¶
Note to RFC-Editor: please remove this entire section prior to publication.¶
Individual draft -00:¶
Individual draft -01:¶
Individual draft -02:¶
Working Group draft -00:¶
Working Group draft -01:¶
Working Group draft -02:¶
Working Group draft -03:¶
Working Group draft -04:¶
Working Group draft -05:¶
Working Group draft -06:¶
Working group draft -07:¶
Working group draft -08:¶
Working group draft -09:¶
Working group draft -10:¶
Working group draft -11:¶
Working group draft -12:¶