| Internet-Draft | Selective RFRAG | March 2020 |
| Thubert | Expires 20 September 2020 | [Page] |
This draft updates RFC 4944 with a simple protocol to recover individual fragments across a route-over mesh network, with a minimal flow control to protect the network against bloat.¶
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 20 September 2020.¶
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.¶
In most Low Power and Lossy Network (LLN) applications, the bulk of the traffic consists of small chunks of data (on the order of a few bytes to a few tens of bytes) at a time. Given that an IEEE Std. 802.15.4 [IEEE.802.15.4] frame can carry a payload of 74 bytes or more, fragmentation is usually not required. However, and though this happens only occasionally, a number of mission critical applications do require the capability to transfer larger chunks of data, for instance to support the firmware upgrade of the LLN nodes or the extraction of logs from LLN nodes.¶
In the former case, the large chunk of data is transferred to the LLN node, whereas in the latter, the large chunk flows away from the LLN node. In both cases, the size can be on the order of 10 kilobytes or more and an end-to-end reliable transport is required.¶
"Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944] defines the original 6LoWPAN datagram fragmentation mechanism for LLNs. One critical issue with this original design is that routing an IPv6 [RFC8200] packet across a route-over mesh requires the reassembly of the packet at each hop. The "6TiSCH Architecture" [I-D.ietf-6tisch-architecture] indicates that this may cause latency along a path and impact critical resources such as memory and battery; to alleviate those undesirable effects it recommends using a 6LoWPAN Fragment Forwarding (6FF) technique .¶
"LLN Minimal Fragment Forwarding" [FRAG-FWD] specifies the generic behavior that all 6FF techniques including this specification follow, and presents the associated caveats. In particular, the routing information is fully indicated in the first fragment, which is always forwarded first. With this specification, the first fragment is identified by a Sequence of 0 as opposed to a dispatch type in [RFC4944]. A state is formed and used to forward all the next fragments along the same path. The Datagram_Tag is locally significant to the Layer-2 source of the packet and is swapped at each hop, more in Section 6. This specification encodes the Datagram_Tag in one byte, which will saturate if more than 256 datagrams transit in fragmented form over a single hop at the same time. This is not realistic at the time of this writing. Should this happen in a new 6LoWPAN technology, a node will need to use several Link-Layer addresses to increase its indexing capacity.¶
"Virtual reassembly buffers in 6LoWPAN" [LWIG-FRAG](VRB) proposes a 6FF technique that is compatible with [RFC4944] without the need to define a new protocol. However, adding that capability alone to the local implementation of the original 6LoWPAN fragmentation would not address the inherent fragility of fragmentation (see [FRAG-ILE]) in particular the issues of resources locked on the reassembling endpoint and the wasted transmissions due to the loss of a single fragment in a whole datagram. [Kent] compares the unreliable delivery of fragments with a mechanism it calls "selective acknowledgements" that recovers the loss of a fragment individually. The paper illustrates the benefits that can be derived from such a method in figures 1, 2 and 3, on pages 6 and 7. [RFC4944] has no selective recovery and the whole datagram fails when one fragment is not delivered to the reassembling endpoint. Constrained memory resources are blocked on the reassembling endpoint until it times out, possibly causing the loss of subsequent packets that cannot be received for the lack of buffers.¶
That problem is exacerbated when forwarding fragments over multiple hops since a loss at an intermediate hop will not be discovered by either the fragmenting and reassembling endpoints, and the source will keep on sending fragments, wasting even more resources in the network since the datagram cannot arrive in its entirety, and possibly contributing to the condition that caused the loss. [RFC4944] is lacking a congestion control to avoid participating in a saturation that may have caused the loss of the fragment. It has no signaling to abort a multi-fragment transmission at any time and from either end, and, if the capability to forward fragments is implemented, clean up the related state in the network.¶
This specification provides a method to forward fragments over typically a few hops in a route-over 6LoWPAN mesh, and a selective acknowledgment to recover individual fragments between 6LoWPAN endpoints. The method can help limit the congestion loss in the network and addresses the requirements in Appendix B. Flow Control is out of scope since the endpoints are expected to be able to store the full datagram. Deployments are expected to be managed and homogeneous, and an incremental transition requires a flag day.¶
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.¶
This document uses 6LoWPAN terms and concepts that are presented in "IPv6 over Low-Power Wireless Personal Area Networks (6LoWPANs): Overview, Assumptions, Problem Statement, and Goals" [RFC4919], "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944], and "Problem Statement and Requirements for IPv6 over Low-Power Wireless Personal Area Network (6LoWPAN) Routing" [RFC6606].¶
"LLN Minimal Fragment Forwarding" [FRAG-FWD] discusses the generic concept of a Virtual Reassembly Buffer (VRB) and specifies behaviors and caveats that are common to a large family of 6FF techniques including the mechanism specified by this document, which fully inherits from that specification. It also defines terms used in this document: Compressed Form, Datagram_Tag, Datagram_Size, Fragment_Offset, and 6LoWPAN Fragment Forwarding endpoint (commonly abbreviated as only "endpoint").¶
Past experience with fragmentation has shown that misassociated or lost fragments can lead to poor network behavior and, occasionally, trouble at the application layer. The reader is encouraged to read "IPv4 Reassembly Errors at High Data Rates" [RFC4963] and follow the references for more information. That experience led to the definition of "Path MTU discovery" [RFC8201] (PMTUD) protocol that limits fragmentation over the Internet. Specifically in the case of UDP, valuable additional information can be found in "UDP Usage Guidelines for Application Designers" [RFC8085].¶
"The Benefits of Using Explicit Congestion Notification (ECN)" [RFC8087] provides useful information on the potential benefits and pitfalls of using ECN.¶
Quoting the "Multiprotocol Label Switching (MPLS) Architecture" [RFC3031]: with MPLS, 'packets are "labeled" before they are forwarded' along a Label Switched Path (LSP). At subsequent hops, there is no further analysis of the packet's network layer header. Rather, the label is used as an index into a table which specifies the next hop, and a new label". [FRAG-FWD] leverages MPLS to forward fragments that actually do not have a network layer header, since the fragmentation occurs below IP, and this specification makes it reversible so the reverse path can be followed as well.¶
This specification uses the following terms:¶
This specification updates the fragmentation mechanism that is specified in "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944] for use in route-over LLNs by providing a model where fragments can be forwarded end-to-end across a 6LoWPAN LLN, and where fragments that are lost on the way can be recovered individually. A new format for fragments is introduced and new dispatch types are defined in Section 5.¶
[RFC8138] allows modifying the size of a packet en route by removing the consumed hops in a compressed Routing Header. This requires that Fragment_Offset and Datagram_Size (see Section 2.3) are also modified en route, which is difficult to do in the uncompressed form. This specification expresses those fields in the Compressed Form and allows modifying them en route (see Section 4.4) easily.¶
Consistently with Section 2 of [RFC6282], for the fragmentation mechanism described in Section 5.3 of [RFC4944], any header that cannot fit within the first fragment MUST NOT be compressed when using the fragmentation mechanism described in this specification.¶
This specification implements the generic 6FF technique defined in "LLN Minimal Fragment Forwarding" [FRAG-FWD], provides end-to-end fragment recovery and congestion control mechanisms.¶
[FRAG-FWD] allows for refragmenting in intermediate nodes, meaning that some bytes from a given fragment may be left in the VRB to be added to the next fragment. The need for more space in the outgoing fragment than was needed for the incoming fragment arises when the 6LoWPAN Header Compression is not as efficient on the outgoing link or the Link MTU is reduced.¶
This specification cannot allow such a refragmentation operation since the fragments are recovered end-to-end based on a sequence number. The Fragment_Size MUST be tailored to fit the minimal MTU along the path, and the first fragment that contains a 6LoWPAN-compressed header MUST have enough slack to enable a less efficient compression in the next hops to still fits within the Link MTU. If the fragmenting endpoint is also the 6LoWPAN compression endpoint, it will elide the IID of the source IPv6 address if it matches the Link-Layer address [RFC6282]. In a network with a consistent MTU, it MUST compute the Fragment_Size as if the MTU was 8 bytes less, so the next hop can expand the IID within the same fragment.¶
[FRAG-FWD] requires that a configurable interval of time is inserted between transmissions to the same next hop and in particular between fragments of a same datagram. In the case of half duplex interfaces, this inter-frame gap ensures that the next hop is done forwarding the previous frame and is capable of receiving the next one.¶
In the case of a mesh operating at a single frequency with omnidirectional antennas, a larger inter-frame gap is required to protect the frame against hidden terminal collisions with the previous frame of the same flow that is still progressing along a common path.¶
The inter-frame gap is useful even for unfragmented datagrams, but it becomes a necessity for fragments that are typically generated in a fast sequence and are all sent over the exact same path.¶
The inter-frame gap is the only protection that [FRAG-FWD] imposes by default. This document enables to group fragments in windows and request intermediate acknowledgements so the number of in-flight fragments can be bounded. This document also adds an ECN mechanism that can be used to to protect the network by adapting the size of the window, the size of the fragments, and/or the inter-frame gap.¶
This specification enables the fragmenting endpoint to apply a congestion control mechanism to tune those parameters, but the mechanism itself is out of scope. In most cases, the expectation is that most datagrams will require only a few fragments, and that only the last fragment will be acknowledged. A basic implementation of the fragmenting endpoint is NOT REQUIRED to vary the size of the window, the duration of the inter-frame gap or the size of a fragment in the middle of the transmission of a datagram, and it MAY ignore the ECN signal or simply reset the window to 1 (see Appendix C for more) until the end of this datagram upon detecting a congestion.¶
An intermediate node that experiences a congestion MAY set the ECN bit in a fragment, and the reassembling endpoint echoes the ECN bit at most once at the next opportunity to acknowledge back.¶
The size of the fragments is typically computed from the Link MTU to maximize the size of the resulting frames. The size of the window and the duration of the inter-frame gap SHOULD be configurable, to reduce the chances of congestion and to follow the general recommendations in [FRAG-FWD], respectively.¶
The compression of the Hop Limit, of the source and destination addresses in the IPv6 Header, and of the Routing Header may change en route in a Route-Over mesh LLN. If the size of the first fragment is modified, then the intermediate node MUST adapt the Datagram_Size, encoded in the Fragment_Size field, to reflect that difference.¶
The intermediate node MUST also save the difference of Datagram_Size of the first fragment in the VRB and add it to the Fragment_Offset of all the subsequent fragments that it forwards for that datagram.¶
This document specifies an alternative to the 6LoWPAN fragmentation sublayer [RFC4944] to emulate an Link MTU up to 2048 bytes for the upper layer, which can be the 6LoWPAN Header Compression sublayer that is defined in the "Compression Format for IPv6 Datagrams" [RFC6282] specification. This specification also provides a reliable transmission of the fragments over a multihop 6LoWPAN route-over mesh network and a minimal congestion control to reduce the chances of congestion loss.¶
A 6LoWPAN Fragment Forwarding [FRAG-FWD] technique derived from MPLS enables the forwarding of individual fragments across a 6LoWPAN route-over mesh without reassembly at each hop. The Datagram_Tag is used as a label; it is locally unique to the node that owns the source Link-Layer address of the fragment, so together the Link-Layer address and the label can identify the fragment globally within the lifetime of the datagram. A node may build the Datagram_Tag in its own locally-significant way, as long as the chosen Datagram_Tag stays unique to the particular datagram for its lifetime. The result is that the label does not need to be globally unique but also that it must be swapped at each hop as the source Link-Layer address changes.¶
In the following sections, a "Datagram_Tag" extends the semantics defined in [RFC4944] Section 5.3."Fragmentation Type and Header". The Datagram_Tag is a locally unique identifier for the datagram from the perspective of the sender. This means that the Datagram_Tag identifies a datagram uniquely in the network when associated with the source of the datagram. As the datagram gets forwarded, the source changes and the Datagram_Tag must be swapped as detailed in [FRAG-FWD].¶
This specification extends RFC 4944 [RFC4944] with 2 new Dispatch types, for Recoverable Fragment (RFRAG) and for the RFRAG Acknowledgment back. The new 6LoWPAN Dispatch types are taken from Page 0 [RFC8025] as indicated in Table 1 in Section 9.¶
In this specification, if the packet is compressed then the size and offset of the fragments are expressed with respect to the Compressed Form of the packet form as opposed to the uncompressed (native) form.¶
The format of the fragment header is shown in Figure 1. It is the same for all fragments though the Fragment_Offset is overloaded. The format has a length and an offset, as well as a Sequence field. This would be redundant if the offset was computed as the product of the Sequence by the length, but this is not the case. The position of a fragment in the reassembly buffer is neither correlated with the value of the Sequence field nor with the order in which the fragments are received. This enables refragmenting to cope with an MTU deduction, see the example of the fragment seq. 5 that is retried end-to-end as smaller fragments seq. 13 and 14 in Section 6.2.¶
The first fragment is recognized by a Sequence of 0; it carries its Fragment_Size and the Datagram_Size of the compressed packet before it is fragmented, whereas the other fragments carry their Fragment_Size and Fragment_Offset. The last fragment for a datagram is recognized when its Fragment_Offset and its Fragment_Size add up to the stored Datagram_Size of the packet identified by the sender Link-Layer address and the Datagram_Tag.¶
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|1 1 1 0 1 0 0|E| Datagram_Tag |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|X| Sequence| Fragment_Size | Fragment_Offset |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
X set == Ack-Request
16-bit unsigned integer.¶
When the Fragment_Offset is set to a non-0 value, its semantics depend on the value of the Sequence field as follows:¶
A Fragment_Offset that is set to a value of 0 indicates an abort condition and all state regarding the datagram should be cleaned up once the processing of the fragment is complete; the processing of the fragment depends on whether there is a VRB already established for this datagram, and the next hop is still reachable:¶
There is no requirement on the reassembling endpoint to check that the received fragments are consecutive and non-overlapping. The fragmenting endpoint knows that the datagram is fully received when the acknowledged fragments cover the whole datagram, which is always the case with a FULL bitmap. This may be useful in particular in the case where the MTU changes and a fragment Sequence is retried with a smaller Fragment_Size, the remainder of the original fragment being retried with new Sequence values.¶
Recoverable Fragments are sequenced and a bitmap is used in the RFRAG Acknowledgment to indicate the received fragments by setting the individual bits that correspond to their sequence.¶
This specification also defines a 4-byte RFRAG Acknowledgment bitmap that is used by the reassembling endpoint to confirm selectively the reception of individual fragments. A given offset in the bitmap maps one-to-one with a given sequence number and indicates which fragment is acknowledged as follows:¶
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| RFRAG Acknowledgment Bitmap |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
^ ^
| | bitmap indicating whether:
| +----- Fragment with Sequence 9 was received
+----------------------- Fragment with Sequence 0 was received
Figure 3 shows an example Acknowledgment bitmap which indicates that all fragments from Sequence 0 to 20 were received, except for fragments 1, 2 and 16 were lost and must be retried.¶
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|1|0|0|1|1|1|1|1|1|1|1|1|1|1|1|1|0|1|1|1|1|0|0|0|0|0|0|0|0|0|0|0|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The RFRAG Acknowledgment Bitmap is included in an RFRAG Acknowledgment header, as follows:¶
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|1 1 1 0 1 0 1|E| Datagram_Tag |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| RFRAG Acknowledgment Bitmap (32 bits) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The Recoverable Fragment header RFRAG is used to transport a fragment and optionally request an RFRAG Acknowledgment RFRAG_ACK that confirms the reception of one or more fragments. An RFRAG_ACK is carried as a standalone fragment header (i.e., with no 6LoWPAN payload) in a message that is propagated back to the fragmenting endpoint. To achieve this, each hop that performed an MPLS-like operation on fragments reverses that operation for the RFRAG_ACK by sending a frame from the next hop to the previous hop as known by its Link-Layer address in the VRB. The Datagram_Tag in the RFRAG_ACK is unique to the reassembling endpoint and is enough information for an intermediate hop to locate the VRB that contains the Datagram_Tag used by the previous hop and the Layer-2 information associated with it (interface and Link-Layer address).¶
The fragmenting endpoint (i.e., the node fragments the packets at the 6LoWPAN level) also controls the number of acknowledgments by setting the Ack-Request flag in the RFRAG packets.¶
The fragmenting endpoint may set the Ack-Request flag on any fragment to perform congestion control by limiting the number of outstanding fragments, which are the fragments that have been sent but for which reception or loss was not positively confirmed by the reassembling endpoint. The maximum number of outstanding fragments is controlled by the Window-Size. It is configurable and may vary in case of ECN notification. When the endpoint that reassembles the packets at the 6LoWPAN level receives a fragment with the Ack-Request flag set, it MUST send an RFRAG_ACK back to the originator to confirm reception of all the fragments it has received so far.¶
The Ack-Request ('X') set in an RFRAG marks the end of a window. This flag MUST be set on the last fragment if the fragmenting endpoint wishes to perform an automatic repeat request (ARQ) process for the datagram, and it MAY be set in any intermediate fragment for the purpose of congestion control.¶
This ARQ process MUST be protected by a Retransmission Time Out (RTO) timer, and the fragment that carries the 'X' flag MAY be retried upon a time out for a configurable number of times (see Section 7.1) with an exponential backoff. Upon exhaustion of the retries the fragmenting endpoint may either abort the transmission of the datagram or resend the first fragment with an 'X' flag set in order to establish a new path for the datagram and obtain the list of fragments that were received over the old path in the acknowledgment bitmap. When the knows that an underlying link-layer mechanism protects the fragments, it may refrain from using the RFRAG Acknowledgment mechanism, and never set the Ack-Request bit.¶
The reassembling endpoint MAY issue unsolicited acknowledgments. An unsolicited acknowledgment signals to the fragmenting endpoint that it can resume sending in case it has reached its maximum number of outstanding fragments. Another use is to inform the fragmenting endpoint that the reassembling endpoint aborted the processing of an individual datagram.¶
The RFRAG Acknowledgment carries an ECN indication for congestion control (see Appendix C). The reassembling endpoint of a fragment with the 'E' (ECN) flag set MUST echo that information at most once by setting the 'E' (ECN) flag in the next RFRAG_ACK.¶
In order to protect the datagram, the fragmenting endpoint transfers a controlled number of fragments and flags the last fragment of a window with an RFRAG Acknowledgment Request. The reassembling endpoint MUST acknowledge a fragment with the acknowledgment request bit set. If any fragment immediately preceding an acknowledgment request is still missing, the reassembling endpoint MAY intentionally delay its acknowledgment to allow in-transit fragments to arrive. Because it might defeat the round-trip time computation, delaying the acknowledgment should be configurable and not enabled by default.¶
When enough fragments are received to cover the whole datagram, the reassembling endpoint reconstructs the packet, passes it to the upper layer, sends an RFRAG_ACK on the reverse path with a FULL bitmap, and arms a short timer, e.g., on the order of an average round-trip time in the network. The FULL bitmap is used as opposed to a bitmap that acknowledges only the received fragments to let the intermediate nodes know that the datagram is fully received. As the timer runs, the reassembling endpoint absorbs the fragments that were still in flight for that datagram without creating a new state, acknowledging the ones that that bear an Ack-Request with an FRAG Acknowledgment and the FULL bitmap. The reassembling endpoint aborts the communication if fragments with matching source and Datagram-Tag continue to be received after the timer expires.¶
Note that acknowledgments might consume precious resources so the use of unsolicited acknowledgments SHOULD be configurable and not enabled by default.¶
An observation is that streamlining forwarding of fragments generally reduces the latency over the LLN mesh, providing room for retries within existing upper-layer reliability mechanisms. The fragmenting endpoint protects the transmission over the LLN mesh with a retry timer that is configured for a use case and may be adapted dynamically, e.g., according to the method detailed in [RFC6298]. It is expected that the upper layer retries obey the recommendations in [RFC8085], in which case a single round of fragment recovery should fit within the upper layer recovery timers.¶
Fragments MUST be sent in a round-robin fashion: the sender MUST send all the fragments for a first time before it retries any lost fragment; lost fragments MUST be retried in sequence, oldest first. This mechanism enables the receiver to acknowledge fragments that were delayed in the network before they are retried.¶
When a single radio frequency is used by contiguous hops, the fragmenting endpoint SHOULD insert a delay between the frames (e.g., carrying fragments) that are sent to the same next hop. The delay SHOULD cover multiple transmissions so as to let a frame progress a few hops and avoid hidden terminal issues. This precaution is not required on channel hopping technologies such as Time Slotted Channel Hopping (TSCH) [RFC6554], where nodes that communicate at Layer-2 are scheduled to send and receive respectively, and different hops operate on different channels.¶
This specification inherits from [FRAG-FWD] and proposes a Virtual Reassembly technique to forward fragments with no intermediate reconstruction of the entire datagram.¶
The IPv6 Header MUST be placed in full in the first fragment to enable the routing decision. The first fragment is routed and creates an LSP from the fragmenting endpoint to the reassembling endpoint. The next fragments are label-switched along that LSP. As a consequence, the next fragments can only follow the path that was set up by the first fragment and cannot follow an alternate route. The Datagram_Tag is used to carry the label, which is swapped in each hop.¶
If the first fragment is too large for the path MTU, it will repeatedly fail and never establish an LSP. In that case, the fragmenting endpoint MAY retry the same datagram with a smaller Fragment_Size, in which case it MUST abort the original attempt and use a new Datagram_Tag for the new attempt.¶
In Route-Over mode, the source and destination Link-Layer addresses in a frame change at each hop. The label that is formed and placed in the Datagram_Tag by the sender is associated with the source Link-Layer address and only valid (and temporarily unique) for that source Link-Layer address.¶
Upon receiving the first fragment (i.e., with a Sequence of 0), an intermediate router creates a VRB and the associated LSP state indexed by the incoming interface, the previous-hop Link-Layer address, and the Datagram_Tag, and forwards the fragment along the IPv6 route that matches the destination IPv6 address in the IPv6 header until it reaches the reassembling endpoint, as prescribed by [FRAG-FWD]. The LSP state enables to match the next incoming fragments of a datagram to the abstract forwarding information of next interface, source and next-hop Link-Layer addresses, and swapped Datagram_Tag.¶
In addition, the router also forms a reverse LSP state indexed by the interface to the next hop, the Link-Layer address the router uses as source for that datagram, and the swapped Datagram_Tag. This reverse LSP state enables matching the tuple (interface, destination Link-Layer address, Datagram_Tag) found in an RFRAG_ACK to the abstract forwarding information (previous interface, previous Link-Layer address, Datagram_Tag) used to forward the RFRAG-ACK back to the fragmenting endpoint.¶
Upon receiving the next fragment (i.e., with a non-zero Sequence), an intermediate router looks up a LSP indexed by the tuple (incoming interface, previous-hop Link-Layer address, Datagram_Tag) found in the fragment. If it is found, the router forwards the fragment using the associated VRB as prescribed by [FRAG-FWD].¶
If the VRB for the tuple is not found, the router builds an RFRAG-ACK to abort the transmission of the packet. The resulting message has the following information:¶
At this point the router is all set and can send the RFRAG-ACK back to the previous router. The RFRAG-ACK should normally be forwarded all the way to the source using the reverse LSP state in the VRBs in the intermediate routers as described in the next section.¶
[FRAG-FWD] indicates that the reassembling endpoint stores "the actual packet data from the fragments received so far, in a form that makes it possible to detect when the whole packet has been received and can be processed or forwarded". How this is computed is implementation specific but relies on receiving all the bytes up to the Datagram_Size indicated in the first fragment. An implementation may receive overlapping fragments as the result of retries after an MTU change.¶
Upon receipt of an RFRAG-ACK, the router looks up a reverse LSP indexed by the interface and destination Link-Layer address of the received frame and the received Datagram_Tag in the RFRAG-ACK. If it is found, the router forwards the fragment using the associated VRB as prescribed by [FRAG-FWD], but using the reverse LSP so that the RFRAG-ACK flows back to the fragmenting endpoint.¶
If the reverse LSP is not found, the router MUST silently drop the RFRAG-ACK message.¶
Either way, if the RFRAG-ACK indicates that the fragment was entirely received (FULL bitmap), it arms a short timer, and upon timeout, the VRB and all the associated state are destroyed. Until the timer elapses, fragments of that datagram may still be received, e.g. if the RFRAG-ACK was lost on the path back and the source retried the last fragment. In that case, the router generates an RFRAG-ACK with a FULL bitmap back to the fragmenting endpoint if an acknowledgement was requested, else it silently drops the fragment.¶
This specification does not provide a method to discover the number of hops or the minimal value of MTU along those hops. In a typical case, the MTU is constant and the same across the network. But should the minimal MTU along the path decrease, it is possible to retry a long fragment (say Sequence of 5) with several shorter fragments with a Sequence that was not used before (e.g., 13 and 14). Fragment 5 is marked as abandoned and will not be retried anymore. Note that when this mechanism is in place, it is hard to predict the total number of fragments that will be needed or the final shape of the bitmap that would cover the whole packet. This is why the FULL bitmap is used when the reassembling endpoint gets the whole datagram regardless of which fragments were actually used to do so. Intermediate nodes will unabiguously know that the process is complete. Note that Path MTU Discovery is out of scope for this document.¶
A reset is signaled on the forward path with a pseudo fragment that has the Fragment_Offset set to 0. The sender of a reset SHOULD also set the Sequence and Fragment_Size field to 0.¶
When the fragmenting endpoint or a router on the path decides that a packet should be dropped and the fragmentation process aborted, it generates a reset pseudo fragment and forwards it down the fragment path.¶
Each router next along the path the way forwards the pseudo fragment based on the VRB state. If an acknowledgment is not requested, the VRB and all associated state are destroyed.¶
Upon reception of the pseudo fragment, the reassembling endpoint cleans up all resources for the packet associated with the Datagram_Tag. If an acknowledgment is requested, the reassembling endpoint responds with a NULL bitmap.¶
The other way around, the reassembling endpoint might need to abort the processing of a fragmented packet for internal reasons, for instance if it is out of reassembly buffers, already uses all 256 possible values of the Datagram_Tag, or if it keeps receiving fragments beyond a reasonable time while it considers that this packet is already fully reassembled and was passed to the upper layer. In that case, the reassembling endpoint SHOULD indicate so to the fragmenting endpoint with a NULL bitmap in an RFRAG_ACK.¶
The RFRAG_ACK is forwarded all the way back to the source of the packet and cleans up all resources on the path. Upon an acknowledgment with a NULL bitmap, the fragmenting endpoint MUST abort the transmission of the fragmented datagram with one exception: In the particular case of the first fragment, it MAY decide to retry via an alternate next hop instead.¶
The text above can be read with the assumption of a serial path between a source and a destination. Section 4.5.3 of the "6TiSCH Architecture" [I-D.ietf-6tisch-architecture] defines the concept of a Track that can be a complex path between a source and a destination with Packet ARQ, Replication, Elimination and Overhearing (PAREO) along the Track. This specification can be used along any subset of the complex Track where the first fragment is flooded. The last RFRAG Acknowledgment is flooded on that same subset in the reverse direction. Intermediate RFRAG Acknowledgments can be flooded on any sub-subset of that reverse subset that reach back to the source.¶
This specification extends "On Forwarding 6LoWPAN Fragments over a Multihop IPv6 Network" [FRAG-FWD] and requires the same parameters in the reassembling endpoint and on intermediate nodes. There is no new parameter as echoing ECN is always on. These parameters typically include the reassembly timeout at the reassembling endpoint and an inactivity clean-up timer on the intermediate nodes, and the number of messages that can be processed in parallel in all nodes.¶
The configuration settings introduced by this specification only apply to the fragmenting endpoint, which is in full control of the transmission. LLNs vary a lot in size (there can be thousands of nodes in a mesh), in speed (from 10 Kbps to several Mbps at the PHY layer), in traffic density, and in optimizations that are desired (e.g., the selection of a RPL [RFC6550] Objective Function [RFC6552] impacts the shape of the routing graph).¶
For that reason, only a very generic guidance can be given on the settings of the fragmenting endpoint and on whether complex algorithms are needed to perform congestion control or estimate the round-trip time. To cover the most complex use cases, this specification enables the fragmenting endpoint to vary the fragment size, the window size, and the inter-frame gap, based on the number of losses, the observed variations of the round-trip time and the setting of the ECN bit.¶
The management system SHOULD be capable of providing the parameters listed in this section and an implementation MUST abide by those parameters and in particular never exceed the minimum and maximum configured boundaries.¶
An implementation must control the rate at which it sends packets over the same path to allow the next hop to forward a packet before it gets the next. In a wireless network that uses the same frequency along a path, more time must be inserted to avoid hidden terminal issues between fragments (more in Section 4.2). An implementation should consider the generic recommendations from the IETF in the matter of congestion control and rate management in [RFC5033]. An implementation may perform a congestion control by using a dynamic value of the window size (Window_Size), adapting the fragment size (Fragment_Size), and may reduce the load by inserting an inter-frame gap that is longer than necessary. In a large network where nodes contend for the bandwidth, a larger Fragment_Size consumes less bandwidth but also reduces fluidity and incurs higher chances of loss in transmission.¶
This is controlled by the following parameters:¶
The Window_Size MUST be at least 1 and less than 33.¶
Otherwise:¶
An implementation may perform its estimate of the RTO or use a configured one. The ARQ process is controlled by the following parameters:¶
An implementation may be capable of performing flow control based on ECN; see in Appendix C. This is controlled by the following parameter:¶
The management system should monitor the number of retries and of ECN settings that can be observed from the perspective of both the fragmenting endpoint and the reassembling endpoint with regards to the other endpoint. It may then tune the optimum size of Fragment_Size and of Window_Size, OptFragmentSize, and OptWindowSize, respectively, at the fragmenting endpoint towards a particular reassembling endpoint, applicable to the next datagrams. It will preferably tune the inter-frame gap to increase the spacing between fragments of the same datagram and reduce the reduce the buffer bloat in intermediate node that holds one or more fragments of that datagram.¶
This document specifies an instantiation of a 6FF technique and inherits from the generic description in [FRAG-FWD]. The considerations in the Security Section of [FRAG-FWD] equally apply to this document.¶
In addition to the threats detailed therein, an attacker that is on-path can prematurely end the transmission of a datagram by sending a RFRAG Acknowledgment to the fragmenting endpoint. It can also cause extra transmissions of fragments by resetting bits in the RFRAG Acknowledgment bitmap, and of RFRAG Acknowledgments by forcing the Ack-Request bit in fragments that it forwards.¶
As indicated in [FRAG-FWD], Secure joining and the Link-Layer security are REQUIRED to protect against those attacks, as the fragmentation protocol does not include any native security mechanisms.¶
This specification does not recommend a particular algorithm for the estimation of the duration of the RTO that covers the detection of the loss of a fragment with the 'X' flag set; regardless, an attacker on the path may slow down or discard packets, which in turn can affect the throughput of fragmented packets.¶
Compared to "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944], this specification reduces the Datagram_Tag to 8 bits and the tag wraps faster than with [RFC4944]. But for a constrained network where a node is expected to be able to hold only one or a few large packets in memory, 256 is still a large number. Also, the acknowledgement mechanism allows cleaning up the state rapidly once the packet is fully transmitted or aborted.¶
The abstract Virtual Recovery Buffer inherited from [FRAG-FWD] may be used to perform a Denial-of-Service (DoS) attack against the intermediate Routers since the routers need to maintain a state per flow. The particular VRB implementation technique described in [LWIG-FRAG] allows realigning which data goes in which fragment, which causes the intermediate node to store a portion of the data, which adds an attack vector that is not present with this specification. With this specification, the data that is transported in each fragment is conserved and the state to keep does not include any data that would not fit in the previous fragment.¶
This document allocates 2 patterns for a total of 4 dispatch values in Page 0 for recoverable fragments from the "Dispatch Type Field" registry that was created by "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944] and reformatted by "6LoWPAN Paging Dispatch" [RFC8025].¶
The suggested patterns (to be confirmed by IANA) are indicated in Table 1.¶
| Bit Pattern | Page | Header Type | Reference |
| 11 10100x | 0 | RFRAG - Recoverable Fragment | THIS RFC |
| 11 10100x | 1-14 | Unassigned | |
| 11 10100x | 15 | Reserved for Experimental Use | RFC 8025 |
| 11 10101x | 0 | RFRAG-ACK - RFRAG Acknowledgment | THIS RFC |
| 11 10101x | 1-14 | Unassigned | |
| 11 10101x | 15 | Reserved for Experimental Use | RFC 8025 |
The author wishes to thank Michel Veillette, Dario Tedeschi, Laurent Toutain, Carles Gomez Montenegro, Thomas Watteyne, and Michael Richardson for in-depth reviews and comments. Also many thanks to Roman Danyliw, Peter Yee, Colin Perkins, Tirumaleswar Reddy Konda, Eric Vyncke, Warren Kumari, Magnus Westerlund, Erik Nordmark, and especially Benjamin Kaduk and Mirja Kuhlewind for their careful reviews and for helping through the IETF Last Call and IESG review process, and to Jonathan Hui, Jay Werb, Christos Polyzois, Soumitri Kolavennu, Pat Kinney, Margaret Wasserman, Richard Kelsey, Carsten Bormann, and Harry Courtice for their various contributions in the long process that lead to this document.¶
There are a number of uses for large packets in Wireless Sensor Networks. Such usages may not be the most typical or represent the largest amount of traffic over the LLN; however, the associated functionality can be critical enough to justify extra care for ensuring effective transport of large packets across the LLN.¶
The list of those usages includes:¶
Uncontrolled firmware download or waveform upload can easily result in a massive increase of the traffic and saturate the network.¶
When a fragment is lost in transmission, the lack of recovery in the original fragmentation system of RFC 4944 implies that all fragments would need to be resent, further contributing to the congestion that caused the initial loss, and potentially leading to congestion collapse.¶
This saturation may lead to excessive radio interference, or random early discard (leaky bucket) in relaying nodes. Additional queuing and memory congestion may result while waiting for a low power next hop to emerge from its sleeping state.¶
Considering that RFC 4944 defines an MTU is 1280 bytes and that in most incarnations (except 802.15.4g) a IEEE Std. 802.15.4 frame can limit the Link-Layer payload to as few as 74 bytes, a packet might be fragmented into at least 18 fragments at the 6LoWPAN shim layer. Taking into account the worst-case header overhead for 6LoWPAN Fragmentation and Mesh Addressing headers will increase the number of required fragments to around 32. This level of fragmentation is much higher than that traditionally experienced over the Internet with IPv4 fragments. At the same time, the use of radios increases the probability of transmission loss and Mesh-Under techniques compound that risk over multiple hops.¶
Mechanisms such as TCP or application-layer segmentation could be used to support end-to-end reliable transport. One option to support bulk data transfer over a frame-size-constrained LLN is to set the Maximum Segment Size to fit within the link maximum frame size. Doing so, however, can add significant header overhead to each 802.15.4 frame and cause extraneous acknowledgements across the LLN compared to the method in this specification.¶
For one-hop communications, a number of Low Power and Lossy Network (LLN) link-layers propose a local acknowledgment mechanism that is enough to detect and recover the loss of fragments. In a multihop environment, an end-to-end fragment recovery mechanism might be a good complement to a hop-by-hop MAC recovery. This draft introduces a simple protocol to recover individual fragments between 6FF endpoints that may be multiple hops away.¶
The method addresses the following requirements of an LLN:¶
Considering that a multi-hop LLN can be a very sensitive environment due to the limited queuing capabilities of a large population of its nodes, this draft recommends a simple and conservative approach to Congestion Control, based on TCP congestion avoidance.¶
Congestion on the forward path is assumed in case of packet loss, and packet loss is assumed upon time out. The draft allows controlling the number of outstanding fragments that have been transmitted but for which an acknowledgment was not received yet and are still covered by the ARQ timer.¶
Congestion on the forward path can also be indicated by an Explicit Congestion Notification (ECN) mechanism. Though whether and how ECN [RFC3168] is carried out over the LoWPAN is out of scope, this draft provides a way for the destination endpoint to echo an ECN indication back to the fragmenting endpoint in an acknowledgment message as represented in Figure 4 in Section 5.2. While the support of echoing the ECN at the reassembling endpoint is mandatory, this specification only provides a minimalistic behaviour on the fragmenting endpoint, that is to reset the window to 1 so the fragments are sent and acknowledged one by one till the end of the datagram.¶
It must be noted that congestion and collision are different topics. In particular, when a mesh operates on the same channel over multiple hops, then the forwarding of a fragment over a certain hop may collide with the forwarding of the next fragment that is following over a previous hop but in the same interference domain. This draft enables end-to-end flow control, but leaves it to the fragmenting endpoint stack to pace individual fragments within a transmit window, so that a given fragment is sent only when the previous fragment has had a chance to progress beyond the interference domain of this hop. In the case of 6TiSCH [I-D.ietf-6tisch-architecture], which operates over the TimeSlotted Channel Hopping [RFC7554] (TSCH) mode of operation of IEEE802.14.5, a fragment is forwarded over a different channel at a different time and it makes full sense to transmit the next fragment as soon as the previous fragment has had its chance to be forwarded at the next hop.¶
From the standpoint of a source 6LoWPAN endpoint, an outstanding fragment is a fragment that was sent but for which no explicit acknowledgment was received yet. This means that the fragment might be on the path, received but not yet acknowledged, or the acknowledgment might be on the path back. It is also possible that either the fragment or the acknowledgment was lost on the way.¶
From the fragmenting endpoint standpoint, all outstanding fragments might still be in the network and contribute to its congestion. There is an assumption, though, that after a certain amount of time, a frame is either received or lost, so it is not causing congestion anymore. This amount of time can be estimated based on the round-trip time between the 6LoWPAN endpoints. For the lack of a more adapted technique, the method detailed in "Computing TCP's Retransmission Timer" [RFC6298] may be used for that computation.¶
The reader is encouraged to read through "Congestion Control Principles" [RFC2914]. Additionally [RFC7567] and [RFC5681] provide deeper information on why this mechanism is needed and how TCP handles Congestion Control. Basically, the goal here is to manage the number of fragments present in the network; this is achieved by to reducing the number of outstanding fragments over a congested path by throttling the sources.¶