| lpwan Working Group | A. Minaburo |
| Internet-Draft | Acklio |
| Intended status: Informational | L. Toutain |
| Expires: September 11, 2017 | IMT-Atlantique |
| C. Gomez | |
| Universitat Politècnica de Catalunya | |
| March 10, 2017 |
LPWAN Static Context Header Compression (SCHC) and fragmentation for IPv6 and UDP
draft-ietf-lpwan-ipv6-static-context-hc-02
This document describes a header compression scheme and fragmentation functionality for IPv6/UDP protocols. These techniques are especially tailored for LPWAN (Low Power Wide Area Network) networks and could be extended to other protocol stacks.
The Static Context Header Compression (SCHC) offers a great level of flexibility when processing the header fields. Static context means that information stored in the context which, describes field values, does not change during the packet transmission, avoiding complex resynchronization mechanisms, incompatible with LPWAN characteristics. In most of the cases, IPv6/UDP headers are reduced to a small identifier.
This document describes the generic compression/decompression process and applies it to IPv6/UDP headers. Similar mechanisms for other protocols such as CoAP will be described in a separate document. Moreover, this document specifies fragmentation and reassembly mechanims for SCHC compressed packets exceeding the L2 pdu size and for the case where the SCHC compression is not possible then the IPv6/UDP packet is sent.
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 http://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 September 11, 2017.
Copyright (c) 2017 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 (http://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.
Header compression is mandatory to efficiently bring Internet connectivity to the node within a LPWAN network [I-D.minaburo-lp-wan-gap-analysis].
Some LPWAN networks properties can be exploited for an efficient header compression:
The Static Context Header Compression (SCHC) is defined for this environment. SCHC uses a context where header information is kept in order, this context is static the values on the header fields do not change during time, avoiding complex resynchronization mechanisms, incompatible with LPWAN characteristics. In most of the cases, IPv6/UDP headers are reduced to a small context identifier.
The SCHC header compression is indedependent of the specific LPWAN technology over which it will be used.
On the other hand, LPWAN technologies are characterized, among others, by a very reduced data unit and/or payload size [I-D.ietf-lpwan-overview]. However, some of these technologies do not support layer two fragmentation, therefore the only option for these to support IPv6 when header compression is not possible (and, in particular, its MTU requirement of 1280 bytes [RFC2460]) is the use of fragmentation mechanism at the adaptation layer below IPv6. This specification defines fragmentation functionality to support the IPv6 MTU requirements over LPWAN technologies.
This section defines the terminology and aconyms used in this document.
Static Context Header Compression (SCHC) avoids context synchronization, which is the most bandwidth-consuming operation in other header compression mechanisms such as RoHC. Based on the fact that the nature of data flows is highly predictable in LPWAN networks, a static context may be stored on the End-System (ES). The context must be stored in both ends. It can also be learned by using a provisionning protocol that is out of the scope of this draft.
End-System Appl Servers
+-----------------+ +---------------+
| APP1 APP2 APP3 | |APP1 APP2 APP3|
| | | |
| UDP | | UDP |
| IPv6 | | IPv6 |
| | | |
| LC (contxt)| | |
+--------+--------+ +-------+-------+
| +--+ +--+ +-----------+ .
+~~ |RG| === |NG| === |LC (contxt)| ... Internet ...
+--+ +--+ +-----+-----+
Figure 1: Architecture
Figure 1 based on [I-D.ietf-lpwan-overview] terminology represents the architecture for compression/decompression. The Thing or End-System is running applications which produce IPv6 or IPv6/UDP flows. These flows are compressed by a LPWAN Compressor (LC) to reduce the headers size. Resulting information is sent on a layer two (L2) frame to the LPWAN Radio Network to a Radio Gateway (RG) which forwards the frame to a Network Gateway. The Network Gateway sends the data to a LC for decompression which shares the same rules with the ES. The LC can be located on the Network Gateway or in another places if a tunnel is established between the NG and the LC. This architecture forms a star topology. After decompression, the packet can be sent on the Internet to one or several LPWAN Application Servers (LA).
The principle is exactly the same in the other direction.
The context contains a list of rules (cf. Figure 2). Each rule contains itself a list of fields descriptions composed of a field identifier (FID), a target value (TV), a matching operator (MO) and a Compression/Decompression Function (CDF).
+-----------------------------------------------------------------+ | Rule N | +----------------------------------------------------------------+ | | Rule i | | +---------------------------------------------------------------+ | | | Rule 1 | | | |+--------+--------------+-------------------+-----------------+| | | ||Field 1 | Target Value | Matching Operator | Comp/Decomp Fct || | | |+--------+--------------+-------------------+-----------------+| | | ||Field 2 | Target Value | Matching Operator | Comp/Decomp Fct || | | |+--------+--------------+-------------------+-----------------+| | | ||... | ... | ... | ... || | | |+--------+--------------+-------------------+-----------------+| |-+ ||Field N | Target Value | Matching Operator | Comp/Decomp Fct || | |+--------+--------------+-------------------+-----------------+|-+ | | +---------------------------------------------------------------+
Figure 2: Compression Decompression Context
The rule does not describe the original packet format which must be known from the compressor/decompressor. The rule just describes the compression/decompression behavior for the header fields. In the rule, it is recommended to describe the header field in the same order they appear in the packet.
The main idea of the compression scheme is to send the rule id to the other end instead of known field values. When a value is known by both ends, it is not necessary to send it on the LPWAN network.
The field description is composed of different entries:
Rule IDs are sent between both compression/decompression elements. The size of the rule ID is not specified in this document and can vary regarding the LPWAN technology, the number of flows,…
Some values in the rule ID space may be reserved for goals other than header compression, for example fragmentation.
Rule IDs are specific to an ES. Two ESs may use the same rule ID for different header compression. The LC needs to combine the rule ID with the ES L2 address to find the appropriate rule.
The compression/decompression process follows several steps:
This document describes basic matching operators (MO)s which must be known by both LC, endpoints involved in the header compression/decompression. They are not typed and can be applied indifferently to integer, string or any other type. The MOs and their definition are provided next:
Matching Operators may need a list of parameters to proceed to the matching. For instance MSB requires an integer indicating the number of bits to test.
The Compression Decompression Functions (CDF) describes the action taken during the compression of headers fields, and inversely, the action taken by the decompressor to restore the original value.
/--------------------+-------------+---------------------------\ | Function | Compression | Decompression | | | | | +--------------------+-------------+---------------------------+ |not-sent |elided |use value stored in ctxt | |value-sent |send |build from received value | |LSB(length) |send LSB |ctxt value OR rcvd value | |compute-length |elided |compute length | |compute-checksum |elided |compute UDP checksum | |ESiid-DID |elided |build IID from L2 ES addr | |LAiid-DID |elided |build IID from L2 LA addr | |mapping-sent |send index |value from index on a table| \--------------------+-------------+---------------------------/
Figure 3: Compression and Decompression Functions
Figure 3 sumarizes the functions defined to compress and decompress a field. The first column gives the function’s name. The second and third columns outlines the compression/decompression behavior.
Compression is done in the rule order and compressed values are sent in that order in the compressed message. The receiver must be able to find the size of each compressed field which can be given by the rule or may be sent with the compressed header.
Not-sent function is generally used when the field value is specified in the rule and therefore known by the both Compressor and Decompressor. This function is generally used with the “equal” MO. If MO is “ignore”, there is a risk to have a decompressed field value different from the compressed field.
The compressor does not send any value on the compressed header for that field on which compression is applied.
The decompressor restores the field value with the target value stored in the matched rule.
The value-sent function is generally used when the field value is not known by both Compressor and Decompressor. The value is sent in the compressed message header. Both Compressor and Decompressor must know the size of the field, either implicitely (the size is known by both sides) or explicitely in the compressed header field by indicating the length. This function is generally used with the “ignore” MO.
The compressor sends the Target Value stored on the rule in the compressed header message. The decompressor restores the field value with the one received from the LPWAN
LSB function is used to send a fixed part of the packet field header to the other end. This function is used together with the “MSB” MO
The compressor sends the “length” Least Significant Bits. The decompressor combines with an OR operator the value received with the Target Value.
These functions are used to process respectively the End System and the LA Device Identifier (DID).
The IID value is computed from the device ID present in the Layer 2 header. The computation depends on the technology and the device ID size.
mapping-sent is used to send a smaller index associated to the field value in the Target Value. This function is used together with the “match-mapping” MO.
The compressor looks in the TV to find the field value and send the corresponding index. The decompressor uses this index to restore the field value.
These functions are used by the decompressor to compute the compressed field value based on received information. Compressed fields are elided during the compression and reconstructed during the decompression.
This section lists the different IPv6 and UDP header fields and how they can be compressed.
This field always holds the same value, therefore the TV is 6, the MO is “equal” and the CDF “not-sent”.
If the DiffServ field identified by the rest of the rule do not vary and is known by both sides, the TV should contain this wellknown value, the MO should be “equal” and the CDF must be “not-sent.
If the DiffServ field identified by the rest of the rule varies over time or is not known by both sides, then there are two possibilities depending on the variability of the value, the first one there is without compression and the original value is sent, or the sencond where the values can be computed by sending only the LSB bits:
If the Flow Label field identified by the rest of the rule does not vary and is known by both sides, the TV should contain this well-known value, the MO should be “equal” and the CDF should be “not-sent”.
If the Flow Label field identified by the rest of the rule varies during time or is not known by both sides, there are two possibilities dpending on the variability of the value, the first one is without compression and then the value is sent and the second where only part of the value is sent and the decompressor needs to compute the original value:
If the LPWAN technology does not add padding, this field can be elided for the transmission on the LPWAN network. The LC recompute the original payload length value. The TV is not set, the MO is set to “ignore” and the CDF is “compute-IPv6-length”.
If the payload is small, the TV can be set to 0x0000, the MO set to “MSB (16-s)” and the CDF to “LSB (s)”. The ‘s’ parameter depends on the maximum packet length.
On other cases, the payload length field must be sent and the CDF is replaced by “value-sent”.
If the Next Header field identified by the rest of the rule does not vary and is known by both sides, the TV should contain this Next Header value, the MO should be “equal” and the CDF should be “not-sent”.
If the Next header field identified by the rest of the rule varies during time or is not known by both sides, then TV is not set, MO is set to “ignore” and CDF is set to “value-sent”.
The End System is generally a host and does not forward packets, therefore the Hop Limit value is constant. So the TV is set with a default value, the MO is set to “equal” and the CDF is set to “not-sent”.
Otherwise the value is sent on the LPWAN: TV is not set, MO is set to ignore and CDF is set to “value-sent”.
As in 6LoWPAN [RFC4944], IPv6 addresses are split into two 64-bit long fields; one for the prefix and one for the Interface Identifier (IID). These fields should be compressed. To allow a single rule, these values are identified by their role (ES or LA) and not by their position in the frame (source or destination). The LC must be aware of the traffic direction (upstream, downstream) to select the appropriate field.
Both ends must be synchronized with the appropriate prefixes. For a specific flow, the source and destination prefix can be unique and stored in the context. It can be either a link-local prefix or a global prefix. In that case, the TV for the source and destination prefixes contains the values, the MO is set to “equal” and the CDF is set to “not-sent”.
In case the rule allows several prefixes, static mapping must be used. The different prefixes are listed in the TV associated with a short ID. The MO is set to “match-mapping” and the CDF is set to “mapping-sent”.
Otherwise the TV contains the prefix, the MO is set to “equal” and the CDF is set to value-sent.
If the ES or LA IID are based on an LPWAN address, then the IID can be reconstructed with information coming from the LPWAN header. In that case, the TV is not set, the MO is set to “ignore” and the CDF is set to “ESiid-DID” or “LAiid-DID”. Note that the LPWAN technology is generally carrying a single device identifier corresponding to the ES. The LC may also not be aware of these values.
For privacy reasons or if the ES address is changing over time, it maybe better to use a static value. In that case, the TV contains the value, the MO operator is set to “equal” and the CDF is set to “not-sent”.
If several IIDs are possible, then the TV contains the list of possible IID, the MO is set to “match-mapping” and the CDF is set to “mapping-sent”.
Otherwise the value variation of the IID may be reduced to few bytes. In that case, the TV is set to the stable part of the IID, the MO is set to MSB and the CDF is set to LSB.
Finally, the IID can be sent on the LPWAN. In that case, the TV is not set, the MO is set to “ignore” and the CDF is set to “value-sent”.
No extension rules are currently defined. They can be based on the MOs and CDFs described above.
To allow a single rule, the UDP port values are identified by their role (ES or LA) and not by their position in the frame (source or destination). The LC must be aware of the traffic direction (upstream, downstream) to select the appropriate field. The following rules apply for ES and LA port numbers.
If both ends knows the port number, it can be elided. The TV contains the port number, the MO is set to “equal” and the CDF is set to “not-sent”.
If the port variation is on few bits, the TV contains the stable part of the port number, the MO is set to “MSB” and the CDF is set to “LSB”.
If some well-known values are used, the TV can contain the list of this values, the MO is set to “match-mapping” and the CDF is set to “mapping-sent”.
Otherwise the port numbers are sent on the LPWAN. The TV is not set, the MO is set to “ignore” and the CDF is set to “value-sent”.
If the LPWAN technology does not introduce padding, the UDP length can be computed from the received data. In that case the TV is not set, the MO is set to “ignore” and the CDF is set to “compute-UDP-length”.
If the payload is small, the TV can be set to 0x0000, the MO set to “MSB” and the CDF to “LSB”.
On other cases, the length must be sent and the CDF is replaced by “value-sent”.
IPv6 mandates a checksum in the protocol above IP. Nevertheless, if a more efficient mechanism such as L2 CRC or MIC is carried by or over the L2 (such as in the LPWAN fragmentation process (see XXXX)), the UDP checksum transmission can be avoided. In that case, the TV is not set, the MO is set to “ignore” and the CDF is set to “compute-UDP-checksum”.
In other cases the checksum must be explicitly sent. The TV is not set, the MO is set to “ignore” and the CDF is set to “value-sent”.
This section gives some scenarios of the compression mechanism for IPv6/UDP. The goal is to illustrate the SCHC behavior.
The most common case using the mechanisms defined in this document will be a LPWAN end-system that embeds some applications running over CoAP. In this example, three flows are considered. The first flow is for the device management based on CoAP using Link Local IPv6 addresses and UDP ports 123 and 124 for ES and LA, respectively. The second flow will be a CoAP server for measurements done by the end-system (using ports 5683) and Global IPv6 Address prefixes alpha::IID/64 to beta::1/64. The last flow is for legacy applications using different ports numbers, the destination IPv6 address prefix is gamma::1/64.
Figure 4 presents the protocol stack for this End-System. IPv6 and UDP are represented with dotted lines since these protocols are compressed on the radio link.
Managment Data
+----------+---------+---------+
| CoAP | CoAP | legacy |
+----||----+---||----+---||----+
. UDP . UDP | UDP |
................................
. IPv6 . IPv6 . IPv6 .
+------------------------------+
| SCHC Header compression |
| and fragmentation |
+------------------------------+
| 6LPWA L2 technologies |
+------------------------------+
End System or LPWA GW
Figure 4: Simplified Protocol Stack for LP-WAN
Note that in some LPWAN technologies, only the End Systems have a device ID. Therefore, when such technologie are used, it is necessary to define statically an IID for the Link Local address for the LPWAN compressor.
Rule 0 +----------------+---------+--------+-------------++------+ | Field | Value | Match | Function || Sent | +----------------+---------+----------------------++------+ |IPv6 version |6 | equal | not-sent || | |IPv6 DiffServ |0 | equal | not-sent || | |IPv6 Flow Label |0 | equal | not-sent || | |IPv6 Length | | ignore | comp-IPv6-l || | |IPv6 Next Header|17 | equal | not-sent || | |IPv6 Hop Limit |255 | ignore | not-sent || | |IPv6 ESprefix |FE80::/64| equal | not-sent || | |IPv6 ESiid | | ignore | ESiid-DID || | |IPv6 LCprefix |FE80::/64| equal | not-sent || | |IPv6 LAiid |::1 | equal | not-sent || | +================+=========+========+=============++======+ |UDP ESport |123 | equal | not-sent || | |UDP LAport |124 | equal | not-sent || | |UDP Length | | ignore | comp-length || | |UDP checksum | | ignore | comp-chk || | +================+=========+========+=============++======+ Rule 1 +----------------+---------+--------+-------------++------+ | Field | Value | Match | Function || Sent | +----------------+---------+--------+-------------++------+ |IPv6 version |6 | equal | not-sent || | |IPv6 DiffServ |0 | equal | not-sent || | |IPv6 Flow Label |0 | equal | not-sent || | |IPv6 Length | | ignore | comp-IPv6-l || | |IPv6 Next Header|17 | equal | not-sent || | |IPv6 Hop Limit |255 | ignore | not-sent || | |IPv6 ESprefix |alpha/64 | equal | not-sent || | |IPv6 ESiid | | ignore | ESiid-DID || | |IPv6 LAprefix |beta/64 | equal | not-sent || | |IPv6 LAiid |::1000 | equal | not-sent || | +================+=========+========+=============++======+ |UDP ESport |5683 | equal | not-sent || | |UDP LAport |5683 | equal | not-sent || | |UDP Length | | ignore | comp-length || | |UDP checksum | | ignore | comp-chk || | +================+=========+========+=============++======+ Rule 2 +----------------+---------+--------+-------------++------+ | Field | Value | Match | Function || Sent | +----------------+---------+--------+-------------++------+ |IPv6 version |6 | equal | not-sent || | |IPv6 DiffServ |0 | equal | not-sent || | |IPv6 Flow Label |0 | equal | not-sent || | |IPv6 Length | | ignore | comp-IPv6-l || | |IPv6 Next Header|17 | equal | not-sent || | |IPv6 Hop Limit |255 | ignore | not-sent || | |IPv6 ESprefix |alpha/64 | equal | not-sent || | |IPv6 ESiid | | ignore | ESiid-DID || | |IPv6 LAprefix |gamma/64 | equal | not-sent || | |IPv6 LAiid |::1000 | equal | not-sent || | +================+=========+========+=============++======+ |UDP ESport |8720 | MSB(12)| LSB(4) || lsb | |UDP LAport |8720 | MSB(12)| LSB(4) || lsb | |UDP Length | | ignore | comp-length || | |UDP checksum | | ignore | comp-chk || | +================+=========+========+=============++======+
Figure 5: Context rules
All the fields described in the three rules Figure 5 are present in the IPv6 and UDP headers. The ESDevice-ID value is found in the L2 header.
The second and third rules use global addresses. The way the ES learns the prefix is not in the scope of the document.
The third rule compresses port numbers to 4 bits.
Fragmentation support in LPWAN is mandatory and it is used if, after SCHC header compression, the size of the resulting packet is larger than the L2 data unit maximum payload. Fragmentation is also used if SCHC header compression has not been able to compress a packet that is larger than the L2 data unit maximum payload. In LPWAN technologies the L2 data unit size typically varies from tens to hundreds of bytes. If the entire IPv6 datagram fits within a single L2 data unit, the fragmentation mechanism is not used and the packet is sent unfragmented.
If the datagram does not fit within a single L2 data unit, it SHALL be broken into fragments.
Moreover, LPWAN technologies impose some strict limitations on traffic; therefore it is desirable to enable optional fragment retransmission, while a single fragment loss should not lead to retransmitting the full datagram. To preserve energy, Things (End Systems) are sleeping most of the time and may receive data during a short period of time after transmission.
In order to adapt to the capabilities of various LPWAN technologies, this specification allows for a gradation of fragment delivery reliability. There are three main options: Unreliable (UnR) mode, Reliable per-Packet (RpP) mode and Reliable per-Window (RpW) mode. Additionally, the specification provides the option to withhold acknowledgments (ACK) in case of success, making effectively the ACK a Negative ACK (NACK). It is up to the underlying LPWAN technology to decide which setting to use and whether the same setting applies to all IPv6 packets. Note that the fragment delivery reliability option to be used is not necessarily tied to the particular characteristics of the underlying L2 LPWAN technology (e.g. UnR may be used on top of an L2 LPWAN technology with symmetric characteristics for uplink and downlink).
The same reliability option MUST be used for all fragments of a packet.
In UnR mode, the receiver MUST NOT issue acknowledgments. In RpP mode, the receiver may transmit one acknowledgment (ACK) after all fragments carrying an IPv6 packet have been transmitted. The ACK informs the sender about received and missing fragments from the IPv6 packet. In RpW mode, an ACK may be transmitted by the fragment receiver after a window of fragments have been sent. A window of fragments is a subset of the full set of fragments needed to carry an IPv6 packet. In this mode, the ACK informs the sender about received and missing fragments from the window of fragments. In either mode, upon receipt of an ACK that informs about any lost fragments, the sender may retransmit the lost fragments. The maximum number of ACK and retransmission rounds is TBD.
Some LPWAN deployments may benefit from conditioning the creation and transmission of an ACK to the detection of at least one fragment loss (per-packet or per-window), thus leading to NACK-oriented behavior, while not having such condition may be preferred for other scenarios.
This document does not make any decision as to whether UnR, RpP or RpW modes are used, or or whether the transmission of ACKs is conditioned to the detection of fragment losses or not. A complete specification of the receiver and sender behaviors that correspond to each acknowledgment policy is also out of scope. Nevertheless, this document does provide examples of the different reliability options described.
A fragment comprises a fragmentation header and a fragment payload, and conforms to the format shown in Figure 6. The fragment payload carries a subset of either the IPv6 packet after header compression or an IPv6 packet which could not be compressed. A fragment is the payload in the L2 protocol data unit (PDU).
+---------------+-----------------------+
| Fragm. Header | Fragment payload |
+---------------+-----------------------+
Figure 6: Fragment format.
Fragments except the last one SHALL
contain the fragmentation header as defined in Figure 7. The total size of this fragmentation header is R bits.
<----------- R ----------->
<-- N -->
+----- ... -----+-- ... --+
| Rule ID | CFN |
+----- ... -----+-- ... --+
Figure 7: Fragmentation Header for Fragments except the Last One
The last fragment SHALL contain a fragmentation header that conforms to the format shown in Figure 8. The total size of this fragmentation header is R+M bits.
<----------- R ---------->
<-- N --> <---- M ----->
+----- ... -----+-- ... --+---- ... ----+
| Rule ID | 11..1 | MIC |
+----- ... -----+-- ... --+---- ... ----+
Figure 8: Fragmentation Header for the Last Fragment
Rule ID: this field has a size of R – N bits in all fragments. Rule ID may be used to signal whether UnR, RpP or RpW mode is in use, and within the latter, whether window mode or packet mode are used.
CFN: CFN stands for Compressed Fragment Number. The size of the CFN field is N bits. In UnR mode, N=1. For RpP or RpW modes, N equal to or greater than 3 is recommended. This field is an unsigned integer that carries a non-absolute fragment number. The CFN MUST be set sequentially decreasing from 2^N - 2 for the first fragment, and MUST wrap from 0 back to 2^N - 2 (e.g. for N=3, the first fragment has CFN=6, subsequent CFNs are set sequentially and in decreasing order, and CFN will wrap from 0 back to 6). The CFN for the last fragment has all bits set to 1. Note that, by this definition, the CFN value of 2^N - 1 is only used to identify a fragment as the last fragment carrying a subset of the IPv6 packet being transported, and thus the CFN does not strictly correspond to the N least significant bits of the actual absolute fragment number. It is also important to note that, for N=1, the last fragment of the packet will carry a CFN equal to 1, while all previous fragments will carry a CFN of 0.
MIC: MIC stands for Message Integrity Check. This field has a size of M bits. It is computed by the sender over the complete IPv6 packet before fragmentation by using the TBD algorithm. The MIC allows to check for errors in the reassembled IPv6 packet, while it also enables compressing the UDP checksum by use of SCHC.
The values for R, N and M are not specified in this document, and have to be determined by the underlying LPWAN technology.
The format of an ACK is shown in Figure 9:
<----- R ---->
+-+-+-+-+-+-+-+-+----- ... ---+
| Rule ID | bitmap |
+-+-+-+-+-+-+-+-+----- ... ---+
Figure 9: Format of an ACK
Rule ID: In all ACKs, Rule ID has a size of R bits and SHALL be set to TBD_ACK to signal that the message is an ACK.
bitmap: size of the bitmap field of an ACK can be equal to 0 or
Ceiling(Number_of_Fragments/8) octets, where Number_of_Fragments denotes the number of fragments of a window (in RpW mode) or the number of fragments that carry the IPv6 packet (in RpP mode). The bitmap is a sequence of bits, where the n-th bit signals whether the n-th fragment transmitted has been correctly received (n-th bit set to 1) or not (n-th bit set to 0). Remaining bits with bit order greater than the number of fragments sent (as determined by the receiver) are set to 0, except for the last bit in the bitmap, which is set to 1 if the last fragment (carrying the MIC) has been correctly received, and 0 otherwise. Absence of the bitmap in an ACK confirms correct reception of all fragments to be acknowledged by means of the ACK.
Figure 10 shows an example of an ACK in packet mode, where the bitmap indicates that the second and the ninth fragments have not been correctly received. In this example, the IPv6 packet is carried by eleven fragments in total, therefore the bitmap has a size of two bytes.
1
<----- R ----> 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Rule ID |1|0|1|1|1|1|1|1|0|1|1|0|0|0|0|1|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 10: Example of the Bitmap in an ACK
Figure 11 shows an example of an ACK in RpW (N=3), where the bitmap indicates that the second and the fifth fragments have not been correctly received.
<----- R ----> 0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Rule ID |1|0|1|1|0|1|1|1|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 11: Example of the bitmap in an ACK (in RpW mode, for N=3)
Figure 12 illustrates an ACK without bitmap.
<----- R ---->
+-+-+-+-+-+-+-+-+
| Rule ID |
+-+-+-+-+-+-+-+-+
Figure 12: Example of an ACK without bitmap
The receiver of link fragments SHALL use (1) the sender’s L2 source address (if present), (2) the destination’s L2 address (if present), and (3) Rule ID to identify all the fragments that belong to a given datagram. The fragment receiver SHALL determine the fragment delivery reliability option in use for the fragment based on the Rule ID field in that fragment.
Upon receipt of a link fragment, the receiver starts constructing the original unfragmented packet. It uses the CFN and the order of arrival of each fragment to determine the location of the individual fragments within the original unfragmented packet. For example, it may place the data payload of the fragments within a payload datagram reassembly buffer at the location determined from the CFN and order of arrival of the fragments, and the fragment payload sizes. Note that the size of the original, unfragmented IPv6 packet cannot be determined from fragmentation headers.
In RpW mode, when a fragment with all CFN bits set to 0 is received, the recipient MAY transmit an ACK for the last window of fragments sent. Note that the first fragment of the window is the one sent with CFN=2^N-2. In RpW mode, the fragment with CFN=0 is considered the last fragment of its window, except for the last fragment of the whole packet (with all CFN bits set to 1), which is also the last fragment of the last window.
Once the recipient has received the last fragment, it checks for the integrity of the reassembled IPv6 datagram, based on the MIC received. In UnR mode, if the integrity check indicates that the reassembled IPv6 datagram does not match the original IPv6 datagram (prior to fragmentation), the reassembled IPv6 datagram MUST be discarded. In RpP or in RpW mode, upon receipt of the last fragment (i.e. with all CFN bits set to 1), the recipient MAY transmit an ACK for the whole set of fragments sent that carry the complete IPv6 packet.
In RpP mode or in RpW mode, the sender retransmits any lost fragments reported in the ACK. A maximum of TBD iterations of ACK and fragment retransmission rounds are allowed per-window or per-IPv6-packet in RpP mode or in RpW mode, respectively. A complete specification of the mechanisms needed to enable the above described fragment delivery reliability options is out of the scope of this document.
If a fragment recipient disassociates from its L2 network, the recipient MUST discard all link fragments of all partially reassembled payload datagrams, and fragment senders MUST discard all not yet transmitted link fragments of all partially transmitted payload (e.g., IPv6) datagrams. Similarly, when a node first receives a fragment of a packet, it starts a reassembly timer. When this time expires, if the entire packet has not been reassembled, the existing fragments MUST be discarded and the reassembly state MUST be flushed. The reassembly timeout MUST be set to a maximum of TBD seconds).
TBD
This subsection describes potential attacks to LPWAN fragmentation and proposes countermeasures, based on existing analysis of attacks to 6LoWPAN fragmentation {HHWH}.
A node can perform a buffer reservation attack by sending a first fragment to a target. Then, the receiver will reserve buffer space for the whole packet on the basis of the datagram size announced in that first fragment. Other incoming fragmented packets will be dropped while the reassembly buffer is occupied during the reassembly timeout. Once that timeout expires, the attacker can repeat the same procedure, and iterate, thus creating a denial of service attack. The (low) cost to mount this attack is linear with the number of buffers at the target node. However, the cost for an attacker can be increased if individual fragments of multiple packets can be stored in the reassembly buffer. To further increase the attack cost, the reassembly buffer can be split into fragment-sized buffer slots. Once a packet is complete, it is processed normally. If buffer overload occurs, a receiver can discard packets based on the sender behavior, which may help identify which fragments have been sent by an attacker.
In another type of attack, the malicious node is required to have overhearing capabilities. If an attacker can overhear a fragment, it can send a spoofed duplicate (e.g. with random payload) to the destination. A receiver cannot distinguish legitimate from spoofed fragments. Therefore, the original IPv6 packet will be considered corrupt and will be dropped. To protect resource-constrained nodes from this attack, it has been proposed to establish a binding among the fragments to be transmitted by a node, by applying content- chaining to the different fragments, based on cryptographic hash functionality. The aim of this technique is to allow a receiver to identify illegitimate fragments.
Further attacks may involve sending overlapped fragments (i.e. comprising some overlapping parts of the original IPv6 datagram). Implementers should make sure that correct operation is not affected by such event.
Thanks to Dominique Barthel, Carsten Bormann, Arunprabhu Kandasamy, Antony Markovski, Alexander Pelov, Pascal Thubert, Juan Carlos Zuniga for useful design consideration.
| [RFC2460] | Deering, S. and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 2460, DOI 10.17487/RFC2460, December 1998. |
| [RFC4944] | Montenegro, G., Kushalnagar, N., Hui, J. and D. Culler, "Transmission of IPv6 Packets over IEEE 802.15.4 Networks", RFC 4944, DOI 10.17487/RFC4944, September 2007. |
| [I-D.ietf-lpwan-overview] | Farrell, S., "LPWAN Overview", Internet-Draft draft-ietf-lpwan-overview-01, February 2017. |
| [I-D.minaburo-lp-wan-gap-analysis] | Minaburo, A., Pelov, A. and L. Toutain, "LP-WAN GAP Analysis", Internet-Draft draft-minaburo-lp-wan-gap-analysis-01, February 2016. |
This section provides examples of different fragment delivery reliability options possible on the basis of this specification.
Figure 13 illustrates the transmission of an IPv6 packet that needs 11 fragments in UnR mode.
Sender Receiver
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=0-------->|
|-------CFN=1-------->|MIC checked =>
Figure 13: Transmission of an IPv6 packet carried by 11 fragments in UnR mode
Figure 14 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpP mode, for N=3, NACK-oriented, without losses.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=3-------->|
|-------CFN=2-------->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=7-------->|MIC checked =>
(no NACK)
Figure 14: Transmission of an IPv6 packet carried by 11 fragments in RpP mode, for N=3, NACK-oriented; no losses.
Figure 15 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpP mode, for N=3, NACK-oriented, with three losses.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=3-------->|
|-------CFN=2---X---->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=7-------->|MIC checked =>
|<-------NACK---------|Bitmap:1101011110100001
|-------CFN=4-------->|
|-------CFN=2-------->|
|-------CFN=4-------->|MIC checked =>
(no NACK)
Figure 15: Transmission of an IPv6 packet carried by 11 fragments in RpP mode, for N=3, NACK-oriented; three losses.
Figure 16 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpW mode, for N=3, without losses. Receiver feedback is NACK-oriented. Note: in RpW mode, an additional bit will be needed to number windows.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=3-------->|
|-------CFN=2-------->|
|-------CFN=1-------->|
|-------CFN=0-------->|
(no NACK)
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=7-------->|MIC checked =>
(no NACK)
Figure 16: Transmission of an IPv6 packet carried by 11 fragments in RpW mode, for N=3, NACK-oriented; without losses.
Figure 17 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpW mode, for N=3, with three losses. Receiver feedback is NACK-oriented. Note: in RpW mode, an additional bit will be needed to number windows.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=3-------->|
|-------CFN=2---X---->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|<-------NACK---------|Bitmap:11010111
|-------CFN=4-------->|
|-------CFN=2-------->|
(no NACK)
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=7-------->|MIC checked =>
|<-------NACK---------|Bitmap:11010001
|-------CFN=4-------->|MIC checked =>
(no NACK)
Figure 17: Transmission of an IPv6 packet carried by 11 fragments in RpW, for N=3, NACK-oriented; three losses.
Figure 18 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpP mode, for N=3, without losses. Receiver feedback is positive-ACK-oriented.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=3-------->|
|-------CFN=2-------->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=7-------->|MIC checked =>
|<-------ACK----------|no bitmap
(End)
Figure 18: Transmission of an IPv6 packet carried by 11 fragments in RpP mode, for N=3, positive-ACK-oriented; no losses.
Figure 19 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpP mode, for N=3, with three losses. Receiver feedback is positive-ACK-oriented.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=3-------->|
|-------CFN=2---X---->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=7-------->|MIC checked =>
|<-------ACK----------|bitmap:1101011110100001
|-------CFN=4-------->|
|-------CFN=2-------->|
|-------CFN=4-------->|MIC checked =>
|<-------ACK----------|no bitmap
(End)
Figure 19: Transmission of an IPv6 packet carried by 11 fragments in RpP, for N=3, positive-ACK-oriented; with three losses.
Figure 20 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpW mode, for N=3, without losses. Receiver feedback is positive-ACK-oriented. Note: in RpW mode, an additional bit will be needed to number windows.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=3-------->|
|-------CFN=2-------->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|<-------ACK----------|no bitmap
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4-------->|
|-------CFN=7-------->|MIC checked =>
|<-------ACK----------|no bitmap
(End)
Figure 20: Transmission of an IPv6 packet carried by 11 fragments in RpW mode, for N=3, positive-ACK-oriented; no losses.
Figure 21 illustrates the transmission of an IPv6 packet that needs 11 fragments in RpW mode, for N=3, with three losses. Receiver feedback is positive-ACK-oriented. Note: in RpW mode, an additional bit will be needed to number windows.
Sender Receiver
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=3-------->|
|-------CFN=2---X---->|
|-------CFN=1-------->|
|-------CFN=0-------->|
|<-------ACK----------|bitmap:11010111
|-------CFN=4-------->|
|-------CFN=2-------->|
|<-------ACK----------|no bitmap
|-------CFN=6-------->|
|-------CFN=5-------->|
|-------CFN=4---X---->|
|-------CFN=7-------->|MIC checked =>
|<-------ACK----------|bitmap:11010001
|-------CFN=4-------->|MIC checked =>
|<-------ACK----------|no bitmap
(End)
Figure 21: Transmission of an IPv6 packet carried by 11 fragments in RpW mode, for N=3, positive-ACK-oriented; with three losses.
Carles Gomez has been funded in part by the Spanish Government (Ministerio de Educacion, Cultura y Deporte) through the Jose Castillejo grant CAS15/00336, and by the ERDF and the Spanish Government through project TEC2016-79988-P. Part of his contribution to this work has been carried out during his stay as a visiting scholar at the Computer Laboratory of the University of Cambridge.