Request for Comments: 4623 Tellabs
Category: Standards Track M. Townsley
Cisco Systems
August 2006
Pseudowire Emulation Edge-to-Edge (PWE3)
Fragmentation and Reassembly
Status of This Memo
This document specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" (STD 1) for the standardization state
and status of this protocol. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2006).
Abstract
This document defines a generalized method of performing
fragmentation for use by Pseudowire Emulation Edge-to-Edge (PWE3)
protocols and services.
Table of Contents
1. Introduction ....................................................3
2. Conventions Used in This Document ...............................4
3. Alternatives to PWE3 Fragmentation/Reassembly ...................5
4. PWE3 Fragmentation with MPLS ....................................5
4.1. Fragment Bit Locations for MPLS ............................6
4.2. Other Considerations .......................................6
5. PWE3 Fragmentation with L2TP ....................................6
5.1. PW-Specific Fragmentation vs. IP fragmentation .............7
5.2. Advertising Reassembly Support in L2TP .....................7
5.3. L2TP Maximum Receive Unit (MRU) AVP ........................8
5.4. L2TP Maximum Reassembled Receive Unit (MRRU) AVP ...........8
5.5. Fragment Bit Locations for L2TPv3 Encapsulation ............9
5.6. Fragment Bit Locations for L2TPv2 Encapsulation ............9
6. Security Considerations ........................................10
7. IANA Considerations ............................................10
7.1. Control Message Attribute Value Pairs (AVPs) ..............11
7.2. Default L2-Specific Sublayer Bits .........................11
7.3. Leading Bits of the L2TPv2 Message Header .................11
8. Acknowledgements ...............................................11
9. Normative References ...........................................12
10. Informative References ........................................12
Appendix A. Relationship Between This Document and RFC 1990 .......14
1. Introduction
The Pseudowire Emulation Edge-to-Edge Architecture Document
[Architecture] defines a network reference model for PWE3:
|<-------------- Emulated Service ---------------->|
| |
| |<------- Pseudowire ------->| |
| | | |
| | |<-- PSN Tunnel -->| | |
| PW End V V V V PW End |
V Service +----+ +----+ Service V
+-----+ | | PE1|==================| PE2| | +-----+
| |----------|............PW1.............|----------| |
| CE1 | | | | | | | | CE2 |
| |----------|............PW2.............|----------| |
+-----+ ^ | | |==================| | | ^ +-----+
^ | +----+ +----+ | | ^
| | Provider Edge 1 Provider Edge 2 | |
| | | |
Customer | | Customer
Edge 1 | | Edge 2
| |
| |
native service native service
Figure 1: PWE3 Network Reference Model
A Pseudowire (PW) payload is normally relayed across the PW as a
single IP or MPLS Packet Switched Network (PSN) Protocol Data Unit
(PDU). However, there are cases where the combined size of the
payload and its associated PWE3 and PSN headers may exceed the PSN
path Maximum Transmission Unit (MTU). When a packet exceeds the MTU
of a given network, fragmentation and reassembly will allow the
packet to traverse the network and reach its intended destination.
The purpose of this document is to define a generalized method of
performing fragmentation for use with all PWE3 protocols and
services. This method should be utilized only in cases where MTU-
management methods fail. Due to the increased processing overhead,
fragmentation and reassembly in core network devices should always be
considered something to avoid whenever possible.
The PWE3 fragmentation and reassembly domain is shown in Figure 2:
|<-------------- Emulated Service ---------------->|
| |<---Fragmentation Domain--->| |
| ||<------- Pseudowire ----->|| |
| || || |
| || |<-- PSN Tunnel -->| || |
| PW End VV V V VV PW End |
V Service +----+ +----+ Service V
+-----+ | | PE1|==================| PE2| | +-----+
| |----------|............PW1.............|----------| |
| CE1 | | | | | | | | CE2 |
| |----------|............PW2.............|----------| |
+-----+ ^ | | |==================| | | ^ +-----+
^ | +----+ +----+ | | ^
| | Provider Edge 1 Provider Edge 2 | |
| | | |
Customer | | Customer
Edge 1 | | Edge 2
| |
| |
native service native service
Figure 2: PWE3 Fragmentation/Reassembly Domain
Fragmentation takes place in the transmitting PE immediately prior to
PW encapsulation, and reassembly takes place in the receiving PE
immediately after PW decapsulation.
Since a sequence number is necessary for the fragmentation and
reassembly procedures, using the Sequence Number field on fragmented
packets is REQUIRED (see Sections 4.1 and 5.5 for the location of the
Sequence Number fields for MPLS and L2TPv3 encapsulations,
respectively). The order of operation is that first fragmentation is
performed, and then the resulting fragments are assigned sequential
sequence numbers.
Depending on the specific PWE3 encapsulation in use, the value 0 may
not be a part of the sequence number space, in which case its use for
fragmentation must follow this same rule: as the sequence number is
incremented, it skips zero and wraps from 65535 to 1. Conversely, if
the value 0 is part of the sequence space, then the same sequence
space is also used for fragmentation and reassembly.
2. Conventions Used in This Document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [KEYWORDS].
3. Alternatives to PWE3 Fragmentation/Reassembly
Fragmentation and reassembly in network equipment generally requires
significantly greater resources than sending a packet as a single
unit. As such, fragmentation and reassembly should be avoided
whenever possible. Ideal solutions for avoiding fragmentation
include proper configuration and management of MTU sizes between the
Customer Edge (CE) router and Provider Edge (PE) router and across
the PSN, as well as adaptive measures that operate with the
originating host (e.g., [PATHMTU], [PATHMTUv6]) to reduce the packet
sizes at the source.
In some cases, a PE may be able to fragment an IP version 4 (IPv4)
[RFC791] packet before it enters a PW. For example, if the PE can
fragment and forward IPv4 packets with the DF bit clear in a manner
that is identical to an IPv4 router, it may fragment packets arriving
from a CE, forwarding the IPv4 fragments with associated framing for
that attachment circuit (AC) over the PW. Architecturally, the IPv4
fragmentation happens before reaching the PW, presenting multiple
frames to the PW to forward in the normal manner for that PWType.
Thus, this method is entirely transparent to the PW encapsulation and
to the remote end of the PW itself. Packet fragments are ultimately
reassembled on the destination IPv4 host in the normal way. IPv6
packets are not to be fragmented in this manner.
4. PWE3 Fragmentation with MPLS
When using the signaling procedures in [MPLS-Control], there is a
Pseudowire Interface Parameter Sub-TLV type used to signal the use of
fragmentation when advertising a VC label [IANA]:
Parameter Length Description
0x09 4 Fragmentation indicator
The presence of this parameter in the VC FEC element indicates that
the receiver is able to reassemble fragments when the control word is
in use for the VC label being advertised. It does not obligate the
sender to use fragmentation; it is simply an indication that the
sender MAY use fragmentation. The sender MUST NOT use fragmentation
if this parameter is not present in the VC FEC element.
If [MPLS-Control] signaling is not in use, then whether or not to use
fragmentation MUST be configured in the sender.
4.1. Fragment Bit Locations for MPLS
MPLS-based PWE3 uses the following control word format
[Control-Word], with the B and E fragmentation bits identified in
position 8 and 9:
0 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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|0 0 0 0| Flags |B|E| Length | Sequence Number |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 3: Preferred PW MPLS Control Word
The B and E bits are defined as follows:
BE
--
00 indicates that the entire (un-fragmented) payload is carried
in a single packet
01 indicates the packet carrying the first fragment
10 indicates the packet carrying the last fragment
11 indicates a packet carrying an intermediate fragment
See Appendix A for a discussion of the derivation of these values for
the B and E bits.
See Section 1 for the description of the use of the Sequence Number
field.
4.2. Other Considerations
Path MTU [PATHMTU] [PATHMTUv6] may be used to dynamically determine
the maximum size for fragments. The application of path MTU to MPLS
is discussed in [LABELSTACK]. The maximum size of the fragments may
also be configured. The signaled Interface MTU parameter in
[MPLS-Control] SHOULD be used to set the maximum size of the
reassembly buffer for received packets to make optimal use of
reassembly buffer resources.
5. PWE3 Fragmentation with L2TP
This section defines the location of the B and E bits for L2TPv3
[L2TPv3] and L2TPv2 [L2TPv2] headers, as well as the signaling
mechanism for advertising MRU (Maximum Receive Unit) values and
support for fragmentation on a given PW. As IP is the most common
PSN used with L2TP, IP PSN fragmentation and reassembly is discussed
as well.
5.1. PW-Specific Fragmentation vs. IP fragmentation
When proper MTU management across a network fails, IP PSN
fragmentation and reassembly may be used to accommodate MTU
mismatches between tunnel endpoints. If the overall traffic
requiring fragmentation and reassembly is very light, or there are
sufficient optimized mechanisms for IP PSN fragmentation and
reassembly available, IP PSN fragmentation and reassembly may be
sufficient.
When facing a large number of PW packets requiring fragmentation and
reassembly, a PW-specific method has properties that potentially
allow for more resource-friendly implementations. Specifically, the
ability to assign buffer usage on a per-PW basis and PW sequencing
may be utilized to gain advantage over a general mechanism applying
to all IP packets across all PWs. Further, PW fragmentation may be
more easily enabled in a selective manner for some or all PWs, rather
than enabling reassembly for all IP traffic arriving at a given node.
Deployments SHOULD avoid a situation that uses a combination of IP
PSN and PW fragmentation and reassembly on the same node. Such
operation clearly defeats the purpose behind the mechanism defined in
this document. This is especially important for L2TPv3 pseudowires,
since potentially fragmentation can take place in three different
places (the IP PSN, the PW, and the encapsulated payload). Care must
be taken to ensure that the MTU/MRU values are set and advertised
properly at each tunnel endpoint to avoid this. When fragmentation
is enabled within a given PW, the DF bit MUST be set on all L2TP over
IP packets for that PW.
L2TPv3 nodes SHOULD participate in Path MTU ([PATHMTU], [PATHMTUv6])
for automatic adjustment of the PSN MTU. When the payload is IP,
Path MTU should be used at they payload level as well.
5.2. Advertising Reassembly Support in L2TP
The constructs defined in this section for advertising fragmentation
support in L2TP are applicable to [L2TPv3] and [L2TPv2].
This document defines two new AVPs to advertise maximum receive unit
values and reassembly support. These AVPs MAY be present in the
Incoming-Call-Request (ICRQ), Incoming-Call-Reply (ICRP), Incoming-
Call-Connected (ICCN), Outgoing-Call-Request (OCRQ), Outgoing-Call-
Reply (OCRP), Outgoing-Call-Connected (OCCN), or Set-Link-Info (SLI)
messages. The most recent value received always takes precedence
over a previous value and MUST be dynamic over the life of the
session if received via the SLI message. One of the two new AVPs
(MRRU) is used to advertise that PWE3 reassembly is supported by the
sender of the AVP. Reassembly support MAY be unidirectional.
5.3. L2TP Maximum Receive Unit (MRU) AVP
0 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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|M|H|0|0|0|0| Length | 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| MRU |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 4: L2TP Maximum Receive Unit (MRU) AVP
MRU (Maximum Receive Unit), attribute number 94, is the maximum size,
in octets, of a fragmented or complete PW frame, including L2TP
encapsulation, receivable by the side of the PW advertising this
value. The advertised MRU does NOT include the PSN header (i.e., the
IP and/or UDP header). This AVP does not imply that PWE3
fragmentation or reassembly is supported. If reassembly is not
enabled or unavailable, this AVP may be used alone to advertise the
MRU for a complete frame.
This AVP MAY be hidden (the H bit MAY be 0 or 1). The mandatory (M)
bit for this AVP SHOULD be set to 0. The Length (before hiding) is
8. The Vendor ID is the IETF Vendor ID of 0.
5.4. L2TP Maximum Reassembled Receive Unit (MRRU) AVP
0 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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|M|H|0|0|0|0| Length | 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| MRRU |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 5: L2TP Maximum Reassembled Receive Unit (MRRU) AVP
MRRU (Maximum Reassembled Receive Unit AVP), attribute number 95, is
the maximum size, in octets, of a reassembled frame, including any PW
framing, but not including the L2TP encapsulation or L2-specific
sublayer. Presence of this AVP signifies the ability to receive PW
fragments and reassemble them. Packet fragments MUST NOT be sent by
a peer that has not received this AVP in a control message. If the
MRRU is present in a message, the MRU AVP MUST be present as well.
The MRRU SHOULD be used to set the maximum size of the reassembly
buffer for received packets to make optimal use of reassembly buffer
resources.
This AVP MAY be hidden (the H bit MAY be 0 or 1). The mandatory (M)
bit for this AVP SHOULD be set to 0. The Length (before hiding) is
8. The Vendor ID is the IETF Vendor ID of 0.
5.5. Fragment Bit Locations for L2TPv3 Encapsulation
The usage of the B and E bits is described in Section 4.1. For
L2TPv3 encapsulation, the B and E bits are defined as bits 2 and 3 in
the leading bits of the Default L2-Specific Sublayer (see Section 7).
0 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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|M|H|0|0|0|0| Length | 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|x|S|B|E|x|x|x|x| Sequence Number |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 6: B and E Bits Location in the Default L2-Specific Sublayer
The S (Sequence) bit is as defined in [L2TPv3]. Location of the B
and E bits for PW-Types that use a variant L2 specific sublayer are
outside the scope of this document.
When fragmentation is used, an L2-Specific Sublayer with B and E bits
defined MUST be present in all data packets for a given session. The
presence and format of the L2-Specific Sublayer is advertised via the
L2-Specific Sublayer AVP, Attribute Type 69, defined in Section 5.4.4
of [L2TPv3].
See Section 1 for the description of the use of the Sequence Number
field.
5.6. Fragment Bit Locations for L2TPv2 Encapsulation
The usage of the B and E bits is described in Section 4.1. For
L2TPv2 encapsulation, the B and E bits are defined as bits 8 and 9 in
the leading bits of the L2TPv2 header as depicted below (see Section
7).
0 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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|M|H|0|0|0|0| Length | 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|T|L|x|x|S|x|O|P|B|E|x|x| Ver | Length (opt) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 7: B and E bits location in the L2TPv2 Message Header
6. Security Considerations
As with any additional protocol construct, each level of complexity
adds the potential to exploit protocol and implementation errors.
Implementers should be especially careful of not tying up an
abundance of resources, even for the most pathological combination of
packet fragments that could be received. Beyond these issues of
general implementation quality, there are no known notable security
issues with using the mechanism defined in this document. It should
be pointed out that RFC 1990, on which this document is based, and
its derivatives have been widely implemented and extensively used in
the Internet and elsewhere.
[IPFRAG-SEC] and [TINYFRAG] describe potential network attacks
associated with IP fragmentation and reassembly. The issues
described in these documents attempt to bypass IP access controls by
sending various carefully formed "tiny fragments", or by exploiting
the IP offset field to cause fragments to overlap and rewrite
interesting portions of an IP packet after access checks have been
performed. The latter is not an issue with the PW-specific
fragmentation method described in this document, as there is no
offset field. However, implementations MUST be sure not to allow
more than one whole fragment to overwrite another in a reconstructed
frame. The former may be a concern if packet filtering and access
controls are being placed on tunneled frames within the PW