Request for Comments: 4023 Motorola, Inc.
Category: Standards Track Y. Rekhter
Juniper Networks, Inc.
E. Rosen, Ed.
Cisco Systems, Inc.
March 2005
Encapsulating MPLS in IP or Generic Routing Encapsulation (GRE)
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 (2005).
Abstract
Various applications of MPLS make use of label stacks with multiple
entries. In some cases, it is possible to replace the top label of
the stack with an IP-based encapsulation, thereby enabling the
application to run over networks that do not have MPLS enabled in
their core routers. This document specifies two IP-based
encapsulations: MPLS-in-IP and MPLS-in-GRE (Generic Routing
Encapsulation). Each of these is applicable in some circumstances.
Table of Contents
1. Motivation .................................................. 2
2. Specification of Requirements ............................... 3
3. Encapsulation in IP ......................................... 3
4. Encapsulation in GRE ........................................ 4
5. Common Procedures ........................................... 5
5.1. Preventing Fragmentation and Reassembly ............... 5
5.2. TTL or Hop Limit ...................................... 6
5.3. Differentiated Services ............................... 7
6. Applicability ............................................... 7
7. IANA Considerations ......................................... 8
8. Security Considerations ..................................... 8
8.1. Securing the Tunnel with IPsec ......................... 8
8.2. In the Absence of IPsec ............................... 10
9. Acknowledgements ............................................. 11
10. Normative References ........................................ 11
11. Informative References ...................................... 12
Authors’ Addresses ............................................... 13
Full Copyright Statement ......................................... 14
1. Motivation
In many applications of MPLS, packets traversing an MPLS backbone
carry label stacks with more than one label. As described in section
3.15 of [RFC3031], each label represents a Label Switched Path (LSP).
For each LSP, there is a Label Switching Router (LSR) that is the
"LSP Ingress", and an LSR that is the "LSP Egress". If LSRs A and B
are the Ingress and Egress, respectively, of the LSP corresponding to
a packet’s top label, then A and B are adjacent LSRs on the LSP
corresponding to the packet’s second label (i.e., the label
immediately beneath the top label).
The purpose (or one of the purposes) of the top label is to get the
packet delivered from A to B, so that B can further process the
packet based on the second label. In this sense, the top label
serves as an encapsulation header for the rest of the packet. In
some cases, other sorts of encapsulation headers can replace the top
label without loss of functionality. For example, an IP header or a
Generic Routing Encapsulation (GRE) header could replace the top
label. As the encapsulated packet would still be an MPLS packet, the
result is an MPLS-in-IP or MPLS-in-GRE encapsulation.
With these encapsulations, it is possible for two LSRs that are
adjacent on an LSP to be separated by an IP network, even if that IP
network does not provide MPLS.
To use either of these encapsulations, the encapsulating LSR must
know
- the IP address of the decapsulating LSR, and
- that the decapsulating LSR actually supports the particular
encapsulation.
This knowledge may be conveyed to the encapsulating LSR by manual
configuration, or by means of some discovery protocol. In
particular, if the tunnel is being used to support a particular
application and that application has a setup or discovery protocol,
then the application’s protocol may convey this knowledge. The means
of conveying this knowledge is outside the scope of the this
document.
2. Specification of Requirements
In this document, the key words "MUST", "MUST NOT", "REQUIRED",
"SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY",
and "OPTIONAL" are to be interpreted as described in [RFC2119].
3. Encapsulation in IP
MPLS-in-IP messages have the following format:
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
| IP Header |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
| MPLS Label Stack |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
| Message Body |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
IP Header
This field contains an IPv4 or an IPv6 datagram header
as defined in [RFC791] or [RFC2460], respectively. The
source and destination addresses are set to addresses
of the encapsulating and decapsulating LSRs, respectively.
MPLS Label Stack
This field contains an MPLS Label Stack as defined in
[RFC3032].
Message Body
This field contains one MPLS message body.
The IPv4 Protocol Number field or the IPv6 Next Header field is set
to 137, indicating an MPLS unicast packet. (The use of the MPLS-in-
IP encapsulation for MPLS multicast packets is not supported by this
specification.)
Following the IP header is an MPLS packet, as specified in [RFC3032].
This encapsulation causes MPLS packets to be sent through "IP
tunnels". When the tunnel’s receive endpoint receives a packet, it
decapsulates the MPLS packet by removing the IP header. The packet
is then processed as a received MPLS packet whose "incoming label"
[RFC3031] is the topmost label of the decapsulated packet.
4. Encapsulation in GRE
The MPLS-in-GRE encapsulation encapsulates an MPLS packet in GRE
[RFC2784]. The packet then consists of an IP header (either IPv4 or
IPv6), followed by a GRE header, followed by an MPLS label stack as
specified in [RFC3032]. The protocol type field in the GRE header
MUST be set to the Ethertype value for MPLS Unicast (0x8847) or
Multicast (0x8848).
This encapsulation causes MPLS packets to be sent through "GRE
tunnels". When the tunnel’s receive endpoint receives a packet, it
decapsulates the MPLS packet by removing the IP and the GRE headers.
The packet is then processed as a received MPLS packet whose
"incoming label" [RFC3031] is the topmost label of the decapsulated
packet.
[RFC2784] specifies an optional GRE checksum, and [RFC2890] specifies
optional GRE key and sequence number fields. These optional fields
are not very useful for the MPLS-in-GRE encapsulation. The sequence
number and checksum fields are not needed, as there are no
corresponding fields in the native MPLS packets being tunneled. The
GRE key field is not needed for demultiplexing, as the top MPLS label
of the encapsulated packet is used for that purpose. The GRE key
field is sometimes considered a security feature, functioning as a
32-bit cleartext password, but this is an extremely weak form of
security. In order (a) to facilitate high-speed implementations of
the encapsulation/decapsulation procedures and (b) to ensure
interoperability, we require that all implementations be able to
operate correctly without these optional fields.
More precisely, an implementation of an MPLS-in-GRE decapsulator MUST
be able to process packets correctly without these optional fields.
It MAY be able to process packets correctly with these optional
fields.
An implementation of an MPLS-in-GRE encapsulator MUST be able to
generate packets without these optional fields. It MAY have the
capability to generate packets with these fields, but the default
state MUST be that packets are generated without these fields. The
encapsulator MUST NOT include any of these optional fields unless it
is known that the decapsulator can process them correctly. Methods
for conveying this knowledge are outside the scope of this
specification.
5. Common Procedures
Certain procedures are common to both the MPLS-in-IP and the MPLS-
in-GRE encapsulations. In the following, the encapsulator, whose
address appears in the IP source address field of the encapsulating
IP header, is known as the "tunnel head". The decapsulator, whose
address appears in the IP destination address field of the
decapsulating IP header, is known as the "tunnel tail".
If IPv6 is being used (for either MPLS-in-IPv6 or MPLS-in-GRE-in-
IPv6), the procedures of [RFC2473] are generally applicable.
5.1. Preventing Fragmentation and Reassembly
If an MPLS-in-IP or MPLS-in-GRE packet were fragmented (due to
"ordinary" IP fragmentation), the tunnel tail would have to
reassemble it before the contained MPLS packet could be decapsulated.
When the tunnel tail is a router, this is likely to be undesirable;
the tunnel tail may not have the ability or the resources to perform
reassembly at the necessary level of performance.
Whether fragmentation of the tunneled packets is allowed MUST be
configurable at the tunnel head. The default value MUST be that
packets are not fragmented. The default value would only be changed
if it were known that the tunnel tail could perform the reassembly
function adequately.
THE PROCEDURES SPECIFIED IN THE REMAINDER OF THIS SECTION ONLY APPLY
IF PACKETS ARE NOT TO BE FRAGMENTED.
Obviously, if packets are not to be fragmented, the tunnel head MUST
NOT fragment a packet before encapsulating it.
If IPv4 is used, then the tunnel MUST set the DF bit. This prevents
intermediate nodes in the tunnel from performing fragmentation. (If
IPv6 is used, intermediate nodes do not perform fragmentation in any
event.)
The tunnel head SHOULD perform Path MTU Discovery ([RFC1191] for
IPv4, or [RFC1981] for IPv6).
The tunnel head MUST maintain a "Tunnel MTU" for each tunnel; this is
the minimum of (a) an administratively configured value, and, if
known, (b) the discovered Path MTU value minus the encapsulation
overhead.
If the tunnel head receives, for encapsulation, an MPLS packet whose
size exceeds the Tunnel MTU, that packet MUST be discarded. However,
silently dropping such packets may cause significant operational
problems; the originator of the packets will notice that his data is
not getting through, but he may not realize that large packets are
causing packet loss. He may therefore continue sending packets that
are discarded. Path MTU discovery can help (if the tunnel head sends
back ICMP errors), but frequently there is insufficient information
available at the tunnel head to identify the originating sender
properly. To minimize problems, it is advised that MTUs be
engineered to be large enough in practice to avoid fragmentation.
In some cases, the tunnel head receives, for encapsulation, an IP
packet, which it first encapsulates in MPLS and then encapsulates in
MPLS-in-IP or MPLS-in-GRE. If the source of the IP packet is
reachable from the tunnel head, and if the result of encapsulating
the packet in MPLS would be a packet whose size exceeds the Tunnel
MTU, then the value that the tunnel head SHOULD use for fragmentation
and PMTU discovery outside the tunnel is the Tunnel MTU value minus
the size of the MPLS encapsulation. (That is, the Tunnel MTU value
minus the size of the MPLS encapsulation is the MTU that is to be
reported in ICMP messages.) The packet will have to be discarded,
but the tunnel head should send the IP source of the discarded packet
the proper ICMP error message as specified in [RFC1191] or [RFC1981].
5.2. TTL or Hop Limit
The tunnel head MAY place the TTL from the MPLS label stack into the
TTL field of the encapsulating IPv4 header or the Hop Limit field of
the encapsulating IPv6 header. The tunnel tail MAY place the TTL
from the encapsulating IPv4 header or the Hop Limit from the
encapsulating IPv6 header into the TTL field of the MPLS header, but
only if this does not increase the TTL value in the MPLS header.
Whether such modifications are made, and the details of how they are
made, will depend on the configuration of the tunnel tail and the
tunnel head.
5.3. Differentiated Services
The procedures specified in this document enable an LSP to be sent
through an IP or GRE tunnel. [RFC2983] details a number of
considerations and procedures that have to be applied to support the
Differentiated Services Architecture properly in the presence of IP-
in-IP tunnels. These considerations and procedures also apply in the
presence of MPLS-in-IP or MPLS-in-GRE tunnels.
Accordingly, when a tunnel head is about to send an MPLS packet into
an MPLS-in-IP or MPLS-in-GRE tunnel, the setting of the DS field of
the encapsulating IPv4 or IPv6 header MAY be determined (at least
partially) by the "Behavior Aggregate" of the MPLS packet.
Procedures for determining the Behavior Aggregate of an MPLS packet
are specified in [RFC3270].
Similarly, at the tunnel tail, the DS field of the encapsulating IPv4
or IPv6 header MAY be used to determine the Behavior Aggregate of the
encapsulated MPLS packet. [RFC3270] specifies the relation between
the Behavior Aggregate and the subsequent disposition of the packet.
6. Applicability
The MPLS-in-IP encapsulation is the more efficient, and it would
generally be regarded as preferable, other things being equal. There
are, however, some situations in which the MPLS-in-GRE encapsulation
may be used:
- Two routers are "adjacent" over a GRE tunnel that exists for
some reason that is outside the scope of this document, and
those two routers have to send MPLS packets over that
adjacency. As all packets sent over this adjacency must have a
GRE encapsulation, the MPLS-in-GRE encapsulation is more
efficient than the alternative, that would be an MPLS-in-IP
encapsulation which is then encapsulated in GRE.
- Implementation considerations may dictate the use of MPLS-in-
GRE. For example, some hardware device might only be able to
handle GRE encapsulations in its fastpath.
7. IANA Considerations
The IANA has allocated IP Protocol Number 137 for MPLS-in-IP
encapsulation, as described in section 3. No future IANA actions
will be required. The MPLS-in-GRE encapsulation does not require any
IANA action.
8. Security Considerations
The main security problem faced when IP or GRE tunnels are used is
the possibility that the tunnel’s receive endpoint will get a packet
that appears to be from the tunnel, but that was not actually put
into the tunnel by the tunnel’s transmit endpoint. (The specified
encapsulations do not by themselves enable the decapsulator to
authenticate the encapsulator.) A second problem is the possibility
that the packet will be altered between the time it enters the tunnel
and the time it leaves. (The specified encapsulations do not by
themselves assure the decapsulator of the packet’s integrity.) A
third problem is the possibility that the packet’s contents will be
seen while the packet is in transit through the tunnel. (The
specification encapsulations do not ensure privacy.) How significant
these issues are in practice depends on the security requirements of
the applications whose traffic is being sent through the tunnel. For
example, lack of privacy for tunneled packets is not a significant
issue if the applications generating the packets do not require
privacy.
Because of the different potential security requirements, deployment
scenarios, and performance considerations of different applications
using the described encapsulation mechanism, this specification
defines IPsec support as OPTIONAL. Basic implementation requirements
if IPsec is implemented are described in section 8.1. If IPsec is
not implemented, additional mechanisms may have to be implemented and
deployed. Those are discussed in section 8.2.
8.1. Securing the Tunnel with IPsec
All of these security issues can be avoided if the MPLS-in-IP or
MPLS-in-GRE tunnels are secured with IPsec. Implementation
requirements defined in this section apply if IPsec is implemented.
When IPsec is used, the tunnel head and the tunnel tail should be
treated as the endpoints of a Security Association. For this
purpose, a single IP address of the tunnel head will be used as the
source IP address, and a single IP address of the tunnel tail will be
used as the destination IP address. The means by which each node
knows the proper address of the other is outside the scope of this
document. If a control protocol is used to set up the tunnels (e.g.,
to inform one tunnel endpoint of the IP address of the other), the
control protocol MUST have an authentication mechanism, and this MUST
be used when the tunnel is set up. If the tunnel is set up
automatically as the result of, for example, information distributed
by BGP, then the use of BGP’s MD5-based authentication mechanism is
satisfactory.
The MPLS-in-IP or MPLS-in-GRE encapsulated packets should be viewed
as originating at the tunnel head and as being destined for the
tunnel tail; IPsec transport mode SHOULD thus be used.
The IP header of the MPLS-in-IP packet becomes the outer IP header of
the resulting packet when the tunnel head uses IPsec transport mode
to secure the MPLS-in-IP packet. This is followed by an IPsec
header, followed by the MPLS label stack. The IPsec header has to
set the payload type to MPLS by using the IP protocol number
specified in section 3. If IPsec transport mode is applied on a
MPLS-in-GRE packet, the GRE header follows the IPsec header.
At the tunnel tail, IPsec outbound processing recovers the contained
MPLS-in-IP/GRE packet. The tunnel tail then strips off the
encapsulating IP/GRE header to recover the MPLS packet, which is then
forwarded according to its label stack.
Note that the tunnel tail and the tunnel head are LSP adjacencies,
which means that the topmost label of any packet sent through the
tunnel must be one that was distributed by the tunnel tail to the
tunnel head. The tunnel tail MUST know precisely which labels it has
distributed to the tunnel heads of IPsec-secured tunnels. Labels in
this set MUST NOT be distributed by the tunnel tail to any LSP
adjacencies other than those that are tunnel heads of IPsec-secured
tunnels. If an MPLS packet is received without an IPsec
encapsulation, and if its topmost label is in this set, then the
packet MUST be discarded.
An IPsec-secured MPLS-in-IP or MPLS-in-GRE tunnel MUST provide
authentication and integrity. (Note that the authentication and
integrity will apply to the entire MPLS packet, including the MPLS
label stack.) Thus, the implementation MUST support ESP will null
encryption. ESP with encryption MAY be supported if a source
requires confidentiality. If ESP is used, the tunnel tail MUST check
that the source IP address of any packet received on a given SA is
the one expected.
Key distribution may be done either manually or automatically by
means of IKE [RFC2409]. Manual keying MUST be supported. If
automatic keying is implemented, IKE in main mode with preshared keys
MUST be supported. A particular application may escalate this
requirement and request implementation of automatic keying.
Manual key distribution is much simpler, but also less scalable, than
automatic key distribution. Therefore, which method of key
distribution is appropriate for a particular tunnel has to be
carefully considered by the administrator (or pair of administrators)
responsible for the tunnel endpoints. If replay protection is
regarded as necessary for a particular tunnel, automatic key
distribution should be configured.
If the MPLS-in-IP encapsulation is being used, the selectors
associated with the SA would be the source and destination addresses
mentioned above, plus the IP protocol number specified in section 3.
If it is desired to secure multiple MPLS-in-IP tunnels between a
given pair of nodes separately, each tunnel must have unique pair of
IP addresses.
If the MPLS-in-GRE encapsulation is being used, the selectors
associated with the SA would be the source and destination addresses
mentioned above, and the IP protocol number representing GRE (47).
If it is desired to secure multiple MPLS-in-GRE tunnels between a
given pair of nodes separately, each tunnel must have unique pair of
IP addresses.
8.2. In the Absence of IPsec
If the tunnels are not secured with IPsec, then some other method
should be used to ensure that packets are decapsulated and forwarded
by the tunnel tail only if those packets were encapsulated by the
tunnel head. If the tunnel lies entirely within a single
administrative domain, address filtering at the boundaries can be
used to ensure that no packet with the IP source address of a tunnel
endpoint or with the IP destination address of a tunnel endpoint can
enter the domain from outside.
However, when the tunnel head and the tunnel tail are not in the same
administrative domain, this may become difficult, and filtering based
on the destination address can even become impossible if the packets
must traverse the public Internet.
Sometimes only source address filtering (but not destination address
filtering) is done at the boundaries of an administrative domain. If
this is the case, the filtering does not provide effective protection
at all unless the decapsulator of an MPLS-in-IP or MPLS-in-GRE
validates the IP source address of the packet. This document does
not require that the decapsulator validate the IP source address of
the tunneled packets, but it should be understood that failure to do
so presupposes that there is effective destination-based (or a