| RFC 8956 | IPv6 Flow Specification | December 2020 |
| Loibl, et al. | Standards Track | [Page] |
"Dissemination of Flow Specification Rules" (RFC 8955) provides a Border Gateway Protocol (BGP) extension for the propagation of traffic flow information for the purpose of rate limiting or filtering IPv4 protocol data packets.¶
This document extends RFC 8955 with IPv6 functionality. It also updates RFC 8955 by changing the IANA Flow Spec Component Types registry.¶
This is an Internet Standards Track document.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc8956.¶
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.¶
The growing amount of IPv6 traffic in private and public networks requires the extension of tools used in IPv4-only networks to also support IPv6 data packets.¶
This document analyzes the differences between describing IPv6 [RFC8200] flows and those of IPv4 packets. It specifies new Border Gateway Protocol [RFC4271] encoding formats to enable "Dissemination of Flow Specification Rules" [RFC8955] for IPv6.¶
This specification is an extension of the base established in [RFC8955]. It only defines the delta changes required to support IPv6, while all other definitions and operation mechanisms of "Dissemination of Flow Specification Rules" will remain in the main specification and will not be repeated here.¶
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.¶
[RFC8955] defines SAFIs 133 (Dissemination of Flow Specification rules) and 134 (L3VPN Dissemination of Flow Specification rules) in order to carry the corresponding Flow Specification.¶
Implementations wishing to exchange IPv6 Flow Specifications MUST use BGP's Capability Advertisement facility to exchange the Multiprotocol Extension Capability Code (Code 1), as defined in [RFC4760]. The (AFI, SAFI) pair carried in the Multiprotocol Extension Capability MUST be (AFI=2, SAFI=133) for IPv6 Flow Specification rules and (AFI=2, SAFI=134) for L3VPN Dissemination of Flow Specification rules.¶
The encoding of each of the components begins with a Type field (1 octet) followed by a variable length parameter. The following sections define component types and parameter encodings for IPv6.¶
Types 4 (Port), 5 (Destination Port), 6 (Source Port), 9 (TCP Flags), 10 (Packet Length), and 11 (DSCP), as defined in [RFC8955], also apply to IPv6. Note that IANA has updated the "Flow Spec Component Types" registry in order to contain both IPv4 and IPv6 Flow Specification component type numbers in a single registry (Section 8).¶
This defines the destination prefix to match. The offset has been defined to allow for flexible matching to portions of an IPv6 address where one is required to skip over the first N bits of the address. (These bits skipped are often indicated as "don't care" bits.) This can be especially useful where part of the IPv6 address consists of an embedded IPv4 address, and matching needs to happen only on the embedded IPv4 address. The encoded pattern contains enough octets for the bits used in matching (length minus offset bits).¶
If length = 0 and offset = 0, this component matches every address; otherwise, length MUST be in the range offset < length < 129 or the component is malformed.¶
Note: This Flow Specification component can be represented by the notation ipv6address/length if offset is 0 or ipv6address/offset-length. The ipv6address in this notation is the textual IPv6 representation of the pattern shifted to the right by the number of offset bits. See also Section 3.8.¶
This defines the source prefix to match. The length, offset, pattern, and padding are the same as in Section 3.1.¶
This contains a list of {numeric_op, value} pairs that are used to match the first Next Header value octet in IPv6 packets that is not an extension header and thus indicates that the next item in the packet is the corresponding upper-layer header (see Section 4 of [RFC8200]).¶
This component uses the Numeric Operator (numeric_op) described in Section 4.2.1.1 of [RFC8955]. Type 3 component values SHOULD be encoded as a single octet (numeric_op len=00).¶
Note: While IPv6 allows for more than one Next Header field in the packet, the main goal of the Type 3 Flow Specification component is to match on the first upper-layer IP protocol value. Therefore, the definition is limited to match only on this specific Next Header field in the packet.¶
This defines a list of {numeric_op, value} pairs used to match the Type field of an ICMPv6 packet (see also Section 2.1 of [RFC4443]).¶
This component uses the Numeric Operator (numeric_op) described in Section 4.2.1.1 of [RFC8955]. Type 7 component values SHOULD be encoded as a single octet (numeric_op len=00).¶
In case of the presence of the ICMPv6 type component, only ICMPv6 packets can match the entire Flow Specification. The ICMPv6 type component, if present, never matches when the packet's upper-layer IP protocol value is not 58 (ICMPv6), if the packet is fragmented and this is not the first fragment, or if the system is unable to locate the transport header. Different implementations may or may not be able to decode the transport header.¶
This defines a list of {numeric_op, value} pairs used to match the code field of an ICMPv6 packet (see also Section 2.1 of [RFC4443]).¶
This component uses the Numeric Operator (numeric_op) described in Section 4.2.1.1 of [RFC8955]. Type 8 component values SHOULD be encoded as a single octet (numeric_op len=00).¶
In case of the presence of the ICMPv6 code component, only ICMPv6 packets can match the entire Flow Specification. The ICMPv6 code component, if present, never matches when the packet's upper-layer IP protocol value is not 58 (ICMPv6), if the packet is fragmented and this is not the first fragment, or if the system is unable to locate the transport header. Different implementations may or may not be able to decode the transport header.¶
This defines a list of {bitmask_op, bitmask} pairs used to match specific IP fragments.¶
This component uses the Bitmask Operator (bitmask_op) described in Section 4.2.1.2 of [RFC8955]. The Type 12 component bitmask MUST be encoded as a single octet bitmask (bitmask_op len=00).¶
0 1 2 3 4 5 6 7
+---+---+---+---+---+---+---+---+
| 0 | 0 | 0 | 0 |LF |FF |IsF| 0 |
+---+---+---+---+---+---+---+---+
Bitmask values:¶
This contains a list of {numeric_op, value} pairs that are used to match the 20-bit Flow Label IPv6 header field (Section 3 of [RFC8200]).¶
This component uses the Numeric Operator (numeric_op) described in Section 4.2.1.1 of [RFC8955]. Type 13 component values SHOULD be encoded as 4-octet quantities (numeric_op len=10).¶
The following example demonstrates the prefix encoding for packets from ::1234:5678:9a00:0/64-104 to 2001:db8::/32 and upper-layer protocol tcp.¶
| len | destination | source | ul-proto |
|---|---|---|---|
| 0x12 | 01 20 00 20 01 0d bb | 02 68 40 12 34 56 78 9a | 03 81 06 |
Decoded:¶
| Value | ||
|---|---|---|
| 0x12 | length | 18 octets (if len<240, 1 octet) |
| 0x01 | type | Type 1 - Dest. IPv6 Prefix |
| 0x20 | length | 32 bits |
| 0x00 | offset | 0 bits |
| 0x20 | pattern | |
| 0x01 | pattern | |
| 0x0d | pattern | |
| 0xb8 | pattern | (no padding needed) |
| 0x02 | type | Type 2 - Source IPv6 Prefix |
| 0x68 | length | 104 bits |
| 0x40 | offset | 64 bits |
| 0x12 | pattern | |
| 0x34 | pattern | |
| 0x56 | pattern | |
| 0x78 | pattern | |
| 0x9a | pattern | (no padding needed) |
| 0x03 | type | Type 3 - Upper-Layer Protocol |
| 0x81 | numeric_op | end-of-list, value size=1, == |
| 0x06 | value | 06 |
This constitutes an NLRI with an NLRI length of 18 octets.¶
Padding is not needed either for the destination prefix pattern (length - offset = 32 bits) or for the source prefix pattern (length - offset = 40 bits), as both patterns end on an octet boundary.¶
The following example demonstrates the prefix encoding for all packets from ::1234:5678:9a00:0/65-104 to 2001:db8::/32.¶
| length | destination | source |
|---|---|---|
| 0x0f | 01 20 00 20 01 0d b8 | 02 68 41 24 68 ac f1 34 |
Decoded:¶
| Value | ||
|---|---|---|
| 0x0f | length | 15 octets (if len<240, 1 octet) |
| 0x01 | type | Type 1 - Dest. IPv6 Prefix |
| 0x20 | length | 32 bits |
| 0x00 | offset | 0 bits |
| 0x20 | pattern | |
| 0x01 | pattern | |
| 0x0d | pattern | |
| 0xb8 | pattern | (no padding needed) |
| 0x02 | type | Type 2 - Source IPv6 Prefix |
| 0x68 | length | 104 bits |
| 0x41 | offset | 65 bits |
| 0x24 | pattern | |
| 0x68 | pattern | |
| 0xac | pattern | |
| 0xf1 | pattern | |
| 0x34 | pattern/pad | (contains 1 bit of padding) |
This constitutes an NLRI with an NLRI length of 15 octets.¶
The source prefix pattern is 104 - 65 = 39 bits in length. After the pattern, one bit of padding needs to be added so that the component ends on an octet boundary. However, only the first 39 bits are actually used for bitwise pattern matching, starting with a 65-bit offset from the topmost bit of the address.¶
The definition for the order of traffic filtering rules from Section 5.1 of [RFC8955] is reused with new consideration for the IPv6 prefix offset. As long as the offsets are equal, the comparison is the same, retaining longest-prefix-match semantics. If the offsets are not equal, the lowest offset has precedence, as this Flow Specification matches the most significant bit.¶
The code in Appendix A shows a Python3 implementation of the resulting comparison algorithm. The full code was tested with Python 3.7.2 and can be obtained at <https://github.com/stoffi92/draft-ietf-idr-flow-spec-v6/tree/master/flowspec-cmp>.¶
The validation procedure is the same as specified in Section 6 of [RFC8955] with the exception that item a) of the validation procedure should now read as follows:¶
- a)
- A destination prefix component with offset=0 is embedded in the Flow Specification¶
Traffic Filtering Actions from Section 7 of [RFC8955] can also be applied to IPv6 Flow Specifications. To allow an IPv6-Address-Specific Route-Target, a new Traffic Filtering Action IPv6-Address-Specific Extended Community is specified in Section 6.1 below.¶
The redirect IPv6-Address-Specific Extended Community allows the traffic to be redirected to a VRF routing instance that lists the specified IPv6-Address-Specific Route-Target in its import policy. If several local instances match this criteria, the choice between them is a local matter (for example, the instance with the lowest Route Distinguisher value can be elected).¶
This IPv6-Address-Specific Extended Community uses the same encoding as the IPv6-Address-Specific Route-Target Extended Community (Section 2 of [RFC5701]) with the Type value always 0x000d.¶
The Local Administrator subfield contains a number from a numbering space that is administered by the organization to which the IP address carried in the Global Administrator subfield has been assigned by an appropriate authority.¶
Interferes with: All BGP Flow Specification redirect Traffic Filtering Actions (with itself and those specified in Section 7.4 of [RFC8955]).¶
This document extends the functionality in [RFC8955] to be applicable to IPv6 data packets. The same security considerations from [RFC8955] now also apply to IPv6 networks.¶
[RFC7112] describes the impact of oversized IPv6 header chains when trying to match on the transport header; Section 4.5 of [RFC8200] also requires that the first fragment must include the upper-layer header, but there could be wrongly formatted packets not respecting [RFC8200]. IPv6 Flow Specification component Type 3 (Section 3.3) will not be enforced for those illegal packets. Moreover, there are hardware limitations in several routers (Section 1 of [RFC8883]) that may make it impossible to enforce a policy signaled by a Type 3 Flow Specification component or Flow Specification components that match on upper-layer properties of the packet.¶
This section complies with [RFC7153].¶
IANA has created and maintains a registry entitled "Flow Spec Component Types". IANA has added this document as a reference for that registry. Furthermore, the registry has been updated to also contain the IPv6 Flow Specification Component Types as described below. The registration procedure remains unchanged.¶
IANA maintains a registry entitled "Transitive IPv6-Address-Specific Extended Community Types". For the purpose of this work, IANA has assigned a new value:¶
| Type Value | Name | Reference |
|---|---|---|
| 0x000d | Flow spec rt-redirect-ipv6 format | RFC 8956 |
<CODE BEGINS>
"""
Copyright (c) 2020 IETF Trust and the persons identified as authors
of the code. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, is permitted pursuant to, and subject to the license
terms contained in, the Simplified BSD License set forth in Section
4.c of the IETF Trust's Legal Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info).
"""
import itertools
import collections
import ipaddress
EQUAL = 0
A_HAS_PRECEDENCE = 1
B_HAS_PRECEDENCE = 2
IP_DESTINATION = 1
IP_SOURCE = 2
FS_component = collections.namedtuple('FS_component',
'component_type value')
class FS_IPv6_prefix_component:
def __init__(self, prefix, offset=0,
component_type=IP_DESTINATION):
self.offset = offset
self.component_type = component_type
# make sure if offset != 0 that none of the
# first offset bits are set in the prefix
self.value = prefix
if offset != 0:
i = ipaddress.IPv6Interface(
(self.value.network_address, offset))
if i.network.network_address != \
ipaddress.ip_address('0::0'):
raise ValueError('Bits set in the offset')
class FS_nlri(object):
"""
FS_nlri class implementation that allows sorting.
By calling .sort() on an array of FS_nlri objects these
will be sorted according to the flow_rule_cmp algorithm.
Example:
nlri = [ FS_nlri(components=[
FS_component(component_type=4,
value=bytearray([0,1,2,3,4,5,6])),
]),
FS_nlri(components=[
FS_component(component_type=5,
value=bytearray([0,1,2,3,4,5,6])),
FS_component(component_type=6,
value=bytearray([0,1,2,3,4,5,6])),
]),
]
nlri.sort() # sorts the array according to the algorithm
"""
def __init__(self, components = None):
"""
components: list of type FS_component
"""
self.components = components
def __lt__(self, other):
# use the below algorithm for sorting
result = flow_rule_cmp_v6(self, other)
if result == B_HAS_PRECEDENCE:
return True
else:
return False
def flow_rule_cmp_v6(a, b):
"""
Implementation of the flowspec sorting algorithm in
RFC 8956.
"""
for comp_a, comp_b in itertools.zip_longest(a.components,
b.components):
# If a component type does not exist in one rule
# this rule has lower precedence
if not comp_a:
return B_HAS_PRECEDENCE
if not comp_b:
return A_HAS_PRECEDENCE
# Higher precedence for lower component type
if comp_a.component_type < comp_b.component_type:
return A_HAS_PRECEDENCE
if comp_a.component_type > comp_b.component_type:
return B_HAS_PRECEDENCE
# component types are equal -> type-specific comparison
if comp_a.component_type in (IP_DESTINATION, IP_SOURCE):
if comp_a.offset < comp_b.offset:
return A_HAS_PRECEDENCE
if comp_a.offset > comp_b.offset:
return B_HAS_PRECEDENCE
# both components have the same offset
# assuming comp_a.value, comp_b.value of type
# ipaddress.IPv6Network
# and the offset bits are reset to 0 (since they are
# not represented in the NLRI)
if comp_a.value.overlaps(comp_b.value):
# longest prefixlen has precedence
if comp_a.value.prefixlen > \
comp_b.value.prefixlen:
return A_HAS_PRECEDENCE
if comp_a.value.prefixlen < \
comp_b.value.prefixlen:
return B_HAS_PRECEDENCE
# components equal -> continue with next
# component
elif comp_a.value > comp_b.value:
return B_HAS_PRECEDENCE
elif comp_a.value < comp_b.value:
return A_HAS_PRECEDENCE
else:
# assuming comp_a.value, comp_b.value of type
# bytearray
if len(comp_a.value) == len(comp_b.value):
if comp_a.value > comp_b.value:
return B_HAS_PRECEDENCE
if comp_a.value < comp_b.value:
return A_HAS_PRECEDENCE
# components equal -> continue with next
# component
else:
common = min(len(comp_a.value),
len(comp_b.value))
if comp_a.value[:common] > \
comp_b.value[:common]:
return B_HAS_PRECEDENCE
elif comp_a.value[:common] < \
comp_b.value[:common]:
return A_HAS_PRECEDENCE
# the first common bytes match
elif len(comp_a.value) > len(comp_b.value):
return A_HAS_PRECEDENCE
else:
return B_HAS_PRECEDENCE
return EQUAL
<CODE ENDS>¶
The authors would like to thank Pedro Marques, Hannes Gredler, Bruno Rijsman, Brian Carpenter, and Thomas Mangin for their valuable input.¶