Internet DRAFT - draft-herbert-eh-inflight-removal
draft-herbert-eh-inflight-removal
Network Working Group T. Herbert
Internet-Draft SiPanda
Intended status: Experimental 23 February 2024
Expires: 26 August 2024
Infight Removal of IPv6 Hop-by-Hop and Routing Headers
draft-herbert-eh-inflight-removal-04
Abstract
This document specifies an experimental method to allow routers to
remove IPv6 Hop-by-Hop Options or Routing headers from packets in-
flight. The goal is to reduce the probability of packets being
dropped because they contain extension headers, without adversely
impacting functionality. An additional goal is to limit visibility
of information in extension headers to those nodes that need to
process the headers.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on 26 August 2024.
Copyright Notice
Copyright (c) 2024 IETF Trust and the persons identified as the
document authors. All rights reserved.
Herbert Expires 26 August 2024 [Page 1]
Internet-Draft Inflight-EH-Removal February 2024
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 Revised BSD License text as
described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Revised BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 3
2. Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Hop-by-Hop Options drop rate . . . . . . . . . . . . . . 3
2.2. Router Header domain firewall . . . . . . . . . . . . . . 4
2.3. Removing extension headers . . . . . . . . . . . . . . . 5
2.3.1. Removal by egress routers . . . . . . . . . . . . . . 5
2.3.2. Removal by ingress routers . . . . . . . . . . . . . 5
2.4. Alternatives to Extension Header removal . . . . . . . . 5
2.4.1. Host routing . . . . . . . . . . . . . . . . . . . . 6
2.4.2. Probing . . . . . . . . . . . . . . . . . . . . . . . 6
2.4.3. IPinIP Encapsulation from source . . . . . . . . . . 7
2.4.4. IPinIP Encapsulation from egress router . . . . . . . 8
3. Arguments against in-flight extension header removal . . . . 9
4. Considerations . . . . . . . . . . . . . . . . . . . . . . . 10
4.1. Reflection of Hop-by-Hop Options . . . . . . . . . . . . 10
4.2. End host processing of Routing Headers . . . . . . . . . 10
4.3. ICMP errors . . . . . . . . . . . . . . . . . . . . . . . 11
5. Requirements . . . . . . . . . . . . . . . . . . . . . . . . 11
6. Procedures . . . . . . . . . . . . . . . . . . . . . . . . . 12
6.1. Removing a Hop-by-Hop Options Header . . . . . . . . . . 12
6.2. Removing a Routing Header . . . . . . . . . . . . . . . . 14
6.3. Removing both a Hop-by-Hop Options and a Routing
header . . . . . . . . . . . . . . . . . . . . . . . . . 17
7. Implementation Considerations . . . . . . . . . . . . . . . . 20
7.1. Copying the IPv6 Header . . . . . . . . . . . . . . . . . 21
7.2. Scatter/gather . . . . . . . . . . . . . . . . . . . . . 21
8. Security Considerations . . . . . . . . . . . . . . . . . . . 21
9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21
10. References . . . . . . . . . . . . . . . . . . . . . . . . . 21
10.1. Normative References . . . . . . . . . . . . . . . . . . 21
10.2. Informative References . . . . . . . . . . . . . . . . . 21
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 23
Herbert Expires 26 August 2024 [Page 2]
Internet-Draft Inflight-EH-Removal February 2024
1. Introduction
This document specifies an experimental protocol for routers to
remove IPv6 Hop-by-Hop Options Routing headers from packets in-
flight.
Current data suggests that there are very high drop rates for packets
with Hop-by-Hop Options sent on the Internet [APNIC-EH] [Cus23a].
The goal of this protocol is to reduce the probability of the packet
being dropped because they contain extension without reducing
functionality, thereby improving the viability and usability for
sending Hop-by-Hop Options.
A secondary goal is to allow removal of Hop-by-Hop Options or Routing
headers when packets egress a limited domain [RFC8799], such as a
segment routing domain, in order to limit exposure of data to only
those nodes that legitimately need to process it.
This specification is limited only to removal of the whole Hop-by-Hop
Options header or Routing header. It does not set requirements for
removing individual Hop-by-Hop options in a Hop-by-Hop Options
header, nor does it specify any method for routers to insert a Hop-
by-Hop Options header, options in a Hop-by-Hop header, or a Routing
header in packets.
1.1. Terminology
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 [RFC2119].
2. Motivation
This section provides the motivations for allowing routers to remove
Hop-by-Hop Options or Routing headers from packets in-flight.
2.1. Hop-by-Hop Options drop rate
Current measurements suggest that packets with Hop-by-Hop headers
have high drop rates when sent over the Internet. From [APNIC-EH]:
The HBH option was experiencing an average packet drop rate of
99.5% across all HBH option sizes
The reported drops rates for Hop-by-Hop Options are greater than that
of packets with Destination Options or Fragment headers. A plausible
explanation for this difference is that Hop-by-Hop Options are
intended to be processed by routers in a network, and hence a network
Herbert Expires 26 August 2024 [Page 3]
Internet-Draft Inflight-EH-Removal February 2024
operator may be motivated to drop packets with Hop-by-Hop options
entering their network from untrusted sources to protect their
network infrastructure. This is mentioned in [RFC9098] as a reason
that packets containing IPv6 Hop-by-Hop Options are dropped:
The Hop-by-Hop Options header has been particularly challenging
since, in most circumstances, the corresponding packet is punted
to the control plane for processing. As a result, many operators
drop IPv6 packets containing this extension header [RFC7872].
[RFC6192] provides advice regarding protection of a router's
control plane.
Given that there doesn't seem to be a easy fix to make Hop-by-Hop
Options work over the Internet, the commonly proposed alternative is
to limit use of Hop-by-Hop Options to limited domains [RFC8799]. It
can be noted that Hop-by-Hop Options are only useful when at least
some of nodes in the path process them, and a network operator would
likely only deploy routers that process Hop-by-Hop Options if they
perceived Hop-by-Hop Options provide some value. An example of such
an option might be FAST [I-D.herbert-fast] which allows the network
infrastructure to provide fine grained QoS and monetize network
services on a per packet basis. If a network supports value add
services that use Hop-by-Hop Options, it stands to reason that
packets with Hop-by-Hop Options wouldn't be dropped while their
within the limited domain of the network operator.
If a destination is outside the limited domain of the source host, a
source host might still desire to use Hop-by-Hop Options to affect
packet processing in the part of the path that is within the limited
domain. In this case, a packet might be created with a Hop-by-Hop
Options header, the packet traverses the local network to an egress
router, and at the egress router the packet is forwarded outside of
the limited domain without Hop-by-Hop Options (presumably by removing
the Hop-by-Hop Options header).
2.2. Router Header domain firewall
When a host sends a packet with a Routing header, for example a
Segment Routing header, the intermediate destinations are considered
to be in the same limited domain; for example, in Segment Routing all
of the intermediate destinations in the Segment Routing header must
be in the same segment routing domain.
The final destination of a Routing header might not be in the routing
domain. It may, in fact, be outside of the limited domain. An
example use case of this would be if routing was used to route the
packet to an egress router of the domain. The egress router would be
Herbert Expires 26 August 2024 [Page 4]
Internet-Draft Inflight-EH-Removal February 2024
the penultimate destination in the segment list such that the
Segments Left field is set to zero and all downstream nodes would
ignore the Routing header. In this case, the packets can be
forwarded beyond the limited domain without a Routing header and with
no adverse impact on functionality.
2.3. Removing extension headers
2.3.1. Removal by egress routers
To contain Hop-by-Hop Options and Routing headers to their limited
domain, this specification proposes that egress routers may remove
the extension headers from packets before forwarding them beyond the
limited domain.
Hop-by-Hop Options would be removed by an egress router in order to
increase the likelihood that packets sent with Hop-by-Hop Options are
successfully delivered to their destination. The assumption is that
the Hop-by-Hop Options are most likely not useful beyond the limited
domain, so removing them from packets when they exit their domain
would have no impact on functionality. Option reflection to affect
processing in the reverse direction of a flow, such as defined in
FAST [I-D.herbert-fast], is one case where it would be useful to send
outside of a limited domain (discussed in Section 4.1).
A Routing header would be removed at an egress router when it's being
used to route a packet beyond the limited domain. Note that when the
penultimate destination processes the Routing header, it sets the
final Destination Address and Segments Left to zero, so at that point
the Routing header can be removed without impacting further
processing of the packet since no downstream routers nor the
destination host processes the Routing header.
2.3.2. Removal by ingress routers
Hop-by-Hop Options could be removed from packets by ingress routers
as an alternative to the current practice of dropping the packets
with Hop-by-Hop Options. In this case, the network operator doesn't
process Hop-by-Hop Options, or it only processes Hop-by-Hop Options
from source hosts in the local domain that it trusts. Removing Hop-
by-Hop Options instead of dropping them allows packets to be
delivered without loss of functionality or risk to the network
infrastructure.
2.4. Alternatives to Extension Header removal
This section discusses some of the alternatives to extension header
removal that have been proposed.
Herbert Expires 26 August 2024 [Page 5]
Internet-Draft Inflight-EH-Removal February 2024
2.4.1. Host routing
It is conceivable that a host network stack could maintain routes to
destinations or networks with an indication that the destination is
within the limited domain. So when a packet is being created, the
routing table could be consulted to determine if it's safe to send
packets with Hop-by-Hop Options to the destination.
The main drawback of this approach is that it requires significant
changes to the host network stack. The routing infrastructure in the
host, the APIs presented to the application trying to set Hop-by-Hop
Options, and probably applications themselves may need to change.
Additionally, it isn't always obvious just given an address whether
the host could determine if the destination is in the same limited
domain as the source host. In some simpler topologies, it might be
possible to configure hosts with all the network prefixes that belong
to the limited domain, however for a more complex topology, hosts may
need to participate in a routing protocol or a discovery protocol
with the network.
2.4.2. Probing
Capabilities probing has been successfully employed in other contexts
such as "Happy Eyeballs" for IPv6. Conceptually, probing could
similarly be used to determine the viability of Hop-by-Hop Options to
a destination. In this case, a host could probe each destination to
determine if Hop-by-Hop Options are viable. An advantage of this
method is that it requires no special assistance from the network.
The main drawback of probing is the complexity in the host network
stack and applications. Probing assumes bidirectional
communications, state needs to be maintained for each destination or
flow, procedures need to be specified for probing, backoff, and
continuous probing in the case of route changes that might affect the
disposition of packet with Hop-by-Hop Options in the network.
Additionally, the implementation for probing would be different for
UDP and TCP: probing in the UDP case would most likely need support
in the application and user space libraries, probing for TCP would
likely need to be supported in the kernel networking stack itself.
Herbert Expires 26 August 2024 [Page 6]
Internet-Draft Inflight-EH-Removal February 2024
2.4.3. IPinIP Encapsulation from source
In order to use Hop-by-Hop Options in the part of the path within a
limited domain, a source host may encapsulate the packet in an IPinIP
encapsulation [RFC2473]. The outer IPv6 header would contain the
Hop-by-Hop Options header and the destination would be the address of
an egress router for the limited domain. At the egress router, the
packet would be decapsulated and the packet can be forwarded without
Hop-by-Hop Options.
The main problem to this approach is that the sending host would need
to know the correct Destination Address to set in the encapsulating
header; that is, the host would need to know the address of the
appropriate egress router for the packet. That information is not
usually available to hosts and might not even be available to
intermediate routers including the first hop router. In a complex,
multi-homed network topology that might support mobile hosts, the
only way to determine the current egress router for a packet may be
to actually route through the network to the external destination
address.
If the network did maintain an association between destinations and
the egress router then conceptually it could share that information
with hosts using a routing protocol or discovery protocol. This
information could be saved in an augmented routing table on the host
similar to that described in Section 2.4.1; but as discussed in that
section, this is significant complexity to implement in hosts.
Another drawback is information exposure. If the network provides
the addresses of egress routers to hosts then it is divulging network
topology information that could be considered a security risk.
To avoid exposing addresses of egress routers, hosts could be
conceivably be configured with a single anycast address to be used as
Destination Address of the egress router when encapsulating. If the
host routing table includes limited domain information, as described
in the Section 2.4.1, then this would be sufficient to route packets
to an egress router. In this case though, the anycast address
represents a default router which might not be the same one had the
packet been routed based on its final destination address-- this
could lead be suboptimal routing or cause out-of-order packets if not
all packets of a flow are encapsulated.
Encapsulation is complex from a host implementation point of view.
An IPinIP encapsulation adds at least forty bytes of overhead to the
packet which reduces the effective MTU for the application and
requires special end host processing that may be prohibitive on low
end devices. Even if an anycast address is configured, a host
Herbert Expires 26 August 2024 [Page 7]
Internet-Draft Inflight-EH-Removal February 2024
network stack will need to maintain routing information to determine
when packets need to be encapsulated. Furthermore, setting the Hop-
by-Hop Options is currently done by the application without regard to
whether the packet is being encapsulated. When a packet is sent and
it needs to encapsulated, the host network stack will need to remove
the Hop-by-Hop Options from the original packet and set them in the
encapsulating IPv6 headers.
2.4.4. IPinIP Encapsulation from egress router
Another solution using IPinIP encapsulation would be for an egress
router to encapsulate a packet containing Hop-by-Hop Options in
IPinIP. The outer IPv6 header would contain no Hop-by-Hop Options
and the inner IPv6 header would contain Hop-by-Hop Options. The
Destination Address in the outer and inner IP headers would be the
same.
This solution is not robust since the encapsulation increases packet
size and reduces the Path MTU seen by the sender which can cause
systematic packet drops. For example, suppose a host sends a packet
with minimum MTU size of 1,280, and an egress router encapsulates the
packet so that its length increases to 1,320 bytes. If a downstream
router has a link MTU of 1,280 then the packet will be dropped since
its length exceeds the link MTU. Since the host sent a minimum MTU
sized packet, it cannot fallback to a smaller MTU using using Path
MTU Discovery [RFC8201], and hence there is no recovery. This
precludes the use of encapsulation when packets egress a limited
since there is no expectation that all the potential paths outside of
the domain have a large enough MTU to accommodate the encapsulation.
Sending encapsulated packets into the Internet requires that they can
successfully transit the Internet. The IPinIP encapsulation protocol
number (41) could be filtered by some networks (similar to how
networks can block packets with Hop-by-Hop Options header). Using a
UDP encapsulation, such as VXLAN, might have better success than
IPinIP.
For this method to be viable, all potential receivers would need to
do decapsulation. This could be modeled as an anonymous
encapsulation. Currently, this is not enabled on commodity host
network stacks, and would be a major change to support in deployment.
Herbert Expires 26 August 2024 [Page 8]
Internet-Draft Inflight-EH-Removal February 2024
Packets to a destination may undergo network address translation such
that the outer addresses might not match the inner addresses of an
encapsulation. If a flow contains a mix of encapsulated and non-
encapsulated packets then the destination may view that a packet is
in different flows. In order to prevent this, a router could
encapsulate all packets, but that would be very costly for what is
currently a narrow use case.
3. Arguments against in-flight extension header removal
Section 4 of [I-D.smith-6man-in-flight-eh-insertion-harmful] presents
the problems of in-flight extension header removal in the context of
extension headers being inserted in-flight. If extension headers are
inserted in-flight then it is expected that those headers are removed
before exiting the domain in which they were inserted. Failure to
remove inserted extension headers could have detrimental behaviors
include systematic packet drop and and leaking sensitive information
outside of a limited domain.
This specification only allows removal of extension headers that were
created by the source host, so the problems related to failing to
remove inserted extension headers are not directly relevant.
However, the effects of failing to remove non-inserted extension
headers that we're intended to be removed by the operator can still
be considered.
[I-D.smith-6man-in-flight-eh-insertion-harmful] describes the
possible causes of extension header removal to fail:
* Implementation bugs
* Partial Node Failure
* Operator Configuration Error
With respect to removing non-inserted extension headers, the effects
of these different failure modes are the same.
Given the current data, the most probable effect when extension
headers are not removed as intended is that those packets will be
dropped in the Internet. Since the primary purpose of dropping Hop-
by-Hop or Routing headers is to avoid packet loss, failure to remove
an extension header does not introduce any new detrimental or
incorrect behavior. If extension headers aren't removed as intended
then they may be processed by the network instead of dropped; this
behavior is also correct and protocol conformant.
Herbert Expires 26 August 2024 [Page 9]
Internet-Draft Inflight-EH-Removal February 2024
The secondary purpose for removing extension headers in-flight is to
avoiding leaking information outside of a limited domain. If an
egress router fails to remove an extension header then sensitive
information may be exposed and this is a security risk. However,
even without extension header removal, a firewall would still be
needed to block packets with Hop-by-Hop Options or Routing headers
from leaving the limited domain in order to enforce security policy.
There is no reason to believe that a firewall that blocks packets
would be no less susceptible to bugs, partial node failures, or
configuration errors than one that removes extension headers and
forwards packets.
4. Considerations
4.1. Reflection of Hop-by-Hop Options
Some Hop-by-Hop options are designed to be reflected by a remote host
back to the sender. IOAM Loopback [RFC9332] is used to report
measurements on the forward path of a sender, the Minimum Path MTU
Hop-by-Hop Option [RFC9268] returns the path MTU of the forward path
to a sender, and FAST [I-D.herbert-fast] allows tickets to be
reflected to affect packet processing in the return path of a flow.
Note that Hop-by-Hop Options reflection is not guaranteed and hence
is an opportunistic mechanism; it cannot be assumed that options will
always be reflected.
In the case that a router removes Hop-by-Hop Options, reflection
won't happen since the destination host does not see the Hop-by-Hop
option to be reflected. In order to preserve the benefits of
reflection, routers should only remove Hop-by-Hop Options headers
that might include options to be reflected as a last resort to
prevent the packets being dropped by a downstream node.
4.2. End host processing of Routing Headers
Per [RFC8200], "If Segments Left is zero, the node must ignore the
Routing header and proceed to process the next header in the packet".
Effectively, this means once the last segment has been processed and
the final destination is set then the Routing header carries no
useful information to any downstream nodes, and removal of the
extension header doesn't affect how the packet is processed.
Herbert Expires 26 August 2024 [Page 10]
Internet-Draft Inflight-EH-Removal February 2024
A possible exception is that the destination host may elect to
validate the Routing header. For instance, the end host may validate
the HMAC TLV in a Segment Routing header. Since Routing headers are
most likely used only in limited domains, which is an explicit
requirement in Segment Routing, the network nodes processing the
Routing header should know if the final destination participates is
required to validate the Routing header-- if it's not then the header
can be removed.
4.3. ICMP errors
When an ICMP error message is sent for a packet with removed
extension headers, the packet headers in the ICMP data will be
different then what the host sent. Operationally, this should not be
an issue since a sender doesn't normally need to correlate packet
with Hop-by-Hop options that were originally sent and the host
network stack doesn't usually maintain sufficient state to make a
precise correlation.
It is possible that a packet may be dropped because it does not have
an expected Hop-by-Hop Options, such as a firewall ticket
[I-D.herbert-fast]. In this case, the ICMP error does contain
relevant information that can be logged and used for debugging.
5. Requirements
An router MAY remove a Hop-by-Hop Options header from a packet if the
following conditions are met:
* The packet does not contain an Authentication header. If the
packet contains and Authentication header then the Hop-by-Hop
Options header MUST NOT be removed
* The Payload Length of the packet is non-zero and the Hop-by-Hop
options does not include a Jumbo Payload Option [RFC2675] (if the
packet contains a Jumbo Payload option then the Payload Length
should be zero).
A router MAY remove a Routing header extension header from a packet
if the following conditions are met:
* The Destination Address has been set to the address of the final
destination and the Segments Left field is zero
* The packet does not contain an Authentication header
* There are no extension headers the precede the Routing header in
Herbert Expires 26 August 2024 [Page 11]
Internet-Draft Inflight-EH-Removal February 2024
the packet. An exception is if the Routing header immediately
follow a Hop-by-Hop Options header that is also being removed
* The final destination is not required to process or validate the
Routing header
* The Routing header does not contain options (segment routing TLVs
for instance), or the destination host doesn't need to process or
validate the options.
6. Procedures
This section describes the procedures for removing a Hop-by-Hop
Options header, removing a Routing header, and removing a Hop-by-Hop
Options Routing header at the same time.
6.1. Removing a Hop-by-Hop Options Header
The procedures for removing a Hop-by-Hop Options header are:
1. Save the value in the Next Header field of the Hop-by-Hop Options
header in a temporary variable
2. Determine the length of the Hop-by-Hop header and save in a
temporary variable. This is equal to the value of the Hdr Ext
Len field times eight plus eight
3. Determine the offset of the first byte in the following the Hop-
by-Hop Options header. This is equal to forty plus the length of
the Hop-by-Hop Options header derived in step 2
4. Copy the IPv6 header with length forty bytes to the offset
derived in set 3 minus forty. Reset the starting offset of the
packet to be the offset of the copied IPv6 header
5. Set the Next Header field in the copied IPv6 header to the value
saved in step 1
6. Subtract the length of the Hop-by-Hop Options header, determined
in step 2, from the Payload Length in the copied IPv6 header.
Set the result as the Payload Length in the copied IPv6 header
An example of removing Hop-by-Hop Options header is shown in the
diagrams below.
The diagram below illustrates shows an example TCP/IPv6 packet with a
Hop-by-Hop Options header; the Payload Length is 1,200 bytes and the
length of the Hop-by-Hop Options header is sixty-four bytes.
Herbert Expires 26 August 2024 [Page 12]
Internet-Draft Inflight-EH-Removal February 2024
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | Traffic Class | Flow Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length = 1200 | Next Hdr = 0 | Hop Limit |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Source Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Destination Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Hdr = 6 | EH Len = 7 | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| |
. .
. Options .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. TCP packet and payload .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The diagram below illustrates the packet after the Hop-by-Hop Options
header has been removed. Note that the Payload Length is now 1,136
bytes which is the original payload length minus the length of the
Hop-by-Hop Options header that was removed.
Herbert Expires 26 August 2024 [Page 13]
Internet-Draft Inflight-EH-Removal February 2024
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | Traffic Class | Flow Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length = 1136 | Next Hdr = 6 | Hop Limit |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Source Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Destination Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. TCP packet and payload .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
6.2. Removing a Routing Header
The procedures for removing a Routing header are:
1. Save the value in the Next Header field of the Routing header in
a temporary variable
2. Determine the length of the Routing header and save in a
temporary variable. This is equal to the value of the Hdr Ext
Len field times eight plus eight
3. Determine the offset of the first byte in the following the
Routing header. This is equal to forty plus the length of the
Hop-by-Hop Options header derived in step 2
4. Copy the IPv6 header with length forty bytes to the offset
derived in set 3 minus forty. Reset the starting offset of the
packet to be the offset of the copied IPv6 header
Herbert Expires 26 August 2024 [Page 14]
Internet-Draft Inflight-EH-Removal February 2024
5. Set the Next Header field in the copied IPv6 header to the value
saved in step 1
6. Subtract the length of the Routing header, determined in step 2,
from the Payload Length in the copied IPv6 header. Set the
result as the Payload Length in the copied IPv6 header
An example of removing a Routing header is shown in the diagrams
below.
The diagram below illustrates shows an example TCP/IPv6 packet with a
Routing header; the Payload Length is 1400 bytes and the length of
the Routing header is 160 bytes. The Segments Left field is set to
zero so that the Routing header may be removed.
Herbert Expires 26 August 2024 [Page 15]
Internet-Draft Inflight-EH-Removal February 2024
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | Traffic Class | Flow Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length = 1400 | Next Hdr = 43| Hop Limit |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Source Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Destination Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Hdr = 6 | EH Len = 19 | Routing Type | Segs Left = 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. type-specific data .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. TCP packet and payload .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The diagram below illustrates the packet after the Routing header has
been removed. Note that the Payload Length is now 1,240 bytes which
is the original payload length minus the length of the Routing header
that was removed.
Herbert Expires 26 August 2024 [Page 16]
Internet-Draft Inflight-EH-Removal February 2024
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | Traffic Class | Flow Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length = 1240 | Next Hdr = 6 | Hop Limit |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Source Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Destination Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. TCP packet and payload .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
6.3. Removing both a Hop-by-Hop Options and a Routing header
The procedures for removing both a Hop-by-Hop Options and a Routing
header are:
1. Save the value in the Next Header field of the Routing header
extension header in a temporary variable
2. Determine the length of the Hop-by-Hop Options header and save in
a temporary variable. This is equal to the value of the Hdr Ext
Len field times eight plus eight
3. Determine the length of the Routing header and save in a
temporary variable. This is equal to the value of the Hdr Ext
Len field times eight plus eight
4. Determine the offset of the first byte in the packet following
the Routing header. This is equal to forty plus the length of
the Hop-by-Hop Options header derived in step 2 plus the length
of the Routing header derived in step 3
Herbert Expires 26 August 2024 [Page 17]
Internet-Draft Inflight-EH-Removal February 2024
5. Copy the IPv6 header with length forty bytes to the offset
derived in set 3 minus forty. Reset the starting offset of the
packet to be the offset of the copied IPv6 header
6. Set the Next Header field in the copied IPv6 header to the value
saved in step 1
7. Subtract the length of the Hop-by-Hop Options header plus the
length of the Routing header (values determined in step 2 and
step 3) from the Payload Length in the copied IPv6 header. Set
the result as the Payload Length in the copied IPv6 header
An example of removing a Hop-by-Hop Options header a Routing header
is shown in the diagrams below.
The diagram below illustrates an example TCP/IPv6 packet with both a
Hop-by-Hop Options and a Routing header; the Payload Length is 1,300
bytes, the length of the Hop-by-Hop Options header is sixty-four
bytes, the length of the Routing header is 160 bytes. The Segments
Left field is set to zero so that the Routing header may be removed.
Herbert Expires 26 August 2024 [Page 18]
Internet-Draft Inflight-EH-Removal February 2024
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | Traffic Class | Flow Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length = 1300 | Next Hdr = 0 | Hop Limit |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Source Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Destination Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Hdr = 43 | EH Len = 7 | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| |
. .
. Options .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Hdr = 6 | EH Len = 19 | Routing Type | Segs Left = 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. type-specific data .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. TCP packet and payload .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Herbert Expires 26 August 2024 [Page 19]
Internet-Draft Inflight-EH-Removal February 2024
The diagram below illustrates the packet after the Hop-by-Hop Options
header and the Routing header have been removed. Note that the
Payload Length is now 1,076 bytes which is the original payload
length minus the length of the Hop-by-Hop Options header and the
Routing header that were removed.
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | Traffic Class | Flow Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length = 1076 | Next Hdr = 6 | Hop Limit |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Source Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| |
+ Destination Address +
| |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
. .
. TCP packet and payload .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
7. Implementation Considerations
Removal of extension headers must be efficient and considered a "fast
path" operation in a router [I-D.ietf-6man-hbh-processing]. The most
computationally complex part of removing extension headers is moving
the IPv6 header. There are two methods to move the octets of the
IPv6 header: memory copy and scatter/gather.
Herbert Expires 26 August 2024 [Page 20]
Internet-Draft Inflight-EH-Removal February 2024
7.1. Copying the IPv6 Header
Extension header removal can be accomplished by performing a data
copy of the IPv4 header (forty bytes) to the offset after the
extension header being removed minus forty bytes. Since the number
of bytes being moved is relatively small and fits within a typical
cache line, the data copy is amenable to efficient implementation in
hardware or software. Once the copy completes, the pointer to the
packet is advanced by the length of data removed. Note that an
implementation may choose to move the link layer header as well.
7.2. Scatter/gather
Scatter/gather allows a packet to be constructed from a list of
memory buffers where each buffer has a data pointer and length. To
use scatter/gather for extension header removal, a receiver might
employ header/data split to store the packet as two buffers in
memory: the first buffer contains the link layer and IPv6 headers,
and the second buffer contains the data following the IPv6 header.
Removing an extension headers entails advancing the pointer to the
second buffer by the length of the extension header being removed.
8. Security Considerations
This specification does not introduce any new security concerns,
9. IANA Considerations
There are no IANA considerations in this specification.
10. References
10.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>.
[RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6
(IPv6) Specification", STD 86, RFC 8200,
DOI 10.17487/RFC8200, July 2017,
<https://www.rfc-editor.org/info/rfc8200>.
10.2. Informative References
Herbert Expires 26 August 2024 [Page 21]
Internet-Draft Inflight-EH-Removal February 2024
[APNIC-EH] Huston, G., "IPv6 extension headers revisited", October
2022, <https://blog.apnic.net/2022/10/13/ipv6-extension-
headers-revisited>.
[Cus23a] Custura, A. and G. Fairhurst, "Internet Measurements: IPv6
Extension Header Edition", IEPG, IETF-116 , March 2023,
<http://www.iepg.org/2023-03-26-ietf116/eh.pdf>.
[I-D.herbert-fast]
Herbert, T., "Firewall and Service Tickets", Work in
Progress, Internet-Draft, draft-herbert-fast-07, 7 October
2023, <https://datatracker.ietf.org/doc/html/draft-
herbert-fast-07>.
[I-D.ietf-6man-hbh-processing]
Hinden, R. M. and G. Fairhurst, "IPv6 Hop-by-Hop Options
Processing Procedures", Work in Progress, Internet-Draft,
draft-ietf-6man-hbh-processing-13, 18 February 2024,
<https://datatracker.ietf.org/doc/html/draft-ietf-6man-
hbh-processing-13>.
[I-D.smith-6man-in-flight-eh-insertion-harmful]
Smith, M., Kottapalli, N., Bonica, R., Gont, F., and T.
Herbert, "In-Flight IPv6 Extension Header Insertion
Considered Harmful", Work in Progress, Internet-Draft,
draft-smith-6man-in-flight-eh-insertion-harmful-02, 30 May
2020, <https://datatracker.ietf.org/doc/html/draft-smith-
6man-in-flight-eh-insertion-harmful-02>.
[RFC2473] Conta, A. and S. Deering, "Generic Packet Tunneling in
IPv6 Specification", RFC 2473, DOI 10.17487/RFC2473,
December 1998, <https://www.rfc-editor.org/info/rfc2473>.
[RFC2675] Borman, D., Deering, S., and R. Hinden, "IPv6 Jumbograms",
RFC 2675, DOI 10.17487/RFC2675, August 1999,
<https://www.rfc-editor.org/info/rfc2675>.
[RFC6192] Dugal, D., Pignataro, C., and R. Dunn, "Protecting the
Router Control Plane", RFC 6192, DOI 10.17487/RFC6192,
March 2011, <https://www.rfc-editor.org/info/rfc6192>.
[RFC7872] Gont, F., Linkova, J., Chown, T., and W. Liu,
"Observations on the Dropping of Packets with IPv6
Extension Headers in the Real World", RFC 7872,
DOI 10.17487/RFC7872, June 2016,
<https://www.rfc-editor.org/info/rfc7872>.
Herbert Expires 26 August 2024 [Page 22]
Internet-Draft Inflight-EH-Removal February 2024
[RFC8201] McCann, J., Deering, S., Mogul, J., and R. Hinden, Ed.,
"Path MTU Discovery for IP version 6", STD 87, RFC 8201,
DOI 10.17487/RFC8201, July 2017,
<https://www.rfc-editor.org/info/rfc8201>.
[RFC8799] Carpenter, B. and B. Liu, "Limited Domains and Internet
Protocols", RFC 8799, DOI 10.17487/RFC8799, July 2020,
<https://www.rfc-editor.org/info/rfc8799>.
[RFC9098] Gont, F., Hilliard, N., Doering, G., Kumari, W., Huston,
G., and W. Liu, "Operational Implications of IPv6 Packets
with Extension Headers", RFC 9098, DOI 10.17487/RFC9098,
September 2021, <https://www.rfc-editor.org/info/rfc9098>.
[RFC9268] Hinden, R. and G. Fairhurst, "IPv6 Minimum Path MTU Hop-
by-Hop Option", RFC 9268, DOI 10.17487/RFC9268, August
2022, <https://www.rfc-editor.org/info/rfc9268>.
[RFC9332] De Schepper, K., Briscoe, B., Ed., and G. White, "Dual-
Queue Coupled Active Queue Management (AQM) for Low
Latency, Low Loss, and Scalable Throughput (L4S)",
RFC 9332, DOI 10.17487/RFC9332, January 2023,
<https://www.rfc-editor.org/info/rfc9332>.
Author's Address
Tom Herbert
SiPanda
Santa Clara, CA,
United States of America
Email: tom@herbertland.com
Herbert Expires 26 August 2024 [Page 23]
