| Internet-Draft | Deep Dive IPv6 EH | October 2022 |
| Elkins, et al. | Expires 24 April 2023 | [Page] |
IPv6 Extension Header testing is a complex area. Studies have shown varying results. This document proposes a methodology for isolating the location and reasons for IPv6 Extension Headers blockage. This document outlines the basic framework and set of documents which will follow.¶
This note is to be removed before publishing as an RFC.¶
Discussion of this document takes place on the IPv6 Operations Working Group mailing list (v6ops@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/v6ops/.¶
Source for this draft and an issue tracker can be found at https://github.com/dhruvdhody/draft-elkins-v6ops-eh-deepdive-fw.¶
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 24 April 2023.¶
Copyright (c) 2022 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 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.¶
IPv6 [RFC8200] Extension Headers (EHs) may be blocked at various points yet show the same symptom. That is, if an EH is blocked at a router, it will appear to the client or measurement technique that the packet is dropped. If an EH is blocked at a load balancer or CDN, the client will see the same symptom -- the packet is dropped.¶
This is a problem because when the same symptom can be the due to multiple factors, incorrect conclusions may be drawn from the results. That is, if, for example, loss of the sense of taste of sweetness can be due either to a brain tumor or a minor neurological problem, then if someone has a tendency towards hypochondria, they may incorrectly bemoan their impending death from a brain tumor when it may really be only a minor issue.¶
This is the same for EH testing. If a packet is seen to be dropped in measurement, it may only be that there is a bug in the load balancer code (for example) and not that EH transmission does not work writ large.¶
This document proposes a framework to isolate the problem of where the packet is being dropped. This is, however, more easily said than done. There are many components and devices in a network. Each may require a different isolation technique.¶
We propose a methdology which starts with the simplest topology and grows towards more and more complex, real-world networks. For example, in today's networks, if one attempts to access a well-known site such as Facebook.com, one is likely to access a cache server managed by a Content Delivery Network (CDN) rather than the origin server managed by Facebook. It is important to isolate the testing so that we can determine the exact component which may be blocking the EH and why.¶
We additionally discuss the potential issues with the type of data sent as well as how the way the data is sent for testing may impact test results.¶
Yet, there are some overriding principles.¶
The blockage may be:¶
Then:¶
Intentional blocks are easier to assess. The block is an administrator or network policy. Of course, the policy may come about as a result of a misunderstanding of the function of EHs or lack of effective guidance but nonetheless, this is an understandable decision.¶
The more troubling causes of EH blocks include:¶
As far as the blocking component itself (whether intentional or unintentional):¶
Note: A DNS or other configuration setting may block IPv6 or prefer IPv4. If not tested correctly, lack of support for IPv6 itself may incorrectly appear as lack of support for EHs.¶
Finally, the blockage or source may differ based on the type of EH. Some EHs are limited to an administrative domain, some EHs are processed at every hop, still others are from the source and destination. Problem isolation techniques will differ for the various classes of EH.¶
Our final document in this series will be a BCP to indicate which EH should be allowed, blocked, encrypted, or authenticated and at which component or platform.¶
Note: [RFC9288] "Recommendations on the Filtering of IPv6 Packets Containing IPv6 Extension Headers at Transit Routers" focuses on the IPv6 EH handling at transit routers only. Our approach is to produce a final BCP with various recommendations across network segments, once the nature of the problems and techniques for isolation are well known.¶
The diagnostic methodology to follow depends on what is being tested. For example, the problem isolation for a CDN would be different than the problem isolation for an internal network.¶
This framework proposes the following set of documents:¶
Note that the server can be owned and operated by the administrator themselves (on-prem) Figure 1, or they could hosted behind a CDN Figure 2, or hosted by the cloud provider Figure 3.¶
+--------------------------+
| |
+--------+ | | +---------+
| client |---------| Internet |-----| server |
+--------+ | | +---------+
| |
+--------------------------+
+-----------------+
+--------------------------+ +-----------------+ |
| | | CDN Network | |
+--------+ | | | +---------+ | |
| client |---------| Internet |-----| | Edge | | |
+--------+ | | | | Server | | |
| | | +---------+ | |
+--------------------------+ | | |
| (LB, Firewall | |
| Cache etc) | |
| | |
| | |
| | |
| |-+
+-----------------+
|
| (*)
|
+---------+
| Origin |
| server |
+---------+
(*) - can be over the internet
+--------------------------+ +-----------------+
| | | |
+--------+ | | | |
| client |---------| Internet |-----| Cloud Network |
+--------+ | | | (LB, Firewall, |
| | | Gateways etc) |
+--------------------------+ | |
| |
| |
| +---------+ |
| | Virtual | |
| | server | |
| +---------+ |
+-----------------+
The first step for all testing is to have a test server, client and / or router enabled to send EHs. There are two basic options to do this, each with its own set of advantages and drawbacks.¶
The options fall broadly into two categories:¶
If crafting packets, then the question arises of which packets to craft. This will be discussed in Section 2.2.¶
For either methodology, the rate of sending may influence results. This will be discussed in Section 3.¶
Possibly the best way to isolate problems with EHs may be to have a server, client or router which has modifications to the OS, interface, driver or other to send real EHs with real application data. This carries the initial cost of creating such modifications since, unfortunately, the current set of host OSs do not natively support this.¶
The reason we find this to be the best method is because the problem of what packet to craft is a field which is rife with land mines. See Section 2.2 for a further discussion of the exact nature of the aforesaid land mines.¶
A number of packages exist which can craft a packet with an EH header. The more interesting and fraught problem may be which exact packet to craft and then which EH to craft.¶
Let's discuss first the options for the type of packet to craft. These include:¶
If crafting a TCP packet, then it is likely that some middlebox will drop a TCP packet which does not have the appropriate ACK and SEQUENCE numbers. One may get around this problem by sending a packet with the SYN flag on and directed to some well-known port such as 443 or 80. But, if many firewalls and other devices such as IDS / IPS, and even some OSs have SYN Flood protection. So, if more than a certain number of these packets, say 10, in some short interval are sent, then they are likely to be dropped for reasons other than blockage of EH headers.¶
One may choose to use UDP, QUIC, or ICMPv6 to attempt to bypass the complexities of TCP. Some enterprise networks are likely to drop UDP and / or QUIC. Testing should be done without EH to make sure that such packets do indeed pass.¶
Certainly, if you have access to all the middleboxes in your domain, you may be able to bypass or stop SYN Flood, UDP, ICMPv6 or other transport layer blocks at all the middleboxes. But, it may be a more difficult effort than might be imagined.¶
You may wish to test in a lab environment first to validate your approach.¶
There are two ways to add an EH to a packet:¶
The best method may be to use an EH which is actually processed at the client, server, and any associated router. This leaves us with a sparse set to choose from. The problem with crafting an EH is that if the data is not "real", one may not be able to fault a network component for blocking it.¶
Having said that, let's discuss crafting an EH. An EH may be crafted by simply using multiple PADN options. One should be careful not to use too many PADNs because then this type of header may be dropped by a middlebox or OS as being a flawed packet. This is likely to distort the results.¶
An EH with all 0's or other data patterns that could be perceived as not "real" may also be dropped by middleboxes which are trying to be helpful.¶
It is also possible that the length of the EH may have some effect on what EHs get dropped and where.¶
You may wish to try this in a lab environment first. If the test suceeds, then you may test on your network.¶
Next, there is a consideration that the type of EH, for example, Destination Options, Hop-by-hop, etc. could get processed differently and may be dropped at different frequencies, by different devices. You may wish to test one at a time and note the results.¶
You may wish to try this in a lab environment first. If the test suceeds, then you may test on your network.¶
Whether you have chosen to send real application data or to craft packets, the rate of sending and sampling may create a false indication of blockage. That is, if you send a great deal of data at frequent intervals, some middlebox in the network is likely to see this traffic as a Denial of Service (DoS) attack and block it. This is more likely if a great many crafted packets are sent but even with real application data such as FTP or HTTP, overly aggressive sending is likely to be counterproductive.¶
You may wish to send only a few packets or one or two CURL or FTPs at any one test.¶
This document has no security considerations.¶
This document has no privacy considerations.¶
This document has no IANA actions.¶
TODO acknowledge.¶
TODO contributors.¶