Can I use several masqueraded subnets?

Yes. This is done all the time. See the discussion in our setup document. The only restriction is that the subnets on the two ends must not overlap. See the next question.

Here is a mailing list message on the topic. The user incorrectly thinks you need a 2.4 kernel for this -- actually various people have been doing it on 2.0 and 2.2 for quite some time -- but he has it right for 2.4.

Subject: Double NAT and freeswan working :)
   Date: Sun, 11 Mar 2001
   From: Paul Wouters <paul@xtdnet.nl>

Just to share my pleasure, and make an entry for people who are searching
the net on how to do this. Here's the very simple solution to have a double
NAT'ed network working with freeswan. (Not sure if this is old news, but I'm
not on the list (too much spam) and I didn't read this in any HOWTO/FAQ/doc
on the freeswan site yet (Sandy, put it in! :)

10.0.0.0/24 --- 10.0.0.1 a.b.c.d  ---- a.b.c.e {internet} ----+
                                                              |
10.0.1.0/24 --- 10.0.1.1 f.g.h.i  ---- f.g.h.j {internet} ----+

the goal is to have the first network do a VPN to the second one, yet also
have NAT in place for connections not destinated for the other side of the
NAT. Here the two Linux security gateways have one real IP number (cable
modem, dialup, whatever.

The problem with NAT is you don't want packets from 10.*.*.* to 10.*.*.*
to be NAT'ed. While with Linux 2.2, you can't, with Linux 2.4 you can.

(This has been tested and works for 2.4.2 with Freeswan snapshot2001mar8b)

relevant parts of /etc/ipsec.conf:

        left=f.g.h.i
        leftsubnet=10.0.1.0/24
        leftnexthop=f.g.h.j
        leftfirewall=yes
        leftid=@firewall.netone.nl
        leftrsasigkey=0x0........
        right=a.b.c.d
        rightsubnet=10.0.0.0/24
        rightnexthop=a.b.c.e
        rightfirewall=yes
        rightid=@firewall.nettwo.nl
        rightrsasigkey=0x0......
        # To authorize this connection, but not actually start it, at startup,
        # uncomment this.
        auto=add

and now the real trick. Setup the NAT correctly on both sites:

iptables -t nat -F
iptables -t nat -A POSTROUTING -o eth0 -d \! 10.0.0.0/8 -j MASQUERADE

This tells the NAT code to only do NAT for packets with destination other then
10.* networks. note the backslash to mask the exclamation mark to protect it
against the shell.

Happy painting :)

Paul

Can I use subnets masqueraded to the same addresses?

No. The notion that IP addresses are unique is one of the fundamental principles of the IP protocol. Messing with it is exceedingly perilous.

Fairly often a situation comes up where a company has several branches, all using the same non-routable addresses, perhaps 192.168.0.0/24. This works fine as long as those nets are kept distinct. The IP masquerading on their firewalls ensures that packets reaching the Internet carry the firewall address, not the private address.

This can break down when IPsec enters the picture. FreeS/WAN builds a tunnel that pokes through both masquerades and delivers packets from leftsubnet to rightsubnet and vice versa. For this to work, the two subnets must be distinct.

There are several solutions to this problem.

Usually, you re-number the subnets. Perhaps the Vancouver office becomes 192.168.101.0/24, Calgary 192.168.102.0/24 and so on. FreeS/WAN can happily handle this. With, for example leftsubnet=192.168.101.0/24 and rightsubnet=192.168.102.0/24 in a connection description, any machine in Calgary can talk to any machine in Vancouver. If you want to be more restrictive and use something like leftsubnet=192.168.101.128/25 and rightsubnet=192.168.102.240/28 so only certain machines on each end have access to the tunnel, that's fine too.

You could also split the subnet into smaller ones, for example using 192.168.1.0/25 in Vancouver and rightsubnet=192.168.0.128/25 in Calgary.

Alternately, you can just give up routing directly to machines on the subnets. Omit the leftsubnet and rightsubnet parameters from your connection descriptions. Your IPsec tunnels will then run between the public interfaces of the two firewalls. Packets will be masqueraded both before they are put into tunnels and after they emerge. Your Vancouver client machines will see only one Calgary machine, the firewall.

Can I assign a road warrior an address on my net (a virtual identity)?

Often it would be convenient to be able to give a Road Warrior an IP address which appears to be on the local network. Some IPsec implementations have support for this, sometimes calling the feature "virtual identity".

Currently (Sept 2002) FreeS/WAN does not support this, and we have no definite plans to add it. The difficulty is that is not yet a standard mechanism for it. There is an Internet Draft for a method of doing it using DHCP which looks promising. FreeS/WAN may support that in a future release.

In the meanwhile, you can do it yourself using the Linux iproute2(8) facilities. Details are in this paper.

Another method has also been discussed on the mailing list.:

• You can use a variant of the extruded subnet procedure.
• You have to avoid having the road warrior's assigned address within the range you actually use at home base. See previous question.
• On the other hand, you want the roadwarrior's address to be within the range that seems to be on your network.

For example, you might have:

leftsubnet=a.b.c.0/25

head office network

rightsubnet=a.b.c.129/32

extruded to a road warrior. Note that this is not in a.b.c.0/25

a.b.c.0/24

whole network, including both the above

You then set up routing so that the office machines use the IPsec gateway as their route to a.b.c.128/25. The leftsubnet parameter tells the road warriors to use tunnels to reach a.b.c.0/25, so you should have two-way communication. Depending or your network and applications, there may be some additional work to do on DNS or Windows configuration

Can I support many road warriors with one gateway?

Yes. This is easily done, using

either RSA authentication

standard in the FreeS/WAN distribution

or X.509 certificates

requires patches to FreeS/WAN

In either case, each Road Warrior must have a different key or certificate.

This cannot be made to work using pre-shared key authentication; see the next question for details.

If you expect to have more than a few dozen Road Warriors connecting simultaneously, you may need a fairly powerful gateway machine. See our document on FreeS/WAN performance.

Can I have many road warriors using shared secret authentication?

No. There is no way to do this securely, and there is no way to fix the problem.

You can have multiple Road Warriors using shared secret authentication only if they all use the same secret. This creates problems:

• If you have many users, it becomes almost certain the secret will leak
• The secret becomes quite valuable to an attacker
• All users authenticate the same way, so the gateway cannot tell them apart for logging or access control purposes
• Changing the secret is difficult. You have to securely notify all users.
• If you find out the secret has been compromised, you can change it, but then what? None of your users can connect without the new secret. How will you notify them all, quickly and securely, without using the VPN?

This is a designed-in limitation of the IKE key negotiation protocol, not a problem with our implementation.

When using shared secrets, the protocol requires that the responding gateway be able to determine which secret to use at a time when all it knows about the initiator is an IP address. This works fine if you know the initiator's address in advance and can use it to look up the appropiriate secret. However, it fails for Road Warriors since the gateway cannot know their IP addresses in advance.

We very strongly recommend that you avoid using shared secret authentication for multiple Road Warriors. Use RSA authentication instead.

With RSA signatures (or certificates) the protocol is slightly different. The initiator provides an identifier early in the exchange and the responder can use that identifier to look up the correct key or certificate. See above.

Can I use Quality of Service routing with FreeS/WAN?

From project technical lead Henry Spencer:

> Do QoS add to FreeS/WAN?
> For example integrating DiffServ and FreeS/WAN?

With a current version of FreeS/WAN, you will have to add hidetos=no to
the config-setup section of your configuration file.  By default, the TOS
field of tunnel packets is zeroed; with hidetos=no, it is copied from the
packet inside.  (This is a modest security hole, which is why it is no
longer the default.)

DiffServ does not interact well with tunneling in general.  Ways of
improving this are being studied.

Copying the TOS (type of service) information from the encapsulated packet to the outer header reveals the TOS information to an eavesdropper. This does not tell him much, but it might be of use in traffic analysis. Since we do not have to give it to him, our default is not to.

See ipsec.conf(5) for more on the hidetos= parameter.

Can I recognise dead tunnels and shut them down?

There is no general mechanism to do this is in the IPsec protocols.

From time to time, there is discussion on the IETF Working Group mailing list of adding a "keep-alive" mechanism (which some say should be called "make-dead"), but it is a fairly complex problem and no consensus has been reached on whether or how it should be done.

The protocol does have optional delete-SA messages which one side can send when it closes a connection in hopes this will cause the other side to do the same. FreeS/WAN does not currently support these. In any case, they would not solve the problem since:

• a gateway that crashes or hangs would not send the messages
• the sender is not required to send them
• they are not authenticated, so any receiver that trusts them leaves itself open to a denial of service attack
• the receiver is not required to do anything about them
• the receiver cannot acknowledge them; the protocol provides no mechanism for that
• since they are not acknowledged, the sender cannot rely on them

However, connections do have limited lifetimes and you can control how many attempts your gateway makes to rekey before giving up. For example, you can set:

conn default
        keyingtries=3
        keylife=30m

With these settings old connections will be cleaned up. Within 30 minutes of the other end dying, rekeying will be attempted. If it succeeds, the new connection replaces the old one. If it fails, no new connection is created. Either way, the old connection is taken down when its lifetime expires.

Here is a mailing list message on the topic from FreeS/WAN tech support person Claudia Schmeing:

You ask how to determine whether a tunnel is redundant:

> Can anybody explain the best way to determine this. Esp when a RW has
> disconnected? I thought 'ipsec auto --status' might be one way.

If a tunnel goes down from one end, Linux FreeS/WAN on the
other end has no way of knowing this until it attempts to rekey.
Once it tries to rekey and fails, it will 'know' that the tunnel is 
down.

Because it doesn't have a way of knowing the state until this point, 
it will also not be able to tell you the state via ipsec auto --status.

> However, comparing output from a working tunnel with that of one that
> was closed 
> did not show clearly show tunnel status.

If your tunnel is down but not 'unrouted' (see man ipsec_auto), you
should not be able to ping the opposite side of the tunnel. You can
use this as an indicator of tunnel status.

On a related note, you may be interested to know that as of 1.7, 
redundant tunnels caused by RW disconnections are likely to be 
less of a pain. From doc/CHANGES:

    There is a new configuration parameter, uniqueids, to control a new Pluto
    option:  when a new connection is negotiated with the same ID as an old
    one, the old one is deleted immediately.  This should help eliminate
    dangling Road Warrior connections when the same Road Warrior reconnects. 
    It thus requires that IDs not be shared by hosts (a previously legal but
    probably useless capability).  NOTE WELL:  the sample ipsec.conf now has
    uniqueids=yes in its config-setup section.


Cheers,

Claudia

Can I build IPsec tunnels over a demand-dialed link?

This is possible, but not easy. FreeS/WAN technical lead Henry Spencer wrote:

> 5. If the ISDN link goes down in between and is reestablished, the SAs
> are still up but the eroute are deleted and the IPsec interface shows
> garbage (with ifconfig)
> 6. Only restarting IPsec will bring the VPN back online.

This one is awkward to solve.  If the real interface that the IPsec
interface is mounted on goes down, it takes most of the IPsec machinery
down with it, and a restart is the only good way to recover. 

The only really clean fix, right now, is to split the machines in two: 

1. A minimal machine serves as the network router, and only it is aware
that the link goes up and down. 

2. The IPsec is done on a separate gateway machine, which thinks it has
a permanent network connection, via the router.

This is clumsy but it does work.  Trying to do both functions within a
single machine is tricky.  There is a software package (diald) which will
give the illusion of a permanent connection for demand-dialed modem
connections; I don't know whether it's usable for ISDN, or whether it can
be made to cooperate properly with FreeS/WAN. 

Doing a restart each time the interface comes up *does* work, although it
is a bit painful.  I did that with PPP when I was running on a modem link;
it wasn't hard to arrange the PPP scripts to bring IPsec up and down at
the right times.  (I'd meant to investigate diald but never found time.)

In principle you don't need to do a complete restart on reconnect, but you
do have to rebuild some things, and we have no nice clean way of doing
only the necessary parts.

In the same thread, one user commented:

Subject: Re: linux-ipsec: IPsec and Dial Up Connections
   Date: Wed, 22 Nov 2000
   From: Andy Bradford <andyb@calderasystems.com>

On Wed, 22 Nov 2000 19:47:11 +0100, Philip Reetz wrote:

> Are there any ideas what might be the cause of the problem and any way
> to work around it.
> Any help is highly appreciated.

On my laptop, when using ppp there is a ip-up script in /etc/ppp that 
will be executed each time that the ppp interface is brought up.  
Likewise there is an ip-down script that is called when it is taken 
down.  You might consider custimzing those to stop and start FreeS/Wan 
with each connection.  I believe that ISDN uses the same files, though 
I could be wrong---there should be something similar though.

Can I build GRE tunnels over IPsec?

This is possible in theory, but we are short on practical details. If you do this, please let us know via the mailing lists.

There is a list message with links to relevant resources.

Life's little mysteries

FreeS/WAN is a fairly complex product. (Neither the networks it runs on nor the protocols it uses are simple, so it could hardly be otherwise.) It therefore sometimes exhibits behaviour which can be somewhat confusing, or has problems which are not easy to diagnose. This section tries to explain those problems.

Setup and configuration of FreeS/WAN are covered in other documentation sections:

• basic setup and configuration
• advanced configuration
• Troubleshooting

However, we also list some of the commonest problems here.

I cannot ping ....

This question is dealt with in the advanced configuration section under the heading multiple tunnels.

The standard subnet-to-subnet tunnel protects traffic only between the subnets. To test it, you must use pings that go from one subnet to the other.

For example, suppose you have:

      subnet a.b.c.0/24
             |
      eth1 = a.b.c.1
         gate1
      eth0 = 1.2.3.4
             |

       ~ internet ~

             |
      eth0 = 4.3.2.1
         gate2
      eth1 = x.y.z.1
              |
       subnet x.y.z.0/24

and the connection description:

conn abc-xyz
     left=1.2.3.4
     leftsubnet=a.b.c.0/24
     right=4.3.2.1
     rightsubnet=x.y.z.0/24

You can test this connection description only by sending a ping that will actually go through the tunnel. Assuming you have machines at addresses a.b.c.2 and x.y.z.2, pings you might consider trying are:

ping from x.y.z.2 to a.b.c.2 or vice versa

Succeeds if tunnel is working. This is the only valid test of the tunnel.

ping from gate2 to a.b.c.2 or vice versa

Does not use tunnel. gate2 is not on protected subnet.

ping from gate1 to x.y.z.2 or vice versa

Does not use tunnel. gate1 is not on protected subnet.

ping from gate1 to gate2 or vice versa

Does not use tunnel. Neither gate is on a protected subnet.

Only the first of these is a useful test of this tunnel. The others do not use the tunnel. Depending on other details of your setup and routing, they:

• either fail, telling you nothing about the tunnel
• or succeed, telling you nothing about the tunnel since these packets use some other route

In some cases, you may be able to get around this. For the example network above, you could use:

        ping -I a.b.c.1 x.y.z.1

Both the adresses given are within protected subnets, so this should go through the tunnel.

If required, you can build additional tunnels so that all the machines involved can talk to all the others. See multiple tunnels in the advanced configuration document for details.

It takes forever to ...

Users fairly often report various problems involving long delays, sometimes on tunnel setup and sometimes on operations done through the tunnel, occasionally on simple things like ping or more often on more complex operations like doing NFS or Samba through the tunnel.

Almost always, these turn out to involve failure of a DNS lookup. The timeouts waiting for DNS are typically set long so that you won't time out when a query involves multiple lookups or long paths. Genuine failures therefore produce long delays before they are detected.

A mailing list message from project technical lead Henry Spencer:

> ... when i run /etc/rc.d/init.d/ipsec start, i get:
> ipsec_setup: Starting FreeS/WAN IPsec 1.5...
> and it just sits there, doesn't give back my bash prompt.

Almost certainly, the problem is that you're using DNS names in your
ipsec.conf, but DNS lookups are not working for some reason.  You will
get your prompt back... eventually.  But the DNS timeouts are long.
Doing something about this is on our list, but it is not easy.

In the meanwhile, we recommend that connection descriptions in ipsec.conf(5) use numeric IP addresses rather than names which will require a DNS lookup.

Names that do not require a lookup are fine. For example:

• a road warrior might use the identity rightid=@lancelot.example.org
• the gateway might use leftid=@camelot.example.org

These are fine. The @ sign prevents any DNS lookup. However, do not attempt to give the gateway address as left=camelot.example.org . That requires a lookup.

A post from one user after solving a problem with long delays:

Subject: Final Answer to Delay!!!
   Date: Mon, 19 Feb 2001
   From: "Felippe Solutions" <felippe@solutionstecnologia.com.br>

Sorry people, but seems like the Delay problem had nothing to do with
freeswan.

The problem was DNS as some people sad from the beginning, but not the way
they thought it was happening. Samba, ssh, telnet and other apps try to
reverse lookup addresses when you use IP numbers (Stupid that ahh).

I could ping very fast because I always ping with "-n" option, but I don't
know the option on the other apps to stop reverse addressing so I don't use
it.

This post is fairly typical. These problems are often tricky and frustrating to diagnose, and most turn out to be DNS-related.

One suggestion for diagnosis: test with both names and addresses if possible. For example, try all of:

• ping address
• ping -n address
• ping name

If these behave differently, the problem must be DNS-related since the three commands do exactly the same thing except for DNS lookups.

I send packets to the tunnel with route(8) but they vanish

IPsec connections are designed to carry only packets travelling between pre-defined connection endpoints. As project technical lead Henry Spencer put it:

IPsec tunnels are not just virtual wires; they are virtual wires with built-in access controls. Negotiation of an IPsec tunnel includes negotiation of access rights for it, which don't include packets to/from other IP addresses. (The protocols themselves are quite inflexible about this, so there are limits to what we can do about it.)

For fairly obvious security reasons, and to comply with the IPsec RFCs, KLIPS drops any packets it receives that are not allowed on the tunnels currently defined. So if you send it packets with route(8), and suitable tunnels are not defined, the packets vanish. Whether this is reported in the logs depends on the setting of klipsdebug in your ipsec.conf(5) file.

To rescue vanishing packets, you must ensure that suitable tunnels for them exist, by editing the connection descriptions in ipsec.conf(5). For example, supposing you have a simple setup:

         leftsubnet -- leftgateway === internet === roadwarrior

If you want to give the roadwarrior access to some resource that is located behind the left gateway but is not in the currently defined left subnet, then the usual procedure is to define an additional tunnel for those packets by creating a new connection description.

In some cases, it may be easier to alter an existing connection description, enlarging the definition of leftsubnet. For example, instead of two connection descriptions with 192.168.8.0/24 and 192.168.9.0/24 as their leftsubnet parameters, you can use a single description with 192.168.8.0/23.

If you have multiple endpoints on each side, you need to ensure that there is a route for each pair of endpoints. See this example.

When a tunnel goes down, packets vanish

This is a special case of the vanishing packet problem described in the previous question. Whenever KLIPS sees packets for which it does not have a tunnel, it drops them.

When a tunnel goes away, either because negotiations with the other gateway failed or because you gave an ipsec auto --down command, the route to its other end is left pointing into KLIPS, and KLIPS will drop packets it has no tunnel for.

This is a documented design decision, not a bug. FreeS/WAN must not automatically adjust things to send packets via another route. The other route might be insecure.

Of course, re-routing may be necessary in many cases. In those cases, you have to do it manually or via scripts. We provide the ipsec auto --unroute command for these cases.

From ipsec_auto(8):

Normally, pluto establishes a route to the destination specified for a connection as part of the --up operation. However, the route and only the route can be established with the --route operation. Until and unless an actual connection is established, this discards any packets sent there, which may be preferable to having them sent elsewhere based on a more general route (e.g., a default route).

Normally, pluto's route to a destination remains in place when a --down operation is used to take the connection down (or if connection setup, or later automatic rekeying, fails). This permits establishing a new connection (perhaps using a different specification; the route is altered as necessary) without having a ``window'' in which packets might go elsewhere based on a more general route. Such a route can be removed using the --unroute operation (and is implicitly removed by --delete).

See also this mailing list message.

The firewall ate my packets!

If firewalls filter out:

• either the UDP port 500 packets used in IKE negotiations
• or the ESP and AH (protocols 50 and 51) packets used to implement the IPsec tunnel

then IPsec cannot work. The first thing to check if packets seem to be vanishing is the firewall rules on the two gateway machines and any other machines along the path that you have access to.

For details, see our document on firewalls .

Some advice from technical lead Henry Spencer on diagnosing such problems:

> > Packets vanishing between the hardware interface and the ipsecN interface
> > is usually the result of firewalls not being configured to let them in...
> 
> Thanks for the suggestion. If only it were that simple! My ipchains startup
> script does take care of that, but just in case I manually inserted rules 
> accepting everything from london on dublin. No difference.

The other thing to check is whether the "RX packets dropped" count on the
ipsecN interface (run "ifconfig ipsecN", for N=1 or whatever, to see the
counts) is rising.  If so, then there's some sort of configuration mismatch
between the two ends, and IPsec itself is rejecting them.  If none of the
ipsecN counts is rising, then the packets are never reaching the IPsec
machinery, and the problem is almost certainly in firewalls etc.

Dropped connections

Networks being what they are, IPsec connections can be broken for any number of reasons, ranging from hardware failures to various software problems such as the path MTU problems discussed elsewhere in the FAQ. Fortunately, various diagnostic tools exist that help you sort out many of the possible problems.

There is one situation, however, where FreeS/WAN (using default settings) may destroy a connection for no readily apparent reason. This occurs when things are misconfigured so that two tunnels from the same gateway expect the same subnet on the far end.

In this situation, the first tunnel comes up fine and works until the second is established. At that point, because of the way we track connections internally, the first tunnel ceases to exist as far as this gateway is concerned. Of course the far end does not know that, and a storm of error messages appears on both systems as it tries to use the tunnel.

If the far end gives up, goes back to square one and negotiates a new tunnel, then that wipes out the second tunnel and ...

The solution is simple. Do not build multiple conn descriptions with the same remote subnet.

This is actually intended to be a feature, rather than a bug. Consider the situation where a single remote system goes down, then comes back up and reconnects to the gateway. It is useful to have the gateway tear down the old tunnel and recover resources when the reconnection is made. It recognises that situation by checking the remote subnet for each tunnel it builds and discarding duplicates. This works fine as long as you don't configure multiple tunnels with the same remote subnet.

If this behaviour is inconvenient for you, you can disable it by setting uniqueids=no in ipsec.conf(5).

TCPdump on the gateway shows strange things

Attempting to look at IPsec packets by running monitoring tools on the IPsec gateway machine can produce silly results. That machine is mangling the packets for IPsec, and possibly for firewall or NAT purposes as well. If the internals of the machine's IP stack are not what the monitoring tool expects, then the tool can misinterpret them and produce nonsense output.

See our testing document for more detail.

Traceroute does not show anything between the gateways

As far as traceroute can see, the two gateways are one hop apart; the data packet goes directly from one to the other through the tunnel. Of course the outer packets that implement the tunnel pass through whatever lies between the gateways, but those packets are built and dismantled by the gateways. Traceroute does not see them and cannot report anything about their path.

Here is a mailing list message with more detail.

Date: Mon, 14 May 2001
To: linux-ipsec@freeswan.org
From: "John S. Denker" <jsd@research.att.com<
Subject: Re: traceroute: one virtual hop

At 02:20 PM 5/14/01 -0400, Claudia Schmeing wrote:
>
>> > A bonus question: traceroute in subnet to subnet enviroment looks like:
>> > 
>> > traceroute to andris.dmz (172.20.24.10), 30 hops max, 38 byte packets
>> > 1  drama (172.20.1.1)  0.716 ms  0.942 ms  0.434 ms
>> > 2  * * *
>> > 3  andris.dmz (172.20.24.10)  73.576 ms  78.858 ms  79.434 ms
>> > 
>> > Why aren't there the other hosts which take part in the delivery during 
>    * * * ?
>
>If there is an ipsec tunnel between GateA and Gate B, this tunnel forms a 
>'virtual wire'.  When it is tunneled, the original packet becomes an inner 
>packet, and new ESP and/or AH headers are added to create an outer packet 
>around it. You can see an example of how this is done for AH at 
>doc/ipsec.html#AH . For ESP it is similar.
>
>Think about the packet's path from the inner packet's perspective.
>It leaves the subnet, goes into the tunnel, and re-emerges in the second
>subnet. This perspective is also the only one available to the
>'traceroute' command when the IPSec tunnel is up.

Claudia got this exactly right.  Let me just expand on a couple of points:

*) GateB is exactly one (virtual) hop away from GateA.  This is how it
would be if there were a physically private wire from A to B.  The
virtually private connection should work the same, and it does.

*) While the information is in transit from GateA to GateB, the hop count
of the outer header (the "envelope") is being decremented.  The hop count
of the inner header (the "contents" of the envelope) is not decremented and
should not be decremented.  The hop count of the outer header is not
derived from and should not be derived from the hop count of the inner header.

Indeed, even if the packets did time out in transit along the tunnel, there
would be no way for traceroute to find out what happened.  Just as
information cannot leak _out_ of the tunnel to the outside, information
cannot leak _into_ the tunnel from outside, and this includes ICMP messages
from routers along the path.

There are some cases where one might wish for information about what is
happening at the IP layer (below the tunnel layer) -- but the protocol
makes no provision for this.  This raises all sorts of conceptual issues.
AFAIK nobody has ever cared enough to really figure out what _should_
happen, let alone implement it and standardize it.

*) I consider the "* * *" to be a slight bug.  One might wish for it to be
replaced by "GateB GateB GateB".  It has to do with treating host-to-subnet
traffic different from subnet-to-subnet traffic (and other gory details).
I fervently hope KLIPS2 will make this problem go away.

*) If you want to ask questions about the link from GateA to GateB at the
IP level (below the tunnel level), you have to ssh to GateA and launch a
traceroute from there.

Testing in stages

It is often useful in debugging to test things one at a time:

• disable IPsec entirely, for example by turning it off with chkconfig(8), and make sure routing works
• Once that works, try a manually keyed connection. This does not require key negotiation between Pluto and the key daemon on the other end.
• Once that works, try automatically keyed connections
• Once IPsec works, add packet compression
• Once everything seems to work, try stress tests with large transfers, many connections, frequent re-keying, ...

FreeS/WAN releases are tested for all of these, so you can be reasonably certain they can do them all. Of course, that does not mean they will on the first try, especially if you have some unusual configuration.

The rest of this section gives information on diagnosing the problem when each of the above steps fails.

Manually keyed connections don't work

Suspect one of:

• mis-configuration of IPsec system in the /etc/ipsec.conf file
  common errors are incorrect interface or next hop information
• mis-configuration of manual connection in the /etc/ipsec.conf file
• routing problems causing IPsec packets to be lost
• bugs in KLIPS
• mismatch between the transforms we support and those another IPsec implementation offers.

One manual connection works, but second one fails

This is a fairly common problem when attempting to configure multiple manually keyed connections from a single gateway.

Each connection must be identified by a unique SPI value. For automatic connections, these values are assigned automatically. For manual connections, you must set them with spi= statements in ipsec.conf(5).

Each manual connection must have a unique SPI value in the range 0x100 to 0x999. Two or more with the same value will fail. For details, see our doc section Using manual keying in production and the man page ipsec.conf(5).

Manual connections work, but automatic keying doesn't

The most common reason for this behaviour is a firewall dropping the UDP port 500 packets used in key negotiation.

Other possibilities:

• mis-configuration of auto connection in the /etc/ipsec.conf file.

  One common configuration error is forgetting that you need auto=add to load the connection description on the receiving end so it recognises the connection when the other end asks for it.

• error in shared secret in /etc/ipsec.secrets
• one gateway lacks a route to the other so Pluto's UDP packets are lost
• bugs in Pluto
• incompatibilities between Pluto's IKE implementation and the IKE at the other end of the tunnel.

  Some possibile problems are discussed in out interoperation document.

IPsec works, but connections using compression fail

When we first added compression, we saw some problems:

• compatibility issues with other implementations. We followed the RFCs and omitted some extra material that many compression libraries add by default. Some other implementations left the extras in
• bugs in assembler compression routines on non-Intel CPUs. The workaround is to use C code instead of possibly problematic assembler.

We have not seen either problem in some time (at least six months as I write in March 2002), but if you have some unusual configuration then you may see them.

Small packets work, but large transfers fail

If tests with ping(1) and a small packet size succeed, but tests or transfers with larger packet sizes fail, suspect problems with packet fragmentation and perhaps path MTU discovery.

Our troubleshooting document covers these problems. Information on the underlying mechanism is in our background document.

Subnet-to-subnet works, but tests from the gateways don't

This is described under I cannot ping... above.

Compilation problems

gmp.h: No such file or directory

Pluto needs the GMP (GNU

Multi-Precision) library for the large integer calculations it uses in public key cryptography. This error message indicates a failure to find the library. You must install it before Pluto will compile.

The GMP library is included in most Linux distributions. Typically, there are two RPMs, libgmp and libgmp-devel, You need to install both, either from your distribution CDs or from your vendor's web site.

For more information and the latest version, see the GMP home page.

... virtual memory exhausted

We have had several reports of this message appearing, all on SPARC Linux. Here is a mailing message on a solution:

> ipsec_sha1.c: In function `SHA1Transform':
> ipsec_sha1.c:95: virtual memory exhausted

I'm seeing exactly the same problem on an Ultra with 256MB ram and 500
MB swap.  Except I am compiling version 1.5 and its Red Hat 6.2.

I can get around this by using -O instead of -O2 for the optimization
level.  So it is probably a bug in the optimizer on the sparc complier. 
I'll try and chase this down on the sparc lists.

Interpreting error messages

route-client (or host) exited with status 7

Here is a discussion of this error from FreeS/WAN "listress" (mailing list tech support person) Claudia Schmeing. The "FAQ on the network unreachable error" which she refers to is the next question below.

> I reached the point where the two boxes (both on dial-up connections, but
> treated as static IPs by getting the IP and editing ipsec.conf after the
> connection is established) to the point where they exchange some info, but I
> get an error like "route-client command exited with status 7 \n internal
> error".
> Where can I find a description of this error?

In general, if the FAQ doesn't cover it, you can search the mailing list 
archives - I like to use
http://www.sandelman.ottawa.on.ca/linux-ipsec/
but you can see doc/mail.html for different archive formats.


Your error comes from the _updown script, which performs some
routing and firewall functions to help Linux FreeS/WAN. More info
is available at doc/firewall.html and man ipsec.conf. Its routing
is integral to the health of Linux FreeS/WAN; it also provides facility
to insert custom firewall rules to be executed when you create or destroy
a connection.

Yours is, of course, a routing error. You can be fairly sure the routing 
machinery is saying "network is unreachable". There's a FAQ on the 
"network is unreachable" error, but more information is available now; read on.

If your _updown script is recent (for example if it shipped with 
Linux FreeS/WAN 1.91), you will see another debugging line in your logs 
that looks something like this:

> output: /usr/local/lib/ipsec/_updown: `route add -net 128.174.253.83 
> netmask 255.255.255.255 dev ipsec0 gw 66.92.93.161' failed

This is, of course, the system route command that exited with status 7, 
(ie. failed). Man route for details. Seeing the command typed out yields 
more information. If your _updown script is older, you may wish to update 
it to show the command explicitly.

Three parameters fed to the route command: net, netmask and gw [gateway] 
are derived from things you've put in ipsec.conf.

Net and netmask are derived from the peer's IP and mask. In more detail:

You may see a routing error when routing to a client (ie. subnet), or 
to a host (IPSec gateway or freestanding host; a box that does IPSec for
itself). In _updown, the "route-client" section  is responsible to set up 
the route for IPSec'd (usually, read 'tunneled') packets headed to a 
peer subnet. Similarly, route-host routes IPSec'd packets to a peer host
or IPSec gateway.

When routing to a 'client', net and netmask are ipsec.conf's left- or 
rightsubnet (whichever is not local). Similarly, when routing to a 
'host' the net is left or right. Host netmask is always /32, indicating a 
single machine.

Gw is nexthop's value. Again, the value in question is left- or rightnexthop,
whichever is local. Where left/right or left-/rightnexthop has the special 
value %defaultroute (described in man ipsec.conf), gw will automagically get
the value of the next hop on the default route.

Q: "What's a nexthop and why do I need one?"

A: 'nexthop' is a routing kluge; its value is the next hop away
   from the machine that's doing IPSec, and toward your IPSec peer. 
   You need it to get the processed packets out of the local system and 
   onto the wire. While we often route other packets through the machine 
   that's now doing IPSec, and are done with it, this does not suffice here. 
   After packets are processed with IPSec, this machine needs to know where 
   they go next. Of course using the 'IPSec gateway' as their routing gateway 
   would cause an infinite loop! [To visualize this, see the packet flow 
   diagram at doc/firewall.html.] To avoid this, we route packets through 
   the next hop down their projected path.

Now that you know the background, consider:
1. Did you test routing between the gateways in the absence of Linux
   FreeS/WAN, as recommended? You need to ensure the two machines that
   will be running Linux FreeS/WAN can route to one another before trying to 
   make a secure connection.
2. Is there anything obviously wrong with the sense of your route command?

Normally, this problem is caused by an incorrect local nexthop parameter.
Check out the use of %defaultroute, described in man ipsec.conf. This is
a simple way to set nexthop for most people. To figure nexthop out by hand,
traceroute in-the-clear to your IPSec peer. Nexthop is the traceroute's 
first hop after your IPSec gateway.

SIOCADDRT:Network is unreachable

This message is not from FreeS/WAN, but from the Linux IP stack itself. That stack is seeing packets it has no route for, either because your routing was broken before FreeS/WAN started or because FreeS/WAN's changes broke it.

Here is a message from Claudia suggesting ways to diagnose and fix such problems:

You write,
> I have correctly installed freeswan-1.8 on RH7.0 kernel 2.2.17, but when 
> I setup a VPN connection with the other machine(RH5.2 Kernel 2.0.36 
> freeswan-1.0, it works well.) it told me that 
> "SIOCADDRT:Network is unreachable"!  But the network connection is no 
> problem.

Often this error is the result of a misconfiguration. 

Be sure that you can route successfully in the absence of Linux
FreeS/WAN. (You say this is no problem, so proceed to the next step.)

Use a custom copy of the default updownscript. Do not change the route 
commands, but add a diagnostic message revealing the exact text of the 
route command. Is there a problem with the sense of the route command
that you can see? If so, then re-examine those ipsec.conf settings
that are being sent to the route command. 

You may wish to use the ipsec auto --route and --unroute commands to 
troubleshoot the problem. See man ipsec_auto for details.

Since the above message was written, we have modified the updown script to provide a better diagnostic for this problem. Check /var/log/messages.

See also the FAQ question route-client (or host) exited with status 7.

ipsec_setup: modprobe: Can't locate module ipsec

ipsec_setup: Fatal error, kernel appears to lack KLIPS

These messages indicate an installation failure. The kernel you are running does not contain the KLIPS (kernel IPsec) code.

Note that the "modprobe: Can't locate module ipsec" message appears even if you are not using modules. If there is no KLIPS in your kernel, FreeS/WAN tries to load it as a module. If that fails, you get this message.

Commands you can quickly try are:

uname -a

to get details, including compilation date and time, of the currently running kernel

ls /

ls /boot

to ensure a new kernel is where it should be. If kernel compilation puts it in / but lilo wants it in /boot , then you should uncomment the INSTALL_PATH=/boot line in the kernel Makefile.

more /etc/lilo.conf

to see that lilo has correct information

lilo

to ensure that information in /etc/lilo.conf has been transferred to the boot sector

If those don't find the problem, you have to go back and check through the install procedure to see what was missed.

Here is one of Claudia's messages on the topic:

> I tried to install freeswan 1.8 on my mandrake 7.2 test box. ...

> It does show version and some output for whack.

Yes, because the Pluto (daemon) part of ipsec is installed correctly, but
as we see below the kernel portion is not.

> However, I get the following from /var/log/messages:
> 
> Mar 11 22:11:55 pavillion ipsec_setup: Starting FreeS/WAN IPsec 1.8...
> Mar 11 22:12:02 pavillion ipsec_setup: modprobe: Can't locate module ipsec
> Mar 11 22:12:02 pavillion ipsec_setup: Fatal error, kernel appears to lack
> KLIPS.

This is your problem. You have not successfully installed a kernel with
IPSec machinery in it. 

Did you build Linux FreeS/WAN as a module? If so, you need to ensure that 
your new module has been installed in the directory where your kernel 
loader normally finds your modules. If not, you need to ensure
that the new IPSec-enabled kernel is being loaded correctly.

See also doc/install.html, and INSTALL in the distro.

ipsec_setup: ... failure to fetch key for ... from DNS

Quoting Henry:

Note that by default, FreeS/WAN is now set up to
     (a) authenticate with RSA keys, and
     (b) fetch the public key of the far end from DNS.
Explicit attention to  ipsec.conf will be needed if you want
to do something different.

and Claudia, responding to the same user:

You write,

>       My current setup in ipsec.conf is leftrsasigkey=%dns I have 
> commented this and authby=rsasig out. I am able to get ipsec running, 
> but what I find is that the documentation only specifies for %dns are 
> there any other values that can be placed in this variable other than 
> %dns and the key? I am also assuming that this is where I would place 
> my public key for the left and right side as well is this correct?

Valid values for authby= are rsasig and secret, which entail authentication
by RSA signature or by shared secret, respectively. Because you have 
commented authby=rsasig out, you are using the default value of authby=secret. 

When using RSA signatures, there are two ways to get the public key for the
IPSec peer: either copy it directly into *rsasigkey= in ipsec.conf, or
fetch it from dns. The magic value %dns for *rsasigkey parameters says to 
try to fetch the peer's key from dns.

For any parameters, you may find their significance and special values in
man ipsec.conf. If you are setting up keys or secrets, be sure also to
reference man ipsec.secrets.

ipsec_setup: ... interfaces ... and ... share address ...

This is a fatal error. FreeS/WAN cannot cope with two or more interfaces using the same IP address. You must re-configure to avoid this.

A mailing list message on the topic from Pluto developer Hugh Redelmeier:

| I'm trying to get freeswan working between two machine where one has a ppp
| interface.
| I've already suceeded with  two machines with ethernet ports but  the ppp
| interface is causing me problems.
|  basically when I run ipsec start  i get
| ipsec_setup: Starting FreeS/WAN IPsec 1.7...
| ipsec_setup: 003 IP interfaces ppp1 and ppp0 share address 192.168.0.10!
| ipsec_setup: 003 IP interfaces ppp1 and ppp2 share address 192.168.0.10!
| ipsec_setup: 003 IP interfaces ppp0 and ppp2 share address 192.168.0.10!
| ipsec_setup: 003 no public interfaces found
|
| followed by lots of cannot work out interface for connection messages
|
| now I can specify the interface in ipsec.conf to be ppp0 , but this does
| not affect the above behaviour. A quick look in server.c indicates that the
| interfaces value  is not used but some sort of raw detect happens.
|
| I guess I could prevent the formation of the extra ppp interfaces or
| allocate them different ip but I'd  rather not. if at all possible. Any
| suggestions please.

Pluto won't touch an interface that shares an IP address with another.
This will eventually change, but it probably won't happen soon.

For now, you will have to give the ppp1 and ppp2 different addresses.

ipsec_setup: Cannot adjust kernel flags

A mailing list message form technical lead Henry Spencer:

> When FreeS/WAN IPsec 1.7 is starting on my 2.0.38 Linux kernel the following
> error message is generated:
> ipsec_setup: Cannot adjust kernel flags, no /proc/sys/net/ipsec directory!
> What is supposed to create this directory and how can I fix this problem?

I think that directory is a 2.2ism, although I'm not certain (I don't have
a 2.0.xx system handy any more for testing).  Without it, some of the
ipsec.conf config-setup flags won't work, but otherwise things should
function. 

You also need to enable the /proc filesystem in your kernel configuration for these operations to work.

Message numbers (MI3, QR1, et cetera) in Pluto messages

Pluto messages often indicate where Pluto is in the IKE protocols. The letters indicate Main mode or Q uick mode and Initiator or Responder. The numerals are message sequence numbers. For more detail, see our IPsec section.

Connection names in Pluto error messages

From Pluto programmer Hugh Redelmeier:

| Jan 17 16:21:10 remus Pluto[13631]: "jumble" #1: responding to Main Mode from Road Warrior 130.205.82.46
| Jan 17 16:21:11 remus Pluto[13631]: "jumble" #1: no suitable connection for peer @banshee.wittsend.com
| 
|     The connection "jumble" has nothing to do with the incoming
| connection requests, which were meant for the connection "banshee".

You are right.  The message tells you which Connection Pluto is
currently using, which need not be the right one.  It need not be the
right one now for the negotiation to eventually succeed!  This is
described in ipsec_pluto(8) in the section "Road Warrior Support".

There are two times when Pluto will consider switching Connections for
a state object.  Both are in response to receiving ID payloads (one in
Phase 1 / Main Mode and one in Phase 2 / Quick Mode).  The second is
not unique to Road Warriors.  In fact, neither is the first any more
(two connections for the same pair of hosts could differ in Phase 1 ID
payload; probably nobody else has tried this).

Pluto: ... can't orient connection

Older versions of FreeS/WAN used this message. The same error now gives the "we have no ipsecN ..." error described just below.

... we have no ipsecN interface for either end of this connection

Each Pluto needs to know whether it is running on the machine which the connection description calls left or on right . It figures that out by:

• looking at the interfaces given in interfaces= lines in the config setup section
• discovering the IP addresses for those interfaces
• searching for a match between those addresses and the ones given in left= or right= lines.

Normally a match is found. Then Pluto knows where it is and can set up other things (for example, if it is left) using parameters such as leftsubnet and leftnexthop, and sending its outgoing packets to right.

If no match is found, it emits the above error message.

Pluto: ... no connection is known

This error message occurs when a remote system attempts to negotiate a connection and Pluto does not have a connection description that matches what the remote system has requested. The most common cause is a configuration error on one end or the other.

Parameters involved in this match are left, right , leftsubnet and rightsubnet.

The match must be exact. For example, if your left subnet is a.b.c.0/24 then neither a single machine in that net nor a smaller subnet such as a.b.c.64/26 will be considered a match.

The message can also occur when an appropriate description exists but Pluto has not loaded it. Use an auto=add statement in the connection description, or an ipsec auto --add <conn_name> command, to correct this.

An explanation from the Pluto developer:

| Jul 12 15:00:22 sohar58 Pluto[574]: "corp_road" #2: cannot respond to IPsec
| SA request because no connection is known for
| 216.112.83.112/32===216.112.83.112...216.67.25.118

This is the first message from the Pluto log showing a problem.  It
means that PGPnet is trying to negotiate a set of SAs with this
topology:

216.112.83.112/32===216.112.83.112...216.67.25.118
^^^^^^^^^^^^^^^^^   ^^^^^^^^^^^^^^   ^^^^^^^^^^^^^
client on our side  our host         PGPnet host, no client

None of the conns you showed look like this.

Use
        ipsec auto --status
to see a snapshot of what connections are in pluto, what
negotiations are going on, and what SAs are established.

The leftsubnet= (client) in your conn is 216.112.83.64/26.  It must
exactly match what pluto is looking for, and it does not.

Pluto: ... no suitable connection ...

This is similar to the no connection known error, but occurs at a different point in Pluto processing.

Here is one of Claudia's messages explaining the problem:

You write,

> What could be the reason of the following error? 
> "no suitable connection for peer '@xforce'"

When a connection is initiated by the peer, Pluto must choose which entry in 
the conf file best matches the incoming connection. A preliminary choice is 
made on the basis of source and destination IPs, since that information is 
available at that time. 

A payload containing an ID arrives later in the negotiation. Based on this
id and the *id= parameters, Pluto refines its conn selection. ...

The message "no suitable connection" indicates that in this refining step,
Pluto does not find a connection that matches that ID.

Please see "Selecting a connection when responding" in man ipsec_pluto for
more details.

See also Connection names in Pluto error messages.

Pluto: ... no connection has been authorized

Here is one of Claudia's messages discussing this problem:

You write,

>  May 22 10:46:31 debian Pluto[25834]: packet from x.y.z.p:10014: 
>  initial Main Mode message from x.y.z.p:10014 
                            but no connection has been authorized

This error occurs early in the connection negotiation process,
at the first step of IKE negotiation (Main Mode), which is itself the 
first of two negotiation phases involved in creating an IPSec connection.

Here, Linux FreeS/WAN receives a packet from a potential peer, which 
requests that they begin discussing a connection.

The "no connection has been authorized" means that there is no connection 
description in Linux FreeS/WAN's internal database that can be used to 
link your ipsec interface with that peer.

"But of course I configured that connection!" 

It may be that the appropriate connection description exists in ipsec.conf 
but has not been added to the database with ipsec auto --add myconn or the 
auto=add method. Or, the connection description may be misconfigured.

The only parameters that are relevant in this decision are left= and right= .
Local and remote ports are also taken into account -- we see that the port 
is printed in the message above -- but there is no way to control these
in ipsec.conf.


Failure at "no connection has been authorized" is similar to the
"no connection is known for..." error in the FAQ, and the "no suitable
connection" error described in the snapshot's FAQ. In all three cases,
Linux FreeS/WAN is trying to match parameters received in the
negotiation with the connection description in the local config file.

As it receives more information, its matches take more parameters into 
account, and become more precise:  first the pair of potential peers,
then the peer IDs, then the endpoints (including any subnets).

The "no suitable connection for peer *" occurs toward the end of IKE 
(Main Mode) negotiation, when the IDs are matched.

"no connection is known for a/b===c...d" is seen at the beginning of IPSec 
(Quick Mode, phase 2) negotiation, when the connections are matched using
left, right, and any information about the subnets.

Pluto: ... OAKLEY_DES_CBC is not supported.

This message occurs when the other system attempts to negotiate a connection using single DES, which we do not support because it is insecure.

Our interoperation document has suggestions for how to deal with systems that attempt to use single DES.

Pluto: ... no acceptable transform

This message means that the other gateway has made a proposal for connection parameters, but nothing they proposed is acceptable to Pluto. Possible causes include:

• misconfiguration on either end
• policy incompatibilities, for example we require encrypted connections but they are trying to create one with just authentication
• interoperation problems, for example they offer only single DES and FreeS/WAN does not support that. See discussion in our interoperation document.

A more detailed explanation, from Pluto programmer Hugh Redelmeier:

Background:

When one IKE system (for example, Pluto) is negotiating with another
to create an SA, the Initiator proposes a bunch of choices and the
Responder replies with one that it has selected.

The structure of the choices is fairly complicated.  An SA payload
contains a list of lists of "Proposals".  The outer list is a set of
choices: the selection must be from one element of this list.

Each of these elements is a list of Proposals.  A selection must be
made from each of the elements of the inner list.  In other words,
*all* of them apply (that is how, for example, both AH and ESP can
apply at once).

Within each of these Proposals is a list of Transforms.  For each
Proposal selected, one Transform must be selected (in other words,
each Proposal provides a choice of Transforms).

Each Transform is made up of a list of Attributes describing, well,
attributes.  Such as lifetime of the SA.  Such as algorithm to be
used.  All the Attributes apply to a Transform.

You will have noticed a pattern here: layers alternate between being
disjunctions ("or") and conjunctions ("and").

For Phase 1 / Main Mode (negotiating an ISAKMP SA), this structure is
cut back.  There must be exactly one Proposal.  So this degenerates to
a list of Transforms, one of which must be chosen.

In your case, no proposal was considered acceptable to Pluto (the
Responder).  So negotiation ceased.  Pluto logs the reason it rejects
each Transform.  So look back in the log to see what is going wrong.

rsasigkey dumps core

On Fri, 29 Jun 2001, Rodrigo Gruppelli wrote:
> ...Well, it seem that there's
> another problem with it. When I try to generate a pair of RSA keys,
> rsasigkey cores dump...

*That* is a neon sign flashing "GMP LIBRARY IS BROKEN".  Rsasigkey calls
GMP a lot, and our own library a little bit, and that's very nearly all it
does.  Barring bugs in its code or our library -- which have happened, but
not very often -- a problem in rsasigkey is a problem in GMP.

See the next question for how to deal with GMP errors.

!Pluto failure!: ... exited with ... signal 4

Pluto has died. Signal 4 is SIGILL, illegal instruction.

The most likely cause is that your GMP (GNU multi-precision) library is compiled for a different processor than what you are running on. Pluto uses that library for its public key calculations.

Try getting the GMP sources and recompile for your processor type. Most Linux distributions will include this source, or you can download it from the GMP home page.

ECONNREFUSED error message

From John Denker, on the mailing list:

1)  The log message
  some IKE message we sent has been rejected with 
  ECONNREFUSED (kernel supplied no details)
is much more suitable than the previous version.  Thanks.

2) Minor suggestion for further improvement: it might be worth mentioning
that the command
  tcpdump -i eth1 icmp[0] != 8 and icmp[0] != 0
is useful for tracking down the details in question.  We shouldn't expect
all IPsec users to figure that out on their own.  The log message might
even provide a hint as to where to look in the docs.

Reply From Pluto developer Hugh Redelmeier

Good idea.

I've added a bit pluto(8)'s BUGS section along these lines.
I didn't have the heart to lengthen this message.

klips_debug: ... no eroute!

This message means KLIPS has received a packet for which no IPsec tunnel has been defined.

Here is a more detailed duscussion from the team's tech support person Claudia Schmeing, responding to a query on the mailing list:

> Why ipsec reports no eroute! ???? IP Masq... is disabled.

In general, more information is required so that people on the list may
give you informed input. See doc/prob.report.

The document she refers to has since been replaced by a section of the troubleshooting document.

However, I can make some general comments on this type of error.

This error usually looks something like this (clipped from an archived
message):

> ttl:64 proto:1 chk:45459 saddr:192.168.1.2 daddr:192.168.100.1
> ... klips_debug:ipsec_findroute: 192.168.1.2->192.168.100.1
> ... klips_debug:rj_match: * See if we match exactly as a host destination
> ... klips_debug:rj_match: ** try to match a leaf, t=0xc1a260b0
> ... klips_debug:rj_match: *** start searching up the tree, t=0xc1a260b0
> ... klips_debug:rj_match: **** t=0xc1a260c8
> ... klips_debug:rj_match: **** t=0xc1fe5960
> ... klips_debug:rj_match: ***** not found.
> ... klips_debug:ipsec_tunnel_start_xmit: Original head/tailroom: 2, 28
> ... klips_debug:ipsec_tunnel_start_xmit: no eroute!: ts=47.3030, dropping.


What does this mean?
- --------------------

"eroute" stands for "extended route", and is a special type of route 
internal to Linux FreeS/WAN. For more information about this type of route, 
see the section of man ipsec_auto on ipsec auto --route.

"no eroute!" here means, roughly, that Linux FreeS/WAN cannot find an 
appropriate tunnel that should have delivered this packet. Linux 
FreeS/WAN therefore drops the packet, with the message "no eroute! ...
dropping", on the assumption that this packet is not a legitimate 
transmission through a properly constructed tunnel.


How does this situation come about?
- -----------------------------------

Linux FreeS/WAN has a number of connection descriptions defined in 
ipsec.conf. These must be successfully brought "up" to form actual tunnels.
(see doc/setup.html's step 15, man ipsec.conf and man ipsec_auto 
for details).

Such connections are often specific to the endpoints' IPs. However, in 
some cases they may be more general, for example in the case of 
Road Warriors where left or right is the special value %any.

When Linux FreeS/WAN receives a packet, it verifies that the packet has
come through a legitimate channel, by checking that there is an
appropriate tunnel through which this packet might legitimately have
arrived. This is the process we see above.

First, it checks for an eroute that exactly matches the packet. In the 
example above, we see it checking for a route that begins at 192.168.1.2
and ends at 192.168.100.1. This search favours the most specific match that
would apply to the route between these IPs. So, if there is a connection 
description exactly matching these IPs, the search will end there. If not, 
the code will search for a more general description matching the IPs.
If there is no match, either specific or general, the packet will be
dropped, as we see, above.

Unless you are working with Road Warriors, only the first, specific part 
of the matching process is likely to be relevant to you.


"But I defined the tunnel, and it came up, why do I have this error?"
- ---------------------------------------------------------------------

One of the most common causes of this error is failure to specify enough
connection descriptions to cover all needed tunnels between any two 
gateways and their respective subnets. As you have noticed, troubleshooting
this error may be complicated by the use of IP Masq. However, this error is
not limited to cases where IP Masq is used. 

See doc/configuration.html#multitunnel for a detailed example of the 
solution to this type of problem.

The documentation section she refers to is now here.

... trouble writing to /dev/ipsec ... SA already in use

This error message occurs when two manual connections are set up with the same SPI value.

See the FAQ for One manual connection works, but second one fails.

... ignoring ... payload

This message is harmless. The IKE protocol provides for a number of optional messages types:

• delete SA
• initial contact
• vendor ID
• ...

An implementation is never required to send these, but they are allowed to. The receiver is not required to do anything with them. FreeS/WAN ignores them, but notifies you via the logs.

For the "ignoring delete SA Payload" message, see also our discussion of cleaning up dead tunnels.

Why don't you restrict the mailing lists to reduce spam?

As a matter of policy, some of our mailing lists need to be open to non-subscribers. Project management feel strongly that maintaining this openness is more important than blocking spam.

• Users should be able to get help or report bugs without subscribing.
• Even a user who is subscribed may not have access to his or her subscribed account when he or she needs help, miles from home base in the middle of setting up a client's gateway.
• There is arguably a legal requirement for this policy. A US resident or citizen could be charged under munitions export laws for providing technical assistance to a foreign cryptographic project. Such a charge would be more easily defended if the discussion takes place in public, on an open list.

This has been discussed several times at some length on the list. See the list archives. Bringing the topic up again is unlikely to be useful. Please don't. Or at the very least, please don't without reading the archives and being certain that whatever you are about to suggest has not yet been discussed.

Project technical lead Henry Spencer summarised one discussion:

For the third and last time: this list *will* *not* do address-based filtering. This is a policy decision, not an implementation problem. The decision is final, and is not open to discussion. This needs to be communicated better to people, and steps are being taken to do that.

Adding this FAQ section is one of the steps he refers to.

You have various options other than just putting up with the spam, filtering it yourself, or unsubscribing:

• subscribe only to one or both of our lists with restricted posting rules:
  • briefs , weekly list summaries
  • announce , project-related announcements
• read the other lists via the archives

A number of tools are available to filter mail.

• Many mail readers include some filtering capability.
• Many Linux distributions include procmail(8) for server-side filtering.
• The Spam Bouncer is a set of procmail(8) filters designed to combat spam.
• Roaring Penguin have a MIME defanger that removes potentially dangerous attachments.

If you use your ISP's mail server rather than running your own, consider suggesting to the ISP that they tag suspected spam as this ISP does. They could just refuse mail from dubious sources, but that is tricky and runs some risk of losing valuable mail or senselessly annoying senders and their admins. However, they can safely tag and deliver dubious mail. The tags can greatly assist your filtering.

For information on tracking down spammers, see these HowTos, or the Sputum site. Sputum have a Linux anti-spam screensaver available for download.

Here is a more detailed message from Henry:

On Mon, 15 Jan 2001, Jay Vaughan wrote:
> I know I'm flogging a dead horse here, but I'm curious as to the reasons for
> an aversion for a subscriber-only mailing list?

Once again:  for legal reasons, it is important that discussions of these
things be held in a public place -- the list -- and we do not want to
force people to subscribe to the list just to ask one question, because
that may be more than merely inconvenient for them.  There are also real
difficulties with people who are temporarily forced to use alternate
addresses; that is precisely the time when they may be most in need of
help, yet a subscribers-only policy shuts them out.

These issues do not apply to most mailing lists, but for a list that is
(necessarily) the primary user support route for a crypto package, they
are very important.  This is *not* an ordinary mailing list; it has to
function under awkward constraints that make various simplistic solutions
inapplicable or undesirable. 

> We're *ALL* sick of hearing about list management problems, not just you
> old-timers, so why don't you DO SOMETHING EFFECTIVE ABOUT IT...

Because it's a lot harder than it looks, and many existing "solutions"
have problems when examined closely.

> A suggestion for you, based on 10 years of experience with management of my
> own mailing lists would be to use mailman, which includes pretty much every
> feature under the sun that you guys need and want, plus some.  The URL for
> mailman...

I assure you, we're aware of mailman.  Along with a whole bunch of others,
including some you almost certainly have never heard of (I hadn't!).

> As for the argument that the list shouldn't be configured to enforce
> subscription - I contend that it *SHOULD* AT LEAST require manual address
> verification in order for posts to be redirected.

You do realize, I hope, that interposing such a manual step might cause
your government to decide that this is not truly a public forum, and thus
you could go to jail if you don't get approval from them before mailing to
it?  If you think this sounds irrational, your government is noted for
making irrational decisions in this area; we can't assume that they will
suddenly start being sensible.  See above about awkward constraints.  You
may be willing to take the risk, but we can't, in good conscience, insist
that all users with problems do so. 

                                                          Henry Spencer
                                                       henry@spsystems.net

and a message on the topic from project leader John Gilmore:

Subject: Re: The linux-ipsec list's topic
   Date: Sat, 30 Dec 2000
   From: John Gilmore <gnu@toad.com>

I'll post this single message, once only, in this discussion, and then
not burden the list with any further off-topic messages.  I encourage
everyone on the list to restrain themself from posting ANY off-topic
messages to the linux-ipsec list.

The topic of the linux-ipsec mailing list is the FreeS/WAN software.

I frequently see "discussions about spam on a list" overwhelm the
volume of "actual spam" on a list. BOTH kinds of messages are
off-topic messages.  Twenty anti-spam messages take just as long to
detect and discard as twenty spam messages.

The Linux-ipsec list encourages on-topic messages from people who have
not joined the list itself.  We will not censor messages to the list
based on where they originate, or what return address they contain.
In other words, non-subscribers ARE allowed to post, and this will not
change.  My own valid contributions have been rejected out-of-hand by
too many other mailing lists for me to want to impose that censorship
on anybody else's contributions.  And every day I see the damage that
anti-spam zeal is causing in many other ways; that zeal is far more
damaging to the culture of the Internet than the nuisance of spam.

In general, it is the responsibility of recipients to filter,
prioritize, or otherwise manage the handling of email that comes to
them.  It is not the responsibility of the rest of the Internet
community to refrain from sending messages to recipients that they
might not want to see.  If your software infrastructure for managing
your incoming email is insufficient, then improve it.  If you think
the signal-to-noise ratio on linux-ipsec is too poor, then please
unsubscribe.  But don't further increase the noise by posting to the
linux-ipsec list about those topics.

        John Gilmore
        founder & sponsor, FreeS/WAN project

FreeS/WAN manual pages

The various components of Linux FreeS/WAN are of course documented in standard Unix manual pages, accessible via the man(1) command.

Links here take you to an HTML version of the man pages.

Files

ipsec.conf(5)

IPsec configuration and connections

ipsec.secrets(5)

secrets for IKE authentication, either pre-shared keys or RSA private keys

These files are also discussed in the configuration section.

Commands

Many users will never give most of the FreeS/WAN commands directly. Configure the files listed above correctly and everything should be automatic.

The exceptions are commands for mainpulating the RSA keys used in Pluto authentication:

ipsec_rsasigkey(8)

generate keys

ipsec_newhostkey(8)

generate keys in a convenient format

ipsec_showhostkey(8)

extract RSA keys from ipsec.secrets(5) (or optionally, another file) and format them for insertion in ipsec.conf(5) or in DNS records

Note that:

The following commands are fairly likely to be used, if only for testing and status checks:

ipsec(8)

invoke IPsec utilities

ipsec_setup(8)

control IPsec subsystem

ipsec_auto(8)

control automatically-keyed IPsec connections

ipsec_manual(8)

take manually-keyed IPsec connections up and down

ipsec_ranbits(8)

generate random bits in ASCII form

ipsec_look(8)

show minimal debugging information

ipsec_barf(8)

spew out collected IPsec debugging information

The lower-level utilities listed below are normally invoked via scripts listed above, but they can also be used directly when required.

ipsec_eroute(8)

manipulate IPsec extended routing tables

ipsec_klipsdebug(8)

set Klips (kernel IPsec support) debug features and level

ipsec_pluto(8)

IPsec IKE keying daemon

ipsec_spi(8)

manage IPsec Security Associations

ipsec_spigrp(8)

group/ungroup IPsec Security Associations

ipsec_tncfg(8)

associate IPsec virtual interface with real interface

ipsec_whack(8)

control interface for IPsec keying daemon

Library routines

ipsec_atoaddr(3)

ipsec_addrtoa(3)

convert Internet addresses to and from ASCII

ipsec_atosubnet(3)

ipsec_subnettoa(3)

convert subnet/mask ASCII form to and from addresses

ipsec_atoasr(3)

convert ASCII to Internet address, subnet, or range

ipsec_rangetoa(3)

convert Internet address range to ASCII

ipsec_atodata(3)

ipsec_datatoa(3)

convert binary data from and to ASCII formats

ipsec_atosa(3)

ipsec_satoa(3)

convert IPsec Security Association IDs to and from ASCII

ipsec_atoul(3)

ipsec_ultoa(3)

convert unsigned-long numbers to and from ASCII

ipsec_goodmask(3)

is this Internet subnet mask a valid one?

ipsec_masktobits(3)

convert Internet subnet mask to bit count

ipsec_bitstomask(3)

convert bit count to Internet subnet mask

ipsec_optionsfrom(3)

read additional ``command-line'' options from file

ipsec_subnetof(3)

given Internet address and subnet mask, return subnet number

ipsec_hostof(3)

given Internet address and subnet mask, return host part

ipsec_broadcastof(3)

given Internet address and subnet mask, return broadcast address

FreeS/WAN and firewalls

FreeS/WAN, or other IPsec implementations, frequently run on gateway machines, the same machines running firewall or packet filtering code. This document discusses the relation between the two.

The firewall code in 2.4 and later kernels is called Netfilter. The user-space utility to manage a firwewall is iptables(8). See the netfilter/iptables web site for details.

Filtering rules for IPsec packets

The basic constraint is that an IPsec gateway must have packet filters that allow IPsec packets, at least when talking to other IPsec gateways:

• UDP port 500 for IKE negotiations
• protocol 50 if you use ESP encryption and/or authentication (the typical case)
• protocol 51 if you use AH packet-level authentication

Your gateway and the other IPsec gateways it communicates with must be able to exchange these packets for IPsec to work. Firewall rules must allow UDP 500 and at least one of AH or ESP on the interface that communicates with the other gateway.

For nearly all FreeS/WAN applications, you must allow UDP port 500 and the ESP protocol.

There are two ways to set this up:

easier but less flexible

Just set up your firewall scripts at boot time to allow IPsec packets to and from your gateway. Let FreeS/WAN reject any bogus packets.

more work, giving you more precise control

Have the ipsec_pluto(8) daemon call scripts to adjust firewall rules dynamically as required. This is done by naming the scripts in the ipsec.conf(5) variables prepluto=, postpluto= , leftupdown= and rightupdown=.

Both methods are described in more detail below.

Firewall configuration at boot

It is possible to set up both firewalling and IPsec with appropriate scripts at boot and then not use leftupdown= and rightupdown=, or use them only for simple up and down operations.

Basically, the technique is

• allow IPsec packets (typically, IKE on UDP port 500 plus ESP, protocol 50)
  • incoming, if the destination address is your gateway (and optionally, only from known senders)
  • outgoing, with the from address of your gateway (and optionally, only to known receivers)
• let Pluto deal with IKE
• let KLIPS deal with ESP

Since Pluto authenticates its partners during the negotiation, and KLIPS drops packets for which no tunnel has been negotiated, this may be all you need.

A simple set of rules

In simple cases, you need only a few rules, as in this example:

# allow IPsec
#
# IKE negotiations
iptables -A INPUT  -p udp --sport 500 --dport 500 -j ACCEPT
iptables -A OUTPUT -p udp --sport 500 --dport 500 -j ACCEPT
# ESP encrypton and authentication
iptables -A INPUT  -p 50 -j ACCEPT
iptables -A OUTPUT -p 50 -j ACCEPT
# uncomment for AH authentication header
# iptables -A INPUT  -p 51 -j ACCEPT
# iptables -A OUTPUT -p 51 -j ACCEPT

Other rules

However, while it is certainly possible to create an elaborate set of rules yourself (please let us know via the mailing list if you do), it may be both easier and more secure to use a set which has already been published and tested.

The published rule sets we know of are described in the next section.

Adding additional rules

reject IPsec packets that are not to or from known gateways

This possibility is discussed in more detail later

allow systems behind your gateway to build IPsec tunnels that pass through the gateway

This possibility is discussed in more detail later

filter incoming packets emerging from KLIPS.

Firewall rules can recognise packets emerging from IPsec. They are marked as arriving on an interface such as ipsec0, rather than eth0, ppp0 or whatever.

It is therefore reasonably straightforward to filter these packets in whatever way suits your situation.

Modifying existing rules

In some cases rules that work fine before you add IPsec may require modification to work with IPsec.

This is especially likely for rules that deal with interfaces on the Internet side of your system. IPsec adds a new interface; often the rules must change to take account of that.

For example, consider the rules given in this section of the Netfilter documentation:

Most people just have a single PPP connection to the Internet, and don't
want anyone coming back into their network, or the firewall:

    ## Insert connection-tracking modules (not needed if built into kernel).
    # insmod ip_conntrack
    # insmod ip_conntrack_ftp

    ## Create chain which blocks new connections, except if coming from inside.
    # iptables -N block
    # iptables -A block -m state --state ESTABLISHED,RELATED -j ACCEPT
    # iptables -A block -m state --state NEW -i ! ppp0 -j ACCEPT
    # iptables -A block -j DROP

    ## Jump to that chain from INPUT and FORWARD chains.
    # iptables -A INPUT -j block
    # iptables -A FORWARD -j block

On an IPsec gateway, those rules may need to be modified. The above allows new connections from anywhere except ppp0. That means new connections from ipsec0 are allowed.

Do you want to allow anyone who can establish an IPsec connection to your gateway to initiate TCP connections to any service on your network? Almost certainly not if you are using opportunistic encryption. Quite possibly not even if you have only explicitly configured connections.

To disallow incoming connections from ipsec0, change the middle section above to:

    ## Create chain which blocks new connections, except if coming from inside.
    # iptables -N block
    # iptables -A block -m state --state ESTABLISHED,RELATED -j ACCEPT
    # iptables -A block -m state --state NEW -i ppp+ -j DROP
    # iptables -A block -m state --state NEW -i ipsec+ -j DROP
    # iptables -A block -m state --state NEW -i -j ACCEPT
    # iptables -A block -j DROP

The original rules accepted NEW connections from anywhere except ppp0. This version drops NEW connections from any PPP interface (ppp+) and from any ipsec interface (ipsec+), then accepts the survivors.

Of course, these are only examples. You will need to adapt them to your own situation.

Published rule sets

Several sets of firewall rules that work with FreeS/WAN are available.

Scripts based on Ranch's work

One user, Rob Hutton, posted his boot time scripts to the mailing list, and we included them in previous versions of this documentation. They are still available from our web site. However, they were for an earlier FreeS/WAN version so we no longer recommend them. Also, they had some bugs. See this message.

Those scripts were based on David Ranch's scripts for his "Trinity OS" for setting up a secure Linux. Check his home page for the latest version and for information on his book on securing Linux. If you are going to base your firewalling on Ranch's scripts, we recommend using his latest version, and sending him any IPsec modifications you make for incorporation into later versions.

The Seattle firewall

We have had several mailing lists reports of good results using FreeS/WAN with Seawall (the Seattle Firewall). See that project's home page on Sourceforge.

The RCF scripts

Another set of firewall scripts with IPsec support are the RCF or rc.firewall scripts. See their home page.

Asgard scripts

Asgard's Realm has set of firewall scripts with FreeS/WAN support, for 2.4 kernels and iptables.

User scripts from the mailing list

One user gave considerable detail on his scripts, including supporting IPX through the tunnel. His message was too long to conveniently be quoted here, so I've put it in a separate file.

Calling firewall scripts, named in ipsec.conf(5)

The ipsec.conf(5) configuration file has three pairs of parameters used to specify an interface between FreeS/WAN and firewalling code.

Note that using these is not required if you have a static firewall setup. In that case, you just set your firewall up at boot time (in a way that permits the IPsec connections you want) and do not change it thereafter. Omit all the FreeS/WAN firewall parameters and FreeS/WAN will not attempt to adjust firewall rules at all. See above for some information on appropriate scripts.

However, if you want your firewall rules to change when IPsec connections change, then you need to use these parameters.

Scripts called at IPsec start and stop

One pair of parmeters are set in the config setup section of the ipsec.conf(5) file and affect all connections:

prepluto=

script to be called before pluto(8) IKE daemon is started.

postpluto=

script to be called after pluto(8) IKE daemon is stopped.

They can also be used in other ways. For example, you might have prepluto add a module to your kernel for the secure network interface or make a dialup connection, and then have postpluto remove the module or take the connection down.

Scripts called at connection up and down

The other parameters are set in connection descriptions. They can be set in individual connection descriptions, and could even call different scripts for each connection for maximum flexibility. In most applications, however, it makes sense to use only one script and to call it from conn %default section so that it applies to all connections.

You can:

either

set leftfirewall=yes or rightfirewall=yes to use our supplied default script

or

assign a name in a leftupdown= or rightupdown= line to use your own script

Note that only one of these should be used. You cannot sensibly use both. Since our default script is obsolete (designed for firewalls using ipfwadm(8) on 2.0 kernels), most users who need this service will need to write a custom script.

The default script

We supply a default script named _updown.

leftfirewall=

rightfirewall=

indicates that the gateway is doing firewalling and that pluto(8) should poke holes in the firewall as required.

Set these to yes and Pluto will call our default script _updown with appropriate arguments whenever it:

• starts or stops IPsec services
• brings a connection up or down

The supplied default _updown script is appropriate for simple cases using the ipfwadm(8) firewalling package.

User-written scripts

You can also write your own script and have Pluto call it. Just put the script's name in one of these ipsec.conf(5) lines:

leftupdown=

rightupdown=

specifies a script to call instead of our default script _updown.

Your script should take the same arguments and use the same environment variables as _updown. See the "updown command" section of the ipsec_pluto(8) man page for details.

Note that you should not modify our _updown script in place . If you did that, then upgraded FreeS/WAN, the upgrade would install a new default script, overwriting your changes.

Scripts for ipchains or iptables

Our _updown is for firewalls using ipfwadm(8), the firewall code for the 2.0 series of Linux kernels. If you are using the more recent packages ipchains(8) (for 2.2 kernels) or iptables(8) (2.4 kernels), then you must do one of:

• use static firewall rules which are set up at boot time as described above and do not need to be changed by Pluto
• limit yourself to ipchains(8)'s ipfwadm(8) emulation mode in order to use our script
• write your own script and call it with leftupdown and rightupdown.

You can write a script to do whatever you need with firewalling. Specify its name in a [left|right]updown= parameter in ipsec.conf(5) and Pluto will automatically call it for you.

The arguments Pluto passes such a script are the same ones it passes to our default _updown script, so the best way to build yours is to copy ours and modify the copy.

Note, however, that you should not modify our _updown script in place. If you did that, then upgraded FreeS/WAN, the upgrade would install a new default script, overwriting your changes.

A complication: IPsec vs. NAT

Network Address Translation, also known as IP masquerading, is a method of allocating IP addresses dynamically, typically in circumstances where the total number of machines which need to access the Internet exceeds the supply of IP addresses.

Any attempt to perform NAT operations on IPsec packets between the IPsec gateways creates a basic conflict:

• IPsec wants to authenticate packets and ensure they are unaltered on a gateway-to-gateway basis
• NAT rewrites packet headers as they go by
• IPsec authentication fails if packets are rewritten anywhere between the IPsec gateways

For AH, which authenticates parts of the packet header including source and destination IP addresses, this is fatal. If NAT changes those fields, AH authentication fails.

For IKE and ESP it is not necessarily fatal, but is certainly an unwelcome complication.

NAT on or behind the IPsec gateway works

This problem can be avoided by having the masquerading take place on or behind the IPsec gateway.

This can be done physically with two machines, one physically behind the other. A picture, using SG to indicate IPsec S ecurity Gateways, is:

      clients --- NAT ----- SG ---------- SG
                  two machines

In this configuration, the actual client addresses need not be given in the leftsubnet= parameter of the FreeS/WAN connection description. The security gateway just delivers packets to the NAT box; it needs only that machine's address. What that machine does with them does not affect FreeS/WAN.

A more common setup has one machine performing both functions:

      clients ----- NAT/SG ---------------SG
                  one machine

Here you have a choice of techniques depending on whether you want to make your client subnet visible to clients on the other end:

• If you want the single gateway to behave like the two shown above, with your clients hidden behind the NAT, then omit the leftsubnet= parameter. It then defaults to the gateway address. Clients on the other end then talk via the tunnel only to your gateway. The gateway takes packets emerging from the tunnel, applies normal masquerading, and forwards them to clients.
• If you want to make your client machines visible, then give the client subnet addresses as the leftsubnet= parameter in the connection description and

  either

  set leftfirewall=yes to use the default updown script

  or

  use your own script by giving its name in a leftupdown= parameter

  These scripts are described in their own section.

  In this case, no masquerading is done. Packets to or from the client subnet are encrypted or decrypted without any change to their client subnet addresses, although of course the encapsulating packets use gateway addresses in their headers. Clients behind the right security gateway see a route via that gateway to the left subnet.

NAT between gateways is problematic

We recommend not trying to build IPsec connections which pass through a NAT machine. This setup poses problems:

      clients --- SG --- NAT ---------- SG

If you must try it, some references are:

• Jean_Francois Nadeau's document on doing IPsec behind NAT
• VPN masquerade patches to make a Linux NAT box handle IPsec packets correctly

Other references on NAT and IPsec

Other documents which may be relevant include:

• an Internet Draft on IPsec and NAT which may eventually evolve into a standard solution for this problem.
• an informational RFC, Security Model with Tunnel-mode IPsec for NAT Domains .
• an article in Cisco's Internet Protocol Journal

Other complications

Of course simply allowing UDP 500 and ESP packets is not the whole story. Various other issues arise in making IPsec and packet filters co-exist and even co-operate. Some of them are summarised below.

IPsec through the gateway

Basic IPsec packet filtering rules deal only with packets addressed to or sent from your IPsec gateway.

It is a separate policy decision whether to permit such packets to pass through the gateway so that client machines can build end-to-end IPsec tunnels of their own. This may not be practical if you are using NAT (IP masquerade) on your gateway, and may conflict with some corporate security policies.

Where possible, allowing this is almost certainly a good idea. Using IPsec on an end-to-end basis is more secure than gateway-to-gateway.

Doing it is quite simple. You just need firewall rules that allow UDP port 500 and protocols 50 and 51 to pass through your gateway. If you wish, you can of course restrict this to certain hosts.

Preventing non-IPsec traffic

One application of this is for the telecommuter who might have:

     Sunset==========West------------------East ================= firewall --- the Internet
         home network      untrusted net        corporate network

The subnet on the right is 0.0.0.0/0, the whole Internet. The West gateway is set up so that it allows only IPsec packets to East in or out.

This configuration is used in AT&T Research's network. For details, see the papers links in our introduction.

Another application would be to set up firewall rules so that an internal machine, such as an employees-only web server, could not talk to the outside world except via specific IPsec tunnels.

Filtering packets from unknown gateways

It is possible to use firewall rules to restrict UDP 500, ESP and AH packets so that these packets are accepted only from known gateways. This is not strictly necessary since FreeS/WAN will discard packets from unknown gateways. You might, however, want to do it for any of a number of reasons. For example:

• Arguably, "belt and suspenders" is the sensible approach to security. If you can block a potential attack in two ways, use both. The only question is whether to look for a third way after implementing the first two.
• Some admins may prefer to use the firewall code this way because they prefer firewall logging to FreeS/WAN's logging.
• You may need it to implement your security policy. Consider an employee working at home, and a policy that says traffic from the home system to the Internet at large must go first via IPsec to the corporate LAN and then out to the Internet via the corporate firewall. One way to do that is to make ipsec0 the default route on the home gateway and provide exceptions only for UDP 500 and ESP to the corporate gateway. Everything else is then routed via the tunnel to the corporate gateway.

It is not possible to use only static firewall rules for this filtering if you do not know the other gateways' IP addresses in advance, for example if you have "road warriors" who may connect from a different address each time or if want to do opportunistic encryption to arbitrary gateways. In these cases, you can accept UDP 500 IKE packets from anywhere, then use the updown script feature of pluto(8) to dynamically adjust firewalling for each negotiated tunnel.

Firewall packet filtering does not much reduce the risk of a denial of service attack on FreeS/WAN. The firewall can drop packets from unknown gateways, but KLIPS does that quite efficiently anyway, so you gain little. The firewall cannot drop otherwise legitmate packets that fail KLIPS authentication, so it cannot protect against an attack designed to exhaust resources by making FreeS/WAN perform many expensive authentication operations.

In summary, firewall filtering of IPsec packets from unknown gateways is possible but not strictly necessary.

Other packet filters

When the IPsec gateway is also acting as your firewall, other packet filtering rules will be in play. In general, those are outside the scope of this document. See our Linux firewall links for information. There are a few types of packet, however, which can affect the operation of FreeS/WAN or of diagnostic tools commonly used with it. These are discussed below.

ICMP filtering

ICMP is the Internet Control Message Protocol. It is used for messages between IP implementations themselves, whereas IP used is used between the clients of those implementations. ICMP is, unsurprisingly, used for control messages. For example, it is used to notify a sender that a desination is not reachable, or to tell a router to reroute certain packets elsewhere.

ICMP handling is tricky for firewalls.

• You definitely want some ICMP messages to get through; things won't work without them. For example, your clients need to know if some destination they ask for is unreachable.
• On the other hand, you do equally definitely do not want untrusted folk sending arbitrary control messages to your machines. Imagine what someone moderately clever and moderately malicious could do to you, given control of your network's routing.

ICMP does not use ports. Messages are distinguished by a "message type" field and, for some types, by an additional "code" field. The definitive list of types and codes is on the IANA site.

One expert uses this definition for ICMP message types to be dropped at the firewall.

# ICMP types which lack socially redeeming value.
#  5     Redirect
#  9     Router Advertisement
# 10     Router Selection
# 15     Information Request
# 16     Information Reply
# 17     Address Mask Request
# 18     Address Mask Reply

badicmp='5 9 10 15 16 17 18'

A more conservative approach would be to make a list of allowed types and drop everything else.

Whichever way you do it, your ICMP filtering rules on a FreeS/WAN gateway should allow at least the following ICMP packet types:

echo (type 8)

echo reply (type 0)

These are used by ping(1). We recommend allowing both types through the tunnel and to or from your gateway's external interface, since ping(1) is an essential testing tool.

It is fairly common for firewalls to drop ICMP echo packets addressed to machines behind the firewall. If that is your policy, please create an exception for such packets arriving via an IPsec tunnel, at least during intial testing of those tunnels.

destination unreachable (type 3)

This is used, with code 4 (Fragmentation Needed and Don't Fragment was Set) in the code field, to control path MTU discovery. Since IPsec processing adds headers, enlarges packets and may cause fragmentation, an IPsec gateway should be able to send and receive these ICMP messages on both inside and outside interfaces.

UDP packets for traceroute

The traceroute(1) utility uses UDP port numbers from 33434 to approximately 33633. Generally, these should be allowed through for troubleshooting.

Some firewalls drop these packets to prevent outsiders exploring the protected network with traceroute(1). If that is your policy, consider creating an exception for such packets arriving via an IPsec tunnel, at least during intial testing of those tunnels.

UDP for L2TP

Windows 2000 does, and products designed for compatibility with it may, build L2TP tunnels over IPsec connections.

For this to work, you must allow UDP protocol 1701 packets coming out of your tunnels to continue to their destination. You can, and probably should, block such packets to or from your external interfaces, but allow them from ipsec0.

See also our Windows 2000 interoperation discussion.

How it all works: IPsec packet details

IPsec uses three main types of packet:

IKE uses the UDP protocol and port 500 .

Unless you are using only (less secure, not recommended) manual keying, you need IKE to negotiate connection parameters, acceptable algorithms, key sizes and key setup. IKE handles everything required to set up, rekey, repair or tear down IPsec connections.

ESP is protocol number 50

This is required for encrypted connections.

AH is protocol number 51

This can be used where only authentication, not encryption, is required.

All of those packets should have appropriate IPsec gateway addresses in both the to and from IP header fields. Firewall rules can check this if you wish, though it is not strictly necessary. This is discussed in more detail later.

IPsec processing of incoming packets authenticates them then removes the ESP or AH header and decrypts if necessary. Successful processing exposes an inner packet which is then delivered back to the firewall machinery, marked as having arrived on an ipsec[0-3] interface. Firewall rules can use that interface label to distinguish these packets from unencrypted packets which are labelled with the physical interface they arrived on (or perhaps with a non-IPsec virtual interface such as ppp0).

One of our users sent a mailing list message with a diagram of the packet flow.

ESP and AH do not have ports

Some protocols, such as TCP and UDP, have the notion of ports. Others protocols, including ESP and AH, do not. Quite a few IPsec newcomers have become confused on this point. There are no ports in the ESP or AH protocols, and no ports used for them. For these protocols, the idea of ports is completely irrelevant.

Header layout

The protocol numbers for ESP or AH are used in the 'next header' field of the IP header. On most non-IPsec packets, that field would have one of:

• 1 for ICMP
• 4 for IP-in-IP encapsulation
• 6 for TCP
• 17 for UDP
• ... or one of about 100 other possibilities listed by IANA

Each header in the sequence tells what the next header will be. IPsec adds headers for ESP or AH near the beginning of the sequence. The original headers are kept and the 'next header' fields adjusted so that all headers can be correctly interpreted.

For example, using [ ] to indicate data protected by ESP and unintelligible to an eavesdropper between the gateways:

• a simple packet might have only IP and TCP headers with:
  • IP header says next header --> TCP
  • TCP header port number --> which process to send data to
  • data
• with ESP transport mode encapsulation, that packet would have:
  • IP header says next header --> ESP
  • ESP header [ says next --> TCP
  • TCP header port number --> which process to send data to
  • data ]
  Note that the IP header is outside ESP protection, visible to an attacker, and that the final destination must be the gateway.
• with ESP in tunnel mode, we might have:
  • IP header says next header --> ESP
  • ESP header [ says next --> IP
  • IP header says next header --> TCP
  • TCP header port number --> which process to send data to
  • data ]
  Here the inner IP header is protected by ESP, unreadable by an attacker. Also, the inner header can have a different IP address than the outer IP header, so the decrypted packet can be routed from the IPsec gateway to a final destination which may be another machine.

Part of the ESP header itself is encrypted, which is why the [ indicating protected data appears in the middle of some lines above. The next header field of the ESP header is protected. This makes traffic analysis more difficult. The next header field would tell an eavesdropper whether your packet was UDP to the gateway, TCP to the gateway, or encapsulated IP. It is better not to give this information away. A clever attacker may deduce some of it from the pattern of packet sizes and timings, but we need not make it easy.

IPsec allows various combinations of these to match local policies, including combinations that use both AH and ESP headers or that nest multiple copies of these headers.

For example, suppose my employer has an IPsec VPN running between two offices so all packets travelling between the gateways for those offices are encrypted. If gateway policies allow it (The admins could block UDP 500 and protocols 50 and 51 to disallow it), I can build an IPsec tunnel from my desktop to a machine in some remote office. Those packets will have one ESP header throughout their life, for my end-to-end tunnel. For part of the route, however, they will also have another ESP layer for the corporate VPN's encapsulation. The whole header scheme for a packet on the Internet might be:

• IP header (with gateway address) says next header --> ESP
• ESP header [ says next --> IP
• IP header (with receiving machine address) says next header --> ESP
• ESP header [ says next --> TCP
• TCP header port number --> which process to send data to
• data ]]

The first ESP (outermost) header is for the corporate VPN. The inner ESP header is for the secure machine-to-machine link.

DHR on the updown script

Here are some mailing list comments from pluto(8) developer Hugh Redelmeier on an earlier draft of this document:

There are many important things left out

- firewalling is important but must reflect (implement) policy.  Since
  policy isn't the same for all our customers, and we're not experts,
  we should concentrate on FW and MASQ interactions with FreeS/WAN.

- we need a diagram to show packet flow WITHIN ONE MACHINE, assuming
  IKE, IPsec, FW, and MASQ are all done on that machine.  The flow is
  obvious if the components are run on different machines (trace the
  cables).

  IKE input:
        + packet appears on public IF, as UDP port 500
        + input firewalling rules are applied (may discard)
        + Pluto sees the packet.

  IKE output:
        + Pluto generates the packet & writes to public IF, UDP port 500
        + output firewalling rules are applied (may discard)
        + packet sent out public IF

  IPsec input, with encapsulated packet, outer destination of this host:
        + packet appears on public IF, protocol 50 or 51.  If this
          packet is the result of decapsulation, it will appear
          instead on the paired ipsec IF.
        + input firewalling rules are applied (but packet is opaque)
        + KLIPS decapsulates it, writes result to paired ipsec IF
        + input firewalling rules are applied to resulting packet
          as input on ipsec IF
        + if the destination of the packet is this machine, the
          packet is passed on to the appropriate protocol handler.
          If the original packet was encapsulated more than once
          and the new outer destination is this machine, that
          handler will be KLIPS.
        + otherwise:
          * routing is done for the resulting packet.  This may well
            direct it into KLIPS for encoding or encrypting.  What
            happens then is described elsewhere.
          * forwarding firewalling rules are applied
          * output firewalling rules are applied
          * the packet is sent where routing specified

 IPsec input, with encapsulated packet, outer destination of another host:
        + packet appears on some IF, protocol 50 or 51
        + input firewalling rules are applied (but packet is opaque)
        + routing selects where to send the packet
        + forwarding firewalling rules are applied (but packet is opaque)
        + packet forwarded, still encapsulated

  IPsec output, from this host or from a client:
        + if from a client, input firewalling rules are applied as the
          packet arrives on the private IF
        + routing directs the packet to an ipsec IF (this is how the
          system decides KLIPS processing is required)
        + if from a client, forwarding firewalling rules are applied
        + KLIPS eroute mechanism matches the source and destination
          to registered eroutes, yielding a SPI group.  This dictates
          processing, and where the resulting packet is to be sent
          (the destinations SG and the nexthop).
        + output firewalling is not applied to the resulting
          encapsulated packet

- Until quite recently, KLIPS would double encapsulate packets that
  didn't strictly need to be.  Firewalling should be prepared for
  those packets showing up as ESP and AH protocol input packets on
  an ipsec IF.

- MASQ processing seems to be done as if it were part of the
  forwarding firewall processing (this should be verified).

- If a firewall is being used, it is likely the case that it needs to
  be adjusted whenever IPsec SAs are added or removed.  Pluto invokes
  a script to do this (and to adjust routing) at suitable times.  The
  default script is only suitable for ipfwadm-managed firewalls.  Under
  LINUX 2.2.x kernels, ipchains can be managed by ipfwadm (emulation),
  but ipchains more powerful if manipulated using the ipchains command.
  In this case, a custom updown script must be used.

  We think that the flexibility of ipchains precludes us supplying an
  updown script that would be widely appropriate.

Linux FreeS/WAN Troubleshooting Guide

Overview

This document covers several general places where you might have a problem:

1. During install.
2. During the negotiation process.
3. Using an established connection.

This document also contains notes which expand on points made in these sections, and tips for problem reporting. If the other end of your connection is not FreeS/WAN, you'll also want to read our interoperation document.

1. During Install

1.1 RPM install gotchas

With the RPM method:

• Be sure you have installed both the userland tools and the kernel components. One will not work without the other. For example, when using FreeS/WAN-produced RPMs for our 1.98 release, you need both:

      freeswan-1.98_2.4.18_3custom-0.i386.rpm
      freeswan-module-1.98_2.4.18_3custom-0.i386.rpm
  

1.2 Problems installing from source

When installing from source, you may find these problems:

• Missing library. See this FAQ.
• Missing utilities required for compile. See this checklist.
• Kernel version incompatibility. See this FAQ.
• Another compile problem. Find information in the out.* files, ie. out.kpatch, out.kbuild, created at compile time in the top-level Linux FreeS/WAN directory. Error messages generated by KLIPS during the boot sequence are accessible with the dmesg command.
  Check the list archives and the List in Brief to see if this is a known issue. If it is not, report it to the bugs list as described in our problem reporting section. In some cases, you may be asked to provide debugging information using gdb; details below.
• If your kernel compiles but you fail to install your new FreeS/WAN-enabled kernel, review the sections on installing the patched kernel, and testing to see if install succeeded.

1.3 Install checks

ipsec verify checks a number of FreeS/WAN essentials. Here are some hints on what do to when your system doesn't check out:

service ipsec start

2. During Negotiation

When you fail to bring up a tunnel, you'll need to find out:

• what your connection state is, and often
• an error message.

before you can diagnose your problem .

2.1 Determine Connection State

Finding current state

You can see connection states (STATE_MAIN_I1 and so on) when you bring up a connection on the command line. If you have missed this, or brought up your connection automatically, use:

ipsec auto --status

The most relevant state is the last one reached.

What's this supposed to look like?

Negotiations should proceed though various states, in the processes of:

1. IKE negotiations (aka Phase 1, Main Mode, STATE_MAIN_*)
2. IPSEC negotiations (aka Phase 2, Quick Mode, STATE_QUICK_*)

These are done and a connection is established when you see messages like:

    000 #21: "myconn" STATE_MAIN_I4 (ISAKMP SA established)...
    000 #2: "myconn" STATE_QUICK_I2 (sent QI2, IPsec SA established)...

Look for the key phrases are "ISAKMP SA established" and "IPSec SA established", with the relevant connection name. Often, this happens at STATE_MAIN_I4 and STATE_QUICK_I2, respectively.

ipsec auto --status will tell you what states have been achieved, rather than the current state. Since determining the current state is rather more difficult to do, current state information is not available from Linux FreeS/WAN. If you are actively bringing a connection up, the status report's last states for that connection likely reflect its current state. Beware, though, of the case where a connection was correctly brought up but is now downed: Linux FreeS/WAN will not notice this until it attempts to rekey. Meanwhile, the last known state indicates that the connection has been established.

If your connection is stuck at STATE_MAIN_I1, skip straight to here.

2.2 Finding error text

Solving most errors will require you to find verbose error text, either on the command line or in the logs.

Verbose start for more information

Note that you can get more detail from ipsec auto using the --verbose flag:

    ipsec auto --verbose --up west-east

More complete information can be gleaned from the log files.

Debug levels count

The amount of description you'll get here depends on ipsec.conf debug settings, klipsdebug= and plutodebug=. When troubleshooting, set at least one of these to all, and when done, reset it to none so your logs don't fill up. Note that you must have enabled the klipsdebug compile-time option for the klipsdebug configuration switch to work.

For negotiation problems plutodebug is most relevant. klipsdebug applies mainly to attempts to use an already-established connection. See also this description of the division of duties within Linux FreeS/WAN.

After raising your debug levels, restart Linux FreeS/WAN to ensure that ipsec.conf is reread, then recreate the error to generate verbose logs.

ipsec barf for lots of debugging information

ipsec barf (8) collects a bunch of useful debugging information, including these logs Use the command

    ipsec barf > barf.west

to generate one.

Find the error

Search out the failure point in your logs. Are there a handful of lines which succinctly describe how things are going wrong or contrary to your expectation? Sometimes the failure point is not immediately obvious: Linux FreeS/WAN's errors are usually not marked "Error". Have a look in the FAQ for what some common failures look like.

Tip: problems snowball. Focus your efforts on the first problem, which is likely to be the cause of later errors.

Play both sides

Also find error text on the peer IPSec box. This gives you two perspectives on the same failure.

At times you will require information which only one side has. The peer can merely indicate the presence of an error, and its approximate point in the negotiations. If one side keeps retrying, it may be because there is a show stopper on the other side. Have a look at the other side and figure out what it doesn't like.

If the other end is not Linux FreeS/WAN, the principle is the same: replicate the error with its most verbose logging on, and capture the output to a file.

2.3 Interpreting a Negotiation Error

Connection stuck at STATE_MAIN_I1

This error commonly happens because IKE (port 500) packets, needed to negotiate an IPSec connection, cannot travel freely between your IPSec gateways. See our firewall document for details.

Other errors

Other errors require a bit more digging. Use the following resources:

• the FAQ . Since this document is constantly updated, the snapshot's FAQ may have a new entry relevant to your problem.
• our background document . Special considerations which, while not central to Linux FreeS/WAN, are often tripped over. Includes problems with packet fragmentation, and considerations for testing opportunism.
• the list archives. Each of the searchable archives works differently, so it's worth checking each. Use a search term which is generic, but identifies your error, for example "No connection is known for".
  Often, you will find that your question has been answered in the past. Finding an archived answer is quicker than asking the list. You may, however, find similar questions without answers. If you do, send their URLs to the list with your trouble report. The additional examples may help the list tech support person find your answer.
• Look into the code where the error is being generated. The pluto code is nicely documented with comments and meaningful variable names.

If you have failed to solve your problem with the help of these resources, send a detailed problem report to the users list, following these guidelines.

3. Using a Connection

3.1 Orienting yourself

How do I know if it works?

Test your connection by sending packets through it. The simplest way to do this is with ping, but the ping needs to test the correct tunnel. See this example scenario if you don't understand this.

If your ping returns, test any other connections you've brought u all check out, great. You may wish to test with large packets for MTU problems.

ipsec barf is useful again

If your ping fails to return, generate an ipsec barf debugging report on each IPSec gateway. On a non-Linux FreeS/WAN implementation, gather equivalent information. Use this, and the tips in the next sections, to troubleshoot. Are you sure that both endpoints are capable of hearing and responding to ping?

3.2 Those pesky configuration errors

IPSec may be dropping your ping packets since they do not belong in the tunnels you have constructed:

• Your ping may not test the tunnel you intend to test. For details, see our "I can't ping" FAQ.
• Alternately, you may have a configuration error. For example, you may have configured one of the four possible tunnels between two gateways, but not the one required to secure the important traffic you're now testing. In this case, add and start the tunnel, and try again.

In either case, you will often see a message like:

klipsdebug... no eroute

which we discuss in this FAQ.

Note:

• NAT and masquerading may have an effect on which tunnels you need to configure.
• When testing a tunnel that protects a multi-node subnet, try several subnet nodes as ping targets, in case one node is routing incorrectly.

3.3 Check Routing and Firewalling

If you've confirmed your configuration assumptions, the problem is almost certainly with routing or firewalling. Isolate the problem using interface statistics, firewall statistics, or a packet sniffer.

Background:

• Linux FreeS/WAN supplies all the special routing it needs; you need only route packets out through your IPSec gateway. Verify that on the subnetted machines you are using for your ping-test, your routing is as expected. I have seen a tunnel "fail" because the subnet machine sending packets out an alternate gateway (not our IPSec gateway) on their return path.
• Linux FreeS/WAN requires particular firewalling considerations. Check the firewall rules on your IPSec gateways and ensure that they allow IPSec traffic through. Be sure that no other machine - for example a router between the gateways - is blocking your IPSec packets.

View Interface and Firewall Statistics

Interface reports and firewall statistics can help you track down lost packets at a glance.

Check any firewall statistics you may be keeping on your IPSec gateways, for dropped packets.

Both cat /proc/net/dev and ifconfig display interface statistics, and both are included in ipsec barf. Use either to check if any interface has dropped packets. If you find that one has, test whether this is related to your ping. While you ping continuously, print that interface's statistics several times. Does its drop count increase in proportion to the ping? If so, check why the packets are dropped there.

To do this, look at the firewall rules that apply to that interface. If the interface is an IPSec interface, more information may be available in the log. Grep for the word "drop" in a log which was created with klipsdebug=all as the error happened.

See also this discussion on interpreting ifconfig statistics.

3.4 When in doubt, sniff it out

If you have checked configuration assumptions, routing, and firewall rules, and your interface statistics yield no clue, it remains for you to investigate the mystery of the lost packet by the most thorough method: with a packet sniffer (providing, of course, that this is legal where you are working).

In order to detect packets on the ipsec virtual interfaces, you will need an up-to-date sniffer (tcpdump, ethereal, ksnuffle) on your IPSec gateway machines. You may also find it useful to sniff the ping endpoints.

Anticipate your packets' path

Ping, and examine each interface along the projected path, checking for your ping's arrival. If it doesn't get to the the next stop, you have narrowed down where to look for it. In this way, you can isolate a problem area, and narrow your troubleshooting focus.

Within a machine running Linux FreeS/WAN, this packet flow diagram will help you anticipate a packet's path.

Note that:

• from the perspective of the tunneled packet, the entire tunnel is one hop. That's explained in this FAQ.
• an encapsulated IPSec packet will look different, when sniffed, from the plaintext packet which generated it. You can see plaintext packets entering an IPSec interface and the resulting cyphertext packets as they emerge from the corresponding physical interface.

Once you isolate where the packet is lost, take a closer look at firewall rules, routing and configuration assumptions as they affect that specific area. If the packet is lost on an IPSec gateway, comb through klipsdebug output for anomalies.

If the packet goes through both gateways successfully and reaches the ping target, but does not return, suspect routing. Check that the ping target routes packets back to the IPSec gateway.

3.5 Check your logs

Here, too, log information can be useful. Start with the guidelines above.

For connection use problems, set klipsdebug=all. Note that you must have enabled the klipsdebug compile-time option to do this. Restart Linux FreeS/WAN so that it rereads ipsec.conf, then recreate the error condition. When searching through klipsdebug data, look especially for the keywords "drop" (as in dropped packets) and "error".

Often the problem with connection use is not software error, but rather that the software is behaving contrary to expectation.

Interpreting log text

To interpret the Linux FreeS/WAN log text you've found, use the same resources as indicated for troubleshooting connection negotiation: the FAQ , our background document, and the list archives. Looking in the KLIPS code is only for the very brave.

If you are still stuck, send a detailed problem report to the users' list.

3.6 More testing for the truly thorough

Large Packets

If each of your connections passed the ping test, you may wish to test by pinging with large packets (2000 bytes or larger). If it does not return, suspect MTU issues, and see this discussion.

Stress Tests

In most users' view, a simple ping test, and perhaps a large-packet ping test suffice to indicate a working IPSec connection.

Some people might like to do additional stress tests prior to production use. They may be interested in this testing protocol we use at interoperation conferences, aka "bakeoffs". We also have a testing directory that ships with the release.

4. Problem Reporting

4.1 How to ask for help

Ask for troubleshooting help on the users' mailing list, users@lists.freeswan.org. While sometimes an initial query with a quick description of your intent and error will twig someone's memory of a similar problem, it's often necessary to send a second mail with a complete problem report.

The essay How to Report Bugs Effectively contains good guidelines.

When reporting problems to the mailing list(s), please include:

• a brief description of the problem
• if it's a compile problem, the actual output from make, showing the problem. Try to edit it down to only the relevant part, but when in doubt, be as complete as you can. If it's a kernel compile problem, any relevant out.* files
• if it's a run-time problem, pointers to where we can find the complete output from "ipsec barf" from BOTH ENDS (not just one of them). Remember that it's common outside the US and Canada to pay for download volume, so if you can't post barfs on the web and send the URL to the mailing list, at least compress them with tar or gzip.
  If you can, try to simplify the case that is causing the problem. In particular, if you clear your logs, start FreeS/WAN with no other connections running, cause the problem to happen, and then do ipsec barf on both ends immediately, that gives the smallest and least cluttered output.
• any other error messages, complaints, etc. that you saw. Please send the complete text of the messages, not just a summary.
• what your network setup is. Include subnets, gateway addresses, etc. A schematic diagram is a good format for this information.
• exactly what you were trying to do with Linux FreeS/WAN, and exactly what went wrong
• a fix, if you have one. But remember, you are sending mail to people all over the world; US residents and US citizens in particular, please read doc/exportlaws.html before sending code -- even small bug fixes -- to the list or to us.
• When in doubt about whether to include some seemingly-trivial item of information, include it. It is rare for problem reports to have too much information, and common for them to have too little.

4.2 Where to ask

To report a problem, send mail about it to the users' list. If you are certain that you have found a bug, report it to the bugs list. If you encounter a problem while doing your own coding on the Linux FreeS/WAN codebase and think it is of interest to the design team, notify the design list. When in doubt, default to the users' list. More information about the mailing lists is found here.

For a number of reasons -- including export-control regulations affecting almost any private discussion of encryption software -- we prefer that problem reports and discussions go to the lists, not directly to the team. Beware that the list goes worldwide; US citizens, read this important information about your export laws. If you're using this software, you really should be on the lists. To get onto them, visit lists.freeswan.org.

If you do send private mail to our coders or want a private reply from them, please make sure that the return address on your mail (From or Reply-To header) is a valid one. They have more important things to do than to unravel addresses that have been mangled in an attempt to confuse spammers.

5. Additional Notes on Troubleshooting

The following sections supplement the Guide: information available on your system; testing between security gateways; ifconfig reports for KLIPS debugging; using GDB on Pluto.

5.1 Information available on your system

Logs used

Linux FreeS/WAN logs to:

• /var/log/secure (or, on Debian, /var/log/auth.log)
• /var/log/messages

Check both places to get full information. If you find nothing, check your syslogd.conf(5) to see where your /etc/syslog.conf or equivalent is directing authpriv messages.

man pages provided

ipsec.conf(5)

Manual page for IPSEC configuration file.

ipsec(8)

Primary man page for ipsec utilities.

Other man pages are on this list and in

• /usr/local/man/man3
• /usr/local/man/man5
• /usr/local/man/man8/ipsec_*

Status information

ipsec auto --status

Command to get status report from running system. Displays Pluto's state. Includes the list of connections which are currently "added" to Pluto's internal database; lists state objects reflecting ISAKMP and IPsec SAs being negotiated or installed.

ipsec look

Brief status info.

ipsec barf

Copious debugging info.

5.2 Testing between security gateways

Sometimes you need to test a subnet-subnet tunnel. This is a tunnel between two security gateways, which protects traffic on behalf of the subnets behind these gateways. On this network:

     Sunset==========West------------------East=========Sunrise
                     IPSec gateway         IPSec gateway
           local net       untrusted net       local net

you might name this tunnel sunset-sunrise. You can test this tunnel by having a machine behind one gateway ping a machine behind the other gateway, but this is not always convenient or even possible.

Simply pinging one gateway from the other is not useful. Such a ping does not normally go through the tunnel. The tunnel handles traffic between the two protected subnets, not between the gateways . Depending on the routing in place, a ping might

• either succeed by finding an unencrypted route
• or fail by finding no route. Packets without an IPSEC eroute are discarded.

Neither event tells you anything about the tunnel. You can explicitly create an eroute to force such packets through the tunnel, or you can create additional tunnels as described in our configuration document, but those may be unnecessary complications in your situation.

The trick is to explicitly test between both gateways' private-side IP addresses. Since the private-side interfaces are on the protected subnets, the resulting packets do go via the tunnel. Use either ping -I or traceroute -i, both of which allow you to specify a source interface. (Note: unsupported on older Linuxes). The same principles apply for a road warrior (or other) case where only one end of your tunnel is a subnet.

5.3 ifconfig reports for KLIPS debugging

When diagnosing problems using ifconfig statistics, you may wonder what type of activity increments a particular counter for an ipsecN device. Here's an index, posted by KLIPS developer Richard Guy Briggs:

Here is a catalogue of the types of errors that can occur for which
statistics are kept when transmitting and receiving packets via klips.
I notice that they are not necessarily logged in the right counter.
. . .

Sources of ifconfig statistics for ipsec devices

rx-errors:
- packet handed to ipsec_rcv that is not an ipsec packet.
- ipsec packet with payload length not modulo 4.
- ipsec packet with bad authenticator length.
- incoming packet with no SA.
- replayed packet.
- incoming authentication failed.
- got esp packet with length not modulo 8.

tx_dropped:
- cannot process ip_options.
- packet ttl expired.
- packet with no eroute.
- eroute with no SA.
- cannot allocate sk_buff.
- cannot allocate kernel memory.
- sk_buff internal error.


The standard counters are:

struct enet_statistics
{
        int        rx_packets;                /* total packets received */
        int        tx_packets;                /* total packets transmitted */
        int        rx_errors;                /* bad packets received */
        int        tx_errors;                /* packet transmit problems */
        int        rx_dropped;                /* no space in linux buffers */
        int        tx_dropped;                /* no space available in linux */
        int        multicast;                /* multicast packets received */
        int        collisions;

        /* detailed rx_errors: */
        int        rx_length_errors;
        int        rx_over_errors;                /* receiver ring buff overflow */
        int        rx_crc_errors;                /* recved pkt with crc error */
        int        rx_frame_errors;        /* recv'd frame alignment error */
        int        rx_fifo_errors;                /* recv'r fifo overrun */
        int        rx_missed_errors;        /* receiver missed packet */

        /* detailed tx_errors */
        int        tx_aborted_errors;
        int        tx_carrier_errors;
        int        tx_fifo_errors;
        int        tx_heartbeat_errors;
        int        tx_window_errors;
};

of which I think only the first 6 are useful.

5.4 Using GDB on Pluto

You may need to use the GNU debugger, gdb(1), on Pluto. This should be necessary only in unusual cases, for example if you encounter a problem which the Pluto developer cannot readily reproduce or if you are modifying Pluto.

Here are the Pluto developer's suggestions for doing this:

Can you get a core dump and use gdb to find out what Pluto was doing
when it died?

To get a core dump, you will have to set dumpdir to point to a
suitable directory (see ipsec.conf(5)).

To get gdb to tell you interesting stuff:
        $ script
        $ cd dump-directory-you-chose
        $ gdb /usr/local/lib/ipsec/pluto core
        (gdb) where
        (gdb) quit
        $ exit

The resulting output will have been captured by the script command in
a file called "typescript".  Send it to the list.

Do not delete the core file.  I may need to ask you to print out some
more relevant stuff.

Note that the dumpdir parameter takes effect only when the IPsec subsystem is restarted -- reboot or ipsec setup restart.

Linux FreeS/WAN Compatibility Guide

Much of this document is quoted directly from the Linux FreeS/WAN mailing list. Thanks very much to the community of testers, patchers and commenters there, especially the ones quoted below but also various contributors we haven't quoted.

Implemented parts of the IPsec Specification

In general, do not expect Linux FreeS/WAN to do everything yet. This is a work-in-progress and some parts of the IPsec specification are not yet implemented.

In Linux FreeS/WAN

Things we do, as of version 1.96:

• key management methods

  manually keyed

  using keys stored in /etc/ipsec.conf

  automatically keyed

  Automatically negotiating session keys as required. All connections are automatically re-keyed periodically. The Pluto daemon implements this using the IKE protocol.

• Methods of authenticating gateways for IKE

  shared secrets

  stored in ipsec.secrets(5)

  RSA signatures

  For details, see pluto(8) .

  looking up RSA authentication keys from DNS.

  Note that this technique cannot be fully secure until secure DNS is widely deployed.

• groups for Diffie-Hellman key negotiation

  group 2, modp 1024-bit

  group 5, modp 1536-bit

  We implement these two groups.

  In negotiating a keying connection (ISAKMP SA, Phase 1) we propose both groups when we are the initiator, and accept either when a peer proposes them. Once the keying connection is made, we propose only the alternative agreed there for data connections (IPsec SA's, Phase 2) negotiated over that keying connection.

• encryption transforms

  DES

  DES is in the source code since it is needed to implement 3DES, but single DES is not made available to users because DES is insecure.

  Triple DES

  implemented, and used as the default encryption in Linux FreeS/WAN.

• authentication transforms

  HMAC using MD5

  implemented, may be used in IKE or by by AH or ESP transforms.

  HMAC using SHA

  implemented, may be used in IKE or by AH or ESP transforms.

  In negotiations, we propose both of these and accept either.

• compression transforms

  IPComp

  IPComp as described in RFC 2393 was added for FreeS/WAN 1.6. Note that Pluto becomes confused if you ask it to do IPComp when the kernel cannot.

All combinations of implemented transforms are supported. Note that some form of packet-level authentication is required whenever encryption is used. Without it, the encryption will not be secure.

Deliberately omitted

Things we deliberately omit which are required in the RFCs are:

• null encryption (to use ESP as an authentication-only service)
• single DES
• DH group 1, a 768-bit modp group

Since these are the only encryption algorithms and DH group the RFCs require, it is possible in theory to have a standards-conforming implementation which will not interpoperate with FreeS/WAN. Such an implementation would be inherently insecure, so we do not consider this a problem.

Anyway, most implementations sensibly include more secure options as well, so dropping null encryption, single DES and Group 1 does not greatly hinder interoperation in practice.

We also do not implement some optional features allowed by the RFCs:

• aggressive mode for negotiation of the keying channel or ISAKMP SA. This mode is a little faster than main mode, but exposes more information to an eavesdropper.

In theory, this should cause no interoperation problems since all implementations are required to support the more secure main mode, whether or not they also allow aggressive mode.

In practice, it does sometimes produce problems with implementations such as Windows 2000 where aggressive mode is the default. Typically, these are easily solved with a configuration change that overrides that default.

Not (yet) in Linux FreeS/WAN

Things we don't yet do, as of version 1.96:

• key management methods
  • authenticate key negotiations via local PKI server, but see links to user patches
  • authenticate key negotiations via secure DNS
  • unauthenticated key management, using Diffie-Hellman key agreement protocol without authentication. Arguably, this would be worth doing since it is secure against all passive attacks. On the other hand, it is vulnerable to an active man-in-the-middle attack.
• encryption transforms

  Currently Triple DES is the only encryption method Pluto will negotiate.

  No additional encryption transforms are implemented, though the RFCs allow them and some other IPsec implementations support various of them. We are not eager to add more. See this FAQ question.

  AES, the successor to the DES standard, is an excellent candidate for inclusion in FreeS/WAN, see links to user patches.

• authentication transforms

  No optional additional authentication transforms are currently implemented. Likely SHA-256, SHA-384 and SHA-512 will be added when AES is.

• Policy checking on decrypted packets

  To fully comply with the RFCs, it is not enough just to accept only packets which survive any firewall rules in place to limit what IPsec packets get in, and then pass KLIPS authentication. That is what FreeS/WAN currently does.

  We should also apply additional tests, for example ensuring that all packets emerging from a particular tunnel have source and destination addresses that fall within the subnets defined for that tunnel, and that packets with those addresses that did not emerge from the appropriate tunnel are disallowed.

  This will be done as part of a KLIPS rewrite. See these links and the design mailing list for discussion.

Our PF-Key implementation

We use PF-key Version Two for communication between the KLIPS kernel code and the Pluto Daemon. PF-Key v2 is defined by RFC 2367.

The "PF" stands for Protocol Family. PF-Inet defines a kernel/userspace interface for the TCP/IP Internet protocols (TCP/IP), and other members of the PF series handle Netware, Appletalk, etc. PF-Key is just a PF for key-related matters.

PF-Key portability

PF-Key came out of Berkeley Unix work and is used in the various BSD IPsec implementations, and in Solaris. This means there is some hope of porting our Pluto(8) to one of the BSD distributions, or of running their photurisd(8) on Linux if you prefer Photuris key management over IKE.

It is, however, more complex than that. The PK-Key RFC deliberately deals only with keying, not policy management. The three PF-Key implementations we have looked at -- ours, OpenBSD and KAME -- all have extensions to deal with security policy, and the extensions are different. There have been discussions aimed at sorting out the differences, perhaps for a version three PF-Key spec. All players are in favour of this, but everyone involved is busy and it is not clear whether or when these discussions might bear fruit.

Kernels other than the latest 2.2.x and 2.4.y

We develop and test on Redhat Linux using the most recent kernel in the 2.2 and 2.4 series. In general, we recommend you use the latest kernel in one of those series. Complications and caveats are discussed below.

2.0.x kernels

Consider upgrading to the 2.2 kernel series. If you want to stay with the 2.0 series, then we strongly recommend 2.0.39. Some useful security patches were added in 2.0.38.

Various versions of the code have run at various times on most 2.0.xx kernels, but the current version is only lightly tested on 2.0.39, and not at all on older kernels.

Some of our patches for older kernels are shipped in 2.0.37 and later, so they are no longer provided in FreeS/WAN. This means recent versions of FreeS/WAN will probably not compile on anything earlier than 2.0.37.

2.2 and 2.4 kernels

FreeS/WAN 1.0

ran only on 2.0 kernels

FreeS/WAN 1.1 to 1.8

ran on 2.0 or 2.2 kernels
ran on some development kernels, 2.3 or 2.4-test

FreeS/WAN 1.9 to 1.96

runs on 2.0, 2.2 or 2.4 kernels

In general, we suggest the latest 2.2 kernel or 2.4 for production use.

Of course no release can be guaranteed to run on kernels more recent than it is, so quite often there will be no stable FreeS/WAN for the absolute latest kernel. See the FAQ for discussion.

Intel Linux distributions other than Redhat

We develop and test on Redhat 6.1 for 2.2 kernels, and on Redhat 7.1 or 7.2 for 2.4, so minor changes may be required for other distributions.

Redhat 7.0

There are some problems with FreeS/WAN on Redhat 7.0. They are soluble, but we recommend you upgrade to a later Redhat instead..

Redhat 7 ships with two compilers.

• Their gcc is version 2.96. Various people, including the GNU compiler developers and Linus, have said fairly emphatically that using this was a mistake. 2.96 is a development version, not intended for production use. In particular, it will not compile a Linux kernel.
• Redhat therefore also ship a separate compiler, which they call kgcc, for compiling kernels.

Kernel Makefiles have gcc as a default, and must be adjusted to use kgcc before a kernel will compile on 7.0. This mailing list message gives details:

Subject: Re: AW: Installing IPsec on Redhat 7.0
   Date: Thu, 1 Feb 2001 14:32:52 -0200 (BRST)
  From: Mads Rasmussen <mads@cit.com.br>
 
> From www.redhat.com/support/docs/gotchas/7.0/gotchas-7-6.html#ss6.1

cd to /usr/src/linux and open the Makefile in your favorite editor. You
will need to look for a line similar to this:

CC = $(CROSS_COMPILE)gcc -D__KERNEL__ -I$(HPATH)

This line specifies which C compiler to use to build the kernel. It should
be changed to:

CC = $(CROSS_COMPILE)kgcc -D__KERNEL__ -I$(HPATH)

for Red Hat Linux 7. The kgcc compiler is egcs 2.91.66. From here you can
proceed with the typical compiling steps.

Check the mailing list archive for more recent news.

SuSE Linux

SuSE 6.3 and later versions, at least in Europe, ship with FreeS/WAN included.

Here are some notes for an earlier SuSE version.

SuSE Linux 5.3

Date: Mon, 30 Nov 1998
From: Peter Onion <ponion@srd.bt.co.uk>

... I got Saturdays snapshot working between my two SUSE5.3 machines at home.

The mods to the install process are quite simple.  From memory and looking at
the files on the SUSE53 machine here at work....

And extra link in each of the /etc/init.d/rc?.d directories called K35ipsec
which SUSE use to shut a service down.

A few mods in /etc/init.d/ipsec  to cope with the different places that SUSE
put config info, and remove the inculsion of /etc/rc.d/init.d/functions and .
/etc/sysconfig/network as they don't exists and 1st one isn't needed anyway.

insert ". /etc/rc.config" to pick up the SUSE config info and use 

  if test -n "$NETCONFIG" -a "$NETCONFIG" != "YAST_ASK" ; then

to replace 

  [ ${NETWORKING} = "no" ] && exit 0

Create /etc/sysconfig  as SUSE doesn't have one.

I think that was all (but I prob forgot something)....

You may also need to fiddle initialisation scripts to ensure that /var/run/pluto.pid is removed when rebooting. If this file is present, Pluto does not come up correctly.

Slackware

Subject: Re: linux-IPsec: Slackware distribution
  Date:  Thu, 15 Apr 1999 12:07:01 -0700
  From:  Evan Brewer <dmessiah@silcon.com>

> Very shortly, I will be needing to install IPsec on at least gateways that
> are running Slackware. . . .

The only trick to getting it up is that on the slackware dist there is no
init.d directory in /etc/rc.d .. so create one.  Then, what I do is take the
IPsec startup script which normally gets put into the init.d directory, and
put it in /etc/rc.d and name ir rc.ipsec .. then I symlink it to the file
in init.d.  The only file in the dist you need to really edit is the
utils/Makefile, setup4:

Everything else should be just fine.

A year or so later:

Subject: Re: HTML Docs- Need some cleanup?
   Date: Mon, 8 Jan 2001
   From: Jody McIntyre <jodym@oeone.com>

I have successfully installed FreeS/WAN on several Slackware 7.1 machines.
FreeS/WAN installed its rc.ipsec file in /etc/rc.d.  I had to manually call
this script from rc.inet2.  This seems to be an easier method than Evan
Brewer's.

Debian

A recent (Nov 2001) mailing list points to a web page on setting up several types of tunnel, including IPsec, on Debian.

Some older information:

Subject: FreeS/WAN 1.0 on Debian 2.1
   Date: Tue, 20 Apr 1999
  From:  Tim Miller <cerebus+counterpane@haybaler.sackheads.org>

        Compiled and installed without error on a Debian 2.1 system
with kernel-source-2.0.36 after pointing RCDIR in utils/Makefile to
/etc/init.d.

        /var/lock/subsys/ doesn't exist on Debian boxen, needs to be
created; not a fatal error.

        Finally, IPsec scripts appear to be dependant on GNU awk
(gawk); the default Debian awk (mawk-1.3.3-2) had fatal difficulties.
With gawk installed and /etc/alternatives/awk linked to /usr/bin/gawk
operation appears flawless.

The scripts in question have been modified since this was posted. Awk versions should no longer be a problem.

Caldera

Subject: Re: HTML Docs- Need some cleanup?
   Date: Mon, 08 Jan 2001
   From: Andy Bradford <andyb@calderasystems.com>

On Sun, 07 Jan 2001 22:59:05 EST, Sandy Harris wrote:

>     Intel Linux distributions other than Redhat 5.x and 6.x 
>         Redhat 7.0 
>         SuSE Linux 
>             SuSE Linux 5.3 
>         Slackware 
>         Debian 

Can you please include Caldera in this list?  I have tested it since 
FreeS/Wan 1.1 and it works great with our systems---provided one 
follows the FreeS/Wan documentation. :-)

Thank you,
Andy

CPUs other than Intel

FreeS/WAN has been run sucessfully on a number of different CPU architectures. If you have tried it on one not listed here, please post to the mailing list.

Corel Netwinder (StrongARM CPU)

Subject: linux-ipsec: Netwinder diffs
Date: Wed, 06 Jan 1999
From: rhatfield@plaintree.com

I had a mistake in my IPsec-auto, so I got things working this morning.

Following are the diffs for my changes.  Probably not the best and cleanest way 
of doing it, but it works. . . . 

These diffs are in the 0.92 and later distributions, so these should work out-of-the-box on Netwinder.

Yellow Dog Linux on Power PC

Subject:  Compiling FreeS/WAN 1.1 on YellowDog Linux (PPC)
   Date:  11 Dec 1999
   From:  Darron Froese <darron@fudgehead.com>

I'm summarizing here for the record - because it's taken me many hours to do
this (multiple times) and because I want to see IPsec on more linuxes than
just x86.

Also, I can't remember if I actually did summarize it before... ;-) I'm
working too many late hours.

That said - here goes.

1. Get your linux kernel and unpack into /usr/src/linux/ - I used 2.2.13.
<http://www.kernel.org/pub/linux/kernel/v2.2/linux-2.2.13.tar.bz2>

2. Get FreeS/WAN and unpack into /usr/src/freeswan-1.1
<ftp://ftp.xs4all.nl/pub/crypto/freeswan/freeswan-1.1.tar.gz>

3. Get the gmp src rpm from here:
<ftp://ftp.yellowdoglinux.com//pub/yellowdog/champion-1.1/SRPMS/SRPMS/gmp-2.0.2-9a.src.rpm>

4. Su to root and do this: rpm --rebuild gmp-2.0.2-9a.src.rpm

You will see a lot of text fly by and when you start to see the rpm
recompiling like this:

Executing: %build
+ umask 022
+ cd /usr/src/redhat/BUILD
+ cd gmp-2.0.2
+ libtoolize --copy --force
Remember to add `AM_PROG_LIBTOOL' to `configure.in'.
You should add the contents of `/usr/share/aclocal/libtool.m4' to
`aclocal.m4'.
+ CFLAGS=-O2 -fsigned-char
+ ./configure --prefix=/usr

Hit Control-C to stop the rebuild. NOTE: We're doing this because for some
reason the gmp source provided with FreeS/WAN 1.1 won't build properly on
ydl.

cd /usr/src/redhat/BUILD/
cp -ar gmp-2.0.2 /usr/src/freeswan-1.1/
cd /usr/src/freeswan-1.1/
rm -rf gmp
mv gmp-2.0.2 gmp

5. Open the freeswan Makefile and change the line that says:
KERNEL=$(b)zimage (or something like that) to
KERNEL=vmlinux

6. cd ../linux/

7. make menuconfig
Select an option or two and then exit - saving your changes.

8. cd ../freeswan-1.1/ ; make menugo

That will start the whole process going - once that's finished compiling,
you have to install your new kernel and reboot.

That should build FreeS/WAN on ydl (I tried it on 1.1).

Subject: Re: FreeS/WAN, PGPnet and E-mail
   Date: Sat, 22 Jan 2000
   From: Darron Froese <darron@fudgehead.com>

on 1/22/00 6:47 PM, Philip Trauring at philip@trauring.com wrote:

> I have a PowerMac G3 ...

The PowerMac G3 can run YDL 1.1 just fine. It should also be able to run
FreeS/WAN 1.2patch1 with a couple minor modifications:

1. In the Makefile it specifies a bzimage for the kernel compile - you have
to change that to vmlinux for the PPC.

2. The gmp source that comes with FreeS/WAN (for whatever reason) fails to
compile. I have gotten around this by getting the gmp src rpm from here:

ftp://ftp.yellowdoglinux.com//pub/yellowdog/champion-1.1/SRPMS/SRPMS/gmp-2.0.2-9a.src.rpm

If you rip the source out of there - and place it where the gmp source
resides it will compile just fine.

FreeS/WAN no longer includes GMP source.

Mklinux

One user reports success on the Mach-based micro kernel Linux.

Subject: Smiles on sparc and ppc
   Date: Fri, 10 Mar 2000
   From: Jake Hill <jah@alien.bt.co.uk>

You may or may not be interested to know that I have successfully built
FreeS/WAN on a number of non intel alpha architectures; namely on ppc
and sparc and also on osfmach3/ppc (MkLinux). I can report that it just
works, mostly, with few changes.

Alpha 64-bit processors

Subject: IT WORKS (again) between intel & alpha :-)))))
   Date: Fri, 29 Jan 1999
   From: Peter Onion <ponion@srd.bt.co.uk>

Well I'm happy to report that I've got an IPsec connection between by intel & alpha machines again :-))

If you look back on this list to 7th of December I wrote...

-On 07-Dec-98 Peter Onion wrote:
-> 
-> I've about had enuf of wandering around inside the kernel trying to find out
-> just what is corrupting outgoing packets...
-
-Its 7:30 in the evening .....
-
-I FIXED IT  :-))))))))))))))))))))))))))))))))
-
-It was my own fault :-((((((((((((((((((
-
-If you ask me very nicly I'll tell you where I was a little too over keen to
-change unsigned long int __u32 :-)  OPSE ...
-
-So tomorrow it will full steam ahead to produce a set of diffs/patches against
-0.91 
-
-Peter Onion.

In general (there have been some glitches), FreeS/WAN has been running on Alphas since then.

Sun SPARC processors

Several users have reported success with FreeS/WAN on SPARC Linux. Here is one mailing list message:

Subject: Smiles on sparc and ppc
   Date: Fri, 10 Mar 2000
   From: Jake Hill <jah@alien.bt.co.uk>

You may or may not be interested to know that I have successfully built
FreeS/WAN on a number of non intel alpha architectures; namely on ppc
and sparc and also on osfmach3/ppc (MkLinux). I can report that it just
works, mostly, with few changes.

I have a question, before I make up some patches. I need to hack
gmp/mpn/powerpc32/*.s to build them. Is this ok? The changes are
trivial, but could I also use a different version of gmp? Is it vanilla
here?

I guess my only real headache is from ipchains, which appears to stop
running when IPsec has been started for a while. This is with 2.2.14 on
sparc.

This message, from a different mailing list, may be relevant for anyone working with FreeS/WAN on Suns:

Subject: UltraSPARC DES assembler
   Date: Thu, 13 Apr 2000
   From: svolaf@inet.uni2.dk (Svend Olaf Mikkelsen)
     To: coderpunks@toad.com

An UltraSPARC assembler version of the LibDES/SSLeay/OpenSSL des_enc.c
file is available at http://inet.uni2.dk/~svolaf/des.htm.

This brings DES on UltraSPARC from slower than Pentium at the same
clock speed to significantly faster.

MIPS processors

We know FreeS/WAN runs on at least some MIPS processors because Lasat manufacture an IPsec box based on an embedded MIPS running Linux with FreeS/WAN. We have no details.

Transmeta Crusoe

The Merilus Firecard, a Linux firewall on a PCI card, is based on a Crusoe processor and supports FreeS/WAN.

Motorola Coldfire

Subject: Re: Crypto hardware support
   Date: Mon, 03 Jul 2000
   From: Dan DeVault <devault@tampabay.rr.com>

.... I have been running
uClinux with FreeS/WAN 1.4 on a system built by Moreton Bay  (
http://www.moretonbay.com )  and it was using a Coldfire processor
and was able to do the Triple DES encryption at just about
1 mbit / sec rate.......  they put a Hi/Fn 7901 hardware encryption
chip on their board and now their system does over 25 mbit of 3DES
encryption........ pretty significant increase if you ask me.

Multiprocessor machines

FreeS/WAN is designed to work on SMP (symmetric multi-processing) Linux machines and is regularly tested on dual processor x86 machines.

We do not know of any testing on multi-processor machines with other CPU architectures or with more than two CPUs. Anyone who does test this, please report results to the mailing list .

The current design does not make particularly efficient use of multiprocessor machines; some of the kernel work is single-threaded.

Support for crypto hardware

Supporting hardware cryptography accelerators has not been a high priority for the development team because it raises a number of fairly complex issues:

• Can you trust the hardware? If it is not Open Source, how do you audit its security? Even if it is, how do you check that the design has no concealed traps?
• If an interface is added for such hardware, can that interface be subverted or misused?
• Is hardware acceleration actually a performance win? It clearly is in many cases, but on a fast machine it might be better to use the CPU for the encryption than to pay the overheads of moving data to and from a crypto board.
• the current KLIPS code does not provide a clean interface for hardware accelerators

That said, we have a report of FreeS/WAN working with one crypto accelerator and some work is going on to modify KLIPS to create a clean generic interface to such products. See this web page for some of the design discussion.

More recently, a patch to support some hardware accelerators has been posted:

Subject: [Design] [PATCH] H/W acceleration patch
   Date: Tue, 18 Sep 2001
   From: "Martin Gadbois" <martin.gadbois@colubris.com>
 
Finally!!
Here's a web site with H/W acceleration patch for FreeS/WAN 1.91, including
S/W and Hifn 7901 crypto support.

http://sources.colubris.com/

Martin Gadbois

Hardware accelerators could take performance well beyond what FreeS/WAN can do in software (discussed here ). Here is some discussion off the IETF IPsec list, October 2001:

 ... Currently shipping chips deliver, 600 mbps throughput on a single
 stream of 3DES IPsec traffic.  There are also chips that use multiple
 cores to do 2.4 gbps.  We (Cavium) and others have announced even faster
 chips. ... Mid 2002 versions will handle at line rate (OC48 and OC192)
 IPsec and SSL/TLS traffic not only 3DES CBC but also AES and arc4.

The patches to date support chips that have been in production for some time, not the state-of-the-art latest-and-greatest devices described in that post. However, they may still outperform software and they almost certainly reduce CPU overhead.

IP version 6 (IPng)

The Internet currently runs on version four of the IP protocols. IPv4 is what is in the standard Linux IP stack, and what FreeS/WAN was built for. In IPv4, IPsec is an optional feature.

The next version of the IP protocol suite is version six, usually abbreviated either as "IPv6" or as "IPng" for "IP: the next generation". For IPv6, IPsec is a required feature. Any machine doing IPv6 is required to support IPsec, much as any machine doing (any version of) IP is required to support ICMP.

There is a Linux implementation of IPv6 in Linux kernels 2.2 and above. For details, see the FAQ. It does not yet support IPsec. The USAGI project are also working on IPv6 for Linux.

FreeS/WAN was originally built for the current standard, IPv4, but we are interested in seeing it work with IPv6. Some progress has been made, and a patched version with IPv6 support is available. For more recent information, check the mailing list.

IPv6 background

IPv6 has been specified by an IETF working group. The group's page lists over 30 RFCs to date, and many Internet Drafts as well. The overview is RFC 2460. Major features include:

• expansion of the address space from 32 to 128 bits,
• changes to improve support for
  • mobile IP
  • automatic network configuration
  • quality of service routing
  • ...
• improved security via IPsec

A number of projects are working on IPv6 implementation. A prominent Open Source effort is KAME, a collaboration among several large Japanese companies to implement IPv6 for Berkeley Unix. Other major players are also working on IPv6. For example, see pages at:

• Sun
• Cisco
• Microsoft

The 6bone (IPv6 backbone) testbed network has been up for some time. There is an active IPv6 user group.

One of the design goals for IPv6 was that it must be possible to convert from v4 to v6 via a gradual transition process. Imagine the mess if there were a "flag day" after which the entire Internet used v6, and all software designed for v4 stopped working. Almost every computer on the planet would need major software changes! There would be huge costs to replace older equipment. Implementers would be worked to death before "the day", systems administrators and technical support would be completely swamped after it. The bugs in every implementation would all bite simultaneously. Large chunks of the net would almost certainly be down for substantial time periods. ...

Fortunately, the design avoids any "flag day". It is therefore a little tricky to tell how quickly IPv6 will take over. The transition has certainly begun. For examples, see announcements from NTT and Nokia. However, it is not yet clear how quickly the process will gain momentum, or when it will be completed. Likely large parts of the Internet will remain with IPv4 for years to come.

Interoperating with FreeS/WAN

Interop at a Glance

Our information comes primarily from mailing list reports and tutorials.

The FreeS/WAN project needs you! We rely on the user community to keep up to date. Mail users@lists.freeswan.org with your interop success stories.

Key

Basic Interop Rules

Longer Stories

For More Compatible Implementations

isakmpd (OpenBSD)

OpenBSD FAQ: Using IPsec
Hans-Joerg Hoexer's interop Linux-OpenBSD (PSK)
Skyper's configuration (PSK)

Kame for FreeBSD, NetBSD

Kame homepage, with FAQ
NetBSD's IPSec FAQ

Itojun's Kame-FreeS/WAN interop tips (PSK)
Ghislaine Labouret's French page with links to matching FreeS/WAN and Kame configs (RSA)
     Ghislaine's post explaining some peculiarities
Frodo's Kame-FreeS/WAN interop (X.509)
Using Kame as a WAVEsec client

PGPNet/McAfee

• Now called McAfee VPN Client.
• PGPNet also came in a freeware version which did not support subnets
• To support dhcp-over-ipsec, you need the X.509 patch, which is included in Super FreeS/WAN.

Rekeying problem with FreeS/WAN and older PGPNets

DHCP over IPSEC HOWTO for FreeS/WAN (requires X.509 and dhcprelay patches)

Microsoft Windows 2000/XP

• Comes with Win2k, and with XP Support Tools. May require High Encryption Pack.
• FreeS/WAN has recently changed its Commit Bit handling, solving an old interop problem.

Jean-Francois Nadeau's Net-net Configuration (PSK)
Telenor's Node-node Config (Transport-mode PSK)
Marcus Mueller's HOWTO using his VPN config tool (X.509). Tool also works with PSK.
Nate Carlson's HOWTO using same tool (Road Warrior with X.509). Unusually, FreeS/WAN is the Road Warrior here.
Oscar Delgado's PDF (X.509, no configs)

Microsoft's Win2k IPsec debugging tips
MS VPN may fall back to 1DES

Safenet SoftPK/SoftRemote

• People recommend SafeNet as a low cost Windows client.
• SoftRemote seems to be the newer name for SoftPK.

Whit Blauvelt's SoftRemote tips

Jean-Francois Nadeau's Practical Configuration (Road Warrior with PSK)
Terradon Communications' PDF (Road Warrior with PSK)
Red Baron Consulting's PDF (Road Warrior with X.509)

SSH Sentinel

• New, popular and well tested.

SSH's Sentinel-FreeSWAN interop PDF (X.509)
Potential problem unless using Legacy Proposal option

For Other Implementations

AshleyLaurent VPCom

Successful interop report, no details

Borderware

• I suspect the Borderware client is a rebranded Safenet. If that's true, our Safenet section will help.

Philip Reetz' configs (PSK)
Borderware server does not support FreeS/WAN road warriors
Older Borderware may not support Diffie Hellman groups 2, 5

Checkpoint VPN-1 or FW-1

Text goes here.

Cisco

Text goes here.

F-Secure

Text goes here.

Gauntlet GVPN

Text goes here.

IBM AS/400

Richard Welty's tips and tricks

Lucent

Text goes here.

Netscreen

Errol Neal's settings

Nortel Conitivity

Text goes here.

Radguard

Text goes here.

Raptor (NT)

Text goes here.

Raptor (Solaris)

Text goes here.

Redcreek Ravlin

Text goes here.

Shiva LANRover

Text goes here.

Sun Solaris

Text goes here.

SonicWall

Text goes here.

Timestep

Text goes here.

Watchguard Firebox

WatchGuard's HOWTO (PSK)
Ronald C. Riviera's Settings (PSK)

Old known issue with auto keying
Tips on key generation and format (Manual)

Xedia Access Point/QVPN

Hybrid IPsec/L2TP connection settings (X.509)
Xedia's LAN-LAN links don't use multiple tunnels
     That explanation, continued

Performance of FreeS/WAN

In normal operation, the main concern is the overhead for encryption, decryption and authentication of the actual IPsec (ESP and/or AH) data packets. Tunnel setup and rekeying occur so much less frequently than packet processing that, in general, their overheads are not worth worrying about.

At startup, however, tunnel setup overheads may be significant. If you reboot a gateway and it needs to establish many tunnels, expect some delay. This and other issues for large gateways are discussed below.

Published material

The University of Wales at Aberystwyth has done quite detailed speed tests and put their results on the web.

Davide Cerri's thesis (in Italian) includes performance results for FreeS/WAN and for TLS. He posted an English summary on the mailing list.

Steve Bellovin used one of AT&T Research's FreeS/WAN gateways as his data source for an analysis of the cache sizes required for key swapping in IPsec. Available as text or PDF slides for a talk on the topic.

See also the NAI work mentioned in the next section.

Estimating CPU overheads

We can come up with a formula that roughly relates CPU speed to the rate of IPsec processing possible. It is far from exact, but should be usable as a first approximation.

An analysis of authentication overheads for high-speed networks, including some tests using FreeS/WAN, is on the NAI Labs site. In particular, see figure 3 in this PDF document. Their estimates of overheads, measured in Pentium II cycles per byte processed are:

Overheads for IPsec with encryption were not tested in the NAI work, but Antoon Bosselaers' web page gives cost for his optimised Triple DES implementation as 928 Pentium cycles per block, or 116 per byte. Adding that to the 24 above, we get 140 cycles per byte for IPsec with encryption.

At 140 cycles per byte, a 140 MHz machine can handle a megabyte -- 8 megabits -- per second. Speeds for other machines will be proportional to this. To saturate a link with capacity C megabits per second, you need a machine running at C * 140/8 = C * 17.5 MHz.

However, that estimate is not precise. It ignores the differences between:

• NAI's test packets and real traffic
• NAI's Pentium II cycles, Bosselaers' Pentium cycles, and your machine's cycles
• different 3DES implementations
• SHA-1 and MD5

and does not account for some overheads you will almost certainly have:

• communication on the client-side interface
• switching between multiple tunnels -- re-keying, cache reloading and so on

so we suggest using C * 25 to get an estimate with a bit of a built-in safety factor.

This covers only IP and IPsec processing. If you have other loads on your gateway -- for example if it is also working as a firewall -- then you will need to add your own safety factor atop that.

This estimate matches empirical data reasonably well. For example, Metheringham's tests, described below, show a 733 topping out between 32 and 36 Mbit/second, pushing data as fast as it can down a 100 Mbit link. Our formula suggests you need at least an 800 to handle a fully loaded 32 Mbit link. The two results are consistent.

Some examples using this estimation method:

Such an estimate is far from exact, but should be usable as minimum requirement for planning. The key observations are:

• older surplus machines are fine for IPsec gateways at loads up to 5 megabits per second or so
• a mid-range new machine can handle IPsec at rates up to 20 megabits per second or more

Higher performance alternatives

AES is a new US government block cipher standard, designed to replace the obsolete DES. If FreeS/WAN using 3DES is not fast enough for your application, the AES patch may help.

To date (March 2002) we have had only one mailing list report of measurements with the patch applied. It indicates that, at least for the tested load on that user's network, AES roughly doubles IPsec throughput. If further testing confirms this, it may prove possible to saturate an OC3 link in software on a high-end box.

Also, some work is being done toward support of hardware IPsec acceleration which might extend the range of requirements FreeS/WAN could meet.

Other considerations

CPU speed may be the main issue for IPsec performance, but of course it isn't the only one.

You need good ethernet cards or other network interface hardware to get the best performance. See this ethernet information page and this Linux network driver page.

The current FreeS/WAN kernel code is largely single-threaded. It is SMP safe, and will run just fine on a multiprocessor machine ( discussion), but the load within the kernel is not shared effectively. This means that, for example to saturate a T3 -- which needs about a 1200 MHz machine -- you cannot expect something like a dual 800 to do the job.

On the other hand, SMP machines do tend to share loads well so -- provided one CPU is fast enough for the IPsec work -- a multiprocessor machine may be ideal for a gateway with a mixed load.

Many tunnels from a single gateway

FreeS/WAN allows a single gateway machine to build tunnels to many others. There may, however, be some problems for large numbers as indicated in this message from the mailing list:

Subject: Re: Maximum number of ipsec tunnels?
   Date: Tue, 18 Apr 2000
   From: "John S. Denker" <jsd@research.att.com>

Christopher Ferris wrote:

>> What are the maximum number ipsec tunnels FreeS/WAN can handle??

Henry Spencer wrote:

>There is no particular limit.  Some of the setup procedures currently
>scale poorly to large numbers of connections, but there are (clumsy)
>workarounds for that now, and proper fixes are coming.

1) "Large" numbers means anything over 50 or so.  I routinely run boxes
with about 200 tunnels.  Once you get more than 50 or so, you need to worry
about several scalability issues:

a) You need to put a "-" sign in syslogd.conf, and rotate the logs daily
not weekly.

b) Processor load per tunnel is small unless the tunnel is not up, in which
case a new half-key gets generated every 90 seconds, which can add up if
you've got a lot of down tunnels.

c) There's other bits of lore you need when running a large number of
tunnels.  For instance, systematically keeping the .conf file free of
conflicts requires tools that aren't shipped with the standard freeswan
package.

d) The pluto startup behavior is quadratic.  With 200 tunnels, this eats up
several minutes at every restart.   I'm told fixes are coming soon.

2) Other than item (1b), the CPU load depends mainly on the size of the
pipe attached, not on the number of tunnels.

It is worth noting that item (1b) applies only to repeated attempts to re-key a data connection (IPsec SA, Phase 2) over an established keying connection (ISAKMP SA, Phase 1). There are two ways to reduce this overhead using settings in ipsec.conf(5):

• set keyingtries to some small value to limit repetitions
• set keylife to a short time so that a failing data connection will be cleaned up when the keying connection is reset.

The overheads for establishing keying connections (ISAKMP SAs, Phase 1) are lower because for these Pluto does not perform expensive operations before receiving a reply from the peer.

A gateway that does a lot of rekeying -- many tunnels and/or low settings for tunnel lifetimes -- will also need a lot of random numbers from the random(4) driver.

Low-end systems

Even a 486 can handle a T1 line, according to this mailing list message:

Subject: Re: linux-ipsec: IPSec Masquerade
   Date: Fri, 15 Jan 1999 11:13:22 -0500
   From: Michael Richardson 

. . . A 486/66 has been clocked by Phil Karn to do
10Mb/s encryption.. that uses all the CPU, so half that to get some CPU,
and you have 5Mb/s. 1/3 that for 3DES and you get 1.6Mb/s....

and a piece of mail from project technical lead Henry Spencer:

Oh yes, and a new timing point for Sandy's docs...  A P60 -- yes, a 60MHz
Pentium, talk about antiques -- running a host-to-host tunnel to another
machine shows an FTP throughput (that is, end-to-end results with a real
protocol) of slightly over 5Mbit/s either way.  (The other machine is much
faster, the network is 100Mbps, and the ether cards are good ones... so
the P60 is pretty definitely the bottleneck.)

From the above, and from general user experience as reported on the list, it seems clear that a cheap surplus machine -- a reasonable 486, a minimal Pentium box, a Sparc 5, ... -- can easily handle a home office or a small company connection using any of:

• ADSL service
• cable modem
• T1
• E1

If available, we suggest using a Pentium 133 or better. This should ensure that, even under maximum load, IPsec will use less than half the CPU cycles. You then have enough left for other things you may want on your gateway -- firewalling, web caching, DNS and such.

Measuring KLIPS

Here is some additional data from the mailing list.

Subject: FreeSWAN (specically KLIPS) performance measurements
   Date: Thu, 01 Feb 2001
   From: Nigel Metheringham <Nigel.Metheringham@intechnology.co.uk>

I've spent a happy morning attempting performance tests against KLIPS 
(this is due to me not being able to work out the CPU usage of KLIPS so 
resorting to the crude measurements of maximum throughput to give a 
baseline to work out loading of a box).

Measurements were done using a set of 4 boxes arranged in a line, each 
connected to the next by 100Mbit duplex ethernet.  The inner 2 had an 
ipsec tunnel between them (shared secret, but I was doing measurements 
when the tunnel was up and running - keying should not be an issue 
here).  The outer pair of boxes were traffic generators or traffic sink.

The crypt boxes are Compaq DL380s - Uniprocessor PIII/733 with 256K 
cache.  They have 128M main memory.  Nothing significant was running on 
the boxes other than freeswan.  The kernel was a 2.2.19pre7 patched 
with freeswan and ext3.

Without an ipsec tunnel in the chain (ie the 2 inner boxes just being 
100BaseT routers), throughput (measured with ttcp) was between 10644 
and 11320 KB/sec

With an ipsec tunnel in place, throughput was between 3268 and 3402 
KB/sec

These measurements are for data pushed across a TCP link, so the 
traffic on the wire between the 2 ipsec boxes would have been higher 
than this....

vmstat (run during some other tests, so not affecting those figures) on 
the encrypting box shows approx 50% system & 50% idle CPU - which I 
don't believe at all.  Interactive feel of the box was significantly 
sluggish.

I also tried running the kernel profiler (see man readprofile) during 
test runs.

A box doing primarily decrypt work showed basically nothing happening - 
I assume interrupts were off.
A box doing encrypt work showed the following:-
 Ticks Function                                   Load
 ~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~    ~~~~~~
   956 total                                      0.0010
   532 des_encrypt2                               0.1330
   110 MD5Transform                               0.0443
    97 kmalloc                                    0.1880
    39 des_encrypt3                               0.1336
    23 speedo_interrupt                           0.0298
    14 skb_copy_expand                            0.0250
    13 ipsec_tunnel_start_xmit                    0.0009
    13 Decode                                     0.1625
    11 handle_IRQ_event                           0.1019
    11 .des_ncbc_encrypt_end                      0.0229
    10 speedo_start_xmit                          0.0188
     9 satoa                                      0.0225
     8 kfree                                      0.0118
     8 ip_fragment                                0.0121
     7 ultoa                                      0.0365
     5 speedo_rx                                  0.0071
     5 .des_encrypt2_end                          5.0000
     4 _stext                                     0.0140
     4 ip_fw_check                                0.0035
     2 rj_match                                   0.0034
     2 ipfw_output_check                          0.0200
     2 inet_addr_type                             0.0156
     2 eth_copy_and_sum                           0.0139
     2 dev_get                                    0.0294
     2 addrtoa                                    0.0143
     1 speedo_tx_buffer_gc                        0.0024
     1 speedo_refill_rx_buf                       0.0022
     1 restore_all                                0.0667
     1 number                                     0.0020
     1 net_bh                                     0.0021
     1 neigh_connected_output                     0.0076
     1 MD5Final                                   0.0083
     1 kmem_cache_free                            0.0016
     1 kmem_cache_alloc                           0.0022
     1 __kfree_skb                                0.0060
     1 ipsec_rcv                                  0.0001
     1 ip_rcv                                     0.0014
     1 ip_options_fragment                        0.0071
     1 ip_local_deliver                           0.0023
     1 ipfw_forward_check                         0.0139
     1 ip_forward                                 0.0011
     1 eth_header                                 0.0040
     1 .des_encrypt3_end                          0.0833
     1 des_decrypt3                               0.0034
     1 csum_partial_copy_generic                  0.0045
     1 call_out_firewall                          0.0125

Hope this data is helpful to someone... however the lack of visibility 
into the decrypt side makes things less clear

Speed with compression

Another user reported some results for connections with and without IP compression:

Subject: [Users] Speed with compression
   Date: Fri, 29 Jun 2001
   From: John McMonagle <johnm@advocap.org>

Did a couple tests with compression using the new 1.91 freeswan.

Running between 2 sites with cable modems.  Both  using approximately
130 mhz pentium.

Transferred files with ncftp.

Compressed file was a 6mb compressed  installation file.
Non compressed was 18mb /var/lib/rpm/packages.rpm

                            Compressed vpn          regular vpn
Compress file                42.59 kBs               42.08 kBs
regular file                110.84 kBs               41.66 kBs

Load  was about 0 either way.
Ping times were very similar  a bit above 9 ms.

Compression looks attractive to me.

> is there a reason not to switch compression on?  I have large gateway boxes
> connecting 3 connections, one of them with a measly DS1 link...

Run some timing tests with and without, with data and loads representative
of what you expect in production.  That's the definitive way to decide. 
If compression is a net loss, then obviously, leave it turned off.  If it
doesn't make much difference, leave it off for simplicity and hence
robustness.  If there's a substantial gain, by all means turn it on. 

If both ends support compression and can successfully negotiate a
compressed connection (trivially true if both are FreeS/WAN 1.91), then
the crucial question is CPU cycles. 

Compression has some overhead, so one question is whether *your* data
compresses well enough to save you more CPU cycles (by reducing the volume
of data going through CPU-intensive encryption/decryption) than it costs
you.  Last time I ran such tests on data that was reasonably compressible
but not deliberately contrived to be so, this generally was not true --
compression cost extra CPU cycles -- so compression was worthwhile only if
the link, not the CPU, was the bottleneck.  However, that was before the
slow-compression bug was fixed.  I haven't had a chance to re-run those
tests yet, but it sounds like I'd probably see a different result. 

Methods of measuring

If you want to measure the loads FreeS/WAN puts on a system, note that tools such as top or measurements such as load average are more-or-less useless for this. They are not designed to measure something that does most of its work inside the kernel.

Here is a message from FreeS/WAN kernel programmer Richard Guy Briggs on this:

> I have a batch of boxes doing Freeswan stuff.
> I want to measure the CPU loading of the Freeswan tunnels, but am 
> having trouble seeing how I get some figures out...
> 
>  - Keying etc is in userspace so will show up on the per-process
>    and load average etc (ie pluto's load)

Correct.

>  - KLIPS is in the kernel space, and does not show up in load average
>    I think also that the KLIPS per-packet processing stuff is running
>    as part of an interrupt handler so it does not show up in the
>    /proc/stat system_cpu or even idle_cpu figures

It is not running in interrupt handler.  It is in the bottom half.
This is somewhere between user context (careful, this is not
userspace!) and hardware interrupt context.

> Is this correct, and is there any means of instrumenting how much the 
> cpu is being loaded - I don't like the idea of a system running out of 
> steam whilst still showing 100% idle CPU :-)

vmstat seems to do a fairly good job, but use a running tally to get a
good idea.  A one-off call to vmstat gives different numbers than a
running stat.  To do this, put an interval on your vmstat command
line.

Subject: Re: Measuring the CPU usage of Freeswan
   Date: Mon, 29 Jan 2001
   From: Patrick Michael Kane <modus@pr.es.to>
 
The only truly accurate way to accurately track FreeSWAN CPU usage is to use
a CPU soaker. You run it on an unloaded system as a benchmark, then start up
FreeSWAN and take the difference to determine how much FreeSWAN is eating.
I believe someone has done this in the past, so you may find something in
the FreeSWAN archives.  If not, someone recently posted a URL to a CPU
soaker benchmark tool on linux-kernel.

Testing FreeS/WAN

Not all types of testing are described here. Other parts of the documentation describe some tests:

installation document

testing for a successful install

configuration document

basic tests for a working configuration

web links document

General information on tests for interoperability between various IPsec implementations. This includes links to several test sites.

interoperation document.

More specific information on FreeS/WAN interoperation with other implementations.

performance document

performance measurements

The test setups and procedures described here can also be used in other testing, but this document focuses on testing the IPsec functionality of FreeS/WAN.

Testing with User Mode Linux

User Mode Linux allows you to run Linux as a user process on another Linux machine.

As of 1.92, the distribution has a new directory named testing. It contains a collection of test scripts and sample configurations. Using these, you can bring up several copies of Linux in user mode and have them build tunnels to each other. This lets you do some testing of a FreeS/WAN configuration on a single machine.

You need a moderately well-endowed machine for this to work well. Each UML wants about 16 megs of memory by default, which is plenty for FreeS/WAN usage. Typical regression testing only occasionally uses as many as 4 UMLs. If one is doing nothing else with the machine (in particular, not running X on it), then 128 megs and a 500MHz CPU are fine.

Configuration for a testbed network

A common test setup is to put a machine with dual Ethernet cards in between two gateways under test. You need at least five machines; two gateways, two clients and a testing machine in the middle.

The central machine both routes packets and provides a place to run diagnostic software for checking IPsec packets. See next section for discussion of using tcpdump(8) for this.

This makes things more complicated than if you just connected the two gateway machines directly to each other, but it also makes your test setup much more like the environment you actually use IPsec in. Those environments nearly always involve routing, and quite a few apparent IPsec failures turn out to be problems with routing or with firewalls dropping packets. This approach lets you deal with those problems on your test setup.

What you end up with looks like:

Testbed network

      subnet a.b.c.0/24
             |
      eth1 = a.b.c.1
         gate1
      eth0 = 192.168.p.1
             |
             |
      eth0 = 192.168.p.2
         route/monitor box
      eth1 = 192.168.q.2
             |
             |
      eth0 = 192.168.q.1
         gate2
      eth1 = x.y.z.1
              |
       subnet x.y.z.0/24

Where p and q are any convenient values that do not interfere with other
routes you may have. The ipsec.conf(5) file then has, among other things:

conn abc-xyz
      left=192.168.p.1
      leftnexthop=192.168.p.2
      right=192.168.q.1
      rightnexthop=192.168.q.2

Once that works, you can remove the "route/monitor box", and connect the two gateways to the Internet. The only parameters in ipsec.conf(5) that need to change are the four shown above. You replace them with values appropriate for your Internet connection, and change the eth0 IP addresses and the default routes on both gateways.

Note that nothing on either subnet needs to change. This lets you test most of your IPsec setup before connecting to the insecure Internet.

Using packet sniffers in testing

A number of tools are available for looking at packets. We will discuss using tcpdump(8), a common Linux tool included in most distributions. Alternatives offerring more-or-less the same functionality include:

Ethereal

Several people on our mailing list report a preference for this over tcpdump.

windump

a Windows version of tcpdump(8), possibly handy if you have Windows boxes in your network

Sniffit

A linux sniffer that we don't know much about. If you use it, please comment on our mailing list.

See also this index of packet sniffers.

tcpdump(8) may misbehave if run on the gateways themselves. It is designed to look into a normal IP stack and may become confused if you ask it to display data from a stack which has IPsec in play.

At one point, the problem was quite severe. Recent versions of tcpdump, however, understand IPsec well enough to be usable on a gateway. You can get the latest version from tcpdump.org.

Even with a recent tcpdump, some care is required. Here is part of a post from Henry on the topic:

> a) data from sunset to sunrise or the other way is not being
> encrypted (I am using tcpdump (ver. 3.4) -x/ping -p to check
> packages) 

What *interface* is tcpdump being applied to?  Use the -i option to
control this.  It matters!  If tcpdump is looking at the ipsecN
interfaces, e.g. ipsec0, then it is seeing the packets before they are
encrypted or after they are decrypted, so of course they don't look
encrypted.  You want to have tcpdump looking at the actual hardware
interfaces, e.g. eth0. 

Actually, the only way to be *sure* what you are sending on the wire is to
have a separate machine eavesdropping on the traffic.  Nothing you can do
on the machines actually running IPsec is 100% guaranteed reliable in this
area (although tcpdump is a lot better now than it used to be).

The most certain way to examine IPsec packets is to look at them on the wire. For security, you need to be certain, so we recommend doing that. To do so, you need a separate sniffer machine located between the two gateways. This machine can be routing IPsec packets, but it must not be an IPsec gateway. Network configuration for such testing is discussed above.

Here's another mailing list message with advice on using tcpdump(8):

Subject: RE: [Users] Encrypted???
   Date: Thu, 29 Nov 2001
   From: "Joe Patterson" <jpatterson@asgardgroup.com>

tcpdump -nl -i $EXT-IF proto 50

-nl tells it not to buffer output or resolve names (if you don't do that it
may confuse you by not outputing anything for a while), -i $EXT-IF (replace
with your external interface) tells it what interface to listen on, and
proto 50 is ESP.  Use "proto 51" if for some odd reason you're using AH, and
"udp port 500" if you want to see the isakmp key exchange/tunnel setup
packets.

You can also run `tcpdump -nl -i ipsec0` to see what traffic is on that
virtual interface.  Anything you see there *should* be either encrypted or
dropped (unless you've turned on some strange options in your ipsec.conf
file)

Another very handy thing is ethereal (http://www.ethereal.com/) which runs
on just about anything, has a nice gui interface (or a nice text-based
interface), and does a great job of protocol  breakdown.  For ESP and AH
it'll basically just tell you that there's a packet of that protocol, and
what the spi is, but for isakmp it'll actually show you a lot of the tunnel
setup information (until it gets to the point in the protocol where isakmp
is encrypted....)

Verifying encryption

The question of how to verify that messages are actually encrypted has been extensively discussed on the mailing list. See this thread.

If you just want to verify that packets are encrypted, look at them with a packet sniffer (see previous section ) located between the gateways. The packets should, except for some of the header information, be utterly unintelligible. The output of good encryption looks exactly like random noise.

A packet sniffer can only tell you that the data you looked at was encrypted. If you have stronger requirements -- for example if your security policy requires verification that plaintext is not leaked during startup or under various anomolous conditions -- then you will need to devise much more thorough tests. If you do that, please post any results or methodological details which your security policy allows you to make public.

You can put recognizable data into ping packets with something like:

        ping -p feedfacedeadbeef 11.0.1.1

"feedfacedeadbeef" is a legal hexadecimal pattern that is easy to pick out of hex dumps.

For other protocols, you may need to check if you have encrypted data or ASCII text. Encrypted data has approximately equal frequencies for all 256 possible characters. ASCII text has most characters in the printable range 0x20-0x7f, a few control characters less than 0x20, and none at all in the range 0x80-0xff. 0x20, space, is a good character to look for. In normal English text space occurs about once in seven characters, versus about once in 256 for random or encrypted data.

One thing to watch for: the output of good compression, like that of good encryption, looks just like random noise. You cannot tell just by looking at a data stream whether it has been compressed, encrypted, or both. You need a little care not to mistake compressed data for encrypted data in your testing.

Note also that weak encryption also produces random-looking output. You cannot tell whether the encryption is strong by looking at the output. To be sure of that, you would need to have both the algorithms and the implementation examined by experts.

For IPsec, you can get partial assurance from interoperability tests. See our interop document. When twenty products all claim to implement 3DES, and they all talk to each other, you can be fairly sure they have it right. Of course, you might wonder whether all the implementers are consipring to trick you or, more plausibly, whether some implementations might have "back doors" so they can get also it wrong when required.. If you're seriously worried about things like that, you need to get the code you use audited (good luck if it is not Open Source), or perhaps to talk to a psychiatrist about treatments for paranoia.

Mailing list pointers

Additional information on testing can be found in these mailing list messages:

• a user's detailed setup diary for his testbed network
• a FreeS/WAN team member's notes from testing at an IPsec interop "bakeoff"

Kernel configuration for FreeS/WAN

This section lists many of the options available when configuring a Linux kernel, and explains how they should be set on a FreeS/WAN IPsec gateway.

Not everyone needs to worry about kernel configuration

Note that in many cases you do not need to mess with these.

You may have a Linux distribution which comes with FreeS/WAN installed (see this list). In that case, you need not do a FreeS/WAN installation or a kernel configuration. Of course, you might still want to configure and rebuild your kernel to improve performance or security. This can be done with standard tools described in the Kernel HowTo.

If you need to install FreeS/WAN, then you do need to configure a kernel. However, you may choose to do that using the simplest procedure:

• Configure, build and test a kernel for your system before adding FreeS/WAN. See the Kernel HowTo for details. This step cannot be skipped . FreeS/WAN needs the results of your configuration.
• Then use FreeS/WAN's make oldgo command. This sets everything FreeS/WAN needs and retains your values everywhere else.

This document is for those who choose to configure their FreeS/WAN kernel themselves.

Assumptions and notation

Help text for most kernel options is included with the kernel files, and is accessible from within the configuration utilities. We assume you will refer to that, and to the Kernel HowTo, as necessary. This document covers only the FreeS/WAN-specific aspects of the problem.

To avoid duplication, this document section does not cover settings for the additional IPsec-related kernel options which become available after you have patched your kernel with FreeS/WAN patches. There is help text for those available from within the configuration utility.

We assume a common configuration in which the FreeS/WAN IPsec gateway is also doing ipchains(8) firewalling for a local network, and possibly masquerading as well.

Some suggestions below are labelled as appropriate for "a true paranoid". By this we mean they may cause inconvenience and it is not entirely clear they are necessary, but they appear to be the safest choice. Not using them might entail some risk. Of course one suggested mantra for security administrators is: "I know I'm paranoid. I wonder if I'm paranoid enough."

Labels used

Six labels are used to indicate how options should be set. We mark the labels with [square brackets]. For two of these labels, you have no choice:

[required]

essential for FreeS/WAN operation.

[incompatible]

incompatible with FreeS/WAN.

those must be set correctly or FreeS/WAN will not work

FreeS/WAN should work with any settings of the others, though of course not all combinations have been tested. We do label these in various ways, but these labels are only suggestions.

[recommended]

useful on most FreeS/WAN gateways

[disable]

an unwelcome complication on a FreeS/WAN gateway.

[optional]

Your choice. We outline issues you might consider.

[anything]

This option has no direct effect on FreeS/WAN and related tools, so you should be able to set it as you please.

Of course complexity is an enemy in any effort to build secure systems. For maximum security, any feature that can reasonably be turned off should be. "If in doubt, leave it out."

Kernel options for FreeS/WAN

Indentation is based on the nesting shown by 'make menuconfig' with a 2.2.16 kernel for the i386 architecture.

Code maturity and level options

Prompt for development ... code/drivers

[optional] If this is no, experimental drivers are not shown in later menus.

For most FreeS/WAN work, no is the preferred setting. Using new or untested components is too risky for a security gateway.

However, for some hardware (such as the author's network cards) the only drivers available are marked new/experimental. In such cases, you must enable this option or your cards will not appear under "network device support". A true paranoid would leave this option off and replace the cards.

Processor type and features

[anything]

Loadable module support

Enable loadable module support

[optional] A true paranoid would disable this. An attacker who has root access to your machine can fairly easily install a bogus module that does awful things, provided modules are enabled. A common tool for attackers is a "rootkit", a set of tools the attacker uses once he or she has become root on your system. The kit introduces assorted additional compromises so that the attacker will continue to "own" your system despite most things you might do to recovery the situation. For Linux, there is a tool called knark which is basically a rootkit packaged as a kernel module.

With modules disabled, an attacker cannot install a bogus module. The only way he can achieve the same effects is to install a new kernel and reboot. This is considerably more likely to be noticed.

Many FreeS/WAN gateways run with modules enabled. This simplifies some administrative tasks and some ipchains features are available only as modules. Once an enemy has root on your machine your security is nil, so arguably defenses which come into play only in that situation are pointless.

Set version information ....

[optional] This provides a check to prevent loading modules compiled for a different kernel.

Kernel module loader

[disable] It gives little benefit on a typical FreeS/WAN gate and entails some risk.

General setup

We list here only the options that matter for FreeS/WAN.

Networking support

[required]

Sysctl interface

[optional] If this option is turned on and the /proc filesystem installed, then you can control various system behaviours by writing to files under /proc/sys. For example:

        echo 1 > /proc/sys/net/ipv4/ipforward

turns IP forwarding on.

Disabling this option breaks many firewall scripts. A true paranoid would disable it anyway since it might conceivably be of use to an attacker.

Plug and Play support

[anything]

Block devices

[anything]

Networking options

Packet socket

[optional] This kernel feature supports tools such as tcpdump(8) which communicate directly with network hardware, bypassing kernel protocols. This is very much a two-edged sword:

• such tools can be very useful to the firewall admin, especially during initial testing
• should an evildoer breach your firewall, such tools could give him or her a great deal of information about the rest of your network

We recommend disabling this option on production gateways.

Kernel/User netlink socket

[optional] Required if you want to use advanced router features.

Routing messages

[optional]

Netlink device emulation

[optional]

Network firewalls

[recommended] You need this if the IPsec gateway also functions as a firewall.

Even if the IPsec gateway is not your primary firewall, we suggest setting this so that you can protect the gateway with at least basic local packet filters.

Socket filtering

[disable] This enables an older filtering interface. We suggest using ipchains(8) instead. To do that, set the "Network firewalls" option just above, and not this one.

Unix domain sockets

[required] These sockets are used for communication between the ipsec(8) commands and the ipsec_pluto(8) daemon.

TCP/IP networking

[required]

IP: multicasting

[anything]

IP: advanced router

[optional] This gives you policy routing, which some people have used to good advantage in their scripts for FreeS/WAN gateway management. It is not used in our distributed scripts, so not required unless you want it for custom scripts. It requires the netlink interface between kernel code and the iproute2(8) command.

IP: kernel level autoconfiguration

[disable] It gives little benefit on a typical FreeS/WAN gate and entails some risk.

IP: firewall packet netlink device

[disable]

IP: transparent proxy support

[optional] This is required in some firewall configurations, but should be disabled unless you have a definite need for it.

IP: masquerading

[optional] Required if you want to use non-routable private IP addresses for your local network.

IP: Optimize as router not host

[recommended]

IP: tunneling

[required]

IP: GRE tunnels over IP

[anything]

IP: aliasing support

[anything]

IP: ARP daemon support (EXPERIMENTAL)

Not required on most systems, but might prove useful on heavily-loaded gateways.

IP: TCP syncookie support

[recommended] It provides a defense against a denial of service attack which uses bogus TCP connection requests to waste resources on the victim machine.

IP: Reverse ARP

IP: large window support

[recommended] unless you have less than 16 meg RAM

IPv6

[optional] FreeS/WAN does not currently support IPv6, though work on integrating FreeS/WAN with the Linux IPv6 stack has begun. Details.

It should be possible to use IPv4 FreeS/WAN on a machine which also does IPv6. This combination is not yet well tested. We would be quite interested in hearing results from anyone expermenting with it, via the mailing list.

We do not recommend using IPv6 on production FreeS/WAN gateways until more testing has been done.

Novell IPX

[disable]

Appletalk

[disable] Quite a few Linux installations use IP but also have some other protocol, such as Appletalk or IPX, for communication with local desktop machines. In theory it should be possible to configure IPsec for the IP side of things without interfering with the second protocol.

We do not recommend this. Keep the software on your gateway as simple as possible. If you need a Linux-based Appletalk or IPX server, use a separate machine.

Telephony support

[anything]

SCSI support

[anything]

I2O device support

[anything]

Network device support

[anything] should work, but there are some points to note.

The development team test almost entirely on 10 or 100 megabit Ethernet and modems. In principle, any device that can do IP should be just fine for IPsec, but in the real world any device that has not been well-tested is somewhat risky. By all means try it, but don't bet your project on it until you have solid test results.

If you disabled experimental drivers in the Code maturity section above, then those drivers will not be shown here. Check that option before going off to hunt for missing drivers.

If you want Linux to automatically find more than one ethernet interface at boot time, you need to:

• compile the appropriate driver(s) into your kernel. Modules will not work for this
• add a line such as

     append="ether=0,0,eth0 ether=0,0,eth1"
  

  to your /etc/lilo.conf file. In some cases you may need to specify parameters such as IRQ or base address. The example uses "0,0" for these, which tells the system to search. If the search does not succeed on your hardware, then you should retry with explicit parameters. See the lilo.conf(5) man page for details.
• run lilo(8)

Having Linux find the cards this way is not necessary, but is usually more convenient than loading modules in your boot scripts.

Amateur radio support

[anything]

IrDA (infrared) support

[anything]

ISDN subsystem

[anything]

Old CDROM drivers

[anything]

Character devices

The only required character device is:

random(4)

[required] This is a source of random numbers which are required for many cryptographic protocols, including several used in IPsec.

If you are comfortable with C source code, it is likely a good idea to go in and adjust the #define lines in /usr/src/linux/drivers/char/random.c to ensure that all sources of randomness are enabled. Relying solely on keyboard and mouse randomness is dubious procedure for a gateway machine. You could also increase the randomness pool size from the default 512 bytes (128 32-bit words).

Filesystems

[anything] should work, but we suggest limiting a gateway machine to the standard Linux ext2 filesystem in most cases.

Network filesystems

[disable] These systems are an unnecessary risk on an IPsec gateway.

Console drivers

[anything]

Sound

[anything] should work, but we suggest enabling sound only if you plan to use audible alarms for firewall problems.

Kernel hacking

[disable] This might be enabled on test machines, but should not be on production gateways.

Other configuration possibilities

This document describes various options for FreeS/WAN configuration which are less used or more complex (often both) than the standard cases described in our quickstart document.

Some rules of thumb about configuration

Tunnels are cheap

Nearly all of the overhead in IPsec processing is in the encryption and authentication of packets. Our performance document discusses these overheads.

Beside those overheads, the cost of managing additional tunnels is trivial. Whether your gateway supports one tunnel or ten just does not matter. A hundred might be a problem; there is a section on this in the performance document.

So, in nearly all cases, if using multiple tunnels gives you a reasonable way to describe what you need to do, you should describe it that way in your configuration files.

For example, one user recently asked on a mailing list about this network configuration:

        netA---gwA---gwB---netB
                            |----netC

   netA and B are secured netC not.
   netA and gwA can not access netC

The user had constructed only one tunnel, netA to netB, and wanted to know how to use ip-route to get netC packets into it. This is entirely unnecessary. One of the replies was:

  The simplest way and indeed the right way to
  solve this problem is to set up two connections:

        leftsubnet=NetA
        left=gwA
        right=gwB
        rightsubnet=NetB
  and
        leftsubnet=NetA
        left=gwA
        right=gwB
        rightsubnet=NetC

Of course another possibility would be to just use one tunnel, with a subnet mask that includes both netB and netC (or B, C, D, ...). See next section.

In general, you can construct as many tunnels as you need. Networks like netC in this example that do not connect directly to the gateway are fine, as long as the gateway can route to them.

The number of tunnels can become an issue if it reaches 50 or so. This is discussed in the performance document. Look there for information on supporting hundreds of Road Warriors from one gateway.

Subnet sizes

The subnets used in leftsubnet and rightsubnet can be of any size that fits your needs, and they need not correspond to physical networks.

You adjust the size by changing the subnet mask , the number after the slash in the subnet description. For example

• in 192.168.100.0/24 the /24 mask says 24 bits are used to designate the network. This leave 8 bits to label machines. This subnet has 256 addresses. .0 and .255 are reserved, so it can have 254 machines.
• A subnet with a /23 mask would be twice as large, 512 addresses.
• A subnet with a /25 mask would be half the size, 128 addresses.
• /0 is the whole Internet
• /32 is a single machine

As an example of using these in connection descriptions, suppose your company's head office has four physical networks using the address ranges:

192.168.100.0/24

development

192.168.101.0/24

production

192.168.102.0/24

marketing

192.168.103.0/24

administration

You can use exactly those subnets in your connection descriptions, or use larger subnets to grant broad access if required:

leftsubnet=192.168.100.0/24

remote hosts can access only development

leftsubnet=192.168.100.0/23

remote hosts can access development or production

leftsubnet=192.168.102.0/23

remote hosts can access marketing or administration

leftsubnet=192.168.100.0/22

remote hosts can access any of the four departments

or use smaller subnets to restrict access:

leftsubnet=192.168.103.0/24

remote hosts can access any machine in administration

leftsubnet=192.168.103.64/28

remote hosts can access only certain machines in administration.

leftsubnet=192.168.103.42/32

remote hosts can access only one particular machine in administration

To be exact, 192.68.103.64/28 means all addresses whose top 28 bits match 192.168.103.64. There are 16 of these because there are 16 possibilities for the remainingg 4 bits. Their addresses are 192.168.103.64 to 192.168.103.79.

Each connection description can use a different subnet if required.

It is possible to use all the examples above on the same FreeS/WAN gateway, each in a different connection description, perhaps for different classes of user or for different remote offices.

It is also possible to have multiple tunnels using different leftsubnet descriptions with the same right. For example, when the marketing manager is on the road he or she might have access to:

leftsubnet=192.168.102.0/24

all machines in marketing

192.168.101.32/29

some machines in production

leftsubnet=192.168.103.42/32

one particular machine in administration

This takes three tunnels, but tunnels are cheap. If the laptop is set up to build all three tunnels automatically, then he or she can access all these machines concurrently, perhaps from different windows.

Other network layouts

Here is the usual network picture for a site-to-site VPN::

     Sunset==========West------------------East=========Sunrise
           local net       untrusted net       local net

and for the Road Warrior::

                                           telecommuter's PC or
                                           traveller's laptop
     Sunset==========West------------------East
         corporate LAN     untrusted net

Other configurations are also possible.

The Internet as a big subnet

A telecommuter might have:

     Sunset==========West------------------East ================= firewall --- the Internet
         home network      untrusted net        corporate network

This can be described as a special case of the general subnet-to-subnet connection. The subnet on the right is 0.0.0.0/0, the whole Internet.

West (the home gateway) can have its firewall rules set up so that only IPsec packets to East are allowed out. It will then behave as if its only connection to the world was a wire to East.

When machines on the home network need to reach the Internet, they do so via the tunnel, East and the corporate firewall. From the viewpoint of the Internet (perhaps of some EvilDoer trying to break in!), those home office machines are behind the firewall and protected by it.

Wireless

Another possible configuration comes up when you do not trust the local network, either because you have very high security standards or because your are using easily-intercepted wireless signals.

Some wireless networks have built-in encryption called WEP, but its security is dubious. It is a fairly common practice to use IPsec instead.

In this case, part of your network may look like this:

          West-----------------------------East == the rest of your network
     workstation   untrusted wireless net

Of course, there would likely be several wireless workstations, each with its own IPsec tunnel to the East gateway.

The connection descriptions look much like Road Warrior descriptions:

• each workstation should have its own unique
  • identifier for IPsec
  • RSA key
  • connection description.
• on the gateway, use left=%any, or the workstation IP address
• on workstations, left=%defaultroute, or the workstation IP address
• leftsubnet= is not used.

The rightsubnet= parameter might be set in any of several ways:

rightsubnet=0.0.0.0/0

allowing workstations to access the entire Internet (see above)

rightsubnet=a.b.c.0/24

allowing access to your entire local network

rightsubnet=a.b.c.d/32

restricting the workstation to connecting to a particular server

Of course you can mix and match these as required. For example, a university might allow faculty full Internet access while letting student laptops connect only to a group of lab machines.

Choosing connection types

One choice you need to make before configuring additional connections is what type or types of connections you will use. There are several options, and you can use more than one concurrently.

Manual vs. automatic keying

IPsec allows two types of connections, with manual or automatic keying. FreeS/WAN starts them with commands such as:

        ipsec manual --start name
        ipsec auto --up name

The difference is in how they are keyed.

Manually keyed connections

use keys stored in ipsec.conf .

Automatically keyed connections

use keys automatically generated by the Pluto key negotiation daemon. The key negotiation protocol, IKE, must authenticate the other system. (It is vulnerable to a man-in-the-middle attack if used without authentication.) We currently support two authentication methods:

• using shared secrets stored in ipsec.secrets.
• RSA public key authentication, with our machine's private key in ipsec.secrets. Public keys for other machines may either be placed in ipsec.conf or provided via DNS.

A third method, using RSA keys embedded in X.509 certtificates, is provided by user patches.

Manually keyed connections provide weaker security than automatically keyed connections. An opponent who reads ipsec.secrets(5) gets your encryption key and can read all data encrypted by it. If he or she has an archive of old messages, all of them back to your last key change are also readable.

With automatically-(re)-keyed connections, an opponent who reads ipsec.secrets(5) gets the key used to authenticate your system in IKE -- the shared secret or your private key, depending what authentication mechanism is in use. However, he or she does not automatically gain access to any encryption keys or any data.

An attacker who has your authentication key can mount a man-in-the-middle attack and, if that succeeds, he or she will get encryption keys and data. This is a serious danger, but it is better than having the attacker read everyting as soon as he or she breaks into ipsec.secrets(5).. Moreover, the keys change often so an opponent who gets one key does not get a large amount of data. To read all your data, he or she would have to do a man-in-the-middle attack at every key change.

We discuss using manual keying in production below, but this is not recommended except in special circumstances, such as needing to communicate with some implementation that offers no auto-keyed mode compatible with FreeS/WAN.

Manual keying may also be useful for testing. There is some discussion of this in our FAQ.

Authentication methods for auto-keying

The IKE protocol which Pluto uses to negotiate connections between gateways must use some form of authentication of peers. A gateway must know who it is talking to before it can create a secure connection. We support two basic methods for this authentication:

• shared secrets, stored in ipsec.secrets(5)
• RSA authentication

There are, howver, several variations on the RSA theme, using different methods of managing the RSA keys:

• our RSA private key in ipsec.secrets(5) with other gateways' public keys

  either

  stored in ipsec.conf(5)

  or

  looked up via DNS

• authentication with x.509 certificates.; See our links section for information on user-contributed patches for this.:

Public keys in ipsec.conf(5 ) give a reasonably straightforward method of specifying keys for explicitly configured connections.

Putting public keys in DNS allows us to support opportunistic encryption. Any two FreeS/WAN gateways can provide secure communication, without either of them having any preset information about the other.

X.509 certificates may be required to interface to various PKIs.

Advantages of public key methods

Authentication with a public key method such as RSA has some important advantages over using shared secrets.

There is also a disadvantage:

• your private key is a single point of attack, extremely valuable to an enemy
  • with shared secrets, an attacker who steals your ipsec.secrets file can impersonate you or try man-in-the-middle attacks, but can only attack connections described in that file
  • an attacker who steals your private key gains the chance to attack not only existing connections but also any future connections created using that key

This is partly counterbalanced by the fact that the key is never transmitted and remains under your control at all times. It is likely necessary, however, to take account of this in setting security policy. For example, you should change gateway keys when an administrator leaves the company, and should change them periodically in any case.

Overall, public key methods are more secure, more easily managed and more flexible. We recommend that they be used for all connections, unless there is a compelling reason to do otherwise.

Using shared secrets in production

Generally, public key methods are preferred for reasons given above, but shared secrets can be used with no loss of security, just more work and perhaps more need to take precautions.

What I call "shared secrets" are sometimes also called "pre-shared keys". They are used only for for authentication, never for encryption. Calling them "pre-shared keys" has confused some users into thinking they were encryption keys, so I prefer to avoid the term..

If you are interoperating with another IPsec implementation, you may find its documentation calling them "passphrases".

Putting secrets in ipsec.secrets(5)

If shared secrets are to be used to authenticate communication for the Diffie-Hellman key exchange in the IKE protocol, then those secrets must be stored in /etc/ipsec.secrets. For details, see the ipsec.secrets(5) man page.

A few considerations are vital:

• make the secrets long and unguessable. Since they need not be remembered by humans, very long ugly strings may be used. We suggest using our ipsec_ranbits(8) utility to generate long (128 bits or more) random strings.
• transmit secrets securely. You have to share them with other systems, but you lose if they are intercepted and used against you. Use PGP, SSH, hand delivery of a floppy disk which is then destroyed, or some other trustworthy method to deliver them.
• store secrets securely, in root-owned files with permissions rw------.
• limit sharing of secrets. Alice, Bob, Carol and Dave may all talk to each other, but only Alice and Bob should know the secret for an Alice-Bob link.
• do not share private keys. The private key for RSA authentication of your system is stored in ipsec.secrets(5), but it is a different class of secret from the pre-shared keys used for the "shared secret" authentication. No-one but you should have the RSA private key.

Each line has the IP addresses of the two gateways plus the secret. It should look something like this:

        10.0.0.1 11.0.0.1 : PSK "jxTR1lnmSjuj33n4W51uW3kTR55luUmSmnlRUuWnkjRj3UuTV4T3USSu23Uk55nWu5TkTUnjT"

PSK indicates the use of a pre-s hared key. The quotes and the whitespace shown are required.

You can use any character string as your secret. For security, it should be both long and extremely hard to guess. We provide a utility to generate such strings, ipsec_ranbits(8).

You want the same secret on the two gateways used, so you create a line with that secret and the two gateway IP addresses. The installation process supplies an example secret, useful only for testing. You must change it for production use.

File security

You must deliver this file, or the relevant part of it, to the other gateway machine by some secure means. Don't just FTP or mail the file! It is vital that the secrets in it remain secret. An attacker who knew those could easily have all the data on your "secure" connection.

This file must be owned by root and should have permissions rw-------.

Shared secrets for road warriors

You can use a shared secret to support a single road warrior connecting to your gateway, and this is a reasonable thing to do in some circumstances. Public key methods have advantages, discussed above, but they are not critical in this case.

To do this, the line in ipsec.secrets(5) is something like:

        10.0.0.1 0.0.0.0 : PSK "jxTR1lnmSjuj33n4W51uW3kTR55luUmSmnlRUuWnkjRj3UuTV4T3USSu23Uk55nWu5TkTUnjT"

For more than one road warrior, shared secrets are not recommended. If shared secrets are used, then when the responder needs to look up the secret, all it knows about the sender is an IP address. This is fine if the sender is at a fixed IP address specified in the config file. It is also fine if only one road warrior uses the wildcard 0.0.0.0 address. However, if you have more than one road warrior using shared secret authentication, then they must all use that wildcard and therefore all road warriors using PSK autentication must use the same secret. Obviously, this is insecure.

For multiple road warriors, use public key authentication. Each roadwarrior can then have its own identity (our leftid= or rightid= parameters), its own public/private key pair, and its own secure connection.

Using manual keying in production

Generally, automatic keying is preferred over manual keying for production use because it is both easier to manage and more secure. Automatic keying frees the admin from much of the burden of managing keys securely, and can provide perfect forward secrecy. This is discussed in more detail above.

However, it is possible to use manual keying in production if that is what you want to do. This might be necessary, for example, in order to interoperate with some device that either does not provide automatic keying or provides it in some version we cannot talk to.

Note that with manual keying all security rests with the keys . If an adversary acquires your keys, you've had it. He or she can read everything ever sent with those keys, including old messages he or she may have archived.

You need to be really paranoid about keys if you're going to rely on manual keying for anything important.

• keep keys in files with 600 permissions, owned by root
• be extremely careful about security of your gateway systems. Anyone who breaks into a gateway and gains root privileges can get all your keys and read everything ever encrypted with those keys, both old messages he has archived and any new ones you may send.
• change keys regularly. This can be a considerable bother, (and provides an excellent reason to consider automatic keying instead), but it is absolutely essential for security. Consider a manually keyed system in which you leave the same key in place for months:
  • an attacker can have a very large sample of text sent with that key to work with. This makes various cryptographic attacks much more likely to succeed.
  • The chances of the key being compromised in some non-cryptographic manner -- a spy finds it on a discarded notepad, someone breaks into your server or your building and steals it, a staff member is bribed, tricked, seduced or coerced into revealing it, etc. -- also increase over time.
  • a successful attacker can read everything ever sent with that key. This makes any successful attack extremely damaging.
  It is clear that you must change keys often to have any useful security. The only question is how often.
• use PGP or SSH for all key transfers
• don't edit files with keys in them when someone can look over your shoulder
• worry about network security; could someone get keys by snooping packets on the LAN between your X desktop and the gateway?
• lock up your backup tapes for the gateway system
• ... and so on

Linux FreeS/WAN provides some facilities to help with this. In particular, it is good policy to keep keys in separate files so you can edit configuration information in /etc/ipsec.conf without exposing keys to "shoulder surfers" or network snoops. We support this with the also= and include syntax in ipsec.conf(5).

See the last example in our examples file. In the /etc/ipsec.conf conn samplesep section, it has the line:

        also=samplesep-keys

which tells the "ipsec manual" script to insert the configuration description labelled "samplesep-keys" if it can find it. The /etc/ipsec.conf file must also have a line such as:

include ipsec.*.conf

which tells it to read other files. One of those other files then might contain the additional data:

conn samplesep-keys
  spi=0x200
  esp=3des-md5-96
  espenckey=0x01234567_89abcdef_02468ace_13579bdf_12345678_9abcdef0
  espauthkey=0x12345678_9abcdef0_2468ace0_13579bdf

The first line matches the label in the "also=" line, so the indented lines are inserted. The net effect is exactly as if the inserted lines had occurred in the original file in place of the "also=" line.

Variables set here are:

spi

A number needed by the manual keying code. Any 3-digit hex number will do, but if you have more than one manual connection then spi must be different for each connection.

esp

Options for ESP (Encapsulated Security Payload), the usual IPsec encryption mode. Settings here are for encryption using triple DES and authentication using MD5. Note that encryption without authentication should not be used; it is insecure.

espenkey

Key for ESP encryption. Here, a 192-bit hex number for triple DES.

espauthkey

Key for ESP authentication. Here, a 128-bit hex number for MD5.

Note that the example keys we supply are intended only for testing. For real use, you should go to automatic keying. If that is not possible, create your own keys for manual mode and keep them secret

Of course, any files containing keys must have 600 permissions and be owned by root.

If you connect in this way to multiple sites, we recommend that you keep keys for each site in a separate file and adopt some naming convention that lets you pick them all up with a single "include" line. This minimizes the risk of losing several keys to one error or attack and of accidentally giving another site admin keys which he or she has no business knowing.

Also note that if you have multiple manually keyed connections on a single machine, then the spi parameter must be different for each one. Any 3-digit hex number is OK, provided they are different for each connection. We reserve the range 0x100 to 0xfff for manual connections. Pluto assigns SPIs from 0x1000 up for automatically keyed connections.

If ipsec.conf(5) contains keys for manual mode connections, then it too must have permissions rw-------. We recommend instead that, if you must manual keying in production, you keep the keys in separate files.

Note also that ipsec.conf is installed with permissions rw-r--r--. If you plan to use manually keyed connections for anything more than initial testing, you must:

• either change permissions to rw-------
• or store keys separately in secure files and access them via include statements in ipsec.conf.

We recommend the latter method for all but the simplest configurations.

Creating keys with ranbits

You can create new random keys with the ranbits(8) utility. For example, the commands:

      umask 177
      ipsec ranbits 192  > temp
      ipsec ranbits 128 >> temp

create keys in the sizes needed for our default algorithms:

• 192-bit key for 3DES encryption
  (only 168 bits are used; parity bits are ignored)
• 128-bit key for keyed MD5 authentication

If you want to use SHA instead of MD5, that requires a 160-bit key

Note that any temporary files used must be kept secure since they contain keys. That is the reason for the umask command above. The temporary file should be deleted as soon as you are done with it. You may also want to change the umask back to its default value after you are finished working on keys.

The ranbits utility may pause for a few seconds if not enough entropy is available immediately. See ipsec_ranbits(8) and random(4) for details. You may wish to provide some activity to feed entropy into the system. For example, you might move the mouse around, type random characters, or do du /usr > /dev/null in the background.

Setting up connections at boot time

You can tell the system to set up connections automatically at boot time by putting suitable stuff in /etc/ipsec.conf on both systems. The relevant section of the file is labelled by a line reading config setup.

Details can be found in the ipsec.conf(5) man page. We also provide a file of example configurations.

The most likely options are something like:

interfaces="ipsec0=eth0 ipsec1=ppp0"

Tells KLIPS which interfaces to use. Up to four interfaces numbered ipsec[0-3] are supported. Each interface can support an arbitrary number of tunnels.

Note that for PPP, you give the ppp[0-9] device name here, not the underlying device such as modem (or eth1 if you are using PPPoE).

interfaces=%defaultroute

Alternative setting, useful in simple cases. KLIPS will pick up both its interface and the next hop information from the settings of the Linux default route.

forwardcontrol=no

Normally "no". Set to "yes" if the IP forwarding option is disabled in your network configuration. (This can be set as a kernel configuration option or later. e.g. on Redhat, it's in /etc/sysconfig/network and on SuSE you can adjust it with Yast.) Linux FreeS/WAN will then enable forwarding when starting up and turn it off when going down. This is used to ensure that no packets will be forwarded before IPsec comes up and takes control.

syslog=daemon.error

Used in messages to the system logging daemon (syslogd) to specify what type of software is sending the messages. If the settings are "daemon.error" as in our example, then syslogd treats the messages as error messages from a daemon.

Note that Pluto does not currently pay attention to this variable. The variable controls setup messages only.

klipsdebug=

Debug settings for KLIPS.

plutodebug=

Debug settings for Pluto.

... for both the above DEBUG settings

Normally, leave empty as shown above for no debugging output.
Use "all" for maximum information.
See ipsec_klipsdebug(8) and ipsec_pluto(8) man page for other options. Beware that if you set /etc/ipsec.conf to enable debug output, your system's log files may get large quickly.

dumpdir=/safe/directory

Normally, programs started by ipsec setup don't crash. If they do, by default, no core dump will be produced because such dumps would contain secrets. If you find you need to debug such crashes, you can set dumpdir to the name of a directory in which to collect the core file.

manualstart=

List of manually keyed connections to be automatically started at boot time. Useful for testing, but not for long term use. Connections which are automatically started should also be automatically re-keyed.

pluto=yes

Whether to start Pluto when ipsec startup is done.
This parameter is optional and defaults to "yes" if not present.

"yes" is strongly recommended for production use so that the keying daemon (Pluto) will automatically re-key the connections regularly. The ipsec-auto parameters ikelifetime, ipseclifetime and reykeywindow give you control over frequency of rekeying.

plutoload="reno-van reno-adam reno-nyc"

List of tunnels (by name, e.g. fred-susan or reno-van in our examples) to be loaded into Pluto's internal database at startup. In this example, Pluto loads three tunnels into its database when it is started.

If plutoload is "%search", Pluto will load any connections whose description includes "auto=add" or "auto=start".

plutostart="reno-van reno-adam reno-nyc"

List of tunnels to attempt to negotiate when Pluto is started.

If plutostart is "%search", Pluto will start any connections whose description includes "auto=start".

Note that, for a connection intended to be permanent, both gateways should be set try to start the tunnel. This allows quick recovery if either gateway is rebooted or has its IPsec restarted. If only one gateway is set to start the tunnel and the other gateway restarts, the tunnel may not be rebuilt.

plutowait=no

Controls whether Pluto waits for one tunnel to be established before starting to negotiate the next. You might set this to "yes"

• if your gateway is a very limited machine and you need to conserve resources.
• for debugging; the logs are clearer if only one connection is brought up at a time

For a busy and resource-laden production gateway, you likely want "no" so that connections are brought up in parallel and the whole process takes less time.

The example assumes you are at the Reno office and will use IPsec to Vancouver, New York City and Amsterdam.

Multiple tunnels between the same two gateways

Consider a pair of subnets, each with a security gateway, connected via the Internet:

         192.168.100.0/24           left subnet
              |
         192.168.100.1
         North Gateway
         101.101.101.101            left
              |
         101.101.101.1              left next hop
         [Internet]
         202.202.202.1              right next hop
              |
         202.202.202.202            right
         South gateway
         192.168.200.1
              |
         192.168.200.0/24           right subnet

A tunnel specification such as:

conn northnet-southnet
      left=101.101.101.101
      leftnexthop=101.101.101.1
      leftsubnet=192.168.100.0/24
      leftfirewall=yes
      right=202.202.202.202
      rightnexthop=202.202.202.1
      rightsubnet=192.168.200.0/24
      rightfirewall=yes

However, this does not cover other traffic you might want to secure. To handle all the possibilities, you might also want these connection descriptions:

conn northgate-southnet
      left=101.101.101.101
      leftnexthop=101.101.101.1
      right=202.202.202.202
      rightnexthop=202.202.202.1
      rightsubnet=192.168.200.0/24
      rightfirewall=yes

conn northnet-southgate
      left=101.101.101.101
      leftnexthop=101.101.101.1
      leftsubnet=192.168.100.0/24
      leftfirewall=yes
      right=202.202.202.202
      rightnexthop=202.202.202.1

Without these, neither gateway can do IPsec to the remote subnet. There is no IPsec tunnel or eroute set up for the traffic.

In our example, with the non-routable 192.168.* addresses used, packets would simply be discarded. In a different configuration, with routable addresses for the remote subnet, they would be sent unencrypted since there would be no IPsec eroute and there would be a normal IP route.

You might also want:

conn northgate-southgate
      left=101.101.101.101
      leftnexthop=101.101.101.1
      right=202.202.202.202
      rightnexthop=202.202.202.1

This is required if you want the two gateways to speak IPsec to each other.

This requires a lot of duplication of details. Judicious use of also= and include can reduce this problem.

Note that, while FreeS/WAN supports all four tunnel types, not all implementations do. In particular, some versions of Windows 2000 and the freely downloadable version of PGP provide only "client" functionality. You cannot use them as gateways with a subnet behind them. To get that functionality, you must upgrade to Windows 2000 server or the commercially available PGP products.

One tunnel plus advanced routing

Subject: Re: linux-ipsec: IPSec packets not entering tunnel?
   Date: Mon, 20 Nov 2000
   From: Justin Guyett <jfg@sonicity.com>

On Mon, 20 Nov 2000, Claudia Schmeing wrote:

> Right                                                         Left
>                      "home"                "office"
> 10.92.10.0/24 ---- 24.93.85.110 ========= 216.175.164.91 ---- 10.91.10.24/24
>
> I've created all four tunnels, and can ping to test each of them,
> *except* homegate-officenet.

I keep wondering why people create all four tunnels.  Why not route
traffic generated from home to 10.91.10.24/24 out ipsec0 with iproute2?
And 99% of the time you don't need to access "office" directly, which
means you can eliminate all but the subnet<->subnet connection.

> I keep wondering why people create all four tunnels.  Why not route
> traffic generated from home to 10.91.10.24/24 out ipsec0 with iproute2?

This is feasible, given some iproute2 attention to source addresses, but
it isn't something we've documented yet... (partly because we're still
making some attempt to support 2.0.xx kernels, which can't do this, but
mostly because we haven't caught up with it yet).

> And 99% of the time you don't need to access "office" directly, which
> means you can eliminate all but the subnet<->subnet connection.

Correct in principle, but people will keep trying to ping to or from the
gateways during testing, and sometimes they want to run services on the
gateway machines too.

Extruded Subnets

What we call extruded subnets are a special case of VPNs.

If your buddy has some unused IP addresses, in his subnet far off at the other side of the Internet, he can loan them to you... provided that the connection between you and him is fast enough to carry all the traffic between your machines and the rest of the Internet. In effect, he "extrudes" a part of his address space over the network to you, with your Internet traffic appearing to originate from behind his Internet gateway.

Suppose your friend has a.b.c.0/24 and wants to give you a.b.c.240/28. The initial situation is:

    subnet           gateway          Internet
  a.b.c.0/24    a.b.c.1    p.q.r.s

Of course it is quite normal for various smaller subnets to exist behind your friend's gateway. For example, your friend's company might have a.b.c.16/28=development, a.b.c.32/28=marketing and so on. The Internet neither knows not cares about this; it just delivers packets to the p.q.r.s and lets the gateway do whatever needs to be done from there.

What we want to do is take a subnet, perhaps a.b.c.240/28, out of your friend's physical location while still having your friend's gateway route to it. As far as the Internet is concerned, you remain behind that gateway.

    subnet           gateway          Internet       your gate  extruded

  a.b.c.0/24   a.b.c.1     p.q.r.s              d.e.f.g         a.b.c.240/28                

                           ========== tunnel ==========

The extruded addresses have to be a complete subnet.

In our example, the friend's security gateway is also his Internet gateway, but this is not necessary. As long as all traffic from the Internet to his addresses passes through the Internet gate, the security gate could be a machine behind that. The IG would need to route all traffic for the extruded subnet to the SG, and the SG could handle the rest.

First, configure your subnet using the extruded addresses. Your security gateway's interface to your subnet needs to have an extruded address (possibly using a Linux virtual interface , if it also has to have a different address). Your gateway needs to have a route to the extruded subnet, pointing to that interface. The other machines at your site need to have addresses in that subnet, and default routes pointing to your gateway.

If any of your friend's machines need to talk to the extruded subnet, they need to have a route for the extruded subnet, pointing at his gateway.

Then set up an IPsec subnet-to-subnet tunnel between your gateway and his, with your subnet specified as the extruded subnet, and his subnet specified as "0.0.0.0/0". Do it with manual keying first for testing, and then with automatic keying for production use.

The tunnel description should be:

conn extruded
        left=p.q.r.s
        leftsubnet=0.0.0.0/0
        right=d.e.f.g
        rightsubnet=a.b.c.0/28

If either side was doing firewalling for the extruded subnet before the IPsec connection is set up, ipsec_manual and ipsec_auto need to know about that (via the {left|right}firewall parameters) so that it can be overridden for the duration of the connection.

And it all just works. Your SG routes traffic for 0.0.0.0/0 -- that is, the whole Internet -- through the tunnel to his SG, which then sends it onward as if it came from his subnet. When traffic for the extruded subnet arrives at his SG, it gets sent through the tunnel to your SG, which passes it to the right machine.

Remember that when ipsec_manual or ipsec_auto takes a connection down, it does not undo the route it made for that connection. This lets you take a connection down and bring up a new one, or a modified version of the old one, without having to rebuild the route it uses and without any risk of packets which should use IPsec accidentally going out in the clear. Because the route always points into KLIPS, the packets will always go there. Because KLIPS temporarily has no idea what to do with them (no eroute for them), they will be discarded.

If you do want to take the route down, this is what the "unroute" operation in manual and auto is for. Just do an unroute after doing the down.

Note that the route for a connection may have replaced an existing non-IPsec route. Nothing in Linux FreeS/WAN will put that pre-IPsec route back. If you need it back, you have to create it with the route command.

Road Warrior with virtual IP address

Here is a mailing list message about another way to configure for road warrior support:

Subject: Re: linux-ipsec: understanding the vpn
   Date: Thu, 28 Oct 1999 10:43:22 -0400
   From: Irving Reid <irving@nevex.com>

>  local-------linux------internet------mobile
>  LAN        box                         user
>  ...

>  now when the mobile user connects to the linux box
>  it is given a virtual IP address, i have configured it to
>  be in the 10.x.x.x range. mobile user and linux box 
>  have a tunnel between them with these IP addresses.

>   Uptil this all is fine.

If it is possible to configure your mobile client software *not* to
use a virtual IP address, that will make your life easier. It is easier
to configure FreeS/WAN to use the actual address the mobile user gets
from its ISP.

Unfortunately, some Windows clients don't let you choose.

>  what i would like to know is that how does the mobile
>  user communicate with other computers on the local
>  LAN , of course with the vpn ?

>   what IP address should the local LAN 
>  computers have ? I guess their default gateway 
>  should be the linux box ? and does the linux box need
>  to be a 2 NIC card box or one is fine.

As someone else stated, yes, the Linux box would usually be the default
IP gateway for the local lan.

However...

If you mobile user has software that *must* use a virtual IP address,
the whole picture changes. Nobody has put much effort into getting
FreeS/WAN to play well in this environment, but here's a sketch of one
approach:

Local Lan 1.0.0.0/24
    |
    +- Linux FreeS/WAN 1.0.0.2
    |
    | 1.0.0.1
 Router
    | 2.0.0.1
    |
Internet
    |
    | 3.0.0.1
Mobile User
      Virtual Address: 1.0.0.3

Note that the Local Lan network (1.0.0.x) can be registered, routable
addresses.

Now, the Mobile User sets up an IPSec security association with the
Linux box (1.0.0.2); it should ESP encapsulate all traffic to the
network 1.0.0.x **EXCEPT** UDP port 500. 500/udp is required for the key
negotiation, which needs to work outside of the IPSec tunnel.

On the Linux side, there's a bunch of stuff you need to do by hand (for
now). FreeS/WAN should correctly handle setting up the IPSec SA and
routes, but I haven't tested it so this may not work...

The FreeS/WAN conn should look like:

conn mobile
        right=1.0.0.2
        rightsubnet=1.0.0.0/24
        rightnexthop=1.0.0.1
        left=0.0.0.0  # The infamous "road warrior"
        leftsubnet=1.0.0.3/32

Note that the left subnet contains *only* the remote host's virtual
address.

Hopefully the routing table on the FreeS/WAN box ends up looking like
this:

% netstat -rn
Kernel IP routing table
Destination     Gateway      Genmask         Flags   MSS Window  irtt Iface
1.0.0.0         0.0.0.0      255.255.255.0   U      1500 0          0 eth0
127.0.0.0       0.0.0.0      255.0.0.0       U      3584 0          0 lo
0.0.0.0         1.0.0.1      0.0.0.0         UG     1500 0          0 eth0
1.0.0.3         1.0.0.1      255.255.255.255 UG     1433 0          0 ipsec0

So, if anybody sends a packet for 1.0.0.3 to the Linux box, it should
get bundled up and sent through the tunnel. To get the packets for
1.0.0.3 to the Linux box in the first place, you need to use "proxy
ARP".

How this works is: when a host or router on the local Ethernet segment
wants to send a packet to 1.0.0.3, it sends out an Ethernet level
broadcast "ARP request". If 1.0.0.3 was on the local LAN, it would
reply, saying "send IP packets for 1.0.0.3 to my Ethernet address".

Instead, you need to set up the Linux box so that _it_ answers ARP
requests for 1.0.0.3, even though that isn't its IP address. That
convinces everyone else on the lan to send 1.0.0.3 packets to the Linux
box, where the usual FreeS/WAN processing and routing take over.

% arp -i eth0 -s 1.0.0.3 -D eth0 pub

This says, if you see an ARP request on interface eth0 asking for
1.0.0.3, respond with the Ethernet address of interface eth0.

Now, as I said at the very beginning, if it is *at all* possible to
configure your client *not* to use the virtual IP address, you can avoid
this whole mess.

Dynamic Network Interfaces

Sometimes you have to cope with a situation where the network interface(s) aren't all there at boot. The common example is notebooks with PCMCIA.

Basics

The key issue here is that the config setup section of the /etc/ipsec.conf configuration file lists the connection between ipsecN and hardware interfaces, in the interfaces= variable. At any time when ipsec setup start or ipsec setup restart is run this variable must correspond to the current real situation. More precisely, it must not mention any hardware interfaces which don't currently exist. The difficulty is that an ipsec setup start command is normally run at boot time so interfaces that are not up then are mis-handled.

Boot Time

Normally, an ipsec setup start is run at boot time. However, if the hardware situation at boot time is uncertain, one of two things must be done.

• One possibility is simply not to have IPsec brought up at boot time. To do this:

          chkconfig --level 2345 ipsec off

  That's for modern Red Hats or other Linuxes with chkconfig. Systems which lack this will require fiddling with symlinks in /etc/rc.d/rc?.d or the equivalent.
• Another possibility is to bring IPsec up with no interfaces, which is less aesthetically satisfying but simpler. Just put

          interfaces=

  in the configuration file. KLIPS and Pluto will be started, but won't do anything.

Change Time

When the hardware *is* in place, IPsec has to be made aware of it. Someday there may be a nice way to do this.

Right now, the way to do it is to fix the /etc/ipsec.conf file appropriately, so interfaces reflects the new situation, and then restart the IPsec subsystem. This does break any existing IPsec connections.

If IPsec wasn't brought up at boot time, do

        ipsec setup start

        ipsec setup restart

If some of the hardware is to be taken out, before doing that, amend the configuration file so interfaces no longer includes it, and do

        ipsec setup restart

Again, this breaks any existing connections.

Unencrypted tunnels

Sometimes you might want to create a tunnel without encryption. Often this is a bad idea, even if you have some data which need not be private. See this discussion.

The IPsec protocols provide two ways to do build such tunnels:

using ESP with null encryption

not supported by FreeS/WAN

using AH without ESP

supported for manually keyed connections

possible with explicit commands via ipsec_whack(8) (see this list message)

not supported in the ipsec_auto(8) scripts.

There are several ways to handle this.

• Just accept the overhead of double encryption. The site admins might choose this if any of the following apply:
  • policy says encrypt everything (usually, it should)
  • they don't entirely trust Alice and Bob (usually, if they don't have to, they shouldn't)
  • if they don't feel the saved cycles are worth the time they'd need to build a non-encrypted tunnel for Alice and Bob's packets (often, they aren't)
• Use a plain IP-in-IP tunnel. These are not well documented. A good starting point is in the Linux kernel source tree, in /usr/src/linux/drivers/net/README.tunnel.
• Use a manually-keyed AH-only tunnel.

Note that if Alice and Bob want end-to-end security, they must build a tunnel end-to-end between their machines or use some other end-to-end tool such as PGP or SSL that suits their data. The only question is whether the admins build some special unencrypted tunnel for those already-encrypted packets.

Installing FreeS/WAN

This document will teach you how to install Linux FreeS/WAN. If your distribution comes with Linux FreeS/WAN, we offer tips to get you started.

Requirements

To install FreeS/WAN you must:

• be running Linux with the 2.4 or 2.2 kernel series. See this kernel compatibility table.
• have root access to your Linux box
• choose the version of FreeS/WAN you wish to install based on mailing list reports or our updates page (coming soon)

Choose your install method

There are three basic ways to get FreeS/WAN onto your system:

• activating and testing a FreeS/WAN that shipped with your Linux distribution
• RPM install
• Install from source

FreeS/WAN ships with some Linuxes

FreeS/WAN comes with these distributions.

If you're running one of these, include FreeS/WAN in the choices you make during installation, or add it later using the distribution's tools.

FreeS/WAN may be altered...

Your distribution may have integrated extra features, such as Andreas Steffen's X.509 patch, into FreeS/WAN. They may also use their own startup script locations or directory names.

You might need to create an authentication keypair

If your FreeS/WAN came with your distribution, and it is pre-1.98, generate an RSA key pair for authentication.

As root type:

    ipsec newhostkey --output /etc/ipsec.secrets --hostname xy.example.com
    chmod 600 /etc/ipsec.secrets

where you replace xy.example.com with your machine's fully-qualified domain name. Generate some randomness, for example by wiggling your mouse, to speed the process.

The resulting ipsec.secrets looks like:

: RSA   {
        # RSA 2192 bits   xy.example.com   Tue Aug 20 17:42:19 2002
        # for signatures only, UNSAFE FOR ENCRYPTION
        #pubkey=0sAQOFppfeE3cC7wqJi...
        #IN KEY 0x4200 4 1 AQOFppfeE3cC7wqJ...
        # (0x4200 = auth-only host-level, 4 = IPSec, 1 = RSA)
        Modulus: 0x85a697de137702ef0...
        # everything after this point is secret
        PrivateExponent: 0x16466ea5033e807...
        Prime1: 0xdfb5003c8947b7cc88759065...
        Prime2: 0x98f199b9149fde11ec956c814...
        Exponent1: 0x9523557db0da7a885af90aee...
        Exponent2: 0x65f6667b63153eb69db8f300dbb...
        Coefficient: 0x90ad00415d3ca17bebff123413fc518...
        }
# do not change the indenting of that "}"

In the actual file, the strings are much longer.

Start and test FreeS/WAN

You can now start FreeS/WAN and test whether it's been successfully installed..

RPM install

These instructions are for a recent Red Hat with a stock Red Hat kernel. We know that Mandrake and SUSE also produce FreeS/WAN RPMs. If you're running either, install using your distribution's tools.

Download RPMs

Decide which functionality you need:

• standard FreeS/WAN RPMs from our FTP site
• or unofficial RPMs from freeswan.ca, which include Andreas Steffen's X.509 patch

Check your kernel version with

    uname -a

Get a kernel module which matches that version. For example:

    freeswan-module-1.99_2.4.18_3-0.i386.rpm

Note: These modules will only work on the Red Hat kernel they were built for, since they are very sensitive to small changes in the kernel.

Get FreeS/WAN utilities to match. For example:

    freeswan-1.99_2.4.18_3-0.i386.rpm

For freeswan.org RPMs: check signatures

While you're at our ftp site, grab the RPM signing key

    freeswan-rpmsign.asc

If you're running RedHat 8.x, import this key into the RPM database:

    rpm --import freeswan-rpmsign.asc

For RedHat 7.x systems, you'll need to add it to your PGP keyring:

    pgp -ka freeswan-rpmsign.asc

Check the signatures on both RPMs using:

    rpm --checksig freeswan-module-1.99_2.4.18_3-0.i386.rpm
    rpm --checksig freeswan-1.99_2.4.18_3-0.i386.rpm

You should see:

    freeswan-module-1.99_2.4.18_3-0.i386.rpm: pgp md5 OK
    freeswan-1.99_2.4.18_3-0.i386.rpm: pgp md5 OK

Install the RPMs

Become root:

    su

Install your RPMs with:

    rpm -ivh freeswan*

Start and Test FreeS/WAN

Now, start FreeS/WAN and test your install.

Install from Source

Decide what functionality you need

Your choices are:

• standard FreeS/WAN
• standard FreeS/WAN plus any of these user-supported patches
• an unofficial FreeS/WAN pre-patched with several of the above. Provides additional algorithms, X.509, and SA deletion

Download FreeS/WAN

Download the source tarball you've chosen, along with any patches.

For freeswan.org source: check its signature

While you're at our ftp site, get our RPM signing key

    freeswan-sigkey.asc

and add it to your PGP keyring:

    pgp -ka freeswan-sigkey.asc

Check the signature using:

    pgp freeswan-1.99.tar.gz.sig freeswan-1.99.tar.gz

You should see something like:

    Good signature from user "Linux FreeS/WAN Software Team (build@freeswan.org)".
    Signature made 2002/06/26 21:04 GMT using 2047-bit key, key ID 46EAFCE1

Untar, unzip

As root, unpack your FreeS/WAN source into /usr/src.

    su
    mv freeswan-1.99.tar.gz /usr/src
    cd /usr/src
    tar -xzf freeswan-1.99.tar.gz

Patch if desired

Now's the time to add any patches. The contributor may have special instructions, or you may simply use the patch command.

... and Make

Either make FreeS/WAN as a module...

Change to your new FreeS/WAN directory:

    cd /usr/src/freeswan-1.99

Make the FreeS/WAN module:

    make oldmod

Install it:

    make minstall

You can directly start FreeS/WAN and test your install.

...or statically linked

Make FreeS/WAN using your old kernel settings:

    make oldgo

Install it:

    make kinstall

Reboot your system and test your install.

Start FreeS/WAN and test your install

Bring FreeS/WAN up with:

    service ipsec start

This is not necessary if you've rebooted.

Test your install

To check that you have a successful install, run:

    ipsec verify

You should see at least:

    Checking your system to see if IPsec got installed and started correctly
    Version check and ipsec on-path                             [OK]
    Checking for KLIPS support in kernel                        [OK]
    Checking for RSA private key (/etc/ipsec.secrets)           [OK]
    Checking that pluto is running                              [OK]

If any of these first four checks fails, see our troubleshooting guide.

Making FreeS/WAN play well with others

There are several things on your system that might interfere with FreeS/WAN, and now's a good time to check these:

• Firewalling. You need to allow UDP 500 through your firewall, plus ESP (protocol 50) and AH (protocol 51). For more information, see our updated firewalls document (coming soon).
• rp_filter. This should be set to 0 on the interfaces FreeS/WAN uses. A command like:

      echo 0 > /proc/sys/net/ipv4/conf/eth0/rp_filter

  may be necessary.
• Do not NAT the packets you will be tunneling.
• ....to be expanded...

Configure for your needs

You'll need to configure FreeS/WAN for your local site. Have a look at our opportunism quickstart guide to see if that easy method is right for your needs. Or, see how to configure a network-to-network or Road Warrior style VPN.

How to configure FreeS/WAN

This page will teach you how to configure a simple network-to-network link or a Road Warrior connection between two Linux FreeS/WAN boxes.

The network-to-network setup allows you to connect two office networks into one Virtual Private Network, while the Road Warrior connection secures a laptop's telecommute to work. Our examples also show the basic procedure on the Linux FreeS/WAN side where another IPsec peer is in play.

Shortcut to net-to-net.
Shortcut to Road Warrior.

Requirements

To configure the network-to-network connection you must have:

• two Linux gateways with static IPs
• a network behind each gate. Networks must have non-overlapping IP ranges.
• Linux FreeS/WAN installed on both gateways
• tcpdump on the local gate, to test the connection

For the Road Warrior you need:

• one Linux box with a static IP
• a Linux laptop with a dynamic IP
• Linux FreeS/WAN installed on both
• for testing, tcpdump on your gateway or laptop

Net-to-Net connection

Gather information

For each gateway, compile the following information:

• gateway IP
• IP range of the subnet you will be protecting. This doesn't have to be your whole physical subnet.
• a name by which that gateway can identify itself for IPsec negotiations. Its form is a Fully Qualified Domain Name preceded by an @ sign, ie. @xy.example.com.
  It does not need to be within a domain that you own. It can be a made-up name.

Get your leftrsasigkey

On your local Linux FreeS/WAN gateway, print your IPsec public key:

    ipsec showhostkey --left

The output should look like this (with the key shortened for easy reading):

    # RSA 2048 bits   xy.example.com   Fri Apr 26 15:01:41 2002
    leftrsasigkey=0sAQOnwiBPt...

Don't have a key? See these instructions.

...and your rightrsasigkey

Get a console on the remote side:

    ssh2 ab.example.com

In that window, type:

    ipsec showhostkey --right

You'll see something like:

    # RSA 2192 bits   ab.example.com   Thu May 16 15:26:20 2002
    rightrsasigkey=0sAQOqH55O...

Edit /etc/ipsec.conf

Back on the local gate, copy our template to /etc/ipsec.conf . (on Mandrake, /etc/freeswan/ipsec.conf). Substitute the information you've gathered for our example data.

conn net-to-net
    left=1.2.3.4                   # Local vitals
    leftsubnet=42.42.42.0/24       # 
    leftid=@xy.example.com         #   
    leftrsasigkey=0s1LgR7/oUM...   #
    leftnexthop=%defaultroute      # correct in many situations 
    right=5.6.7.8                  # Remote vitals
    rightsubnet=42.42.42.1/24      #
    rightid=@ab.example.com        # 
    rightrsasigkey=0sAQOqH55O...   #
    rightnexthop=%defaultroute     # correct in many situations
    auto=add                       # authorizes but doesn't start this 
                                   # connection at startup

Copy conn net-to-net to the remote-side /etc/ipsec.conf. If you've made no other modifications to either ipsec.conf, simply:

    scp2 ipsec.conf root@ab.example.com:/etc/ipsec.conf

Start your connection

Locally, type:

    ipsec auto --start net-to-net

You should see:

    104 "net-net" #223: STATE_MAIN_I1: initiate
    106 "net-net" #223: STATE_MAIN_I2: sent MI2, expecting MR2
    108 "net-net" #223: STATE_MAIN_I3: sent MI3, expecting MR3
    004 "net-net" #223: STATE_MAIN_I4: ISAKMP SA established
    112 "net-net" #224: STATE_QUICK_I1: initiate
    004 "net-net" #224: STATE_QUICK_I2: sent QI2, IPsec SA established

The important thing is IPsec SA established. If you're unsuccessful, see our troubleshooting tips.

Do not MASQ or NAT packets to be tunneled

If you are masquerading or NATting packets on either gateway, you must now exempt the packets you wish to tunnel from this treatment. If you have a rule like:

    iptables -A FORWARD -s 42.42.42.0/255.255.255.0 -j MASQ

change it to something like:

    iptables -A FORWARD -s 42.42.42.0/255.255.255.0 -d ! 42.42.42.1/255.255.255.0 -j MASQ

This may be necessary on both gateways.

Test your connection

Sit at one of your local subnet nodes (not the gateway), and ping a subnet node on the other (again, not the gateway).

    ping fileserver.toledo.example.com

While still pinging, go to the local gateway and snoop your outgoing interface, for example:

    tcpdump -i ppp0

You want to see ESP (Encapsulating Security Payload) packets moving back and forth between the two gateways at the same frequency as your pings:

    19:16:32.046220 1.2.3.4 > 5.6.7.8: ESP(spi=0x3be6c4dc,seq=0x3)
    19:16:32.085630 5.6.7.8 > 1.2.3.4: ESP(spi=0x5fdd1cf8,seq=0x6)

If you see this, congratulations are in order! You have a tunnel which will protect any IP data from one subnet to the other, as it passes between the two gates. If not, go and troubleshoot.

Note: your new tunnel protects only net-net traffic, not gateway-gateway, or gateway-subnet. If you need this (for example, if machines on one net need to securely contact a fileserver on the IPsec gateway), you'll need to create extra connections.

Finishing touches

Now that your connection works, name it something sensible, like:

conn winstonnet-toledonet

To have the tunnel come up on-boot, replace

    auto=add

with:

    auto=start

Copy these changes to the other side, for example:

    scp2 ipsec.conf root@ab.example.com:/etc/ipsec.conf

Enjoy!

Road Warrior Configuration

Gather information

You'll need to know:

• the gateway's static IP
• the IP range of the subnet behind that gateway
• a name by which each side can identify itself for IPsec negotiations. Its form is a Fully Qualified Domain Name preceded by an @ sign, ie. @road.example.com.
  It does not need to be within a domain that you own. It can be a made-up name.

Get your leftrsasigkey...

On your laptop, print your IPsec public key:

    ipsec showhostkey --left

The output should look like this (with the key shortened for easy reading):

    # RSA 2192 bits   road.example.com   Sun Jun  9 02:45:02 2002
    leftrsasigkey=0sAQPIPN9uI...

Don't have a key? See these instructions.

...and your rightrsasigkey

Get a console on the gateway:

    ssh2 xy.example.com

View the gateway's public key with:

    ipsec showhostkey --right

This will yield something like

    # RSA 2048 bits   xy.example.com   Fri Apr 26 15:01:41 2002
    rightrsasigkey=0sAQOnwiBPt...

Customize /etc/ipsec.conf

On your laptop, copy this template to /etc/ipsec.conf. (on Mandrake, /etc/freeswan/ipsec.conf). Substitute the information you've gathered for our example data.

conn road
    left=%defaultroute             # Picks up our dynamic IP 
    leftnexthop=%defaultroute      # 
    leftid=@road.example.com       # Local information
    leftrsasigkey=0sAQPIPN9uI...   #
    right=1.2.3.4                  # Remote information
    rightsubnet=42.42.42.0/24      #
    rightid=@xy.example.com        # 
    rightrsasigkey=0sAQOnwiBPt...  #
    auto=add                       # authorizes but doesn't start this
                                   # connection at startup

The template for the gateway is different. Notice how it reverses left and right, in keeping with our convention that Left is Local, Right Remote. Be sure to switch your rsasigkeys in keeping with this.

    ssh2 xy.example.com
    vi /etc/ipsec.conf

and add:

conn road
    left=1.2.3.4                   # Gateway's information
    leftid=@xy.example.com         #
    leftsubnet=42.42.42.0/24       #
    leftrsasigkey=0sAQOnwiBPt...   #
    rightnexthop=%defaultroute     # correct in many situations
    right=%any                     # Wildcard: we don't know the laptop's IP
    rightid=@road.example.com      #
    rightrsasigkey=0sAQPIPN9uI...  #
    auto=add                       # authorizes but doesn't start this
                                   # connection at startup

Start your connection

You must start the connection from the Road Warrior side. On your laptop, type:

    ipsec auto --start net-to-net

You should see:

104 "net-net" #223: STATE_MAIN_I1: initiate
106 "road" #301: STATE_MAIN_I2: sent MI2, expecting MR2
108 "road" #301: STATE_MAIN_I3: sent MI3, expecting MR3
004 "road" #301: STATE_MAIN_I4: ISAKMP SA established
112 "road" #302: STATE_QUICK_I1: initiate
004 "road" #302: STATE_QUICK_I2: sent QI2, IPsec SA established

Look for IPsec SA established. If you're unsuccessful, see our troubleshooting tips.

Do not MASQ or NAT packets to be tunneled

If you are masquerading or NATting packets on your office gateway, you must now exempt the packets you wish to tunnel from this treatment. If you have a rule like:

    iptables -A FORWARD -s 42.42.42.0/255.255.255.0 -j MASQ

change it to something like:

    iptables -A FORWARD -s 42.42.42.0/255.255.255.0 -d ! 42.42.42.1/255.255.255.0 -j MASQ

Test your connection

From your laptop, ping a subnet node behind the remote gateway. Do not choose the gateway itself for this test.

    ping ns.winston.example.com

Snoop the packets exiting the laptop, with a command like:

    tcpdump -i wlan0

You have success if you see (Encapsulating Security Payload) packets travelling in both directions:

22:57:31.507992 7.7.7.7 > 1.2.3.4: ESP(spi=0x3a0cdf1e,seq=0x1a)
22:57:31.624772 1.2.3.4 > 7.7.7.7: ESP(spi=0x5fdd1d0e,seq=0x1c)

If you do, great! Traffic between your Road Warrior and the net behind your gateway is protected. If not, see our troubleshooting hints.

Your new tunnel protects only traffic addressed to the net, not to the IPsec gateway itself. If you need the latter, you'll want to make an extra tunnel..

Finishing touches

On both ends, name your connection wisely, like:

conn mike-to-office

On the laptop only, replace

    auto=add

with:

    auto=start

so that you'll be connected on-boot.

Happy telecommuting!

What next?

Using the principles illustrated here, you can try variations such as:

• a telecommuter with a static IP
• a road warrior with a subnet behind it

Or, look at some of our more complex configuration examples..

Linux FreeS/WAN background

This section discusses a number of issues which have three things in common:

• They are not specifically FreeS/WAN problems
• You may have to understand them to get FreeS/WAN working right
• They are not simple questions

Grouping them here lets us provide the explanations some users will need without unduly complicating the main text.

The explanations here are intended to be adequate for FreeS/WAN purposes (please comment to the users mailing list if you don't find them so), but they are not trying to be complete or definitive. If you need more information, see the references provided in each section.

Some DNS background

Opportunistic encryption requires that the gateway systems be able to fetch public keys, and other IPsec-related information, from each other's DNS (Domain Name Service) records.

DNS is a distributed database that maps names to IP addresses and vice versa.

Much good reference material is available for DNS, including:

• the DNS HowTo
• the standard DNS reference book
• Linux Network Administrator's Guide
• BIND overview
• BIND 9 Administrator's Reference

We give only a brief overview here, intended to help you use DNS for FreeS/WAN purposes.

Forward and reverse maps

Although the implementation is distributed, it is often useful to speak of DNS as if it were just two enormous tables:

• the forward map: look up a name, get an IP address
• the reverse map: look up an IP address, get a name

Both maps can optionally contain additional data. For opportunistic encryption, we insert the data need for IPsec authentication.

A system named gateway.example.com with IP address 10.20.30.40 should have at least two DNS records, one in each map:

gateway.example.com. IN A 10.20.30.40

used to look up the name and get an IP address

40.30.20.10.in-addr.arpa. IN PTR gateway.example.com.

used for reverse lookups, looking up an address to get the associated name. Notice that the digits here are in reverse order; the actual address is 10.20.30.40 but we use 40.30.20.10 here.

Hierarchy and delegation

For both maps there is a hierarchy of DNS servers and a system of delegating authority so that, for example:

• the DNS administrator for example.com can create entries of the form name.example.com
• the example.com admin cannot create an entry for counterexample.com; only someone with authority for .com can do that
• an admin might have authority for 20.10.in-addr.arpa.
• in either map, authority can be delegated
  • the example.com admin could give you authority for westcoast.example.com
  • the 20.10.in-addr.arpa admin could give you authority for 30.20.10.in-addr.arpa

DNS zones are the units of delegation. There is a hierarchy of zones.

Syntax of DNS records

Returning to the example records:

        gateway.example.com. IN A 10.20.30.40
        40.30.20.10.in-addr.arpa. IN PTR gateway.example.com.

some syntactic details are:

• the IN indicates that these records are for In ternet addresses
• The final periods in '.com.' and '.arpa.' are required. They indicate the root of the domain name system.

The capitalised strings after IN indicate the type of record. Possible types include:

• Address, for forward lookups
• PoinTeR, for reverse lookups
• Canonical NAME, records to support aliasing, multiple names for one address
• Mail eXchange, used in mail routing
• SIGnature, used in secure DNS
• KEY, used in secure DNS
• TeXT, a multi-purpose record type

To set up for opportunistic encryption, you add some KEY and TXT records to your DNS data. Details are in our quickstart document.

Cacheing, TTL and propagation delay

DNS information is extensively cached. With no caching, a lookup by your system of "www.freeswan.org" might involve:

• your system asks your nameserver for "www.freeswan.org"
• local nameserver asks root server about ".org", gets reply
• local nameserver asks .org nameserver about "freeswan.org", gets reply
• local nameserver asks freeswan.org nameserver about "www.freeswan.org", gets reply

However, this can be a bit inefficient. For example, if you are in the Phillipines, the closest a root server is in Japan. That might send you to a .org server in the US, and then to freeswan.org in Holland. If everyone did all those lookups every time they clicked on a web link, the net would grind to a halt.

Nameservers therefore cache information they look up. When you click on another link at www.freeswan.org, your local nameserver has the IP address for that server in its cache, and no further lookups are required.

Intermediate results are also cached. If you next go to lists.freeswan.org, your nameserver can just ask the freeswan.org nameserver for that address; it does not need to query the root or .org nameservers because it has a cached address for the freeswan.org zone server.

Of course, like any cacheing mechanism, this can create problems of consistency. What if the administrator for freeswan.org changes the IP address, or the authentication key, for www.freeswan.org? If you use old information from the cache, you may get it wrong. On the other hand, you cannot afford to look up fresh information every time. Nor can you expect the freeswan.org server to notify you; that isn't in the protocols.

The solution that is in the protocols is fairly simple. Cacheable records are marked with Time To Live (TTL) information. When the time expires, the caching server discards the record. The next time someone asks for it, the server fetches a fresh copy. Of course, a server may also discard records before their TTL expires if it is running out of cache space.

This implies that there will be some delay before the new version of a changed record propagates around the net. Until the TTLs on all copies of the old record expire, some users will see it because that is what is in their cache. Other users may see the new record immediately because they don't have an old one cached.

Problems with packet fragmentation

It seems, from mailing list reports, to be moderately common for problems to crop up in which small packets pass through the IPsec tunnels just fine but larger packets fail.

These problems are caused by various devices along the way mis-handling either packet fragments or path MTU discovery.

IPsec makes packets larger by adding an ESP or AH header. This can tickle assorted bugs in fragment handling in routers and firewalls, or in path MTU discovery mechanisms, and cause a variety of symptoms which are both annoying and, often, quite hard to diagnose.

An explanation from project technical lead Henry Spencer:

The problem is IP fragmentation; more precisely, the problem is that the
second, third, etc. fragments of an IP packet are often difficult for
filtering mechanisms to classify.

Routers cannot rely on reassembling the packet, or remembering what was in
earlier fragments, because the fragments may be out of order or may even
follow different routes.  So any general, worst-case filtering decision
pretty much has to be made on each fragment independently.  (If the router
knows that it is the only route to the destination, so all fragments
*must* pass through it, reassembly would be possible... but most routers
don't want to bother with the complications of that.)

All fragments carry roughly the original IP header, but any higher-level
header is (for IP purposes) just the first part of the packet data... so
only the first fragment carries that.  So, for example, on examining the
second fragment of a TCP packet, you could tell that it's TCP, but not
what port number it is destined for -- that information is in the TCP
header, which appears in the first fragment only. 

The result of this classification difficulty is that stupid routers and
over-paranoid firewalls may just throw fragments away.  To get through
them, you must reduce your MTU enough that fragmentation will not occur.
(In some cases, they might be willing to attempt reassembly, but have very
limited resources to devote to it, meaning that packets must be small and
fragments few in number, leading to the same conclusion:  smaller MTU.)

In addition to the problem Henry describes, you may also have trouble with path MTU discovery.

By default, FreeS/WAN uses a large MTU for the ipsec device. This avoids some problems, but may complicate others. Here's an explanation from Claudia:

Here are a couple of pieces of background information. Apologies if you
have seen these already. An excerpt from one of my old posts:

    An MTU of 16260 on ipsec0 is usual. The IPSec device defaults to this 
    high MTU so that it does not fragment incoming packets before encryption 
    and encapsulation. If after IPSec processing packets are larger than 1500,
    [ie. the mtu of eth0] then eth0 will fragment them. 

    Adding IPSec headers adds a certain number of bytes to each packet. 
    The MTU of the IPSec interface refers to the maximum size of the packet
    before the IPSec headers are added. In some cases, people find it helpful 
    to set ipsec0's MTU to 1500-(IPSec header size), which IIRC is about 1430.

    That way, the resulting encapsulated packets don't exceed 1500. On most 
    networks, packets less than 1500 will not need to be fragmented.

and... (from Henry Spencer)

    The way it *ought* to work is that the MTU advertised by the ipsecN
    interface should be that of the underlying hardware interface, less a
    pinch for the extra headers needed. 

    Unfortunately, in certain situations this breaks many applications.
    There is a widespread implicit assumption that the smallest MTUs are 
    at the ends of paths, not in the middle, and another that MTUs are 
    never less than 1500.  A lot of code is unprepared to handle paths 
    where there is an "interior minimum" in the MTU, especially when it's 
    less than 1500. So we advertise a big MTU and just let the resulting 
    big packets fragment.

This usually works, but we do get bitten in cases where some intermediate
point can't handle all that fragmentation.  We can't win on this one.

The MTU can be changed with an overridemtu= statement in the config setup section of ipsec.conf.5.

For a discussion of MTU issues and some possible solutions using Linux advanced routing facilities, see the Linux 2.4 Advanced Routing HOWTO.

Network address translation (NAT)

Network Address T ranslation is a service provided by some gateway machines. Calling it NAPT (adding the word Port) would be more precise, but we will follow the widespread usage.

A gateway doing NAT rewrites the headers of packets it is forwarding, changing one or more of:

• source address
• source port
• destination address
• destination port

On Linux 2.4, NAT services are provided by the netfilter(8) firewall code. There are several Netfilter HowTos including one on NAT.

For older versions of Linux, this was referred to as "IP masquerade" and different tools were used. See this resource page.

Putting an IPsec gateway behind a NAT gateway is not recommended. See our firewalls document.

NAT to non-routable addresses

The most common application of NAT uses private non-routable addresses.

Often a home or small office network will have:

• one connection to the Internet
• one assigned publicly visible IP address
• several machines that all need access to the net

Of course this poses a problem since several machines cannot use one address. The best solution might be to obtain more addresses, but often this is impractical or uneconomical.

A common solution is to have:

• non-routable addresses on the local network
• the gateway machine doing NAT
• all packets going outside the LAN rewritten to have the gateway as their source address

The client machines are set up with reserved non-routable IP addresses defined in RFC 1918. The masquerading gateway, the machine with the actual link to the Internet, rewrites packet headers so that all packets going onto the Internet appear to come from one IP address, that of its Internet interface. It then gets all the replies, does some table lookups and more header rewriting, and delivers the replies to the appropriate client machines.

As far as anyone else on the Internet is concerned, the systems behind the gateway are completely hidden. Only one machine with one IP address is visible.

For IPsec on such a gateway, you can entirely ignore the NAT in:

• ipsec.conf(5)
• firewall rules affecting your Internet-side interface

Those can be set up exactly as they would be if your gateway had no other systems behind it.

You do, however, have to take account of the NAT in firewall rules which affect packet forwarding.

NAT to routable addresses

NAT to routable addresses is also possible, but is less common and may make for rather tricky routing problems. We will not discuss it here. See the Netfilter HowTos.

FreeS/WAN script examples

So far it has only one entry.

Poltorak's Firewall script

From: Poltorak Serguei <poltorak@dataforce.net>
Subject: [Users] Using FreeS/WAN
Date: Tue, 16 Oct 2001

Hello.

I'm using FreeS/WAN IPsec for half a year. I learned a lot of things about
it and I think it would be interesting for someone to see the result of my
experiments and usage of FreeS/WAN. If you find a mistake in this
file, please e-mail me. And excuse me for my english... I'm learning.. :)

I'll talk about vary simple configuration:

addresses prefix = 192.168

    lan1          sgw1     .0.0/24 (Internet)       sgw2            lan2
  .1.0/24---[ .1.1 ; .0.1 ]===================[ .0.10 ; . 2.10 ]---.2.0/24


We need to let lan1 see lan2 across Internet like it is behind sgw1. The
same for lan2. And we need to do IPX bridge for Novel Clients and NDS
synchronization.

my config:
------------------- ipsec.conf -------------------
conn lan1-lan2
        type=tunnel
        compress=yes
        #-------------------
        left=192.168.0.1
        leftsubnet=192.168.1.0/24
        #-------------------
        right=192.168.0.10
        rightsubnet=192.168.2.0/24
        #-------------------
        auth=esp
        authby=secret
--------------- end of ipsec.conf ----------------

ping .2.x from .1.y   (y != 1)
It works?? Fine. Let's continue...

Why y != 1 ?? Because kernel of sgw1 have 2 IP addresses and it will choose
the first IP (which is used to go to Internet) .0.1 and the packet won't go
through IPsec tunnel :(  But if do ping on .1.1 kernel will respond from
that address (.1.1) and the packet will be tunneled. The same problem occurred then
.2.x sends a packet to .1.2 which is down at the moment. What happens? .1.1
sends ARP requesting .1.2... after 3 tries it send to .2.x an destunreach,
but from his "natural" IP or .0.1 . So the error message won't be delivered!
It's a big problem...

Resolution... One can manipulate with ipsec0 or ipsec0:0 to solve the
problem (if ipsec0 has .1.1 kernel will send packets correctly), but there
are powerful and elegant iproute2 :) We simply need to change source address
of packet that goes to other secure lan. This is done with

ip route replace 192.168.2.0/24 via 192.168.0.10 dev ipsec0 src 192.168.1.1

Cool!! Now it works!!

The second step. We want install firewall on sgw1 and sgw2. Encryption of 
traffic without security isn't a good idea. I don't use {left|right}firewall, 
because I'm running firewall from init scripts.

We want IPsec data between lan1-lan2, some ICMP errors (destination
unreachable, TTL exceeded, parameter problem and source quench), replying on 
pings from both lans and Internet, ipxtunnel data for IPX and of course SSH
between sgw1 and sgw2 and from/to one specified host.

I'm using ipchains. With iptables there are some changes.

---------------- rc.firewall ---------------------
#!/bin/sh
#
# Firewall for IPsec lan1-lan2
#

IPC=/sbin/ipchains
ANY=0.0.0.0/0

# left
SGW1_EXT=192.168.0.1
SGW1_INT=192.168.1.1
LAN1=192.168.1.0/24

# right
SGW2_EXT=192.168.0.10
SGW2_INT=192.168.2.10
LAN2=192.168.2.0/24

# SSH from and to this host
SSH_PEER_HOST=_SOME_HOST_

# this is for left. exchange these values for right.
MY_EXT=$SGW1_EXT
MY_INT=$SGW1_INT
PEER_EXT=$SGW2_EXT
PEER_INT=$SGW2_INT
INT_IF=eth1
EXT_IF=eth0
IPSEC_IF=ipsec0
MY_LAN=$LAN1
PEER_LAN=$LAN2

$IPC -F
$IPC -P input DENY
$IPC -P forward DENY
$IPC -P output DENY

# Loopback traffic
$IPC -A input -i lo -j ACCEPT
$IPC -A output -i lo -j ACCEPT

# for IPsec SGW1-SGW2
## IKE
$IPC -A input -p udp -s $PEER_EXT 500 -d $MY_EXT 500 -i $EXT_IF -j ACCEPT
$IPC -A output -p udp -s $MY_EXT 500 -d $PEER_EXT 500 -i $EXT_IF -j ACCEPT
## ESP
$IPC -A input -p 50 -s $PEER_EXT -d $MY_EXT -i $EXT_IF -j ACCEPT
### we don't need this line ### $IPC -A output -p 50 -s $MY_EXT -d $PEER_EXT -i $EXT_IF -j ACCEPT
## forward LAN1-LAN2
$IPC -A forward -s $MY_LAN -d $PEER_LAN -i $IPSEC_IF -j ACCEPT
$IPC -A forward -s $PEER_LAN -d $MY_LAN -i $INT_IF -j ACCEPT
$IPC -A output -s $PEER_LAN -d $MY_LAN -i $INT_IF -j ACCEPT
$IPC -A input -s $PEER_LAN -d $MY_LAN -i $IPSEC_IF -j ACCEPT
$IPC -A input -s $MY_LAN -d $PEER_LAN -i $INT_IF -j ACCEPT
$IPC -A output -s $MY_LAN -d $PEER_LAN -i $IPSEC_IF -j ACCEPT

# ICMP
#
## Dest unreachable
### from/to Internet
$IPC -A input -p icmp --icmp-type destination-unreachable -s $ANY -d $MY_EXT -i $EXT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type destination-unreachable -s $MY_EXT -d $ANY -i $EXT_IF -j ACCEPT
### from/to Lan
$IPC -A input -p icmp --icmp-type destination-unreachable -s $ANY -d $MY_INT -i $INT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type destination-unreachable -s $MY_INT -d $ANY -i $INT_IF -j ACCEPT
### from/to Peer Lan
$IPC -A input -p icmp --icmp-type destination-unreachable -s $ANY -d $MY_INT -i $IPSEC_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type destination-unreachable -s $MY_INT -d $ANY -i $IPSEC_IF -j ACCEPT
#
## Source quench
### from/to Internet
$IPC -A input -p icmp --icmp-type source-quench -s $ANY -d $MY_EXT -i $EXT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type source-quench -s $MY_EXT -d $ANY -i $EXT_IF -j ACCEPT
### from/to Lan
$IPC -A input -p icmp --icmp-type source-quench -s $ANY -d $MY_INT -i $INT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type source-quench -s $MY_INT -d $ANY -i $INT_IF -j ACCEPT
### from/to Peer Lan
$IPC -A input -p icmp --icmp-type source-quench -s $ANY -d $MY_INT -i $IPSEC_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type source-quench -s $MY_INT -d $ANY -i $IPSEC_IF -j ACCEPT
#
## Parameter problem
### from/to Internet
$IPC -A input -p icmp --icmp-type parameter-problem -s $ANY -d $MY_EXT -i $EXT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type parameter-problem -s $MY_EXT -d $ANY -i $EXT_IF -j ACCEPT
### from/to Lan
$IPC -A input -p icmp --icmp-type parameter-problem -s $ANY -d $MY_INT -i $INT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type parameter-problem -s $MY_INT -d $ANY -i $INT_IF -j ACCEPT
### from/to Peer Lan
$IPC -A input -p icmp --icmp-type parameter-problem -s $ANY -d $MY_INT -i $IPSEC_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type parameter-problem -s $MY_INT -d $ANY -i $IPSEC_IF -j ACCEPT
#
## Time To Live exceeded
### from/to Internet
$IPC -A input -p icmp --icmp-type time-exceeded -s $ANY -d $MY_EXT -i $EXT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type time-exceeded -s $MY_EXT -d $ANY -i $EXT_IF -j ACCEPT
### to Lan
$IPC -A input -p icmp --icmp-type time-exceeded -s $ANY -d $MY_INT -i $INT_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type time-exceeded -s $MY_INT -d $ANY -i $INT_IF -j ACCEPT
### to Peer Lan
$IPC -A input -p icmp --icmp-type time-exceeded -s $ANY -d $MY_INT -i $IPSEC_IF -j ACCEPT
$IPC -A output -p icmp --icmp-type time-exceeded -s $MY_INT -d $ANY -i $IPSEC_IF -j ACCEPT

# ICMP PINGs
## from Internet
$IPC -A input -p icmp -s $ANY -d $MY_EXT --icmp-type echo-request  -i $EXT_IF -j ACCEPT
$IPC -A output -p icmp -s $MY_EXT -d $ANY --icmp-type echo-reply  -i $EXT_IF -j ACCEPT
## from LAN
$IPC -A input -p icmp -s $ANY -d $MY_INT --icmp-type echo-request -i $INT_IF -j ACCEPT
$IPC -A output -p icmp -s $MY_INT -d $ANY --icmp-type echo-reply  -i $INT_IF -j ACCEPT
## from Peer LAN
$IPC -A input -p icmp -s $ANY -d $MY_INT --icmp-type echo-request -i $IPSEC_IF -j ACCEPT
$IPC -A output -p icmp -s $MY_INT -d $ANY --icmp-type echo-reply  -i $IPSEC_IF -j ACCEPT

# SSH
## from SSH_PEER_HOST
$IPC -A input -p tcp -s $SSH_PEER_HOST -d $MY_EXT 22 -i $EXT_IF -j ACCEPT
$IPC -A output -p tcp \! -y -s $MY_EXT 22 -d $SSH_PEER_HOST -i $EXT_IF -j ACCEPT
## to SSH_PEER_HOST
$IPC -A input -p tcp \! -y -s $SSH_PEER_HOST 22 -d $MY_EXT -i $EXT_IF -j ACCEPT
$IPC -A output -p tcp -s $MY_EXT -d $SSH_PEER_HOST 22 -i $EXT_IF -j ACCEPT
## from PEER
$IPC -A input -p tcp -s $PEER_EXT -d $MY_EXT 22 -i $EXT_IF -j ACCEPT
$IPC -A output -p tcp \! -y -s $MY_EXT 22 -d $PEER_EXT -i $EXT_IF -j ACCEPT
## to PEER
$IPC -A input -p tcp \! -y -s $PEER_EXT 22 -d $MY_EXT -i $EXT_IF -j ACCEPT
$IPC -A output -p tcp -s $MY_EXT -d $PEER_EXT 22 -i $EXT_IF -j ACCEPT

# ipxtunnel
$IPC -A input -p udp -s $PEER_INT 2005 -d $MY_INT 2005 -i $IPSEC_IF -j ACCEPT
$IPC -A output -p udp -s $MY_INT 2005 -d $PEER_INT 2005 -i $IPSEC_IF -j ACCEPT

---------------- end of rc.firewall ----------------------

To understand this we need to look on this scheme:

           ++-----------------------<----------------------------+
           || ipsec0                                             |
           \/                                                    |
 eth0  +--------+    /---------/ yes  /---------/ yes +-----------------------+
------>| INPUT  |-->/ ?local? /----->/ ?IPsec? /----->| decrypt decapsulate |
 eth1  +--------+  /---------/      /---------/       +-----------------------+
                       || no            || no
                       \/               \/
                  +----------+      +---------+        +-------+
                  | routing  |      |  local  |        | local |
                  | decision |      | deliver |        | send  |
                  +----------+      +---------+        +-------+
                       ||                                 ||
                       \/                                 \/
                   +---------+                       +----------+
                   | forward |                       | routing  |
                   +---------+                       | decision |
                       ||                            +----------+
                       ||                                  ||
                       ++----------------<-----------------++
                       ||
                       \/
                   +--------+ eth0
                   | OUTPUT | eth1
                   +--------+ ipsec0
                       ||
                       \/
                   /---------/ yes +-----------------------+
                  / ?IPsec? /----->| encrypt encapsulate |
                 /---------/       +-----------------------+
                      || no                    ||
                      ||                       ||
                      ||                       \/   eth0, eth1
                      ++-----------------------++-------------->

This explain how a packet traverse TCP/IP stack in IPsec capable kernel.

FIX ME, please, if there are any errors

Test the new firewall now.


Now about IPX. I tried 3 programs for tunneling IPX: tipxd, SIB and ipxtunnel

tipxd didn't send packets.. :(
SIB and ipxtunnel worked fine :)
With ipxtunnel there was a little problem. In sources there are an error.

--------------------- in main.c ------------------------
<       bytes += p.len;
---
>       bytes += len;
--------------------------------------------------------

After this FIX everything goes right...

------------------- /etc/ipxtunnel.conf ----------------
port    2005
remote  192.168.101.97    2005
interface eth1
--------------- end of /etc/ipxtunnel.conf -------------

I use IPX tunnel between .1.1 and .2.10 so we don't need to encrypt nor
authenticate encapsulated IPX packets, it is done with IPsec.

If you don't wont to use iproute2 to change source IP you need to use SIB
(it is able to bind local address) or establish tunnel between .0.1 and
.0.10 (external IPs, you need to do encryption in the program, but it isn't
strong).

For now I'm using ipxtunnel.

I think that's all for the moment. If there are any error, please e-mail me: 
poltorak@df.ru . It would be cool if someone puts the scheme of TCP/IP in
kernel and firewall example on FreeS/WAN's manual pages.

PoltoS

History and politics of cryptography

Cryptography has a long and interesting history, and has been the subject of considerable political controversy.

Introduction

History

The classic book on the history of cryptography is David Kahn's The Codebreakers. It traces codes and codebreaking from ancient Egypt to the 20th century.

Diffie and Landau Privacy on the Line: The Politics of Wiretapping and Encryption covers the history from the First World War to the 1990s, with an emphasis on the US.

World War II

During the Second World War, the British "Ultra" project achieved one of the greatest intelligence triumphs in the history of warfare, breaking many Axis codes. One major target was the Enigma cipher machine, a German device whose users were convinced it was unbreakable. The American "Magic" project had some similar triumphs against Japanese codes.

There are many books on this period. See our bibliography for several. Two I particularly like are:

• Andrew Hodges has done a superb biography of Alan Turing, a key player among the Ultra codebreakers. Turing was also an important computer pioneer. The terms Turing test and Turing machine are named for him, as is the ACM's highest technical award.
• Neal Stephenson's Cryptonomicon is a novel with cryptography central to the plot. Parts of it take place during WW II, other parts today.

Bletchley Park, where much of the Ultra work was done, now has a museum and a web site.

The Ultra work introduced three major innovations.

• The first break of Enigma was achieved by Polish Intelligence in 1931. Until then most code-breakers had been linguists, but a different approach was needed to break machine ciphers. Polish Intelligence recruited bright young mathematicians to crack the "unbreakable" Enigma. When war came in 1939, the Poles told their allies about this, putting Britain on the road to Ultra. The British also adopted a mathematical approach.
• Machines were extensively used in the attacks. First the Polish "Bombe" for attacking Enigma, then British versions of it, then machines such as Collosus for attacking other codes. By the end of the war, some of these machines were beginning to closely resemble digital computers. After the war, a team at Manchester University, several old Ultra hands included, built one of the world's first actual general-purpose digital computers.
• Ultra made codebreaking a large-scale enterprise, producing intelligence on an industrial scale. This was not a "black chamber", not a hidden room in some obscure government building with a small crew of code-breakers. The whole operation -- from wholesale interception of enemy communications by stations around the world, through large-scale code-breaking and analysis of the decrypted material (with an enormous set of files for cross-referencing), to delivery of intelligence to field commanders -- was huge, and very carefully managed.

So by the end of the war, Allied code-breakers were expert at large-scale mechanised code-breaking. The payoffs were enormous.

Postwar and Cold War

The wartime innovations were enthusiastically adopted by post-war and Cold War signals intelligence agencies. Presumably many nations now have some agency capable of sophisticated attacks on communications security, and quite a few engage in such activity on a large scale.

America's NSA, for example, is said to be both the world's largest employer of mathematicians and the world's largest purchaser of computer equipment. Such claims may be somewhat exaggerated, but beyond doubt the NSA -- and similar agencies in other countries -- have some excellent mathematicians, lots of powerful computers, sophisticated software, and the organisation and funding to apply them on a large scale. Details of the NSA budget are secret, but there are some published estimates.

Changes in the world's communications systems since WW II have provided these agencies with new targets. Cracking the codes used on an enemy's military or diplomatic communications has been common practice for centuries. Extensive use of radio in war made large-scale attacks such as Ultra possible. Modern communications make it possible to go far beyond that. Consider listening in on cell phones, or intercepting electronic mail, or tapping into the huge volumes of data on new media such as fiber optics or satellite links. None of these targets existed in 1950. All of them can be attacked today, and almost certainly are being attacked.

The Ultra story was not made public until the 1970s. Much of the recent history of codes and code-breaking has not been made public, and some of it may never be. Two important books are:

• Bamford's The Puzzle Palace, a history of the NSA
• Hager's Secret Power, about the Echelon system -- the US, UK, Canada, Australia and New Zealand co-operating to monitor much of the world's communications.

Note that these books cover only part of what is actually going on, and then only the activities of nations open and democratic enough that (some of) what they are doing can be discovered. A full picture, including:

• actions of the English-speaking democracies not covered in those books
• actions of other more-or-less sane governments
• the activities of various more-or-less insane governments
• possibilities for unauthorized action by government employees
• possible actions by large non-government organisations: corporations, criminals, or conspiracies

might be really frightening.

Recent history -- the crypto wars

Until quite recently, cryptography was primarily a concern of governments, especially of the military, of spies, and of diplomats. Much of it was extremely secret.

In recent years, that has changed a great deal. With computers and networking becoming ubiquitous, cryptography is now important to almost everyone. Among the developments since the 1970s:

• The US gov't established the Data Encryption Standard, DES, a block cipher for cryptographic protection of unclassfied documents.
• DES also became widely used in industry, especially regulated industries such as banking.
• Other nations produced their own standards, such as GOST in the Soviet Union.
• Public key cryptography was invented by Diffie and Hellman.
• Academic conferences such as Crypto and Eurocrypt began.
• Several companies began offerring cryptographic products: RSA, PGP, the many vendors with PKI products, ...
• Cryptography appeared in other products: operating systems, word processors, ...
• Network protocols based on crypto were developed: SSH , SSL, IPsec, ...
• Crytography came into widespread use to secure bank cards, terminals, ...
• The US government replaced DES with the much stronger Advanced Encryption Standard, AES

This has led to a complex ongoing battle between various mainly government groups wanting to control the spread of crypto and various others, notably the computer industry and the cypherpunk crypto advocates, wanting to encourage widespread use.

Steven Levy has written a fine history of much of this, called Crypto: How the Code rebels Beat the Government -- Saving Privacy in the Digital Age.

The FreeS/WAN project is to a large extent an outgrowth of cypherpunk ideas. Our reasons for doing the project can be seen in these quotes from the Cypherpunk Manifesto:

Privacy is necessary for an open society in the electronic age. ...

We cannot expect governments, corporations, or other large, faceless organizations to grant us privacy out of their beneficence. It is to their advantage to speak of us, and we should expect that they will speak. ...

We must defend our own privacy if we expect to have any. ...

Cypherpunks write code. We know that someone has to write software to defend privacy, and since we can't get privacy unless we all do, we're going to write it. We publish our code so that our fellow Cypherpunks may practice and play with it. Our code is free for all to use, worldwide. We don't much care if you don't approve of the software we write. We know that software can't be destroyed and that a widely dispersed system can't be shut down.

Cypherpunks deplore regulations on cryptography, for encryption is fundamentally a private act. ...

For privacy to be widespread it must be part of a social contract. People must come and together deploy these systems for the common good. ...

To quote project leader John Gilmore:

We are literally in a race between our ability to build and deploy technology, and their ability to build and deploy laws and treaties. Neither side is likely to back down or wise up until it has definitively lost the race.

If FreeS/WAN reaches its goal of making opportunistic encryption widespread so that secure communication can become the default for a large part of the net, we will have struck a major blow.

Politics

The political problem is that nearly all governments want to monitor their enemies' communications, and some want to monitor their citizens. They may be very interested in protecting some of their own communications, and often some types of business communication, but not in having everyone able to communicate securely. They therefore attempt to restrict availability of strong cryptography as much as possible.

Things various governments have tried or are trying include:

• Echelon, a monitor-the-world project of the US, UK, NZ, Australian and Canadian signals intelligence agencies. See this collection of links and this story on the French Parliament's reaction.
• Others governments may well have their own Echelon-like projects. To quote the Dutch Minister of Defense, as reported in a German magazine:

  The government believes not only the governments associated with Echelon are able to intercept communication systems, but that it is an activity of the investigative authorities and intelligence services of many countries with governments of different political signature.

  Even if they have nothing on the scale of Echelon, most intelligence agencies and police forces certainly have some interception capability.
• NSA tapping of submarine communication cables, described in this article
• A proposal for international co-operation on Internet surveillance.
• Alleged sabotage of security products by the NSA (the US signals intelligence agency).
• The German armed forces and some government departments will stop using American software for fear of NSA "back doors", according to this news story .
• The British Regulation of Investigatory Powers bill. See this web page. and perhaps this cartoon.
• A Russian ban on cryptography
• Chinese controls on net use.
• The FBI's carnivore system for covert searches of email. See this news coverage and this risk assessment. The government had an external review of some aspects of this system done. See this analysis of that review. Possible defenses against Carnivore include:
  • PGP for end-to-end mail encryption
  • secure sendmail for server-to-server encryption
  • IPsec encryption on the underlying IP network
• export laws restricting strong cryptography as a munition. See discussion below.
• various attempts to convince people that fundamentally flawed cryptography, such as encryption with a back door for government access to data or with inadequate key lengths, was adequate for their needs.

Of course governments are by no means the only threat to privacy and security on the net. Other threats include:

• industrial espionage, as for example in this news story
• attacks by organised criminals, as in this large-scale attack
• collection of personal data by various companies.
  • for example, consider the various corporate winners of Privacy International's Big Brother Awards.
  • Zero Knowledge sell tools to defend against this
• individuals may also be a threat in a variety of ways and for a variety of reasons
• in particular, an individual with access to government or industry data collections could do considerable damage using that data in unauthorized ways.

One study enumerates threats and possible responses for small and medium businesses. VPNs are a key part of the suggested strategy.

We consider privacy a human right. See the UN's Universal Declaration of Human Rights, article twelve:

No one shall be subjected to arbitrary interference with his privacy, family, home or correspondence, nor to attacks upon his honor and reputation. Everyone has the right to the protection of the law against such interference or attacks.

Our objective is to help make privacy possible on the Internet using cryptography strong enough not even those well-funded government agencies are likely to break it. If we can do that, the chances of anyone else breaking it are negliible.

Links

Many groups are working in different ways to defend privacy on the net and elsewhere. Please consider contributing to one or more of these groups:

• the EFF's Privacy Now! campaign
• the Global Internet Liberty Campaign
• Computer Professionals for Social Responsibility

For more on these issues see:

• Steven Levy (Newsweek's chief technology writer and author of the classic "Hackers") new book Crypto: How the Code Rebels Beat the Government--Saving Privacy in the Digital Age
• Simson Garfinkel (Boston Globe columnist and author of books on PGP and Unix Security) book Database Nation: the death of privacy in the 21st century

There are several collections of crypto quotes on the net.

See also the bibliography and our list of web references on cryptography law and policy.

Outline of this section

The remainder of this section includes two pieces of writing by our project leader

• his rationale for starting this
• another discussion of project goals

and discussions of:

• why we do not use DES
• cryptography export laws
• why government access to keys is not a good idea
• the myth that short keys are adequate for some security requirements

and a section on press coverage of FreeS/WAN.

From our project leader

FreeS/WAN project founder John Gilmore wrote a web page about why we are doing this. The version below is slightly edited, to fit this format and to update some links. For a version without these edits, see his home page.

My project for 1996 was to secure 5% of the Internet traffic against passive wiretapping. It didn't happen in 1996, so I'm still working on it in 1997, 1998, and 1999! If we get 5% in 1999 or 2000, we can secure 20% the next year, against both active and passive attacks; and 80% the following year. Soon the whole Internet will be private and secure. The project is called S/WAN or S/Wan or Swan for Secure Wide Area Network; since it's free software, we call it FreeSwan to distinguish it from various commercial implementations. RSA came up with the term "S/WAN". Our main web site is at http://www.freeswan.org/. Want to help?

The idea is to deploy PC-based boxes that will sit between your local area network and the Internet (near your firewall or router) which opportunistically encrypt your Internet packets. Whenever you talk to a machine (like a Web site) that doesn't support encryption, your traffic goes out "in the clear" as usual. Whenever you connect to a machine that does support this kind of encryption, this box automatically encrypts all your packets, and decrypts the ones that come in. In effect, each packet gets put into an "envelope" on one side of the net, and removed from the envelope when it reaches its destination. This works for all kinds of Internet traffic, including Web access, Telnet, FTP, email, IRC, Usenet, etc.

The encryption boxes are standard PC's that use freely available Linux software that you can download over the Internet or install from a cheap CDROM.

This wasn't just my idea; lots of people have been working on it for years. The encryption protocols for these boxes are called IPSEC (IP Security). They have been developed by the IP Security Working Group of the Internet Engineering Task Force, and will be a standard part of the next major version of the Internet protocols ( IPv6). For today's (IP version 4) Internet, they are an option.

The Internet Architecture Board and Internet Engineering Steering Group have taken a strong stand that the Internet should use powerful encryption to provide security and privacy. I think these protocols are the best chance to do that, because they can be deployed very easily, without changing your hardware or software or retraining your users. They offer the best security we know how to build, using the Triple-DES, RSA, and Diffie-Hellman algorithms.

This "opportunistic encryption box" offers the "fax effect". As each person installs one for their own use, it becomes more valuable for their neighbors to install one too, because there's one more person to use it with. The software automatically notices each newly installed box, and doesn't require a network administrator to reconfigure it. Instead of "virtual private networks" we have a "REAL private network"; we add privacy to the real network instead of layering a manually-maintained virtual network on top of an insecure Internet.

Deployment of IPSEC

The US government would like to control the deployment of IP Security with its crypto export laws. This isn't a problem for my effort, because the cryptographic work is happening outside the United States. A foreign philanthropist, and others, have donated the resources required to add these protocols to the Linux operating system. Linux is a complete, freely available operating system for IBM PC's and several kinds of workstation, which is compatible with Unix. It was written by Linus Torvalds, and is still maintained by a talented team of expert programmers working all over the world and coordinating over the Internet. Linux is distributed under the GNU Public License, which gives everyone the right to copy it, improve it, give it to their friends, sell it commercially, or do just about anything else with it, without paying anyone for the privilege.

Organizations that want to secure their network will be able to put two Ethernet cards into an IBM PC, install Linux on it from a $30 CDROM or by downloading it over the net, and plug it in between their Ethernet and their Internet link or firewall. That's all they'll have to do to encrypt their Internet traffic everywhere outside their own local area network.

Travelers will be able to run Linux on their laptops, to secure their connection back to their home network (and to everywhere else that they connect to, such as customer sites). Anyone who runs Linux on a standalone PC will also be able to secure their network connections, without changing their application software or how they operate their computer from day to day.

There will also be numerous commercially available firewalls that use this technology. RSA Data Security is coordinating the S/Wan (Secure Wide Area Network) project among more than a dozen vendors who use these protocols. There's a compatability chart that shows which vendors have tested their boxes against which other vendors to guarantee interoperatility.

Eventually it will also move into the operating systems and networking protocol stacks of major vendors. This will probably take longer, because those vendors will have to figure out what they want to do about the export controls.

Current status

My initial goal of securing 5% of the net by Christmas '96 was not met. It was an ambitious goal, and inspired me and others to work hard, but was ultimately too ambitious. The protocols were in an early stage of development, and needed a lot more protocol design before they could be implemented. As of April 1999, we have released version 1.0 of the software ( freeswan-1.0.tar.gz), which is suitable for setting up Virtual Private Networks using shared secrets for authentication. It does not yet do opportunistic encryption, or use DNSSEC for authentication; those features are coming in a future release.

Protocols

The low-level encrypted packet formats are defined. The system for publishing keys and providing secure domain name service is defined. The IP Security working group has settled on an NSA-sponsored protocol for key agreement (called ISAKMP/Oakley), but it is still being worked on, as the protocol and its documentation is too complex and incomplete. There are prototype implementations of ISAKMP. The protocol is not yet defined to enable opportunistic encryption or the use of DNSSEC keys.

Linux Implementation

The Linux implementation has reached its first major release and is ready for production use in manually-configured networks, using Linux kernel version 2.0.36.

Domain Name System Security

There is now a release of BIND 8.2 that includes most DNS Security features.

The first prototype implementation of Domain Name System Security was funded by DARPA as part of their Information Survivability program. Trusted Information Systems wrote a modified version of BIND, the widely-used Berkeley implementation of the Domain Name System.

TIS, ISC, and I merged the prototype into the standard version of BIND. The first production version that supports KEY and SIG records is bind-4.9.5. This or any later version of BIND will do for publishing keys. It is available from the Internet Software Consortium. This version of BIND is not export-controlled since it does not contain any cryptography. Later releases starting with BIND 8.2 include cryptography for authenticating DNS records, which is also exportable. Better documentation is needed.

Why?

Because I can. I have made enough money from several successful startup companies, that for a while I don't have to work to support myself. I spend my energies and money creating the kind of world that I'd like to live in and that I'd like my (future) kids to live in. Keeping and improving on the civil rights we have in the United States, as we move more of our lives into cyberspace, is a particular goal of mine.

What You Can Do

Install the latest BIND at your site.

You won't be able to publish any keys for your domain, until you have upgraded your copy of BIND. The thing you really need from it is the new version of named, the Name Daemon, which knows about the new KEY and SIG record types. So, download it from the Internet Software Consortium and install it on your name server machine (or get your system administrator, or Internet Service Provider, to install it). Both your primary DNS site and all of your secondary DNS sites will need the new release before you will be able to publish your keys. You can tell which sites this is by running the Unix command "dig MYDOMAIN ns" and seeing which sites are mentioned in your NS (name server) records.

Set up a Linux system and run a 2.0.x kernel on it

Get a machine running Linux (say the 5.2 release from Red Hat). Give the machine two Ethernet cards.

Install the Linux IPSEC (Freeswan) software

If you're an experienced sysadmin or Linux hacker, install the freeswan-1.0 release, or any later release or snapshot. These releases do NOT provide automated "opportunistic" operation; they must be manually configured for each site you wish to encrypt with.

Get on the linux-ipsec mailing list

The discussion forum for people working on the project, and testing the code and documentation, is: linux-ipsec@clinet.fi. To join this mailing list, send email to linux-ipsec-REQUEST@clinet.fi containing a line of text that says "subscribe linux-ipsec". (You can later get off the mailing list the same way -- just send "unsubscribe linux-ipsec").

Check back at this web page every once in a while

I update this page periodically, and there may be new information in it that you haven't seen. My intent is to send email to the mailing list when I update the page in any significant way, so subscribing to the list is an alternative.

Would you like to help? I can use people who are willing to write documentation, install early releases for testing, write cryptographic code outside the United States, sell pre-packaged software or systems including this technology, and teach classes for network administrators who want to install this technology. To offer to help, send me email at gnu@toad.com. Tell me what country you live in and what your citizenship is (it matters due to the export control laws; personally I don't care). Include a copy of your resume and the URL of your home page. Describe what you'd like to do for the project, and what you're uniquely qualified for. Mention what other volunteer projects you've been involved in (and how they worked out). Helping out will require that you be able to commit to doing particular things, meet your commitments, and be responsive by email. Volunteer projects just don't work without those things.

Related projects

IPSEC for NetBSD

This prototype implementation of the IP Security protocols is for another free operating system. Download BSDipsec.tar.gz.

IPSEC for OpenBSD

This prototype implementation of the IP Security protocols is for yet another free operating system. It is directly integrated into the OS release, since the OS is maintained in Canada, which has freedom of speech in software.

Stopping wholesale monitoring

From a message project leader John Gilmore posted to the mailing list:

John Denker wrote:

> Indeed there are several ways in which the documentation overstates the 
> scope of what this project does -- starting with the name 
> FreeS/WAN.  There's a big difference between having an encrypted IP tunnel 
> versus having a Secure Wide-Area Network.  This software does a fine job of 
> the former, which is necessary but not sufficient for the latter.

The goal of the project is to make it very hard to tap your wide area
communications.  The current system provides very good protection
against passive attacks (wiretapping and those big antenna farms).
Active attacks, which involve the intruder sending packets to your
system (like packets that break into sendmail and give them a root
shell :-) are much harder to guard against.  Active attacks that
involve sending people (breaking into your house and replacing parts
of your computer with ones that transmit what you're doing) are also
much harder to guard against.  Though we are putting effort into
protecting against active attacks, it's a much bigger job than merely
providing strong encryption.  It involves general computer security,
and general physical security, which are two very expensive problems
for even a site to solve, let alone to build into a whole society.

The societal benefit of building an infrastructure that protects
well against passive attacks is that it makes it much harder to do
undetected bulk monitoring of the population.  It's a defense against
police-states, not against policemen.

Policemen can put in the effort required to actively attack sites that
they have strong suspicions about.  But police states won't be able to
build systems that automatically monitor everyone's communications.
Either they will be able to monitor only a small subset of the
populace (by targeting those who screwed up their passive security),
or their monitoring activities will be detectable by those monitored
(active attacks leave packet traces or footprints), which can then be
addressed through the press and through political means if they become
too widespread.

FreeS/WAN does not protect very well against traffic analysis, which
is a kind of widespread police-state style monitoring that still
reveals significant information (who's talking to who) without
revealing the contents of what was said.  Defenses against traffic
analysis are an open research problem.  Zero Knowledge Systems is
actively deploying a system designed to thwart it, designed by Ian
Goldberg.  The jury is out on whether it actually works; a lot more
experience with it will be needed.

Notes on things mentioned in that message:

• Denker is a co-author of a paper on a large FreeS/WAN application.
• Information on Zero Knowledge is on their web site. Their Freedom product, designed to provide untracable pseudonyms for use on the net, is no longer marketed.
• Another section of our documentation discusses ways to resist traffic analysis.

Government promotion of weak crypto

Various groups, especially governments and especially the US government, have a long history of advocating various forms of bogus security.

We regard bogus security as extremely dangerous. If users are deceived into relying on bogus security, then they may be exposed to large risks. They would be better off having no security and knowing it. At least then they would be careful about what they said.

Avoiding bogus security is a key design criterion for everything we do in FreeS/WAN. The most conspicuous example is our refusal to support single DES. Other IPsec "features" which we do not implement are discussed in our compatibility document.

Escrowed encryption

Various governments have made persistent attempts to encourage or mandate "escrowed encrytion", also called "key recovery", or GAK for "government access to keys". The idea is that cryptographic keys be held by some third party and turned over to law enforcement or security agencies under some conditions.

  Mary had a little key - she kept it in escrow,
  and every thing that Mary said,
  the feds were sure to know.

A crypto quotes page attributes this to Sam Simpson.

There is an excellent paper available on Risks of Escrowed Encryption, from a group of cryptographic luminaries which included our project leader.

Like any unnecessary complication, GAK tends to weaken security of any design it infects. For example:

• Matt Blaze found a fatal flaw in the US government's Clipper chip shortly after design information became public. See his paper "Protocol Failure in the Escrowed Encryption Standard" on his papers page.
• a rather nasty bug was found in the "additional decryption keys" "feature" of some releases of PGP

FreeS/WAN does not support escrowed encryption, and never will.

Limited key lengths

Various governments, and some vendors, have also made persistent attempts to convince people that:

• weak systems are sufficient for some data
• strong cryptography should be reserved for cases where the extra overheads are justified

This is utter nonsense.

Weak systems touted include:

• the ludicrously weak (deliberately crippled) 40-bit ciphers that until recently were all various export laws allowed
• 56-bit single DES, discussed below
• 64-bit symmetric ciphers and 512-bit RSA, the maximums for unrestricted export under various current laws

The notion that choice of ciphers or keysize should be determined by a trade-off between security requirements and overheads is pure bafflegab.

• For most symmetric ciphers, it is simply a lie. Any block cipher has some natural maximum keysize inherent in the design -- 128 bits for IDEA or CAST-128, 256 for Serpent or Twofish, 448 for Blowfish and 2048 for RC4. Using a key size smaller than that limit gives exactly zero savings in overhead. The crippled 40-bit or 64-bit version of the cipher provides no advantage whatsoever.
• AES uses 10 rounds with 128-bit keys, 12 rounds for 192-bit and 14 rounds for 256-bit, so there actually is a small difference in overhead, but not enough to matter in most applications.
• For triple DES there is a grain of truth in the argument. 3DES is indeed three times slower than single DES. However, the solution is not to use the insecure single DES, but to pick a faster secure cipher. CAST-128, Blowfish and the AES candidate ciphers are are all considerably faster in software than DES (let alone 3DES!), and apparently secure.
• For public key techniques, there are extra overheads for larger keys, but they generally do not affect overall performance significantly. Practical public key applications are usually hybrid systems in which the bulk of the work is done by a symmetric cipher. The effect of increasing the cost of the public key operations is typically negligible because the public key operations use only a tiny fraction of total resources.

  For example, suppose public key operations use use 1% of the time in a hybrid system and you triple the cost of public key operations. The cost of symmetric cipher operations is unchanged at 99% of the original total cost, so the overall effect is a jump from 99 + 1 = 100 to 99 + 3 = 102, a 2% rise in system cost.

In short, there has never been any technical reason to use inadequate ciphers. The only reason there has ever been for anyone to use such ciphers is that government agencies want weak ciphers used so that they can crack them. The alleged savings are simply propaganda.

   Mary had a little key (It's all she could export),
   and all the email that she sent was opened at the Fort.

A crypto quotes page attributes this to Ron Rivest. NSA headquarters is at Fort Meade, Maryland.

Our policy in FreeS/WAN is to use only cryptographic components with adequate keylength and no known weaknesses.

• We do not implement single DES because it is clearly insecure, so implemeting it would violate our policy of avoiding bogus security. Our default cipher is 3DES
• Similarly, we do not implement the 768-bit Group 1 for Diffie-Hellman key negotiation. We provide only the 1024-bit Group 2 and 1536-bit Group 5.

Detailed discussion of which IPsec features we implement or omit is in out compatibility document.

These decisions imply that we cannot fully conform to the IPsec RFCs, since those have DES as the only required cipher and Group 1 as the only required DH group. (In our view, the standards were subverted into offerring bogus security.) Fortunately, we can still interoperate with most other IPsec implementations since nearly all implementers provide at least 3DES and Group 2 as well.

We hope that eventually the RFCs will catch up with our (and others') current practice and reject dubious components. Some of our team and a number of others are working on this in IETF working groups.

Some real trade-offs

Of course, making systems secure does involve costs, and trade-offs can be made between cost and security. However, the real trade-offs have nothing to do with using weaker ciphers.

There can be substantial hardware and software costs. There are often substantial training costs, both to train administrators and to increase user awareness of security issues and procedures. There are almost always substantial staff or contracting costs.

Security takes staff time for planning, implementation, testing and auditing. Some of the issues are subtle; you need good (hence often expensive) people for this. You also need people to monitor your systems and respond to problems. The best safe ever built is insecure if an attacker can work on it for days without anyone noticing. Any computer is insecure if the administrator is "too busy" to check the logs.

Moreover, someone in your organisation (or on contract to it) needs to spend considerable time keeping up with new developments. EvilDoers will know about new attacks shortly after they are found. You need to know about them before your systems are attacked. If your vendor provides a patch, you need to apply it. If the vendor does nothing, you need to complain or start looking for another vendor.

For a fairly awful example, see this report. In that case over a million credit card numbers were taken from e-commerce sites, using security flaws in Windows NT servers. Microsoft had long since released patches for most or all of the flaws, but the site administrators had not applied them.

At an absolute minimum, you must do something about such issues before an exploitation tool is posted to the net for downloading by dozens of "script kiddies". Such a tool might appear at any time from the announcement of the security hole to several months later. Once it appears, anyone with a browser and an attitude can break any system whose administrators have done nothing about the flaw.

Compared to those costs, cipher overheads are an insignificant factor in the cost of security.

The only thing using a weak cipher can do for you is to cause all your other investment to be wasted.

Cryptography Export Laws

Many nations restrict the export of cryptography and some restrict its use by their citizens or others within their borders.

US Law

US laws, as currently interpreted by the US government, forbid export of most cryptographic software from the US in machine-readable form without government permission. In general, the restrictions apply even if the software is widely-disseminated or public-domain and even if it came from outside the US originally. Cryptography is legally a munition and export is tightly controlled under the EAR Export Administration Regulations.

If you are a US citizen, your brain is considered US territory no matter where it is physically located at the moment. The US believes that its laws apply to its citizens everywhere, not just within the US. Providing technical assistance or advice to foreign "munitions" projects is illegal. The US government has very little sense of humor about this issue and does not consider good intentions to be sufficient excuse. Beware.

The official website for these regulations is run by the Commerce Department's Bureau of Export Administration (BXA).

The Bernstein case challenges the export restrictions on Constitutional grounds. Code is speech so restrictions on export of code violate the First Amendment's free speech provisions. This argument has succeeded in two levels of court so far. It is quite likely to go on to the Supreme Court.

The regulations were changed substantially in January 2000, apparently as a government attempt to get off the hook in the Bernstein case. It is now legal to export public domain source code for encryption, provided you notify the BXA.

There are, however, still restrictions in force. Moreover, the regulations can still be changed again whenever the government chooses to do so. Short of a Supreme Court ruling (in the Berstein case or another) that overturns the regulations completely, the problem of export regulation is not likely to go away in the forseeable future.

US contributions to FreeS/WAN

The FreeS/WAN project cannot accept software contributions, not even small bug fixes, from US citizens or residents. We want it to be absolutely clear that our distribution is not subject to US export law. Any contribution from an American might open that question to a debate we'd prefer to avoid. It might also put the contributor at serious legal risk.

Of course Americans can still make valuable contributions (many already have) by reporting bugs, or otherwise contributing to discussions, on the project mailing list. Since the list is public, this is clearly constitutionally protected free speech.

Note, however, that the export laws restrict Americans from providing technical assistance to foreign "munitions" projects. The government might claim that private discussions or correspondence with FreeS/WAN developers were covered by this. It is not clear what the courts would do with such a claim, so we strongly encourage Americans to use the list rather than risk the complications.

What's wrong with restrictions on cryptography

Some quotes from prominent cryptography experts:

The real aim of current policy is to ensure the continued effectiveness of US information warfare assets against individuals, businesses and governments in Europe and elsewhere.
Ross Anderson, Cambridge University

If the government were honest about its motives, then the debate about crypto export policy would have ended years ago.
Bruce Schneier, Counterpane Systems

The NSA regularly lies to people who ask it for advice on export control. They have no reason not to; accomplishing their goal by any legal means is fine by them. Lying by government employees is legal.
John Gilmore.

The Internet Architecture Board (IAB) and the Internet Engineering Steering Group (IESG) made a strong statement in favour of worldwide access to strong cryptography. Essentially the same statement is in the appropriately numbered RFC 1984. Two critical paragraphs are:

... various governments have actual or proposed policies on access to cryptographic technology ...

(a) ... export controls ...
(b) ... short cryptographic keys ...
(c) ... keys should be in the hands of the government or ...
(d) prohibit the use of cryptology ...

We believe that such policies are against the interests of consumers and the business community, are largely irrelevant to issues of military security, and provide only a marginal or illusory benefit to law enforcement agencies, ...

The IAB and IESG would like to encourage policies that allow ready access to uniform strong cryptographic technology for all Internet users in all countries.

Our goal in the FreeS/WAN project is to build just such "strong cryptographic technology" and to distribute it "for all Internet users in all countries".

More recently, the same two bodies (IESG and IAB) have issued RFC 2804 on why the IETF should not build wiretapping capabilities into protocols for the convenience of security or law enforcement agenicies. The abstract from that document is:

The Internet Engineering Task Force (IETF) has been asked to take a position on the inclusion into IETF standards-track documents of functionality designed to facilitate wiretapping.

This memo explains what the IETF thinks the question means, why its answer is "no", and what that answer means.

We should not be building surveillance technology into standards. Law enforcement was not supposed to be easy. Where it is easy, it's called a police state.
Jeff Schiller of MIT, in a discussion of FBI demands for wiretap capability on the net, as quoted by Wired.

The Raven mailing list was set up for this IETF discussion.

Our goal is to go beyond that RFC and prevent Internet wiretapping entirely.

The Wassenaar Arrangement

Restrictions on the export of cryptography are not just US policy, though some consider the US at least partly to blame for the policies of other nations in this area.

A number of countries:

Argentina, Australia, Austria, Belgium, Bulgaria, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Luxembourg, Netherlands, New Zealand, Norway, Poland, Portugal, Republic of Korea, Romania, Russian Federation, Slovak Republic, Spain, Sweden, Switzerland, Turkey, Ukraine, United Kingdom and United States

have signed the Wassenaar Arrangement which restricts export of munitions and other tools of war. Cryptographic sofware is covered there.

Wassenaar details are available from the Wassenaar Secretariat, and elsewhere in a more readable HTML version.

For a critique see the GILC site:

The Global Internet Liberty Campaign (GILC) has begun a campaign calling for the removal of cryptography controls from the Wassenaar Arrangement.

The aim of the Wassenaar Arrangement is to prevent the build up of military capabilities that threaten regional and international security and stability . . .

There is no sound basis within the Wassenaar Arrangement for the continuation of any export controls on cryptographic products.

We agree entirely.

An interesting analysis of Wassenaar can be found on the cyber-rights.org site.

Export status of Linux FreeS/WAN

We believe our software is entirely exempt from these controls since the Wassenaar General Software Note says:

The Lists do not control "software" which is either:

1. Generally available to the public by . . . retail . . . or
2. "In the public domain".

There is a note restricting some of this, but it is a sub-heading under point 1, so it appears not to apply to public domain software.

Their glossary defines "In the public domain" as:

. . . "technology" or "software" which has been made available without restrictions upon its further dissemination.

N.B. Copyright restrictions do not remove "technology" or "software" from being "in the public domain".

We therefore believe that software freely distributed under the GNU Public License, such as Linux FreeS/WAN, is exempt from Wassenaar restrictions.

Most of the development work is being done in Canada. Our understanding is that the Canadian government accepts this interpretation.

• A web statement of Canadian policy is available from the Department of Foreign Affairs and International Trade.
• Another document from that department states that public domain software is exempt from the export controls.
• A researcher's analysis of Canadian policy is also available.

Recent copies of the freely modifiable and distributable source code exist in many countries. Citizens all over the world participate in its use and evolution, and guard its ongoing distribution. Even if Canadian policy were to change, the software would continue to evolve in countries which do not restrict exports, and would continue to be imported from there into unfree countries. "The Net culture treats censorship as damage, and routes around it."

Help spread IPsec around

You can help. If you don't know of a Linux FreeS/WAN archive in your own country, please download it now to your personal machine, and consider making it publicly accessible if that doesn't violate your own laws. If you have the resources, consider going one step further and setting up a mirror site for the whole munitions Linux crypto software archive.

If you make Linux CD-ROMs, please consider including this code, in a way that violates no laws (in a free country, or in a domestic-only CD product).

Please send a note about any new archive mirror sites or CD distributions to linux-ipsec@clinet.fi so we can update the documentation.

Lists of current mirror sites and of distributions which include FreeS/WAN are in our introduction section.

DES is Not Secure

DES, the Data Encryption S tandard, can no longer be considered secure. While no major flaws in its innards are known, it is fundamentally inadequate because its 56-bit key is too short. It is vulnerable to brute-force search of the whole key space, either by large collections of general-purpose machines or even more quickly by specialized hardware. Of course this also applies to any other cipher with only a 56-bit key. The only reason anyone could have for using a 56 or 64-bit key is to comply with various export laws intended to ensure the use of breakable ciphers.

Non-government cryptologists have been saying DES's 56-bit key was too short for some time -- some of them were saying it in the 70's when DES became a standard -- but the US government has consistently ridiculed such suggestions.

A group of well-known cryptographers looked at key lengths in a 1996 paper. They suggested a minimum of 75 bits to consider an existing cipher secure and a minimum of 90 bits for new ciphers. More recent papers, covering both symmetric and public key systems are at cryptosavvy.com and rsa.com. For all algorithms, the minimum keylengths recommended in such papers are significantly longer than the maximums allowed by various export laws.

In a 1998 ruling, a German court described DES as "out-of-date and not safe enough" and held a bank liable for using it.

Dedicated hardware breaks DES in a few days

The question of DES security has now been settled once and for all. In early 1998, the Electronic Frontier Foundation built a DES-cracking machine. It can find a DES key in an average of a few days' search. The details of all this, including complete code listings and complete plans for the machine, have been published in Cracking DES, by the Electronic Frontier Foundation.

That machine cost just over $200,000 to design and build. "Moore's Law" is that machines get faster (or cheaper, for the same speed) by roughly a factor of two every 18 months. At that rate, their $200,000 in 1998 becomes $50,000 in 2001.

However, Moore's Law is not exact and the $50,000 estimate does not allow for the fact that a copy based on the published EFF design would cost far less than the original. We cannot say exactly what such a cracker would cost today, but it would likely be somewhere between $10,000 and $100,000.

A large corporation could build one of these out of petty cash. The cost is low enough for a senior manager to hide it in a departmental budget and avoid having to announce or justify the project. Any government agency, from a major municipal police force up, could afford one. Or any other group with a respectable budget -- criminal organisations, political groups, labour unions, religious groups, ... Or any millionaire with an obsession or a grudge, or just strange taste in toys.

One might wonder if a private security or detective agency would have one for rent. They wouldn't need many clients to pay off that investment.

Spooks may break DES faster yet

As for the security and intelligence agencies of various nations, they may have had DES crackers for years, and theirs may be much faster. It is difficult to make most computer applications work well on parallel machines, or to design specialised hardware to accelerate them. Cipher-cracking is one of the very few exceptions. It is entirely straightforward to speed up cracking by just adding hardware. Within very broad limits, you can make it as fast as you like if you have the budget. The EFF's $200,000 machine breaks DES in a few days. An aviation website gives the cost of a B1 bomber as $200,000,000. Spending that much, an intelligence agency could break DES in an average time of six and a half minutes.

That estimate assumes they use the EFF's 1998 technology and just spend more money. They may have an attack that is superior to brute force, they quite likely have better chip technology (Moore's law, a bigger budget, and whatever secret advances they may have made) and of course they may have spent the price of an aircraft carrier, not just one aircraft.

In short, we have no idea how quickly these organisations can break DES. Unless they're spectacularly incompetent or horribly underfunded, they can certainly break it, but we cannot guess how quickly. Pick any time unit between days and milliseconds; none is entirely unbelievable. More to the point, none of them is of any comfort if you don't want such organisations reading your communications.

Note that this may be a concern even if nothing you do is a threat to anyone's national security. An intelligence agency might well consider it to be in their national interest for certain companies to do well. If you're competing against such companies in a world market and that agency can read your secrets, you have a serious problem.

One might wonder about technology the former Soviet Union and its allies developed for cracking DES during the Cold War. They must have tried; the cipher was an American standard and widely used. Certainly those countries have some fine mathematicians, and those agencies had budget. How well did they succeed? Is their technology now for sale or rent?

Networks break DES in a few weeks

Before the definitive EFF effort, DES had been cracked several times by people using many machines. See this press release for example.

A major corporation, university, or government department could break DES by using spare cycles on their existing collection of computers, by dedicating a group of otherwise surplus machines to the problem, or by combining the two approaches. It might take them weeks or months, rather than the days required for the EFF machine, but they could do it.

What about someone working alone, without the resources of a large organisation? For them, cracking DES will not be easy, but it may be possible. A few thousand dollars buys a lot of surplus workstations. A pile of such machines will certainly heat your garage nicely and might break DES in a few months or years. Or enroll at a university and use their machines. Or use an employer's machines. Or crack security somewhere and steal the resources to crack a DES key. Or write a virus that steals small amounts of resources on many machines. Or . . .

None of these approaches are easy or break DES really quickly, but an attacker only needs to find one that is feasible and breaks DES quickly enough to be dangerous. How much would you care to bet that this will be impossible if the attacker is clever and determined? How valuable is your data? Are you authorised to risk it on a dubious bet?

We disable DES

In short, it is now absolutely clear that DES is not secure against

• any well-funded opponent
• any opponent (even a penniless one) with access (even stolen access) to enough general purpose computers

That is why Linux FreeS/WAN disables all transforms which use plain DES for encryption.

DES is in the source code, because we need DES to implement our default encryption transform, Triple DES. We urge you not to use single DES. We do not provide any easy way to enable it in FreeS/WAN, and our policy is to provide no assistance to anyone wanting to do so.

40-bits is laughably weak

The same is true, in spades, of ciphers -- DES or others -- crippled by 40-bit keys, as many ciphers were required to be until recently under various export laws. A brute force search of such a cipher's keyspace is 2¹⁶ times faster than a similar search against DES. The EFF's machine can do a brute-force search of a 40-bit key space in seconds. One contest to crack a 40-bit cipher was won by a student using a few hundred idle machines at his university. It took only three and half hours.

We do not, and will not, implement any 40-bit cipher.

Triple DES is almost certainly secure

Triple DES, usually abbreviated 3DES, applies DES three times, with three different keys. DES seems to be basically an excellent cipher design; it has withstood several decades of intensive analysis without any disastrous flaws being found. It's only major flaw is that the small keyspace allows brute force attacks to succeeed. Triple DES enlarges the key space to 168 bits, making brute-force search a ridiculous impossibility.

3DES is currently the only block cipher implemented in FreeS/WAN. 3DES is, unfortunately, about 1/3 the speed of DES, but modern CPUs still do it at quite respectable speeds. Some speed measurements for our code are available.

AES in IPsec

The AES project has chosen a replacement for DES, a new standard cipher for use in non-classified US government work and in regulated industries such as banking. This cipher will almost certainly become widely used for many applications, including IPsec.

The winner, announced in October 2000 after several years of analysis and discussion, was the Rijndael cipher from two Belgian designers.

It is almost certain that FreeS/WAN will add AES support. AES patches are already available.

Press coverage of Linux FreeS/WAN:

FreeS/WAN 1.0 press

• Wired "Linux-Based Crypto Stops Snoops", James Glave April 15 1999
• Slashdot
• DGL, Damar Group Limited; looking at FreeS/WAN from a perspective of business computing
• Linux Today
• TBTF, Tasty Bits from the Technology Front
• Salon Magazine "Free Encryption Takes a Big Step"

Press release for version 1.0

        Strong Internet Privacy Software Free for Linux Users Worldwide

Toronto, ON, April 14, 1999 - 

The Linux FreeS/WAN project today released free software to protect
the privacy of Internet communications using strong encryption codes.
FreeS/WAN automatically encrypts data as it crosses the Internet, to
prevent unauthorized people from receiving or modifying it.  One
ordinary PC per site runs this free software under Linux to become a
secure gateway in a Virtual Private Network, without having to modify
users' operating systems or application software.  The project built
and released the software outside the United States, avoiding US
government regulations which prohibit good privacy protection.
FreeS/WAN version 1.0 is available immediately for downloading at
http://www.xs4all.nl/~freeswan/.

"Today's FreeS/WAN release allows network administrators to build
excellent secure gateways out of old PCs at no cost, or using a cheap
new PC," said John Gilmore, the entrepreneur who instigated the
project in 1996.  "They can build operational experience with strong
network encryption and protect their users' most important
communications worldwide."

"The software was written outside the United States, and we do not
accept contributions from US citizens or residents, so that it can be
freely published for use in every country," said Henry Spencer, who
built the release in Toronto, Canada.  "Similar products based in the
US require hard-to-get government export licenses before they can be
provided to non-US users, and can never be simply published on a Web
site.  Our product is freely available worldwide for immediate
downloading, at no cost."

FreeS/WAN provides privacy against both quiet eavesdropping (such as
"packet sniffing") and active attempts to compromise communications
(such as impersonating participating computers).  Secure "tunnels" carry
information safely across the Internet between locations such as a
company's main office, distant sales offices, and roaming laptops.  This
protects the privacy and integrity of all information sent among those
locations, including sensitive intra-company email, financial transactions
such as mergers and acquisitions, business negotiations, personal medical
records, privileged correspondence with lawyers, and information about
crimes or civil rights violations.  The software will be particularly
useful to frequent wiretapping targets such as private companies competing
with government-owned companies, civil rights groups and lawyers,
opposition political parties, and dissidents. 

FreeS/WAN provides privacy for Internet packets using the proposed
standard Internet Protocol Security (IPSEC) protocols.  FreeS/WAN
negotiates strong keys using Diffie-Hellman key agreement with 1024-bit
keys, and encrypts each packet with 168-bit Triple-DES (3DES).  A modern
$500 PC can set up a tunnel in less than a second, and can encrypt
6 megabits of packets per second, easily handling the whole available
bandwidth at the vast majority of Internet sites.  In preliminary testing,
FreeS/WAN interoperated with 3DES IPSEC products from OpenBSD, PGP, SSH,
Cisco, Raptor, and Xedia.  Since FreeS/WAN is distributed as source code,
its innards are open to review by outside experts and sophisticated users,
reducing the chance of undetected bugs or hidden security compromises.

The software has been in development for several years.  It has been
funded by several philanthropists interested in increased privacy on
the Internet, including John Gilmore, co-founder of the Electronic
Frontier Foundation, a leading online civil rights group.

Press contacts:
Hugh Daniel,   +1 408 353 8124, hugh@toad.com
Henry Spencer, +1 416 690 6561, henry@spsystems.net

* FreeS/WAN derives its name from S/WAN, which is a trademark of RSA Data
  Security, Inc; used by permission.

The IPsec protocols

This section provides information on the IPsec protocols which FreeS/WAN implements. For more detail, see the RFCs .

The basic idea of IPsec is to provide security functions, authentication and encryption, at the IP (Internet Protocol) level. This requires a higher-level protocol (IKE) to set things up for the IP-level services (ESP and AH).

Protocols and phases

Three protocols are used in an IPsec implementation:

ESP, Encapsulating Security Payload

Encrypts and/or authenticates data

AH, Authentication Header

Provides a packet authentication service

IKE, Internet Key Exchange

Negotiates connection parameters, including keys, for the other two

The term "IPsec" (also written as IPSEC) is slightly ambiguous. In some contexts, it includes all three of the above but in other contexts it refers only to AH and ESP.

There is more detail below, but a quick summary of how the whole thing works is:

Phase one IKE (main mode exchange)

sets up a keying channel (ISAKMP SA) between the two gateways

Phase two IKE (quick mode exchange)

sets up data channels (IPsec SAs)

IPsec proper

exchanges data using AH or ESP

Both phases of IKE are repeated periodically to automate re-keying.

Applying IPsec

Authentication and encryption functions for network data can, of course, be provided at other levels. Many security protocols work at levels above IP.

• PGP encrypts and authenticates mail messages
• SSH authenticates remote logins and then encrypts the session
• SSL or TLS provides security at the sockets layer, e.g. for secure web browsing

and so on. Other techniques work at levels below IP. For example, data on a communications circuit or an entire network can be encrypted by specialised hardware. This is common practice in high-security applications.

Advantages of IPsec

There are, however, advantages to doing it at the IP level instead of, or as well as, at other levels.

IPsec is the most general way to provide these services for the Internet.

• Higher-level services protect a single protocol; for example PGP protects mail.
• Lower level services protect a single medium; for example a pair of encryption boxes on the ends of a line make wiretaps on that line useless unless the attacker is capable of breaking the encryption.

IPsec, however, can protect any protocol running above IP and any medium which IP runs over. More to the point, it can protect a mixture of application protocols running over a complex combination of media. This is the normal situation for Internet communication; IPsec is the only general solution.

IPsec can also provide some security services "in the background", with no visible impact on users. To use PGP encryption and signatures on mail, for example, the user must at least:

• remember his or her passphrase,
• keep it secure
• follow procedures to validate correspondents' keys

These systems can be designed so that the burden on users is not onerous, but any system will place some requirements on users. No such system can hope to be secure if users are sloppy about meeting those requirements. The author has seen username and password stuck on terminals with post-it notes in an allegedly secure environment, for example.

Limitations of IPsec

IPsec is designed to secure IP links between machines. It does that well, but it is important to remember that there are many things it does not do. Some of the important limitations are:

IPsec cannot be secure if your system isn't

System security on IPsec gateway machines is an essential requirement if IPsec is to function as designed. No system can be trusted if the underlying machine has been subverted. See books on Unix security such as Garfinkel and Spafford or our web references for Linux security or more general computer security.

Of course, there is another side to this. IPsec can be a powerful tool for improving system and network security. For example, requiring packet authentication makes various spoofing attacks harder and IPsec tunnels can be extremely useful for secure remote administration of various things.

IPsec is not end-to-end

IPsec cannot provide the same end-to-end security as systems working at higher levels. IPsec encrypts an IP connection between two machines, which is quite a different thing than encrypting messages between users or between applications.

For example, if you need mail encrypted from the sender's desktop to the recipient's desktop and decryptable only by the recipient, use PGP or another such system. IPsec can encrypt any or all of the links involved -- between the two mail servers, or between either server and its clients. It could even be used to secure a direct IP link from the sender's desktop machine to the recipient's, cutting out any sort of network snoop. What it cannot ensure is end-to-end user-to-user security. If only IPsec is used to secure mail, then anyone with appropriate privileges on any machine where that mail is stored (at either end or on any store-and-forward servers in the path) can read it.

In another common setup, IPsec encrypts packets at a security gateway machine as they leave the sender's site and decrypts them on arrival at the gateway to the recipient's site. This does provide a useful security service -- only encrypted data is passed over the Internet -- but it does not even come close to providing an end-to-end service. In particular, anyone with appropriate privileges on either site's LAN can intercept the message in unencrypted form.

IPsec cannot do everything

IPsec also cannot provide all the functions of systems working at higher levels of the protocol stack. If you need a document electronically signed by a particular person, then you need his or her digital signature and a public key cryptosystem to verify it with.

Note, however, that IPsec authentication of the underlying communication can make various attacks on higher-level protocols more difficult. In particular, authentication prevents man-in-the-middle attacks.

IPsec authenticates machines, not users

IPsec uses strong authentication mechanisms to control which messages go to which machines, but it does not have the concept of user ID, which is vital to many other security mechansims and policies. This means some care must be taken in fitting the various security mechansims on a network together. For example, if you need to control which users access your database server, you need some non-IPsec mechansim for that. IPsec can control which machines connect to the server, and can ensure that data transfer to those machines is done securely, but that is all. Either the machines themselves must control user access or there must be some form of user authentication to the database, independent of IPsec.

IPsec does not stop denial of service attacks

Denial of service attacks aim at causing a system to crash, overload, or become confused so that legitimate users cannot get whatever services the system is supposed to provide. These are quite different from attacks in which the attacker seeks either to use the service himself or to subvert the service into delivering incorrect results.

IPsec shifts the ground for DoS attacks; the attacks possible against systems using IPsec are different than those that might be used against other systems. It does not, however, eliminate the possibility of such attacks.

IPsec does not stop traffic analysis

Traffic analysis is the attempt to derive intelligence from messages without regard for their contents. In the case of IPsec, it would mean analysis based on things visible in the unencrypted headers of encrypted packets -- source and destination gateway addresses, packet size, et cetera. Given the resources to acquire such data and some skill in analysing it (both of which any national intelligence agency should have), this can be a very powerful technique.

IPsec is not designed to defend against this. Partial defenses are certainly possible, and some are described below, but it is not clear that any complete defense can be provided.

IPsec is a general mechanism for securing IP

While IPsec does not provide all functions of a mail encryption package, it can encrypt your mail. In particular, it can ensure that all mail passing between a pair or a group of sites is encrypted. An attacker looking only at external traffic, without access to anything on or behind the IPsec gateway, cannot read your mail. He or she is stymied by IPsec just as he or she would be by PGP.

The advantage is that IPsec can provide the same protection for anything transmitted over IP. In a corporate network example, PGP lets the branch offices exchange secure mail with head office. SSL and SSH allow them to securely view web pages, connect as terminals to machines, and so on. IPsec can support all those applications, plus database queries, file sharing (NFS or Windows), other protocols encapsulated in IP (Netware, Appletalk, ...), phone-over-IP, video-over-IP, ... anything-over-IP. The only limitation is that IP Multicast is not yet supported, though there are Internet Draft documents for that.

IPsec creates secure tunnels through untrusted networks . Sites connected by these tunnels form VPNs, Virtual Private Networks.

IPsec gateways can be installed wherever they are required.

• One organisation might choose to install IPsec only on firewalls between their LANs and the Internet. This would allow them to create a VPN linking several offices. It would provide protection against anyone outside their sites.
• Another might install IPsec on departmental servers so everything on the corporate backbone net was encrypted. This would protect messages on that net from everyone except the sending and receiving department.
• Another might be less concerned with information secrecy and more with controlling access to certain resources. They might use IPsec packet authentication as part of an access control mechanism, with or without also using the IPsec encryption service.
• It is even possible (assuming adequate processing power and an IPsec implementation in each node) to make every machine its own IPsec gateway so that everything on a LAN is encrypted. This protects information from everyone outside the sending and receiving machine.
• These techniques can be combined in various ways. One might, for example, require authentication everywhere on a network while using encryption only for a few links.

Which of these, or of the many other possible variants, to use is up to you. IPsec provides mechanisms; you provide the policy .

No end user action is required for IPsec security to be used; they don't even have to know about it. The site administrators, of course, do have to know about it and to put some effort into making it work. Poor administration can compromise IPsec as badly as the post-it notes mentioned above. It seems reasonable, though, for organisations to hope their system administrators are generally both more security-conscious than end users and more able to follow computer security procedures. If not, at least there are fewer of them to educate or replace.

IPsec can be, and often should be, used with along with security protocols at other levels. If two sites communicate with each other via the Internet, then IPsec is the obvious way to protect that communication. If two others have a direct link between them, either link encryption or IPsec would make sense. Choose one or use both. Whatever you use at and below the IP level, use other things as required above that level. Whatever you use above the IP level, consider what can be done with IPsec to make attacks on the higher levels harder. For example, man-in-the-middle attacks on various protocols become difficult if authentication at packet level is in use on the potential victims' communication channel.

Using authentication without encryption

Where appropriate, IPsec can provide authentication without encryption. One might do this, for example:

• where the data is public but one wants to be sure of getting the right data, for example on some web sites
• where encryption is judged unnecessary, for example on some company or department LANs
• where strong encryption is provided at link level, below IP
• where strong encryption is provided in other protocols, above IP
  Note that IPsec authentication may make some attacks on those protocols harder.

Authentication has lower overheads than encryption.

The protocols provide four ways to build such connections, using either an AH-only connection or ESP using null encryption, and in either manually or automatically keyed mode. FreeS/WAN supports only one of these, manually keyed AH-only connections, and we do not recommend using that. Our reasons are discussed under Resisting traffic analysis a few sections further along.

Encryption without authentication is dangerous

Originally, the IPsec encryption protocol ESP didn't do integrity checking. It only did encryption. Steve Bellovin found many ways to attack ESP used without authentication. See his paper Problem areas for the IP Security Protocols. To make a secure connection, you had to add an AH Authentication Header as well as ESP. Rather than incur the overhead of several layers (and rather than provide an ESP layer that didn't actually protect the traffic), the IPsec working group built integrity and replay checking directly into ESP.

Today, typical usage is one of:

• ESP for encryption and authentication
• AH for authentication alone

Other variants are allowed by the standard, but not much used:

ESP encryption without authentication

Bellovin has demonstrated fatal flaws in this. Do not use.

ESP encryption with AH authentication

This has higher overheads than using the authentication in ESP, and no obvious benefit in most cases. The exception might be a network where AH authentication was widely or universally used. If you're going to do AH to conform with network policy, why authenticate again in the ESP layer?

Authenticate twice, with AH and with ESP

Why? Of course, some folk consider "belt and suspenders" the sensible approach to security. If you're among them, you might use both protocols here. You might also use both to satisfy different parts of a security policy. For example, an organisation might require AH authentication everywhere but two users within the organisation might use ESP as well.

ESP authentication without encryption

The standard allows this, calling it "null encryption". FreeS/WAN does not support it. We recommend that you use AH instead if authentication is all you require. AH authenticates parts of the IP header, which ESP-null does not do.

Some of these variants cannot be used with FreeS/WAN because we do not support ESP-null and do not support automatic keying of AH-only connections.

There are fairly frequent suggestions that AH be dropped entirely from the IPsec specifications since ESP and null encryption can handle that situation. It is not clear whether this will occur. My guess is that it is unlikely.

Multiple layers of IPsec processing are possible

The above describes combinations possible on a single IPsec connection. In a complex network you may have several layers of IPsec in play, with any of the above combinations at each layer.

For example, a connection from a desktop machine to a database server might require AH authentication. Working with other host, network and database security measures, AH might be just the thing for access control. You might decide not to use ESP encryption on such packets, since it uses resources and might complicate network debugging. Within the site where the server is, then, only AH would be used on those packets.

Users at another office, however, might have their whole connection (AH headers and all) passing over an IPsec tunnel connecting their office to the one with the database server. Such a tunnel should use ESP encryption and authentication. You need authentication in this layer because without authentication the encryption is vulnerable and the gateway cannot verify the AH authentication. The AH is between client and database server; the gateways aren't party to it.

In this situation, some packets would get multiple layers of IPsec applied to them, AH on an end-to-end client-to-server basis and ESP from one office's security gateway to the other.

Resisting traffic analysis

Traffic analysis is the attempt to derive useful intelligence from encrypted traffic without breaking the encryption.

Is your CEO exchanging email with a venture capital firm? With bankruptcy trustees? With an executive recruiting agency? With the holder of some important patents? If an eavesdropper learns about any of those, then he has interesting intelligence on your company, whether or not he can read the messages themselves.

Even just knowing that there is network traffic between two sites may tell an analyst something useful, especially when combined with whatever other information he or she may have. For example, if you know Company A is having cashflow problems and Company B is looking for aquisitions, then knowing that packets are passing between the two is interesting. It is more interesting if you can tell it is email, and perhaps yet more if you know the sender and recipient.

Except in the simplest cases, traffic analysis is hard to do well. It requires both considerable resources and considerable analytic skill. However, intelligence agencies of various nations have been doing it for centuries and many of them are likely quite good at it by now. Various commercial organisations, especially those working on "targeted marketing" may also be quite good at analysing certain types of traffic.

In general, defending against traffic analysis is also difficult. Inventing a really good defense could get you a PhD and some interesting job offers.

IPsec is not designed to stop traffic analysis and we know of no plausible method of extending it to do so. That said, there are ways to make traffic analysis harder. This section describes them.

Using "unnecessary" encryption

One might choose to use encryption even where it appears unnecessary in order to make analysis more difficult. Consider two offices which pass a small volume of business data between them using IPsec and also transfer large volumes of Usenet news. At first glance, it would seem silly to encrypt the newsfeed, except possibly for any newsgroups that are internal to the company. Why encrypt data that is all publicly available from many sites?

However, if we encrypt a lot of news and send it down the same connection as our business data, we make traffic analysis much harder. A snoop cannot now make inferences based on patterns in the volume, direction, sizes, sender, destination, or timing of our business messages. Those messages are hidden in a mass of news messages encapsulated in the same way.

If we're going to do this we need to ensure that keys change often enough to remain secure even with high volumes and with the adversary able to get plaintext of much of the data. We also need to look at other attacks this might open up. For example, can the adversary use a chosen plaintext attack, deliberately posting news articles which, when we receive and encrypt them, will help break our encryption? Or can he block our business data transmission by flooding us with silly news articles? Or ...

Also, note that this does not provide complete protection against traffic analysis. A clever adversary might still deduce useful intelligence from statistical analysis (perhaps comparing the input newsfeed to encrypted output, or comparing the streams we send to different branch offices), or by looking for small packets which might indicate establishment of TCP connections, or ...

As a general rule, though, to improve resistance to traffic analysis, you should encrypt as much traffic as possible, not just as much as seems necessary.

Using multiple encryption

This also applies to using multiple layers of encryption. If you have an IPsec tunnel between two branch offices, it might appear silly to send PGP-encrypted email through that tunnel. However, if you suspect someone is snooping your traffic, then it does make sense:

• it protects the mail headers; they cannot even see who is mailing who
• it protects against user bungles or software malfunctions that accidentally send messages in the clear
• it makes any attack on the mail encryption much harder; they have to break IPsec or break into your network before they can start on the mail encryption

Similar arguments apply for SSL-encrypted web traffic or SSH-encrypted remote login sessions, even for end-to-end IPsec tunnels between systems in the two offices.

Using fewer tunnels

It may also help to use fewer tunnels. For example, if all you actually need encrypted is connections between:

• mail servers at branch and head offices
• a few branch office users and the head office database server

You might build one tunnel per mail server and one per remote database user, restricting traffic to those applications. This gives the traffic analyst some information, however. He or she can distinguish the tunnels by looking at information in the ESP header and, given that distinction and the patterns of tunnel usage, might be able to figure out something useful. Perhaps not, but why take the risk?

We suggest instead that you build one tunnel per branch office, encrypting everything passing from head office to branches. This has a number of advantages:

• it is easier to build and administer
• it resists traffic analysis somewhat better
• it provides security for whatever you forgot. For example, if some user at a remote office browses proprietary company data on some head office web page (that the security people may not even know about!), then that data is encrypted before it reaches the Internet.

Of course you might also want to add additional tunnels. For example, if some of the database data is confidential and should not be exposed even within the company, then you need protection from the user's desktop to the database server. We suggest you do that in whatever way seems appropriate -- IPsec, SSH or SSL might fit -- but, whatever you choose, pass it between locations via a gateway-to-gateway IPsec tunnel to provide some resistance to traffic analysis.

Cryptographic components

IPsec combines a number of cryptographic techniques, all of them well-known and well-analyzed. The overall design approach was conservative; no new or poorly-understood components were included.

This section gives a brief overview of each technique. It is intended only as an introduction. There is more information, and links to related topics, in our glossary. See also our bibliography and cryptography web links.

Block ciphers

The encryption in the ESP encapsulation protocol is done with a block cipher .

We do not implement single DES. It is insecure. Our default, and currently only, block cipher is triple DES.

The Rijndael block cipher has won the AES competition to choose a relacement for DES. It will almost certainly be added to FreeS/WAN and to other IPsec implementations. Patches are already available.

Hash functions

The HMAC construct

IPsec packet authentication is done with the HMAC construct. This is not just a hash of the packet data, but a more complex operation which uses both a hashing algorithm and a key. It therefore does more than a simple hash would. A simple hash would only tell you that the packet data was not changed in transit, or that whoever changed it also regenerated the hash. An HMAC also tells you that the sender knew the HMAC key.

For IPsec HMAC, the output of the hash algorithm is truncated to 96 bits. This saves some space in the packets. More important, it prevents an attacker from seeing all the hash output bits and perhaps creating some sort of attack based on that knowledge.

Choice of hash algorithm

The IPsec RFCs require two hash algorithms -- MD5 and SHA-1 -- both of which FreeS/WAN implements.

Various other algorithms -- such as RIPEMD and Tiger -- are listed in the RFCs as optional. None of these are in the FreeS/WAN distribution, or are likely to be added, although user patches exist for several of them.

Additional hash algorithms -- SHA-256, SHA-384 and SHA-512 -- may be required to give hash strength matching the strength of AES. These are likely to be added to FreeS/WAN along with AES.

Diffie-Hellman key agreement

The Diffie-Hellman key agreement protocol allows two parties (A and B or Alice and Bob) to agree on a key in such a way that an eavesdropper who intercepts the entire conversation cannot learn the key.

The protocol is based on the discrete logarithm problem and is therefore thought to be secure. Mathematicians have been working on that problem for years and seem no closer to a solution, though there is no proof that an efficient solution is impossible.

RSA authentication

The RSA algorithm (named for its inventors -- Rivest, Shamir and Adleman) is a very widely used public key cryptographic technique. It is used in IPsec as one method of authenticating gateways for Diffie-Hellman key negotiation.

Structure of IPsec

There are three protocols used in an IPsec implementation:

ESP, Encapsulating Security Payload

Encrypts and/or authenticates data

AH, Authentication Header

Provides a packet authentication service

IKE, Internet Key Exchange

Negotiates connection parameters, including keys, for the other two

The term "IPsec" is slightly ambiguous. In some contexts, it includes all three of the above but in other contexts it refers only to AH and ESP.

IKE (Internet Key Exchange)

The IKE protocol sets up IPsec (ESP or AH) connections after negotiating appropriate parameters (algorithms to be used, keys, connection lifetimes) for them. This is done by exchanging packets on UDP port 500 between the two gateways.

IKE (RFC 2409) was the outcome of a long, complex process in which quite a number of protocols were proposed and debated. Oversimplifying mildly, IKE combines:

ISAKMP (RFC 2408)

The Internet Security A ssociation and Key Management Protocol manages negotiation of connections and defines SAs (Security Associations) as a means of describing connection properties.

IPsec DOI for ISAKMP (RFC 2407)

A Domain Of I nterpretation fills in the details necessary to turn the rather abstract ISAKMP protocol into a more tightly specified protocol, so it becomes applicable in a particular domain.

Oakley key determination protocol (RFC 2412)

Oakley creates keys using the Diffie-Hellman key agreement protocol.

For all the details, you would need to read the four RFCs just mentioned (over 200 pages) and a number of others. We give a summary below, but it is far from complete.

Phases of IKE

IKE negotiations have two phases.

Phase one

The two gateways negotiate and set up a two-way ISAKMP SA which they can then use to handle phase two negotiations. One such SA between a pair of gateways can handle negotiations for multiple tunnels.

Phase two

Using the ISAKMP SA, the gateways negotiate IPsec (ESP and/or AH) SAs as required. IPsec SAs are unidirectional (a different key is used in each direction) and are always negotiated in pairs to handle two-way traffic. There may be more than one pair defined between two gateways.

Both of these phases use the UDP protocol and port 500 for their negotiations.

After both IKE phases are complete, you have IPsec SAs to carry your encrypted data. These use the ESP or AH protocols. These protocols do not have ports. Ports apply only to UDP or TCP.

The IKE protocol is designed to be extremely flexible. Among the things that can be negotiated (separately for each SA) are:

• SA lifetime before rekeying
• encryption algorithm used. We currently support only triple DES. Single DES is insecure. The RFCs say you MUST do DES, SHOULD do 3DES and MAY do various others. We do not do any of the others.
• authentication algorithms. We support MD5 and SHA. These are the two the RFCs require.
• choice of group for Diffie-Hellman key agreement. We currently support Groups 2 and 5 (which are defined modulo primes of various lengths) and do not support Group 1 (defined modulo a shorter prime, and therefore cryptographically weak) or groups 3 and 4 (defined using elliptic curves). The RFCs require only Group 1.

The protocol also allows implementations to add their own encryption algorithms, authentication algorithms or Diffie-Hellman groups. We do not support any such extensions, but there are some patches that do.

There are a number of complications:

• The gateways must be able to authenticate each other's identities before they can create a secure connection. This host authentication is part of phase one negotiations, and is a required prerequisite for packet authentication used later. Host authentication can be done in a variety of ways. Those supported by FreeS/WAN are discussed in our advanced configuration document.
• Phase one can be done in two ways.
  • Main Mode is required by the RFCs and supported in FreeS/WAN. It uses a 6-packet exzchange.
  • Aggressive Mode is somewhat faster (only 3 packets) but reveals more to an eavesdropper. This is optional in the RFCs, not currently supported by FreeS/WAN, and not likely to be.
• A new group exchange may take place after phase one but before phase two, defining an additional group for use in the Diffie-Hellman key agreement part of phase two. FreeS/WAN does not currently support this.
• Phase two always uses Quick Mode, but there are two variants of that:
  • One variant provides Perfect Forward Secrecy (PFS) . An attacker that obtains your long-term host authentication key does not immediately get any of your short-term packet encryption of packet authentication keys. He must conduct another successful attack each time you rekey to get the short-term keys. Having some short-term keys does not help him learn others. In particular, breaking your system today does not let him read messages he archived yestarday, assuming you've changed short-term keys in the meanwhile. We enable PFS as the default.
  • The other variant disables PFS and is therefore slightly faster. We do not recommend this since it is less secure, but FreeS/WAN does support it. You can enable it with a pfs=no statement in ipsec.conf(5).
  • The protocol provides no way to negotiate which variant will be used. If one gateway is set for PFS and the other is not, the negotiation fails. This has proved a fairly common source of interoperation problems.
• Several types of notification message may be sent by either side during either phase, or later. FreeS/WAN does not currently support these, but they are a likely addition in future releases.
• There is a commit flag which may optionally be set on some messages. The errata page for the RFCs includes two changes related to this, one to clarify the description of its use and one to block a denial of service attack which uses it. We currently do not implement this feature.

These complications can of course lead to problems, particularly when two different implementations attempt to interoperate. For example, we have seen problems such as:

• Some implementations (often products crippled by export laws) have the insecure DES algorithm as their only supported encryption method. Other parts of our documentation discuss the reasons we do not implement single DES, and how to cope with crippled products .
• Windows 2000 IPsec tries to negotiate using Aggressive Mode, which we don't support. Later on, it uses the commit bit, which we also don't support.
• Various implementations disable PFS by default, and therefore will not talk to FreeS/WAN until you either turn on PFS on their end or turn it off in FreeS/WAN with a pfs=no entry in the connection description.
• FreeS/WAN's interaction with PGPnet is complicated by their use of notification messages we do not yet support.

Despite this, we do interoperate successfully with many implementations, including both Windows 2000 and PGPnet. Details are in our interoperability document.

Sequence of messages in IKE

Each phase (see previous section)of IKE involves a series of messages. In Pluto error messages, these are abbreviated using:

M

Main mode, settting up the keying channel (ISAKMP SA)

Q

Quick mode, setting up the data channel (IPsec SA)

I

Initiator, the machine that starts the negotiation

R

Responder

For example, the six messages of a main mode negotiation, in sequence, are labelled:

       MI1 ---------->
           <---------- MR1
       MI2 ----------> 
           <---------- MR2
       MI3 ---------->
           <---------- MR3

Structure of IKE messages

Here is our Pluto developer explaining some of this on the mailing list:

When one IKE system (for example, Pluto) is negotiating with another
to create an SA, the Initiator proposes a bunch of choices and the
Responder replies with one that it has selected.

The structure of the choices is fairly complicated.  An SA payload
contains a list of lists of "Proposals".  The outer list is a set of
choices: the selection must be from one element of this list.

Each of these elements is a list of Proposals.  A selection must be
made from each of the elements of the inner list.  In other words,
*all* of them apply (that is how, for example, both AH and ESP can
apply at once).

Within each of these Proposals is a list of Transforms.  For each
Proposal selected, one Transform must be selected (in other words,
each Proposal provides a choice of Transforms).

Each Transform is made up of a list of Attributes describing, well,
attributes.  Such as lifetime of the SA.  Such as algorithm to be
used.  All the Attributes apply to a Transform.

You will have noticed a pattern here: layers alternate between being
disjunctions ("or") and conjunctions ("and").

For Phase 1 / Main Mode (negotiating an ISAKMP SA), this structure is
cut back.  There must be exactly one Proposal.  So this degenerates to
a list of Transforms, one of which must be chosen.

IPsec Services, AH and ESP

IPsec offers two services, authentication and encryption. These can be used separately but are often used together.

Authentication

Packet-level authentication allows you to be confident that a packet came from a particular machine and that its contents were not altered en route to you. No attempt is made to conceal or protect the contents, only to assure their integrity. Packet authentication can be provided separately using an Authentication Header, described just below, or it can be included as part of the ESP (Encapsulated Security Payload) service, described in the following section. That service offers encryption as well as authentication. In either case, the HMAC construct is used as the authentication mechanism.

There is a separate authentication operation at the IKE level, in which each gateway authenticates the other. This can be done in a variety of ways.

Encryption

Encryption allows you to conceal the contents of a message from eavesdroppers.

In IPsec this is done using a block cipher (normally Triple DES for Linux). In the most used setup, keys are automatically negotiated, and periodically re-negotiated, using the IKE (Internet Key Exchange) protocol. In Linux FreeS/WAN this is handled by the Pluto Daemon.

The IPsec protocol offering encryption is ESP, Encapsulated Security Payload. It can also include a packet authentication service.

Note that encryption should always be used with some packet authentication service. Unauthenticated encryption is vulnerable to man-in-the-middle attacks. Also note that encryption does not prevent traffic analysis.