VPN(8) BSD System Manager's Manual VPN(8)NAMEvpn - configuring the system for virtual private networks
DESCRIPTION
A Virtual Private Network (VPN) is used to securely connect two or more
subnets over the internet. For each subnet there is a security gateway
which is linked via a cryptographically secured tunnel to the security
gateway of the other subnet. ipsec(4) is used to provide the necessary
network-layer cryptographic services. This document describes the confi-
guration process for setting up a VPN.
Briefly, creating a VPN consists of the following steps:
1. Enable packet forwarding.
2. Choose a key exchange method: manual or automated.
3. For manual keying, generate the keys.
4. For manual keying, create the Security Associations (SA).
5. For manual keying, create the appropriate IPsec flows.
6. For automated keying, configure the keying daemon.
7. Configure firewall rules appropriately.
8. Enable the packet filter.
9. For automated keying, start the keying daemon.
10. Test the setup.
About this page
It is recommended that a test setup be created before attempting to de-
ploy a VPN on the internet. The examples in this page can be done using
two machines directly connected to each other, and a little imagination.
The IP address of each machine represents a gateway address; the alias
(see below) is simply a hook into a fictitious network.
The following steps are only necessary if the VPN is being set up as a
test VPN, on an internal LAN.
The VPN can be represented using two machines (A and B). An alias should
be added to each machine, to give it the appearance of being in another
network.
On machine A:
# ifconfig ne0 192.168.1.13 description "Machine A"
# ifconfig ne0 alias 10.0.50.1
On machine B:
# ifconfig bge0 192.168.1.15 description "Machine B"
# ifconfig bge0 alias 10.0.99.1
For all other (non-test) cases, ifconfig(8) should be used to configure
machines as normal.
Additionally, the GATEWAY_* and NETWORK_* variables used in the following
sections are defined below in Configuring Firewall Rules. Please see that
section for the correct values for these variables.
Enabling Packet Forwarding
For security gateways, proper operation often requires packet forwarding
to be enabled using sysctl(8):
# sysctl net.inet.ip.forwarding=1
# sysctl net.inet6.ip6.forwarding=1
Packet forwarding defaults to 'off'.
Additionally, if net.inet.ip.forwarding is set to 2, IP forwarding is
restricted to IPsec traffic only. These and other IPsec related options
are documented in sysctl(3).
For more permanent operation, the appropriate option(s) can be enabled in
sysctl.conf(5).
Choosing a Key Exchange Method
There are currently two key exchange methods available:
+ manual keying: ipsecadm(8)
+ automated keying: isakmpd(8)
Generating Manual Keys [manual keying]
The shared secret symmetric keys used to create a VPN can be any hexade-
cimal value, so long as both sides of the connection use the same values.
Since the security of the VPN is based on these keys being unguessable,
it is very important that the keys be chosen using a strong random
source. One practical method of generating them is by using the random(4)
device. To produce 160 bits (20 bytes) of randomness, for example, do:
# openssl rand 20 | hexdump -ve '20/1 "%02X"'
or:
# openssl rand 20 | perl -pe 's/./unpack("H2",$&)/ges'
Different cipher types may require different sized keys.
Cipher Key Length
DES 56 bits
3DES 168 bits
AES Variable (128 bits recommended)
BLF Variable (160 bits recommended)
CAST Variable (128 bits maximum and recommended)
SKIPJACK 80 bits
Use of DES or SKIPJACK as an encryption algorithm is not recommended (ex-
cept for backwards compatibility) due to their short key length. Further-
more, recent attacks on SKIPJACK have shown severe weaknesses in its
structure.
Note that DES requires 8 bytes to form a 56-bit key and 3DES requires 24
bytes to form its 168-bit key. This is because the most significant bit
of each byte is ignored by both algorithms.
The following would create suitable keys for a 3DES encryption key and
SHA-1 authentication key:
# openssl rand 24 | hexdump -ve '24/1 "%02X"' > enc_key
# openssl rand 20 | hexdump -ve '20/1 "%02X"' > auth_key
The 3DES encryption key needs 192 bits (3x64), or 24 bytes. The SHA-1 au-
thentication key needs 160 bits, or 20 bytes.
Creating Security Associations [manual keying]
Before the IPsec flows can be defined, two Security Associations (SAs)
must be defined on each end of the VPN e.g.:
# ipsecadm new esp -src $GATEWAY_A -dst $GATEWAY_B \
-spi $SPI_AB -forcetunnel -enc 3des -auth sha1 \
-keyfile $ENCRYPTION_KEY_FILE \
-authkeyfile $AUTHENTICATION_KEY_FILE
# ipsecadm new esp -src $GATEWAY_B -dst $GATEWAY_A \
-spi $SPI_BA -forcetunnel -enc 3des -auth sha1 \
-keyfile $ENCRYPTION_KEY_FILE \
-authkeyfile $AUTHENTICATION_KEY_FILE
Note that the -key and -authkey options may be used to specify the keys
directly in the ipsecadm(8) command line. However, another user could
view the keys by using the ps(1) command at the appropriate time (or use
a program for doing so).
Creating IPsec Flows [manual keying]
Both IPsec gateways need to configure ipsec(4) routes (flows) with the
ipsecadm(8) tool. Two flows are created on each machine: the first is for
outbound flows, the second is the ingress filter for the incoming securi-
ty association.
On the security gateway of subnet A:
# ipsecadm flow -out -require -proto esp \
-src $GATEWAY_A -dst $GATEWAY_B \
-addr $NETWORK_A $NETWORK_B
# ipsecadm flow -in -require -proto esp \
-src $GATEWAY_A -dst $GATEWAY_B \
-addr $NETWORK_B $NETWORK_A
On the security gateway of subnet B:
# ipsecadm flow -out -require -proto esp \
-src $GATEWAY_B -dst $GATEWAY_A \
-addr $NETWORK_B $NETWORK_A
# ipsecadm flow -in -require -proto esp \
-src $GATEWAY_B -dst $GATEWAY_A \
-addr $NETWORK_A $NETWORK_B
Configuring the Keying Daemon [automated keying]
Unless manual keying is used, both security gateways need to use the
isakmpd(8) key management daemon. isakmpd(8) implements security policy
using the KeyNote trust management system.
To create a VPN between the same two C class networks as the example
above, using isakmpd(8):
1. Create /etc/isakmpd/isakmpd.conf for machine A:
# Filter incoming phase 1 negotiations so they are only
# valid if negotiating with this local address.
[General]
Listen-On= 192.168.1.13
# Incoming phase 1 negotiations are multiplexed on the
# source IP address. Phase 1 is used to set up a protected
# channel just between the two gateway machines.
# This channel is then used for the phase 2 negotiation
# traffic (i.e. encrypted & authenticated).
[Phase 1]
192.168.1.15= peer-machineB
# 'Phase 2' defines which connections the daemon
# should establish. These connections contain the actual
# "IPsec VPN" information.
[Phase 2]
Connections= VPN-A-B
# ISAKMP phase 1 peers (from [Phase 1])
[peer-machineB]
Phase= 1
Transport= udp
Address= 192.168.1.15
Configuration= Default-main-mode
Authentication= yoursharedsecret
# IPSEC phase 2 connections (from [Phase 2])
[VPN-A-B]
Phase= 2
ISAKMP-peer= peer-machineB
Configuration= Default-quick-mode
Local-ID= machineA-internal-network
Remote-ID= machineB-internal-network
# ID sections (as used in [VPN-A-B])
[machineA-internal-network]
ID-type= IPV4_ADDR_SUBNET
Network= 10.0.50.0
Netmask= 255.255.255.0
[machineB-internal-network]
ID-type= IPV4_ADDR_SUBNET
Network= 10.0.99.0
Netmask= 255.255.255.0
# Main and Quick Mode descriptions
# (as used by peers and connections).
[Default-main-mode]
DOI= IPSEC
EXCHANGE_TYPE= ID_PROT
Transforms= 3DES-SHA,BLF-SHA
[Default-quick-mode]
DOI= IPSEC
EXCHANGE_TYPE= QUICK_MODE
Suites= QM-ESP-3DES-SHA-SUITE
2. Create /etc/isakmpd/isakmpd.conf for machine B:
# Filter incoming phase 1 negotiations so they are only
# valid if negotiating with this local address.
[General]
Listen-On= 192.168.1.15
# Incoming phase 1 negotiations are multiplexed on the
# source IP address. Phase 1 is used to set up a protected
# channel just between the two gateway machines.
# This channel is then used for the phase 2 negotiation
# traffic (i.e. encrypted & authenticated).
[Phase 1]
192.168.1.13= peer-machineA
# 'Phase 2' defines which connections the daemon
# should establish. These connections contain the actual
# "IPsec VPN" information.
[Phase 2]
Connections= VPN-B-A
# ISAKMP phase 1 peers (from [Phase 1])
[peer-machineA]
Phase= 1
Transport= udp
Address= 192.168.1.13
Configuration= Default-main-mode
Authentication= yoursharedsecret
# IPSEC phase 2 connections (from [Phase 2])
[VPN-B-A]
Phase= 2
ISAKMP-peer= peer-machineA
Configuration= Default-quick-mode
Local-ID= machineB-internal-network
Remote-ID= machineA-internal-network
# ID sections (as used in [VPN-A-B])
[machineA-internal-network]
ID-type= IPV4_ADDR_SUBNET
Network= 10.0.50.0
Netmask= 255.255.255.0
[machineB-internal-network]
ID-type= IPV4_ADDR_SUBNET
Network= 10.0.99.0
Netmask= 255.255.255.0
# Main and Quick Mode descriptions
# (as used by peers and connections).
[Default-main-mode]
DOI= IPSEC
EXCHANGE_TYPE= ID_PROT
Transforms= 3DES-SHA,BLF-SHA
[Default-quick-mode]
DOI= IPSEC
EXCHANGE_TYPE= QUICK_MODE
Suites= QM-ESP-3DES-SHA-SUITE
3. Read through the configuration one more time. The only real differ-
ences between the two files in this example are the IP addresses,
and ordering of Local-ID and Remote-ID for the VPN itself. Note that
the shared secret (the Authentication tag) must match between
machineA and machineB.
Due to the sensitive information contained in the configuration
file, it must be owned by root and installed without any permissions
for "group" or "other".
# chown root:wheel /etc/isakmpd/isakmpd.conf
# chmod 0600 /etc/isakmpd/isakmpd.conf
4. Create a simple /etc/isakmpd/isakmpd.policy file for both machine A
and machine B (identical):
Keynote-version: 2
Authorizer: "POLICY"
Conditions: app_domain == "IPsec policy" &&
esp_present == "yes" &&
esp_enc_alg != "null" -> "true";
Due to the sensitive information contained in the policy file, it
must be owned by root and installed without any permissions for
"group" or "other".
# chown root:wheel /etc/isakmpd/isakmpd.policy
# chmod 0600 /etc/isakmpd/isakmpd.policy
Configuring Firewall Rules
pf(4) needs to be configured such that all packets from the outside are
blocked by default. Only successfully IPsec-processed packets (those on
the enc(4) interface) or key management packets (for automated keying,
UDP packets with source and destination ports of 500) should be allowed
to pass.
Additional filter rules may be present for other traffic, though care
should be taken that other rules do not leak IPsec traffic. NAT rules can
also be used on the enc(4) interface.
Note: The examples in this page describe a test setup on an internal LAN,
using private (non-routable) IP addresses. In a typical setup, at least
GATEWAY_A and GATEWAY_B would be configured using public (routable) IP
addresses. NETWORK_A and NETWORK_B may or may not use public IP ad-
dresses, depending on the network.
The pf.conf(5) rules for a tunnel which uses encryption (the ESP IPsec
protocol) and isakmpd(8) on security gateway A might look like this:
GATEWAY_A = "192.168.1.13"
GATEWAY_B = "192.168.1.15"
NETWORK_A = "10.0.50.0/24"
NETWORK_B = "10.0.99.0/24"
ext_if="ne0"
# default deny
# $ext_if is the only interface going to the outside.
block log on { enc0, $ext_if } all
# Passing in encrypted traffic from security gateways
pass in proto esp from $GATEWAY_B to $GATEWAY_A
pass out proto esp from $GATEWAY_A to $GATEWAY_B
# Need to allow ipencap traffic on enc0.
pass in on enc0 proto ipencap from $GATEWAY_B to $GATEWAY_A
# Passing in traffic from the designated subnets.
pass in on enc0 from $NETWORK_B to $NETWORK_A
pass out on enc0 from $NETWORK_A to $NETWORK_B
# Passing in isakmpd(8) traffic from the security gateways
pass in on $ext_if proto udp from $GATEWAY_B port = 500 \
to $GATEWAY_A port = 500
pass out on $ext_if proto udp from $GATEWAY_A port = 500 \
to $GATEWAY_B port = 500
The pf.conf(5) rules on security gateway B might look like this:
GATEWAY_A = "192.168.1.13"
GATEWAY_B = "192.168.1.15"
NETWORK_A = "10.0.50.0/24"
NETWORK_B = "10.0.99.0/24"
ext_if="bge0"
# default deny
# $ext_if is the only interface going to the outside.
block log on { enc0, $ext_if } all
# Passing in encrypted traffic from security gateways
pass in proto esp from $GATEWAY_A to $GATEWAY_B
pass out proto esp from $GATEWAY_B to $GATEWAY_A
# Need to allow ipencap traffic on enc0.
pass in on enc0 proto ipencap from $GATEWAY_A to $GATEWAY_B
# Passing in traffic from the designated subnets.
pass in on enc0 from $NETWORK_A to $NETWORK_B
pass out on enc0 from $NETWORK_B to $NETWORK_A
# Passing in isakmpd(8) traffic from the security gateways
pass in on $ext_if proto udp from $GATEWAY_A port = 500 \
to $GATEWAY_B port = 500
pass out on $ext_if proto udp from $GATEWAY_B port = 500 \
to $GATEWAY_A port = 500
Enabling the Packet Filter
Enable the packet filter and load the ruleset:
# pfctl -e
# pfctl -f /etc/pf.conf
Starting the Keying Daemon [automated keying]
Start isakmpd(8)
On both machines, run:
# /sbin/isakmpd
To run with verbose debugging enabled, instead start with:
# /sbin/isakmpd -d -DA=99
Testing the Setup
It is important to check the setup is working correctly. Remember that
the following examples illustrate a test setup only, and therefore tests
carried out on GATEWAY_A and NETWORK_A will be carried out on the same
machine (Machine A). If this were a real setup, GATEWAY_A and a machine
on NETWORK_A would be different machines.
Using the test setup, first check the routing table shows the routes
between the two gateways.
On GATEWAY_A:
$ netstat -rn -f encap
Routing tables
Encap:
Source Port Destination Port Proto SA(Address/Proto/Type/Direction)
10.0.99/24 0 10.0.50/24 0 0 192.168.1.15/50/use/in
10.0.50/24 0 10.0.99/24 0 0 192.168.1.15/50/require/out
This shows that anything with source address 10.0.99.0/24 (NETWORK_B) is
routed to destination 10.0.50.0/24 (NETWORK_A), and vice versa. The oppo-
site would be true if netstat(1) were run on GATEWAY_B.
Note that the routing table above is given for an automated keying ses-
sion. SA information for a manual keying session would differ slightly:
the "Type" field would be "require" for both directions.
Next check that you can ping(8) the networks:
On NETWORK_A:
$ ping -I 10.0.50.1 10.0.99.1
Note the -I option passed to ping(8): this is necessary to specify a
source address from the network. Check that the ping(8) works from both
NETWORK_A and NETWORK_B, changing the arguments as necessary.
Check that the traffic between the two networks really is ESP encapsulat-
ed. On GATEWAY_A:
# tcpdump -n -i ne0 esp
On NETWORK_A:
$ ping -I 10.0.50.1 10.0.99.1
Check that tcpdump(8) shows ESP packets whilst the ping is in progress.
That shows that the traffic is IPsec encapsulated.
If both networks are pingable, the routing tables look as described
above, and tcpdump(8) is working as described, it means the VPN is work-
ing correctly. However, it is also important to check that no IPsec
traffic is being leaked, either by badly designed firewall rules or by a
misconfigured VPN setup.
On GATEWAY_A:
# tcpdump -n -i ne0 not esp and host 192.168.1.15
On NETWORK_A:
$ ping -I 10.0.50.1 10.0.99.1
This time tcpdump(8) has been instructed to ignore ESP packets going to
host 192.168.1.15 (GATEWAY_B), and no traffic should be seen whilst the
ping is running. One exception to this is if the automated keying setup
has been followed, in which case isakmpd(8) key management packets on UDP
port 500 may be seen. This is perfectly normal. If any traffic is being
leaked i.e. the last ping detailed above is showing traffic, it is sug-
gested that the administrator review the steps above, paying particular
notice to the firewall configuration procedures.
FILES
/etc/isakmpd/isakmpd.conf isakmpd(8) configuration file.
/etc/isakmpd/isakmpd.policy isakmpd(8) policy file.
/etc/pf.conf Firewall configuration file.
/usr/share/ipsec/rc.vpn Sample VPN configuration file.
SEE ALSOnetstat(1), openssl(1), sysctl(3), enc(4), ipsec(4), keynote(4),
isakmpd.conf(5), isakmpd.policy(5), pf.conf(5), ifconfig(8), ipsecadm(8),
isakmpd(8), pfctl(8), ping(8), sysctl(8), tcpdump(8)MirOS BSD #10-current February 9, 1999 7
