ipsec.conf - Man Page

The following parameters are relevant to IKE automatic keying. Unless otherwise noted, for a connection to work, in general it is necessary for the two ends to agree exactly on the values of these parameters.

keyexchange

Whether to use IKEv2 (RFC 7296) or IKEv1 (RFC 4301). Currently the accepted values are ikev2 (the default) and ikev1.

This option replaces ikev2=yes|no. Old values yes, propose and instist all map to keyexchange=ikev2. Values no and permit are map to keyexchange=ikev1.

hostaddrfamily

the address family of the hosts; currently the accepted values are ipv4 and ipv6. The default is to detect this based on the IP addresses specified or the IP addresses resolved, so this option is not needed, unless you specify hostnames that resolve to both IPv4 and IPv6.

This option used to be named connaddrfamily but its use was broken so it was obsoleted in favour or using the new hostaddrfamily and clientaddrfamily.

clientaddrfamily

the address family of the clients (subnets); currently the accepted values are ipv4 and ipv6. The default is to detect this based on the network IP addresses specified or the network IP addresses resolved, so this option is not needed, unless you specify names that resolve to both IPv4 and IPv6.

type

the type of the connection; currently the accepted values are tunnel (the default) signifying a host-to-host, host-to-subnet, or subnet-to-subnet tunnel; transport, signifying host-to-host transport mode; passthrough, signifying that no IPsec processing should be done at all; drop, signifying that packets should be discarded; and reject, signifying that packets should be discarded and a diagnostic ICMP returned.

left

(required) the IP address or DNS hostname of the left participant's public-network interface, Currently, IPv4 and IPv6 IP addresses are supported. If a DNS hostname is used, it will be resolved to an IP address on load time, and whenever a connection is rekeying or restarting (such as when restarted via a DPD failure detection). This allows one to use a DNS hostname when the endpoint is on a dynamic IP address.

There are several magic values. If it is %defaultroute, left will be filled in automatically with the local address of the default-route interface (as determined at IPsec startup time); this also overrides any value supplied for leftnexthop. (Either left or right may be %defaultroute, but not both.) The value %any signifies an address to be filled in (by automatic keying) during negotiation. The value %opportunistic signifies that both left and leftnexthop are to be filled in (by automatic keying) from DNS data for left's client. The value can also contain the interface name, which will then later be used to obtain the IP address from to fill in. For example %ppp0.

The values %group and %opportunisticgroup makes this a policy group conn: one that will be instantiated into a regular or opportunistic conn for each CIDR block listed in the policy group file with the same name as the conn.

If using IP addresses in combination with NAT, always use the actual local machine's (NATed) IP address, and if the remote (eg right=) is NATed as well, the remote's public (not NATed) IP address. Note that this makes the configuration no longer symmetrical on both sides, so you cannot use an identical configuration file on both hosts.

leftsubnet

A comma separated list of traffic selectors behind the left participant. Each expressed as: network-prefix / netmask/ protocol/ port where trailing elements may be omitted. For instance: leftsubnet=1.2.3.0/24,1:2::/64, leftsubnet=1.2.3.4/32/tcp,1:2::/128/tcp, leftsubnet=1.2.3.4/31/tcp/22,1:2::/128/tcp/22.

If both leftsubnet= and rightsubnet= are specified, all combinations will be established as a single IPsec tunnel.

When omitted, essentially assumed to be left, signifying that the left end of the connection goes to the left participant only.

Note
Support for specifying multiple selectors and the protocol and port was added in Libreswan version 5.

IKEv1
In IKEv1 only a single selector is allowed and it is limited to specifying a subnet as in: network / netmask

IKEv1
IKEv1 supports two magic shorthands vhost: and vnet:, which can list subnets in the same syntax as virtual-private. The value %priv expands to the networks specified in virtual-private. The value %no means no subnet. A common use for allowing roadwarriors to come in on public IPs or via accepted NATed networks from RFC1918 is to use leftsubnet=vhost:%no,%priv. The vnet: option can be used to allow RFC1918 subnets without hardcoding them. When using vnet the connection will instantiate, allowing for multiple tunnels with different subnets.

Use with leftprotport and leftsubnets
Compatible with libreswan version 4, a simple leftsubnet (specifying a single network prefix and netmask) may be combined with leftprotoport and leftsubnets. Anything more complicated should be converted to use the selector syntax.

leftsubnets

specify multiple private subnets behind the left participant, expressed as networkA/netmaskA,networkB/netmaskB[,...] If both a leftsubnets= and rightsubnets= are defined, all combinations of subnet tunnels will be established as separate IPsec tunnels. You cannot use leftsubnet= and leftsubnets= together. For examples see testing/pluto/multinet-*. Be aware that when using spaces as separator, that the entire option value needs to be in double quotes.

leftvti

the address/mask to configure on the VTI interface when vti-interface is set. It takes the form of network/netmask Currently, IPv4 and IPv6 addresses with cidr netmask are supported. This option is often used in combination with routed based VPNs.

leftaddresspool

address pool from where the IKEv1 ModeCFG or IKEv2 server can assign IP addresses to clients. When configured as a server, using leftxauthserver=yes this option specifies the address pool from which IP addresses are taken to assign the clients. The syntax of the address pool specifies a range (not a CIDR) for IPv4 and CIDR for IPv6, in the following syntax: rightaddresspool=192.168.1.100-192.168.1.200 or rightaddresspool=2001:db8:0:3:1::/97 Generally, the rightaddresspool= option will be accompanied by rightxauthclient=yes, leftxauthserver=yes and leftsubnet=0.0.0.0/0 option.

When leftaddresspool= is specified, the connection may not specify either leftsubnet= or leftsubnets=. Address pools are fully allocated when the connection is loaded, so the ranges should be sane. For example, specifying a range rightaddresspool=10.0.0.0-11.0.0.0 will lead to massive memory allocation. Address pools specifying the exact same range are shared between different connections. Different addresspools should not be defined to partially overlap.

leftprotoport

allowed protocols and ports over connection, also called Port Selectors. The argument is in the form protocol, which can be a number or a name that will be looked up in /etc/protocols, such as leftprotoport=icmp, or in the form of protocol/port, such as tcp/smtp. Ports can be defined as a number (eg. 25) or as a name (eg smtp) which will be looked up in /etc/services. A special keyword %any can be used to allow all ports of a certain protocol. The most common use of this option is for L2TP connections to only allow l2tp packets (UDP port 1701), eg: leftprotoport=17/1701.

To filter on specific icmp type and code, use the higher 8 bits for type and the lower 8 bits for port. For example, to allow icmp echo packets (type 8, code 0) the 'port' would be 0x0800, or 2048 in decimal, so you configure leftprotoport=icmp/2048. Similarly, to allow ipv6-icmp Neighbour Discovery which has type 136 (0x88) and code 0(0x00) this becomes 0x8800 or in decimal 34816 resulting in leftprotoport=ipv6-icmp/34816.

Some clients, notably older Windows XP and some Mac OSX clients, use a random high port as source port. In those cases rightprotoport=17/%any can be used to allow all UDP traffic on the connection. Note that this option is part of the proposal, so it cannot be arbitrarily left out if one end does not care about the traffic selection over this connection - both peers have to agree. The Port Selectors show up in the output of ipsec status eg:"l2tp": 193.110.157.131[@aivd.libreswan.org]:7/1701...%any:17/1701 This option only filters outbound traffic. Inbound traffic selection must still be based on firewall rules activated by an updown script. The variables $PLUTO_MY_PROTOCOL, $PLUTO_PEER_PROTOCOL, $PLUTO_MY_PORT, and $PLUTO_PEER_PORT are available for use in updown scripts. Older workarounds for bugs involved a setting of 17/0 to denote any single UDP port (not UDP port 0). Some clients, most notably OSX, uses a random high port, instead of port 1701 for L2TP.

use with leftsubnet
With IKEv2, the leftsubnet specification can include the protocol and port. Combining that syntax with protoport is not supported.

leftnexthop

next-hop gateway IP address for the left participant's connection to the public network; defaults to %direct (meaning right). If the value is to be overridden by the left=%defaultroute method (see above), an explicit value must not be given. If that method is not being used, but leftnexthop is %defaultroute, the next-hop gateway address of the default-route interface will be used. The magic value %direct signifies a value to be filled in (by automatic keying) with the peer's address. Relevant only locally, other end need not agree on it.

leftsourceip

the IP address for this host to use when transmitting a packet to the other side of this link. Relevant only locally, the other end need not agree. This option is used to make the gateway itself use its internal IP, which is part of the leftsubnet, to communicate to the rightsubnet or right. Otherwise, it will use its nearest IP address, which is its public IP address. This option is mostly used when defining subnet-subnet connections, so that the gateways can talk to each other and the subnet at the other end, without the need to build additional host-subnet, subnet-host and host-host tunnels. Both IPv4 and IPv6 addresses are supported.

leftupdown

The script to run to adjust routing and/or firewalling when the status of the connection changes (default ipsec _updown). May include positional parameters separated by white space (although this requires enclosing the whole string in quotes); including shell metacharacters is unwise. An example to enable routing when using the XFRM stack, one can use:

updown="ipsec _updown --route yes"

To disable calling an updown script, set it to the empty string, eg leftupdown="" or leftupdown="%disabled".

Connections with type= set to passthrough, reject or drop never run updown.

See libreswan(7) for details.

Relevant only locally, other end need not agree on it.

leftcat

Whether to perform Client Address Translation ("CAT") when using Opportunistic IPsec behind NAT. Accepted values are no (the default) and yes. This option should only be enabled on the special Opportunistic IPsec connections, usually called "private" and "private-or-clear". When set, this option causes the given addresspool IP from the remote peer to be NATed with nftables or iptables. It will also install an additional IPsec SA policy to cover the pre-NAT IP. See the Opportunistic IPsec information on the libreswan website for more information and examples.

leftfirewall

This option is obsolete and should not used anymore.

If one or both security gateways are doing forwarding firewalling (possibly including masquerading), and this is specified using the firewall parameters, tunnels established with IPsec are exempted from it so that packets can flow unchanged through the tunnels. (This means that all subnets connected in this manner must have distinct, non-overlapping subnet address blocks.) This is done by the default updown script (see ipsec_pluto(8)).

The implementation of this makes certain assumptions about firewall setup, and the availability of the Linux Advanced Routing tools. In situations calling for more control, it may be preferable for the user to supply his own updown script, which makes the appropriate adjustments for his system.

The following parameters are relevant to automatic keying via IKE. Unless otherwise noted, for a connection to work, in general it is necessary for the two ends to agree exactly on the values of these parameters.

auto=

what operation, if any, should be done automatically at IPsec startup; currently-accepted values are:

auto=add

equivalent to:

ipsec add connection      

auto=ondemand

equivalent to:

ipsec add connection      

ipsec route connection      

(auto=route is an alias)

auto=up, auto=start

equivalent to:

ipsec add connection      

ipsec up connection      

auto=ignore

signifying no automatic operation when Libreswan is starting (also the default)

auto=keep

signifying an add plus an attempt to keep the connection up once the remote peer brought it up

Relevant only locally, other end need not agree on it (but in general, for an intended-to-be-permanent connection, both ends should use auto=up to ensure that any reboot causes immediate renegotiation).

See the config setup discussion below.

authby=

How the two security gateways should authenticate each other. A comma separated list of one or more of the following:

ecdsa, ecdsa-sha2

ECDSA digital signature based authentication with any of the hash algorithms SHA2_256, SHA2_385, or SHA2_512

ecdsa-sha2_256, ecdsa-sha2_384, ecdsa-sha2_512

ECDSA digital signature based authentication with the specified SHA2_256, SHA2_385, or SHA2_512 hash algorithm

rsa-sha1

RSA-PKCSv1.5 digital signature based authentication with SHA1

rsa-sha2

RSASSA-PSS digital signature based authentication with any of the hash algorithms SHA2_256, SHA2_385, or SHA2_512

rsa-sha2_256, rsa-sha2_384, rsa-sha2_512

RSASSA-PSS digital signature based authentication with the specified SHA2_256, SHA2_385, or SHA2_512 hash algorithm

secret

shared secrets (PSK) authentication; cannot be combined with other authentication methods

never

negotiation is never to be attempted or accepted; cannot be combined with other authentication methods

null

null-authentication; cannot be combined with other authentication methods

For IKEv1, SHA2 based signatures are not defined and ECDSA is not implemented, so the default authby value is rsa-sha1. Using authby=rsasig results in only rsa-sha1 as well.

For IKEv2, rsasig,ecdsa which allows ECDSA with SHA-2 and RSA with SHA2 or SHA1. Using authby=rsasig means using rsa-sha2_512, rsa-sha2_384, rsa-sha2_256 and rsa-sha1, where rsa-sha1 will used only if RFC 7427 is not supported by the peer.

Note that authby cannot be used for asymmetric authentication. Instead, IKEv2 must be enabled and the options leftauth and rightauth used.

As per RFC 8221, authby=rsa-sha1 is only supported in the old style, meaning RSA-PKCSv1.5. The SHA2 variants are only supported for the new style of RFC 7427, so authby=rsa-sha2 will use the new style. The default authby= will remove rsa-sha1 in the near future. It is strongly recommended that if certificates are used, the certificates and the authby= signature methods used are the same, as it increases interoperability and keeps the authentication of everything within one digital signature system.

Digital signatures are superior in every way to shared secrets. Especially IKEv1 in Aggressive Mode is vulnerable to offline dictionary attacks and is performed routinely by at least the NSA on monitored internet traffic globally. The never option is only used for connections that do not actually start an IKE negotiation, such as type=passthrough connections. The auth method null is used for "anonymous opportunistic IPsec" and should not be used for regular pre-configured IPsec VPNs.

ike

IKE encryption/authentication algorithm to be used for the connection (phase 1 aka ISAKMP SA or IKE SA). If this option is not set, the builtin defaults will be used. This is the preferred method, and allows for gradual automatic updates using the same configuration. Some distributions, such as Fedora and RHEL/CentOS, use a System Wide Crypto Policy that sets the default ike= (and esp=) lines. Specifying your own ike= line means overriding all these system or software recommended defaults, but can be necessary at times. Note that libreswan does not support using a PRF algorithm that is different from the INTEGRITY (hash) algorithm by design.

The format is "cipher-hash-modpgroup, cipher-hash-modpgroup, ..." Any omitited option will be filled in with all allowed default values. Multiple proposals are separated by a comma. If an ike= line is specified, no other received proposals will be accepted than those specified on the IKE line. Some examples are ike=3des-sha1,aes-sha1, ike=aes, ike=aes_ctr, ike=aes_gcm256;sha2_256, ike=aes128-sha1;modp2048, ike=aes256-sha2;dh19, ike=aes128-sha1;dh22, ike=3des-md5;modp1024,aes-sha1;modp1536.

IKEv2 allows combining elements into a single proposal. These can be specified by using the + symbol. An example is: ike=aes_gcm+chacha20_poly1305;dh14+dh19,aes+3des-sha2+sha1;dh14. Note that AEAD algorithms (aes_gcm, aes_ccm, chacha20_poly1305) and non-AEAD algorithms (aes, 3des) cannot be combined into a single proposal. To support aes and aes_gcm, two proposals separated by a comma must be used.

The default IKE proposal depends on the version of libreswan used. It follow the recommendations of RFC4306, RFC7321 and as of this writing their successor draft documents RFC4306bis and RFC7321bis. As for libreswan 3.32, SHA1 and MODP1536(dh5) are still allowed per default for backwards compatibility, but 3DES and MODP1024(dh2) are not allowed per default. As of libreswan 4.x, modp1024(dh2) support is no longer compiled in at all. For IKEv2, the defaults include AES, AES-GCM, DH14 and stronger, and SHA2. The default key size is 256 bits. The default AES_GCM ICV is 16 bytes.

Note that AES-GCM is an AEAD algorithm, meaning that it performs encryption+authentication in one step. This means that AES-GCM must not specify an authentication algorithm. However, for IKE it does require a PRF function, so the second argument to an AEAD algorithm denotes the PRF. So ike=aes_gcm-sha2_256 means propose AES_GCM with SHA2_256 as the prf. Note that for esp, there is no prf, so AES-GCM is specified for ESP as esp=aes_gcm. The AES-GCM and AES-CCM algorithms support 8,12 and 16 byte ICV's. These can be specified using a postfix, for example aes_gcm_a (for 8), aes_gcm_b (for 12) and aes_gcm_c (for 16). The default (aes_gcm without postfix) refers to the 16 byte ICV version. RFC 8247 only mandates the 16 byte ICV version is implemented, so it is recommended to NOT use the 8 or 12 byte versions of GCM or CCM. These versions are NOT included in the default proposal list and will be removed in a future version.

Weak algorithms are regularly removed from libreswan. Currently, 1DES and modp768(DH1) have been removed and modp1024(DH2) has been disabled at compile time. Additionally, MD5 and SHA1 will be removed within the next few years. Null encryption is available, and should only be used for testing or benchmarking purposes. Please do not request for insecure algorithms to be re-added to libreswan. IKEv1 has been disabled per default, and will soon no longer be compiled in by default.

For all Diffie-Hellman groups, the "dh" keyword can be used instead of the "modp" keyword. For example ike=3des-sha1-dh19. Diffie-Hellman groups 19,20 and 21 from RFC-5903 are supported. Curve25519 from RFC-8031 is supported as "dh31". Curve448 and GOST DH groups are not yet supported in libreswan because these are not supported yet in the NSS crypto library.

Diffie-Hellman groups 22, 23 and 24 from RFC-5114 are implemented but not compiled in by default. These DH groups are extremely controversial and MUST NOT be used unless forced (administratively) by the other party. This is further documented in RFC 8247, but the summary is that it cannot be proven that these DH groups do not contain a cryptographic trapdoor (a backdoor by the USG who provided these primes without revealing the seeds and generation process used).

The modp syntax will be removed in favour of the dh syntax in the future

phase2

Sets the type of SA that will be produced. Valid options are: esp for encryption (the default), ah for authentication only.

The very first IPsec designs called for use of AH plus ESP to offer authentication, integrity and confidentiality. That dual protocol use was a significant burden, so ESP was extended to offer all three services, and AH remained as an auth/integ. The old mode of ah+esp is no longer supported in compliance with RFC 8221 Section 4. Additionally, AH does not play well with NATs, so it is strongly recommended to use ESP with the null cipher if you require unencrypted authenticated transport.

sha2-truncbug

The default ESP hash truncation for sha2_256 is 128 bits. Some IPsec implementations (Linux before 2.6.33, some Cisco (2811?) routers) implement the draft version which stated 96 bits. If a draft implementation communicates with an RFC implementation, both ends will reject encrypted packets from each other.

This option enables using the draft 96 bits version to interop with those implementations. Currently the accepted values are no, (the default) signifying default RFC truncation of 128 bits, or yes, signifying the draft 96 bits truncation.

Another workaround is to switch from sha2_256 to sha2_128 or sha2_512.

ms-dh-downgrade

Whether to allow a downgrade of Diffie-Hellman group during rekey (using CREATE_CHILD_SA).

Microsoft Windows (at the time of writing, Feb 2018) defaults to using the very weak modp1024 (DH2). This can be changed using a Windows registry setting to use modp2048 (DH14). However, at rekey times, it will shamelessly use modp1024 again and the connection might fail. Setting ms-dh-downgrade=yes (and adding modp1024 proposals to the ike line) will allow this downgrade attack to happen. This should only be used to support Windows that feature this bug.

The accepted values are no, (the default) or yes.

dns-match-id

Whether to perform an additional DNS lookup and confirm the remote ID payload with the DNS name in the reverse DNS PTR record. Accepted values are no (the default) or yes. This check should be enabled when Opportunistic IPsec is enabled in a mode that is based on packet triggers (on-demand) using IPSECKEY records in DNS. Since in that case the IKE daemon pluto does not know the remote ID, it only knows the remote IP address, this option forces it to confirm the peer's proposed ID (and thus its public/private key) with its actual IP address as listed in the DNS. This prevents attacks where mail.example.com's IP address is taken over by a neighbour machine with a valid web.example.com setup. This check is not needed for certificate based Opportunistic IPsec, as "mail.example.com"s certificate does not have an entry for "web.example.com". It is also not needed for DNS server triggered Opportunistic IPsec, as in that case the IKE daemon pluto is informed of both the IP address, and the hostname/public key.

require-id-on-certificate

When using certificates, check whether the IKE peer ID is present as a subjectAltName (SAN) on the peer certificate. Accepted values are yes (the default) or no. This check should only be disabled when intentionally using certificates that do not have their peer ID specified as a SAN on the certificate. These certificates violate RFC 4945 Section 3.1 and are normally rejected to prevent a compromised host from assuming the IKE identity of another host. The SAN limits the IDs that the peer is able to assume.

ppk

EXPERIMENTAL: Post-quantum preshared keys (PPKs) to be used. Currently the accepted values are propose or yes (the default), signifying we propose to use PPK for this connection; insist, signifying we allow communication only if PPK is used for key derivation; never or no, signifying that PPK should not be used for key derivation. PPKs can be used in connections that allow only IKEv2. In libreswan that would mean that ikev2 option must have value insist. This feature is based on RFC 8784.

ppk-ids

EXPERIMENTAL: Specify which PPK_ID's to look for in .secrets file. The value must be a string (enclosed in quotes) containing a comma-separated list with PPK_ID's. Whitespaces cannot be a part of PPK_ID, i.e., they will be seen as a delimiter. If some PPK_ID's are specified and PPK feature is enabled (ppk=), one of the PPK secrets must have an ID matching one of the specified PPK_ID's. If this option is not set and the PPK feature is enabled, any matching PPK secret will be used.

nat-ikev1-method

NAT Traversal in IKEv1 is negotiated via Vendor ID options as specified in RFC 3947. However, many implementations only support the draft version of the RFC. Libreswan sends both the RFC and the most common draft versions (02, 02_n and 03) to maximize interoperability. Unfortunately, there are known broken implementations of RFC 3947, notably Cisco routers that have not been updated to the latest firmware. As the NAT-T payload is sent in the very first packet of the initiator, there is no method to auto-detect this problem and initiate a workaround.

This option allows fine tuning which of the NAT-T payloads to consider for sending and processing. Currently the accepted values are drafts, rfc, both (the default) and none. To interoperate with known broken devices, use nat-ikev1-method=drafts. To prevent the other end from triggering IKEv1 NAT-T encapsulation, set this to none. This will omit the NAT-T payloads used to determine NAT, forcing the other end not to use encapsulation.

esp

Specifies the algorithms that will be offered/accepted when negotiating a a Child SA. The general syntax is:

ESP = PROPOSAL[,PROPOSAL...]
PROPOSAL = ENCRYPT_ALGS[-INTEG_ALGS[-DH_ALGS]]
ENCRYPT_ALGS = ENCRYPT_ALG[+ENCRYPT_ALG...]
INTEG_ALGS = INTEG_ALG[+INTEG_ALG...]
DH_ALGS = DH_ALG[+DH_ALG...]

During startup, ipsec_pluto(8) will log all supported ESP algorithms.

Specifying the DH algorithms explicitly is not recommended. When PFS is enabled, and the DH algorithms are omitted, each PROPOSAL will automatically include the DH algorithm negotiated during the IKE exchange.

AEAD algorithms such as AES_GCM and AES_CCM do not not require a separate integrity algorithm. For example esp=aes_gcm256 or esp=aes_ccm.

Note that AES_GCM and AES_CCM for ESP come in 8, 12 and 16 byte ICV versions. RFC 8221 only requires AES_GCM with 16 byte ICV and AES_CCM with 8 byte ICV to be implemented, and "aes_gcm" and "aes_ccm" refer to these variants. The other variants can be specified using an _a (8), _b(12) or _c(16) postfix, eg esp=aes_gcm_a for the 8 byte ICV and esp=aes_gcm_b for the 12 byte ICV.

For instance:

esp=aes_gcm;aes128+aes256-sha2_512+sha2_256
esp=aes128-sha2_512;dh19

If not specified, a secure set of defaults will be used. The program ipsec algparse can be used to query these defaults for instance: ipsec algparse esp= (see ipsec_algparse(8)).

This option replaces phase2alg= option for esp.

ah

A comma separated list of AH algorithms that will be offered/accepted when negotiating the Child SA. The general syntax is:

AH = PROPOSAL[,PROPOSAL...]
PROPOSAL = INTEG_ALGS[-DH_ALGS]
INTEG_ALGS = INTEG_ALG[+INTEG_ALG...]
DH_ALGS = DH_ALG[+DH_ALG...]

During startup, ipsec_pluto(8) will log all supported AH algorithms.

Specifying the DH algorithms explicitly is not recommended. When PFS is enabled, and the DH algorithms are omitted, each PROPOSAL will automatically include the DH algorithm negotiated during the IKE exchange.

The default is not to use AH. If for some (invalid) reason you still think you need AH, please use esp with the null encryption cipher instead.

For instance:

ah=sha2_256+sha2_512
ah=sha2_256+sha2_512-dh14+dh19

If not specified, a secure set of defaults will be used. The command ipsec algparse ah=... can be used to query these defaults.

This option replaces phase2alg= option for ah.

fragmentation

Whether or not to allow IKE fragmentation. Valid values are yes (the default) and no. In addition IKEv1 allows force.

IKEv2 fragmentation support is implemented using RFC 7383.

IKEv1 fragmentation capabilities are negotiated via a well-known private vendor id. If pluto does not receive the fragmentation payload, no IKE fragments will be sent, regardless of the fragmentation= setting. When set to yes, IKE fragmentation will be attempted on the first re-transmit of an IKE packet of a size larger then 576 bytes for IPv4 and 1280 bytes for IPv6. If fragmentation is set to force, IKE fragmentation is used on initial transmits of such sized packets as well. When we have received IKE fragments for a connection, pluto behaves as if in force mode.

ikepad=

Whether or not to pad IKEv1 messages to a multiple of 4 bytes. Valid values are ikepad=yes, (the default) and ikepad=no.

IKE padding is allowed in IKEv1 but has been known to cause interoperability issues. The ikepad= option can be used to disable IKEv1 padding. This used to be required for some devices (such as Checkpoint in Aggressive Mode) that reject padded IKEv1 packets. A bug was fixed in libreswan 3.25 that applied wrong IKE padding in XAUTH, so it is suspected that Checkpoint padding issue bas been resolved. And this option should not be needed by anyone. In IKEv2, no padding is allowed, and this option has no effect. If you find a device that seems to require IKE padding, please contact the libreswan developers. This option should almost never be enabled and might be removed in a future version.

mobike

Whether to allow MOBIKE (RFC 4555) to enable a connection to migrate its endpoint without needing to restart the connection from scratch. This is used on mobile devices that switch between wired, wireless or mobile data connections. Current values are no (the default) or yes, Only connection acting as modecfgclient will allow the initiator to migrate using mobike. Only connections acting as modecfgserver will allow clients to migrate.

VTI and MOBIKE might not work well when used together.

intermediate

EXPERIMENTAL: Whether to enable the INTERMEDIATE exchange in IKEv2 (RFC 9242). Currently the accepted values are yes, signifying we propose to use the INTERMEDIATE exchange for this connection; no (the default), signifying we will not use the INTERMEDIATE exchange for this connection.

esn

Whether or not to enable Extended Sequence Number (ESN) for the IPsec SA. This option is only implemented for IKEv2. ESN is typically used for very high-speed links (10Gbps or faster) where the standard 32 bit sequence number is exhausted too quickly, causing IPsec SA's rekeys to happen too often. Accepted values are either (the default), yes and no. If either is specified as an initiator, the responder will make the choice. As a responder, if either is received, yes is picked.

If replay-window is set to 0, ESN is disabled as some (most?) IPsec stacks won't support ESN in such a configuration.

decap-dscp

Enable decapsulating the Differentiated Services Code Point (DSCP, formerly known as Terms Of Service (TOS)) bits. If these bits are set on the inner (encrypted) IP packets, these bits are set on the decrypted IP packets. Acceptable values are no (the default) or yes. Currently this feature is only implemented for the Linux XFRM stack.

encap-dscp

Enable encapsulating the Differentiated Services Code Point (DSCP, formerly known as Terms Of Service (TOS)) bits. The extra xfrm flag "dont-encap-dscp" prevents these bits from being copied from the unencrypted IP packets to the encrypted IP packets. Acceptable values are yes (the default) or no. Currently this feature is only implemented for the Linux XFRM stack.

nopmtudisc

Disable Path MTU discovery for the IPsec SA. Acceptable values are no (the default) or yes. Currently this feature is only implemented for the Linux XFRM stack.

narrowing

IKEv2 (RFC5996) Section 2.9 Traffic Selector narrowing options. Currently the accepted values are no, (the default) signifying no narrowing will be proposed or accepted, or yes, signifying IKEv2 negotiation may allow establishing an IPsec connection with narrowed down traffic selectors. This option is ignored for IKEv1.

There are security implications in allowing narrowing down the proposal. For one, what should be done with packets that we hoped to tunnel, but cannot. Should these be dropped or send in the clear? Second, this could cause thousands of narrowed down Child SAs to be created if the conn has a broad policy (eg 0/0 to 0/0). One possible good use case scenario is that a remote end (that you fully trust) allows you to define a 0/0 to them, while adjusting what traffic you route via them, and what traffic remains outside the tunnel. However, it is always preferred to setup the exact tunnel policy you want, as this will be much clearer to the user.

sareftrack

Set the method of tracking reply packets with SArefs when using an SAref compatible stack. Currently only the mast stack supports this. Acceptable values are yes (the default), no or conntrack. This option is ignored when SArefs are not supported. This option is passed as PLUTO_SAREF_TRACKING to the updown script which makes the actual decisions whether to perform any iptables/ip_conntrack manipulation. A value of yes means that an IPSEC mangle table will be created. This table will be used to match reply packets. A value of conntrack means that additionally, subsequent packets using this connection will be marked as well, reducing the lookups needed to find the proper SAref by using the ip_conntrack state. A value of no means no IPSEC mangle table is created, and SAref tracking is left to a third-party (kernel) module. In case of a third party module, the SArefs can be relayed using the statsbin= notification helper.

nic-offload

Set the method of Network Interface Controller (NIC) hardware offload for ESP/AH packet processing. Acceptable values are no (the default), crypto or packet. The value yes is a backwards compatible value for crypto. The nic-offload option is separate from any CPU hardware offload available. When set to crypto, only cryptographic operations are offloaded to the NIC card. When set to packet, the entire packet processing including the encryption/decryption is offloaded to the NIC card.

Crypto nic-offload is available starting Linux 4.13 using the XFRM IPsec stack. Packet nic-offload is available starting Linux 6.3. Both require that the Linux kernel is compiled with the options CONFIG_XFRM_OFFLOAD, CONFIG_INET_ESP_OFFLOAD and CONFIG_INET6_ESP_OFFLOAD. Network card support can be seen by the presence of the esp-hw-offload capability using the ethtool -S command. The Linux kernel attempts to fall back from crypto hardware offload to software, but only for some algorithms (AEADs only?). There is no fallback from packet offload to crypto offload. At the time of libreswan 5.0, we are only aware of the Nvidia/Mellanox ConnectX-7 (and to some extend ConnectX-6) cards supporting packet offload.

In general, it makes no sense to try to offload older (non-AEAD) cryptographic algorithms such as AES-CBC or 3DES, as these algorithms are so much slower than AEAD algorithms (such as AES-GCM) that one would gain more performance by switching the algorithm to AEAD than by offloading. As such, AES-CBC tends to not be implemented in offload hardware. This option has also no effect on IKE packets, which are never offloaded, although IKE encryption does use supported CPU hardware instructions, such as AESNI.

leftid

how the left participant should be identified for authentication; defaults to left. Can be an IP address or a fully-qualified domain name which will be resolved. If preceded by @, the value is used as a literal string and will not be resolved. To support opaque identifiers (usually of type ID_KEY_ID, such as used by Cisco to specify Group Name, use square brackets, eg rightid=@[GroupName]. The magic value %fromcert causes the ID to be set to a DN taken from a certificate that is loaded. Prior to 2.5.16, this was the default if a certificate was specified. The magic value %none sets the ID to no ID. This is included for completeness, as the ID may have been set in the default conn, and one wishes for it to default instead of being explicitly set.

When using certificate based ID's, one need to specify the full RDN, optionally using wildcard matching (eg CN='*'). If the RDN contains a comma, this can be masked using a backslash (eg OU='Foo\, Bar and associates')

leftrsasigkey

the left participant's public key for RSA signature authentication, in base-64 encoded RFC 2537 format (with 0s prepended), or one of the following:

%none

the same as not specifying a value (useful to override a default)

%dnsondemand

(the default) the key is to be fetched from DNS at the time it is needed.

%dnsonload

the key is to be fetched from DNS at the time the connection description is read from ipsec.conf; currently this will be treated as %none if right=%any or right=%opportunistic.

%dns

currently treated as %dnsonload but will change to %dnsondemand in the future. The identity used for the left participant must be a specific host, not %any or another magic value.

%cert

the information required from a certificate defined in %leftcert and automatically define leftid for you
Caution
If two connection descriptions specify different public keys for the same leftid, confusion and madness will ensue.

leftcert

If you are using leftrsasigkey=%cert this defines the certificate nickname of your certificate in the NSS database. This can be on software or hardware security device.

leftckaid

The CKAID of the X.509 certificate or host key.

For X.509 certificates, the CKAID is either the certificate's SubjectKeyIdentifier or the public key's SHA1 fingerprint (when the SubjectKeyIdentifier isn't specified). For host keys the CKAID is the SHA1 fingerprint of the public key.

leftauth

How the security gateways will authenticate to the other side in the case of asymmetric authentication; acceptable values are rsasig or rsa for RSA Authentication with SHA-1, rsa-sha2 for RSA-PSS digital signatures based authentication with SHA2-256, rsa-sha2_384 for RSA-PSS digital signatures based authentication with SHA2-384, rsa-sha2_512 for RSA-PSS digital signatures based authentication with SHA2-512, ecdsa for ECDSA digital signatures based authentication, secret for shared secrets (PSK) authentication and null for null-authentication. There is no default value - if unset, the symmetrical authby= keyword is used to determine the authentication policy of the connection.

Asymmetric authentication is only supported with IKEv2. If symmetric authentication is required, use authby= instead of leftauth and rightauth. If leftauth is set, rightauth must also be set and authby= must not be set. Asymmetric authentication cannot use secret (psk) on one side and null on the other side - use psk on both ends instead.

When using EAPONLY authentication, which omits the regular IKEv2 AUTH payload, leftauth= (or rightauth=) should be set to eaponly.

Be aware that the symmetric keyword is authby= but the asymmetric keyword is leftauth and rightauth (without the "by").

leftautheap

Whether the security gateways will authenticate uing an EAP method. Acceptable values are none (the default) and tls for EAPTLS. If EAP is the only authentication method, set leftauth=none in addition to leftautheap=tls=.

The EAP authentication mechanisms are only available for IKEv2 based connections.

leftca

specifies the authorized Certificate Authority (CA) that signed the certificate of the peer. If undefined, it defaults to the CA that signed the certificate specified in leftcert. The special rightca=%same is implied when not specifying a rightca and means that only peers with certificates signed by the same CA as the leftca will be allowed. This option is only useful in complex multi CA certificate situations. When using a single CA, it can be safely omitted for both left and right.

leftikeport

The UDP IKE port to listen on or send data to. This port cannot be 0 or 500. For TCP, see tcp-remoteport=

leftsendcert

This option configures when Libreswan will send X.509 certificates to the remote host. Acceptable values are yes|always (signifying that we should always send a certificate), sendifasked (signifying that we should send a certificate if the remote end asks for it), and no|never (signifying that we will never send a X.509 certificate). The default for this option is sendifasked which may break compatibility with other vendor's IPsec implementations, such as Cisco and SafeNet. If you find that you are getting errors about no ID/Key found, you likely need to set this to always. This per-conn option replaces the obsolete global nocrsend option.

leftxauthserver

Left is an XAUTH server. This can use PAM for authentication or md5 passwords in /etc/ipsec.d/passwd. These are additional credentials to verify the user identity, and should not be confused with the XAUTH group secret, which is just a regular PSK defined in ipsec.secrets. The other side of the connection should be configured as rightxauthclient. XAUTH connections cannot rekey, so rekey=no should be specified in this conn. For further details on how to compile and use XAUTH, see README.XAUTH. Acceptable values are yes or no (the default).

leftxauthclient

Left is an XAUTH client. The xauth connection will have to be started interactively and cannot be configured using auto=start. Instead, it has to be started from the commandline using ipsec up connection. You will then be prompted for the username and password. To setup an XAUTH connection non-interactively, which defeats the whole purpose of XAUTH, but is regularly requested by users, it is possible to use a whack command - ipsec whack --name baduser --ipsecgroup-xauth --xauthname badusername --xauthpass password --initiate The other side of the connection should be configured as rightxauthserver. Acceptable values are yes or no (the default).

leftusername

The username associated with this connection. The username can be the IKEv2 XAUTH username, a GSSAPI username or IKEv2 CP username. For the XAUTH username, the XAUTH password can be configured in the ipsec.secrets file. This option was previously called leftxauthusername.

leftmodecfgserver

Left is a Mode Config server. It can push network configuration to the client. Acceptable values are yes or no (the default).

leftmodecfgclient

Left is a Mode Config client. It can receive network configuration from the server. Acceptable values are yes or no (the default).

xauthby

When IKEv1 XAUTH support is available, set the method used by XAUTH to authenticate the user with IKEv1. The currently supported values are file (the default), pam or alwaysok. The password file is located at /etc/ipsec.d/passwd, and follows a syntax similar to the Apache htpasswd file, except an additional connection name argument (and optional static IP address) are also present:

username:password:conname:ipaddress

For supported password hashing methods, see crypt(3). If pluto is running in FIPS mode, some hash methods, such as MD5, might not be available. Threads are used to launch an xauth authentication helper for file as well as PAM methods.

The alwaysok should only be used if the XAUTH user authentication is not really used, but is required for interoperability, as it defeats the whole point of XAUTH which is to rely on a secret only known by a human. See also pam-authorize=yes

xauthfail

When XAUTH support is available, set the failure method desired when authentication has failed. The currently supported values are hard (the default) and soft. A soft failure means the IPsec SA is allowed to be established, as if authentication had passed successfully, but the XAUTH_FAILED environment variable will be set to 1 for the updown script, which can then be used to redirect the user into a walled garden, for example a payment portal.

pam-authorize

IKEv2 does not support receiving a plaintext username and password. Libreswan does not yet support EAP authentication methods for IKE. The pam-authorize=yes option performs an authorization call via PAM, but only includes the remote ID (not username or password). This allows for backends to disallow an ID based on non-password situations, such as "user disabled" or "user over quota".

See also the IKEv1 option xauthby=pam

modecfgpull

Pull the Mode Config network information from the server. Acceptable values are yes or no (the default).

modecfgdns, modecfgdomains, modecfgbanner

When configured as IKEv1 ModeCFG or IKEv2 server, specifying any of these options will cause those options and values to be sent to the connecting client. Libreswan as a client will use these received options to either update /etc/resolv.conf or the running unbound DNS server. When the connection is brought down, the previous DNS resolving state is restored.

The modecfgdns option takes a comma or space separated list of IP addresses that can be used for DNS resolution. The modecfgdomains option takes a comma or space separated list of internal domain names that are reachable via the supplied modecfgdns DNS servers.

The IKEv1 split tunnel directive will be sent automatically if the xauth server side has configured a network other than 0.0.0.0/0. For IKEv2, this is automated via narrowing.

remote-peer-type

Set the remote peer type. This can enable additional processing during the IKEv1 negotiation. Acceptable values are cisco or ietf (the default). When set to cisco, support for Cisco IPsec gateway redirection and Cisco obtained DNS and domainname are enabled. This includes automatically updating (and restoring) /etc/resolv.conf. These options require that XAUTH is also enabled on this connection.

nm-configured

Mark this connection as controlled by Network Manager. Acceptable values are yes or no (the default). Currently, setting this to yes will cause libreswan to skip reconfiguring resolv.conf when used with XAUTH and ModeConfig.

encapsulation

In some cases, for example when ESP packets are filtered or when a broken IPsec peer does not properly recognise NAT, it can be useful to force RFC-3948 encapsulation. In other cases, where IKE is NAT'ed but ESP packets can or should flow without encapsulation, it can be useful to ignore the NAT-Traversal auto-detection. encapsulation=yes forces the NAT detection code to lie and tell the remote peer that RFC-3948 encapsulation (ESP in port 4500 packets) is required. encapsulation=no ignores the NAT detection causing ESP packets to send send without encapsulation. The default value of encapsulation=auto follows the regular outcome of the NAT auto-detection code performed in IKE. This option replaced the obsoleted forceencaps option.

enable-tcp

Normally, IKE negotiation and ESP encapsulation happens over UDP. This option enables support for IKE and ESP over TCP as per RFC 8229. Acceptable values are no(the default), yes meaning only TCP will be used, or fallback meaning that TCP will be attempted only after negotiation over UDP failed. Since performance over TCP is much less, and TCP sessions are vulnerable to simply RST resets and MITM attacks causing the TCP connection to close, this option should really only be used in fallback mode. If used in fallback mode, it is recommend to reduce the retransmit-timeout from the default 60s to a much shorter value such as 10s, so that one does not have to wait a minute for the TCP fallback to be attempted. To receive incoming TCP connections, the config setup option listen-tcp needs to be enabled.

tcp-remoteport

Which remote TCP port to use when IKE over TCP is attempted. The default value is to use the NAT-T IKE port (4500). This value is not negotiated and should be configured properly on all endpoints. When opening a TCP socket to the remote host in this port, a regular ephemeral source port is obtained from the OS. For changing the UDP ports, see leftikeport=

nat-keepalive

whether to send any NAT-T keep-alives. These one byte packets prevent the NAT router from closing its port when there is not enough traffic on the IPsec connection. Acceptable values are: yes (the default) and no.

initial-contact

whether to send an INITIAL_CONTACT payload to the peer we are initiating to, if we currently have no IPsec SAs up with that peer. Acceptable values are: yes (the default) and no. It is recommended to leave this option set, unless multiple clients with the same identity are expected to connect using the same subnets and should operate at the same time. Or if a reconnecting client should not delete its old instance (eg perhaps it is still running). This is unlikely to be true.

cisco-unity

whether to send a CISCO_UNITY payload to the peer. Acceptable values are: no (the default) and yes.

It is recommended to leave this option unset, unless the remote peer (Cisco client or server) requires it. This option does not modify local behaviour. It can be needed to connect as a client to a Cisco server. It can also be needed to act as a server for a Cisco client, which otherwise might send back an error DEL_REASON_NON_UNITY_PEER.

ignore-peer-dns

whether to ignore received DNS configuration. Acceptable values are: no (the default) and yes. Normally, when a roadwarrior connects to a remote VPN, the remote VPN server sends a list of DNS domains and DNS nameserver IP addresses that the roadwarrior can use to reach internal only resources through the VPN. This option allows the roadwarrior to ignore the server's suggestion. The roadwarrior will normally use this information to update the DNS resolving process. What is changed depends on the detected DNS configuration. It can modify /etc/resolv.conf directly, or reconfigure a locally running DNS server (unbound, knot, stubby or systemd-resolved) or inform NetworkManager.

accept-redirect-to

Specify the comma separated list of addresses we accept being redirected to. Both IPv4 and IPv6 addresses are supported as well the FQDNs. The value %any, as well as not specifying any address, signifes that we will redirect to any address gateway sends us in REDIRECT notify payload.

The value of this option is not considered at all if accept-redirect is set to no.

accept-redirect

Whether requests of the remote peer to redirect IKE/IPsec SA's are accepted. Valid options are no (the default) and yes. See also accept-redirect-to.

redirect-to

Where to send remote peers to via the send-redirect option. This can be an IP address or hostname (FQDN).

send-redirect

Whether to send requests for the remote peer to redirect IKE/IPsec SA's during IKE_AUTH. Valid options are no (the default) and yes. If set, the option redirect-to= must also be set to indicate where to redirect peers to. For redirection during IKE_SA_INIT exchange, see the global-redirect= and global-redirect-to= options. Runtime redirects can be triggered via the ipsec redirect command.

fake-strongswan

whether to send a STRONGSWAN Vendor ID payload to the peer. Acceptable values are: no (the default) and yes. This used to be required because strongswan rejects certain proposals with private use numbers such as esp=twofish or esp=serpent unless it receives a strongswan vendorid by the peer. This option sends such an (unversioned) vendor id. Note that libreswan and strongswan no longer support twofish or serpent, so enabling this option likely will no longer do anything.

send-vendorid

whether to send our Vendor ID during IKE. Acceptable values are: no (the default) and yes. The vendor id sent can be configured using the "config setup" option myvendorid=. It defaults to OE-Libreswan-VERSION.

Vendor ID's can be useful in tracking interoperability problems. However, specific vendor identification and software versions can be useful to an attacker when there are known vulnerabilities to a specific vendor/version.

The prefix OE stands for "Opportunistic Encryption". This prefix was historically used by The FreeS/WAN Project and The Openswan Project (openswan up to version 2.6.38) and in one Xeleranized openswan versions (2.6.39). Further Xeleranized openswan's use the prefix OSW.

overlapip

a boolean (yes/no) that determines, when (left|right)subnet=vhost: is used, if the virtual IP claimed by this states created from this connection can with states created from other connections.

Note that connection instances created by the Opportunistic Encryption or PKIX (x.509) instantiation system are distinct internally. They will inherit this policy bit.

The default is no.

This feature is only available with kernel drivers that support SAs to overlapping conns. At present only the (klips) mast protocol stack supports this feature.

reqid

a unique identifier used to match IPsec SAs using iptables with XFRM. This identifier is normally automatically allocated in groups of 4. It is exported to the _updown script as REQID. On Linux, reqids are supported with IP Connection Tracking and NAT (iptables). Automatically generated values use the range 16380 and higher. Manually specified reqid values therefore must be between 1 and 16379.

Automatically generated reqids use a range of 0-3 (eg 16380-16383 for the first reqid). These are used depending on the exact policy (AH, AH+ESP, IPCOMP, etc).

WARNING: Manually assigned reqids are all identical. Instantiations of connections (those using %any wildcards) will all use the same reqid. If you use manual assigning you should make sure your connections only match single road warrior only or you break multiple road warriors behind same NAT router because this feature requires unique reqids to work.

dpddelay

Set the delay (in time units, defaults to seconds) between Dead Peer Detection (IKEv1 RFC 3706) or IKEv2 Liveness keepalives that are sent for this connection (default 0 seconds). Set to enable checking.

If dpddelay is set, you might need to adjust retransmit-timeout for IKEv2 and you must set dpdtimeout for IKEv1.

dpdtimeout

Set the length of time (in time units, defaults to seconds) that pluto will idle without hearing back from the peer. After this period has elapsed with no response and no traffic, pluto will declare the peer dead, and remove the SA (default 0 meaning disabled). Set value bigger than dpddelay to enable. If dpdtimeout is set, dpddelay also needs to be set.

This option is only valid for IKEv1. For IKEv2, this value is always the same as the retransmit-timeout, as IKEv2 is blocked from sending further IKE messages if an answer is not received.

pfs

whether Perfect Forward Secrecy of keys is desired on the connection's keying channel (with PFS, penetration of the key-exchange protocol does not compromise keys negotiated earlier); Acceptable values are yes (the default) and no.

aggressive

Use IKEv1 Aggressive Mode instead of IKEv1 Main Mode. This option has no effect when IKEv2 is used. Acceptable values are no (the default) or yes. When this option is enabled, IKEv1 Main Mode will no longer be allowed for this connection. The old name of this option was aggrmode.

Aggressive Mode is less secure, and more vulnerable to Denial Of Service attacks. It is also vulnerable to brute force attacks with software such as ikecrack. It should not be used, and it should especially not be used with XAUTH and group secrets (PSK). If the remote system administrator insists on staying irresponsible, enable this option.

Aggressive Mode is further limited to only proposals with one DH group as there is no room to negotiate the DH group. Therefore it is mandatory for Aggressive Mode connections that both ike= and esp= options are specified with only one fully specified proposal using one DH group.

The KE payload is created in the first exchange packet when using aggressive mode. The KE payload depends on the DH group used. This is why there cannot be multiple DH groups in IKEv1 aggressive mode. In IKEv2, which uses a similar method to IKEv1 Aggressive Mode, there is an INVALID_KE response payload that can inform the initiator of the responder's desired DH group and so an IKEv2 connection can actually recover from picking the wrong DH group by restarting its negotiation.

salifetime

how long a particular instance of a connection (a set of encryption/authentication keys for user packets) should last, from successful negotiation to expiry; acceptable values are an integer optionally followed by s (a time in seconds) or a decimal number followed by m, h, or d (a time in minutes, hours, or days respectively) (default 8h, maximum 24h). Normally, the connection is renegotiated (via the keying channel) before it expires. The two ends need not exactly agree on salifetime, although if they do not, there will be some clutter of superseded connections on the end which thinks the lifetime is longer.

The keywords "keylife" and "lifetime" are obsoleted aliases for "salifetime." Change your configs to use "salifetime" instead.

ipsec-max-bytes

how many bytes can be sent, or how many bytes can be received on an IPsec SA instance for a connection; acceptable values are an integer optionally followed by KiB, MiB, GiB, TiB, PiB or EiB to signify kilobytes, megabytes, gigabytes, terabytes, petabytes or exabytes.

An IPsec SA contains two keys, one for inbound and one for outbound traffic. The ipsec-max-bytes sets two limits on each of these keys: the hard limit which is the total number of bytes that a given key can encrypt, and the soft limit which is the number of bytes that can be encrypted before a renegotiation of the IPsec SA is initiated. Normally the renegotiation (via the IKE SA) is completed before the ipsec-max-bytes value is reached.

Pluto sets the the original initiator's soft limit to 25% of ipsec-max-bytes (with 12% fuzz) and on the original responder's soft limit to 50% of ipsec-max-bytes (with 12% fuzz). This way the original initiator hopefully is the one initiating the renegotiation of the IPsec SA.

This option is not negotiated between IKE peers. Each end of the IPsec SA sets their own limits independently.

The default (hard limit) is 2^63 bytes. The original initiator's soft limit is 2^61 bytes (approx.) and the original responder's soft limit is 2^62 bytes (approx.).

ipsec-max-packets

how many packets can be sent/received on a particular instance of a connection (a set of encryption/authentication keys for user packets) , from successful negotiation to expiry.

Default values and caveats are the same as for ipsec-max-bytes. This option uses a prefix without "B" for bytes.

replay-window

The size of the IPsec SA replay window protection in packets. Kernels (Linux, and most BSDs) support a window size of at least 2040 packets. The default replay window size is 128 packets.

A value of 0 disables replay protection. Disabling of replay protection is sometimes used on a pair of IPsec servers in a High Availability setup, or on servers with very unpredictable latency, such as mobile networks, which can cause an excessive amount of out of order packets.

Disabling replay protection will also disable Extended Sequence Numbers (esn=no), as advise from RFC 4303 caused some stacks to not support ESN without a replay-window.

Note: on Linux, sequence errors can be seen in /proc/net/xfrm_stat.

Note: the BSD setkey utility displays the replay window size in bytes (8 packets per byte) and not packets.

Technically, at least the Linux kernel can install IPsec SA's with an IPsec SA Sequence Number, but this is currently not supported by libreswan.

rekey

whether a connection should be renegotiated when it is about to expire; acceptable values are yes (the default) and no. The two ends need not agree, but while a value of no prevents Pluto from requesting renegotiation, it does not prevent responding to renegotiation requested from the other end, so no will be largely ineffective unless both ends agree on it.

rekeymargin

how long before connection expiry or keying-channel expiry should attempts to negotiate a replacement begin; acceptable values as for salifetime (default 9m). Relevant only locally, other end need not agree on it.

rekeyfuzz

maximum percentage by which rekeymargin should be randomly increased to randomize rekeying intervals (important for hosts with many connections); acceptable values are an integer, which may exceed 100, followed by a `%' (default set by ipsec_pluto(8), currently 100%). The value of rekeymargin, after this random increase, must not exceed salifetime. The value 0% will suppress time randomization. Relevant only locally, other end need not agree on it.

ikelifetime

how long the keying channel of a connection (buzzphrase: “IKE SA” or “ISAKMP SA”) should last before being renegotiated; acceptable values as for salifetime. The default as of version 4.2 is 8h, before that it was 1h. The maximum is 24h. The two-ends-disagree case is similar to that of salifetime.

retransmit-timeout

how long a single packet, including retransmits of that packet, may take before the IKE attempt is aborted. If rekeying is enabled, a new IKE attempt is started. The default set by ipsec_pluto(8), currently is 60s.

See also: retransmit-interval and rekey.

retransmit-interval

the initial interval time period, specified in msecs, that pluto waits before retransmitting an IKE packet. This interval is doubled for each attempt (exponential back-off). The default set by ipsec_pluto(8), currently is 500.

See also: retransmit-timeout and rekey

compress

whether IPComp compression of content is proposed on the connection (link-level compression does not work on encrypted data, so to be effective, compression must be done before encryption); acceptable values are yes and no (the default).

For IKEv1, compress settings on both peers must match. For IKEv2, compression can only be suggested and a mismatched compress setting results in connection without compression.

When set to yes, compression is negotiated for the DEFLATE compression algorithm.

metric

Set the metric for added routes. This value is passed to the _updown scripts as PLUTO_METRIC.

Acceptable values are positive numbers, with the default being 1.

mtu

Set the MTU for the route(s) to the remote endpoint and/or subnets. This is sometimes required when the overhead of the IPsec encapsulation would cause the packet the become too big for a router on the path. Since IPsec cannot trust any unauthenticated ICMP messages, PATH MTU discovery does not work. This can also be needed when using "6to4" IPV6 deployments, which adds another header on the packet size.

Acceptable values are positive numbers. There is no default.

tfc

Enable Traffic Flow Confidentiality ("TFC") (RFC-4303) for outgoing ESP packets in Tunnel Mode. When enabled, ESP packets are padded to the specified size (up to the PMTU size) to prevent leaking information based on ESP packet size. This option is ignored for AH and for ESP in Transport Mode as those always leak traffic characteristics and applying TFC will not do anything. Acceptable values are positive numbers. The value 0 means TFC padding is not performed. Currently this feature is only implemented for the Linux XFRM stack. In IKEv2, when the notify payload ESP_TFC_PADDING_NOT_SUPPORTED is received, TFC padding is disabled. The default is not to do any TFC padding, but this might change in the near future.

send-no-esp-tfc

Whether or not to tell the remote peer that we do not support Traffic Flow Confidentiality ("TFC") (RFC-4303). Possible values are no (the default) which allows the peer to use TFC or yes which prevents to peer from using TFC. This does not affect whether this endpoint uses TFC, which only depends on the local tfc setting. This option is only valid for IKEv2.

nflog

If set, the NFLOG group number to log this connection's pre-crypt and post-decrypt traffic to. The default value of 0 means no logging at all. This option is only available on linux kernel 2.6.14 and later. It allows common network utilities such as tcpdump, wireshark and dumpcap, to use nflog:XXX pseudo interfaces where XXX is the nflog group number. During the updown phase of a connection, iptables will be used to add and remove the source/destination pair to the nflog group specified. The rules are setup with the nflog-prefix matching the connection name. See also the global nflog-all option.

mark

If set, the MARK to set for the IPsec SA of this connection. The format of a CONNMARK is mark/mask. If the mask is left out, a default mask of 0xffffffff is used. A mark value of -1 means to assign a new global unique mark number for each instance of the connection. Global marks start at 1001. This option is only available on linux XFRM kernels. It can be used with iptables to create custom iptables rules using CONNMARK. It can also be used with Virtual Tunnel Interfaces ("VTI") to direct marked traffic to specific vtiXX devices.

mark-in

The same as mark, but mark-in only applies to the inbound half of the IPsec SA. It overrides any mark= setting.

mark-out

The same as mark, but mark-out only applies to the outbound half of the IPsec SA. It overrides any mark= setting.

vti-interface

This option is deprecated. See ipsec-interface instead.

This option is used to create "Routing based VPNs" (as opposed to "Policy based VPNs"). It will create a new interface that can be used to route traffic in for encryption/decryption. The Virtual Tunnel Interface ("VTI") interface name is used to for all IPsec SA's created by this connection. This requires that the connection also enables either the mark= or mark-in= / mark-out- option(s). All traffic marked with the proper MARKs will be automatically encrypted if there is an IPsec SA policy covering the source/destination traffic. Tools such as tcpdump and iptables can be used on all cleartext pre-encrypt and post-decrypt traffic on the device. See the libreswan wiki for example configurations that use VTI.

VTI interfaces are currently only supported on Linux with XFRM. The _updown script handles certain Linux specific interfaces settings required for proper functioning (disable_policy, rp_filter, forwarding, etc). Interface names are limited to 16 characters and may not allow all characters to be used. If marking and vti-routing=yes is used, no manual iptables should be required. However, administrators can use the iptables mangle table to mark traffic manually if desired.

vti-routing

Whether or not to add network rules or routes for IPsec SA's to the respective VTI devices. Valid values are yes (the default) or no. When using "routing based VPNs" with a subnets policy of 0.0.0.0/0, this setting needs to set to no to prevent imploding the tunnel, and the administrator is expected to manually add ip rules and ip routes to configure what traffic must be encrypted. When set to yes, the _updown script will automatically route the leftsubnet/rightsubnet traffic into the VTI device specified with vti-interface

vti-shared

Whether or not the VTI device is shared amongst connections. Valid values are no (the default) or yes. When set to no, the VTI device is automatically deleted if the connection is a single non-instantiated connection. If a connection instantiates (eg right=%any) then this option has no effect, as the VTI device is not removed as it is shared with multiple roadwarriors.

ipsec-interface

Create a new or use an existing Virtual Interface for "Routing based IPsec VPNs" (as opposed to "Policy based VPNs").

On Linux, FreeBSD, and NetBSD the Virtual Interface device name is ipsec; OpenBSD uses the name sec.

Valid options are yes, no or a number. When using a number, the IPsec interface created and/or used will use that number as part of the interface name. For example setting ipsec-interface=5 will create and/or use the ipsec5 interface. The value 0 cannot be used and is interpreted as no. The value yes is interpreted as the number 1, and thus will use the interface named ipsec1. An IP address can be configured for this interface using the interface-ip= option.

The Virtual Interface is used to route outbound traffic that needs to be encrypted, and will decrypt inbound traffic that arrives on this interface. All traffic that is routed to this interface will be automatically encrypted providing the IPsec SA policy covers this traffic. Traffic not matching the IPsec SA will be dropped. Tools such as tcpdump, iptables, ifconfig and tools that need traffic counters can be used on all cleartext pre-encrypt and post-decrypt traffic on the device. When leftsubnet= is equal to rightsubnet=, the routing needs to be managed by an external routing daemon or manually by the administrator.

Support:

Linux (since 4.19)

This option is currently only supported on Linux kernels when compiled with XFRMi support (CONFIG_XFRM_INTERFACE). The number of the ipsecX device corresponds with the XFRM IF_ID policy option of the IPsec SA in the Linux kernel. On Linux, XFRMi interfaces can be managed by libreswan automatically or can be preconfigured on the system using the existing init system or via networking tools such as systemd-networkd and NetworkManager. The _updown script handles certain Linux specific interfaces settings required for proper functioning, such as forwarding and routing rules for IPsec traffic.

Netbsd (since 8.0), OpenBSD (since 7.4), FreeBSD (since 11.0)

See the release notes for current level of support in Libreswan

interface-ip=

The IP address and netmask to configure on a virtual device (eg ipsecXXX). This is often used when building Routing based IPsec tunnels using transport mode and GRE, but can also be useful in other scenarios. Currently requires ipsec-interface=. See also leftvti= for configuring IP addresses when using deprecated VTI.

priority

The priority in the kernel SPD/SAD database, when matching up packets. Each kernel (XFRM, OSX, etc) has its own mechanism for setting the priority. Setting this option to non-zero passes the priority to the kernel stack unmodified. The maximum value depends on the stack. It is recommended not to exceed 65536

XFRM use a priority system based on "most specific match first". It uses an internal algorithm to calculate these based on network prefix length, protocol and port selectors. A lower value means a higher priority.

Typical values are about the 2000 range. These can be seen on the XFRM stack using ip xfrm policy when the connection is up. For "anonymous IPsec" or Opportunistic Encryption based connections, a much lower priority (65535) is used to ensure administrator configured IPsec always takes precedence over opportunistic IPsec.

sendca

How much of our available X.509 trust chain to send with the End certificate, excluding any root CA's. Specifying issuer sends just the issuing intermediate CA, while all will send the entire chain of intermediate CA's.none (the default) will not send any CA certs.

policy-label

The string representation of an access control security label that is interpreted by the Linux Security Module (e.g. SELinux) for use with Labeled IPsec.

For example, policy-label=system_u:object_r:ipsec_spd_t:s0-s15:c0.c1023

Labeled IPsec support is IKEv2 only.

failureshunt

what to do with packets when negotiation fails. The default is none: no shunt; passthrough, drop, and reject have the obvious meanings.

negotiationshunt

What to do with packets during the IKE negotiation. Valid options are hold (the default) or passthrough. This should almost always be left to the default hold value to avoid cleartext packet leaking. The only reason to set this to passthrough is if plaintext service availability is more important than service security or privacy, a scenario that also implies failureshunt=passthrough and most likely authby=%null using Opportunistic Encryption.

debug

Enable per-connection debug logging. For more information see ipsec-whack(8) and the --debug option.

Note
This feature is used by Libreswan developers. The default logs should provide sufficient information to diagnose configuration and connection problems.