XFRM Interface Development Notes

From Libreswan
Jump to navigation Jump to search

libreswan/pluto design choices

pluto keywords

ipsec-interface=no|yes|<n> (n = user configured device, systemd-networkd, OpenWRT..)
leftinterface-ip=<ip>/<mask> get configured on xfrmi or loopback /* to be implemented */

possibly need split of mark to in and out in the future, which means 3 key words for mask.


pluto discarded keywords

vti-routing is discarded. Pluto would add route as longs the leftsubnet != rightsubnet. If they are same such as to no route will be added.

  • allow names ipsec1 ... ipsecx.
  • Should we allow names other than ipsecX ?
  • initial thought is keep "xfrm interface id" and "xfrm output mark" consistent.
  • interface creation is inside pluto.
  • create bugzilla entry for 4.18 support.
  • XFRMi code is compile time option. If the kernel headers do no't support it won't compile. Think RHEL 6 or Debian Weezy.

Kernel XFRM - related


Commit cover letter from Steffen 20180612, Merged in 4.19:


Steffen Klassert June 12, 2018, 7:56 a.m.

This patchset introduces new virtual xfrm interfaces.
The design of virtual xfrm interfaces interfaces was
discussed at the Linux IPsec workshop 2018. This patchset
implements these interfaces as the IPsec userspace and
kernel developers agreed. The purpose of these interfaces
is to overcome the design limitations that the existing
VTI devices have.

The main limitations that we see with the current VTI are the

- VTI interfaces are L3 tunnels with configurable endpoints.
  For xfrm, the tunnel endpoint are already determined by the SA.
  So the VTI tunnel endpoints must be either the same as on the
  SA or wildcards. In case VTI tunnel endpoints are same as on
  the SA, we get a one to one correlation between the SA and
  the tunnel. So each SA needs its own tunnel interface.

  On the other hand, we can have only one VTI tunnel with
  wildcard src/dst tunnel endpoints in the system because the
  lookup is based on the tunnel endpoints. The existing tunnel
  lookup won't work with multiple tunnels with wildcard
  tunnel endpoints. Some usecases require more than on
  VTI tunnel of this type, for example if somebody has multiple
  namespaces and every namespace requires such a VTI.

- VTI needs separate interfaces for IPv4 and IPv6 tunnels.
  So when routing to a VTI, we have to know to which address
  family this traffic class is going to be encapsulated.
  This is a lmitation because it makes routing more complex
  and it is not always possible to know what happens behind the
  VTI, e.g. when the VTI is move to some namespace.

- VTI works just with tunnel mode SAs. We need generic interfaces
  that ensures transfomation, regardless of the xfrm mode and
  the encapsulated address family.

- VTI is configured with a combination GRE keys and xfrm marks.
  With this we have to deal with some extra cases in the generic
  tunnel lookup because the GRE keys on the VTI are actually
  not GRE keys, the GRE keys were just reused for something else.
  All extensions to the VTI interfaces would require to add
  even more complexity to the generic tunnel lookup.

To overcome this, we started with the following design goal:

- It should be possible to tunnel IPv4 and IPv6 through the same

- No limitation on xfrm mode (tunnel, transport and beet).

- Should be a generic virtual interface that ensures IPsec
  transformation, no need to know what happens behind the

- Interfaces should be configured with a new key that must match a
  new policy/SA lookup key.

- The lookup logic should stay in the xfrm codebase, no need to
  change or extend generic routing and tunnel lookups.

- Should be possible to use IPsec hardware offloads of the underlying

Initial xfrmi kernel commits




Add mask and rename to XFRMA_SET_MARK, Steffen

Initial Commit, Lorenzo

4.18 cherry pick the following commits to get xfrmi










Which mark? there as so many now!

There are wo marks relevant to IPsec , xfrm in Linux kernel. XFRMA_MARK and XFRMA_SET_MARK(aka XFRMA_OUTPUT_MARK). Now on I use XFRM_OUTPUT_MARK mark which as of 4.20 mean XFRM_SET_MARK/XFRM_SET_MARK_MASK

A quick summary of difference between the two marks. XFRM_OUTPUT_MARK is for routing a packet after XFRM(think as ESP out or Clear text in) a look up key for routing rule, while XFRMA_MARK is for routing a clear text packet into xfrm sub system; XFRMA_MARK is a lookup key in SPDB and SADB.

XFRMA_MARK (along with mask) is used inside xfrm code to find policy and state on for an outgoing, clear text, packet Jamal's notes (clear text packet out only. Then I wonder why set XFRMA_MARK on incoming SA? See libreswan commit. What is the use case of it. Is it supposed to copy mark from incoming ESP to the incoming clear text? I don't thing so yet! I think that is done by the new XFRMA_OUTPUT_MARK on the in SA.

To be clear, both marks could be used for routing a packet at its different stages, while it goes through Linux stack. or in a more complicated case where a packet get marked when arrive at the host application replies. Now the magic is the incoming packet's mark is used for two things to decide source address of the respons, the mark get copied to related outgoing clear text packet. Then it is used to route in multi homed situation. While XFRM_OUTPUT_MARK is used for routing ESP packet on the output side and clear text packet on the input side. You see a bit more details on XFRMA_MARK see Jamal's explanation

XFRMA_MARK - mark(u32)/mask(u32). XFRM_OUTPUT_MARK - mark(u32)/[mask(u32). So one difference is for XFRM_OUTPUT_MARK mask is optional as of 4.19, 4.18 did not support mask for OUTPUT_MARK.

IPsec and routing has to share same mark. That why there is mask. One part for IPsec/XFRM and other part for the rest of the system use. XFRM stack should pass on the mark set by the system when correct mask is used. Masked part is opaque to xfrm.

use case of marks

  • Simple use case XFRMI interface.

XFRM_OUTPUT_MARK by libreswan when the the other/peer end is inside the extruded tunnel. In other words. Say /32-to-/32 tunnel without NAT or tunnel. Note it is adding rules to rout

XFRMi and RW

XFRMi need "ip rule" when the peer's ip is covered by ipsec policy and for /32-to-/32 IPsec policies. Antony's initial attempt was the following rule and route 50 table. Because of "table 50" we don't need 0/1 and 128/1 split rules when the the other end is 0/0.

ip rule add from to fwmark 0x1/0xffffff lookup main
ip rule add from to lookup 50
ip route add default dev ipsec1 src table 50

= New xfrmi rule attempt

# with not rule

0:      from all lookup local
100:    not from all fwmark 0x1 lookup 50
32766:  from all lookup main
32767:  from all lookup default

# ip route list table main
default via dev eth0 dev eth0 proto kernel scope link src             

# ip route list table 50
default dev ipsec1 scope link src 

# With yes rule
ip route dev ipsec1 scope link src
default via dev eth0 proto static dev ipsec1 scope link src dev eth0 proto kernel scope link src

ip rule
0:      from all lookup local
100:    from all fwmark 0x1 lookup 50
32766:  from all lookup main
32767:  from all lookup default

ip route show table 50
#esp traffic via dev eth0

XFRMi and RW Tuomo's suggestion

20190212 Tuomo thinks the above rule is negation rule with "not" which is not friendly. We should have two rules and a route entry. A further refined form of this is below.

ip rule add prio 99 fwmark 6/0xffffffff lookup main
ip rule add prio 100 to lookup 50
ip route add  table 50

201810 proff of concept rule

ip rule add prio 100 to not fwmark 1/0xffffffff lookup 50
ip route add dev ipsec1 src table 50
ip route add default dev ipsec1 src table 50
  • XFRMI interface and multihoming. Yet to test this. It would bet interesting to test. If it all work the idea is mark on incoming ESP, set by iptable rule, would get copied first to incoming clear packet, then to outgoing clear text packet and finally to the outgoing ESP packet. With XFRMi at every stage there are could routlookup involving fwmark with mark. I wonder if it works. Any one tested it?
  • Simple case IPsec with IP_VTI, only libreswan using marks on the ssytem.

Then you can pick any mark you want. in libreswan config you set mark=2/0xffffffff. The rest is transparent to users, magic happens.

  • more complicated case VTI and multihoming

Lets say you have two upstreams. Both route packets to your host. Then you would start to use iptable to mark the incoming traffic, ESP. First the ESP will get decrypted and new clear text packet will retain the mark. The mark will be used to get source address for response packet. Then it will go through xfrm. Get encrypted again retainng the mark. Then get routed to via the interface the the packet came in.

Now with VTI. A portion of mark clear text outgoing response is over written (using mask). If you are familiar with the commands bellow you get an idea what is going on. I herd it works. I wonder if the incoming ESP packet's mark, say set by iptable before hitting XFRM input code, would end up on the outgoing ESP too or just on the clear text response.

iptables -t mangle -N MARK-ISP1
iptables -t mangle -A MARK-ISP1 -j MARK --set-mark 1
iptables -t mangle -A MARK-ISP1 -j CONNMARK --save-mark

iptables -t mangle -N MARK-ISP2
iptables -t mangle -A MARK-ISP2 -j MARK --set-mark 2
iptables -t mangle -A MARK-ISP2 -j CONNMARK --save-mark

ip rule add fwmark 1 table ISP1
ip rule add fwmark 2 table ISP2
ip route add table ISP1 dev eth1 src
ip route add table ISP2 dev eth2 src

Track bugs and related issue

Strongswan support


if you delete the namespace before deleting xfrm interface, xfrm interface will not go away

Message from syslogd@swantest at Aug 28 18:38:15 ...
 kernel:unregister_netdevice: waiting for eth1 to become free. Usage count = 1

routed vpn

  • same destination, difference source routes to two different gateways.
conn north-east

conn north-west
        rightsubnet= #same subnet 
	leftsubnet=  #different 

more configuration options

  • Currently the new routing table is id "50" is hardcoded
  • same route to two different gateways. this will need an option to configure "ip rule priority"

xfrm interface with hardware offload

systemd-networkd man page specify, for the hardware offload, network interface must match NIC used.