An introduction to Linux virtual interfaces: Tunnels
Linux has supported many kinds of tunnels, but new users may be confused by their differences and unsure which one is best suited for a given use case. In this article, I will give a brief introduction for commonly used tunnel interfaces in the Linux kernel. There is no code analysis, only a brief introduction to the interfaces and their usage on Linux. Anyone with a network background might be interested in this information. A list of tunnel interfaces, as well as help on specific tunnel configuration, can be obtained by issuing the iproute2 command
ip link help.
This post covers the following frequently used interfaces:
- IPIP Tunnel
- SIT Tunnel
- ip6tnl Tunnel
- VTI and VTI6
- GRE and GRETAP
- GRE6 and GRE6TAP
- ERSPAN and IP6ERSPAN
After reading this article, you will know what these interfaces are, the differences between them, when to use them, and how to create them.
IPIP tunnel, just as the name suggests, is an IP over IP tunnel, defined in RFC 2003. The IPIP tunnel header looks like:
It’s typically used to connect two internal IPv4 subnets through public IPv4 internet. It has the lowest overhead but can only transmit IPv4 unicast traffic. That means you cannot send multicast via IPIP tunnel.
IPIP tunnel supports both IP over IP and MPLS over IP.
Note: When the
ipip module is loaded, or an IPIP device is created for the first time, the Linux kernel will create a
tunl0 default device in each namespace, with attributes
remote=any. When receiving IPIP protocol packets, the kernel will forward them to
tunl0 as a fallback device if it can’t find another device whose local/remote attributes match their source or destination address more closely.
Here is how to create an IPIP tunnel:
On Server A: # ip link add name ipip0 type ipip local remote # ip link set ipip0 up # ip addr add /24 dev ipip0 Add a remote internal subnet route if the endpoints don't belong to the same subnet # ip route add /24 dev ipip0 On Server B: # ip link add name ipip0 type ipip local remote # ip link set ipip0 up # ip addr add /24 dev ipip0 # ip route add /24 dev ipip0
Everything you need to grow your career.
With your free Red Hat Developer program membership, unlock our library of cheat sheets and ebooks on next-generation application development.SIGN UP
SIT stands for Simple Internet Transition. The main purpose is to interconnect isolated IPv6 networks, located in global IPv4 internet.
Initially, it only had an IPv6 over IPv4 tunneling mode. After years of development, however, it acquired support for several different modes, such as
ipip (the same with IPIP tunnel),
any is used to accept both IP and IPv6 traffic, which may prove useful in some deployments. SIT tunnel also supports ISATA, and here is a usage example.
The SIT tunnel header looks like:
sit module is loaded, the Linux kernel will create a default device, named
Here is how to create a SIT tunnel:
On Server A: # ip link add name sit1 type sit local remote mode any # ip link set sit1 up # ip addr add /24 dev sit1
Then, perform the same steps on the remote side.
ip6tnl is an IPv4/IPv6 over IPv6 tunnel interface, which looks like an IPv6 version of the SIT tunnel. The tunnel header looks like:
ip6tnl supports modes
ipip6 is IPv4 over IPv6, and mode
ip6ip6 is IPv6 over IPv6, and mode
any supports both IPv4/IPv6 over IPv6.
ip6tnl module is loaded, the Linux kernel will create a default device, named
Here is how to create an ip6tnl tunnel:
# ip link add name ipip6 type ip6tnl local remote mode any
Virtual Tunnel Interface (VTI) on Linux is similar to Cisco’s VTI and Juniper’s implementation of secure tunnel (st.xx).
This particular tunneling driver implements IP encapsulations, which can be used with xfrm to give the notion of a secure tunnel and then use kernel routing on top.
In general, VTI tunnels operate in almost the same way as ipip or sit tunnels, except that they add a fwmark and IPsec encapsulation/decapsulation.
VTI6 is the IPv6 equivalent of VTI.
Here is how to create a VTI tunnel:
# ip link add name vti1 type vti key local remote # ip link set vti1 up # ip addr add /24 dev vti1 # ip xfrm state add src dst spi mode tunnel # ip xfrm state add src dst spi mode tunnel # ip xfrm policy add dir in tmpl src dst mode tunnel mark # ip xfrm policy add dir out tmpl src dst mode tunnel mark
Generic Routing Encapsulation, also known as GRE, is defined in RFC 2784
GRE tunneling adds an additional GRE header between the inside and outside IP headers. In theory, GRE could encapsulate any Layer 3 protocol with a valid Ethernet type, unlike IPIP, which can only encapsulate IP. The GRE header looks like:
Note that you can transport multicast traffic and IPv6 through a GRE tunnel.
gre module is loaded, the Linux kernel will create a default device, named
Here is how to create a GRE tunnel:
# ip link add name gre1 type gre local remote [seq] key
While GRE tunnels operate at OSI Layer 3, GRETAP works at OSI Layer 2, which means there is an Ethernet header in the inner header.
Here is how to create a GRETAP tunnel:
# ip link add name gretap1 type gretap local remote
GRE6 is the IPv6 equivalent of GRE, which allows us to encapsulate any Layer 3 protocol over IPv6. The tunnel header looks like:
GRE6TAP, just like GRETAP, has an Ethernet header in the inner header:
Here is how to create a GRE tunnel:
# ip link add name gre1 type gre6 local remote # ip link add name gretap1 type gretap6 local remote
Tunneling can happen at multiple levels in the networking stack. IPIP, SIT, GRE tunnels are at the IP level, while FOU (foo over UDP) is UDP-level tunneling.
There are some advantages of using UDP tunneling as UDP works with existing HW infrastructure, like RSS in NICs, ECMP in switches, and checksum offload. The developer’s patch set shows significant performance increases for the SIT and IPIP protocols.
Currently, the FOU tunnel supports encapsulation protocol based on IPIP, SIT, GRE. An example FOU header looks like:
Here is how to create a FOU tunnel:
# ip fou add port 5555 ipproto 4 # ip link add name tun1 type ipip remote 192.168.1.1 local 192.168.1.2 ttl 225 encap fou encap-sport auto encap-dport 5555
The first command configured a FOU receive port for IPIP bound to 5555; for GRE, you need to set
ipproto 47. The second command set up a new IPIP virtual interface (tun1) configured for FOU encapsulation, with dest port 5555.
NOTE: FOU is not supported in Red Hat Enterprise Linux.
Generic UDP Encapsulation (GUE) is another kind of UDP tunneling. The difference between FOU and GUE is that GUE has its own encapsulation header, which contains the protocol info and other data.
Currently, GUE tunnel supports inner IPIP, SIT, GRE encapsulation. An example GUE header looks like:
Here is how to create a GUE tunnel:
# ip fou add port 5555 gue # ip link add name tun1 type ipip remote 192.168.1.1 local 192.168.1.2 ttl 225 encap gue encap-sport auto encap-dport 5555
This will set up a GUE receive port for IPIP bound to 5555, and an IPIP tunnel configured for GUE encapsulation.
NOTE: GUE is not supported in Red Hat Enterprise Linux.
Generic Network Virtualization Encapsulation (GENEVE) supports all of the capabilities of VXLAN, NVGRE, and STT and was designed to overcome their perceived limitations. Many believe GENEVE could eventually replace these earlier formats entirely. The tunnel header looks like:
which looks very similar to VXLAN. The main difference is that the GENEVE header is flexible. It’s very easy to add new features by extending the header with a new Type-Length-Value (TLV) field. For more details, you can see the latest geneve ietf draft or refer to this What is GENEVE? article.
Open Virtual Network (OVN) uses GENEVE as default encapsulation. Here is how to create a GENEVE tunnel:
# ip link add name geneve0 type geneve id remote
Encapsulated Remote Switched Port Analyzer (ERSPAN) uses GRE encapsulation to extend the basic port mirroring capability from Layer 2 to Layer 3, which allows the mirrored traffic to be sent through a routable IP network. The ERSPAN header looks like:
The ERSPAN tunnel allows a Linux host to act as an ERSPAN traffic source and send the ERSPAN mirrored traffic to either a remote host or to an ERSPAN destination, which receives and parses the ERSPAN packets generated from Cisco or other ERSPAN-capable switches. This setup could be used to analyze, diagnose, and detect malicious traffic.
Linux currently supports most features of two ERSPAN versions: v1 (type II) and v2 (type III).
Here is how to create an ERSPAN tunnel:
# ip link add dev erspan1 type erspan local remote seq key erspan_ver 1 erspan or # ip link add dev erspan1 type erspan local remote seq key erspan_ver 2 erspan_dir erspan_hwid Add tc filter to monitor traffic # tc qdisc add dev handle ffff: ingress # tc filter add dev parent ffff: matchall skip_hw action mirred egress mirror dev erspan1
Here is a summary of all the tunnels we introduced.
|Tunnel/Link Type||Outer Header||Encapsulate Header||Inner Header|
|gretap||IPv4||GRE||Ether + IPv4/IPv6|
|gre6tap||IPv6||GRE||Ether + IPv4/IPv6|
|gue||IPv4/IPv6||UDP + GUE||IPv4/IPv6/GRE|
|geneve||IPv4/IPv6||UDP + Geneve||Ether + IPv4/IPv6|
|erspan||IPv4||GRE + ERSPAN||IPv4/IPv6|
|ip6erspan||IPv6||GRE + ERSPAN||IPv4/IPv6|
Note: All configurations in this tutorial are volatile and won’t survive to a server reboot. If you want to make the configuration persistent across reboots, please consider using a networking configuration daemon, such as NetworkManager, or distribution-specific mechanisms.
Take advantage of your Red Hat Developers membership and download RHEL today at no cost.