1.. SPDX-License-Identifier: GPL-2.0
2
3====
4L2TP
5====
6
7This document describes how to use the kernel's L2TP drivers to
8provide L2TP functionality. L2TP is a protocol that tunnels one or
9more sessions over an IP tunnel. It is commonly used for VPNs
10(L2TP/IPSec) and by ISPs to tunnel subscriber PPP sessions over an IP
11network infrastructure. With L2TPv3, it is also useful as a Layer-2
12tunneling infrastructure.
13
14Features
15========
16
17L2TPv2 (PPP over L2TP (UDP tunnels)).
18L2TPv3 ethernet pseudowires.
19L2TPv3 PPP pseudowires.
20L2TPv3 IP encapsulation.
21Netlink sockets for L2TPv3 configuration management.
22
23History
24=======
25
26The original pppol2tp driver was introduced in 2.6.23 and provided
27L2TPv2 functionality (rfc2661). L2TPv2 is used to tunnel one or more PPP
28sessions over a UDP tunnel.
29
30L2TPv3 (rfc3931) changes the protocol to allow different frame types
31to be passed over an L2TP tunnel by moving the PPP-specific parts of
32the protocol out of the core L2TP packet headers. Each frame type is
33known as a pseudowire type. Ethernet, PPP, HDLC, Frame Relay and ATM
34pseudowires for L2TP are defined in separate RFC standards. Another
35change for L2TPv3 is that it can be carried directly over IP with no
36UDP header (UDP is optional). It is also possible to create static
37unmanaged L2TPv3 tunnels manually without a control protocol
38(userspace daemon) to manage them.
39
40To support L2TPv3, the original pppol2tp driver was split up to
41separate the L2TP and PPP functionality. Existing L2TPv2 userspace
42apps should be unaffected as the original pppol2tp sockets API is
43retained. L2TPv3, however, uses netlink to manage L2TPv3 tunnels and
44sessions.
45
46Design
47======
48
49The L2TP protocol separates control and data frames.  The L2TP kernel
50drivers handle only L2TP data frames; control frames are always
51handled by userspace. L2TP control frames carry messages between L2TP
52clients/servers and are used to setup / teardown tunnels and
53sessions. An L2TP client or server is implemented in userspace.
54
55Each L2TP tunnel is implemented using a UDP or L2TPIP socket; L2TPIP
56provides L2TPv3 IP encapsulation (no UDP) and is implemented using a
57new l2tpip socket family. The tunnel socket is typically created by
58userspace, though for unmanaged L2TPv3 tunnels, the socket can also be
59created by the kernel. Each L2TP session (pseudowire) gets a network
60interface instance. In the case of PPP, these interfaces are created
61indirectly by pppd using a pppol2tp socket. In the case of ethernet,
62the netdevice is created upon a netlink request to create an L2TPv3
63ethernet pseudowire.
64
65For PPP, the PPPoL2TP driver, net/l2tp/l2tp_ppp.c, provides a
66mechanism by which PPP frames carried through an L2TP session are
67passed through the kernel's PPP subsystem. The standard PPP daemon,
68pppd, handles all PPP interaction with the peer. PPP network
69interfaces are created for each local PPP endpoint. The kernel's PPP
70subsystem arranges for PPP control frames to be delivered to pppd,
71while data frames are forwarded as usual.
72
73For ethernet, the L2TPETH driver, net/l2tp/l2tp_eth.c, implements a
74netdevice driver, managing virtual ethernet devices, one per
75pseudowire. These interfaces can be managed using standard Linux tools
76such as "ip" and "ifconfig". If only IP frames are passed over the
77tunnel, the interface can be given an IP addresses of itself and its
78peer. If non-IP frames are to be passed over the tunnel, the interface
79can be added to a bridge using brctl. All L2TP datapath protocol
80functions are handled by the L2TP core driver.
81
82Each tunnel and session within a tunnel is assigned a unique tunnel_id
83and session_id. These ids are carried in the L2TP header of every
84control and data packet. (Actually, in L2TPv3, the tunnel_id isn't
85present in data frames - it is inferred from the IP connection on
86which the packet was received.) The L2TP driver uses the ids to lookup
87internal tunnel and/or session contexts to determine how to handle the
88packet. Zero tunnel / session ids are treated specially - zero ids are
89never assigned to tunnels or sessions in the network. In the driver,
90the tunnel context keeps a reference to the tunnel UDP or L2TPIP
91socket. The session context holds data that lets the driver interface
92to the kernel's network frame type subsystems, i.e. PPP, ethernet.
93
94Userspace Programming
95=====================
96
97For L2TPv2, there are a number of requirements on the userspace L2TP
98daemon in order to use the pppol2tp driver.
99
1001. Use a UDP socket per tunnel.
101
1022. Create a single PPPoL2TP socket per tunnel bound to a special null
103   session id. This is used only for communicating with the driver but
104   must remain open while the tunnel is active. Opening this tunnel
105   management socket causes the driver to mark the tunnel socket as an
106   L2TP UDP encapsulation socket and flags it for use by the
107   referenced tunnel id. This hooks up the UDP receive path via
108   udp_encap_rcv() in net/ipv4/udp.c. PPP data frames are never passed
109   in this special PPPoX socket.
110
1113. Create a PPPoL2TP socket per L2TP session. This is typically done
112   by starting pppd with the pppol2tp plugin and appropriate
113   arguments. A PPPoL2TP tunnel management socket (Step 2) must be
114   created before the first PPPoL2TP session socket is created.
115
116When creating PPPoL2TP sockets, the application provides information
117to the driver about the socket in a socket connect() call. Source and
118destination tunnel and session ids are provided, as well as the file
119descriptor of a UDP socket. See struct pppol2tp_addr in
120include/linux/if_pppol2tp.h. Note that zero tunnel / session ids are
121treated specially. When creating the per-tunnel PPPoL2TP management
122socket in Step 2 above, zero source and destination session ids are
123specified, which tells the driver to prepare the supplied UDP file
124descriptor for use as an L2TP tunnel socket.
125
126Userspace may control behavior of the tunnel or session using
127setsockopt and ioctl on the PPPoX socket. The following socket
128options are supported:-
129
130=========   ===========================================================
131DEBUG       bitmask of debug message categories. See below.
132SENDSEQ     - 0 => don't send packets with sequence numbers
133	    - 1 => send packets with sequence numbers
134RECVSEQ     - 0 => receive packet sequence numbers are optional
135	    - 1 => drop receive packets without sequence numbers
136LNSMODE     - 0 => act as LAC.
137	    - 1 => act as LNS.
138REORDERTO   reorder timeout (in millisecs). If 0, don't try to reorder.
139=========   ===========================================================
140
141Only the DEBUG option is supported by the special tunnel management
142PPPoX socket.
143
144In addition to the standard PPP ioctls, a PPPIOCGL2TPSTATS is provided
145to retrieve tunnel and session statistics from the kernel using the
146PPPoX socket of the appropriate tunnel or session.
147
148For L2TPv3, userspace must use the netlink API defined in
149include/linux/l2tp.h to manage tunnel and session contexts. The
150general procedure to create a new L2TP tunnel with one session is:-
151
1521. Open a GENL socket using L2TP_GENL_NAME for configuring the kernel
153   using netlink.
154
1552. Create a UDP or L2TPIP socket for the tunnel.
156
1573. Create a new L2TP tunnel using a L2TP_CMD_TUNNEL_CREATE
158   request. Set attributes according to desired tunnel parameters,
159   referencing the UDP or L2TPIP socket created in the previous step.
160
1614. Create a new L2TP session in the tunnel using a
162   L2TP_CMD_SESSION_CREATE request.
163
164The tunnel and all of its sessions are closed when the tunnel socket
165is closed. The netlink API may also be used to delete sessions and
166tunnels. Configuration and status info may be set or read using netlink.
167
168The L2TP driver also supports static (unmanaged) L2TPv3 tunnels. These
169are where there is no L2TP control message exchange with the peer to
170setup the tunnel; the tunnel is configured manually at each end of the
171tunnel. There is no need for an L2TP userspace application in this
172case -- the tunnel socket is created by the kernel and configured
173using parameters sent in the L2TP_CMD_TUNNEL_CREATE netlink
174request. The "ip" utility of iproute2 has commands for managing static
175L2TPv3 tunnels; do "ip l2tp help" for more information.
176
177Debugging
178=========
179
180The driver supports a flexible debug scheme where kernel trace
181messages may be optionally enabled per tunnel and per session. Care is
182needed when debugging a live system since the messages are not
183rate-limited and a busy system could be swamped. Userspace uses
184setsockopt on the PPPoX socket to set a debug mask.
185
186The following debug mask bits are available:
187
188================  ==============================
189L2TP_MSG_DEBUG    verbose debug (if compiled in)
190L2TP_MSG_CONTROL  userspace - kernel interface
191L2TP_MSG_SEQ      sequence numbers handling
192L2TP_MSG_DATA     data packets
193================  ==============================
194
195If enabled, files under a l2tp debugfs directory can be used to dump
196kernel state about L2TP tunnels and sessions. To access it, the
197debugfs filesystem must first be mounted::
198
199	# mount -t debugfs debugfs /debug
200
201Files under the l2tp directory can then be accessed::
202
203	# cat /debug/l2tp/tunnels
204
205The debugfs files should not be used by applications to obtain L2TP
206state information because the file format is subject to change. It is
207implemented to provide extra debug information to help diagnose
208problems.) Users should use the netlink API.
209
210/proc/net/pppol2tp is also provided for backwards compatibility with
211the original pppol2tp driver. It lists information about L2TPv2
212tunnels and sessions only. Its use is discouraged.
213
214Unmanaged L2TPv3 Tunnels
215========================
216
217Some commercial L2TP products support unmanaged L2TPv3 ethernet
218tunnels, where there is no L2TP control protocol; tunnels are
219configured at each side manually. New commands are available in
220iproute2's ip utility to support this.
221
222To create an L2TPv3 ethernet pseudowire between local host 192.168.1.1
223and peer 192.168.1.2, using IP addresses 10.5.1.1 and 10.5.1.2 for the
224tunnel endpoints::
225
226	# ip l2tp add tunnel tunnel_id 1 peer_tunnel_id 1 udp_sport 5000 \
227	  udp_dport 5000 encap udp local 192.168.1.1 remote 192.168.1.2
228	# ip l2tp add session tunnel_id 1 session_id 1 peer_session_id 1
229	# ip -s -d show dev l2tpeth0
230	# ip addr add 10.5.1.2/32 peer 10.5.1.1/32 dev l2tpeth0
231	# ip li set dev l2tpeth0 up
232
233Choose IP addresses to be the address of a local IP interface and that
234of the remote system. The IP addresses of the l2tpeth0 interface can be
235anything suitable.
236
237Repeat the above at the peer, with ports, tunnel/session ids and IP
238addresses reversed.  The tunnel and session IDs can be any non-zero
23932-bit number, but the values must be reversed at the peer.
240
241========================       ===================
242Host 1                         Host2
243========================       ===================
244udp_sport=5000                 udp_sport=5001
245udp_dport=5001                 udp_dport=5000
246tunnel_id=42                   tunnel_id=45
247peer_tunnel_id=45              peer_tunnel_id=42
248session_id=128                 session_id=5196755
249peer_session_id=5196755        peer_session_id=128
250========================       ===================
251
252When done at both ends of the tunnel, it should be possible to send
253data over the network. e.g.::
254
255	# ping 10.5.1.1
256
257
258Sample Userspace Code
259=====================
260
2611. Create tunnel management PPPoX socket::
262
263	kernel_fd = socket(AF_PPPOX, SOCK_DGRAM, PX_PROTO_OL2TP);
264	if (kernel_fd >= 0) {
265		struct sockaddr_pppol2tp sax;
266		struct sockaddr_in const *peer_addr;
267
268		peer_addr = l2tp_tunnel_get_peer_addr(tunnel);
269		memset(&sax, 0, sizeof(sax));
270		sax.sa_family = AF_PPPOX;
271		sax.sa_protocol = PX_PROTO_OL2TP;
272		sax.pppol2tp.fd = udp_fd;       /* fd of tunnel UDP socket */
273		sax.pppol2tp.addr.sin_addr.s_addr = peer_addr->sin_addr.s_addr;
274		sax.pppol2tp.addr.sin_port = peer_addr->sin_port;
275		sax.pppol2tp.addr.sin_family = AF_INET;
276		sax.pppol2tp.s_tunnel = tunnel_id;
277		sax.pppol2tp.s_session = 0;     /* special case: mgmt socket */
278		sax.pppol2tp.d_tunnel = 0;
279		sax.pppol2tp.d_session = 0;     /* special case: mgmt socket */
280
281		if(connect(kernel_fd, (struct sockaddr *)&sax, sizeof(sax) ) < 0 ) {
282			perror("connect failed");
283			result = -errno;
284			goto err;
285		}
286	}
287
2882. Create session PPPoX data socket::
289
290	struct sockaddr_pppol2tp sax;
291	int fd;
292
293	/* Note, the target socket must be bound already, else it will not be ready */
294	sax.sa_family = AF_PPPOX;
295	sax.sa_protocol = PX_PROTO_OL2TP;
296	sax.pppol2tp.fd = tunnel_fd;
297	sax.pppol2tp.addr.sin_addr.s_addr = addr->sin_addr.s_addr;
298	sax.pppol2tp.addr.sin_port = addr->sin_port;
299	sax.pppol2tp.addr.sin_family = AF_INET;
300	sax.pppol2tp.s_tunnel  = tunnel_id;
301	sax.pppol2tp.s_session = session_id;
302	sax.pppol2tp.d_tunnel  = peer_tunnel_id;
303	sax.pppol2tp.d_session = peer_session_id;
304
305	/* session_fd is the fd of the session's PPPoL2TP socket.
306	 * tunnel_fd is the fd of the tunnel UDP socket.
307	 */
308	fd = connect(session_fd, (struct sockaddr *)&sax, sizeof(sax));
309	if (fd < 0 )    {
310		return -errno;
311	}
312	return 0;
313
314Internal Implementation
315=======================
316
317The driver keeps a struct l2tp_tunnel context per L2TP tunnel and a
318struct l2tp_session context for each session. The l2tp_tunnel is
319always associated with a UDP or L2TP/IP socket and keeps a list of
320sessions in the tunnel. The l2tp_session context keeps kernel state
321about the session. It has private data which is used for data specific
322to the session type. With L2TPv2, the session always carried PPP
323traffic. With L2TPv3, the session can also carry ethernet frames
324(ethernet pseudowire) or other data types such as ATM, HDLC or Frame
325Relay.
326
327When a tunnel is first opened, the reference count on the socket is
328increased using sock_hold(). This ensures that the kernel socket
329cannot be removed while L2TP's data structures reference it.
330
331Some L2TP sessions also have a socket (PPP pseudowires) while others
332do not (ethernet pseudowires). We can't use the socket reference count
333as the reference count for session contexts. The L2TP implementation
334therefore has its own internal reference counts on the session
335contexts.
336
337To Do
338=====
339
340Add L2TP tunnel switching support. This would route tunneled traffic
341from one L2TP tunnel into another. Specified in
342http://tools.ietf.org/html/draft-ietf-l2tpext-tunnel-switching-08
343
344Add L2TPv3 VLAN pseudowire support.
345
346Add L2TPv3 IP pseudowire support.
347
348Add L2TPv3 ATM pseudowire support.
349
350Miscellaneous
351=============
352
353The L2TP drivers were developed as part of the OpenL2TP project by
354Katalix Systems Ltd. OpenL2TP is a full-featured L2TP client / server,
355designed from the ground up to have the L2TP datapath in the
356kernel. The project also implemented the pppol2tp plugin for pppd
357which allows pppd to use the kernel driver. Details can be found at
358http://www.openl2tp.org.
359