xref: /linux/Documentation/networking/tuntap.rst (revision 3a2c4d55e32ad65efebdb6de44eef3bfa08bb49d)
1.. SPDX-License-Identifier: GPL-2.0
2.. include:: <isonum.txt>
3
4===============================
5Universal TUN/TAP device driver
6===============================
7
8Copyright |copy| 1999-2000 Maxim Krasnyansky <max_mk@yahoo.com>
9
10  Linux, Solaris drivers
11  Copyright |copy| 1999-2000 Maxim Krasnyansky <max_mk@yahoo.com>
12
13  FreeBSD TAP driver
14  Copyright |copy| 1999-2000 Maksim Yevmenkin <m_evmenkin@yahoo.com>
15
16  Revision of this document 2002 by Florian Thiel <florian.thiel@gmx.net>
17
181. Description
19==============
20
21  TUN/TAP provides packet reception and transmission for user space programs.
22  It can be seen as a simple Point-to-Point or Ethernet device, which,
23  instead of receiving packets from physical media, receives them from
24  user space program and instead of sending packets via physical media
25  writes them to the user space program.
26
27  In order to use the driver a program has to open /dev/net/tun and issue a
28  corresponding ioctl() to register a network device with the kernel. A network
29  device will appear as tunXX or tapXX, depending on the options chosen. When
30  the program closes the file descriptor, the network device and all
31  corresponding routes will disappear.
32
33  Depending on the type of device chosen the userspace program has to read/write
34  IP packets (with tun) or ethernet frames (with tap). Which one is being used
35  depends on the flags given with the ioctl().
36
37  The package from http://vtun.sourceforge.net/tun contains two simple examples
38  for how to use tun and tap devices. Both programs work like a bridge between
39  two network interfaces.
40  br_select.c - bridge based on select system call.
41  br_sigio.c  - bridge based on async io and SIGIO signal.
42  However, the best example is VTun http://vtun.sourceforge.net :))
43
442. Configuration
45================
46
47  Create device node::
48
49     mkdir /dev/net (if it doesn't exist already)
50     mknod /dev/net/tun c 10 200
51
52  Set permissions::
53
54     e.g. chmod 0666 /dev/net/tun
55
56  There's no harm in allowing the device to be accessible by non-root users,
57  since CAP_NET_ADMIN is required for creating network devices or for
58  connecting to network devices which aren't owned by the user in question.
59  If you want to create persistent devices and give ownership of them to
60  unprivileged users, then you need the /dev/net/tun device to be usable by
61  those users.
62
63  Driver module autoloading
64
65     Make sure that "Kernel module loader" - module auto-loading
66     support is enabled in your kernel.  The kernel should load it on
67     first access.
68
69  Manual loading
70
71     insert the module by hand::
72
73	modprobe tun
74
75  If you do it the latter way, you have to load the module every time you
76  need it, if you do it the other way it will be automatically loaded when
77  /dev/net/tun is being opened.
78
793. Program interface
80====================
81
823.1 Network device allocation
83-----------------------------
84
85``char *dev`` should be the name of the device with a format string (e.g.
86"tun%d"), but (as far as I can see) this can be any valid network device name.
87Note that the character pointer becomes overwritten with the real device name
88(e.g. "tun0")::
89
90  #include <linux/if.h>
91  #include <linux/if_tun.h>
92
93  int tun_alloc(char *dev)
94  {
95      struct ifreq ifr;
96      int fd, err;
97
98      if( (fd = open("/dev/net/tun", O_RDWR)) < 0 )
99	 return tun_alloc_old(dev);
100
101      memset(&ifr, 0, sizeof(ifr));
102
103      /* Flags: IFF_TUN   - TUN device (no Ethernet headers)
104       *        IFF_TAP   - TAP device
105       *
106       *        IFF_NO_PI - Do not provide packet information
107       */
108      ifr.ifr_flags = IFF_TUN;
109      if( *dev )
110	 strscpy_pad(ifr.ifr_name, dev, IFNAMSIZ);
111
112      if( (err = ioctl(fd, TUNSETIFF, (void *) &ifr)) < 0 ){
113	 close(fd);
114	 return err;
115      }
116      strcpy(dev, ifr.ifr_name);
117      return fd;
118  }
119
1203.2 Frame format
121----------------
122
123If flag IFF_NO_PI is not set each frame format is::
124
125     Flags [2 bytes]
126     Proto [2 bytes]
127     Raw protocol(IP, IPv6, etc) frame.
128
1293.3 Multiqueue tuntap interface
130-------------------------------
131
132From version 3.8, Linux supports multiqueue tuntap which can uses multiple
133file descriptors (queues) to parallelize packets sending or receiving. The
134device allocation is the same as before, and if user wants to create multiple
135queues, TUNSETIFF with the same device name must be called many times with
136IFF_MULTI_QUEUE flag.
137
138``char *dev`` should be the name of the device, queues is the number of queues
139to be created, fds is used to store and return the file descriptors (queues)
140created to the caller. Each file descriptor were served as the interface of a
141queue which could be accessed by userspace.
142
143::
144
145  #include <linux/if.h>
146  #include <linux/if_tun.h>
147
148  int tun_alloc_mq(char *dev, int queues, int *fds)
149  {
150      struct ifreq ifr;
151      int fd, err, i;
152
153      if (!dev)
154	  return -1;
155
156      memset(&ifr, 0, sizeof(ifr));
157      /* Flags: IFF_TUN   - TUN device (no Ethernet headers)
158       *        IFF_TAP   - TAP device
159       *
160       *        IFF_NO_PI - Do not provide packet information
161       *        IFF_MULTI_QUEUE - Create a queue of multiqueue device
162       */
163      ifr.ifr_flags = IFF_TAP | IFF_NO_PI | IFF_MULTI_QUEUE;
164      strcpy(ifr.ifr_name, dev);
165
166      for (i = 0; i < queues; i++) {
167	  if ((fd = open("/dev/net/tun", O_RDWR)) < 0)
168	     goto err;
169	  err = ioctl(fd, TUNSETIFF, (void *)&ifr);
170	  if (err) {
171	     close(fd);
172	     goto err;
173	  }
174	  fds[i] = fd;
175      }
176
177      return 0;
178  err:
179      for (--i; i >= 0; i--)
180	  close(fds[i]);
181      return err;
182  }
183
184A new ioctl(TUNSETQUEUE) were introduced to enable or disable a queue. When
185calling it with IFF_DETACH_QUEUE flag, the queue were disabled. And when
186calling it with IFF_ATTACH_QUEUE flag, the queue were enabled. The queue were
187enabled by default after it was created through TUNSETIFF.
188
189fd is the file descriptor (queue) that we want to enable or disable, when
190enable is true we enable it, otherwise we disable it::
191
192  #include <linux/if.h>
193  #include <linux/if_tun.h>
194
195  int tun_set_queue(int fd, int enable)
196  {
197      struct ifreq ifr;
198
199      memset(&ifr, 0, sizeof(ifr));
200
201      if (enable)
202	 ifr.ifr_flags = IFF_ATTACH_QUEUE;
203      else
204	 ifr.ifr_flags = IFF_DETACH_QUEUE;
205
206      return ioctl(fd, TUNSETQUEUE, (void *)&ifr);
207  }
208
2093.4 qdisc backpressure
210----------------------
211
212IFF_BACKPRESSURE can be set to enable qdisc backpressure. Without it, TX
213drops occur when the internal ring buffer is full, so any attached qdisc
214is effectively bypassed and applications only learn about congestion
215through those drops.
216
217With it, the kernel stops the queue instead, letting the qdisc hold and
218schedule packets, so its AQM, shaping and fairness actually apply. This
219helps protocols like TCP, which cut throughput in reaction to packet
220drops. With IFF_BACKPRESSURE, drops then only occur as a rare race.
221Backpressure requires a qdisc to be attached and has no effect with
222noqueue.
223
224The flag is a property of the TUN/TAP device rather than of the file
225descriptor it was set on, so it applies to all queues of the device,
226regardless of which process opened which queue.
227
228The flag can only be changed while the device has at most one queue. On a
229multiqueue device that already has a second queue attached or detached, a
230later TUNSETIFF succeeds but leaves the flag as it is. All the other
231TUNSETIFF flags behave the same way.
232
233The txqueuelen can be reduced alongside this flag to further shift
234buffering into the qdisc and reduce bufferbloat, at a possible
235performance cost.
236
237When running multiple network streams in parallel through a single
238TUN/TAP queue, the flag may reduce performance due to the extra overhead
239of the backpressure mechanism.
240
241Universal TUN/TAP device driver Frequently Asked Question
242=========================================================
243
2441. What platforms are supported by TUN/TAP driver ?
245
246Currently driver has been written for 3 Unices:
247
248  - Linux kernels 2.2.x, 2.4.x
249  - FreeBSD 3.x, 4.x, 5.x
250  - Solaris 2.6, 7.0, 8.0
251
2522. What is TUN/TAP driver used for?
253
254As mentioned above, main purpose of TUN/TAP driver is tunneling.
255It is used by VTun (http://vtun.sourceforge.net).
256
257Another interesting application using TUN/TAP is pipsecd
258(http://perso.enst.fr/~beyssac/pipsec/), a userspace IPSec
259implementation that can use complete kernel routing (unlike FreeS/WAN).
260
2613. How does Virtual network device actually work ?
262
263Virtual network device can be viewed as a simple Point-to-Point or
264Ethernet device, which instead of receiving packets from a physical
265media, receives them from user space program and instead of sending
266packets via physical media sends them to the user space program.
267
268Let's say that you configured IPv6 on the tap0, then whenever
269the kernel sends an IPv6 packet to tap0, it is passed to the application
270(VTun for example). The application encrypts, compresses and sends it to
271the other side over TCP or UDP. The application on the other side decompresses
272and decrypts the data received and writes the packet to the TAP device,
273the kernel handles the packet like it came from real physical device.
274
2754. What is the difference between TUN driver and TAP driver?
276
277TUN works with IP frames. TAP works with Ethernet frames.
278
279This means that you have to read/write IP packets when you are using tun and
280ethernet frames when using tap.
281
2825. What is the difference between BPF and TUN/TAP driver?
283
284BPF is an advanced packet filter. It can be attached to existing
285network interface. It does not provide a virtual network interface.
286A TUN/TAP driver does provide a virtual network interface and it is possible
287to attach BPF to this interface.
288
2896. Does TAP driver support kernel Ethernet bridging?
290
291Yes. Linux and FreeBSD drivers support Ethernet bridging.
292