blob: 7bea9c65119e339326155c45e4af6e268bda0c6c [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
16 */
17
18/*
19 * Changes:
20 *
Mike Kershawff4cc3a2005-09-01 17:40:05 -070021 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
23 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 * Mark Smith <markzzzsmith@yahoo.com.au>
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -070025 * Use random_ether_addr() for tap MAC address.
Linus Torvalds1da177e2005-04-16 15:20:36 -070026 *
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
30 * Added ethtool API.
31 * Minor cleanups
32 *
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
35 */
36
Joe Perches6b8a66e2011-03-02 07:18:10 +000037#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#define DRV_NAME "tun"
40#define DRV_VERSION "1.6"
41#define DRV_DESCRIPTION "Universal TUN/TAP device driver"
42#define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/module.h>
45#include <linux/errno.h>
46#include <linux/kernel.h>
47#include <linux/major.h>
48#include <linux/slab.h>
49#include <linux/poll.h>
50#include <linux/fcntl.h>
51#include <linux/init.h>
52#include <linux/skbuff.h>
53#include <linux/netdevice.h>
54#include <linux/etherdevice.h>
55#include <linux/miscdevice.h>
56#include <linux/ethtool.h>
57#include <linux/rtnetlink.h>
Arnd Bergmann50857e22009-11-06 22:52:32 -080058#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/if.h>
60#include <linux/if_arp.h>
61#include <linux/if_ether.h>
62#include <linux/if_tun.h>
63#include <linux/crc32.h>
Pavel Emelyanovd647a592008-04-16 00:41:16 -070064#include <linux/nsproxy.h>
Rusty Russellf43798c2008-07-03 03:48:02 -070065#include <linux/virtio_net.h>
Michael S. Tsirkin99405162010-02-14 01:01:10 +000066#include <linux/rcupdate.h>
Eric W. Biederman881d9662007-09-17 11:56:21 -070067#include <net/net_namespace.h>
Pavel Emelyanov79d17602008-04-16 00:40:46 -070068#include <net/netns/generic.h>
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -080069#include <net/rtnetlink.h>
Herbert Xu33dccbb2009-02-05 21:25:32 -080070#include <net/sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071
72#include <asm/system.h>
73#include <asm/uaccess.h>
74
Rusty Russell14daa022008-04-12 18:48:58 -070075/* Uncomment to enable debugging */
76/* #define TUN_DEBUG 1 */
77
Linus Torvalds1da177e2005-04-16 15:20:36 -070078#ifdef TUN_DEBUG
79static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070080
Joe Perches6b8a66e2011-03-02 07:18:10 +000081#define tun_debug(level, tun, fmt, args...) \
82do { \
83 if (tun->debug) \
84 netdev_printk(level, tun->dev, fmt, ##args); \
85} while (0)
86#define DBG1(level, fmt, args...) \
87do { \
88 if (debug == 2) \
89 printk(level fmt, ##args); \
90} while (0)
Rusty Russell14daa022008-04-12 18:48:58 -070091#else
Joe Perches6b8a66e2011-03-02 07:18:10 +000092#define tun_debug(level, tun, fmt, args...) \
93do { \
94 if (0) \
95 netdev_printk(level, tun->dev, fmt, ##args); \
96} while (0)
97#define DBG1(level, fmt, args...) \
98do { \
99 if (0) \
100 printk(level fmt, ##args); \
101} while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#endif
103
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700104#define FLT_EXACT_COUNT 8
105struct tap_filter {
106 unsigned int count; /* Number of addrs. Zero means disabled */
107 u32 mask[2]; /* Mask of the hashed addrs */
108 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
109};
110
Eric W. Biederman631ab462009-01-20 11:00:40 +0000111struct tun_file {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000112 atomic_t count;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000113 struct tun_struct *tun;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +0000114 struct net *net;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000115};
116
Herbert Xu33dccbb2009-02-05 21:25:32 -0800117struct tun_sock;
118
Rusty Russell14daa022008-04-12 18:48:58 -0700119struct tun_struct {
Eric W. Biederman631ab462009-01-20 11:00:40 +0000120 struct tun_file *tfile;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700121 unsigned int flags;
Rusty Russell14daa022008-04-12 18:48:58 -0700122 uid_t owner;
123 gid_t group;
124
Rusty Russell14daa022008-04-12 18:48:58 -0700125 struct net_device *dev;
Michał Mirosław88255372011-04-19 06:13:10 +0000126 u32 set_features;
127#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
128 NETIF_F_TSO6|NETIF_F_UFO)
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700129 struct fasync_struct *fasync;
Rusty Russell14daa022008-04-12 18:48:58 -0700130
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700131 struct tap_filter txflt;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800132 struct socket socket;
Eric Dumazet43815482010-04-29 11:01:49 +0000133 struct socket_wq wq;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200134
135 int vnet_hdr_sz;
136
Rusty Russell14daa022008-04-12 18:48:58 -0700137#ifdef TUN_DEBUG
138 int debug;
139#endif
140};
141
Herbert Xu33dccbb2009-02-05 21:25:32 -0800142struct tun_sock {
143 struct sock sk;
144 struct tun_struct *tun;
145};
146
147static inline struct tun_sock *tun_sk(struct sock *sk)
148{
149 return container_of(sk, struct tun_sock, sk);
150}
151
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000152static int tun_attach(struct tun_struct *tun, struct file *file)
153{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000154 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000155 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000156
157 ASSERT_RTNL();
158
Eric W. Biederman38231b72009-01-20 11:02:28 +0000159 netif_tx_lock_bh(tun->dev);
160
161 err = -EINVAL;
162 if (tfile->tun)
163 goto out;
164
165 err = -EBUSY;
166 if (tun->tfile)
167 goto out;
168
169 err = 0;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000170 tfile->tun = tun;
171 tun->tfile = tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000172 tun->socket.file = file;
Nolan Leakebee31362010-07-27 13:53:43 +0000173 netif_carrier_on(tun->dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000174 dev_hold(tun->dev);
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000175 sock_hold(tun->socket.sk);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000176 atomic_inc(&tfile->count);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000177
Eric W. Biederman38231b72009-01-20 11:02:28 +0000178out:
179 netif_tx_unlock_bh(tun->dev);
180 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000181}
182
Eric W. Biederman631ab462009-01-20 11:00:40 +0000183static void __tun_detach(struct tun_struct *tun)
184{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000185 /* Detach from net device */
Eric W. Biederman38231b72009-01-20 11:02:28 +0000186 netif_tx_lock_bh(tun->dev);
Nolan Leakebee31362010-07-27 13:53:43 +0000187 netif_carrier_off(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000188 tun->tfile = NULL;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000189 tun->socket.file = NULL;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000190 netif_tx_unlock_bh(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000191
192 /* Drop read queue */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000193 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000194
195 /* Drop the extra count on the net device */
196 dev_put(tun->dev);
197}
198
199static void tun_detach(struct tun_struct *tun)
200{
201 rtnl_lock();
202 __tun_detach(tun);
203 rtnl_unlock();
Eric W. Biederman631ab462009-01-20 11:00:40 +0000204}
205
206static struct tun_struct *__tun_get(struct tun_file *tfile)
207{
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000208 struct tun_struct *tun = NULL;
209
210 if (atomic_inc_not_zero(&tfile->count))
211 tun = tfile->tun;
212
213 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000214}
215
216static struct tun_struct *tun_get(struct file *file)
217{
218 return __tun_get(file->private_data);
219}
220
221static void tun_put(struct tun_struct *tun)
222{
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000223 struct tun_file *tfile = tun->tfile;
224
225 if (atomic_dec_and_test(&tfile->count))
226 tun_detach(tfile->tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000227}
228
Joe Perches6b8a66e2011-03-02 07:18:10 +0000229/* TAP filtering */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700230static void addr_hash_set(u32 *mask, const u8 *addr)
231{
232 int n = ether_crc(ETH_ALEN, addr) >> 26;
233 mask[n >> 5] |= (1 << (n & 31));
234}
235
236static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
237{
238 int n = ether_crc(ETH_ALEN, addr) >> 26;
239 return mask[n >> 5] & (1 << (n & 31));
240}
241
242static int update_filter(struct tap_filter *filter, void __user *arg)
243{
244 struct { u8 u[ETH_ALEN]; } *addr;
245 struct tun_filter uf;
246 int err, alen, n, nexact;
247
248 if (copy_from_user(&uf, arg, sizeof(uf)))
249 return -EFAULT;
250
251 if (!uf.count) {
252 /* Disabled */
253 filter->count = 0;
254 return 0;
255 }
256
257 alen = ETH_ALEN * uf.count;
258 addr = kmalloc(alen, GFP_KERNEL);
259 if (!addr)
260 return -ENOMEM;
261
262 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
263 err = -EFAULT;
264 goto done;
265 }
266
267 /* The filter is updated without holding any locks. Which is
268 * perfectly safe. We disable it first and in the worst
269 * case we'll accept a few undesired packets. */
270 filter->count = 0;
271 wmb();
272
273 /* Use first set of addresses as an exact filter */
274 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
275 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
276
277 nexact = n;
278
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800279 /* Remaining multicast addresses are hashed,
280 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700281 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800282 for (; n < uf.count; n++) {
283 if (!is_multicast_ether_addr(addr[n].u)) {
284 err = 0; /* no filter */
285 goto done;
286 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700287 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800288 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700289
290 /* For ALLMULTI just set the mask to all ones.
291 * This overrides the mask populated above. */
292 if ((uf.flags & TUN_FLT_ALLMULTI))
293 memset(filter->mask, ~0, sizeof(filter->mask));
294
295 /* Now enable the filter */
296 wmb();
297 filter->count = nexact;
298
299 /* Return the number of exact filters */
300 err = nexact;
301
302done:
303 kfree(addr);
304 return err;
305}
306
307/* Returns: 0 - drop, !=0 - accept */
308static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
309{
310 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
311 * at this point. */
312 struct ethhdr *eh = (struct ethhdr *) skb->data;
313 int i;
314
315 /* Exact match */
316 for (i = 0; i < filter->count; i++)
317 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
318 return 1;
319
320 /* Inexact match (multicast only) */
321 if (is_multicast_ether_addr(eh->h_dest))
322 return addr_hash_test(filter->mask, eh->h_dest);
323
324 return 0;
325}
326
327/*
328 * Checks whether the packet is accepted or not.
329 * Returns: 0 - drop, !=0 - accept
330 */
331static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
332{
333 if (!filter->count)
334 return 1;
335
336 return run_filter(filter, skb);
337}
338
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339/* Network device part of the driver */
340
Jeff Garzik7282d492006-09-13 14:30:00 -0400341static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000343/* Net device detach from fd. */
344static void tun_net_uninit(struct net_device *dev)
345{
346 struct tun_struct *tun = netdev_priv(dev);
347 struct tun_file *tfile = tun->tfile;
348
349 /* Inform the methods they need to stop using the dev.
350 */
351 if (tfile) {
Eric Dumazet43815482010-04-29 11:01:49 +0000352 wake_up_all(&tun->wq.wait);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000353 if (atomic_dec_and_test(&tfile->count))
354 __tun_detach(tun);
355 }
356}
357
Herbert Xu9c3fea62009-04-18 14:15:52 +0000358static void tun_free_netdev(struct net_device *dev)
359{
360 struct tun_struct *tun = netdev_priv(dev);
361
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000362 sock_put(tun->socket.sk);
Herbert Xu9c3fea62009-04-18 14:15:52 +0000363}
364
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365/* Net device open. */
366static int tun_net_open(struct net_device *dev)
367{
368 netif_start_queue(dev);
369 return 0;
370}
371
372/* Net device close. */
373static int tun_net_close(struct net_device *dev)
374{
375 netif_stop_queue(dev);
376 return 0;
377}
378
379/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000380static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381{
382 struct tun_struct *tun = netdev_priv(dev);
383
Joe Perches6b8a66e2011-03-02 07:18:10 +0000384 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
386 /* Drop packet if interface is not attached */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000387 if (!tun->tfile)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 goto drop;
389
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700390 /* Drop if the filter does not like it.
391 * This is a noop if the filter is disabled.
392 * Filter can be enabled only for the TAP devices. */
393 if (!check_filter(&tun->txflt, skb))
394 goto drop;
395
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000396 if (tun->socket.sk->sk_filter &&
397 sk_filter(tun->socket.sk, skb))
398 goto drop;
399
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000400 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 if (!(tun->flags & TUN_ONE_QUEUE)) {
402 /* Normal queueing mode. */
403 /* Packet scheduler handles dropping of further packets. */
404 netif_stop_queue(dev);
405
406 /* We won't see all dropped packets individually, so overrun
407 * error is more appropriate. */
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700408 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 } else {
410 /* Single queue mode.
411 * Driver handles dropping of all packets itself. */
412 goto drop;
413 }
414 }
415
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000416 /* Orphan the skb - required as we might hang on to it
417 * for indefinite time. */
418 skb_orphan(skb);
419
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700420 /* Enqueue packet */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000421 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422
423 /* Notify and wake up reader process */
424 if (tun->flags & TUN_FASYNC)
425 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
Eric Dumazet43815482010-04-29 11:01:49 +0000426 wake_up_interruptible_poll(&tun->wq.wait, POLLIN |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000427 POLLRDNORM | POLLRDBAND);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000428 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
430drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700431 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000433 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434}
435
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700436static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700438 /*
439 * This callback is supposed to deal with mc filter in
440 * _rx_ path and has nothing to do with the _tx_ path.
441 * In rx path we always accept everything userspace gives us.
442 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443}
444
Ed Swierk4885a502007-09-16 12:21:38 -0700445#define MIN_MTU 68
446#define MAX_MTU 65535
447
448static int
449tun_net_change_mtu(struct net_device *dev, int new_mtu)
450{
451 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
452 return -EINVAL;
453 dev->mtu = new_mtu;
454 return 0;
455}
456
Michał Mirosław88255372011-04-19 06:13:10 +0000457static u32 tun_net_fix_features(struct net_device *dev, u32 features)
458{
459 struct tun_struct *tun = netdev_priv(dev);
460
461 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
462}
Neil Hormanbebd0972011-06-15 05:25:01 +0000463#ifdef CONFIG_NET_POLL_CONTROLLER
464static void tun_poll_controller(struct net_device *dev)
465{
466 /*
467 * Tun only receives frames when:
468 * 1) the char device endpoint gets data from user space
469 * 2) the tun socket gets a sendmsg call from user space
470 * Since both of those are syncronous operations, we are guaranteed
471 * never to have pending data when we poll for it
472 * so theres nothing to do here but return.
473 * We need this though so netpoll recognizes us as an interface that
474 * supports polling, which enables bridge devices in virt setups to
475 * still use netconsole
476 */
477 return;
478}
479#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800480static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000481 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800482 .ndo_open = tun_net_open,
483 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800484 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800485 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000486 .ndo_fix_features = tun_net_fix_features,
Neil Hormanbebd0972011-06-15 05:25:01 +0000487#ifdef CONFIG_NET_POLL_CONTROLLER
488 .ndo_poll_controller = tun_poll_controller,
489#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800490};
491
492static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000493 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800494 .ndo_open = tun_net_open,
495 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800496 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800497 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000498 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +0000499 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800500 .ndo_set_mac_address = eth_mac_addr,
501 .ndo_validate_addr = eth_validate_addr,
Neil Hormanbebd0972011-06-15 05:25:01 +0000502#ifdef CONFIG_NET_POLL_CONTROLLER
503 .ndo_poll_controller = tun_poll_controller,
504#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800505};
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507/* Initialize net device. */
508static void tun_net_init(struct net_device *dev)
509{
510 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 switch (tun->flags & TUN_TYPE_MASK) {
513 case TUN_TUN_DEV:
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800514 dev->netdev_ops = &tun_netdev_ops;
515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 /* Point-to-Point TUN Device */
517 dev->hard_header_len = 0;
518 dev->addr_len = 0;
519 dev->mtu = 1500;
520
521 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400522 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
524 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
525 break;
526
527 case TUN_TAP_DEV:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -0800528 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700530 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +0000531 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700533 random_ether_addr(dev->dev_addr);
Brian Braunstein36226a82007-04-26 01:00:55 -0700534
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
536 break;
537 }
538}
539
540/* Character device part */
541
542/* Poll */
543static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400544{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +0000545 struct tun_file *tfile = file->private_data;
546 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000547 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800548 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
550 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +0000551 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000553 sk = tun->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000554
Joe Perches6b8a66e2011-03-02 07:18:10 +0000555 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
Eric Dumazet43815482010-04-29 11:01:49 +0000557 poll_wait(file, &tun->wq.wait, wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400558
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000559 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 mask |= POLLIN | POLLRDNORM;
561
Herbert Xu33dccbb2009-02-05 21:25:32 -0800562 if (sock_writeable(sk) ||
563 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
564 sock_writeable(sk)))
565 mask |= POLLOUT | POLLWRNORM;
566
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000567 if (tun->dev->reg_state != NETREG_REGISTERED)
568 mask = POLLERR;
569
Eric W. Biederman631ab462009-01-20 11:00:40 +0000570 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 return mask;
572}
573
Rusty Russellf42157c2008-08-15 15:15:10 -0700574/* prepad is the amount to reserve at front. len is length after that.
575 * linear is a hint as to how much to copy (usually headers). */
stephen hemminger6f7c1562011-06-08 14:33:08 +0000576static struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
577 size_t prepad, size_t len,
578 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -0700579{
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000580 struct sock *sk = tun->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -0700581 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800582 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -0700583
Herbert Xu82862742010-05-24 00:14:10 -0700584 sock_update_classid(sk);
585
Rusty Russellf42157c2008-08-15 15:15:10 -0700586 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -0700587 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800588 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -0700589
Herbert Xu33dccbb2009-02-05 21:25:32 -0800590 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
591 &err);
Rusty Russellf42157c2008-08-15 15:15:10 -0700592 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800593 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -0700594
595 skb_reserve(skb, prepad);
596 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800597 skb->data_len = len - linear;
598 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -0700599
600 return skb;
601}
602
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603/* Get packet from user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +0000604static ssize_t tun_get_user(struct tun_struct *tun,
605 const struct iovec *iv, size_t count,
606 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607{
Harvey Harrison09640e62009-02-01 00:45:17 -0800608 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 struct sk_buff *skb;
stephen hemmingera504b862011-06-08 14:33:07 +0000610 size_t len = count, align = NET_SKB_PAD;
Rusty Russellf43798c2008-07-03 03:48:02 -0700611 struct virtio_net_hdr gso = { 0 };
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000612 int offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
614 if (!(tun->flags & TUN_NO_PI)) {
615 if ((len -= sizeof(pi)) > count)
616 return -EINVAL;
617
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000618 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 return -EFAULT;
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000620 offset += sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 }
622
Rusty Russellf43798c2008-07-03 03:48:02 -0700623 if (tun->flags & TUN_VNET_HDR) {
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200624 if ((len -= tun->vnet_hdr_sz) > count)
Rusty Russellf43798c2008-07-03 03:48:02 -0700625 return -EINVAL;
626
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000627 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
Rusty Russellf43798c2008-07-03 03:48:02 -0700628 return -EFAULT;
629
Herbert Xu49091222009-06-08 00:20:01 -0700630 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
631 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
632 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
633
Rusty Russellf43798c2008-07-03 03:48:02 -0700634 if (gso.hdr_len > len)
635 return -EINVAL;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200636 offset += tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -0700637 }
638
Rusty Russelle01bf1c2008-04-12 18:49:30 -0700639 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
stephen hemmingera504b862011-06-08 14:33:07 +0000640 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -0700641 if (unlikely(len < ETH_HLEN ||
642 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -0700643 return -EINVAL;
644 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400645
Herbert Xu33dccbb2009-02-05 21:25:32 -0800646 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
647 if (IS_ERR(skb)) {
648 if (PTR_ERR(skb) != -EAGAIN)
649 tun->dev->stats.rx_dropped++;
650 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 }
652
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000653 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700654 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -0800655 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -0800657 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Rusty Russellf43798c2008-07-03 03:48:02 -0700659 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
660 if (!skb_partial_csum_set(skb, gso.csum_start,
661 gso.csum_offset)) {
662 tun->dev->stats.rx_frame_errors++;
663 kfree_skb(skb);
664 return -EINVAL;
665 }
Michał Mirosław88255372011-04-19 06:13:10 +0000666 }
Rusty Russellf43798c2008-07-03 03:48:02 -0700667
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 switch (tun->flags & TUN_TYPE_MASK) {
669 case TUN_TUN_DEV:
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -0700670 if (tun->flags & TUN_NO_PI) {
671 switch (skb->data[0] & 0xf0) {
672 case 0x40:
673 pi.proto = htons(ETH_P_IP);
674 break;
675 case 0x60:
676 pi.proto = htons(ETH_P_IPV6);
677 break;
678 default:
679 tun->dev->stats.rx_dropped++;
680 kfree_skb(skb);
681 return -EINVAL;
682 }
683 }
684
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -0700685 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -0700687 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 break;
689 case TUN_TAP_DEV:
690 skb->protocol = eth_type_trans(skb, tun->dev);
691 break;
Joe Perches6403eab2011-06-03 11:51:20 +0000692 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
Rusty Russellf43798c2008-07-03 03:48:02 -0700694 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
695 pr_debug("GSO!\n");
696 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
697 case VIRTIO_NET_HDR_GSO_TCPV4:
698 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
699 break;
700 case VIRTIO_NET_HDR_GSO_TCPV6:
701 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
702 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +0000703 case VIRTIO_NET_HDR_GSO_UDP:
704 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
705 break;
Rusty Russellf43798c2008-07-03 03:48:02 -0700706 default:
707 tun->dev->stats.rx_frame_errors++;
708 kfree_skb(skb);
709 return -EINVAL;
710 }
711
712 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
713 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
714
715 skb_shinfo(skb)->gso_size = gso.gso_size;
716 if (skb_shinfo(skb)->gso_size == 0) {
717 tun->dev->stats.rx_frame_errors++;
718 kfree_skb(skb);
719 return -EINVAL;
720 }
721
722 /* Header must be checked, and gso_segs computed. */
723 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
724 skb_shinfo(skb)->gso_segs = 0;
725 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400726
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400728
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700729 tun->dev->stats.rx_packets++;
730 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
732 return count;
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400733}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700735static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
736 unsigned long count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737{
Herbert Xu33dccbb2009-02-05 21:25:32 -0800738 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -0800739 struct tun_struct *tun = tun_get(file);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000740 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
742 if (!tun)
743 return -EBADFD;
744
Joe Perches6b8a66e2011-03-02 07:18:10 +0000745 tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000747 result = tun_get_user(tun, iv, iov_length(iv, count),
Herbert Xu33dccbb2009-02-05 21:25:32 -0800748 file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000749
750 tun_put(tun);
751 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752}
753
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +0000755static ssize_t tun_put_user(struct tun_struct *tun,
756 struct sk_buff *skb,
757 const struct iovec *iv, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758{
759 struct tun_pi pi = { 0, skb->protocol };
760 ssize_t total = 0;
761
762 if (!(tun->flags & TUN_NO_PI)) {
763 if ((len -= sizeof(pi)) < 0)
764 return -EINVAL;
765
766 if (len < skb->len) {
767 /* Packet will be striped */
768 pi.flags |= TUN_PKT_STRIP;
769 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400770
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000771 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 return -EFAULT;
773 total += sizeof(pi);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400774 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Rusty Russellf43798c2008-07-03 03:48:02 -0700776 if (tun->flags & TUN_VNET_HDR) {
777 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200778 if ((len -= tun->vnet_hdr_sz) < 0)
Rusty Russellf43798c2008-07-03 03:48:02 -0700779 return -EINVAL;
780
781 if (skb_is_gso(skb)) {
782 struct skb_shared_info *sinfo = skb_shinfo(skb);
783
784 /* This is a hint as to how much should be linear. */
785 gso.hdr_len = skb_headlen(skb);
786 gso.gso_size = sinfo->gso_size;
787 if (sinfo->gso_type & SKB_GSO_TCPV4)
788 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
789 else if (sinfo->gso_type & SKB_GSO_TCPV6)
790 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Sridhar Samudralae36aa252009-07-14 14:21:04 +0000791 else if (sinfo->gso_type & SKB_GSO_UDP)
792 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +0000793 else {
Joe Perches6b8a66e2011-03-02 07:18:10 +0000794 pr_err("unexpected GSO type: "
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +0000795 "0x%x, gso_size %d, hdr_len %d\n",
796 sinfo->gso_type, gso.gso_size,
797 gso.hdr_len);
798 print_hex_dump(KERN_ERR, "tun: ",
799 DUMP_PREFIX_NONE,
800 16, 1, skb->head,
801 min((int)gso.hdr_len, 64), true);
802 WARN_ON_ONCE(1);
803 return -EINVAL;
804 }
Rusty Russellf43798c2008-07-03 03:48:02 -0700805 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
806 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
807 } else
808 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
809
810 if (skb->ip_summed == CHECKSUM_PARTIAL) {
811 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
Michał Mirosław55508d62010-12-14 15:24:08 +0000812 gso.csum_start = skb_checksum_start_offset(skb);
Rusty Russellf43798c2008-07-03 03:48:02 -0700813 gso.csum_offset = skb->csum_offset;
Jason Wang10a8d942011-06-10 00:56:17 +0000814 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
815 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
Rusty Russellf43798c2008-07-03 03:48:02 -0700816 } /* else everything is zero */
817
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000818 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
819 sizeof(gso))))
Rusty Russellf43798c2008-07-03 03:48:02 -0700820 return -EFAULT;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200821 total += tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -0700822 }
823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 len = min_t(int, skb->len, len);
825
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000826 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000827 total += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700829 tun->dev->stats.tx_packets++;
830 tun->dev->stats.tx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
832 return total;
833}
834
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000835static ssize_t tun_do_read(struct tun_struct *tun,
836 struct kiocb *iocb, const struct iovec *iv,
837 ssize_t len, int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 DECLARE_WAITQUEUE(wait, current);
840 struct sk_buff *skb;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000841 ssize_t ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842
Joe Perches6b8a66e2011-03-02 07:18:10 +0000843 tun_debug(KERN_INFO, tun, "tun_chr_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844
Amos Kong61a5ff12011-06-09 00:27:10 -0700845 if (unlikely(!noblock))
846 add_wait_queue(&tun->wq.wait, &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 while (len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 current->state = TASK_INTERRUPTIBLE;
849
850 /* Read frames from the queue */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000851 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000852 if (noblock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 ret = -EAGAIN;
854 break;
855 }
856 if (signal_pending(current)) {
857 ret = -ERESTARTSYS;
858 break;
859 }
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000860 if (tun->dev->reg_state != NETREG_REGISTERED) {
861 ret = -EIO;
862 break;
863 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865 /* Nothing to read, let's sleep */
866 schedule();
867 continue;
868 }
869 netif_wake_queue(tun->dev);
870
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000871 ret = tun_put_user(tun, skb, iv, len);
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700872 kfree_skb(skb);
873 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 }
875
876 current->state = TASK_RUNNING;
Amos Kong61a5ff12011-06-09 00:27:10 -0700877 if (unlikely(!noblock))
878 remove_wait_queue(&tun->wq.wait, &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000880 return ret;
881}
882
883static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
884 unsigned long count, loff_t pos)
885{
886 struct file *file = iocb->ki_filp;
887 struct tun_file *tfile = file->private_data;
888 struct tun_struct *tun = __tun_get(tfile);
889 ssize_t len, ret;
890
891 if (!tun)
892 return -EBADFD;
893 len = iov_length(iv, count);
894 if (len < 0) {
895 ret = -EINVAL;
896 goto out;
897 }
898
899 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
900 ret = min_t(ssize_t, ret, len);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000901out:
902 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 return ret;
904}
905
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906static void tun_setup(struct net_device *dev)
907{
908 struct tun_struct *tun = netdev_priv(dev);
909
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 tun->owner = -1;
Guido Guenther8c644622007-07-02 22:50:25 -0700911 tun->group = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 dev->ethtool_ops = &tun_ethtool_ops;
Herbert Xu9c3fea62009-04-18 14:15:52 +0000914 dev->destructor = tun_free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -0800917/* Trivial set of netlink ops to allow deleting tun or tap
918 * device with netlink.
919 */
920static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
921{
922 return -EINVAL;
923}
924
925static struct rtnl_link_ops tun_link_ops __read_mostly = {
926 .kind = DRV_NAME,
927 .priv_size = sizeof(struct tun_struct),
928 .setup = tun_setup,
929 .validate = tun_validate,
930};
931
Herbert Xu33dccbb2009-02-05 21:25:32 -0800932static void tun_sock_write_space(struct sock *sk)
933{
934 struct tun_struct *tun;
Eric Dumazet43815482010-04-29 11:01:49 +0000935 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800936
937 if (!sock_writeable(sk))
938 return;
939
Herbert Xu33dccbb2009-02-05 21:25:32 -0800940 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
941 return;
942
Eric Dumazet43815482010-04-29 11:01:49 +0000943 wqueue = sk_sleep(sk);
944 if (wqueue && waitqueue_active(wqueue))
945 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000946 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -0700947
Vitaliy Gusev80924e52009-12-25 07:17:43 +0000948 tun = tun_sk(sk)->tun;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800949 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
950}
951
952static void tun_sock_destruct(struct sock *sk)
953{
Vitaliy Gusev80924e52009-12-25 07:17:43 +0000954 free_netdev(tun_sk(sk)->tun->dev);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800955}
956
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000957static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
958 struct msghdr *m, size_t total_len)
959{
960 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
961 return tun_get_user(tun, m->msg_iov, total_len,
962 m->msg_flags & MSG_DONTWAIT);
963}
964
965static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
966 struct msghdr *m, size_t total_len,
967 int flags)
968{
969 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
970 int ret;
971 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
972 return -EINVAL;
973 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
974 flags & MSG_DONTWAIT);
975 if (ret > total_len) {
976 m->msg_flags |= MSG_TRUNC;
977 ret = flags & MSG_TRUNC ? ret : total_len;
978 }
979 return ret;
980}
981
982/* Ops structure to mimic raw sockets with tun */
983static const struct proto_ops tun_socket_ops = {
984 .sendmsg = tun_sendmsg,
985 .recvmsg = tun_recvmsg,
986};
987
Herbert Xu33dccbb2009-02-05 21:25:32 -0800988static struct proto tun_proto = {
989 .name = "tun",
990 .owner = THIS_MODULE,
991 .obj_size = sizeof(struct tun_sock),
992};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -0800993
David Woodhouse980c9e82009-05-09 22:54:21 -0700994static int tun_flags(struct tun_struct *tun)
995{
996 int flags = 0;
997
998 if (tun->flags & TUN_TUN_DEV)
999 flags |= IFF_TUN;
1000 else
1001 flags |= IFF_TAP;
1002
1003 if (tun->flags & TUN_NO_PI)
1004 flags |= IFF_NO_PI;
1005
1006 if (tun->flags & TUN_ONE_QUEUE)
1007 flags |= IFF_ONE_QUEUE;
1008
1009 if (tun->flags & TUN_VNET_HDR)
1010 flags |= IFF_VNET_HDR;
1011
1012 return flags;
1013}
1014
1015static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1016 char *buf)
1017{
1018 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1019 return sprintf(buf, "0x%x\n", tun_flags(tun));
1020}
1021
1022static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1023 char *buf)
1024{
1025 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1026 return sprintf(buf, "%d\n", tun->owner);
1027}
1028
1029static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1030 char *buf)
1031{
1032 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1033 return sprintf(buf, "%d\n", tun->group);
1034}
1035
1036static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1037static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1038static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1039
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001040static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001042 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 struct tun_struct *tun;
1044 struct net_device *dev;
1045 int err;
1046
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001047 dev = __dev_get_by_name(net, ifr->ifr_name);
1048 if (dev) {
Paul Moore2b980db2009-08-28 18:12:43 -04001049 const struct cred *cred = current_cred();
1050
David Woodhousef85ba782009-04-27 03:23:54 -07001051 if (ifr->ifr_flags & IFF_TUN_EXCL)
1052 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001053 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1054 tun = netdev_priv(dev);
1055 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1056 tun = netdev_priv(dev);
1057 else
1058 return -EINVAL;
1059
Paul Moore2b980db2009-08-28 18:12:43 -04001060 if (((tun->owner != -1 && cred->euid != tun->owner) ||
1061 (tun->group != -1 && !in_egroup_p(tun->group))) &&
1062 !capable(CAP_NET_ADMIN))
1063 return -EPERM;
Linus Torvaldsd7e96602009-09-14 10:37:28 -07001064 err = security_tun_dev_attach(tun->socket.sk);
Paul Moore2b980db2009-08-28 18:12:43 -04001065 if (err < 0)
1066 return err;
1067
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001068 err = tun_attach(tun, file);
1069 if (err < 0)
1070 return err;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001071 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 else {
1073 char *name;
1074 unsigned long flags = 0;
1075
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001076 if (!capable(CAP_NET_ADMIN))
1077 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001078 err = security_tun_dev_create();
1079 if (err < 0)
1080 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001081
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 /* Set dev type */
1083 if (ifr->ifr_flags & IFF_TUN) {
1084 /* TUN device */
1085 flags |= TUN_TUN_DEV;
1086 name = "tun%d";
1087 } else if (ifr->ifr_flags & IFF_TAP) {
1088 /* TAP device */
1089 flags |= TUN_TAP_DEV;
1090 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001091 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001092 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001093
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 if (*ifr->ifr_name)
1095 name = ifr->ifr_name;
1096
1097 dev = alloc_netdev(sizeof(struct tun_struct), name,
1098 tun_setup);
1099 if (!dev)
1100 return -ENOMEM;
1101
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001102 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001103 dev->rtnl_link_ops = &tun_link_ops;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001104
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 tun = netdev_priv(dev);
1106 tun->dev = dev;
1107 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001108 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001109 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
Herbert Xu33dccbb2009-02-05 21:25:32 -08001111 err = -ENOMEM;
1112 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
1113 if (!sk)
1114 goto err_free_dev;
1115
Eric Dumazet43815482010-04-29 11:01:49 +00001116 tun->socket.wq = &tun->wq;
1117 init_waitqueue_head(&tun->wq.wait);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001118 tun->socket.ops = &tun_socket_ops;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001119 sock_init_data(&tun->socket, sk);
1120 sk->sk_write_space = tun_sock_write_space;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001121 sk->sk_sndbuf = INT_MAX;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001122
Vitaliy Gusev80924e52009-12-25 07:17:43 +00001123 tun_sk(sk)->tun = tun;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001124
Paul Moore2b980db2009-08-28 18:12:43 -04001125 security_tun_dev_post_create(sk);
1126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 tun_net_init(dev);
1128
Michał Mirosław88255372011-04-19 06:13:10 +00001129 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
1130 TUN_USER_FEATURES;
1131 dev->features = dev->hw_features;
1132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 err = register_netdevice(tun->dev);
1134 if (err < 0)
Herbert Xu9c3fea62009-04-18 14:15:52 +00001135 goto err_free_sk;
1136
David Woodhouse980c9e82009-05-09 22:54:21 -07001137 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1138 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1139 device_create_file(&tun->dev->dev, &dev_attr_group))
Joe Perches6b8a66e2011-03-02 07:18:10 +00001140 pr_err("Failed to create tun sysfs files\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001141
Herbert Xu9c3fea62009-04-18 14:15:52 +00001142 sk->sk_destruct = tun_sock_destruct;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001143
1144 err = tun_attach(tun, file);
1145 if (err < 0)
Herbert Xu9c3fea62009-04-18 14:15:52 +00001146 goto failed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 }
1148
Joe Perches6b8a66e2011-03-02 07:18:10 +00001149 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
1151 if (ifr->ifr_flags & IFF_NO_PI)
1152 tun->flags |= TUN_NO_PI;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001153 else
1154 tun->flags &= ~TUN_NO_PI;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
1156 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1157 tun->flags |= TUN_ONE_QUEUE;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001158 else
1159 tun->flags &= ~TUN_ONE_QUEUE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160
Rusty Russellf43798c2008-07-03 03:48:02 -07001161 if (ifr->ifr_flags & IFF_VNET_HDR)
1162 tun->flags |= TUN_VNET_HDR;
1163 else
1164 tun->flags &= ~TUN_VNET_HDR;
1165
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001166 /* Make sure persistent devices do not get stuck in
1167 * xoff state.
1168 */
1169 if (netif_running(tun->dev))
1170 netif_wake_queue(tun->dev);
1171
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 strcpy(ifr->ifr_name, tun->dev->name);
1173 return 0;
1174
Herbert Xu33dccbb2009-02-05 21:25:32 -08001175 err_free_sk:
1176 sock_put(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 err_free_dev:
1178 free_netdev(dev);
1179 failed:
1180 return err;
1181}
1182
Herbert Xu876bfd42009-08-06 14:22:44 +00001183static int tun_get_iff(struct net *net, struct tun_struct *tun,
1184 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001185{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001186 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001187
1188 strcpy(ifr->ifr_name, tun->dev->name);
1189
David Woodhouse980c9e82009-05-09 22:54:21 -07001190 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001191
1192 return 0;
1193}
1194
Rusty Russell5228ddc2008-07-03 03:46:16 -07001195/* This is like a cut-down ethtool ops, except done via tun fd so no
1196 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001197static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001198{
Michał Mirosław88255372011-04-19 06:13:10 +00001199 u32 features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001200
1201 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001202 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001203 arg &= ~TUN_F_CSUM;
1204
1205 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1206 if (arg & TUN_F_TSO_ECN) {
1207 features |= NETIF_F_TSO_ECN;
1208 arg &= ~TUN_F_TSO_ECN;
1209 }
1210 if (arg & TUN_F_TSO4)
1211 features |= NETIF_F_TSO;
1212 if (arg & TUN_F_TSO6)
1213 features |= NETIF_F_TSO6;
1214 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1215 }
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001216
1217 if (arg & TUN_F_UFO) {
1218 features |= NETIF_F_UFO;
1219 arg &= ~TUN_F_UFO;
1220 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001221 }
1222
1223 /* This gives the user a way to test for new features in future by
1224 * trying to set them. */
1225 if (arg)
1226 return -EINVAL;
1227
Michał Mirosław88255372011-04-19 06:13:10 +00001228 tun->set_features = features;
1229 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001230
1231 return 0;
1232}
1233
Arnd Bergmann50857e22009-11-06 22:52:32 -08001234static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1235 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001237 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001238 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 void __user* argp = (void __user*)arg;
Michael S. Tsirkin99405162010-02-14 01:01:10 +00001240 struct sock_fprog fprog;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 struct ifreq ifr;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001242 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001243 int vnet_hdr_sz;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001244 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245
1246 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
Arnd Bergmann50857e22009-11-06 22:52:32 -08001247 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 return -EFAULT;
1249
Eric W. Biederman631ab462009-01-20 11:00:40 +00001250 if (cmd == TUNGETFEATURES) {
1251 /* Currently this just means: "what IFF flags are valid?".
1252 * This is needed because we never checked for invalid flags on
1253 * TUNSETIFF. */
1254 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1255 IFF_VNET_HDR,
1256 (unsigned int __user*)argp);
1257 }
1258
Herbert Xu876bfd42009-08-06 14:22:44 +00001259 rtnl_lock();
1260
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001261 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1264
Herbert Xu876bfd42009-08-06 14:22:44 +00001265 ret = tun_set_iff(tfile->net, file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Herbert Xu876bfd42009-08-06 14:22:44 +00001267 if (ret)
1268 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269
Arnd Bergmann50857e22009-11-06 22:52:32 -08001270 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001271 ret = -EFAULT;
1272 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 }
1274
Herbert Xu876bfd42009-08-06 14:22:44 +00001275 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001277 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278
Joe Perches6b8a66e2011-03-02 07:18:10 +00001279 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %d\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
Eric W. Biederman631ab462009-01-20 11:00:40 +00001281 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001283 case TUNGETIFF:
Herbert Xu876bfd42009-08-06 14:22:44 +00001284 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001285 if (ret)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001286 break;
Mark McLoughline3b99552008-08-15 15:09:56 -07001287
Arnd Bergmann50857e22009-11-06 22:52:32 -08001288 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001289 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001290 break;
1291
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 case TUNSETNOCSUM:
1293 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294
Michał Mirosław88255372011-04-19 06:13:10 +00001295 /* [unimplemented] */
1296 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00001297 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 break;
1299
1300 case TUNSETPERSIST:
1301 /* Disable/Enable persist mode */
1302 if (arg)
1303 tun->flags |= TUN_PERSIST;
1304 else
1305 tun->flags &= ~TUN_PERSIST;
1306
Joe Perches6b8a66e2011-03-02 07:18:10 +00001307 tun_debug(KERN_INFO, tun, "persist %s\n",
1308 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 break;
1310
1311 case TUNSETOWNER:
1312 /* Set owner of the device */
1313 tun->owner = (uid_t) arg;
1314
Joe Perches6b8a66e2011-03-02 07:18:10 +00001315 tun_debug(KERN_INFO, tun, "owner set to %d\n", tun->owner);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 break;
1317
Guido Guenther8c644622007-07-02 22:50:25 -07001318 case TUNSETGROUP:
1319 /* Set group of the device */
1320 tun->group= (gid_t) arg;
1321
Joe Perches6b8a66e2011-03-02 07:18:10 +00001322 tun_debug(KERN_INFO, tun, "group set to %d\n", tun->group);
Guido Guenther8c644622007-07-02 22:50:25 -07001323 break;
1324
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001325 case TUNSETLINK:
1326 /* Only allow setting the type when the interface is down */
1327 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001328 tun_debug(KERN_INFO, tun,
1329 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07001330 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001331 } else {
1332 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00001333 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1334 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07001335 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001336 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001337 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001338
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339#ifdef TUN_DEBUG
1340 case TUNSETDEBUG:
1341 tun->debug = arg;
1342 break;
1343#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07001344 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00001345 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001346 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001347
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001348 case TUNSETTXFILTER:
1349 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00001350 ret = -EINVAL;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001351 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001352 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07001353 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001354 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
1356 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04001357 /* Get hw address */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001358 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1359 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08001360 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001361 ret = -EFAULT;
1362 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
1364 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001365 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00001366 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
1367 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08001368
Kim B. Heino40102372008-02-29 12:26:21 -08001369 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001370 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001371
1372 case TUNGETSNDBUF:
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001373 sndbuf = tun->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001374 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1375 ret = -EFAULT;
1376 break;
1377
1378 case TUNSETSNDBUF:
1379 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1380 ret = -EFAULT;
1381 break;
1382 }
1383
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001384 tun->socket.sk->sk_sndbuf = sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001385 break;
1386
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001387 case TUNGETVNETHDRSZ:
1388 vnet_hdr_sz = tun->vnet_hdr_sz;
1389 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1390 ret = -EFAULT;
1391 break;
1392
1393 case TUNSETVNETHDRSZ:
1394 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1395 ret = -EFAULT;
1396 break;
1397 }
1398 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1399 ret = -EINVAL;
1400 break;
1401 }
1402
1403 tun->vnet_hdr_sz = vnet_hdr_sz;
1404 break;
1405
Michael S. Tsirkin99405162010-02-14 01:01:10 +00001406 case TUNATTACHFILTER:
1407 /* Can be set only for TAPs */
1408 ret = -EINVAL;
1409 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1410 break;
1411 ret = -EFAULT;
1412 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1413 break;
1414
1415 ret = sk_attach_filter(&fprog, tun->socket.sk);
1416 break;
1417
1418 case TUNDETACHFILTER:
1419 /* Can be set only for TAPs */
1420 ret = -EINVAL;
1421 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1422 break;
1423 ret = sk_detach_filter(tun->socket.sk);
1424 break;
1425
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00001427 ret = -EINVAL;
1428 break;
Joe Perchesee289b62010-05-17 22:47:34 -07001429 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
Herbert Xu876bfd42009-08-06 14:22:44 +00001431unlock:
1432 rtnl_unlock();
1433 if (tun)
1434 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001435 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436}
1437
Arnd Bergmann50857e22009-11-06 22:52:32 -08001438static long tun_chr_ioctl(struct file *file,
1439 unsigned int cmd, unsigned long arg)
1440{
1441 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1442}
1443
1444#ifdef CONFIG_COMPAT
1445static long tun_chr_compat_ioctl(struct file *file,
1446 unsigned int cmd, unsigned long arg)
1447{
1448 switch (cmd) {
1449 case TUNSETIFF:
1450 case TUNGETIFF:
1451 case TUNSETTXFILTER:
1452 case TUNGETSNDBUF:
1453 case TUNSETSNDBUF:
1454 case SIOCGIFHWADDR:
1455 case SIOCSIFHWADDR:
1456 arg = (unsigned long)compat_ptr(arg);
1457 break;
1458 default:
1459 arg = (compat_ulong_t)arg;
1460 break;
1461 }
1462
1463 /*
1464 * compat_ifreq is shorter than ifreq, so we must not access beyond
1465 * the end of that structure. All fields that are used in this
1466 * driver are compatible though, we don't need to convert the
1467 * contents.
1468 */
1469 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1470}
1471#endif /* CONFIG_COMPAT */
1472
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473static int tun_chr_fasync(int fd, struct file *file, int on)
1474{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001475 struct tun_struct *tun = tun_get(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 int ret;
1477
1478 if (!tun)
1479 return -EBADFD;
1480
Joe Perches6b8a66e2011-03-02 07:18:10 +00001481 tun_debug(KERN_INFO, tun, "tun_chr_fasync %d\n", on);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482
1483 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06001484 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001485
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 if (on) {
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001487 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 if (ret)
Jonathan Corbet9d319522008-06-19 15:50:37 -06001489 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 tun->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001491 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 tun->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06001493 ret = 0;
1494out:
Eric W. Biederman631ab462009-01-20 11:00:40 +00001495 tun_put(tun);
Jonathan Corbet9d319522008-06-19 15:50:37 -06001496 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497}
1498
1499static int tun_chr_open(struct inode *inode, struct file * file)
1500{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001501 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07001502
Joe Perches6b8a66e2011-03-02 07:18:10 +00001503 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00001504
1505 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1506 if (!tfile)
1507 return -ENOMEM;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001508 atomic_set(&tfile->count, 0);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001509 tfile->tun = NULL;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001510 tfile->net = get_net(current->nsproxy->net_ns);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001511 file->private_data = tfile;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 return 0;
1513}
1514
1515static int tun_chr_close(struct inode *inode, struct file *file)
1516{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001517 struct tun_file *tfile = file->private_data;
Eric W. Biedermanf0a4d0e2009-06-08 00:44:31 -07001518 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
Eric W. Biedermanf0a4d0e2009-06-08 00:44:31 -07001520 tun = __tun_get(tfile);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001521 if (tun) {
Herbert Xud23e4362009-07-02 23:03:55 +00001522 struct net_device *dev = tun->dev;
1523
Joe Perches6b8a66e2011-03-02 07:18:10 +00001524 tun_debug(KERN_INFO, tun, "tun_chr_close\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525
Eric W. Biederman631ab462009-01-20 11:00:40 +00001526 __tun_detach(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Daniel Mack3ad2f3f2010-02-03 08:01:28 +08001528 /* If desirable, unregister the netdevice. */
Herbert Xud23e4362009-07-02 23:03:55 +00001529 if (!(tun->flags & TUN_PERSIST)) {
1530 rtnl_lock();
1531 if (dev->reg_state == NETREG_REGISTERED)
1532 unregister_netdevice(dev);
1533 rtnl_unlock();
1534 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001535 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
Herbert Xu9c3fea62009-04-18 14:15:52 +00001537 tun = tfile->tun;
1538 if (tun)
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001539 sock_put(tun->socket.sk);
Herbert Xu9c3fea62009-04-18 14:15:52 +00001540
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001541 put_net(tfile->net);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001542 kfree(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543
1544 return 0;
1545}
1546
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08001547static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001548 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 .llseek = no_llseek,
Badari Pulavartyee0b3e62006-09-30 23:28:47 -07001550 .read = do_sync_read,
1551 .aio_read = tun_chr_aio_read,
1552 .write = do_sync_write,
1553 .aio_write = tun_chr_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08001555 .unlocked_ioctl = tun_chr_ioctl,
1556#ifdef CONFIG_COMPAT
1557 .compat_ioctl = tun_chr_compat_ioctl,
1558#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 .open = tun_chr_open,
1560 .release = tun_chr_close,
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001561 .fasync = tun_chr_fasync
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562};
1563
1564static struct miscdevice tun_miscdev = {
1565 .minor = TUN_MINOR,
1566 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02001567 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569};
1570
1571/* ethtool interface */
1572
1573static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1574{
1575 cmd->supported = 0;
1576 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00001577 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 cmd->duplex = DUPLEX_FULL;
1579 cmd->port = PORT_TP;
1580 cmd->phy_address = 0;
1581 cmd->transceiver = XCVR_INTERNAL;
1582 cmd->autoneg = AUTONEG_DISABLE;
1583 cmd->maxtxpkt = 0;
1584 cmd->maxrxpkt = 0;
1585 return 0;
1586}
1587
1588static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1589{
1590 struct tun_struct *tun = netdev_priv(dev);
1591
1592 strcpy(info->driver, DRV_NAME);
1593 strcpy(info->version, DRV_VERSION);
1594 strcpy(info->fw_version, "N/A");
1595
1596 switch (tun->flags & TUN_TYPE_MASK) {
1597 case TUN_TUN_DEV:
1598 strcpy(info->bus_info, "tun");
1599 break;
1600 case TUN_TAP_DEV:
1601 strcpy(info->bus_info, "tap");
1602 break;
1603 }
1604}
1605
1606static u32 tun_get_msglevel(struct net_device *dev)
1607{
1608#ifdef TUN_DEBUG
1609 struct tun_struct *tun = netdev_priv(dev);
1610 return tun->debug;
1611#else
1612 return -EOPNOTSUPP;
1613#endif
1614}
1615
1616static void tun_set_msglevel(struct net_device *dev, u32 value)
1617{
1618#ifdef TUN_DEBUG
1619 struct tun_struct *tun = netdev_priv(dev);
1620 tun->debug = value;
1621#endif
1622}
1623
Jeff Garzik7282d492006-09-13 14:30:00 -04001624static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 .get_settings = tun_get_settings,
1626 .get_drvinfo = tun_get_drvinfo,
1627 .get_msglevel = tun_get_msglevel,
1628 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00001629 .get_link = ethtool_op_get_link,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630};
1631
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001632
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633static int __init tun_init(void)
1634{
1635 int ret = 0;
1636
Joe Perches6b8a66e2011-03-02 07:18:10 +00001637 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1638 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001640 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001641 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001642 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001643 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001644 }
1645
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001647 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001648 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001649 goto err_misc;
1650 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001651 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001652err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001653 rtnl_link_unregister(&tun_link_ops);
1654err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 return ret;
1656}
1657
1658static void tun_cleanup(void)
1659{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001660 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001661 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662}
1663
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001664/* Get an underlying socket object from tun file. Returns error unless file is
1665 * attached to a device. The returned object works like a packet socket, it
1666 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1667 * holding a reference to the file for as long as the socket is in use. */
1668struct socket *tun_get_socket(struct file *file)
1669{
1670 struct tun_struct *tun;
1671 if (file->f_op != &tun_fops)
1672 return ERR_PTR(-EINVAL);
1673 tun = tun_get(file);
1674 if (!tun)
1675 return ERR_PTR(-EBADFD);
1676 tun_put(tun);
1677 return &tun->socket;
1678}
1679EXPORT_SYMBOL_GPL(tun_get_socket);
1680
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681module_init(tun_init);
1682module_exit(tun_cleanup);
1683MODULE_DESCRIPTION(DRV_DESCRIPTION);
1684MODULE_AUTHOR(DRV_COPYRIGHT);
1685MODULE_LICENSE("GPL");
1686MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02001687MODULE_ALIAS("devname:net/tun");