blob: 01e99f22210e0710b8e1bfe77565b051aa849b54 [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
37#define DRV_NAME "tun"
38#define DRV_VERSION "1.6"
39#define DRV_DESCRIPTION "Universal TUN/TAP device driver"
40#define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
41
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <linux/module.h>
43#include <linux/errno.h>
44#include <linux/kernel.h>
45#include <linux/major.h>
46#include <linux/slab.h>
47#include <linux/poll.h>
48#include <linux/fcntl.h>
49#include <linux/init.h>
50#include <linux/skbuff.h>
51#include <linux/netdevice.h>
52#include <linux/etherdevice.h>
53#include <linux/miscdevice.h>
54#include <linux/ethtool.h>
55#include <linux/rtnetlink.h>
Arnd Bergmann50857e22009-11-06 22:52:32 -080056#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/if.h>
58#include <linux/if_arp.h>
59#include <linux/if_ether.h>
60#include <linux/if_tun.h>
61#include <linux/crc32.h>
Pavel Emelyanovd647a592008-04-16 00:41:16 -070062#include <linux/nsproxy.h>
Rusty Russellf43798c2008-07-03 03:48:02 -070063#include <linux/virtio_net.h>
Eric W. Biederman881d9662007-09-17 11:56:21 -070064#include <net/net_namespace.h>
Pavel Emelyanov79d17602008-04-16 00:40:46 -070065#include <net/netns/generic.h>
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -080066#include <net/rtnetlink.h>
Herbert Xu33dccbb2009-02-05 21:25:32 -080067#include <net/sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
69#include <asm/system.h>
70#include <asm/uaccess.h>
71
Rusty Russell14daa022008-04-12 18:48:58 -070072/* Uncomment to enable debugging */
73/* #define TUN_DEBUG 1 */
74
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#ifdef TUN_DEBUG
76static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070077
78#define DBG if(tun->debug)printk
79#define DBG1 if(debug==2)printk
80#else
81#define DBG( a... )
82#define DBG1( a... )
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#endif
84
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -070085#define FLT_EXACT_COUNT 8
86struct tap_filter {
87 unsigned int count; /* Number of addrs. Zero means disabled */
88 u32 mask[2]; /* Mask of the hashed addrs */
89 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
90};
91
Eric W. Biederman631ab462009-01-20 11:00:40 +000092struct tun_file {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +000093 atomic_t count;
Eric W. Biederman631ab462009-01-20 11:00:40 +000094 struct tun_struct *tun;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +000095 struct net *net;
Eric W. Biederman631ab462009-01-20 11:00:40 +000096};
97
Herbert Xu33dccbb2009-02-05 21:25:32 -080098struct tun_sock;
99
Rusty Russell14daa022008-04-12 18:48:58 -0700100struct tun_struct {
Eric W. Biederman631ab462009-01-20 11:00:40 +0000101 struct tun_file *tfile;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700102 unsigned int flags;
Rusty Russell14daa022008-04-12 18:48:58 -0700103 uid_t owner;
104 gid_t group;
105
Rusty Russell14daa022008-04-12 18:48:58 -0700106 struct net_device *dev;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700107 struct fasync_struct *fasync;
Rusty Russell14daa022008-04-12 18:48:58 -0700108
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700109 struct tap_filter txflt;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800110 struct socket socket;
Rusty Russell14daa022008-04-12 18:48:58 -0700111
112#ifdef TUN_DEBUG
113 int debug;
114#endif
115};
116
Herbert Xu33dccbb2009-02-05 21:25:32 -0800117struct tun_sock {
118 struct sock sk;
119 struct tun_struct *tun;
120};
121
122static inline struct tun_sock *tun_sk(struct sock *sk)
123{
124 return container_of(sk, struct tun_sock, sk);
125}
126
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000127static int tun_attach(struct tun_struct *tun, struct file *file)
128{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000129 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000130 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000131
132 ASSERT_RTNL();
133
Eric W. Biederman38231b72009-01-20 11:02:28 +0000134 netif_tx_lock_bh(tun->dev);
135
136 err = -EINVAL;
137 if (tfile->tun)
138 goto out;
139
140 err = -EBUSY;
141 if (tun->tfile)
142 goto out;
143
144 err = 0;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000145 tfile->tun = tun;
146 tun->tfile = tfile;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000147 dev_hold(tun->dev);
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000148 sock_hold(tun->socket.sk);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000149 atomic_inc(&tfile->count);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000150
Eric W. Biederman38231b72009-01-20 11:02:28 +0000151out:
152 netif_tx_unlock_bh(tun->dev);
153 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000154}
155
Eric W. Biederman631ab462009-01-20 11:00:40 +0000156static void __tun_detach(struct tun_struct *tun)
157{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000158 /* Detach from net device */
Eric W. Biederman38231b72009-01-20 11:02:28 +0000159 netif_tx_lock_bh(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000160 tun->tfile = NULL;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000161 netif_tx_unlock_bh(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000162
163 /* Drop read queue */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000164 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000165
166 /* Drop the extra count on the net device */
167 dev_put(tun->dev);
168}
169
170static void tun_detach(struct tun_struct *tun)
171{
172 rtnl_lock();
173 __tun_detach(tun);
174 rtnl_unlock();
Eric W. Biederman631ab462009-01-20 11:00:40 +0000175}
176
177static struct tun_struct *__tun_get(struct tun_file *tfile)
178{
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000179 struct tun_struct *tun = NULL;
180
181 if (atomic_inc_not_zero(&tfile->count))
182 tun = tfile->tun;
183
184 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000185}
186
187static struct tun_struct *tun_get(struct file *file)
188{
189 return __tun_get(file->private_data);
190}
191
192static void tun_put(struct tun_struct *tun)
193{
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000194 struct tun_file *tfile = tun->tfile;
195
196 if (atomic_dec_and_test(&tfile->count))
197 tun_detach(tfile->tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000198}
199
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700200/* TAP filterting */
201static void addr_hash_set(u32 *mask, const u8 *addr)
202{
203 int n = ether_crc(ETH_ALEN, addr) >> 26;
204 mask[n >> 5] |= (1 << (n & 31));
205}
206
207static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
208{
209 int n = ether_crc(ETH_ALEN, addr) >> 26;
210 return mask[n >> 5] & (1 << (n & 31));
211}
212
213static int update_filter(struct tap_filter *filter, void __user *arg)
214{
215 struct { u8 u[ETH_ALEN]; } *addr;
216 struct tun_filter uf;
217 int err, alen, n, nexact;
218
219 if (copy_from_user(&uf, arg, sizeof(uf)))
220 return -EFAULT;
221
222 if (!uf.count) {
223 /* Disabled */
224 filter->count = 0;
225 return 0;
226 }
227
228 alen = ETH_ALEN * uf.count;
229 addr = kmalloc(alen, GFP_KERNEL);
230 if (!addr)
231 return -ENOMEM;
232
233 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
234 err = -EFAULT;
235 goto done;
236 }
237
238 /* The filter is updated without holding any locks. Which is
239 * perfectly safe. We disable it first and in the worst
240 * case we'll accept a few undesired packets. */
241 filter->count = 0;
242 wmb();
243
244 /* Use first set of addresses as an exact filter */
245 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
246 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
247
248 nexact = n;
249
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800250 /* Remaining multicast addresses are hashed,
251 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700252 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800253 for (; n < uf.count; n++) {
254 if (!is_multicast_ether_addr(addr[n].u)) {
255 err = 0; /* no filter */
256 goto done;
257 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700258 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800259 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700260
261 /* For ALLMULTI just set the mask to all ones.
262 * This overrides the mask populated above. */
263 if ((uf.flags & TUN_FLT_ALLMULTI))
264 memset(filter->mask, ~0, sizeof(filter->mask));
265
266 /* Now enable the filter */
267 wmb();
268 filter->count = nexact;
269
270 /* Return the number of exact filters */
271 err = nexact;
272
273done:
274 kfree(addr);
275 return err;
276}
277
278/* Returns: 0 - drop, !=0 - accept */
279static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
280{
281 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
282 * at this point. */
283 struct ethhdr *eh = (struct ethhdr *) skb->data;
284 int i;
285
286 /* Exact match */
287 for (i = 0; i < filter->count; i++)
288 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
289 return 1;
290
291 /* Inexact match (multicast only) */
292 if (is_multicast_ether_addr(eh->h_dest))
293 return addr_hash_test(filter->mask, eh->h_dest);
294
295 return 0;
296}
297
298/*
299 * Checks whether the packet is accepted or not.
300 * Returns: 0 - drop, !=0 - accept
301 */
302static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
303{
304 if (!filter->count)
305 return 1;
306
307 return run_filter(filter, skb);
308}
309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310/* Network device part of the driver */
311
Jeff Garzik7282d492006-09-13 14:30:00 -0400312static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000314/* Net device detach from fd. */
315static void tun_net_uninit(struct net_device *dev)
316{
317 struct tun_struct *tun = netdev_priv(dev);
318 struct tun_file *tfile = tun->tfile;
319
320 /* Inform the methods they need to stop using the dev.
321 */
322 if (tfile) {
Herbert Xuc40af842009-04-19 22:35:50 +0000323 wake_up_all(&tun->socket.wait);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000324 if (atomic_dec_and_test(&tfile->count))
325 __tun_detach(tun);
326 }
327}
328
Herbert Xu9c3fea62009-04-18 14:15:52 +0000329static void tun_free_netdev(struct net_device *dev)
330{
331 struct tun_struct *tun = netdev_priv(dev);
332
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000333 sock_put(tun->socket.sk);
Herbert Xu9c3fea62009-04-18 14:15:52 +0000334}
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336/* Net device open. */
337static int tun_net_open(struct net_device *dev)
338{
339 netif_start_queue(dev);
340 return 0;
341}
342
343/* Net device close. */
344static int tun_net_close(struct net_device *dev)
345{
346 netif_stop_queue(dev);
347 return 0;
348}
349
350/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000351static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352{
353 struct tun_struct *tun = netdev_priv(dev);
354
355 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
356
357 /* Drop packet if interface is not attached */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000358 if (!tun->tfile)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 goto drop;
360
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700361 /* Drop if the filter does not like it.
362 * This is a noop if the filter is disabled.
363 * Filter can be enabled only for the TAP devices. */
364 if (!check_filter(&tun->txflt, skb))
365 goto drop;
366
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000367 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 if (!(tun->flags & TUN_ONE_QUEUE)) {
369 /* Normal queueing mode. */
370 /* Packet scheduler handles dropping of further packets. */
371 netif_stop_queue(dev);
372
373 /* We won't see all dropped packets individually, so overrun
374 * error is more appropriate. */
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700375 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 } else {
377 /* Single queue mode.
378 * Driver handles dropping of all packets itself. */
379 goto drop;
380 }
381 }
382
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700383 /* Enqueue packet */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000384 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 dev->trans_start = jiffies;
386
387 /* Notify and wake up reader process */
388 if (tun->flags & TUN_FASYNC)
389 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
Herbert Xuc40af842009-04-19 22:35:50 +0000390 wake_up_interruptible(&tun->socket.wait);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000391 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700394 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000396 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397}
398
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700399static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700401 /*
402 * This callback is supposed to deal with mc filter in
403 * _rx_ path and has nothing to do with the _tx_ path.
404 * In rx path we always accept everything userspace gives us.
405 */
406 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407}
408
Ed Swierk4885a502007-09-16 12:21:38 -0700409#define MIN_MTU 68
410#define MAX_MTU 65535
411
412static int
413tun_net_change_mtu(struct net_device *dev, int new_mtu)
414{
415 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
416 return -EINVAL;
417 dev->mtu = new_mtu;
418 return 0;
419}
420
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800421static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000422 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800423 .ndo_open = tun_net_open,
424 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800425 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800426 .ndo_change_mtu = tun_net_change_mtu,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800427};
428
429static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000430 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800431 .ndo_open = tun_net_open,
432 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800433 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800434 .ndo_change_mtu = tun_net_change_mtu,
435 .ndo_set_multicast_list = tun_net_mclist,
436 .ndo_set_mac_address = eth_mac_addr,
437 .ndo_validate_addr = eth_validate_addr,
438};
439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440/* Initialize net device. */
441static void tun_net_init(struct net_device *dev)
442{
443 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400444
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 switch (tun->flags & TUN_TYPE_MASK) {
446 case TUN_TUN_DEV:
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800447 dev->netdev_ops = &tun_netdev_ops;
448
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 /* Point-to-Point TUN Device */
450 dev->hard_header_len = 0;
451 dev->addr_len = 0;
452 dev->mtu = 1500;
453
454 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400455 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
457 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
458 break;
459
460 case TUN_TAP_DEV:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -0800461 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700463 ether_setup(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700465 random_ether_addr(dev->dev_addr);
Brian Braunstein36226a82007-04-26 01:00:55 -0700466
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
468 break;
469 }
470}
471
472/* Character device part */
473
474/* Poll */
475static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400476{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +0000477 struct tun_file *tfile = file->private_data;
478 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000479 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800480 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481
482 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +0000483 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000485 sk = tun->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
488
Herbert Xuc40af842009-04-19 22:35:50 +0000489 poll_wait(file, &tun->socket.wait, wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400490
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000491 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 mask |= POLLIN | POLLRDNORM;
493
Herbert Xu33dccbb2009-02-05 21:25:32 -0800494 if (sock_writeable(sk) ||
495 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
496 sock_writeable(sk)))
497 mask |= POLLOUT | POLLWRNORM;
498
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000499 if (tun->dev->reg_state != NETREG_REGISTERED)
500 mask = POLLERR;
501
Eric W. Biederman631ab462009-01-20 11:00:40 +0000502 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 return mask;
504}
505
Rusty Russellf42157c2008-08-15 15:15:10 -0700506/* prepad is the amount to reserve at front. len is length after that.
507 * linear is a hint as to how much to copy (usually headers). */
Herbert Xu33dccbb2009-02-05 21:25:32 -0800508static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
509 size_t prepad, size_t len,
510 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -0700511{
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000512 struct sock *sk = tun->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -0700513 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800514 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -0700515
516 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -0700517 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800518 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -0700519
Herbert Xu33dccbb2009-02-05 21:25:32 -0800520 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
521 &err);
Rusty Russellf42157c2008-08-15 15:15:10 -0700522 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800523 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -0700524
525 skb_reserve(skb, prepad);
526 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800527 skb->data_len = len - linear;
528 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -0700529
530 return skb;
531}
532
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533/* Get packet from user space buffer */
Herbert Xu33dccbb2009-02-05 21:25:32 -0800534static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000535 const struct iovec *iv, size_t count,
Herbert Xu33dccbb2009-02-05 21:25:32 -0800536 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
Harvey Harrison09640e62009-02-01 00:45:17 -0800538 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 struct sk_buff *skb;
540 size_t len = count, align = 0;
Rusty Russellf43798c2008-07-03 03:48:02 -0700541 struct virtio_net_hdr gso = { 0 };
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000542 int offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543
544 if (!(tun->flags & TUN_NO_PI)) {
545 if ((len -= sizeof(pi)) > count)
546 return -EINVAL;
547
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000548 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 return -EFAULT;
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000550 offset += sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 }
552
Rusty Russellf43798c2008-07-03 03:48:02 -0700553 if (tun->flags & TUN_VNET_HDR) {
554 if ((len -= sizeof(gso)) > count)
555 return -EINVAL;
556
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000557 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
Rusty Russellf43798c2008-07-03 03:48:02 -0700558 return -EFAULT;
559
Herbert Xu49091222009-06-08 00:20:01 -0700560 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
561 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
562 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
563
Rusty Russellf43798c2008-07-03 03:48:02 -0700564 if (gso.hdr_len > len)
565 return -EINVAL;
Sridhar Samudrala6f536f42009-06-08 00:27:28 -0700566 offset += sizeof(gso);
Rusty Russellf43798c2008-07-03 03:48:02 -0700567 }
568
Rusty Russelle01bf1c2008-04-12 18:49:30 -0700569 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 align = NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -0700571 if (unlikely(len < ETH_HLEN ||
572 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -0700573 return -EINVAL;
574 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400575
Herbert Xu33dccbb2009-02-05 21:25:32 -0800576 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
577 if (IS_ERR(skb)) {
578 if (PTR_ERR(skb) != -EAGAIN)
579 tun->dev->stats.rx_dropped++;
580 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 }
582
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000583 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700584 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -0800585 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -0800587 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
Rusty Russellf43798c2008-07-03 03:48:02 -0700589 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
590 if (!skb_partial_csum_set(skb, gso.csum_start,
591 gso.csum_offset)) {
592 tun->dev->stats.rx_frame_errors++;
593 kfree_skb(skb);
594 return -EINVAL;
595 }
596 } else if (tun->flags & TUN_NOCHECKSUM)
597 skb->ip_summed = CHECKSUM_UNNECESSARY;
598
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 switch (tun->flags & TUN_TYPE_MASK) {
600 case TUN_TUN_DEV:
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -0700601 if (tun->flags & TUN_NO_PI) {
602 switch (skb->data[0] & 0xf0) {
603 case 0x40:
604 pi.proto = htons(ETH_P_IP);
605 break;
606 case 0x60:
607 pi.proto = htons(ETH_P_IPV6);
608 break;
609 default:
610 tun->dev->stats.rx_dropped++;
611 kfree_skb(skb);
612 return -EINVAL;
613 }
614 }
615
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -0700616 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -0700618 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 break;
620 case TUN_TAP_DEV:
621 skb->protocol = eth_type_trans(skb, tun->dev);
622 break;
623 };
624
Rusty Russellf43798c2008-07-03 03:48:02 -0700625 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
626 pr_debug("GSO!\n");
627 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
628 case VIRTIO_NET_HDR_GSO_TCPV4:
629 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
630 break;
631 case VIRTIO_NET_HDR_GSO_TCPV6:
632 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
633 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +0000634 case VIRTIO_NET_HDR_GSO_UDP:
635 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
636 break;
Rusty Russellf43798c2008-07-03 03:48:02 -0700637 default:
638 tun->dev->stats.rx_frame_errors++;
639 kfree_skb(skb);
640 return -EINVAL;
641 }
642
643 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
644 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
645
646 skb_shinfo(skb)->gso_size = gso.gso_size;
647 if (skb_shinfo(skb)->gso_size == 0) {
648 tun->dev->stats.rx_frame_errors++;
649 kfree_skb(skb);
650 return -EINVAL;
651 }
652
653 /* Header must be checked, and gso_segs computed. */
654 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
655 skb_shinfo(skb)->gso_segs = 0;
656 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400657
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400659
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700660 tun->dev->stats.rx_packets++;
661 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662
663 return count;
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400664}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700666static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
667 unsigned long count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668{
Herbert Xu33dccbb2009-02-05 21:25:32 -0800669 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -0800670 struct tun_struct *tun = tun_get(file);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000671 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
673 if (!tun)
674 return -EBADFD;
675
676 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
677
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000678 result = tun_get_user(tun, iv, iov_length(iv, count),
Herbert Xu33dccbb2009-02-05 21:25:32 -0800679 file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000680
681 tun_put(tun);
682 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683}
684
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685/* Put packet to the user space buffer */
686static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
687 struct sk_buff *skb,
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000688 const struct iovec *iv, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689{
690 struct tun_pi pi = { 0, skb->protocol };
691 ssize_t total = 0;
692
693 if (!(tun->flags & TUN_NO_PI)) {
694 if ((len -= sizeof(pi)) < 0)
695 return -EINVAL;
696
697 if (len < skb->len) {
698 /* Packet will be striped */
699 pi.flags |= TUN_PKT_STRIP;
700 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400701
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000702 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return -EFAULT;
704 total += sizeof(pi);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400705 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
Rusty Russellf43798c2008-07-03 03:48:02 -0700707 if (tun->flags & TUN_VNET_HDR) {
708 struct virtio_net_hdr gso = { 0 }; /* no info leak */
709 if ((len -= sizeof(gso)) < 0)
710 return -EINVAL;
711
712 if (skb_is_gso(skb)) {
713 struct skb_shared_info *sinfo = skb_shinfo(skb);
714
715 /* This is a hint as to how much should be linear. */
716 gso.hdr_len = skb_headlen(skb);
717 gso.gso_size = sinfo->gso_size;
718 if (sinfo->gso_type & SKB_GSO_TCPV4)
719 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
720 else if (sinfo->gso_type & SKB_GSO_TCPV6)
721 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Sridhar Samudralae36aa252009-07-14 14:21:04 +0000722 else if (sinfo->gso_type & SKB_GSO_UDP)
723 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
Rusty Russellf43798c2008-07-03 03:48:02 -0700724 else
725 BUG();
726 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
727 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
728 } else
729 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
730
731 if (skb->ip_summed == CHECKSUM_PARTIAL) {
732 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
733 gso.csum_start = skb->csum_start - skb_headroom(skb);
734 gso.csum_offset = skb->csum_offset;
735 } /* else everything is zero */
736
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000737 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
738 sizeof(gso))))
Rusty Russellf43798c2008-07-03 03:48:02 -0700739 return -EFAULT;
740 total += sizeof(gso);
741 }
742
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 len = min_t(int, skb->len, len);
744
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000745 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 total += len;
747
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700748 tun->dev->stats.tx_packets++;
749 tun->dev->stats.tx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751 return total;
752}
753
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700754static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
755 unsigned long count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756{
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700757 struct file *file = iocb->ki_filp;
Eric W. Biedermanb2430de2009-01-20 11:03:21 +0000758 struct tun_file *tfile = file->private_data;
759 struct tun_struct *tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 DECLARE_WAITQUEUE(wait, current);
761 struct sk_buff *skb;
762 ssize_t len, ret = 0;
763
764 if (!tun)
765 return -EBADFD;
766
767 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
768
Akinobu Mita52427c92007-11-19 22:46:51 -0800769 len = iov_length(iv, count);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000770 if (len < 0) {
771 ret = -EINVAL;
772 goto out;
773 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774
Herbert Xuc40af842009-04-19 22:35:50 +0000775 add_wait_queue(&tun->socket.wait, &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 while (len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 current->state = TASK_INTERRUPTIBLE;
778
779 /* Read frames from the queue */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000780 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 if (file->f_flags & O_NONBLOCK) {
782 ret = -EAGAIN;
783 break;
784 }
785 if (signal_pending(current)) {
786 ret = -ERESTARTSYS;
787 break;
788 }
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000789 if (tun->dev->reg_state != NETREG_REGISTERED) {
790 ret = -EIO;
791 break;
792 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 /* Nothing to read, let's sleep */
795 schedule();
796 continue;
797 }
798 netif_wake_queue(tun->dev);
799
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000800 ret = tun_put_user(tun, skb, iv, len);
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700801 kfree_skb(skb);
802 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 }
804
805 current->state = TASK_RUNNING;
Herbert Xuc40af842009-04-19 22:35:50 +0000806 remove_wait_queue(&tun->socket.wait, &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
Eric W. Biederman631ab462009-01-20 11:00:40 +0000808out:
809 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 return ret;
811}
812
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813static void tun_setup(struct net_device *dev)
814{
815 struct tun_struct *tun = netdev_priv(dev);
816
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 tun->owner = -1;
Guido Guenther8c644622007-07-02 22:50:25 -0700818 tun->group = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 dev->ethtool_ops = &tun_ethtool_ops;
Herbert Xu9c3fea62009-04-18 14:15:52 +0000821 dev->destructor = tun_free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822}
823
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -0800824/* Trivial set of netlink ops to allow deleting tun or tap
825 * device with netlink.
826 */
827static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
828{
829 return -EINVAL;
830}
831
832static struct rtnl_link_ops tun_link_ops __read_mostly = {
833 .kind = DRV_NAME,
834 .priv_size = sizeof(struct tun_struct),
835 .setup = tun_setup,
836 .validate = tun_validate,
837};
838
Herbert Xu33dccbb2009-02-05 21:25:32 -0800839static void tun_sock_write_space(struct sock *sk)
840{
841 struct tun_struct *tun;
842
843 if (!sock_writeable(sk))
844 return;
845
Herbert Xu33dccbb2009-02-05 21:25:32 -0800846 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
847 return;
848
Herbert Xuc722c622009-06-03 21:45:55 -0700849 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
850 wake_up_interruptible_sync(sk->sk_sleep);
851
Herbert Xu33dccbb2009-02-05 21:25:32 -0800852 tun = container_of(sk, struct tun_sock, sk)->tun;
853 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
854}
855
856static void tun_sock_destruct(struct sock *sk)
857{
Herbert Xu9c3fea62009-04-18 14:15:52 +0000858 free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800859}
860
861static struct proto tun_proto = {
862 .name = "tun",
863 .owner = THIS_MODULE,
864 .obj_size = sizeof(struct tun_sock),
865};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -0800866
David Woodhouse980c9e82009-05-09 22:54:21 -0700867static int tun_flags(struct tun_struct *tun)
868{
869 int flags = 0;
870
871 if (tun->flags & TUN_TUN_DEV)
872 flags |= IFF_TUN;
873 else
874 flags |= IFF_TAP;
875
876 if (tun->flags & TUN_NO_PI)
877 flags |= IFF_NO_PI;
878
879 if (tun->flags & TUN_ONE_QUEUE)
880 flags |= IFF_ONE_QUEUE;
881
882 if (tun->flags & TUN_VNET_HDR)
883 flags |= IFF_VNET_HDR;
884
885 return flags;
886}
887
888static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
889 char *buf)
890{
891 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
892 return sprintf(buf, "0x%x\n", tun_flags(tun));
893}
894
895static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
896 char *buf)
897{
898 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
899 return sprintf(buf, "%d\n", tun->owner);
900}
901
902static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
903 char *buf)
904{
905 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
906 return sprintf(buf, "%d\n", tun->group);
907}
908
909static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
910static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
911static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
912
Pavel Emelyanovd647a592008-04-16 00:41:16 -0700913static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914{
Herbert Xu33dccbb2009-02-05 21:25:32 -0800915 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 struct tun_struct *tun;
917 struct net_device *dev;
918 int err;
919
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +0000920 dev = __dev_get_by_name(net, ifr->ifr_name);
921 if (dev) {
Paul Moore2b980db2009-08-28 18:12:43 -0400922 const struct cred *cred = current_cred();
923
David Woodhousef85ba782009-04-27 03:23:54 -0700924 if (ifr->ifr_flags & IFF_TUN_EXCL)
925 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +0000926 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
927 tun = netdev_priv(dev);
928 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
929 tun = netdev_priv(dev);
930 else
931 return -EINVAL;
932
Paul Moore2b980db2009-08-28 18:12:43 -0400933 if (((tun->owner != -1 && cred->euid != tun->owner) ||
934 (tun->group != -1 && !in_egroup_p(tun->group))) &&
935 !capable(CAP_NET_ADMIN))
936 return -EPERM;
Linus Torvaldsd7e96602009-09-14 10:37:28 -0700937 err = security_tun_dev_attach(tun->socket.sk);
Paul Moore2b980db2009-08-28 18:12:43 -0400938 if (err < 0)
939 return err;
940
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000941 err = tun_attach(tun, file);
942 if (err < 0)
943 return err;
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400944 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 else {
946 char *name;
947 unsigned long flags = 0;
948
David Woodhouseca6bb5d2006-06-22 16:07:52 -0700949 if (!capable(CAP_NET_ADMIN))
950 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -0400951 err = security_tun_dev_create();
952 if (err < 0)
953 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -0700954
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 /* Set dev type */
956 if (ifr->ifr_flags & IFF_TUN) {
957 /* TUN device */
958 flags |= TUN_TUN_DEV;
959 name = "tun%d";
960 } else if (ifr->ifr_flags & IFF_TAP) {
961 /* TAP device */
962 flags |= TUN_TAP_DEV;
963 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400964 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +0000965 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400966
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 if (*ifr->ifr_name)
968 name = ifr->ifr_name;
969
970 dev = alloc_netdev(sizeof(struct tun_struct), name,
971 tun_setup);
972 if (!dev)
973 return -ENOMEM;
974
Pavel Emelyanovfc54c652008-04-16 00:41:53 -0700975 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -0800976 dev->rtnl_link_ops = &tun_link_ops;
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800977
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 tun = netdev_priv(dev);
979 tun->dev = dev;
980 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700981 tun->txflt.count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Herbert Xu33dccbb2009-02-05 21:25:32 -0800983 err = -ENOMEM;
984 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
985 if (!sk)
986 goto err_free_dev;
987
Herbert Xuc40af842009-04-19 22:35:50 +0000988 init_waitqueue_head(&tun->socket.wait);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800989 sock_init_data(&tun->socket, sk);
990 sk->sk_write_space = tun_sock_write_space;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800991 sk->sk_sndbuf = INT_MAX;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800992
Herbert Xu33dccbb2009-02-05 21:25:32 -0800993 container_of(sk, struct tun_sock, sk)->tun = tun;
994
Paul Moore2b980db2009-08-28 18:12:43 -0400995 security_tun_dev_post_create(sk);
996
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 tun_net_init(dev);
998
999 if (strchr(dev->name, '%')) {
1000 err = dev_alloc_name(dev, dev->name);
1001 if (err < 0)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001002 goto err_free_sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 }
1004
1005 err = register_netdevice(tun->dev);
1006 if (err < 0)
Herbert Xu9c3fea62009-04-18 14:15:52 +00001007 goto err_free_sk;
1008
David Woodhouse980c9e82009-05-09 22:54:21 -07001009 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1010 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1011 device_create_file(&tun->dev->dev, &dev_attr_group))
1012 printk(KERN_ERR "Failed to create tun sysfs files\n");
1013
Herbert Xu9c3fea62009-04-18 14:15:52 +00001014 sk->sk_destruct = tun_sock_destruct;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001015
1016 err = tun_attach(tun, file);
1017 if (err < 0)
Herbert Xu9c3fea62009-04-18 14:15:52 +00001018 goto failed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 }
1020
1021 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
1022
1023 if (ifr->ifr_flags & IFF_NO_PI)
1024 tun->flags |= TUN_NO_PI;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001025 else
1026 tun->flags &= ~TUN_NO_PI;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
1028 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1029 tun->flags |= TUN_ONE_QUEUE;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001030 else
1031 tun->flags &= ~TUN_ONE_QUEUE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Rusty Russellf43798c2008-07-03 03:48:02 -07001033 if (ifr->ifr_flags & IFF_VNET_HDR)
1034 tun->flags |= TUN_VNET_HDR;
1035 else
1036 tun->flags &= ~TUN_VNET_HDR;
1037
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001038 /* Make sure persistent devices do not get stuck in
1039 * xoff state.
1040 */
1041 if (netif_running(tun->dev))
1042 netif_wake_queue(tun->dev);
1043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 strcpy(ifr->ifr_name, tun->dev->name);
1045 return 0;
1046
Herbert Xu33dccbb2009-02-05 21:25:32 -08001047 err_free_sk:
1048 sock_put(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 err_free_dev:
1050 free_netdev(dev);
1051 failed:
1052 return err;
1053}
1054
Herbert Xu876bfd42009-08-06 14:22:44 +00001055static int tun_get_iff(struct net *net, struct tun_struct *tun,
1056 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001057{
Mark McLoughline3b99552008-08-15 15:09:56 -07001058 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1059
1060 strcpy(ifr->ifr_name, tun->dev->name);
1061
David Woodhouse980c9e82009-05-09 22:54:21 -07001062 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001063
1064 return 0;
1065}
1066
Rusty Russell5228ddc2008-07-03 03:46:16 -07001067/* This is like a cut-down ethtool ops, except done via tun fd so no
1068 * privs required. */
1069static int set_offload(struct net_device *dev, unsigned long arg)
1070{
1071 unsigned int old_features, features;
1072
1073 old_features = dev->features;
1074 /* Unset features, set them as we chew on the arg. */
1075 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001076 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6
1077 |NETIF_F_UFO));
Rusty Russell5228ddc2008-07-03 03:46:16 -07001078
1079 if (arg & TUN_F_CSUM) {
1080 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1081 arg &= ~TUN_F_CSUM;
1082
1083 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1084 if (arg & TUN_F_TSO_ECN) {
1085 features |= NETIF_F_TSO_ECN;
1086 arg &= ~TUN_F_TSO_ECN;
1087 }
1088 if (arg & TUN_F_TSO4)
1089 features |= NETIF_F_TSO;
1090 if (arg & TUN_F_TSO6)
1091 features |= NETIF_F_TSO6;
1092 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1093 }
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001094
1095 if (arg & TUN_F_UFO) {
1096 features |= NETIF_F_UFO;
1097 arg &= ~TUN_F_UFO;
1098 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001099 }
1100
1101 /* This gives the user a way to test for new features in future by
1102 * trying to set them. */
1103 if (arg)
1104 return -EINVAL;
1105
1106 dev->features = features;
1107 if (old_features != dev->features)
1108 netdev_features_change(dev);
1109
1110 return 0;
1111}
1112
Arnd Bergmann50857e22009-11-06 22:52:32 -08001113static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1114 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001116 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001117 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 void __user* argp = (void __user*)arg;
1119 struct ifreq ifr;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001120 int sndbuf;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001121 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122
1123 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
Arnd Bergmann50857e22009-11-06 22:52:32 -08001124 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 return -EFAULT;
1126
Eric W. Biederman631ab462009-01-20 11:00:40 +00001127 if (cmd == TUNGETFEATURES) {
1128 /* Currently this just means: "what IFF flags are valid?".
1129 * This is needed because we never checked for invalid flags on
1130 * TUNSETIFF. */
1131 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1132 IFF_VNET_HDR,
1133 (unsigned int __user*)argp);
1134 }
1135
Herbert Xu876bfd42009-08-06 14:22:44 +00001136 rtnl_lock();
1137
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001138 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1141
Herbert Xu876bfd42009-08-06 14:22:44 +00001142 ret = tun_set_iff(tfile->net, file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143
Herbert Xu876bfd42009-08-06 14:22:44 +00001144 if (ret)
1145 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Arnd Bergmann50857e22009-11-06 22:52:32 -08001147 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001148 ret = -EFAULT;
1149 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 }
1151
Herbert Xu876bfd42009-08-06 14:22:44 +00001152 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001154 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
1156 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1157
Eric W. Biederman631ab462009-01-20 11:00:40 +00001158 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001160 case TUNGETIFF:
Herbert Xu876bfd42009-08-06 14:22:44 +00001161 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001162 if (ret)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001163 break;
Mark McLoughline3b99552008-08-15 15:09:56 -07001164
Arnd Bergmann50857e22009-11-06 22:52:32 -08001165 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001166 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001167 break;
1168
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 case TUNSETNOCSUM:
1170 /* Disable/Enable checksum */
1171 if (arg)
1172 tun->flags |= TUN_NOCHECKSUM;
1173 else
1174 tun->flags &= ~TUN_NOCHECKSUM;
1175
1176 DBG(KERN_INFO "%s: checksum %s\n",
1177 tun->dev->name, arg ? "disabled" : "enabled");
1178 break;
1179
1180 case TUNSETPERSIST:
1181 /* Disable/Enable persist mode */
1182 if (arg)
1183 tun->flags |= TUN_PERSIST;
1184 else
1185 tun->flags &= ~TUN_PERSIST;
1186
1187 DBG(KERN_INFO "%s: persist %s\n",
Toyo Abec6e991d2007-12-24 21:29:35 -08001188 tun->dev->name, arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 break;
1190
1191 case TUNSETOWNER:
1192 /* Set owner of the device */
1193 tun->owner = (uid_t) arg;
1194
1195 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1196 break;
1197
Guido Guenther8c644622007-07-02 22:50:25 -07001198 case TUNSETGROUP:
1199 /* Set group of the device */
1200 tun->group= (gid_t) arg;
1201
1202 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1203 break;
1204
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001205 case TUNSETLINK:
1206 /* Only allow setting the type when the interface is down */
1207 if (tun->dev->flags & IFF_UP) {
1208 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1209 tun->dev->name);
David S. Miller48abfe02008-04-23 19:37:58 -07001210 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001211 } else {
1212 tun->dev->type = (int) arg;
1213 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07001214 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001215 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001216 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218#ifdef TUN_DEBUG
1219 case TUNSETDEBUG:
1220 tun->debug = arg;
1221 break;
1222#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07001223 case TUNSETOFFLOAD:
Rusty Russell5228ddc2008-07-03 03:46:16 -07001224 ret = set_offload(tun->dev, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001225 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001226
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001227 case TUNSETTXFILTER:
1228 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00001229 ret = -EINVAL;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001230 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001231 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07001232 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001233 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
1235 case SIOCGIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001236 /* Get hw addres */
1237 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1238 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08001239 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001240 ret = -EFAULT;
1241 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
1243 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001244 /* Set hw address */
Johannes Berge1749612008-10-27 15:59:26 -07001245 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1246 tun->dev->name, ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08001247
Kim B. Heino40102372008-02-29 12:26:21 -08001248 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001249 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001250
1251 case TUNGETSNDBUF:
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001252 sndbuf = tun->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001253 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1254 ret = -EFAULT;
1255 break;
1256
1257 case TUNSETSNDBUF:
1258 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1259 ret = -EFAULT;
1260 break;
1261 }
1262
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001263 tun->socket.sk->sk_sndbuf = sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001264 break;
1265
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00001267 ret = -EINVAL;
1268 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 };
1270
Herbert Xu876bfd42009-08-06 14:22:44 +00001271unlock:
1272 rtnl_unlock();
1273 if (tun)
1274 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001275 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276}
1277
Arnd Bergmann50857e22009-11-06 22:52:32 -08001278static long tun_chr_ioctl(struct file *file,
1279 unsigned int cmd, unsigned long arg)
1280{
1281 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1282}
1283
1284#ifdef CONFIG_COMPAT
1285static long tun_chr_compat_ioctl(struct file *file,
1286 unsigned int cmd, unsigned long arg)
1287{
1288 switch (cmd) {
1289 case TUNSETIFF:
1290 case TUNGETIFF:
1291 case TUNSETTXFILTER:
1292 case TUNGETSNDBUF:
1293 case TUNSETSNDBUF:
1294 case SIOCGIFHWADDR:
1295 case SIOCSIFHWADDR:
1296 arg = (unsigned long)compat_ptr(arg);
1297 break;
1298 default:
1299 arg = (compat_ulong_t)arg;
1300 break;
1301 }
1302
1303 /*
1304 * compat_ifreq is shorter than ifreq, so we must not access beyond
1305 * the end of that structure. All fields that are used in this
1306 * driver are compatible though, we don't need to convert the
1307 * contents.
1308 */
1309 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1310}
1311#endif /* CONFIG_COMPAT */
1312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313static int tun_chr_fasync(int fd, struct file *file, int on)
1314{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001315 struct tun_struct *tun = tun_get(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 int ret;
1317
1318 if (!tun)
1319 return -EBADFD;
1320
1321 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1322
1323 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06001324 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001325
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 if (on) {
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001327 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 if (ret)
Jonathan Corbet9d319522008-06-19 15:50:37 -06001329 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 tun->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001331 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 tun->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06001333 ret = 0;
1334out:
Eric W. Biederman631ab462009-01-20 11:00:40 +00001335 tun_put(tun);
Jonathan Corbet9d319522008-06-19 15:50:37 -06001336 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337}
1338
1339static int tun_chr_open(struct inode *inode, struct file * file)
1340{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001341 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07001342
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 DBG1(KERN_INFO "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00001344
1345 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1346 if (!tfile)
1347 return -ENOMEM;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001348 atomic_set(&tfile->count, 0);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001349 tfile->tun = NULL;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001350 tfile->net = get_net(current->nsproxy->net_ns);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001351 file->private_data = tfile;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 return 0;
1353}
1354
1355static int tun_chr_close(struct inode *inode, struct file *file)
1356{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001357 struct tun_file *tfile = file->private_data;
Eric W. Biedermanf0a4d0e2009-06-08 00:44:31 -07001358 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
Eric W. Biedermanf0a4d0e2009-06-08 00:44:31 -07001360 tun = __tun_get(tfile);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001361 if (tun) {
Herbert Xud23e4362009-07-02 23:03:55 +00001362 struct net_device *dev = tun->dev;
1363
1364 DBG(KERN_INFO "%s: tun_chr_close\n", dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Eric W. Biederman631ab462009-01-20 11:00:40 +00001366 __tun_detach(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Eric W. Biederman631ab462009-01-20 11:00:40 +00001368 /* If desireable, unregister the netdevice. */
Herbert Xud23e4362009-07-02 23:03:55 +00001369 if (!(tun->flags & TUN_PERSIST)) {
1370 rtnl_lock();
1371 if (dev->reg_state == NETREG_REGISTERED)
1372 unregister_netdevice(dev);
1373 rtnl_unlock();
1374 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001375 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
Herbert Xu9c3fea62009-04-18 14:15:52 +00001377 tun = tfile->tun;
1378 if (tun)
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001379 sock_put(tun->socket.sk);
Herbert Xu9c3fea62009-04-18 14:15:52 +00001380
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001381 put_net(tfile->net);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001382 kfree(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
1384 return 0;
1385}
1386
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08001387static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001388 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 .llseek = no_llseek,
Badari Pulavartyee0b3e62006-09-30 23:28:47 -07001390 .read = do_sync_read,
1391 .aio_read = tun_chr_aio_read,
1392 .write = do_sync_write,
1393 .aio_write = tun_chr_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08001395 .unlocked_ioctl = tun_chr_ioctl,
1396#ifdef CONFIG_COMPAT
1397 .compat_ioctl = tun_chr_compat_ioctl,
1398#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 .open = tun_chr_open,
1400 .release = tun_chr_close,
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001401 .fasync = tun_chr_fasync
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402};
1403
1404static struct miscdevice tun_miscdev = {
1405 .minor = TUN_MINOR,
1406 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02001407 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409};
1410
1411/* ethtool interface */
1412
1413static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1414{
1415 cmd->supported = 0;
1416 cmd->advertising = 0;
1417 cmd->speed = SPEED_10;
1418 cmd->duplex = DUPLEX_FULL;
1419 cmd->port = PORT_TP;
1420 cmd->phy_address = 0;
1421 cmd->transceiver = XCVR_INTERNAL;
1422 cmd->autoneg = AUTONEG_DISABLE;
1423 cmd->maxtxpkt = 0;
1424 cmd->maxrxpkt = 0;
1425 return 0;
1426}
1427
1428static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1429{
1430 struct tun_struct *tun = netdev_priv(dev);
1431
1432 strcpy(info->driver, DRV_NAME);
1433 strcpy(info->version, DRV_VERSION);
1434 strcpy(info->fw_version, "N/A");
1435
1436 switch (tun->flags & TUN_TYPE_MASK) {
1437 case TUN_TUN_DEV:
1438 strcpy(info->bus_info, "tun");
1439 break;
1440 case TUN_TAP_DEV:
1441 strcpy(info->bus_info, "tap");
1442 break;
1443 }
1444}
1445
1446static u32 tun_get_msglevel(struct net_device *dev)
1447{
1448#ifdef TUN_DEBUG
1449 struct tun_struct *tun = netdev_priv(dev);
1450 return tun->debug;
1451#else
1452 return -EOPNOTSUPP;
1453#endif
1454}
1455
1456static void tun_set_msglevel(struct net_device *dev, u32 value)
1457{
1458#ifdef TUN_DEBUG
1459 struct tun_struct *tun = netdev_priv(dev);
1460 tun->debug = value;
1461#endif
1462}
1463
1464static u32 tun_get_link(struct net_device *dev)
1465{
1466 struct tun_struct *tun = netdev_priv(dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001467 return !!tun->tfile;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468}
1469
1470static u32 tun_get_rx_csum(struct net_device *dev)
1471{
1472 struct tun_struct *tun = netdev_priv(dev);
1473 return (tun->flags & TUN_NOCHECKSUM) == 0;
1474}
1475
1476static int tun_set_rx_csum(struct net_device *dev, u32 data)
1477{
1478 struct tun_struct *tun = netdev_priv(dev);
1479 if (data)
1480 tun->flags &= ~TUN_NOCHECKSUM;
1481 else
1482 tun->flags |= TUN_NOCHECKSUM;
1483 return 0;
1484}
1485
Jeff Garzik7282d492006-09-13 14:30:00 -04001486static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 .get_settings = tun_get_settings,
1488 .get_drvinfo = tun_get_drvinfo,
1489 .get_msglevel = tun_get_msglevel,
1490 .set_msglevel = tun_set_msglevel,
1491 .get_link = tun_get_link,
1492 .get_rx_csum = tun_get_rx_csum,
1493 .set_rx_csum = tun_set_rx_csum
1494};
1495
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001496
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497static int __init tun_init(void)
1498{
1499 int ret = 0;
1500
1501 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1502 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1503
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001504 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001505 if (ret) {
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001506 printk(KERN_ERR "tun: Can't register link_ops\n");
1507 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001508 }
1509
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001511 if (ret) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001513 goto err_misc;
1514 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001515 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001516err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001517 rtnl_link_unregister(&tun_link_ops);
1518err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 return ret;
1520}
1521
1522static void tun_cleanup(void)
1523{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001524 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001525 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526}
1527
1528module_init(tun_init);
1529module_exit(tun_cleanup);
1530MODULE_DESCRIPTION(DRV_DESCRIPTION);
1531MODULE_AUTHOR(DRV_COPYRIGHT);
1532MODULE_LICENSE("GPL");
1533MODULE_ALIAS_MISCDEV(TUN_MINOR);