blob: 0873cdcf39bebb09d75b1cadbffce4c2d614f593 [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>
Joe Perches344dc8e2012-07-12 19:33:09 +000025 * Use eth_random_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>
Daniel Wagnerf3419802012-09-12 16:12:01 +020071#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#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
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000104#define GOODCOPY_LEN 128
105
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700106#define FLT_EXACT_COUNT 8
107struct tap_filter {
108 unsigned int count; /* Number of addrs. Zero means disabled */
109 u32 mask[2]; /* Mask of the hashed addrs */
110 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
111};
112
Eric W. Biederman631ab462009-01-20 11:00:40 +0000113struct tun_file {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000114 atomic_t count;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000115 struct tun_struct *tun;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +0000116 struct net *net;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000117};
118
Herbert Xu33dccbb2009-02-05 21:25:32 -0800119struct tun_sock;
120
Rusty Russell14daa022008-04-12 18:48:58 -0700121struct tun_struct {
Eric W. Biederman631ab462009-01-20 11:00:40 +0000122 struct tun_file *tfile;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700123 unsigned int flags;
Eric W. Biederman0625c882012-02-07 16:48:55 -0800124 kuid_t owner;
125 kgid_t group;
Rusty Russell14daa022008-04-12 18:48:58 -0700126
Rusty Russell14daa022008-04-12 18:48:58 -0700127 struct net_device *dev;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000128 netdev_features_t set_features;
Michał Mirosław88255372011-04-19 06:13:10 +0000129#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
130 NETIF_F_TSO6|NETIF_F_UFO)
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700131 struct fasync_struct *fasync;
Rusty Russell14daa022008-04-12 18:48:58 -0700132
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700133 struct tap_filter txflt;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800134 struct socket socket;
Eric Dumazet43815482010-04-29 11:01:49 +0000135 struct socket_wq wq;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200136
137 int vnet_hdr_sz;
138
Rusty Russell14daa022008-04-12 18:48:58 -0700139#ifdef TUN_DEBUG
140 int debug;
141#endif
142};
143
Herbert Xu33dccbb2009-02-05 21:25:32 -0800144struct tun_sock {
145 struct sock sk;
146 struct tun_struct *tun;
147};
148
149static inline struct tun_sock *tun_sk(struct sock *sk)
150{
151 return container_of(sk, struct tun_sock, sk);
152}
153
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000154static int tun_attach(struct tun_struct *tun, struct file *file)
155{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000156 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000157 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000158
159 ASSERT_RTNL();
160
Eric W. Biederman38231b72009-01-20 11:02:28 +0000161 netif_tx_lock_bh(tun->dev);
162
163 err = -EINVAL;
164 if (tfile->tun)
165 goto out;
166
167 err = -EBUSY;
168 if (tun->tfile)
169 goto out;
170
171 err = 0;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000172 tfile->tun = tun;
173 tun->tfile = tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000174 tun->socket.file = file;
Nolan Leakebee31362010-07-27 13:53:43 +0000175 netif_carrier_on(tun->dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000176 dev_hold(tun->dev);
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000177 sock_hold(tun->socket.sk);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000178 atomic_inc(&tfile->count);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000179
Eric W. Biederman38231b72009-01-20 11:02:28 +0000180out:
181 netif_tx_unlock_bh(tun->dev);
182 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000183}
184
Eric W. Biederman631ab462009-01-20 11:00:40 +0000185static void __tun_detach(struct tun_struct *tun)
186{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000187 /* Detach from net device */
Eric W. Biederman38231b72009-01-20 11:02:28 +0000188 netif_tx_lock_bh(tun->dev);
Nolan Leakebee31362010-07-27 13:53:43 +0000189 netif_carrier_off(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000190 tun->tfile = NULL;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000191 netif_tx_unlock_bh(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000192
193 /* Drop read queue */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000194 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000195
196 /* Drop the extra count on the net device */
197 dev_put(tun->dev);
198}
199
200static void tun_detach(struct tun_struct *tun)
201{
202 rtnl_lock();
203 __tun_detach(tun);
204 rtnl_unlock();
Eric W. Biederman631ab462009-01-20 11:00:40 +0000205}
206
207static struct tun_struct *__tun_get(struct tun_file *tfile)
208{
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000209 struct tun_struct *tun = NULL;
210
211 if (atomic_inc_not_zero(&tfile->count))
212 tun = tfile->tun;
213
214 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000215}
216
217static struct tun_struct *tun_get(struct file *file)
218{
219 return __tun_get(file->private_data);
220}
221
222static void tun_put(struct tun_struct *tun)
223{
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000224 struct tun_file *tfile = tun->tfile;
225
226 if (atomic_dec_and_test(&tfile->count))
227 tun_detach(tfile->tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000228}
229
Joe Perches6b8a66e2011-03-02 07:18:10 +0000230/* TAP filtering */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700231static void addr_hash_set(u32 *mask, const u8 *addr)
232{
233 int n = ether_crc(ETH_ALEN, addr) >> 26;
234 mask[n >> 5] |= (1 << (n & 31));
235}
236
237static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
238{
239 int n = ether_crc(ETH_ALEN, addr) >> 26;
240 return mask[n >> 5] & (1 << (n & 31));
241}
242
243static int update_filter(struct tap_filter *filter, void __user *arg)
244{
245 struct { u8 u[ETH_ALEN]; } *addr;
246 struct tun_filter uf;
247 int err, alen, n, nexact;
248
249 if (copy_from_user(&uf, arg, sizeof(uf)))
250 return -EFAULT;
251
252 if (!uf.count) {
253 /* Disabled */
254 filter->count = 0;
255 return 0;
256 }
257
258 alen = ETH_ALEN * uf.count;
259 addr = kmalloc(alen, GFP_KERNEL);
260 if (!addr)
261 return -ENOMEM;
262
263 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
264 err = -EFAULT;
265 goto done;
266 }
267
268 /* The filter is updated without holding any locks. Which is
269 * perfectly safe. We disable it first and in the worst
270 * case we'll accept a few undesired packets. */
271 filter->count = 0;
272 wmb();
273
274 /* Use first set of addresses as an exact filter */
275 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
276 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
277
278 nexact = n;
279
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800280 /* Remaining multicast addresses are hashed,
281 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700282 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800283 for (; n < uf.count; n++) {
284 if (!is_multicast_ether_addr(addr[n].u)) {
285 err = 0; /* no filter */
286 goto done;
287 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700288 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800289 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700290
291 /* For ALLMULTI just set the mask to all ones.
292 * This overrides the mask populated above. */
293 if ((uf.flags & TUN_FLT_ALLMULTI))
294 memset(filter->mask, ~0, sizeof(filter->mask));
295
296 /* Now enable the filter */
297 wmb();
298 filter->count = nexact;
299
300 /* Return the number of exact filters */
301 err = nexact;
302
303done:
304 kfree(addr);
305 return err;
306}
307
308/* Returns: 0 - drop, !=0 - accept */
309static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
310{
311 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
312 * at this point. */
313 struct ethhdr *eh = (struct ethhdr *) skb->data;
314 int i;
315
316 /* Exact match */
317 for (i = 0; i < filter->count; i++)
Joe Perches2e42e472012-05-09 17:17:46 +0000318 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700319 return 1;
320
321 /* Inexact match (multicast only) */
322 if (is_multicast_ether_addr(eh->h_dest))
323 return addr_hash_test(filter->mask, eh->h_dest);
324
325 return 0;
326}
327
328/*
329 * Checks whether the packet is accepted or not.
330 * Returns: 0 - drop, !=0 - accept
331 */
332static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
333{
334 if (!filter->count)
335 return 1;
336
337 return run_filter(filter, skb);
338}
339
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340/* Network device part of the driver */
341
Jeff Garzik7282d492006-09-13 14:30:00 -0400342static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000344/* Net device detach from fd. */
345static void tun_net_uninit(struct net_device *dev)
346{
347 struct tun_struct *tun = netdev_priv(dev);
348 struct tun_file *tfile = tun->tfile;
349
350 /* Inform the methods they need to stop using the dev.
351 */
352 if (tfile) {
Eric Dumazet43815482010-04-29 11:01:49 +0000353 wake_up_all(&tun->wq.wait);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000354 if (atomic_dec_and_test(&tfile->count))
355 __tun_detach(tun);
356 }
357}
358
Herbert Xu9c3fea62009-04-18 14:15:52 +0000359static void tun_free_netdev(struct net_device *dev)
360{
361 struct tun_struct *tun = netdev_priv(dev);
362
Mikulas Patockab09e7862012-07-19 06:13:36 +0000363 BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED, &tun->socket.flags));
364
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +0000365 sk_release_kernel(tun->socket.sk);
Herbert Xu9c3fea62009-04-18 14:15:52 +0000366}
367
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368/* Net device open. */
369static int tun_net_open(struct net_device *dev)
370{
371 netif_start_queue(dev);
372 return 0;
373}
374
375/* Net device close. */
376static int tun_net_close(struct net_device *dev)
377{
378 netif_stop_queue(dev);
379 return 0;
380}
381
382/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000383static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384{
385 struct tun_struct *tun = netdev_priv(dev);
386
Joe Perches6b8a66e2011-03-02 07:18:10 +0000387 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389 /* Drop packet if interface is not attached */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000390 if (!tun->tfile)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 goto drop;
392
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700393 /* Drop if the filter does not like it.
394 * This is a noop if the filter is disabled.
395 * Filter can be enabled only for the TAP devices. */
396 if (!check_filter(&tun->txflt, skb))
397 goto drop;
398
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000399 if (tun->socket.sk->sk_filter &&
400 sk_filter(tun->socket.sk, skb))
401 goto drop;
402
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000403 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 if (!(tun->flags & TUN_ONE_QUEUE)) {
405 /* Normal queueing mode. */
406 /* Packet scheduler handles dropping of further packets. */
407 netif_stop_queue(dev);
408
409 /* We won't see all dropped packets individually, so overrun
410 * error is more appropriate. */
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700411 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 } else {
413 /* Single queue mode.
414 * Driver handles dropping of all packets itself. */
415 goto drop;
416 }
417 }
418
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000419 /* Orphan the skb - required as we might hang on to it
420 * for indefinite time. */
Michael S. Tsirkin868eefe2012-07-20 09:23:14 +0000421 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
422 goto drop;
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000423 skb_orphan(skb);
424
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700425 /* Enqueue packet */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000426 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
428 /* Notify and wake up reader process */
429 if (tun->flags & TUN_FASYNC)
430 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
Eric Dumazet43815482010-04-29 11:01:49 +0000431 wake_up_interruptible_poll(&tun->wq.wait, POLLIN |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000432 POLLRDNORM | POLLRDBAND);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000433 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
435drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700436 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000438 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439}
440
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700441static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700443 /*
444 * This callback is supposed to deal with mc filter in
445 * _rx_ path and has nothing to do with the _tx_ path.
446 * In rx path we always accept everything userspace gives us.
447 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448}
449
Ed Swierk4885a502007-09-16 12:21:38 -0700450#define MIN_MTU 68
451#define MAX_MTU 65535
452
453static int
454tun_net_change_mtu(struct net_device *dev, int new_mtu)
455{
456 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
457 return -EINVAL;
458 dev->mtu = new_mtu;
459 return 0;
460}
461
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000462static netdev_features_t tun_net_fix_features(struct net_device *dev,
463 netdev_features_t features)
Michał Mirosław88255372011-04-19 06:13:10 +0000464{
465 struct tun_struct *tun = netdev_priv(dev);
466
467 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
468}
Neil Hormanbebd0972011-06-15 05:25:01 +0000469#ifdef CONFIG_NET_POLL_CONTROLLER
470static void tun_poll_controller(struct net_device *dev)
471{
472 /*
473 * Tun only receives frames when:
474 * 1) the char device endpoint gets data from user space
475 * 2) the tun socket gets a sendmsg call from user space
476 * Since both of those are syncronous operations, we are guaranteed
477 * never to have pending data when we poll for it
478 * so theres nothing to do here but return.
479 * We need this though so netpoll recognizes us as an interface that
480 * supports polling, which enables bridge devices in virt setups to
481 * still use netconsole
482 */
483 return;
484}
485#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800486static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000487 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800488 .ndo_open = tun_net_open,
489 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800490 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800491 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000492 .ndo_fix_features = tun_net_fix_features,
Neil Hormanbebd0972011-06-15 05:25:01 +0000493#ifdef CONFIG_NET_POLL_CONTROLLER
494 .ndo_poll_controller = tun_poll_controller,
495#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800496};
497
498static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000499 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800500 .ndo_open = tun_net_open,
501 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800502 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800503 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000504 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +0000505 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800506 .ndo_set_mac_address = eth_mac_addr,
507 .ndo_validate_addr = eth_validate_addr,
Neil Hormanbebd0972011-06-15 05:25:01 +0000508#ifdef CONFIG_NET_POLL_CONTROLLER
509 .ndo_poll_controller = tun_poll_controller,
510#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800511};
512
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513/* Initialize net device. */
514static void tun_net_init(struct net_device *dev)
515{
516 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400517
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 switch (tun->flags & TUN_TYPE_MASK) {
519 case TUN_TUN_DEV:
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800520 dev->netdev_ops = &tun_netdev_ops;
521
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 /* Point-to-Point TUN Device */
523 dev->hard_header_len = 0;
524 dev->addr_len = 0;
525 dev->mtu = 1500;
526
527 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400528 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
530 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
531 break;
532
533 case TUN_TAP_DEV:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -0800534 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700536 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +0000537 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Danny Kukawkaf2cedb62012-02-15 06:45:39 +0000539 eth_hw_addr_random(dev);
Brian Braunstein36226a82007-04-26 01:00:55 -0700540
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
542 break;
543 }
544}
545
546/* Character device part */
547
548/* Poll */
549static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400550{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +0000551 struct tun_file *tfile = file->private_data;
552 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000553 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800554 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555
556 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +0000557 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000559 sk = tun->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000560
Joe Perches6b8a66e2011-03-02 07:18:10 +0000561 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
Eric Dumazet43815482010-04-29 11:01:49 +0000563 poll_wait(file, &tun->wq.wait, wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400564
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000565 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 mask |= POLLIN | POLLRDNORM;
567
Herbert Xu33dccbb2009-02-05 21:25:32 -0800568 if (sock_writeable(sk) ||
569 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
570 sock_writeable(sk)))
571 mask |= POLLOUT | POLLWRNORM;
572
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000573 if (tun->dev->reg_state != NETREG_REGISTERED)
574 mask = POLLERR;
575
Eric W. Biederman631ab462009-01-20 11:00:40 +0000576 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 return mask;
578}
579
Rusty Russellf42157c2008-08-15 15:15:10 -0700580/* prepad is the amount to reserve at front. len is length after that.
581 * linear is a hint as to how much to copy (usually headers). */
stephen hemminger6f7c1562011-06-08 14:33:08 +0000582static struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
583 size_t prepad, size_t len,
584 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -0700585{
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000586 struct sock *sk = tun->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -0700587 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800588 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -0700589
Herbert Xu82862742010-05-24 00:14:10 -0700590 sock_update_classid(sk);
591
Rusty Russellf42157c2008-08-15 15:15:10 -0700592 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -0700593 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800594 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -0700595
Herbert Xu33dccbb2009-02-05 21:25:32 -0800596 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
597 &err);
Rusty Russellf42157c2008-08-15 15:15:10 -0700598 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800599 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -0700600
601 skb_reserve(skb, prepad);
602 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800603 skb->data_len = len - linear;
604 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -0700605
606 return skb;
607}
608
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000609/* set skb frags from iovec, this can move to core network code for reuse */
610static int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *from,
611 int offset, size_t count)
612{
613 int len = iov_length(from, count) - offset;
614 int copy = skb_headlen(skb);
615 int size, offset1 = 0;
616 int i = 0;
617
618 /* Skip over from offset */
619 while (count && (offset >= from->iov_len)) {
620 offset -= from->iov_len;
621 ++from;
622 --count;
623 }
624
625 /* copy up to skb headlen */
626 while (count && (copy > 0)) {
627 size = min_t(unsigned int, copy, from->iov_len - offset);
628 if (copy_from_user(skb->data + offset1, from->iov_base + offset,
629 size))
630 return -EFAULT;
631 if (copy > size) {
632 ++from;
633 --count;
634 offset = 0;
635 } else
636 offset += size;
637 copy -= size;
638 offset1 += size;
639 }
640
641 if (len == offset1)
642 return 0;
643
644 while (count--) {
645 struct page *page[MAX_SKB_FRAGS];
646 int num_pages;
647 unsigned long base;
648 unsigned long truesize;
649
650 len = from->iov_len - offset;
651 if (!len) {
652 offset = 0;
653 ++from;
654 continue;
655 }
656 base = (unsigned long)from->iov_base + offset;
657 size = ((base & ~PAGE_MASK) + len + ~PAGE_MASK) >> PAGE_SHIFT;
658 if (i + size > MAX_SKB_FRAGS)
659 return -EMSGSIZE;
660 num_pages = get_user_pages_fast(base, size, 0, &page[i]);
661 if (num_pages != size) {
662 for (i = 0; i < num_pages; i++)
663 put_page(page[i]);
664 return -EFAULT;
665 }
666 truesize = size * PAGE_SIZE;
667 skb->data_len += len;
668 skb->len += len;
669 skb->truesize += truesize;
670 atomic_add(truesize, &skb->sk->sk_wmem_alloc);
671 while (len) {
672 int off = base & ~PAGE_MASK;
673 int size = min_t(int, len, PAGE_SIZE - off);
674 __skb_fill_page_desc(skb, i, page[i], off, size);
675 skb_shinfo(skb)->nr_frags++;
676 /* increase sk_wmem_alloc */
677 base += size;
678 len -= size;
679 i++;
680 }
681 offset = 0;
682 ++from;
683 }
684 return 0;
685}
686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687/* Get packet from user space buffer */
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000688static ssize_t tun_get_user(struct tun_struct *tun, void *msg_control,
689 const struct iovec *iv, size_t total_len,
690 size_t count, int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691{
Harvey Harrison09640e62009-02-01 00:45:17 -0800692 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 struct sk_buff *skb;
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000694 size_t len = total_len, align = NET_SKB_PAD;
Rusty Russellf43798c2008-07-03 03:48:02 -0700695 struct virtio_net_hdr gso = { 0 };
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000696 int offset = 0;
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000697 int copylen;
698 bool zerocopy = false;
699 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
701 if (!(tun->flags & TUN_NO_PI)) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000702 if ((len -= sizeof(pi)) > total_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return -EINVAL;
704
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000705 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 return -EFAULT;
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000707 offset += sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 }
709
Rusty Russellf43798c2008-07-03 03:48:02 -0700710 if (tun->flags & TUN_VNET_HDR) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000711 if ((len -= tun->vnet_hdr_sz) > total_len)
Rusty Russellf43798c2008-07-03 03:48:02 -0700712 return -EINVAL;
713
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +0000714 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
Rusty Russellf43798c2008-07-03 03:48:02 -0700715 return -EFAULT;
716
Herbert Xu49091222009-06-08 00:20:01 -0700717 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
718 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
719 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
720
Rusty Russellf43798c2008-07-03 03:48:02 -0700721 if (gso.hdr_len > len)
722 return -EINVAL;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200723 offset += tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -0700724 }
725
Rusty Russelle01bf1c2008-04-12 18:49:30 -0700726 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
stephen hemmingera504b862011-06-08 14:33:07 +0000727 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -0700728 if (unlikely(len < ETH_HLEN ||
729 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -0700730 return -EINVAL;
731 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400732
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000733 if (msg_control)
734 zerocopy = true;
735
736 if (zerocopy) {
737 /* Userspace may produce vectors with count greater than
738 * MAX_SKB_FRAGS, so we need to linearize parts of the skb
739 * to let the rest of data to be fit in the frags.
740 */
741 if (count > MAX_SKB_FRAGS) {
742 copylen = iov_length(iv, count - MAX_SKB_FRAGS);
743 if (copylen < offset)
744 copylen = 0;
745 else
746 copylen -= offset;
747 } else
748 copylen = 0;
749 /* There are 256 bytes to be copied in skb, so there is enough
750 * room for skb expand head in case it is used.
751 * The rest of the buffer is mapped from userspace.
752 */
753 if (copylen < gso.hdr_len)
754 copylen = gso.hdr_len;
755 if (!copylen)
756 copylen = GOODCOPY_LEN;
757 } else
758 copylen = len;
759
760 skb = tun_alloc_skb(tun, align, copylen, gso.hdr_len, noblock);
Herbert Xu33dccbb2009-02-05 21:25:32 -0800761 if (IS_ERR(skb)) {
762 if (PTR_ERR(skb) != -EAGAIN)
763 tun->dev->stats.rx_dropped++;
764 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 }
766
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000767 if (zerocopy)
768 err = zerocopy_sg_from_iovec(skb, iv, offset, count);
769 else
770 err = skb_copy_datagram_from_iovec(skb, 0, iv, offset, len);
771
772 if (err) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700773 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -0800774 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -0800776 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777
Rusty Russellf43798c2008-07-03 03:48:02 -0700778 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
779 if (!skb_partial_csum_set(skb, gso.csum_start,
780 gso.csum_offset)) {
781 tun->dev->stats.rx_frame_errors++;
782 kfree_skb(skb);
783 return -EINVAL;
784 }
Michał Mirosław88255372011-04-19 06:13:10 +0000785 }
Rusty Russellf43798c2008-07-03 03:48:02 -0700786
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 switch (tun->flags & TUN_TYPE_MASK) {
788 case TUN_TUN_DEV:
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -0700789 if (tun->flags & TUN_NO_PI) {
790 switch (skb->data[0] & 0xf0) {
791 case 0x40:
792 pi.proto = htons(ETH_P_IP);
793 break;
794 case 0x60:
795 pi.proto = htons(ETH_P_IPV6);
796 break;
797 default:
798 tun->dev->stats.rx_dropped++;
799 kfree_skb(skb);
800 return -EINVAL;
801 }
802 }
803
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -0700804 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -0700806 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 break;
808 case TUN_TAP_DEV:
809 skb->protocol = eth_type_trans(skb, tun->dev);
810 break;
Joe Perches6403eab2011-06-03 11:51:20 +0000811 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Rusty Russellf43798c2008-07-03 03:48:02 -0700813 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
814 pr_debug("GSO!\n");
815 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
816 case VIRTIO_NET_HDR_GSO_TCPV4:
817 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
818 break;
819 case VIRTIO_NET_HDR_GSO_TCPV6:
820 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
821 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +0000822 case VIRTIO_NET_HDR_GSO_UDP:
823 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
824 break;
Rusty Russellf43798c2008-07-03 03:48:02 -0700825 default:
826 tun->dev->stats.rx_frame_errors++;
827 kfree_skb(skb);
828 return -EINVAL;
829 }
830
831 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
832 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
833
834 skb_shinfo(skb)->gso_size = gso.gso_size;
835 if (skb_shinfo(skb)->gso_size == 0) {
836 tun->dev->stats.rx_frame_errors++;
837 kfree_skb(skb);
838 return -EINVAL;
839 }
840
841 /* Header must be checked, and gso_segs computed. */
842 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
843 skb_shinfo(skb)->gso_segs = 0;
844 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400845
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000846 /* copy skb_ubuf_info for callback when skb has no error */
847 if (zerocopy) {
848 skb_shinfo(skb)->destructor_arg = msg_control;
849 skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
850 }
851
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400853
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700854 tun->dev->stats.rx_packets++;
855 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000857 return total_len;
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400858}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700860static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
861 unsigned long count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862{
Herbert Xu33dccbb2009-02-05 21:25:32 -0800863 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -0800864 struct tun_struct *tun = tun_get(file);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000865 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
867 if (!tun)
868 return -EBADFD;
869
Joe Perches6b8a66e2011-03-02 07:18:10 +0000870 tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000872 result = tun_get_user(tun, NULL, iv, iov_length(iv, count), count,
Herbert Xu33dccbb2009-02-05 21:25:32 -0800873 file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000874
875 tun_put(tun);
876 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877}
878
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +0000880static ssize_t tun_put_user(struct tun_struct *tun,
881 struct sk_buff *skb,
882 const struct iovec *iv, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883{
884 struct tun_pi pi = { 0, skb->protocol };
885 ssize_t total = 0;
886
887 if (!(tun->flags & TUN_NO_PI)) {
888 if ((len -= sizeof(pi)) < 0)
889 return -EINVAL;
890
891 if (len < skb->len) {
892 /* Packet will be striped */
893 pi.flags |= TUN_PKT_STRIP;
894 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400895
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000896 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 return -EFAULT;
898 total += sizeof(pi);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400899 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900
Rusty Russellf43798c2008-07-03 03:48:02 -0700901 if (tun->flags & TUN_VNET_HDR) {
902 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200903 if ((len -= tun->vnet_hdr_sz) < 0)
Rusty Russellf43798c2008-07-03 03:48:02 -0700904 return -EINVAL;
905
906 if (skb_is_gso(skb)) {
907 struct skb_shared_info *sinfo = skb_shinfo(skb);
908
909 /* This is a hint as to how much should be linear. */
910 gso.hdr_len = skb_headlen(skb);
911 gso.gso_size = sinfo->gso_size;
912 if (sinfo->gso_type & SKB_GSO_TCPV4)
913 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
914 else if (sinfo->gso_type & SKB_GSO_TCPV6)
915 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Sridhar Samudralae36aa252009-07-14 14:21:04 +0000916 else if (sinfo->gso_type & SKB_GSO_UDP)
917 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +0000918 else {
Joe Perches6b8a66e2011-03-02 07:18:10 +0000919 pr_err("unexpected GSO type: "
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +0000920 "0x%x, gso_size %d, hdr_len %d\n",
921 sinfo->gso_type, gso.gso_size,
922 gso.hdr_len);
923 print_hex_dump(KERN_ERR, "tun: ",
924 DUMP_PREFIX_NONE,
925 16, 1, skb->head,
926 min((int)gso.hdr_len, 64), true);
927 WARN_ON_ONCE(1);
928 return -EINVAL;
929 }
Rusty Russellf43798c2008-07-03 03:48:02 -0700930 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
931 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
932 } else
933 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
934
935 if (skb->ip_summed == CHECKSUM_PARTIAL) {
936 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
Michał Mirosław55508d62010-12-14 15:24:08 +0000937 gso.csum_start = skb_checksum_start_offset(skb);
Rusty Russellf43798c2008-07-03 03:48:02 -0700938 gso.csum_offset = skb->csum_offset;
Jason Wang10a8d942011-06-10 00:56:17 +0000939 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
940 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
Rusty Russellf43798c2008-07-03 03:48:02 -0700941 } /* else everything is zero */
942
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000943 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
944 sizeof(gso))))
Rusty Russellf43798c2008-07-03 03:48:02 -0700945 return -EFAULT;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200946 total += tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -0700947 }
948
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 len = min_t(int, skb->len, len);
950
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000951 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000952 total += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700954 tun->dev->stats.tx_packets++;
955 tun->dev->stats.tx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956
957 return total;
958}
959
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000960static ssize_t tun_do_read(struct tun_struct *tun,
961 struct kiocb *iocb, const struct iovec *iv,
962 ssize_t len, int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 DECLARE_WAITQUEUE(wait, current);
965 struct sk_buff *skb;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000966 ssize_t ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967
Joe Perches6b8a66e2011-03-02 07:18:10 +0000968 tun_debug(KERN_INFO, tun, "tun_chr_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
Amos Kong61a5ff12011-06-09 00:27:10 -0700970 if (unlikely(!noblock))
971 add_wait_queue(&tun->wq.wait, &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 while (len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 current->state = TASK_INTERRUPTIBLE;
974
975 /* Read frames from the queue */
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000976 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
Michael S. Tsirkin05c28282010-01-14 06:17:09 +0000977 if (noblock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 ret = -EAGAIN;
979 break;
980 }
981 if (signal_pending(current)) {
982 ret = -ERESTARTSYS;
983 break;
984 }
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000985 if (tun->dev->reg_state != NETREG_REGISTERED) {
986 ret = -EIO;
987 break;
988 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
990 /* Nothing to read, let's sleep */
991 schedule();
992 continue;
993 }
994 netif_wake_queue(tun->dev);
995
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +0000996 ret = tun_put_user(tun, skb, iv, len);
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700997 kfree_skb(skb);
998 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 }
1000
1001 current->state = TASK_RUNNING;
Amos Kong61a5ff12011-06-09 00:27:10 -07001002 if (unlikely(!noblock))
1003 remove_wait_queue(&tun->wq.wait, &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001005 return ret;
1006}
1007
1008static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
1009 unsigned long count, loff_t pos)
1010{
1011 struct file *file = iocb->ki_filp;
1012 struct tun_file *tfile = file->private_data;
1013 struct tun_struct *tun = __tun_get(tfile);
1014 ssize_t len, ret;
1015
1016 if (!tun)
1017 return -EBADFD;
1018 len = iov_length(iv, count);
1019 if (len < 0) {
1020 ret = -EINVAL;
1021 goto out;
1022 }
1023
1024 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
1025 ret = min_t(ssize_t, ret, len);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001026out:
1027 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 return ret;
1029}
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031static void tun_setup(struct net_device *dev)
1032{
1033 struct tun_struct *tun = netdev_priv(dev);
1034
Eric W. Biederman0625c882012-02-07 16:48:55 -08001035 tun->owner = INVALID_UID;
1036 tun->group = INVALID_GID;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 dev->ethtool_ops = &tun_ethtool_ops;
Herbert Xu9c3fea62009-04-18 14:15:52 +00001039 dev->destructor = tun_free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040}
1041
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001042/* Trivial set of netlink ops to allow deleting tun or tap
1043 * device with netlink.
1044 */
1045static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
1046{
1047 return -EINVAL;
1048}
1049
1050static struct rtnl_link_ops tun_link_ops __read_mostly = {
1051 .kind = DRV_NAME,
1052 .priv_size = sizeof(struct tun_struct),
1053 .setup = tun_setup,
1054 .validate = tun_validate,
1055};
1056
Herbert Xu33dccbb2009-02-05 21:25:32 -08001057static void tun_sock_write_space(struct sock *sk)
1058{
1059 struct tun_struct *tun;
Eric Dumazet43815482010-04-29 11:01:49 +00001060 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001061
1062 if (!sock_writeable(sk))
1063 return;
1064
Herbert Xu33dccbb2009-02-05 21:25:32 -08001065 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
1066 return;
1067
Eric Dumazet43815482010-04-29 11:01:49 +00001068 wqueue = sk_sleep(sk);
1069 if (wqueue && waitqueue_active(wqueue))
1070 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001071 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -07001072
Vitaliy Gusev80924e52009-12-25 07:17:43 +00001073 tun = tun_sk(sk)->tun;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001074 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
1075}
1076
1077static void tun_sock_destruct(struct sock *sk)
1078{
Vitaliy Gusev80924e52009-12-25 07:17:43 +00001079 free_netdev(tun_sk(sk)->tun->dev);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001080}
1081
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001082static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
1083 struct msghdr *m, size_t total_len)
1084{
1085 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001086 return tun_get_user(tun, m->msg_control, m->msg_iov, total_len,
1087 m->msg_iovlen, m->msg_flags & MSG_DONTWAIT);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001088}
1089
1090static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
1091 struct msghdr *m, size_t total_len,
1092 int flags)
1093{
1094 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
1095 int ret;
1096 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
1097 return -EINVAL;
1098 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
1099 flags & MSG_DONTWAIT);
1100 if (ret > total_len) {
1101 m->msg_flags |= MSG_TRUNC;
1102 ret = flags & MSG_TRUNC ? ret : total_len;
1103 }
1104 return ret;
1105}
1106
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001107static int tun_release(struct socket *sock)
1108{
1109 if (sock->sk)
1110 sock_put(sock->sk);
1111 return 0;
1112}
1113
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001114/* Ops structure to mimic raw sockets with tun */
1115static const struct proto_ops tun_socket_ops = {
1116 .sendmsg = tun_sendmsg,
1117 .recvmsg = tun_recvmsg,
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001118 .release = tun_release,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001119};
1120
Herbert Xu33dccbb2009-02-05 21:25:32 -08001121static struct proto tun_proto = {
1122 .name = "tun",
1123 .owner = THIS_MODULE,
1124 .obj_size = sizeof(struct tun_sock),
1125};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001126
David Woodhouse980c9e82009-05-09 22:54:21 -07001127static int tun_flags(struct tun_struct *tun)
1128{
1129 int flags = 0;
1130
1131 if (tun->flags & TUN_TUN_DEV)
1132 flags |= IFF_TUN;
1133 else
1134 flags |= IFF_TAP;
1135
1136 if (tun->flags & TUN_NO_PI)
1137 flags |= IFF_NO_PI;
1138
1139 if (tun->flags & TUN_ONE_QUEUE)
1140 flags |= IFF_ONE_QUEUE;
1141
1142 if (tun->flags & TUN_VNET_HDR)
1143 flags |= IFF_VNET_HDR;
1144
1145 return flags;
1146}
1147
1148static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1149 char *buf)
1150{
1151 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1152 return sprintf(buf, "0x%x\n", tun_flags(tun));
1153}
1154
1155static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1156 char *buf)
1157{
1158 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001159 return uid_valid(tun->owner)?
1160 sprintf(buf, "%u\n",
1161 from_kuid_munged(current_user_ns(), tun->owner)):
1162 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001163}
1164
1165static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1166 char *buf)
1167{
1168 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001169 return gid_valid(tun->group) ?
1170 sprintf(buf, "%u\n",
1171 from_kgid_munged(current_user_ns(), tun->group)):
1172 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001173}
1174
1175static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1176static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1177static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1178
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001179static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001181 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 struct tun_struct *tun;
1183 struct net_device *dev;
1184 int err;
1185
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001186 dev = __dev_get_by_name(net, ifr->ifr_name);
1187 if (dev) {
Paul Moore2b980db2009-08-28 18:12:43 -04001188 const struct cred *cred = current_cred();
1189
David Woodhousef85ba782009-04-27 03:23:54 -07001190 if (ifr->ifr_flags & IFF_TUN_EXCL)
1191 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001192 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1193 tun = netdev_priv(dev);
1194 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1195 tun = netdev_priv(dev);
1196 else
1197 return -EINVAL;
1198
Eric W. Biederman0625c882012-02-07 16:48:55 -08001199 if (((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
1200 (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
Paul Moore2b980db2009-08-28 18:12:43 -04001201 !capable(CAP_NET_ADMIN))
1202 return -EPERM;
Linus Torvaldsd7e96602009-09-14 10:37:28 -07001203 err = security_tun_dev_attach(tun->socket.sk);
Paul Moore2b980db2009-08-28 18:12:43 -04001204 if (err < 0)
1205 return err;
1206
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001207 err = tun_attach(tun, file);
1208 if (err < 0)
1209 return err;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001210 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 else {
1212 char *name;
1213 unsigned long flags = 0;
1214
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001215 if (!capable(CAP_NET_ADMIN))
1216 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001217 err = security_tun_dev_create();
1218 if (err < 0)
1219 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001220
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 /* Set dev type */
1222 if (ifr->ifr_flags & IFF_TUN) {
1223 /* TUN device */
1224 flags |= TUN_TUN_DEV;
1225 name = "tun%d";
1226 } else if (ifr->ifr_flags & IFF_TAP) {
1227 /* TAP device */
1228 flags |= TUN_TAP_DEV;
1229 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001230 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001231 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001232
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 if (*ifr->ifr_name)
1234 name = ifr->ifr_name;
1235
1236 dev = alloc_netdev(sizeof(struct tun_struct), name,
1237 tun_setup);
1238 if (!dev)
1239 return -ENOMEM;
1240
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001241 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001242 dev->rtnl_link_ops = &tun_link_ops;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001243
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 tun = netdev_priv(dev);
1245 tun->dev = dev;
1246 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001247 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001248 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Mikulas Patockab09e7862012-07-19 06:13:36 +00001249 set_bit(SOCK_EXTERNALLY_ALLOCATED, &tun->socket.flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250
Herbert Xu33dccbb2009-02-05 21:25:32 -08001251 err = -ENOMEM;
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001252 sk = sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001253 if (!sk)
1254 goto err_free_dev;
1255
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001256 sk_change_net(sk, net);
Eric Dumazet43815482010-04-29 11:01:49 +00001257 tun->socket.wq = &tun->wq;
1258 init_waitqueue_head(&tun->wq.wait);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001259 tun->socket.ops = &tun_socket_ops;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001260 sock_init_data(&tun->socket, sk);
1261 sk->sk_write_space = tun_sock_write_space;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001262 sk->sk_sndbuf = INT_MAX;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001263 sock_set_flag(sk, SOCK_ZEROCOPY);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001264
Vitaliy Gusev80924e52009-12-25 07:17:43 +00001265 tun_sk(sk)->tun = tun;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001266
Paul Moore2b980db2009-08-28 18:12:43 -04001267 security_tun_dev_post_create(sk);
1268
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 tun_net_init(dev);
1270
Michał Mirosław88255372011-04-19 06:13:10 +00001271 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
1272 TUN_USER_FEATURES;
1273 dev->features = dev->hw_features;
1274
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 err = register_netdevice(tun->dev);
1276 if (err < 0)
Herbert Xu9c3fea62009-04-18 14:15:52 +00001277 goto err_free_sk;
1278
David Woodhouse980c9e82009-05-09 22:54:21 -07001279 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1280 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1281 device_create_file(&tun->dev->dev, &dev_attr_group))
Joe Perches6b8a66e2011-03-02 07:18:10 +00001282 pr_err("Failed to create tun sysfs files\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001283
Herbert Xu9c3fea62009-04-18 14:15:52 +00001284 sk->sk_destruct = tun_sock_destruct;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001285
1286 err = tun_attach(tun, file);
1287 if (err < 0)
Herbert Xu9c3fea62009-04-18 14:15:52 +00001288 goto failed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 }
1290
Joe Perches6b8a66e2011-03-02 07:18:10 +00001291 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
1293 if (ifr->ifr_flags & IFF_NO_PI)
1294 tun->flags |= TUN_NO_PI;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001295 else
1296 tun->flags &= ~TUN_NO_PI;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
1298 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1299 tun->flags |= TUN_ONE_QUEUE;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001300 else
1301 tun->flags &= ~TUN_ONE_QUEUE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302
Rusty Russellf43798c2008-07-03 03:48:02 -07001303 if (ifr->ifr_flags & IFF_VNET_HDR)
1304 tun->flags |= TUN_VNET_HDR;
1305 else
1306 tun->flags &= ~TUN_VNET_HDR;
1307
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001308 /* Make sure persistent devices do not get stuck in
1309 * xoff state.
1310 */
1311 if (netif_running(tun->dev))
1312 netif_wake_queue(tun->dev);
1313
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 strcpy(ifr->ifr_name, tun->dev->name);
1315 return 0;
1316
Herbert Xu33dccbb2009-02-05 21:25:32 -08001317 err_free_sk:
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001318 tun_free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 err_free_dev:
1320 free_netdev(dev);
1321 failed:
1322 return err;
1323}
1324
Herbert Xu876bfd42009-08-06 14:22:44 +00001325static int tun_get_iff(struct net *net, struct tun_struct *tun,
1326 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001327{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001328 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001329
1330 strcpy(ifr->ifr_name, tun->dev->name);
1331
David Woodhouse980c9e82009-05-09 22:54:21 -07001332 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001333
1334 return 0;
1335}
1336
Rusty Russell5228ddc2008-07-03 03:46:16 -07001337/* This is like a cut-down ethtool ops, except done via tun fd so no
1338 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001339static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001340{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001341 netdev_features_t features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001342
1343 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001344 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001345 arg &= ~TUN_F_CSUM;
1346
1347 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1348 if (arg & TUN_F_TSO_ECN) {
1349 features |= NETIF_F_TSO_ECN;
1350 arg &= ~TUN_F_TSO_ECN;
1351 }
1352 if (arg & TUN_F_TSO4)
1353 features |= NETIF_F_TSO;
1354 if (arg & TUN_F_TSO6)
1355 features |= NETIF_F_TSO6;
1356 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1357 }
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001358
1359 if (arg & TUN_F_UFO) {
1360 features |= NETIF_F_UFO;
1361 arg &= ~TUN_F_UFO;
1362 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001363 }
1364
1365 /* This gives the user a way to test for new features in future by
1366 * trying to set them. */
1367 if (arg)
1368 return -EINVAL;
1369
Michał Mirosław88255372011-04-19 06:13:10 +00001370 tun->set_features = features;
1371 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001372
1373 return 0;
1374}
1375
Arnd Bergmann50857e22009-11-06 22:52:32 -08001376static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1377 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001379 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001380 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 void __user* argp = (void __user*)arg;
Michael S. Tsirkin99405162010-02-14 01:01:10 +00001382 struct sock_fprog fprog;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 struct ifreq ifr;
Eric W. Biederman0625c882012-02-07 16:48:55 -08001384 kuid_t owner;
1385 kgid_t group;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001386 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001387 int vnet_hdr_sz;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001388 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389
Mathias Krausea117dac2012-07-29 19:45:14 +00001390 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89) {
Arnd Bergmann50857e22009-11-06 22:52:32 -08001391 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 return -EFAULT;
David S. Miller8bbb1812012-07-30 14:52:48 -07001393 } else {
Mathias Krausea117dac2012-07-29 19:45:14 +00001394 memset(&ifr, 0, sizeof(ifr));
David S. Miller8bbb1812012-07-30 14:52:48 -07001395 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001396 if (cmd == TUNGETFEATURES) {
1397 /* Currently this just means: "what IFF flags are valid?".
1398 * This is needed because we never checked for invalid flags on
1399 * TUNSETIFF. */
1400 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1401 IFF_VNET_HDR,
1402 (unsigned int __user*)argp);
1403 }
1404
Herbert Xu876bfd42009-08-06 14:22:44 +00001405 rtnl_lock();
1406
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001407 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1410
Herbert Xu876bfd42009-08-06 14:22:44 +00001411 ret = tun_set_iff(tfile->net, file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
Herbert Xu876bfd42009-08-06 14:22:44 +00001413 if (ret)
1414 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415
Arnd Bergmann50857e22009-11-06 22:52:32 -08001416 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001417 ret = -EFAULT;
1418 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 }
1420
Herbert Xu876bfd42009-08-06 14:22:44 +00001421 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001423 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
Joe Perches6b8a66e2011-03-02 07:18:10 +00001425 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %d\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Eric W. Biederman631ab462009-01-20 11:00:40 +00001427 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001429 case TUNGETIFF:
Herbert Xu876bfd42009-08-06 14:22:44 +00001430 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001431 if (ret)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001432 break;
Mark McLoughline3b99552008-08-15 15:09:56 -07001433
Arnd Bergmann50857e22009-11-06 22:52:32 -08001434 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001435 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001436 break;
1437
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 case TUNSETNOCSUM:
1439 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440
Michał Mirosław88255372011-04-19 06:13:10 +00001441 /* [unimplemented] */
1442 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00001443 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 break;
1445
1446 case TUNSETPERSIST:
1447 /* Disable/Enable persist mode */
1448 if (arg)
1449 tun->flags |= TUN_PERSIST;
1450 else
1451 tun->flags &= ~TUN_PERSIST;
1452
Joe Perches6b8a66e2011-03-02 07:18:10 +00001453 tun_debug(KERN_INFO, tun, "persist %s\n",
1454 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 break;
1456
1457 case TUNSETOWNER:
1458 /* Set owner of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001459 owner = make_kuid(current_user_ns(), arg);
1460 if (!uid_valid(owner)) {
1461 ret = -EINVAL;
1462 break;
1463 }
1464 tun->owner = owner;
1465 tun_debug(KERN_INFO, tun, "owner set to %d\n",
1466 from_kuid(&init_user_ns, tun->owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 break;
1468
Guido Guenther8c644622007-07-02 22:50:25 -07001469 case TUNSETGROUP:
1470 /* Set group of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001471 group = make_kgid(current_user_ns(), arg);
1472 if (!gid_valid(group)) {
1473 ret = -EINVAL;
1474 break;
1475 }
1476 tun->group = group;
1477 tun_debug(KERN_INFO, tun, "group set to %d\n",
1478 from_kgid(&init_user_ns, tun->group));
Guido Guenther8c644622007-07-02 22:50:25 -07001479 break;
1480
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001481 case TUNSETLINK:
1482 /* Only allow setting the type when the interface is down */
1483 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001484 tun_debug(KERN_INFO, tun,
1485 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07001486 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001487 } else {
1488 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00001489 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1490 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07001491 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001492 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001493 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495#ifdef TUN_DEBUG
1496 case TUNSETDEBUG:
1497 tun->debug = arg;
1498 break;
1499#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07001500 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00001501 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001502 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001503
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001504 case TUNSETTXFILTER:
1505 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00001506 ret = -EINVAL;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001507 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001508 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07001509 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001510 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511
1512 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04001513 /* Get hw address */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001514 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1515 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08001516 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001517 ret = -EFAULT;
1518 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
1520 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001521 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00001522 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
1523 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08001524
Kim B. Heino40102372008-02-29 12:26:21 -08001525 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001526 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001527
1528 case TUNGETSNDBUF:
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001529 sndbuf = tun->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001530 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1531 ret = -EFAULT;
1532 break;
1533
1534 case TUNSETSNDBUF:
1535 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1536 ret = -EFAULT;
1537 break;
1538 }
1539
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001540 tun->socket.sk->sk_sndbuf = sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001541 break;
1542
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001543 case TUNGETVNETHDRSZ:
1544 vnet_hdr_sz = tun->vnet_hdr_sz;
1545 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1546 ret = -EFAULT;
1547 break;
1548
1549 case TUNSETVNETHDRSZ:
1550 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1551 ret = -EFAULT;
1552 break;
1553 }
1554 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1555 ret = -EINVAL;
1556 break;
1557 }
1558
1559 tun->vnet_hdr_sz = vnet_hdr_sz;
1560 break;
1561
Michael S. Tsirkin99405162010-02-14 01:01:10 +00001562 case TUNATTACHFILTER:
1563 /* Can be set only for TAPs */
1564 ret = -EINVAL;
1565 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1566 break;
1567 ret = -EFAULT;
1568 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1569 break;
1570
1571 ret = sk_attach_filter(&fprog, tun->socket.sk);
1572 break;
1573
1574 case TUNDETACHFILTER:
1575 /* Can be set only for TAPs */
1576 ret = -EINVAL;
1577 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1578 break;
1579 ret = sk_detach_filter(tun->socket.sk);
1580 break;
1581
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00001583 ret = -EINVAL;
1584 break;
Joe Perchesee289b62010-05-17 22:47:34 -07001585 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586
Herbert Xu876bfd42009-08-06 14:22:44 +00001587unlock:
1588 rtnl_unlock();
1589 if (tun)
1590 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001591 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592}
1593
Arnd Bergmann50857e22009-11-06 22:52:32 -08001594static long tun_chr_ioctl(struct file *file,
1595 unsigned int cmd, unsigned long arg)
1596{
1597 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1598}
1599
1600#ifdef CONFIG_COMPAT
1601static long tun_chr_compat_ioctl(struct file *file,
1602 unsigned int cmd, unsigned long arg)
1603{
1604 switch (cmd) {
1605 case TUNSETIFF:
1606 case TUNGETIFF:
1607 case TUNSETTXFILTER:
1608 case TUNGETSNDBUF:
1609 case TUNSETSNDBUF:
1610 case SIOCGIFHWADDR:
1611 case SIOCSIFHWADDR:
1612 arg = (unsigned long)compat_ptr(arg);
1613 break;
1614 default:
1615 arg = (compat_ulong_t)arg;
1616 break;
1617 }
1618
1619 /*
1620 * compat_ifreq is shorter than ifreq, so we must not access beyond
1621 * the end of that structure. All fields that are used in this
1622 * driver are compatible though, we don't need to convert the
1623 * contents.
1624 */
1625 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1626}
1627#endif /* CONFIG_COMPAT */
1628
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629static int tun_chr_fasync(int fd, struct file *file, int on)
1630{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001631 struct tun_struct *tun = tun_get(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 int ret;
1633
1634 if (!tun)
1635 return -EBADFD;
1636
Joe Perches6b8a66e2011-03-02 07:18:10 +00001637 tun_debug(KERN_INFO, tun, "tun_chr_fasync %d\n", on);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638
1639 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06001640 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001641
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 if (on) {
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001643 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 if (ret)
Jonathan Corbet9d319522008-06-19 15:50:37 -06001645 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 tun->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001647 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 tun->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06001649 ret = 0;
1650out:
Eric W. Biederman631ab462009-01-20 11:00:40 +00001651 tun_put(tun);
Jonathan Corbet9d319522008-06-19 15:50:37 -06001652 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653}
1654
1655static int tun_chr_open(struct inode *inode, struct file * file)
1656{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001657 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07001658
Joe Perches6b8a66e2011-03-02 07:18:10 +00001659 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00001660
1661 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1662 if (!tfile)
1663 return -ENOMEM;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001664 atomic_set(&tfile->count, 0);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001665 tfile->tun = NULL;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001666 tfile->net = get_net(current->nsproxy->net_ns);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001667 file->private_data = tfile;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 return 0;
1669}
1670
1671static int tun_chr_close(struct inode *inode, struct file *file)
1672{
Eric W. Biederman631ab462009-01-20 11:00:40 +00001673 struct tun_file *tfile = file->private_data;
Eric W. Biedermanf0a4d0e2009-06-08 00:44:31 -07001674 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675
Eric W. Biedermanf0a4d0e2009-06-08 00:44:31 -07001676 tun = __tun_get(tfile);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001677 if (tun) {
Herbert Xud23e4362009-07-02 23:03:55 +00001678 struct net_device *dev = tun->dev;
1679
Joe Perches6b8a66e2011-03-02 07:18:10 +00001680 tun_debug(KERN_INFO, tun, "tun_chr_close\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681
Eric W. Biederman631ab462009-01-20 11:00:40 +00001682 __tun_detach(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683
Daniel Mack3ad2f3f2010-02-03 08:01:28 +08001684 /* If desirable, unregister the netdevice. */
Herbert Xud23e4362009-07-02 23:03:55 +00001685 if (!(tun->flags & TUN_PERSIST)) {
1686 rtnl_lock();
1687 if (dev->reg_state == NETREG_REGISTERED)
1688 unregister_netdevice(dev);
1689 rtnl_unlock();
1690 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001691 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692
Herbert Xu9c3fea62009-04-18 14:15:52 +00001693 tun = tfile->tun;
1694 if (tun)
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001695 sock_put(tun->socket.sk);
Herbert Xu9c3fea62009-04-18 14:15:52 +00001696
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001697 put_net(tfile->net);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001698 kfree(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
1700 return 0;
1701}
1702
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08001703static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001704 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 .llseek = no_llseek,
Badari Pulavartyee0b3e62006-09-30 23:28:47 -07001706 .read = do_sync_read,
1707 .aio_read = tun_chr_aio_read,
1708 .write = do_sync_write,
1709 .aio_write = tun_chr_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08001711 .unlocked_ioctl = tun_chr_ioctl,
1712#ifdef CONFIG_COMPAT
1713 .compat_ioctl = tun_chr_compat_ioctl,
1714#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 .open = tun_chr_open,
1716 .release = tun_chr_close,
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001717 .fasync = tun_chr_fasync
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718};
1719
1720static struct miscdevice tun_miscdev = {
1721 .minor = TUN_MINOR,
1722 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02001723 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725};
1726
1727/* ethtool interface */
1728
1729static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1730{
1731 cmd->supported = 0;
1732 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00001733 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 cmd->duplex = DUPLEX_FULL;
1735 cmd->port = PORT_TP;
1736 cmd->phy_address = 0;
1737 cmd->transceiver = XCVR_INTERNAL;
1738 cmd->autoneg = AUTONEG_DISABLE;
1739 cmd->maxtxpkt = 0;
1740 cmd->maxrxpkt = 0;
1741 return 0;
1742}
1743
1744static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1745{
1746 struct tun_struct *tun = netdev_priv(dev);
1747
Rick Jones33a5ba12011-11-15 14:59:53 +00001748 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
1749 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
1751 switch (tun->flags & TUN_TYPE_MASK) {
1752 case TUN_TUN_DEV:
Rick Jones33a5ba12011-11-15 14:59:53 +00001753 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 break;
1755 case TUN_TAP_DEV:
Rick Jones33a5ba12011-11-15 14:59:53 +00001756 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 break;
1758 }
1759}
1760
1761static u32 tun_get_msglevel(struct net_device *dev)
1762{
1763#ifdef TUN_DEBUG
1764 struct tun_struct *tun = netdev_priv(dev);
1765 return tun->debug;
1766#else
1767 return -EOPNOTSUPP;
1768#endif
1769}
1770
1771static void tun_set_msglevel(struct net_device *dev, u32 value)
1772{
1773#ifdef TUN_DEBUG
1774 struct tun_struct *tun = netdev_priv(dev);
1775 tun->debug = value;
1776#endif
1777}
1778
Jeff Garzik7282d492006-09-13 14:30:00 -04001779static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 .get_settings = tun_get_settings,
1781 .get_drvinfo = tun_get_drvinfo,
1782 .get_msglevel = tun_get_msglevel,
1783 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00001784 .get_link = ethtool_op_get_link,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785};
1786
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001787
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788static int __init tun_init(void)
1789{
1790 int ret = 0;
1791
Joe Perches6b8a66e2011-03-02 07:18:10 +00001792 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1793 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001795 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001796 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001797 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001798 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001799 }
1800
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001802 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001803 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001804 goto err_misc;
1805 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001806 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07001807err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001808 rtnl_link_unregister(&tun_link_ops);
1809err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 return ret;
1811}
1812
1813static void tun_cleanup(void)
1814{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001815 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001816 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817}
1818
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001819/* Get an underlying socket object from tun file. Returns error unless file is
1820 * attached to a device. The returned object works like a packet socket, it
1821 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1822 * holding a reference to the file for as long as the socket is in use. */
1823struct socket *tun_get_socket(struct file *file)
1824{
1825 struct tun_struct *tun;
1826 if (file->f_op != &tun_fops)
1827 return ERR_PTR(-EINVAL);
1828 tun = tun_get(file);
1829 if (!tun)
1830 return ERR_PTR(-EBADFD);
1831 tun_put(tun);
1832 return &tun->socket;
1833}
1834EXPORT_SYMBOL_GPL(tun_get_socket);
1835
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836module_init(tun_init);
1837module_exit(tun_cleanup);
1838MODULE_DESCRIPTION(DRV_DESCRIPTION);
1839MODULE_AUTHOR(DRV_COPYRIGHT);
1840MODULE_LICENSE("GPL");
1841MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02001842MODULE_ALIAS("devname:net/tun");